1 2018-04-17 Jim Wilson <jimw@sifive.com>
4 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
5 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
6 Set arg_pointer_offset after using pretend_args_size.
8 2018-04-17 Jakub Jelinek <jakub@redhat.com>
10 PR rtl-optimization/85431
11 * dse.c (record_store): Ignore zero width stores.
14 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
15 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
16 __builtin_stack_restore rather than after it.
17 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
18 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
19 argument instead of virtual_dynamic_stack_rtx.
21 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
23 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
25 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
26 Add note to error message to explain internal mapping of overloaded
27 built-in function name to non-overloaded built-in function name.
28 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
31 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
34 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
35 where the inputs overlap with the output.
37 2018-04-17 Jakub Jelinek <jakub@redhat.com>
39 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
40 (=v, v) alternative and explicit "memory" attribute.
41 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
42 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
44 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
45 "sselog1" type instead of "sselog".
46 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
47 "sselog". Remove explicit "memory" attribute.
48 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
49 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
51 (vec_extract_hi_v32hi): Merge all alternatives into one, use
52 "sselog1" type instead of "sselog". Remove explicit "memory"
54 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
55 use "sselog1" type instead of "sselog". Remove explicit "memory"
57 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
58 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
60 (vec_extract_hi_v64qi): Merge all alternatives into one, use
61 "sselog1" type instead of "sselog". Remove explicit "memory"
63 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
64 use "sselog1" type instead of "sselog". Remove explicit "memory"
68 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
71 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
74 2018-04-17 Martin Jambor <mjambor@suse.cz>
77 * ipa-cp.c (create_specialized_node): Call
78 expand_all_artificial_thunks if necessary.
80 2018-04-17 Martin Liska <mliska@suse.cz>
83 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
84 in message, remote space in between '_G' and '('.
86 2018-04-17 Jakub Jelinek <jakub@redhat.com>
89 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
90 avx512f_vmcmp<mode>3<round_saeonly_name>,
91 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
92 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
93 avx512f_rndscale<mode><round_saeonly_name>,
94 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
95 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
96 Use %<iptr>2 instead of %2 for -masm=intel.
97 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
98 avx512f_vcvttss2usi<round_saeonly_name>,
99 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
101 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
102 avx512f_vcvttsd2usi<round_saeonly_name>,
103 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
104 Use %q1 instead of %1 for -masm=intel.
105 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
106 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
107 of %3 for -masm=intel.
108 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
110 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
112 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
114 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
115 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
117 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
119 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
120 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
121 %g1 and one with %0 and %1.
122 (avx512er_vmrcp28<mode><round_saeonly_name>,
123 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
125 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
126 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
127 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
128 of %0 and %{%4%} for -masm=intel.
129 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
130 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
131 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
132 order of %0 and %{%5%}%{z%} for -masm=intel.
134 2018-04-17 Jan Hubicka <jh@suse.cz>
137 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
139 2018-04-17 Martin Liska <mliska@suse.cz>
142 * multiple_target.c (create_dispatcher_calls): Set apostrophes
143 for target_clone error message. Make default implementation
144 clone to be a local declaration.
145 (separate_attrs): Add new argument and check for an empty
147 (expand_target_clones): Handle it.
148 (ipa_target_clone): Make redirection just for target_clones
151 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
152 Tom de Vries <tom@codesourcery.com>
155 * omp-expand.c (expand_oacc_for): Add dummy false branch for
156 tiled basic blocks without omp continue statements.
158 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
161 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
162 vec_extract expression as having side effects to make sure it gets
165 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
168 * config/i386/i386.c (get_builtin_code_for_version): Check
171 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
174 * gcc/config.gcc: Support "skylake".
175 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
177 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
178 (processor_target_table): Add "skylake".
179 (ix86_option_override_internal): Add "skylake".
180 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
181 PROCESSOR_CANNONLAKE.
182 (get_builtin_code_for_version): Fix priority for
183 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
184 PROCESSOR_SKYLAKE-AVX512.
185 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
186 (processor_type): Add PROCESSOR_SKYLAKE.
188 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
189 Jason Merrill <jason@redhat.com>
192 * convert.c (convert_to_integer_1): Use direct recursion for
193 enumeral types and types with a precision less than the number
194 of bits in their mode.
196 2018-04-16 Julia Koval <julia.koval@intel.com>
199 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
200 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
202 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
205 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
206 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
207 and -mno-direct-move.
209 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
212 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
213 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
214 Ensure negative shifts result in {0}.
216 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
218 PR rtl-optimization/79916
219 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
220 regs (if any) to define how to gnerate SD moves when LRA is in
223 2018-04-13 Jakub Jelinek <jakub@redhat.com>
225 PR rtl-optimization/85393
226 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
227 * except.c (expand_dw2_landing_pad_for_region): Make static.
228 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
229 a label and unconditional jump to old_bb, rather than
230 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
233 PR rtl-optimization/85376
234 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
235 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
236 instead of a specific value.
238 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
239 Bin Cheng <bin.cheng@arm.com>
241 PR tree-optimization/82965
242 PR tree-optimization/83991
243 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
244 by_profile_only parameter.
245 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
246 information if the loop was predicted to iterate too many times.
247 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
249 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
252 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
255 2018-04-13 Martin Liska <mliska@suse.cz>
256 Jakub Jelinek <jakub@redhat.com>
259 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
260 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
261 * builtins.c (expand_builtin_memory_copy_args): Use
262 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
263 handle dest_addr == pc_rtx.
265 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
268 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
269 asked to not generate direct moves.
270 (fix_trunc<mode>si2_stfiwx): Similar.
271 (fix_trunc<mode>si2_internal): Similar.
273 2018-04-12 Jakub Jelinek <jakub@redhat.com>
276 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
277 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
278 lookup if dest in some wider mode is known to be const0_rtx and
279 if so, record permanent equivalence for it to be ZERO_EXTEND of
280 the narrower mode destination.
282 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
284 * lto-streamer-out.c (output_function): Revert 259346.
285 * omp-expand.c (expand_oacc_for): Likewise.
287 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
289 PR rtl-optimization/85354
290 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
291 * sel-sched.c (sel_global_init): ... here.
293 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
296 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
297 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
298 mode for PE-COFF targets.
299 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
300 (i386_pe_asm_lto_end): Likewise.
301 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
302 (TARGET_ASM_LTO_END): Likewise.
303 * config/i386/winnt.c (saved_debug_info_level): New static variable.
304 (i386_pe_asm_lto_start): New function.
305 (i386_pe_asm_lto_end): Likewise.
307 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
308 Richard Biener <rguenther@suse.de>
311 * lto-streamer-out.c (output_function): Fix CFG loop state before
313 * omp-expand.c (expand_oacc_for): Handle calls to internal
314 functions like regular functions.
316 2018-04-12 Richard Biener <rguenther@suse.de>
319 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
320 for the early LTO debug to properly generate references to it
321 during DIE emission. Do not re-use that for the skeleton for
323 (dwarf2out_early_finish): Likewise.
325 2018-04-12 Jakub Jelinek <jakub@redhat.com>
329 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
330 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
331 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
332 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
333 and output is a reg, avoid creating invalid lowpart subreg, but
334 instead split into a 512-bit move. Don't split if not AVX512VL,
335 input is xmm16+ reg and output is a mem.
336 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
337 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
338 xmm16+ reg and output is a mem.
340 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
342 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
343 also for flag_dwarf2_cfi_asm.
345 2018-04-12 Jakub Jelinek <jakub@redhat.com>
347 PR rtl-optimization/85342
348 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
349 a bool scalar var inside of the loop instead. Don't try to update
350 recog_data.operand after failed apply_change_group.
352 2018-04-12 Tom de Vries <tom@codesourcery.com>
355 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
356 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
357 array with flexible array member as array without given dimension.
358 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
359 argument for undefined param to true.
361 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
364 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
365 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
366 from PowerPC section.
367 * config/rs6000/sysv4.opt (mcall-): Improve help text.
368 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
369 help text that is too long.
370 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
371 help text that is too long.
372 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
373 help text that is too long.
375 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
377 * config/alpha/alpha.md (stack_probe_internal): Rename
378 from "probe_stack". Update all callers.
380 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
382 PR rtl-optimization/84566
383 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
384 sched_macro_fuse_insns.
386 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
389 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
390 (compute_block_dependences): ... from here.
392 2018-04-11 Jakub Jelinek <jakub@redhat.com>
394 PR tree-optimization/85331
395 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
396 from int to HOST_WIDE_INT.
398 2018-04-11 Martin Jambor <mjambor@suse.cz>
401 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
402 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
403 not the same as the source val.
404 (cgraph_edge_brings_value_p): New parameter.
405 (gather_edges_for_value): Pass destination value to
406 cgraph_edge_brings_value_p.
407 (perhaps_add_new_callers): Likewise.
408 (get_info_about_necessary_edges): Likewise and exclude values brought
409 only by self-recursive edges.
410 (create_specialized_node): Redirect only clones of self-calling edges.
411 (+self_recursive_pass_through_p): New function.
412 (find_more_scalar_values_for_callers_subset): Use it.
413 (find_aggregate_values_for_callers_subset): Likewise.
414 (known_aggs_to_agg_replacement_list): Removed.
415 (decide_whether_version_node): Re-calculate known constants for all
416 remaining context clones.
418 2018-04-11 Richard Biener <rguenther@suse.de>
421 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
422 from early DWARF output.
423 (dwarf2out_early_finish): Output line info unconditionally into
424 early DWARF and add reference to it.
426 2018-04-11 Jakub Jelinek <jakub@redhat.com>
429 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
430 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
431 other than V2DFmode using iptr mode attribute.
432 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
434 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
436 PR rtl-optimization/84659
437 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
439 2018-04-11 Jakub Jelinek <jakub@redhat.com>
442 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
444 (output_loc_list): Pass address of a dummy size variable even in the
445 locview handling loop.
446 (index_location_lists): Add comment on why skip_loc_list_entry can't
449 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
452 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
455 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
458 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
459 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
460 and -mstring-compare-inline-limit.
462 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
465 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
466 for stack clash protection in a register whenever we need it to be in
469 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
471 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
472 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
474 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
477 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
479 (mlong-double-): Ditto.
480 * config/rs6000/sysv4.opt (msdata=): Ditto.
483 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
485 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
486 erroneous entries for
487 "vector int vec_ldl (int, long int *)", and
488 "vector unsigned int vec_ldl (int, unsigned long int *)".
489 Add comments and entries for
490 "vector bool char vec_ldl (int, bool char *)",
491 "vector bool short vec_ldl (int, bool short *)",
492 "vector bool int vec_ldl (int, bool int *)",
493 "vector bool long long vec_ldl (int, bool long long *)",
494 "vector pixel vec_ldl (int, pixel *)",
495 "vector long long vec_ldl (int, long long *)",
496 "vector unsigned long long vec_ldl (int, unsigned long long *)".
497 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
498 type tree bool_long_long_type_node and correct definition of
499 bool_V2DI_type_node to make reference to this new type tree.
500 (rs6000_mangle_type): Replace erroneous reference to
501 bool_long_type_node with bool_long_long_type_node.
502 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
503 comments to emphasize sign distinctions for char and int types and
504 replace RS6000_BTI_bool_long constant with
505 RS6000_BTI_bool_long_long constant. Also add comment to restrict
506 use of RS6000_BTI_pixel.
507 (bool_long_type_node): Remove this macro definition.
508 (bool_long_long_type_node): New macro definition
510 2018-04-10 Jakub Jelinek <jakub@redhat.com>
512 PR rtl-optimization/85300
513 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
514 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
515 simplify_unary_operation fails.
517 2018-04-10 Martin Liska <mliska@suse.cz>
519 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
520 cgraph_edge and ipa_ref.
522 2018-04-10 Jakub Jelinek <jakub@redhat.com>
527 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
528 computation of the VEC_MERGE selector from mask.
529 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
530 Fix decoding of the VEC_MERGE selector into mask.
532 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
534 PR tree-optimization/85286
535 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
537 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
539 * final.c (final_1): Set insn_last_address as well as
540 insn_current_address.
542 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
545 * explow.c (emit_stack_probe): Call validize_mem on memory location
546 before passing it to gen_probe_stack. Create address operand and
547 legitimize it for the probe_stack_address case.
549 2018-04-09 Jan Hubicka <jh@suse.cz>
552 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
553 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
554 * tree.c (free_lang_data_in_type): Fix handling of binfos;
556 (free_lang_data): Rebuild type inheritance graph.
558 2018-04-09 Martin Sebor <msebor@redhat.com>
560 * invoke.texi (-finline-small-functions): Mention other optimization
562 (-findirect-inlining, -fpartial-inlining): Same.
563 (-finline-functions-called-once): Same.
564 (-freorder-blocks-and-partition): Same.
566 2018-04-09 Jan Hubicka <jh@suse.cz>
569 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
570 jumps; choose last target that matches the criteria (i.e.
571 no partition changes for non-crossing jumps).
572 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
573 support for redirecting crossing jumps to non-crossing.
575 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
577 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
578 also for naked functions.
580 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
582 * config/arc/arc.md (add_shift): New pattern.
583 (add_shift2): Likewise.
584 (sub_shift): Likewise.
585 (sub_shift_cmp0_noout): Likewise.
586 (compare_si_ashiftsi): Likewise.
587 (xbfu_cmp0_noout): New combine pattern.
588 (xbfu_cmp0"): Likewise.
589 (movsi_set_cc_insn): Place the predicable variant first.
590 (commutative_binary_cmp0_noout): Remove clobber.
591 (commutative_binary_cmp0): New pattern.
592 (noncommutative_binary_cmp0): Likewise.
593 (noncommutative_binary_cmp0_noout): Likewise.
594 (noncommutative_binary_comparison_result_used): Removed.
595 (rsub_cmp0): New pattern.
596 (rsub_cmp0_noout): Likewise.
597 (extzvsi): Changed, keep only meaningful variants.
598 (SQH, SEZ): New iterators.
599 (SQH_postfix): New mode attribute.
600 (SEZ_prefix): New code attribute.
601 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
602 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
603 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
605 (noncommutative_operator): Check the availability of barrel
608 2018-04-09 Richard Biener <rguenther@suse.de>
610 PR tree-optimization/85284
611 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
612 Only use the niter constraining form of simple_iv when the exit
615 2018-04-09 Tom de Vries <tom@codesourcery.com>
618 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
619 (define_expand "*memory_barrier"): New define_expand.
620 (define_insn "memory_barrier"): New insn.
622 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
624 PR rtl-optimization/80463
625 PR rtl-optimization/83972
626 PR rtl-optimization/83480
628 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
629 correct producer for the insn.
630 (tidy_control_flow): Fixup seqnos in case of debug insns.
632 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
634 PR rtl-optimization/83913
636 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
637 different sched-times when merging exprs.
639 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
641 PR rtl-optimization/83962
643 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
644 tidy_fallthru_edge and tidy_control_flow.
646 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
648 PR rtl-optimization/83530
650 * sel-sched.c (force_next_insn): New global variable.
651 (remove_insn_for_debug): When force_next_insn is true, also leave only
652 next insn in the ready list.
653 (sel_sched_region): When the region wasn't scheduled, make another pass
654 over it with force_next_insn set to 1.
656 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
658 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
660 * config/nds32/constants.md (unspec_volatile_element): Add enum values
661 for interrupt control.
662 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
663 functions for interrupt control.
664 * config/nds32/nds32-intrinsic.md: Likewise.
665 * config/nds32/nds32_intrinsic.h: Likewise.
666 * config/nds32/nds32.h (nds32_builtins): Likewise.
668 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
670 * config/nds32/nds32.c (nds32_init_machine_status,
671 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
672 strict_aligned_p field.
673 (nds32_expand_to_rtl_hook): New function.
674 (TARGET_EXPAND_TO_RTL_HOOK): Define.
675 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
677 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
678 Chung-Ju Wu <jasonwucj@gmail.com>
680 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
681 * config/nds32/nds32-n7.md: New file.
682 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
683 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
685 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
686 * config/nds32/nds32.md (pipeline_model): Add n7.
687 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
688 * config/nds32/pipelines.md: Include n7 settings.
690 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
691 Chung-Ju Wu <jasonwucj@gmail.com>
693 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
694 * config/nds32/nds32-e8.md: New file.
695 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
696 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
698 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
699 * config/nds32/nds32.md (pipeline_model): Add e8.
700 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
701 * config/nds32/pipelines.md: Include e8 settings.
703 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
704 Chung-Ju Wu <jasonwucj@gmail.com>
706 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
707 * config/nds32/nds32-n8.md: New file.
708 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
709 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
711 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
712 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
713 * config/nds32/nds32.md (pipeline_model): Add n8.
714 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
715 * config/nds32/pipelines.md: Include n8 settings.
717 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
718 Chung-Ju Wu <jasonwucj@gmail.com>
720 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
721 * config/nds32/nds32-n9-2r1w.md: New file.
722 * config/nds32/nds32-n9-3r2w.md: New file.
723 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
724 nds32_register_ports): New or modify for cpu n9.
725 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
727 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
728 * config/nds32/nds32-utils.c: New file.
729 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
730 TARGET_MUL_SLOW): Define.
731 * config/nds32/nds32.md (pipeline_model): New attribute.
732 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
733 New options that support cpu n9.
734 * config/nds32/pipelines.md: Include n9 settings.
735 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
737 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
739 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
740 information if necessary.
741 (output_cond_branch_compare_zero): Likewise.
742 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
743 (nds32_target_alignment): Refine for alignment.
744 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
745 (FUNCTION_BOUNDARY): Modify.
746 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
748 * config/nds32/nds32.opt (malways-align, malign-functions): New.
750 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
752 * config/nds32/constants.md (unspec_volatile_element): Add values for
753 TLB operation and data prefetch.
754 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
755 functions for TLB operation and data prefetch.
756 * config/nds32/nds32-intrinsic.md: Likewise.
757 * config/nds32/nds32_intrinsic.h: Likewise.
758 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
759 (nds32_print_operand): Likewise.
760 * config/nds32/nds32.h (nds32_builtins): Likewise.
762 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
763 Andrew Pinski <pinsika@gcc.gnu.org>
766 * match.pd: Use constant_boolean_node of correct type instead of
767 boolean_true_node or boolean_false_node for simplifying
768 pointer comparisons to zero.
770 2018-04-07 Jakub Jelinek <jakub@redhat.com>
772 PR tree-optimization/80021
773 * tree.c (verify_type_variant): Make error call in verify_variant_match
774 translatable and remove final full stop.
776 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
778 * config/nds32/constants.md (unspec_volatile_element): Add
779 UNSPEC_VOLATILE_EH_RETURN.
780 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
781 nds32_output_stack_pop): Support dwarf exception handling process.
782 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
783 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
784 exception handling process.
785 (nds32_compute_stack_frame): Likewise.
786 (nds32_return_addr_rtx): Likewise.
787 (nds32_initial_elimination_offset): Likewise.
788 (nds32_expand_prologue): Likewise.
789 (nds32_expand_epilogue): Likewise.
790 (nds32_dynamic_chain_address): New function.
791 * config/nds32/nds32.h (machine_function): Add fields for dwarf
793 (DYNAMIC_CHAIN_ADDRESS): Define.
794 (EH_RETURN_DATA_REGNO): Define.
795 (EH_RETURN_STACKADJ_RTX): Define.
796 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
797 patterns for dwarf exception handling.
799 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
801 * config/nds32/nds32.h: Clean up obsolete macros.
803 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
805 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
806 Add enum values for particular instructions.
807 * config/nds32/nds32-intrinsic.c: Implementation of expanding
808 particular intrinsic functions.
809 * config/nds32/nds32-intrinsic.md: Likewise.
810 * config/nds32/nds32_intrinsic.h: Likewise.
811 * config/nds32/nds32.h (nds32_builtins): Likewise.
812 * config/nds32/nds32.md (type): Add pbsad and pbsada.
813 (btst, ave): New patterns for particular instructions.
815 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
817 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
818 Add enum values for atomic load/store and memory sync.
819 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
821 * config/nds32/nds32-intrinsic.md: Likewise.
822 * config/nds32/nds32_intrinsic.h: Likewise.
823 * config/nds32/nds32.h (nds32_builtins): Likewise.
825 2018-04-07 Jakub Jelinek <jakub@redhat.com>
827 PR tree-optimization/85257
828 * fold-const.c (native_encode_vector): If not all elts could fit
829 and off is -1, return 0 rather than offset.
830 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
831 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
832 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
833 adjust buffer in native_interpret_expr call.
835 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
837 * config/nds32/constants.md (unspec_volatile_element): Add cache
839 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
840 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
841 * config/nds32/nds32.c (nds32_cctl_names): New.
842 (nds32_print_operand): Handle cache control register names.
843 * config/nds32/nds32.h (nds32_builtins): New enum values.
844 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
846 * config/nds32/nds32.md (type): Add mmu.
847 * config/nds32/pipelines.md (simple_insn): Add mmu.
849 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
851 * config/nds32/nds32.md (type): Remove call.
852 * config/nds32/pipelines.md (simple_insn): Likewise.
854 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
856 * config/nds32/constants.md (unspec_volatile_element): Add
857 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
858 UNSPEC_VOLATILE_FMFCFG.
859 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
860 description for fmfcfg and fmfcsr.
861 (bdesc_1arg): Add fmtcsr.
862 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
863 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
864 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
865 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
866 unspec_fmfcfg): New patterns.
867 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
868 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
869 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
870 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
871 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
872 __nds32__fmfcfg): Define.
874 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
876 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
877 intrinsic register names.
878 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
879 intrinsic register enum values and macros.
881 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
883 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
884 for load/store addressing form.
885 (nds32_print_operand_address): Likewise.
887 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
890 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
891 based on LABEL_REF. Remove useless assertion.
892 (pic_address_needs_scratch): Fix formatting.
893 (sparc_legitimize_pic_address): Minor tweaks.
894 (sparc_delegitimize_address): Adjust assertion accordingly.
895 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
896 into symbolic_operand.
897 (movsi_high_pic_label_ref): Likewise.
898 (movsi_lo_sum_pic_label_ref): Likewise.
899 (movdi_pic_label_ref): Likewise.
900 (movdi_high_pic_label_ref): Likewise.
901 (movdi_lo_sum_pic_label_ref): Likewise.
903 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
905 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
906 custom LIB_SPEC setup.
908 2018-04-06 Ruslan Bukin <br@bsdpad.com>
909 Kito Cheng <kito.cheng@gmail.com>
911 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
912 * config/riscv/freebsd.h: New.
914 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
916 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
917 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
920 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
921 Kito Cheng <kito.cheng@gmail.com>
923 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
924 nds32_output_call, nds32_symbol_binds_local_p): New functions.
925 * config/nds32/nds32-protos.h (nds32_output_call,
926 nds32_output_return): Declare.
927 * config/nds32/nds32.md: Refine all the call and return patterns.
929 2018-04-06 Jakub Jelinek <jakub@redhat.com>
932 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
933 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
935 PR rtl-optimization/84872
936 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
937 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
940 2018-04-06 Tamar Christina <tamar.christina@arm.com>
942 * expr.c (copy_blkmode_to_reg): Revert 254862.
943 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
945 2018-04-06 Richard Biener <rguenther@suse.de>
948 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
949 after seeing a component reference with an adjacent field. Treat
950 refs to arrays at struct end of external decls similar to
951 refs to unconstrained commons.
953 2018-04-06 Jakub Jelinek <jakub@redhat.com>
956 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
957 look through SAVE_EXPRs with non-side-effects argument. Adjust
959 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
962 2018-04-06 Richard Biener <rguenther@suse.de>
965 * alias.c (find_base_term): New wrapper around find_base_term
966 unwinding CSELIB_VAL_PTR changes.
967 (find_base_term): Do not restore CSELIB_VAL_PTR during the
970 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
972 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
974 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
975 constant definitions.
976 ("nop"): lr 0,0 -> nopr r0
977 ("nop_lr0", "nop_lr1"): New insn definitions.
979 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
981 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
982 NDS32_V3PUSH_AVAILABLE_P macro.
984 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
985 Chung-Ju Wu <jasonwucj@gmail.com>
987 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
988 (nds32*-*-*): Add float and fpu_config into supported_defaults.
989 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
990 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
991 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
992 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
993 * config/nds32/constraints.md: New constraints and checking for hard
995 * config/nds32/iterators.md: New mode iterator and attribute for hard
997 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
999 * config/nds32/nds32-fpu.md: New file.
1000 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1001 deal with hard float code generation.
1002 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1004 (abi_type, float_reg_number): New enum type.
1005 * config/nds32/nds32-predicates.c: New predicates for hard float.
1006 * config/nds32/nds32-protos.h: Declare functions for hard float.
1007 * config/nds32/nds32.c: Implementation for hard float configuration.
1008 * config/nds32/nds32.h: Definitions for hard float configuration.
1009 * config/nds32/nds32.md: Include hard float machine description and
1010 modify patterns for hard float configuration.
1011 * config/nds32/nds32.opt: New options for hard float configuration.
1012 * config/nds32/predicates.md: New predicates for hard float
1015 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1017 * common/config/nds32/nds32-common.c
1018 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1020 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1023 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1024 CONSTRUCTOR_ELT (ctor, ...)->value.
1026 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1029 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1031 2018-04-05 Tom de Vries <tom@codesourcery.com>
1034 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1037 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1038 Kito Cheng <kito.cheng@gmail.com>
1040 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1041 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1042 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1043 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1045 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1046 Kito Cheng <kito.cheng@gmail.com>
1048 * config/nds32/constraints.md (Ufe): New memory constraint.
1049 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1050 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1051 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1053 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1054 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1056 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1058 * config/nds32/nds32.md: Use optimize_size in the condition for
1059 alu-shift instructions.
1061 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1063 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1065 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1067 * config/nds32/nds32.md (negsi2): Refine pattern.
1069 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1070 Chung-Ju Wu <jasonwucj@gmail.com>
1072 * config/nds32/iterators.md (shift_rotate): New code iterator.
1073 (shift): New code attribute.
1074 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1075 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1076 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1077 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1078 bit-wise operations.
1079 (andsi3, *andsi3): Ditto.
1080 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1081 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1082 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1083 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1084 nds32_ior_operand, nds32_xor_operand): New predicates.
1086 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1088 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1089 (addsi3, subsi3): ... this.
1091 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1093 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1095 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1097 * config/nds32/nds32.md: Adjust indention.
1099 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1101 * config/nds32/nds32.md (feature): New attribute.
1103 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1105 * config/nds32/nds32.md (subtype): New attribute.
1107 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1110 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1111 expansion to perform a bitwise AND of the argument followed by a
1112 boolean negation of the result.
1114 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1116 PR rtl-optimization/84878
1117 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1118 the basic block. Assert the use reference is not artificial and that
1119 it has an associated insn.
1121 2018-04-04 Michael Matz <matz@suse.de>
1123 * builtins.c (compute_objsize): Pass correct operand
1124 to array_at_struct_end_p.
1126 2018-04-04 Richard Biener <rguenther@suse.de>
1129 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1130 from contexts for DINFO_LEVEL_TERSE and below.
1132 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1134 * config/nds32/nds32-doubleword.md (move_<mode>): Require
1135 resiter_operand condition.
1136 * config/nds32/nds32.md (*move<mode>): Ditto.
1138 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1139 Monk Chiang <sh.chiang04@gmail.com>
1141 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1143 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1145 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1147 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1148 Kito Cheng <kito.cheng@gmail.com>
1150 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1151 nds32_cond_code_str, output_cond_branch,
1152 output_cond_branch_compare_zero, nds32_expand_cbranch,
1153 nds32_expand_cstore, nds32_expand_movcc,
1154 nds32_output_cbranchsi4_equality_zero,
1155 nds32_output_cbranchsi4_equality_reg,
1156 nds32_output_cbranchsi4_equality_reg_or_const_int,
1157 nds32_output_cbranchsi4_greater_less_zero: New functions.
1158 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1159 nds32_expand_cstore, nds32_expand_movcc,
1160 nds32_output_cbranchsi4_equality_zero,
1161 nds32_output_cbranchsi4_equality_reg,
1162 nds32_output_cbranchsi4_equality_reg_or_const_int,
1163 nds32_output_cbranchsi4_greater_less_zero): Declare.
1164 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1165 nds32_rimm11s_operand): New predicates.
1166 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1167 * config/nds32/nds32.md: Rewrite all the branch and conditional move
1170 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1172 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
1173 * config/nds32/nds32.md: Ditto.
1174 * config/nds32/pipelines.md: Ditto.
1176 2018-04-04 Richard Biener <rguenther@suse.de>
1178 PR tree-optimization/85168
1179 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
1180 propagating abnormals.
1182 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1184 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
1186 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1187 Kito Cheng <kito.cheng@gmail.com>
1189 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
1190 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
1191 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
1192 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
1193 * config/nds32/nds32.md (sibcall_internal): New.
1194 (sibcall_register): Remove.
1195 (sibcall_immediate): Remove.
1196 (sibcall_value_internal): New.
1197 (sibcall_value_register): Remove.
1198 (sibcall_value_immediate): Remove.
1199 * config/nds32/predicates.md (nds32_general_register_operand): New.
1200 (nds32_call_address_operand): New.
1202 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1204 PR rtl-optimization/85167
1205 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1206 bb_defs if *split_p, instead preinitialize it to NULL.
1208 PR tree-optimization/85156
1209 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1210 evaluating the argument multiple times.
1212 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
1214 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1216 (_mm_cvtpd_ps): Likewise.
1217 (_mm_cvttpd_epi32): Likewise.
1218 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1219 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1220 vector, pixel, and bool following altivec.h include.
1222 2018-04-03 Martin Sebor <msebor@redhat.com>
1224 * doc/extend.texi (Common Function Attributes): Clarify.
1225 (const attribute): Likewise.
1226 (pure attribute): Likewise.
1228 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1231 * config/i386/i386.c (ix86_expand_vector_set): Use
1232 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
1234 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
1236 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1237 instructions when changing rounding bits to preserve precision bits
1238 in the x87 control word.
1240 2018-04-03 Martin Liska <mliska@suse.cz>
1242 PR tree-optimization/82491
1243 * rtl.h (strip_offset_and_add): Replace += suboffset with
1244 poly_uint64 () + suboffset.
1246 2018-03-29 Martin Liska <mliska@suse.cz>
1247 Martin Jambor <mjambor@suse.cz>
1250 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1251 param_type is not an integral or pointer type.
1253 2018-04-03 Richard Biener <rguenther@suse.de>
1255 * sese.h (recompute_all_dominators): Remove.
1257 2018-04-02 Martin Sebor <msebor@redhat.com>
1259 * doc/invoke.texi (-Wrestrict): Fix typos.
1261 2018-04-02 Jim Wilson <jimw@sifive.com>
1263 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1264 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1265 (<optab>di3, <optab>si3_extend): Likewise.
1266 (<optab>si3_mask, <optab>si3_mask_1): New.
1267 (<optab>di3_mask, <optab>di3_mask_1): New.
1268 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1269 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1270 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1272 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1274 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1277 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1279 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1280 (nds32_canonicalize_comparison): New function.
1282 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1283 Kito Cheng <kito.cheng@gmail.com>
1284 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1286 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1287 * config/nds32/constants.md (unspec_volatile_element): Add
1288 UNSPEC_VOLATILE_RELAX_GROUP.
1289 * config/nds32/nds32-relax-opt.c: New file.
1290 * config/nds32/nds32-predicates.c
1291 (nds32_symbol_load_store_p): New function.
1292 * config/nds32/nds32-protos.h
1293 (nds32_symbol_load_store_p): Declare function.
1294 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1295 * config/nds32/nds32.c
1296 (nds32_register_pass): New function to register pass.
1297 (nds32_register_passes): New function to register passes.
1298 * config/nds32/nds32.md (relax_group): New pattern.
1299 * config/nds32/nds32.opt (mrelax-hint): New option.
1300 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1302 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
1304 * config/nds32/t-nds32: Modify files dependency.
1306 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1308 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1309 (PROFILE_HOOK): Define its implementation.
1311 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1313 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1314 type and 32-bit size.
1316 2018-04-01 Jakub Jelinek <jakub@redhat.com>
1319 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1320 (V_128_256): New mode iterator.
1321 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1322 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1323 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1325 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1326 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1328 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1331 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1332 NaN inputs correctly.
1334 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1337 * config/rs6000/vsx.md (??r): New mode attribute.
1338 (*vsx_mov<mode>_64bit): Use it.
1339 (*vsx_mov<mode>_32bit): Likewise.
1341 2018-03-30 Martin Sebor <msebor@redhat.com>
1343 PR tree-optimization/84818
1344 * builtins.c (check_access): Use warning_n.
1346 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1349 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1351 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1354 2018-03-30 Julia Koval <julia.koval@intel.com>
1357 * x86-tune.def (movx, partial_reg_dependency): Enable for
1360 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
1363 * lra-constraints.c (process_alt_operands): Move setting
1364 this_alternative_matches below.
1366 2018-03-29 Martin Liska <mliska@suse.cz>
1369 * doc/invoke.texi: Document how LTO works with debug info.
1370 Describe auto-load support of binutils. Mention 'x86-64'
1371 as valid option value of -march option.
1373 2018-03-29 Jakub Jelinek <jakub@redhat.com>
1375 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
1378 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
1379 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
1380 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
1383 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1386 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
1387 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
1388 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
1389 for RS6000_BTM_POWERPC64.
1390 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
1391 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
1392 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
1397 2018-03-28 Carl Love <cel@us.ibm.com>
1400 2017-09-27 Carl Love <cel@us.ibm.com>
1402 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
1403 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
1404 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
1407 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1409 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
1410 instead of __vector bool.
1411 (_mm_max_pu8): Likewise.
1412 (_mm_min_pi16): Likewise.
1414 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1417 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
1418 (DIVWEUO): Likewise.
1420 (DIVDEUO): Likewise.
1421 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
1422 DIVWEUO and DIVDEUO.
1423 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
1424 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
1425 (div_extend): Likewise.
1426 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
1428 (__builtin_divweuo): Likewise.
1429 (__builtin_divdeo): Likewise.
1430 (__builtin_divdeuo): Likewise.
1432 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1435 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
1436 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
1438 PR tree-optimization/82004
1439 * gimple-match-head.c (optimize_pow_to_exp): New function.
1440 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
1441 Don't fold to exp if optimize_pow_to_exp is false.
1443 2018-03-28 Martin Liska <mliska@suse.cz>
1446 * calls.c (initialize_argument_information): Fix trailing space.
1447 * common.opt: Fix typo and provide better explanation for
1448 -fsanitize-coverage option.
1449 * config/i386/i386.opt: Fix typo.
1451 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1452 Martin Liska <mliska@suse.cz>
1455 * gimplify.c (asan_poison_variable): Don't do the check for
1456 gimplify_omp_ctxp here.
1457 (gimplify_decl_expr): Do it here.
1458 (gimplify_target_expr): Likewise.
1460 2018-03-28 Martin Liska <mliska@suse.cz>
1463 * config/i386/i386.c (ix86_function_arg_advance): Do not call
1464 chkp_type_bounds_count if MPX is not enabled.
1466 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
1468 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
1470 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
1473 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
1474 function to create the function decl for complex long double
1475 multiply and divide for -mabi=ieeelongdouble.
1476 (init_float128_ieee): Call it.
1478 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
1481 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
1482 -fcf-protection=branch -mibt.
1483 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
1485 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1488 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
1490 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
1493 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
1494 extern array declarations.
1496 2018-03-27 Richard Biener <rguenther@suse.de>
1499 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
1500 explicit single_use checks.
1502 2018-03-27 Richard Biener <rguenther@suse.de>
1504 PR tree-optimization/85082
1505 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
1508 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1510 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
1511 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
1512 Turn on fasynchronous-unwind-tables and funwind-tables.
1514 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
1517 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
1518 (*bmi_blsr_<mode>_ccz): Ditto.
1520 2018-03-26 Tom de Vries <tom@codesourcery.com>
1522 PR tree-optimization/85063
1523 * omp-general.c (offloading_function_p): New function. Factor out
1525 * omp-offload.c (pass_omp_target_link::gate): ... here.
1526 * omp-general.h (offloading_function_p): Declare.
1527 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
1528 with attribute omp declare target for offloading functions.
1530 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
1532 PR tree-optimization/84005
1533 * tree-data-ref.h (get_base_for_alignment): Declare.
1534 * tree-data-ref.c (get_base_for_alignment_1): New function.
1535 (get_base_for_alignment): Likewise.
1536 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
1537 get_base_for_alignment to find a suitable base object, instead
1538 of always using drb->base_address.
1540 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1543 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
1544 known size by default.
1546 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
1549 * lra-constraints.c (process_alt_operands): Don't match BLKmode
1550 and non BLKmode operands.
1552 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1555 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
1556 Clean up attributes.
1558 2018-03-23 Richard Biener <rguenther@suse.de>
1561 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
1562 we are going to emit early debug for LTO.
1564 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1567 * function.c (match_asm_constraints_1): Don't optimize if input
1568 doesn't satisfy general_operand predicate for output's mode.
1571 * alias.c (write_dependence_p): Don't require for x_canonicalized
1572 non-VOIDmode if x has VOIDmode.
1575 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
1576 just don't try to optimize it rather than assert it never happens.
1578 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1580 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
1581 macro expansions for definition of ST_INTERNAL_<mode> and
1582 LD_INTERNAL_<mode> builtins.
1583 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
1585 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
1587 (altivec_expand_st_builtin): Likewise.
1588 (altivec_expand_builtin): Remove calls to deleted functions.
1589 (rs6000_address_for_altivec): Delete this function.
1590 * config/rs6000/vector.md: Remove expands for
1591 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
1593 2018-03-22 Sudakshina Das <sudi.das@arm.com>
1596 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
1597 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
1598 re-computing once computed.
1599 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
1600 (arm_init_machine_status): Initialize
1601 machine->static_chain_stack_bytes.
1603 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1606 * doc/extend.texi: Add four new prototypes for vec_ld.
1607 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
1608 definitions for more logical presentation.
1609 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
1610 entries for V1TI variants of __builtin_altivec_ld builtin.
1611 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
1612 handling of V1TI variant of LVX icode pattern.
1613 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
1614 (rs6000_gimple_fold_builtin): Likewise.
1615 (altivec_init_builtins): Add code to define
1616 __builtin_altivec_lvx_v1ti function.
1618 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1621 * function.c (match_asm_constraints_1): Don't do the optimization
1622 if input isn't a REG, SUBREG, MEM or constant.
1624 2018-03-22 Tom de Vries <tom@codesourcery.com>
1626 PR tree-optimization/84956
1627 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
1628 bb_has_abnormal_pred.
1630 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1633 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
1634 DECL_INITIAL (decl) to decl at the end.
1635 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
1638 2018-03-21 Joseph Myers <joseph@codesourcery.com>
1640 * doc/extend.texi (__builtin_tgmath): Document when complex
1641 integer types are treated as _Complex _Float64.
