1 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
5 * config/riscv/linux.h (LINK_SPEC): Ditto.
7 2018-04-20 Martin Sebor <msebor@redhat.com>
10 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
12 (gimple_fold_builtin_stxcpy_chk): Same.
13 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
15 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
18 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
19 __powikf2 when long double is IEEE 128-bit.
21 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
23 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
24 step to make sure stack always aligned.
26 2018-04-20 Carl Love <cel@us.ibm.com>
29 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
32 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
33 Tom de Vries <tom@codesourcery.com>
36 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
37 Emit insns for calls too.
38 (nvptx_find_par): Always look for worker-level predecessor insn.
39 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
41 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
42 (nvptx_process_pars): Propagate frames for calls.
44 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
47 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
49 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
50 (ix86_handle_option): Don't handle OPT_mibt.
51 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
53 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
55 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
56 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
57 (ix86_target_macros): Define __CET__ with flag_cf_protection
59 * config/i386/i386.c (isa2_opts): Remove -mibt.
60 * config/i386/i386.h (TARGET_IBT): Removed.
61 (TARGET_IBT_P): Likewise.
62 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
63 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
64 * config/i386/i386.opt (mcet): Update help message.
67 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
68 -mcet as an alias for -mshstk.
70 2018-04-20 Richard Biener <rguenther@suse.de>
73 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
74 complexity by forcing a single use of the multiply operand.
76 2018-04-20 Martin Jambor <mjambor@suse.cz>
79 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
80 recursion dependency to only apply to non-clones.
82 2018-04-20 Martin Jambor <mjambor@suse.cz>
85 * ipa-cp.c (create_specialized_node): Check that clones of
86 self-recursive edges exist during IPA-CP.
88 2018-04-19 Toon Moene <toon@moene.org>
90 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
93 2018-04-19 Jakub Jelinek <jakub@redhat.com>
95 PR tree-optimization/85467
96 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
97 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
98 VECTOR_CST element to type.
100 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
103 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
104 * config/i386/i386.md (builtin_setjmp_setup): Removed.
105 (builtin_longjmp): Likewise.
106 (save_stack_nonlocal): New pattern.
107 (restore_stack_nonlocal): Likewise.
109 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
112 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
113 Replace ASM_OUTPUT_LABEL with fprintf.
115 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
118 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
119 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
120 * config/i386/i386-c.c (ix86_target_macros_internal): Also
121 define __IBT__ and __SHSTK__ for -fcf-protection.
122 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
124 (ix86_trampoline_init): Likewise.
125 (x86_output_mi_thunk): Likewise.
126 (ix86_notrack_prefixed_insn_p): Likewise.
127 (ix86_option_override_internal): Don't disallow -fcf-protection.
128 * config/i386/i386.md (rdssp<mode>): Also enable for
130 (incssp<mode>): Likewise.
131 (nop_endbr): Likewise.
132 * config/i386/i386.opt (mcet): Change help message to built-in
136 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
137 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
138 enable CET built-in functions.
140 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
142 * common/config/i386/i386-common.c
143 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
144 OPTION_MASK_ISA_MOVDIRI_UNSET,
145 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
146 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
147 * config.gcc (movdirintrin.h): New header.
148 * config/i386/cpuid.h (bit_MOVDIRI,
149 bit_MOVDIR64B): New bits.
150 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
152 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
153 (VOID, PVOID, PCVOID)): New function types.
154 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
155 __builtin_ia32_directstoreu_u64,
156 __builtin_ia32_movdir64b): New builtins.
157 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
158 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
160 (ix86_valid_target_attribute_inner_p): Ditto.
161 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
162 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
163 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
164 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
165 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
166 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
167 (movdiri<mode>, movdir64b_<mode>): New.
168 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
169 * config/i386/immintrin.h: Include movdirintrin.h.
170 * config/i386/movdirintrin.h: New file.
171 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
173 2018-04-19 Richard Biener <rguenther@suse.de>
176 * cfg.c (clear_bb_flags): When loop state says we have
177 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
179 2018-04-19 Richard Biener <rguenther@suse.de>
181 PR tree-optimization/84737
182 * tree-vect-data-refs.c (vect_copy_ref_info): New function
183 copying restrict info.
184 (vect_setup_realignment): Use it.
185 * tree-vectorizer.h (vect_copy_ref_info): Declare.
186 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
187 the first DR to all generated stores.
188 (vectorizable_load): Likewise for loads.
190 2018-04-19 Jakub Jelinek <jakub@redhat.com>
192 PR tree-optimization/85446
193 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
194 the integral and pointer types to have the same precision.
196 * doc/install.texi: Document --disable-cet being the default and
199 2018-04-18 Martin Liska <mliska@suse.cz>
201 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
204 2018-04-18 Martin Liska <mliska@suse.cz>
207 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
210 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
211 arguments if they are comparable.
213 2018-04-18 Martin Liska <mliska@suse.cz>
216 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
219 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
222 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
225 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
226 ENDBR after calling __morestack.
228 2018-04-18 David Malcolm <dmalcolm@redhat.com>
231 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
232 by using gcc_base_ver to generate a gcc_driver_version, and use
233 it when generating GCC_DRIVER_NAME.
234 * configure: Regenerate.
236 2018-04-18 Jakub Jelinek <jakub@redhat.com>
239 * config.gcc: Obsolete powerpc*-*-*spe*.
241 2018-04-17 Jakub Jelinek <jakub@redhat.com>
244 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
245 (stabstr_D): Change type of unum from unsigned int to
246 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
249 2018-04-17 Jim Wilson <jimw@sifive.com>
252 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
253 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
254 Set arg_pointer_offset after using pretend_args_size.
256 2018-04-17 Jakub Jelinek <jakub@redhat.com>
258 PR rtl-optimization/85431
259 * dse.c (record_store): Ignore zero width stores.
262 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
263 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
264 __builtin_stack_restore rather than after it.
265 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
266 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
267 argument instead of virtual_dynamic_stack_rtx.
269 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
271 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
273 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
274 Add note to error message to explain internal mapping of overloaded
275 built-in function name to non-overloaded built-in function name.
276 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
279 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
282 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
283 where the inputs overlap with the output.
285 2018-04-17 Jakub Jelinek <jakub@redhat.com>
287 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
288 (=v, v) alternative and explicit "memory" attribute.
289 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
290 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
292 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
293 "sselog1" type instead of "sselog".
294 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
295 "sselog". Remove explicit "memory" attribute.
296 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
297 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
299 (vec_extract_hi_v32hi): Merge all alternatives into one, use
300 "sselog1" type instead of "sselog". Remove explicit "memory"
302 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
303 use "sselog1" type instead of "sselog". Remove explicit "memory"
305 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
306 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
308 (vec_extract_hi_v64qi): Merge all alternatives into one, use
309 "sselog1" type instead of "sselog". Remove explicit "memory"
311 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
312 use "sselog1" type instead of "sselog". Remove explicit "memory"
316 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
319 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
322 2018-04-17 Martin Jambor <mjambor@suse.cz>
325 * ipa-cp.c (create_specialized_node): Call
326 expand_all_artificial_thunks if necessary.
328 2018-04-17 Martin Liska <mliska@suse.cz>
331 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
332 in message, remote space in between '_G' and '('.
334 2018-04-17 Jakub Jelinek <jakub@redhat.com>
337 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
338 avx512f_vmcmp<mode>3<round_saeonly_name>,
339 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
340 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
341 avx512f_rndscale<mode><round_saeonly_name>,
342 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
343 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
344 Use %<iptr>2 instead of %2 for -masm=intel.
345 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
346 avx512f_vcvttss2usi<round_saeonly_name>,
347 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
349 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
350 avx512f_vcvttsd2usi<round_saeonly_name>,
351 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
352 Use %q1 instead of %1 for -masm=intel.
353 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
354 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
355 of %3 for -masm=intel.
356 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
358 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
360 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
362 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
363 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
365 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
367 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
368 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
369 %g1 and one with %0 and %1.
370 (avx512er_vmrcp28<mode><round_saeonly_name>,
371 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
373 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
374 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
375 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
376 of %0 and %{%4%} for -masm=intel.
377 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
378 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
379 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
380 order of %0 and %{%5%}%{z%} for -masm=intel.
382 2018-04-17 Jan Hubicka <jh@suse.cz>
385 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
387 2018-04-17 Martin Liska <mliska@suse.cz>
390 * multiple_target.c (create_dispatcher_calls): Set apostrophes
391 for target_clone error message. Make default implementation
392 clone to be a local declaration.
393 (separate_attrs): Add new argument and check for an empty
395 (expand_target_clones): Handle it.
396 (ipa_target_clone): Make redirection just for target_clones
399 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
400 Tom de Vries <tom@codesourcery.com>
403 * omp-expand.c (expand_oacc_for): Add dummy false branch for
404 tiled basic blocks without omp continue statements.
406 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
409 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
410 vec_extract expression as having side effects to make sure it gets
413 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
416 * config/i386/i386.c (get_builtin_code_for_version): Check
419 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
422 * gcc/config.gcc: Support "skylake".
423 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
425 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
426 (processor_target_table): Add "skylake".
427 (ix86_option_override_internal): Add "skylake".
428 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
429 PROCESSOR_CANNONLAKE.
430 (get_builtin_code_for_version): Fix priority for
431 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
432 PROCESSOR_SKYLAKE-AVX512.
433 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
434 (processor_type): Add PROCESSOR_SKYLAKE.
436 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
437 Jason Merrill <jason@redhat.com>
440 * convert.c (convert_to_integer_1): Use direct recursion for
441 enumeral types and types with a precision less than the number
442 of bits in their mode.
444 2018-04-16 Julia Koval <julia.koval@intel.com>
447 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
448 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
450 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
453 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
454 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
455 and -mno-direct-move.
457 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
460 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
461 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
462 Ensure negative shifts result in {0}.
464 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
466 PR rtl-optimization/79916
467 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
468 regs (if any) to define how to gnerate SD moves when LRA is in
471 2018-04-13 Jakub Jelinek <jakub@redhat.com>
473 PR rtl-optimization/85393
474 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
475 * except.c (expand_dw2_landing_pad_for_region): Make static.
476 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
477 a label and unconditional jump to old_bb, rather than
478 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
481 PR rtl-optimization/85376
482 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
483 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
484 instead of a specific value.
486 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
487 Bin Cheng <bin.cheng@arm.com>
489 PR tree-optimization/82965
490 PR tree-optimization/83991
491 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
492 by_profile_only parameter.
493 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
494 information if the loop was predicted to iterate too many times.
495 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
497 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
500 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
503 2018-04-13 Martin Liska <mliska@suse.cz>
504 Jakub Jelinek <jakub@redhat.com>
507 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
508 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
509 * builtins.c (expand_builtin_memory_copy_args): Use
510 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
511 handle dest_addr == pc_rtx.
513 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
516 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
517 asked to not generate direct moves.
518 (fix_trunc<mode>si2_stfiwx): Similar.
519 (fix_trunc<mode>si2_internal): Similar.
521 2018-04-12 Jakub Jelinek <jakub@redhat.com>
524 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
525 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
526 lookup if dest in some wider mode is known to be const0_rtx and
527 if so, record permanent equivalence for it to be ZERO_EXTEND of
528 the narrower mode destination.
530 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
532 * lto-streamer-out.c (output_function): Revert 259346.
533 * omp-expand.c (expand_oacc_for): Likewise.
535 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
537 PR rtl-optimization/85354
538 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
539 * sel-sched.c (sel_global_init): ... here.
541 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
544 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
545 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
546 mode for PE-COFF targets.
547 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
548 (i386_pe_asm_lto_end): Likewise.
549 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
550 (TARGET_ASM_LTO_END): Likewise.
551 * config/i386/winnt.c (saved_debug_info_level): New static variable.
552 (i386_pe_asm_lto_start): New function.
553 (i386_pe_asm_lto_end): Likewise.
555 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
556 Richard Biener <rguenther@suse.de>
559 * lto-streamer-out.c (output_function): Fix CFG loop state before
561 * omp-expand.c (expand_oacc_for): Handle calls to internal
562 functions like regular functions.
564 2018-04-12 Richard Biener <rguenther@suse.de>
567 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
568 for the early LTO debug to properly generate references to it
569 during DIE emission. Do not re-use that for the skeleton for
571 (dwarf2out_early_finish): Likewise.
573 2018-04-12 Jakub Jelinek <jakub@redhat.com>
577 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
578 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
579 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
580 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
581 and output is a reg, avoid creating invalid lowpart subreg, but
582 instead split into a 512-bit move. Don't split if not AVX512VL,
583 input is xmm16+ reg and output is a mem.
584 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
585 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
586 xmm16+ reg and output is a mem.
588 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
590 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
591 also for flag_dwarf2_cfi_asm.
593 2018-04-12 Jakub Jelinek <jakub@redhat.com>
595 PR rtl-optimization/85342
596 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
597 a bool scalar var inside of the loop instead. Don't try to update
598 recog_data.operand after failed apply_change_group.
600 2018-04-12 Tom de Vries <tom@codesourcery.com>
603 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
604 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
605 array with flexible array member as array without given dimension.
606 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
607 argument for undefined param to true.
609 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
612 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
613 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
614 from PowerPC section.
615 * config/rs6000/sysv4.opt (mcall-): Improve help text.
616 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
617 help text that is too long.
618 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
619 help text that is too long.
620 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
621 help text that is too long.
623 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
625 * config/alpha/alpha.md (stack_probe_internal): Rename
626 from "probe_stack". Update all callers.
628 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
630 PR rtl-optimization/84566
631 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
632 sched_macro_fuse_insns.
634 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
637 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
638 (compute_block_dependences): ... from here.
640 2018-04-11 Jakub Jelinek <jakub@redhat.com>
642 PR tree-optimization/85331
643 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
644 from int to HOST_WIDE_INT.
646 2018-04-11 Martin Jambor <mjambor@suse.cz>
649 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
650 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
651 not the same as the source val.
652 (cgraph_edge_brings_value_p): New parameter.
653 (gather_edges_for_value): Pass destination value to
654 cgraph_edge_brings_value_p.
655 (perhaps_add_new_callers): Likewise.
656 (get_info_about_necessary_edges): Likewise and exclude values brought
657 only by self-recursive edges.
658 (create_specialized_node): Redirect only clones of self-calling edges.
659 (+self_recursive_pass_through_p): New function.
660 (find_more_scalar_values_for_callers_subset): Use it.
661 (find_aggregate_values_for_callers_subset): Likewise.
662 (known_aggs_to_agg_replacement_list): Removed.
663 (decide_whether_version_node): Re-calculate known constants for all
664 remaining context clones.
666 2018-04-11 Richard Biener <rguenther@suse.de>
669 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
670 from early DWARF output.
671 (dwarf2out_early_finish): Output line info unconditionally into
672 early DWARF and add reference to it.
674 2018-04-11 Jakub Jelinek <jakub@redhat.com>
677 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
678 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
679 other than V2DFmode using iptr mode attribute.
680 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
682 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
684 PR rtl-optimization/84659
685 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
687 2018-04-11 Jakub Jelinek <jakub@redhat.com>
690 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
692 (output_loc_list): Pass address of a dummy size variable even in the
693 locview handling loop.
694 (index_location_lists): Add comment on why skip_loc_list_entry can't
697 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
700 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
703 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
706 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
707 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
708 and -mstring-compare-inline-limit.
710 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
713 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
714 for stack clash protection in a register whenever we need it to be in
717 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
719 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
720 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
722 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
725 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
727 (mlong-double-): Ditto.
728 * config/rs6000/sysv4.opt (msdata=): Ditto.
731 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
733 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
734 erroneous entries for
735 "vector int vec_ldl (int, long int *)", and
736 "vector unsigned int vec_ldl (int, unsigned long int *)".
737 Add comments and entries for
738 "vector bool char vec_ldl (int, bool char *)",
739 "vector bool short vec_ldl (int, bool short *)",
740 "vector bool int vec_ldl (int, bool int *)",
741 "vector bool long long vec_ldl (int, bool long long *)",
742 "vector pixel vec_ldl (int, pixel *)",
743 "vector long long vec_ldl (int, long long *)",
744 "vector unsigned long long vec_ldl (int, unsigned long long *)".
745 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
746 type tree bool_long_long_type_node and correct definition of
747 bool_V2DI_type_node to make reference to this new type tree.
748 (rs6000_mangle_type): Replace erroneous reference to
749 bool_long_type_node with bool_long_long_type_node.
750 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
751 comments to emphasize sign distinctions for char and int types and
752 replace RS6000_BTI_bool_long constant with
753 RS6000_BTI_bool_long_long constant. Also add comment to restrict
754 use of RS6000_BTI_pixel.
755 (bool_long_type_node): Remove this macro definition.
756 (bool_long_long_type_node): New macro definition
758 2018-04-10 Jakub Jelinek <jakub@redhat.com>
760 PR rtl-optimization/85300
761 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
762 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
763 simplify_unary_operation fails.
765 2018-04-10 Martin Liska <mliska@suse.cz>
767 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
768 cgraph_edge and ipa_ref.
770 2018-04-10 Jakub Jelinek <jakub@redhat.com>
775 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
776 computation of the VEC_MERGE selector from mask.
777 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
778 Fix decoding of the VEC_MERGE selector into mask.
780 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
782 PR tree-optimization/85286
783 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
785 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
787 * final.c (final_1): Set insn_last_address as well as
788 insn_current_address.
790 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
793 * explow.c (emit_stack_probe): Call validize_mem on memory location
794 before passing it to gen_probe_stack. Create address operand and
795 legitimize it for the probe_stack_address case.
797 2018-04-09 Jan Hubicka <jh@suse.cz>
800 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
801 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
802 * tree.c (free_lang_data_in_type): Fix handling of binfos;
804 (free_lang_data): Rebuild type inheritance graph.
806 2018-04-09 Martin Sebor <msebor@redhat.com>
808 * invoke.texi (-finline-small-functions): Mention other optimization
810 (-findirect-inlining, -fpartial-inlining): Same.
811 (-finline-functions-called-once): Same.
812 (-freorder-blocks-and-partition): Same.
814 2018-04-09 Jan Hubicka <jh@suse.cz>
817 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
818 jumps; choose last target that matches the criteria (i.e.
819 no partition changes for non-crossing jumps).
820 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
821 support for redirecting crossing jumps to non-crossing.
823 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
825 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
826 also for naked functions.
828 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
830 * config/arc/arc.md (add_shift): New pattern.
831 (add_shift2): Likewise.
832 (sub_shift): Likewise.
833 (sub_shift_cmp0_noout): Likewise.
834 (compare_si_ashiftsi): Likewise.
835 (xbfu_cmp0_noout): New combine pattern.
836 (xbfu_cmp0"): Likewise.
837 (movsi_set_cc_insn): Place the predicable variant first.
838 (commutative_binary_cmp0_noout): Remove clobber.
839 (commutative_binary_cmp0): New pattern.
840 (noncommutative_binary_cmp0): Likewise.
841 (noncommutative_binary_cmp0_noout): Likewise.
842 (noncommutative_binary_comparison_result_used): Removed.
843 (rsub_cmp0): New pattern.
844 (rsub_cmp0_noout): Likewise.
845 (extzvsi): Changed, keep only meaningful variants.
846 (SQH, SEZ): New iterators.
847 (SQH_postfix): New mode attribute.
848 (SEZ_prefix): New code attribute.
849 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
850 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
851 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
853 (noncommutative_operator): Check the availability of barrel
856 2018-04-09 Richard Biener <rguenther@suse.de>
858 PR tree-optimization/85284
859 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
860 Only use the niter constraining form of simple_iv when the exit
863 2018-04-09 Tom de Vries <tom@codesourcery.com>
866 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
867 (define_expand "*memory_barrier"): New define_expand.
868 (define_insn "memory_barrier"): New insn.
870 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
872 PR rtl-optimization/80463
873 PR rtl-optimization/83972
874 PR rtl-optimization/83480
876 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
877 correct producer for the insn.
878 (tidy_control_flow): Fixup seqnos in case of debug insns.
880 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
882 PR rtl-optimization/83913
884 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
885 different sched-times when merging exprs.
887 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
889 PR rtl-optimization/83962
891 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
892 tidy_fallthru_edge and tidy_control_flow.
894 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
896 PR rtl-optimization/83530
898 * sel-sched.c (force_next_insn): New global variable.
899 (remove_insn_for_debug): When force_next_insn is true, also leave only
900 next insn in the ready list.
901 (sel_sched_region): When the region wasn't scheduled, make another pass
902 over it with force_next_insn set to 1.
904 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
906 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
908 * config/nds32/constants.md (unspec_volatile_element): Add enum values
909 for interrupt control.
910 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
911 functions for interrupt control.
912 * config/nds32/nds32-intrinsic.md: Likewise.
913 * config/nds32/nds32_intrinsic.h: Likewise.
914 * config/nds32/nds32.h (nds32_builtins): Likewise.
916 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
918 * config/nds32/nds32.c (nds32_init_machine_status,
919 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
920 strict_aligned_p field.
921 (nds32_expand_to_rtl_hook): New function.
922 (TARGET_EXPAND_TO_RTL_HOOK): Define.
923 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
925 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
926 Chung-Ju Wu <jasonwucj@gmail.com>
928 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
929 * config/nds32/nds32-n7.md: New file.
930 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
931 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
933 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
934 * config/nds32/nds32.md (pipeline_model): Add n7.
935 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
936 * config/nds32/pipelines.md: Include n7 settings.
938 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
939 Chung-Ju Wu <jasonwucj@gmail.com>
941 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
942 * config/nds32/nds32-e8.md: New file.
943 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
944 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
946 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
947 * config/nds32/nds32.md (pipeline_model): Add e8.
948 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
949 * config/nds32/pipelines.md: Include e8 settings.
951 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
952 Chung-Ju Wu <jasonwucj@gmail.com>
954 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
955 * config/nds32/nds32-n8.md: New file.
956 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
957 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
959 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
960 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
961 * config/nds32/nds32.md (pipeline_model): Add n8.
962 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
963 * config/nds32/pipelines.md: Include n8 settings.
965 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
966 Chung-Ju Wu <jasonwucj@gmail.com>
968 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
969 * config/nds32/nds32-n9-2r1w.md: New file.
970 * config/nds32/nds32-n9-3r2w.md: New file.
971 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
972 nds32_register_ports): New or modify for cpu n9.
973 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
975 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
976 * config/nds32/nds32-utils.c: New file.
977 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
978 TARGET_MUL_SLOW): Define.
979 * config/nds32/nds32.md (pipeline_model): New attribute.
980 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
981 New options that support cpu n9.
982 * config/nds32/pipelines.md: Include n9 settings.
983 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
985 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
987 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
988 information if necessary.
989 (output_cond_branch_compare_zero): Likewise.
990 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
991 (nds32_target_alignment): Refine for alignment.
992 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
993 (FUNCTION_BOUNDARY): Modify.
994 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
996 * config/nds32/nds32.opt (malways-align, malign-functions): New.
998 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1000 * config/nds32/constants.md (unspec_volatile_element): Add values for
1001 TLB operation and data prefetch.
1002 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1003 functions for TLB operation and data prefetch.
1004 * config/nds32/nds32-intrinsic.md: Likewise.
1005 * config/nds32/nds32_intrinsic.h: Likewise.
1006 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1007 (nds32_print_operand): Likewise.
1008 * config/nds32/nds32.h (nds32_builtins): Likewise.
1010 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1011 Andrew Pinski <pinsika@gcc.gnu.org>
1014 * match.pd: Use constant_boolean_node of correct type instead of
1015 boolean_true_node or boolean_false_node for simplifying
1016 pointer comparisons to zero.
1018 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1020 PR tree-optimization/80021
1021 * tree.c (verify_type_variant): Make error call in verify_variant_match
1022 translatable and remove final full stop.
1024 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1026 * config/nds32/constants.md (unspec_volatile_element): Add
1027 UNSPEC_VOLATILE_EH_RETURN.
1028 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1029 nds32_output_stack_pop): Support dwarf exception handling process.
1030 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1031 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1032 exception handling process.
1033 (nds32_compute_stack_frame): Likewise.
1034 (nds32_return_addr_rtx): Likewise.
1035 (nds32_initial_elimination_offset): Likewise.
1036 (nds32_expand_prologue): Likewise.
1037 (nds32_expand_epilogue): Likewise.
1038 (nds32_dynamic_chain_address): New function.
1039 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1041 (DYNAMIC_CHAIN_ADDRESS): Define.
1042 (EH_RETURN_DATA_REGNO): Define.
1043 (EH_RETURN_STACKADJ_RTX): Define.
1044 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1045 patterns for dwarf exception handling.
1047 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1049 * config/nds32/nds32.h: Clean up obsolete macros.
1051 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1053 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1054 Add enum values for particular instructions.
1055 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1056 particular intrinsic functions.
1057 * config/nds32/nds32-intrinsic.md: Likewise.
1058 * config/nds32/nds32_intrinsic.h: Likewise.
1059 * config/nds32/nds32.h (nds32_builtins): Likewise.
1060 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1061 (btst, ave): New patterns for particular instructions.
1063 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1065 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1066 Add enum values for atomic load/store and memory sync.
1067 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1069 * config/nds32/nds32-intrinsic.md: Likewise.
1070 * config/nds32/nds32_intrinsic.h: Likewise.
1071 * config/nds32/nds32.h (nds32_builtins): Likewise.
1073 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1075 PR tree-optimization/85257
1076 * fold-const.c (native_encode_vector): If not all elts could fit
1077 and off is -1, return 0 rather than offset.
1078 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1079 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1080 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1081 adjust buffer in native_interpret_expr call.
1083 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1085 * config/nds32/constants.md (unspec_volatile_element): Add cache
1086 control enum values.
1087 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1088 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1089 * config/nds32/nds32.c (nds32_cctl_names): New.
1090 (nds32_print_operand): Handle cache control register names.
1091 * config/nds32/nds32.h (nds32_builtins): New enum values.
1092 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1094 * config/nds32/nds32.md (type): Add mmu.
1095 * config/nds32/pipelines.md (simple_insn): Add mmu.
1097 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1099 * config/nds32/nds32.md (type): Remove call.
1100 * config/nds32/pipelines.md (simple_insn): Likewise.
1102 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1104 * config/nds32/constants.md (unspec_volatile_element): Add
1105 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1106 UNSPEC_VOLATILE_FMFCFG.
1107 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1108 description for fmfcfg and fmfcsr.
1109 (bdesc_1arg): Add fmtcsr.
1110 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1111 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1112 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1113 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1114 unspec_fmfcfg): New patterns.
1115 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1116 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1117 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1118 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1119 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1120 __nds32__fmfcfg): Define.
1122 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1124 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1125 intrinsic register names.
1126 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1127 intrinsic register enum values and macros.
1129 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1131 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1132 for load/store addressing form.
1133 (nds32_print_operand_address): Likewise.
1135 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1138 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1139 based on LABEL_REF. Remove useless assertion.
1140 (pic_address_needs_scratch): Fix formatting.
1141 (sparc_legitimize_pic_address): Minor tweaks.
1142 (sparc_delegitimize_address): Adjust assertion accordingly.
1143 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1144 into symbolic_operand.
1145 (movsi_high_pic_label_ref): Likewise.
1146 (movsi_lo_sum_pic_label_ref): Likewise.
1147 (movdi_pic_label_ref): Likewise.
1148 (movdi_high_pic_label_ref): Likewise.
1149 (movdi_lo_sum_pic_label_ref): Likewise.
1151 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1153 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1154 custom LIB_SPEC setup.
1156 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1157 Kito Cheng <kito.cheng@gmail.com>
1159 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1160 * config/riscv/freebsd.h: New.
1162 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1164 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
1165 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
1168 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1169 Kito Cheng <kito.cheng@gmail.com>
1171 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
1172 nds32_output_call, nds32_symbol_binds_local_p): New functions.
1173 * config/nds32/nds32-protos.h (nds32_output_call,
1174 nds32_output_return): Declare.
1175 * config/nds32/nds32.md: Refine all the call and return patterns.
1177 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1180 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
1181 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
1183 PR rtl-optimization/84872
1184 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
1185 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
1188 2018-04-06 Tamar Christina <tamar.christina@arm.com>
1190 * expr.c (copy_blkmode_to_reg): Revert 254862.
1191 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
1193 2018-04-06 Richard Biener <rguenther@suse.de>
1196 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
1197 after seeing a component reference with an adjacent field. Treat
1198 refs to arrays at struct end of external decls similar to
1199 refs to unconstrained commons.
1201 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1204 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1205 look through SAVE_EXPRs with non-side-effects argument. Adjust
1207 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1210 2018-04-06 Richard Biener <rguenther@suse.de>
1213 * alias.c (find_base_term): New wrapper around find_base_term
1214 unwinding CSELIB_VAL_PTR changes.
1215 (find_base_term): Do not restore CSELIB_VAL_PTR during the
1218 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1220 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1222 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1223 constant definitions.
1224 ("nop"): lr 0,0 -> nopr r0
1225 ("nop_lr0", "nop_lr1"): New insn definitions.
1227 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1229 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1230 NDS32_V3PUSH_AVAILABLE_P macro.
1232 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
1233 Chung-Ju Wu <jasonwucj@gmail.com>
1235 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1236 (nds32*-*-*): Add float and fpu_config into supported_defaults.
1237 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1238 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1239 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1240 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1241 * config/nds32/constraints.md: New constraints and checking for hard
1242 float configuration.
1243 * config/nds32/iterators.md: New mode iterator and attribute for hard
1244 float configuration.
1245 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1247 * config/nds32/nds32-fpu.md: New file.
1248 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1249 deal with hard float code generation.
1250 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1252 (abi_type, float_reg_number): New enum type.
1253 * config/nds32/nds32-predicates.c: New predicates for hard float.
1254 * config/nds32/nds32-protos.h: Declare functions for hard float.
1255 * config/nds32/nds32.c: Implementation for hard float configuration.
1256 * config/nds32/nds32.h: Definitions for hard float configuration.
1257 * config/nds32/nds32.md: Include hard float machine description and
1258 modify patterns for hard float configuration.
1259 * config/nds32/nds32.opt: New options for hard float configuration.
1260 * config/nds32/predicates.md: New predicates for hard float
1263 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1265 * common/config/nds32/nds32-common.c
1266 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1268 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1271 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1272 CONSTRUCTOR_ELT (ctor, ...)->value.
1274 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1277 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1279 2018-04-05 Tom de Vries <tom@codesourcery.com>
1282 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1285 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1286 Kito Cheng <kito.cheng@gmail.com>
1288 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1289 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1290 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1291 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1293 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1294 Kito Cheng <kito.cheng@gmail.com>
1296 * config/nds32/constraints.md (Ufe): New memory constraint.
1297 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1298 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1299 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1301 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1302 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1304 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1306 * config/nds32/nds32.md: Use optimize_size in the condition for
1307 alu-shift instructions.
1309 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1311 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1313 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1315 * config/nds32/nds32.md (negsi2): Refine pattern.
1317 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1318 Chung-Ju Wu <jasonwucj@gmail.com>
1320 * config/nds32/iterators.md (shift_rotate): New code iterator.
1321 (shift): New code attribute.
1322 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1323 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1324 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1325 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1326 bit-wise operations.
1327 (andsi3, *andsi3): Ditto.
1328 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1329 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1330 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1331 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1332 nds32_ior_operand, nds32_xor_operand): New predicates.
1334 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1336 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1337 (addsi3, subsi3): ... this.
1339 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1341 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1343 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1345 * config/nds32/nds32.md: Adjust indention.
1347 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1349 * config/nds32/nds32.md (feature): New attribute.
1351 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1353 * config/nds32/nds32.md (subtype): New attribute.
1355 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1358 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1359 expansion to perform a bitwise AND of the argument followed by a
1360 boolean negation of the result.
1362 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1364 PR rtl-optimization/84878
1365 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1366 the basic block. Assert the use reference is not artificial and that
1367 it has an associated insn.
1369 2018-04-04 Michael Matz <matz@suse.de>
1371 * builtins.c (compute_objsize): Pass correct operand
1372 to array_at_struct_end_p.
1374 2018-04-04 Richard Biener <rguenther@suse.de>
1377 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1378 from contexts for DINFO_LEVEL_TERSE and below.
1380 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1382 * config/nds32/nds32-doubleword.md (move_<mode>): Require
1383 resiter_operand condition.
1384 * config/nds32/nds32.md (*move<mode>): Ditto.
1386 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1387 Monk Chiang <sh.chiang04@gmail.com>
1389 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1391 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1393 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1395 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1396 Kito Cheng <kito.cheng@gmail.com>
1398 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1399 nds32_cond_code_str, output_cond_branch,
1400 output_cond_branch_compare_zero, nds32_expand_cbranch,
1401 nds32_expand_cstore, nds32_expand_movcc,
1402 nds32_output_cbranchsi4_equality_zero,
1403 nds32_output_cbranchsi4_equality_reg,
1404 nds32_output_cbranchsi4_equality_reg_or_const_int,
1405 nds32_output_cbranchsi4_greater_less_zero: New functions.
1406 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1407 nds32_expand_cstore, nds32_expand_movcc,
1408 nds32_output_cbranchsi4_equality_zero,
1409 nds32_output_cbranchsi4_equality_reg,
1410 nds32_output_cbranchsi4_equality_reg_or_const_int,
1411 nds32_output_cbranchsi4_greater_less_zero): Declare.
1412 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1413 nds32_rimm11s_operand): New predicates.
1414 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1415 * config/nds32/nds32.md: Rewrite all the branch and conditional move
1418 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1420 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
1421 * config/nds32/nds32.md: Ditto.
1422 * config/nds32/pipelines.md: Ditto.
1424 2018-04-04 Richard Biener <rguenther@suse.de>
1426 PR tree-optimization/85168
1427 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
1428 propagating abnormals.
1430 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1432 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
1434 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1435 Kito Cheng <kito.cheng@gmail.com>
1437 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
1438 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
1439 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
1440 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
1441 * config/nds32/nds32.md (sibcall_internal): New.
1442 (sibcall_register): Remove.
1443 (sibcall_immediate): Remove.
1444 (sibcall_value_internal): New.
1445 (sibcall_value_register): Remove.
1446 (sibcall_value_immediate): Remove.
1447 * config/nds32/predicates.md (nds32_general_register_operand): New.
1448 (nds32_call_address_operand): New.
1450 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1452 PR rtl-optimization/85167
1453 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1454 bb_defs if *split_p, instead preinitialize it to NULL.
1456 PR tree-optimization/85156
1457 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1458 evaluating the argument multiple times.
1460 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
1462 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1464 (_mm_cvtpd_ps): Likewise.
1465 (_mm_cvttpd_epi32): Likewise.
1466 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1467 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1468 vector, pixel, and bool following altivec.h include.
1470 2018-04-03 Martin Sebor <msebor@redhat.com>
1472 * doc/extend.texi (Common Function Attributes): Clarify.
1473 (const attribute): Likewise.
1474 (pure attribute): Likewise.
1476 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1479 * config/i386/i386.c (ix86_expand_vector_set): Use
1480 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
1482 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
1484 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1485 instructions when changing rounding bits to preserve precision bits
1486 in the x87 control word.
1488 2018-04-03 Martin Liska <mliska@suse.cz>
1490 PR tree-optimization/82491
1491 * rtl.h (strip_offset_and_add): Replace += suboffset with
1492 poly_uint64 () + suboffset.
1494 2018-03-29 Martin Liska <mliska@suse.cz>
1495 Martin Jambor <mjambor@suse.cz>
1498 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1499 param_type is not an integral or pointer type.
1501 2018-04-03 Richard Biener <rguenther@suse.de>
1503 * sese.h (recompute_all_dominators): Remove.
1505 2018-04-02 Martin Sebor <msebor@redhat.com>
1507 * doc/invoke.texi (-Wrestrict): Fix typos.
1509 2018-04-02 Jim Wilson <jimw@sifive.com>
1511 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1512 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1513 (<optab>di3, <optab>si3_extend): Likewise.
1514 (<optab>si3_mask, <optab>si3_mask_1): New.
1515 (<optab>di3_mask, <optab>di3_mask_1): New.
1516 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1517 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1518 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1520 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1522 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1525 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1527 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1528 (nds32_canonicalize_comparison): New function.
1530 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1531 Kito Cheng <kito.cheng@gmail.com>
1532 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1534 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1535 * config/nds32/constants.md (unspec_volatile_element): Add
1536 UNSPEC_VOLATILE_RELAX_GROUP.
1537 * config/nds32/nds32-relax-opt.c: New file.
1538 * config/nds32/nds32-predicates.c
1539 (nds32_symbol_load_store_p): New function.
1540 * config/nds32/nds32-protos.h
1541 (nds32_symbol_load_store_p): Declare function.
1542 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1543 * config/nds32/nds32.c
1544 (nds32_register_pass): New function to register pass.
1545 (nds32_register_passes): New function to register passes.
1546 * config/nds32/nds32.md (relax_group): New pattern.
1547 * config/nds32/nds32.opt (mrelax-hint): New option.
1548 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1550 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
1552 * config/nds32/t-nds32: Modify files dependency.
1554 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1556 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1557 (PROFILE_HOOK): Define its implementation.
1559 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1561 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1562 type and 32-bit size.
1564 2018-04-01 Jakub Jelinek <jakub@redhat.com>
1567 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1568 (V_128_256): New mode iterator.
1569 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1570 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1571 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1573 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1574 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1576 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1579 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1580 NaN inputs correctly.
1582 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1585 * config/rs6000/vsx.md (??r): New mode attribute.
1586 (*vsx_mov<mode>_64bit): Use it.
1587 (*vsx_mov<mode>_32bit): Likewise.
1589 2018-03-30 Martin Sebor <msebor@redhat.com>
1591 PR tree-optimization/84818
1592 * builtins.c (check_access): Use warning_n.
1594 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1597 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1599 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1602 2018-03-30 Julia Koval <julia.koval@intel.com>
1605 * x86-tune.def (movx, partial_reg_dependency): Enable for
1608 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
1611 * lra-constraints.c (process_alt_operands): Move setting
1612 this_alternative_matches below.
1614 2018-03-29 Martin Liska <mliska@suse.cz>
1617 * doc/invoke.texi: Document how LTO works with debug info.
1618 Describe auto-load support of binutils. Mention 'x86-64'
1619 as valid option value of -march option.
1621 2018-03-29 Jakub Jelinek <jakub@redhat.com>
1623 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
1626 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
1627 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
1628 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
1631 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1634 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
1635 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
1636 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
1637 for RS6000_BTM_POWERPC64.
1638 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
1639 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
1640 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
1645 2018-03-28 Carl Love <cel@us.ibm.com>
1648 2017-09-27 Carl Love <cel@us.ibm.com>
1650 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
1651 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
1652 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
1655 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1657 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
1658 instead of __vector bool.
1659 (_mm_max_pu8): Likewise.
