1 2018-04-18 Martin Liska <mliska@suse.cz>
4 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8 arguments if they are comparable.
10 2018-04-18 Martin Liska <mliska@suse.cz>
13 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
16 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
19 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
22 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
23 ENDBR after calling __morestack.
25 2018-04-18 David Malcolm <dmalcolm@redhat.com>
28 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
29 by using gcc_base_ver to generate a gcc_driver_version, and use
30 it when generating GCC_DRIVER_NAME.
31 * configure: Regenerate.
33 2018-04-18 Jakub Jelinek <jakub@redhat.com>
36 * config.gcc: Obsolete powerpc*-*-*spe*.
38 2018-04-17 Jakub Jelinek <jakub@redhat.com>
41 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
42 (stabstr_D): Change type of unum from unsigned int to
43 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
46 2018-04-17 Jim Wilson <jimw@sifive.com>
49 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
50 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
51 Set arg_pointer_offset after using pretend_args_size.
53 2018-04-17 Jakub Jelinek <jakub@redhat.com>
55 PR rtl-optimization/85431
56 * dse.c (record_store): Ignore zero width stores.
59 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
60 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
61 __builtin_stack_restore rather than after it.
62 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
63 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
64 argument instead of virtual_dynamic_stack_rtx.
66 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
68 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
70 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
71 Add note to error message to explain internal mapping of overloaded
72 built-in function name to non-overloaded built-in function name.
73 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
76 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
79 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
80 where the inputs overlap with the output.
82 2018-04-17 Jakub Jelinek <jakub@redhat.com>
84 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
85 (=v, v) alternative and explicit "memory" attribute.
86 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
87 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
89 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
90 "sselog1" type instead of "sselog".
91 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
92 "sselog". Remove explicit "memory" attribute.
93 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
94 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
96 (vec_extract_hi_v32hi): Merge all alternatives into one, use
97 "sselog1" type instead of "sselog". Remove explicit "memory"
99 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
100 use "sselog1" type instead of "sselog". Remove explicit "memory"
102 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
103 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
105 (vec_extract_hi_v64qi): Merge all alternatives into one, use
106 "sselog1" type instead of "sselog". Remove explicit "memory"
108 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
109 use "sselog1" type instead of "sselog". Remove explicit "memory"
113 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
116 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
119 2018-04-17 Martin Jambor <mjambor@suse.cz>
122 * ipa-cp.c (create_specialized_node): Call
123 expand_all_artificial_thunks if necessary.
125 2018-04-17 Martin Liska <mliska@suse.cz>
128 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
129 in message, remote space in between '_G' and '('.
131 2018-04-17 Jakub Jelinek <jakub@redhat.com>
134 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
135 avx512f_vmcmp<mode>3<round_saeonly_name>,
136 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
137 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
138 avx512f_rndscale<mode><round_saeonly_name>,
139 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
140 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
141 Use %<iptr>2 instead of %2 for -masm=intel.
142 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
143 avx512f_vcvttss2usi<round_saeonly_name>,
144 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
146 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
147 avx512f_vcvttsd2usi<round_saeonly_name>,
148 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
149 Use %q1 instead of %1 for -masm=intel.
150 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
151 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
152 of %3 for -masm=intel.
153 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
155 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
157 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
159 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
160 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
162 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
164 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
165 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
166 %g1 and one with %0 and %1.
167 (avx512er_vmrcp28<mode><round_saeonly_name>,
168 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
170 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
171 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
172 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
173 of %0 and %{%4%} for -masm=intel.
174 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
175 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
176 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
177 order of %0 and %{%5%}%{z%} for -masm=intel.
179 2018-04-17 Jan Hubicka <jh@suse.cz>
182 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
184 2018-04-17 Martin Liska <mliska@suse.cz>
187 * multiple_target.c (create_dispatcher_calls): Set apostrophes
188 for target_clone error message. Make default implementation
189 clone to be a local declaration.
190 (separate_attrs): Add new argument and check for an empty
192 (expand_target_clones): Handle it.
193 (ipa_target_clone): Make redirection just for target_clones
196 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
197 Tom de Vries <tom@codesourcery.com>
200 * omp-expand.c (expand_oacc_for): Add dummy false branch for
201 tiled basic blocks without omp continue statements.
203 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
206 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
207 vec_extract expression as having side effects to make sure it gets
210 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
213 * config/i386/i386.c (get_builtin_code_for_version): Check
216 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
219 * gcc/config.gcc: Support "skylake".
220 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
222 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
223 (processor_target_table): Add "skylake".
224 (ix86_option_override_internal): Add "skylake".
225 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
226 PROCESSOR_CANNONLAKE.
227 (get_builtin_code_for_version): Fix priority for
228 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
229 PROCESSOR_SKYLAKE-AVX512.
230 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
231 (processor_type): Add PROCESSOR_SKYLAKE.
233 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
234 Jason Merrill <jason@redhat.com>
237 * convert.c (convert_to_integer_1): Use direct recursion for
238 enumeral types and types with a precision less than the number
239 of bits in their mode.
241 2018-04-16 Julia Koval <julia.koval@intel.com>
244 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
245 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
247 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
250 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
251 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
252 and -mno-direct-move.
254 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
257 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
258 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
259 Ensure negative shifts result in {0}.
261 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
263 PR rtl-optimization/79916
264 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
265 regs (if any) to define how to gnerate SD moves when LRA is in
268 2018-04-13 Jakub Jelinek <jakub@redhat.com>
270 PR rtl-optimization/85393
271 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
272 * except.c (expand_dw2_landing_pad_for_region): Make static.
273 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
274 a label and unconditional jump to old_bb, rather than
275 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
278 PR rtl-optimization/85376
279 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
280 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
281 instead of a specific value.
283 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
284 Bin Cheng <bin.cheng@arm.com>
286 PR tree-optimization/82965
287 PR tree-optimization/83991
288 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
289 by_profile_only parameter.
290 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
291 information if the loop was predicted to iterate too many times.
292 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
294 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
297 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
300 2018-04-13 Martin Liska <mliska@suse.cz>
301 Jakub Jelinek <jakub@redhat.com>
304 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
305 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
306 * builtins.c (expand_builtin_memory_copy_args): Use
307 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
308 handle dest_addr == pc_rtx.
310 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
313 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
314 asked to not generate direct moves.
315 (fix_trunc<mode>si2_stfiwx): Similar.
316 (fix_trunc<mode>si2_internal): Similar.
318 2018-04-12 Jakub Jelinek <jakub@redhat.com>
321 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
322 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
323 lookup if dest in some wider mode is known to be const0_rtx and
324 if so, record permanent equivalence for it to be ZERO_EXTEND of
325 the narrower mode destination.
327 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
329 * lto-streamer-out.c (output_function): Revert 259346.
330 * omp-expand.c (expand_oacc_for): Likewise.
332 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
334 PR rtl-optimization/85354
335 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
336 * sel-sched.c (sel_global_init): ... here.
338 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
341 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
342 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
343 mode for PE-COFF targets.
344 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
345 (i386_pe_asm_lto_end): Likewise.
346 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
347 (TARGET_ASM_LTO_END): Likewise.
348 * config/i386/winnt.c (saved_debug_info_level): New static variable.
349 (i386_pe_asm_lto_start): New function.
350 (i386_pe_asm_lto_end): Likewise.
352 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
353 Richard Biener <rguenther@suse.de>
356 * lto-streamer-out.c (output_function): Fix CFG loop state before
358 * omp-expand.c (expand_oacc_for): Handle calls to internal
359 functions like regular functions.
361 2018-04-12 Richard Biener <rguenther@suse.de>
364 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
365 for the early LTO debug to properly generate references to it
366 during DIE emission. Do not re-use that for the skeleton for
368 (dwarf2out_early_finish): Likewise.
370 2018-04-12 Jakub Jelinek <jakub@redhat.com>
374 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
375 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
376 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
377 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
378 and output is a reg, avoid creating invalid lowpart subreg, but
379 instead split into a 512-bit move. Don't split if not AVX512VL,
380 input is xmm16+ reg and output is a mem.
381 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
382 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
383 xmm16+ reg and output is a mem.
385 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
387 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
388 also for flag_dwarf2_cfi_asm.
390 2018-04-12 Jakub Jelinek <jakub@redhat.com>
392 PR rtl-optimization/85342
393 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
394 a bool scalar var inside of the loop instead. Don't try to update
395 recog_data.operand after failed apply_change_group.
397 2018-04-12 Tom de Vries <tom@codesourcery.com>
400 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
401 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
402 array with flexible array member as array without given dimension.
403 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
404 argument for undefined param to true.
406 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
409 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
410 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
411 from PowerPC section.
412 * config/rs6000/sysv4.opt (mcall-): Improve help text.
413 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
414 help text that is too long.
415 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
416 help text that is too long.
417 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
418 help text that is too long.
420 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
422 * config/alpha/alpha.md (stack_probe_internal): Rename
423 from "probe_stack". Update all callers.
425 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
427 PR rtl-optimization/84566
428 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
429 sched_macro_fuse_insns.
431 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
434 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
435 (compute_block_dependences): ... from here.
437 2018-04-11 Jakub Jelinek <jakub@redhat.com>
439 PR tree-optimization/85331
440 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
441 from int to HOST_WIDE_INT.
443 2018-04-11 Martin Jambor <mjambor@suse.cz>
446 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
447 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
448 not the same as the source val.
449 (cgraph_edge_brings_value_p): New parameter.
450 (gather_edges_for_value): Pass destination value to
451 cgraph_edge_brings_value_p.
452 (perhaps_add_new_callers): Likewise.
453 (get_info_about_necessary_edges): Likewise and exclude values brought
454 only by self-recursive edges.
455 (create_specialized_node): Redirect only clones of self-calling edges.
456 (+self_recursive_pass_through_p): New function.
457 (find_more_scalar_values_for_callers_subset): Use it.
458 (find_aggregate_values_for_callers_subset): Likewise.
459 (known_aggs_to_agg_replacement_list): Removed.
460 (decide_whether_version_node): Re-calculate known constants for all
461 remaining context clones.
463 2018-04-11 Richard Biener <rguenther@suse.de>
466 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
467 from early DWARF output.
468 (dwarf2out_early_finish): Output line info unconditionally into
469 early DWARF and add reference to it.
471 2018-04-11 Jakub Jelinek <jakub@redhat.com>
474 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
475 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
476 other than V2DFmode using iptr mode attribute.
477 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
479 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
481 PR rtl-optimization/84659
482 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
484 2018-04-11 Jakub Jelinek <jakub@redhat.com>
487 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
489 (output_loc_list): Pass address of a dummy size variable even in the
490 locview handling loop.
491 (index_location_lists): Add comment on why skip_loc_list_entry can't
494 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
497 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
500 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
503 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
504 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
505 and -mstring-compare-inline-limit.
507 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
510 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
511 for stack clash protection in a register whenever we need it to be in
514 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
516 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
517 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
519 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
522 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
524 (mlong-double-): Ditto.
525 * config/rs6000/sysv4.opt (msdata=): Ditto.
528 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
530 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
531 erroneous entries for
532 "vector int vec_ldl (int, long int *)", and
533 "vector unsigned int vec_ldl (int, unsigned long int *)".
534 Add comments and entries for
535 "vector bool char vec_ldl (int, bool char *)",
536 "vector bool short vec_ldl (int, bool short *)",
537 "vector bool int vec_ldl (int, bool int *)",
538 "vector bool long long vec_ldl (int, bool long long *)",
539 "vector pixel vec_ldl (int, pixel *)",
540 "vector long long vec_ldl (int, long long *)",
541 "vector unsigned long long vec_ldl (int, unsigned long long *)".
542 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
543 type tree bool_long_long_type_node and correct definition of
544 bool_V2DI_type_node to make reference to this new type tree.
545 (rs6000_mangle_type): Replace erroneous reference to
546 bool_long_type_node with bool_long_long_type_node.
547 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
548 comments to emphasize sign distinctions for char and int types and
549 replace RS6000_BTI_bool_long constant with
550 RS6000_BTI_bool_long_long constant. Also add comment to restrict
551 use of RS6000_BTI_pixel.
552 (bool_long_type_node): Remove this macro definition.
553 (bool_long_long_type_node): New macro definition
555 2018-04-10 Jakub Jelinek <jakub@redhat.com>
557 PR rtl-optimization/85300
558 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
559 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
560 simplify_unary_operation fails.
562 2018-04-10 Martin Liska <mliska@suse.cz>
564 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
565 cgraph_edge and ipa_ref.
567 2018-04-10 Jakub Jelinek <jakub@redhat.com>
572 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
573 computation of the VEC_MERGE selector from mask.
574 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
575 Fix decoding of the VEC_MERGE selector into mask.
577 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
579 PR tree-optimization/85286
580 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
582 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
584 * final.c (final_1): Set insn_last_address as well as
585 insn_current_address.
587 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
590 * explow.c (emit_stack_probe): Call validize_mem on memory location
591 before passing it to gen_probe_stack. Create address operand and
592 legitimize it for the probe_stack_address case.
594 2018-04-09 Jan Hubicka <jh@suse.cz>
597 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
598 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
599 * tree.c (free_lang_data_in_type): Fix handling of binfos;
601 (free_lang_data): Rebuild type inheritance graph.
603 2018-04-09 Martin Sebor <msebor@redhat.com>
605 * invoke.texi (-finline-small-functions): Mention other optimization
607 (-findirect-inlining, -fpartial-inlining): Same.
608 (-finline-functions-called-once): Same.
609 (-freorder-blocks-and-partition): Same.
611 2018-04-09 Jan Hubicka <jh@suse.cz>
614 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
615 jumps; choose last target that matches the criteria (i.e.
616 no partition changes for non-crossing jumps).
617 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
618 support for redirecting crossing jumps to non-crossing.
620 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
622 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
623 also for naked functions.
625 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
627 * config/arc/arc.md (add_shift): New pattern.
628 (add_shift2): Likewise.
629 (sub_shift): Likewise.
630 (sub_shift_cmp0_noout): Likewise.
631 (compare_si_ashiftsi): Likewise.
632 (xbfu_cmp0_noout): New combine pattern.
633 (xbfu_cmp0"): Likewise.
634 (movsi_set_cc_insn): Place the predicable variant first.
635 (commutative_binary_cmp0_noout): Remove clobber.
636 (commutative_binary_cmp0): New pattern.
637 (noncommutative_binary_cmp0): Likewise.
638 (noncommutative_binary_cmp0_noout): Likewise.
639 (noncommutative_binary_comparison_result_used): Removed.
640 (rsub_cmp0): New pattern.
641 (rsub_cmp0_noout): Likewise.
642 (extzvsi): Changed, keep only meaningful variants.
643 (SQH, SEZ): New iterators.
644 (SQH_postfix): New mode attribute.
645 (SEZ_prefix): New code attribute.
646 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
647 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
648 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
650 (noncommutative_operator): Check the availability of barrel
653 2018-04-09 Richard Biener <rguenther@suse.de>
655 PR tree-optimization/85284
656 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
657 Only use the niter constraining form of simple_iv when the exit
660 2018-04-09 Tom de Vries <tom@codesourcery.com>
663 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
664 (define_expand "*memory_barrier"): New define_expand.
665 (define_insn "memory_barrier"): New insn.
667 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
669 PR rtl-optimization/80463
670 PR rtl-optimization/83972
671 PR rtl-optimization/83480
673 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
674 correct producer for the insn.
675 (tidy_control_flow): Fixup seqnos in case of debug insns.
677 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
679 PR rtl-optimization/83913
681 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
682 different sched-times when merging exprs.
684 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
686 PR rtl-optimization/83962
688 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
689 tidy_fallthru_edge and tidy_control_flow.
691 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
693 PR rtl-optimization/83530
695 * sel-sched.c (force_next_insn): New global variable.
696 (remove_insn_for_debug): When force_next_insn is true, also leave only
697 next insn in the ready list.
698 (sel_sched_region): When the region wasn't scheduled, make another pass
699 over it with force_next_insn set to 1.
701 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
703 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
705 * config/nds32/constants.md (unspec_volatile_element): Add enum values
706 for interrupt control.
707 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
708 functions for interrupt control.
709 * config/nds32/nds32-intrinsic.md: Likewise.
710 * config/nds32/nds32_intrinsic.h: Likewise.
711 * config/nds32/nds32.h (nds32_builtins): Likewise.
713 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
715 * config/nds32/nds32.c (nds32_init_machine_status,
716 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
717 strict_aligned_p field.
718 (nds32_expand_to_rtl_hook): New function.
719 (TARGET_EXPAND_TO_RTL_HOOK): Define.
720 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
722 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
723 Chung-Ju Wu <jasonwucj@gmail.com>
725 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
726 * config/nds32/nds32-n7.md: New file.
727 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
728 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
730 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
731 * config/nds32/nds32.md (pipeline_model): Add n7.
732 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
733 * config/nds32/pipelines.md: Include n7 settings.
735 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
736 Chung-Ju Wu <jasonwucj@gmail.com>
738 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
739 * config/nds32/nds32-e8.md: New file.
740 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
741 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
743 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
744 * config/nds32/nds32.md (pipeline_model): Add e8.
745 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
746 * config/nds32/pipelines.md: Include e8 settings.
748 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
749 Chung-Ju Wu <jasonwucj@gmail.com>
751 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
752 * config/nds32/nds32-n8.md: New file.
753 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
754 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
756 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
757 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
758 * config/nds32/nds32.md (pipeline_model): Add n8.
759 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
760 * config/nds32/pipelines.md: Include n8 settings.
762 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
763 Chung-Ju Wu <jasonwucj@gmail.com>
765 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
766 * config/nds32/nds32-n9-2r1w.md: New file.
767 * config/nds32/nds32-n9-3r2w.md: New file.
768 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
769 nds32_register_ports): New or modify for cpu n9.
770 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
772 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
773 * config/nds32/nds32-utils.c: New file.
774 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
775 TARGET_MUL_SLOW): Define.
776 * config/nds32/nds32.md (pipeline_model): New attribute.
777 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
778 New options that support cpu n9.
779 * config/nds32/pipelines.md: Include n9 settings.
780 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
782 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
784 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
785 information if necessary.
786 (output_cond_branch_compare_zero): Likewise.
787 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
788 (nds32_target_alignment): Refine for alignment.