1643 2018-03-21 Tom de Vries <tom@codesourcery.com>
1645 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
1647 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1649 PR tree-optimization/84960
1650 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
1651 if it is ENTRY block, move them into single succ of ENTRY in that case.
1653 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
1655 PR tree-optimization/84811
1656 * poly-int.h (poly_span_traits): Remove the T3 parameter and
1657 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
1658 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
1659 (known_subrange_p): Update accordingly. Cast each value involved
1660 in the size comparison, rather than casting the result of the
1663 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1665 PR tree-optimization/84982
1666 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
1667 by flipping the least significant bit rather than all bits from
1668 bitpos to bitpos + bitsize - 1.
1670 2018-03-21 Nathan Sidwell <nathan@acm.org>
1672 * doc/extend.texi (Deprecated Features): Remove mention of
1673 long-deleted deprecations.
1675 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1678 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
1679 * configure: Regenerate.
1681 2018-03-21 Tom de Vries <tom@codesourcery.com>
1683 PR tree-optimization/83126
1684 * tree-parloops.c (num_phis): New function.
1685 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
1687 2018-03-21 Nathan Sidwell <nathan@acm.org>
1689 * doc/extend.texi (Deprecated Features): Update deprecated flags,
1690 mention anon-struct/union members and trailing attributes.
1692 2018-03-21 Bin Cheng <bin.cheng@arm.com>
1694 PR tree-optimization/84969
1695 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
1696 builtin memset partitions if they set different rhs values.
1698 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1700 PR rtl-optimization/84989
1701 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
1702 VEC_DUPLICATE with scalar result mode.
1704 2018-03-21 Martin Liska <mliska@suse.cz>
1707 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
1708 not intended return statement.
1710 2018-03-21 Martin Liska <mliska@suse.cz>
1713 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
1714 (chkp_find_bound_slots_1): Limit number of iterations.
1716 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
1719 * Minor grammar fixes for x86 options.
1721 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1724 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
1725 holding REG_CFA_RESTORE notes, instead turn them into a USE.
1727 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
1730 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
1731 (altivec_lvx_<mode>_1op): Likewise.
1732 (altivec_stvx_<mode>_2op): Likewise.
1733 (altivec_stvx_<mode>_1op): Likewise.
1734 (altivec_lvx_<VM2:mode>): New define_expand.
1735 (altivec_stvx_<VM2:mode>): Likewise.
1736 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
1737 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1738 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
1739 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1740 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
1741 (rs6000_gen_lvx): Likewise.
1742 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
1743 (altivec_expand_stv_builtin): Likewise.
1744 (altivec_expand_builtin): Likewise.
1745 * config/rs6000/vector.md: Likewise.
1747 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1750 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
1753 2018-03-20 Richard Biener <rguenther@suse.de>
1756 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
1757 sign-conversions as zero, fall back to standard scalar_stmt
1760 2018-03-20 Martin Liska <mliska@suse.cz>
1763 * predict.c (rebuild_frequencies): Handle case when we have
1764 PROFILE_ABSENT, but flag_guess_branch_prob is false.
1766 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1769 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
1770 flag_section_anchors.
1771 * varasm.c (use_blocks_for_decl_p): Remove hack for
1772 dw2_force_const_mem.
1775 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
1777 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
1778 be created, use lowpart_subreg of operands[0] rather than operands[0]
1780 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
1781 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
1782 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
1783 and n constraint instead of aarch64_shift_imm_di and Usd.
1784 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
1785 (*aarch64_<optab>_reg_minus<mode>3): ... this.
1787 2018-03-20 Sudakshina Das <sudi.das@arm.com>
1790 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
1791 to favor GPR over NEON registers.
1792 (<shift>di3_neon): Likewise.
1794 2018-03-20 Tom de Vries <tom@codesourcery.com>
1797 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
1798 (nvptx_process_pars): Emit bar.sync asap and alap.
1800 2018-03-20 Tom de Vries <tom@codesourcery.com>
1803 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
1804 seen_label if seen_label is already set.
1806 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1809 * config/i386/i386.c (fold_builtin_cpu): For features above 31
1810 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
1811 Use 1U instead of 1. Formatting fixes.
1814 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
1815 instead of TREE_TYPE (s1) for the return value.
1817 2018-03-19 Jakub Jelinek <jakub@redhat.com>
1819 PR tree-optimization/84946
1820 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
1821 bitsize + bitsize in poly_uint64 rather than poly_int64.
1824 * dwarf2asm.c: Include fold-const.c.
1825 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
1826 of decl rather than decl itself.
1828 PR rtl-optimization/84643
1829 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
1831 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
1834 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
1835 calling assemble_variable.
1837 2018-03-19 Sudakshina Das <sudi.das@arm.com>
1840 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
1841 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
1843 2018-03-19 Jim Wilson <jimw@sifive.com>
1846 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
1847 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
1848 (riscv_first_stack_step): Likewise.
1849 (riscv_option_override): Use STACK_BOUNDARY instead of
1851 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
1853 (BIGGEST_ALIGNMENT): Set to 128.
1854 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
1855 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
1858 2018-03-19 Richard Biener <rguenther@suse.de>
1860 PR tree-optimization/84933
1861 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
1862 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
1864 2018-03-19 Richard Biener <rguenther@suse.de>
1866 PR tree-optimization/84859
1867 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
1868 (cond_if_else_store_replacement): Perform sinking operation on
1869 single-store BBs regardless of MAX_STORES_TO_SINK setting.
1870 Generalize what a BB with a single eligible store is.
1872 2018-03-19 Richard Biener <rguenther@suse.de>
1874 PR tree-optimization/84929
1875 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
1876 chrec_is_positive against non-chrec arg.
1878 2018-03-19 Tamar Christina <tamar.christina@arm.com>
1881 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
1883 2018-03-18 Martin Liska <mliska@suse.cz>
1885 PR rtl-optimization/84635
1886 * regrename.c (build_def_use): Use matches_mode only when
1889 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
1891 PR tree-optimization/84913
1892 * tree-vect-loop.c (vectorizable_reduction): Don't try to
1893 vectorize chains of COND_EXPRs.
1895 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1897 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
1899 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1901 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
1903 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1905 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
1907 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1908 Kito Cheng <kito.cheng@gmail.com>
1910 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
1911 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
1912 (nds32_adjust_reg_alloc_order): New function.
1913 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
1915 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
1917 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
1918 nds32_print_operand, nds32_print_operand_address): Use
1919 HOST_WIDE_INT_PRINT_DEC instead.
1921 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1923 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
1925 2018-03-17 Jakub Jelinek <jakub@redhat.com>
1928 * config/i386/i386.c (initial_ix86_tune_features,
1929 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
1931 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
1932 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
1933 rather than 1u << ix86_tune. Formatting fix.
1934 (ix86_option_override_internal): Change ix86_arch_mask from
1935 unsigned int to unsigned HOST_WIDE_INT, initialize to
1936 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
1937 (ix86_function_specific_restore): Likewise.
1939 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1942 * postreload.c (reload_combine_recognize_pattern): Perform
1943 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
1944 truncate_int_for_mode the result for the destination's mode.
1947 * hsa-gen.c (mem_type_for_type): Fix comment typo.
1948 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
1950 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1953 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
1956 * lra-assigns.c (lra_split_hard_reg_for): Don't use
1957 regno_allocno_class_array and sorted_pseudos.
1958 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
1959 insns where regno is used.
1961 2018-03-16 Martin Liska <mliska@suse.cz>
1964 * multiple_target.c (create_dispatcher_calls): Redirect
1965 reference in the symbol table.
1967 2018-03-16 Martin Liska <mliska@suse.cz>
1970 * multiple_target.c (create_dispatcher_calls): Redirect also
1973 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1977 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
1978 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
1979 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
1981 2018-03-16 Julia Koval <julia.koval@intel.com>
1983 * doc/invoke.texi (Skylake Server): Add CLWB.
1984 Cannonlake): Remove CLWB.
1986 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1988 PR tree-optimization/84841
1989 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
1991 (FLOAT_ONE_CONST_TYPE): Define.
1992 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
1993 (sort_by_operand_rank): Put entries with higher constant_type last
1994 rather than first to match comments.
1996 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
1998 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2001 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2004 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2005 check and use address of target always.
2007 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2010 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2011 (indirect_thunk_bnd_needed): Likewise.
2012 (indirect_thunks_used): Likewise.
2013 (indirect_thunks_bnd_used): Likewise.
2014 (indirect_return_needed): New.
2015 (indirect_return_bnd_needed): Likewise.
2016 (output_indirect_thunk_function): Add a bool argument for
2018 (output_indirect_thunk_function): Don't generate alias for
2019 function return thunk.
2020 (ix86_code_end): Call output_indirect_thunk_function to generate
2021 function return thunks.
2022 (ix86_output_function_return): Set indirect_return_bnd_needed
2023 and indirect_return_needed instead of indirect_thunk_bnd_needed
2024 and indirect_thunk_needed.
2026 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2028 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2029 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2030 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2032 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2033 Paul Hua <paul.hua.gm@gmail.com>
2036 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2038 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2040 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2041 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2044 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2047 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2048 instead of GET_MODE_SIZE when comparing Units.
2050 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2053 * varasm.c (hash_section): Return an unchangeble hash value
2054 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2055 Return !aarch64_can_use_per_function_literal_pools_p ().
2057 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2060 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2061 rather than address of cmode as last argument to prepare_cmp_insn.
2063 2018-03-15 Julia Koval <julia.koval@intel.com>
2065 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2066 F_AVX512VNNI, F_AVX512BITALG): New.
2068 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2071 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2072 insn for floating-point loads and stores.
2074 2018-03-14 Carl Love <cel@us.ibm.com>
2076 * config/rs6000/rs6000-c.c: Add macro definitions for
2077 ALTIVEC_BUILTIN_VEC_PERMXOR.
2078 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2079 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2080 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2081 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2083 * config/doc/extend.texi: Add prototypes for vec_permxor.
2085 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2088 * diagnostic-show-locus.c (class layout_point): Convert m_line
2089 from int to linenum_type.
2090 (line_span::comparator): Use linenum "compare" function when
2091 comparing line numbers.
2092 (test_line_span): New function.
2093 (layout_range::contains_point): Convert param "row" from int to
2095 (layout_range::intersects_line_p): Likewise.
2096 (layout::will_show_line_p): Likewise.
2097 (layout::print_source_line): Likewise.
2098 (layout::should_print_annotation_line_p): Likewise.
2099 (layout::print_annotation_line): Likewise.
2100 (layout::print_leading_fixits): Likewise.
2101 (layout::annotation_line_showed_range_p): Likewise.
2102 (struct line_corrections): Likewise for field m_row.
2103 (line_corrections::line_corrections): Likewise for param "row".
2104 (layout::print_trailing_fixits): Likewise.
2105 (layout::get_state_at_point): Likewise.
2106 (layout::get_x_bound_for_row): Likewise.
2107 (layout::print_line): Likewise.
2108 (diagnostic_show_locus): Likewise for locals "last_line" and
2110 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2111 * input.c (selftest::test_linenum_comparisons): New function.
2112 (selftest::input_c_tests): Call it.
2113 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2114 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2115 * selftest.h (ASSERT_GT): New macro.
2116 (ASSERT_GT_AT): New macro.
2117 (ASSERT_LT): New macro.
2118 (ASSERT_LT_AT): New macro.
2120 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
2122 PR rtl-optimization/84780
2123 * combine.c (distribute_links): Don't make a link based on pc_rtx.
2125 2018-03-14 Martin Liska <mliska@suse.cz>
2127 * tree.c (record_node_allocation_statistics): Use
2128 get_stats_node_kind.
2129 (get_stats_node_kind): New function extracted from
2130 record_node_allocation_statistics.
2131 (free_node): Use get_stats_node_kind.
2133 2018-03-14 Richard Biener <rguenther@suse.de>
2135 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2136 that the value-set of ANTIC_IN doesn't grow.
2139 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2141 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2142 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2144 2018-03-14 Julia Koval <julia.koval@intel.com>
2146 * config.gcc (icelake-client, icelake-server): New.
2148 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2149 (initial_ix86_arch_features): Ditto.
2150 (PTA_SKYLAKE): Add SGX.
2151 (PTA_ICELAKE): Remove.
2152 (PTA_ICELAKE_CLIENT): New.
2153 (PTA_ICELAKE_SERVER): New.
2154 (ix86_option_override_internal): Split up icelake on icelake client and
2156 (get_builtin_code_for_version): Ditto.
2157 (fold_builtin_cpu): Ditto.
2158 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2159 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2160 * config/i386/i386.h (processor_type): Ditto.
2161 * doc/invoke.texi: Ditto.
2163 2018-03-14 Jakub Jelinek <jakub@redhat.com>
2166 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2167 INTEGER_CST offset, add it together with bitpos / 8 and
2168 sign extend based on POINTER_SIZE.
2172 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
2175 * config/i386/constraints.md (Yc): New register constraint.
2176 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
2177 Use Yc constraint for alternative 2 of operand 0. Remove
2178 preferred_for_speed attribute.
2180 2018-03-14 Richard Biener <rguenther@suse.de>
2182 PR tree-optimization/84830
2183 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
2184 with the old one to avoid oscillations.
2186 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
2189 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
2191 (assign_by_spills): Return a flag of reload assignment failure.
2192 Do not process the reload assignment failures. Do not spill other
2193 reload pseudos if they has the same reg class. Update n if
2195 (lra_assign): Add a return arg. Set up from the result of
2196 assign_by_spills call.
2197 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2198 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2199 usage_insns if it is not NULL.
2200 (spill_hard_reg_in_range): New function.
2201 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2202 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2203 function prototypes.
2204 (lra_assign): Change prototype.
2205 * lra.c (lra): Add code to deal with fails by splitting hard reg
2208 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
2210 * config/riscv/riscv.opt (mrelax): New option.
2211 * config/riscv/riscv.c (riscv_file_start): Emit ".option
2212 "norelax" when riscv_mrelax is disabled.
2213 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2215 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2218 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2219 reassociation for int modes.
2221 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2223 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
2224 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
2226 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
2227 * config/aarch64/aarch64-sve.md
2228 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
2229 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2230 (*extend<mode><Vwide>2): Rename to...
2231 (aarch64_sve_extend<mode><Vwide>2): ...this.
2232 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2233 renaming the old pattern to...
2234 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
2236 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2237 define_expand, renaming the old pattern to...
2238 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2239 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2240 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2241 account when deciding which SVE instruction the optab should use.
2242 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2244 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2246 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2247 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2248 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2249 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2250 (tlsdesc_small_<mode>): Turn a define_expand and use
2251 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
2252 (tlsdesc_small_advsimd_<mode>): ...this.
2253 (tlsdesc_small_sve_<mode>): New pattern.
2255 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2257 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2258 (UNSPEC_UMUL_HIGHPART): New constants.
2259 (MUL_HIGHPART): New int iteraor.
2260 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2261 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2263 (*<su>mul<mode>3_highpart): New define_insn.
2265 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2268 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2271 2018-03-13 Martin Liska <mliska@suse.cz>
2274 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2276 (sem_item_optimizer::~sem_item_optimizer): Release it.
2277 (sem_item_optimizer::merge_classes): Register variable aliases.
2278 (sem_item_optimizer::fixup_pt_set): New function.
2279 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2280 * ipa-icf.h: Declare new variables and functions.
2282 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2285 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2286 integer_pow2p@2 and test integer_pow2p in condition.
2287 (A < 0 ? C : 0): Similarly for @1.
2290 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2291 characters starting at p contain '\0' character, don't look beyond
2295 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2296 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2299 * reg-stack.c (change_stack): Change update_end var from int to
2300 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2301 also call set_block_for_insn on the newly added insns and rescan.
2304 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2305 on the last operand.
2308 * tree.c (stabilize_reference_1): Return save_expr (e) for
2309 STATEMENT_LIST even if it doesn't have side-effects.
2311 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
2313 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2315 2018-03-12 Renlin Li <renlin.li@arm.com>
2317 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2318 aarch64_output_scalar_simd_mov_immediate.
2320 2018-03-12 Martin Sebor <msebor@redhat.com>
2322 PR tree-optimization/83456
2323 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2324 for perfectly overlapping calls to memcpy.
2325 (gimple_fold_builtin_memory_chk): Same.
2326 (gimple_fold_builtin_strcpy): Handle no-warning.
2327 (gimple_fold_builtin_stxcpy_chk): Same.
2328 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2330 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2332 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2333 parameter. Use it for SFmode.
2334 (rs6000_function_arg_advance_1): Adjust.
2335 (rs6000_function_arg): Adjust.
2336 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2338 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2340 PR rtl-optimization/84169
2341 PR rtl-optimization/84780
2342 * combine.c (can_combine_p): Check for a 2-insn combination whether
2343 the destination register is used between the two insns, too.
2345 2018-03-12 Richard Biener <rguenther@suse.de>
2347 PR tree-optimization/84803
2348 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2349 for refs DR analysis didn't process.
2351 2018-03-12 Richard Biener <rguenther@suse.de>
2353 PR tree-optimization/84777
2354 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2355 force-vectorize loops ignore whether we are optimizing for size.
2357 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
2359 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2360 (TARGET_MD_ASM_ADJUST): Define.
2362 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
2363 Kito Cheng <kito.cheng@gmail.com>
2364 Chung-Ju Wu <jasonwucj@gmail.com>
2366 * config/nds32/nds32.c (nds32_compute_stack_frame,
2367 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2368 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2369 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2370 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2371 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2372 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2373 * config/nds32/nds32.md (prologue, epilogue): Use macro
2374 NDS32_V3PUSH_AVAILABLE_P to do checking.
2376 2018-03-11 Jakub Jelinek <jakub@redhat.com>
2379 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
2380 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
2381 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
2382 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
2383 addition of most attributes on !orig_type_die or the attribute not
2384 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
2386 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2387 Chung-Ju Wu <jasonwucj@gmail.com>
2389 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
2391 * config/nds32/nds32.opt (mvh): New option.
2393 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2394 Chung-Ju Wu <jasonwucj@gmail.com>
2396 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
2398 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
2399 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
2402 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2403 Chung-Ju Wu <jasonwucj@gmail.com>
2405 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
2407 * config/nds32/nds32-multiple.md (strlensi): New pattern.
2408 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
2410 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
2411 Kito Cheng <kito.cheng@gmail.com>
2412 Chung-Ju Wu <jasonwucj@gmail.com>
2414 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
2415 UNSPEC_FFMISM and UNSPEC_FLMISM.
2416 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
2417 for ffb, ffmism and flmism.
2418 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
2419 (unspec_ffmism): Ditto.
2420 (unspec_flmism): Ditto.
2421 (nds32_expand_builtin_impl): Check if string extension is available.
2422 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
2423 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
2425 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
2428 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2431 * lra-assigns.c (assign_by_spills): Return a flag of reload
2432 assignment failure. Do not process the reload assignment
2433 failures. Do not spill other reload pseudos if they has the same
2435 (lra_assign): Add a return arg. Set up from the result of
2436 assign_by_spills call.
2437 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2438 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2439 usage_insns if it is not NULL.
2440 (spill_hard_reg_in_range): New function.
2441 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2442 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2443 function prototypes.
2444 (lra_assign): Change prototype.
2445 * lra.c (lra): Add code to deal with fails by splitting hard reg
2448 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
2451 * config/i386/i386.opt: Replace Enforcment with Enforcement.
2453 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
2456 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
2457 (dw_val_node): Add val_symbolic_view.
2458 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
2459 (symview_upper_bound): New.
2460 (new_line_info_table): Initialize symviews_since_reset.
2461 (dwarf2out_source_line): Count symviews_since_reset and set
2462 symview_upper_bound.
2463 (dw_val_equal_p): Handle symview.
2464 (add_AT_symview): New.
2465 (print_dw_val): Handle symview.
2466 (attr_checksum, attr_checksum_ordered): Likewise.
2467 (same_dw_val_p, size_of_die): Likewise.
2468 (value_format, output_die): Likewise.
2469 (add_high_low_attributes): Use add_AT_symview for entry_view.
2470 (dwarf2out_finish): Reset symview_upper_bound, clear
2473 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
2476 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
2477 Add strict argument and use it.
2478 (rs6000_split_multireg_move): Update for new strict argument.
2479 (mem_operand_gpr): Disallow all non-offsettable addresses.
2480 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
2482 2018-03-09 Jakub Jelinek <jakub@redhat.com>
2485 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
2486 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
2487 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
2490 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
2491 decl, use remap_type if we want to use the type.
2493 2018-03-09 Martin Sebor <msebor@redhat.com>
2495 PR tree-optimization/84526
2496 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2498 (builtin_access::generic_overlap): Be prepared to handle non-array
2501 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
2503 PR rtl-optimization/84682
2504 * lra-constraints.c (process_address_1): Check is_address flag
2505 for address constraints.
2506 (process_alt_operands): Likewise.
2507 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
2508 preprocess_constraints.
2509 * recog.h (preprocess_constraints): Add oploc parameter.
2511 * recog.c (preprocess_constraints): Test address_operand for
2512 CT_ADDRESS constraints.
2514 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2517 * lra-assigns.c (assign_by_spills): Return a flag of reload
2518 assignment failure. Do not process the reload assignment
2519 failures. Do not spill other reload pseudos if they has the same
2521 (lra_assign): Add a return arg. Set up from the result of
2522 assign_by_spills call.
2523 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2524 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2525 usage_insns if it is not NULL.
2526 (spill_hard_reg_in_range): New function.
2527 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2528 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2529 function prototypes.
2530 (lra_assign): Change prototype.
2531 * lra.c (lra): Add code to deal with fails by splitting hard reg
2534 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2537 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
2538 Accept complain bool parameter. Only emit errors if it is true.
2539 (arm_parse_cpu_option_name): Likewise.
2540 (arm_target_thumb_only): Adjust callers of the above.
2541 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
2542 prototype to take a default true bool parameter.
2543 (arm_parse_arch_option_name): Likewise.
2545 2018-03-09 David Malcolm <dmalcolm@redhat.com>
2546 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2550 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
2551 * configure: Regenerate.
2552 * configure.ac ("linker --version-script option"): New.
2553 ("linker soname option"): New.
2555 2018-03-09 Richard Biener <rguenther@suse.de>
2557 PR tree-optimization/84775
2558 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
2559 immediate uses of predicate stmts and mark them modified.
2562 PR tree-optimization/84178
2563 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2565 (version_loop_for_if_conversion): Delay update_ssa call.
2566 (tree_if_conversion): Delay update_ssa until after predicate
2569 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
2572 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
2573 when the function accesses prior frames.
2575 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2578 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
2579 gen_llsym, otherwise call maybe_gen_llsym.
2582 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
2583 has ',' character inside of it.
2585 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2588 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
2589 as clobbering CC_REGNUM.
2591 2018-03-08 Richard Biener <rguenther@suse.de>
2594 * tree-scalar-evolution.c: Include tree-into-ssa.h.
2595 (follow_copies_to_constant): Do not follow SSA names registered
2598 2018-03-08 Richard Biener <rguenther@suse.de>
2600 PR tree-optimization/84178
2601 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2603 (version_loop_for_if_conversion): Delay update_ssa call.
2604 (tree_if_conversion): Delay update_ssa until after predicate
2607 2018-03-08 David Malcolm <dmalcolm@redhat.com>
2609 PR tree-optimization/84178
2610 * tree-if-conv.c (release_bb_predicate): Remove the
2611 the assertion that the stmts have NULL use_ops.
2612 Discard the statements, asserting that they haven't
2613 yet been added to a BB.
2615 2018-03-08 Richard Biener <rguenther@suse.de>
2617 PR tree-optimization/84746
2618 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
2619 (phi_translate): Pass in destination ANTIC_OUT set.
2620 (phi_translate_1): Likewise. For a simplified result lookup
2621 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
2622 (phi_translate_set): Adjust.
2623 (do_pre_regular_insertion): Likewise.
2624 (do_pre_partial_partial_insertion): Likewise.
2626 2018-03-08 Martin Liska <mliska@suse.cz>
2628 PR gcov-profile/84735
2629 * doc/gcov.texi: Document usage of profile files.
2630 * gcov-io.h: Document changes in the format.
2632 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
2636 * dwarf2out.c (struct dw_line_info_table): Update comments for
2638 (FORCE_RESET_NEXT_VIEW): New.
2639 (FORCE_RESETTING_VIEW_P): New.
2640 (RESETTING_VIEW_P): Check for -1 too.
2641 (ZERO_VIEW_P): Likewise.
2642 (new_line_info_table): Force-reset next view.
2643 (dwarf2out_begin_function): Likewise.
2644 (dwarf2out_source_line): Simplify zero_view_p initialization.
2645 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
2646 view directly. Omit view when omitting .loc at line 0.
2648 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2650 PR tree-optimization/84740
2651 * tree-switch-conversion.c (process_switch): Call build_constructors
2652 only if info.phi_count is non-zero.
2654 PR tree-optimization/84739
2655 * tree-tailcall.c (find_tail_calls): Check call arguments against
2656 DECL_ARGUMENTS (current_function_decl) rather than
2657 DECL_ARGUMENTS (func) when checking for tail recursion.
2659 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2661 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
2662 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
2663 Volker Reichelt's entry and add entries for people that perform
2664 GCC fuzzy testing and report numerous bugs.
2666 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
2669 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
2670 readonly data in sdata, if that is disabled.
2671 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
2672 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2673 -mreadonly-in-sdata option.
2675 2018-03-07 Martin Sebor <msebor@redhat.com>
2677 PR tree-optimization/84468
2678 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
2679 basic block when looking for nul assignment.
2681 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
2684 * except.h (output_function_exception_table): Adjust prototype.
2685 * except.c (output_function_exception_table): Remove FNNAME parameter
2686 and add SECTION parameter. Ouput one part of the table at a time.
2687 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
2688 the first part of the exception table and emit unwind directives.
2689 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
2690 (i386_pe_seh_cold_init): Likewise.
2691 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
2692 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
2693 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
2694 (ix86_output_call_insn): Emit a nop in one more case for SEH.
2695 * config/i386/winnt.c: Include except.h.
2696 (struct seh_frame_state): Add reg_offset, after_prologue and
2697 in_cold_section fields.
2698 (i386_pe_seh_end_prologue): Set seh->after_prologue.
2699 (i386_pe_seh_cold_init): New function.
2700 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
2701 to seh->in_cold_section.
2702 (seh_emit_push): Record the offset of the push.
2703 (seh_emit_save): Record the offet of the save.
2704 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
2705 Test seh->after_prologue to disregard the epilogue.
2706 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
2707 (i386_pe_end_cold_function): New function.
2709 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2712 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
2713 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
2716 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
2718 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
2719 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
2722 * multiple_target.c: Include tree-inline.h and intl.h.
2723 (expand_target_clones): Diagnose and fail if node->definition and
2724 !tree_versionable_function_p (node->decl).
2726 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
2728 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
2730 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
2731 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
2732 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
2734 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2737 * combine.c (try_combine): Use reg_or_subregno instead of handling
2738 just paradoxical SUBREGs and REGs.
2740 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
2742 * config/arc/arc.c (arc_finalize_pic): Remove function.
2743 (arc_must_save_register): We use single base PIC register, remove
2744 checks to save/restore the PIC register.
2745 (arc_expand_prologue): Likewise.
2746 * config/arc/arc-protos.h (arc_set_default_type_attributes):
2748 (arc_verify_short): Likewise.
2749 (arc_attr_type): Likewise.
2750 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
2751 (walk_stores): Likewise.
2752 (arc_address_cost): Make it static.
2753 (arc_verify_short): Likewise.
2754 (branch_dest): Likewise.
2755 (arc_attr_type): Likewise.
2756 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
2757 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
2758 (arc_final_prescan_insn): Remove inserting the nops due to
2759 hardware hazards. It is done in reorg step.
2760 (insn_length_variant_t): Remove.
2761 (insn_length_parameters_t): Likewise.
2762 (arc_insn_length_parameters): Likewise.
2763 (arc_get_insn_variants): Likewise.
2764 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
2766 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2769 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
2772 PR tree-optimization/84687
2773 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
2775 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
2777 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2779 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
2780 Rename to ppc_speculation_barrier.
2781 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
2782 __builtin_ppc_speculation_barrier.
2784 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2787 * combine.c (combine_simplify_rtx): Don't try to simplify if
2788 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
2791 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
2793 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
2794 to 32 bytes when compiling for POWER9.
2796 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2799 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
2800 regparm >= 3 with no arg reg available also for calls with
2801 flag_force_indirect_call. Pass decl to ix86_function_regparm.
2804 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
2806 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
2808 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
2811 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
2813 2018-03-05 Richard Biener <rguenther@suse.de>
2815 PR tree-optimization/84486
2816 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
2817 When inserting a __builtin_assume_aligned call set the LHS
2818 SSA name alignment info accordingly.
2820 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
2822 PR tree-optimization/84114
2823 * config/aarch64/aarch64.c (aarch64_reassociation_width)
2824 Avoid reassociation of FLOAT_MODE addition.
2826 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
2828 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
2829 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
2830 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
2831 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
2832 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
2833 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
2834 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
2836 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
2837 __builtin_ia32_wbinvd): New builtins.
2838 (SPECIAL_ARGS2): New.
2839 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
2840 (SPECIAL_ARGS2): New.
2841 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
2842 (ix86_valid_target_attribute_inner_p): Ditto.
2843 (ix86_init_mmx_sse_builtins): Add special_args2.
2844 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
2845 TARGET_WBNOINVD_P): New.
2846 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
2847 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
2848 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
2849 * config/i386/immintrin.h (_wbinvd): New intrinsic.
2850 * config/i386/pconfigintrin.h: New file.
2851 * config/i386/wbnoinvdintrin.h: Ditto.
2852 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
2853 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
2855 2018-03-05 Richard Biener <rguenther@suse.de>
2857 PR tree-optimization/84670
2858 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2860 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2861 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2862 (compute_antic_aux): Only assert the number of values in ANTIC_IN
2863 doesn't grow if all successors (recursively) were visited at least
2866 2018-03-05 Richard Biener <rguenther@suse.de>
2868 PR tree-optimization/84650
2869 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
2870 if executed in the loop pipeline.
2872 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
2874 * doc/configfiles.texi (Configuration Files): Move info about
2875 conditionalizing $target-protos.h to...
2876 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
2877 differs from $target-protos.h.
2879 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
2880 Chung-Ju Wu <jasonwucj@gmail.com>
2882 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
2883 * config/nds32/nds32-multiple.md (setmemsi): Define.
2884 * config/nds32/nds32-memory-manipulation.c
2885 (nds32_gen_dup_4_byte_to_word_value): New.
2886 (emit_setmem_word_loop): New.
2887 (emit_setmem_byte_loop): New.
2888 (nds32_expand_setmem_loop): New.
2889 (nds32_expand_setmem_loop_v3m): New.
2890 (nds32_expand_setmem_unroll): New.
2891 (nds32_expand_setmem): New.
2893 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2894 Chung-Ju Wu <jasonwucj@gmail.com>
2896 * config/nds32/nds32-memory-manipulation.c
2897 (nds32_emit_load_store): New.
2898 (nds32_emit_post_inc_load_store): New.
2899 (nds32_emit_mem_move): New.
2900 (nds32_emit_mem_move_block): New.
2901 (nds32_expand_movmemsi_loop_unknown_size): New.
2902 (nds32_expand_movmemsi_loop_known_size): New.
2903 (nds32_expand_movmemsi_loop): New.
2904 (nds32_expand_movmemsi_unroll): New.
2905 (nds32_expand_movmemqi): Rename ...
2906 (nds32_expand_movmemsi): ... to this.
2907 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
2908 (movmemsi): ... to this.
2909 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
2910 (nds32_expand_movmemsi): ... to this.
2912 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2913 Monk Chiang <sh.chiang04@gmail.com>
2914 Chung-Ju Wu <jasonwucj@gmail.com>
2916 * config/nds32/nds32-protos.h
2917 (nds32_expand_load_multiple): New arguments.
2918 (nds32_expand_store_multiple): Ditto.
2919 (nds32_valid_multiple_load_store): Rename ...
2920 (nds32_valid_multiple_load_store_p): ... to this.
2921 * config/nds32/nds32-memory-manipulation.c
2922 (nds32_expand_load_multiple): Refine implementation.
2923 (nds32_expand_store_multiple): Ditto.
2924 * config/nds32/nds32-multiple.md
2925 (load_multiple): Update nds32_expand_load_multiple interface.
2926 (store_multiple): Update nds32_expand_store_multiple interface.
2927 * config/nds32/nds32-predicates.c
2928 (nds32_valid_multiple_load_store): Rename ...
2929 (nds32_valid_multiple_load_store_p): ... to this and refine
2931 * config/nds32/predicates.md
2932 (nds32_load_multiple_and_update_address_operation): New predicate.
2933 (nds32_store_multiple_and_update_address_operation): New predicate.
2935 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2936 Chung-Ju Wu <jasonwucj@gmail.com>
2938 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
2939 (combo): New attribute.
2940 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
2942 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
2944 * config/nds32/nds32.opt: Change -mcmodel= default value.
2946 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
2947 Monk Chiang <sh.chiang04@gmail.com>
2948 Chung-Ju Wu <jasonwucj@gmail.com>
2950 * config/nds32/constants.md (unspec_element): New enum.
2951 * config/nds32/constraints.md (Umw): New constraint.
2952 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
2953 * config/nds32/nds32-intrinsic.md: Likewise.
2954 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
2955 (nds32_valid_smw_lwm_base_p): New.
2956 (nds32_output_smw_single_word): New.
2957 (nds32_output_lmw_single_word): New.
2958 (nds32_expand_unaligned_load): New.
2959 (nds32_expand_unaligned_store): New.
2960 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
2961 (nds32_output_smw_single_word): Declare.
2962 (nds32_output_lmw_single_word): Declare.
2963 (nds32_expand_unaligned_load): Declare.
2964 (nds32_expand_unaligned_store): Declare.
2965 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
2966 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
2967 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
2968 NDS32_BUILTIN_UASTORE_DW.
2969 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
2972 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
2973 Kito Cheng <kito.cheng@gmail.com>
2974 Chung-Ju Wu <jasonwucj@gmail.com>
2976 * config/nds32/nds32-intrinsic.c
2977 (nds32_expand_builtin_null_ftype_reg): Delete.
2978 (nds32_expand_builtin_reg_ftype_imm): Ditto.
2979 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
2980 (nds32_read_argument): New.
2981 (nds32_legitimize_target): Ditto.
2982 (nds32_legitimize_argument): Ditto.
2983 (nds32_check_constant_argument): Ditto.
2984 (nds32_expand_unop_builtin): Ditto.
2985 (nds32_expand_unopimm_builtin): Ditto.
2986 (nds32_expand_binop_builtin): Ditto.
2987 (nds32_builtin_decl_impl): Ditto.
2988 (builtin_description): Ditto.
2989 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
2990 (nds32_init_builtins_impl): Ditto.
2991 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
2992 (nds32_builtin_decl): New.
2993 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
2994 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
2996 2018-03-02 Jeff Law <law@redhat.com>
2998 * reorg.c (stop_search_p): Handle DEBUG_INSN.
2999 (redundant_insn, fill_simple_delay_slots): Likewise.
3000 (fill_slots_from_thread): Likewise.
3001 * resource.c (mark_referenced_resources): Likewise.
3002 (mark_set_resources, find_dead_or_set_registers): Likewise.
3004 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3006 * substring-locations.h (format_warning_va): Formatting fix for
3008 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3010 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3011 * substring-locations.c: Include intl.h.
3012 (format_warning_va): Turned into small wrapper around
3013 format_warning_n_va, renamed to ...
3014 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3015 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3017 (format_warning_at_substring_n): New function.
3018 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3019 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3020 format_warning_at_substring with just a shorter name instead of
3021 const function pointer.
3022 (fmtwarn_n): New function.
3023 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3024 appropriate, get rid of all the fmtstr temporaries, move conditionals
3025 with G_() wrapped string literals directly into fmtwarn arguments,
3026 cast dir.len to (int), formatting fixes.
3028 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3030 * doc/invoke.texi: Remove "Cilk Plus" references.
3032 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3033 Richard Biener <rguenther@suse.de>
3036 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3037 for error or warning attributes if CALL_FROM_THUNK_P is set.
3040 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3043 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3044 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3047 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3048 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3050 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3051 instead of -1U in last predictors element's probability member.
3053 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3056 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3057 arguments if they are comparable.
3059 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3061 PR tree-optimization/84634
3062 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3063 masks and masked_loop_p with a single loop_masks, making sure it's
3064 null for bb vectorization.
3066 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3068 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3069 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3070 loop->force_vectorize to check whether there is no alias.
3072 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3075 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3077 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3079 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3080 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3081 instead of a loop around prev_real_insn.
3082 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3086 * config/i386/i386.c (ix86_print_operand): Use conditional
3087 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3090 2018-03-02 Richard Biener <rguenther@suse.de>
3092 PR tree-optimization/84427
3093 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3094 (bitmap_set_subtract_values): Rewrite to handle multiple
3097 (prune_clobbered_mems): Likewise.
3098 (phi_translate): Take edge instead of pred/phiblock.
3099 (phi_translate_1): Likewise.
3100 (phi_translate_set): Likewise. Insert all translated
3101 exprs for a value into the set, keeping possibly multiple
3102 expressions per value.
3103 (compute_antic_aux): Adjust for phi_translate changes.
3104 When intersecting union the expressions and prune those
3105 not in the final value set, keeping possibly multiple
3106 expressions per value. Do not use value-insertion
3107 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3108 all expressions. Add verification that the value-sets
3109 only shrink during iteration.
3110 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3111 (do_pre_regular_insertion): Likewise.
3112 (do_pre_partial_partial_insertion): Likewise.
3114 2018-03-02 Richard Biener <rguenther@suse.de>
3117 * config/darwin.c (saved_debug_info_level): New static global.
3118 (darwin_asm_lto_start): Disable debug info generation for LTO out.
3119 (darwin_asm_lto_end): Restore debug info generation settings.
3121 2018-03-01 Martin Liska <mliska@suse.cz>
3124 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3127 2018-03-01 Richard Biener <rguenther@suse.de>
3130 * dwarf2out.c (gen_variable_die): Properly handle late VLA
3131 type annotation with LTO when debug was disabled at compile-time.
3133 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3135 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3137 (mips_final_postscan_insn): Likewise.
3139 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
3141 PR rtl-optimization/84528
3142 * alias.c (init_alias_target): Add commentary.
3143 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3144 a unique base value if the frame pointer is not eliminated
3145 to the stack pointer.
3147 2018-03-01 Tom de Vries <tom@codesourcery.com>
3149 PR rtl-optimization/83327
3150 * lra-int.h (hard_regs_spilled_into): Declare.
3151 * lra.c (hard_regs_spilled_into): Define.