1660 (_mm_min_pi16): Likewise.
1662 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1665 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
1666 (DIVWEUO): Likewise.
1668 (DIVDEUO): Likewise.
1669 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
1670 DIVWEUO and DIVDEUO.
1671 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
1672 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
1673 (div_extend): Likewise.
1674 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
1676 (__builtin_divweuo): Likewise.
1677 (__builtin_divdeo): Likewise.
1678 (__builtin_divdeuo): Likewise.
1680 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1683 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
1684 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
1686 PR tree-optimization/82004
1687 * gimple-match-head.c (optimize_pow_to_exp): New function.
1688 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
1689 Don't fold to exp if optimize_pow_to_exp is false.
1691 2018-03-28 Martin Liska <mliska@suse.cz>
1694 * calls.c (initialize_argument_information): Fix trailing space.
1695 * common.opt: Fix typo and provide better explanation for
1696 -fsanitize-coverage option.
1697 * config/i386/i386.opt: Fix typo.
1699 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1700 Martin Liska <mliska@suse.cz>
1703 * gimplify.c (asan_poison_variable): Don't do the check for
1704 gimplify_omp_ctxp here.
1705 (gimplify_decl_expr): Do it here.
1706 (gimplify_target_expr): Likewise.
1708 2018-03-28 Martin Liska <mliska@suse.cz>
1711 * config/i386/i386.c (ix86_function_arg_advance): Do not call
1712 chkp_type_bounds_count if MPX is not enabled.
1714 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
1716 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
1718 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
1721 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
1722 function to create the function decl for complex long double
1723 multiply and divide for -mabi=ieeelongdouble.
1724 (init_float128_ieee): Call it.
1726 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
1729 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
1730 -fcf-protection=branch -mibt.
1731 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
1733 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1736 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
1738 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
1741 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
1742 extern array declarations.
1744 2018-03-27 Richard Biener <rguenther@suse.de>
1747 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
1748 explicit single_use checks.
1750 2018-03-27 Richard Biener <rguenther@suse.de>
1752 PR tree-optimization/85082
1753 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
1756 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1758 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
1759 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
1760 Turn on fasynchronous-unwind-tables and funwind-tables.
1762 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
1765 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
1766 (*bmi_blsr_<mode>_ccz): Ditto.
1768 2018-03-26 Tom de Vries <tom@codesourcery.com>
1770 PR tree-optimization/85063
1771 * omp-general.c (offloading_function_p): New function. Factor out
1773 * omp-offload.c (pass_omp_target_link::gate): ... here.
1774 * omp-general.h (offloading_function_p): Declare.
1775 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
1776 with attribute omp declare target for offloading functions.
1778 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
1780 PR tree-optimization/84005
1781 * tree-data-ref.h (get_base_for_alignment): Declare.
1782 * tree-data-ref.c (get_base_for_alignment_1): New function.
1783 (get_base_for_alignment): Likewise.
1784 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
1785 get_base_for_alignment to find a suitable base object, instead
1786 of always using drb->base_address.
1788 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1791 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
1792 known size by default.
1794 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
1797 * lra-constraints.c (process_alt_operands): Don't match BLKmode
1798 and non BLKmode operands.
1800 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1803 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
1804 Clean up attributes.
1806 2018-03-23 Richard Biener <rguenther@suse.de>
1809 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
1810 we are going to emit early debug for LTO.
1812 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1815 * function.c (match_asm_constraints_1): Don't optimize if input
1816 doesn't satisfy general_operand predicate for output's mode.
1819 * alias.c (write_dependence_p): Don't require for x_canonicalized
1820 non-VOIDmode if x has VOIDmode.
1823 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
1824 just don't try to optimize it rather than assert it never happens.
1826 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1828 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
1829 macro expansions for definition of ST_INTERNAL_<mode> and
1830 LD_INTERNAL_<mode> builtins.
1831 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
1833 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
1835 (altivec_expand_st_builtin): Likewise.
1836 (altivec_expand_builtin): Remove calls to deleted functions.
1837 (rs6000_address_for_altivec): Delete this function.
1838 * config/rs6000/vector.md: Remove expands for
1839 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
1841 2018-03-22 Sudakshina Das <sudi.das@arm.com>
1844 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
1845 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
1846 re-computing once computed.
1847 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
1848 (arm_init_machine_status): Initialize
1849 machine->static_chain_stack_bytes.
1851 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1854 * doc/extend.texi: Add four new prototypes for vec_ld.
1855 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
1856 definitions for more logical presentation.
1857 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
1858 entries for V1TI variants of __builtin_altivec_ld builtin.
1859 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
1860 handling of V1TI variant of LVX icode pattern.
1861 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
1862 (rs6000_gimple_fold_builtin): Likewise.
1863 (altivec_init_builtins): Add code to define
1864 __builtin_altivec_lvx_v1ti function.
1866 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1869 * function.c (match_asm_constraints_1): Don't do the optimization
1870 if input isn't a REG, SUBREG, MEM or constant.
1872 2018-03-22 Tom de Vries <tom@codesourcery.com>
1874 PR tree-optimization/84956
1875 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
1876 bb_has_abnormal_pred.
1878 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1881 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
1882 DECL_INITIAL (decl) to decl at the end.
1883 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
1886 2018-03-21 Joseph Myers <joseph@codesourcery.com>
1888 * doc/extend.texi (__builtin_tgmath): Document when complex
1889 integer types are treated as _Complex _Float64.
1891 2018-03-21 Tom de Vries <tom@codesourcery.com>
1893 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
1895 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1897 PR tree-optimization/84960
1898 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
1899 if it is ENTRY block, move them into single succ of ENTRY in that case.
1901 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
1903 PR tree-optimization/84811
1904 * poly-int.h (poly_span_traits): Remove the T3 parameter and
1905 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
1906 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
1907 (known_subrange_p): Update accordingly. Cast each value involved
1908 in the size comparison, rather than casting the result of the
1911 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1913 PR tree-optimization/84982
1914 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
1915 by flipping the least significant bit rather than all bits from
1916 bitpos to bitpos + bitsize - 1.
1918 2018-03-21 Nathan Sidwell <nathan@acm.org>
1920 * doc/extend.texi (Deprecated Features): Remove mention of
1921 long-deleted deprecations.
1923 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1926 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
1927 * configure: Regenerate.
1929 2018-03-21 Tom de Vries <tom@codesourcery.com>
1931 PR tree-optimization/83126
1932 * tree-parloops.c (num_phis): New function.
1933 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
1935 2018-03-21 Nathan Sidwell <nathan@acm.org>
1937 * doc/extend.texi (Deprecated Features): Update deprecated flags,
1938 mention anon-struct/union members and trailing attributes.
1940 2018-03-21 Bin Cheng <bin.cheng@arm.com>
1942 PR tree-optimization/84969
1943 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
1944 builtin memset partitions if they set different rhs values.
1946 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1948 PR rtl-optimization/84989
1949 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
1950 VEC_DUPLICATE with scalar result mode.
1952 2018-03-21 Martin Liska <mliska@suse.cz>
1955 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
1956 not intended return statement.
1958 2018-03-21 Martin Liska <mliska@suse.cz>
1961 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
1962 (chkp_find_bound_slots_1): Limit number of iterations.
1964 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
1967 * Minor grammar fixes for x86 options.
1969 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1972 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
1973 holding REG_CFA_RESTORE notes, instead turn them into a USE.
1975 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
1978 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
1979 (altivec_lvx_<mode>_1op): Likewise.
1980 (altivec_stvx_<mode>_2op): Likewise.
1981 (altivec_stvx_<mode>_1op): Likewise.
1982 (altivec_lvx_<VM2:mode>): New define_expand.
1983 (altivec_stvx_<VM2:mode>): Likewise.
1984 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
1985 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1986 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
1987 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1988 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
1989 (rs6000_gen_lvx): Likewise.
1990 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
1991 (altivec_expand_stv_builtin): Likewise.
1992 (altivec_expand_builtin): Likewise.
1993 * config/rs6000/vector.md: Likewise.
1995 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1998 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2001 2018-03-20 Richard Biener <rguenther@suse.de>
2004 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2005 sign-conversions as zero, fall back to standard scalar_stmt
2008 2018-03-20 Martin Liska <mliska@suse.cz>
2011 * predict.c (rebuild_frequencies): Handle case when we have
2012 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2014 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2017 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2018 flag_section_anchors.
2019 * varasm.c (use_blocks_for_decl_p): Remove hack for
2020 dw2_force_const_mem.
2023 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2025 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2026 be created, use lowpart_subreg of operands[0] rather than operands[0]
2028 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2029 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2030 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2031 and n constraint instead of aarch64_shift_imm_di and Usd.
2032 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2033 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2035 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2038 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2039 to favor GPR over NEON registers.
2040 (<shift>di3_neon): Likewise.
2042 2018-03-20 Tom de Vries <tom@codesourcery.com>
2045 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2046 (nvptx_process_pars): Emit bar.sync asap and alap.
2048 2018-03-20 Tom de Vries <tom@codesourcery.com>
2051 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2052 seen_label if seen_label is already set.
2054 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2057 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2058 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2059 Use 1U instead of 1. Formatting fixes.
2062 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2063 instead of TREE_TYPE (s1) for the return value.
2065 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2067 PR tree-optimization/84946
2068 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2069 bitsize + bitsize in poly_uint64 rather than poly_int64.
2072 * dwarf2asm.c: Include fold-const.c.
2073 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2074 of decl rather than decl itself.
2076 PR rtl-optimization/84643
2077 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2079 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2082 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2083 calling assemble_variable.
2085 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2088 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2089 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2091 2018-03-19 Jim Wilson <jimw@sifive.com>
2094 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2095 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2096 (riscv_first_stack_step): Likewise.
2097 (riscv_option_override): Use STACK_BOUNDARY instead of
2099 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2101 (BIGGEST_ALIGNMENT): Set to 128.
2102 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2103 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2106 2018-03-19 Richard Biener <rguenther@suse.de>
2108 PR tree-optimization/84933
2109 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2110 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2112 2018-03-19 Richard Biener <rguenther@suse.de>
2114 PR tree-optimization/84859
2115 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2116 (cond_if_else_store_replacement): Perform sinking operation on
2117 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2118 Generalize what a BB with a single eligible store is.
2120 2018-03-19 Richard Biener <rguenther@suse.de>
2122 PR tree-optimization/84929
2123 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2124 chrec_is_positive against non-chrec arg.
2126 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2129 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2131 2018-03-18 Martin Liska <mliska@suse.cz>
2133 PR rtl-optimization/84635
2134 * regrename.c (build_def_use): Use matches_mode only when
2137 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2139 PR tree-optimization/84913
2140 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2141 vectorize chains of COND_EXPRs.
2143 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2145 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2147 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2149 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2151 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2153 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2155 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2156 Kito Cheng <kito.cheng@gmail.com>
2158 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2159 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2160 (nds32_adjust_reg_alloc_order): New function.
2161 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
2163 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
2165 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
2166 nds32_print_operand, nds32_print_operand_address): Use
2167 HOST_WIDE_INT_PRINT_DEC instead.
2169 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2171 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
2173 2018-03-17 Jakub Jelinek <jakub@redhat.com>
2176 * config/i386/i386.c (initial_ix86_tune_features,
2177 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
2179 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
2180 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
2181 rather than 1u << ix86_tune. Formatting fix.
2182 (ix86_option_override_internal): Change ix86_arch_mask from
2183 unsigned int to unsigned HOST_WIDE_INT, initialize to
2184 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
2185 (ix86_function_specific_restore): Likewise.
2187 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2190 * postreload.c (reload_combine_recognize_pattern): Perform
2191 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
2192 truncate_int_for_mode the result for the destination's mode.
2195 * hsa-gen.c (mem_type_for_type): Fix comment typo.
2196 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
2198 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2201 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2204 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2205 regno_allocno_class_array and sorted_pseudos.
2206 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2207 insns where regno is used.
2209 2018-03-16 Martin Liska <mliska@suse.cz>
2212 * multiple_target.c (create_dispatcher_calls): Redirect
2213 reference in the symbol table.
2215 2018-03-16 Martin Liska <mliska@suse.cz>
2218 * multiple_target.c (create_dispatcher_calls): Redirect also
2221 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2225 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2226 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2227 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2229 2018-03-16 Julia Koval <julia.koval@intel.com>
2231 * doc/invoke.texi (Skylake Server): Add CLWB.
2232 Cannonlake): Remove CLWB.
2234 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2236 PR tree-optimization/84841
2237 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2239 (FLOAT_ONE_CONST_TYPE): Define.
2240 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2241 (sort_by_operand_rank): Put entries with higher constant_type last
2242 rather than first to match comments.
2244 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
2246 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2249 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2252 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2253 check and use address of target always.
2255 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2258 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2259 (indirect_thunk_bnd_needed): Likewise.
2260 (indirect_thunks_used): Likewise.
2261 (indirect_thunks_bnd_used): Likewise.
2262 (indirect_return_needed): New.
2263 (indirect_return_bnd_needed): Likewise.
2264 (output_indirect_thunk_function): Add a bool argument for
2266 (output_indirect_thunk_function): Don't generate alias for
2267 function return thunk.
2268 (ix86_code_end): Call output_indirect_thunk_function to generate
2269 function return thunks.
2270 (ix86_output_function_return): Set indirect_return_bnd_needed
2271 and indirect_return_needed instead of indirect_thunk_bnd_needed
2272 and indirect_thunk_needed.
2274 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2276 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2277 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2278 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2280 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2281 Paul Hua <paul.hua.gm@gmail.com>
2284 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2286 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2288 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2289 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2292 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2295 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2296 instead of GET_MODE_SIZE when comparing Units.
2298 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2301 * varasm.c (hash_section): Return an unchangeble hash value
2302 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2303 Return !aarch64_can_use_per_function_literal_pools_p ().
2305 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2308 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2309 rather than address of cmode as last argument to prepare_cmp_insn.
2311 2018-03-15 Julia Koval <julia.koval@intel.com>
2313 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2314 F_AVX512VNNI, F_AVX512BITALG): New.
2316 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2319 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2320 insn for floating-point loads and stores.
2322 2018-03-14 Carl Love <cel@us.ibm.com>
2324 * config/rs6000/rs6000-c.c: Add macro definitions for
2325 ALTIVEC_BUILTIN_VEC_PERMXOR.
2326 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2327 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2328 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2329 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2331 * config/doc/extend.texi: Add prototypes for vec_permxor.
2333 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2336 * diagnostic-show-locus.c (class layout_point): Convert m_line
2337 from int to linenum_type.
2338 (line_span::comparator): Use linenum "compare" function when
2339 comparing line numbers.
2340 (test_line_span): New function.
2341 (layout_range::contains_point): Convert param "row" from int to
2343 (layout_range::intersects_line_p): Likewise.
2344 (layout::will_show_line_p): Likewise.
2345 (layout::print_source_line): Likewise.
2346 (layout::should_print_annotation_line_p): Likewise.
2347 (layout::print_annotation_line): Likewise.
2348 (layout::print_leading_fixits): Likewise.
2349 (layout::annotation_line_showed_range_p): Likewise.
2350 (struct line_corrections): Likewise for field m_row.
2351 (line_corrections::line_corrections): Likewise for param "row".
2352 (layout::print_trailing_fixits): Likewise.
2353 (layout::get_state_at_point): Likewise.
2354 (layout::get_x_bound_for_row): Likewise.
2355 (layout::print_line): Likewise.
2356 (diagnostic_show_locus): Likewise for locals "last_line" and
2358 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2359 * input.c (selftest::test_linenum_comparisons): New function.
2360 (selftest::input_c_tests): Call it.
2361 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2362 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2363 * selftest.h (ASSERT_GT): New macro.
2364 (ASSERT_GT_AT): New macro.
2365 (ASSERT_LT): New macro.
2366 (ASSERT_LT_AT): New macro.
2368 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
2370 PR rtl-optimization/84780
2371 * combine.c (distribute_links): Don't make a link based on pc_rtx.
2373 2018-03-14 Martin Liska <mliska@suse.cz>
2375 * tree.c (record_node_allocation_statistics): Use
2376 get_stats_node_kind.
2377 (get_stats_node_kind): New function extracted from
2378 record_node_allocation_statistics.
2379 (free_node): Use get_stats_node_kind.
2381 2018-03-14 Richard Biener <rguenther@suse.de>
2383 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2384 that the value-set of ANTIC_IN doesn't grow.
2387 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2389 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2390 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2392 2018-03-14 Julia Koval <julia.koval@intel.com>
2394 * config.gcc (icelake-client, icelake-server): New.
2396 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2397 (initial_ix86_arch_features): Ditto.
2398 (PTA_SKYLAKE): Add SGX.
2399 (PTA_ICELAKE): Remove.
2400 (PTA_ICELAKE_CLIENT): New.
2401 (PTA_ICELAKE_SERVER): New.
2402 (ix86_option_override_internal): Split up icelake on icelake client and
2404 (get_builtin_code_for_version): Ditto.
2405 (fold_builtin_cpu): Ditto.
2406 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2407 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2408 * config/i386/i386.h (processor_type): Ditto.
2409 * doc/invoke.texi: Ditto.
2411 2018-03-14 Jakub Jelinek <jakub@redhat.com>
2414 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2415 INTEGER_CST offset, add it together with bitpos / 8 and
2416 sign extend based on POINTER_SIZE.
2420 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
2423 * config/i386/constraints.md (Yc): New register constraint.
2424 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
2425 Use Yc constraint for alternative 2 of operand 0. Remove
2426 preferred_for_speed attribute.
2428 2018-03-14 Richard Biener <rguenther@suse.de>
2430 PR tree-optimization/84830
2431 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
2432 with the old one to avoid oscillations.
2434 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
2437 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
2439 (assign_by_spills): Return a flag of reload assignment failure.
2440 Do not process the reload assignment failures. Do not spill other
2441 reload pseudos if they has the same reg class. Update n if
2443 (lra_assign): Add a return arg. Set up from the result of
2444 assign_by_spills call.
2445 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2446 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2447 usage_insns if it is not NULL.
2448 (spill_hard_reg_in_range): New function.
2449 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2450 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2451 function prototypes.
2452 (lra_assign): Change prototype.
2453 * lra.c (lra): Add code to deal with fails by splitting hard reg
2456 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
2458 * config/riscv/riscv.opt (mrelax): New option.
2459 * config/riscv/riscv.c (riscv_file_start): Emit ".option
2460 "norelax" when riscv_mrelax is disabled.
2461 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2463 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2466 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2467 reassociation for int modes.
2469 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2471 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
2472 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
2474 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
2475 * config/aarch64/aarch64-sve.md
2476 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
2477 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2478 (*extend<mode><Vwide>2): Rename to...
2479 (aarch64_sve_extend<mode><Vwide>2): ...this.
2480 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2481 renaming the old pattern to...
2482 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
2484 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2485 define_expand, renaming the old pattern to...
2486 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2487 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2488 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2489 account when deciding which SVE instruction the optab should use.
2490 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2492 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2494 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2495 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2496 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2497 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2498 (tlsdesc_small_<mode>): Turn a define_expand and use
2499 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
2500 (tlsdesc_small_advsimd_<mode>): ...this.
2501 (tlsdesc_small_sve_<mode>): New pattern.
2503 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2505 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2506 (UNSPEC_UMUL_HIGHPART): New constants.
2507 (MUL_HIGHPART): New int iteraor.
2508 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2509 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2511 (*<su>mul<mode>3_highpart): New define_insn.
2513 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2516 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2519 2018-03-13 Martin Liska <mliska@suse.cz>
2522 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2524 (sem_item_optimizer::~sem_item_optimizer): Release it.
2525 (sem_item_optimizer::merge_classes): Register variable aliases.
2526 (sem_item_optimizer::fixup_pt_set): New function.
2527 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2528 * ipa-icf.h: Declare new variables and functions.
2530 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2533 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2534 integer_pow2p@2 and test integer_pow2p in condition.
2535 (A < 0 ? C : 0): Similarly for @1.
2538 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2539 characters starting at p contain '\0' character, don't look beyond
2543 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2544 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2547 * reg-stack.c (change_stack): Change update_end var from int to
2548 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2549 also call set_block_for_insn on the newly added insns and rescan.
2552 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2553 on the last operand.
2556 * tree.c (stabilize_reference_1): Return save_expr (e) for
2557 STATEMENT_LIST even if it doesn't have side-effects.
2559 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
2561 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2563 2018-03-12 Renlin Li <renlin.li@arm.com>
2565 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2566 aarch64_output_scalar_simd_mov_immediate.
2568 2018-03-12 Martin Sebor <msebor@redhat.com>
2570 PR tree-optimization/83456
2571 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2572 for perfectly overlapping calls to memcpy.
2573 (gimple_fold_builtin_memory_chk): Same.
2574 (gimple_fold_builtin_strcpy): Handle no-warning.
2575 (gimple_fold_builtin_stxcpy_chk): Same.
2576 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2578 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2580 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2581 parameter. Use it for SFmode.
2582 (rs6000_function_arg_advance_1): Adjust.
2583 (rs6000_function_arg): Adjust.
2584 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2586 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2588 PR rtl-optimization/84169
2589 PR rtl-optimization/84780
2590 * combine.c (can_combine_p): Check for a 2-insn combination whether
2591 the destination register is used between the two insns, too.
2593 2018-03-12 Richard Biener <rguenther@suse.de>
2595 PR tree-optimization/84803
2596 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2597 for refs DR analysis didn't process.
2599 2018-03-12 Richard Biener <rguenther@suse.de>
2601 PR tree-optimization/84777
2602 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2603 force-vectorize loops ignore whether we are optimizing for size.
2605 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
2607 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2608 (TARGET_MD_ASM_ADJUST): Define.
2610 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
2611 Kito Cheng <kito.cheng@gmail.com>
2612 Chung-Ju Wu <jasonwucj@gmail.com>
2614 * config/nds32/nds32.c (nds32_compute_stack_frame,
2615 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2616 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2617 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2618 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2619 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2620 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2621 * config/nds32/nds32.md (prologue, epilogue): Use macro
2622 NDS32_V3PUSH_AVAILABLE_P to do checking.
2624 2018-03-11 Jakub Jelinek <jakub@redhat.com>
2627 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
2628 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
2629 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
2630 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
2631 addition of most attributes on !orig_type_die or the attribute not
2632 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
2634 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2635 Chung-Ju Wu <jasonwucj@gmail.com>
2637 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
2639 * config/nds32/nds32.opt (mvh): New option.
2641 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2642 Chung-Ju Wu <jasonwucj@gmail.com>
2644 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
2646 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
2647 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
2650 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2651 Chung-Ju Wu <jasonwucj@gmail.com>
2653 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
2655 * config/nds32/nds32-multiple.md (strlensi): New pattern.
2656 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
2658 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
2659 Kito Cheng <kito.cheng@gmail.com>
2660 Chung-Ju Wu <jasonwucj@gmail.com>
2662 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
2663 UNSPEC_FFMISM and UNSPEC_FLMISM.
2664 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
2665 for ffb, ffmism and flmism.
2666 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
2667 (unspec_ffmism): Ditto.
2668 (unspec_flmism): Ditto.
2669 (nds32_expand_builtin_impl): Check if string extension is available.
2670 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
2671 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
2673 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
2676 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2679 * lra-assigns.c (assign_by_spills): Return a flag of reload
2680 assignment failure. Do not process the reload assignment
2681 failures. Do not spill other reload pseudos if they has the same
2683 (lra_assign): Add a return arg. Set up from the result of
2684 assign_by_spills call.
2685 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2686 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2687 usage_insns if it is not NULL.
2688 (spill_hard_reg_in_range): New function.
2689 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2690 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2691 function prototypes.
2692 (lra_assign): Change prototype.
2693 * lra.c (lra): Add code to deal with fails by splitting hard reg
2696 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
2699 * config/i386/i386.opt: Replace Enforcment with Enforcement.
2701 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
2704 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
2705 (dw_val_node): Add val_symbolic_view.
2706 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
2707 (symview_upper_bound): New.
2708 (new_line_info_table): Initialize symviews_since_reset.
2709 (dwarf2out_source_line): Count symviews_since_reset and set
2710 symview_upper_bound.
2711 (dw_val_equal_p): Handle symview.
2712 (add_AT_symview): New.
2713 (print_dw_val): Handle symview.
2714 (attr_checksum, attr_checksum_ordered): Likewise.
2715 (same_dw_val_p, size_of_die): Likewise.
2716 (value_format, output_die): Likewise.
2717 (add_high_low_attributes): Use add_AT_symview for entry_view.
2718 (dwarf2out_finish): Reset symview_upper_bound, clear
2721 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
2724 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
2725 Add strict argument and use it.
2726 (rs6000_split_multireg_move): Update for new strict argument.
2727 (mem_operand_gpr): Disallow all non-offsettable addresses.
2728 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
2730 2018-03-09 Jakub Jelinek <jakub@redhat.com>
2733 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
2734 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
2735 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
2738 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
2739 decl, use remap_type if we want to use the type.
2741 2018-03-09 Martin Sebor <msebor@redhat.com>
2743 PR tree-optimization/84526
2744 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2746 (builtin_access::generic_overlap): Be prepared to handle non-array
2749 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
2751 PR rtl-optimization/84682
2752 * lra-constraints.c (process_address_1): Check is_address flag
2753 for address constraints.
2754 (process_alt_operands): Likewise.
2755 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
2756 preprocess_constraints.
2757 * recog.h (preprocess_constraints): Add oploc parameter.
2759 * recog.c (preprocess_constraints): Test address_operand for
2760 CT_ADDRESS constraints.
2762 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2765 * lra-assigns.c (assign_by_spills): Return a flag of reload
2766 assignment failure. Do not process the reload assignment
2767 failures. Do not spill other reload pseudos if they has the same
2769 (lra_assign): Add a return arg. Set up from the result of
2770 assign_by_spills call.
2771 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2772 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2773 usage_insns if it is not NULL.
2774 (spill_hard_reg_in_range): New function.
2775 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2776 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2777 function prototypes.
2778 (lra_assign): Change prototype.
2779 * lra.c (lra): Add code to deal with fails by splitting hard reg
2782 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2785 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
2786 Accept complain bool parameter. Only emit errors if it is true.
2787 (arm_parse_cpu_option_name): Likewise.
2788 (arm_target_thumb_only): Adjust callers of the above.
2789 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
2790 prototype to take a default true bool parameter.
2791 (arm_parse_arch_option_name): Likewise.
2793 2018-03-09 David Malcolm <dmalcolm@redhat.com>
2794 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2798 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
2799 * configure: Regenerate.
2800 * configure.ac ("linker --version-script option"): New.
2801 ("linker soname option"): New.
2803 2018-03-09 Richard Biener <rguenther@suse.de>
2805 PR tree-optimization/84775
2806 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
2807 immediate uses of predicate stmts and mark them modified.
2810 PR tree-optimization/84178
2811 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2813 (version_loop_for_if_conversion): Delay update_ssa call.
2814 (tree_if_conversion): Delay update_ssa until after predicate
2817 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
2820 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
2821 when the function accesses prior frames.
2823 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2826 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
2827 gen_llsym, otherwise call maybe_gen_llsym.
2830 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
2831 has ',' character inside of it.
2833 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2836 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
2837 as clobbering CC_REGNUM.
2839 2018-03-08 Richard Biener <rguenther@suse.de>
2842 * tree-scalar-evolution.c: Include tree-into-ssa.h.
2843 (follow_copies_to_constant): Do not follow SSA names registered
2846 2018-03-08 Richard Biener <rguenther@suse.de>
2848 PR tree-optimization/84178
2849 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2851 (version_loop_for_if_conversion): Delay update_ssa call.
2852 (tree_if_conversion): Delay update_ssa until after predicate
2855 2018-03-08 David Malcolm <dmalcolm@redhat.com>
2857 PR tree-optimization/84178
2858 * tree-if-conv.c (release_bb_predicate): Remove the
2859 the assertion that the stmts have NULL use_ops.
2860 Discard the statements, asserting that they haven't
2861 yet been added to a BB.
2863 2018-03-08 Richard Biener <rguenther@suse.de>
2865 PR tree-optimization/84746
2866 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
2867 (phi_translate): Pass in destination ANTIC_OUT set.
2868 (phi_translate_1): Likewise. For a simplified result lookup
2869 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
2870 (phi_translate_set): Adjust.
2871 (do_pre_regular_insertion): Likewise.
2872 (do_pre_partial_partial_insertion): Likewise.
2874 2018-03-08 Martin Liska <mliska@suse.cz>
2876 PR gcov-profile/84735
2877 * doc/gcov.texi: Document usage of profile files.
2878 * gcov-io.h: Document changes in the format.
2880 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
2884 * dwarf2out.c (struct dw_line_info_table): Update comments for
2886 (FORCE_RESET_NEXT_VIEW): New.
2887 (FORCE_RESETTING_VIEW_P): New.
2888 (RESETTING_VIEW_P): Check for -1 too.
2889 (ZERO_VIEW_P): Likewise.
2890 (new_line_info_table): Force-reset next view.
2891 (dwarf2out_begin_function): Likewise.
2892 (dwarf2out_source_line): Simplify zero_view_p initialization.
2893 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
2894 view directly. Omit view when omitting .loc at line 0.
2896 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2898 PR tree-optimization/84740
2899 * tree-switch-conversion.c (process_switch): Call build_constructors
2900 only if info.phi_count is non-zero.
2902 PR tree-optimization/84739
2903 * tree-tailcall.c (find_tail_calls): Check call arguments against
2904 DECL_ARGUMENTS (current_function_decl) rather than
2905 DECL_ARGUMENTS (func) when checking for tail recursion.
2907 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2909 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
2910 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
2911 Volker Reichelt's entry and add entries for people that perform
2912 GCC fuzzy testing and report numerous bugs.
2914 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
2917 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
2918 readonly data in sdata, if that is disabled.
2919 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
2920 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2921 -mreadonly-in-sdata option.
2923 2018-03-07 Martin Sebor <msebor@redhat.com>
2925 PR tree-optimization/84468
2926 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
2927 basic block when looking for nul assignment.
2929 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
2932 * except.h (output_function_exception_table): Adjust prototype.
2933 * except.c (output_function_exception_table): Remove FNNAME parameter
2934 and add SECTION parameter. Ouput one part of the table at a time.
2935 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
2936 the first part of the exception table and emit unwind directives.
2937 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
2938 (i386_pe_seh_cold_init): Likewise.
2939 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
2940 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
2941 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
2942 (ix86_output_call_insn): Emit a nop in one more case for SEH.
2943 * config/i386/winnt.c: Include except.h.
2944 (struct seh_frame_state): Add reg_offset, after_prologue and
2945 in_cold_section fields.
2946 (i386_pe_seh_end_prologue): Set seh->after_prologue.
2947 (i386_pe_seh_cold_init): New function.
2948 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
2949 to seh->in_cold_section.
2950 (seh_emit_push): Record the offset of the push.
2951 (seh_emit_save): Record the offet of the save.
2952 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
2953 Test seh->after_prologue to disregard the epilogue.
2954 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
2955 (i386_pe_end_cold_function): New function.
2957 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2960 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
2961 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
2964 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
2966 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
2967 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
2970 * multiple_target.c: Include tree-inline.h and intl.h.
2971 (expand_target_clones): Diagnose and fail if node->definition and
2972 !tree_versionable_function_p (node->decl).
2974 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
2976 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
2978 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
2979 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
2980 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
2982 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2985 * combine.c (try_combine): Use reg_or_subregno instead of handling
2986 just paradoxical SUBREGs and REGs.
2988 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
2990 * config/arc/arc.c (arc_finalize_pic): Remove function.
2991 (arc_must_save_register): We use single base PIC register, remove
2992 checks to save/restore the PIC register.
2993 (arc_expand_prologue): Likewise.
2994 * config/arc/arc-protos.h (arc_set_default_type_attributes):
2996 (arc_verify_short): Likewise.
2997 (arc_attr_type): Likewise.
2998 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
2999 (walk_stores): Likewise.
3000 (arc_address_cost): Make it static.
3001 (arc_verify_short): Likewise.
3002 (branch_dest): Likewise.
3003 (arc_attr_type): Likewise.
3004 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3005 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3006 (arc_final_prescan_insn): Remove inserting the nops due to
3007 hardware hazards. It is done in reorg step.
3008 (insn_length_variant_t): Remove.
3009 (insn_length_parameters_t): Likewise.
3010 (arc_insn_length_parameters): Likewise.
3011 (arc_get_insn_variants): Likewise.
3012 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3014 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3017 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3020 PR tree-optimization/84687
3021 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3023 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3025 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3027 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3028 Rename to ppc_speculation_barrier.
3029 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3030 __builtin_ppc_speculation_barrier.
3032 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3035 * combine.c (combine_simplify_rtx): Don't try to simplify if
3036 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3039 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3041 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3042 to 32 bytes when compiling for POWER9.
3044 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3047 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3048 regparm >= 3 with no arg reg available also for calls with
3049 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3052 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3054 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3056 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3059 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3061 2018-03-05 Richard Biener <rguenther@suse.de>
3063 PR tree-optimization/84486
3064 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3065 When inserting a __builtin_assume_aligned call set the LHS
3066 SSA name alignment info accordingly.
3068 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3070 PR tree-optimization/84114
3071 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3072 Avoid reassociation of FLOAT_MODE addition.
3074 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3076 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3077 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3078 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3079 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3080 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3081 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3082 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3084 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3085 __builtin_ia32_wbinvd): New builtins.
3086 (SPECIAL_ARGS2): New.
3087 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3088 (SPECIAL_ARGS2): New.
3089 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3090 (ix86_valid_target_attribute_inner_p): Ditto.
3091 (ix86_init_mmx_sse_builtins): Add special_args2.
3092 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3093 TARGET_WBNOINVD_P): New.
3094 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3095 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3096 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3097 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3098 * config/i386/pconfigintrin.h: New file.
3099 * config/i386/wbnoinvdintrin.h: Ditto.
3100 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3101 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3103 2018-03-05 Richard Biener <rguenther@suse.de>
3105 PR tree-optimization/84670
3106 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3108 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3109 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3110 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3111 doesn't grow if all successors (recursively) were visited at least
3114 2018-03-05 Richard Biener <rguenther@suse.de>
3116 PR tree-optimization/84650
3117 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3118 if executed in the loop pipeline.
3120 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3122 * doc/configfiles.texi (Configuration Files): Move info about
3123 conditionalizing $target-protos.h to...
3124 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3125 differs from $target-protos.h.
3127 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3128 Chung-Ju Wu <jasonwucj@gmail.com>
3130 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3131 * config/nds32/nds32-multiple.md (setmemsi): Define.
3132 * config/nds32/nds32-memory-manipulation.c
3133 (nds32_gen_dup_4_byte_to_word_value): New.
3134 (emit_setmem_word_loop): New.
3135 (emit_setmem_byte_loop): New.
3136 (nds32_expand_setmem_loop): New.
3137 (nds32_expand_setmem_loop_v3m): New.
3138 (nds32_expand_setmem_unroll): New.
3139 (nds32_expand_setmem): New.
3141 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3142 Chung-Ju Wu <jasonwucj@gmail.com>
3144 * config/nds32/nds32-memory-manipulation.c
3145 (nds32_emit_load_store): New.
3146 (nds32_emit_post_inc_load_store): New.
3147 (nds32_emit_mem_move): New.
3148 (nds32_emit_mem_move_block): New.
3149 (nds32_expand_movmemsi_loop_unknown_size): New.
3150 (nds32_expand_movmemsi_loop_known_size): New.
3151 (nds32_expand_movmemsi_loop): New.
3152 (nds32_expand_movmemsi_unroll): New.
3153 (nds32_expand_movmemqi): Rename ...
3154 (nds32_expand_movmemsi): ... to this.
3155 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3156 (movmemsi): ... to this.
3157 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3158 (nds32_expand_movmemsi): ... to this.
3160 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3161 Monk Chiang <sh.chiang04@gmail.com>
3162 Chung-Ju Wu <jasonwucj@gmail.com>
3164 * config/nds32/nds32-protos.h
3165 (nds32_expand_load_multiple): New arguments.
3166 (nds32_expand_store_multiple): Ditto.
3167 (nds32_valid_multiple_load_store): Rename ...
3168 (nds32_valid_multiple_load_store_p): ... to this.
3169 * config/nds32/nds32-memory-manipulation.c
3170 (nds32_expand_load_multiple): Refine implementation.
3171 (nds32_expand_store_multiple): Ditto.
3172 * config/nds32/nds32-multiple.md
3173 (load_multiple): Update nds32_expand_load_multiple interface.
3174 (store_multiple): Update nds32_expand_store_multiple interface.
3175 * config/nds32/nds32-predicates.c
3176 (nds32_valid_multiple_load_store): Rename ...
3177 (nds32_valid_multiple_load_store_p): ... to this and refine
3179 * config/nds32/predicates.md
3180 (nds32_load_multiple_and_update_address_operation): New predicate.
3181 (nds32_store_multiple_and_update_address_operation): New predicate.
3183 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3184 Chung-Ju Wu <jasonwucj@gmail.com>
3186 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
3187 (combo): New attribute.
3188 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
3190 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3192 * config/nds32/nds32.opt: Change -mcmodel= default value.
3194 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
3195 Monk Chiang <sh.chiang04@gmail.com>
3196 Chung-Ju Wu <jasonwucj@gmail.com>
3198 * config/nds32/constants.md (unspec_element): New enum.
3199 * config/nds32/constraints.md (Umw): New constraint.
3200 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3201 * config/nds32/nds32-intrinsic.md: Likewise.
3202 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3203 (nds32_valid_smw_lwm_base_p): New.
3204 (nds32_output_smw_single_word): New.
3205 (nds32_output_lmw_single_word): New.
3206 (nds32_expand_unaligned_load): New.
3207 (nds32_expand_unaligned_store): New.
3208 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3209 (nds32_output_smw_single_word): Declare.
3210 (nds32_output_lmw_single_word): Declare.
3211 (nds32_expand_unaligned_load): Declare.
3212 (nds32_expand_unaligned_store): Declare.
3213 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3214 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3215 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3216 NDS32_BUILTIN_UASTORE_DW.
3217 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3220 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
3221 Kito Cheng <kito.cheng@gmail.com>
3222 Chung-Ju Wu <jasonwucj@gmail.com>
3224 * config/nds32/nds32-intrinsic.c
3225 (nds32_expand_builtin_null_ftype_reg): Delete.
3226 (nds32_expand_builtin_reg_ftype_imm): Ditto.
3227 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3228 (nds32_read_argument): New.
3229 (nds32_legitimize_target): Ditto.
3230 (nds32_legitimize_argument): Ditto.
3231 (nds32_check_constant_argument): Ditto.
3232 (nds32_expand_unop_builtin): Ditto.
3233 (nds32_expand_unopimm_builtin): Ditto.
3234 (nds32_expand_binop_builtin): Ditto.