789 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
790 (FUNCTION_BOUNDARY): Modify.
791 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
793 * config/nds32/nds32.opt (malways-align, malign-functions): New.
795 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
797 * config/nds32/constants.md (unspec_volatile_element): Add values for
798 TLB operation and data prefetch.
799 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
800 functions for TLB operation and data prefetch.
801 * config/nds32/nds32-intrinsic.md: Likewise.
802 * config/nds32/nds32_intrinsic.h: Likewise.
803 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
804 (nds32_print_operand): Likewise.
805 * config/nds32/nds32.h (nds32_builtins): Likewise.
807 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
808 Andrew Pinski <pinsika@gcc.gnu.org>
811 * match.pd: Use constant_boolean_node of correct type instead of
812 boolean_true_node or boolean_false_node for simplifying
813 pointer comparisons to zero.
815 2018-04-07 Jakub Jelinek <jakub@redhat.com>
817 PR tree-optimization/80021
818 * tree.c (verify_type_variant): Make error call in verify_variant_match
819 translatable and remove final full stop.
821 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
823 * config/nds32/constants.md (unspec_volatile_element): Add
824 UNSPEC_VOLATILE_EH_RETURN.
825 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
826 nds32_output_stack_pop): Support dwarf exception handling process.
827 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
828 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
829 exception handling process.
830 (nds32_compute_stack_frame): Likewise.
831 (nds32_return_addr_rtx): Likewise.
832 (nds32_initial_elimination_offset): Likewise.
833 (nds32_expand_prologue): Likewise.
834 (nds32_expand_epilogue): Likewise.
835 (nds32_dynamic_chain_address): New function.
836 * config/nds32/nds32.h (machine_function): Add fields for dwarf
838 (DYNAMIC_CHAIN_ADDRESS): Define.
839 (EH_RETURN_DATA_REGNO): Define.
840 (EH_RETURN_STACKADJ_RTX): Define.
841 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
842 patterns for dwarf exception handling.
844 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
846 * config/nds32/nds32.h: Clean up obsolete macros.
848 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
850 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
851 Add enum values for particular instructions.
852 * config/nds32/nds32-intrinsic.c: Implementation of expanding
853 particular intrinsic functions.
854 * config/nds32/nds32-intrinsic.md: Likewise.
855 * config/nds32/nds32_intrinsic.h: Likewise.
856 * config/nds32/nds32.h (nds32_builtins): Likewise.
857 * config/nds32/nds32.md (type): Add pbsad and pbsada.
858 (btst, ave): New patterns for particular instructions.
860 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
862 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
863 Add enum values for atomic load/store and memory sync.
864 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
866 * config/nds32/nds32-intrinsic.md: Likewise.
867 * config/nds32/nds32_intrinsic.h: Likewise.
868 * config/nds32/nds32.h (nds32_builtins): Likewise.
870 2018-04-07 Jakub Jelinek <jakub@redhat.com>
872 PR tree-optimization/85257
873 * fold-const.c (native_encode_vector): If not all elts could fit
874 and off is -1, return 0 rather than offset.
875 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
876 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
877 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
878 adjust buffer in native_interpret_expr call.
880 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
882 * config/nds32/constants.md (unspec_volatile_element): Add cache
884 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
885 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
886 * config/nds32/nds32.c (nds32_cctl_names): New.
887 (nds32_print_operand): Handle cache control register names.
888 * config/nds32/nds32.h (nds32_builtins): New enum values.
889 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
891 * config/nds32/nds32.md (type): Add mmu.
892 * config/nds32/pipelines.md (simple_insn): Add mmu.
894 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
896 * config/nds32/nds32.md (type): Remove call.
897 * config/nds32/pipelines.md (simple_insn): Likewise.
899 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
901 * config/nds32/constants.md (unspec_volatile_element): Add
902 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
903 UNSPEC_VOLATILE_FMFCFG.
904 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
905 description for fmfcfg and fmfcsr.
906 (bdesc_1arg): Add fmtcsr.
907 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
908 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
909 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
910 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
911 unspec_fmfcfg): New patterns.
912 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
913 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
914 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
915 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
916 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
917 __nds32__fmfcfg): Define.
919 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
921 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
922 intrinsic register names.
923 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
924 intrinsic register enum values and macros.
926 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
928 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
929 for load/store addressing form.
930 (nds32_print_operand_address): Likewise.
932 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
935 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
936 based on LABEL_REF. Remove useless assertion.
937 (pic_address_needs_scratch): Fix formatting.
938 (sparc_legitimize_pic_address): Minor tweaks.
939 (sparc_delegitimize_address): Adjust assertion accordingly.
940 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
941 into symbolic_operand.
942 (movsi_high_pic_label_ref): Likewise.
943 (movsi_lo_sum_pic_label_ref): Likewise.
944 (movdi_pic_label_ref): Likewise.
945 (movdi_high_pic_label_ref): Likewise.
946 (movdi_lo_sum_pic_label_ref): Likewise.
948 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
950 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
951 custom LIB_SPEC setup.
953 2018-04-06 Ruslan Bukin <br@bsdpad.com>
954 Kito Cheng <kito.cheng@gmail.com>
956 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
957 * config/riscv/freebsd.h: New.
959 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
961 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
962 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
965 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
966 Kito Cheng <kito.cheng@gmail.com>
968 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
969 nds32_output_call, nds32_symbol_binds_local_p): New functions.
970 * config/nds32/nds32-protos.h (nds32_output_call,
971 nds32_output_return): Declare.
972 * config/nds32/nds32.md: Refine all the call and return patterns.
974 2018-04-06 Jakub Jelinek <jakub@redhat.com>
977 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
978 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
980 PR rtl-optimization/84872
981 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
982 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
985 2018-04-06 Tamar Christina <tamar.christina@arm.com>
987 * expr.c (copy_blkmode_to_reg): Revert 254862.
988 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
990 2018-04-06 Richard Biener <rguenther@suse.de>
993 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
994 after seeing a component reference with an adjacent field. Treat
995 refs to arrays at struct end of external decls similar to
996 refs to unconstrained commons.
998 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1001 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1002 look through SAVE_EXPRs with non-side-effects argument. Adjust
1004 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1007 2018-04-06 Richard Biener <rguenther@suse.de>
1010 * alias.c (find_base_term): New wrapper around find_base_term
1011 unwinding CSELIB_VAL_PTR changes.
1012 (find_base_term): Do not restore CSELIB_VAL_PTR during the
1015 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1017 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1019 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1020 constant definitions.
1021 ("nop"): lr 0,0 -> nopr r0
1022 ("nop_lr0", "nop_lr1"): New insn definitions.
1024 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1026 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1027 NDS32_V3PUSH_AVAILABLE_P macro.
1029 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
1030 Chung-Ju Wu <jasonwucj@gmail.com>
1032 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1033 (nds32*-*-*): Add float and fpu_config into supported_defaults.
1034 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1035 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1036 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1037 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1038 * config/nds32/constraints.md: New constraints and checking for hard
1039 float configuration.
1040 * config/nds32/iterators.md: New mode iterator and attribute for hard
1041 float configuration.
1042 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1044 * config/nds32/nds32-fpu.md: New file.
1045 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1046 deal with hard float code generation.
1047 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1049 (abi_type, float_reg_number): New enum type.
1050 * config/nds32/nds32-predicates.c: New predicates for hard float.
1051 * config/nds32/nds32-protos.h: Declare functions for hard float.
1052 * config/nds32/nds32.c: Implementation for hard float configuration.
1053 * config/nds32/nds32.h: Definitions for hard float configuration.
1054 * config/nds32/nds32.md: Include hard float machine description and
1055 modify patterns for hard float configuration.
1056 * config/nds32/nds32.opt: New options for hard float configuration.
1057 * config/nds32/predicates.md: New predicates for hard float
1060 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1062 * common/config/nds32/nds32-common.c
1063 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1065 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1068 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1069 CONSTRUCTOR_ELT (ctor, ...)->value.
1071 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1074 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1076 2018-04-05 Tom de Vries <tom@codesourcery.com>
1079 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1082 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1083 Kito Cheng <kito.cheng@gmail.com>
1085 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1086 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1087 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1088 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1090 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1091 Kito Cheng <kito.cheng@gmail.com>
1093 * config/nds32/constraints.md (Ufe): New memory constraint.
1094 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1095 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1096 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1098 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1099 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1101 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1103 * config/nds32/nds32.md: Use optimize_size in the condition for
1104 alu-shift instructions.
1106 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1108 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1110 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1112 * config/nds32/nds32.md (negsi2): Refine pattern.
1114 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1115 Chung-Ju Wu <jasonwucj@gmail.com>
1117 * config/nds32/iterators.md (shift_rotate): New code iterator.
1118 (shift): New code attribute.
1119 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1120 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1121 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1122 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1123 bit-wise operations.
1124 (andsi3, *andsi3): Ditto.
1125 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1126 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1127 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1128 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1129 nds32_ior_operand, nds32_xor_operand): New predicates.
1131 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1133 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1134 (addsi3, subsi3): ... this.
1136 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1138 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1140 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1142 * config/nds32/nds32.md: Adjust indention.
1144 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1146 * config/nds32/nds32.md (feature): New attribute.
1148 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1150 * config/nds32/nds32.md (subtype): New attribute.
1152 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1155 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1156 expansion to perform a bitwise AND of the argument followed by a
1157 boolean negation of the result.
1159 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1161 PR rtl-optimization/84878
1162 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1163 the basic block. Assert the use reference is not artificial and that
1164 it has an associated insn.
1166 2018-04-04 Michael Matz <matz@suse.de>
1168 * builtins.c (compute_objsize): Pass correct operand
1169 to array_at_struct_end_p.
1171 2018-04-04 Richard Biener <rguenther@suse.de>
1174 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1175 from contexts for DINFO_LEVEL_TERSE and below.
1177 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1179 * config/nds32/nds32-doubleword.md (move_<mode>): Require
1180 resiter_operand condition.
1181 * config/nds32/nds32.md (*move<mode>): Ditto.
1183 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1184 Monk Chiang <sh.chiang04@gmail.com>
1186 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1188 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1190 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1192 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1193 Kito Cheng <kito.cheng@gmail.com>
1195 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1196 nds32_cond_code_str, output_cond_branch,
1197 output_cond_branch_compare_zero, nds32_expand_cbranch,
1198 nds32_expand_cstore, nds32_expand_movcc,
1199 nds32_output_cbranchsi4_equality_zero,
1200 nds32_output_cbranchsi4_equality_reg,
1201 nds32_output_cbranchsi4_equality_reg_or_const_int,
1202 nds32_output_cbranchsi4_greater_less_zero: New functions.
1203 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1204 nds32_expand_cstore, nds32_expand_movcc,
1205 nds32_output_cbranchsi4_equality_zero,
1206 nds32_output_cbranchsi4_equality_reg,
1207 nds32_output_cbranchsi4_equality_reg_or_const_int,
1208 nds32_output_cbranchsi4_greater_less_zero): Declare.
1209 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1210 nds32_rimm11s_operand): New predicates.
1211 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1212 * config/nds32/nds32.md: Rewrite all the branch and conditional move
1215 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1217 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
1218 * config/nds32/nds32.md: Ditto.
1219 * config/nds32/pipelines.md: Ditto.
1221 2018-04-04 Richard Biener <rguenther@suse.de>
1223 PR tree-optimization/85168
1224 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
1225 propagating abnormals.
1227 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1229 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
1231 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1232 Kito Cheng <kito.cheng@gmail.com>
1234 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
1235 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
1236 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
1237 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
1238 * config/nds32/nds32.md (sibcall_internal): New.
1239 (sibcall_register): Remove.
1240 (sibcall_immediate): Remove.
1241 (sibcall_value_internal): New.
1242 (sibcall_value_register): Remove.
1243 (sibcall_value_immediate): Remove.
1244 * config/nds32/predicates.md (nds32_general_register_operand): New.
1245 (nds32_call_address_operand): New.
1247 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1249 PR rtl-optimization/85167
1250 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1251 bb_defs if *split_p, instead preinitialize it to NULL.
1253 PR tree-optimization/85156
1254 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1255 evaluating the argument multiple times.
1257 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
1259 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1261 (_mm_cvtpd_ps): Likewise.
1262 (_mm_cvttpd_epi32): Likewise.
1263 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1264 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1265 vector, pixel, and bool following altivec.h include.
1267 2018-04-03 Martin Sebor <msebor@redhat.com>
1269 * doc/extend.texi (Common Function Attributes): Clarify.
1270 (const attribute): Likewise.
1271 (pure attribute): Likewise.
1273 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1276 * config/i386/i386.c (ix86_expand_vector_set): Use
1277 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
1279 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
1281 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1282 instructions when changing rounding bits to preserve precision bits
1283 in the x87 control word.
1285 2018-04-03 Martin Liska <mliska@suse.cz>
1287 PR tree-optimization/82491
1288 * rtl.h (strip_offset_and_add): Replace += suboffset with
1289 poly_uint64 () + suboffset.
1291 2018-03-29 Martin Liska <mliska@suse.cz>
1292 Martin Jambor <mjambor@suse.cz>
1295 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1296 param_type is not an integral or pointer type.
1298 2018-04-03 Richard Biener <rguenther@suse.de>
1300 * sese.h (recompute_all_dominators): Remove.
1302 2018-04-02 Martin Sebor <msebor@redhat.com>
1304 * doc/invoke.texi (-Wrestrict): Fix typos.
1306 2018-04-02 Jim Wilson <jimw@sifive.com>
1308 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1309 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1310 (<optab>di3, <optab>si3_extend): Likewise.
1311 (<optab>si3_mask, <optab>si3_mask_1): New.
1312 (<optab>di3_mask, <optab>di3_mask_1): New.
1313 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1314 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1315 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1317 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1319 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1322 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1324 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1325 (nds32_canonicalize_comparison): New function.
1327 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1328 Kito Cheng <kito.cheng@gmail.com>
1329 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1331 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1332 * config/nds32/constants.md (unspec_volatile_element): Add
1333 UNSPEC_VOLATILE_RELAX_GROUP.
1334 * config/nds32/nds32-relax-opt.c: New file.
1335 * config/nds32/nds32-predicates.c
1336 (nds32_symbol_load_store_p): New function.
1337 * config/nds32/nds32-protos.h
1338 (nds32_symbol_load_store_p): Declare function.
1339 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1340 * config/nds32/nds32.c
1341 (nds32_register_pass): New function to register pass.
1342 (nds32_register_passes): New function to register passes.
1343 * config/nds32/nds32.md (relax_group): New pattern.
1344 * config/nds32/nds32.opt (mrelax-hint): New option.
1345 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1347 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
1349 * config/nds32/t-nds32: Modify files dependency.
1351 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1353 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1354 (PROFILE_HOOK): Define its implementation.
1356 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1358 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1359 type and 32-bit size.
1361 2018-04-01 Jakub Jelinek <jakub@redhat.com>
1364 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1365 (V_128_256): New mode iterator.
1366 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1367 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1368 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1370 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1371 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1373 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1376 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1377 NaN inputs correctly.
1379 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1382 * config/rs6000/vsx.md (??r): New mode attribute.
1383 (*vsx_mov<mode>_64bit): Use it.
1384 (*vsx_mov<mode>_32bit): Likewise.
1386 2018-03-30 Martin Sebor <msebor@redhat.com>
1388 PR tree-optimization/84818
1389 * builtins.c (check_access): Use warning_n.
1391 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1394 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1396 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1399 2018-03-30 Julia Koval <julia.koval@intel.com>
1402 * x86-tune.def (movx, partial_reg_dependency): Enable for
1405 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
1408 * lra-constraints.c (process_alt_operands): Move setting
1409 this_alternative_matches below.
1411 2018-03-29 Martin Liska <mliska@suse.cz>
1414 * doc/invoke.texi: Document how LTO works with debug info.
1415 Describe auto-load support of binutils. Mention 'x86-64'
1416 as valid option value of -march option.
1418 2018-03-29 Jakub Jelinek <jakub@redhat.com>
1420 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
1423 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
1424 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
1425 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
1428 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1431 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
1432 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
1433 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
1434 for RS6000_BTM_POWERPC64.
1435 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
1436 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
1437 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
1442 2018-03-28 Carl Love <cel@us.ibm.com>
1445 2017-09-27 Carl Love <cel@us.ibm.com>
1447 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
1448 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
1449 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
1452 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1454 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
1455 instead of __vector bool.
1456 (_mm_max_pu8): Likewise.
1457 (_mm_min_pi16): Likewise.
1459 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1462 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
1463 (DIVWEUO): Likewise.
1465 (DIVDEUO): Likewise.
1466 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
1467 DIVWEUO and DIVDEUO.
1468 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
1469 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
1470 (div_extend): Likewise.
1471 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
1473 (__builtin_divweuo): Likewise.
1474 (__builtin_divdeo): Likewise.
1475 (__builtin_divdeuo): Likewise.
1477 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1480 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
1481 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
1483 PR tree-optimization/82004
1484 * gimple-match-head.c (optimize_pow_to_exp): New function.
1485 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
1486 Don't fold to exp if optimize_pow_to_exp is false.
1488 2018-03-28 Martin Liska <mliska@suse.cz>
1491 * calls.c (initialize_argument_information): Fix trailing space.
1492 * common.opt: Fix typo and provide better explanation for
1493 -fsanitize-coverage option.
1494 * config/i386/i386.opt: Fix typo.
1496 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1497 Martin Liska <mliska@suse.cz>
1500 * gimplify.c (asan_poison_variable): Don't do the check for
1501 gimplify_omp_ctxp here.
1502 (gimplify_decl_expr): Do it here.
1503 (gimplify_target_expr): Likewise.
1505 2018-03-28 Martin Liska <mliska@suse.cz>
1508 * config/i386/i386.c (ix86_function_arg_advance): Do not call
1509 chkp_type_bounds_count if MPX is not enabled.
1511 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
1513 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
1515 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
1518 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
1519 function to create the function decl for complex long double
1520 multiply and divide for -mabi=ieeelongdouble.
1521 (init_float128_ieee): Call it.
1523 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
1526 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
1527 -fcf-protection=branch -mibt.
1528 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
1530 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1533 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
1535 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
1538 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
1539 extern array declarations.
1541 2018-03-27 Richard Biener <rguenther@suse.de>
1544 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
1545 explicit single_use checks.