3152 (init_reg_info): Init hard_regs_spilled_into.
3153 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3154 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3155 (process_bb_lives): Handle hard_regs_spilled_into.
3156 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3157 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3159 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3161 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3162 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3163 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3164 * config/rs6000/aix72.h: New file.
3166 2018-02-28 Jakub Jelinek <jakub@redhat.com>
3168 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3169 instead of warning_at with conditional singular and plural messages
3173 * stor-layout.c (update_alignment_for_field): For
3174 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
3175 && !DECL_PACKED (field), do the alignment update, just use
3176 only desired_align instead of MAX (type_align, desired_align)
3178 (place_field): Don't do known_align < desired_align handling
3179 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
3180 is non-NULL, instead do it after rli->prev_field handling and
3181 only if not within a bitfield word. For DECL_PACKED (field)
3182 use type_align of BITS_PER_UNIT.
3184 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
3186 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
3187 superfluous parentheses and trailing spaces.
3189 2018-02-28 Richard Biener <rguenther@suse.de>
3191 PR tree-optimization/84584
3192 * graphite-scop-detection.c (scop_detection::add_scop): Discard
3193 SCoPs with fake exit edge.
3195 2018-02-28 Martin Liska <mliska@suse.cz>
3198 * timevar.c (timer::print): Fix format to properly print 100%
3201 2018-02-28 Richard Biener <rguenther@suse.de>
3204 * genmatch.c (capture_info::walk_match): Do not mark
3205 captured expressions without operands as expr_p given
3206 they act more like predicates and should be subject to
3207 "lost tail" side-effect preserving.
3209 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
3211 PR rtl-optimization/81611
3212 * auto-inc-dec.c (attempt_change): Move dead note from
3213 mem_insn if it's the next use of regno
3214 (find_address): Take address use of reg holding
3215 non-incremented value. Add parm to limit search to the named
3217 (merge_in_block): Attempt to use a mem insn that is the next
3218 use of the original regno.
3220 2018-02-27 Martin Sebor <msebor@redhat.com>
3223 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
3224 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
3226 2018-02-27 Martin Sebor <msebor@redhat.com>
3228 PR translation/84207
3229 * diagnostic-core.h (warning_n, error_n, inform_n): Change
3230 n argument to unsigned HOST_WIDE_INT.
3231 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3232 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
3233 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3234 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3236 2018-02-27 Richard Biener <rguenther@suse.de>
3238 PR tree-optimization/84512
3239 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3240 Do not use the estimate returned from record_stmt_cost for
3241 the scalar iteration cost but sum properly using add_stmt_cost.
3243 2018-02-27 Richard Biener <rguenther@suse.de>
3245 PR tree-optimization/84466
3246 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3247 Adjust last change to less strictly validate use operands.
3249 2018-02-27 Martin Liska <mliska@suse.cz>
3251 PR gcov-profile/84548
3252 * gcov.c (process_file): Allow partial overlap and consider it
3253 also as group functions.
3254 (output_lines): Properly calculate range of lines for a group.
3256 2018-02-27 Martin Liska <mliska@suse.cz>
3258 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3259 'ggc' suffixes. Change first column width.
3260 (timer::print): Fix formatting of the column.
3262 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3264 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3265 preserve inline entry blocks for the sake of debug inline
3266 entry point markers alone.
3267 (remove_unused_locals): Suggest in comments a better place to
3268 force the preservation of inline entry blocks that are
3269 otherwise unused, but do not preserve them.
3271 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3273 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3275 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3278 * config/i386/constraints.md (Bs): Replace
3279 ix86_indirect_branch_register with
3280 TARGET_INDIRECT_BRANCH_REGISTER.
3282 * config/i386/i386.md (indirect_jump): Likewise.
3283 (tablejump): Likewise.
3284 (*sibcall_memory): Likewise.
3285 (*sibcall_value_memory): Likewise.
3286 Peepholes of indirect call and jump via memory: Likewise.
3287 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3288 (*sibcall_value_GOT_32): Likewise.
3289 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3290 (GOT_memory_operand): Likewise.
3291 (call_insn_operand): Likewise.
3292 (sibcall_insn_operand): Likewise.
3293 (GOT32_symbol_operand): Likewise.
3294 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3296 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3298 PR rtl-optimization/83496
3299 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3300 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3301 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3302 redundant insn, if any.
3303 (relax_delay_slots): Likewise.
3304 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3306 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
3308 PR tree-optimization/83965
3309 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3310 that grouped statements are part of a reduction chain. Return
3311 true if the statement is not marked as a reduction itself but
3313 (vect_recog_dot_prod_pattern): Don't check whether the statement
3314 is part of a group here.
3315 (vect_recog_sad_pattern): Likewise.
3316 (vect_recog_widen_sum_pattern): Likewise.
3318 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3321 * final.c (rest_of_clean_state): Also look for calls inside sequences.
3323 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3326 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3328 (ix86_output_indirect_function_return): New prototype.
3329 (ix86_split_simple_return_pop_internal): Likewise.
3330 * config/i386/i386.c (indirect_return_via_cx): New.
3331 (indirect_return_via_cx_bnd): Likewise.
3332 (indirect_thunk_name): Handle return va CX_REG.
3333 (output_indirect_thunk_function): Create alias for
3334 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3335 (ix86_output_indirect_jmp): Remove the bool argument.
3336 (ix86_output_indirect_function_return): New function.
3337 (ix86_split_simple_return_pop_internal): Likewise.
3338 * config/i386/i386.md (*indirect_jump): Don't pass false
3339 to ix86_output_indirect_jmp.
3340 (*tablejump_1): Likewise.
3341 (simple_return_pop_internal): Change it to define_insn_and_split.
3342 Call ix86_split_simple_return_pop_internal to split it for
3344 (simple_return_indirect_internal): Call
3345 ix86_output_indirect_function_return instead of
3346 ix86_output_indirect_jmp.
3348 2018-02-26 Jakub Jelinek <jakub@redhat.com>
3351 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3352 memset and value initialization afterwards.
3354 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
3356 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3358 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3361 * common/config/aarch64/aarch64-common.c
3362 (aarch_option_optimization_table[]): Switch
3363 off fomit-frame-pointer
3365 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3366 Chung-Ju Wu <jasonwucj@gmail.com>
3368 * config/nds32/nds32-multiple.md (load_multiple): Disallow
3370 (store_multiple): Ditto.
3372 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3374 * config.gcc: Add --with-cpu support for nds32 target.
3375 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
3376 * config/nds32/nds32.opt: Add -mcpu= option.
3378 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
3380 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
3381 isel=yes): Warn for these deprecated options.
3383 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
3385 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
3386 ISA_2_5_MASKS_EMBEDDED.
3388 2018-02-23 Jakub Jelinek <jakub@redhat.com>
3390 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
3391 p->max as pointers rather than using iterative_hash_expr.
3393 2018-02-23 Carl Love <cel@us.ibm.com>
3395 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
3396 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
3398 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
3399 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
3400 P8V_BUILTIN_VEC_VUNSIGNED2.
3402 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
3405 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
3406 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
3408 * lra-constraints.c (curr_insn_transform): Set up
3409 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
3411 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
3412 * lra-eliminations.c (spill_pseudos): Ditto.
3413 (process_insn_for_elimination): Ditto.
3414 * lra-lives.c (reg_early_clobber_p): Use the new macros.
3415 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
3416 LRA_NON_CLOBBERED_ALT.
3418 2018-02-22 Martin Sebor <msebor@redhat.com>
3420 PR tree-optimization/84480
3421 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
3422 to maybe_diag_stxncpy_trunc. Call it.
3423 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
3424 from gimple_fold_builtin_strcpy. Print inlining stack.
3425 (handle_builtin_stxncpy): Print inlining stack.
3426 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
3428 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
3431 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
3432 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
3433 and -fcheck-pointer-bounds are used together.
3434 (indirect_thunk_prefix): New enum.
3435 (indirect_thunk_need_prefix): New function.
3436 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
3437 "_nt" instead of "_bnd" for NOTRACK prefix.
3438 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
3439 (output_indirect_thunk_function): Likewise.
3441 (ix86_code_end): Update output_indirect_thunk_function calls.
3442 (ix86_output_indirect_branch_via_reg): Replace
3443 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
3444 (ix86_output_indirect_branch_via_push): Likewise.
3445 (ix86_output_function_return): Likewise.
3446 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
3447 incompatible with -fcf-protection=branch and
3448 -fcheck-pointer-bounds.
3450 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3453 * config/aarch64/aarch64.c (aarch64_print_address_internal):
3454 Change gcc_assert call to output_operand_lossage.
3456 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3458 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
3460 2018-02-22 DJ Delorie <dj@redhat.com>
3461 Sebastian Perta <sebastian.perta@renesas.com>
3462 Oleg Endo <olegendo@gcc.gnu.org>
3464 * config/rx/rx.c (rx_rtx_costs): New function.
3465 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
3467 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
3469 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
3471 2018-02-22 Martin Liska <mliska@suse.cz>
3474 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
3475 Add "native" as a possible value.
3477 2018-02-22 Martin Liska <mliska@suse.cz>
3480 * config/i386/i386.c (ix86_option_override_internal):
3481 Add "native" as a possible value for -march and -mtune.
3483 2018-02-22 Jakub Jelinek <jakub@redhat.com>
3486 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
3487 to all type variants.
3489 PR tree-optimization/84503
3490 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
3491 width as info->bitpos + info->bitsize - start.
3492 (merged_store_group::merge_overlapping): Simplify width computation.
3493 (check_no_overlap): New function.
3494 (imm_store_chain_info::try_coalesce_bswap): Compute expected
3495 start + width and last_order of the group, fail if check_no_overlap
3497 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
3498 to group if check_no_overlap fails.
3500 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3502 * config/rs6000/altivec.md: Delete contraint arguments to
3503 define_expand, define_split, and define_peephole2, and in
3504 define_insn_and_split if always unused.
3505 * config/rs6000/darwin.md: Ditto.
3506 * config/rs6000/dfp.md: Ditto.
3507 * config/rs6000/rs6000.md: Ditto.
3508 * config/rs6000/sync.md: Ditto.
3509 * config/rs6000/vector.md: Ditto.
3510 * config/rs6000/vsx.md: Ditto.
3512 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3514 * config/rs6000/altivec.md: Write output control strings as braced
3515 blocks instead of double-quoted strings.
3516 * config/rs6000/darwin.md: Ditto.
3517 * config/rs6000/rs6000.md: Ditto.
3518 * config/rs6000/vector.md: Ditto.
3519 * config/rs6000/vsx.md: Ditto.
3521 2018-02-21 Jason Merrill <jason@redhat.com>
3523 PR c++/84314 - ICE with templates and fastcall attribute.
3524 * attribs.c (build_type_attribute_qual_variant): Remove assert.
3526 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3528 * ipa-cp.c (determine_versionability): Fix comment typos.
3530 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3533 * ipa-cp.c (determine_versionability): Do not version functions caling
3536 2018-02-21 Martin Liska <mliska@suse.cz>
3539 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
3540 Add "native" as a possible value.
3541 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
3542 the macro when native cpu detection is available.
3544 2018-02-21 Martin Liska <mliska@suse.cz>
3547 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
3548 Add "native" as a possible value.
3549 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
3550 when native cpu detection is available.
3552 2018-02-21 Jakub Jelinek <jakub@redhat.com>
3553 Martin Sebor <msebor@redhat.com>
3555 PR tree-optimization/84478
3556 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
3558 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
3559 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
3560 support which is conservatively correct, for 2 only stay conservative
3561 for maxlen. Formatting and comment capitalization fixes. Add STRICT
3562 argument to the 2 argument get_range_strlen, adjust 6 arg
3563 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
3565 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
3566 (gimple_fold_builtin_strlen): Pass true as last argument to
3569 2018-02-20 Martin Sebor <msebor@redhat.com>
3572 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
3573 (builtin_memref::set_base_and_offset): Same. Handle inner references.
3574 (builtin_memref::builtin_memref): Factor out parts into
3575 set_base_and_offset and call it.
3577 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3580 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
3581 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
3582 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
3583 search at the associated MODE_INT.
3585 2018-02-20 Jeff Law <law@redhat.com>
3588 PR tree-optimization/81592
3590 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
3591 for range data rather than using global data.
3592 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
3593 range data rather than using global data.
3594 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
3595 pass it to children as needed.
3596 (struct directive::fmtresult): Similarly.
3597 (struct directive::set_width): Similarly.
3598 (struct directive::set_precision): Similarly.
3599 (format_integer, format_directive, parse_directive): Similarly.
3600 (format_none): Accept unnamed vr_values parameter.
3601 (format_percent, format_floating, format_character): Similarly.
3602 (format_string, format_plain): Similarly.
3603 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
3604 the EVRP range analyzer for range data rather than using global data.
3605 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
3606 gimple-ssa-evrp-analyze.h
3607 (class sprintf_dom_walker): Add after_dom_children member function.
3608 Add evrp_range_analyzer member.
3609 (sprintf_dom_walker::before_dom_children): Call into the EVRP
3610 range analyzer as needed.
3611 (sprintf_dom_walker::after_dom_children): New member function.
3612 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
3614 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
3615 (evrp_range_analyzer::pop_to_marker): Likewise.
3617 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3619 PR tree-optimization/84419
3620 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
3621 with the required type if its current type is compatible but
3624 2018-02-20 Jakub Jelinek <jakub@redhat.com>
3627 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
3628 after vectorization.
3630 2018-02-20 Martin Liska <mliska@suse.cz>
3633 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
3634 possible values if we don't have a hint.
3636 2018-02-20 Martin Liska <mliska@suse.cz>
3640 * final.c (shorten_branches): Build align_tab array with one
3642 * opts.c (finish_options): Add alignment option limit check.
3643 (MAX_CODE_ALIGN): Likewise.
3644 (MAX_CODE_ALIGN_VALUE): Likewise.
3645 * doc/invoke.texi: Document maximum allowed option value for
3646 all -falign-* options.
3648 2018-02-19 Jakub Jelinek <jakub@redhat.com>
3651 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
3652 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
3653 * var-tracking.c (emit_note_insn_var_location): Remove all references
3654 to NOTE_INSN_CALL_ARG_LOCATION.
3655 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
3656 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
3657 Use copy_rtx_if_shared.
3658 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
3659 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
3660 (dwarf2out_var_location): Remove handling of
3661 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
3663 * final.c (final_scan_insn): Remove all references to
3664 NOTE_INSN_CALL_ARG_LOCATION.
3665 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
3666 before dumping final insns.
3667 * except.c (emit_note_eh_region_end): Remove all references to
3668 NOTE_INSN_CALL_ARG_LOCATION.
3669 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
3670 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
3671 * config/arc/arc.c (hwloop_optimize): Likewise.
3672 * config/arm/arm.c (create_fix_barrier): Likewise.
3673 * config/s390/s390.c (s390_chunkify_start): Likewise.
3674 * config/sh/sh.c (find_barrier): Likewise.
3675 * config/i386/i386.c (rest_of_insert_endbranch,
3676 ix86_seh_fixup_eh_fallthru): Likewise.
3677 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
3678 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
3679 * config/frv/frv.c (frv_function_prologue): Likewise.
3680 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
3682 (note_outside_basic_block_p): Remove all references to
3683 NOTE_INSN_CALL_ARG_LOCATION.
3684 * gengtype.c (adjust_field_rtx_def): Likewise.
3685 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
3687 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
3688 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
3691 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
3694 PR tree-optimization/84452
3695 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
3696 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
3699 2018-02-19 Martin Liska <mliska@suse.cz>
3702 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
3704 2018-02-19 Martin Liska <mliska@suse.cz>
3705 Richard Sandiford <richard.sandiford@linaro.org>
3707 PR tree-optimization/82491
3708 * gimple-fold.c (get_base_constructor): Make earlier bail out
3711 2018-02-19 Carl Love <cel@us.ibm.com>
3713 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
3714 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
3716 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
3717 P8V_BUILTIN_VEC_NEG.
3719 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
3721 * config/rl78/rl78.md (movdf): New define expand.
3723 2018-02-19 Martin Liska <mliska@suse.cz>
3726 * doc/invoke.texi: Fix typo.
3727 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
3729 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
3731 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3732 handle rs6000_single_float and rs6000_double_float specially for
3735 2018-02-16 Jeff Law <law@redhat.com>
3737 * config/rx/rx.c (add_pop_cfi_notes): New function.;
3740 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3743 * ipa-inline.c (inline_small_functions): Fix a typo.
3745 2018-02-16 Nathan Sidwell <nathan@acm.org>
3747 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
3749 2018-02-16 Carl Love <cel@us.ibm.com>
3751 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
3752 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
3753 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
3754 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
3755 expansion to P8V_BUILTIN_VEC_FLOAT2.
3757 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
3759 PR rtl-optimization/70023
3760 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
3761 src_regno into account.
3763 2018-02-16 Carl Love <cel@us.ibm.com>
3765 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
3766 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
3767 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
3768 * config/rs6000/rs6000.c: Remove case statements for
3769 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
3770 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
3771 and P9V_BUILTIN_VEC_VINSERT4B.
3772 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
3773 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
3774 * config/rs6000/vsx.md:
3775 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
3778 2018-02-16 Carl Love <cel@us.ibm.com>
3780 * config/rs6000/altivec.h: Add builtin names vec_extract4b
3782 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
3784 * config/rs6000/rs6000-c.c: Add the definitions for
3785 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
3786 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
3787 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
3788 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
3789 definition for insert4b and define insn *insert3b_internal.
3790 * doc/extend.texi: Add documentation for vec_extract4b.
3792 2018-02-16 Nathan Sidwell <nathan@acm.org>
3794 * doc/extend.texi (Backwards Compatibility): Mention friend
3795 injection. Note for-scope is deprecated.
3796 * doc/invoke.texi (-ffriend-injection): Deprecate.
3798 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
3800 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
3801 that moved to I2, also allow destinations that are a paradoxical
3802 subreg (instead of a normal reg).
3804 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
3807 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
3810 2018-02-16 Richard Biener <rguenther@suse.de>
3812 PR tree-optimization/84037
3813 PR tree-optimization/84016
3815 * config/i386/i386.c (ix86_builtin_vectorization_cost):
3816 Adjust vec_construct for the fact we need additional higher latency
3817 128bit inserts for AVX256 and AVX512 vector builds.
3818 (ix86_add_stmt_cost): Scale vector construction cost for
3821 2018-02-16 Richard Biener <rguenther@suse.de>
3823 PR tree-optimization/84417
3824 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
3825 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
3826 (non_rewritable_lvalue_p): Likewise, use poly-ints.
3828 2018-02-16 Martin Liska <mliska@suse.cz>
3831 * internal-fn.def (ASAN_CHECK): Set proper flags.
3832 (ASAN_MARK): Likewise.
3834 2018-02-16 Julia Koval <julia.koval@intel.com>
3836 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
3837 from PTA_CANNONLAKE.
3839 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3842 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
3843 Use ++iter rather than iter++ for std::list iterators.
3844 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
3845 defer deleting them until all nodes in the forest are processed. Do
3846 free even leaf nodes. Change to_process into auto_vec.
3849 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
3850 * vec.h (vec_default_construct): Use memset instead of placement new
3851 if BROKEN_VALUE_INITIALIZATION is defined.
3852 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
3853 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
3856 PR rtl-optimization/83723
3857 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
3858 * lra.c (lra_substitute_pseudo): Likewise. If true, use
3859 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
3861 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
3863 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
3865 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
3867 PR rtl-optimization/81443
3868 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
3869 from inner REGs to paradoxical SUBREGs.
3871 2018-02-16 Richard Biener <rguenther@suse.de>
3873 PR tree-optimization/84399
3874 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3875 For operands we can analyze at their definition make sure we can
3876 analyze them at each use as well.
3878 2018-02-16 Richard Biener <rguenther@suse.de>
3880 PR tree-optimization/84190
3881 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
3882 volatile accesses if the decl isn't volatile.
3884 2018-02-15 Jason Merrill <jason@redhat.com>
3886 PR c++/84314 - ICE with templates and fastcall attribute.
3887 * attribs.c (build_type_attribute_qual_variant): Don't clobber
3888 TYPE_CANONICAL on an existing type.
3890 2018-02-15 Jakub Jelinek <jakub@redhat.com>
3892 PR tree-optimization/84383
3893 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
3894 dstoff nor call operand_equal_p if dstbase is NULL.
3896 PR tree-optimization/84334
3897 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
3898 also a CONSTANT_CLASS_P, punt.
3900 2018-02-14 Jim Wilson <jimw@sifive.com>
3902 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
3903 first SMALL_OPERAND check. New local min_second_step. Move assert
3904 to where locals are set. Add TARGET_RVC support.
3905 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
3907 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
3909 * doc/invoke.texi: Correct -Wformat-overflow code sample.
3911 2018-02-14 Martin Sebor <msebor@redhat.com>
3913 PR tree-optimization/83698
3914 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
3915 arrays constrain the offset range to their bounds.
3916 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
3917 (builtin_access::overlap): Avoid setting the size of overlap if it's
3919 (maybe_diag_overlap): Also consider arrays when deciding what values
3920 of offsets to include in diagnostics.
3922 2018-02-14 Martin Sebor <msebor@redhat.com>
3925 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
3926 that correspond to the kind of a declaration.
3928 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
3931 * config/pa/pa.md: Load address of PIC label using the linkage table
3932 if the label is nonlocal.
3934 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
3936 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
3937 warning message if user requests -maltivec=be.
3938 * doc/invoke.texi: Document deprecation of -maltivec=be.
3940 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
3943 * config/rs6000/rs6000-c.c: Update definitions for
3944 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
3945 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
3947 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3950 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
3951 add _get_ssp intrinsics. Remove argument from
3952 __builtin_ia32_rdssp[d|q].
3953 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
3954 * config/i386/i386-builtin.def: Remove argument from
3955 __builtin_ia32_rdssp[d|q].
3956 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
3957 ix86_expand_special_args_builtin for _rdssp[d|q].
3958 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
3959 Clear register before usage.
3960 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
3961 Add documentation for new _get_ssp and _inc_ssp intrinsics.
3963 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
3965 PR tree-optimization/84357
3966 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
3967 operand 1 of an ARRAY_REF too.
3969 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
3972 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
3973 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3975 (set_of_reg): New struct.
3976 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
3977 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
3978 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3980 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
3981 Split into bitclr, bitset, bitinvert patterns if appropriate.
3982 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
3983 use rx_fuse_in_memory_bitop.
3984 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
3985 to named insn, correct maximum insn length.
3987 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
3990 * machmode.def: Define a complex mode for PARTIAL_INT.
3991 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
3993 * doc/rtl.texi: Document CSPImode.
3994 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
3996 (msp430_hard_regno_nregs_with_padding): Likewise.
3998 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4001 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4003 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4005 PR rtl-optimization/84169
4006 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4007 we generated a parallel as new i3 and we split that to new i2 and i3
4008 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4009 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4010 those to i2, not i1. Partially rewrite this scan code.
4012 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4015 * stor-layout.c (place_field): For variable length fields, adjust
4016 offset_align afterwards not just based on the field's alignment,
4017 but also on the size.
4020 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4021 of exps and logs in the use_exp2 case.
4023 2018-02-13 Jeff Law <law@redhat.com>
4025 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4028 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4031 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4035 * common.opt (gas-loc-support, gas-locview-support): New.
4036 (ginline-points, ginternal-reset-location-views): New.
4037 * doc/invoke.texi: Document them. Use @itemx where intended.
4038 (gvariable-location-views): Adjust.
4039 * target.def (reset_location_view): New.
4040 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4041 (TARGET_RESET_LOCATION_VIEW): New.
4042 * doc/tm.texi: Rebuilt.
4043 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4044 (dwarf2out_default_as_locview_support): New.
4045 (output_asm_line_debug_info): Use option variables.
4046 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4047 (output_loc_list): Likewise.
4048 (add_high_low_attributes): Check option variables.
4049 Don't output entry view attribute in strict mode.
4050 (gen_inlined_subroutine_die): Check option variables.
4051 (dwarf2out_inline_entry): Likewise.
4052 (init_sections_and_labels): Likewise.
4053 (dwarf2out_early_finish): Likewise.
4054 (maybe_reset_location_view): New, from...
4055 (dwarf2out_var_location): ... here. Call it.
4056 * debug.h (dwarf2out_default_as_loc_support): Declare.
4057 (dwarf2out_default_as_locview_support): Declare.
4058 * hooks.c (hook_int_rtx_insn_0): New.
4059 * hooks.h (hook_int_rtx_insn_0): Declare.
4060 * toplev.c (process_options): Take -gas-loc-support and
4061 -gas-locview-support from dwarf2out. Enable
4062 -gvariable-location-views by default only with locview
4063 assembler support. Enable -ginternal-reset-location-views by
4064 default only if the target defines the corresponding hook.
4065 Enable -ginline-points by default if location views are
4066 enabled; force it disabled if statement frontiers are
4068 * tree-inline.c (expand_call_inline): Check option variables.
4069 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4071 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4073 PR tree-optimization/84321
4074 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4075 handling. Also check whether the anti-range contains any values
4076 that satisfy the mask; switch to a VR_RANGE if not.
4078 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4081 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4083 2018-02-13 Martin Jambor <mjambor@suse.cz>
4086 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4087 of call statements, also set location of a load to a temporary.
4089 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4091 * config/rl78/rl78.c (add_vector_labels): New function.
4092 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4093 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4094 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4095 which checks that no arguments are passed.
4096 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4097 * doc/extend.texi: Documentation for the new attribute.
4099 2018-02-13 Andreas Schwab <schwab@suse.de>
4101 * config/riscv/linux.h (CPP_SPEC): Define.
4103 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4106 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4107 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4108 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4109 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4110 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4111 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4112 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4113 temporarily for AES and PCLMUL builtins.
4115 PR tree-optimization/84339
4116 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4117 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4121 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4122 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4123 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4125 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4127 * omp-simd-clone.h: New file.
4128 * omp-simd-clone.c: Include omp-simd-clone.h.
4129 (expand_simd_clones): No longer static.
4130 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4131 cgraph.h and omp-simd-clone.h.
4132 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4133 (vect_recog_widen_shift_pattern): Formatting fix.
4134 (vect_pattern_recog_1): Don't check optab for calls.
4137 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4138 operands[2] into a REG before using gen_lowpart on it.
4140 2018-02-12 Jeff Law <law@redhat.com>
4143 * config/sh/sh.c (find_barrier): Consider a sibling call
4146 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4147 successfully back substituting a reg.
4149 2018-02-12 Richard Biener <rguenther@suse.de>
4151 PR tree-optimization/84037
4152 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4153 parameter, move visited init to caller.
4154 (vect_slp_analyze_operations): Separate cost from validity
4155 check, initialize visited once for all instances.
4156 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4158 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4160 (vect_model_promotion_demotion_cost): Likewise.
4161 (vectorizable_bswap): Guard cost modeling with !slp_node
4162 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4164 (vectorizable_call): Likewise.
4165 (vectorizable_conversion): Likewise.
4166 (vectorizable_assignment): Likewise.
4167 (vectorizable_shift): Likewise.
4168 (vectorizable_operation): Likewise.
4169 (vectorizable_store): Likewise.
4170 (vectorizable_load): Likewise.
4171 (vectorizable_condition): Likewise.
4172 (vectorizable_comparison): Likewise.
4174 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
4177 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
4178 (ASAN_MARK): Fix fnspec to account for return value, change pointer
4179 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
4181 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4184 * params.def (inline-min-speedup): Increase from 8 to 15.
4185 (max-inline-insns-auto): Decrease from 40 to 30.
4186 * ipa-split.c (consider_split): Add some buffer for function to
4187 be considered inlining candidate.
4188 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
4191 2018-02-12 Richard Biener <rguenther@suse.de>
4193 PR tree-optimization/84037
4194 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
4195 matched stmts if we cannot swap the non-matched ones.
4197 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4199 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
4200 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4201 _mm_maskz_scalef_round_ss): New intrinsics.
4202 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4203 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4204 __builtin_ia32_scalefss_round): Remove.
4205 (__builtin_ia32_scalefsd_mask_round,
4206 __builtin_ia32_scalefss_mask_round): New intrinsics.
4207 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4208 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4209 ((match_operand:VF_128 2 "<round_nimm_predicate>"
4210 "<round_constraint>")): Changed to ...
4211 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4212 "<round_scalar_constraint>")): ... this.
4213 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4214 %0, %1, %2<round_op3>}"): Changed to ...
4215 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4216 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4217 %2<round_scalar_mask_op3>}"): ... this.
4218 * config/i386/subst.md (round_scalar_nimm_predicate): New.
4220 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4222 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4223 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
4224 (_mm_maskz_sqrt_round_ss): New intrinsics.
4225 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
4226 (__builtin_ia32_sqrtsd_mask_round)
4227 (__builtin_ia32_sqrtss_mask_round): New builtins.
4228 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
4229 (__builtin_ia32_sqrtss_round): Remove.
4230 (__builtin_ia32_sqrtsd_mask_round)
4231 (__builtin_ia32_sqrtss_mask_round): New builtins.
4232 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4233 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4234 ((match_operand:VF_128 1 "vector_operand"
4235 "xBm,<round_constraint>")): Changed to ...
4236 ((match_operand:VF_128 1 "vector_operand"
4237 "xBm,<round_scalar_constraint>")): ... this.
4238 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4239 %0, %2, %<iptr>1<round_op3>}): Changed to ...
4240 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4241 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4242 %<iptr>1<round_scalar_mask_op3>}): ... this.
4243 ((set_attr "prefix" "<round_prefix>")): Changed to ...
4244 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4246 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
4249 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4250 Cast vec_cmpeq result to correct type.
4251 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4252 Cast vec_cmpgt result to correct type.
4254 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4256 * final.c (final_scan_insn_1): Renamed from...
4257 (final_scan_insn): ... this. New wrapper, to recover
4258 seen from the outermost call in recursive ones.
4259 * config/sparc/sparc.c (output_return): Drop seen from call.
4260 (output_sibcall): Likewise.
4261 * config/visium/visium.c (output_branch): Likewise.
4263 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4265 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4268 2018-02-10 Alan Modra <amodra@gmail.com>
4271 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4272 Specify LR as an input.
4274 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4277 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4278 remove_member_access_dummy_vars): New functions.
4279 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4280 lower_omp_1, execute_lower_omp): Use them.
4282 PR rtl-optimization/84308
4283 * shrink-wrap.c (spread_components): Release todo vector.
4285 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4287 PR rtl-optimization/57193
4288 * ira-color.c (struct allocno_color_data): Add member
4289 conflict_allocno_hard_prefs.
4290 (update_conflict_allocno_hard_prefs): New.
4291 (bucket_allocno_compare_func): Add a preference based on
4292 conflict_allocno_hard_prefs.
4293 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4294 (color_allocnos): Remove a dead code. Initiate
4295 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4297 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4300 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4301 constraint from =wa to wa. Avoid a subreg on the output operand,
4302 instead use a pseudo and subreg it in a move.
4303 (p9_xxbrd_<mode>): Changed to ...
4304 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4305 (p9_xxbrd_v2df): New expander.
4306 (p9_xxbrw_<mode>): Changed to ...
4307 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4308 (p9_xxbrw_v4sf): New expander.
4310 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4312 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4314 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
4317 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4318 multiply in 32-bit mode.
4319 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4320 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4323 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4325 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4326 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4327 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4328 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4330 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4333 * dwarf2out.c (is_trivial_indirect_ref): New function.
4334 (dwarf2out_late_global_decl): Do not generate a location
4335 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4336 and that are not defined in the current unit.
4338 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4340 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4341 instead of a libcall for UNORDERED.
4343 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4346 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4347 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4349 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4352 * config/s390/s390.c (s390_set_current_function): Invoke
4353 s390_indirect_branch_settings also if fndecl didn't change.
4355 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4357 * config/rs6000/rs6000.md (blockage): Set length to zero.
4359 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4361 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4363 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4366 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4367 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4371 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4372 PARALLEL incoming that failed vt_get_decl_and_offset check.
4375 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
4376 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
4378 (get_variable_section): For decls in named .bss* sections pass true as
4379 second argument to bss_initializer_p.
4381 2018-02-09 Marek Polacek <polacek@redhat.com>
4382 Jakub Jelinek <jakub@redhat.com>
4385 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
4386 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
4387 Sync some changes from cxx_fold_indirect_ref.
4389 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4391 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
4393 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
4394 (BLOCK_INLINE_ENTRY_LABEL): New.
4395 (dwarf2out_var_location): Disregard inline entry markers.
4396 (inline_entry_data): New struct.
4397 (inline_entry_data_hasher): New hashtable type.
4398 (inline_entry_data_hasher::hash): New.
4399 (inline_entry_data_hasher::equal): New.
4400 (inline_entry_data_table): New variable.
4401 (add_high_low_attributes): Add DW_AT_entry_pc and
4402 DW_AT_GNU_entry_view attributes if a pending entry is found
4403 in inline_entry_data_table. Add old entry_pc attribute only
4404 if debug nonbinding markers are disabled.
4405 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
4406 markers are enabled.
4407 (block_within_block_p, dwarf2out_inline_entry): New.
4408 (dwarf2out_finish): Check that no entries remained in
4409 inline_entry_data_table.
4410 * final.c (reemit_insn_block_notes): Handle inline entry notes.
4411 (final_scan_insn, notice_source_line): Likewise.
4412 (rest_of_clean_state): Skip inline entry markers.
4413 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
4415 * gimple.c (gimple_build_debug_inline_entry): New.
4416 * gimple.h (enum gimple_debug_subcode): Add
4417 GIMPLE_DEBUG_INLINE_ENTRY.
4418 (gimple_build_debug_inline_entry): Declare.
4419 (gimple_debug_inline_entry_p): New.
4420 (gimple_debug_nonbind_marker_p): Adjust.
4421 * insn-notes.def (INLINE_ENTRY): New.
4422 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4423 inline entry marker notes.
4424 (print_insn): Likewise.
4425 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
4426 (INSN_DEBUG_MARKER_KIND): Likewise.
4427 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
4428 * tree-inline.c (expand_call_inline): Build and insert
4429 debug_inline_entry stmt.
4430 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
4431 inline entry blocks early, if nonbind markers are enabled.
4432 (dump_scope_block): Dump fragment info.
4433 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
4434 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
4435 (gimple_build_debug_inline_entry): New.
4436 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
4437 Enable/disable inline entry points too.
4438 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
4439 (DEBUG_INSN): Describe inline entry markers.
4441 * common.opt (gvariable-location-views): New.
4442 (gvariable-location-views=incompat5): New.
4443 * config.in: Rebuilt.
4444 * configure: Rebuilt.
4445 * configure.ac: Test assembler for view support.
4446 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
4447 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
4448 * dwarf2out.c (var_loc_view): New typedef.
4449 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
4450 (dwarf2out_locviews_in_attribute): New.
4451 (dwarf2out_locviews_in_loclist): New.
4452 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
4453 (enum dw_line_info_opcode): Add LI_adv_address.
4454 (struct dw_line_info_table): Add view.
4455 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
4456 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
4457 (zero_view_p): New variable.
4458 (ZERO_VIEW_P): New macro.
4459 (output_asm_line_debug_info): New.
4460 (struct var_loc_node): Add view.
4461 (add_AT_view_list, AT_loc_list): New.
4462 (add_var_loc_to_decl): Add view param. Test it against last.
4463 (new_loc_list): Add view params. Record them.
4464 (AT_loc_list_ptr): Handle loc and view lists.
4465 (view_list_to_loc_list_val_node): New.
4466 (print_dw_val): Handle dw_val_class_view_list.
4467 (size_of_die): Likewise.
4468 (value_format): Likewise.
4469 (loc_list_has_views): New.
4470 (gen_llsym): Set vl_symbol too.
4471 (maybe_gen_llsym, skip_loc_list_entry): New.
4472 (dwarf2out_maybe_output_loclist_view_pair): New.
4473 (output_loc_list): Output view list or entries too.
4474 (output_view_list_offset): New.
4475 (output_die): Handle dw_val_class_view_list.
4476 (output_dwarf_version): New.
4477 (output_compilation_unit_header): Use it.
4478 (output_skeleton_debug_sections): Likewise.
4479 (output_rnglists, output_line_info): Likewise.
4480 (output_pubnames, output_aranges): Update version comments.
4481 (output_one_line_info_table): Output view numbers in asm comments.
4482 (dw_loc_list): Determine current endview, pass it to new_loc_list.
4483 Call maybe_gen_llsym.
4484 (loc_list_from_tree_1): Adjust.
4485 (add_AT_location_description): Create view list attribute if
4486 needed, check it's absent otherwise.
4487 (convert_cfa_to_fb_loc_list): Adjust.
4488 (maybe_emit_file): Call output_asm_line_debug_info for test.
4489 (dwarf2out_var_location): Reset views as needed. Precompute
4490 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
4491 attribute. Set view.
4492 (new_line_info_table): Reset next view.
4493 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
4494 (dwarf2out_source_line): Likewise. Output view resets and labels to
4495 the assembler, or select appropriate line info opcodes.
4496 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
4497 (optimize_string_length): Catch it. Adjust.
4498 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
4499 dw_val_class_view_list, and remove it if no longer needed.
4500 (hash_loc_list): Hash view numbers.
4501 (loc_list_hasher::equal): Compare them.
4502 (optimize_location_lists): Check whether a view list symbol is
4503 needed, and whether the locview attribute is present, and
4504 whether they match. Remove the locview attribute if no longer
4506 (index_location_lists): Call skip_loc_list_entry for test.
4507 (dwarf2out_finish): Call output_asm_line_debug_info for test.
4508 Use output_dwarf_version.
4509 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
4510 (struct dw_val_node): Add val_view_list.
4511 * final.c (SEEN_NEXT_VIEW): New.
4512 (set_next_view_needed): New.
4513 (clear_next_view_needed): New.
4514 (maybe_output_next_view): New.
4515 (final_start_function): Rename to...
4516 (final_start_function_1): ... this. Take pointer to FIRST,
4517 add SEEN parameter. Emit param bindings in the initial view.
4518 (final_start_function): Reintroduce SEEN-less interface.
4519 (final): Rename to...
4520 (final_1): ... this. Take SEEN parameter. Output final pending
4521 next view at the end.
4522 (final): Reintroduce seen-less interface.
4523 (final_scan_insn): Output pending next view before switching
4524 sections or ending a block. Mark the next view as needed when
4525 outputting variable locations. Notify debug backend of section
4526 changes, and of location view changes.
4527 (rest_of_handle_final): Adjust.
4528 * toplev.c (process_options): Autodetect value for debug variable
4529 location views option. Warn on incompat5 without -gdwarf-5.
4530 * doc/invoke.texi (gvariable-location-views): New.
4531 (gvariable-location-views=incompat5): New.
4532 (gno-variable-location-views): New.
4534 2018-02-08 David Malcolm <dmalcolm@redhat.com>
4536 PR tree-optimization/84136
4537 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
4538 that the result of find_edge is non-NULL.