3235 (nds32_builtin_decl_impl): Ditto.
3236 (builtin_description): Ditto.
3237 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3238 (nds32_init_builtins_impl): Ditto.
3239 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3240 (nds32_builtin_decl): New.
3241 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3242 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3244 2018-03-02 Jeff Law <law@redhat.com>
3246 * reorg.c (stop_search_p): Handle DEBUG_INSN.
3247 (redundant_insn, fill_simple_delay_slots): Likewise.
3248 (fill_slots_from_thread): Likewise.
3249 * resource.c (mark_referenced_resources): Likewise.
3250 (mark_set_resources, find_dead_or_set_registers): Likewise.
3252 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3254 * substring-locations.h (format_warning_va): Formatting fix for
3256 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3258 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3259 * substring-locations.c: Include intl.h.
3260 (format_warning_va): Turned into small wrapper around
3261 format_warning_n_va, renamed to ...
3262 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3263 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3265 (format_warning_at_substring_n): New function.
3266 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3267 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3268 format_warning_at_substring with just a shorter name instead of
3269 const function pointer.
3270 (fmtwarn_n): New function.
3271 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3272 appropriate, get rid of all the fmtstr temporaries, move conditionals
3273 with G_() wrapped string literals directly into fmtwarn arguments,
3274 cast dir.len to (int), formatting fixes.
3276 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3278 * doc/invoke.texi: Remove "Cilk Plus" references.
3280 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3281 Richard Biener <rguenther@suse.de>
3284 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3285 for error or warning attributes if CALL_FROM_THUNK_P is set.
3288 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3291 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3292 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3295 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3296 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3298 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3299 instead of -1U in last predictors element's probability member.
3301 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3304 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3305 arguments if they are comparable.
3307 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3309 PR tree-optimization/84634
3310 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3311 masks and masked_loop_p with a single loop_masks, making sure it's
3312 null for bb vectorization.
3314 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3316 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3317 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3318 loop->force_vectorize to check whether there is no alias.
3320 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3323 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3325 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3327 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3328 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3329 instead of a loop around prev_real_insn.
3330 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3334 * config/i386/i386.c (ix86_print_operand): Use conditional
3335 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3338 2018-03-02 Richard Biener <rguenther@suse.de>
3340 PR tree-optimization/84427
3341 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3342 (bitmap_set_subtract_values): Rewrite to handle multiple
3345 (prune_clobbered_mems): Likewise.
3346 (phi_translate): Take edge instead of pred/phiblock.
3347 (phi_translate_1): Likewise.
3348 (phi_translate_set): Likewise. Insert all translated
3349 exprs for a value into the set, keeping possibly multiple
3350 expressions per value.
3351 (compute_antic_aux): Adjust for phi_translate changes.
3352 When intersecting union the expressions and prune those
3353 not in the final value set, keeping possibly multiple
3354 expressions per value. Do not use value-insertion
3355 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3356 all expressions. Add verification that the value-sets
3357 only shrink during iteration.
3358 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3359 (do_pre_regular_insertion): Likewise.
3360 (do_pre_partial_partial_insertion): Likewise.
3362 2018-03-02 Richard Biener <rguenther@suse.de>
3365 * config/darwin.c (saved_debug_info_level): New static global.
3366 (darwin_asm_lto_start): Disable debug info generation for LTO out.
3367 (darwin_asm_lto_end): Restore debug info generation settings.
3369 2018-03-01 Martin Liska <mliska@suse.cz>
3372 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3375 2018-03-01 Richard Biener <rguenther@suse.de>
3378 * dwarf2out.c (gen_variable_die): Properly handle late VLA
3379 type annotation with LTO when debug was disabled at compile-time.
3381 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3383 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3385 (mips_final_postscan_insn): Likewise.
3387 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
3389 PR rtl-optimization/84528
3390 * alias.c (init_alias_target): Add commentary.
3391 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3392 a unique base value if the frame pointer is not eliminated
3393 to the stack pointer.
3395 2018-03-01 Tom de Vries <tom@codesourcery.com>
3397 PR rtl-optimization/83327
3398 * lra-int.h (hard_regs_spilled_into): Declare.
3399 * lra.c (hard_regs_spilled_into): Define.
3400 (init_reg_info): Init hard_regs_spilled_into.
3401 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3402 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3403 (process_bb_lives): Handle hard_regs_spilled_into.
3404 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3405 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3407 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3409 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3410 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3411 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3412 * config/rs6000/aix72.h: New file.
3414 2018-02-28 Jakub Jelinek <jakub@redhat.com>
3416 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3417 instead of warning_at with conditional singular and plural messages
3421 * stor-layout.c (update_alignment_for_field): For
3422 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
3423 && !DECL_PACKED (field), do the alignment update, just use
3424 only desired_align instead of MAX (type_align, desired_align)
3426 (place_field): Don't do known_align < desired_align handling
3427 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
3428 is non-NULL, instead do it after rli->prev_field handling and
3429 only if not within a bitfield word. For DECL_PACKED (field)
3430 use type_align of BITS_PER_UNIT.
3432 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
3434 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
3435 superfluous parentheses and trailing spaces.
3437 2018-02-28 Richard Biener <rguenther@suse.de>
3439 PR tree-optimization/84584
3440 * graphite-scop-detection.c (scop_detection::add_scop): Discard
3441 SCoPs with fake exit edge.
3443 2018-02-28 Martin Liska <mliska@suse.cz>
3446 * timevar.c (timer::print): Fix format to properly print 100%
3449 2018-02-28 Richard Biener <rguenther@suse.de>
3452 * genmatch.c (capture_info::walk_match): Do not mark
3453 captured expressions without operands as expr_p given
3454 they act more like predicates and should be subject to
3455 "lost tail" side-effect preserving.
3457 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
3459 PR rtl-optimization/81611
3460 * auto-inc-dec.c (attempt_change): Move dead note from
3461 mem_insn if it's the next use of regno
3462 (find_address): Take address use of reg holding
3463 non-incremented value. Add parm to limit search to the named
3465 (merge_in_block): Attempt to use a mem insn that is the next
3466 use of the original regno.
3468 2018-02-27 Martin Sebor <msebor@redhat.com>
3471 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
3472 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
3474 2018-02-27 Martin Sebor <msebor@redhat.com>
3476 PR translation/84207
3477 * diagnostic-core.h (warning_n, error_n, inform_n): Change
3478 n argument to unsigned HOST_WIDE_INT.
3479 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3480 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
3481 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3482 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3484 2018-02-27 Richard Biener <rguenther@suse.de>
3486 PR tree-optimization/84512
3487 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3488 Do not use the estimate returned from record_stmt_cost for
3489 the scalar iteration cost but sum properly using add_stmt_cost.
3491 2018-02-27 Richard Biener <rguenther@suse.de>
3493 PR tree-optimization/84466
3494 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3495 Adjust last change to less strictly validate use operands.
3497 2018-02-27 Martin Liska <mliska@suse.cz>
3499 PR gcov-profile/84548
3500 * gcov.c (process_file): Allow partial overlap and consider it
3501 also as group functions.
3502 (output_lines): Properly calculate range of lines for a group.
3504 2018-02-27 Martin Liska <mliska@suse.cz>
3506 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3507 'ggc' suffixes. Change first column width.
3508 (timer::print): Fix formatting of the column.
3510 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3512 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3513 preserve inline entry blocks for the sake of debug inline
3514 entry point markers alone.
3515 (remove_unused_locals): Suggest in comments a better place to
3516 force the preservation of inline entry blocks that are
3517 otherwise unused, but do not preserve them.
3519 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3521 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3523 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3526 * config/i386/constraints.md (Bs): Replace
3527 ix86_indirect_branch_register with
3528 TARGET_INDIRECT_BRANCH_REGISTER.
3530 * config/i386/i386.md (indirect_jump): Likewise.
3531 (tablejump): Likewise.
3532 (*sibcall_memory): Likewise.
3533 (*sibcall_value_memory): Likewise.
3534 Peepholes of indirect call and jump via memory: Likewise.
3535 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3536 (*sibcall_value_GOT_32): Likewise.
3537 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3538 (GOT_memory_operand): Likewise.
3539 (call_insn_operand): Likewise.
3540 (sibcall_insn_operand): Likewise.
3541 (GOT32_symbol_operand): Likewise.
3542 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3544 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3546 PR rtl-optimization/83496
3547 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3548 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3549 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3550 redundant insn, if any.
3551 (relax_delay_slots): Likewise.
3552 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3554 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
3556 PR tree-optimization/83965
3557 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3558 that grouped statements are part of a reduction chain. Return
3559 true if the statement is not marked as a reduction itself but
3561 (vect_recog_dot_prod_pattern): Don't check whether the statement
3562 is part of a group here.
3563 (vect_recog_sad_pattern): Likewise.
3564 (vect_recog_widen_sum_pattern): Likewise.
3566 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3569 * final.c (rest_of_clean_state): Also look for calls inside sequences.
3571 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3574 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3576 (ix86_output_indirect_function_return): New prototype.
3577 (ix86_split_simple_return_pop_internal): Likewise.
3578 * config/i386/i386.c (indirect_return_via_cx): New.
3579 (indirect_return_via_cx_bnd): Likewise.
3580 (indirect_thunk_name): Handle return va CX_REG.
3581 (output_indirect_thunk_function): Create alias for
3582 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3583 (ix86_output_indirect_jmp): Remove the bool argument.
3584 (ix86_output_indirect_function_return): New function.
3585 (ix86_split_simple_return_pop_internal): Likewise.
3586 * config/i386/i386.md (*indirect_jump): Don't pass false
3587 to ix86_output_indirect_jmp.
3588 (*tablejump_1): Likewise.
3589 (simple_return_pop_internal): Change it to define_insn_and_split.
3590 Call ix86_split_simple_return_pop_internal to split it for
3592 (simple_return_indirect_internal): Call
3593 ix86_output_indirect_function_return instead of
3594 ix86_output_indirect_jmp.
3596 2018-02-26 Jakub Jelinek <jakub@redhat.com>
3599 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3600 memset and value initialization afterwards.
3602 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
3604 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3606 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3609 * common/config/aarch64/aarch64-common.c
3610 (aarch_option_optimization_table[]): Switch
3611 off fomit-frame-pointer
3613 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3614 Chung-Ju Wu <jasonwucj@gmail.com>
3616 * config/nds32/nds32-multiple.md (load_multiple): Disallow
3618 (store_multiple): Ditto.
3620 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3622 * config.gcc: Add --with-cpu support for nds32 target.
3623 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
3624 * config/nds32/nds32.opt: Add -mcpu= option.
3626 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
3628 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
3629 isel=yes): Warn for these deprecated options.
3631 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
3633 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
3634 ISA_2_5_MASKS_EMBEDDED.
3636 2018-02-23 Jakub Jelinek <jakub@redhat.com>
3638 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
3639 p->max as pointers rather than using iterative_hash_expr.
3641 2018-02-23 Carl Love <cel@us.ibm.com>
3643 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
3644 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
3646 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
3647 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
3648 P8V_BUILTIN_VEC_VUNSIGNED2.
3650 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
3653 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
3654 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
3656 * lra-constraints.c (curr_insn_transform): Set up
3657 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
3659 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
3660 * lra-eliminations.c (spill_pseudos): Ditto.
3661 (process_insn_for_elimination): Ditto.
3662 * lra-lives.c (reg_early_clobber_p): Use the new macros.
3663 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
3664 LRA_NON_CLOBBERED_ALT.
3666 2018-02-22 Martin Sebor <msebor@redhat.com>
3668 PR tree-optimization/84480
3669 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
3670 to maybe_diag_stxncpy_trunc. Call it.
3671 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
3672 from gimple_fold_builtin_strcpy. Print inlining stack.
3673 (handle_builtin_stxncpy): Print inlining stack.
3674 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
3676 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
3679 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
3680 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
3681 and -fcheck-pointer-bounds are used together.
3682 (indirect_thunk_prefix): New enum.
3683 (indirect_thunk_need_prefix): New function.
3684 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
3685 "_nt" instead of "_bnd" for NOTRACK prefix.
3686 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
3687 (output_indirect_thunk_function): Likewise.
3689 (ix86_code_end): Update output_indirect_thunk_function calls.
3690 (ix86_output_indirect_branch_via_reg): Replace
3691 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
3692 (ix86_output_indirect_branch_via_push): Likewise.
3693 (ix86_output_function_return): Likewise.
3694 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
3695 incompatible with -fcf-protection=branch and
3696 -fcheck-pointer-bounds.
3698 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3701 * config/aarch64/aarch64.c (aarch64_print_address_internal):
3702 Change gcc_assert call to output_operand_lossage.
3704 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3706 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
3708 2018-02-22 DJ Delorie <dj@redhat.com>
3709 Sebastian Perta <sebastian.perta@renesas.com>
3710 Oleg Endo <olegendo@gcc.gnu.org>
3712 * config/rx/rx.c (rx_rtx_costs): New function.
3713 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
3715 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
3717 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
3719 2018-02-22 Martin Liska <mliska@suse.cz>
3722 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
3723 Add "native" as a possible value.
3725 2018-02-22 Martin Liska <mliska@suse.cz>
3728 * config/i386/i386.c (ix86_option_override_internal):
3729 Add "native" as a possible value for -march and -mtune.
3731 2018-02-22 Jakub Jelinek <jakub@redhat.com>
3734 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
3735 to all type variants.
3737 PR tree-optimization/84503
3738 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
3739 width as info->bitpos + info->bitsize - start.
3740 (merged_store_group::merge_overlapping): Simplify width computation.
3741 (check_no_overlap): New function.
3742 (imm_store_chain_info::try_coalesce_bswap): Compute expected
3743 start + width and last_order of the group, fail if check_no_overlap
3745 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
3746 to group if check_no_overlap fails.
3748 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3750 * config/rs6000/altivec.md: Delete contraint arguments to
3751 define_expand, define_split, and define_peephole2, and in
3752 define_insn_and_split if always unused.
3753 * config/rs6000/darwin.md: Ditto.
3754 * config/rs6000/dfp.md: Ditto.
3755 * config/rs6000/rs6000.md: Ditto.
3756 * config/rs6000/sync.md: Ditto.
3757 * config/rs6000/vector.md: Ditto.
3758 * config/rs6000/vsx.md: Ditto.
3760 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3762 * config/rs6000/altivec.md: Write output control strings as braced
3763 blocks instead of double-quoted strings.
3764 * config/rs6000/darwin.md: Ditto.
3765 * config/rs6000/rs6000.md: Ditto.
3766 * config/rs6000/vector.md: Ditto.
3767 * config/rs6000/vsx.md: Ditto.
3769 2018-02-21 Jason Merrill <jason@redhat.com>
3771 PR c++/84314 - ICE with templates and fastcall attribute.
3772 * attribs.c (build_type_attribute_qual_variant): Remove assert.
3774 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3776 * ipa-cp.c (determine_versionability): Fix comment typos.
3778 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3781 * ipa-cp.c (determine_versionability): Do not version functions caling
3784 2018-02-21 Martin Liska <mliska@suse.cz>
3787 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
3788 Add "native" as a possible value.
3789 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
3790 the macro when native cpu detection is available.
3792 2018-02-21 Martin Liska <mliska@suse.cz>
3795 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
3796 Add "native" as a possible value.
3797 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
3798 when native cpu detection is available.
3800 2018-02-21 Jakub Jelinek <jakub@redhat.com>
3801 Martin Sebor <msebor@redhat.com>
3803 PR tree-optimization/84478
3804 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
3806 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
3807 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
3808 support which is conservatively correct, for 2 only stay conservative
3809 for maxlen. Formatting and comment capitalization fixes. Add STRICT
3810 argument to the 2 argument get_range_strlen, adjust 6 arg
3811 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
3813 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
3814 (gimple_fold_builtin_strlen): Pass true as last argument to
3817 2018-02-20 Martin Sebor <msebor@redhat.com>
3820 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
3821 (builtin_memref::set_base_and_offset): Same. Handle inner references.
3822 (builtin_memref::builtin_memref): Factor out parts into
3823 set_base_and_offset and call it.
3825 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3828 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
3829 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
3830 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
3831 search at the associated MODE_INT.
3833 2018-02-20 Jeff Law <law@redhat.com>
3836 PR tree-optimization/81592
3838 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
3839 for range data rather than using global data.
3840 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
3841 range data rather than using global data.
3842 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
3843 pass it to children as needed.
3844 (struct directive::fmtresult): Similarly.
3845 (struct directive::set_width): Similarly.
3846 (struct directive::set_precision): Similarly.
3847 (format_integer, format_directive, parse_directive): Similarly.
3848 (format_none): Accept unnamed vr_values parameter.
3849 (format_percent, format_floating, format_character): Similarly.
3850 (format_string, format_plain): Similarly.
3851 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
3852 the EVRP range analyzer for range data rather than using global data.
3853 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
3854 gimple-ssa-evrp-analyze.h
3855 (class sprintf_dom_walker): Add after_dom_children member function.
3856 Add evrp_range_analyzer member.
3857 (sprintf_dom_walker::before_dom_children): Call into the EVRP
3858 range analyzer as needed.
3859 (sprintf_dom_walker::after_dom_children): New member function.
3860 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
3862 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
3863 (evrp_range_analyzer::pop_to_marker): Likewise.
3865 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3867 PR tree-optimization/84419
3868 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
3869 with the required type if its current type is compatible but
3872 2018-02-20 Jakub Jelinek <jakub@redhat.com>
3875 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
3876 after vectorization.
3878 2018-02-20 Martin Liska <mliska@suse.cz>
3881 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
3882 possible values if we don't have a hint.
3884 2018-02-20 Martin Liska <mliska@suse.cz>
3888 * final.c (shorten_branches): Build align_tab array with one
3890 * opts.c (finish_options): Add alignment option limit check.
3891 (MAX_CODE_ALIGN): Likewise.
3892 (MAX_CODE_ALIGN_VALUE): Likewise.
3893 * doc/invoke.texi: Document maximum allowed option value for
3894 all -falign-* options.
3896 2018-02-19 Jakub Jelinek <jakub@redhat.com>
3899 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
3900 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
3901 * var-tracking.c (emit_note_insn_var_location): Remove all references
3902 to NOTE_INSN_CALL_ARG_LOCATION.
3903 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
3904 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
3905 Use copy_rtx_if_shared.
3906 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
3907 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
3908 (dwarf2out_var_location): Remove handling of
3909 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
3911 * final.c (final_scan_insn): Remove all references to
3912 NOTE_INSN_CALL_ARG_LOCATION.
3913 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
3914 before dumping final insns.
3915 * except.c (emit_note_eh_region_end): Remove all references to
3916 NOTE_INSN_CALL_ARG_LOCATION.
3917 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
3918 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
3919 * config/arc/arc.c (hwloop_optimize): Likewise.
3920 * config/arm/arm.c (create_fix_barrier): Likewise.
3921 * config/s390/s390.c (s390_chunkify_start): Likewise.
3922 * config/sh/sh.c (find_barrier): Likewise.
3923 * config/i386/i386.c (rest_of_insert_endbranch,
3924 ix86_seh_fixup_eh_fallthru): Likewise.
3925 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
3926 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
3927 * config/frv/frv.c (frv_function_prologue): Likewise.
3928 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
3930 (note_outside_basic_block_p): Remove all references to
3931 NOTE_INSN_CALL_ARG_LOCATION.
3932 * gengtype.c (adjust_field_rtx_def): Likewise.
3933 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
3935 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
3936 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
3939 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
3942 PR tree-optimization/84452
3943 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
3944 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
3947 2018-02-19 Martin Liska <mliska@suse.cz>
3950 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
3952 2018-02-19 Martin Liska <mliska@suse.cz>
3953 Richard Sandiford <richard.sandiford@linaro.org>
3955 PR tree-optimization/82491
3956 * gimple-fold.c (get_base_constructor): Make earlier bail out
3959 2018-02-19 Carl Love <cel@us.ibm.com>
3961 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
3962 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
3964 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
3965 P8V_BUILTIN_VEC_NEG.
3967 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
3969 * config/rl78/rl78.md (movdf): New define expand.
3971 2018-02-19 Martin Liska <mliska@suse.cz>
3974 * doc/invoke.texi: Fix typo.
3975 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
3977 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
3979 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3980 handle rs6000_single_float and rs6000_double_float specially for
3983 2018-02-16 Jeff Law <law@redhat.com>
3985 * config/rx/rx.c (add_pop_cfi_notes): New function.;
3988 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3991 * ipa-inline.c (inline_small_functions): Fix a typo.
3993 2018-02-16 Nathan Sidwell <nathan@acm.org>
3995 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
3997 2018-02-16 Carl Love <cel@us.ibm.com>
3999 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4000 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4001 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4002 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4003 expansion to P8V_BUILTIN_VEC_FLOAT2.
4005 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4007 PR rtl-optimization/70023
4008 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4009 src_regno into account.
4011 2018-02-16 Carl Love <cel@us.ibm.com>
4013 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4014 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4015 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4016 * config/rs6000/rs6000.c: Remove case statements for
4017 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4018 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4019 and P9V_BUILTIN_VEC_VINSERT4B.
4020 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4021 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4022 * config/rs6000/vsx.md:
4023 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4026 2018-02-16 Carl Love <cel@us.ibm.com>
4028 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4030 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4032 * config/rs6000/rs6000-c.c: Add the definitions for
4033 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4034 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4035 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4036 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4037 definition for insert4b and define insn *insert3b_internal.
4038 * doc/extend.texi: Add documentation for vec_extract4b.
4040 2018-02-16 Nathan Sidwell <nathan@acm.org>
4042 * doc/extend.texi (Backwards Compatibility): Mention friend
4043 injection. Note for-scope is deprecated.
4044 * doc/invoke.texi (-ffriend-injection): Deprecate.
4046 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4048 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4049 that moved to I2, also allow destinations that are a paradoxical
4050 subreg (instead of a normal reg).
4052 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4055 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4058 2018-02-16 Richard Biener <rguenther@suse.de>
4060 PR tree-optimization/84037
4061 PR tree-optimization/84016
4063 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4064 Adjust vec_construct for the fact we need additional higher latency
4065 128bit inserts for AVX256 and AVX512 vector builds.
4066 (ix86_add_stmt_cost): Scale vector construction cost for
4069 2018-02-16 Richard Biener <rguenther@suse.de>
4071 PR tree-optimization/84417
4072 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4073 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4074 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4076 2018-02-16 Martin Liska <mliska@suse.cz>
4079 * internal-fn.def (ASAN_CHECK): Set proper flags.
4080 (ASAN_MARK): Likewise.
4082 2018-02-16 Julia Koval <julia.koval@intel.com>
4084 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4085 from PTA_CANNONLAKE.
4087 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4090 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4091 Use ++iter rather than iter++ for std::list iterators.
4092 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4093 defer deleting them until all nodes in the forest are processed. Do
4094 free even leaf nodes. Change to_process into auto_vec.
4097 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4098 * vec.h (vec_default_construct): Use memset instead of placement new
4099 if BROKEN_VALUE_INITIALIZATION is defined.
4100 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4101 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4104 PR rtl-optimization/83723
4105 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4106 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4107 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4109 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4111 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4113 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4115 PR rtl-optimization/81443
4116 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4117 from inner REGs to paradoxical SUBREGs.
4119 2018-02-16 Richard Biener <rguenther@suse.de>
4121 PR tree-optimization/84399
4122 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4123 For operands we can analyze at their definition make sure we can
4124 analyze them at each use as well.
4126 2018-02-16 Richard Biener <rguenther@suse.de>
4128 PR tree-optimization/84190
4129 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4130 volatile accesses if the decl isn't volatile.
4132 2018-02-15 Jason Merrill <jason@redhat.com>
4134 PR c++/84314 - ICE with templates and fastcall attribute.
4135 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4136 TYPE_CANONICAL on an existing type.
4138 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4140 PR tree-optimization/84383
4141 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4142 dstoff nor call operand_equal_p if dstbase is NULL.
4144 PR tree-optimization/84334
4145 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4146 also a CONSTANT_CLASS_P, punt.
4148 2018-02-14 Jim Wilson <jimw@sifive.com>
4150 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4151 first SMALL_OPERAND check. New local min_second_step. Move assert
4152 to where locals are set. Add TARGET_RVC support.
4153 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4155 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4157 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4159 2018-02-14 Martin Sebor <msebor@redhat.com>
4161 PR tree-optimization/83698
4162 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
4163 arrays constrain the offset range to their bounds.
4164 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
4165 (builtin_access::overlap): Avoid setting the size of overlap if it's
4167 (maybe_diag_overlap): Also consider arrays when deciding what values
4168 of offsets to include in diagnostics.
4170 2018-02-14 Martin Sebor <msebor@redhat.com>
4173 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
4174 that correspond to the kind of a declaration.
4176 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
4179 * config/pa/pa.md: Load address of PIC label using the linkage table
4180 if the label is nonlocal.
4182 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
4184 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
4185 warning message if user requests -maltivec=be.
4186 * doc/invoke.texi: Document deprecation of -maltivec=be.
4188 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
4191 * config/rs6000/rs6000-c.c: Update definitions for
4192 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
4193 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
4195 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4198 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4199 add _get_ssp intrinsics. Remove argument from
4200 __builtin_ia32_rdssp[d|q].
4201 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4202 * config/i386/i386-builtin.def: Remove argument from
4203 __builtin_ia32_rdssp[d|q].
4204 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4205 ix86_expand_special_args_builtin for _rdssp[d|q].
4206 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4207 Clear register before usage.
4208 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4209 Add documentation for new _get_ssp and _inc_ssp intrinsics.
4211 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
4213 PR tree-optimization/84357
4214 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4215 operand 1 of an ARRAY_REF too.
4217 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
4220 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4221 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4223 (set_of_reg): New struct.
4224 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4225 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4226 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4228 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4229 Split into bitclr, bitset, bitinvert patterns if appropriate.
4230 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4231 use rx_fuse_in_memory_bitop.
4232 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4233 to named insn, correct maximum insn length.
4235 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
4238 * machmode.def: Define a complex mode for PARTIAL_INT.
4239 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4241 * doc/rtl.texi: Document CSPImode.
4242 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4244 (msp430_hard_regno_nregs_with_padding): Likewise.
4246 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4249 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4251 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4253 PR rtl-optimization/84169
4254 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4255 we generated a parallel as new i3 and we split that to new i2 and i3
4256 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4257 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4258 those to i2, not i1. Partially rewrite this scan code.
4260 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4263 * stor-layout.c (place_field): For variable length fields, adjust
4264 offset_align afterwards not just based on the field's alignment,
4265 but also on the size.
4268 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4269 of exps and logs in the use_exp2 case.
4271 2018-02-13 Jeff Law <law@redhat.com>
4273 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4276 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4279 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4283 * common.opt (gas-loc-support, gas-locview-support): New.
4284 (ginline-points, ginternal-reset-location-views): New.
4285 * doc/invoke.texi: Document them. Use @itemx where intended.
4286 (gvariable-location-views): Adjust.
4287 * target.def (reset_location_view): New.
4288 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4289 (TARGET_RESET_LOCATION_VIEW): New.
4290 * doc/tm.texi: Rebuilt.
4291 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4292 (dwarf2out_default_as_locview_support): New.
4293 (output_asm_line_debug_info): Use option variables.
4294 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4295 (output_loc_list): Likewise.
4296 (add_high_low_attributes): Check option variables.
4297 Don't output entry view attribute in strict mode.
4298 (gen_inlined_subroutine_die): Check option variables.
4299 (dwarf2out_inline_entry): Likewise.
4300 (init_sections_and_labels): Likewise.
4301 (dwarf2out_early_finish): Likewise.
4302 (maybe_reset_location_view): New, from...
4303 (dwarf2out_var_location): ... here. Call it.
4304 * debug.h (dwarf2out_default_as_loc_support): Declare.
4305 (dwarf2out_default_as_locview_support): Declare.
4306 * hooks.c (hook_int_rtx_insn_0): New.
4307 * hooks.h (hook_int_rtx_insn_0): Declare.
4308 * toplev.c (process_options): Take -gas-loc-support and
4309 -gas-locview-support from dwarf2out. Enable
4310 -gvariable-location-views by default only with locview
4311 assembler support. Enable -ginternal-reset-location-views by
4312 default only if the target defines the corresponding hook.
4313 Enable -ginline-points by default if location views are
4314 enabled; force it disabled if statement frontiers are
4316 * tree-inline.c (expand_call_inline): Check option variables.
4317 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4319 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4321 PR tree-optimization/84321
4322 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4323 handling. Also check whether the anti-range contains any values
4324 that satisfy the mask; switch to a VR_RANGE if not.
4326 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4329 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4331 2018-02-13 Martin Jambor <mjambor@suse.cz>
4334 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4335 of call statements, also set location of a load to a temporary.
4337 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4339 * config/rl78/rl78.c (add_vector_labels): New function.
4340 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4341 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4342 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4343 which checks that no arguments are passed.
4344 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4345 * doc/extend.texi: Documentation for the new attribute.
4347 2018-02-13 Andreas Schwab <schwab@suse.de>
4349 * config/riscv/linux.h (CPP_SPEC): Define.
4351 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4354 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4355 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4356 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4357 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4358 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4359 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4360 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4361 temporarily for AES and PCLMUL builtins.
4363 PR tree-optimization/84339
4364 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4365 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4369 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4370 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4371 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4373 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4375 * omp-simd-clone.h: New file.
4376 * omp-simd-clone.c: Include omp-simd-clone.h.
4377 (expand_simd_clones): No longer static.
4378 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4379 cgraph.h and omp-simd-clone.h.
4380 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4381 (vect_recog_widen_shift_pattern): Formatting fix.
4382 (vect_pattern_recog_1): Don't check optab for calls.
4385 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4386 operands[2] into a REG before using gen_lowpart on it.
4388 2018-02-12 Jeff Law <law@redhat.com>
4391 * config/sh/sh.c (find_barrier): Consider a sibling call
4394 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4395 successfully back substituting a reg.
4397 2018-02-12 Richard Biener <rguenther@suse.de>
4399 PR tree-optimization/84037
4400 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4401 parameter, move visited init to caller.
4402 (vect_slp_analyze_operations): Separate cost from validity
4403 check, initialize visited once for all instances.
4404 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4406 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4408 (vect_model_promotion_demotion_cost): Likewise.
4409 (vectorizable_bswap): Guard cost modeling with !slp_node
4410 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4412 (vectorizable_call): Likewise.
4413 (vectorizable_conversion): Likewise.
4414 (vectorizable_assignment): Likewise.
4415 (vectorizable_shift): Likewise.
4416 (vectorizable_operation): Likewise.
4417 (vectorizable_store): Likewise.
4418 (vectorizable_load): Likewise.
4419 (vectorizable_condition): Likewise.
4420 (vectorizable_comparison): Likewise.
4422 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
4425 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
4426 (ASAN_MARK): Fix fnspec to account for return value, change pointer
4427 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
4429 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4432 * params.def (inline-min-speedup): Increase from 8 to 15.
4433 (max-inline-insns-auto): Decrease from 40 to 30.
4434 * ipa-split.c (consider_split): Add some buffer for function to
4435 be considered inlining candidate.
4436 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
4439 2018-02-12 Richard Biener <rguenther@suse.de>
4441 PR tree-optimization/84037
4442 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
4443 matched stmts if we cannot swap the non-matched ones.
4445 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4447 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
4448 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4449 _mm_maskz_scalef_round_ss): New intrinsics.
4450 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4451 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4452 __builtin_ia32_scalefss_round): Remove.
4453 (__builtin_ia32_scalefsd_mask_round,
4454 __builtin_ia32_scalefss_mask_round): New intrinsics.
4455 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4456 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4457 ((match_operand:VF_128 2 "<round_nimm_predicate>"
4458 "<round_constraint>")): Changed to ...
4459 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4460 "<round_scalar_constraint>")): ... this.
4461 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4462 %0, %1, %2<round_op3>}"): Changed to ...
4463 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4464 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4465 %2<round_scalar_mask_op3>}"): ... this.
4466 * config/i386/subst.md (round_scalar_nimm_predicate): New.
4468 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4470 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4471 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
4472 (_mm_maskz_sqrt_round_ss): New intrinsics.
4473 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
4474 (__builtin_ia32_sqrtsd_mask_round)
4475 (__builtin_ia32_sqrtss_mask_round): New builtins.
4476 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
4477 (__builtin_ia32_sqrtss_round): Remove.
4478 (__builtin_ia32_sqrtsd_mask_round)
4479 (__builtin_ia32_sqrtss_mask_round): New builtins.
4480 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4481 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4482 ((match_operand:VF_128 1 "vector_operand"
4483 "xBm,<round_constraint>")): Changed to ...
4484 ((match_operand:VF_128 1 "vector_operand"
4485 "xBm,<round_scalar_constraint>")): ... this.
4486 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4487 %0, %2, %<iptr>1<round_op3>}): Changed to ...
4488 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4489 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4490 %<iptr>1<round_scalar_mask_op3>}): ... this.
4491 ((set_attr "prefix" "<round_prefix>")): Changed to ...
4492 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4494 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
4497 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4498 Cast vec_cmpeq result to correct type.
4499 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4500 Cast vec_cmpgt result to correct type.
4502 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4504 * final.c (final_scan_insn_1): Renamed from...
4505 (final_scan_insn): ... this. New wrapper, to recover
4506 seen from the outermost call in recursive ones.
4507 * config/sparc/sparc.c (output_return): Drop seen from call.
4508 (output_sibcall): Likewise.
4509 * config/visium/visium.c (output_branch): Likewise.
4511 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4513 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4516 2018-02-10 Alan Modra <amodra@gmail.com>
4519 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4520 Specify LR as an input.
4522 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4525 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4526 remove_member_access_dummy_vars): New functions.
4527 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4528 lower_omp_1, execute_lower_omp): Use them.
4530 PR rtl-optimization/84308
4531 * shrink-wrap.c (spread_components): Release todo vector.
4533 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4535 PR rtl-optimization/57193
4536 * ira-color.c (struct allocno_color_data): Add member
4537 conflict_allocno_hard_prefs.
4538 (update_conflict_allocno_hard_prefs): New.
4539 (bucket_allocno_compare_func): Add a preference based on
4540 conflict_allocno_hard_prefs.
4541 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4542 (color_allocnos): Remove a dead code. Initiate
4543 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4545 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4548 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4549 constraint from =wa to wa. Avoid a subreg on the output operand,
4550 instead use a pseudo and subreg it in a move.
4551 (p9_xxbrd_<mode>): Changed to ...
4552 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4553 (p9_xxbrd_v2df): New expander.
4554 (p9_xxbrw_<mode>): Changed to ...
4555 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4556 (p9_xxbrw_v4sf): New expander.
4558 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4560 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4562 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
4565 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4566 multiply in 32-bit mode.
4567 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4568 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4571 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4573 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4574 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4575 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4576 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4578 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4581 * dwarf2out.c (is_trivial_indirect_ref): New function.
4582 (dwarf2out_late_global_decl): Do not generate a location
4583 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4584 and that are not defined in the current unit.
4586 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4588 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4589 instead of a libcall for UNORDERED.
4591 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4594 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4595 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4597 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4600 * config/s390/s390.c (s390_set_current_function): Invoke
4601 s390_indirect_branch_settings also if fndecl didn't change.
4603 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4605 * config/rs6000/rs6000.md (blockage): Set length to zero.
4607 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4609 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4611 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4614 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4615 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4619 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4620 PARALLEL incoming that failed vt_get_decl_and_offset check.
4623 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
4624 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
4626 (get_variable_section): For decls in named .bss* sections pass true as
4627 second argument to bss_initializer_p.
4629 2018-02-09 Marek Polacek <polacek@redhat.com>
4630 Jakub Jelinek <jakub@redhat.com>
4633 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
4634 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
4635 Sync some changes from cxx_fold_indirect_ref.
4637 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4639 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
4641 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
4642 (BLOCK_INLINE_ENTRY_LABEL): New.
4643 (dwarf2out_var_location): Disregard inline entry markers.
4644 (inline_entry_data): New struct.
4645 (inline_entry_data_hasher): New hashtable type.
4646 (inline_entry_data_hasher::hash): New.
4647 (inline_entry_data_hasher::equal): New.
4648 (inline_entry_data_table): New variable.
4649 (add_high_low_attributes): Add DW_AT_entry_pc and
4650 DW_AT_GNU_entry_view attributes if a pending entry is found
4651 in inline_entry_data_table. Add old entry_pc attribute only
4652 if debug nonbinding markers are disabled.
4653 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
4654 markers are enabled.
4655 (block_within_block_p, dwarf2out_inline_entry): New.
4656 (dwarf2out_finish): Check that no entries remained in
4657 inline_entry_data_table.
4658 * final.c (reemit_insn_block_notes): Handle inline entry notes.
4659 (final_scan_insn, notice_source_line): Likewise.
4660 (rest_of_clean_state): Skip inline entry markers.
4661 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
4663 * gimple.c (gimple_build_debug_inline_entry): New.
4664 * gimple.h (enum gimple_debug_subcode): Add
4665 GIMPLE_DEBUG_INLINE_ENTRY.
4666 (gimple_build_debug_inline_entry): Declare.
4667 (gimple_debug_inline_entry_p): New.
4668 (gimple_debug_nonbind_marker_p): Adjust.
4669 * insn-notes.def (INLINE_ENTRY): New.
4670 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4671 inline entry marker notes.
4672 (print_insn): Likewise.
4673 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
4674 (INSN_DEBUG_MARKER_KIND): Likewise.
4675 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
4676 * tree-inline.c (expand_call_inline): Build and insert
4677 debug_inline_entry stmt.
4678 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
4679 inline entry blocks early, if nonbind markers are enabled.
4680 (dump_scope_block): Dump fragment info.
4681 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
4682 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
4683 (gimple_build_debug_inline_entry): New.
4684 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
4685 Enable/disable inline entry points too.
4686 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
4687 (DEBUG_INSN): Describe inline entry markers.
4689 * common.opt (gvariable-location-views): New.
4690 (gvariable-location-views=incompat5): New.
4691 * config.in: Rebuilt.
4692 * configure: Rebuilt.
4693 * configure.ac: Test assembler for view support.
4694 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
4695 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
4696 * dwarf2out.c (var_loc_view): New typedef.
4697 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
4698 (dwarf2out_locviews_in_attribute): New.
4699 (dwarf2out_locviews_in_loclist): New.
4700 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
4701 (enum dw_line_info_opcode): Add LI_adv_address.
4702 (struct dw_line_info_table): Add view.
4703 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
4704 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
4705 (zero_view_p): New variable.
4706 (ZERO_VIEW_P): New macro.
4707 (output_asm_line_debug_info): New.
4708 (struct var_loc_node): Add view.
4709 (add_AT_view_list, AT_loc_list): New.
4710 (add_var_loc_to_decl): Add view param. Test it against last.
4711 (new_loc_list): Add view params. Record them.