1547 2018-03-27 Richard Biener <rguenther@suse.de>
1549 PR tree-optimization/85082
1550 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
1553 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1555 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
1556 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
1557 Turn on fasynchronous-unwind-tables and funwind-tables.
1559 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
1562 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
1563 (*bmi_blsr_<mode>_ccz): Ditto.
1565 2018-03-26 Tom de Vries <tom@codesourcery.com>
1567 PR tree-optimization/85063
1568 * omp-general.c (offloading_function_p): New function. Factor out
1570 * omp-offload.c (pass_omp_target_link::gate): ... here.
1571 * omp-general.h (offloading_function_p): Declare.
1572 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
1573 with attribute omp declare target for offloading functions.
1575 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
1577 PR tree-optimization/84005
1578 * tree-data-ref.h (get_base_for_alignment): Declare.
1579 * tree-data-ref.c (get_base_for_alignment_1): New function.
1580 (get_base_for_alignment): Likewise.
1581 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
1582 get_base_for_alignment to find a suitable base object, instead
1583 of always using drb->base_address.
1585 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1588 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
1589 known size by default.
1591 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
1594 * lra-constraints.c (process_alt_operands): Don't match BLKmode
1595 and non BLKmode operands.
1597 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1600 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
1601 Clean up attributes.
1603 2018-03-23 Richard Biener <rguenther@suse.de>
1606 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
1607 we are going to emit early debug for LTO.
1609 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1612 * function.c (match_asm_constraints_1): Don't optimize if input
1613 doesn't satisfy general_operand predicate for output's mode.
1616 * alias.c (write_dependence_p): Don't require for x_canonicalized
1617 non-VOIDmode if x has VOIDmode.
1620 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
1621 just don't try to optimize it rather than assert it never happens.
1623 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1625 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
1626 macro expansions for definition of ST_INTERNAL_<mode> and
1627 LD_INTERNAL_<mode> builtins.
1628 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
1630 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
1632 (altivec_expand_st_builtin): Likewise.
1633 (altivec_expand_builtin): Remove calls to deleted functions.
1634 (rs6000_address_for_altivec): Delete this function.
1635 * config/rs6000/vector.md: Remove expands for
1636 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
1638 2018-03-22 Sudakshina Das <sudi.das@arm.com>
1641 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
1642 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
1643 re-computing once computed.
1644 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
1645 (arm_init_machine_status): Initialize
1646 machine->static_chain_stack_bytes.
1648 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1651 * doc/extend.texi: Add four new prototypes for vec_ld.
1652 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
1653 definitions for more logical presentation.
1654 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
1655 entries for V1TI variants of __builtin_altivec_ld builtin.
1656 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
1657 handling of V1TI variant of LVX icode pattern.
1658 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
1659 (rs6000_gimple_fold_builtin): Likewise.
1660 (altivec_init_builtins): Add code to define
1661 __builtin_altivec_lvx_v1ti function.
1663 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1666 * function.c (match_asm_constraints_1): Don't do the optimization
1667 if input isn't a REG, SUBREG, MEM or constant.
1669 2018-03-22 Tom de Vries <tom@codesourcery.com>
1671 PR tree-optimization/84956
1672 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
1673 bb_has_abnormal_pred.
1675 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1678 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
1679 DECL_INITIAL (decl) to decl at the end.
1680 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
1683 2018-03-21 Joseph Myers <joseph@codesourcery.com>
1685 * doc/extend.texi (__builtin_tgmath): Document when complex
1686 integer types are treated as _Complex _Float64.
1688 2018-03-21 Tom de Vries <tom@codesourcery.com>
1690 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
1692 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1694 PR tree-optimization/84960
1695 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
1696 if it is ENTRY block, move them into single succ of ENTRY in that case.
1698 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
1700 PR tree-optimization/84811
1701 * poly-int.h (poly_span_traits): Remove the T3 parameter and
1702 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
1703 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
1704 (known_subrange_p): Update accordingly. Cast each value involved
1705 in the size comparison, rather than casting the result of the
1708 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1710 PR tree-optimization/84982
1711 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
1712 by flipping the least significant bit rather than all bits from
1713 bitpos to bitpos + bitsize - 1.
1715 2018-03-21 Nathan Sidwell <nathan@acm.org>
1717 * doc/extend.texi (Deprecated Features): Remove mention of
1718 long-deleted deprecations.
1720 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1723 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
1724 * configure: Regenerate.
1726 2018-03-21 Tom de Vries <tom@codesourcery.com>
1728 PR tree-optimization/83126
1729 * tree-parloops.c (num_phis): New function.
1730 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
1732 2018-03-21 Nathan Sidwell <nathan@acm.org>
1734 * doc/extend.texi (Deprecated Features): Update deprecated flags,
1735 mention anon-struct/union members and trailing attributes.
1737 2018-03-21 Bin Cheng <bin.cheng@arm.com>
1739 PR tree-optimization/84969
1740 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
1741 builtin memset partitions if they set different rhs values.
1743 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1745 PR rtl-optimization/84989
1746 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
1747 VEC_DUPLICATE with scalar result mode.
1749 2018-03-21 Martin Liska <mliska@suse.cz>
1752 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
1753 not intended return statement.
1755 2018-03-21 Martin Liska <mliska@suse.cz>
1758 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
1759 (chkp_find_bound_slots_1): Limit number of iterations.
1761 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
1764 * Minor grammar fixes for x86 options.
1766 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1769 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
1770 holding REG_CFA_RESTORE notes, instead turn them into a USE.
1772 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
1775 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
1776 (altivec_lvx_<mode>_1op): Likewise.
1777 (altivec_stvx_<mode>_2op): Likewise.
1778 (altivec_stvx_<mode>_1op): Likewise.
1779 (altivec_lvx_<VM2:mode>): New define_expand.
1780 (altivec_stvx_<VM2:mode>): Likewise.
1781 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
1782 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1783 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
1784 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1785 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
1786 (rs6000_gen_lvx): Likewise.
1787 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
1788 (altivec_expand_stv_builtin): Likewise.
1789 (altivec_expand_builtin): Likewise.
1790 * config/rs6000/vector.md: Likewise.
1792 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1795 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
1798 2018-03-20 Richard Biener <rguenther@suse.de>
1801 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
1802 sign-conversions as zero, fall back to standard scalar_stmt
1805 2018-03-20 Martin Liska <mliska@suse.cz>
1808 * predict.c (rebuild_frequencies): Handle case when we have
1809 PROFILE_ABSENT, but flag_guess_branch_prob is false.
1811 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1814 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
1815 flag_section_anchors.
1816 * varasm.c (use_blocks_for_decl_p): Remove hack for
1817 dw2_force_const_mem.
1820 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
1822 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
1823 be created, use lowpart_subreg of operands[0] rather than operands[0]
1825 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
1826 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
1827 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
1828 and n constraint instead of aarch64_shift_imm_di and Usd.
1829 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
1830 (*aarch64_<optab>_reg_minus<mode>3): ... this.
1832 2018-03-20 Sudakshina Das <sudi.das@arm.com>
1835 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
1836 to favor GPR over NEON registers.
1837 (<shift>di3_neon): Likewise.
1839 2018-03-20 Tom de Vries <tom@codesourcery.com>
1842 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
1843 (nvptx_process_pars): Emit bar.sync asap and alap.
1845 2018-03-20 Tom de Vries <tom@codesourcery.com>
1848 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
1849 seen_label if seen_label is already set.
1851 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1854 * config/i386/i386.c (fold_builtin_cpu): For features above 31
1855 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
1856 Use 1U instead of 1. Formatting fixes.
1859 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
1860 instead of TREE_TYPE (s1) for the return value.
1862 2018-03-19 Jakub Jelinek <jakub@redhat.com>
1864 PR tree-optimization/84946
1865 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
1866 bitsize + bitsize in poly_uint64 rather than poly_int64.
1869 * dwarf2asm.c: Include fold-const.c.
1870 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
1871 of decl rather than decl itself.
1873 PR rtl-optimization/84643
1874 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
1876 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
1879 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
1880 calling assemble_variable.
1882 2018-03-19 Sudakshina Das <sudi.das@arm.com>
1885 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
1886 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
1888 2018-03-19 Jim Wilson <jimw@sifive.com>
1891 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
1892 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
1893 (riscv_first_stack_step): Likewise.
1894 (riscv_option_override): Use STACK_BOUNDARY instead of
1896 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
1898 (BIGGEST_ALIGNMENT): Set to 128.
1899 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
1900 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
1903 2018-03-19 Richard Biener <rguenther@suse.de>
1905 PR tree-optimization/84933
1906 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
1907 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
1909 2018-03-19 Richard Biener <rguenther@suse.de>
1911 PR tree-optimization/84859
1912 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
1913 (cond_if_else_store_replacement): Perform sinking operation on
1914 single-store BBs regardless of MAX_STORES_TO_SINK setting.
1915 Generalize what a BB with a single eligible store is.
1917 2018-03-19 Richard Biener <rguenther@suse.de>
1919 PR tree-optimization/84929
1920 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
1921 chrec_is_positive against non-chrec arg.
1923 2018-03-19 Tamar Christina <tamar.christina@arm.com>
1926 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
1928 2018-03-18 Martin Liska <mliska@suse.cz>
1930 PR rtl-optimization/84635
1931 * regrename.c (build_def_use): Use matches_mode only when
1934 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
1936 PR tree-optimization/84913
1937 * tree-vect-loop.c (vectorizable_reduction): Don't try to
1938 vectorize chains of COND_EXPRs.
1940 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1942 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
1944 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1946 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
1948 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1950 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
1952 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1953 Kito Cheng <kito.cheng@gmail.com>
1955 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
1956 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
1957 (nds32_adjust_reg_alloc_order): New function.
1958 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
1960 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
1962 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
1963 nds32_print_operand, nds32_print_operand_address): Use
1964 HOST_WIDE_INT_PRINT_DEC instead.
1966 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1968 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
1970 2018-03-17 Jakub Jelinek <jakub@redhat.com>
1973 * config/i386/i386.c (initial_ix86_tune_features,
1974 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
1976 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
1977 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
1978 rather than 1u << ix86_tune. Formatting fix.
1979 (ix86_option_override_internal): Change ix86_arch_mask from
1980 unsigned int to unsigned HOST_WIDE_INT, initialize to
1981 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
1982 (ix86_function_specific_restore): Likewise.
1984 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1987 * postreload.c (reload_combine_recognize_pattern): Perform
1988 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
1989 truncate_int_for_mode the result for the destination's mode.
1992 * hsa-gen.c (mem_type_for_type): Fix comment typo.
1993 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
1995 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1998 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2001 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2002 regno_allocno_class_array and sorted_pseudos.
2003 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2004 insns where regno is used.
2006 2018-03-16 Martin Liska <mliska@suse.cz>
2009 * multiple_target.c (create_dispatcher_calls): Redirect
2010 reference in the symbol table.
2012 2018-03-16 Martin Liska <mliska@suse.cz>
2015 * multiple_target.c (create_dispatcher_calls): Redirect also
2018 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2022 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2023 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2024 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2026 2018-03-16 Julia Koval <julia.koval@intel.com>
2028 * doc/invoke.texi (Skylake Server): Add CLWB.
2029 Cannonlake): Remove CLWB.
2031 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2033 PR tree-optimization/84841
2034 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2036 (FLOAT_ONE_CONST_TYPE): Define.
2037 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2038 (sort_by_operand_rank): Put entries with higher constant_type last
2039 rather than first to match comments.
2041 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
2043 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2046 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2049 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2050 check and use address of target always.
2052 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2055 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2056 (indirect_thunk_bnd_needed): Likewise.
2057 (indirect_thunks_used): Likewise.
2058 (indirect_thunks_bnd_used): Likewise.
2059 (indirect_return_needed): New.
2060 (indirect_return_bnd_needed): Likewise.
2061 (output_indirect_thunk_function): Add a bool argument for
2063 (output_indirect_thunk_function): Don't generate alias for
2064 function return thunk.
2065 (ix86_code_end): Call output_indirect_thunk_function to generate
2066 function return thunks.
2067 (ix86_output_function_return): Set indirect_return_bnd_needed
2068 and indirect_return_needed instead of indirect_thunk_bnd_needed
2069 and indirect_thunk_needed.
2071 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2073 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2074 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2075 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2077 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2078 Paul Hua <paul.hua.gm@gmail.com>
2081 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2083 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2085 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2086 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2089 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2092 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2093 instead of GET_MODE_SIZE when comparing Units.
2095 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2098 * varasm.c (hash_section): Return an unchangeble hash value
2099 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2100 Return !aarch64_can_use_per_function_literal_pools_p ().
2102 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2105 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2106 rather than address of cmode as last argument to prepare_cmp_insn.
2108 2018-03-15 Julia Koval <julia.koval@intel.com>
2110 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2111 F_AVX512VNNI, F_AVX512BITALG): New.
2113 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2116 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2117 insn for floating-point loads and stores.
2119 2018-03-14 Carl Love <cel@us.ibm.com>
2121 * config/rs6000/rs6000-c.c: Add macro definitions for
2122 ALTIVEC_BUILTIN_VEC_PERMXOR.
2123 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2124 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2125 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2126 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2128 * config/doc/extend.texi: Add prototypes for vec_permxor.
2130 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2133 * diagnostic-show-locus.c (class layout_point): Convert m_line
2134 from int to linenum_type.
2135 (line_span::comparator): Use linenum "compare" function when
2136 comparing line numbers.
2137 (test_line_span): New function.
2138 (layout_range::contains_point): Convert param "row" from int to
2140 (layout_range::intersects_line_p): Likewise.
2141 (layout::will_show_line_p): Likewise.
2142 (layout::print_source_line): Likewise.
2143 (layout::should_print_annotation_line_p): Likewise.
2144 (layout::print_annotation_line): Likewise.
2145 (layout::print_leading_fixits): Likewise.
2146 (layout::annotation_line_showed_range_p): Likewise.
2147 (struct line_corrections): Likewise for field m_row.
2148 (line_corrections::line_corrections): Likewise for param "row".
2149 (layout::print_trailing_fixits): Likewise.
2150 (layout::get_state_at_point): Likewise.
2151 (layout::get_x_bound_for_row): Likewise.
2152 (layout::print_line): Likewise.
2153 (diagnostic_show_locus): Likewise for locals "last_line" and
2155 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2156 * input.c (selftest::test_linenum_comparisons): New function.
2157 (selftest::input_c_tests): Call it.
2158 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2159 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2160 * selftest.h (ASSERT_GT): New macro.
2161 (ASSERT_GT_AT): New macro.
2162 (ASSERT_LT): New macro.
2163 (ASSERT_LT_AT): New macro.
2165 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
2167 PR rtl-optimization/84780
2168 * combine.c (distribute_links): Don't make a link based on pc_rtx.
2170 2018-03-14 Martin Liska <mliska@suse.cz>
2172 * tree.c (record_node_allocation_statistics): Use
2173 get_stats_node_kind.
2174 (get_stats_node_kind): New function extracted from
2175 record_node_allocation_statistics.
2176 (free_node): Use get_stats_node_kind.
2178 2018-03-14 Richard Biener <rguenther@suse.de>
2180 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2181 that the value-set of ANTIC_IN doesn't grow.
2184 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2186 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2187 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2189 2018-03-14 Julia Koval <julia.koval@intel.com>
2191 * config.gcc (icelake-client, icelake-server): New.
2193 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2194 (initial_ix86_arch_features): Ditto.
2195 (PTA_SKYLAKE): Add SGX.
2196 (PTA_ICELAKE): Remove.
2197 (PTA_ICELAKE_CLIENT): New.
2198 (PTA_ICELAKE_SERVER): New.
2199 (ix86_option_override_internal): Split up icelake on icelake client and
2201 (get_builtin_code_for_version): Ditto.
2202 (fold_builtin_cpu): Ditto.
2203 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2204 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2205 * config/i386/i386.h (processor_type): Ditto.
2206 * doc/invoke.texi: Ditto.
2208 2018-03-14 Jakub Jelinek <jakub@redhat.com>
2211 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2212 INTEGER_CST offset, add it together with bitpos / 8 and
2213 sign extend based on POINTER_SIZE.
2217 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
2220 * config/i386/constraints.md (Yc): New register constraint.
2221 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
2222 Use Yc constraint for alternative 2 of operand 0. Remove
2223 preferred_for_speed attribute.
2225 2018-03-14 Richard Biener <rguenther@suse.de>
2227 PR tree-optimization/84830
2228 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
2229 with the old one to avoid oscillations.
2231 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
2234 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
2236 (assign_by_spills): Return a flag of reload assignment failure.
2237 Do not process the reload assignment failures. Do not spill other
2238 reload pseudos if they has the same reg class. Update n if
2240 (lra_assign): Add a return arg. Set up from the result of
2241 assign_by_spills call.
2242 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2243 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2244 usage_insns if it is not NULL.
2245 (spill_hard_reg_in_range): New function.
2246 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2247 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2248 function prototypes.
2249 (lra_assign): Change prototype.
2250 * lra.c (lra): Add code to deal with fails by splitting hard reg
2253 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
2255 * config/riscv/riscv.opt (mrelax): New option.
2256 * config/riscv/riscv.c (riscv_file_start): Emit ".option
2257 "norelax" when riscv_mrelax is disabled.
2258 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2260 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2263 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2264 reassociation for int modes.
2266 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2268 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
2269 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
2271 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
2272 * config/aarch64/aarch64-sve.md
2273 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
2274 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2275 (*extend<mode><Vwide>2): Rename to...
2276 (aarch64_sve_extend<mode><Vwide>2): ...this.
2277 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2278 renaming the old pattern to...
2279 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
2281 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2282 define_expand, renaming the old pattern to...
2283 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2284 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2285 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2286 account when deciding which SVE instruction the optab should use.
2287 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2289 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2291 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2292 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2293 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2294 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2295 (tlsdesc_small_<mode>): Turn a define_expand and use
2296 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
2297 (tlsdesc_small_advsimd_<mode>): ...this.
2298 (tlsdesc_small_sve_<mode>): New pattern.
2300 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2302 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2303 (UNSPEC_UMUL_HIGHPART): New constants.
2304 (MUL_HIGHPART): New int iteraor.
2305 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2306 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2308 (*<su>mul<mode>3_highpart): New define_insn.