4540 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4543 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
4544 storing integer register in SImode. Fix cost of 256 and 512
4545 byte aligned SSE register store.
4547 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4549 * config/i386/i386.c (ix86_multiplication_cost): Fix
4550 multiplication cost for TARGET_AVX512DQ.
4552 2018-02-08 Marek Polacek <polacek@redhat.com>
4554 PR tree-optimization/84238
4555 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
4558 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4560 PR tree-optimization/84265
4561 * tree-vect-stmts.c (vectorizable_store): Don't treat
4562 VMAT_CONTIGUOUS accesses as grouped.
4563 (vectorizable_load): Likewise.
4565 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4567 PR tree-optimization/81635
4568 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
4569 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
4570 (test_round_for_mask): New functions.
4571 (wide_int_cc_tests): Call test_round_for_mask.
4572 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
4573 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
4574 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
4575 range returned by get_range_info.
4577 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4580 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
4581 * symtab.c: Include builtins.h
4582 (symtab_node::output_to_lto_symbol_table_p): Move here
4583 from lto-streamer-out.c:output_symbol_p.
4584 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
4585 (output_symbol_p): Move all logic to symtab.c
4586 (produce_symtab): Update.
4588 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4590 * config/s390/s390-opts.h (enum indirect_branch): Define.
4591 * config/s390/s390-protos.h (s390_return_addr_from_memory)
4592 (s390_indirect_branch_via_thunk)
4593 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
4594 (enum s390_indirect_branch_type): Define.
4595 * config/s390/s390.c (struct s390_frame_layout, struct
4596 machine_function): Remove.
4597 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
4598 (indirect_branch_table_label_no, indirect_branch_table_name):
4600 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
4601 (enum s390_indirect_branch_option): Define.
4602 (s390_return_addr_from_memory): New function.
4603 (s390_handle_string_attribute): New function.
4604 (s390_attribute_table): Add new attribute handler.
4605 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
4606 (s390_indirect_branch_via_thunk): New function.
4607 (s390_indirect_branch_via_inline_thunk): New function.
4608 (s390_function_ok_for_sibcall): When jumping via thunk disallow
4609 sibling call optimization for non z10 compiles.
4610 (s390_emit_call): Force indirect branch target to be a single
4611 register. Add r1 clobber for non-z10 compiles.
4612 (s390_emit_epilogue): Emit return jump via return_use expander.
4613 (s390_reorg): Handle JUMP_INSNs as execute targets.
4614 (s390_option_override_internal): Perform validity checks for the
4615 new command line options.
4616 (s390_indirect_branch_attrvalue): New function.
4617 (s390_indirect_branch_settings): New function.
4618 (s390_set_current_function): Invoke s390_indirect_branch_settings.
4619 (s390_output_indirect_thunk_function): New function.
4620 (s390_code_end): Implement target hook.
4621 (s390_case_values_threshold): Implement target hook.
4622 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
4624 * config/s390/s390.h (struct s390_frame_layout)
4625 (struct machine_function): Move here from s390.c.
4626 (TARGET_INDIRECT_BRANCH_NOBP_RET)
4627 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
4628 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
4629 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
4630 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
4631 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
4632 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
4633 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
4634 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
4635 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
4636 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
4637 (mnemonic attribute): Add values which aren't recognized
4639 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
4640 pattern for branch conversion. Fix mnemonic attribute.
4641 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
4642 indirect branch via thunk if requested.
4643 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
4644 ("*indirect_jump"): Disable for branch conversion using out of
4646 ("indirect_jump_via_thunk<mode>_z10")
4647 ("indirect_jump_via_thunk<mode>")
4648 ("indirect_jump_via_inlinethunk<mode>_z10")
4649 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
4650 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
4651 ("casesi_jump_via_inlinethunk<mode>_z10")
4652 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
4653 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
4654 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
4655 ("*indirect2_jump"): Disable for branch conversion.
4656 ("casesi_jump"): Turn into expander and expand patterns for branch
4658 ("return_use"): New expander.
4659 ("*return"): Emit return via thunk and rename it to ...
4660 ("*return<mode>"): ... this one.
4661 * config/s390/s390.opt: Add new options and and enum for the
4664 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4666 * lra-constraints.c (match_reload): Unconditionally use
4667 gen_lowpart_SUBREG, rather than selecting between that
4668 and equivalent gen_rtx_SUBREG code.
4670 2018-02-08 Richard Biener <rguenther@suse.de>
4672 PR tree-optimization/84233
4673 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
4674 changed flag instead of boguously re-using phi_inserted.
4676 2018-02-08 Martin Jambor <mjambor@suse.cz>
4678 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
4679 static local variables.
4681 2018-02-08 Richard Biener <rguenther@suse.de>
4683 PR tree-optimization/84278
4684 * tree-vect-stmts.c (vectorizable_store): When looking for
4685 smaller vector types to perform grouped strided loads/stores
4686 make sure the mode is supported by the target.
4687 (vectorizable_load): Likewise.
4689 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4691 * config/aarch64/aarch64.c (aarch64_components_for_bb):
4692 Increase LDP/STP opportunities by adding adjacent callee-saves.
4694 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4696 PR rtl-optimization/84068
4697 PR rtl-optimization/83459
4698 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
4700 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
4702 PR tree-optimization/84224
4703 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
4704 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
4707 2018-02-07 Iain Sandoe <iain@codesourcery.com>
4710 * config/rs6000/altivec.md (*restore_world): Remove LR use.
4711 * config/rs6000/predicates.md (restore_world_operation): Adjust op
4712 count, remove one USE.
4714 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
4716 * doc/install.texi (Configuration): Document the
4717 --with-long-double-format={ibm,ieee} PowerPC configuration
4721 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
4722 Convert from define_expand to be define_insn_and_split. Rework
4723 float/double/_Float128 conversions to QI/HI/SImode to work with
4724 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
4725 conversions to QI/HImode types did a store and then a load to
4726 truncate the value. For conversions to VSX registers, don't split
4727 the insn, instead emit the code directly. Use the code iterator
4728 any_fix to combine signed and unsigned conversions.
4729 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
4730 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
4731 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4732 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
4733 (fix_<mode>di2_hw): Likewise.
4734 (fixuns_<mode>di2_hw): Likewise.
4735 (fix_<mode>si2_hw): Likewise.
4736 (fixuns_<mode>si2_hw): Likewise.
4737 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
4738 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4739 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
4740 fix<uns>_trunc<SFDF:mode>si2_p8.
4741 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
4743 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
4744 (fix<uns>_<mode>_mem): Likewise.
4745 (fctiw<u>z_<mode>_mem): Likewise.
4746 (fix<uns>_<mode>_mem): Likewise.
4747 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
4748 the register allocator from doing a direct move to the GPRs to do
4749 a store, and instead use the ISA 3.0 store byte/half-word from
4750 vector register instruction. For IEEE 128-bit floating point,
4751 also optimize stores of 32-bit ints.
4752 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
4754 2018-02-07 Alan Hayward <alan.hayward@arm.com>
4756 * genextract.c (push_pathstr_operand): New function to support
4758 (walk_rtx): Call push_pathstr_operand.
4759 (print_path): Support [a-zA-Z].
4761 2018-02-07 Richard Biener <rguenther@suse.de>
4763 PR tree-optimization/84037
4764 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
4765 (cse_and_gimplify_to_preheader): Declare.
4766 (vect_get_place_in_interleaving_chain): Likewise.
4767 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4769 (_loop_vec_info::~_loop_vec_info): Delete it.
4770 (cse_and_gimplify_to_preheader): New function.
4771 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
4772 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
4773 (vectorizable_load): Likewise. For grouped stores always base
4774 the IV on the first element.
4775 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
4776 condition before gimplifying.
4778 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4780 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
4781 *DIV_EXPR and *MOD_EXPR.
4783 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
4786 * config/i386/i386.c (ix86_option_override_internal): Mask out
4787 the CF_SET bit when checking -fcf-protection.
4789 2018-02-07 Tom de Vries <tom@codesourcery.com>
4792 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
4795 2018-02-07 Richard Biener <rguenther@suse.de>
4797 PR tree-optimization/84204
4798 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4801 PR tree-optimization/84205
4802 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
4803 special-case isl_ast_op_zdiv_r.
4805 PR tree-optimization/84223
4806 * graphite-scop-detection.c (gather_bbs::before_dom_children):
4807 Only add conditions from within the region.
4808 (gather_bbs::after_dom_children): Adjust.
4810 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
4813 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
4814 * config/avr/avr.md: Only post-reload split REG-REG moves if
4815 either register is GENERAL_REG_P.
4817 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4819 PR tree-optimization/84235
4820 * tree-ssa-scopedtables.c
4821 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
4822 if the subtraction is performed in floating point type where NaNs are
4823 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
4824 build 1. Formatting fix.
4826 2018-02-06 Jakub Jelinek <jakub@redhat.com>
4829 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
4830 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
4831 and skip it regardless of bb boundaries. Use CALL_P macro,
4832 don't test INSN_P (insn) together with CALL_P or JUMP_P check
4833 unnecessarily, formatting fix.
4835 2018-02-06 Michael Collison <michael.collison@arm.com>
4837 * config/arm/thumb2.md:
4838 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
4839 (*thumb_mov_notscc): Ditto.
4841 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
4844 * config/rs6000/rs6000.md (su code attribute): Use "u" for
4845 unsigned_fix, not "s".
4847 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4849 * configure.ac (gcc_fn_eh_frame_ro): New function.
4850 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
4851 correct .eh_frame permissions.
4852 * configure: Regenerate.
4854 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
4856 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
4859 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4861 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4862 Display warning message for -mno-speculate-indirect-jumps.
4864 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
4866 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
4868 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
4870 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
4872 PR tree-optimization/84225
4873 * tree-eh.c (find_trapping_overflow): Only call
4874 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
4876 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4879 * config/i386/i386.c: Reimplement the check of possible options
4880 -mibt/-mshstk conbination. Change error messages.
4881 * doc/invoke.texi: Fix a typo: remove extra '='.
4883 2018-02-06 Marek Polacek <polacek@redhat.com>
4885 PR tree-optimization/84228
4886 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
4888 2018-02-06 Tamar Christina <tamar.christina@arm.com>
4891 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
4892 emitted arch directives.
4893 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
4894 __ARM_FEATURE_COPROC before changing architectures.
4896 2018-02-06 Richard Biener <rguenther@suse.de>
4898 * config/i386/i386.c (print_reg): Fix typo.
4899 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
4901 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
4903 * configure: Regenerate.
4905 2018-02-05 Martin Sebor <msebor@redhat.com>
4907 PR tree-optimization/83369
4908 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
4911 2018-02-05 Martin Liska <mliska@suse.cz>
4913 * doc/invoke.texi: Cherry-pick upstream r323995.
4915 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
4917 * ira.c (ira_init_register_move_cost): Adjust comment.
4919 2018-02-05 Martin Liska <mliska@suse.cz>
4921 PR gcov-profile/84137
4922 * doc/gcov.texi: Fix typo in documentation.
4924 2018-02-05 Martin Liska <mliska@suse.cz>
4926 PR gcov-profile/83879
4927 * doc/gcov.texi: Document necessity of --dynamic-list-data when
4928 using dlopen functionality.
4930 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
4932 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
4933 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
4934 _mm_maskz_range_ss, _mm_mask_range_round_ss,
4935 _mm_maskz_range_round_ss): New intrinsics.
4936 (__builtin_ia32_rangesd128_round)
4937 (__builtin_ia32_rangess128_round): Remove.
4938 (__builtin_ia32_rangesd128_mask_round,
4939 __builtin_ia32_rangess128_mask_round): New builtins.
4940 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
4941 __builtin_ia32_rangess128_round): Remove.
4942 (__builtin_ia32_rangesd128_mask_round,
4943 __builtin_ia32_rangess128_mask_round): New builtins.
4944 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
4945 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
4946 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
4947 "<round_saeonly_constraint>")): Changed to ...
4948 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
4949 "<round_saeonly_scalar_constraint>")): ... this.
4950 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
4951 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
4952 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
4953 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
4954 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
4956 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
4958 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
4960 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
4961 Remove all values except native, 8540 and 8548.
4963 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
4965 * config/i386/i386.c (ix86_output_function_return): Pass
4966 INVALID_REGNUM, instead of -1, as invalid register number to
4967 indirect_thunk_name and output_indirect_thunk.
4969 2018-02-02 Julia Koval <julia.koval@intel.com>
4971 * config.gcc: Add -march=icelake.
4972 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
4973 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
4974 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
4975 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
4976 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
4977 (processor_target_table): Add icelake.
4978 (ix86_option_override_internal): Handle new PTAs.
4979 (get_builtin_code_for_version): Handle icelake.
4980 (M_INTEL_COREI7_ICELAKE): New.
4981 (fold_builtin_cpu): Handle icelake.
4982 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
4983 * doc/invoke.texi: Add -march=icelake.
4985 2018-02-02 Julia Koval <julia.koval@intel.com>
4987 * config/i386/i386.c (ix86_option_override_internal): Change flags type
4988 to wide_int_bitmask.
4989 * wide-int-bitmask.h: New.
4991 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4994 * config/i386/i386.md: Replace Pmode with word_mode in
4995 builtin_setjmp_setup and builtin_longjmp to support x32.
4997 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5001 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5003 (rs6000_supported_cpu_names): New static variable.
5004 (linux_cpu_translation_table): Likewise.
5005 (elf_platform) <cpu>: Define new static variable and use it.
5006 Translate kernel AT_PLATFORM name to canonical name if needed.
5007 Error if platform name is unknown.
5009 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
5012 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5014 2018-02-01 Jeff Law <law@redhat.com>
5017 * config/i386/i386.c (release_scratch_register_on_entry): Add new
5018 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
5019 the scratch if RELEASE_VIA_POP is false.
5020 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5021 If we have to save a temporary register, decrement SIZE appropriately.
5022 Pass new arguments to release_scratch_register_on_entry.
5023 (ix86_adjust_stack_and_probe): Likewise.
5024 (ix86_emit_probe_stack_range): Pass new arguments to
5025 release_scratch_register_on_entry.
5027 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5029 PR rtl-optimization/84157
5030 * combine.c (change_zero_ext): Use REG_P predicate in
5031 front of HARD_REGISTER_P predicate.
5033 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5035 * config/avr/avr.c (avr_option_override): Move disabling of
5036 -fdelete-null-pointer-checks to...
5037 * common/config/avr/avr-common.c (avr_option_optimization_table):
5040 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5042 PR tree-optimization/81635
5043 * tree-data-ref.c (split_constant_offset_1): For types that
5044 wrap on overflow, try to use range info to prove that wrapping
5047 2018-02-01 Renlin Li <renlin.li@arm.com>
5050 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5052 (aarch64_register_move_cost): Likewise.
5053 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5055 (REG_CLASS_NAMES): Likewise.
5056 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5057 TAILCALL_ADDR_REGS. Remove IP registers.
5058 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5060 2018-02-01 Richard Biener <rguenther@suse.de>
5062 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5063 for specifying RPO order and allow NULL for that.
5064 * domwalk.c (dom_walker::dom_walker): Likewise.
5065 (dom_walker::walk): Handle NULL RPO order.
5066 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5068 (rewrite_update_dom_walker): Likewise.
5069 (mark_def_dom_walker): Likewise.
5071 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5073 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5074 (aarch64_maybe_expand_sve_subreg_move): Declare.
5075 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5076 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5078 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5079 that are semantically a reverse operation.
5080 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5081 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5082 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5084 (aarch64_can_change_mode_class): For big-endian, forbid changes
5085 between two SVE modes if they have different element sizes.
5087 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5089 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5090 the TImode handling for big-endian targets.
5092 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5094 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5095 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5097 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5098 Remove BSWAP handing for big-endian targets and use the form of
5099 LD1RQ appropariate for the mode.
5101 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5103 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5104 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5107 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5110 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5111 check for operands that need to go through aarch64_sve_reload_be.
5113 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5115 PR tree-optimization/81661
5116 PR tree-optimization/84117
5117 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5118 * tree-eh.c: Include gimplify.h.
5119 (find_trapping_overflow, replace_trapping_overflow,
5120 rewrite_to_non_trapping_overflow): New functions.
5121 * tree-vect-loop.c: Include tree-eh.h.
5122 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5123 * tree-data-ref.c: Include tree-eh.h.
5124 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5126 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
5128 PR rtl-optimization/84123
5129 * combine.c (change_zero_ext): Check if hard register satisfies
5130 can_change_dest_mode before calling gen_lowpart_SUBREG.
5132 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
5135 * ira.c (ira_init_register_move_cost): Remove assert.
5137 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5139 PR rtl-optimization/84071
5140 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5141 * doc/tm.texi: Regenerate.
5143 2018-01-31 Richard Biener <rguenther@suse.de>
5145 PR tree-optimization/84132
5146 * tree-data-ref.c (analyze_miv_subscript): Properly
5147 check whether evolution_function_is_affine_multivariate_p
5148 before calling gcd_of_steps_may_divide_p.
5150 2018-01-31 Julia Koval <julia.koval@intel.com>
5153 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5154 * config/i386/i386.md (rdpid_rex64) New.
5155 (rdpid): Make 32bit only.
5157 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5160 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5161 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5163 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5166 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5168 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5170 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5172 PR rtl-optimization/84071
5173 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
5174 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
5176 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5178 * config/arc/arc.c (arc_handle_aux_attribute): New function.
5179 (arc_attribute_table): Add 'aux' attribute.
5180 (arc_in_small_data_p): Consider aux like variables.
5181 (arc_is_aux_reg_p): New function.
5182 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
5183 (arc_get_aux_arg): New function.
5184 (prepare_move_operands): Handle aux-register access.
5185 (arc_handle_aux_attribute): New function.
5186 * doc/extend.texi (ARC Variable attributes): Add subsection.
5188 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5190 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
5191 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
5192 (arc_attribute_table): Add 'uncached' attribute.
5193 (arc_print_operand): Print '.di' flag for uncached memory
5195 (arc_in_small_data_p): Do not consider for small data the uncached
5197 (arc_is_uncached_mem_p): New function.
5198 * config/arc/predicates.md (compact_store_memory_operand): Check
5199 for uncached memory accesses.
5200 (nonvol_nonimm_operand): Likewise.
5201 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5203 2018-01-31 Jakub Jelinek <jakub@redhat.com>
5206 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5207 falign-loops=): Add Optimization flag.
5209 2018-01-30 Jeff Law <law@redhat.com>
5212 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5213 INT_REGISTERS_SAVED. Check it prior to calling
5214 get_scratch_register_on_entry.
5215 (ix86_adjust_stack_and_probe): Similarly.
5216 (ix86_emit_probe_stack_range): Similarly.
5217 (ix86_expand_prologue): Corresponding changes.
5219 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5222 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5223 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
5225 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
5228 * lra-constraints.c (curr_insn_transform): Process AND in the
5231 2018-01-30 Jakub Jelinek <jakub@redhat.com>
5233 PR rtl-optimization/83986
5234 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5235 dependence against last_pending_memory_flush in addition to
5238 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
5240 PR tree-optimization/81611
5241 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5244 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5247 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5250 2018-01-30 Richard Biener <rguenther@suse.de>
5251 Jakub Jelinek <jakub@redhat.com>
5253 PR tree-optimization/84111
5254 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5255 inner loops added during recursion, as they don't have up-to-date
5258 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5261 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5262 (can_inline_edge_by_limits_p): ... here.
5263 (can_early_inline_edge_p, check_callers,
5264 update_caller_keys, update_callee_keys, recursive_inlining,
5265 add_new_edges_to_heap, speculation_useful_p,
5266 inline_small_functions,
5267 inline_small_functions, flatten_function,
5268 inline_to_all_callers_1): Update.
5270 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5272 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5275 2018-01-30 Richard Biener <rguenther@suse.de>
5277 PR tree-optimization/83008
5278 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5279 invariant and constant vector uses in stmts when they need
5282 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5285 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5286 * configure: Regenerate.
5288 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5290 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5292 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5293 Use gen_rtx_REG rather than gen_lowpart.
5295 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5297 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5298 rather than 0 when creating partial subregs.
5300 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5302 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5305 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
5308 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5309 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5310 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5311 flags. This restores the settings used before the 2017-07-24.
5312 Turning off pre increment/decrement/modify allows IVOPTS to
5313 optimize DF/SF loops where the index is an int.
5315 2018-01-29 Richard Biener <rguenther@suse.de>
5316 Kelvin Nilsen <kelvin@gcc.gnu.org>
5319 * tree-vect-stmts.c (vectorizable_call): Don't call
5320 targetm.vectorize_builtin_md_vectorized_function if callee is
5323 2018-01-22 Carl Love <cel@us.ibm.com>
5325 * doc/extend.tex: Fix typo in second arg in
5326 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5328 2018-01-29 Richard Biener <rguenther@suse.de>
5330 PR tree-optimization/84086
5331 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5332 (flush_ssaname_freelist): When SSA names were released reset
5333 the SCEV hash table.
5335 2018-01-29 Richard Biener <rguenther@suse.de>
5337 PR tree-optimization/84057
5338 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5339 removed paths when removing edges.
5341 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
5343 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5344 -mfunction-return=@var{choice}.
5346 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
5349 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5351 (layout::print_source_line): Likewise.
5352 (test_get_line_width_without_trailing_whitespace): Add test cases.
5354 2018-01-27 Jakub Jelinek <jakub@redhat.com>
5357 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5360 2018-01-26 Jim Wilson <jimw@sifive.com>
5362 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5364 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5367 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5369 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5370 and CMP + SUB-immediate -> SUBS.
5372 2018-01-26 Martin Sebor <msebor@redhat.com>
5374 PR tree-optimization/83896
5375 * tree-ssa-strlen.c (get_string_len): Rename...
5376 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
5377 Avoid assuming length is constant.
5378 (handle_char_store): Use HOST_WIDE_INT for string length.
5380 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
5383 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
5384 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
5386 2018-01-26 Richard Biener <rguenther@suse.de>
5388 PR rtl-optimization/84003
5389 * dse.c (record_store): Only record redundant stores when
5390 the earlier store aliases at least all accesses the later one does.
5392 2018-01-26 Jakub Jelinek <jakub@redhat.com>
5394 PR rtl-optimization/83985
5395 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
5396 REG_CFA_RESTORE insns.
5397 (delete_unmarked_insns): Don't ignore separate shrink wrapping
5398 REG_CFA_RESTORE insns here.
5401 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
5402 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
5404 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5406 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
5407 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
5408 (arc_init): Likewise.
5409 (arc_override_options): Likewise.
5410 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
5412 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
5414 * config/arc/arc.h (TARGET_DBNZ): Define.
5415 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
5416 properly set the tune attribute.
5417 (dbnz): Use TARGET_DBNZ guard.
5418 * config/arc/arc.opt (mtune): Add core3 option.
5420 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5422 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
5423 recognize new pic like addresses.
5424 (arc_delegitimize_address): Clean up.
5426 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5428 * config/arc/arc-arches.def: Option mrf16 valid for all
5430 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
5431 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
5432 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
5433 * config/arc/arc-tables.opt: Regenerate.
5434 * config/arc/arc.c (arc_conditional_register_usage): Handle
5435 reduced register file case.
5436 (arc_file_start): Set must have build attributes.
5437 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
5439 * config/arc/arc.opt (mrf16): Add new option.
5440 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
5441 * config/arc/genmultilib.awk: Handle new mrf16 option.
5442 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
5443 * config/arc/t-multilib: Regenerate.
5444 * doc/invoke.texi (ARC Options): Document mrf16 option.
5446 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5448 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
5449 * config/arc/arc.c (arc_handle_secure_attribute): New function.
5450 (arc_attribute_table): Add 'secure_call' attribute.
5451 (arc_print_operand): Print secure call operand.
5452 (arc_function_ok_for_sibcall): Don't optimize tail calls when
5454 (arc_is_secure_call_p): New function. * config/arc/arc.md
5455 (call_i): Add support for sjli instruction.
5456 (call_value_i): Likewise.
5457 * config/arc/constraints.md (Csc): New constraint.
5459 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5460 John Eric Martin <John.Martin@emmicro-us.com>
5462 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
5463 * config/arc/arc.c (_arc_jli_section): New struct.
5464 (arc_jli_section): New type.
5465 (rc_jli_sections): New static variable.
5466 (arc_handle_jli_attribute): New function.
5467 (arc_attribute_table): Add jli_always and jli_fixed attribute.
5468 (arc_file_end): New function.
5469 (TARGET_ASM_FILE_END): Define.
5470 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
5471 (arc_add_jli_section): New function.
5472 (jli_call_scan): Likewise.
5473 (arc_reorg): Call jli_call_scan.
5474 (arc_output_addsi): Remove 'S' from printing asm operand.
5475 (arc_is_jli_call_p): New function.
5476 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
5478 (movhi_insn): Likewise.
5479 (movsi_insn): Likewise.
5480 (movsi_set_cc_insn): Likewise.
5481 (loadqi_update): Likewise.
5482 (load_zeroextendqisi_update): Likewise.
5483 (load_signextendqisi_update): Likewise.
5484 (loadhi_update): Likewise.
5485 (load_zeroextendhisi_update): Likewise.
5486 (load_signextendhisi_update): Likewise.
5487 (loadsi_update): Likewise.
5488 (loadsf_update): Likewise.
5489 (movsicc_insn): Likewise.
5490 (bset_insn): Likewise.
5491 (bxor_insn): Likewise.
5492 (bclr_insn): Likewise.
5493 (bmsk_insn): Likewise.
5494 (bicsi3_insn): Likewise.
5495 (cmpsi_cc_c_insn): Likewise.
5496 (movsi_ne): Likewise.
5497 (movsi_cond_exec): Likewise.
5498 (clrsbsi2): Likewise.
5509 (call_i): Remove 'S' asm letter, add jli instruction.
5510 (call_value_i): Likewise.
5511 * config/arc/arc.op (mjli-always): New option.
5512 * config/arc/constraints.md (Cji): New constraint.
5513 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
5515 (subsf3_fpx): Likewise.
5516 (mulsf3_fpx): Likewise.
5517 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
5519 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
5521 * doc/invoke.texi (ARC): Document mjli-always option.
5523 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
5525 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
5526 avoid addition with 0 and use incw and decw where possible.
5528 2018-01-26 Richard Biener <rguenther@suse.de>
5530 PR tree-optimization/81082
5531 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
5532 association if it requires casting to unsigned.
5533 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
5534 from fold_plusminus_mult_expr to catch important cases late when
5535 range info is available.
5537 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5539 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
5540 * configure.ac (hidden_linkonce): New test.
5541 * configure: Regenerate.
5542 * config.in: Regenerate.
5544 2018-01-26 Julia Koval <julia.koval@intel.com>
5546 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
5547 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
5548 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
5549 _mm_mask_bitshuffle_epi64_mask): Fix type.
5550 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
5551 USI_FTYPE_V4DI_V4DI_USI): Remove.
5552 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
5553 __builtin_ia32_vpshufbitqmb256_mask,
5554 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
5555 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
5556 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
5558 2018-01-26 Alan Modra <amodra@gmail.com>
5561 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
5562 UNSPEC_VBPERMQ. Sort other unspecs.
5564 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
5566 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
5568 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
5571 * predict.c (drop_profile): Do not push/pop cfun; update also
5573 (handle_missing_profiles): Fix logic looking for zero profiles.
5575 2018-01-25 Jakub Jelinek <jakub@redhat.com>
5578 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
5579 on functions with #pragma omp declare simd or functions with simd
5581 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
5582 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5583 Remove trailing \n from warning_at calls.
5585 2018-01-25 Tom de Vries <tom@codesourcery.com>
5588 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5589 for neutered workers.
5591 2018-01-24 Joseph Myers <joseph@codesourcery.com>
5594 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
5595 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
5597 2018-01-24 Jeff Law <law@redhat.com>
5600 * i386.c (get_probe_interval): Move to earlier point.
5601 (ix86_compute_frame_layout): If -fstack-clash-protection and
5602 the frame is larger than the probe interval, then use pushes
5603 to save registers rather than reg->mem moves.
5604 (ix86_expand_prologue): Remove conditional for int_registers_saved
5607 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
5610 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
5611 min/max for never referenced object.
5613 2018-01-24 Jakub Jelinek <jakub@redhat.com>
5616 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
5618 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
5619 attributes from DECL_ATTRIBUTES (decl) without affecting
5620 DECL_ATTRIBUTES (current_function_decl).
5621 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
5622 functions with non-NULL DECL_ABSTRACT_ORIGIN.
5624 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
5626 PR tree-optimization/83979
5627 * fold-const.c (fold_comparison): Use constant_boolean_node
5628 instead of boolean_{true,false}_node.
5630 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
5632 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
5635 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5637 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5638 Simplify the clause that sets the length attribute.
5639 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5640 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
5641 clause that sets the length attribute.
5642 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5644 2018-01-24 Tom de Vries <tom@codesourcery.com>
5647 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
5648 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
5649 Add strict parameter.
5650 (prevent_branch_around_nothing): Insert dummy insn between branch to
5651 label and label with no ptx insn inbetween.
5652 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
5654 2018-01-24 Tom de Vries <tom@codesourcery.com>
5657 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5658 for neutered threads in warp.
5659 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
5661 2018-01-24 Richard Biener <rguenther@suse.de>
5663 PR tree-optimization/83176
5664 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
5667 2018-01-24 Richard Biener <rguenther@suse.de>
5669 PR tree-optimization/82819
5670 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
5671 code generating pluses that are no-ops in the target precision.
5673 2018-01-24 Richard Biener <rguenther@suse.de>
5676 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
5678 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5680 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
5681 to merge probabilities.
5682 * predict.c (probably_never_executed): Also mark as cold functions
5683 with global 0 profile and guessed local profile.
5684 * profile-count.c (profile_probability::combine_with_count): New
5686 * profile-count.h (profile_probability::operator*,
5687 profile_probability::operator*=, profile_probability::operator/,
5688 profile_probability::operator/=): Reduce precision to adjusted
5689 and set value to guessed on contradictory divisions.
5690 (profile_probability::combine_with_freq): Remove.
5691 (profile_probability::combine_wiht_count): Declare.
5692 (profile_count::force_nonzero):: Set to adjusted.
5693 (profile_count::probability_in):: Set quality to adjusted.
5694 * tree-ssa-tail-merge.c (replace_block_by): Use
5697 2018-01-23 Andrew Waterman <andrew@sifive.com>
5698 Jim Wilson <jimw@sifive.com>
5700 * config/riscv/riscv.c (riscv_stack_boundary): New.
5701 (riscv_option_override): Set riscv_stack_boundary. Handle
5702 riscv_preferred_stack_boundary_arg.
5703 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
5704 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
5705 (STACK_BOUNDARY): Set to riscv_stack_boundary.
5706 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
5707 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
5708 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
5710 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
5713 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
5714 of struct ix86_frame.
5715 (ix86_expand_epilogue): Likewise. Add a local variable for
5716 the reg_save_offset field in struct ix86_frame.
5718 2018-01-23 Bin Cheng <bin.cheng@arm.com>
5720 PR tree-optimization/82604
5721 * tree-loop-distribution.c (enum partition_kind): New enum item
5722 PKIND_PARTIAL_MEMSET.
5723 (partition_builtin_p): Support above new enum item.
5724 (generate_code_for_partition): Ditto.
5725 (compute_access_range): Differentiate cases that equality can be
5726 proven at all loops, the innermost loops or no loops.
5727 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
5728 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
5729 (finalize_partitions, distribute_loop): Don't fuse partition of
5730 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
5731 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
5734 2018-01-23 Martin Liska <mliska@suse.cz>
5736 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
5737 order to ignore the predictor.
5738 (PRED_POLYMORPHIC_CALL): Likewise.
5739 (PRED_RECURSIVE_CALL): Likewise.
5741 2018-01-23 Martin Liska <mliska@suse.cz>
5743 * tree-profile.c (tree_profiling): Print function header to
5744 aware reader which function we are working on.
5745 * value-prof.c (gimple_find_values_to_profile): Do not print
5746 not interesting value histograms.
5748 2018-01-23 Martin Liska <mliska@suse.cz>
5750 * profile-count.h (enum profile_quality): Add
5751 profile_uninitialized as the first value. Do not number values
5752 as they are zero based.
5753 (profile_count::verify): Update sanity check.
5754 (profile_probability::verify): Likewise.
5756 2018-01-23 Nathan Sidwell <nathan@acm.org>
5758 * doc/invoke.texi (ffor-scope): Deprecate.
5760 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5762 PR tree-optimization/83510
5763 * domwalk.c (set_all_edges_as_executable): New function.
5764 (dom_walker::dom_walker): Convert bool param
5765 "skip_unreachable_blocks" to enum reachability. Move setup of
5766 edge flags to set_all_edges_as_executable and only do it when
5767 reachability is REACHABLE_BLOCKS.
5768 * domwalk.h (enum dom_walker::reachability): New enum.
5769 (dom_walker::dom_walker): Convert bool param
5770 "skip_unreachable_blocks" to enum reachability.
5771 (set_all_edges_as_executable): New decl.
5772 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
5773 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
5775 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
5776 but converting true to REACHABLE_BLOCKS.
5777 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
5779 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
5780 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
5781 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
5783 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
5784 if check_all_array_refs will be called.
5786 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5788 * tree.c (selftest::test_location_wrappers): Add more test
5791 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5793 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
5794 (selftest::test_bit_in_range): Likewise.
5796 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5799 * doc/sourcebuild.texi (vect_float): Say that the selector
5800 only describes the situation when -funsafe-math-optimizations is on.
5801 (vect_float_strict): Document.
5803 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5805 PR tree-optimization/83965
5806 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5807 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
5808 instead of checking only for a reduction.
5809 (vect_recog_widen_sum_pattern): Likewise.
5811 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5813 * predict.c (probably_never_executed): Only use precise profile info.
5814 (compute_function_frequency): Skip after inlining hack since we now
5815 have quality checking.
5817 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5819 * profile-count.h (profile_probability::very_unlikely,
5820 profile_probability::unlikely, profile_probability::even): Set
5821 precision to guessed.
5823 2018-01-23 Richard Biener <rguenther@suse.de>
5825 PR tree-optimization/83963
5826 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
5827 Properly terminate dominator walk when crossing the exit edge not
5828 when visiting its source block.
5830 2018-01-23 Jakub Jelinek <jakub@redhat.com>
5833 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
5834 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
5836 2018-01-22 Jakub Jelinek <jakub@redhat.com>
5838 PR tree-optimization/83957
5839 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
5840 semicolon after for body surrounded by braces.
5842 PR tree-optimization/83081
5843 * profile-count.h (profile_probability::split): New method.
5844 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
5845 Use profile_probability::split.
5846 (do_compare_rtx_and_jump): Fix adjustment of probabilities
5847 when splitting a single conditional jump into 2.
5849 2018-01-22 David Malcolm <dmalcolm@redhat.com>
5851 PR tree-optimization/69452
5852 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
5855 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5857 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
5858 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
5859 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
5861 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5863 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
5864 * config/rl78/rl78.md (movdi): New define_expand.
5865 * config/rl78/rl78.c (rl78_split_movdi): New function.
5867 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
5870 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
5872 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
5873 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
5874 128-bit to produce an UNSPEC move to get the double word with the
5875 signbit and then a shift directly to do signbit.
5876 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
5877 implementation with a new version that just does either a direct
5878 move or a regular move. Move memory interface to separate insns.
5879 Move insns so they are next to the expander.
5880 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
5881 with signbit move. Split big and little endian case.
5882 (signbit<mode>2_dm_mem_le): Likewise.
5883 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
5884 (signbit<mode>2_dm2): Likewise.
5886 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5888 * config/rl78/rl78.md (anddi3): New define_expand.
5890 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5892 * config/rl78/rl78.md (umindi3): New define_expand.
5894 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5896 * config/rl78/rl78.md (smindi3): New define_expand.
5898 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5900 * config/rl78/rl78.md (smaxdi3): New define_expand.
5902 2018-01-22 Carl Love <cel@us.ibm.com>
5904 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
5905 LVX_V1TI): Add macro expansion.
5906 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
5907 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
5908 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
5909 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
5910 Change check to determine if the instruction is a byte reversing
5911 entry. Fix typo in comment.
5912 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
5913 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
5914 Add def_builtin calls for new builtins.
5915 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
5916 Add define_insn expansion.
5918 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5920 * config/rl78/rl78.md (umaxdi3): New define_expand.
5922 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5924 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
5925 for non-QImode registers.
5927 2018-01-22 Richard Biener <rguenther@suse.de>
5929 PR tree-optimization/83963
5930 * graphite-scop-detection.c (scop_detection::get_sese): Delay
5931 including the loop exit block.
5932 (scop_detection::merge_sese): Likewise.
5933 (scop_detection::add_scop): Do it here instead.
5935 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5937 * doc/sourcebuild.texi (arm_softfloat): Document.
5939 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
5942 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
5943 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
5944 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
5946 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5947 David Edelsohn <dje.gcc@gmail.com>
5950 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5951 Change "crset eq" to "crset 2".
5952 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5953 (*call_indirect_aix<mode>_nospec): Likewise.
5954 (*call_value_indirect_aix<mode>_nospec): Likewise.
5955 (*call_indirect_elfv2<mode>_nospec): Likewise.
5956 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
5957 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
5958 change assembly output from . to $.
5959 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5960 (indirect_jump<mode>_nospec): Change assembly output from . to $.
5961 (*tablejump<mode>_internal1_nospec): Likewise.
5963 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
5966 * config/sh/sh_optimize_sett_clrt.cc:
5967 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
5969 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
5971 PR tree-optimization/83940
5972 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
5973 offset_dt to vect_constant_def rather than vect_unknown_def_type.
5974 (vect_check_load_store_mask): Add a mask_dt_out parameter and
5975 use it to pass back the definition type.
5976 (vect_check_store_rhs): Likewise rhs_dt_out.
5977 (vect_build_gather_load_calls): Add a mask_dt argument and use
5978 it instead of a call to vect_is_simple_use.
5979 (vectorizable_store): Update calls to vect_check_load_store_mask
5980 and vect_check_store_rhs. Use the dt returned by the latter instead
5981 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
5982 instead of calls to vect_is_simple_use. Pass the scalar rather
5983 than the vector operand to vect_is_simple_use when handling
5984 second and subsequent copies of an rhs value.
5985 (vectorizable_load): Update calls to vect_check_load_store_mask
5986 and vect_build_gather_load_calls. Use the cached mask_dt and
5987 gs_info.offset_dt instead of calls to vect_is_simple_use.
5989 2018-01-20 Jakub Jelinek <jakub@redhat.com>
5992 * tree-emutls.c: Include gimplify.h.
5993 (lower_emutls_2): New function.
5994 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
5995 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
5996 it before further processing.
5999 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6000 UINTVAL (trueop1) instead of INTVAL (op1).