4712 (AT_loc_list_ptr): Handle loc and view lists.
4713 (view_list_to_loc_list_val_node): New.
4714 (print_dw_val): Handle dw_val_class_view_list.
4715 (size_of_die): Likewise.
4716 (value_format): Likewise.
4717 (loc_list_has_views): New.
4718 (gen_llsym): Set vl_symbol too.
4719 (maybe_gen_llsym, skip_loc_list_entry): New.
4720 (dwarf2out_maybe_output_loclist_view_pair): New.
4721 (output_loc_list): Output view list or entries too.
4722 (output_view_list_offset): New.
4723 (output_die): Handle dw_val_class_view_list.
4724 (output_dwarf_version): New.
4725 (output_compilation_unit_header): Use it.
4726 (output_skeleton_debug_sections): Likewise.
4727 (output_rnglists, output_line_info): Likewise.
4728 (output_pubnames, output_aranges): Update version comments.
4729 (output_one_line_info_table): Output view numbers in asm comments.
4730 (dw_loc_list): Determine current endview, pass it to new_loc_list.
4731 Call maybe_gen_llsym.
4732 (loc_list_from_tree_1): Adjust.
4733 (add_AT_location_description): Create view list attribute if
4734 needed, check it's absent otherwise.
4735 (convert_cfa_to_fb_loc_list): Adjust.
4736 (maybe_emit_file): Call output_asm_line_debug_info for test.
4737 (dwarf2out_var_location): Reset views as needed. Precompute
4738 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
4739 attribute. Set view.
4740 (new_line_info_table): Reset next view.
4741 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
4742 (dwarf2out_source_line): Likewise. Output view resets and labels to
4743 the assembler, or select appropriate line info opcodes.
4744 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
4745 (optimize_string_length): Catch it. Adjust.
4746 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
4747 dw_val_class_view_list, and remove it if no longer needed.
4748 (hash_loc_list): Hash view numbers.
4749 (loc_list_hasher::equal): Compare them.
4750 (optimize_location_lists): Check whether a view list symbol is
4751 needed, and whether the locview attribute is present, and
4752 whether they match. Remove the locview attribute if no longer
4754 (index_location_lists): Call skip_loc_list_entry for test.
4755 (dwarf2out_finish): Call output_asm_line_debug_info for test.
4756 Use output_dwarf_version.
4757 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
4758 (struct dw_val_node): Add val_view_list.
4759 * final.c (SEEN_NEXT_VIEW): New.
4760 (set_next_view_needed): New.
4761 (clear_next_view_needed): New.
4762 (maybe_output_next_view): New.
4763 (final_start_function): Rename to...
4764 (final_start_function_1): ... this. Take pointer to FIRST,
4765 add SEEN parameter. Emit param bindings in the initial view.
4766 (final_start_function): Reintroduce SEEN-less interface.
4767 (final): Rename to...
4768 (final_1): ... this. Take SEEN parameter. Output final pending
4769 next view at the end.
4770 (final): Reintroduce seen-less interface.
4771 (final_scan_insn): Output pending next view before switching
4772 sections or ending a block. Mark the next view as needed when
4773 outputting variable locations. Notify debug backend of section
4774 changes, and of location view changes.
4775 (rest_of_handle_final): Adjust.
4776 * toplev.c (process_options): Autodetect value for debug variable
4777 location views option. Warn on incompat5 without -gdwarf-5.
4778 * doc/invoke.texi (gvariable-location-views): New.
4779 (gvariable-location-views=incompat5): New.
4780 (gno-variable-location-views): New.
4782 2018-02-08 David Malcolm <dmalcolm@redhat.com>
4784 PR tree-optimization/84136
4785 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
4786 that the result of find_edge is non-NULL.
4788 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4791 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
4792 storing integer register in SImode. Fix cost of 256 and 512
4793 byte aligned SSE register store.
4795 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4797 * config/i386/i386.c (ix86_multiplication_cost): Fix
4798 multiplication cost for TARGET_AVX512DQ.
4800 2018-02-08 Marek Polacek <polacek@redhat.com>
4802 PR tree-optimization/84238
4803 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
4806 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4808 PR tree-optimization/84265
4809 * tree-vect-stmts.c (vectorizable_store): Don't treat
4810 VMAT_CONTIGUOUS accesses as grouped.
4811 (vectorizable_load): Likewise.
4813 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4815 PR tree-optimization/81635
4816 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
4817 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
4818 (test_round_for_mask): New functions.
4819 (wide_int_cc_tests): Call test_round_for_mask.
4820 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
4821 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
4822 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
4823 range returned by get_range_info.
4825 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4828 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
4829 * symtab.c: Include builtins.h
4830 (symtab_node::output_to_lto_symbol_table_p): Move here
4831 from lto-streamer-out.c:output_symbol_p.
4832 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
4833 (output_symbol_p): Move all logic to symtab.c
4834 (produce_symtab): Update.
4836 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4838 * config/s390/s390-opts.h (enum indirect_branch): Define.
4839 * config/s390/s390-protos.h (s390_return_addr_from_memory)
4840 (s390_indirect_branch_via_thunk)
4841 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
4842 (enum s390_indirect_branch_type): Define.
4843 * config/s390/s390.c (struct s390_frame_layout, struct
4844 machine_function): Remove.
4845 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
4846 (indirect_branch_table_label_no, indirect_branch_table_name):
4848 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
4849 (enum s390_indirect_branch_option): Define.
4850 (s390_return_addr_from_memory): New function.
4851 (s390_handle_string_attribute): New function.
4852 (s390_attribute_table): Add new attribute handler.
4853 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
4854 (s390_indirect_branch_via_thunk): New function.
4855 (s390_indirect_branch_via_inline_thunk): New function.
4856 (s390_function_ok_for_sibcall): When jumping via thunk disallow
4857 sibling call optimization for non z10 compiles.
4858 (s390_emit_call): Force indirect branch target to be a single
4859 register. Add r1 clobber for non-z10 compiles.
4860 (s390_emit_epilogue): Emit return jump via return_use expander.
4861 (s390_reorg): Handle JUMP_INSNs as execute targets.
4862 (s390_option_override_internal): Perform validity checks for the
4863 new command line options.
4864 (s390_indirect_branch_attrvalue): New function.
4865 (s390_indirect_branch_settings): New function.
4866 (s390_set_current_function): Invoke s390_indirect_branch_settings.
4867 (s390_output_indirect_thunk_function): New function.
4868 (s390_code_end): Implement target hook.
4869 (s390_case_values_threshold): Implement target hook.
4870 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
4872 * config/s390/s390.h (struct s390_frame_layout)
4873 (struct machine_function): Move here from s390.c.
4874 (TARGET_INDIRECT_BRANCH_NOBP_RET)
4875 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
4876 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
4877 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
4878 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
4879 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
4880 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
4881 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
4882 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
4883 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
4884 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
4885 (mnemonic attribute): Add values which aren't recognized
4887 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
4888 pattern for branch conversion. Fix mnemonic attribute.
4889 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
4890 indirect branch via thunk if requested.
4891 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
4892 ("*indirect_jump"): Disable for branch conversion using out of
4894 ("indirect_jump_via_thunk<mode>_z10")
4895 ("indirect_jump_via_thunk<mode>")
4896 ("indirect_jump_via_inlinethunk<mode>_z10")
4897 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
4898 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
4899 ("casesi_jump_via_inlinethunk<mode>_z10")
4900 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
4901 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
4902 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
4903 ("*indirect2_jump"): Disable for branch conversion.
4904 ("casesi_jump"): Turn into expander and expand patterns for branch
4906 ("return_use"): New expander.
4907 ("*return"): Emit return via thunk and rename it to ...
4908 ("*return<mode>"): ... this one.
4909 * config/s390/s390.opt: Add new options and and enum for the
4912 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4914 * lra-constraints.c (match_reload): Unconditionally use
4915 gen_lowpart_SUBREG, rather than selecting between that
4916 and equivalent gen_rtx_SUBREG code.
4918 2018-02-08 Richard Biener <rguenther@suse.de>
4920 PR tree-optimization/84233
4921 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
4922 changed flag instead of boguously re-using phi_inserted.
4924 2018-02-08 Martin Jambor <mjambor@suse.cz>
4926 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
4927 static local variables.
4929 2018-02-08 Richard Biener <rguenther@suse.de>
4931 PR tree-optimization/84278
4932 * tree-vect-stmts.c (vectorizable_store): When looking for
4933 smaller vector types to perform grouped strided loads/stores
4934 make sure the mode is supported by the target.
4935 (vectorizable_load): Likewise.
4937 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4939 * config/aarch64/aarch64.c (aarch64_components_for_bb):
4940 Increase LDP/STP opportunities by adding adjacent callee-saves.
4942 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4944 PR rtl-optimization/84068
4945 PR rtl-optimization/83459
4946 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
4948 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
4950 PR tree-optimization/84224
4951 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
4952 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
4955 2018-02-07 Iain Sandoe <iain@codesourcery.com>
4958 * config/rs6000/altivec.md (*restore_world): Remove LR use.
4959 * config/rs6000/predicates.md (restore_world_operation): Adjust op
4960 count, remove one USE.
4962 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
4964 * doc/install.texi (Configuration): Document the
4965 --with-long-double-format={ibm,ieee} PowerPC configuration
4969 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
4970 Convert from define_expand to be define_insn_and_split. Rework
4971 float/double/_Float128 conversions to QI/HI/SImode to work with
4972 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
4973 conversions to QI/HImode types did a store and then a load to
4974 truncate the value. For conversions to VSX registers, don't split
4975 the insn, instead emit the code directly. Use the code iterator
4976 any_fix to combine signed and unsigned conversions.
4977 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
4978 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
4979 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4980 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
4981 (fix_<mode>di2_hw): Likewise.
4982 (fixuns_<mode>di2_hw): Likewise.
4983 (fix_<mode>si2_hw): Likewise.
4984 (fixuns_<mode>si2_hw): Likewise.
4985 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
4986 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4987 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
4988 fix<uns>_trunc<SFDF:mode>si2_p8.
4989 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
4991 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
4992 (fix<uns>_<mode>_mem): Likewise.
4993 (fctiw<u>z_<mode>_mem): Likewise.
4994 (fix<uns>_<mode>_mem): Likewise.
4995 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
4996 the register allocator from doing a direct move to the GPRs to do
4997 a store, and instead use the ISA 3.0 store byte/half-word from
4998 vector register instruction. For IEEE 128-bit floating point,
4999 also optimize stores of 32-bit ints.
5000 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5002 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5004 * genextract.c (push_pathstr_operand): New function to support
5006 (walk_rtx): Call push_pathstr_operand.
5007 (print_path): Support [a-zA-Z].
5009 2018-02-07 Richard Biener <rguenther@suse.de>
5011 PR tree-optimization/84037
5012 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5013 (cse_and_gimplify_to_preheader): Declare.
5014 (vect_get_place_in_interleaving_chain): Likewise.
5015 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5017 (_loop_vec_info::~_loop_vec_info): Delete it.
5018 (cse_and_gimplify_to_preheader): New function.
5019 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5020 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5021 (vectorizable_load): Likewise. For grouped stores always base
5022 the IV on the first element.
5023 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5024 condition before gimplifying.
5026 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5028 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5029 *DIV_EXPR and *MOD_EXPR.
5031 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5034 * config/i386/i386.c (ix86_option_override_internal): Mask out
5035 the CF_SET bit when checking -fcf-protection.
5037 2018-02-07 Tom de Vries <tom@codesourcery.com>
5040 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5043 2018-02-07 Richard Biener <rguenther@suse.de>
5045 PR tree-optimization/84204
5046 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5049 PR tree-optimization/84205
5050 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5051 special-case isl_ast_op_zdiv_r.
5053 PR tree-optimization/84223
5054 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5055 Only add conditions from within the region.
5056 (gather_bbs::after_dom_children): Adjust.
5058 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5061 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5062 * config/avr/avr.md: Only post-reload split REG-REG moves if
5063 either register is GENERAL_REG_P.
5065 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5067 PR tree-optimization/84235
5068 * tree-ssa-scopedtables.c
5069 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5070 if the subtraction is performed in floating point type where NaNs are
5071 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5072 build 1. Formatting fix.
5074 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5077 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5078 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5079 and skip it regardless of bb boundaries. Use CALL_P macro,
5080 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5081 unnecessarily, formatting fix.
5083 2018-02-06 Michael Collison <michael.collison@arm.com>
5085 * config/arm/thumb2.md:
5086 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5087 (*thumb_mov_notscc): Ditto.
5089 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5092 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5093 unsigned_fix, not "s".
5095 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5097 * configure.ac (gcc_fn_eh_frame_ro): New function.
5098 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5099 correct .eh_frame permissions.
5100 * configure: Regenerate.
5102 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5104 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5107 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5109 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5110 Display warning message for -mno-speculate-indirect-jumps.
5112 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5114 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5116 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5118 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5120 PR tree-optimization/84225
5121 * tree-eh.c (find_trapping_overflow): Only call
5122 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5124 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5127 * config/i386/i386.c: Reimplement the check of possible options
5128 -mibt/-mshstk conbination. Change error messages.
5129 * doc/invoke.texi: Fix a typo: remove extra '='.
5131 2018-02-06 Marek Polacek <polacek@redhat.com>
5133 PR tree-optimization/84228
5134 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5136 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5139 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5140 emitted arch directives.
5141 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5142 __ARM_FEATURE_COPROC before changing architectures.
5144 2018-02-06 Richard Biener <rguenther@suse.de>
5146 * config/i386/i386.c (print_reg): Fix typo.
5147 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5149 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5151 * configure: Regenerate.
5153 2018-02-05 Martin Sebor <msebor@redhat.com>
5155 PR tree-optimization/83369
5156 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5159 2018-02-05 Martin Liska <mliska@suse.cz>
5161 * doc/invoke.texi: Cherry-pick upstream r323995.
5163 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
5165 * ira.c (ira_init_register_move_cost): Adjust comment.
5167 2018-02-05 Martin Liska <mliska@suse.cz>
5169 PR gcov-profile/84137
5170 * doc/gcov.texi: Fix typo in documentation.
5172 2018-02-05 Martin Liska <mliska@suse.cz>
5174 PR gcov-profile/83879
5175 * doc/gcov.texi: Document necessity of --dynamic-list-data when
5176 using dlopen functionality.
5178 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
5180 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
5181 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
5182 _mm_maskz_range_ss, _mm_mask_range_round_ss,
5183 _mm_maskz_range_round_ss): New intrinsics.
5184 (__builtin_ia32_rangesd128_round)
5185 (__builtin_ia32_rangess128_round): Remove.
5186 (__builtin_ia32_rangesd128_mask_round,
5187 __builtin_ia32_rangess128_mask_round): New builtins.
5188 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
5189 __builtin_ia32_rangess128_round): Remove.
5190 (__builtin_ia32_rangesd128_mask_round,
5191 __builtin_ia32_rangess128_mask_round): New builtins.
5192 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
5193 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
5194 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
5195 "<round_saeonly_constraint>")): Changed to ...
5196 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
5197 "<round_saeonly_scalar_constraint>")): ... this.
5198 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5199 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5200 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5201 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5202 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5204 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
5206 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5208 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5209 Remove all values except native, 8540 and 8548.
5211 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
5213 * config/i386/i386.c (ix86_output_function_return): Pass
5214 INVALID_REGNUM, instead of -1, as invalid register number to
5215 indirect_thunk_name and output_indirect_thunk.
5217 2018-02-02 Julia Koval <julia.koval@intel.com>
5219 * config.gcc: Add -march=icelake.
5220 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5221 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5222 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5223 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5224 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5225 (processor_target_table): Add icelake.
5226 (ix86_option_override_internal): Handle new PTAs.
5227 (get_builtin_code_for_version): Handle icelake.
5228 (M_INTEL_COREI7_ICELAKE): New.
5229 (fold_builtin_cpu): Handle icelake.
5230 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5231 * doc/invoke.texi: Add -march=icelake.
5233 2018-02-02 Julia Koval <julia.koval@intel.com>
5235 * config/i386/i386.c (ix86_option_override_internal): Change flags type
5236 to wide_int_bitmask.
5237 * wide-int-bitmask.h: New.
5239 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5242 * config/i386/i386.md: Replace Pmode with word_mode in
5243 builtin_setjmp_setup and builtin_longjmp to support x32.
5245 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5249 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5251 (rs6000_supported_cpu_names): New static variable.
5252 (linux_cpu_translation_table): Likewise.
5253 (elf_platform) <cpu>: Define new static variable and use it.
5254 Translate kernel AT_PLATFORM name to canonical name if needed.
5255 Error if platform name is unknown.
5257 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
5260 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5262 2018-02-01 Jeff Law <law@redhat.com>
5265 * config/i386/i386.c (release_scratch_register_on_entry): Add new
5266 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
5267 the scratch if RELEASE_VIA_POP is false.
5268 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5269 If we have to save a temporary register, decrement SIZE appropriately.
5270 Pass new arguments to release_scratch_register_on_entry.
5271 (ix86_adjust_stack_and_probe): Likewise.
5272 (ix86_emit_probe_stack_range): Pass new arguments to
5273 release_scratch_register_on_entry.
5275 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5277 PR rtl-optimization/84157
5278 * combine.c (change_zero_ext): Use REG_P predicate in
5279 front of HARD_REGISTER_P predicate.
5281 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5283 * config/avr/avr.c (avr_option_override): Move disabling of
5284 -fdelete-null-pointer-checks to...
5285 * common/config/avr/avr-common.c (avr_option_optimization_table):
5288 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5290 PR tree-optimization/81635
5291 * tree-data-ref.c (split_constant_offset_1): For types that
5292 wrap on overflow, try to use range info to prove that wrapping
5295 2018-02-01 Renlin Li <renlin.li@arm.com>
5298 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5300 (aarch64_register_move_cost): Likewise.
5301 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5303 (REG_CLASS_NAMES): Likewise.
5304 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5305 TAILCALL_ADDR_REGS. Remove IP registers.
5306 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5308 2018-02-01 Richard Biener <rguenther@suse.de>
5310 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5311 for specifying RPO order and allow NULL for that.
5312 * domwalk.c (dom_walker::dom_walker): Likewise.
5313 (dom_walker::walk): Handle NULL RPO order.
5314 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5316 (rewrite_update_dom_walker): Likewise.
5317 (mark_def_dom_walker): Likewise.
5319 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5321 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5322 (aarch64_maybe_expand_sve_subreg_move): Declare.
5323 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5324 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5326 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5327 that are semantically a reverse operation.
5328 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5329 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5330 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5332 (aarch64_can_change_mode_class): For big-endian, forbid changes
5333 between two SVE modes if they have different element sizes.
5335 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5337 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5338 the TImode handling for big-endian targets.
5340 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5342 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5343 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5345 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5346 Remove BSWAP handing for big-endian targets and use the form of
5347 LD1RQ appropariate for the mode.
5349 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5351 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5352 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5355 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5358 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5359 check for operands that need to go through aarch64_sve_reload_be.
5361 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5363 PR tree-optimization/81661
5364 PR tree-optimization/84117
5365 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5366 * tree-eh.c: Include gimplify.h.
5367 (find_trapping_overflow, replace_trapping_overflow,
5368 rewrite_to_non_trapping_overflow): New functions.
5369 * tree-vect-loop.c: Include tree-eh.h.
5370 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5371 * tree-data-ref.c: Include tree-eh.h.
5372 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5374 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
5376 PR rtl-optimization/84123
5377 * combine.c (change_zero_ext): Check if hard register satisfies
5378 can_change_dest_mode before calling gen_lowpart_SUBREG.
5380 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
5383 * ira.c (ira_init_register_move_cost): Remove assert.
5385 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5387 PR rtl-optimization/84071
5388 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5389 * doc/tm.texi: Regenerate.
5391 2018-01-31 Richard Biener <rguenther@suse.de>
5393 PR tree-optimization/84132
5394 * tree-data-ref.c (analyze_miv_subscript): Properly
5395 check whether evolution_function_is_affine_multivariate_p
5396 before calling gcd_of_steps_may_divide_p.
5398 2018-01-31 Julia Koval <julia.koval@intel.com>
5401 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5402 * config/i386/i386.md (rdpid_rex64) New.
5403 (rdpid): Make 32bit only.
5405 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5408 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5409 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5411 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5414 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5416 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5418 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5420 PR rtl-optimization/84071
5421 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
5422 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
5424 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5426 * config/arc/arc.c (arc_handle_aux_attribute): New function.
5427 (arc_attribute_table): Add 'aux' attribute.
5428 (arc_in_small_data_p): Consider aux like variables.
5429 (arc_is_aux_reg_p): New function.
5430 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
5431 (arc_get_aux_arg): New function.
5432 (prepare_move_operands): Handle aux-register access.
5433 (arc_handle_aux_attribute): New function.
5434 * doc/extend.texi (ARC Variable attributes): Add subsection.
5436 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5438 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
5439 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
5440 (arc_attribute_table): Add 'uncached' attribute.
5441 (arc_print_operand): Print '.di' flag for uncached memory
5443 (arc_in_small_data_p): Do not consider for small data the uncached
5445 (arc_is_uncached_mem_p): New function.
5446 * config/arc/predicates.md (compact_store_memory_operand): Check
5447 for uncached memory accesses.
5448 (nonvol_nonimm_operand): Likewise.
5449 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5451 2018-01-31 Jakub Jelinek <jakub@redhat.com>
5454 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5455 falign-loops=): Add Optimization flag.
5457 2018-01-30 Jeff Law <law@redhat.com>
5460 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5461 INT_REGISTERS_SAVED. Check it prior to calling
5462 get_scratch_register_on_entry.
5463 (ix86_adjust_stack_and_probe): Similarly.
5464 (ix86_emit_probe_stack_range): Similarly.
5465 (ix86_expand_prologue): Corresponding changes.
5467 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5470 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5471 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
5473 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
5476 * lra-constraints.c (curr_insn_transform): Process AND in the
5479 2018-01-30 Jakub Jelinek <jakub@redhat.com>
5481 PR rtl-optimization/83986
5482 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5483 dependence against last_pending_memory_flush in addition to
5486 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
5488 PR tree-optimization/81611
5489 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5492 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5495 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5498 2018-01-30 Richard Biener <rguenther@suse.de>
5499 Jakub Jelinek <jakub@redhat.com>
5501 PR tree-optimization/84111
5502 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5503 inner loops added during recursion, as they don't have up-to-date
5506 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5509 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5510 (can_inline_edge_by_limits_p): ... here.
5511 (can_early_inline_edge_p, check_callers,
5512 update_caller_keys, update_callee_keys, recursive_inlining,
5513 add_new_edges_to_heap, speculation_useful_p,
5514 inline_small_functions,
5515 inline_small_functions, flatten_function,
5516 inline_to_all_callers_1): Update.
5518 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5520 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5523 2018-01-30 Richard Biener <rguenther@suse.de>
5525 PR tree-optimization/83008
5526 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5527 invariant and constant vector uses in stmts when they need
5530 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5533 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5534 * configure: Regenerate.
5536 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5538 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5540 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5541 Use gen_rtx_REG rather than gen_lowpart.
5543 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5545 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5546 rather than 0 when creating partial subregs.
5548 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5550 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5553 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
5556 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5557 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5558 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5559 flags. This restores the settings used before the 2017-07-24.
5560 Turning off pre increment/decrement/modify allows IVOPTS to
5561 optimize DF/SF loops where the index is an int.
5563 2018-01-29 Richard Biener <rguenther@suse.de>
5564 Kelvin Nilsen <kelvin@gcc.gnu.org>
5567 * tree-vect-stmts.c (vectorizable_call): Don't call
5568 targetm.vectorize_builtin_md_vectorized_function if callee is
5571 2018-01-22 Carl Love <cel@us.ibm.com>
5573 * doc/extend.tex: Fix typo in second arg in
5574 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5576 2018-01-29 Richard Biener <rguenther@suse.de>
5578 PR tree-optimization/84086
5579 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5580 (flush_ssaname_freelist): When SSA names were released reset
5581 the SCEV hash table.
5583 2018-01-29 Richard Biener <rguenther@suse.de>
5585 PR tree-optimization/84057
5586 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5587 removed paths when removing edges.
5589 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
5591 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5592 -mfunction-return=@var{choice}.
5594 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
5597 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5599 (layout::print_source_line): Likewise.
5600 (test_get_line_width_without_trailing_whitespace): Add test cases.
5602 2018-01-27 Jakub Jelinek <jakub@redhat.com>
5605 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5608 2018-01-26 Jim Wilson <jimw@sifive.com>
5610 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5612 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5615 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5617 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5618 and CMP + SUB-immediate -> SUBS.
5620 2018-01-26 Martin Sebor <msebor@redhat.com>
5622 PR tree-optimization/83896
5623 * tree-ssa-strlen.c (get_string_len): Rename...
5624 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
5625 Avoid assuming length is constant.
5626 (handle_char_store): Use HOST_WIDE_INT for string length.
5628 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
5631 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
5632 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
5634 2018-01-26 Richard Biener <rguenther@suse.de>
5636 PR rtl-optimization/84003
5637 * dse.c (record_store): Only record redundant stores when
5638 the earlier store aliases at least all accesses the later one does.
5640 2018-01-26 Jakub Jelinek <jakub@redhat.com>
5642 PR rtl-optimization/83985
5643 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
5644 REG_CFA_RESTORE insns.
5645 (delete_unmarked_insns): Don't ignore separate shrink wrapping
5646 REG_CFA_RESTORE insns here.
5649 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
5650 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
5652 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5654 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
5655 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
5656 (arc_init): Likewise.
5657 (arc_override_options): Likewise.
5658 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
5660 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
5662 * config/arc/arc.h (TARGET_DBNZ): Define.
5663 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
5664 properly set the tune attribute.
5665 (dbnz): Use TARGET_DBNZ guard.
5666 * config/arc/arc.opt (mtune): Add core3 option.
5668 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5670 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
5671 recognize new pic like addresses.
5672 (arc_delegitimize_address): Clean up.
5674 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5676 * config/arc/arc-arches.def: Option mrf16 valid for all
5678 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
5679 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
5680 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
5681 * config/arc/arc-tables.opt: Regenerate.
5682 * config/arc/arc.c (arc_conditional_register_usage): Handle
5683 reduced register file case.
5684 (arc_file_start): Set must have build attributes.
5685 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
5687 * config/arc/arc.opt (mrf16): Add new option.
5688 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
5689 * config/arc/genmultilib.awk: Handle new mrf16 option.
5690 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
5691 * config/arc/t-multilib: Regenerate.
5692 * doc/invoke.texi (ARC Options): Document mrf16 option.
5694 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5696 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
5697 * config/arc/arc.c (arc_handle_secure_attribute): New function.
5698 (arc_attribute_table): Add 'secure_call' attribute.
5699 (arc_print_operand): Print secure call operand.
5700 (arc_function_ok_for_sibcall): Don't optimize tail calls when
5702 (arc_is_secure_call_p): New function. * config/arc/arc.md
5703 (call_i): Add support for sjli instruction.
5704 (call_value_i): Likewise.
5705 * config/arc/constraints.md (Csc): New constraint.
5707 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5708 John Eric Martin <John.Martin@emmicro-us.com>
5710 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
5711 * config/arc/arc.c (_arc_jli_section): New struct.
5712 (arc_jli_section): New type.
5713 (rc_jli_sections): New static variable.
5714 (arc_handle_jli_attribute): New function.
5715 (arc_attribute_table): Add jli_always and jli_fixed attribute.
5716 (arc_file_end): New function.
5717 (TARGET_ASM_FILE_END): Define.
5718 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
5719 (arc_add_jli_section): New function.
5720 (jli_call_scan): Likewise.
5721 (arc_reorg): Call jli_call_scan.
5722 (arc_output_addsi): Remove 'S' from printing asm operand.
5723 (arc_is_jli_call_p): New function.
5724 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
5726 (movhi_insn): Likewise.
5727 (movsi_insn): Likewise.
5728 (movsi_set_cc_insn): Likewise.
5729 (loadqi_update): Likewise.
5730 (load_zeroextendqisi_update): Likewise.
5731 (load_signextendqisi_update): Likewise.
5732 (loadhi_update): Likewise.
5733 (load_zeroextendhisi_update): Likewise.
5734 (load_signextendhisi_update): Likewise.
5735 (loadsi_update): Likewise.
5736 (loadsf_update): Likewise.
5737 (movsicc_insn): Likewise.
5738 (bset_insn): Likewise.
5739 (bxor_insn): Likewise.
5740 (bclr_insn): Likewise.
5741 (bmsk_insn): Likewise.
5742 (bicsi3_insn): Likewise.
5743 (cmpsi_cc_c_insn): Likewise.
5744 (movsi_ne): Likewise.
5745 (movsi_cond_exec): Likewise.
5746 (clrsbsi2): Likewise.
5757 (call_i): Remove 'S' asm letter, add jli instruction.
5758 (call_value_i): Likewise.
5759 * config/arc/arc.op (mjli-always): New option.
5760 * config/arc/constraints.md (Cji): New constraint.
5761 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
5763 (subsf3_fpx): Likewise.
5764 (mulsf3_fpx): Likewise.
5765 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
5767 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
5769 * doc/invoke.texi (ARC): Document mjli-always option.
5771 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
5773 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
5774 avoid addition with 0 and use incw and decw where possible.
5776 2018-01-26 Richard Biener <rguenther@suse.de>
5778 PR tree-optimization/81082
5779 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
5780 association if it requires casting to unsigned.
5781 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
5782 from fold_plusminus_mult_expr to catch important cases late when
5783 range info is available.
5785 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5787 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
5788 * configure.ac (hidden_linkonce): New test.
5789 * configure: Regenerate.
5790 * config.in: Regenerate.
5792 2018-01-26 Julia Koval <julia.koval@intel.com>
5794 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
5795 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
5796 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
5797 _mm_mask_bitshuffle_epi64_mask): Fix type.
5798 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
5799 USI_FTYPE_V4DI_V4DI_USI): Remove.
5800 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
5801 __builtin_ia32_vpshufbitqmb256_mask,
5802 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
5803 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
5804 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
5806 2018-01-26 Alan Modra <amodra@gmail.com>
5809 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
5810 UNSPEC_VBPERMQ. Sort other unspecs.
5812 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
5814 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
5816 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
5819 * predict.c (drop_profile): Do not push/pop cfun; update also
5821 (handle_missing_profiles): Fix logic looking for zero profiles.
5823 2018-01-25 Jakub Jelinek <jakub@redhat.com>
5826 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
5827 on functions with #pragma omp declare simd or functions with simd
5829 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
5830 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5831 Remove trailing \n from warning_at calls.
5833 2018-01-25 Tom de Vries <tom@codesourcery.com>
5836 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5837 for neutered workers.
5839 2018-01-24 Joseph Myers <joseph@codesourcery.com>
5842 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
5843 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
5845 2018-01-24 Jeff Law <law@redhat.com>
5848 * i386.c (get_probe_interval): Move to earlier point.
5849 (ix86_compute_frame_layout): If -fstack-clash-protection and
5850 the frame is larger than the probe interval, then use pushes
5851 to save registers rather than reg->mem moves.
5852 (ix86_expand_prologue): Remove conditional for int_registers_saved
5855 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
5858 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
5859 min/max for never referenced object.
5861 2018-01-24 Jakub Jelinek <jakub@redhat.com>
5864 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
5866 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
5867 attributes from DECL_ATTRIBUTES (decl) without affecting
5868 DECL_ATTRIBUTES (current_function_decl).
5869 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
5870 functions with non-NULL DECL_ABSTRACT_ORIGIN.
5872 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
5874 PR tree-optimization/83979
5875 * fold-const.c (fold_comparison): Use constant_boolean_node
5876 instead of boolean_{true,false}_node.
5878 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
5880 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
5883 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5885 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5886 Simplify the clause that sets the length attribute.
5887 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5888 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
5889 clause that sets the length attribute.
5890 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5892 2018-01-24 Tom de Vries <tom@codesourcery.com>
5895 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
5896 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
5897 Add strict parameter.
5898 (prevent_branch_around_nothing): Insert dummy insn between branch to
5899 label and label with no ptx insn inbetween.
5900 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
5902 2018-01-24 Tom de Vries <tom@codesourcery.com>
5905 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5906 for neutered threads in warp.
5907 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
5909 2018-01-24 Richard Biener <rguenther@suse.de>
5911 PR tree-optimization/83176
5912 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
5915 2018-01-24 Richard Biener <rguenther@suse.de>
5917 PR tree-optimization/82819
5918 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
5919 code generating pluses that are no-ops in the target precision.
5921 2018-01-24 Richard Biener <rguenther@suse.de>
5924 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
5926 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5928 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
5929 to merge probabilities.
5930 * predict.c (probably_never_executed): Also mark as cold functions
5931 with global 0 profile and guessed local profile.
5932 * profile-count.c (profile_probability::combine_with_count): New
5934 * profile-count.h (profile_probability::operator*,
5935 profile_probability::operator*=, profile_probability::operator/,
5936 profile_probability::operator/=): Reduce precision to adjusted
5937 and set value to guessed on contradictory divisions.
5938 (profile_probability::combine_with_freq): Remove.
5939 (profile_probability::combine_wiht_count): Declare.
5940 (profile_count::force_nonzero):: Set to adjusted.
5941 (profile_count::probability_in):: Set quality to adjusted.
5942 * tree-ssa-tail-merge.c (replace_block_by): Use
5945 2018-01-23 Andrew Waterman <andrew@sifive.com>
5946 Jim Wilson <jimw@sifive.com>
5948 * config/riscv/riscv.c (riscv_stack_boundary): New.
5949 (riscv_option_override): Set riscv_stack_boundary. Handle
5950 riscv_preferred_stack_boundary_arg.
5951 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
5952 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
5953 (STACK_BOUNDARY): Set to riscv_stack_boundary.
5954 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
5955 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
5956 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
5958 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
5961 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
5962 of struct ix86_frame.
5963 (ix86_expand_epilogue): Likewise. Add a local variable for
5964 the reg_save_offset field in struct ix86_frame.
5966 2018-01-23 Bin Cheng <bin.cheng@arm.com>
5968 PR tree-optimization/82604
5969 * tree-loop-distribution.c (enum partition_kind): New enum item
5970 PKIND_PARTIAL_MEMSET.
5971 (partition_builtin_p): Support above new enum item.
5972 (generate_code_for_partition): Ditto.
5973 (compute_access_range): Differentiate cases that equality can be
5974 proven at all loops, the innermost loops or no loops.
5975 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
5976 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
5977 (finalize_partitions, distribute_loop): Don't fuse partition of
5978 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
5979 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
5982 2018-01-23 Martin Liska <mliska@suse.cz>
5984 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
5985 order to ignore the predictor.
5986 (PRED_POLYMORPHIC_CALL): Likewise.
5987 (PRED_RECURSIVE_CALL): Likewise.
5989 2018-01-23 Martin Liska <mliska@suse.cz>
5991 * tree-profile.c (tree_profiling): Print function header to
5992 aware reader which function we are working on.
5993 * value-prof.c (gimple_find_values_to_profile): Do not print
5994 not interesting value histograms.
5996 2018-01-23 Martin Liska <mliska@suse.cz>
5998 * profile-count.h (enum profile_quality): Add
5999 profile_uninitialized as the first value. Do not number values
6000 as they are zero based.
6001 (profile_count::verify): Update sanity check.
6002 (profile_probability::verify): Likewise.
6004 2018-01-23 Nathan Sidwell <nathan@acm.org>
6006 * doc/invoke.texi (ffor-scope): Deprecate.
6008 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6010 PR tree-optimization/83510
6011 * domwalk.c (set_all_edges_as_executable): New function.
6012 (dom_walker::dom_walker): Convert bool param
6013 "skip_unreachable_blocks" to enum reachability. Move setup of
6014 edge flags to set_all_edges_as_executable and only do it when
6015 reachability is REACHABLE_BLOCKS.
6016 * domwalk.h (enum dom_walker::reachability): New enum.
6017 (dom_walker::dom_walker): Convert bool param
6018 "skip_unreachable_blocks" to enum reachability.
6019 (set_all_edges_as_executable): New decl.
6020 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6021 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6023 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6024 but converting true to REACHABLE_BLOCKS.
6025 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6027 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6028 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6029 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6031 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6032 if check_all_array_refs will be called.
6034 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6036 * tree.c (selftest::test_location_wrappers): Add more test
6039 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6041 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6042 (selftest::test_bit_in_range): Likewise.
6044 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6047 * doc/sourcebuild.texi (vect_float): Say that the selector
6048 only describes the situation when -funsafe-math-optimizations is on.
6049 (vect_float_strict): Document.
6051 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6053 PR tree-optimization/83965
6054 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6055 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6056 instead of checking only for a reduction.
6057 (vect_recog_widen_sum_pattern): Likewise.
6059 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6061 * predict.c (probably_never_executed): Only use precise profile info.
6062 (compute_function_frequency): Skip after inlining hack since we now
6063 have quality checking.
6065 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6067 * profile-count.h (profile_probability::very_unlikely,
6068 profile_probability::unlikely, profile_probability::even): Set
6069 precision to guessed.
6071 2018-01-23 Richard Biener <rguenther@suse.de>
6073 PR tree-optimization/83963
6074 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6075 Properly terminate dominator walk when crossing the exit edge not
6076 when visiting its source block.
6078 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6081 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6082 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6084 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6086 PR tree-optimization/83957
6087 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6088 semicolon after for body surrounded by braces.
6090 PR tree-optimization/83081
6091 * profile-count.h (profile_probability::split): New method.
6092 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6093 Use profile_probability::split.
6094 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6095 when splitting a single conditional jump into 2.
6097 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6099 PR tree-optimization/69452
6100 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6103 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6105 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6106 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6107 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6109 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6111 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6112 * config/rl78/rl78.md (movdi): New define_expand.
6113 * config/rl78/rl78.c (rl78_split_movdi): New function.
6115 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6118 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6120 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6121 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6122 128-bit to produce an UNSPEC move to get the double word with the
6123 signbit and then a shift directly to do signbit.
6124 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6125 implementation with a new version that just does either a direct
6126 move or a regular move. Move memory interface to separate insns.
6127 Move insns so they are next to the expander.
6128 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6129 with signbit move. Split big and little endian case.
6130 (signbit<mode>2_dm_mem_le): Likewise.
6131 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6132 (signbit<mode>2_dm2): Likewise.
6134 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6136 * config/rl78/rl78.md (anddi3): New define_expand.
6138 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6140 * config/rl78/rl78.md (umindi3): New define_expand.
6142 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6144 * config/rl78/rl78.md (smindi3): New define_expand.
6146 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6148 * config/rl78/rl78.md (smaxdi3): New define_expand.
6150 2018-01-22 Carl Love <cel@us.ibm.com>
6152 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6153 LVX_V1TI): Add macro expansion.