2310 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2313 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2316 2018-03-13 Martin Liska <mliska@suse.cz>
2319 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2321 (sem_item_optimizer::~sem_item_optimizer): Release it.
2322 (sem_item_optimizer::merge_classes): Register variable aliases.
2323 (sem_item_optimizer::fixup_pt_set): New function.
2324 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2325 * ipa-icf.h: Declare new variables and functions.
2327 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2330 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2331 integer_pow2p@2 and test integer_pow2p in condition.
2332 (A < 0 ? C : 0): Similarly for @1.
2335 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2336 characters starting at p contain '\0' character, don't look beyond
2340 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2341 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2344 * reg-stack.c (change_stack): Change update_end var from int to
2345 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2346 also call set_block_for_insn on the newly added insns and rescan.
2349 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2350 on the last operand.
2353 * tree.c (stabilize_reference_1): Return save_expr (e) for
2354 STATEMENT_LIST even if it doesn't have side-effects.
2356 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
2358 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2360 2018-03-12 Renlin Li <renlin.li@arm.com>
2362 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2363 aarch64_output_scalar_simd_mov_immediate.
2365 2018-03-12 Martin Sebor <msebor@redhat.com>
2367 PR tree-optimization/83456
2368 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2369 for perfectly overlapping calls to memcpy.
2370 (gimple_fold_builtin_memory_chk): Same.
2371 (gimple_fold_builtin_strcpy): Handle no-warning.
2372 (gimple_fold_builtin_stxcpy_chk): Same.
2373 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2375 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2377 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2378 parameter. Use it for SFmode.
2379 (rs6000_function_arg_advance_1): Adjust.
2380 (rs6000_function_arg): Adjust.
2381 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2383 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2385 PR rtl-optimization/84169
2386 PR rtl-optimization/84780
2387 * combine.c (can_combine_p): Check for a 2-insn combination whether
2388 the destination register is used between the two insns, too.
2390 2018-03-12 Richard Biener <rguenther@suse.de>
2392 PR tree-optimization/84803
2393 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2394 for refs DR analysis didn't process.
2396 2018-03-12 Richard Biener <rguenther@suse.de>
2398 PR tree-optimization/84777
2399 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2400 force-vectorize loops ignore whether we are optimizing for size.
2402 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
2404 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2405 (TARGET_MD_ASM_ADJUST): Define.
2407 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
2408 Kito Cheng <kito.cheng@gmail.com>
2409 Chung-Ju Wu <jasonwucj@gmail.com>
2411 * config/nds32/nds32.c (nds32_compute_stack_frame,
2412 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2413 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2414 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2415 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2416 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2417 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2418 * config/nds32/nds32.md (prologue, epilogue): Use macro
2419 NDS32_V3PUSH_AVAILABLE_P to do checking.
2421 2018-03-11 Jakub Jelinek <jakub@redhat.com>
2424 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
2425 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
2426 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
2427 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
2428 addition of most attributes on !orig_type_die or the attribute not
2429 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
2431 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2432 Chung-Ju Wu <jasonwucj@gmail.com>
2434 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
2436 * config/nds32/nds32.opt (mvh): New option.
2438 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2439 Chung-Ju Wu <jasonwucj@gmail.com>
2441 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
2443 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
2444 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
2447 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2448 Chung-Ju Wu <jasonwucj@gmail.com>
2450 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
2452 * config/nds32/nds32-multiple.md (strlensi): New pattern.
2453 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
2455 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
2456 Kito Cheng <kito.cheng@gmail.com>
2457 Chung-Ju Wu <jasonwucj@gmail.com>
2459 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
2460 UNSPEC_FFMISM and UNSPEC_FLMISM.
2461 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
2462 for ffb, ffmism and flmism.
2463 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
2464 (unspec_ffmism): Ditto.
2465 (unspec_flmism): Ditto.
2466 (nds32_expand_builtin_impl): Check if string extension is available.
2467 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
2468 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
2470 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
2473 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2476 * lra-assigns.c (assign_by_spills): Return a flag of reload
2477 assignment failure. Do not process the reload assignment
2478 failures. Do not spill other reload pseudos if they has the same
2480 (lra_assign): Add a return arg. Set up from the result of
2481 assign_by_spills call.
2482 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2483 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2484 usage_insns if it is not NULL.
2485 (spill_hard_reg_in_range): New function.
2486 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2487 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2488 function prototypes.
2489 (lra_assign): Change prototype.
2490 * lra.c (lra): Add code to deal with fails by splitting hard reg
2493 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
2496 * config/i386/i386.opt: Replace Enforcment with Enforcement.
2498 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
2501 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
2502 (dw_val_node): Add val_symbolic_view.
2503 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
2504 (symview_upper_bound): New.
2505 (new_line_info_table): Initialize symviews_since_reset.
2506 (dwarf2out_source_line): Count symviews_since_reset and set
2507 symview_upper_bound.
2508 (dw_val_equal_p): Handle symview.
2509 (add_AT_symview): New.
2510 (print_dw_val): Handle symview.
2511 (attr_checksum, attr_checksum_ordered): Likewise.
2512 (same_dw_val_p, size_of_die): Likewise.
2513 (value_format, output_die): Likewise.
2514 (add_high_low_attributes): Use add_AT_symview for entry_view.
2515 (dwarf2out_finish): Reset symview_upper_bound, clear
2518 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
2521 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
2522 Add strict argument and use it.
2523 (rs6000_split_multireg_move): Update for new strict argument.
2524 (mem_operand_gpr): Disallow all non-offsettable addresses.
2525 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
2527 2018-03-09 Jakub Jelinek <jakub@redhat.com>
2530 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
2531 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
2532 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
2535 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
2536 decl, use remap_type if we want to use the type.
2538 2018-03-09 Martin Sebor <msebor@redhat.com>
2540 PR tree-optimization/84526
2541 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2543 (builtin_access::generic_overlap): Be prepared to handle non-array
2546 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
2548 PR rtl-optimization/84682
2549 * lra-constraints.c (process_address_1): Check is_address flag
2550 for address constraints.
2551 (process_alt_operands): Likewise.
2552 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
2553 preprocess_constraints.
2554 * recog.h (preprocess_constraints): Add oploc parameter.
2556 * recog.c (preprocess_constraints): Test address_operand for
2557 CT_ADDRESS constraints.
2559 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2562 * lra-assigns.c (assign_by_spills): Return a flag of reload
2563 assignment failure. Do not process the reload assignment
2564 failures. Do not spill other reload pseudos if they has the same
2566 (lra_assign): Add a return arg. Set up from the result of
2567 assign_by_spills call.
2568 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2569 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2570 usage_insns if it is not NULL.
2571 (spill_hard_reg_in_range): New function.
2572 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2573 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2574 function prototypes.
2575 (lra_assign): Change prototype.
2576 * lra.c (lra): Add code to deal with fails by splitting hard reg
2579 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2582 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
2583 Accept complain bool parameter. Only emit errors if it is true.
2584 (arm_parse_cpu_option_name): Likewise.
2585 (arm_target_thumb_only): Adjust callers of the above.
2586 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
2587 prototype to take a default true bool parameter.
2588 (arm_parse_arch_option_name): Likewise.
2590 2018-03-09 David Malcolm <dmalcolm@redhat.com>
2591 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2595 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
2596 * configure: Regenerate.
2597 * configure.ac ("linker --version-script option"): New.
2598 ("linker soname option"): New.
2600 2018-03-09 Richard Biener <rguenther@suse.de>
2602 PR tree-optimization/84775
2603 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
2604 immediate uses of predicate stmts and mark them modified.
2607 PR tree-optimization/84178
2608 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2610 (version_loop_for_if_conversion): Delay update_ssa call.
2611 (tree_if_conversion): Delay update_ssa until after predicate
2614 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
2617 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
2618 when the function accesses prior frames.
2620 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2623 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
2624 gen_llsym, otherwise call maybe_gen_llsym.
2627 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
2628 has ',' character inside of it.
2630 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2633 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
2634 as clobbering CC_REGNUM.
2636 2018-03-08 Richard Biener <rguenther@suse.de>
2639 * tree-scalar-evolution.c: Include tree-into-ssa.h.
2640 (follow_copies_to_constant): Do not follow SSA names registered
2643 2018-03-08 Richard Biener <rguenther@suse.de>
2645 PR tree-optimization/84178
2646 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2648 (version_loop_for_if_conversion): Delay update_ssa call.
2649 (tree_if_conversion): Delay update_ssa until after predicate
2652 2018-03-08 David Malcolm <dmalcolm@redhat.com>
2654 PR tree-optimization/84178
2655 * tree-if-conv.c (release_bb_predicate): Remove the
2656 the assertion that the stmts have NULL use_ops.
2657 Discard the statements, asserting that they haven't
2658 yet been added to a BB.
2660 2018-03-08 Richard Biener <rguenther@suse.de>
2662 PR tree-optimization/84746
2663 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
2664 (phi_translate): Pass in destination ANTIC_OUT set.
2665 (phi_translate_1): Likewise. For a simplified result lookup
2666 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
2667 (phi_translate_set): Adjust.
2668 (do_pre_regular_insertion): Likewise.
2669 (do_pre_partial_partial_insertion): Likewise.
2671 2018-03-08 Martin Liska <mliska@suse.cz>
2673 PR gcov-profile/84735
2674 * doc/gcov.texi: Document usage of profile files.
2675 * gcov-io.h: Document changes in the format.
2677 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
2681 * dwarf2out.c (struct dw_line_info_table): Update comments for
2683 (FORCE_RESET_NEXT_VIEW): New.
2684 (FORCE_RESETTING_VIEW_P): New.
2685 (RESETTING_VIEW_P): Check for -1 too.
2686 (ZERO_VIEW_P): Likewise.
2687 (new_line_info_table): Force-reset next view.
2688 (dwarf2out_begin_function): Likewise.
2689 (dwarf2out_source_line): Simplify zero_view_p initialization.
2690 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
2691 view directly. Omit view when omitting .loc at line 0.
2693 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2695 PR tree-optimization/84740
2696 * tree-switch-conversion.c (process_switch): Call build_constructors
2697 only if info.phi_count is non-zero.
2699 PR tree-optimization/84739
2700 * tree-tailcall.c (find_tail_calls): Check call arguments against
2701 DECL_ARGUMENTS (current_function_decl) rather than
2702 DECL_ARGUMENTS (func) when checking for tail recursion.
2704 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2706 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
2707 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
2708 Volker Reichelt's entry and add entries for people that perform
2709 GCC fuzzy testing and report numerous bugs.
2711 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
2714 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
2715 readonly data in sdata, if that is disabled.
2716 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
2717 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2718 -mreadonly-in-sdata option.
2720 2018-03-07 Martin Sebor <msebor@redhat.com>
2722 PR tree-optimization/84468
2723 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
2724 basic block when looking for nul assignment.
2726 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
2729 * except.h (output_function_exception_table): Adjust prototype.
2730 * except.c (output_function_exception_table): Remove FNNAME parameter
2731 and add SECTION parameter. Ouput one part of the table at a time.
2732 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
2733 the first part of the exception table and emit unwind directives.
2734 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
2735 (i386_pe_seh_cold_init): Likewise.
2736 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
2737 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
2738 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
2739 (ix86_output_call_insn): Emit a nop in one more case for SEH.
2740 * config/i386/winnt.c: Include except.h.
2741 (struct seh_frame_state): Add reg_offset, after_prologue and
2742 in_cold_section fields.
2743 (i386_pe_seh_end_prologue): Set seh->after_prologue.
2744 (i386_pe_seh_cold_init): New function.
2745 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
2746 to seh->in_cold_section.
2747 (seh_emit_push): Record the offset of the push.
2748 (seh_emit_save): Record the offet of the save.
2749 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
2750 Test seh->after_prologue to disregard the epilogue.
2751 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
2752 (i386_pe_end_cold_function): New function.
2754 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2757 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
2758 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
2761 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
2763 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
2764 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
2767 * multiple_target.c: Include tree-inline.h and intl.h.
2768 (expand_target_clones): Diagnose and fail if node->definition and
2769 !tree_versionable_function_p (node->decl).
2771 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
2773 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
2775 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
2776 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
2777 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
2779 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2782 * combine.c (try_combine): Use reg_or_subregno instead of handling
2783 just paradoxical SUBREGs and REGs.
2785 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
2787 * config/arc/arc.c (arc_finalize_pic): Remove function.
2788 (arc_must_save_register): We use single base PIC register, remove
2789 checks to save/restore the PIC register.
2790 (arc_expand_prologue): Likewise.
2791 * config/arc/arc-protos.h (arc_set_default_type_attributes):
2793 (arc_verify_short): Likewise.
2794 (arc_attr_type): Likewise.
2795 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
2796 (walk_stores): Likewise.
2797 (arc_address_cost): Make it static.
2798 (arc_verify_short): Likewise.
2799 (branch_dest): Likewise.
2800 (arc_attr_type): Likewise.
2801 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
2802 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
2803 (arc_final_prescan_insn): Remove inserting the nops due to
2804 hardware hazards. It is done in reorg step.
2805 (insn_length_variant_t): Remove.
2806 (insn_length_parameters_t): Likewise.
2807 (arc_insn_length_parameters): Likewise.
2808 (arc_get_insn_variants): Likewise.
2809 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
2811 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2814 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
2817 PR tree-optimization/84687
2818 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
2820 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
2822 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2824 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
2825 Rename to ppc_speculation_barrier.
2826 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
2827 __builtin_ppc_speculation_barrier.
2829 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2832 * combine.c (combine_simplify_rtx): Don't try to simplify if
2833 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
2836 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
2838 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
2839 to 32 bytes when compiling for POWER9.
2841 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2844 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
2845 regparm >= 3 with no arg reg available also for calls with
2846 flag_force_indirect_call. Pass decl to ix86_function_regparm.
2849 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
2851 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
2853 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
2856 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
2858 2018-03-05 Richard Biener <rguenther@suse.de>
2860 PR tree-optimization/84486
2861 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
2862 When inserting a __builtin_assume_aligned call set the LHS
2863 SSA name alignment info accordingly.
2865 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
2867 PR tree-optimization/84114
2868 * config/aarch64/aarch64.c (aarch64_reassociation_width)
2869 Avoid reassociation of FLOAT_MODE addition.
2871 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
2873 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
2874 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
2875 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
2876 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
2877 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
2878 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
2879 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
2881 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
2882 __builtin_ia32_wbinvd): New builtins.
2883 (SPECIAL_ARGS2): New.
2884 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
2885 (SPECIAL_ARGS2): New.
2886 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
2887 (ix86_valid_target_attribute_inner_p): Ditto.
2888 (ix86_init_mmx_sse_builtins): Add special_args2.
2889 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
2890 TARGET_WBNOINVD_P): New.
2891 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
2892 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
2893 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
2894 * config/i386/immintrin.h (_wbinvd): New intrinsic.
2895 * config/i386/pconfigintrin.h: New file.
2896 * config/i386/wbnoinvdintrin.h: Ditto.
2897 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
2898 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
2900 2018-03-05 Richard Biener <rguenther@suse.de>
2902 PR tree-optimization/84670
2903 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2905 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2906 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2907 (compute_antic_aux): Only assert the number of values in ANTIC_IN
2908 doesn't grow if all successors (recursively) were visited at least
2911 2018-03-05 Richard Biener <rguenther@suse.de>
2913 PR tree-optimization/84650
2914 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
2915 if executed in the loop pipeline.
2917 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
2919 * doc/configfiles.texi (Configuration Files): Move info about
2920 conditionalizing $target-protos.h to...
2921 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
2922 differs from $target-protos.h.
2924 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
2925 Chung-Ju Wu <jasonwucj@gmail.com>
2927 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
2928 * config/nds32/nds32-multiple.md (setmemsi): Define.
2929 * config/nds32/nds32-memory-manipulation.c
2930 (nds32_gen_dup_4_byte_to_word_value): New.
2931 (emit_setmem_word_loop): New.
2932 (emit_setmem_byte_loop): New.
2933 (nds32_expand_setmem_loop): New.
2934 (nds32_expand_setmem_loop_v3m): New.
2935 (nds32_expand_setmem_unroll): New.
2936 (nds32_expand_setmem): New.
2938 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2939 Chung-Ju Wu <jasonwucj@gmail.com>
2941 * config/nds32/nds32-memory-manipulation.c
2942 (nds32_emit_load_store): New.
2943 (nds32_emit_post_inc_load_store): New.
2944 (nds32_emit_mem_move): New.
2945 (nds32_emit_mem_move_block): New.
2946 (nds32_expand_movmemsi_loop_unknown_size): New.
2947 (nds32_expand_movmemsi_loop_known_size): New.
2948 (nds32_expand_movmemsi_loop): New.
2949 (nds32_expand_movmemsi_unroll): New.
2950 (nds32_expand_movmemqi): Rename ...
2951 (nds32_expand_movmemsi): ... to this.
2952 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
2953 (movmemsi): ... to this.
2954 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
2955 (nds32_expand_movmemsi): ... to this.
2957 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2958 Monk Chiang <sh.chiang04@gmail.com>
2959 Chung-Ju Wu <jasonwucj@gmail.com>
2961 * config/nds32/nds32-protos.h
2962 (nds32_expand_load_multiple): New arguments.
2963 (nds32_expand_store_multiple): Ditto.
2964 (nds32_valid_multiple_load_store): Rename ...
2965 (nds32_valid_multiple_load_store_p): ... to this.
2966 * config/nds32/nds32-memory-manipulation.c
2967 (nds32_expand_load_multiple): Refine implementation.
2968 (nds32_expand_store_multiple): Ditto.
2969 * config/nds32/nds32-multiple.md
2970 (load_multiple): Update nds32_expand_load_multiple interface.
2971 (store_multiple): Update nds32_expand_store_multiple interface.
2972 * config/nds32/nds32-predicates.c
2973 (nds32_valid_multiple_load_store): Rename ...
2974 (nds32_valid_multiple_load_store_p): ... to this and refine
2976 * config/nds32/predicates.md
2977 (nds32_load_multiple_and_update_address_operation): New predicate.
2978 (nds32_store_multiple_and_update_address_operation): New predicate.
2980 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2981 Chung-Ju Wu <jasonwucj@gmail.com>
2983 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
2984 (combo): New attribute.