6002 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6006 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6007 INCOMING_FRAME_SP_OFFSET if not defined.
6008 (scan_trace): Add ENTRY argument. If true and
6009 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6010 emit a note to adjust the CFA offset.
6011 (create_cfi_notes): Adjust scan_trace callers.
6012 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6013 INCOMING_FRAME_SP_OFFSET in the CIE.
6014 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6015 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6017 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6018 * doc/tm.texi: Regenerated.
6020 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6022 PR rtl-optimization/83147
6023 * lra-constraints.c (remove_inheritance_pseudos): Use
6024 lra_substitute_pseudo_within_insn.
6026 2018-01-19 Tom de Vries <tom@codesourcery.com>
6027 Cesar Philippidis <cesar@codesourcery.com>
6030 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6032 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6035 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6036 spaces for function labels.
6038 2018-01-19 Martin Liska <mliska@suse.cz>
6040 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6041 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6042 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6043 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6044 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6045 (PRED_CONST_RETURN): Change from 69 to 65.
6046 (PRED_NULL_RETURN): Change from 91 to 71.
6047 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6048 (PRED_LOOP_GUARD): Change from 66 to 73.
6050 2018-01-19 Martin Liska <mliska@suse.cz>
6052 * predict.c (predict_insn_def): Add new assert.
6053 (struct branch_predictor): Change type to signed integer.
6054 (test_prediction_value_range): Amend test to cover
6056 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6057 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6058 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6059 (PRED_LOOP_IV_COMPARE): Likewise.
6060 * predict.h (PROB_UNINITIALIZED): Define new constant.
6062 2018-01-19 Martin Liska <mliska@suse.cz>
6064 * predict.c (dump_prediction): Add new format for
6065 analyze_brprob.py script which is enabled with -details
6067 * profile-count.h (precise_p): New function.
6069 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6071 PR tree-optimization/83922
6072 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6073 there are no statements that need masking.
6074 (vect_active_double_reduction_p): New function.
6075 (vect_analyze_loop_operations): Use it when handling phis that
6076 are not in the loop header.
6078 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6080 PR tree-optimization/83914
6081 * tree-vect-loop.c (vectorizable_induction): Don't convert
6082 init_expr or apply the peeling adjustment for inductions
6083 that are nested within the vectorized loop.
6085 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6087 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6090 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6094 * function.h (gimplify_parameters): Add gimple_seq * argument.
6095 * function.c: Include gimple.h and options.h.
6096 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6097 for the added local temporaries if needed.
6098 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6099 if there are any parameter cleanups, wrap whole body into a
6100 try/finally with the cleanups.
6102 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6105 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6106 Use GET_MODE_CLASS for scalar floating point.
6108 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6112 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6113 Fix call of call_cgraph_insertion_hooks.
6115 2018-01-18 Martin Sebor <msebor@redhat.com>
6117 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6119 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6122 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6125 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
6128 * common.opt: (-ffile-prefix-map): New option.
6129 * opts.c (common_handle_option): Defer it.
6130 * opts-global.c (handle_common_deferred_options): Handle it.
6131 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6132 * file-prefix-map.h: New file.
6133 (remap_debug_filename, add_debug_prefix_map): ...here.
6134 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6135 * final.c (debug_prefix_map, add_debug_prefix_map
6136 remap_debug_filename): Move to...
6137 * file-prefix-map.c: New file.
6138 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6139 generalize, get rid of alloca(), use strrchr() instead of strchr().
6140 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6141 Implement in terms of add_prefix_map().
6142 (remap_macro_filename, remap_debug_filename): Implement in term of
6144 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6145 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6146 * dbxout.c: Include file-prefix-map.h.
6147 * varasm.c: Likewise.
6148 * vmsdbgout.c: Likewise.
6149 * xcoffout.c: Likewise.
6150 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6151 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6152 * doc/invoke.texi (-ffile-prefix-map): Document.
6153 (-fdebug-prefix-map): Update description.
6155 2018-01-18 Martin Liska <mliska@suse.cz>
6157 * config/i386/i386.c (indirect_thunk_name): Document that also
6159 (output_indirect_thunk): Document why both instructions
6160 (pause and lfence) are generated.
6162 2018-01-18 Richard Biener <rguenther@suse.de>
6164 PR tree-optimization/83887
6165 * graphite-scop-detection.c
6166 (scop_detection::get_nearest_dom_with_single_entry): Remove.
6167 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6168 (scop_detection::merge_sese): Re-implement with a flood-fill
6169 algorithm that properly finds a SESE region if it exists.
6171 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6174 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
6175 pointer_diff optimizations use view_convert instead of convert.
6177 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6179 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6180 Generate different code for -mno-speculate-indirect-jumps.
6181 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6182 (*call_indirect_aix<mode>): Disable for
6183 -mno-speculate-indirect-jumps.
6184 (*call_indirect_aix<mode>_nospec): New define_insn.
6185 (*call_value_indirect_aix<mode>): Disable for
6186 -mno-speculate-indirect-jumps.
6187 (*call_value_indirect_aix<mode>_nospec): New define_insn.
6188 (*sibcall_nonlocal_sysv<mode>): Generate different code for
6189 -mno-speculate-indirect-jumps.
6190 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6192 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6194 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
6195 long double type, set the flags for noting the default long double
6196 type, even if we don't pass or return a long double type.
6198 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
6201 * ipa-inline.c (flatten_function): Do not overwrite final inlining
6204 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
6206 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6207 support for merge[hl].
6208 (fold_mergehl_helper): New helper function.
6209 (tree-vector-builder.h): New #include for tree_vector_builder usage.
6210 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6211 (altivec_vmrglw_direct): Add xxmrglw insn.
6213 2018-01-17 Andrew Waterman <andrew@sifive.com>
6215 * config/riscv/riscv.c (riscv_conditional_register_usage): If
6216 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6218 2018-01-17 David Malcolm <dmalcolm@redhat.com>
6221 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6222 call the lto_location_cache before reading the
6223 DECL_SOURCE_LOCATION of the types.
6225 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
6226 Richard Sandiford <richard.sandiford@linaro.org>
6228 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
6229 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6230 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6231 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6232 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6234 * config/aarch64/constraints.md (aarch64_movti_operand):
6236 * config/aarch64/predicates.md (Uti): Add new constraint.
6238 2018-01-17 Carl Love <cel@us.ibm.com>
6240 * config/rs6000/vsx.md (define_expand xl_len_r,
6241 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6242 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6244 (define_expand, define_insn): Move the shift left from the
6245 define_insn to the define_expand for lxvl and stxvl instructions.
6246 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6247 and XL_LEN_R definitions to PURE.
6249 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
6251 * config/i386/i386.c (indirect_thunk_name): Declare regno
6252 as unsigned int. Compare regno with INVALID_REGNUM.
6253 (output_indirect_thunk): Ditto.
6254 (output_indirect_thunk_function): Ditto.
6255 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6256 in the call to output_indirect_thunk_function.
6258 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6261 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6262 rather than the size of inner_type to determine the stack slot size
6263 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6265 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6268 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6271 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6273 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6274 endian Linux systems to optionally enable multilibs for selecting
6275 the long double type if the user configured an explicit type.
6276 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6277 have no long double multilibs if not defined.
6278 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6279 warn if the user used -mabi={ieee,ibm}longdouble and we built
6280 multilibs for long double.
6281 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6282 appropriate multilib option.
6283 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6285 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6286 for building long double multilibs.
6287 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6289 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6291 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6294 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6296 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6299 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6302 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6305 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6307 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6308 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6310 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6312 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6313 different rtl trees depending on TARGET_64BIT.
6314 (rs6000_gen_lvx): Likewise.
6316 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6318 * config/visium/visium.md (nop): Tweak comment.
6319 (hazard_nop): Likewise.
6321 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6323 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6324 -mspeculate-indirect-jumps.
6325 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6326 for -mno-speculate-indirect-jumps.
6327 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6328 (*call_value_indirect_elfv2<mode>): Disable for
6329 -mno-speculate-indirect-jumps.
6330 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6331 (indirect_jump): Emit different RTL for
6332 -mno-speculate-indirect-jumps.
6333 (*indirect_jump<mode>): Disable for
6334 -mno-speculate-indirect-jumps.
6335 (*indirect_jump<mode>_nospec): New define_insn.
6336 (tablejump): Emit different RTL for
6337 -mno-speculate-indirect-jumps.
6338 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6339 (tablejumpsi_nospec): New define_expand.
6340 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6341 (tablejumpdi_nospec): New define_expand.
6342 (*tablejump<mode>_internal1): Disable for
6343 -mno-speculate-indirect-jumps.
6344 (*tablejump<mode>_internal1_nospec): New define_insn.
6345 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6348 2018-01-16 Artyom Skrobov tyomitch@gmail.com
6350 * caller-save.c (insert_save): Drop unnecessary parameter. All
6353 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6354 Richard Biener <rguenth@suse.de>
6357 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6358 return early, inline manually is_gimple_sizepos. Make sure if we
6359 call gimplify_expr we don't end up with a gimple constant.
6360 * tree.c (variably_modified_type_p): Don't return true for
6361 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
6362 * gimplify.h (is_gimple_sizepos): Remove.
6364 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6366 PR tree-optimization/83857
6367 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6368 vectorizable_live_operation for pure SLP statements.
6369 (vectorizable_live_operation): Handle PHIs.
6371 2018-01-16 Richard Biener <rguenther@suse.de>
6373 PR tree-optimization/83867
6374 * tree-vect-stmts.c (vect_transform_stmt): Precompute
6375 nested_in_vect_loop_p since the scalar stmt may get invalidated.
6377 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6380 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
6381 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
6382 If off is not INTEGER_CST, issue a may not be aligned warning
6383 rather than isn't aligned. Use isn%'t rather than isn't.
6384 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
6386 <case MULT_EXPR>: Improve the case when bottom and one of the
6387 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
6388 operand, in that case check if the other operand is multiple of
6389 bottom divided by the INTEGER_CST operand.
6391 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6394 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
6395 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
6396 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
6397 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
6398 * config/pa/pa.c (pa_function_arg_advance): Likewise.
6399 (pa_function_arg, pa_arg_partial_bytes): Likewise.
6400 (pa_function_arg_size): New function.
6402 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6404 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
6405 in a separate statement.
6407 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6409 PR tree-optimization/83847
6410 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
6411 group gathers and scatters.
6413 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6415 PR rtl-optimization/86620
6416 * params.def (max-sched-ready-insns): Bump minimum value to 1.
6418 PR rtl-optimization/83213
6419 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
6420 to last if both are JUMP_INSNs.
6422 PR tree-optimization/83843
6423 * gimple-ssa-store-merging.c
6424 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
6425 store_immediate_info for bswap/nop orig_stores.
6427 2018-01-15 Andrew Waterman <andrew@sifive.com>
6429 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
6431 <UDIV>: Increase cost if !TARGET_DIV.
6433 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
6435 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
6436 (define_attr "cr_logical_3op"): New.
6437 (cceq_ior_compare): Adjust.
6438 (cceq_ior_compare_complement): Adjust.
6439 (*cceq_rev_compare): Adjust.
6440 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
6441 (is_cracked_insn): Adjust.
6442 (insn_must_be_first_in_group): Adjust.
6443 * config/rs6000/40x.md: Adjust.
6444 * config/rs6000/440.md: Adjust.
6445 * config/rs6000/476.md: Adjust.
6446 * config/rs6000/601.md: Adjust.
6447 * config/rs6000/603.md: Adjust.
6448 * config/rs6000/6xx.md: Adjust.
6449 * config/rs6000/7450.md: Adjust.
6450 * config/rs6000/7xx.md: Adjust.
6451 * config/rs6000/8540.md: Adjust.
6452 * config/rs6000/cell.md: Adjust.
6453 * config/rs6000/e300c2c3.md: Adjust.
6454 * config/rs6000/e500mc.md: Adjust.
6455 * config/rs6000/e500mc64.md: Adjust.
6456 * config/rs6000/e5500.md: Adjust.
6457 * config/rs6000/e6500.md: Adjust.
6458 * config/rs6000/mpc.md: Adjust.
6459 * config/rs6000/power4.md: Adjust.
6460 * config/rs6000/power5.md: Adjust.
6461 * config/rs6000/power6.md: Adjust.
6462 * config/rs6000/power7.md: Adjust.
6463 * config/rs6000/power8.md: Adjust.
6464 * config/rs6000/power9.md: Adjust.
6465 * config/rs6000/rs64.md: Adjust.
6466 * config/rs6000/titan.md: Adjust.
6468 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6470 * config/i386/predicates.md (indirect_branch_operand): Rewrite
6471 ix86_indirect_branch_register logic.
6473 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6475 * config/i386/constraints.md (Bs): Update
6476 ix86_indirect_branch_register check. Don't check
6477 ix86_indirect_branch_register with GOT_memory_operand.
6479 * config/i386/predicates.md (GOT_memory_operand): Don't check
6480 ix86_indirect_branch_register here.
6481 (GOT32_symbol_operand): Likewise.
6483 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6485 * config/i386/predicates.md (constant_call_address_operand):
6486 Rewrite ix86_indirect_branch_register logic.
6487 (sibcall_insn_operand): Likewise.
6489 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6491 * config/i386/constraints.md (Bs): Replace
6492 ix86_indirect_branch_thunk_register with
6493 ix86_indirect_branch_register.
6495 * config/i386/i386.md (indirect_jump): Likewise.
6496 (tablejump): Likewise.
6497 (*sibcall_memory): Likewise.
6498 (*sibcall_value_memory): Likewise.
6499 Peepholes of indirect call and jump via memory: Likewise.
6500 * config/i386/i386.opt: Likewise.
6501 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6502 (GOT_memory_operand): Likewise.
6503 (call_insn_operand): Likewise.
6504 (sibcall_insn_operand): Likewise.
6505 (GOT32_symbol_operand): Likewise.
6507 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6510 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
6511 type rather than type addr's type points to.
6512 (expand_omp_atomic_mutex): Likewise.
6513 (expand_omp_atomic): Likewise.
6515 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6518 * config/i386/i386.c (output_indirect_thunk_function): Use
6519 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
6520 for __x86_return_thunk.
6522 2018-01-15 Richard Biener <rguenther@suse.de>
6525 * expmed.c (extract_bit_field_1): Fix typo.
6527 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6530 * config/arm/iterators.md (VF): New mode iterator.
6531 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
6532 Remove integer-related logic from pattern.
6533 (neon_vabd<mode>_3): Likewise.
6535 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6538 * common.opt (fstrict-overflow): No longer an alias.
6539 (fwrapv-pointer): New option.
6540 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
6541 also for pointer types based on flag_wrapv_pointer.
6542 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
6543 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
6544 opts->x_flag_wrapv got set.
6545 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
6546 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
6547 POINTER_TYPE_OVERFLOW_UNDEFINED.
6548 * match.pd: Likewise in address comparison pattern.
6549 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
6551 2018-01-15 Richard Biener <rguenther@suse.de>
6554 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
6555 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
6556 Reset type names to their identifier if their TYPE_DECL doesn't
6557 have linkage (and thus is used for ODR and devirt).
6558 (save_debug_info_for_decl): Remove.
6559 (save_debug_info_for_type): Likewise.
6560 (add_tree_to_fld_list): Adjust.
6561 * tree-pretty-print.c (dump_generic_node): Make dumping of
6562 type names more robust.
6564 2018-01-15 Richard Biener <rguenther@suse.de>
6566 * BASE-VER: Bump to 8.0.1.
6568 2018-01-14 Martin Sebor <msebor@redhat.com>
6571 * builtins.c (check_access): Avoid warning when the no-warning bit
6574 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
6576 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
6577 * ira-color (allocno_hard_regs_compare): Likewise.
6579 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
6582 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
6583 Use .pushsection/.popsection.
6585 2018-01-14 Martin Sebor <msebor@redhat.com>
6588 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
6590 2018-01-14 Jakub Jelinek <jakub@redhat.com>
6592 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
6593 entry from extra_headers.
6594 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
6595 extra_headers, make the list bitwise identical to the i?86-*-* one.
6597 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6599 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
6600 -mcmodel=large with -mindirect-branch=thunk,
6601 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
6602 -mfunction-return=thunk-extern.
6603 * doc/invoke.texi: Document -mcmodel=large is incompatible with
6604 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
6605 -mfunction-return=thunk and -mfunction-return=thunk-extern.
6607 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6609 * config/i386/i386.c (print_reg): Print the name of the full
6610 integer register without '%'.
6611 (ix86_print_operand): Handle 'V'.
6612 * doc/extend.texi: Document 'V' modifier.
6614 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6616 * config/i386/constraints.md (Bs): Disallow memory operand for
6617 -mindirect-branch-register.
6619 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6620 (GOT_memory_operand): Likewise.
6621 (call_insn_operand): Likewise.
6622 (sibcall_insn_operand): Likewise.
6623 (GOT32_symbol_operand): Likewise.
6624 * config/i386/i386.md (indirect_jump): Call convert_memory_address
6625 for -mindirect-branch-register.
6626 (tablejump): Likewise.
6627 (*sibcall_memory): Likewise.
6628 (*sibcall_value_memory): Likewise.
6629 Disallow peepholes of indirect call and jump via memory for
6630 -mindirect-branch-register.
6631 (*call_pop): Replace m with Bw.
6632 (*call_value_pop): Likewise.
6633 (*sibcall_pop_memory): Replace m with Bs.
6634 * config/i386/i386.opt (mindirect-branch-register): New option.
6635 * doc/invoke.texi: Document -mindirect-branch-register option.
6637 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6639 * config/i386/i386-protos.h (ix86_output_function_return): New.
6640 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
6641 set function_return_type.
6642 (indirect_thunk_name): Add ret_p to indicate thunk for function
6644 (output_indirect_thunk_function): Pass false to
6645 indirect_thunk_name.
6646 (ix86_output_indirect_branch_via_reg): Likewise.
6647 (ix86_output_indirect_branch_via_push): Likewise.
6648 (output_indirect_thunk_function): Create alias for function
6649 return thunk if regno < 0.
6650 (ix86_output_function_return): New function.
6651 (ix86_handle_fndecl_attribute): Handle function_return.
6652 (ix86_attribute_table): Add function_return.
6653 * config/i386/i386.h (machine_function): Add
6654 function_return_type.
6655 * config/i386/i386.md (simple_return_internal): Use
6656 ix86_output_function_return.
6657 (simple_return_internal_long): Likewise.
6658 * config/i386/i386.opt (mfunction-return=): New option.
6659 (indirect_branch): Mention -mfunction-return=.
6660 * doc/extend.texi: Document function_return function attribute.
6661 * doc/invoke.texi: Document -mfunction-return= option.
6663 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6665 * config/i386/i386-opts.h (indirect_branch): New.
6666 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
6667 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
6668 with local indirect jump when converting indirect call and jump.
6669 (ix86_set_indirect_branch_type): New.
6670 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
6671 (indirectlabelno): New.
6672 (indirect_thunk_needed): Likewise.
6673 (indirect_thunk_bnd_needed): Likewise.
6674 (indirect_thunks_used): Likewise.
6675 (indirect_thunks_bnd_used): Likewise.
6676 (INDIRECT_LABEL): Likewise.
6677 (indirect_thunk_name): Likewise.
6678 (output_indirect_thunk): Likewise.
6679 (output_indirect_thunk_function): Likewise.
6680 (ix86_output_indirect_branch_via_reg): Likewise.
6681 (ix86_output_indirect_branch_via_push): Likewise.
6682 (ix86_output_indirect_branch): Likewise.
6683 (ix86_output_indirect_jmp): Likewise.
6684 (ix86_code_end): Call output_indirect_thunk_function if needed.
6685 (ix86_output_call_insn): Call ix86_output_indirect_branch if
6687 (ix86_handle_fndecl_attribute): Handle indirect_branch.
6688 (ix86_attribute_table): Add indirect_branch.
6689 * config/i386/i386.h (machine_function): Add indirect_branch_type
6690 and has_local_indirect_jump.
6691 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
6693 (tablejump): Likewise.
6694 (*indirect_jump): Use ix86_output_indirect_jmp.
6695 (*tablejump_1): Likewise.
6696 (simple_return_indirect_internal): Likewise.
6697 * config/i386/i386.opt (mindirect-branch=): New option.
6698 (indirect_branch): New.
6701 (thunk-inline): Likewise.
6702 (thunk-extern): Likewise.
6703 * doc/extend.texi: Document indirect_branch function attribute.
6704 * doc/invoke.texi: Document -mindirect-branch= option.
6706 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
6709 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
6711 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
6713 * ipa-inline.c (want_inline_small_function_p): Return false if
6714 inlining has already failed with CIF_FINAL_ERROR.
6715 (update_caller_keys): Call want_inline_small_function_p before
6717 (update_callee_keys): Likewise.
6719 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6721 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
6723 (rs6000_quadword_masked_address_p): Likewise.
6724 (quad_aligned_load_p): Likewise.
6725 (quad_aligned_store_p): Likewise.
6726 (const_load_sequence_p): Add comment to describe the outer-most loop.
6727 (mimic_memory_attributes_and_flags): New function.
6728 (rs6000_gen_stvx): Likewise.
6729 (replace_swapped_aligned_store): Likewise.
6730 (rs6000_gen_lvx): Likewise.
6731 (replace_swapped_aligned_load): Likewise.
6732 (replace_swapped_load_constant): Capitalize argument name in
6733 comment describing this function.
6734 (rs6000_analyze_swaps): Add a third pass to search for vector loads
6735 and stores that access quad-word aligned addresses and replace
6736 with stvx or lvx instructions when appropriate.
6737 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
6738 New function prototype.
6739 (rs6000_quadword_masked_address_p): Likewise.
6740 (rs6000_gen_lvx): Likewise.
6741 (rs6000_gen_stvx): Likewise.
6742 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
6743 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
6744 when memory address is aligned.
6745 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
6746 this split to select lvx instruction when memory address is aligned.
6747 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
6748 instruction when memory address is aligned.
6749 (*vsx_le_perm_load_v16qi): Likewise.
6750 (four unnamed splitters): Modify to select the stvx instruction
6751 when memory is aligned.
6753 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
6755 * predict.c (determine_unlikely_bbs): Handle correctly BBs
6756 which appears in the queue multiple times.
6758 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6759 Alan Hayward <alan.hayward@arm.com>
6760 David Sherwood <david.sherwood@arm.com>
6762 * tree-vectorizer.h (vec_lower_bound): New structure.
6763 (_loop_vec_info): Add check_nonzero and lower_bounds.
6764 (LOOP_VINFO_CHECK_NONZERO): New macro.
6765 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
6766 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
6767 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
6768 fields. Make seg_len the distance travelled, not including the
6770 (dr_direction_indicator): Declare.
6771 (dr_zero_step_indicator): Likewise.
6772 (dr_known_forward_stride_p): Likewise.
6773 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
6775 (runtime_alias_check_p): Allow runtime alias checks with
6777 (operator ==): Compare access_size and align.
6778 (prune_runtime_alias_test_list): Rework for new distinction between
6779 the access_size and seg_len.
6780 (create_intersect_range_checks_index): Likewise. Cope with polynomial
6782 (get_segment_min_max): New function.
6783 (create_intersect_range_checks): Use it.
6784 (dr_step_indicator): New function.
6785 (dr_direction_indicator): Likewise.
6786 (dr_zero_step_indicator): Likewise.
6787 (dr_known_forward_stride_p): Likewise.
6788 * tree-loop-distribution.c (data_ref_segment_size): Return
6789 DR_STEP * (niters - 1).
6790 (compute_alias_check_pairs): Update call to the dr_with_seg_len
6792 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
6793 (vect_preserves_scalar_order_p): New function, split out from...
6794 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
6795 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
6796 (vect_vfa_access_size): New function.
6797 (vect_vfa_align): Likewise.
6798 (vect_compile_time_alias): Take access_size_a and access_b arguments.
6799 (dump_lower_bound): New function.
6800 (vect_check_lower_bound): Likewise.
6801 (vect_small_gap_p): Likewise.
6802 (vectorizable_with_step_bound_p): Likewise.
6803 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
6804 depencies if the vectorization factor is 1. Convert the checks
6805 for nonzero steps into checks on the bounds of DR_STEP. Try using
6806 a bunds check for variable steps if the minimum required step is
6807 relatively small. Update calls to the dr_with_seg_len
6808 constructor and to vect_compile_time_alias.
6809 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
6811 (vect_loop_versioning): Call it.
6812 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
6814 (vect_estimate_min_profitable_iters): Account for any bounds checks.
6816 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6817 Alan Hayward <alan.hayward@arm.com>
6818 David Sherwood <david.sherwood@arm.com>
6820 * doc/sourcebuild.texi (vect_scatter_store): Document.
6821 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
6823 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
6825 * genopinit.c (main): Add supports_vec_scatter_store and
6826 supports_vec_scatter_store_cached to target_optabs.
6827 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
6828 IFN_MASK_SCATTER_STORE.
6829 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
6831 * internal-fn.h (internal_store_fn_p): Declare.
6832 (internal_fn_stored_value_index): Likewise.
6833 * internal-fn.c (scatter_store_direct): New macro.
6834 (expand_scatter_store_optab_fn): New function.
6835 (direct_scatter_store_optab_supported_p): New macro.
6836 (internal_store_fn_p): New function.
6837 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
6838 IFN_MASK_SCATTER_STORE.
6839 (internal_fn_mask_index): Likewise.
6840 (internal_fn_stored_value_index): New function.
6841 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
6843 * optabs-query.h (supports_vec_scatter_store_p): Declare.
6844 * optabs-query.c (supports_vec_scatter_store_p): New function.
6845 * tree-vectorizer.h (vect_get_store_rhs): Declare.
6846 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
6847 true for scatter stores.
6848 (vect_gather_scatter_fn_p): Handle scatter stores too.
6849 (vect_check_gather_scatter): Consider using scatter stores if
6850 supports_vec_scatter_store_p.
6851 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
6853 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6854 internal_fn_stored_value_index.
6855 (check_load_store_masking): Handle scatter stores too.
6856 (vect_get_store_rhs): Make public.
6857 (vectorizable_call): Use internal_store_fn_p.
6858 (vectorizable_store): Handle scatter store internal functions.
6859 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
6860 when deciding whether the end of the group has been reached.
6861 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
6862 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
6863 (mask_scatter_store<mode>): New insns.
6865 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6866 Alan Hayward <alan.hayward@arm.com>
6867 David Sherwood <david.sherwood@arm.com>
6869 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
6870 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
6871 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
6873 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
6874 Use vect_truncate_gather_scatter_offset if we can't treat the
6875 operation as a normal gather load or scatter store.
6876 (get_group_load_store_type): Take the gather_scatter_info
6877 as argument. Try using a gather load or scatter store for
6878 single-element groups.
6879 (get_load_store_type): Update calls to get_group_load_store_type
6880 and vect_use_strided_gather_scatters_p.
6882 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6883 Alan Hayward <alan.hayward@arm.com>
6884 David Sherwood <david.sherwood@arm.com>
6886 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
6887 optional tree argument.
6888 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
6890 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
6891 but continue to use the current value as a fallback.
6892 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
6893 to compare the updates.
6894 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
6895 (get_load_store_type): Use it when handling a strided access.
6896 (vect_get_strided_load_store_ops): New function.
6897 (vect_get_data_ptr_increment): Likewise.
6898 (vectorizable_load): Handle strided gather loads. Always pass
6899 a step to vect_create_data_ref_ptr and bump_vector_ptr.
6901 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6902 Alan Hayward <alan.hayward@arm.com>
6903 David Sherwood <david.sherwood@arm.com>
6905 * doc/md.texi (gather_load@var{m}): Document.
6906 (mask_gather_load@var{m}): Likewise.
6907 * genopinit.c (main): Add supports_vec_gather_load and
6908 supports_vec_gather_load_cached to target_optabs.
6909 * optabs-tree.c (init_tree_optimization_optabs): Use
6910 ggc_cleared_alloc to allocate target_optabs.
6911 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
6912 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
6914 * internal-fn.h (internal_load_fn_p): Declare.
6915 (internal_gather_scatter_fn_p): Likewise.
6916 (internal_fn_mask_index): Likewise.
6917 (internal_gather_scatter_fn_supported_p): Likewise.
6918 * internal-fn.c (gather_load_direct): New macro.
6919 (expand_gather_load_optab_fn): New function.
6920 (direct_gather_load_optab_supported_p): New macro.
6921 (direct_internal_fn_optab): New function.
6922 (internal_load_fn_p): Likewise.
6923 (internal_gather_scatter_fn_p): Likewise.
6924 (internal_fn_mask_index): Likewise.
6925 (internal_gather_scatter_fn_supported_p): Likewise.
6926 * optabs-query.c (supports_at_least_one_mode_p): New function.
6927 (supports_vec_gather_load_p): Likewise.
6928 * optabs-query.h (supports_vec_gather_load_p): Declare.
6929 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
6930 and memory_type field.
6931 (NUM_PATTERNS): Bump to 15.
6932 * tree-vect-data-refs.c: Include internal-fn.h.
6933 (vect_gather_scatter_fn_p): New function.
6934 (vect_describe_gather_scatter_call): Likewise.
6935 (vect_check_gather_scatter): Try using internal functions for
6936 gather loads. Recognize existing calls to a gather load function.
6937 (vect_analyze_data_refs): Consider using gather loads if
6938 supports_vec_gather_load_p.
6939 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
6940 (vect_get_gather_scatter_offset_type): Likewise.
6941 (vect_convert_mask_for_vectype): Likewise.
6942 (vect_add_conversion_to_patterm): Likewise.
6943 (vect_try_gather_scatter_pattern): Likewise.
6944 (vect_recog_gather_scatter_pattern): New pattern recognizer.
6945 (vect_vect_recog_func_ptrs): Add it.
6946 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6947 internal_fn_mask_index and internal_gather_scatter_fn_p.
6948 (check_load_store_masking): Take the gather_scatter_info as an
6949 argument and handle gather loads.
6950 (vect_get_gather_scatter_ops): New function.
6951 (vectorizable_call): Check internal_load_fn_p.
6952 (vectorizable_load): Likewise. Handle gather load internal
6954 (vectorizable_store): Update call to check_load_store_masking.
6955 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
6956 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
6957 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
6958 (aarch64_gather_scale_operand_d): New predicates.
6959 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
6960 (mask_gather_load<mode>): New insns.
6962 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6963 Alan Hayward <alan.hayward@arm.com>
6964 David Sherwood <david.sherwood@arm.com>
6966 * optabs.def (fold_left_plus_optab): New optab.
6967 * doc/md.texi (fold_left_plus_@var{m}): Document.
6968 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
6969 * internal-fn.c (fold_left_direct): Define.
6970 (expand_fold_left_optab_fn): Likewise.
6971 (direct_fold_left_optab_supported_p): Likewise.
6972 * fold-const-call.c (fold_const_fold_left): New function.
6973 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
6974 * tree-parloops.c (valid_reduction_p): New function.
6975 (gather_scalar_reductions): Use it.
6976 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
6977 (vect_finish_replace_stmt): Declare.
6978 * tree-vect-loop.c (fold_left_reduction_fn): New function.
6979 (needs_fold_left_reduction_p): New function, split out from...
6980 (vect_is_simple_reduction): ...here. Accept reductions that
6981 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
6982 (vect_force_simple_reduction): Also store the reduction type in
6983 the assignment's STMT_VINFO_REDUC_TYPE.
6984 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
6985 (merge_with_identity): New function.
6986 (vect_expand_fold_left): Likewise.
6987 (vectorize_fold_left_reduction): Likewise.
6988 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
6989 scalar phi in place for it. Check for target support and reject
6990 cases that would reassociate the operation. Defer the transform
6991 phase to vectorize_fold_left_reduction.
6992 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
6993 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
6994 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
6996 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6998 * tree-if-conv.c (predicate_mem_writes): Remove redundant
6999 call to ifc_temp_var.
7001 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7002 Alan Hayward <alan.hayward@arm.com>
7003 David Sherwood <david.sherwood@arm.com>
7005 * target.def (legitimize_address_displacement): Take the original
7006 offset as a poly_int.
7007 * targhooks.h (default_legitimize_address_displacement): Update
7009 * targhooks.c (default_legitimize_address_displacement): Likewise.
7010 * doc/tm.texi: Regenerate.
7011 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7012 as an argument, moving assert of ad->disp == ad->disp_term to...
7013 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
7014 Try calling targetm.legitimize_address_displacement before expanding
7015 the address rather than afterwards, and adjust for the new interface.
7016 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7017 Match the new hook interface. Handle SVE addresses.
7018 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7021 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7023 * Makefile.in (OBJS): Add early-remat.o.
7024 * target.def (select_early_remat_modes): New hook.
7025 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7026 * doc/tm.texi: Regenerate.
7027 * targhooks.h (default_select_early_remat_modes): Declare.
7028 * targhooks.c (default_select_early_remat_modes): New function.
7029 * timevar.def (TV_EARLY_REMAT): New timevar.
7030 * passes.def (pass_early_remat): New pass.
7031 * tree-pass.h (make_pass_early_remat): Declare.
7032 * early-remat.c: New file.
7033 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7035 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7037 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7038 Alan Hayward <alan.hayward@arm.com>
7039 David Sherwood <david.sherwood@arm.com>
7041 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7042 vfm1 with a bound_epilog parameter.
7043 (vect_do_peeling): Update calls accordingly, and move the prologue
7044 call earlier in the function. Treat the base bound_epilog as 0 for
7045 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7046 this base when peeling for gaps.
7047 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7048 with fully-masked loops.
7049 (vect_estimate_min_profitable_iters): Handle the single peeled
7050 iteration in that case.
7052 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7053 Alan Hayward <alan.hayward@arm.com>
7054 David Sherwood <david.sherwood@arm.com>
7056 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7057 single-element interleaving even if the size is not a power of 2.
7058 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7059 accesses for single-element interleaving if the group size is
7062 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7063 Alan Hayward <alan.hayward@arm.com>
7064 David Sherwood <david.sherwood@arm.com>
7066 * doc/md.texi (fold_extract_last_@var{m}): Document.
7067 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7068 * optabs.def (fold_extract_last_optab): New optab.
7069 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7070 * internal-fn.c (fold_extract_direct): New macro.
7071 (expand_fold_extract_optab_fn): Likewise.
7072 (direct_fold_extract_optab_supported_p): Likewise.
7073 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7074 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7075 EXTRACT_LAST_REDUCTION.
7076 (get_initial_def_for_reduction): Do not create an initial vector
7077 for EXTRACT_LAST_REDUCTION reductions.
7078 (vectorizable_reduction): Leave the scalar phi in place for
7079 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7080 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7081 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7082 transform phase to vectorizable_condition.
7083 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7085 (vect_finish_stmt_generation): ...here.
7086 (vect_finish_replace_stmt): New function.
7087 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7088 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7090 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7092 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7093 Alan Hayward <alan.hayward@arm.com>
7094 David Sherwood <david.sherwood@arm.com>
7096 * doc/md.texi (extract_last_@var{m}): Document.
7097 * optabs.def (extract_last_optab): New optab.
7098 * internal-fn.def (EXTRACT_LAST): New internal function.
7099 * internal-fn.c (cond_unary_direct): New macro.
7100 (expand_cond_unary_optab_fn): Likewise.
7101 (direct_cond_unary_optab_supported_p): Likewise.
7102 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7103 loops using EXTRACT_LAST.
7104 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7105 (extract_last_<mode>): ...this optab.
7106 (vec_extract<mode><Vel>): Update accordingly.
7108 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7109 Alan Hayward <alan.hayward@arm.com>
7110 David Sherwood <david.sherwood@arm.com>
7112 * target.def (empty_mask_is_expensive): New hook.
7113 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7114 * doc/tm.texi: Regenerate.
7115 * targhooks.h (default_empty_mask_is_expensive): Declare.
7116 * targhooks.c (default_empty_mask_is_expensive): New function.
7117 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7118 if the target says that empty masks are expensive.
7119 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7121 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
7123 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7124 Alan Hayward <alan.hayward@arm.com>
7125 David Sherwood <david.sherwood@arm.com>
7127 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7128 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7129 (vect_use_loop_mask_for_alignment_p): New function.
7130 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7131 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7132 niters_skip argument. Make sure that the first niters_skip elements
7133 of the first iteration are inactive.
7134 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7135 Update call to vect_set_loop_masks_directly.
7136 (get_misalign_in_elems): New function, split out from...
7137 (vect_gen_prolog_loop_niters): ...here.
7138 (vect_update_init_of_dr): Take a code argument that specifies whether
7139 the adjustment should be added or subtracted.
7140 (vect_update_init_of_drs): Likewise.
7141 (vect_prepare_for_masked_peels): New function.
7142 (vect_do_peeling): Skip prologue peeling if we're using a mask
7143 instead. Update call to vect_update_inits_of_drs.
7144 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7146 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7147 alignment. Do not include the number of peeled iterations in
7148 the minimum threshold in that case.
7149 (vectorizable_induction): Adjust the start value down by
7150 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7151 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7152 Take the number of skipped iterations into account when calculating
7154 * tree-vect-stmts.c (vect_gen_while_not): New function.
7156 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7157 Alan Hayward <alan.hayward@arm.com>
7158 David Sherwood <david.sherwood@arm.com>
7160 * doc/sourcebuild.texi (vect_fully_masked): Document.
7161 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7163 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7165 (vect_analyze_loop_2): ...here. Don't check the vectorization
7166 factor against the number of loop iterations if the loop is
7169 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7170 Alan Hayward <alan.hayward@arm.com>
7171 David Sherwood <david.sherwood@arm.com>
7173 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
7174 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
7175 (dump_groups): Update accordingly.
7176 (iv_use::mem_type): New member variable.
7177 (address_p): New function.
7178 (record_use): Add a mem_type argument and initialize the new
7180 (record_group_use): Add a mem_type argument. Use address_p.
7181 Remove obsolete null checks of base_object. Update call to record_use.
7182 (find_interesting_uses_op): Update call to record_group_use.
7183 (find_interesting_uses_cond): Likewise.
7184 (find_interesting_uses_address): Likewise.
7185 (get_mem_type_for_internal_fn): New function.
7186 (find_address_like_use): Likewise.
7187 (find_interesting_uses_stmt): Try find_address_like_use before
7188 calling find_interesting_uses_op.
7189 (addr_offset_valid_p): Use the iv mem_type field as the type
7190 of the addressed memory.
7191 (add_autoinc_candidates): Likewise.
7192 (get_address_cost): Likewise.
7193 (split_small_address_groups_p): Use address_p.
7194 (split_address_groups): Likewise.
7195 (add_iv_candidate_for_use): Likewise.
7196 (autoinc_possible_for_pair): Likewise.
7197 (rewrite_groups): Likewise.
7198 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
7199 (determine_group_iv_cost): Update after split of USE_ADDRESS.
7200 (get_alias_ptr_type_for_ptr_address): New function.