6154 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6155 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6156 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6157 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6158 Change check to determine if the instruction is a byte reversing
6159 entry. Fix typo in comment.
6160 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6161 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
6162 Add def_builtin calls for new builtins.
6163 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
6164 Add define_insn expansion.
6166 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6168 * config/rl78/rl78.md (umaxdi3): New define_expand.
6170 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6172 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
6173 for non-QImode registers.
6175 2018-01-22 Richard Biener <rguenther@suse.de>
6177 PR tree-optimization/83963
6178 * graphite-scop-detection.c (scop_detection::get_sese): Delay
6179 including the loop exit block.
6180 (scop_detection::merge_sese): Likewise.
6181 (scop_detection::add_scop): Do it here instead.
6183 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6185 * doc/sourcebuild.texi (arm_softfloat): Document.
6187 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
6190 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
6191 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
6192 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
6194 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6195 David Edelsohn <dje.gcc@gmail.com>
6198 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6199 Change "crset eq" to "crset 2".
6200 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6201 (*call_indirect_aix<mode>_nospec): Likewise.
6202 (*call_value_indirect_aix<mode>_nospec): Likewise.
6203 (*call_indirect_elfv2<mode>_nospec): Likewise.
6204 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6205 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6206 change assembly output from . to $.
6207 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6208 (indirect_jump<mode>_nospec): Change assembly output from . to $.
6209 (*tablejump<mode>_internal1_nospec): Likewise.
6211 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
6214 * config/sh/sh_optimize_sett_clrt.cc:
6215 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6217 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
6219 PR tree-optimization/83940
6220 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6221 offset_dt to vect_constant_def rather than vect_unknown_def_type.
6222 (vect_check_load_store_mask): Add a mask_dt_out parameter and
6223 use it to pass back the definition type.
6224 (vect_check_store_rhs): Likewise rhs_dt_out.
6225 (vect_build_gather_load_calls): Add a mask_dt argument and use
6226 it instead of a call to vect_is_simple_use.
6227 (vectorizable_store): Update calls to vect_check_load_store_mask
6228 and vect_check_store_rhs. Use the dt returned by the latter instead
6229 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
6230 instead of calls to vect_is_simple_use. Pass the scalar rather
6231 than the vector operand to vect_is_simple_use when handling
6232 second and subsequent copies of an rhs value.
6233 (vectorizable_load): Update calls to vect_check_load_store_mask
6234 and vect_build_gather_load_calls. Use the cached mask_dt and
6235 gs_info.offset_dt instead of calls to vect_is_simple_use.
6237 2018-01-20 Jakub Jelinek <jakub@redhat.com>
6240 * tree-emutls.c: Include gimplify.h.
6241 (lower_emutls_2): New function.
6242 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6243 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6244 it before further processing.
6247 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6248 UINTVAL (trueop1) instead of INTVAL (op1).
6250 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6254 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6255 INCOMING_FRAME_SP_OFFSET if not defined.
6256 (scan_trace): Add ENTRY argument. If true and
6257 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6258 emit a note to adjust the CFA offset.
6259 (create_cfi_notes): Adjust scan_trace callers.
6260 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6261 INCOMING_FRAME_SP_OFFSET in the CIE.
6262 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6263 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6265 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6266 * doc/tm.texi: Regenerated.
6268 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6270 PR rtl-optimization/83147
6271 * lra-constraints.c (remove_inheritance_pseudos): Use
6272 lra_substitute_pseudo_within_insn.
6274 2018-01-19 Tom de Vries <tom@codesourcery.com>
6275 Cesar Philippidis <cesar@codesourcery.com>
6278 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6280 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6283 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6284 spaces for function labels.
6286 2018-01-19 Martin Liska <mliska@suse.cz>
6288 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6289 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6290 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6291 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6292 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6293 (PRED_CONST_RETURN): Change from 69 to 65.
6294 (PRED_NULL_RETURN): Change from 91 to 71.
6295 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6296 (PRED_LOOP_GUARD): Change from 66 to 73.
6298 2018-01-19 Martin Liska <mliska@suse.cz>
6300 * predict.c (predict_insn_def): Add new assert.
6301 (struct branch_predictor): Change type to signed integer.
6302 (test_prediction_value_range): Amend test to cover
6304 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6305 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6306 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6307 (PRED_LOOP_IV_COMPARE): Likewise.
6308 * predict.h (PROB_UNINITIALIZED): Define new constant.
6310 2018-01-19 Martin Liska <mliska@suse.cz>
6312 * predict.c (dump_prediction): Add new format for
6313 analyze_brprob.py script which is enabled with -details
6315 * profile-count.h (precise_p): New function.
6317 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6319 PR tree-optimization/83922
6320 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6321 there are no statements that need masking.
6322 (vect_active_double_reduction_p): New function.
6323 (vect_analyze_loop_operations): Use it when handling phis that
6324 are not in the loop header.
6326 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6328 PR tree-optimization/83914
6329 * tree-vect-loop.c (vectorizable_induction): Don't convert
6330 init_expr or apply the peeling adjustment for inductions
6331 that are nested within the vectorized loop.
6333 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6335 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6338 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6342 * function.h (gimplify_parameters): Add gimple_seq * argument.
6343 * function.c: Include gimple.h and options.h.
6344 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6345 for the added local temporaries if needed.
6346 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6347 if there are any parameter cleanups, wrap whole body into a
6348 try/finally with the cleanups.
6350 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6353 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6354 Use GET_MODE_CLASS for scalar floating point.
6356 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6360 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6361 Fix call of call_cgraph_insertion_hooks.
6363 2018-01-18 Martin Sebor <msebor@redhat.com>
6365 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6367 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6370 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6373 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
6376 * common.opt: (-ffile-prefix-map): New option.
6377 * opts.c (common_handle_option): Defer it.
6378 * opts-global.c (handle_common_deferred_options): Handle it.
6379 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6380 * file-prefix-map.h: New file.
6381 (remap_debug_filename, add_debug_prefix_map): ...here.
6382 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6383 * final.c (debug_prefix_map, add_debug_prefix_map
6384 remap_debug_filename): Move to...
6385 * file-prefix-map.c: New file.
6386 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6387 generalize, get rid of alloca(), use strrchr() instead of strchr().
6388 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6389 Implement in terms of add_prefix_map().
6390 (remap_macro_filename, remap_debug_filename): Implement in term of
6392 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6393 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6394 * dbxout.c: Include file-prefix-map.h.
6395 * varasm.c: Likewise.
6396 * vmsdbgout.c: Likewise.
6397 * xcoffout.c: Likewise.
6398 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6399 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6400 * doc/invoke.texi (-ffile-prefix-map): Document.
6401 (-fdebug-prefix-map): Update description.
6403 2018-01-18 Martin Liska <mliska@suse.cz>
6405 * config/i386/i386.c (indirect_thunk_name): Document that also
6407 (output_indirect_thunk): Document why both instructions
6408 (pause and lfence) are generated.
6410 2018-01-18 Richard Biener <rguenther@suse.de>
6412 PR tree-optimization/83887
6413 * graphite-scop-detection.c
6414 (scop_detection::get_nearest_dom_with_single_entry): Remove.
6415 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6416 (scop_detection::merge_sese): Re-implement with a flood-fill
6417 algorithm that properly finds a SESE region if it exists.
6419 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6422 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
6423 pointer_diff optimizations use view_convert instead of convert.
6425 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6427 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6428 Generate different code for -mno-speculate-indirect-jumps.
6429 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6430 (*call_indirect_aix<mode>): Disable for
6431 -mno-speculate-indirect-jumps.
6432 (*call_indirect_aix<mode>_nospec): New define_insn.
6433 (*call_value_indirect_aix<mode>): Disable for
6434 -mno-speculate-indirect-jumps.
6435 (*call_value_indirect_aix<mode>_nospec): New define_insn.
6436 (*sibcall_nonlocal_sysv<mode>): Generate different code for
6437 -mno-speculate-indirect-jumps.
6438 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6440 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6442 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
6443 long double type, set the flags for noting the default long double
6444 type, even if we don't pass or return a long double type.
6446 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
6449 * ipa-inline.c (flatten_function): Do not overwrite final inlining
6452 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
6454 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6455 support for merge[hl].
6456 (fold_mergehl_helper): New helper function.
6457 (tree-vector-builder.h): New #include for tree_vector_builder usage.
6458 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6459 (altivec_vmrglw_direct): Add xxmrglw insn.
6461 2018-01-17 Andrew Waterman <andrew@sifive.com>
6463 * config/riscv/riscv.c (riscv_conditional_register_usage): If
6464 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6466 2018-01-17 David Malcolm <dmalcolm@redhat.com>
6469 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6470 call the lto_location_cache before reading the
6471 DECL_SOURCE_LOCATION of the types.
6473 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
6474 Richard Sandiford <richard.sandiford@linaro.org>
6476 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
6477 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6478 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6479 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6480 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6482 * config/aarch64/constraints.md (aarch64_movti_operand):
6484 * config/aarch64/predicates.md (Uti): Add new constraint.
6486 2018-01-17 Carl Love <cel@us.ibm.com>
6488 * config/rs6000/vsx.md (define_expand xl_len_r,
6489 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6490 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6492 (define_expand, define_insn): Move the shift left from the
6493 define_insn to the define_expand for lxvl and stxvl instructions.
6494 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6495 and XL_LEN_R definitions to PURE.
6497 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
6499 * config/i386/i386.c (indirect_thunk_name): Declare regno
6500 as unsigned int. Compare regno with INVALID_REGNUM.
6501 (output_indirect_thunk): Ditto.
6502 (output_indirect_thunk_function): Ditto.
6503 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6504 in the call to output_indirect_thunk_function.
6506 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6509 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6510 rather than the size of inner_type to determine the stack slot size
6511 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6513 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6516 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6519 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6521 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6522 endian Linux systems to optionally enable multilibs for selecting
6523 the long double type if the user configured an explicit type.
6524 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6525 have no long double multilibs if not defined.
6526 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6527 warn if the user used -mabi={ieee,ibm}longdouble and we built
6528 multilibs for long double.
6529 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6530 appropriate multilib option.
6531 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6533 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6534 for building long double multilibs.
6535 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6537 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6539 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6542 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6544 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6547 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6550 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6553 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6555 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6556 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6558 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6560 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6561 different rtl trees depending on TARGET_64BIT.
6562 (rs6000_gen_lvx): Likewise.
6564 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6566 * config/visium/visium.md (nop): Tweak comment.
6567 (hazard_nop): Likewise.
6569 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6571 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6572 -mspeculate-indirect-jumps.
6573 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6574 for -mno-speculate-indirect-jumps.
6575 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6576 (*call_value_indirect_elfv2<mode>): Disable for
6577 -mno-speculate-indirect-jumps.
6578 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6579 (indirect_jump): Emit different RTL for
6580 -mno-speculate-indirect-jumps.
6581 (*indirect_jump<mode>): Disable for
6582 -mno-speculate-indirect-jumps.
6583 (*indirect_jump<mode>_nospec): New define_insn.
6584 (tablejump): Emit different RTL for
6585 -mno-speculate-indirect-jumps.
6586 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6587 (tablejumpsi_nospec): New define_expand.
6588 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6589 (tablejumpdi_nospec): New define_expand.
6590 (*tablejump<mode>_internal1): Disable for
6591 -mno-speculate-indirect-jumps.
6592 (*tablejump<mode>_internal1_nospec): New define_insn.
6593 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6596 2018-01-16 Artyom Skrobov tyomitch@gmail.com
6598 * caller-save.c (insert_save): Drop unnecessary parameter. All
6601 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6602 Richard Biener <rguenth@suse.de>
6605 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6606 return early, inline manually is_gimple_sizepos. Make sure if we
6607 call gimplify_expr we don't end up with a gimple constant.
6608 * tree.c (variably_modified_type_p): Don't return true for
6609 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
6610 * gimplify.h (is_gimple_sizepos): Remove.
6612 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6614 PR tree-optimization/83857
6615 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6616 vectorizable_live_operation for pure SLP statements.
6617 (vectorizable_live_operation): Handle PHIs.
6619 2018-01-16 Richard Biener <rguenther@suse.de>
6621 PR tree-optimization/83867
6622 * tree-vect-stmts.c (vect_transform_stmt): Precompute
6623 nested_in_vect_loop_p since the scalar stmt may get invalidated.
6625 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6628 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
6629 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
6630 If off is not INTEGER_CST, issue a may not be aligned warning
6631 rather than isn't aligned. Use isn%'t rather than isn't.
6632 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
6634 <case MULT_EXPR>: Improve the case when bottom and one of the
6635 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
6636 operand, in that case check if the other operand is multiple of
6637 bottom divided by the INTEGER_CST operand.
6639 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6642 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
6643 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
6644 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
6645 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
6646 * config/pa/pa.c (pa_function_arg_advance): Likewise.
6647 (pa_function_arg, pa_arg_partial_bytes): Likewise.
6648 (pa_function_arg_size): New function.
6650 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6652 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
6653 in a separate statement.
6655 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6657 PR tree-optimization/83847
6658 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
6659 group gathers and scatters.
6661 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6663 PR rtl-optimization/86620
6664 * params.def (max-sched-ready-insns): Bump minimum value to 1.
6666 PR rtl-optimization/83213
6667 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
6668 to last if both are JUMP_INSNs.
6670 PR tree-optimization/83843
6671 * gimple-ssa-store-merging.c
6672 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
6673 store_immediate_info for bswap/nop orig_stores.
6675 2018-01-15 Andrew Waterman <andrew@sifive.com>
6677 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
6679 <UDIV>: Increase cost if !TARGET_DIV.
6681 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
6683 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
6684 (define_attr "cr_logical_3op"): New.
6685 (cceq_ior_compare): Adjust.
6686 (cceq_ior_compare_complement): Adjust.
6687 (*cceq_rev_compare): Adjust.
6688 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
6689 (is_cracked_insn): Adjust.
6690 (insn_must_be_first_in_group): Adjust.
6691 * config/rs6000/40x.md: Adjust.
6692 * config/rs6000/440.md: Adjust.
6693 * config/rs6000/476.md: Adjust.
6694 * config/rs6000/601.md: Adjust.
6695 * config/rs6000/603.md: Adjust.
6696 * config/rs6000/6xx.md: Adjust.
6697 * config/rs6000/7450.md: Adjust.
6698 * config/rs6000/7xx.md: Adjust.
6699 * config/rs6000/8540.md: Adjust.
6700 * config/rs6000/cell.md: Adjust.
6701 * config/rs6000/e300c2c3.md: Adjust.
6702 * config/rs6000/e500mc.md: Adjust.
6703 * config/rs6000/e500mc64.md: Adjust.
6704 * config/rs6000/e5500.md: Adjust.
6705 * config/rs6000/e6500.md: Adjust.
6706 * config/rs6000/mpc.md: Adjust.
6707 * config/rs6000/power4.md: Adjust.
6708 * config/rs6000/power5.md: Adjust.
6709 * config/rs6000/power6.md: Adjust.
6710 * config/rs6000/power7.md: Adjust.
6711 * config/rs6000/power8.md: Adjust.
6712 * config/rs6000/power9.md: Adjust.
6713 * config/rs6000/rs64.md: Adjust.
6714 * config/rs6000/titan.md: Adjust.
6716 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6718 * config/i386/predicates.md (indirect_branch_operand): Rewrite
6719 ix86_indirect_branch_register logic.
6721 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6723 * config/i386/constraints.md (Bs): Update
6724 ix86_indirect_branch_register check. Don't check
6725 ix86_indirect_branch_register with GOT_memory_operand.
6727 * config/i386/predicates.md (GOT_memory_operand): Don't check
6728 ix86_indirect_branch_register here.
6729 (GOT32_symbol_operand): Likewise.
6731 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6733 * config/i386/predicates.md (constant_call_address_operand):
6734 Rewrite ix86_indirect_branch_register logic.
6735 (sibcall_insn_operand): Likewise.
6737 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6739 * config/i386/constraints.md (Bs): Replace
6740 ix86_indirect_branch_thunk_register with
6741 ix86_indirect_branch_register.
6743 * config/i386/i386.md (indirect_jump): Likewise.
6744 (tablejump): Likewise.
6745 (*sibcall_memory): Likewise.
6746 (*sibcall_value_memory): Likewise.
6747 Peepholes of indirect call and jump via memory: Likewise.
6748 * config/i386/i386.opt: Likewise.
6749 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6750 (GOT_memory_operand): Likewise.
6751 (call_insn_operand): Likewise.
6752 (sibcall_insn_operand): Likewise.
6753 (GOT32_symbol_operand): Likewise.
6755 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6758 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
6759 type rather than type addr's type points to.
6760 (expand_omp_atomic_mutex): Likewise.
6761 (expand_omp_atomic): Likewise.
6763 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6766 * config/i386/i386.c (output_indirect_thunk_function): Use
6767 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
6768 for __x86_return_thunk.
6770 2018-01-15 Richard Biener <rguenther@suse.de>
6773 * expmed.c (extract_bit_field_1): Fix typo.
6775 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6778 * config/arm/iterators.md (VF): New mode iterator.
6779 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
6780 Remove integer-related logic from pattern.
6781 (neon_vabd<mode>_3): Likewise.
6783 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6786 * common.opt (fstrict-overflow): No longer an alias.
6787 (fwrapv-pointer): New option.
6788 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
6789 also for pointer types based on flag_wrapv_pointer.
6790 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
6791 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
6792 opts->x_flag_wrapv got set.
6793 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
6794 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
6795 POINTER_TYPE_OVERFLOW_UNDEFINED.
6796 * match.pd: Likewise in address comparison pattern.
6797 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
6799 2018-01-15 Richard Biener <rguenther@suse.de>
6802 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
6803 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
6804 Reset type names to their identifier if their TYPE_DECL doesn't
6805 have linkage (and thus is used for ODR and devirt).
6806 (save_debug_info_for_decl): Remove.
6807 (save_debug_info_for_type): Likewise.
6808 (add_tree_to_fld_list): Adjust.
6809 * tree-pretty-print.c (dump_generic_node): Make dumping of
6810 type names more robust.
6812 2018-01-15 Richard Biener <rguenther@suse.de>
6814 * BASE-VER: Bump to 8.0.1.
6816 2018-01-14 Martin Sebor <msebor@redhat.com>
6819 * builtins.c (check_access): Avoid warning when the no-warning bit
6822 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
6824 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
6825 * ira-color (allocno_hard_regs_compare): Likewise.
6827 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
6830 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
6831 Use .pushsection/.popsection.
6833 2018-01-14 Martin Sebor <msebor@redhat.com>
6836 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
6838 2018-01-14 Jakub Jelinek <jakub@redhat.com>
6840 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
6841 entry from extra_headers.
6842 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
6843 extra_headers, make the list bitwise identical to the i?86-*-* one.
6845 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6847 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
6848 -mcmodel=large with -mindirect-branch=thunk,
6849 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
6850 -mfunction-return=thunk-extern.
6851 * doc/invoke.texi: Document -mcmodel=large is incompatible with
6852 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
6853 -mfunction-return=thunk and -mfunction-return=thunk-extern.
6855 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6857 * config/i386/i386.c (print_reg): Print the name of the full
6858 integer register without '%'.
6859 (ix86_print_operand): Handle 'V'.
6860 * doc/extend.texi: Document 'V' modifier.
6862 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6864 * config/i386/constraints.md (Bs): Disallow memory operand for
6865 -mindirect-branch-register.
6867 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6868 (GOT_memory_operand): Likewise.
6869 (call_insn_operand): Likewise.
6870 (sibcall_insn_operand): Likewise.
6871 (GOT32_symbol_operand): Likewise.
6872 * config/i386/i386.md (indirect_jump): Call convert_memory_address
6873 for -mindirect-branch-register.
6874 (tablejump): Likewise.
6875 (*sibcall_memory): Likewise.
6876 (*sibcall_value_memory): Likewise.
6877 Disallow peepholes of indirect call and jump via memory for
6878 -mindirect-branch-register.
6879 (*call_pop): Replace m with Bw.
6880 (*call_value_pop): Likewise.
6881 (*sibcall_pop_memory): Replace m with Bs.
6882 * config/i386/i386.opt (mindirect-branch-register): New option.
6883 * doc/invoke.texi: Document -mindirect-branch-register option.
6885 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6887 * config/i386/i386-protos.h (ix86_output_function_return): New.
6888 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
6889 set function_return_type.
6890 (indirect_thunk_name): Add ret_p to indicate thunk for function
6892 (output_indirect_thunk_function): Pass false to
6893 indirect_thunk_name.
6894 (ix86_output_indirect_branch_via_reg): Likewise.
6895 (ix86_output_indirect_branch_via_push): Likewise.
6896 (output_indirect_thunk_function): Create alias for function
6897 return thunk if regno < 0.
6898 (ix86_output_function_return): New function.
6899 (ix86_handle_fndecl_attribute): Handle function_return.
6900 (ix86_attribute_table): Add function_return.
6901 * config/i386/i386.h (machine_function): Add
6902 function_return_type.
6903 * config/i386/i386.md (simple_return_internal): Use
6904 ix86_output_function_return.
6905 (simple_return_internal_long): Likewise.
6906 * config/i386/i386.opt (mfunction-return=): New option.
6907 (indirect_branch): Mention -mfunction-return=.
6908 * doc/extend.texi: Document function_return function attribute.
6909 * doc/invoke.texi: Document -mfunction-return= option.
6911 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6913 * config/i386/i386-opts.h (indirect_branch): New.
6914 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
6915 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
6916 with local indirect jump when converting indirect call and jump.
6917 (ix86_set_indirect_branch_type): New.
6918 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
6919 (indirectlabelno): New.
6920 (indirect_thunk_needed): Likewise.
6921 (indirect_thunk_bnd_needed): Likewise.
6922 (indirect_thunks_used): Likewise.
6923 (indirect_thunks_bnd_used): Likewise.
6924 (INDIRECT_LABEL): Likewise.
6925 (indirect_thunk_name): Likewise.
6926 (output_indirect_thunk): Likewise.
6927 (output_indirect_thunk_function): Likewise.
6928 (ix86_output_indirect_branch_via_reg): Likewise.
6929 (ix86_output_indirect_branch_via_push): Likewise.
6930 (ix86_output_indirect_branch): Likewise.
6931 (ix86_output_indirect_jmp): Likewise.
6932 (ix86_code_end): Call output_indirect_thunk_function if needed.
6933 (ix86_output_call_insn): Call ix86_output_indirect_branch if
6935 (ix86_handle_fndecl_attribute): Handle indirect_branch.
6936 (ix86_attribute_table): Add indirect_branch.
6937 * config/i386/i386.h (machine_function): Add indirect_branch_type
6938 and has_local_indirect_jump.
6939 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
6941 (tablejump): Likewise.
6942 (*indirect_jump): Use ix86_output_indirect_jmp.
6943 (*tablejump_1): Likewise.
6944 (simple_return_indirect_internal): Likewise.
6945 * config/i386/i386.opt (mindirect-branch=): New option.
6946 (indirect_branch): New.
6949 (thunk-inline): Likewise.
6950 (thunk-extern): Likewise.
6951 * doc/extend.texi: Document indirect_branch function attribute.
6952 * doc/invoke.texi: Document -mindirect-branch= option.
6954 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
6957 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
6959 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
6961 * ipa-inline.c (want_inline_small_function_p): Return false if
6962 inlining has already failed with CIF_FINAL_ERROR.
6963 (update_caller_keys): Call want_inline_small_function_p before
6965 (update_callee_keys): Likewise.
6967 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6969 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
6971 (rs6000_quadword_masked_address_p): Likewise.
6972 (quad_aligned_load_p): Likewise.
6973 (quad_aligned_store_p): Likewise.
6974 (const_load_sequence_p): Add comment to describe the outer-most loop.
6975 (mimic_memory_attributes_and_flags): New function.
6976 (rs6000_gen_stvx): Likewise.
6977 (replace_swapped_aligned_store): Likewise.
6978 (rs6000_gen_lvx): Likewise.
6979 (replace_swapped_aligned_load): Likewise.
6980 (replace_swapped_load_constant): Capitalize argument name in
6981 comment describing this function.
6982 (rs6000_analyze_swaps): Add a third pass to search for vector loads
6983 and stores that access quad-word aligned addresses and replace
6984 with stvx or lvx instructions when appropriate.
6985 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
6986 New function prototype.
6987 (rs6000_quadword_masked_address_p): Likewise.
6988 (rs6000_gen_lvx): Likewise.
6989 (rs6000_gen_stvx): Likewise.
6990 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
6991 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
6992 when memory address is aligned.
6993 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
6994 this split to select lvx instruction when memory address is aligned.
6995 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
6996 instruction when memory address is aligned.
6997 (*vsx_le_perm_load_v16qi): Likewise.
6998 (four unnamed splitters): Modify to select the stvx instruction
6999 when memory is aligned.
7001 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7003 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7004 which appears in the queue multiple times.
7006 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7007 Alan Hayward <alan.hayward@arm.com>
7008 David Sherwood <david.sherwood@arm.com>
7010 * tree-vectorizer.h (vec_lower_bound): New structure.
7011 (_loop_vec_info): Add check_nonzero and lower_bounds.
7012 (LOOP_VINFO_CHECK_NONZERO): New macro.
7013 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7014 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7015 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7016 fields. Make seg_len the distance travelled, not including the
7018 (dr_direction_indicator): Declare.
7019 (dr_zero_step_indicator): Likewise.
7020 (dr_known_forward_stride_p): Likewise.
7021 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7023 (runtime_alias_check_p): Allow runtime alias checks with
7025 (operator ==): Compare access_size and align.
7026 (prune_runtime_alias_test_list): Rework for new distinction between
7027 the access_size and seg_len.
7028 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7030 (get_segment_min_max): New function.
7031 (create_intersect_range_checks): Use it.
7032 (dr_step_indicator): New function.
7033 (dr_direction_indicator): Likewise.
7034 (dr_zero_step_indicator): Likewise.
7035 (dr_known_forward_stride_p): Likewise.
7036 * tree-loop-distribution.c (data_ref_segment_size): Return
7037 DR_STEP * (niters - 1).
7038 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7040 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7041 (vect_preserves_scalar_order_p): New function, split out from...
7042 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7043 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7044 (vect_vfa_access_size): New function.
7045 (vect_vfa_align): Likewise.
7046 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7047 (dump_lower_bound): New function.
7048 (vect_check_lower_bound): Likewise.
7049 (vect_small_gap_p): Likewise.
7050 (vectorizable_with_step_bound_p): Likewise.
7051 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7052 depencies if the vectorization factor is 1. Convert the checks
7053 for nonzero steps into checks on the bounds of DR_STEP. Try using
7054 a bunds check for variable steps if the minimum required step is
7055 relatively small. Update calls to the dr_with_seg_len
7056 constructor and to vect_compile_time_alias.
7057 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7059 (vect_loop_versioning): Call it.
7060 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7062 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7064 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7065 Alan Hayward <alan.hayward@arm.com>
7066 David Sherwood <david.sherwood@arm.com>
7068 * doc/sourcebuild.texi (vect_scatter_store): Document.
7069 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7071 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7073 * genopinit.c (main): Add supports_vec_scatter_store and
7074 supports_vec_scatter_store_cached to target_optabs.
7075 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7076 IFN_MASK_SCATTER_STORE.
7077 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7079 * internal-fn.h (internal_store_fn_p): Declare.
7080 (internal_fn_stored_value_index): Likewise.
7081 * internal-fn.c (scatter_store_direct): New macro.
7082 (expand_scatter_store_optab_fn): New function.
7083 (direct_scatter_store_optab_supported_p): New macro.
7084 (internal_store_fn_p): New function.
7085 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7086 IFN_MASK_SCATTER_STORE.
7087 (internal_fn_mask_index): Likewise.
7088 (internal_fn_stored_value_index): New function.
7089 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7091 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7092 * optabs-query.c (supports_vec_scatter_store_p): New function.
7093 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7094 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7095 true for scatter stores.
7096 (vect_gather_scatter_fn_p): Handle scatter stores too.
7097 (vect_check_gather_scatter): Consider using scatter stores if
7098 supports_vec_scatter_store_p.
7099 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7101 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7102 internal_fn_stored_value_index.
7103 (check_load_store_masking): Handle scatter stores too.
7104 (vect_get_store_rhs): Make public.
7105 (vectorizable_call): Use internal_store_fn_p.
7106 (vectorizable_store): Handle scatter store internal functions.
7107 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7108 when deciding whether the end of the group has been reached.
7109 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7110 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7111 (mask_scatter_store<mode>): New insns.
7113 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7114 Alan Hayward <alan.hayward@arm.com>
7115 David Sherwood <david.sherwood@arm.com>
7117 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
7118 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7119 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7121 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7122 Use vect_truncate_gather_scatter_offset if we can't treat the
7123 operation as a normal gather load or scatter store.
7124 (get_group_load_store_type): Take the gather_scatter_info
7125 as argument. Try using a gather load or scatter store for
7126 single-element groups.
7127 (get_load_store_type): Update calls to get_group_load_store_type
7128 and vect_use_strided_gather_scatters_p.
7130 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7131 Alan Hayward <alan.hayward@arm.com>
7132 David Sherwood <david.sherwood@arm.com>
7134 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7135 optional tree argument.
7136 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7138 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7139 but continue to use the current value as a fallback.
7140 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7141 to compare the updates.
7142 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7143 (get_load_store_type): Use it when handling a strided access.
7144 (vect_get_strided_load_store_ops): New function.
7145 (vect_get_data_ptr_increment): Likewise.
7146 (vectorizable_load): Handle strided gather loads. Always pass
7147 a step to vect_create_data_ref_ptr and bump_vector_ptr.
7149 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7150 Alan Hayward <alan.hayward@arm.com>
7151 David Sherwood <david.sherwood@arm.com>
7153 * doc/md.texi (gather_load@var{m}): Document.
7154 (mask_gather_load@var{m}): Likewise.
7155 * genopinit.c (main): Add supports_vec_gather_load and
7156 supports_vec_gather_load_cached to target_optabs.
7157 * optabs-tree.c (init_tree_optimization_optabs): Use
7158 ggc_cleared_alloc to allocate target_optabs.
7159 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7160 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
7162 * internal-fn.h (internal_load_fn_p): Declare.
7163 (internal_gather_scatter_fn_p): Likewise.
7164 (internal_fn_mask_index): Likewise.
7165 (internal_gather_scatter_fn_supported_p): Likewise.
7166 * internal-fn.c (gather_load_direct): New macro.
7167 (expand_gather_load_optab_fn): New function.
7168 (direct_gather_load_optab_supported_p): New macro.
7169 (direct_internal_fn_optab): New function.
7170 (internal_load_fn_p): Likewise.
7171 (internal_gather_scatter_fn_p): Likewise.
7172 (internal_fn_mask_index): Likewise.
7173 (internal_gather_scatter_fn_supported_p): Likewise.
7174 * optabs-query.c (supports_at_least_one_mode_p): New function.
7175 (supports_vec_gather_load_p): Likewise.
7176 * optabs-query.h (supports_vec_gather_load_p): Declare.
7177 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
7178 and memory_type field.
7179 (NUM_PATTERNS): Bump to 15.
7180 * tree-vect-data-refs.c: Include internal-fn.h.
7181 (vect_gather_scatter_fn_p): New function.
7182 (vect_describe_gather_scatter_call): Likewise.
7183 (vect_check_gather_scatter): Try using internal functions for
7184 gather loads. Recognize existing calls to a gather load function.
7185 (vect_analyze_data_refs): Consider using gather loads if
7186 supports_vec_gather_load_p.
7187 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
7188 (vect_get_gather_scatter_offset_type): Likewise.
7189 (vect_convert_mask_for_vectype): Likewise.
7190 (vect_add_conversion_to_patterm): Likewise.
7191 (vect_try_gather_scatter_pattern): Likewise.
7192 (vect_recog_gather_scatter_pattern): New pattern recognizer.
7193 (vect_vect_recog_func_ptrs): Add it.
7194 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7195 internal_fn_mask_index and internal_gather_scatter_fn_p.
7196 (check_load_store_masking): Take the gather_scatter_info as an
7197 argument and handle gather loads.
7198 (vect_get_gather_scatter_ops): New function.
7199 (vectorizable_call): Check internal_load_fn_p.
7200 (vectorizable_load): Likewise. Handle gather load internal
7202 (vectorizable_store): Update call to check_load_store_masking.
7203 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7204 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7205 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7206 (aarch64_gather_scale_operand_d): New predicates.
7207 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7208 (mask_gather_load<mode>): New insns.
7210 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7211 Alan Hayward <alan.hayward@arm.com>
7212 David Sherwood <david.sherwood@arm.com>
7214 * optabs.def (fold_left_plus_optab): New optab.
7215 * doc/md.texi (fold_left_plus_@var{m}): Document.
7216 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7217 * internal-fn.c (fold_left_direct): Define.
7218 (expand_fold_left_optab_fn): Likewise.
7219 (direct_fold_left_optab_supported_p): Likewise.
7220 * fold-const-call.c (fold_const_fold_left): New function.
7221 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7222 * tree-parloops.c (valid_reduction_p): New function.
7223 (gather_scalar_reductions): Use it.
7224 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7225 (vect_finish_replace_stmt): Declare.
7226 * tree-vect-loop.c (fold_left_reduction_fn): New function.
7227 (needs_fold_left_reduction_p): New function, split out from...
7228 (vect_is_simple_reduction): ...here. Accept reductions that
7229 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7230 (vect_force_simple_reduction): Also store the reduction type in
7231 the assignment's STMT_VINFO_REDUC_TYPE.
7232 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7233 (merge_with_identity): New function.
7234 (vect_expand_fold_left): Likewise.
7235 (vectorize_fold_left_reduction): Likewise.
7236 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
7237 scalar phi in place for it. Check for target support and reject
7238 cases that would reassociate the operation. Defer the transform
7239 phase to vectorize_fold_left_reduction.
7240 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7241 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7242 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7244 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7246 * tree-if-conv.c (predicate_mem_writes): Remove redundant
7247 call to ifc_temp_var.
7249 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7250 Alan Hayward <alan.hayward@arm.com>
7251 David Sherwood <david.sherwood@arm.com>
7253 * target.def (legitimize_address_displacement): Take the original
7254 offset as a poly_int.
7255 * targhooks.h (default_legitimize_address_displacement): Update
7257 * targhooks.c (default_legitimize_address_displacement): Likewise.
7258 * doc/tm.texi: Regenerate.
7259 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7260 as an argument, moving assert of ad->disp == ad->disp_term to...
7261 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
7262 Try calling targetm.legitimize_address_displacement before expanding
7263 the address rather than afterwards, and adjust for the new interface.
7264 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7265 Match the new hook interface. Handle SVE addresses.
7266 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7269 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7271 * Makefile.in (OBJS): Add early-remat.o.
7272 * target.def (select_early_remat_modes): New hook.
7273 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7274 * doc/tm.texi: Regenerate.
7275 * targhooks.h (default_select_early_remat_modes): Declare.
7276 * targhooks.c (default_select_early_remat_modes): New function.
7277 * timevar.def (TV_EARLY_REMAT): New timevar.
7278 * passes.def (pass_early_remat): New pass.
7279 * tree-pass.h (make_pass_early_remat): Declare.
7280 * early-remat.c: New file.
7281 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7283 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7285 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7286 Alan Hayward <alan.hayward@arm.com>
7287 David Sherwood <david.sherwood@arm.com>
7289 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7290 vfm1 with a bound_epilog parameter.
7291 (vect_do_peeling): Update calls accordingly, and move the prologue
7292 call earlier in the function. Treat the base bound_epilog as 0 for
7293 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7294 this base when peeling for gaps.
7295 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7296 with fully-masked loops.
7297 (vect_estimate_min_profitable_iters): Handle the single peeled
7298 iteration in that case.
7300 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7301 Alan Hayward <alan.hayward@arm.com>
7302 David Sherwood <david.sherwood@arm.com>
7304 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7305 single-element interleaving even if the size is not a power of 2.
7306 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7307 accesses for single-element interleaving if the group size is
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 (fold_extract_last_@var{m}): Document.
7315 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7316 * optabs.def (fold_extract_last_optab): New optab.
7317 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7318 * internal-fn.c (fold_extract_direct): New macro.
7319 (expand_fold_extract_optab_fn): Likewise.
7320 (direct_fold_extract_optab_supported_p): Likewise.
7321 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7322 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7323 EXTRACT_LAST_REDUCTION.
7324 (get_initial_def_for_reduction): Do not create an initial vector
7325 for EXTRACT_LAST_REDUCTION reductions.
7326 (vectorizable_reduction): Leave the scalar phi in place for
7327 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7328 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7329 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7330 transform phase to vectorizable_condition.
7331 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7333 (vect_finish_stmt_generation): ...here.
7334 (vect_finish_replace_stmt): New function.
7335 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7336 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7338 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7340 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7341 Alan Hayward <alan.hayward@arm.com>
7342 David Sherwood <david.sherwood@arm.com>
7344 * doc/md.texi (extract_last_@var{m}): Document.
7345 * optabs.def (extract_last_optab): New optab.
7346 * internal-fn.def (EXTRACT_LAST): New internal function.
7347 * internal-fn.c (cond_unary_direct): New macro.
7348 (expand_cond_unary_optab_fn): Likewise.
7349 (direct_cond_unary_optab_supported_p): Likewise.
7350 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7351 loops using EXTRACT_LAST.
7352 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7353 (extract_last_<mode>): ...this optab.
7354 (vec_extract<mode><Vel>): Update accordingly.
7356 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7357 Alan Hayward <alan.hayward@arm.com>
7358 David Sherwood <david.sherwood@arm.com>
7360 * target.def (empty_mask_is_expensive): New hook.
7361 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7362 * doc/tm.texi: Regenerate.
7363 * targhooks.h (default_empty_mask_is_expensive): Declare.
7364 * targhooks.c (default_empty_mask_is_expensive): New function.
7365 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7366 if the target says that empty masks are expensive.
7367 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7369 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
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 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7376 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7377 (vect_use_loop_mask_for_alignment_p): New function.
7378 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7379 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7380 niters_skip argument. Make sure that the first niters_skip elements
7381 of the first iteration are inactive.
7382 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7383 Update call to vect_set_loop_masks_directly.
7384 (get_misalign_in_elems): New function, split out from...
7385 (vect_gen_prolog_loop_niters): ...here.
7386 (vect_update_init_of_dr): Take a code argument that specifies whether
7387 the adjustment should be added or subtracted.
7388 (vect_update_init_of_drs): Likewise.
7389 (vect_prepare_for_masked_peels): New function.
7390 (vect_do_peeling): Skip prologue peeling if we're using a mask
7391 instead. Update call to vect_update_inits_of_drs.
7392 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7394 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7395 alignment. Do not include the number of peeled iterations in
7396 the minimum threshold in that case.