2985 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
2987 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
2989 * config/nds32/nds32.opt: Change -mcmodel= default value.
2991 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
2992 Monk Chiang <sh.chiang04@gmail.com>
2993 Chung-Ju Wu <jasonwucj@gmail.com>
2995 * config/nds32/constants.md (unspec_element): New enum.
2996 * config/nds32/constraints.md (Umw): New constraint.
2997 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
2998 * config/nds32/nds32-intrinsic.md: Likewise.
2999 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3000 (nds32_valid_smw_lwm_base_p): New.
3001 (nds32_output_smw_single_word): New.
3002 (nds32_output_lmw_single_word): New.
3003 (nds32_expand_unaligned_load): New.
3004 (nds32_expand_unaligned_store): New.
3005 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3006 (nds32_output_smw_single_word): Declare.
3007 (nds32_output_lmw_single_word): Declare.
3008 (nds32_expand_unaligned_load): Declare.
3009 (nds32_expand_unaligned_store): Declare.
3010 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3011 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3012 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3013 NDS32_BUILTIN_UASTORE_DW.
3014 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3017 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
3018 Kito Cheng <kito.cheng@gmail.com>
3019 Chung-Ju Wu <jasonwucj@gmail.com>
3021 * config/nds32/nds32-intrinsic.c
3022 (nds32_expand_builtin_null_ftype_reg): Delete.
3023 (nds32_expand_builtin_reg_ftype_imm): Ditto.
3024 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3025 (nds32_read_argument): New.
3026 (nds32_legitimize_target): Ditto.
3027 (nds32_legitimize_argument): Ditto.
3028 (nds32_check_constant_argument): Ditto.
3029 (nds32_expand_unop_builtin): Ditto.
3030 (nds32_expand_unopimm_builtin): Ditto.
3031 (nds32_expand_binop_builtin): Ditto.
3032 (nds32_builtin_decl_impl): Ditto.
3033 (builtin_description): Ditto.
3034 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3035 (nds32_init_builtins_impl): Ditto.
3036 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3037 (nds32_builtin_decl): New.
3038 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3039 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3041 2018-03-02 Jeff Law <law@redhat.com>
3043 * reorg.c (stop_search_p): Handle DEBUG_INSN.
3044 (redundant_insn, fill_simple_delay_slots): Likewise.
3045 (fill_slots_from_thread): Likewise.
3046 * resource.c (mark_referenced_resources): Likewise.
3047 (mark_set_resources, find_dead_or_set_registers): Likewise.
3049 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3051 * substring-locations.h (format_warning_va): Formatting fix for
3053 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3055 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3056 * substring-locations.c: Include intl.h.
3057 (format_warning_va): Turned into small wrapper around
3058 format_warning_n_va, renamed to ...
3059 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3060 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3062 (format_warning_at_substring_n): New function.
3063 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3064 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3065 format_warning_at_substring with just a shorter name instead of
3066 const function pointer.
3067 (fmtwarn_n): New function.
3068 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3069 appropriate, get rid of all the fmtstr temporaries, move conditionals
3070 with G_() wrapped string literals directly into fmtwarn arguments,
3071 cast dir.len to (int), formatting fixes.
3073 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3075 * doc/invoke.texi: Remove "Cilk Plus" references.
3077 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3078 Richard Biener <rguenther@suse.de>
3081 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3082 for error or warning attributes if CALL_FROM_THUNK_P is set.
3085 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3088 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3089 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3092 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3093 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3095 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3096 instead of -1U in last predictors element's probability member.
3098 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3101 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3102 arguments if they are comparable.
3104 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3106 PR tree-optimization/84634
3107 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3108 masks and masked_loop_p with a single loop_masks, making sure it's
3109 null for bb vectorization.
3111 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3113 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3114 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3115 loop->force_vectorize to check whether there is no alias.
3117 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3120 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3122 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3124 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3125 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3126 instead of a loop around prev_real_insn.
3127 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3131 * config/i386/i386.c (ix86_print_operand): Use conditional
3132 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3135 2018-03-02 Richard Biener <rguenther@suse.de>
3137 PR tree-optimization/84427
3138 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3139 (bitmap_set_subtract_values): Rewrite to handle multiple
3142 (prune_clobbered_mems): Likewise.
3143 (phi_translate): Take edge instead of pred/phiblock.
3144 (phi_translate_1): Likewise.
3145 (phi_translate_set): Likewise. Insert all translated
3146 exprs for a value into the set, keeping possibly multiple
3147 expressions per value.
3148 (compute_antic_aux): Adjust for phi_translate changes.
3149 When intersecting union the expressions and prune those
3150 not in the final value set, keeping possibly multiple
3151 expressions per value. Do not use value-insertion
3152 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3153 all expressions. Add verification that the value-sets
3154 only shrink during iteration.
3155 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3156 (do_pre_regular_insertion): Likewise.
3157 (do_pre_partial_partial_insertion): Likewise.
3159 2018-03-02 Richard Biener <rguenther@suse.de>
3162 * config/darwin.c (saved_debug_info_level): New static global.
3163 (darwin_asm_lto_start): Disable debug info generation for LTO out.
3164 (darwin_asm_lto_end): Restore debug info generation settings.
3166 2018-03-01 Martin Liska <mliska@suse.cz>
3169 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3172 2018-03-01 Richard Biener <rguenther@suse.de>
3175 * dwarf2out.c (gen_variable_die): Properly handle late VLA
3176 type annotation with LTO when debug was disabled at compile-time.
3178 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3180 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3182 (mips_final_postscan_insn): Likewise.
3184 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
3186 PR rtl-optimization/84528
3187 * alias.c (init_alias_target): Add commentary.
3188 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3189 a unique base value if the frame pointer is not eliminated
3190 to the stack pointer.
3192 2018-03-01 Tom de Vries <tom@codesourcery.com>
3194 PR rtl-optimization/83327
3195 * lra-int.h (hard_regs_spilled_into): Declare.
3196 * lra.c (hard_regs_spilled_into): Define.
3197 (init_reg_info): Init hard_regs_spilled_into.
3198 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3199 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3200 (process_bb_lives): Handle hard_regs_spilled_into.
3201 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3202 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3204 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3206 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3207 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3208 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3209 * config/rs6000/aix72.h: New file.
3211 2018-02-28 Jakub Jelinek <jakub@redhat.com>
3213 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3214 instead of warning_at with conditional singular and plural messages
3218 * stor-layout.c (update_alignment_for_field): For
3219 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
3220 && !DECL_PACKED (field), do the alignment update, just use
3221 only desired_align instead of MAX (type_align, desired_align)
3223 (place_field): Don't do known_align < desired_align handling
3224 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
3225 is non-NULL, instead do it after rli->prev_field handling and
3226 only if not within a bitfield word. For DECL_PACKED (field)
3227 use type_align of BITS_PER_UNIT.
3229 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
3231 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
3232 superfluous parentheses and trailing spaces.
3234 2018-02-28 Richard Biener <rguenther@suse.de>
3236 PR tree-optimization/84584
3237 * graphite-scop-detection.c (scop_detection::add_scop): Discard
3238 SCoPs with fake exit edge.
3240 2018-02-28 Martin Liska <mliska@suse.cz>
3243 * timevar.c (timer::print): Fix format to properly print 100%
3246 2018-02-28 Richard Biener <rguenther@suse.de>
3249 * genmatch.c (capture_info::walk_match): Do not mark
3250 captured expressions without operands as expr_p given
3251 they act more like predicates and should be subject to
3252 "lost tail" side-effect preserving.
3254 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
3256 PR rtl-optimization/81611
3257 * auto-inc-dec.c (attempt_change): Move dead note from
3258 mem_insn if it's the next use of regno
3259 (find_address): Take address use of reg holding
3260 non-incremented value. Add parm to limit search to the named
3262 (merge_in_block): Attempt to use a mem insn that is the next
3263 use of the original regno.
3265 2018-02-27 Martin Sebor <msebor@redhat.com>
3268 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
3269 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
3271 2018-02-27 Martin Sebor <msebor@redhat.com>
3273 PR translation/84207
3274 * diagnostic-core.h (warning_n, error_n, inform_n): Change
3275 n argument to unsigned HOST_WIDE_INT.
3276 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3277 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
3278 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3279 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3281 2018-02-27 Richard Biener <rguenther@suse.de>
3283 PR tree-optimization/84512
3284 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3285 Do not use the estimate returned from record_stmt_cost for
3286 the scalar iteration cost but sum properly using add_stmt_cost.
3288 2018-02-27 Richard Biener <rguenther@suse.de>
3290 PR tree-optimization/84466
3291 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3292 Adjust last change to less strictly validate use operands.
3294 2018-02-27 Martin Liska <mliska@suse.cz>
3296 PR gcov-profile/84548
3297 * gcov.c (process_file): Allow partial overlap and consider it
3298 also as group functions.
3299 (output_lines): Properly calculate range of lines for a group.
3301 2018-02-27 Martin Liska <mliska@suse.cz>
3303 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3304 'ggc' suffixes. Change first column width.
3305 (timer::print): Fix formatting of the column.
3307 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3309 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3310 preserve inline entry blocks for the sake of debug inline
3311 entry point markers alone.
3312 (remove_unused_locals): Suggest in comments a better place to
3313 force the preservation of inline entry blocks that are
3314 otherwise unused, but do not preserve them.
3316 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3318 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3320 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3323 * config/i386/constraints.md (Bs): Replace
3324 ix86_indirect_branch_register with
3325 TARGET_INDIRECT_BRANCH_REGISTER.
3327 * config/i386/i386.md (indirect_jump): Likewise.
3328 (tablejump): Likewise.
3329 (*sibcall_memory): Likewise.
3330 (*sibcall_value_memory): Likewise.
3331 Peepholes of indirect call and jump via memory: Likewise.
3332 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3333 (*sibcall_value_GOT_32): Likewise.
3334 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3335 (GOT_memory_operand): Likewise.
3336 (call_insn_operand): Likewise.
3337 (sibcall_insn_operand): Likewise.
3338 (GOT32_symbol_operand): Likewise.
3339 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3341 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3343 PR rtl-optimization/83496
3344 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3345 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3346 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3347 redundant insn, if any.
3348 (relax_delay_slots): Likewise.
3349 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3351 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
3353 PR tree-optimization/83965
3354 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3355 that grouped statements are part of a reduction chain. Return
3356 true if the statement is not marked as a reduction itself but
3358 (vect_recog_dot_prod_pattern): Don't check whether the statement
3359 is part of a group here.
3360 (vect_recog_sad_pattern): Likewise.
3361 (vect_recog_widen_sum_pattern): Likewise.
3363 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3366 * final.c (rest_of_clean_state): Also look for calls inside sequences.
3368 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3371 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3373 (ix86_output_indirect_function_return): New prototype.
3374 (ix86_split_simple_return_pop_internal): Likewise.
3375 * config/i386/i386.c (indirect_return_via_cx): New.
3376 (indirect_return_via_cx_bnd): Likewise.
3377 (indirect_thunk_name): Handle return va CX_REG.
3378 (output_indirect_thunk_function): Create alias for
3379 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3380 (ix86_output_indirect_jmp): Remove the bool argument.
3381 (ix86_output_indirect_function_return): New function.
3382 (ix86_split_simple_return_pop_internal): Likewise.
3383 * config/i386/i386.md (*indirect_jump): Don't pass false
3384 to ix86_output_indirect_jmp.
3385 (*tablejump_1): Likewise.
3386 (simple_return_pop_internal): Change it to define_insn_and_split.
3387 Call ix86_split_simple_return_pop_internal to split it for
3389 (simple_return_indirect_internal): Call
3390 ix86_output_indirect_function_return instead of
3391 ix86_output_indirect_jmp.
3393 2018-02-26 Jakub Jelinek <jakub@redhat.com>
3396 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3397 memset and value initialization afterwards.
3399 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
3401 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3403 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3406 * common/config/aarch64/aarch64-common.c
3407 (aarch_option_optimization_table[]): Switch
3408 off fomit-frame-pointer
3410 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3411 Chung-Ju Wu <jasonwucj@gmail.com>
3413 * config/nds32/nds32-multiple.md (load_multiple): Disallow
3415 (store_multiple): Ditto.
3417 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3419 * config.gcc: Add --with-cpu support for nds32 target.
3420 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
3421 * config/nds32/nds32.opt: Add -mcpu= option.
3423 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
3425 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
3426 isel=yes): Warn for these deprecated options.
3428 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
3430 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
3431 ISA_2_5_MASKS_EMBEDDED.
3433 2018-02-23 Jakub Jelinek <jakub@redhat.com>
3435 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
3436 p->max as pointers rather than using iterative_hash_expr.
3438 2018-02-23 Carl Love <cel@us.ibm.com>
3440 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
3441 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
3443 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
3444 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
3445 P8V_BUILTIN_VEC_VUNSIGNED2.
3447 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
3450 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
3451 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
3453 * lra-constraints.c (curr_insn_transform): Set up
3454 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
3456 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
3457 * lra-eliminations.c (spill_pseudos): Ditto.
3458 (process_insn_for_elimination): Ditto.
3459 * lra-lives.c (reg_early_clobber_p): Use the new macros.
3460 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
3461 LRA_NON_CLOBBERED_ALT.
3463 2018-02-22 Martin Sebor <msebor@redhat.com>
3465 PR tree-optimization/84480
3466 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
3467 to maybe_diag_stxncpy_trunc. Call it.
3468 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
3469 from gimple_fold_builtin_strcpy. Print inlining stack.
3470 (handle_builtin_stxncpy): Print inlining stack.
3471 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
3473 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
3476 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
3477 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
3478 and -fcheck-pointer-bounds are used together.
3479 (indirect_thunk_prefix): New enum.
3480 (indirect_thunk_need_prefix): New function.
3481 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
3482 "_nt" instead of "_bnd" for NOTRACK prefix.
3483 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
3484 (output_indirect_thunk_function): Likewise.
3486 (ix86_code_end): Update output_indirect_thunk_function calls.
3487 (ix86_output_indirect_branch_via_reg): Replace
3488 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
3489 (ix86_output_indirect_branch_via_push): Likewise.
3490 (ix86_output_function_return): Likewise.
3491 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
3492 incompatible with -fcf-protection=branch and
3493 -fcheck-pointer-bounds.
3495 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3498 * config/aarch64/aarch64.c (aarch64_print_address_internal):
3499 Change gcc_assert call to output_operand_lossage.
3501 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3503 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
3505 2018-02-22 DJ Delorie <dj@redhat.com>
3506 Sebastian Perta <sebastian.perta@renesas.com>
3507 Oleg Endo <olegendo@gcc.gnu.org>
3509 * config/rx/rx.c (rx_rtx_costs): New function.
3510 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
3512 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
3514 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
3516 2018-02-22 Martin Liska <mliska@suse.cz>
3519 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
3520 Add "native" as a possible value.
3522 2018-02-22 Martin Liska <mliska@suse.cz>
3525 * config/i386/i386.c (ix86_option_override_internal):
3526 Add "native" as a possible value for -march and -mtune.
3528 2018-02-22 Jakub Jelinek <jakub@redhat.com>
3531 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
3532 to all type variants.
3534 PR tree-optimization/84503
3535 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
3536 width as info->bitpos + info->bitsize - start.
3537 (merged_store_group::merge_overlapping): Simplify width computation.
3538 (check_no_overlap): New function.
3539 (imm_store_chain_info::try_coalesce_bswap): Compute expected
3540 start + width and last_order of the group, fail if check_no_overlap
3542 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
3543 to group if check_no_overlap fails.
3545 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3547 * config/rs6000/altivec.md: Delete contraint arguments to
3548 define_expand, define_split, and define_peephole2, and in
3549 define_insn_and_split if always unused.
3550 * config/rs6000/darwin.md: Ditto.
3551 * config/rs6000/dfp.md: Ditto.
3552 * config/rs6000/rs6000.md: Ditto.
3553 * config/rs6000/sync.md: Ditto.
3554 * config/rs6000/vector.md: Ditto.
3555 * config/rs6000/vsx.md: Ditto.
3557 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3559 * config/rs6000/altivec.md: Write output control strings as braced
3560 blocks instead of double-quoted strings.
3561 * config/rs6000/darwin.md: Ditto.
3562 * config/rs6000/rs6000.md: Ditto.
3563 * config/rs6000/vector.md: Ditto.
3564 * config/rs6000/vsx.md: Ditto.
3566 2018-02-21 Jason Merrill <jason@redhat.com>
3568 PR c++/84314 - ICE with templates and fastcall attribute.
3569 * attribs.c (build_type_attribute_qual_variant): Remove assert.
3571 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3573 * ipa-cp.c (determine_versionability): Fix comment typos.
3575 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3578 * ipa-cp.c (determine_versionability): Do not version functions caling
3581 2018-02-21 Martin Liska <mliska@suse.cz>
3584 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
3585 Add "native" as a possible value.
3586 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
3587 the macro when native cpu detection is available.
3589 2018-02-21 Martin Liska <mliska@suse.cz>
3592 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
3593 Add "native" as a possible value.
3594 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
3595 when native cpu detection is available.
3597 2018-02-21 Jakub Jelinek <jakub@redhat.com>
3598 Martin Sebor <msebor@redhat.com>
3600 PR tree-optimization/84478
3601 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
3603 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
3604 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
3605 support which is conservatively correct, for 2 only stay conservative
3606 for maxlen. Formatting and comment capitalization fixes. Add STRICT
3607 argument to the 2 argument get_range_strlen, adjust 6 arg
3608 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
3610 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
3611 (gimple_fold_builtin_strlen): Pass true as last argument to
3614 2018-02-20 Martin Sebor <msebor@redhat.com>
3617 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
3618 (builtin_memref::set_base_and_offset): Same. Handle inner references.
3619 (builtin_memref::builtin_memref): Factor out parts into
3620 set_base_and_offset and call it.
3622 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3625 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
3626 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
3627 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
3628 search at the associated MODE_INT.
3630 2018-02-20 Jeff Law <law@redhat.com>
3633 PR tree-optimization/81592
3635 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
3636 for range data rather than using global data.
3637 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
3638 range data rather than using global data.
3639 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
3640 pass it to children as needed.
3641 (struct directive::fmtresult): Similarly.
3642 (struct directive::set_width): Similarly.
3643 (struct directive::set_precision): Similarly.