7201 (rewrite_use_address): Rewrite address uses in calls that were
7202 identified by find_address_like_use.
7204 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7205 Alan Hayward <alan.hayward@arm.com>
7206 David Sherwood <david.sherwood@arm.com>
7208 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7210 * gimple-expr.h (is_gimple_addressable: Likewise.
7211 * gimple-expr.c (is_gimple_address): Likewise.
7212 * internal-fn.c (expand_call_mem_ref): New function.
7213 (expand_mask_load_optab_fn): Use it.
7214 (expand_mask_store_optab_fn): Likewise.
7216 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7217 Alan Hayward <alan.hayward@arm.com>
7218 David Sherwood <david.sherwood@arm.com>
7220 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7221 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7222 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7223 (cond_umax@var{mode}): Document.
7224 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
7225 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
7226 (cond_umin_optab, cond_umax_optab): New optabs.
7227 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
7228 (COND_IOR, COND_XOR): New internal functions.
7229 * internal-fn.h (get_conditional_internal_fn): Declare.
7230 * internal-fn.c (cond_binary_direct): New macro.
7231 (expand_cond_binary_optab_fn): Likewise.
7232 (direct_cond_binary_optab_supported_p): Likewise.
7233 (get_conditional_internal_fn): New function.
7234 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7235 Cope with reduction statements that are vectorized as calls rather
7237 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7238 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7239 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7240 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7241 (UNSPEC_COND_EOR): New unspecs.
7242 (optab): Add mappings for them.
7243 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7244 (sve_int_op, sve_fp_op): New int attributes.
7246 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7247 Alan Hayward <alan.hayward@arm.com>
7248 David Sherwood <david.sherwood@arm.com>
7250 * optabs.def (while_ult_optab): New optab.
7251 * doc/md.texi (while_ult@var{m}@var{n}): Document.
7252 * internal-fn.def (WHILE_ULT): New internal function.
7253 * internal-fn.h (direct_internal_fn_supported_p): New override
7254 that takes two types as argument.
7255 * internal-fn.c (while_direct): New macro.
7256 (expand_while_optab_fn): New function.
7257 (convert_optab_supported_p): Likewise.
7258 (direct_while_optab_supported_p): New macro.
7259 * wide-int.h (wi::udiv_ceil): New function.
7260 * tree-vectorizer.h (rgroup_masks): New structure.
7261 (vec_loop_masks): New typedef.
7262 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7264 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7265 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7266 (vect_max_vf): New function.
7267 (slpeel_make_loop_iterate_ntimes): Delete.
7268 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7269 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7270 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7271 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7272 internal-fn.h, stor-layout.h and optabs-query.h.
7273 (vect_set_loop_mask): New function.
7274 (add_preheader_seq): Likewise.
7275 (add_header_seq): Likewise.
7276 (interleave_supported_p): Likewise.
7277 (vect_maybe_permute_loop_masks): Likewise.
7278 (vect_set_loop_masks_directly): Likewise.
7279 (vect_set_loop_condition_masked): Likewise.
7280 (vect_set_loop_condition_unmasked): New function, split out from
7281 slpeel_make_loop_iterate_ntimes.
7282 (slpeel_make_loop_iterate_ntimes): Rename to..
7283 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7284 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7285 (vect_do_peeling): Update call accordingly.
7286 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7288 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7289 mask_compare_type, can_fully_mask_p and fully_masked_p.
7290 (release_vec_loop_masks): New function.
7291 (_loop_vec_info): Use it to free the loop masks.
7292 (can_produce_all_loop_masks_p): New function.
7293 (vect_get_max_nscalars_per_iter): Likewise.
7294 (vect_verify_full_masking): Likewise.
7295 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7296 retries, and free the mask rgroups before retrying. Check loop-wide
7297 reasons for disallowing fully-masked loops. Make the final decision
7298 about whether use a fully-masked loop or not.
7299 (vect_estimate_min_profitable_iters): Do not assume that peeling
7300 for the number of iterations will be needed for fully-masked loops.
7301 (vectorizable_reduction): Disable fully-masked loops.
7302 (vectorizable_live_operation): Likewise.
7303 (vect_halve_mask_nunits): New function.
7304 (vect_double_mask_nunits): Likewise.
7305 (vect_record_loop_mask): Likewise.
7306 (vect_get_loop_mask): Likewise.
7307 (vect_transform_loop): Handle the case in which the final loop
7308 iteration might handle a partial vector. Call vect_set_loop_condition
7309 instead of slpeel_make_loop_iterate_ntimes.
7310 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7311 (check_load_store_masking): New function.
7312 (prepare_load_store_mask): Likewise.
7313 (vectorizable_store): Handle fully-masked loops.
7314 (vectorizable_load): Likewise.
7315 (supportable_widening_operation): Use vect_halve_mask_nunits for
7317 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7318 (vect_gen_while): New function.
7319 * config/aarch64/aarch64.md (umax<mode>3): New expander.
7320 (aarch64_uqdec<mode>): New insn.
7322 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7323 Alan Hayward <alan.hayward@arm.com>
7324 David Sherwood <david.sherwood@arm.com>
7326 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7327 (reduc_xor_scal_optab): New optabs.
7328 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7329 (reduc_xor_scal_@var{m}): Document.
7330 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7331 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7333 * fold-const-call.c (fold_const_call): Handle them.
7334 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7335 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7336 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7337 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7338 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7339 (UNSPEC_XORV): New unspecs.
7340 (optab): Add entries for them.
7341 (BITWISEV): New int iterator.
7342 (bit_reduc_op): New int attributes.
7344 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7345 Alan Hayward <alan.hayward@arm.com>
7346 David Sherwood <david.sherwood@arm.com>
7348 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7349 * internal-fn.def (VEC_SHL_INSERT): New internal function.
7350 * optabs.def (vec_shl_insert_optab): New optab.
7351 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7352 (duplicate_and_interleave): Likewise.
7353 * tree-vect-loop.c: Include internal-fn.h.
7354 (neutral_op_for_slp_reduction): New function, split out from
7355 get_initial_defs_for_reduction.
7356 (get_initial_def_for_reduction): Handle option 2 for variable-length
7357 vectors by loading the neutral value into a vector and then shifting
7358 the initial value into element 0.
7359 (get_initial_defs_for_reduction): Replace the code argument with
7360 the neutral value calculated by neutral_op_for_slp_reduction.
7361 Use gimple_build_vector for constant-length vectors.
7362 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7363 but the first group_size elements have a neutral value.
7364 Use duplicate_and_interleave otherwise.
7365 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7366 Update call to get_initial_defs_for_reduction. Handle SLP
7367 reductions for variable-length vectors by creating one vector
7368 result for each scalar result, with the elements associated
7369 with other scalar results stubbed out with the neutral value.
7370 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7371 Require IFN_VEC_SHL_INSERT for double reductions on
7372 variable-length vectors, or SLP reductions that have
7373 a neutral value. Require can_duplicate_and_interleave_p
7374 support for variable-length unchained SLP reductions if there
7375 is no neutral value, such as for MIN/MAX reductions. Also require
7376 the number of vector elements to be a multiple of the number of
7377 SLP statements when doing variable-length unchained SLP reductions.
7378 Update call to vect_create_epilog_for_reduction.
7379 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
7380 and remove initial values.
7381 (duplicate_and_interleave): Make public.
7382 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
7383 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
7385 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7386 Alan Hayward <alan.hayward@arm.com>
7387 David Sherwood <david.sherwood@arm.com>
7389 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
7390 (can_duplicate_and_interleave_p): New function.
7391 (vect_get_and_check_slp_defs): Take the vector of statements
7392 rather than just the current one. Remove excess parentheses.
7393 Restriction rejectinon of vect_constant_def and vect_external_def
7394 for variable-length vectors to boolean types, or types for which
7395 can_duplicate_and_interleave_p is false.
7396 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
7397 (duplicate_and_interleave): New function.
7398 (vect_get_constant_vectors): Use gimple_build_vector for
7399 constant-length vectors and suitable variable-length constant
7400 vectors. Use duplicate_and_interleave for other variable-length
7401 vectors. Don't defer the update when inserting new statements.
7403 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7404 Alan Hayward <alan.hayward@arm.com>
7405 David Sherwood <david.sherwood@arm.com>
7407 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
7408 min_profitable_iters doesn't go negative.
7410 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7411 Alan Hayward <alan.hayward@arm.com>
7412 David Sherwood <david.sherwood@arm.com>
7414 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
7415 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
7416 * optabs.def (vec_mask_load_lanes_optab): New optab.
7417 (vec_mask_store_lanes_optab): Likewise.
7418 * internal-fn.def (MASK_LOAD_LANES): New internal function.
7419 (MASK_STORE_LANES): Likewise.
7420 * internal-fn.c (mask_load_lanes_direct): New macro.
7421 (mask_store_lanes_direct): Likewise.
7422 (expand_mask_load_optab_fn): Handle masked operations.
7423 (expand_mask_load_lanes_optab_fn): New macro.
7424 (expand_mask_store_optab_fn): Handle masked operations.
7425 (expand_mask_store_lanes_optab_fn): New macro.
7426 (direct_mask_load_lanes_optab_supported_p): Likewise.
7427 (direct_mask_store_lanes_optab_supported_p): Likewise.
7428 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
7430 (vect_load_lanes_supported): Likewise.
7431 * tree-vect-data-refs.c (strip_conversion): New function.
7432 (can_group_stmts_p): Likewise.
7433 (vect_analyze_data_ref_accesses): Use it instead of checking
7434 for a pair of assignments.
7435 (vect_store_lanes_supported): Take a masked_p parameter.
7436 (vect_load_lanes_supported): Likewise.
7437 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
7438 vect_store_lanes_supported and vect_load_lanes_supported.
7439 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7440 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
7441 parameter. Don't allow gaps for masked accesses.
7442 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
7443 and vect_load_lanes_supported.
7444 (get_load_store_type): Take a masked_p parameter and update
7445 call to get_group_load_store_type.
7446 (vectorizable_store): Update call to get_load_store_type.
7447 Handle IFN_MASK_STORE_LANES.
7448 (vectorizable_load): Update call to get_load_store_type.
7449 Handle IFN_MASK_LOAD_LANES.
7451 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7452 Alan Hayward <alan.hayward@arm.com>
7453 David Sherwood <david.sherwood@arm.com>
7455 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
7457 * config/aarch64/aarch64-protos.h
7458 (aarch64_sve_struct_memory_operand_p): Declare.
7459 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
7460 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
7461 (VPRED, vpred): Handle SVE structure modes.
7462 * config/aarch64/constraints.md (Utx): New constraint.
7463 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
7464 (aarch64_sve_struct_nonimmediate_operand): New predicates.
7465 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
7466 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
7467 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
7468 structure modes. Split into pieces after RA.
7469 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
7470 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
7472 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
7473 SVE structure modes.
7474 (aarch64_classify_address): Likewise.
7475 (sizetochar): Move earlier in file.
7476 (aarch64_print_operand): Handle SVE register lists.
7477 (aarch64_array_mode): New function.
7478 (aarch64_sve_struct_memory_operand_p): Likewise.
7479 (TARGET_ARRAY_MODE): Redefine.
7481 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7482 Alan Hayward <alan.hayward@arm.com>
7483 David Sherwood <david.sherwood@arm.com>
7485 * target.def (array_mode): New target hook.
7486 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
7487 * doc/tm.texi: Regenerate.
7488 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
7489 * hooks.c (hook_optmode_mode_uhwi_none): New function.
7490 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
7492 * stor-layout.c (mode_for_array): Likewise. Support polynomial
7495 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7496 Alan Hayward <alan.hayward@arm.com>
7497 David Sherwood <david.sherwood@arm.com>
7499 * fold-const.c (fold_binary_loc): Check the argument types
7500 rather than the result type when testing for a vector operation.
7502 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7504 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
7505 * doc/tm.texi: Regenerate.
7507 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7508 Alan Hayward <alan.hayward@arm.com>
7509 David Sherwood <david.sherwood@arm.com>
7511 * doc/invoke.texi (-msve-vector-bits=): Document new option.
7512 (sve): Document new AArch64 extension.
7513 * doc/md.texi (w): Extend the description of the AArch64
7514 constraint to include SVE vectors.
7515 (Upl, Upa): Document new AArch64 predicate constraints.
7516 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
7518 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
7519 (msve-vector-bits=): New option.
7520 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
7521 SVE when these are disabled.
7522 (sve): New extension.
7523 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
7524 modes. Adjust their number of units based on aarch64_sve_vg.
7525 (MAX_BITSIZE_MODE_ANY_MODE): Define.
7526 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
7527 aarch64_addr_query_type.
7528 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
7529 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
7530 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
7531 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
7532 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
7533 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
7534 (aarch64_simd_imm_zero_p): Delete.
7535 (aarch64_check_zero_based_sve_index_immediate): Declare.
7536 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7537 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7538 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7539 (aarch64_sve_float_mul_immediate_p): Likewise.
7540 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7542 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
7543 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
7544 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
7545 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
7546 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
7547 (aarch64_regmode_natural_size): Likewise.
7548 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
7549 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
7551 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
7552 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
7553 for VG and the SVE predicate registers.
7554 (V_ALIASES): Add a "z"-prefixed alias.
7555 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
7556 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
7557 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
7558 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
7559 (REG_CLASS_NAMES): Add entries for them.
7560 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
7561 and the predicate registers.
7562 (aarch64_sve_vg): Declare.
7563 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
7564 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
7565 (REGMODE_NATURAL_SIZE): Define.
7566 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
7568 * config/aarch64/aarch64.c: Include cfgrtl.h.
7569 (simd_immediate_info): Add a constructor for series vectors,
7570 and an associated step field.
7571 (aarch64_sve_vg): New variable.
7572 (aarch64_dbx_register_number): Handle VG and the predicate registers.
7573 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
7574 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
7575 (VEC_ANY_DATA, VEC_STRUCT): New constants.
7576 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
7577 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
7578 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
7579 (aarch64_get_mask_mode): New functions.
7580 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
7581 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
7582 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
7583 predicate modes and predicate registers. Explicitly restrict
7584 GPRs to modes of 16 bytes or smaller. Only allow FP registers
7585 to store a vector mode if it is recognized by
7586 aarch64_classify_vector_mode.
7587 (aarch64_regmode_natural_size): New function.
7588 (aarch64_hard_regno_caller_save_mode): Return the original mode
7590 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
7591 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
7592 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
7593 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
7595 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
7596 does not overlap dest if the function is frame-related. Handle
7598 (aarch64_split_add_offset): New function.
7599 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
7600 them aarch64_add_offset.
7601 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
7602 and update call to aarch64_sub_sp.
7603 (aarch64_add_cfa_expression): New function.
7604 (aarch64_expand_prologue): Pass extra temporary registers to the
7605 functions above. Handle the case in which we need to emit new
7606 DW_CFA_expressions for registers that were originally saved
7607 relative to the stack pointer, but now have to be expressed
7608 relative to the frame pointer.
7609 (aarch64_output_mi_thunk): Pass extra temporary registers to the
7611 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
7612 IP0 and IP1 values for SVE frames.
7613 (aarch64_expand_vec_series): New function.
7614 (aarch64_expand_sve_widened_duplicate): Likewise.
7615 (aarch64_expand_sve_const_vector): Likewise.
7616 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
7617 Handle SVE constants. Use emit_move_insn to move a force_const_mem
7618 into the register, rather than emitting a SET directly.
7619 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
7620 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
7621 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
7622 (offset_9bit_signed_scaled_p): New functions.
7623 (aarch64_replicate_bitmask_imm): New function.
7624 (aarch64_bitmask_imm): Use it.
7625 (aarch64_cannot_force_const_mem): Reject expressions involving
7626 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
7627 (aarch64_classify_index): Handle SVE indices, by requiring
7628 a plain register index with a scale that matches the element size.
7629 (aarch64_classify_address): Handle SVE addresses. Assert that
7630 the mode of the address is VOIDmode or an integer mode.
7631 Update call to aarch64_classify_symbol.
7632 (aarch64_classify_symbolic_expression): Update call to
7633 aarch64_classify_symbol.
7634 (aarch64_const_vec_all_in_range_p): New function.
7635 (aarch64_print_vector_float_operand): Likewise.
7636 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
7637 "vN" for FP registers with SVE modes. Handle (const ...) vectors
7638 and the FP immediates 1.0 and 0.5.
7639 (aarch64_print_address_internal): Handle SVE addresses.
7640 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
7641 (aarch64_regno_regclass): Handle predicate registers.
7642 (aarch64_secondary_reload): Handle big-endian reloads of SVE
7644 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
7645 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
7646 (aarch64_convert_sve_vector_bits): New function.
7647 (aarch64_override_options): Use it to handle -msve-vector-bits=.
7648 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7650 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
7651 Handle SVE vector and predicate modes. Accept VL-based constants
7652 that need only one temporary register, and VL offsets that require
7653 no temporary registers.
7654 (aarch64_conditional_register_usage): Mark the predicate registers
7655 as fixed if SVE isn't available.
7656 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
7657 Return true for SVE vector and predicate modes.
7658 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
7659 rather than an unsigned int. Handle SVE modes.
7660 (aarch64_preferred_simd_mode): Update call accordingly. Handle
7662 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
7664 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7665 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7666 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7667 (aarch64_sve_float_mul_immediate_p): New functions.
7668 (aarch64_sve_valid_immediate): New function.
7669 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
7670 Explicitly reject structure modes. Check for INDEX constants.
7671 Handle PTRUE and PFALSE constants.
7672 (aarch64_check_zero_based_sve_index_immediate): New function.
7673 (aarch64_simd_imm_zero_p): Delete.
7674 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
7675 vector modes. Accept constants in the range of CNT[BHWD].
7676 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
7677 ask for an Advanced SIMD mode.
7678 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
7679 (aarch64_simd_vector_alignment): Handle SVE predicates.
7680 (aarch64_vectorize_preferred_vector_alignment): New function.
7681 (aarch64_simd_vector_alignment_reachable): Use it instead of
7683 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
7684 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
7686 (MAX_VECT_LEN): Delete.
7687 (expand_vec_perm_d): Add a vec_flags field.
7688 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
7689 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
7690 (aarch64_evpc_ext): Don't apply a big-endian lane correction
7692 (aarch64_evpc_rev): Rename to...
7693 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
7694 (aarch64_evpc_rev_global): New function.
7695 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
7696 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
7698 (aarch64_evpc_sve_tbl): New function.
7699 (aarch64_expand_vec_perm_const_1): Update after rename of
7700 aarch64_evpc_rev. Handle SVE permutes too, trying
7701 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
7702 than aarch64_evpc_tbl.
7703 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
7704 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
7705 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
7706 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
7707 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
7708 (aarch64_expand_sve_vcond): New functions.
7709 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
7710 of aarch64_vector_mode_p.
7711 (aarch64_dwarf_poly_indeterminate_value): New function.
7712 (aarch64_compute_pressure_classes): Likewise.
7713 (aarch64_can_change_mode_class): Likewise.
7714 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
7715 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
7716 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
7717 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
7718 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
7719 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
7720 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
7721 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
7723 (Dn, Dl, Dr): Accept const as well as const_vector.
7724 (Dz): Likewise. Compare against CONST0_RTX.
7725 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
7726 of "vector" where appropriate.
7727 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
7728 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
7729 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
7730 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
7731 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
7732 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
7733 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
7734 (v_int_equiv): Extend to SVE modes.
7735 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
7737 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
7738 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
7739 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
7740 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
7741 (SVE_COND_FP_CMP): New int iterators.
7742 (perm_hilo): Handle the new unpack unspecs.
7743 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
7745 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
7746 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
7747 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
7748 (aarch64_equality_operator, aarch64_constant_vector_operand)
7749 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
7750 (aarch64_sve_nonimmediate_operand): Likewise.
7751 (aarch64_sve_general_operand): Likewise.
7752 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
7753 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
7754 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
7755 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
7756 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
7757 (aarch64_sve_float_arith_immediate): Likewise.
7758 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
7759 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
7760 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
7761 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
7762 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
7763 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
7764 (aarch64_sve_float_arith_operand): Likewise.
7765 (aarch64_sve_float_arith_with_sub_operand): Likewise.
7766 (aarch64_sve_float_mul_operand): Likewise.
7767 (aarch64_sve_vec_perm_operand): Likewise.
7768 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
7769 (aarch64_mov_operand): Accept const_poly_int and const_vector.
7770 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
7771 as well as const_vector.
7772 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
7773 in file. Use CONST0_RTX and CONSTM1_RTX.
7774 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
7775 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
7776 Use aarch64_simd_imm_zero.
7777 * config/aarch64/aarch64-sve.md: New file.
7778 * config/aarch64/aarch64.md: Include it.
7779 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
7780 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
7781 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
7782 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
7783 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
7784 (sve): New attribute.
7785 (enabled): Disable instructions with the sve attribute unless
7787 (movqi, movhi): Pass CONST_POLY_INT operaneds through
7788 aarch64_expand_mov_immediate.
7789 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
7790 CNT[BHSD] immediates.
7791 (movti): Split CONST_POLY_INT moves into two halves.
7792 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
7793 Split additions that need a temporary here if the destination
7794 is the stack pointer.
7795 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
7796 (*add<mode>3_poly_1): New instruction.
7797 (set_clobber_cc): New expander.
7799 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7801 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
7802 parameter and use it instead of GET_MODE_SIZE (innermode). Use
7803 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
7804 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
7805 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
7806 Change innermode from fixed_mode_size to machine_mode.
7807 (simplify_subreg): Update call accordingly. Handle a constant-sized
7808 subreg of a variable-length CONST_VECTOR.
7810 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7811 Alan Hayward <alan.hayward@arm.com>
7812 David Sherwood <david.sherwood@arm.com>
7814 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
7815 (add_offset_to_base): New function, split out from...
7816 (create_mem_ref): ...here. When handling a scale other than 1,
7817 check first whether the address is valid without the offset.
7818 Add it into the base if so, leaving the index and scale as-is.
7820 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7823 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
7824 fold_for_warn before checking if arg2 is INTEGER_CST.
7826 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
7828 * config/rs6000/predicates.md (load_multiple_operation): Delete.
7829 (store_multiple_operation): Delete.
7830 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
7831 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
7832 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
7833 guarded by TARGET_STRING.
7834 (rs6000_output_load_multiple): Delete.
7835 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
7836 OPTION_MASK_STRING / TARGET_STRING handling.
7837 (print_operand) <'N', 'O'>: Add comment that these are unused now.
7838 (const rs6000_opt_masks) <"string">: Change mask to 0.
7839 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
7840 (MASK_STRING): Delete.
7841 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
7843 (load_multiple): Delete.
7850 (store_multiple): Delete.
7857 (movmemsi_8reg): Delete.
7858 (corresponding unnamed define_insn): Delete.
7859 (movmemsi_6reg): Delete.
7860 (corresponding unnamed define_insn): Delete.
7861 (movmemsi_4reg): Delete.
7862 (corresponding unnamed define_insn): Delete.
7863 (movmemsi_2reg): Delete.
7864 (corresponding unnamed define_insn): Delete.
7865 (movmemsi_1reg): Delete.
7866 (corresponding unnamed define_insn): Delete.
7867 * config/rs6000/rs6000.opt (mno-string): New.
7868 (mstring): Replace by deprecation warning stub.
7869 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
7871 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7873 * regrename.c (regrename_do_replace): If replacing the same
7874 reg multiple times, try to reuse last created gen_raw_REG.
7877 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
7878 main to workaround a bug in GDB.
7880 2018-01-12 Tom de Vries <tom@codesourcery.com>
7883 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
7885 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
7887 PR rtl-optimization/80481
7888 * ira-color.c (get_cap_member): New function.
7889 (allocnos_conflict_by_live_ranges_p): Use it.
7890 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
7891 (setup_slot_coalesced_allocno_live_ranges): Ditto.
7893 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
7896 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
7897 (*saddl_se_1): Ditto.
7899 (*ssubl_se_1): Ditto.
7901 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7903 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
7904 rather than wi::to_widest for DR_INITs.
7905 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
7906 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
7907 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
7909 (vect_analyze_group_access_1): Note that here.
7911 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7913 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
7914 polynomial type sizes.
7916 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7918 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
7919 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
7920 (gimple_add_tmp_var): Likewise.
7922 2018-01-12 Martin Liska <mliska@suse.cz>
7924 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
7925 (gimple_alloc_sizes): Likewise.
7926 (dump_gimple_statistics): Use PRIu64 in printf format.
7927 * gimple.h: Change uint64_t to int.
7929 2018-01-12 Martin Liska <mliska@suse.cz>
7931 * tree-core.h: Use uint64_t instead of int.
7932 * tree.c (tree_node_counts): Likewise.
7933 (tree_node_sizes): Likewise.
7934 (dump_tree_statistics): Use PRIu64 in printf format.
7936 2018-01-12 Martin Liska <mliska@suse.cz>
7938 * Makefile.in: As qsort_chk is implemented in vec.c, add
7939 vec.o to linkage of gencfn-macros.
7940 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
7941 passing the info to record_node_allocation_statistics.
7942 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
7944 * ggc-common.c (struct ggc_usage): Add operator== and use
7945 it in operator< and compare function.
7946 * mem-stats.h (struct mem_usage): Likewise.
7947 * vec.c (struct vec_usage): Remove operator< and compare
7948 function. Can be simply inherited.
7950 2018-01-12 Martin Jambor <mjambor@suse.cz>
7953 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
7954 * tree-ssa-math-opts.c: Include domwalk.h.
7955 (convert_mult_to_fma_1): New function.
7956 (fma_transformation_info): New type.
7957 (fma_deferring_state): Likewise.
7958 (cancel_fma_deferring): New function.
7959 (result_of_phi): Likewise.
7960 (last_fma_candidate_feeds_initial_phi): Likewise.
7961 (convert_mult_to_fma): Added deferring logic, split actual
7962 transformation to convert_mult_to_fma_1.
7963 (math_opts_dom_walker): New type.
7964 (math_opts_dom_walker::after_dom_children): New method, body moved
7965 here from pass_optimize_widening_mul::execute, added deferring logic
7967 (pass_optimize_widening_mul::execute): Moved most of code to
7968 math_opts_dom_walker::after_dom_children.
7969 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
7970 * config/i386/i386.c (ix86_option_override_internal): Added
7971 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
7973 2018-01-12 Richard Biener <rguenther@suse.de>
7976 * dwarf2out.c (gen_variable_die): Do not reset old_die for
7977 inline instance vars.
7979 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
7982 * config/rx/rx.c (rx_is_restricted_memory_address):
7985 2018-01-12 Richard Biener <rguenther@suse.de>
7987 PR tree-optimization/80846
7988 * target.def (split_reduction): New target hook.
7989 * targhooks.c (default_split_reduction): New function.
7990 * targhooks.h (default_split_reduction): Declare.
7991 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
7992 target requests first reduce vectors by combining low and high
7994 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
7995 (get_vectype_for_scalar_type_and_size): Export.
7996 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
7997 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
7998 * doc/tm.texi: Regenerate.
7999 * config/i386/i386.c (ix86_split_reduction): Implement
8000 TARGET_VECTORIZE_SPLIT_REDUCTION.
8002 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8005 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8006 in PIC mode except for TARGET_VXWORKS_RTP.
8007 * config/sparc/sparc.c: Include cfgrtl.h.
8008 (TARGET_INIT_PIC_REG): Define.
8009 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8010 (sparc_pic_register_p): New predicate.
8011 (sparc_legitimate_address_p): Use it.
8012 (sparc_legitimize_pic_address): Likewise.
8013 (sparc_delegitimize_address): Likewise.
8014 (sparc_mode_dependent_address_p): Likewise.
8015 (gen_load_pcrel_sym): Remove 4th parameter.
8016 (load_got_register): Adjust call to above. Remove obsolete stuff.
8017 (sparc_expand_prologue): Do not call load_got_register here.
8018 (sparc_flat_expand_prologue): Likewise.
8019 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8020 (sparc_use_pseudo_pic_reg): New function.
8021 (sparc_init_pic_reg): Likewise.
8022 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8023 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8025 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8027 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8028 Add item for branch_cost.
8030 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8032 PR rtl-optimization/83565
8033 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8034 not extend the result to a larger mode for rotate operations.
8035 (num_sign_bit_copies1): Likewise.
8037 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8040 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8042 Use values-Xc.o for -pedantic.
8043 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8045 2018-01-12 Martin Liska <mliska@suse.cz>
8048 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8050 (possible_polymorphic_call_targets): Use it.
8051 (ipa_devirt): Likewise.
8053 2018-01-12 Martin Liska <mliska@suse.cz>
8055 * profile-count.h (enum profile_quality): Use 0 as invalid
8056 enum value of profile_quality.
8058 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8060 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8061 -mext-string options.
8063 2018-01-12 Richard Biener <rguenther@suse.de>
8065 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8066 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8067 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8069 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8071 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8073 * configure.ac (--with-long-double-format): Add support for the
8074 configuration option to change the default long double format on
8076 * config.gcc (powerpc*-linux*-*): Likewise.
8077 * configure: Regenerate.
8078 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8079 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8080 used without modification.
8082 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8084 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8085 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8086 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8087 MISC_BUILTIN_SPEC_BARRIER.
8088 (rs6000_init_builtins): Likewise.
8089 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8091 (speculation_barrier): New define_insn.
8092 * doc/extend.texi: Document __builtin_speculation_barrier.
8094 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8097 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8098 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8099 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8101 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8102 integral modes instead of "ss" and "sd".
8103 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8104 vectors with 32-bit and 64-bit elements.
8105 (vecdupssescalarmodesuffix): New mode attribute.
8106 (vec_dup<mode>): Use it.
8108 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8111 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8112 frame if argument is passed on stack.
8114 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8117 * ree.c (combine_reaching_defs): Optimize also
8118 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8119 reg2=any_extend(exp); reg1=reg2;, formatting fix.
8121 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8124 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8126 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8129 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8130 after they are computed.
8132 2018-01-11 Bin Cheng <bin.cheng@arm.com>
8134 PR tree-optimization/83695
8135 * gimple-loop-linterchange.cc
8136 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8137 reset cached scev information after interchange.
8138 (pass_linterchange::execute): Remove call to scev_reset_htab.
8140 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8142 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8143 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8144 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8145 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8146 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8147 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8148 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8149 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8150 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8151 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8152 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8153 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8154 (V_lane_reg): Likewise.
8155 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8157 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8158 (vfmal_lane_low<mode>_intrinsic,
8159 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8160 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8161 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8162 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8163 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8164 vfmsl_lane_high<mode>_intrinsic): New define_insns.
8166 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8168 * config/arm/arm-cpus.in (fp16fml): New feature.
8169 (ALL_SIMD): Add fp16fml.
8170 (armv8.2-a): Add fp16fml as an option.
8171 (armv8.3-a): Likewise.
8172 (armv8.4-a): Add fp16fml as part of fp16.
8173 * config/arm/arm.h (TARGET_FP16FML): Define.
8174 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
8176 * config/arm/arm-modes.def (V2HF): Define.
8177 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
8178 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
8179 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
8180 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
8181 vfmsl_low, vfmsl_high): New set of builtins.
8182 * config/arm/iterators.md (PLUSMINUS): New code iterator.
8183 (vfml_op): New code attribute.
8184 (VFMLHALVES): New int iterator.
8185 (VFML, VFMLSEL): New mode attributes.
8186 (V_reg): Define mapping for V2HF.
8187 (V_hi, V_lo): New mode attributes.
8188 (VF_constraint): Likewise.
8189 (vfml_half, vfml_half_selector): New int attributes.
8190 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
8192 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
8193 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
8195 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
8196 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
8197 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
8198 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
8200 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8201 Document new effective target and option set.
8203 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8205 * config/arm/arm-cpus.in (armv8_4): New feature.
8206 (ARMv8_4a): New fgroup.
8207 (armv8.4-a): New arch.
8208 * config/arm/arm-tables.opt: Regenerate.
8209 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8210 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8211 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8212 Add matching rules for -march=armv8.4-a and extensions.
8213 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8215 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
8218 * config/rx/rx.md (BW): New mode attribute.
8219 (sync_lock_test_and_setsi): Add mode suffix to insn output.
8221 2018-01-11 Richard Biener <rguenther@suse.de>
8223 PR tree-optimization/83435
8224 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
8225 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
8226 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
8228 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8229 Alan Hayward <alan.hayward@arm.com>
8230 David Sherwood <david.sherwood@arm.com>
8232 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8234 (aarch64_classify_address): Initialize it. Track polynomial offsets.
8235 (aarch64_print_address_internal): Use it to check for a zero offset.
8237 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8238 Alan Hayward <alan.hayward@arm.com>
8239 David Sherwood <david.sherwood@arm.com>
8241 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8242 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8243 Return a poly_int64 rather than a HOST_WIDE_INT.
8244 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8245 rather than a HOST_WIDE_INT.
8246 * config/aarch64/aarch64.h (aarch64_frame): Protect with
8247 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
8248 hard_fp_offset, frame_size, initial_adjust, callee_offset and
8249 final_offset from HOST_WIDE_INT to poly_int64.
8250 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8251 to_constant when getting the number of units in an Advanced SIMD
8253 (aarch64_builtin_vectorized_function): Check for a constant number
8255 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8257 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8258 attribute instead of GET_MODE_NUNITS.
8259 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8260 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8261 GET_MODE_SIZE for fixed-size registers.
8262 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8263 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8264 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8265 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8266 (aarch64_print_operand, aarch64_print_address_internal)
8267 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8268 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8269 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8270 Handle polynomial GET_MODE_SIZE.
8271 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8272 wider than SImode without modification.
8273 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8274 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8275 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8276 passing and returning SVE modes.
8277 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8278 rather than GEN_INT.
8279 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8280 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8281 (aarch64_allocate_and_probe_stack_space): Likewise.
8282 (aarch64_layout_frame): Cope with polynomial offsets.
8283 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8284 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8286 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8287 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8288 poly_int64 rather than a HOST_WIDE_INT.
8289 (aarch64_get_separate_components, aarch64_process_components)
8290 (aarch64_expand_prologue, aarch64_expand_epilogue)
8291 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8292 (aarch64_anchor_offset): New function, split out from...
8293 (aarch64_legitimize_address): ...here.
8294 (aarch64_builtin_vectorization_cost): Handle polynomial
8295 TYPE_VECTOR_SUBPARTS.
8296 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8298 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8299 number of elements from the PARALLEL rather than the mode.
8300 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8301 rather than GET_MODE_BITSIZE.
8302 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8303 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8304 (aarch64_expand_vec_perm_const_1): Handle polynomial
8305 d->perm.length () and d->perm elements.
8306 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
8307 Apply to_constant to d->perm elements.
8308 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8309 polynomial CONST_VECTOR_NUNITS.
8310 (aarch64_move_pointer): Take amount as a poly_int64 rather
8312 (aarch64_progress_pointer): Avoid temporary variable.
8313 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8314 the mode attribute instead of GET_MODE.
8316 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8317 Alan Hayward <alan.hayward@arm.com>
8318 David Sherwood <david.sherwood@arm.com>
8320 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8321 x exists before using it.
8322 (aarch64_add_constant_internal): Rename to...
8323 (aarch64_add_offset_1): ...this. Replace regnum with separate
8324 src and dest rtxes. Handle the case in which they're different,
8325 including when the offset is zero. Replace scratchreg with an rtx.
8326 Use 2 additions if there is no spare register into which we can
8327 move a 16-bit constant.
8328 (aarch64_add_constant): Delete.
8329 (aarch64_add_offset): Replace reg with separate src and dest
8330 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8331 Use aarch64_add_offset_1.
8332 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8333 an rtx rather than an int. Take the delta as a poly_int64
8334 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8335 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8336 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8337 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8338 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8340 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8341 aarch64_add_constant.
8343 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8345 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8346 Use scalar_float_mode.
8348 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8350 * config/aarch64/aarch64-simd.md
8351 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8352 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8353 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8354 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8355 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8356 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8357 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8358 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8359 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8360 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8362 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8365 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8366 targ_options->x_arm_arch_string is non NULL.
8368 2018-01-11 Tamar Christina <tamar.christina@arm.com>
8370 * config/aarch64/aarch64.h
8371 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
8373 2018-01-11 Sudakshina Das <sudi.das@arm.com>
8376 * expmed.c (emit_store_flag_force): Swap if const op0
8377 and change VOIDmode to mode of op0.
8379 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8381 PR rtl-optimization/83761
8382 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
8383 than bytes to mode_for_size.
8385 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8388 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
8389 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
8392 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8395 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
8396 when in layout mode.
8397 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
8398 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
8401 2018-01-10 Michael Collison <michael.collison@arm.com>
8403 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
8404 * config/aarch64/aarch64-option-extension.def: Add
8405 AARCH64_OPT_EXTENSION of 'fp16fml'.
8406 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8407 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
8408 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
8409 * config/aarch64/constraints.md (Ui7): New constraint.
8410 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
8411 (VFMLA_SEL_W): Ditto.
8414 (VFMLA16_LOW): New int iterator.
8415 (VFMLA16_HIGH): Ditto.
8416 (UNSPEC_FMLAL): New unspec.
8417 (UNSPEC_FMLSL): Ditto.
8418 (UNSPEC_FMLAL2): Ditto.
8419 (UNSPEC_FMLSL2): Ditto.
8420 (f16mac): New code attribute.
8421 * config/aarch64/aarch64-simd-builtins.def
8422 (aarch64_fmlal_lowv2sf): Ditto.
8423 (aarch64_fmlsl_lowv2sf): Ditto.
8424 (aarch64_fmlalq_lowv4sf): Ditto.
8425 (aarch64_fmlslq_lowv4sf): Ditto.
8426 (aarch64_fmlal_highv2sf): Ditto.
8427 (aarch64_fmlsl_highv2sf): Ditto.
8428 (aarch64_fmlalq_highv4sf): Ditto.
8429 (aarch64_fmlslq_highv4sf): Ditto.
8430 (aarch64_fmlal_lane_lowv2sf): Ditto.
8431 (aarch64_fmlsl_lane_lowv2sf): Ditto.
8432 (aarch64_fmlal_laneq_lowv2sf): Ditto.
8433 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
8434 (aarch64_fmlalq_lane_lowv4sf): Ditto.
8435 (aarch64_fmlsl_lane_lowv4sf): Ditto.
8436 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
8437 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
8438 (aarch64_fmlal_lane_highv2sf): Ditto.
8439 (aarch64_fmlsl_lane_highv2sf): Ditto.
8440 (aarch64_fmlal_laneq_highv2sf): Ditto.
8441 (aarch64_fmlsl_laneq_highv2sf): Ditto.
8442 (aarch64_fmlalq_lane_highv4sf): Ditto.
8443 (aarch64_fmlsl_lane_highv4sf): Ditto.
8444 (aarch64_fmlalq_laneq_highv4sf): Ditto.