7397 (vectorizable_induction): Adjust the start value down by
7398 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7399 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7400 Take the number of skipped iterations into account when calculating
7402 * tree-vect-stmts.c (vect_gen_while_not): New function.
7404 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7405 Alan Hayward <alan.hayward@arm.com>
7406 David Sherwood <david.sherwood@arm.com>
7408 * doc/sourcebuild.texi (vect_fully_masked): Document.
7409 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7411 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7413 (vect_analyze_loop_2): ...here. Don't check the vectorization
7414 factor against the number of loop iterations if the loop is
7417 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7418 Alan Hayward <alan.hayward@arm.com>
7419 David Sherwood <david.sherwood@arm.com>
7421 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
7422 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
7423 (dump_groups): Update accordingly.
7424 (iv_use::mem_type): New member variable.
7425 (address_p): New function.
7426 (record_use): Add a mem_type argument and initialize the new
7428 (record_group_use): Add a mem_type argument. Use address_p.
7429 Remove obsolete null checks of base_object. Update call to record_use.
7430 (find_interesting_uses_op): Update call to record_group_use.
7431 (find_interesting_uses_cond): Likewise.
7432 (find_interesting_uses_address): Likewise.
7433 (get_mem_type_for_internal_fn): New function.
7434 (find_address_like_use): Likewise.
7435 (find_interesting_uses_stmt): Try find_address_like_use before
7436 calling find_interesting_uses_op.
7437 (addr_offset_valid_p): Use the iv mem_type field as the type
7438 of the addressed memory.
7439 (add_autoinc_candidates): Likewise.
7440 (get_address_cost): Likewise.
7441 (split_small_address_groups_p): Use address_p.
7442 (split_address_groups): Likewise.
7443 (add_iv_candidate_for_use): Likewise.
7444 (autoinc_possible_for_pair): Likewise.
7445 (rewrite_groups): Likewise.
7446 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
7447 (determine_group_iv_cost): Update after split of USE_ADDRESS.
7448 (get_alias_ptr_type_for_ptr_address): New function.
7449 (rewrite_use_address): Rewrite address uses in calls that were
7450 identified by find_address_like_use.
7452 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7453 Alan Hayward <alan.hayward@arm.com>
7454 David Sherwood <david.sherwood@arm.com>
7456 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7458 * gimple-expr.h (is_gimple_addressable: Likewise.
7459 * gimple-expr.c (is_gimple_address): Likewise.
7460 * internal-fn.c (expand_call_mem_ref): New function.
7461 (expand_mask_load_optab_fn): Use it.
7462 (expand_mask_store_optab_fn): Likewise.
7464 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7465 Alan Hayward <alan.hayward@arm.com>
7466 David Sherwood <david.sherwood@arm.com>
7468 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7469 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7470 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7471 (cond_umax@var{mode}): Document.
7472 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
7473 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
7474 (cond_umin_optab, cond_umax_optab): New optabs.
7475 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
7476 (COND_IOR, COND_XOR): New internal functions.
7477 * internal-fn.h (get_conditional_internal_fn): Declare.
7478 * internal-fn.c (cond_binary_direct): New macro.
7479 (expand_cond_binary_optab_fn): Likewise.
7480 (direct_cond_binary_optab_supported_p): Likewise.
7481 (get_conditional_internal_fn): New function.
7482 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7483 Cope with reduction statements that are vectorized as calls rather
7485 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7486 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7487 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7488 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7489 (UNSPEC_COND_EOR): New unspecs.
7490 (optab): Add mappings for them.
7491 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7492 (sve_int_op, sve_fp_op): New int attributes.
7494 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7495 Alan Hayward <alan.hayward@arm.com>
7496 David Sherwood <david.sherwood@arm.com>
7498 * optabs.def (while_ult_optab): New optab.
7499 * doc/md.texi (while_ult@var{m}@var{n}): Document.
7500 * internal-fn.def (WHILE_ULT): New internal function.
7501 * internal-fn.h (direct_internal_fn_supported_p): New override
7502 that takes two types as argument.
7503 * internal-fn.c (while_direct): New macro.
7504 (expand_while_optab_fn): New function.
7505 (convert_optab_supported_p): Likewise.
7506 (direct_while_optab_supported_p): New macro.
7507 * wide-int.h (wi::udiv_ceil): New function.
7508 * tree-vectorizer.h (rgroup_masks): New structure.
7509 (vec_loop_masks): New typedef.
7510 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7512 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7513 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7514 (vect_max_vf): New function.
7515 (slpeel_make_loop_iterate_ntimes): Delete.
7516 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7517 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7518 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7519 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7520 internal-fn.h, stor-layout.h and optabs-query.h.
7521 (vect_set_loop_mask): New function.
7522 (add_preheader_seq): Likewise.
7523 (add_header_seq): Likewise.
7524 (interleave_supported_p): Likewise.
7525 (vect_maybe_permute_loop_masks): Likewise.
7526 (vect_set_loop_masks_directly): Likewise.
7527 (vect_set_loop_condition_masked): Likewise.
7528 (vect_set_loop_condition_unmasked): New function, split out from
7529 slpeel_make_loop_iterate_ntimes.
7530 (slpeel_make_loop_iterate_ntimes): Rename to..
7531 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7532 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7533 (vect_do_peeling): Update call accordingly.
7534 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7536 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7537 mask_compare_type, can_fully_mask_p and fully_masked_p.
7538 (release_vec_loop_masks): New function.
7539 (_loop_vec_info): Use it to free the loop masks.
7540 (can_produce_all_loop_masks_p): New function.
7541 (vect_get_max_nscalars_per_iter): Likewise.
7542 (vect_verify_full_masking): Likewise.
7543 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7544 retries, and free the mask rgroups before retrying. Check loop-wide
7545 reasons for disallowing fully-masked loops. Make the final decision
7546 about whether use a fully-masked loop or not.
7547 (vect_estimate_min_profitable_iters): Do not assume that peeling
7548 for the number of iterations will be needed for fully-masked loops.
7549 (vectorizable_reduction): Disable fully-masked loops.
7550 (vectorizable_live_operation): Likewise.
7551 (vect_halve_mask_nunits): New function.
7552 (vect_double_mask_nunits): Likewise.
7553 (vect_record_loop_mask): Likewise.
7554 (vect_get_loop_mask): Likewise.
7555 (vect_transform_loop): Handle the case in which the final loop
7556 iteration might handle a partial vector. Call vect_set_loop_condition
7557 instead of slpeel_make_loop_iterate_ntimes.
7558 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7559 (check_load_store_masking): New function.
7560 (prepare_load_store_mask): Likewise.
7561 (vectorizable_store): Handle fully-masked loops.
7562 (vectorizable_load): Likewise.
7563 (supportable_widening_operation): Use vect_halve_mask_nunits for
7565 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7566 (vect_gen_while): New function.
7567 * config/aarch64/aarch64.md (umax<mode>3): New expander.
7568 (aarch64_uqdec<mode>): New insn.
7570 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7571 Alan Hayward <alan.hayward@arm.com>
7572 David Sherwood <david.sherwood@arm.com>
7574 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7575 (reduc_xor_scal_optab): New optabs.
7576 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7577 (reduc_xor_scal_@var{m}): Document.
7578 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7579 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7581 * fold-const-call.c (fold_const_call): Handle them.
7582 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7583 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7584 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7585 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7586 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7587 (UNSPEC_XORV): New unspecs.
7588 (optab): Add entries for them.
7589 (BITWISEV): New int iterator.
7590 (bit_reduc_op): New int attributes.
7592 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7593 Alan Hayward <alan.hayward@arm.com>
7594 David Sherwood <david.sherwood@arm.com>
7596 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7597 * internal-fn.def (VEC_SHL_INSERT): New internal function.
7598 * optabs.def (vec_shl_insert_optab): New optab.
7599 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7600 (duplicate_and_interleave): Likewise.
7601 * tree-vect-loop.c: Include internal-fn.h.
7602 (neutral_op_for_slp_reduction): New function, split out from
7603 get_initial_defs_for_reduction.
7604 (get_initial_def_for_reduction): Handle option 2 for variable-length
7605 vectors by loading the neutral value into a vector and then shifting
7606 the initial value into element 0.
7607 (get_initial_defs_for_reduction): Replace the code argument with
7608 the neutral value calculated by neutral_op_for_slp_reduction.
7609 Use gimple_build_vector for constant-length vectors.
7610 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7611 but the first group_size elements have a neutral value.
7612 Use duplicate_and_interleave otherwise.
7613 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7614 Update call to get_initial_defs_for_reduction. Handle SLP
7615 reductions for variable-length vectors by creating one vector
7616 result for each scalar result, with the elements associated
7617 with other scalar results stubbed out with the neutral value.
7618 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7619 Require IFN_VEC_SHL_INSERT for double reductions on
7620 variable-length vectors, or SLP reductions that have
7621 a neutral value. Require can_duplicate_and_interleave_p
7622 support for variable-length unchained SLP reductions if there
7623 is no neutral value, such as for MIN/MAX reductions. Also require
7624 the number of vector elements to be a multiple of the number of
7625 SLP statements when doing variable-length unchained SLP reductions.
7626 Update call to vect_create_epilog_for_reduction.
7627 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
7628 and remove initial values.
7629 (duplicate_and_interleave): Make public.
7630 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
7631 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
7633 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7634 Alan Hayward <alan.hayward@arm.com>
7635 David Sherwood <david.sherwood@arm.com>
7637 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
7638 (can_duplicate_and_interleave_p): New function.
7639 (vect_get_and_check_slp_defs): Take the vector of statements
7640 rather than just the current one. Remove excess parentheses.
7641 Restriction rejectinon of vect_constant_def and vect_external_def
7642 for variable-length vectors to boolean types, or types for which
7643 can_duplicate_and_interleave_p is false.
7644 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
7645 (duplicate_and_interleave): New function.
7646 (vect_get_constant_vectors): Use gimple_build_vector for
7647 constant-length vectors and suitable variable-length constant
7648 vectors. Use duplicate_and_interleave for other variable-length
7649 vectors. Don't defer the update when inserting new statements.
7651 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7652 Alan Hayward <alan.hayward@arm.com>
7653 David Sherwood <david.sherwood@arm.com>
7655 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
7656 min_profitable_iters doesn't go negative.
7658 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7659 Alan Hayward <alan.hayward@arm.com>
7660 David Sherwood <david.sherwood@arm.com>
7662 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
7663 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
7664 * optabs.def (vec_mask_load_lanes_optab): New optab.
7665 (vec_mask_store_lanes_optab): Likewise.
7666 * internal-fn.def (MASK_LOAD_LANES): New internal function.
7667 (MASK_STORE_LANES): Likewise.
7668 * internal-fn.c (mask_load_lanes_direct): New macro.
7669 (mask_store_lanes_direct): Likewise.
7670 (expand_mask_load_optab_fn): Handle masked operations.
7671 (expand_mask_load_lanes_optab_fn): New macro.
7672 (expand_mask_store_optab_fn): Handle masked operations.
7673 (expand_mask_store_lanes_optab_fn): New macro.
7674 (direct_mask_load_lanes_optab_supported_p): Likewise.
7675 (direct_mask_store_lanes_optab_supported_p): Likewise.
7676 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
7678 (vect_load_lanes_supported): Likewise.
7679 * tree-vect-data-refs.c (strip_conversion): New function.
7680 (can_group_stmts_p): Likewise.
7681 (vect_analyze_data_ref_accesses): Use it instead of checking
7682 for a pair of assignments.
7683 (vect_store_lanes_supported): Take a masked_p parameter.
7684 (vect_load_lanes_supported): Likewise.
7685 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
7686 vect_store_lanes_supported and vect_load_lanes_supported.
7687 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7688 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
7689 parameter. Don't allow gaps for masked accesses.
7690 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
7691 and vect_load_lanes_supported.
7692 (get_load_store_type): Take a masked_p parameter and update
7693 call to get_group_load_store_type.
7694 (vectorizable_store): Update call to get_load_store_type.
7695 Handle IFN_MASK_STORE_LANES.
7696 (vectorizable_load): Update call to get_load_store_type.
7697 Handle IFN_MASK_LOAD_LANES.
7699 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7700 Alan Hayward <alan.hayward@arm.com>
7701 David Sherwood <david.sherwood@arm.com>
7703 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
7705 * config/aarch64/aarch64-protos.h
7706 (aarch64_sve_struct_memory_operand_p): Declare.
7707 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
7708 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
7709 (VPRED, vpred): Handle SVE structure modes.
7710 * config/aarch64/constraints.md (Utx): New constraint.
7711 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
7712 (aarch64_sve_struct_nonimmediate_operand): New predicates.
7713 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
7714 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
7715 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
7716 structure modes. Split into pieces after RA.
7717 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
7718 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
7720 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
7721 SVE structure modes.
7722 (aarch64_classify_address): Likewise.
7723 (sizetochar): Move earlier in file.
7724 (aarch64_print_operand): Handle SVE register lists.
7725 (aarch64_array_mode): New function.
7726 (aarch64_sve_struct_memory_operand_p): Likewise.
7727 (TARGET_ARRAY_MODE): Redefine.
7729 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7730 Alan Hayward <alan.hayward@arm.com>
7731 David Sherwood <david.sherwood@arm.com>
7733 * target.def (array_mode): New target hook.
7734 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
7735 * doc/tm.texi: Regenerate.
7736 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
7737 * hooks.c (hook_optmode_mode_uhwi_none): New function.
7738 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
7740 * stor-layout.c (mode_for_array): Likewise. Support polynomial
7743 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7744 Alan Hayward <alan.hayward@arm.com>
7745 David Sherwood <david.sherwood@arm.com>
7747 * fold-const.c (fold_binary_loc): Check the argument types
7748 rather than the result type when testing for a vector operation.
7750 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7752 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
7753 * doc/tm.texi: Regenerate.
7755 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7756 Alan Hayward <alan.hayward@arm.com>
7757 David Sherwood <david.sherwood@arm.com>
7759 * doc/invoke.texi (-msve-vector-bits=): Document new option.
7760 (sve): Document new AArch64 extension.
7761 * doc/md.texi (w): Extend the description of the AArch64
7762 constraint to include SVE vectors.
7763 (Upl, Upa): Document new AArch64 predicate constraints.
7764 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
7766 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
7767 (msve-vector-bits=): New option.
7768 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
7769 SVE when these are disabled.
7770 (sve): New extension.
7771 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
7772 modes. Adjust their number of units based on aarch64_sve_vg.
7773 (MAX_BITSIZE_MODE_ANY_MODE): Define.
7774 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
7775 aarch64_addr_query_type.
7776 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
7777 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
7778 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
7779 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
7780 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
7781 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
7782 (aarch64_simd_imm_zero_p): Delete.
7783 (aarch64_check_zero_based_sve_index_immediate): Declare.
7784 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7785 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7786 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7787 (aarch64_sve_float_mul_immediate_p): Likewise.
7788 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7790 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
7791 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
7792 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
7793 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
7794 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
7795 (aarch64_regmode_natural_size): Likewise.
7796 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
7797 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
7799 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
7800 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
7801 for VG and the SVE predicate registers.
7802 (V_ALIASES): Add a "z"-prefixed alias.
7803 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
7804 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
7805 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
7806 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
7807 (REG_CLASS_NAMES): Add entries for them.
7808 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
7809 and the predicate registers.
7810 (aarch64_sve_vg): Declare.
7811 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
7812 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
7813 (REGMODE_NATURAL_SIZE): Define.
7814 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
7816 * config/aarch64/aarch64.c: Include cfgrtl.h.
7817 (simd_immediate_info): Add a constructor for series vectors,
7818 and an associated step field.
7819 (aarch64_sve_vg): New variable.
7820 (aarch64_dbx_register_number): Handle VG and the predicate registers.
7821 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
7822 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
7823 (VEC_ANY_DATA, VEC_STRUCT): New constants.
7824 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
7825 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
7826 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
7827 (aarch64_get_mask_mode): New functions.
7828 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
7829 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
7830 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
7831 predicate modes and predicate registers. Explicitly restrict
7832 GPRs to modes of 16 bytes or smaller. Only allow FP registers
7833 to store a vector mode if it is recognized by
7834 aarch64_classify_vector_mode.
7835 (aarch64_regmode_natural_size): New function.
7836 (aarch64_hard_regno_caller_save_mode): Return the original mode
7838 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
7839 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
7840 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
7841 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
7843 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
7844 does not overlap dest if the function is frame-related. Handle
7846 (aarch64_split_add_offset): New function.
7847 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
7848 them aarch64_add_offset.
7849 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
7850 and update call to aarch64_sub_sp.
7851 (aarch64_add_cfa_expression): New function.
7852 (aarch64_expand_prologue): Pass extra temporary registers to the
7853 functions above. Handle the case in which we need to emit new
7854 DW_CFA_expressions for registers that were originally saved
7855 relative to the stack pointer, but now have to be expressed
7856 relative to the frame pointer.
7857 (aarch64_output_mi_thunk): Pass extra temporary registers to the
7859 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
7860 IP0 and IP1 values for SVE frames.
7861 (aarch64_expand_vec_series): New function.
7862 (aarch64_expand_sve_widened_duplicate): Likewise.
7863 (aarch64_expand_sve_const_vector): Likewise.
7864 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
7865 Handle SVE constants. Use emit_move_insn to move a force_const_mem
7866 into the register, rather than emitting a SET directly.
7867 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
7868 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
7869 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
7870 (offset_9bit_signed_scaled_p): New functions.
7871 (aarch64_replicate_bitmask_imm): New function.
7872 (aarch64_bitmask_imm): Use it.
7873 (aarch64_cannot_force_const_mem): Reject expressions involving
7874 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
7875 (aarch64_classify_index): Handle SVE indices, by requiring
7876 a plain register index with a scale that matches the element size.
7877 (aarch64_classify_address): Handle SVE addresses. Assert that
7878 the mode of the address is VOIDmode or an integer mode.
7879 Update call to aarch64_classify_symbol.
7880 (aarch64_classify_symbolic_expression): Update call to
7881 aarch64_classify_symbol.
7882 (aarch64_const_vec_all_in_range_p): New function.
7883 (aarch64_print_vector_float_operand): Likewise.
7884 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
7885 "vN" for FP registers with SVE modes. Handle (const ...) vectors
7886 and the FP immediates 1.0 and 0.5.
7887 (aarch64_print_address_internal): Handle SVE addresses.
7888 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
7889 (aarch64_regno_regclass): Handle predicate registers.
7890 (aarch64_secondary_reload): Handle big-endian reloads of SVE
7892 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
7893 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
7894 (aarch64_convert_sve_vector_bits): New function.
7895 (aarch64_override_options): Use it to handle -msve-vector-bits=.
7896 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7898 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
7899 Handle SVE vector and predicate modes. Accept VL-based constants
7900 that need only one temporary register, and VL offsets that require
7901 no temporary registers.
7902 (aarch64_conditional_register_usage): Mark the predicate registers
7903 as fixed if SVE isn't available.
7904 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
7905 Return true for SVE vector and predicate modes.
7906 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
7907 rather than an unsigned int. Handle SVE modes.
7908 (aarch64_preferred_simd_mode): Update call accordingly. Handle
7910 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
7912 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7913 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7914 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7915 (aarch64_sve_float_mul_immediate_p): New functions.
7916 (aarch64_sve_valid_immediate): New function.
7917 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
7918 Explicitly reject structure modes. Check for INDEX constants.
7919 Handle PTRUE and PFALSE constants.
7920 (aarch64_check_zero_based_sve_index_immediate): New function.
7921 (aarch64_simd_imm_zero_p): Delete.
7922 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
7923 vector modes. Accept constants in the range of CNT[BHWD].
7924 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
7925 ask for an Advanced SIMD mode.
7926 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
7927 (aarch64_simd_vector_alignment): Handle SVE predicates.
7928 (aarch64_vectorize_preferred_vector_alignment): New function.
7929 (aarch64_simd_vector_alignment_reachable): Use it instead of
7931 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
7932 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
7934 (MAX_VECT_LEN): Delete.
7935 (expand_vec_perm_d): Add a vec_flags field.
7936 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
7937 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
7938 (aarch64_evpc_ext): Don't apply a big-endian lane correction
7940 (aarch64_evpc_rev): Rename to...
7941 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
7942 (aarch64_evpc_rev_global): New function.
7943 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
7944 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
7946 (aarch64_evpc_sve_tbl): New function.
7947 (aarch64_expand_vec_perm_const_1): Update after rename of
7948 aarch64_evpc_rev. Handle SVE permutes too, trying
7949 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
7950 than aarch64_evpc_tbl.
7951 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
7952 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
7953 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
7954 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
7955 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
7956 (aarch64_expand_sve_vcond): New functions.
7957 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
7958 of aarch64_vector_mode_p.
7959 (aarch64_dwarf_poly_indeterminate_value): New function.
7960 (aarch64_compute_pressure_classes): Likewise.
7961 (aarch64_can_change_mode_class): Likewise.
7962 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
7963 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
7964 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
7965 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
7966 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
7967 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
7968 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
7969 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
7971 (Dn, Dl, Dr): Accept const as well as const_vector.
7972 (Dz): Likewise. Compare against CONST0_RTX.
7973 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
7974 of "vector" where appropriate.
7975 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
7976 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
7977 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
7978 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
7979 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
7980 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
7981 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
7982 (v_int_equiv): Extend to SVE modes.
7983 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
7985 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
7986 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
7987 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
7988 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
7989 (SVE_COND_FP_CMP): New int iterators.
7990 (perm_hilo): Handle the new unpack unspecs.
7991 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
7993 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
7994 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
7995 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
7996 (aarch64_equality_operator, aarch64_constant_vector_operand)
7997 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
7998 (aarch64_sve_nonimmediate_operand): Likewise.
7999 (aarch64_sve_general_operand): Likewise.
8000 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8001 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8002 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8003 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8004 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8005 (aarch64_sve_float_arith_immediate): Likewise.
8006 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8007 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8008 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8009 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8010 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8011 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8012 (aarch64_sve_float_arith_operand): Likewise.
8013 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8014 (aarch64_sve_float_mul_operand): Likewise.
8015 (aarch64_sve_vec_perm_operand): Likewise.
8016 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8017 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8018 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8019 as well as const_vector.
8020 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8021 in file. Use CONST0_RTX and CONSTM1_RTX.
8022 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8023 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8024 Use aarch64_simd_imm_zero.
8025 * config/aarch64/aarch64-sve.md: New file.
8026 * config/aarch64/aarch64.md: Include it.
8027 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8028 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8029 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8030 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8031 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8032 (sve): New attribute.
8033 (enabled): Disable instructions with the sve attribute unless
8035 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8036 aarch64_expand_mov_immediate.
8037 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8038 CNT[BHSD] immediates.
8039 (movti): Split CONST_POLY_INT moves into two halves.
8040 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8041 Split additions that need a temporary here if the destination
8042 is the stack pointer.
8043 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8044 (*add<mode>3_poly_1): New instruction.
8045 (set_clobber_cc): New expander.
8047 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8049 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8050 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8051 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8052 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8053 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8054 Change innermode from fixed_mode_size to machine_mode.
8055 (simplify_subreg): Update call accordingly. Handle a constant-sized
8056 subreg of a variable-length CONST_VECTOR.
8058 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8059 Alan Hayward <alan.hayward@arm.com>
8060 David Sherwood <david.sherwood@arm.com>
8062 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8063 (add_offset_to_base): New function, split out from...
8064 (create_mem_ref): ...here. When handling a scale other than 1,
8065 check first whether the address is valid without the offset.
8066 Add it into the base if so, leaving the index and scale as-is.
8068 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8071 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8072 fold_for_warn before checking if arg2 is INTEGER_CST.
8074 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8076 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8077 (store_multiple_operation): Delete.
8078 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8079 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8080 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8081 guarded by TARGET_STRING.
8082 (rs6000_output_load_multiple): Delete.
8083 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8084 OPTION_MASK_STRING / TARGET_STRING handling.
8085 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8086 (const rs6000_opt_masks) <"string">: Change mask to 0.
8087 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8088 (MASK_STRING): Delete.
8089 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8091 (load_multiple): Delete.
8098 (store_multiple): Delete.
8105 (movmemsi_8reg): Delete.
8106 (corresponding unnamed define_insn): Delete.
8107 (movmemsi_6reg): Delete.
8108 (corresponding unnamed define_insn): Delete.
8109 (movmemsi_4reg): Delete.
8110 (corresponding unnamed define_insn): Delete.
8111 (movmemsi_2reg): Delete.
8112 (corresponding unnamed define_insn): Delete.
8113 (movmemsi_1reg): Delete.
8114 (corresponding unnamed define_insn): Delete.
8115 * config/rs6000/rs6000.opt (mno-string): New.
8116 (mstring): Replace by deprecation warning stub.
8117 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8119 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8121 * regrename.c (regrename_do_replace): If replacing the same
8122 reg multiple times, try to reuse last created gen_raw_REG.
8125 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8126 main to workaround a bug in GDB.
8128 2018-01-12 Tom de Vries <tom@codesourcery.com>
8131 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8133 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8135 PR rtl-optimization/80481
8136 * ira-color.c (get_cap_member): New function.
8137 (allocnos_conflict_by_live_ranges_p): Use it.
8138 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8139 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8141 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8144 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8145 (*saddl_se_1): Ditto.
8147 (*ssubl_se_1): Ditto.
8149 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8151 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8152 rather than wi::to_widest for DR_INITs.
8153 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8154 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8155 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8157 (vect_analyze_group_access_1): Note that here.
8159 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8161 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
8162 polynomial type sizes.
8164 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8166 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
8167 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
8168 (gimple_add_tmp_var): Likewise.
8170 2018-01-12 Martin Liska <mliska@suse.cz>
8172 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
8173 (gimple_alloc_sizes): Likewise.
8174 (dump_gimple_statistics): Use PRIu64 in printf format.
8175 * gimple.h: Change uint64_t to int.
8177 2018-01-12 Martin Liska <mliska@suse.cz>
8179 * tree-core.h: Use uint64_t instead of int.
8180 * tree.c (tree_node_counts): Likewise.
8181 (tree_node_sizes): Likewise.
8182 (dump_tree_statistics): Use PRIu64 in printf format.
8184 2018-01-12 Martin Liska <mliska@suse.cz>
8186 * Makefile.in: As qsort_chk is implemented in vec.c, add
8187 vec.o to linkage of gencfn-macros.
8188 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
8189 passing the info to record_node_allocation_statistics.
8190 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
8192 * ggc-common.c (struct ggc_usage): Add operator== and use
8193 it in operator< and compare function.
8194 * mem-stats.h (struct mem_usage): Likewise.
8195 * vec.c (struct vec_usage): Remove operator< and compare
8196 function. Can be simply inherited.
8198 2018-01-12 Martin Jambor <mjambor@suse.cz>
8201 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8202 * tree-ssa-math-opts.c: Include domwalk.h.
8203 (convert_mult_to_fma_1): New function.
8204 (fma_transformation_info): New type.
8205 (fma_deferring_state): Likewise.
8206 (cancel_fma_deferring): New function.
8207 (result_of_phi): Likewise.
8208 (last_fma_candidate_feeds_initial_phi): Likewise.
8209 (convert_mult_to_fma): Added deferring logic, split actual
8210 transformation to convert_mult_to_fma_1.
8211 (math_opts_dom_walker): New type.
8212 (math_opts_dom_walker::after_dom_children): New method, body moved
8213 here from pass_optimize_widening_mul::execute, added deferring logic
8215 (pass_optimize_widening_mul::execute): Moved most of code to
8216 math_opts_dom_walker::after_dom_children.
8217 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8218 * config/i386/i386.c (ix86_option_override_internal): Added
8219 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8221 2018-01-12 Richard Biener <rguenther@suse.de>
8224 * dwarf2out.c (gen_variable_die): Do not reset old_die for
8225 inline instance vars.
8227 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
8230 * config/rx/rx.c (rx_is_restricted_memory_address):
8233 2018-01-12 Richard Biener <rguenther@suse.de>
8235 PR tree-optimization/80846
8236 * target.def (split_reduction): New target hook.
8237 * targhooks.c (default_split_reduction): New function.
8238 * targhooks.h (default_split_reduction): Declare.
8239 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8240 target requests first reduce vectors by combining low and high
8242 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8243 (get_vectype_for_scalar_type_and_size): Export.
8244 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8245 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8246 * doc/tm.texi: Regenerate.
8247 * config/i386/i386.c (ix86_split_reduction): Implement
8248 TARGET_VECTORIZE_SPLIT_REDUCTION.
8250 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8253 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8254 in PIC mode except for TARGET_VXWORKS_RTP.
8255 * config/sparc/sparc.c: Include cfgrtl.h.
8256 (TARGET_INIT_PIC_REG): Define.
8257 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8258 (sparc_pic_register_p): New predicate.
8259 (sparc_legitimate_address_p): Use it.
8260 (sparc_legitimize_pic_address): Likewise.
8261 (sparc_delegitimize_address): Likewise.
8262 (sparc_mode_dependent_address_p): Likewise.
8263 (gen_load_pcrel_sym): Remove 4th parameter.
8264 (load_got_register): Adjust call to above. Remove obsolete stuff.
8265 (sparc_expand_prologue): Do not call load_got_register here.
8266 (sparc_flat_expand_prologue): Likewise.
8267 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8268 (sparc_use_pseudo_pic_reg): New function.
8269 (sparc_init_pic_reg): Likewise.
8270 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8271 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8273 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8275 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8276 Add item for branch_cost.
8278 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8280 PR rtl-optimization/83565
8281 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8282 not extend the result to a larger mode for rotate operations.
8283 (num_sign_bit_copies1): Likewise.
8285 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8288 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8290 Use values-Xc.o for -pedantic.
8291 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8293 2018-01-12 Martin Liska <mliska@suse.cz>
8296 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8298 (possible_polymorphic_call_targets): Use it.
8299 (ipa_devirt): Likewise.
8301 2018-01-12 Martin Liska <mliska@suse.cz>
8303 * profile-count.h (enum profile_quality): Use 0 as invalid
8304 enum value of profile_quality.
8306 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8308 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8309 -mext-string options.
8311 2018-01-12 Richard Biener <rguenther@suse.de>
8313 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8314 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8315 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8317 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8319 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8321 * configure.ac (--with-long-double-format): Add support for the
8322 configuration option to change the default long double format on
8324 * config.gcc (powerpc*-linux*-*): Likewise.
8325 * configure: Regenerate.
8326 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8327 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8328 used without modification.
8330 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8332 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8333 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8334 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8335 MISC_BUILTIN_SPEC_BARRIER.
8336 (rs6000_init_builtins): Likewise.
8337 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8339 (speculation_barrier): New define_insn.
8340 * doc/extend.texi: Document __builtin_speculation_barrier.
8342 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8345 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8346 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8347 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8349 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8350 integral modes instead of "ss" and "sd".
8351 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8352 vectors with 32-bit and 64-bit elements.
8353 (vecdupssescalarmodesuffix): New mode attribute.
8354 (vec_dup<mode>): Use it.
8356 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8359 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8360 frame if argument is passed on stack.
8362 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8365 * ree.c (combine_reaching_defs): Optimize also
8366 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8367 reg2=any_extend(exp); reg1=reg2;, formatting fix.
8369 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8372 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8374 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8377 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8378 after they are computed.
8380 2018-01-11 Bin Cheng <bin.cheng@arm.com>
8382 PR tree-optimization/83695
8383 * gimple-loop-linterchange.cc
8384 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8385 reset cached scev information after interchange.
8386 (pass_linterchange::execute): Remove call to scev_reset_htab.
8388 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8390 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8391 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8392 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8393 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8394 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8395 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8396 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8397 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8398 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8399 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8400 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8401 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8402 (V_lane_reg): Likewise.
8403 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8405 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8406 (vfmal_lane_low<mode>_intrinsic,
8407 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8408 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8409 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8410 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8411 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8412 vfmsl_lane_high<mode>_intrinsic): New define_insns.
8414 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8416 * config/arm/arm-cpus.in (fp16fml): New feature.
8417 (ALL_SIMD): Add fp16fml.
8418 (armv8.2-a): Add fp16fml as an option.
8419 (armv8.3-a): Likewise.
8420 (armv8.4-a): Add fp16fml as part of fp16.
8421 * config/arm/arm.h (TARGET_FP16FML): Define.
8422 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
8424 * config/arm/arm-modes.def (V2HF): Define.
8425 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
8426 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
8427 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
8428 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
8429 vfmsl_low, vfmsl_high): New set of builtins.
8430 * config/arm/iterators.md (PLUSMINUS): New code iterator.
8431 (vfml_op): New code attribute.
8432 (VFMLHALVES): New int iterator.
8433 (VFML, VFMLSEL): New mode attributes.
8434 (V_reg): Define mapping for V2HF.
8435 (V_hi, V_lo): New mode attributes.
8436 (VF_constraint): Likewise.
8437 (vfml_half, vfml_half_selector): New int attributes.
8438 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
8440 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
8441 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
8443 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
8444 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
8445 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
8446 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
8448 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8449 Document new effective target and option set.
8451 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8453 * config/arm/arm-cpus.in (armv8_4): New feature.
8454 (ARMv8_4a): New fgroup.
8455 (armv8.4-a): New arch.
8456 * config/arm/arm-tables.opt: Regenerate.
8457 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8458 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8459 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8460 Add matching rules for -march=armv8.4-a and extensions.
8461 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8463 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
8466 * config/rx/rx.md (BW): New mode attribute.
8467 (sync_lock_test_and_setsi): Add mode suffix to insn output.
8469 2018-01-11 Richard Biener <rguenther@suse.de>
8471 PR tree-optimization/83435
8472 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
8473 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
8474 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
8476 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8477 Alan Hayward <alan.hayward@arm.com>
8478 David Sherwood <david.sherwood@arm.com>
8480 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8482 (aarch64_classify_address): Initialize it. Track polynomial offsets.
8483 (aarch64_print_address_internal): Use it to check for a zero offset.
8485 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8486 Alan Hayward <alan.hayward@arm.com>
8487 David Sherwood <david.sherwood@arm.com>
8489 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8490 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8491 Return a poly_int64 rather than a HOST_WIDE_INT.
8492 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8493 rather than a HOST_WIDE_INT.
8494 * config/aarch64/aarch64.h (aarch64_frame): Protect with
8495 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
8496 hard_fp_offset, frame_size, initial_adjust, callee_offset and
8497 final_offset from HOST_WIDE_INT to poly_int64.
8498 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8499 to_constant when getting the number of units in an Advanced SIMD
8501 (aarch64_builtin_vectorized_function): Check for a constant number
8503 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8505 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8506 attribute instead of GET_MODE_NUNITS.
8507 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8508 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8509 GET_MODE_SIZE for fixed-size registers.
8510 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8511 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8512 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8513 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8514 (aarch64_print_operand, aarch64_print_address_internal)
8515 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8516 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8517 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8518 Handle polynomial GET_MODE_SIZE.
8519 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8520 wider than SImode without modification.
8521 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8522 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8523 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8524 passing and returning SVE modes.
8525 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8526 rather than GEN_INT.
8527 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8528 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8529 (aarch64_allocate_and_probe_stack_space): Likewise.
8530 (aarch64_layout_frame): Cope with polynomial offsets.
8531 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8532 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8534 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8535 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8536 poly_int64 rather than a HOST_WIDE_INT.
8537 (aarch64_get_separate_components, aarch64_process_components)
8538 (aarch64_expand_prologue, aarch64_expand_epilogue)
8539 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8540 (aarch64_anchor_offset): New function, split out from...
8541 (aarch64_legitimize_address): ...here.
8542 (aarch64_builtin_vectorization_cost): Handle polynomial
8543 TYPE_VECTOR_SUBPARTS.
8544 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8546 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8547 number of elements from the PARALLEL rather than the mode.
8548 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8549 rather than GET_MODE_BITSIZE.
8550 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8551 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8552 (aarch64_expand_vec_perm_const_1): Handle polynomial
8553 d->perm.length () and d->perm elements.
8554 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
8555 Apply to_constant to d->perm elements.
8556 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8557 polynomial CONST_VECTOR_NUNITS.
8558 (aarch64_move_pointer): Take amount as a poly_int64 rather
8560 (aarch64_progress_pointer): Avoid temporary variable.
8561 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8562 the mode attribute instead of GET_MODE.
8564 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8565 Alan Hayward <alan.hayward@arm.com>
8566 David Sherwood <david.sherwood@arm.com>
8568 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8569 x exists before using it.
8570 (aarch64_add_constant_internal): Rename to...
8571 (aarch64_add_offset_1): ...this. Replace regnum with separate
8572 src and dest rtxes. Handle the case in which they're different,
8573 including when the offset is zero. Replace scratchreg with an rtx.
8574 Use 2 additions if there is no spare register into which we can
8575 move a 16-bit constant.
8576 (aarch64_add_constant): Delete.
8577 (aarch64_add_offset): Replace reg with separate src and dest
8578 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8579 Use aarch64_add_offset_1.
8580 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8581 an rtx rather than an int. Take the delta as a poly_int64
8582 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8583 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8584 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8585 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8586 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8588 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8589 aarch64_add_constant.
8591 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8593 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8594 Use scalar_float_mode.
8596 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8598 * config/aarch64/aarch64-simd.md
8599 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8600 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8601 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8602 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8603 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8604 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8605 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8606 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8607 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8608 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8610 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8613 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8614 targ_options->x_arm_arch_string is non NULL.
8616 2018-01-11 Tamar Christina <tamar.christina@arm.com>
8618 * config/aarch64/aarch64.h
8619 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
8621 2018-01-11 Sudakshina Das <sudi.das@arm.com>
8624 * expmed.c (emit_store_flag_force): Swap if const op0
8625 and change VOIDmode to mode of op0.
8627 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8629 PR rtl-optimization/83761
8630 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
8631 than bytes to mode_for_size.
8633 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8636 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
8637 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
8640 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8643 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
8644 when in layout mode.
8645 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
8646 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
8649 2018-01-10 Michael Collison <michael.collison@arm.com>
8651 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
8652 * config/aarch64/aarch64-option-extension.def: Add
8653 AARCH64_OPT_EXTENSION of 'fp16fml'.
8654 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8655 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
8656 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
8657 * config/aarch64/constraints.md (Ui7): New constraint.
8658 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
8659 (VFMLA_SEL_W): Ditto.
8662 (VFMLA16_LOW): New int iterator.
8663 (VFMLA16_HIGH): Ditto.
8664 (UNSPEC_FMLAL): New unspec.
8665 (UNSPEC_FMLSL): Ditto.
8666 (UNSPEC_FMLAL2): Ditto.
8667 (UNSPEC_FMLSL2): Ditto.
8668 (f16mac): New code attribute.
8669 * config/aarch64/aarch64-simd-builtins.def
8670 (aarch64_fmlal_lowv2sf): Ditto.
8671 (aarch64_fmlsl_lowv2sf): Ditto.