3644 (format_integer, format_directive, parse_directive): Similarly.
3645 (format_none): Accept unnamed vr_values parameter.
3646 (format_percent, format_floating, format_character): Similarly.
3647 (format_string, format_plain): Similarly.
3648 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
3649 the EVRP range analyzer for range data rather than using global data.
3650 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
3651 gimple-ssa-evrp-analyze.h
3652 (class sprintf_dom_walker): Add after_dom_children member function.
3653 Add evrp_range_analyzer member.
3654 (sprintf_dom_walker::before_dom_children): Call into the EVRP
3655 range analyzer as needed.
3656 (sprintf_dom_walker::after_dom_children): New member function.
3657 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
3659 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
3660 (evrp_range_analyzer::pop_to_marker): Likewise.
3662 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3664 PR tree-optimization/84419
3665 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
3666 with the required type if its current type is compatible but
3669 2018-02-20 Jakub Jelinek <jakub@redhat.com>
3672 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
3673 after vectorization.
3675 2018-02-20 Martin Liska <mliska@suse.cz>
3678 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
3679 possible values if we don't have a hint.
3681 2018-02-20 Martin Liska <mliska@suse.cz>
3685 * final.c (shorten_branches): Build align_tab array with one
3687 * opts.c (finish_options): Add alignment option limit check.
3688 (MAX_CODE_ALIGN): Likewise.
3689 (MAX_CODE_ALIGN_VALUE): Likewise.
3690 * doc/invoke.texi: Document maximum allowed option value for
3691 all -falign-* options.
3693 2018-02-19 Jakub Jelinek <jakub@redhat.com>
3696 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
3697 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
3698 * var-tracking.c (emit_note_insn_var_location): Remove all references
3699 to NOTE_INSN_CALL_ARG_LOCATION.
3700 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
3701 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
3702 Use copy_rtx_if_shared.
3703 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
3704 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
3705 (dwarf2out_var_location): Remove handling of
3706 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
3708 * final.c (final_scan_insn): Remove all references to
3709 NOTE_INSN_CALL_ARG_LOCATION.
3710 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
3711 before dumping final insns.
3712 * except.c (emit_note_eh_region_end): Remove all references to
3713 NOTE_INSN_CALL_ARG_LOCATION.
3714 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
3715 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
3716 * config/arc/arc.c (hwloop_optimize): Likewise.
3717 * config/arm/arm.c (create_fix_barrier): Likewise.
3718 * config/s390/s390.c (s390_chunkify_start): Likewise.
3719 * config/sh/sh.c (find_barrier): Likewise.
3720 * config/i386/i386.c (rest_of_insert_endbranch,
3721 ix86_seh_fixup_eh_fallthru): Likewise.
3722 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
3723 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
3724 * config/frv/frv.c (frv_function_prologue): Likewise.
3725 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
3727 (note_outside_basic_block_p): Remove all references to
3728 NOTE_INSN_CALL_ARG_LOCATION.
3729 * gengtype.c (adjust_field_rtx_def): Likewise.
3730 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
3732 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
3733 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
3736 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
3739 PR tree-optimization/84452
3740 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
3741 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
3744 2018-02-19 Martin Liska <mliska@suse.cz>
3747 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
3749 2018-02-19 Martin Liska <mliska@suse.cz>
3750 Richard Sandiford <richard.sandiford@linaro.org>
3752 PR tree-optimization/82491
3753 * gimple-fold.c (get_base_constructor): Make earlier bail out
3756 2018-02-19 Carl Love <cel@us.ibm.com>
3758 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
3759 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
3761 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
3762 P8V_BUILTIN_VEC_NEG.
3764 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
3766 * config/rl78/rl78.md (movdf): New define expand.
3768 2018-02-19 Martin Liska <mliska@suse.cz>
3771 * doc/invoke.texi: Fix typo.
3772 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
3774 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
3776 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3777 handle rs6000_single_float and rs6000_double_float specially for
3780 2018-02-16 Jeff Law <law@redhat.com>
3782 * config/rx/rx.c (add_pop_cfi_notes): New function.;
3785 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3788 * ipa-inline.c (inline_small_functions): Fix a typo.
3790 2018-02-16 Nathan Sidwell <nathan@acm.org>
3792 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
3794 2018-02-16 Carl Love <cel@us.ibm.com>
3796 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
3797 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
3798 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
3799 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
3800 expansion to P8V_BUILTIN_VEC_FLOAT2.
3802 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
3804 PR rtl-optimization/70023
3805 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
3806 src_regno into account.
3808 2018-02-16 Carl Love <cel@us.ibm.com>
3810 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
3811 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
3812 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
3813 * config/rs6000/rs6000.c: Remove case statements for
3814 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
3815 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
3816 and P9V_BUILTIN_VEC_VINSERT4B.
3817 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
3818 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
3819 * config/rs6000/vsx.md:
3820 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
3823 2018-02-16 Carl Love <cel@us.ibm.com>
3825 * config/rs6000/altivec.h: Add builtin names vec_extract4b
3827 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
3829 * config/rs6000/rs6000-c.c: Add the definitions for
3830 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
3831 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
3832 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
3833 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
3834 definition for insert4b and define insn *insert3b_internal.
3835 * doc/extend.texi: Add documentation for vec_extract4b.
3837 2018-02-16 Nathan Sidwell <nathan@acm.org>
3839 * doc/extend.texi (Backwards Compatibility): Mention friend
3840 injection. Note for-scope is deprecated.
3841 * doc/invoke.texi (-ffriend-injection): Deprecate.
3843 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
3845 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
3846 that moved to I2, also allow destinations that are a paradoxical
3847 subreg (instead of a normal reg).
3849 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
3852 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
3855 2018-02-16 Richard Biener <rguenther@suse.de>
3857 PR tree-optimization/84037
3858 PR tree-optimization/84016
3860 * config/i386/i386.c (ix86_builtin_vectorization_cost):
3861 Adjust vec_construct for the fact we need additional higher latency
3862 128bit inserts for AVX256 and AVX512 vector builds.
3863 (ix86_add_stmt_cost): Scale vector construction cost for
3866 2018-02-16 Richard Biener <rguenther@suse.de>
3868 PR tree-optimization/84417
3869 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
3870 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
3871 (non_rewritable_lvalue_p): Likewise, use poly-ints.
3873 2018-02-16 Martin Liska <mliska@suse.cz>
3876 * internal-fn.def (ASAN_CHECK): Set proper flags.
3877 (ASAN_MARK): Likewise.
3879 2018-02-16 Julia Koval <julia.koval@intel.com>
3881 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
3882 from PTA_CANNONLAKE.
3884 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3887 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
3888 Use ++iter rather than iter++ for std::list iterators.
3889 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
3890 defer deleting them until all nodes in the forest are processed. Do
3891 free even leaf nodes. Change to_process into auto_vec.
3894 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
3895 * vec.h (vec_default_construct): Use memset instead of placement new
3896 if BROKEN_VALUE_INITIALIZATION is defined.
3897 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
3898 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
3901 PR rtl-optimization/83723
3902 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
3903 * lra.c (lra_substitute_pseudo): Likewise. If true, use
3904 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
3906 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
3908 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
3910 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
3912 PR rtl-optimization/81443
3913 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
3914 from inner REGs to paradoxical SUBREGs.
3916 2018-02-16 Richard Biener <rguenther@suse.de>
3918 PR tree-optimization/84399
3919 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3920 For operands we can analyze at their definition make sure we can
3921 analyze them at each use as well.
3923 2018-02-16 Richard Biener <rguenther@suse.de>
3925 PR tree-optimization/84190
3926 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
3927 volatile accesses if the decl isn't volatile.
3929 2018-02-15 Jason Merrill <jason@redhat.com>
3931 PR c++/84314 - ICE with templates and fastcall attribute.
3932 * attribs.c (build_type_attribute_qual_variant): Don't clobber
3933 TYPE_CANONICAL on an existing type.
3935 2018-02-15 Jakub Jelinek <jakub@redhat.com>
3937 PR tree-optimization/84383
3938 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
3939 dstoff nor call operand_equal_p if dstbase is NULL.
3941 PR tree-optimization/84334
3942 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
3943 also a CONSTANT_CLASS_P, punt.
3945 2018-02-14 Jim Wilson <jimw@sifive.com>
3947 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
3948 first SMALL_OPERAND check. New local min_second_step. Move assert
3949 to where locals are set. Add TARGET_RVC support.
3950 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
3952 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
3954 * doc/invoke.texi: Correct -Wformat-overflow code sample.
3956 2018-02-14 Martin Sebor <msebor@redhat.com>
3958 PR tree-optimization/83698
3959 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
3960 arrays constrain the offset range to their bounds.
3961 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
3962 (builtin_access::overlap): Avoid setting the size of overlap if it's
3964 (maybe_diag_overlap): Also consider arrays when deciding what values
3965 of offsets to include in diagnostics.
3967 2018-02-14 Martin Sebor <msebor@redhat.com>
3970 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
3971 that correspond to the kind of a declaration.
3973 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
3976 * config/pa/pa.md: Load address of PIC label using the linkage table
3977 if the label is nonlocal.
3979 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
3981 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
3982 warning message if user requests -maltivec=be.
3983 * doc/invoke.texi: Document deprecation of -maltivec=be.
3985 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
3988 * config/rs6000/rs6000-c.c: Update definitions for
3989 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
3990 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
3992 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3995 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
3996 add _get_ssp intrinsics. Remove argument from
3997 __builtin_ia32_rdssp[d|q].
3998 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
3999 * config/i386/i386-builtin.def: Remove argument from
4000 __builtin_ia32_rdssp[d|q].
4001 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4002 ix86_expand_special_args_builtin for _rdssp[d|q].
4003 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4004 Clear register before usage.
4005 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4006 Add documentation for new _get_ssp and _inc_ssp intrinsics.
4008 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
4010 PR tree-optimization/84357
4011 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4012 operand 1 of an ARRAY_REF too.
4014 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
4017 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4018 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4020 (set_of_reg): New struct.
4021 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4022 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4023 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4025 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4026 Split into bitclr, bitset, bitinvert patterns if appropriate.
4027 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4028 use rx_fuse_in_memory_bitop.
4029 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4030 to named insn, correct maximum insn length.
4032 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
4035 * machmode.def: Define a complex mode for PARTIAL_INT.
4036 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4038 * doc/rtl.texi: Document CSPImode.
4039 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4041 (msp430_hard_regno_nregs_with_padding): Likewise.
4043 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4046 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4048 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4050 PR rtl-optimization/84169
4051 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4052 we generated a parallel as new i3 and we split that to new i2 and i3
4053 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4054 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4055 those to i2, not i1. Partially rewrite this scan code.
4057 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4060 * stor-layout.c (place_field): For variable length fields, adjust
4061 offset_align afterwards not just based on the field's alignment,
4062 but also on the size.
4065 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4066 of exps and logs in the use_exp2 case.
4068 2018-02-13 Jeff Law <law@redhat.com>
4070 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4073 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4076 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4080 * common.opt (gas-loc-support, gas-locview-support): New.
4081 (ginline-points, ginternal-reset-location-views): New.
4082 * doc/invoke.texi: Document them. Use @itemx where intended.
4083 (gvariable-location-views): Adjust.
4084 * target.def (reset_location_view): New.
4085 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4086 (TARGET_RESET_LOCATION_VIEW): New.
4087 * doc/tm.texi: Rebuilt.
4088 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4089 (dwarf2out_default_as_locview_support): New.
4090 (output_asm_line_debug_info): Use option variables.
4091 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4092 (output_loc_list): Likewise.
4093 (add_high_low_attributes): Check option variables.
4094 Don't output entry view attribute in strict mode.
4095 (gen_inlined_subroutine_die): Check option variables.
4096 (dwarf2out_inline_entry): Likewise.
4097 (init_sections_and_labels): Likewise.
4098 (dwarf2out_early_finish): Likewise.
4099 (maybe_reset_location_view): New, from...
4100 (dwarf2out_var_location): ... here. Call it.
4101 * debug.h (dwarf2out_default_as_loc_support): Declare.
4102 (dwarf2out_default_as_locview_support): Declare.
4103 * hooks.c (hook_int_rtx_insn_0): New.
4104 * hooks.h (hook_int_rtx_insn_0): Declare.
4105 * toplev.c (process_options): Take -gas-loc-support and
4106 -gas-locview-support from dwarf2out. Enable
4107 -gvariable-location-views by default only with locview
4108 assembler support. Enable -ginternal-reset-location-views by
4109 default only if the target defines the corresponding hook.
4110 Enable -ginline-points by default if location views are
4111 enabled; force it disabled if statement frontiers are
4113 * tree-inline.c (expand_call_inline): Check option variables.
4114 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4116 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4118 PR tree-optimization/84321
4119 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4120 handling. Also check whether the anti-range contains any values
4121 that satisfy the mask; switch to a VR_RANGE if not.
4123 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4126 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4128 2018-02-13 Martin Jambor <mjambor@suse.cz>
4131 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4132 of call statements, also set location of a load to a temporary.
4134 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4136 * config/rl78/rl78.c (add_vector_labels): New function.
4137 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4138 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4139 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4140 which checks that no arguments are passed.
4141 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4142 * doc/extend.texi: Documentation for the new attribute.
4144 2018-02-13 Andreas Schwab <schwab@suse.de>
4146 * config/riscv/linux.h (CPP_SPEC): Define.
4148 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4151 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4152 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4153 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4154 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4155 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4156 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4157 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4158 temporarily for AES and PCLMUL builtins.
4160 PR tree-optimization/84339
4161 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4162 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4166 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4167 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4168 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4170 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4172 * omp-simd-clone.h: New file.
4173 * omp-simd-clone.c: Include omp-simd-clone.h.
4174 (expand_simd_clones): No longer static.
4175 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4176 cgraph.h and omp-simd-clone.h.
4177 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4178 (vect_recog_widen_shift_pattern): Formatting fix.
4179 (vect_pattern_recog_1): Don't check optab for calls.
4182 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4183 operands[2] into a REG before using gen_lowpart on it.
4185 2018-02-12 Jeff Law <law@redhat.com>
4188 * config/sh/sh.c (find_barrier): Consider a sibling call
4191 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4192 successfully back substituting a reg.
4194 2018-02-12 Richard Biener <rguenther@suse.de>
4196 PR tree-optimization/84037
4197 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4198 parameter, move visited init to caller.
4199 (vect_slp_analyze_operations): Separate cost from validity
4200 check, initialize visited once for all instances.
4201 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4203 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4205 (vect_model_promotion_demotion_cost): Likewise.
4206 (vectorizable_bswap): Guard cost modeling with !slp_node
4207 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4209 (vectorizable_call): Likewise.
4210 (vectorizable_conversion): Likewise.
4211 (vectorizable_assignment): Likewise.
4212 (vectorizable_shift): Likewise.
4213 (vectorizable_operation): Likewise.
4214 (vectorizable_store): Likewise.
4215 (vectorizable_load): Likewise.
4216 (vectorizable_condition): Likewise.
4217 (vectorizable_comparison): Likewise.
4219 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
4222 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
4223 (ASAN_MARK): Fix fnspec to account for return value, change pointer
4224 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
4226 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4229 * params.def (inline-min-speedup): Increase from 8 to 15.
4230 (max-inline-insns-auto): Decrease from 40 to 30.
4231 * ipa-split.c (consider_split): Add some buffer for function to
4232 be considered inlining candidate.
4233 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
4236 2018-02-12 Richard Biener <rguenther@suse.de>
4238 PR tree-optimization/84037
4239 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
4240 matched stmts if we cannot swap the non-matched ones.
4242 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4244 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
4245 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4246 _mm_maskz_scalef_round_ss): New intrinsics.
4247 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4248 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4249 __builtin_ia32_scalefss_round): Remove.
4250 (__builtin_ia32_scalefsd_mask_round,
4251 __builtin_ia32_scalefss_mask_round): New intrinsics.
4252 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4253 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4254 ((match_operand:VF_128 2 "<round_nimm_predicate>"
4255 "<round_constraint>")): Changed to ...
4256 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4257 "<round_scalar_constraint>")): ... this.
4258 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4259 %0, %1, %2<round_op3>}"): Changed to ...
4260 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4261 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4262 %2<round_scalar_mask_op3>}"): ... this.
4263 * config/i386/subst.md (round_scalar_nimm_predicate): New.
4265 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4267 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4268 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
4269 (_mm_maskz_sqrt_round_ss): New intrinsics.
4270 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
4271 (__builtin_ia32_sqrtsd_mask_round)
4272 (__builtin_ia32_sqrtss_mask_round): New builtins.
4273 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
4274 (__builtin_ia32_sqrtss_round): Remove.
4275 (__builtin_ia32_sqrtsd_mask_round)
4276 (__builtin_ia32_sqrtss_mask_round): New builtins.
4277 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4278 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4279 ((match_operand:VF_128 1 "vector_operand"
4280 "xBm,<round_constraint>")): Changed to ...
4281 ((match_operand:VF_128 1 "vector_operand"
4282 "xBm,<round_scalar_constraint>")): ... this.
4283 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4284 %0, %2, %<iptr>1<round_op3>}): Changed to ...
4285 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4286 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4287 %<iptr>1<round_scalar_mask_op3>}): ... this.
4288 ((set_attr "prefix" "<round_prefix>")): Changed to ...
4289 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4291 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
4294 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4295 Cast vec_cmpeq result to correct type.
4296 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4297 Cast vec_cmpgt result to correct type.
4299 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4301 * final.c (final_scan_insn_1): Renamed from...
4302 (final_scan_insn): ... this. New wrapper, to recover
4303 seen from the outermost call in recursive ones.
4304 * config/sparc/sparc.c (output_return): Drop seen from call.
4305 (output_sibcall): Likewise.
4306 * config/visium/visium.c (output_branch): Likewise.
4308 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4310 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4313 2018-02-10 Alan Modra <amodra@gmail.com>
4316 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4317 Specify LR as an input.
4319 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4322 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4323 remove_member_access_dummy_vars): New functions.
4324 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4325 lower_omp_1, execute_lower_omp): Use them.
4327 PR rtl-optimization/84308
4328 * shrink-wrap.c (spread_components): Release todo vector.
4330 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4332 PR rtl-optimization/57193
4333 * ira-color.c (struct allocno_color_data): Add member
4334 conflict_allocno_hard_prefs.