8445 (aarch64_fmlsl_laneq_highv4sf): Ditto.
8446 * config/aarch64/aarch64-simd.md:
8447 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
8448 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8449 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
8450 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8451 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
8452 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
8453 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
8454 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
8455 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
8456 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
8457 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
8458 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
8459 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
8460 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
8461 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
8462 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
8463 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
8464 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
8465 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
8466 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
8467 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
8468 (vfmlsl_low_u32): Ditto.
8469 (vfmlalq_low_u32): Ditto.
8470 (vfmlslq_low_u32): Ditto.
8471 (vfmlal_high_u32): Ditto.
8472 (vfmlsl_high_u32): Ditto.
8473 (vfmlalq_high_u32): Ditto.
8474 (vfmlslq_high_u32): Ditto.
8475 (vfmlal_lane_low_u32): Ditto.
8476 (vfmlsl_lane_low_u32): Ditto.
8477 (vfmlal_laneq_low_u32): Ditto.
8478 (vfmlsl_laneq_low_u32): Ditto.
8479 (vfmlalq_lane_low_u32): Ditto.
8480 (vfmlslq_lane_low_u32): Ditto.
8481 (vfmlalq_laneq_low_u32): Ditto.
8482 (vfmlslq_laneq_low_u32): Ditto.
8483 (vfmlal_lane_high_u32): Ditto.
8484 (vfmlsl_lane_high_u32): Ditto.
8485 (vfmlal_laneq_high_u32): Ditto.
8486 (vfmlsl_laneq_high_u32): Ditto.
8487 (vfmlalq_lane_high_u32): Ditto.
8488 (vfmlslq_lane_high_u32): Ditto.
8489 (vfmlalq_laneq_high_u32): Ditto.
8490 (vfmlslq_laneq_high_u32): Ditto.
8491 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
8492 (AARCH64_FL_FOR_ARCH8_4): New.
8493 (AARCH64_ISA_F16FML): New ISA flag.
8494 (TARGET_F16FML): New feature flag for fp16fml.
8495 (doc/invoke.texi): Document new fp16fml option.
8497 2018-01-10 Michael Collison <michael.collison@arm.com>
8499 * config/aarch64/aarch64-builtins.c:
8500 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
8501 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8502 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
8503 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
8504 (AARCH64_ISA_SHA3): New ISA flag.
8505 (TARGET_SHA3): New feature flag for sha3.
8506 * config/aarch64/iterators.md (sha512_op): New int attribute.
8507 (CRYPTO_SHA512): New int iterator.
8508 (UNSPEC_SHA512H): New unspec.
8509 (UNSPEC_SHA512H2): Ditto.
8510 (UNSPEC_SHA512SU0): Ditto.
8511 (UNSPEC_SHA512SU1): Ditto.
8512 * config/aarch64/aarch64-simd-builtins.def
8513 (aarch64_crypto_sha512hqv2di): New builtin.
8514 (aarch64_crypto_sha512h2qv2di): Ditto.
8515 (aarch64_crypto_sha512su0qv2di): Ditto.
8516 (aarch64_crypto_sha512su1qv2di): Ditto.
8517 (aarch64_eor3qv8hi): Ditto.
8518 (aarch64_rax1qv2di): Ditto.
8519 (aarch64_xarqv2di): Ditto.
8520 (aarch64_bcaxqv8hi): Ditto.
8521 * config/aarch64/aarch64-simd.md:
8522 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
8523 (aarch64_crypto_sha512su0qv2di): Ditto.
8524 (aarch64_crypto_sha512su1qv2di): Ditto.
8525 (aarch64_eor3qv8hi): Ditto.
8526 (aarch64_rax1qv2di): Ditto.
8527 (aarch64_xarqv2di): Ditto.
8528 (aarch64_bcaxqv8hi): Ditto.
8529 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
8530 (vsha512h2q_u64): Ditto.
8531 (vsha512su0q_u64): Ditto.
8532 (vsha512su1q_u64): Ditto.
8533 (veor3q_u16): Ditto.
8534 (vrax1q_u64): Ditto.
8536 (vbcaxq_u16): Ditto.
8537 * config/arm/types.md (crypto_sha512): New type attribute.
8538 (crypto_sha3): Ditto.
8539 (doc/invoke.texi): Document new sha3 option.
8541 2018-01-10 Michael Collison <michael.collison@arm.com>
8543 * config/aarch64/aarch64-builtins.c:
8544 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
8545 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8546 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
8547 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
8548 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
8549 (AARCH64_ISA_SM4): New ISA flag.
8550 (TARGET_SM4): New feature flag for sm4.
8551 * config/aarch64/aarch64-simd-builtins.def
8552 (aarch64_sm3ss1qv4si): Ditto.
8553 (aarch64_sm3tt1aq4si): Ditto.
8554 (aarch64_sm3tt1bq4si): Ditto.
8555 (aarch64_sm3tt2aq4si): Ditto.
8556 (aarch64_sm3tt2bq4si): Ditto.
8557 (aarch64_sm3partw1qv4si): Ditto.
8558 (aarch64_sm3partw2qv4si): Ditto.
8559 (aarch64_sm4eqv4si): Ditto.
8560 (aarch64_sm4ekeyqv4si): Ditto.
8561 * config/aarch64/aarch64-simd.md:
8562 (aarch64_sm3ss1qv4si): Ditto.
8563 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
8564 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
8565 (aarch64_sm4eqv4si): Ditto.
8566 (aarch64_sm4ekeyqv4si): Ditto.
8567 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
8568 (sm3part_op): Ditto.
8569 (CRYPTO_SM3TT): Ditto.
8570 (CRYPTO_SM3PART): Ditto.
8571 (UNSPEC_SM3SS1): New unspec.
8572 (UNSPEC_SM3TT1A): Ditto.
8573 (UNSPEC_SM3TT1B): Ditto.
8574 (UNSPEC_SM3TT2A): Ditto.
8575 (UNSPEC_SM3TT2B): Ditto.
8576 (UNSPEC_SM3PARTW1): Ditto.
8577 (UNSPEC_SM3PARTW2): Ditto.
8578 (UNSPEC_SM4E): Ditto.
8579 (UNSPEC_SM4EKEY): Ditto.
8580 * config/aarch64/constraints.md (Ui2): New constraint.
8581 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
8582 * config/arm/types.md (crypto_sm3): New type attribute.
8583 (crypto_sm4): Ditto.
8584 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
8585 (vsm3tt1aq_u32): Ditto.
8586 (vsm3tt1bq_u32): Ditto.
8587 (vsm3tt2aq_u32): Ditto.
8588 (vsm3tt2bq_u32): Ditto.
8589 (vsm3partw1q_u32): Ditto.
8590 (vsm3partw2q_u32): Ditto.
8591 (vsm4eq_u32): Ditto.
8592 (vsm4ekeyq_u32): Ditto.
8593 (doc/invoke.texi): Document new sm4 option.
8595 2018-01-10 Michael Collison <michael.collison@arm.com>
8597 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
8598 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
8599 (AARCH64_FL_FOR_ARCH8_4): New.
8600 (AARCH64_FL_V8_4): New flag.
8601 (doc/invoke.texi): Document new armv8.4-a option.
8603 2018-01-10 Michael Collison <michael.collison@arm.com>
8605 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8606 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
8607 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
8608 * config/aarch64/aarch64-option-extension.def: Add
8609 AARCH64_OPT_EXTENSION of 'sha2'.
8610 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
8611 (crypto): Disable sha2 and aes if crypto disabled.
8612 (crypto): Enable aes and sha2 if enabled.
8613 (simd): Disable sha2 and aes if simd disabled.
8614 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
8616 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
8617 (TARGET_SHA2): New feature flag for sha2.
8618 (TARGET_AES): New feature flag for aes.
8619 * config/aarch64/aarch64-simd.md:
8620 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
8621 conditional on TARGET_AES.
8622 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
8623 (aarch64_crypto_sha1hsi): Make pattern conditional
8625 (aarch64_crypto_sha1hv4si): Ditto.
8626 (aarch64_be_crypto_sha1hv4si): Ditto.
8627 (aarch64_crypto_sha1su1v4si): Ditto.
8628 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
8629 (aarch64_crypto_sha1su0v4si): Ditto.
8630 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
8631 (aarch64_crypto_sha256su0v4si): Ditto.
8632 (aarch64_crypto_sha256su1v4si): Ditto.
8633 (doc/invoke.texi): Document new aes and sha2 options.
8635 2018-01-10 Martin Sebor <msebor@redhat.com>
8637 PR tree-optimization/83781
8638 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
8641 2018-01-11 Martin Sebor <msebor@gmail.com>
8642 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8644 PR tree-optimization/83501
8645 PR tree-optimization/81703
8647 * tree-ssa-strlen.c (get_string_cst): Rename...
8648 (get_string_len): ...to this. Handle global constants.
8649 (handle_char_store): Adjust.
8651 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
8652 Jim Wilson <jimw@sifive.com>
8654 * config/riscv/riscv-protos.h (riscv_output_return): New.
8655 * config/riscv/riscv.c (struct machine_function): New naked_p field.
8656 (riscv_attribute_table, riscv_output_return),
8657 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
8658 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
8659 (riscv_compute_frame_info): Only compute frame->mask if not a naked
8661 (riscv_expand_prologue): Add early return for naked function.
8662 (riscv_expand_epilogue): Likewise.
8663 (riscv_function_ok_for_sibcall): Return false for naked function.
8664 (riscv_set_current_function): New.
8665 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
8666 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
8667 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
8668 * doc/extend.texi (RISC-V Function Attributes): New.
8670 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8672 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
8673 check for 128-bit long double before checking TCmode.
8674 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
8675 128-bit long doubles before checking TFmode or TCmode.
8676 (FLOAT128_IBM_P): Likewise.
8678 2018-01-10 Martin Sebor <msebor@redhat.com>
8680 PR tree-optimization/83671
8681 * builtins.c (c_strlen): Unconditionally return zero for the empty
8683 Use -Warray-bounds for warnings.
8684 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
8685 for non-constant array indices with COMPONENT_REF, arrays of
8686 arrays, and pointers to arrays.
8687 (gimple_fold_builtin_strlen): Determine and set length range for
8688 non-constant character arrays.
8690 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
8693 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
8696 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
8698 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
8700 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8703 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
8704 VECTOR_MEM_ALTIVEC_OR_VSX_P.
8705 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
8706 indexed_or_indirect_operand predicate.
8707 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
8708 (*vsx_le_perm_load_v8hi): Likewise.
8709 (*vsx_le_perm_load_v16qi): Likewise.
8710 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
8711 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
8712 (*vsx_le_perm_store_v8hi): Likewise.
8713 (*vsx_le_perm_store_v16qi): Likewise.
8714 (eight unnamed splitters): Likewise.
8716 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8718 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
8719 * config/rs6000/emmintrin.h: Likewise.
8720 * config/rs6000/mmintrin.h: Likewise.
8721 * config/rs6000/xmmintrin.h: Likewise.
8723 2018-01-10 David Malcolm <dmalcolm@redhat.com>
8726 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
8728 * tree.c (tree_nop_conversion): Return true for location wrapper
8730 (maybe_wrap_with_location): New function.
8731 (selftest::check_strip_nops): New function.
8732 (selftest::test_location_wrappers): New function.
8733 (selftest::tree_c_tests): Call it.
8734 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
8735 (maybe_wrap_with_location): New decl.
8736 (EXPR_LOCATION_WRAPPER_P): New macro.
8737 (location_wrapper_p): New inline function.
8738 (tree_strip_any_location_wrapper): New inline function.
8740 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
8743 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
8744 stack_realign_offset for the largest alignment of stack slot
8746 (ix86_find_max_used_stack_alignment): New function.
8747 (ix86_finalize_stack_frame_flags): Use it. Set
8748 max_used_stack_alignment if we don't realign stack.
8749 * config/i386/i386.h (machine_function): Add
8750 max_used_stack_alignment.
8752 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
8754 * config/arm/arm.opt (-mbranch-cost): New option.
8755 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
8758 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
8761 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
8762 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
8764 2018-01-10 Richard Biener <rguenther@suse.de>
8767 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
8768 early out so it also covers the case where we have a non-NULL
8771 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8773 PR tree-optimization/83753
8774 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
8775 for non-strided grouped accesses if the number of elements is 1.
8777 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8780 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
8781 * i386.h (TARGET_USE_GATHER): Define.
8782 * x86-tune.def (X86_TUNE_USE_GATHER): New.
8784 2018-01-10 Martin Liska <mliska@suse.cz>
8787 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
8788 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
8790 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
8791 CLEANUP_NO_PARTITIONING is not set.
8793 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8795 * doc/rtl.texi: Remove documentation of (const ...) wrappers
8796 for vectors, as a partial revert of r254296.
8797 * rtl.h (const_vec_p): Delete.
8798 (const_vec_duplicate_p): Don't test for vector CONSTs.
8799 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
8800 * expmed.c (make_tree): Likewise.
8803 * common.md (E, F): Use CONSTANT_P instead of checking for
8805 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
8806 checking for CONST_VECTOR.
8808 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8811 * predict.c (force_edge_cold): Handle in more sane way edges
8814 2018-01-09 Carl Love <cel@us.ibm.com>
8816 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
8818 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
8819 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
8820 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
8821 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
8822 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
8823 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
8824 * config/rs6000/rs6000-protos.h: Add extern defition for
8825 rs6000_generate_float2_double_code.
8826 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
8828 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
8829 (float2_v2df): Add define_expand.
8831 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
8834 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
8835 op_mode in the force_to_mode call.
8837 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8839 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
8840 instead of checking each element individually.
8841 (aarch64_evpc_uzp): Likewise.
8842 (aarch64_evpc_zip): Likewise.
8843 (aarch64_evpc_ext): Likewise.
8844 (aarch64_evpc_rev): Likewise.
8845 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
8846 instead of checking each element individually. Return true without
8848 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
8849 whether all selected elements come from the same input, instead of
8850 checking each element individually. Remove calls to gen_rtx_REG,
8851 start_sequence and end_sequence and instead assert that no rtl is
8854 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8856 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
8857 order of HIGH and CONST checks.
8859 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8861 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
8862 if the destination isn't an SSA_NAME.
8864 2018-01-09 Richard Biener <rguenther@suse.de>
8866 PR tree-optimization/83668
8867 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
8869 (canonicalize_loop_form): ... here, renamed from ...
8870 (canonicalize_loop_closed_ssa_form): ... this and amended to
8871 swap successor edges for loop exit blocks to make us use
8872 the RPO order we need for initial schedule generation.
8874 2018-01-09 Joseph Myers <joseph@codesourcery.com>
8876 PR tree-optimization/64811
8877 * match.pd: When optimizing comparisons with Inf, avoid
8878 introducing or losing exceptions from comparisons with NaN.
8880 2018-01-09 Martin Liska <mliska@suse.cz>
8883 * asan.c (shadow_mem_size): Add gcc_assert.
8885 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
8887 Don't save registers in main().
8890 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
8891 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
8892 * config/avr/avr.c (avr_set_current_function): Don't error if
8893 naked, OS_task or OS_main are specified at the same time.
8894 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
8896 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
8898 * common/config/avr/avr-common.c (avr_option_optimization_table):
8899 Switch on -mmain-is-OS_task for optimizing compilations.
8901 2018-01-09 Richard Biener <rguenther@suse.de>
8903 PR tree-optimization/83572
8904 * graphite.c: Include cfganal.h.
8905 (graphite_transform_loops): Connect infinite loops to exit
8906 and remove fake edges at the end.
8908 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8910 * ipa-inline.c (edge_badness): Revert accidental checkin.
8912 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8915 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
8916 symbols; not inline clones.
8918 2018-01-09 Jakub Jelinek <jakub@redhat.com>
8921 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
8922 hard registers. Formatting fixes.
8924 PR preprocessor/83722
8925 * gcc.c (try_generate_repro): Pass
8926 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
8927 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
8930 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
8931 Kito Cheng <kito.cheng@gmail.com>
8933 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
8934 (riscv_leaf_function_p): Delete.
8935 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
8937 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8939 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
8941 (do_ifelse): New function.
8942 (do_isel): New function.
8943 (do_sub3): New function.
8944 (do_add3): New function.
8945 (do_load_mask_compare): New function.
8946 (do_overlap_load_compare): New function.
8947 (expand_compare_loop): New function.
8948 (expand_block_compare): Call expand_compare_loop() when appropriate.
8949 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
8951 (-mblock-compare-inline-loop-limit): New option.
8953 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8956 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
8957 Reverse order of second and third operands in first alternative.
8958 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
8959 of first and second elements in UNSPEC_VPERMR vector.
8960 (altivec_expand_vec_perm_le): Likewise.
8962 2018-01-08 Jeff Law <law@redhat.com>
8964 PR rtl-optimizatin/81308
8965 * tree-switch-conversion.c (cfg_altered): New file scoped static.
8966 (process_switch): If group_case_labels makes a change, then set
8968 (pass_convert_switch::execute): If a switch is converted, then
8969 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
8971 PR rtl-optimization/81308
8972 * recog.c (split_all_insns): Conditionally cleanup the CFG after
8975 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
8977 PR target/83663 - Revert r255946
8978 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
8979 generation for cases where splatting a value is not useful.
8980 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8981 across a vec_duplicate and a paradoxical subreg forming a vector
8982 mode to a vec_concat.
8984 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8986 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
8987 -march=armv8.3-a variants.
8988 * config/arm/t-multilib: Likewise.
8989 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
8991 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8993 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
8995 (cceq_ior_compare_complement): Give it a name so I can use it, and
8996 change boolean_or_operator predicate to boolean_operator so it can
8997 be used to generate a crand.
8998 (eqne): New code iterator.
8999 (bd/bd_neg): New code_attrs.
9000 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9001 a single define_insn.
9002 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9003 decrement (bdnzt/bdnzf/bdzt/bdzf).
9004 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9005 with the new names of the branch decrement patterns, and added the
9006 names of the branch decrement conditional patterns.
9008 2018-01-08 Richard Biener <rguenther@suse.de>
9010 PR tree-optimization/83563
9011 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9014 2018-01-08 Richard Biener <rguenther@suse.de>
9017 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9019 2018-01-08 Richard Biener <rguenther@suse.de>
9021 PR tree-optimization/83685
9022 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9023 references to abnormals.
9025 2018-01-08 Richard Biener <rguenther@suse.de>
9028 * dwarf2out.c (output_indirect_strings): Handle empty
9029 skeleton_debug_str_hash.
9030 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9032 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9034 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9035 (emit_store_direct): Likewise.
9036 (arc_trampoline_adjust_address): Likewise.
9037 (arc_asm_trampoline_template): New function.
9038 (arc_initialize_trampoline): Use asm_trampoline_template.
9039 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9040 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9041 * config/arc/arc.md (flush_icache): Delete pattern.
9043 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9045 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9046 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9049 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9052 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9053 by not USED_FOR_TARGET.
9054 (make_pass_resolve_sw_modes): Likewise.
9056 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9058 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9061 2018-01-08 Richard Biener <rguenther@suse.de>
9064 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9066 2018-01-08 Richard Biener <rguenther@suse.de>
9069 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9071 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9074 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9075 basic blocks with a small number of successors.
9076 (convert_control_dep_chain_into_preds): Improve handling of
9078 (dump_predicates): Split apart into...
9079 (dump_pred_chain): ...here...
9080 (dump_pred_info): ...and here.
9081 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9082 (can_chain_union_be_invalidated_p): Improve check for invalidation
9084 (uninit_uses_cannot_happen): Avoid unnecessary if
9085 convert_control_dep_chain_into_preds yielded nothing.
9087 2018-01-06 Martin Sebor <msebor@redhat.com>
9089 PR tree-optimization/83640
9090 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9091 subtracting negative offset from size.
9092 (builtin_access::overlap): Adjust offset bounds of the access to fall
9093 within the size of the object if possible.
9095 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9097 PR rtl-optimization/83699
9098 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9099 extract_bit_field_as_subreg to cases in which the extracted
9100 value is also a vector.
9102 * lra-constraints.c (process_alt_operands): Test for the equivalence
9103 substitutions when detecting a possible reload cycle.
9105 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9108 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9109 by default if flag_selective_schedling{,2}. Formatting fixes.
9111 PR rtl-optimization/83682
9112 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9113 if it has non-VECTOR_MODE element mode.
9114 (vec_duplicate_p): Likewise.
9117 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9118 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9120 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9123 * config/i386/i386-builtin.def
9124 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9125 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9126 Require also OPTION_MASK_ISA_AVX512F in addition to
9127 OPTION_MASK_ISA_GFNI.
9128 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9129 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9130 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9131 to OPTION_MASK_ISA_GFNI.
9132 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9133 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9134 OPTION_MASK_ISA_AVX512BW.
9135 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9136 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9137 addition to OPTION_MASK_ISA_GFNI.
9138 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9139 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9140 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9141 to OPTION_MASK_ISA_GFNI.
9142 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9143 a requirement for all ISAs rather than any of them with a few
9145 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9147 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9148 bitmasks to be enabled with 3 exceptions, instead of requiring any
9149 enabled ISA with lots of exceptions.
9150 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9151 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9152 Change avx512bw in isa attribute to avx512f.
9153 * config/i386/sgxintrin.h: Add license boilerplate.
9154 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9155 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9156 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9157 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9159 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9160 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9161 temporarily sse2 rather than sse if not enabled already.
9164 * config/i386/sse.md (VI248_VLBW): Rename to ...
9165 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
9166 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9167 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9168 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9169 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9170 mode iterator instead of VI248_VLBW.
9172 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
9174 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
9175 (record_modified): Skip clobbers; add debug output.
9176 (param_change_prob): Use sreal frequencies.
9178 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9180 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
9181 punt for user-aligned variables.
9183 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9185 * tree-chrec.c (chrec_contains_symbols): Return true for
9188 2018-01-05 Sudakshina Das <sudi.das@arm.com>
9191 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
9192 of (x|y) == x for BICS pattern.
9194 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9196 PR tree-optimization/83605
9197 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
9198 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9201 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
9203 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9204 * config/epiphany/rtems.h: New file.
9206 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9207 Uros Bizjak <ubizjak@gmail.com>
9210 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9211 QIreg_operand instead of register_operand predicate.
9212 * config/i386/i386.c (ix86_rop_should_change_byte_p,
9213 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9214 comments instead of -fmitigate[-_]rop.
9216 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9219 * cgraphunit.c (symbol_table::compile): Switch to text_section
9220 before calling assembly_start debug hook.
9221 * run-rtl-passes.c (run_rtl_passes): Likewise.
9224 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9226 * tree-vrp.c (extract_range_from_binary_expr_1): Check
9227 range_int_cst_p rather than !symbolic_range_p before calling
9228 extract_range_from_multiplicative_op_1.
9230 2018-01-04 Jeff Law <law@redhat.com>
9232 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9233 redundant test in assertion.
9235 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9237 * doc/rtl.texi: Document machine_mode wrapper classes.
9239 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9241 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9244 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9246 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9247 the VEC_PERM_EXPR fold to fail.
9249 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9252 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9253 to switched_sections.
9255 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9258 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9261 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9264 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9265 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9267 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9270 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9271 is BLKmode and bitpos not zero or mode change is needed.
9273 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9276 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9279 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9282 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9283 instead of MULT rtx. Update all corresponding splitters.
9285 (*ssub<modesuffix>): Ditto.
9287 (*cmp_sadd_di): Update split patterns.
9288 (*cmp_sadd_si): Ditto.
9289 (*cmp_sadd_sidi): Ditto.
9290 (*cmp_ssub_di): Ditto.
9291 (*cmp_ssub_si): Ditto.
9292 (*cmp_ssub_sidi): Ditto.
9293 * config/alpha/predicates.md (const23_operand): New predicate.
9294 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9295 Look for ASHIFT, not MULT inner operand.
9296 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9298 2018-01-04 Martin Liska <mliska@suse.cz>
9300 PR gcov-profile/83669
9301 * gcov.c (output_intermediate_file): Add version to intermediate
9303 * doc/gcov.texi: Document new field 'version' in intermediate
9304 file format. Fix location of '-k' option of gcov command.
9306 2018-01-04 Martin Liska <mliska@suse.cz>
9309 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9311 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9313 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9315 2018-01-03 Martin Sebor <msebor@redhat.com>
9317 PR tree-optimization/83655
9318 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9319 checking calls with invalid arguments.
9321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9323 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9324 (vectorizable_mask_load_store): Delete.
9325 (vectorizable_call): Return false for masked loads and stores.
9326 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9327 instead of gimple_assign_rhs1.
9328 (vectorizable_load): Handle IFN_MASK_LOAD.
9329 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9331 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9333 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9335 (vectorizable_mask_load_store): ...here.
9336 (vectorizable_load): ...and here.
9338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9340 * tree-vect-stmts.c (vect_build_all_ones_mask)
9341 (vect_build_zero_merge_argument): New functions, split out from...
9342 (vectorizable_load): ...here.
9344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9346 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9348 (vectorizable_mask_load_store): ...here.
9349 (vectorizable_store): ...and here.
9351 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9353 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9355 (vectorizable_mask_load_store): ...here.
9357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9359 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9360 (vect_model_store_cost): Take a vec_load_store_type instead of a
9362 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9363 (vect_model_store_cost): Take a vec_load_store_type instead of a
9365 (vectorizable_mask_load_store): Update accordingly.
9366 (vectorizable_store): Likewise.
9367 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9369 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9371 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9372 IFN_MASK_LOAD calls here rather than...
9373 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9376 Alan Hayward <alan.hayward@arm.com>
9377 David Sherwood <david.sherwood@arm.com>
9379 * expmed.c (extract_bit_field_1): For vector extracts,
9380 fall back to extract_bit_field_as_subreg if vec_extract
9383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9384 Alan Hayward <alan.hayward@arm.com>
9385 David Sherwood <david.sherwood@arm.com>
9387 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
9388 they are variable or constant sized.
9389 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
9390 slots for constant-sized data.
9392 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9393 Alan Hayward <alan.hayward@arm.com>
9394 David Sherwood <david.sherwood@arm.com>
9396 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
9397 handling COND_EXPRs with boolean comparisons, try to find a better
9398 basis for the mask type than the boolean itself.
9400 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9402 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
9403 is calculated and how it can be overridden.
9404 * genmodes.c (max_bitsize_mode_any_mode): New variable.
9405 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
9407 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
9410 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9411 Alan Hayward <alan.hayward@arm.com>
9412 David Sherwood <david.sherwood@arm.com>
9414 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
9415 Remove the mode argument.
9416 (aarch64_simd_valid_immediate): Remove the mode and inverse
9418 * config/aarch64/iterators.md (bitsize): New iterator.
9419 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
9420 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
9421 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
9422 aarch64_simd_valid_immediate.
9423 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
9424 (aarch64_reg_or_bic_imm): Likewise.
9425 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
9426 with an insn_type enum and msl with a modifier_type enum.
9427 Replace element_width with a scalar_mode. Change the shift
9428 to unsigned int. Add constructors for scalar_float_mode and
9429 scalar_int_mode elements.
9430 (aarch64_vect_float_const_representable_p): Delete.
9431 (aarch64_can_const_movi_rtx_p)
9432 (aarch64_simd_scalar_immediate_valid_for_move)
9433 (aarch64_simd_make_constant): Update call to
9434 aarch64_simd_valid_immediate.
9435 (aarch64_advsimd_valid_immediate_hs): New function.
9436 (aarch64_advsimd_valid_immediate): Likewise.
9437 (aarch64_simd_valid_immediate): Remove mode and inverse
9438 arguments. Rewrite to use the above. Use const_vec_duplicate_p
9439 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
9440 and aarch64_float_const_representable_p on the result.
9441 (aarch64_output_simd_mov_immediate): Remove mode argument.
9442 Update call to aarch64_simd_valid_immediate and use of
9443 simd_immediate_info.
9444 (aarch64_output_scalar_simd_mov_immediate): Update call
9447 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9448 Alan Hayward <alan.hayward@arm.com>
9449 David Sherwood <david.sherwood@arm.com>
9451 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
9452 (mode_nunits): Likewise CONST_MODE_NUNITS.
9453 * machmode.def (ADJUST_NUNITS): Document.
9454 * genmodes.c (mode_data::need_nunits_adj): New field.
9455 (blank_mode): Update accordingly.
9456 (adj_nunits): New variable.
9457 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
9459 (emit_mode_size_inline): Set need_bytesize_adj for all modes
9460 listed in adj_nunits.
9461 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
9462 listed in adj_nunits. Don't emit case statements for such modes.
9463 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
9464 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
9465 nothing if adj_nunits is nonnull.
9466 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
9467 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
9468 (emit_mode_fbit): Update use of print_maybe_const_decl.
9469 (emit_move_size): Likewise. Treat the array as non-const
9471 (emit_mode_adjustments): Handle adj_nunits.
9473 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9475 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
9476 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
9477 (VECTOR_MODES): Use it.
9478 (make_vector_modes): Take the prefix as an argument.
9480 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9481 Alan Hayward <alan.hayward@arm.com>
9482 David Sherwood <david.sherwood@arm.com>
9484 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
9485 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
9486 for MODE_VECTOR_BOOL.
9487 * machmode.def (VECTOR_BOOL_MODE): Document.
9488 * genmodes.c (VECTOR_BOOL_MODE): New macro.
9489 (make_vector_bool_mode): New function.
9490 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
9492 * lto-streamer-in.c (lto_input_mode_table): Likewise.
9493 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
9495 * stor-layout.c (int_mode_for_mode): Likewise.
9496 * tree.c (build_vector_type_for_mode): Likewise.
9497 * varasm.c (output_constant_pool_2): Likewise.
9498 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
9499 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
9500 for MODE_VECTOR_BOOL.
9501 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
9502 of mode class checks.
9503 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
9504 instead of a list of mode class checks.
9505 (expand_vector_scalar_condition): Likewise.
9506 (type_for_widest_vector_mode): Handle BImode as an inner mode.
9508 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9509 Alan Hayward <alan.hayward@arm.com>
9510 David Sherwood <david.sherwood@arm.com>
9512 * machmode.h (mode_size): Change from unsigned short to
9514 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
9515 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9516 or if measurement_type is not polynomial.
9517 (fixed_size_mode::includes_p): Check for constant-sized modes.
9518 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
9519 return a poly_uint16 rather than an unsigned short.
9520 (emit_mode_size): Change the type of mode_size from unsigned short
9521 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
9522 (emit_mode_adjustments): Cope with polynomial vector sizes.
9523 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9525 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9527 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
9528 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
9529 * caller-save.c (setup_save_areas): Likewise.
9530 (replace_reg_with_saved_mem): Likewise.
9531 * calls.c (emit_library_call_value_1): Likewise.
9532 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
9533 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
9534 (gen_lowpart_for_combine): Likewise.
9535 * convert.c (convert_to_integer_1): Likewise.
9536 * cse.c (equiv_constant, cse_insn): Likewise.
9537 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
9538 (cselib_subst_to_values): Likewise.
9539 * dce.c (word_dce_process_block): Likewise.
9540 * df-problems.c (df_word_lr_mark_ref): Likewise.
9541 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
9542 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
9543 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
9544 (rtl_for_decl_location): Likewise.
9545 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
9546 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
9547 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
9548 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
9549 (expand_expr_real_1): Likewise.
9550 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
9551 (pad_below): Likewise.
9552 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9553 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
9554 * ira.c (get_subreg_tracking_sizes): Likewise.
9555 * ira-build.c (ira_create_allocno_objects): Likewise.
9556 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
9557 (ira_sort_regnos_for_alter_reg): Likewise.
9558 * ira-costs.c (record_operand_costs): Likewise.
9559 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
9560 (resolve_simple_move): Likewise.
9561 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
9562 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
9563 (lra_constraints): Likewise.
9564 (CONST_POOL_OK_P): Reject variable-sized modes.
9565 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
9566 (add_pseudo_to_slot, lra_spill): Likewise.
9567 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9568 * optabs-query.c (get_best_extraction_insn): Likewise.
9569 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9570 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
9571 (expand_mult_highpart, valid_multiword_target_p): Likewise.
9572 * recog.c (offsettable_address_addr_space_p): Likewise.
9573 * regcprop.c (maybe_mode_change): Likewise.
9574 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
9575 * regrename.c (build_def_use): Likewise.
9576 * regstat.c (dump_reg_info): Likewise.
9577 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
9578 (find_reloads, find_reloads_subreg_address): Likewise.
9579 * reload1.c (eliminate_regs_1): Likewise.
9580 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
9581 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
9582 (simplify_binary_operation_1, simplify_subreg): Likewise.
9583 * targhooks.c (default_function_arg_padding): Likewise.
9584 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
9585 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
9586 (verify_gimple_assign_ternary): Likewise.
9587 * tree-inline.c (estimate_move_cost): Likewise.
9588 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9589 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
9590 (get_address_cost_ainc): Likewise.
9591 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9592 (vect_supportable_dr_alignment): Likewise.
9593 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9594 (vectorizable_reduction): Likewise.
9595 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
9596 (vectorizable_operation, vectorizable_load): Likewise.
9597 * tree.c (build_same_sized_truth_vector_type): Likewise.
9598 * valtrack.c (cleanup_auto_inc_dec): Likewise.
9599 * var-tracking.c (emit_note_insn_var_location): Likewise.
9600 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
9601 (ADDR_VEC_ALIGN): Likewise.
9603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9604 Alan Hayward <alan.hayward@arm.com>
9605 David Sherwood <david.sherwood@arm.com>
9607 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
9609 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9610 or if measurement_type is polynomial.
9611 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
9612 * combine.c (make_extraction): Likewise.
9613 * dse.c (find_shift_sequence): Likewise.
9614 * dwarf2out.c (mem_loc_descriptor): Likewise.
9615 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
9616 (extract_bit_field, extract_low_bits): Likewise.
9617 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
9618 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
9619 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
9620 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
9621 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9622 * reload.c (find_reloads): Likewise.
9623 * reload1.c (alter_reg): Likewise.
9624 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
9625 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
9626 * tree-if-conv.c (predicate_mem_writes): Likewise.
9627 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
9628 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
9629 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9630 * valtrack.c (dead_debug_insert_temp): Likewise.
9631 * varasm.c (mergeable_constant_section): Likewise.
9632 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
9634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9635 Alan Hayward <alan.hayward@arm.com>
9636 David Sherwood <david.sherwood@arm.com>
9638 * expr.c (expand_assignment): Cope with polynomial mode sizes
9639 when assigning to a CONCAT.
9641 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9642 Alan Hayward <alan.hayward@arm.com>
9643 David Sherwood <david.sherwood@arm.com>
9645 * machmode.h (mode_precision): Change from unsigned short to
9647 (mode_to_precision): Return a poly_uint16 rather than an unsigned
9649 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
9650 or if measurement_type is not polynomial.
9651 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
9652 in which the mode is already known to be a scalar_int_mode.
9653 * genmodes.c (emit_mode_precision): Change the type of mode_precision
9654 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
9656 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9657 for GET_MODE_PRECISION.
9658 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9659 for GET_MODE_PRECISION.
9660 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
9662 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
9663 (expand_field_assignment, make_extraction): Likewise.
9664 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
9665 (get_last_value): Likewise.
9666 * convert.c (convert_to_integer_1): Likewise.
9667 * cse.c (cse_insn): Likewise.
9668 * expr.c (expand_expr_real_1): Likewise.
9669 * lra-constraints.c (simplify_operand_subreg): Likewise.
9670 * optabs-query.c (can_atomic_load_p): Likewise.
9671 * optabs.c (expand_atomic_load): Likewise.
9672 (expand_atomic_store): Likewise.
9673 * ree.c (combine_reaching_defs): Likewise.
9674 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
9675 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
9676 * tree.h (type_has_mode_precision_p): Likewise.
9677 * ubsan.c (instrument_si_overflow): Likewise.
9679 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9680 Alan Hayward <alan.hayward@arm.com>
9681 David Sherwood <david.sherwood@arm.com>
9683 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
9684 polynomial numbers of units.
9685 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
9686 (valid_vector_subparts_p): New function.
9687 (build_vector_type): Remove temporary shim and take the number
9688 of units as a poly_uint64 rather than an int.
9689 (build_opaque_vector_type): Take the number of units as a
9690 poly_uint64 rather than an int.
9691 * tree.c (build_vector_from_ctor): Handle polynomial
9692 TYPE_VECTOR_SUBPARTS.
9693 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
9694 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
9695 (build_vector_from_val): If the number of units is variable,
9696 use build_vec_duplicate_cst for constant operands and
9697 VEC_DUPLICATE_EXPR otherwise.
9698 (make_vector_type): Remove temporary is_constant ().
9699 (build_vector_type, build_opaque_vector_type): Take the number of
9700 units as a poly_uint64 rather than an int.
9701 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
9703 * cfgexpand.c (expand_debug_expr): Likewise.
9704 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
9705 (store_constructor, expand_expr_real_1): Likewise.
9706 (const_scalar_mask_from_tree): Likewise.
9707 * fold-const-call.c (fold_const_reduction): Likewise.
9708 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
9709 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
9710 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
9711 (fold_relational_const): Likewise.
9712 (native_interpret_vector): Likewise. Change the size from an
9713 int to an unsigned int.
9714 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
9715 TYPE_VECTOR_SUBPARTS.
9716 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
9717 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
9718 duplicating a non-constant operand into a variable-length vector.
9719 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
9720 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
9721 * ipa-icf.c (sem_variable::equals): Likewise.
9722 * match.pd: Likewise.
9723 * omp-simd-clone.c (simd_clone_subparts): Likewise.
9724 * print-tree.c (print_node): Likewise.
9725 * stor-layout.c (layout_type): Likewise.
9726 * targhooks.c (default_builtin_vectorization_cost): Likewise.
9727 * tree-cfg.c (verify_gimple_comparison): Likewise.
9728 (verify_gimple_assign_binary): Likewise.
9729 (verify_gimple_assign_ternary): Likewise.
9730 (verify_gimple_assign_single): Likewise.
9731 * tree-pretty-print.c (dump_generic_node): Likewise.
9732 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9733 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
9734 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
9735 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
9736 (vect_shift_permute_load_chain): Likewise.
9737 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
9738 (expand_vector_condition, optimize_vector_constructor): Likewise.
9739 (lower_vec_perm, get_compute_type): Likewise.
9740 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9741 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
9742 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
9743 (vect_recog_mask_conversion_pattern): Likewise.
9744 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
9745 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
9746 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9747 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
9748 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
9749 (vectorizable_shift, vectorizable_operation, vectorizable_store)
9750 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
9751 (supportable_widening_operation): Likewise.
9752 (supportable_narrowing_operation): Likewise.
9753 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
9755 * varasm.c (output_constant): Likewise.
9757 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9758 Alan Hayward <alan.hayward@arm.com>
9759 David Sherwood <david.sherwood@arm.com>
9761 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
9762 so that both the length == 3 and length != 3 cases set up their
9763 own permute vectors. Add comments explaining why we know the
9764 number of elements is constant.