8672 (aarch64_fmlalq_lowv4sf): Ditto.
8673 (aarch64_fmlslq_lowv4sf): Ditto.
8674 (aarch64_fmlal_highv2sf): Ditto.
8675 (aarch64_fmlsl_highv2sf): Ditto.
8676 (aarch64_fmlalq_highv4sf): Ditto.
8677 (aarch64_fmlslq_highv4sf): Ditto.
8678 (aarch64_fmlal_lane_lowv2sf): Ditto.
8679 (aarch64_fmlsl_lane_lowv2sf): Ditto.
8680 (aarch64_fmlal_laneq_lowv2sf): Ditto.
8681 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
8682 (aarch64_fmlalq_lane_lowv4sf): Ditto.
8683 (aarch64_fmlsl_lane_lowv4sf): Ditto.
8684 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
8685 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
8686 (aarch64_fmlal_lane_highv2sf): Ditto.
8687 (aarch64_fmlsl_lane_highv2sf): Ditto.
8688 (aarch64_fmlal_laneq_highv2sf): Ditto.
8689 (aarch64_fmlsl_laneq_highv2sf): Ditto.
8690 (aarch64_fmlalq_lane_highv4sf): Ditto.
8691 (aarch64_fmlsl_lane_highv4sf): Ditto.
8692 (aarch64_fmlalq_laneq_highv4sf): Ditto.
8693 (aarch64_fmlsl_laneq_highv4sf): Ditto.
8694 * config/aarch64/aarch64-simd.md:
8695 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
8696 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8697 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
8698 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8699 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
8700 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
8701 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
8702 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
8703 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
8704 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
8705 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
8706 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
8707 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
8708 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
8709 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
8710 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
8711 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
8712 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
8713 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
8714 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
8715 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
8716 (vfmlsl_low_u32): Ditto.
8717 (vfmlalq_low_u32): Ditto.
8718 (vfmlslq_low_u32): Ditto.
8719 (vfmlal_high_u32): Ditto.
8720 (vfmlsl_high_u32): Ditto.
8721 (vfmlalq_high_u32): Ditto.
8722 (vfmlslq_high_u32): Ditto.
8723 (vfmlal_lane_low_u32): Ditto.
8724 (vfmlsl_lane_low_u32): Ditto.
8725 (vfmlal_laneq_low_u32): Ditto.
8726 (vfmlsl_laneq_low_u32): Ditto.
8727 (vfmlalq_lane_low_u32): Ditto.
8728 (vfmlslq_lane_low_u32): Ditto.
8729 (vfmlalq_laneq_low_u32): Ditto.
8730 (vfmlslq_laneq_low_u32): Ditto.
8731 (vfmlal_lane_high_u32): Ditto.
8732 (vfmlsl_lane_high_u32): Ditto.
8733 (vfmlal_laneq_high_u32): Ditto.
8734 (vfmlsl_laneq_high_u32): Ditto.
8735 (vfmlalq_lane_high_u32): Ditto.
8736 (vfmlslq_lane_high_u32): Ditto.
8737 (vfmlalq_laneq_high_u32): Ditto.
8738 (vfmlslq_laneq_high_u32): Ditto.
8739 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
8740 (AARCH64_FL_FOR_ARCH8_4): New.
8741 (AARCH64_ISA_F16FML): New ISA flag.
8742 (TARGET_F16FML): New feature flag for fp16fml.
8743 (doc/invoke.texi): Document new fp16fml option.
8745 2018-01-10 Michael Collison <michael.collison@arm.com>
8747 * config/aarch64/aarch64-builtins.c:
8748 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
8749 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8750 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
8751 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
8752 (AARCH64_ISA_SHA3): New ISA flag.
8753 (TARGET_SHA3): New feature flag for sha3.
8754 * config/aarch64/iterators.md (sha512_op): New int attribute.
8755 (CRYPTO_SHA512): New int iterator.
8756 (UNSPEC_SHA512H): New unspec.
8757 (UNSPEC_SHA512H2): Ditto.
8758 (UNSPEC_SHA512SU0): Ditto.
8759 (UNSPEC_SHA512SU1): Ditto.
8760 * config/aarch64/aarch64-simd-builtins.def
8761 (aarch64_crypto_sha512hqv2di): New builtin.
8762 (aarch64_crypto_sha512h2qv2di): Ditto.
8763 (aarch64_crypto_sha512su0qv2di): Ditto.
8764 (aarch64_crypto_sha512su1qv2di): Ditto.
8765 (aarch64_eor3qv8hi): Ditto.
8766 (aarch64_rax1qv2di): Ditto.
8767 (aarch64_xarqv2di): Ditto.
8768 (aarch64_bcaxqv8hi): Ditto.
8769 * config/aarch64/aarch64-simd.md:
8770 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
8771 (aarch64_crypto_sha512su0qv2di): Ditto.
8772 (aarch64_crypto_sha512su1qv2di): Ditto.
8773 (aarch64_eor3qv8hi): Ditto.
8774 (aarch64_rax1qv2di): Ditto.
8775 (aarch64_xarqv2di): Ditto.
8776 (aarch64_bcaxqv8hi): Ditto.
8777 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
8778 (vsha512h2q_u64): Ditto.
8779 (vsha512su0q_u64): Ditto.
8780 (vsha512su1q_u64): Ditto.
8781 (veor3q_u16): Ditto.
8782 (vrax1q_u64): Ditto.
8784 (vbcaxq_u16): Ditto.
8785 * config/arm/types.md (crypto_sha512): New type attribute.
8786 (crypto_sha3): Ditto.
8787 (doc/invoke.texi): Document new sha3 option.
8789 2018-01-10 Michael Collison <michael.collison@arm.com>
8791 * config/aarch64/aarch64-builtins.c:
8792 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
8793 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8794 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
8795 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
8796 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
8797 (AARCH64_ISA_SM4): New ISA flag.
8798 (TARGET_SM4): New feature flag for sm4.
8799 * config/aarch64/aarch64-simd-builtins.def
8800 (aarch64_sm3ss1qv4si): Ditto.
8801 (aarch64_sm3tt1aq4si): Ditto.
8802 (aarch64_sm3tt1bq4si): Ditto.
8803 (aarch64_sm3tt2aq4si): Ditto.
8804 (aarch64_sm3tt2bq4si): Ditto.
8805 (aarch64_sm3partw1qv4si): Ditto.
8806 (aarch64_sm3partw2qv4si): Ditto.
8807 (aarch64_sm4eqv4si): Ditto.
8808 (aarch64_sm4ekeyqv4si): Ditto.
8809 * config/aarch64/aarch64-simd.md:
8810 (aarch64_sm3ss1qv4si): Ditto.
8811 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
8812 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
8813 (aarch64_sm4eqv4si): Ditto.
8814 (aarch64_sm4ekeyqv4si): Ditto.
8815 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
8816 (sm3part_op): Ditto.
8817 (CRYPTO_SM3TT): Ditto.
8818 (CRYPTO_SM3PART): Ditto.
8819 (UNSPEC_SM3SS1): New unspec.
8820 (UNSPEC_SM3TT1A): Ditto.
8821 (UNSPEC_SM3TT1B): Ditto.
8822 (UNSPEC_SM3TT2A): Ditto.
8823 (UNSPEC_SM3TT2B): Ditto.
8824 (UNSPEC_SM3PARTW1): Ditto.
8825 (UNSPEC_SM3PARTW2): Ditto.
8826 (UNSPEC_SM4E): Ditto.
8827 (UNSPEC_SM4EKEY): Ditto.
8828 * config/aarch64/constraints.md (Ui2): New constraint.
8829 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
8830 * config/arm/types.md (crypto_sm3): New type attribute.
8831 (crypto_sm4): Ditto.
8832 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
8833 (vsm3tt1aq_u32): Ditto.
8834 (vsm3tt1bq_u32): Ditto.
8835 (vsm3tt2aq_u32): Ditto.
8836 (vsm3tt2bq_u32): Ditto.
8837 (vsm3partw1q_u32): Ditto.
8838 (vsm3partw2q_u32): Ditto.
8839 (vsm4eq_u32): Ditto.
8840 (vsm4ekeyq_u32): Ditto.
8841 (doc/invoke.texi): Document new sm4 option.
8843 2018-01-10 Michael Collison <michael.collison@arm.com>
8845 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
8846 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
8847 (AARCH64_FL_FOR_ARCH8_4): New.
8848 (AARCH64_FL_V8_4): New flag.
8849 (doc/invoke.texi): Document new armv8.4-a option.
8851 2018-01-10 Michael Collison <michael.collison@arm.com>
8853 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8854 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
8855 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
8856 * config/aarch64/aarch64-option-extension.def: Add
8857 AARCH64_OPT_EXTENSION of 'sha2'.
8858 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
8859 (crypto): Disable sha2 and aes if crypto disabled.
8860 (crypto): Enable aes and sha2 if enabled.
8861 (simd): Disable sha2 and aes if simd disabled.
8862 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
8864 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
8865 (TARGET_SHA2): New feature flag for sha2.
8866 (TARGET_AES): New feature flag for aes.
8867 * config/aarch64/aarch64-simd.md:
8868 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
8869 conditional on TARGET_AES.
8870 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
8871 (aarch64_crypto_sha1hsi): Make pattern conditional
8873 (aarch64_crypto_sha1hv4si): Ditto.
8874 (aarch64_be_crypto_sha1hv4si): Ditto.
8875 (aarch64_crypto_sha1su1v4si): Ditto.
8876 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
8877 (aarch64_crypto_sha1su0v4si): Ditto.
8878 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
8879 (aarch64_crypto_sha256su0v4si): Ditto.
8880 (aarch64_crypto_sha256su1v4si): Ditto.
8881 (doc/invoke.texi): Document new aes and sha2 options.
8883 2018-01-10 Martin Sebor <msebor@redhat.com>
8885 PR tree-optimization/83781
8886 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
8889 2018-01-11 Martin Sebor <msebor@gmail.com>
8890 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8892 PR tree-optimization/83501
8893 PR tree-optimization/81703
8895 * tree-ssa-strlen.c (get_string_cst): Rename...
8896 (get_string_len): ...to this. Handle global constants.
8897 (handle_char_store): Adjust.
8899 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
8900 Jim Wilson <jimw@sifive.com>
8902 * config/riscv/riscv-protos.h (riscv_output_return): New.
8903 * config/riscv/riscv.c (struct machine_function): New naked_p field.
8904 (riscv_attribute_table, riscv_output_return),
8905 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
8906 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
8907 (riscv_compute_frame_info): Only compute frame->mask if not a naked
8909 (riscv_expand_prologue): Add early return for naked function.
8910 (riscv_expand_epilogue): Likewise.
8911 (riscv_function_ok_for_sibcall): Return false for naked function.
8912 (riscv_set_current_function): New.
8913 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
8914 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
8915 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
8916 * doc/extend.texi (RISC-V Function Attributes): New.
8918 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8920 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
8921 check for 128-bit long double before checking TCmode.
8922 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
8923 128-bit long doubles before checking TFmode or TCmode.
8924 (FLOAT128_IBM_P): Likewise.
8926 2018-01-10 Martin Sebor <msebor@redhat.com>
8928 PR tree-optimization/83671
8929 * builtins.c (c_strlen): Unconditionally return zero for the empty
8931 Use -Warray-bounds for warnings.
8932 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
8933 for non-constant array indices with COMPONENT_REF, arrays of
8934 arrays, and pointers to arrays.
8935 (gimple_fold_builtin_strlen): Determine and set length range for
8936 non-constant character arrays.
8938 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
8941 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
8944 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
8946 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
8948 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8951 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
8952 VECTOR_MEM_ALTIVEC_OR_VSX_P.
8953 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
8954 indexed_or_indirect_operand predicate.
8955 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
8956 (*vsx_le_perm_load_v8hi): Likewise.
8957 (*vsx_le_perm_load_v16qi): Likewise.
8958 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
8959 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
8960 (*vsx_le_perm_store_v8hi): Likewise.
8961 (*vsx_le_perm_store_v16qi): Likewise.
8962 (eight unnamed splitters): Likewise.
8964 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8966 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
8967 * config/rs6000/emmintrin.h: Likewise.
8968 * config/rs6000/mmintrin.h: Likewise.
8969 * config/rs6000/xmmintrin.h: Likewise.
8971 2018-01-10 David Malcolm <dmalcolm@redhat.com>
8974 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
8976 * tree.c (tree_nop_conversion): Return true for location wrapper
8978 (maybe_wrap_with_location): New function.
8979 (selftest::check_strip_nops): New function.
8980 (selftest::test_location_wrappers): New function.
8981 (selftest::tree_c_tests): Call it.
8982 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
8983 (maybe_wrap_with_location): New decl.
8984 (EXPR_LOCATION_WRAPPER_P): New macro.
8985 (location_wrapper_p): New inline function.
8986 (tree_strip_any_location_wrapper): New inline function.
8988 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
8991 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
8992 stack_realign_offset for the largest alignment of stack slot
8994 (ix86_find_max_used_stack_alignment): New function.
8995 (ix86_finalize_stack_frame_flags): Use it. Set
8996 max_used_stack_alignment if we don't realign stack.
8997 * config/i386/i386.h (machine_function): Add
8998 max_used_stack_alignment.
9000 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9002 * config/arm/arm.opt (-mbranch-cost): New option.
9003 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9006 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9009 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9010 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9012 2018-01-10 Richard Biener <rguenther@suse.de>
9015 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9016 early out so it also covers the case where we have a non-NULL
9019 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9021 PR tree-optimization/83753
9022 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9023 for non-strided grouped accesses if the number of elements is 1.
9025 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9028 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9029 * i386.h (TARGET_USE_GATHER): Define.
9030 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9032 2018-01-10 Martin Liska <mliska@suse.cz>
9035 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9036 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9038 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9039 CLEANUP_NO_PARTITIONING is not set.
9041 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9043 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9044 for vectors, as a partial revert of r254296.
9045 * rtl.h (const_vec_p): Delete.
9046 (const_vec_duplicate_p): Don't test for vector CONSTs.
9047 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9048 * expmed.c (make_tree): Likewise.
9051 * common.md (E, F): Use CONSTANT_P instead of checking for
9053 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9054 checking for CONST_VECTOR.
9056 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9059 * predict.c (force_edge_cold): Handle in more sane way edges
9062 2018-01-09 Carl Love <cel@us.ibm.com>
9064 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9066 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9067 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9068 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9069 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9070 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9071 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9072 * config/rs6000/rs6000-protos.h: Add extern defition for
9073 rs6000_generate_float2_double_code.
9074 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9076 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9077 (float2_v2df): Add define_expand.
9079 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9082 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9083 op_mode in the force_to_mode call.
9085 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9087 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9088 instead of checking each element individually.
9089 (aarch64_evpc_uzp): Likewise.
9090 (aarch64_evpc_zip): Likewise.
9091 (aarch64_evpc_ext): Likewise.
9092 (aarch64_evpc_rev): Likewise.
9093 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9094 instead of checking each element individually. Return true without
9096 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9097 whether all selected elements come from the same input, instead of
9098 checking each element individually. Remove calls to gen_rtx_REG,
9099 start_sequence and end_sequence and instead assert that no rtl is
9102 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9104 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9105 order of HIGH and CONST checks.
9107 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9109 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9110 if the destination isn't an SSA_NAME.
9112 2018-01-09 Richard Biener <rguenther@suse.de>
9114 PR tree-optimization/83668
9115 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9117 (canonicalize_loop_form): ... here, renamed from ...
9118 (canonicalize_loop_closed_ssa_form): ... this and amended to
9119 swap successor edges for loop exit blocks to make us use
9120 the RPO order we need for initial schedule generation.
9122 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9124 PR tree-optimization/64811
9125 * match.pd: When optimizing comparisons with Inf, avoid
9126 introducing or losing exceptions from comparisons with NaN.
9128 2018-01-09 Martin Liska <mliska@suse.cz>
9131 * asan.c (shadow_mem_size): Add gcc_assert.
9133 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9135 Don't save registers in main().
9138 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9139 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9140 * config/avr/avr.c (avr_set_current_function): Don't error if
9141 naked, OS_task or OS_main are specified at the same time.
9142 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9144 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9146 * common/config/avr/avr-common.c (avr_option_optimization_table):
9147 Switch on -mmain-is-OS_task for optimizing compilations.
9149 2018-01-09 Richard Biener <rguenther@suse.de>
9151 PR tree-optimization/83572
9152 * graphite.c: Include cfganal.h.
9153 (graphite_transform_loops): Connect infinite loops to exit
9154 and remove fake edges at the end.
9156 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9158 * ipa-inline.c (edge_badness): Revert accidental checkin.
9160 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9163 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
9164 symbols; not inline clones.
9166 2018-01-09 Jakub Jelinek <jakub@redhat.com>
9169 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
9170 hard registers. Formatting fixes.
9172 PR preprocessor/83722
9173 * gcc.c (try_generate_repro): Pass
9174 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
9175 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
9178 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
9179 Kito Cheng <kito.cheng@gmail.com>
9181 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
9182 (riscv_leaf_function_p): Delete.
9183 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
9185 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9187 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
9189 (do_ifelse): New function.
9190 (do_isel): New function.
9191 (do_sub3): New function.
9192 (do_add3): New function.
9193 (do_load_mask_compare): New function.
9194 (do_overlap_load_compare): New function.
9195 (expand_compare_loop): New function.
9196 (expand_block_compare): Call expand_compare_loop() when appropriate.
9197 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9199 (-mblock-compare-inline-loop-limit): New option.
9201 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9204 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9205 Reverse order of second and third operands in first alternative.
9206 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9207 of first and second elements in UNSPEC_VPERMR vector.
9208 (altivec_expand_vec_perm_le): Likewise.
9210 2018-01-08 Jeff Law <law@redhat.com>
9212 PR rtl-optimizatin/81308
9213 * tree-switch-conversion.c (cfg_altered): New file scoped static.
9214 (process_switch): If group_case_labels makes a change, then set
9216 (pass_convert_switch::execute): If a switch is converted, then
9217 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
9219 PR rtl-optimization/81308
9220 * recog.c (split_all_insns): Conditionally cleanup the CFG after
9223 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
9225 PR target/83663 - Revert r255946
9226 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9227 generation for cases where splatting a value is not useful.
9228 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9229 across a vec_duplicate and a paradoxical subreg forming a vector
9230 mode to a vec_concat.
9232 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9234 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9235 -march=armv8.3-a variants.
9236 * config/arm/t-multilib: Likewise.
9237 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
9239 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9241 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9243 (cceq_ior_compare_complement): Give it a name so I can use it, and
9244 change boolean_or_operator predicate to boolean_operator so it can
9245 be used to generate a crand.
9246 (eqne): New code iterator.
9247 (bd/bd_neg): New code_attrs.
9248 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9249 a single define_insn.
9250 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9251 decrement (bdnzt/bdnzf/bdzt/bdzf).
9252 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9253 with the new names of the branch decrement patterns, and added the
9254 names of the branch decrement conditional patterns.
9256 2018-01-08 Richard Biener <rguenther@suse.de>
9258 PR tree-optimization/83563
9259 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9262 2018-01-08 Richard Biener <rguenther@suse.de>
9265 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9267 2018-01-08 Richard Biener <rguenther@suse.de>
9269 PR tree-optimization/83685
9270 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9271 references to abnormals.
9273 2018-01-08 Richard Biener <rguenther@suse.de>
9276 * dwarf2out.c (output_indirect_strings): Handle empty
9277 skeleton_debug_str_hash.
9278 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9280 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9282 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9283 (emit_store_direct): Likewise.
9284 (arc_trampoline_adjust_address): Likewise.
9285 (arc_asm_trampoline_template): New function.
9286 (arc_initialize_trampoline): Use asm_trampoline_template.
9287 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9288 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9289 * config/arc/arc.md (flush_icache): Delete pattern.
9291 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9293 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9294 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9297 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9300 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9301 by not USED_FOR_TARGET.
9302 (make_pass_resolve_sw_modes): Likewise.
9304 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9306 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9309 2018-01-08 Richard Biener <rguenther@suse.de>
9312 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9314 2018-01-08 Richard Biener <rguenther@suse.de>
9317 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9319 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9322 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9323 basic blocks with a small number of successors.
9324 (convert_control_dep_chain_into_preds): Improve handling of
9326 (dump_predicates): Split apart into...
9327 (dump_pred_chain): ...here...
9328 (dump_pred_info): ...and here.
9329 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9330 (can_chain_union_be_invalidated_p): Improve check for invalidation
9332 (uninit_uses_cannot_happen): Avoid unnecessary if
9333 convert_control_dep_chain_into_preds yielded nothing.
9335 2018-01-06 Martin Sebor <msebor@redhat.com>
9337 PR tree-optimization/83640
9338 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9339 subtracting negative offset from size.
9340 (builtin_access::overlap): Adjust offset bounds of the access to fall
9341 within the size of the object if possible.
9343 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9345 PR rtl-optimization/83699
9346 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9347 extract_bit_field_as_subreg to cases in which the extracted
9348 value is also a vector.
9350 * lra-constraints.c (process_alt_operands): Test for the equivalence
9351 substitutions when detecting a possible reload cycle.
9353 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9356 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9357 by default if flag_selective_schedling{,2}. Formatting fixes.
9359 PR rtl-optimization/83682
9360 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9361 if it has non-VECTOR_MODE element mode.
9362 (vec_duplicate_p): Likewise.
9365 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9366 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9368 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9371 * config/i386/i386-builtin.def
9372 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9373 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9374 Require also OPTION_MASK_ISA_AVX512F in addition to
9375 OPTION_MASK_ISA_GFNI.
9376 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9377 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9378 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9379 to OPTION_MASK_ISA_GFNI.
9380 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9381 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9382 OPTION_MASK_ISA_AVX512BW.
9383 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9384 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9385 addition to OPTION_MASK_ISA_GFNI.
9386 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9387 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9388 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9389 to OPTION_MASK_ISA_GFNI.
9390 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9391 a requirement for all ISAs rather than any of them with a few
9393 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9395 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9396 bitmasks to be enabled with 3 exceptions, instead of requiring any
9397 enabled ISA with lots of exceptions.
9398 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9399 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9400 Change avx512bw in isa attribute to avx512f.
9401 * config/i386/sgxintrin.h: Add license boilerplate.
9402 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9403 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9404 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9405 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9407 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9408 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9409 temporarily sse2 rather than sse if not enabled already.
9412 * config/i386/sse.md (VI248_VLBW): Rename to ...
9413 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
9414 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9415 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9416 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9417 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9418 mode iterator instead of VI248_VLBW.
9420 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
9422 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
9423 (record_modified): Skip clobbers; add debug output.
9424 (param_change_prob): Use sreal frequencies.
9426 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9428 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
9429 punt for user-aligned variables.
9431 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9433 * tree-chrec.c (chrec_contains_symbols): Return true for
9436 2018-01-05 Sudakshina Das <sudi.das@arm.com>
9439 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
9440 of (x|y) == x for BICS pattern.
9442 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9444 PR tree-optimization/83605
9445 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
9446 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9449 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
9451 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9452 * config/epiphany/rtems.h: New file.
9454 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9455 Uros Bizjak <ubizjak@gmail.com>
9458 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9459 QIreg_operand instead of register_operand predicate.
9460 * config/i386/i386.c (ix86_rop_should_change_byte_p,
9461 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9462 comments instead of -fmitigate[-_]rop.
9464 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9467 * cgraphunit.c (symbol_table::compile): Switch to text_section
9468 before calling assembly_start debug hook.
9469 * run-rtl-passes.c (run_rtl_passes): Likewise.
9472 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9474 * tree-vrp.c (extract_range_from_binary_expr_1): Check
9475 range_int_cst_p rather than !symbolic_range_p before calling
9476 extract_range_from_multiplicative_op_1.
9478 2018-01-04 Jeff Law <law@redhat.com>
9480 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9481 redundant test in assertion.
9483 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9485 * doc/rtl.texi: Document machine_mode wrapper classes.
9487 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9489 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9492 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9494 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9495 the VEC_PERM_EXPR fold to fail.
9497 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9500 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9501 to switched_sections.
9503 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9506 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9509 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9512 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9513 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9515 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9518 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9519 is BLKmode and bitpos not zero or mode change is needed.
9521 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9524 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9527 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9530 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9531 instead of MULT rtx. Update all corresponding splitters.
9533 (*ssub<modesuffix>): Ditto.
9535 (*cmp_sadd_di): Update split patterns.
9536 (*cmp_sadd_si): Ditto.
9537 (*cmp_sadd_sidi): Ditto.
9538 (*cmp_ssub_di): Ditto.
9539 (*cmp_ssub_si): Ditto.
9540 (*cmp_ssub_sidi): Ditto.
9541 * config/alpha/predicates.md (const23_operand): New predicate.
9542 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9543 Look for ASHIFT, not MULT inner operand.
9544 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9546 2018-01-04 Martin Liska <mliska@suse.cz>
9548 PR gcov-profile/83669
9549 * gcov.c (output_intermediate_file): Add version to intermediate
9551 * doc/gcov.texi: Document new field 'version' in intermediate
9552 file format. Fix location of '-k' option of gcov command.
9554 2018-01-04 Martin Liska <mliska@suse.cz>
9557 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9559 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9561 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9563 2018-01-03 Martin Sebor <msebor@redhat.com>
9565 PR tree-optimization/83655
9566 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9567 checking calls with invalid arguments.
9569 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9571 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9572 (vectorizable_mask_load_store): Delete.
9573 (vectorizable_call): Return false for masked loads and stores.
9574 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9575 instead of gimple_assign_rhs1.
9576 (vectorizable_load): Handle IFN_MASK_LOAD.
9577 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9579 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9581 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9583 (vectorizable_mask_load_store): ...here.
9584 (vectorizable_load): ...and here.
9586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9588 * tree-vect-stmts.c (vect_build_all_ones_mask)
9589 (vect_build_zero_merge_argument): New functions, split out from...
9590 (vectorizable_load): ...here.
9592 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9594 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9596 (vectorizable_mask_load_store): ...here.
9597 (vectorizable_store): ...and here.
9599 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9601 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9603 (vectorizable_mask_load_store): ...here.
9605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9607 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9608 (vect_model_store_cost): Take a vec_load_store_type instead of a
9610 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9611 (vect_model_store_cost): Take a vec_load_store_type instead of a
9613 (vectorizable_mask_load_store): Update accordingly.
9614 (vectorizable_store): Likewise.
9615 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9617 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9619 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9620 IFN_MASK_LOAD calls here rather than...
9621 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9624 Alan Hayward <alan.hayward@arm.com>
9625 David Sherwood <david.sherwood@arm.com>
9627 * expmed.c (extract_bit_field_1): For vector extracts,
9628 fall back to extract_bit_field_as_subreg if vec_extract
9631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9632 Alan Hayward <alan.hayward@arm.com>
9633 David Sherwood <david.sherwood@arm.com>
9635 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
9636 they are variable or constant sized.
9637 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
9638 slots for constant-sized data.
9640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9641 Alan Hayward <alan.hayward@arm.com>
9642 David Sherwood <david.sherwood@arm.com>
9644 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
9645 handling COND_EXPRs with boolean comparisons, try to find a better
9646 basis for the mask type than the boolean itself.
9648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9650 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
9651 is calculated and how it can be overridden.
9652 * genmodes.c (max_bitsize_mode_any_mode): New variable.
9653 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
9655 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
9658 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9659 Alan Hayward <alan.hayward@arm.com>
9660 David Sherwood <david.sherwood@arm.com>
9662 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
9663 Remove the mode argument.
9664 (aarch64_simd_valid_immediate): Remove the mode and inverse
9666 * config/aarch64/iterators.md (bitsize): New iterator.
9667 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
9668 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
9669 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
9670 aarch64_simd_valid_immediate.
9671 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
9672 (aarch64_reg_or_bic_imm): Likewise.
9673 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
9674 with an insn_type enum and msl with a modifier_type enum.
9675 Replace element_width with a scalar_mode. Change the shift
9676 to unsigned int. Add constructors for scalar_float_mode and
9677 scalar_int_mode elements.
9678 (aarch64_vect_float_const_representable_p): Delete.
9679 (aarch64_can_const_movi_rtx_p)
9680 (aarch64_simd_scalar_immediate_valid_for_move)
9681 (aarch64_simd_make_constant): Update call to
9682 aarch64_simd_valid_immediate.
9683 (aarch64_advsimd_valid_immediate_hs): New function.
9684 (aarch64_advsimd_valid_immediate): Likewise.
9685 (aarch64_simd_valid_immediate): Remove mode and inverse
9686 arguments. Rewrite to use the above. Use const_vec_duplicate_p
9687 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
9688 and aarch64_float_const_representable_p on the result.
9689 (aarch64_output_simd_mov_immediate): Remove mode argument.
9690 Update call to aarch64_simd_valid_immediate and use of
9691 simd_immediate_info.
9692 (aarch64_output_scalar_simd_mov_immediate): Update call
9695 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9696 Alan Hayward <alan.hayward@arm.com>
9697 David Sherwood <david.sherwood@arm.com>
9699 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
9700 (mode_nunits): Likewise CONST_MODE_NUNITS.
9701 * machmode.def (ADJUST_NUNITS): Document.
9702 * genmodes.c (mode_data::need_nunits_adj): New field.
9703 (blank_mode): Update accordingly.
9704 (adj_nunits): New variable.
9705 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
9707 (emit_mode_size_inline): Set need_bytesize_adj for all modes
9708 listed in adj_nunits.
9709 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
9710 listed in adj_nunits. Don't emit case statements for such modes.
9711 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
9712 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
9713 nothing if adj_nunits is nonnull.
9714 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
9715 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
9716 (emit_mode_fbit): Update use of print_maybe_const_decl.
9717 (emit_move_size): Likewise. Treat the array as non-const
9719 (emit_mode_adjustments): Handle adj_nunits.
9721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9723 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
9724 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
9725 (VECTOR_MODES): Use it.
9726 (make_vector_modes): Take the prefix as an argument.
9728 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9729 Alan Hayward <alan.hayward@arm.com>
9730 David Sherwood <david.sherwood@arm.com>
9732 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
9733 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
9734 for MODE_VECTOR_BOOL.
9735 * machmode.def (VECTOR_BOOL_MODE): Document.
9736 * genmodes.c (VECTOR_BOOL_MODE): New macro.
9737 (make_vector_bool_mode): New function.
9738 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
9740 * lto-streamer-in.c (lto_input_mode_table): Likewise.
9741 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
9743 * stor-layout.c (int_mode_for_mode): Likewise.
9744 * tree.c (build_vector_type_for_mode): Likewise.
9745 * varasm.c (output_constant_pool_2): Likewise.
9746 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
9747 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
9748 for MODE_VECTOR_BOOL.
9749 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
9750 of mode class checks.
9751 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
9752 instead of a list of mode class checks.
9753 (expand_vector_scalar_condition): Likewise.
9754 (type_for_widest_vector_mode): Handle BImode as an inner mode.
9756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9757 Alan Hayward <alan.hayward@arm.com>
9758 David Sherwood <david.sherwood@arm.com>
9760 * machmode.h (mode_size): Change from unsigned short to
9762 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
9763 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9764 or if measurement_type is not polynomial.
9765 (fixed_size_mode::includes_p): Check for constant-sized modes.
9766 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
9767 return a poly_uint16 rather than an unsigned short.
9768 (emit_mode_size): Change the type of mode_size from unsigned short
9769 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
9770 (emit_mode_adjustments): Cope with polynomial vector sizes.
9771 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9773 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9775 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
9776 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
9777 * caller-save.c (setup_save_areas): Likewise.
9778 (replace_reg_with_saved_mem): Likewise.
9779 * calls.c (emit_library_call_value_1): Likewise.
9780 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
9781 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
9782 (gen_lowpart_for_combine): Likewise.
9783 * convert.c (convert_to_integer_1): Likewise.
9784 * cse.c (equiv_constant, cse_insn): Likewise.
9785 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
9786 (cselib_subst_to_values): Likewise.
9787 * dce.c (word_dce_process_block): Likewise.
9788 * df-problems.c (df_word_lr_mark_ref): Likewise.
9789 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
9790 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
9791 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
9792 (rtl_for_decl_location): Likewise.
9793 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
9794 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
9795 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
9796 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
9797 (expand_expr_real_1): Likewise.
9798 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
9799 (pad_below): Likewise.
9800 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9801 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
9802 * ira.c (get_subreg_tracking_sizes): Likewise.
9803 * ira-build.c (ira_create_allocno_objects): Likewise.
9804 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
9805 (ira_sort_regnos_for_alter_reg): Likewise.
9806 * ira-costs.c (record_operand_costs): Likewise.
9807 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
9808 (resolve_simple_move): Likewise.
9809 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
9810 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
9811 (lra_constraints): Likewise.
9812 (CONST_POOL_OK_P): Reject variable-sized modes.
9813 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
9814 (add_pseudo_to_slot, lra_spill): Likewise.
9815 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9816 * optabs-query.c (get_best_extraction_insn): Likewise.
9817 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9818 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
9819 (expand_mult_highpart, valid_multiword_target_p): Likewise.
9820 * recog.c (offsettable_address_addr_space_p): Likewise.
9821 * regcprop.c (maybe_mode_change): Likewise.
9822 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
9823 * regrename.c (build_def_use): Likewise.
9824 * regstat.c (dump_reg_info): Likewise.
9825 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
9826 (find_reloads, find_reloads_subreg_address): Likewise.
9827 * reload1.c (eliminate_regs_1): Likewise.
9828 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
9829 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
9830 (simplify_binary_operation_1, simplify_subreg): Likewise.
9831 * targhooks.c (default_function_arg_padding): Likewise.
9832 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
9833 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
9834 (verify_gimple_assign_ternary): Likewise.
9835 * tree-inline.c (estimate_move_cost): Likewise.
9836 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9837 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
9838 (get_address_cost_ainc): Likewise.
9839 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9840 (vect_supportable_dr_alignment): Likewise.
9841 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9842 (vectorizable_reduction): Likewise.
9843 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
9844 (vectorizable_operation, vectorizable_load): Likewise.
9845 * tree.c (build_same_sized_truth_vector_type): Likewise.
9846 * valtrack.c (cleanup_auto_inc_dec): Likewise.
9847 * var-tracking.c (emit_note_insn_var_location): Likewise.
9848 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
9849 (ADDR_VEC_ALIGN): Likewise.
9851 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9852 Alan Hayward <alan.hayward@arm.com>
9853 David Sherwood <david.sherwood@arm.com>
9855 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
9857 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9858 or if measurement_type is polynomial.
9859 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
9860 * combine.c (make_extraction): Likewise.
9861 * dse.c (find_shift_sequence): Likewise.
9862 * dwarf2out.c (mem_loc_descriptor): Likewise.
9863 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
9864 (extract_bit_field, extract_low_bits): Likewise.
9865 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
9866 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
9867 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
9868 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
9869 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9870 * reload.c (find_reloads): Likewise.
9871 * reload1.c (alter_reg): Likewise.
9872 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
9873 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
9874 * tree-if-conv.c (predicate_mem_writes): Likewise.
9875 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
9876 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
9877 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9878 * valtrack.c (dead_debug_insert_temp): Likewise.
9879 * varasm.c (mergeable_constant_section): Likewise.
9880 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
9882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9883 Alan Hayward <alan.hayward@arm.com>
9884 David Sherwood <david.sherwood@arm.com>
9886 * expr.c (expand_assignment): Cope with polynomial mode sizes
9887 when assigning to a CONCAT.
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 * machmode.h (mode_precision): Change from unsigned short to
9895 (mode_to_precision): Return a poly_uint16 rather than an unsigned
9897 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
9898 or if measurement_type is not polynomial.
9899 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
9900 in which the mode is already known to be a scalar_int_mode.
9901 * genmodes.c (emit_mode_precision): Change the type of mode_precision
9902 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
9904 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9905 for GET_MODE_PRECISION.
9906 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9907 for GET_MODE_PRECISION.
9908 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
9910 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
9911 (expand_field_assignment, make_extraction): Likewise.
9912 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
9913 (get_last_value): Likewise.
9914 * convert.c (convert_to_integer_1): Likewise.
9915 * cse.c (cse_insn): Likewise.
9916 * expr.c (expand_expr_real_1): Likewise.
9917 * lra-constraints.c (simplify_operand_subreg): Likewise.
9918 * optabs-query.c (can_atomic_load_p): Likewise.
9919 * optabs.c (expand_atomic_load): Likewise.
9920 (expand_atomic_store): Likewise.
9921 * ree.c (combine_reaching_defs): Likewise.
9922 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
9923 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
9924 * tree.h (type_has_mode_precision_p): Likewise.
9925 * ubsan.c (instrument_si_overflow): Likewise.
9927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9928 Alan Hayward <alan.hayward@arm.com>
9929 David Sherwood <david.sherwood@arm.com>
9931 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
9932 polynomial numbers of units.
9933 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
9934 (valid_vector_subparts_p): New function.
9935 (build_vector_type): Remove temporary shim and take the number
9936 of units as a poly_uint64 rather than an int.
9937 (build_opaque_vector_type): Take the number of units as a
9938 poly_uint64 rather than an int.
9939 * tree.c (build_vector_from_ctor): Handle polynomial
9940 TYPE_VECTOR_SUBPARTS.
9941 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
9942 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
9943 (build_vector_from_val): If the number of units is variable,
9944 use build_vec_duplicate_cst for constant operands and
9945 VEC_DUPLICATE_EXPR otherwise.
9946 (make_vector_type): Remove temporary is_constant ().
9947 (build_vector_type, build_opaque_vector_type): Take the number of
9948 units as a poly_uint64 rather than an int.
9949 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
9951 * cfgexpand.c (expand_debug_expr): Likewise.
9952 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
9953 (store_constructor, expand_expr_real_1): Likewise.
9954 (const_scalar_mask_from_tree): Likewise.
9955 * fold-const-call.c (fold_const_reduction): Likewise.
9956 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
9957 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
9958 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
9959 (fold_relational_const): Likewise.
9960 (native_interpret_vector): Likewise. Change the size from an
9961 int to an unsigned int.
9962 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
9963 TYPE_VECTOR_SUBPARTS.
9964 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
9965 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
9966 duplicating a non-constant operand into a variable-length vector.
9967 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
9968 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
9969 * ipa-icf.c (sem_variable::equals): Likewise.
9970 * match.pd: Likewise.
9971 * omp-simd-clone.c (simd_clone_subparts): Likewise.
9972 * print-tree.c (print_node): Likewise.
9973 * stor-layout.c (layout_type): Likewise.
9974 * targhooks.c (default_builtin_vectorization_cost): Likewise.
9975 * tree-cfg.c (verify_gimple_comparison): Likewise.
9976 (verify_gimple_assign_binary): Likewise.
9977 (verify_gimple_assign_ternary): Likewise.
9978 (verify_gimple_assign_single): Likewise.