4335 (update_conflict_allocno_hard_prefs): New.
4336 (bucket_allocno_compare_func): Add a preference based on
4337 conflict_allocno_hard_prefs.
4338 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4339 (color_allocnos): Remove a dead code. Initiate
4340 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4342 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4345 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4346 constraint from =wa to wa. Avoid a subreg on the output operand,
4347 instead use a pseudo and subreg it in a move.
4348 (p9_xxbrd_<mode>): Changed to ...
4349 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4350 (p9_xxbrd_v2df): New expander.
4351 (p9_xxbrw_<mode>): Changed to ...
4352 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4353 (p9_xxbrw_v4sf): New expander.
4355 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4357 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4359 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
4362 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4363 multiply in 32-bit mode.
4364 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4365 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4368 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4370 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4371 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4372 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4373 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4375 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4378 * dwarf2out.c (is_trivial_indirect_ref): New function.
4379 (dwarf2out_late_global_decl): Do not generate a location
4380 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4381 and that are not defined in the current unit.
4383 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4385 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4386 instead of a libcall for UNORDERED.
4388 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4391 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4392 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4394 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4397 * config/s390/s390.c (s390_set_current_function): Invoke
4398 s390_indirect_branch_settings also if fndecl didn't change.
4400 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4402 * config/rs6000/rs6000.md (blockage): Set length to zero.
4404 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4406 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4408 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4411 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4412 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4416 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4417 PARALLEL incoming that failed vt_get_decl_and_offset check.
4420 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
4421 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
4423 (get_variable_section): For decls in named .bss* sections pass true as
4424 second argument to bss_initializer_p.
4426 2018-02-09 Marek Polacek <polacek@redhat.com>
4427 Jakub Jelinek <jakub@redhat.com>
4430 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
4431 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
4432 Sync some changes from cxx_fold_indirect_ref.
4434 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4436 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
4438 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
4439 (BLOCK_INLINE_ENTRY_LABEL): New.
4440 (dwarf2out_var_location): Disregard inline entry markers.
4441 (inline_entry_data): New struct.
4442 (inline_entry_data_hasher): New hashtable type.
4443 (inline_entry_data_hasher::hash): New.
4444 (inline_entry_data_hasher::equal): New.
4445 (inline_entry_data_table): New variable.
4446 (add_high_low_attributes): Add DW_AT_entry_pc and
4447 DW_AT_GNU_entry_view attributes if a pending entry is found
4448 in inline_entry_data_table. Add old entry_pc attribute only
4449 if debug nonbinding markers are disabled.
4450 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
4451 markers are enabled.
4452 (block_within_block_p, dwarf2out_inline_entry): New.
4453 (dwarf2out_finish): Check that no entries remained in
4454 inline_entry_data_table.
4455 * final.c (reemit_insn_block_notes): Handle inline entry notes.
4456 (final_scan_insn, notice_source_line): Likewise.
4457 (rest_of_clean_state): Skip inline entry markers.
4458 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
4460 * gimple.c (gimple_build_debug_inline_entry): New.
4461 * gimple.h (enum gimple_debug_subcode): Add
4462 GIMPLE_DEBUG_INLINE_ENTRY.
4463 (gimple_build_debug_inline_entry): Declare.
4464 (gimple_debug_inline_entry_p): New.
4465 (gimple_debug_nonbind_marker_p): Adjust.
4466 * insn-notes.def (INLINE_ENTRY): New.
4467 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4468 inline entry marker notes.
4469 (print_insn): Likewise.
4470 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
4471 (INSN_DEBUG_MARKER_KIND): Likewise.
4472 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
4473 * tree-inline.c (expand_call_inline): Build and insert
4474 debug_inline_entry stmt.
4475 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
4476 inline entry blocks early, if nonbind markers are enabled.
4477 (dump_scope_block): Dump fragment info.
4478 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
4479 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
4480 (gimple_build_debug_inline_entry): New.
4481 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
4482 Enable/disable inline entry points too.
4483 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
4484 (DEBUG_INSN): Describe inline entry markers.
4486 * common.opt (gvariable-location-views): New.
4487 (gvariable-location-views=incompat5): New.
4488 * config.in: Rebuilt.
4489 * configure: Rebuilt.
4490 * configure.ac: Test assembler for view support.
4491 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
4492 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
4493 * dwarf2out.c (var_loc_view): New typedef.
4494 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
4495 (dwarf2out_locviews_in_attribute): New.
4496 (dwarf2out_locviews_in_loclist): New.
4497 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
4498 (enum dw_line_info_opcode): Add LI_adv_address.
4499 (struct dw_line_info_table): Add view.
4500 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
4501 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
4502 (zero_view_p): New variable.
4503 (ZERO_VIEW_P): New macro.
4504 (output_asm_line_debug_info): New.
4505 (struct var_loc_node): Add view.
4506 (add_AT_view_list, AT_loc_list): New.
4507 (add_var_loc_to_decl): Add view param. Test it against last.
4508 (new_loc_list): Add view params. Record them.
4509 (AT_loc_list_ptr): Handle loc and view lists.
4510 (view_list_to_loc_list_val_node): New.
4511 (print_dw_val): Handle dw_val_class_view_list.
4512 (size_of_die): Likewise.
4513 (value_format): Likewise.
4514 (loc_list_has_views): New.
4515 (gen_llsym): Set vl_symbol too.
4516 (maybe_gen_llsym, skip_loc_list_entry): New.
4517 (dwarf2out_maybe_output_loclist_view_pair): New.
4518 (output_loc_list): Output view list or entries too.
4519 (output_view_list_offset): New.
4520 (output_die): Handle dw_val_class_view_list.
4521 (output_dwarf_version): New.
4522 (output_compilation_unit_header): Use it.
4523 (output_skeleton_debug_sections): Likewise.
4524 (output_rnglists, output_line_info): Likewise.
4525 (output_pubnames, output_aranges): Update version comments.
4526 (output_one_line_info_table): Output view numbers in asm comments.
4527 (dw_loc_list): Determine current endview, pass it to new_loc_list.
4528 Call maybe_gen_llsym.
4529 (loc_list_from_tree_1): Adjust.
4530 (add_AT_location_description): Create view list attribute if
4531 needed, check it's absent otherwise.
4532 (convert_cfa_to_fb_loc_list): Adjust.
4533 (maybe_emit_file): Call output_asm_line_debug_info for test.
4534 (dwarf2out_var_location): Reset views as needed. Precompute
4535 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
4536 attribute. Set view.
4537 (new_line_info_table): Reset next view.
4538 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
4539 (dwarf2out_source_line): Likewise. Output view resets and labels to
4540 the assembler, or select appropriate line info opcodes.
4541 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
4542 (optimize_string_length): Catch it. Adjust.
4543 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
4544 dw_val_class_view_list, and remove it if no longer needed.
4545 (hash_loc_list): Hash view numbers.
4546 (loc_list_hasher::equal): Compare them.
4547 (optimize_location_lists): Check whether a view list symbol is
4548 needed, and whether the locview attribute is present, and
4549 whether they match. Remove the locview attribute if no longer
4551 (index_location_lists): Call skip_loc_list_entry for test.
4552 (dwarf2out_finish): Call output_asm_line_debug_info for test.
4553 Use output_dwarf_version.
4554 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
4555 (struct dw_val_node): Add val_view_list.
4556 * final.c (SEEN_NEXT_VIEW): New.
4557 (set_next_view_needed): New.
4558 (clear_next_view_needed): New.
4559 (maybe_output_next_view): New.
4560 (final_start_function): Rename to...
4561 (final_start_function_1): ... this. Take pointer to FIRST,
4562 add SEEN parameter. Emit param bindings in the initial view.
4563 (final_start_function): Reintroduce SEEN-less interface.
4564 (final): Rename to...
4565 (final_1): ... this. Take SEEN parameter. Output final pending
4566 next view at the end.
4567 (final): Reintroduce seen-less interface.
4568 (final_scan_insn): Output pending next view before switching
4569 sections or ending a block. Mark the next view as needed when
4570 outputting variable locations. Notify debug backend of section
4571 changes, and of location view changes.
4572 (rest_of_handle_final): Adjust.
4573 * toplev.c (process_options): Autodetect value for debug variable
4574 location views option. Warn on incompat5 without -gdwarf-5.
4575 * doc/invoke.texi (gvariable-location-views): New.
4576 (gvariable-location-views=incompat5): New.
4577 (gno-variable-location-views): New.
4579 2018-02-08 David Malcolm <dmalcolm@redhat.com>
4581 PR tree-optimization/84136
4582 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
4583 that the result of find_edge is non-NULL.
4585 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4588 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
4589 storing integer register in SImode. Fix cost of 256 and 512
4590 byte aligned SSE register store.
4592 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4594 * config/i386/i386.c (ix86_multiplication_cost): Fix
4595 multiplication cost for TARGET_AVX512DQ.
4597 2018-02-08 Marek Polacek <polacek@redhat.com>
4599 PR tree-optimization/84238
4600 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
4603 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4605 PR tree-optimization/84265
4606 * tree-vect-stmts.c (vectorizable_store): Don't treat
4607 VMAT_CONTIGUOUS accesses as grouped.
4608 (vectorizable_load): Likewise.
4610 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4612 PR tree-optimization/81635
4613 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
4614 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
4615 (test_round_for_mask): New functions.
4616 (wide_int_cc_tests): Call test_round_for_mask.
4617 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
4618 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
4619 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
4620 range returned by get_range_info.
4622 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4625 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
4626 * symtab.c: Include builtins.h
4627 (symtab_node::output_to_lto_symbol_table_p): Move here
4628 from lto-streamer-out.c:output_symbol_p.
4629 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
4630 (output_symbol_p): Move all logic to symtab.c
4631 (produce_symtab): Update.
4633 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4635 * config/s390/s390-opts.h (enum indirect_branch): Define.
4636 * config/s390/s390-protos.h (s390_return_addr_from_memory)
4637 (s390_indirect_branch_via_thunk)
4638 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
4639 (enum s390_indirect_branch_type): Define.
4640 * config/s390/s390.c (struct s390_frame_layout, struct
4641 machine_function): Remove.
4642 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
4643 (indirect_branch_table_label_no, indirect_branch_table_name):
4645 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
4646 (enum s390_indirect_branch_option): Define.
4647 (s390_return_addr_from_memory): New function.
4648 (s390_handle_string_attribute): New function.
4649 (s390_attribute_table): Add new attribute handler.
4650 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
4651 (s390_indirect_branch_via_thunk): New function.
4652 (s390_indirect_branch_via_inline_thunk): New function.
4653 (s390_function_ok_for_sibcall): When jumping via thunk disallow
4654 sibling call optimization for non z10 compiles.
4655 (s390_emit_call): Force indirect branch target to be a single
4656 register. Add r1 clobber for non-z10 compiles.
4657 (s390_emit_epilogue): Emit return jump via return_use expander.
4658 (s390_reorg): Handle JUMP_INSNs as execute targets.
4659 (s390_option_override_internal): Perform validity checks for the
4660 new command line options.
4661 (s390_indirect_branch_attrvalue): New function.
4662 (s390_indirect_branch_settings): New function.
4663 (s390_set_current_function): Invoke s390_indirect_branch_settings.
4664 (s390_output_indirect_thunk_function): New function.
4665 (s390_code_end): Implement target hook.
4666 (s390_case_values_threshold): Implement target hook.
4667 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
4669 * config/s390/s390.h (struct s390_frame_layout)
4670 (struct machine_function): Move here from s390.c.
4671 (TARGET_INDIRECT_BRANCH_NOBP_RET)
4672 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
4673 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
4674 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
4675 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
4676 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
4677 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
4678 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
4679 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
4680 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
4681 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
4682 (mnemonic attribute): Add values which aren't recognized
4684 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
4685 pattern for branch conversion. Fix mnemonic attribute.
4686 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
4687 indirect branch via thunk if requested.
4688 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
4689 ("*indirect_jump"): Disable for branch conversion using out of
4691 ("indirect_jump_via_thunk<mode>_z10")
4692 ("indirect_jump_via_thunk<mode>")
4693 ("indirect_jump_via_inlinethunk<mode>_z10")
4694 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
4695 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
4696 ("casesi_jump_via_inlinethunk<mode>_z10")
4697 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
4698 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
4699 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
4700 ("*indirect2_jump"): Disable for branch conversion.
4701 ("casesi_jump"): Turn into expander and expand patterns for branch
4703 ("return_use"): New expander.
4704 ("*return"): Emit return via thunk and rename it to ...
4705 ("*return<mode>"): ... this one.
4706 * config/s390/s390.opt: Add new options and and enum for the
4709 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4711 * lra-constraints.c (match_reload): Unconditionally use
4712 gen_lowpart_SUBREG, rather than selecting between that
4713 and equivalent gen_rtx_SUBREG code.
4715 2018-02-08 Richard Biener <rguenther@suse.de>
4717 PR tree-optimization/84233
4718 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
4719 changed flag instead of boguously re-using phi_inserted.
4721 2018-02-08 Martin Jambor <mjambor@suse.cz>
4723 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
4724 static local variables.
4726 2018-02-08 Richard Biener <rguenther@suse.de>
4728 PR tree-optimization/84278
4729 * tree-vect-stmts.c (vectorizable_store): When looking for
4730 smaller vector types to perform grouped strided loads/stores
4731 make sure the mode is supported by the target.
4732 (vectorizable_load): Likewise.
4734 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4736 * config/aarch64/aarch64.c (aarch64_components_for_bb):
4737 Increase LDP/STP opportunities by adding adjacent callee-saves.
4739 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4741 PR rtl-optimization/84068
4742 PR rtl-optimization/83459
4743 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
4745 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
4747 PR tree-optimization/84224
4748 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
4749 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
4752 2018-02-07 Iain Sandoe <iain@codesourcery.com>
4755 * config/rs6000/altivec.md (*restore_world): Remove LR use.
4756 * config/rs6000/predicates.md (restore_world_operation): Adjust op
4757 count, remove one USE.
4759 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
4761 * doc/install.texi (Configuration): Document the
4762 --with-long-double-format={ibm,ieee} PowerPC configuration
4766 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
4767 Convert from define_expand to be define_insn_and_split. Rework
4768 float/double/_Float128 conversions to QI/HI/SImode to work with
4769 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
4770 conversions to QI/HImode types did a store and then a load to
4771 truncate the value. For conversions to VSX registers, don't split
4772 the insn, instead emit the code directly. Use the code iterator
4773 any_fix to combine signed and unsigned conversions.
4774 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
4775 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
4776 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4777 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
4778 (fix_<mode>di2_hw): Likewise.
4779 (fixuns_<mode>di2_hw): Likewise.
4780 (fix_<mode>si2_hw): Likewise.
4781 (fixuns_<mode>si2_hw): Likewise.
4782 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
4783 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4784 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
4785 fix<uns>_trunc<SFDF:mode>si2_p8.
4786 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
4788 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
4789 (fix<uns>_<mode>_mem): Likewise.
4790 (fctiw<u>z_<mode>_mem): Likewise.
4791 (fix<uns>_<mode>_mem): Likewise.
4792 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
4793 the register allocator from doing a direct move to the GPRs to do
4794 a store, and instead use the ISA 3.0 store byte/half-word from
4795 vector register instruction. For IEEE 128-bit floating point,
4796 also optimize stores of 32-bit ints.
4797 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
4799 2018-02-07 Alan Hayward <alan.hayward@arm.com>
4801 * genextract.c (push_pathstr_operand): New function to support
4803 (walk_rtx): Call push_pathstr_operand.
4804 (print_path): Support [a-zA-Z].
4806 2018-02-07 Richard Biener <rguenther@suse.de>
4808 PR tree-optimization/84037
4809 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
4810 (cse_and_gimplify_to_preheader): Declare.
4811 (vect_get_place_in_interleaving_chain): Likewise.
4812 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4814 (_loop_vec_info::~_loop_vec_info): Delete it.
4815 (cse_and_gimplify_to_preheader): New function.
4816 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
4817 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
4818 (vectorizable_load): Likewise. For grouped stores always base
4819 the IV on the first element.
4820 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
4821 condition before gimplifying.
4823 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4825 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
4826 *DIV_EXPR and *MOD_EXPR.
4828 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
4831 * config/i386/i386.c (ix86_option_override_internal): Mask out
4832 the CF_SET bit when checking -fcf-protection.
4834 2018-02-07 Tom de Vries <tom@codesourcery.com>
4837 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
4840 2018-02-07 Richard Biener <rguenther@suse.de>
4842 PR tree-optimization/84204
4843 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4846 PR tree-optimization/84205
4847 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
4848 special-case isl_ast_op_zdiv_r.
4850 PR tree-optimization/84223
4851 * graphite-scop-detection.c (gather_bbs::before_dom_children):
4852 Only add conditions from within the region.
4853 (gather_bbs::after_dom_children): Adjust.
4855 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
4858 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
4859 * config/avr/avr.md: Only post-reload split REG-REG moves if
4860 either register is GENERAL_REG_P.
4862 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4864 PR tree-optimization/84235
4865 * tree-ssa-scopedtables.c
4866 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
4867 if the subtraction is performed in floating point type where NaNs are
4868 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
4869 build 1. Formatting fix.
4871 2018-02-06 Jakub Jelinek <jakub@redhat.com>
4874 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
4875 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
4876 and skip it regardless of bb boundaries. Use CALL_P macro,
4877 don't test INSN_P (insn) together with CALL_P or JUMP_P check
4878 unnecessarily, formatting fix.
4880 2018-02-06 Michael Collison <michael.collison@arm.com>
4882 * config/arm/thumb2.md:
4883 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
4884 (*thumb_mov_notscc): Ditto.
4886 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
4889 * config/rs6000/rs6000.md (su code attribute): Use "u" for
4890 unsigned_fix, not "s".
4892 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4894 * configure.ac (gcc_fn_eh_frame_ro): New function.
4895 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
4896 correct .eh_frame permissions.
4897 * configure: Regenerate.
4899 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
4901 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
4904 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4906 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4907 Display warning message for -mno-speculate-indirect-jumps.
4909 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
4911 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
4913 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
4915 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
4917 PR tree-optimization/84225
4918 * tree-eh.c (find_trapping_overflow): Only call
4919 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
4921 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4924 * config/i386/i386.c: Reimplement the check of possible options
4925 -mibt/-mshstk conbination. Change error messages.