9765 (vect_permute_load_chain): Likewise.
9767 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9768 Alan Hayward <alan.hayward@arm.com>
9769 David Sherwood <david.sherwood@arm.com>
9771 * machmode.h (mode_nunits): Change from unsigned char to
9773 (ONLY_FIXED_SIZE_MODES): New macro.
9774 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
9775 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
9776 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
9778 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
9779 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
9780 or if measurement_type is not polynomial.
9781 * genmodes.c (ZERO_COEFFS): New macro.
9782 (emit_mode_nunits_inline): Make mode_nunits_inline return a
9784 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
9785 Use ZERO_COEFFS when emitting initializers.
9786 * data-streamer.h (bp_pack_poly_value): New function.
9787 (bp_unpack_poly_value): Likewise.
9788 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9789 for GET_MODE_NUNITS.
9790 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9791 for GET_MODE_NUNITS.
9792 * tree.c (make_vector_type): Remove temporary shim and make
9793 the real function take the number of units as a poly_uint64
9795 (build_vector_type_for_mode): Handle polynomial nunits.
9796 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
9797 * emit-rtl.c (const_vec_series_p_1): Likewise.
9798 (gen_rtx_CONST_VECTOR): Likewise.
9799 * fold-const.c (test_vec_duplicate_folding): Likewise.
9800 * genrecog.c (validate_pattern): Likewise.
9801 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
9802 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9803 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
9804 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
9805 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
9806 * rtlanal.c (subreg_get_info): Likewise.
9807 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9808 (vect_grouped_load_supported): Likewise.
9809 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
9810 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
9811 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
9812 (simplify_const_unary_operation, simplify_binary_operation_1)
9813 (simplify_const_binary_operation, simplify_ternary_operation)
9814 (test_vector_ops_duplicate, test_vector_ops): Likewise.
9815 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
9816 instead of CONST_VECTOR_NUNITS.
9817 * varasm.c (output_constant_pool_2): Likewise.
9818 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
9819 explicit-encoded elements in the XVEC for variable-length vectors.
9821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9823 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
9825 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9826 Alan Hayward <alan.hayward@arm.com>
9827 David Sherwood <david.sherwood@arm.com>
9829 * coretypes.h (fixed_size_mode): Declare.
9830 (fixed_size_mode_pod): New typedef.
9831 * builtins.h (target_builtins::x_apply_args_mode)
9832 (target_builtins::x_apply_result_mode): Change type to
9833 fixed_size_mode_pod.
9834 * builtins.c (apply_args_size, apply_result_size, result_vector)
9835 (expand_builtin_apply_args_1, expand_builtin_apply)
9836 (expand_builtin_return): Update accordingly.
9838 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9840 * cse.c (hash_rtx_cb): Hash only the encoded elements.
9841 * cselib.c (cselib_hash_rtx): Likewise.
9842 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
9843 CONST_VECTOR encoding.
9845 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9846 Jeff Law <law@redhat.com>
9849 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
9850 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
9851 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
9852 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
9855 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
9856 explicitly probe *sp in a noreturn function if there were any callee
9857 register saves or frame pointer is needed.
9859 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9862 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
9863 BLKmode for ternary, binary or unary expressions.
9866 * var-tracking.c (delete_vta_debug_insn): New inline function.
9867 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
9868 insns from get_insns () to NULL instead of each bb separately.
9869 Use delete_vta_debug_insn. No longer static.
9870 (vt_debug_insns_local, variable_tracking_main_1): Adjust
9871 delete_vta_debug_insns callers.
9872 * rtl.h (delete_vta_debug_insns): Declare.
9873 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
9874 instead of variable_tracking_main.
9876 2018-01-03 Martin Sebor <msebor@redhat.com>
9878 PR tree-optimization/83603
9879 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
9880 arguments past the endof the argument list in functions declared
9881 without a prototype.
9882 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
9883 Avoid checking when arguments are null.
9885 2018-01-03 Martin Sebor <msebor@redhat.com>
9888 * doc/extend.texi (attribute const): Fix a typo.
9889 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
9890 issuing -Wsuggest-attribute for void functions.
9892 2018-01-03 Martin Sebor <msebor@redhat.com>
9894 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
9895 offset_int::from instead of wide_int::to_shwi.
9896 (maybe_diag_overlap): Remove assertion.
9897 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
9898 * gimple-ssa-sprintf.c (format_directive): Same.
9899 (parse_directive): Same.
9900 (sprintf_dom_walker::compute_format_length): Same.
9901 (try_substitute_return_value): Same.
9903 2018-01-03 Jeff Law <law@redhat.com>
9906 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
9907 non-constant residual for zero at runtime and avoid probing in
9908 that case. Reorganize code for trailing problem to mirror handling
9911 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9913 PR tree-optimization/83501
9914 * tree-ssa-strlen.c (get_string_cst): New.
9915 (handle_char_store): Call get_string_cst.
9917 2018-01-03 Martin Liska <mliska@suse.cz>
9919 PR tree-optimization/83593
9920 * tree-ssa-strlen.c: Include tree-cfg.h.
9921 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
9922 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
9923 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
9925 (strlen_dom_walker::before_dom_children): Call
9926 gimple_purge_dead_eh_edges. Dump tranformation with details
9928 (strlen_dom_walker::before_dom_children): Update call by adding
9929 new argument cleanup_eh.
9930 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
9932 2018-01-03 Martin Liska <mliska@suse.cz>
9935 * cif-code.def (VARIADIC_THUNK): New enum value.
9936 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
9939 2018-01-03 Jan Beulich <jbeulich@suse.com>
9941 * sse.md (mov<mode>_internal): Tighten condition for when to use
9942 vmovdqu<ssescalarsize> for TI and OI modes.
9944 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9946 Update copyright years.
9948 2018-01-03 Martin Liska <mliska@suse.cz>
9951 * ipa-visibility.c (function_and_variable_visibility): Skip
9952 functions with noipa attribure.
9954 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9956 * gcc.c (process_command): Update copyright notice dates.
9957 * gcov-dump.c (print_version): Ditto.
9958 * gcov.c (print_version): Ditto.
9959 * gcov-tool.c (print_version): Ditto.
9960 * gengtype.c (create_file): Ditto.
9961 * doc/cpp.texi: Bump @copying's copyright year.
9962 * doc/cppinternals.texi: Ditto.
9963 * doc/gcc.texi: Ditto.
9964 * doc/gccint.texi: Ditto.
9965 * doc/gcov.texi: Ditto.
9966 * doc/install.texi: Ditto.
9967 * doc/invoke.texi: Ditto.
9969 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9971 * vector-builder.h (vector_builder::m_full_nelts): Change from
9972 unsigned int to poly_uint64.
9973 (vector_builder::full_nelts): Update prototype accordingly.
9974 (vector_builder::new_vector): Likewise.
9975 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
9976 (vector_builder::operator ==): Likewise.
9977 (vector_builder::finalize): Likewise.
9978 * int-vector-builder.h (int_vector_builder::int_vector_builder):
9979 Take the number of elements as a poly_uint64 rather than an
9981 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
9982 from unsigned int to poly_uint64.
9983 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
9984 (vec_perm_indices::new_vector): Likewise.
9985 (vec_perm_indices::length): Likewise.
9986 (vec_perm_indices::nelts_per_input): Likewise.
9987 (vec_perm_indices::input_nelts): Likewise.
9988 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
9989 number of elements per input as a poly_uint64 rather than an
9990 unsigned int. Use the original encoding for variable-length
9991 vectors, rather than clamping each individual element.
9992 For the second and subsequent elements in each pattern,
9993 clamp the step and base before clamping their sum.
9994 (vec_perm_indices::series_p): Handle polynomial element counts.
9995 (vec_perm_indices::all_in_range_p): Likewise.
9996 (vec_perm_indices_to_tree): Likewise.
9997 (vec_perm_indices_to_rtx): Likewise.
9998 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
9999 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10000 (tree_vector_builder::new_binary_operation): Handle polynomial
10001 element counts. Return false if we need to know the number
10002 of elements at compile time.
10003 * fold-const.c (fold_vec_perm): Punt if the number of elements
10004 isn't known at compile time.
10006 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10008 * vec-perm-indices.h (vec_perm_builder): Change element type
10009 from HOST_WIDE_INT to poly_int64.
10010 (vec_perm_indices::element_type): Update accordingly.
10011 (vec_perm_indices::clamp): Handle polynomial element_types.
10012 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10013 (vec_perm_indices::all_in_range_p): Likewise.
10014 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10016 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10017 polynomial vec_perm_indices element types.
10018 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10019 * fold-const.c (fold_vec_perm): Likewise.
10020 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10021 * tree-vect-generic.c (lower_vec_perm): Likewise.
10022 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10023 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10024 element type to HOST_WIDE_INT.
10026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10027 Alan Hayward <alan.hayward@arm.com>
10028 David Sherwood <david.sherwood@arm.com>
10030 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10031 rather than an int. Use plus_constant.
10032 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10033 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10035 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10036 Alan Hayward <alan.hayward@arm.com>
10037 David Sherwood <david.sherwood@arm.com>
10039 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10040 a HOST_WIDE_INT to a poly_int64.
10042 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10043 Alan Hayward <alan.hayward@arm.com>
10044 David Sherwood <david.sherwood@arm.com>
10046 * calls.c (load_register_parameters): Cope with polynomial
10047 mode sizes. Require a constant size for BLKmode parameters
10048 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10049 forces a parameter to be padded at the lsb end in order to
10050 fill a complete number of words, require the parameter size
10051 to be ordered wrt UNITS_PER_WORD.
10053 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10054 Alan Hayward <alan.hayward@arm.com>
10055 David Sherwood <david.sherwood@arm.com>
10057 * reload1.c (spill_stack_slot_width): Change element type
10058 from unsigned int to poly_uint64_pod.
10059 (alter_reg): Treat mode sizes as polynomial.
10061 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10062 Alan Hayward <alan.hayward@arm.com>
10063 David Sherwood <david.sherwood@arm.com>
10065 * reload.c (complex_word_subreg_p): New function.
10066 (reload_inner_reg_of_subreg, push_reload): Use it.
10068 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10069 Alan Hayward <alan.hayward@arm.com>
10070 David Sherwood <david.sherwood@arm.com>
10072 * lra-constraints.c (process_alt_operands): Reject matched
10073 operands whose sizes aren't ordered.
10074 (match_reload): Refer to this check here.
10076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10077 Alan Hayward <alan.hayward@arm.com>
10078 David Sherwood <david.sherwood@arm.com>
10080 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10081 that the mode size is in the set {1, 2, 4, 8, 16}.
10083 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10084 Alan Hayward <alan.hayward@arm.com>
10085 David Sherwood <david.sherwood@arm.com>
10087 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10088 Use plus_constant instead of gen_rtx_PLUS.
10090 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10091 Alan Hayward <alan.hayward@arm.com>
10092 David Sherwood <david.sherwood@arm.com>
10094 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10095 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10096 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10097 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10098 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10099 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10100 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10101 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10102 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10103 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10105 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10106 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10107 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10108 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10109 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10110 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10111 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10112 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10113 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10114 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10116 * expr.c (emit_move_resolve_push): Treat the input and result
10117 of PUSH_ROUNDING as a poly_int64.
10118 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10119 (emit_push_insn): Likewise.
10120 * lra-eliminations.c (mark_not_eliminable): Likewise.
10121 * recog.c (push_operand): Likewise.
10122 * reload1.c (elimination_effects): Likewise.
10123 * rtlanal.c (nonzero_bits1): Likewise.
10124 * calls.c (store_one_arg): Likewise. Require the padding to be
10125 known at compile time.
10127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10128 Alan Hayward <alan.hayward@arm.com>
10129 David Sherwood <david.sherwood@arm.com>
10131 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10132 Use plus_constant instead of gen_rtx_PLUS.
10134 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10135 Alan Hayward <alan.hayward@arm.com>
10136 David Sherwood <david.sherwood@arm.com>
10138 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10139 rather than an int.
10141 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10142 Alan Hayward <alan.hayward@arm.com>
10143 David Sherwood <david.sherwood@arm.com>
10145 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10146 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10147 via stack temporaries. Treat the mode size as polynomial too.
10149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10150 Alan Hayward <alan.hayward@arm.com>
10151 David Sherwood <david.sherwood@arm.com>
10153 * expr.c (expand_expr_real_2): When handling conversions involving
10154 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10155 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10156 as a poly_uint64 too.
10158 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10159 Alan Hayward <alan.hayward@arm.com>
10160 David Sherwood <david.sherwood@arm.com>
10162 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
10164 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10165 Alan Hayward <alan.hayward@arm.com>
10166 David Sherwood <david.sherwood@arm.com>
10168 * combine.c (can_change_dest_mode): Handle polynomial
10169 REGMODE_NATURAL_SIZE.
10170 * expmed.c (store_bit_field_1): Likewise.
10171 * expr.c (store_constructor): Likewise.
10172 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10173 and polynomial REGMODE_NATURAL_SIZE.
10174 (gen_lowpart_common): Likewise.
10175 * reginfo.c (record_subregs_of_mode): Likewise.
10176 * rtlanal.c (read_modify_subreg_p): Likewise.
10178 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10179 Alan Hayward <alan.hayward@arm.com>
10180 David Sherwood <david.sherwood@arm.com>
10182 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
10183 numbers of elements.
10185 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10186 Alan Hayward <alan.hayward@arm.com>
10187 David Sherwood <david.sherwood@arm.com>
10189 * match.pd: Cope with polynomial numbers of vector elements.
10191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10192 Alan Hayward <alan.hayward@arm.com>
10193 David Sherwood <david.sherwood@arm.com>
10195 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
10196 in a POINTER_PLUS_EXPR.
10198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10199 Alan Hayward <alan.hayward@arm.com>
10200 David Sherwood <david.sherwood@arm.com>
10202 * omp-simd-clone.c (simd_clone_subparts): New function.
10203 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10204 (ipa_simd_modify_function_body): Likewise.
10206 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10207 Alan Hayward <alan.hayward@arm.com>
10208 David Sherwood <david.sherwood@arm.com>
10210 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
10211 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10212 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10213 (expand_vector_condition, vector_element): Likewise.
10214 (subparts_gt): New function.
10215 (get_compute_type): Use subparts_gt.
10216 (count_type_subparts): Delete.
10217 (expand_vector_operations_1): Use subparts_gt instead of
10218 count_type_subparts.
10220 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10221 Alan Hayward <alan.hayward@arm.com>
10222 David Sherwood <david.sherwood@arm.com>
10224 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
10225 (vect_compile_time_alias): ...this new function. Do the calculation
10226 on poly_ints rather than trees.
10227 (vect_prune_runtime_alias_test_list): Update call accordingly.
10229 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10230 Alan Hayward <alan.hayward@arm.com>
10231 David Sherwood <david.sherwood@arm.com>
10233 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
10235 (vect_schedule_slp_instance): Likewise.
10237 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10238 Alan Hayward <alan.hayward@arm.com>
10239 David Sherwood <david.sherwood@arm.com>
10241 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
10242 constant and extern definitions for variable-length vectors.
10243 (vect_get_constant_vectors): Note that the number of units
10244 is known to be constant.
10246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10247 Alan Hayward <alan.hayward@arm.com>
10248 David Sherwood <david.sherwood@arm.com>
10250 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
10251 of units as polynomial. Choose between WIDE and NARROW based
10254 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10255 Alan Hayward <alan.hayward@arm.com>
10256 David Sherwood <david.sherwood@arm.com>
10258 * tree-vect-stmts.c (simd_clone_subparts): New function.
10259 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10261 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10262 Alan Hayward <alan.hayward@arm.com>
10263 David Sherwood <david.sherwood@arm.com>
10265 * tree-vect-stmts.c (vectorizable_call): Treat the number of
10266 vectors as polynomial. Use build_index_vector for
10267 IFN_GOMP_SIMD_LANE.
10269 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10270 Alan Hayward <alan.hayward@arm.com>
10271 David Sherwood <david.sherwood@arm.com>
10273 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10274 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10275 for variable-length vectors.
10276 (vectorizable_mask_load_store): Treat the number of units as
10277 polynomial, asserting that it is constant if the condition has
10278 already been enforced.
10279 (vectorizable_store, vectorizable_load): Likewise.
10281 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10282 Alan Hayward <alan.hayward@arm.com>
10283 David Sherwood <david.sherwood@arm.com>
10285 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10286 of units as polynomial. Punt if we can't tell at compile time
10287 which vector contains the final result.
10289 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10290 Alan Hayward <alan.hayward@arm.com>
10291 David Sherwood <david.sherwood@arm.com>
10293 * tree-vect-loop.c (vectorizable_induction): Treat the number
10294 of units as polynomial. Punt on SLP inductions. Use an integer
10295 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10296 cast of such a series for variable-length floating-point
10299 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10300 Alan Hayward <alan.hayward@arm.com>
10301 David Sherwood <david.sherwood@arm.com>
10303 * tree.h (build_index_vector): Declare.
10304 * tree.c (build_index_vector): New function.
10305 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10306 of units as polynomial, forcibly converting it to a constant if
10307 vectorizable_reduction has already enforced the condition.
10308 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
10309 to create a {1,2,3,...} vector.
10310 (vectorizable_reduction): Treat the number of units as polynomial.
10311 Choose vectype_in based on the largest scalar element size rather
10312 than the smallest number of units. Enforce the restrictions
10315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10316 Alan Hayward <alan.hayward@arm.com>
10317 David Sherwood <david.sherwood@arm.com>
10319 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10320 number of units as polynomial.
10322 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10323 Alan Hayward <alan.hayward@arm.com>
10324 David Sherwood <david.sherwood@arm.com>
10326 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10327 * target.def (autovectorize_vector_sizes): Return the vector sizes
10328 by pointer, using vector_sizes rather than a bitmask.
10329 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10330 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10331 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10333 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10334 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10335 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10336 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10337 * omp-general.c (omp_max_vf): Likewise.
10338 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10339 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10340 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10341 * tree-vect-slp.c (vect_slp_bb): Likewise.
10342 * doc/tm.texi: Regenerate.
10343 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10345 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10346 the vector size as a poly_uint64 rather than an unsigned int.
10347 (current_vector_size): Change from an unsigned int to a poly_uint64.
10348 (get_vectype_for_scalar_type): Update accordingly.
10349 * tree.h (build_truth_vector_type): Take the size and number of
10350 units as a poly_uint64 rather than an unsigned int.
10351 (build_vector_type): Add a temporary overload that takes
10352 the number of units as a poly_uint64 rather than an unsigned int.
10353 * tree.c (make_vector_type): Likewise.
10354 (build_truth_vector_type): Take the number of units as a poly_uint64
10355 rather than an unsigned int.
10357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10358 Alan Hayward <alan.hayward@arm.com>
10359 David Sherwood <david.sherwood@arm.com>
10361 * target.def (get_mask_mode): Take the number of units and length
10362 as poly_uint64s rather than unsigned ints.
10363 * targhooks.h (default_get_mask_mode): Update accordingly.
10364 * targhooks.c (default_get_mask_mode): Likewise.
10365 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10366 * doc/tm.texi: Regenerate.
10368 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10369 Alan Hayward <alan.hayward@arm.com>
10370 David Sherwood <david.sherwood@arm.com>
10372 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10373 * omp-general.c (omp_max_vf): Likewise.
10374 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10375 (expand_omp_simd): Handle polynomial safelen.
10376 * omp-low.c (omplow_simd_context): Add a default constructor.
10377 (omplow_simd_context::max_vf): Change from int to poly_uint64.
10378 (lower_rec_simd_input_clauses): Update accordingly.
10379 (lower_rec_input_clauses): Likewise.
10381 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10382 Alan Hayward <alan.hayward@arm.com>
10383 David Sherwood <david.sherwood@arm.com>
10385 * tree-vectorizer.h (vect_nunits_for_cost): New function.
10386 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
10387 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
10388 (vect_analyze_slp_cost): Likewise.
10389 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
10390 (vect_model_load_cost): Likewise.
10392 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10393 Alan Hayward <alan.hayward@arm.com>
10394 David Sherwood <david.sherwood@arm.com>
10396 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
10397 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
10398 from an unsigned int * to a poly_uint64_pod *.
10399 (calculate_unrolling_factor): New function.
10400 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
10402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10403 Alan Hayward <alan.hayward@arm.com>
10404 David Sherwood <david.sherwood@arm.com>
10406 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
10407 from an unsigned int to a poly_uint64.
10408 (_loop_vec_info::slp_unrolling_factor): Likewise.
10409 (_loop_vec_info::vectorization_factor): Change from an int
10411 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
10412 (vect_get_num_vectors): New function.
10413 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
10414 (vect_get_num_copies): Use vect_get_num_vectors.
10415 (vect_analyze_data_ref_dependences): Change max_vf from an int *
10416 to an unsigned int *.
10417 (vect_analyze_data_refs): Change min_vf from an int * to a
10419 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10420 than an unsigned HOST_WIDE_INT.
10421 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10422 (vect_analyze_data_ref_dependence): Change max_vf from an int *
10423 to an unsigned int *.
10424 (vect_analyze_data_ref_dependences): Likewise.
10425 (vect_compute_data_ref_alignment): Handle polynomial vf.
10426 (vect_enhance_data_refs_alignment): Likewise.
10427 (vect_prune_runtime_alias_test_list): Likewise.
10428 (vect_shift_permute_load_chain): Likewise.
10429 (vect_supportable_dr_alignment): Likewise.
10430 (dependence_distance_ge_vf): Take the vectorization factor as a
10431 poly_uint64 rather than an unsigned HOST_WIDE_INT.
10432 (vect_analyze_data_refs): Change min_vf from an int * to a
10434 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
10435 vfm1 as a poly_uint64 rather than an int. Make the same change
10436 for the returned bound_scalar.
10437 (vect_gen_vector_loop_niters): Handle polynomial vf.
10438 (vect_do_peeling): Likewise. Update call to
10439 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
10440 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
10442 * tree-vect-loop.c (vect_determine_vectorization_factor)
10443 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
10444 (vect_get_known_peeling_cost): Likewise.
10445 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
10446 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
10447 (vect_transform_loop): Likewise. Use the lowest possible VF when
10448 updating the upper bounds of the loop.
10449 (vect_min_worthwhile_factor): Make static. Return an unsigned int
10450 rather than an int.
10451 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
10452 polynomial unroll factors.
10453 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
10454 (vect_make_slp_decision): Likewise.
10455 (vect_supported_load_permutation_p): Likewise, and polynomial
10457 (vect_analyze_slp_cost): Handle polynomial vf.
10458 (vect_slp_analyze_node_operations): Likewise.
10459 (vect_slp_analyze_bb_1): Likewise.
10460 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10461 than an unsigned HOST_WIDE_INT.
10462 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
10463 (vectorizable_load): Handle polynomial vf.
10464 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
10466 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
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 * match.pd: Handle bit operations involving three constants
10473 and try to fold one pair.
10475 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10477 * tree-vect-loop-manip.c: Include gimple-fold.h.
10478 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
10479 niters_maybe_zero parameters. Handle other cases besides a step of 1.
10480 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
10481 Add a path that uses a step of VF instead of 1, but disable it
10483 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
10484 and niters_no_overflow parameters. Update calls to
10485 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
10486 Create a new SSA name if the latter choses to use a ste other
10487 than zero, and return it via niters_vector_mult_vf_var.
10488 * tree-vect-loop.c (vect_transform_loop): Update calls to
10489 vect_do_peeling, vect_gen_vector_loop_niters and
10490 slpeel_make_loop_iterate_ntimes.
10491 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
10492 (vect_gen_vector_loop_niters): Update declarations after above changes.
10494 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
10496 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
10497 128-bit round to integer instructions.
10498 (ceil<mode>2): Likewise.
10499 (btrunc<mode>2): Likewise.
10500 (round<mode>2): Likewise.
10502 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10504 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
10505 unaligned VSX load/store on P8/P9.
10506 (expand_block_clear): Allow the use of unaligned VSX
10507 load/store on P8/P9.
10509 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10511 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
10513 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
10514 swap associated with both a load and a store.
10516 2018-01-02 Andrew Waterman <andrew@sifive.com>
10518 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
10519 * config/riscv/riscv.md (clear_cache): Use it.
10521 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
10523 * web.c: Remove out-of-date comment.
10525 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10527 * expr.c (fixup_args_size_notes): Check that any existing
10528 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
10529 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
10530 (emit_single_push_insn): ...here.
10532 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10534 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
10535 (const_vector_encoded_nelts): New function.
10536 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
10537 (const_vector_int_elt, const_vector_elt): Declare.
10538 * emit-rtl.c (const_vector_int_elt_1): New function.
10539 (const_vector_elt): Likewise.
10540 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
10541 of CONST_VECTOR_ELT.
10543 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10545 * expr.c: Include rtx-vector-builder.h.
10546 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
10547 directly on the tree encoding.
10548 (const_vector_from_tree): Likewise.
10549 * optabs.c: Include rtx-vector-builder.h.
10550 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
10551 sequence of "u" values.
10552 * vec-perm-indices.c: Include rtx-vector-builder.h.
10553 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
10554 directly on the vec_perm_indices encoding.
10556 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10558 * doc/rtl.texi (const_vector): Describe new encoding scheme.
10559 * Makefile.in (OBJS): Add rtx-vector-builder.o.
10560 * rtx-vector-builder.h: New file.
10561 * rtx-vector-builder.c: Likewise.
10562 * rtl.h (rtx_def::u2): Add a const_vector field.
10563 (CONST_VECTOR_NPATTERNS): New macro.
10564 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
10565 (CONST_VECTOR_DUPLICATE_P): Likewise.
10566 (CONST_VECTOR_STEPPED_P): Likewise.
10567 (CONST_VECTOR_ENCODED_ELT): Likewise.
10568 (const_vec_duplicate_p): Check for a duplicated vector encoding.
10569 (unwrap_const_vec_duplicate): Likewise.
10570 (const_vec_series_p): Check for a non-duplicated vector encoding.
10571 Say that the function only returns true for integer vectors.
10572 * emit-rtl.c: Include rtx-vector-builder.h.
10573 (gen_const_vec_duplicate_1): Delete.
10574 (gen_const_vector): Call gen_const_vec_duplicate instead of
10575 gen_const_vec_duplicate_1.
10576 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
10577 (gen_const_vec_duplicate): Use rtx_vector_builder.
10578 (gen_const_vec_series): Likewise.
10579 (gen_rtx_CONST_VECTOR): Likewise.
10580 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
10581 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10582 Build a new vector rather than modifying a CONST_VECTOR in-place.
10583 (handle_special_swappables): Update call accordingly.
10584 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
10585 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10586 Build a new vector rather than modifying a CONST_VECTOR in-place.
10587 (handle_special_swappables): Update call accordingly.
10589 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10591 * simplify-rtx.c (simplify_const_binary_operation): Use
10592 CONST_VECTOR_ELT instead of XVECEXP.
10594 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10596 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
10597 the selector elements to be different from the data elements
10598 if the selector is a VECTOR_CST.
10599 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
10600 ssizetype for the selector.
10602 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10604 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
10605 before testing each element individually.
10606 * tree-vect-generic.c (lower_vec_perm): Likewise.
10608 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10610 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
10611 * selftest-run-tests.c (selftest::run_tests): Call it.
10612 * vector-builder.h (vector_builder::operator ==): New function.
10613 (vector_builder::operator !=): Likewise.
10614 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
10615 (vec_perm_indices::all_from_input_p): New function.
10616 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10617 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
10618 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
10619 instead of reading the VECTOR_CST directly. Detect whether both
10620 vector inputs are the same before constructing the vec_perm_indices,
10621 and update the number of inputs argument accordingly. Use the
10622 utility functions added above. Only construct sel2 if we need to.
10624 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10626 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
10627 the broadcast of the low byte.
10628 (expand_mult_highpart): Use an explicit encoding for the permutes.
10629 * optabs-query.c (can_mult_highpart_p): Likewise.
10630 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
10631 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10632 (vectorizable_bswap): Likewise.
10633 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
10634 explicit encoding for the power-of-2 permutes.
10635 (vect_permute_store_chain): Likewise.
10636 (vect_grouped_load_supported): Likewise.
10637 (vect_permute_load_chain): Likewise.
10639 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10641 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
10642 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
10643 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
10644 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10645 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
10646 (vect_gen_perm_mask_any): Likewise.
10648 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10650 * int-vector-builder.h: New file.
10651 * vec-perm-indices.h: Include int-vector-builder.h.
10652 (vec_perm_indices): Redefine as an int_vector_builder.
10653 (auto_vec_perm_indices): Delete.
10654 (vec_perm_builder): Redefine as a stand-alone class.
10655 (vec_perm_indices::vec_perm_indices): New function.
10656 (vec_perm_indices::clamp): Likewise.
10657 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
10658 (vec_perm_indices::new_vector): New function.
10659 (vec_perm_indices::new_expanded_vector): Update for new
10660 vec_perm_indices class.
10661 (vec_perm_indices::rotate_inputs): New function.
10662 (vec_perm_indices::all_in_range_p): Operate directly on the
10663 encoded form, without computing elided elements.
10664 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
10665 encoding. Update for new vec_perm_indices class.
10666 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
10667 the given vec_perm_builder.
10668 (expand_vec_perm_var): Update vec_perm_builder constructor.
10669 (expand_mult_highpart): Use vec_perm_builder instead of
10670 auto_vec_perm_indices.
10671 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
10672 vec_perm_indices instead of auto_vec_perm_indices. Use a single
10673 or double series encoding as appropriate.
10674 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
10675 vec_perm_indices instead of auto_vec_perm_indices.
10676 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10677 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10678 (vect_permute_store_chain): Likewise.
10679 (vect_grouped_load_supported): Likewise.
10680 (vect_permute_load_chain): Likewise.
10681 (vect_shift_permute_load_chain): Likewise.
10682 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10683 (vect_transform_slp_perm_load): Likewise.
10684 (vect_schedule_slp_instance): Likewise.
10685 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10686 (vectorizable_mask_load_store): Likewise.
10687 (vectorizable_bswap): Likewise.
10688 (vectorizable_store): Likewise.
10689 (vectorizable_load): Likewise.
10690 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
10691 vec_perm_indices instead of auto_vec_perm_indices. Use
10692 tree_to_vec_perm_builder to read the vector from a tree.
10693 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
10694 vec_perm_builder instead of a vec_perm_indices.
10695 (have_whole_vector_shift): Use vec_perm_builder and
10696 vec_perm_indices instead of auto_vec_perm_indices. Leave the
10697 truncation to calc_vec_perm_mask_for_shift.
10698 (vect_create_epilog_for_reduction): Likewise.
10699 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
10700 from auto_vec_perm_indices to vec_perm_indices.
10701 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10702 instead of changing individual elements.
10703 (aarch64_vectorize_vec_perm_const): Use new_vector to install
10704 the vector in d.perm.
10705 * config/arm/arm.c (expand_vec_perm_d::perm): Change
10706 from auto_vec_perm_indices to vec_perm_indices.
10707 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10708 instead of changing individual elements.
10709 (arm_vectorize_vec_perm_const): Use new_vector to install
10710 the vector in d.perm.
10711 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
10712 Update vec_perm_builder constructor.
10713 (rs6000_expand_interleave): Likewise.
10714 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
10715 (rs6000_expand_interleave): Likewise.
10717 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10719 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
10720 to qimode could truncate the indices.
10721 * optabs.c (expand_vec_perm_var): Likewise.
10723 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10725 * Makefile.in (OBJS): Add vec-perm-indices.o.
10726 * vec-perm-indices.h: New file.
10727 * vec-perm-indices.c: Likewise.
10728 * target.h (vec_perm_indices): Replace with a forward class
10730 (auto_vec_perm_indices): Move to vec-perm-indices.h.
10731 * optabs.h: Include vec-perm-indices.h.
10732 (expand_vec_perm): Delete.
10733 (selector_fits_mode_p, expand_vec_perm_var): Declare.
10734 (expand_vec_perm_const): Declare.
10735 * target.def (vec_perm_const_ok): Replace with...
10736 (vec_perm_const): ...this new hook.
10737 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
10738 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
10739 * doc/tm.texi: Regenerate.
10740 * optabs.def (vec_perm_const): Delete.
10741 * doc/md.texi (vec_perm_const): Likewise.
10742 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
10743 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
10744 expand_vec_perm for constant permutation vectors. Assert that
10745 the mode of variable permutation vectors is the integer equivalent
10746 of the mode that is being permuted.
10747 * optabs-query.h (selector_fits_mode_p): Declare.
10748 * optabs-query.c: Include vec-perm-indices.h.
10749 (selector_fits_mode_p): New function.
10750 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
10751 is defined, instead of checking whether the vec_perm_const_optab
10752 exists. Use targetm.vectorize.vec_perm_const instead of
10753 targetm.vectorize.vec_perm_const_ok. Check whether the indices
10754 fit in the vector mode before using a variable permute.
10755 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
10756 vec_perm_indices instead of an rtx.
10757 (expand_vec_perm): Replace with...
10758 (expand_vec_perm_const): ...this new function. Take the selector
10759 as a vec_perm_indices rather than an rtx. Also take the mode of
10760 the selector. Update call to shift_amt_for_vec_perm_mask.
10761 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
10762 Use vec_perm_indices::new_expanded_vector to expand the original
10763 selector into bytes. Check whether the indices fit in the vector
10764 mode before using a variable permute.
10765 (expand_vec_perm_var): Make global.
10766 (expand_mult_highpart): Use expand_vec_perm_const.
10767 * fold-const.c: Includes vec-perm-indices.h.
10768 * tree-ssa-forwprop.c: Likewise.
10769 * tree-vect-data-refs.c: Likewise.
10770 * tree-vect-generic.c: Likewise.
10771 * tree-vect-loop.c: Likewise.
10772 * tree-vect-slp.c: Likewise.
10773 * tree-vect-stmts.c: Likewise.
10774 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
10776 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
10777 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
10778 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
10779 (aarch64_vectorize_vec_perm_const): ...this new function.
10780 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10781 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10782 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
10783 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
10784 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10785 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10786 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
10788 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
10789 check for NEON modes.
10790 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
10791 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
10792 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
10793 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
10795 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
10796 the old VEC_PERM_CONST conditions.
10797 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
10798 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
10799 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
10800 (ia64_vectorize_vec_perm_const_ok): Merge into...
10801 (ia64_vectorize_vec_perm_const): ...this new function.
10802 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
10803 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
10804 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
10805 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
10806 * config/mips/mips.c (mips_expand_vec_perm_const)
10807 (mips_vectorize_vec_perm_const_ok): Merge into...
10808 (mips_vectorize_vec_perm_const): ...this new function.
10809 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
10810 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
10811 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
10812 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
10813 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
10814 (rs6000_expand_vec_perm_const): Delete.
10815 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
10817 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10818 (altivec_expand_vec_perm_const_le): Take each operand individually.
10819 Operate on constant selectors rather than rtxes.
10820 (altivec_expand_vec_perm_const): Likewise. Update call to
10821 altivec_expand_vec_perm_const_le.
10822 (rs6000_expand_vec_perm_const): Delete.
10823 (rs6000_vectorize_vec_perm_const_ok): Delete.
10824 (rs6000_vectorize_vec_perm_const): New function.
10825 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10826 an element count and rtx array.
10827 (rs6000_expand_extract_even): Update call accordingly.
10828 (rs6000_expand_interleave): Likewise.
10829 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
10830 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
10831 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
10832 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
10833 (rs6000_expand_vec_perm_const): Delete.
10834 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10835 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10836 (altivec_expand_vec_perm_const_le): Take each operand individually.
10837 Operate on constant selectors rather than rtxes.
10838 (altivec_expand_vec_perm_const): Likewise. Update call to
10839 altivec_expand_vec_perm_const_le.
10840 (rs6000_expand_vec_perm_const): Delete.
10841 (rs6000_vectorize_vec_perm_const_ok): Delete.
10842 (rs6000_vectorize_vec_perm_const): New function. Remove stray
10843 reference to the SPE evmerge intructions.
10844 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10845 an element count and rtx array.
10846 (rs6000_expand_extract_even): Update call accordingly.
10847 (rs6000_expand_interleave): Likewise.
10848 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
10849 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
10851 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10853 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10855 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
10856 vector mode and that that mode matches the mode of the data
10858 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
10859 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
10860 directly using expand_vec_perm_1 when forcing selectors into
10862 (expand_vec_perm_var): New function, split out from expand_vec_perm.
10864 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10866 * optabs-query.h (can_vec_perm_p): Delete.
10867 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
10868 * optabs-query.c (can_vec_perm_p): Split into...
10869 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
10870 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
10871 particular selector is valid.
10872 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10873 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10874 (vect_grouped_load_supported): Likewise.
10875 (vect_shift_permute_load_chain): Likewise.
10876 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10877 (vect_transform_slp_perm_load): Likewise.
10878 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10879 (vectorizable_bswap): Likewise.
10880 (vect_gen_perm_mask_checked): Likewise.
10881 * fold-const.c (fold_ternary_loc): Likewise. Don't take
10882 implementations of variable permutation vectors into account
10883 when deciding which selector to use.
10884 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
10885 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
10886 with a false third argument.
10887 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
10888 to test whether the constant selector is valid and can_vec_perm_var_p
10889 to test whether a variable selector is valid.
10891 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10893 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
10894 * optabs-query.c (can_vec_perm_p): Likewise.
10895 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
10896 instead of vec_perm_indices.
10897 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
10898 (vect_gen_perm_mask_checked): Likewise,
10899 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
10900 (vect_gen_perm_mask_checked): Likewise,
10902 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10904 * optabs-query.h (qimode_for_vec_perm): Declare.
10905 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
10906 (qimode_for_vec_perm): ...this new function.
10907 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
10909 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10911 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
10912 does not have a conditional at the top.
10914 2018-01-02 Richard Biener <rguenther@suse.de>
10916 * ipa-inline.c (big_speedup_p): Fix expression.
10918 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10921 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
10924 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10928 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
10929 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
10930 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
10931 cond_taken_branch_cost 3->4.
10933 2018-01-01 Jakub Jelinek <jakub@redhat.com>
10935 PR tree-optimization/83581
10936 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
10937 TODO_cleanup_cfg if any changes have been made.
10939 PR middle-end/83608
10940 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
10941 convert_modes if target mode has the right side, but different mode
10944 PR middle-end/83609
10945 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
10946 last argument when extracting from CONCAT. If either from_real or
10947 from_imag is NULL, use expansion through memory. If result is not
10948 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
10949 the parts directly to inner mode, if even that fails, use expansion
10952 PR middle-end/83623
10953 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
10954 check for bswap in mode rather than HImode and use that in expand_unop
10957 Copyright (C) 2018 Free Software Foundation, Inc.
10959 Copying and distribution of this file, with or without modification,
10960 are permitted in any medium without royalty provided the copyright
10961 notice and this notice are preserved.