9979 * tree-pretty-print.c (dump_generic_node): Likewise.
9980 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9981 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
9982 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
9983 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
9984 (vect_shift_permute_load_chain): Likewise.
9985 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
9986 (expand_vector_condition, optimize_vector_constructor): Likewise.
9987 (lower_vec_perm, get_compute_type): Likewise.
9988 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9989 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
9990 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
9991 (vect_recog_mask_conversion_pattern): Likewise.
9992 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
9993 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
9994 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9995 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
9996 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
9997 (vectorizable_shift, vectorizable_operation, vectorizable_store)
9998 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
9999 (supportable_widening_operation): Likewise.
10000 (supportable_narrowing_operation): Likewise.
10001 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10003 * varasm.c (output_constant): Likewise.
10005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10006 Alan Hayward <alan.hayward@arm.com>
10007 David Sherwood <david.sherwood@arm.com>
10009 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10010 so that both the length == 3 and length != 3 cases set up their
10011 own permute vectors. Add comments explaining why we know the
10012 number of elements is constant.
10013 (vect_permute_load_chain): Likewise.
10015 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10016 Alan Hayward <alan.hayward@arm.com>
10017 David Sherwood <david.sherwood@arm.com>
10019 * machmode.h (mode_nunits): Change from unsigned char to
10021 (ONLY_FIXED_SIZE_MODES): New macro.
10022 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10023 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10024 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10026 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10027 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10028 or if measurement_type is not polynomial.
10029 * genmodes.c (ZERO_COEFFS): New macro.
10030 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10032 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10033 Use ZERO_COEFFS when emitting initializers.
10034 * data-streamer.h (bp_pack_poly_value): New function.
10035 (bp_unpack_poly_value): Likewise.
10036 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10037 for GET_MODE_NUNITS.
10038 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10039 for GET_MODE_NUNITS.
10040 * tree.c (make_vector_type): Remove temporary shim and make
10041 the real function take the number of units as a poly_uint64
10042 rather than an int.
10043 (build_vector_type_for_mode): Handle polynomial nunits.
10044 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10045 * emit-rtl.c (const_vec_series_p_1): Likewise.
10046 (gen_rtx_CONST_VECTOR): Likewise.
10047 * fold-const.c (test_vec_duplicate_folding): Likewise.
10048 * genrecog.c (validate_pattern): Likewise.
10049 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10050 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10051 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10052 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10053 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10054 * rtlanal.c (subreg_get_info): Likewise.
10055 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10056 (vect_grouped_load_supported): Likewise.
10057 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10058 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10059 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10060 (simplify_const_unary_operation, simplify_binary_operation_1)
10061 (simplify_const_binary_operation, simplify_ternary_operation)
10062 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10063 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10064 instead of CONST_VECTOR_NUNITS.
10065 * varasm.c (output_constant_pool_2): Likewise.
10066 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10067 explicit-encoded elements in the XVEC for variable-length vectors.
10069 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10071 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10073 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10074 Alan Hayward <alan.hayward@arm.com>
10075 David Sherwood <david.sherwood@arm.com>
10077 * coretypes.h (fixed_size_mode): Declare.
10078 (fixed_size_mode_pod): New typedef.
10079 * builtins.h (target_builtins::x_apply_args_mode)
10080 (target_builtins::x_apply_result_mode): Change type to
10081 fixed_size_mode_pod.
10082 * builtins.c (apply_args_size, apply_result_size, result_vector)
10083 (expand_builtin_apply_args_1, expand_builtin_apply)
10084 (expand_builtin_return): Update accordingly.
10086 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10088 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10089 * cselib.c (cselib_hash_rtx): Likewise.
10090 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10091 CONST_VECTOR encoding.
10093 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10094 Jeff Law <law@redhat.com>
10097 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10098 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10099 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10100 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10103 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10104 explicitly probe *sp in a noreturn function if there were any callee
10105 register saves or frame pointer is needed.
10107 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10110 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10111 BLKmode for ternary, binary or unary expressions.
10114 * var-tracking.c (delete_vta_debug_insn): New inline function.
10115 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10116 insns from get_insns () to NULL instead of each bb separately.
10117 Use delete_vta_debug_insn. No longer static.
10118 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10119 delete_vta_debug_insns callers.
10120 * rtl.h (delete_vta_debug_insns): Declare.
10121 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10122 instead of variable_tracking_main.
10124 2018-01-03 Martin Sebor <msebor@redhat.com>
10126 PR tree-optimization/83603
10127 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10128 arguments past the endof the argument list in functions declared
10129 without a prototype.
10130 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10131 Avoid checking when arguments are null.
10133 2018-01-03 Martin Sebor <msebor@redhat.com>
10136 * doc/extend.texi (attribute const): Fix a typo.
10137 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10138 issuing -Wsuggest-attribute for void functions.
10140 2018-01-03 Martin Sebor <msebor@redhat.com>
10142 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10143 offset_int::from instead of wide_int::to_shwi.
10144 (maybe_diag_overlap): Remove assertion.
10145 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10146 * gimple-ssa-sprintf.c (format_directive): Same.
10147 (parse_directive): Same.
10148 (sprintf_dom_walker::compute_format_length): Same.
10149 (try_substitute_return_value): Same.
10151 2018-01-03 Jeff Law <law@redhat.com>
10153 PR middle-end/83654
10154 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10155 non-constant residual for zero at runtime and avoid probing in
10156 that case. Reorganize code for trailing problem to mirror handling
10159 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10161 PR tree-optimization/83501
10162 * tree-ssa-strlen.c (get_string_cst): New.
10163 (handle_char_store): Call get_string_cst.
10165 2018-01-03 Martin Liska <mliska@suse.cz>
10167 PR tree-optimization/83593
10168 * tree-ssa-strlen.c: Include tree-cfg.h.
10169 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
10170 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
10171 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
10173 (strlen_dom_walker::before_dom_children): Call
10174 gimple_purge_dead_eh_edges. Dump tranformation with details
10176 (strlen_dom_walker::before_dom_children): Update call by adding
10177 new argument cleanup_eh.
10178 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
10180 2018-01-03 Martin Liska <mliska@suse.cz>
10183 * cif-code.def (VARIADIC_THUNK): New enum value.
10184 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
10187 2018-01-03 Jan Beulich <jbeulich@suse.com>
10189 * sse.md (mov<mode>_internal): Tighten condition for when to use
10190 vmovdqu<ssescalarsize> for TI and OI modes.
10192 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10194 Update copyright years.
10196 2018-01-03 Martin Liska <mliska@suse.cz>
10199 * ipa-visibility.c (function_and_variable_visibility): Skip
10200 functions with noipa attribure.
10202 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10204 * gcc.c (process_command): Update copyright notice dates.
10205 * gcov-dump.c (print_version): Ditto.
10206 * gcov.c (print_version): Ditto.
10207 * gcov-tool.c (print_version): Ditto.
10208 * gengtype.c (create_file): Ditto.
10209 * doc/cpp.texi: Bump @copying's copyright year.
10210 * doc/cppinternals.texi: Ditto.
10211 * doc/gcc.texi: Ditto.
10212 * doc/gccint.texi: Ditto.
10213 * doc/gcov.texi: Ditto.
10214 * doc/install.texi: Ditto.
10215 * doc/invoke.texi: Ditto.
10217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10219 * vector-builder.h (vector_builder::m_full_nelts): Change from
10220 unsigned int to poly_uint64.
10221 (vector_builder::full_nelts): Update prototype accordingly.
10222 (vector_builder::new_vector): Likewise.
10223 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10224 (vector_builder::operator ==): Likewise.
10225 (vector_builder::finalize): Likewise.
10226 * int-vector-builder.h (int_vector_builder::int_vector_builder):
10227 Take the number of elements as a poly_uint64 rather than an
10229 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10230 from unsigned int to poly_uint64.
10231 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10232 (vec_perm_indices::new_vector): Likewise.
10233 (vec_perm_indices::length): Likewise.
10234 (vec_perm_indices::nelts_per_input): Likewise.
10235 (vec_perm_indices::input_nelts): Likewise.
10236 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10237 number of elements per input as a poly_uint64 rather than an
10238 unsigned int. Use the original encoding for variable-length
10239 vectors, rather than clamping each individual element.
10240 For the second and subsequent elements in each pattern,
10241 clamp the step and base before clamping their sum.
10242 (vec_perm_indices::series_p): Handle polynomial element counts.
10243 (vec_perm_indices::all_in_range_p): Likewise.
10244 (vec_perm_indices_to_tree): Likewise.
10245 (vec_perm_indices_to_rtx): Likewise.
10246 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10247 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10248 (tree_vector_builder::new_binary_operation): Handle polynomial
10249 element counts. Return false if we need to know the number
10250 of elements at compile time.
10251 * fold-const.c (fold_vec_perm): Punt if the number of elements
10252 isn't known at compile time.
10254 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10256 * vec-perm-indices.h (vec_perm_builder): Change element type
10257 from HOST_WIDE_INT to poly_int64.
10258 (vec_perm_indices::element_type): Update accordingly.
10259 (vec_perm_indices::clamp): Handle polynomial element_types.
10260 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10261 (vec_perm_indices::all_in_range_p): Likewise.
10262 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10264 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10265 polynomial vec_perm_indices element types.
10266 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10267 * fold-const.c (fold_vec_perm): Likewise.
10268 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10269 * tree-vect-generic.c (lower_vec_perm): Likewise.
10270 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10271 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10272 element type to HOST_WIDE_INT.
10274 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10275 Alan Hayward <alan.hayward@arm.com>
10276 David Sherwood <david.sherwood@arm.com>
10278 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10279 rather than an int. Use plus_constant.
10280 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10281 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10284 Alan Hayward <alan.hayward@arm.com>
10285 David Sherwood <david.sherwood@arm.com>
10287 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10288 a HOST_WIDE_INT to a poly_int64.
10290 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10291 Alan Hayward <alan.hayward@arm.com>
10292 David Sherwood <david.sherwood@arm.com>
10294 * calls.c (load_register_parameters): Cope with polynomial
10295 mode sizes. Require a constant size for BLKmode parameters
10296 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10297 forces a parameter to be padded at the lsb end in order to
10298 fill a complete number of words, require the parameter size
10299 to be ordered wrt UNITS_PER_WORD.
10301 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10302 Alan Hayward <alan.hayward@arm.com>
10303 David Sherwood <david.sherwood@arm.com>
10305 * reload1.c (spill_stack_slot_width): Change element type
10306 from unsigned int to poly_uint64_pod.
10307 (alter_reg): Treat mode sizes as polynomial.
10309 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10310 Alan Hayward <alan.hayward@arm.com>
10311 David Sherwood <david.sherwood@arm.com>
10313 * reload.c (complex_word_subreg_p): New function.
10314 (reload_inner_reg_of_subreg, push_reload): Use it.
10316 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10317 Alan Hayward <alan.hayward@arm.com>
10318 David Sherwood <david.sherwood@arm.com>
10320 * lra-constraints.c (process_alt_operands): Reject matched
10321 operands whose sizes aren't ordered.
10322 (match_reload): Refer to this check here.
10324 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10325 Alan Hayward <alan.hayward@arm.com>
10326 David Sherwood <david.sherwood@arm.com>
10328 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10329 that the mode size is in the set {1, 2, 4, 8, 16}.
10331 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10332 Alan Hayward <alan.hayward@arm.com>
10333 David Sherwood <david.sherwood@arm.com>
10335 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10336 Use plus_constant instead of gen_rtx_PLUS.
10338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10339 Alan Hayward <alan.hayward@arm.com>
10340 David Sherwood <david.sherwood@arm.com>
10342 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10343 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10344 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10345 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10346 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10347 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10348 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10349 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10350 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10351 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10353 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10354 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10355 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10356 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10357 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10358 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10359 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10360 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10361 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10362 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10364 * expr.c (emit_move_resolve_push): Treat the input and result
10365 of PUSH_ROUNDING as a poly_int64.
10366 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10367 (emit_push_insn): Likewise.
10368 * lra-eliminations.c (mark_not_eliminable): Likewise.
10369 * recog.c (push_operand): Likewise.
10370 * reload1.c (elimination_effects): Likewise.
10371 * rtlanal.c (nonzero_bits1): Likewise.
10372 * calls.c (store_one_arg): Likewise. Require the padding to be
10373 known at compile time.
10375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10376 Alan Hayward <alan.hayward@arm.com>
10377 David Sherwood <david.sherwood@arm.com>
10379 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10380 Use plus_constant instead of gen_rtx_PLUS.
10382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10383 Alan Hayward <alan.hayward@arm.com>
10384 David Sherwood <david.sherwood@arm.com>
10386 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10387 rather than an int.
10389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10390 Alan Hayward <alan.hayward@arm.com>
10391 David Sherwood <david.sherwood@arm.com>
10393 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10394 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10395 via stack temporaries. Treat the mode size as polynomial too.
10397 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10398 Alan Hayward <alan.hayward@arm.com>
10399 David Sherwood <david.sherwood@arm.com>
10401 * expr.c (expand_expr_real_2): When handling conversions involving
10402 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10403 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10404 as a poly_uint64 too.
10406 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10407 Alan Hayward <alan.hayward@arm.com>
10408 David Sherwood <david.sherwood@arm.com>
10410 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
10412 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10413 Alan Hayward <alan.hayward@arm.com>
10414 David Sherwood <david.sherwood@arm.com>
10416 * combine.c (can_change_dest_mode): Handle polynomial
10417 REGMODE_NATURAL_SIZE.
10418 * expmed.c (store_bit_field_1): Likewise.
10419 * expr.c (store_constructor): Likewise.
10420 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10421 and polynomial REGMODE_NATURAL_SIZE.
10422 (gen_lowpart_common): Likewise.
10423 * reginfo.c (record_subregs_of_mode): Likewise.
10424 * rtlanal.c (read_modify_subreg_p): Likewise.
10426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10427 Alan Hayward <alan.hayward@arm.com>
10428 David Sherwood <david.sherwood@arm.com>
10430 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
10431 numbers of elements.
10433 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10434 Alan Hayward <alan.hayward@arm.com>
10435 David Sherwood <david.sherwood@arm.com>
10437 * match.pd: Cope with polynomial numbers of vector elements.
10439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10440 Alan Hayward <alan.hayward@arm.com>
10441 David Sherwood <david.sherwood@arm.com>
10443 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
10444 in a POINTER_PLUS_EXPR.
10446 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10447 Alan Hayward <alan.hayward@arm.com>
10448 David Sherwood <david.sherwood@arm.com>
10450 * omp-simd-clone.c (simd_clone_subparts): New function.
10451 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10452 (ipa_simd_modify_function_body): Likewise.
10454 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10455 Alan Hayward <alan.hayward@arm.com>
10456 David Sherwood <david.sherwood@arm.com>
10458 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
10459 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10460 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10461 (expand_vector_condition, vector_element): Likewise.
10462 (subparts_gt): New function.
10463 (get_compute_type): Use subparts_gt.
10464 (count_type_subparts): Delete.
10465 (expand_vector_operations_1): Use subparts_gt instead of
10466 count_type_subparts.
10468 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10469 Alan Hayward <alan.hayward@arm.com>
10470 David Sherwood <david.sherwood@arm.com>
10472 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
10473 (vect_compile_time_alias): ...this new function. Do the calculation
10474 on poly_ints rather than trees.
10475 (vect_prune_runtime_alias_test_list): Update call accordingly.
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 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
10483 (vect_schedule_slp_instance): Likewise.
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 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
10490 constant and extern definitions for variable-length vectors.
10491 (vect_get_constant_vectors): Note that the number of units
10492 is known to be constant.
10494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10495 Alan Hayward <alan.hayward@arm.com>
10496 David Sherwood <david.sherwood@arm.com>
10498 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
10499 of units as polynomial. Choose between WIDE and NARROW based
10502 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10503 Alan Hayward <alan.hayward@arm.com>
10504 David Sherwood <david.sherwood@arm.com>
10506 * tree-vect-stmts.c (simd_clone_subparts): New function.
10507 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10509 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10510 Alan Hayward <alan.hayward@arm.com>
10511 David Sherwood <david.sherwood@arm.com>
10513 * tree-vect-stmts.c (vectorizable_call): Treat the number of
10514 vectors as polynomial. Use build_index_vector for
10515 IFN_GOMP_SIMD_LANE.
10517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10518 Alan Hayward <alan.hayward@arm.com>
10519 David Sherwood <david.sherwood@arm.com>
10521 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10522 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10523 for variable-length vectors.
10524 (vectorizable_mask_load_store): Treat the number of units as
10525 polynomial, asserting that it is constant if the condition has
10526 already been enforced.
10527 (vectorizable_store, vectorizable_load): Likewise.
10529 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10530 Alan Hayward <alan.hayward@arm.com>
10531 David Sherwood <david.sherwood@arm.com>
10533 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10534 of units as polynomial. Punt if we can't tell at compile time
10535 which vector contains the final result.
10537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10538 Alan Hayward <alan.hayward@arm.com>
10539 David Sherwood <david.sherwood@arm.com>
10541 * tree-vect-loop.c (vectorizable_induction): Treat the number
10542 of units as polynomial. Punt on SLP inductions. Use an integer
10543 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10544 cast of such a series for variable-length floating-point
10547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10548 Alan Hayward <alan.hayward@arm.com>
10549 David Sherwood <david.sherwood@arm.com>
10551 * tree.h (build_index_vector): Declare.
10552 * tree.c (build_index_vector): New function.
10553 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10554 of units as polynomial, forcibly converting it to a constant if
10555 vectorizable_reduction has already enforced the condition.
10556 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
10557 to create a {1,2,3,...} vector.
10558 (vectorizable_reduction): Treat the number of units as polynomial.
10559 Choose vectype_in based on the largest scalar element size rather
10560 than the smallest number of units. Enforce the restrictions
10563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10564 Alan Hayward <alan.hayward@arm.com>
10565 David Sherwood <david.sherwood@arm.com>
10567 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10568 number of units as polynomial.
10570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10571 Alan Hayward <alan.hayward@arm.com>
10572 David Sherwood <david.sherwood@arm.com>
10574 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10575 * target.def (autovectorize_vector_sizes): Return the vector sizes
10576 by pointer, using vector_sizes rather than a bitmask.
10577 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10578 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10579 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10581 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10582 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10583 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10584 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10585 * omp-general.c (omp_max_vf): Likewise.
10586 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10587 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10588 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10589 * tree-vect-slp.c (vect_slp_bb): Likewise.
10590 * doc/tm.texi: Regenerate.
10591 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10593 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10594 the vector size as a poly_uint64 rather than an unsigned int.
10595 (current_vector_size): Change from an unsigned int to a poly_uint64.
10596 (get_vectype_for_scalar_type): Update accordingly.
10597 * tree.h (build_truth_vector_type): Take the size and number of
10598 units as a poly_uint64 rather than an unsigned int.
10599 (build_vector_type): Add a temporary overload that takes
10600 the number of units as a poly_uint64 rather than an unsigned int.
10601 * tree.c (make_vector_type): Likewise.
10602 (build_truth_vector_type): Take the number of units as a poly_uint64
10603 rather than an unsigned int.
10605 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10606 Alan Hayward <alan.hayward@arm.com>
10607 David Sherwood <david.sherwood@arm.com>
10609 * target.def (get_mask_mode): Take the number of units and length
10610 as poly_uint64s rather than unsigned ints.
10611 * targhooks.h (default_get_mask_mode): Update accordingly.
10612 * targhooks.c (default_get_mask_mode): Likewise.
10613 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10614 * doc/tm.texi: Regenerate.
10616 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10617 Alan Hayward <alan.hayward@arm.com>
10618 David Sherwood <david.sherwood@arm.com>
10620 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10621 * omp-general.c (omp_max_vf): Likewise.
10622 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10623 (expand_omp_simd): Handle polynomial safelen.
10624 * omp-low.c (omplow_simd_context): Add a default constructor.
10625 (omplow_simd_context::max_vf): Change from int to poly_uint64.
10626 (lower_rec_simd_input_clauses): Update accordingly.
10627 (lower_rec_input_clauses): Likewise.
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-vectorizer.h (vect_nunits_for_cost): New function.
10634 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
10635 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
10636 (vect_analyze_slp_cost): Likewise.
10637 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
10638 (vect_model_load_cost): Likewise.
10640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10641 Alan Hayward <alan.hayward@arm.com>
10642 David Sherwood <david.sherwood@arm.com>
10644 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
10645 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
10646 from an unsigned int * to a poly_uint64_pod *.
10647 (calculate_unrolling_factor): New function.
10648 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
10650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10651 Alan Hayward <alan.hayward@arm.com>
10652 David Sherwood <david.sherwood@arm.com>
10654 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
10655 from an unsigned int to a poly_uint64.
10656 (_loop_vec_info::slp_unrolling_factor): Likewise.
10657 (_loop_vec_info::vectorization_factor): Change from an int
10659 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
10660 (vect_get_num_vectors): New function.
10661 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
10662 (vect_get_num_copies): Use vect_get_num_vectors.
10663 (vect_analyze_data_ref_dependences): Change max_vf from an int *
10664 to an unsigned int *.
10665 (vect_analyze_data_refs): Change min_vf from an int * to a
10667 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10668 than an unsigned HOST_WIDE_INT.
10669 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10670 (vect_analyze_data_ref_dependence): Change max_vf from an int *
10671 to an unsigned int *.
10672 (vect_analyze_data_ref_dependences): Likewise.
10673 (vect_compute_data_ref_alignment): Handle polynomial vf.
10674 (vect_enhance_data_refs_alignment): Likewise.
10675 (vect_prune_runtime_alias_test_list): Likewise.
10676 (vect_shift_permute_load_chain): Likewise.
10677 (vect_supportable_dr_alignment): Likewise.
10678 (dependence_distance_ge_vf): Take the vectorization factor as a
10679 poly_uint64 rather than an unsigned HOST_WIDE_INT.
10680 (vect_analyze_data_refs): Change min_vf from an int * to a
10682 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
10683 vfm1 as a poly_uint64 rather than an int. Make the same change
10684 for the returned bound_scalar.
10685 (vect_gen_vector_loop_niters): Handle polynomial vf.
10686 (vect_do_peeling): Likewise. Update call to
10687 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
10688 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
10690 * tree-vect-loop.c (vect_determine_vectorization_factor)
10691 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
10692 (vect_get_known_peeling_cost): Likewise.
10693 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
10694 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
10695 (vect_transform_loop): Likewise. Use the lowest possible VF when
10696 updating the upper bounds of the loop.
10697 (vect_min_worthwhile_factor): Make static. Return an unsigned int
10698 rather than an int.
10699 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
10700 polynomial unroll factors.
10701 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
10702 (vect_make_slp_decision): Likewise.
10703 (vect_supported_load_permutation_p): Likewise, and polynomial
10705 (vect_analyze_slp_cost): Handle polynomial vf.
10706 (vect_slp_analyze_node_operations): Likewise.
10707 (vect_slp_analyze_bb_1): Likewise.
10708 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10709 than an unsigned HOST_WIDE_INT.
10710 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
10711 (vectorizable_load): Handle polynomial vf.
10712 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
10714 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
10716 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10717 Alan Hayward <alan.hayward@arm.com>
10718 David Sherwood <david.sherwood@arm.com>
10720 * match.pd: Handle bit operations involving three constants
10721 and try to fold one pair.
10723 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10725 * tree-vect-loop-manip.c: Include gimple-fold.h.
10726 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
10727 niters_maybe_zero parameters. Handle other cases besides a step of 1.
10728 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
10729 Add a path that uses a step of VF instead of 1, but disable it
10731 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
10732 and niters_no_overflow parameters. Update calls to
10733 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
10734 Create a new SSA name if the latter choses to use a ste other
10735 than zero, and return it via niters_vector_mult_vf_var.
10736 * tree-vect-loop.c (vect_transform_loop): Update calls to
10737 vect_do_peeling, vect_gen_vector_loop_niters and
10738 slpeel_make_loop_iterate_ntimes.
10739 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
10740 (vect_gen_vector_loop_niters): Update declarations after above changes.
10742 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
10744 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
10745 128-bit round to integer instructions.
10746 (ceil<mode>2): Likewise.
10747 (btrunc<mode>2): Likewise.
10748 (round<mode>2): Likewise.
10750 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10752 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
10753 unaligned VSX load/store on P8/P9.
10754 (expand_block_clear): Allow the use of unaligned VSX
10755 load/store on P8/P9.
10757 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10759 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
10761 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
10762 swap associated with both a load and a store.
10764 2018-01-02 Andrew Waterman <andrew@sifive.com>
10766 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
10767 * config/riscv/riscv.md (clear_cache): Use it.
10769 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
10771 * web.c: Remove out-of-date comment.
10773 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10775 * expr.c (fixup_args_size_notes): Check that any existing
10776 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
10777 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
10778 (emit_single_push_insn): ...here.
10780 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10782 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
10783 (const_vector_encoded_nelts): New function.
10784 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
10785 (const_vector_int_elt, const_vector_elt): Declare.
10786 * emit-rtl.c (const_vector_int_elt_1): New function.
10787 (const_vector_elt): Likewise.
10788 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
10789 of CONST_VECTOR_ELT.
10791 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10793 * expr.c: Include rtx-vector-builder.h.
10794 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
10795 directly on the tree encoding.
10796 (const_vector_from_tree): Likewise.
10797 * optabs.c: Include rtx-vector-builder.h.
10798 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
10799 sequence of "u" values.
10800 * vec-perm-indices.c: Include rtx-vector-builder.h.
10801 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
10802 directly on the vec_perm_indices encoding.
10804 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10806 * doc/rtl.texi (const_vector): Describe new encoding scheme.
10807 * Makefile.in (OBJS): Add rtx-vector-builder.o.
10808 * rtx-vector-builder.h: New file.
10809 * rtx-vector-builder.c: Likewise.
10810 * rtl.h (rtx_def::u2): Add a const_vector field.
10811 (CONST_VECTOR_NPATTERNS): New macro.
10812 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
10813 (CONST_VECTOR_DUPLICATE_P): Likewise.
10814 (CONST_VECTOR_STEPPED_P): Likewise.
10815 (CONST_VECTOR_ENCODED_ELT): Likewise.
10816 (const_vec_duplicate_p): Check for a duplicated vector encoding.
10817 (unwrap_const_vec_duplicate): Likewise.
10818 (const_vec_series_p): Check for a non-duplicated vector encoding.
10819 Say that the function only returns true for integer vectors.
10820 * emit-rtl.c: Include rtx-vector-builder.h.
10821 (gen_const_vec_duplicate_1): Delete.
10822 (gen_const_vector): Call gen_const_vec_duplicate instead of
10823 gen_const_vec_duplicate_1.
10824 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
10825 (gen_const_vec_duplicate): Use rtx_vector_builder.
10826 (gen_const_vec_series): Likewise.
10827 (gen_rtx_CONST_VECTOR): Likewise.
10828 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
10829 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10830 Build a new vector rather than modifying a CONST_VECTOR in-place.
10831 (handle_special_swappables): Update call accordingly.
10832 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
10833 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10834 Build a new vector rather than modifying a CONST_VECTOR in-place.
10835 (handle_special_swappables): Update call accordingly.
10837 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10839 * simplify-rtx.c (simplify_const_binary_operation): Use
10840 CONST_VECTOR_ELT instead of XVECEXP.
10842 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10844 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
10845 the selector elements to be different from the data elements
10846 if the selector is a VECTOR_CST.
10847 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
10848 ssizetype for the selector.
10850 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10852 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
10853 before testing each element individually.
10854 * tree-vect-generic.c (lower_vec_perm): Likewise.
10856 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10858 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
10859 * selftest-run-tests.c (selftest::run_tests): Call it.
10860 * vector-builder.h (vector_builder::operator ==): New function.
10861 (vector_builder::operator !=): Likewise.
10862 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
10863 (vec_perm_indices::all_from_input_p): New function.
10864 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10865 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
10866 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
10867 instead of reading the VECTOR_CST directly. Detect whether both
10868 vector inputs are the same before constructing the vec_perm_indices,
10869 and update the number of inputs argument accordingly. Use the
10870 utility functions added above. Only construct sel2 if we need to.
10872 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10874 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
10875 the broadcast of the low byte.
10876 (expand_mult_highpart): Use an explicit encoding for the permutes.
10877 * optabs-query.c (can_mult_highpart_p): Likewise.
10878 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
10879 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10880 (vectorizable_bswap): Likewise.
10881 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
10882 explicit encoding for the power-of-2 permutes.
10883 (vect_permute_store_chain): Likewise.
10884 (vect_grouped_load_supported): Likewise.
10885 (vect_permute_load_chain): Likewise.
10887 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10889 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
10890 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
10891 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
10892 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10893 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
10894 (vect_gen_perm_mask_any): Likewise.
10896 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10898 * int-vector-builder.h: New file.
10899 * vec-perm-indices.h: Include int-vector-builder.h.
10900 (vec_perm_indices): Redefine as an int_vector_builder.
10901 (auto_vec_perm_indices): Delete.
10902 (vec_perm_builder): Redefine as a stand-alone class.
10903 (vec_perm_indices::vec_perm_indices): New function.
10904 (vec_perm_indices::clamp): Likewise.
10905 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
10906 (vec_perm_indices::new_vector): New function.
10907 (vec_perm_indices::new_expanded_vector): Update for new
10908 vec_perm_indices class.
10909 (vec_perm_indices::rotate_inputs): New function.
10910 (vec_perm_indices::all_in_range_p): Operate directly on the
10911 encoded form, without computing elided elements.
10912 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
10913 encoding. Update for new vec_perm_indices class.
10914 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
10915 the given vec_perm_builder.
10916 (expand_vec_perm_var): Update vec_perm_builder constructor.
10917 (expand_mult_highpart): Use vec_perm_builder instead of
10918 auto_vec_perm_indices.
10919 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
10920 vec_perm_indices instead of auto_vec_perm_indices. Use a single
10921 or double series encoding as appropriate.
10922 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
10923 vec_perm_indices instead of auto_vec_perm_indices.
10924 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10925 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10926 (vect_permute_store_chain): Likewise.
10927 (vect_grouped_load_supported): Likewise.
10928 (vect_permute_load_chain): Likewise.
10929 (vect_shift_permute_load_chain): Likewise.
10930 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10931 (vect_transform_slp_perm_load): Likewise.
10932 (vect_schedule_slp_instance): Likewise.
10933 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10934 (vectorizable_mask_load_store): Likewise.
10935 (vectorizable_bswap): Likewise.
10936 (vectorizable_store): Likewise.
10937 (vectorizable_load): Likewise.
10938 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
10939 vec_perm_indices instead of auto_vec_perm_indices. Use
10940 tree_to_vec_perm_builder to read the vector from a tree.
10941 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
10942 vec_perm_builder instead of a vec_perm_indices.
10943 (have_whole_vector_shift): Use vec_perm_builder and
10944 vec_perm_indices instead of auto_vec_perm_indices. Leave the
10945 truncation to calc_vec_perm_mask_for_shift.
10946 (vect_create_epilog_for_reduction): Likewise.
10947 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
10948 from auto_vec_perm_indices to vec_perm_indices.
10949 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10950 instead of changing individual elements.
10951 (aarch64_vectorize_vec_perm_const): Use new_vector to install
10952 the vector in d.perm.
10953 * config/arm/arm.c (expand_vec_perm_d::perm): Change
10954 from auto_vec_perm_indices to vec_perm_indices.
10955 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10956 instead of changing individual elements.
10957 (arm_vectorize_vec_perm_const): Use new_vector to install
10958 the vector in d.perm.
10959 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
10960 Update vec_perm_builder constructor.
10961 (rs6000_expand_interleave): Likewise.
10962 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
10963 (rs6000_expand_interleave): Likewise.
10965 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10967 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
10968 to qimode could truncate the indices.
10969 * optabs.c (expand_vec_perm_var): Likewise.
10971 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10973 * Makefile.in (OBJS): Add vec-perm-indices.o.
10974 * vec-perm-indices.h: New file.
10975 * vec-perm-indices.c: Likewise.
10976 * target.h (vec_perm_indices): Replace with a forward class
10978 (auto_vec_perm_indices): Move to vec-perm-indices.h.
10979 * optabs.h: Include vec-perm-indices.h.
10980 (expand_vec_perm): Delete.
10981 (selector_fits_mode_p, expand_vec_perm_var): Declare.
10982 (expand_vec_perm_const): Declare.
10983 * target.def (vec_perm_const_ok): Replace with...
10984 (vec_perm_const): ...this new hook.
10985 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
10986 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
10987 * doc/tm.texi: Regenerate.
10988 * optabs.def (vec_perm_const): Delete.
10989 * doc/md.texi (vec_perm_const): Likewise.
10990 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
10991 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
10992 expand_vec_perm for constant permutation vectors. Assert that
10993 the mode of variable permutation vectors is the integer equivalent
10994 of the mode that is being permuted.
10995 * optabs-query.h (selector_fits_mode_p): Declare.
10996 * optabs-query.c: Include vec-perm-indices.h.
10997 (selector_fits_mode_p): New function.
10998 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
10999 is defined, instead of checking whether the vec_perm_const_optab
11000 exists. Use targetm.vectorize.vec_perm_const instead of
11001 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11002 fit in the vector mode before using a variable permute.
11003 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11004 vec_perm_indices instead of an rtx.
11005 (expand_vec_perm): Replace with...
11006 (expand_vec_perm_const): ...this new function. Take the selector
11007 as a vec_perm_indices rather than an rtx. Also take the mode of
11008 the selector. Update call to shift_amt_for_vec_perm_mask.
11009 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11010 Use vec_perm_indices::new_expanded_vector to expand the original
11011 selector into bytes. Check whether the indices fit in the vector
11012 mode before using a variable permute.
11013 (expand_vec_perm_var): Make global.
11014 (expand_mult_highpart): Use expand_vec_perm_const.
11015 * fold-const.c: Includes vec-perm-indices.h.
11016 * tree-ssa-forwprop.c: Likewise.
11017 * tree-vect-data-refs.c: Likewise.
11018 * tree-vect-generic.c: Likewise.
11019 * tree-vect-loop.c: Likewise.
11020 * tree-vect-slp.c: Likewise.
11021 * tree-vect-stmts.c: Likewise.
11022 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11024 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11025 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11026 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11027 (aarch64_vectorize_vec_perm_const): ...this new function.
11028 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11029 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11030 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11031 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11032 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11033 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11034 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11036 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11037 check for NEON modes.
11038 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11039 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11040 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11041 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11043 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11044 the old VEC_PERM_CONST conditions.
11045 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11046 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11047 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11048 (ia64_vectorize_vec_perm_const_ok): Merge into...
11049 (ia64_vectorize_vec_perm_const): ...this new function.
11050 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11051 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11052 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11053 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11054 * config/mips/mips.c (mips_expand_vec_perm_const)
11055 (mips_vectorize_vec_perm_const_ok): Merge into...
11056 (mips_vectorize_vec_perm_const): ...this new function.
11057 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11058 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11059 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11060 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11061 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11062 (rs6000_expand_vec_perm_const): Delete.
11063 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11065 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11066 (altivec_expand_vec_perm_const_le): Take each operand individually.
11067 Operate on constant selectors rather than rtxes.
11068 (altivec_expand_vec_perm_const): Likewise. Update call to
11069 altivec_expand_vec_perm_const_le.
11070 (rs6000_expand_vec_perm_const): Delete.
11071 (rs6000_vectorize_vec_perm_const_ok): Delete.
11072 (rs6000_vectorize_vec_perm_const): New function.
11073 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11074 an element count and rtx array.
11075 (rs6000_expand_extract_even): Update call accordingly.
11076 (rs6000_expand_interleave): Likewise.
11077 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11078 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11079 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11080 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11081 (rs6000_expand_vec_perm_const): Delete.
11082 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11083 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11084 (altivec_expand_vec_perm_const_le): Take each operand individually.
11085 Operate on constant selectors rather than rtxes.
11086 (altivec_expand_vec_perm_const): Likewise. Update call to
11087 altivec_expand_vec_perm_const_le.
11088 (rs6000_expand_vec_perm_const): Delete.
11089 (rs6000_vectorize_vec_perm_const_ok): Delete.
11090 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11091 reference to the SPE evmerge intructions.
11092 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11093 an element count and rtx array.
11094 (rs6000_expand_extract_even): Update call accordingly.
11095 (rs6000_expand_interleave): Likewise.
11096 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11097 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11099 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11101 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11103 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11104 vector mode and that that mode matches the mode of the data
11106 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11107 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11108 directly using expand_vec_perm_1 when forcing selectors into
11110 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11112 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11114 * optabs-query.h (can_vec_perm_p): Delete.
11115 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11116 * optabs-query.c (can_vec_perm_p): Split into...
11117 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11118 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11119 particular selector is valid.
11120 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11121 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11122 (vect_grouped_load_supported): Likewise.
11123 (vect_shift_permute_load_chain): Likewise.
11124 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11125 (vect_transform_slp_perm_load): Likewise.
11126 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11127 (vectorizable_bswap): Likewise.
11128 (vect_gen_perm_mask_checked): Likewise.
11129 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11130 implementations of variable permutation vectors into account
11131 when deciding which selector to use.
11132 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11133 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11134 with a false third argument.
11135 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11136 to test whether the constant selector is valid and can_vec_perm_var_p
11137 to test whether a variable selector is valid.
11139 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11141 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11142 * optabs-query.c (can_vec_perm_p): Likewise.
11143 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11144 instead of vec_perm_indices.
11145 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11146 (vect_gen_perm_mask_checked): Likewise,
11147 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11148 (vect_gen_perm_mask_checked): Likewise,
11150 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11152 * optabs-query.h (qimode_for_vec_perm): Declare.
11153 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11154 (qimode_for_vec_perm): ...this new function.
11155 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11157 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11159 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11160 does not have a conditional at the top.
11162 2018-01-02 Richard Biener <rguenther@suse.de>
11164 * ipa-inline.c (big_speedup_p): Fix expression.
11166 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11169 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
11172 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11176 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
11177 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
11178 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
11179 cond_taken_branch_cost 3->4.
11181 2018-01-01 Jakub Jelinek <jakub@redhat.com>
11183 PR tree-optimization/83581
11184 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
11185 TODO_cleanup_cfg if any changes have been made.
11187 PR middle-end/83608
11188 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
11189 convert_modes if target mode has the right side, but different mode
11192 PR middle-end/83609
11193 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
11194 last argument when extracting from CONCAT. If either from_real or
11195 from_imag is NULL, use expansion through memory. If result is not
11196 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
11197 the parts directly to inner mode, if even that fails, use expansion
11200 PR middle-end/83623
11201 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11202 check for bswap in mode rather than HImode and use that in expand_unop
11205 Copyright (C) 2018 Free Software Foundation, Inc.
11207 Copying and distribution of this file, with or without modification,
11208 are permitted in any medium without royalty provided the copyright
11209 notice and this notice are preserved.