4926 * doc/invoke.texi: Fix a typo: remove extra '='.
4928 2018-02-06 Marek Polacek <polacek@redhat.com>
4930 PR tree-optimization/84228
4931 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
4933 2018-02-06 Tamar Christina <tamar.christina@arm.com>
4936 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
4937 emitted arch directives.
4938 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
4939 __ARM_FEATURE_COPROC before changing architectures.
4941 2018-02-06 Richard Biener <rguenther@suse.de>
4943 * config/i386/i386.c (print_reg): Fix typo.
4944 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
4946 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
4948 * configure: Regenerate.
4950 2018-02-05 Martin Sebor <msebor@redhat.com>
4952 PR tree-optimization/83369
4953 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
4956 2018-02-05 Martin Liska <mliska@suse.cz>
4958 * doc/invoke.texi: Cherry-pick upstream r323995.
4960 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
4962 * ira.c (ira_init_register_move_cost): Adjust comment.
4964 2018-02-05 Martin Liska <mliska@suse.cz>
4966 PR gcov-profile/84137
4967 * doc/gcov.texi: Fix typo in documentation.
4969 2018-02-05 Martin Liska <mliska@suse.cz>
4971 PR gcov-profile/83879
4972 * doc/gcov.texi: Document necessity of --dynamic-list-data when
4973 using dlopen functionality.
4975 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
4977 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
4978 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
4979 _mm_maskz_range_ss, _mm_mask_range_round_ss,
4980 _mm_maskz_range_round_ss): New intrinsics.
4981 (__builtin_ia32_rangesd128_round)
4982 (__builtin_ia32_rangess128_round): Remove.
4983 (__builtin_ia32_rangesd128_mask_round,
4984 __builtin_ia32_rangess128_mask_round): New builtins.
4985 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
4986 __builtin_ia32_rangess128_round): Remove.
4987 (__builtin_ia32_rangesd128_mask_round,
4988 __builtin_ia32_rangess128_mask_round): New builtins.
4989 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
4990 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
4991 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
4992 "<round_saeonly_constraint>")): Changed to ...
4993 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
4994 "<round_saeonly_scalar_constraint>")): ... this.
4995 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
4996 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
4997 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
4998 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
4999 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5001 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
5003 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5005 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5006 Remove all values except native, 8540 and 8548.
5008 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
5010 * config/i386/i386.c (ix86_output_function_return): Pass
5011 INVALID_REGNUM, instead of -1, as invalid register number to
5012 indirect_thunk_name and output_indirect_thunk.
5014 2018-02-02 Julia Koval <julia.koval@intel.com>
5016 * config.gcc: Add -march=icelake.
5017 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5018 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5019 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5020 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5021 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5022 (processor_target_table): Add icelake.
5023 (ix86_option_override_internal): Handle new PTAs.
5024 (get_builtin_code_for_version): Handle icelake.
5025 (M_INTEL_COREI7_ICELAKE): New.
5026 (fold_builtin_cpu): Handle icelake.
5027 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5028 * doc/invoke.texi: Add -march=icelake.
5030 2018-02-02 Julia Koval <julia.koval@intel.com>
5032 * config/i386/i386.c (ix86_option_override_internal): Change flags type
5033 to wide_int_bitmask.
5034 * wide-int-bitmask.h: New.
5036 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5039 * config/i386/i386.md: Replace Pmode with word_mode in
5040 builtin_setjmp_setup and builtin_longjmp to support x32.
5042 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5046 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5048 (rs6000_supported_cpu_names): New static variable.
5049 (linux_cpu_translation_table): Likewise.
5050 (elf_platform) <cpu>: Define new static variable and use it.
5051 Translate kernel AT_PLATFORM name to canonical name if needed.
5052 Error if platform name is unknown.
5054 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
5057 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5059 2018-02-01 Jeff Law <law@redhat.com>
5062 * config/i386/i386.c (release_scratch_register_on_entry): Add new
5063 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
5064 the scratch if RELEASE_VIA_POP is false.
5065 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5066 If we have to save a temporary register, decrement SIZE appropriately.
5067 Pass new arguments to release_scratch_register_on_entry.
5068 (ix86_adjust_stack_and_probe): Likewise.
5069 (ix86_emit_probe_stack_range): Pass new arguments to
5070 release_scratch_register_on_entry.
5072 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5074 PR rtl-optimization/84157
5075 * combine.c (change_zero_ext): Use REG_P predicate in
5076 front of HARD_REGISTER_P predicate.
5078 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5080 * config/avr/avr.c (avr_option_override): Move disabling of
5081 -fdelete-null-pointer-checks to...
5082 * common/config/avr/avr-common.c (avr_option_optimization_table):
5085 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5087 PR tree-optimization/81635
5088 * tree-data-ref.c (split_constant_offset_1): For types that
5089 wrap on overflow, try to use range info to prove that wrapping
5092 2018-02-01 Renlin Li <renlin.li@arm.com>
5095 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5097 (aarch64_register_move_cost): Likewise.
5098 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5100 (REG_CLASS_NAMES): Likewise.
5101 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5102 TAILCALL_ADDR_REGS. Remove IP registers.
5103 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5105 2018-02-01 Richard Biener <rguenther@suse.de>
5107 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5108 for specifying RPO order and allow NULL for that.
5109 * domwalk.c (dom_walker::dom_walker): Likewise.
5110 (dom_walker::walk): Handle NULL RPO order.
5111 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5113 (rewrite_update_dom_walker): Likewise.
5114 (mark_def_dom_walker): Likewise.
5116 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5118 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5119 (aarch64_maybe_expand_sve_subreg_move): Declare.
5120 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5121 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5123 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5124 that are semantically a reverse operation.
5125 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5126 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5127 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5129 (aarch64_can_change_mode_class): For big-endian, forbid changes
5130 between two SVE modes if they have different element sizes.
5132 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5134 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5135 the TImode handling for big-endian targets.
5137 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5139 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5140 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5142 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5143 Remove BSWAP handing for big-endian targets and use the form of
5144 LD1RQ appropariate for the mode.
5146 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5148 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5149 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5152 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5155 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5156 check for operands that need to go through aarch64_sve_reload_be.
5158 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5160 PR tree-optimization/81661
5161 PR tree-optimization/84117
5162 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5163 * tree-eh.c: Include gimplify.h.
5164 (find_trapping_overflow, replace_trapping_overflow,
5165 rewrite_to_non_trapping_overflow): New functions.
5166 * tree-vect-loop.c: Include tree-eh.h.
5167 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5168 * tree-data-ref.c: Include tree-eh.h.
5169 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5171 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
5173 PR rtl-optimization/84123
5174 * combine.c (change_zero_ext): Check if hard register satisfies
5175 can_change_dest_mode before calling gen_lowpart_SUBREG.
5177 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
5180 * ira.c (ira_init_register_move_cost): Remove assert.
5182 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5184 PR rtl-optimization/84071
5185 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5186 * doc/tm.texi: Regenerate.
5188 2018-01-31 Richard Biener <rguenther@suse.de>
5190 PR tree-optimization/84132
5191 * tree-data-ref.c (analyze_miv_subscript): Properly
5192 check whether evolution_function_is_affine_multivariate_p
5193 before calling gcd_of_steps_may_divide_p.
5195 2018-01-31 Julia Koval <julia.koval@intel.com>
5198 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5199 * config/i386/i386.md (rdpid_rex64) New.
5200 (rdpid): Make 32bit only.
5202 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5205 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5206 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5208 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5211 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5213 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5215 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5217 PR rtl-optimization/84071
5218 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
5219 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
5221 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5223 * config/arc/arc.c (arc_handle_aux_attribute): New function.
5224 (arc_attribute_table): Add 'aux' attribute.
5225 (arc_in_small_data_p): Consider aux like variables.
5226 (arc_is_aux_reg_p): New function.
5227 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
5228 (arc_get_aux_arg): New function.
5229 (prepare_move_operands): Handle aux-register access.
5230 (arc_handle_aux_attribute): New function.
5231 * doc/extend.texi (ARC Variable attributes): Add subsection.
5233 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5235 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
5236 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
5237 (arc_attribute_table): Add 'uncached' attribute.
5238 (arc_print_operand): Print '.di' flag for uncached memory
5240 (arc_in_small_data_p): Do not consider for small data the uncached
5242 (arc_is_uncached_mem_p): New function.
5243 * config/arc/predicates.md (compact_store_memory_operand): Check
5244 for uncached memory accesses.
5245 (nonvol_nonimm_operand): Likewise.
5246 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5248 2018-01-31 Jakub Jelinek <jakub@redhat.com>
5251 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5252 falign-loops=): Add Optimization flag.
5254 2018-01-30 Jeff Law <law@redhat.com>
5257 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5258 INT_REGISTERS_SAVED. Check it prior to calling
5259 get_scratch_register_on_entry.
5260 (ix86_adjust_stack_and_probe): Similarly.
5261 (ix86_emit_probe_stack_range): Similarly.
5262 (ix86_expand_prologue): Corresponding changes.
5264 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5267 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5268 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
5270 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
5273 * lra-constraints.c (curr_insn_transform): Process AND in the
5276 2018-01-30 Jakub Jelinek <jakub@redhat.com>
5278 PR rtl-optimization/83986
5279 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5280 dependence against last_pending_memory_flush in addition to
5283 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
5285 PR tree-optimization/81611
5286 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5289 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5292 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5295 2018-01-30 Richard Biener <rguenther@suse.de>
5296 Jakub Jelinek <jakub@redhat.com>
5298 PR tree-optimization/84111
5299 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5300 inner loops added during recursion, as they don't have up-to-date
5303 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5306 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5307 (can_inline_edge_by_limits_p): ... here.
5308 (can_early_inline_edge_p, check_callers,
5309 update_caller_keys, update_callee_keys, recursive_inlining,
5310 add_new_edges_to_heap, speculation_useful_p,
5311 inline_small_functions,
5312 inline_small_functions, flatten_function,
5313 inline_to_all_callers_1): Update.
5315 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5317 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5320 2018-01-30 Richard Biener <rguenther@suse.de>
5322 PR tree-optimization/83008
5323 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5324 invariant and constant vector uses in stmts when they need
5327 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5330 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5331 * configure: Regenerate.
5333 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5335 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5337 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5338 Use gen_rtx_REG rather than gen_lowpart.
5340 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5342 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5343 rather than 0 when creating partial subregs.
5345 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5347 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5350 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
5353 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5354 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5355 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5356 flags. This restores the settings used before the 2017-07-24.
5357 Turning off pre increment/decrement/modify allows IVOPTS to
5358 optimize DF/SF loops where the index is an int.
5360 2018-01-29 Richard Biener <rguenther@suse.de>
5361 Kelvin Nilsen <kelvin@gcc.gnu.org>
5364 * tree-vect-stmts.c (vectorizable_call): Don't call
5365 targetm.vectorize_builtin_md_vectorized_function if callee is
5368 2018-01-22 Carl Love <cel@us.ibm.com>
5370 * doc/extend.tex: Fix typo in second arg in
5371 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5373 2018-01-29 Richard Biener <rguenther@suse.de>
5375 PR tree-optimization/84086
5376 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5377 (flush_ssaname_freelist): When SSA names were released reset
5378 the SCEV hash table.
5380 2018-01-29 Richard Biener <rguenther@suse.de>
5382 PR tree-optimization/84057
5383 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5384 removed paths when removing edges.
5386 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
5388 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5389 -mfunction-return=@var{choice}.
5391 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
5394 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5396 (layout::print_source_line): Likewise.
5397 (test_get_line_width_without_trailing_whitespace): Add test cases.
5399 2018-01-27 Jakub Jelinek <jakub@redhat.com>
5402 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5405 2018-01-26 Jim Wilson <jimw@sifive.com>
5407 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5409 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5412 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5414 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5415 and CMP + SUB-immediate -> SUBS.
5417 2018-01-26 Martin Sebor <msebor@redhat.com>
5419 PR tree-optimization/83896
5420 * tree-ssa-strlen.c (get_string_len): Rename...
5421 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
5422 Avoid assuming length is constant.
5423 (handle_char_store): Use HOST_WIDE_INT for string length.
5425 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
5428 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
5429 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
5431 2018-01-26 Richard Biener <rguenther@suse.de>
5433 PR rtl-optimization/84003
5434 * dse.c (record_store): Only record redundant stores when
5435 the earlier store aliases at least all accesses the later one does.
5437 2018-01-26 Jakub Jelinek <jakub@redhat.com>
5439 PR rtl-optimization/83985
5440 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
5441 REG_CFA_RESTORE insns.
5442 (delete_unmarked_insns): Don't ignore separate shrink wrapping
5443 REG_CFA_RESTORE insns here.
5446 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
5447 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
5449 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5451 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
5452 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
5453 (arc_init): Likewise.
5454 (arc_override_options): Likewise.
5455 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
5457 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
5459 * config/arc/arc.h (TARGET_DBNZ): Define.
5460 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
5461 properly set the tune attribute.
5462 (dbnz): Use TARGET_DBNZ guard.
5463 * config/arc/arc.opt (mtune): Add core3 option.
5465 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5467 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
5468 recognize new pic like addresses.
5469 (arc_delegitimize_address): Clean up.
5471 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5473 * config/arc/arc-arches.def: Option mrf16 valid for all
5475 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
5476 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
5477 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
5478 * config/arc/arc-tables.opt: Regenerate.
5479 * config/arc/arc.c (arc_conditional_register_usage): Handle
5480 reduced register file case.
5481 (arc_file_start): Set must have build attributes.
5482 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
5484 * config/arc/arc.opt (mrf16): Add new option.
5485 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
5486 * config/arc/genmultilib.awk: Handle new mrf16 option.
5487 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
5488 * config/arc/t-multilib: Regenerate.
5489 * doc/invoke.texi (ARC Options): Document mrf16 option.
5491 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5493 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
5494 * config/arc/arc.c (arc_handle_secure_attribute): New function.
5495 (arc_attribute_table): Add 'secure_call' attribute.
5496 (arc_print_operand): Print secure call operand.
5497 (arc_function_ok_for_sibcall): Don't optimize tail calls when
5499 (arc_is_secure_call_p): New function. * config/arc/arc.md
5500 (call_i): Add support for sjli instruction.
5501 (call_value_i): Likewise.
5502 * config/arc/constraints.md (Csc): New constraint.
5504 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5505 John Eric Martin <John.Martin@emmicro-us.com>
5507 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
5508 * config/arc/arc.c (_arc_jli_section): New struct.
5509 (arc_jli_section): New type.
5510 (rc_jli_sections): New static variable.
5511 (arc_handle_jli_attribute): New function.
5512 (arc_attribute_table): Add jli_always and jli_fixed attribute.
5513 (arc_file_end): New function.
5514 (TARGET_ASM_FILE_END): Define.
5515 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
5516 (arc_add_jli_section): New function.
5517 (jli_call_scan): Likewise.
5518 (arc_reorg): Call jli_call_scan.
5519 (arc_output_addsi): Remove 'S' from printing asm operand.
5520 (arc_is_jli_call_p): New function.
5521 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
5523 (movhi_insn): Likewise.
5524 (movsi_insn): Likewise.
5525 (movsi_set_cc_insn): Likewise.
5526 (loadqi_update): Likewise.
5527 (load_zeroextendqisi_update): Likewise.
5528 (load_signextendqisi_update): Likewise.
5529 (loadhi_update): Likewise.
5530 (load_zeroextendhisi_update): Likewise.
5531 (load_signextendhisi_update): Likewise.
5532 (loadsi_update): Likewise.
5533 (loadsf_update): Likewise.
5534 (movsicc_insn): Likewise.
5535 (bset_insn): Likewise.
5536 (bxor_insn): Likewise.
5537 (bclr_insn): Likewise.
5538 (bmsk_insn): Likewise.
5539 (bicsi3_insn): Likewise.
5540 (cmpsi_cc_c_insn): Likewise.
5541 (movsi_ne): Likewise.
5542 (movsi_cond_exec): Likewise.
5543 (clrsbsi2): Likewise.
5554 (call_i): Remove 'S' asm letter, add jli instruction.
5555 (call_value_i): Likewise.
5556 * config/arc/arc.op (mjli-always): New option.
5557 * config/arc/constraints.md (Cji): New constraint.
5558 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
5560 (subsf3_fpx): Likewise.
5561 (mulsf3_fpx): Likewise.
5562 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
5564 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
5566 * doc/invoke.texi (ARC): Document mjli-always option.
5568 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
5570 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
5571 avoid addition with 0 and use incw and decw where possible.
5573 2018-01-26 Richard Biener <rguenther@suse.de>
5575 PR tree-optimization/81082
5576 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
5577 association if it requires casting to unsigned.
5578 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
5579 from fold_plusminus_mult_expr to catch important cases late when
5580 range info is available.
5582 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5584 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
5585 * configure.ac (hidden_linkonce): New test.
5586 * configure: Regenerate.
5587 * config.in: Regenerate.
5589 2018-01-26 Julia Koval <julia.koval@intel.com>
5591 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
5592 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
5593 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
5594 _mm_mask_bitshuffle_epi64_mask): Fix type.
5595 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
5596 USI_FTYPE_V4DI_V4DI_USI): Remove.
5597 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
5598 __builtin_ia32_vpshufbitqmb256_mask,
5599 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
5600 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
5601 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
5603 2018-01-26 Alan Modra <amodra@gmail.com>
5606 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
5607 UNSPEC_VBPERMQ. Sort other unspecs.
5609 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
5611 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
5613 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
5616 * predict.c (drop_profile): Do not push/pop cfun; update also
5618 (handle_missing_profiles): Fix logic looking for zero profiles.
5620 2018-01-25 Jakub Jelinek <jakub@redhat.com>
5623 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
5624 on functions with #pragma omp declare simd or functions with simd
5626 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
5627 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5628 Remove trailing \n from warning_at calls.
5630 2018-01-25 Tom de Vries <tom@codesourcery.com>
5633 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5634 for neutered workers.
5636 2018-01-24 Joseph Myers <joseph@codesourcery.com>
5639 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
5640 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
5642 2018-01-24 Jeff Law <law@redhat.com>
5645 * i386.c (get_probe_interval): Move to earlier point.
5646 (ix86_compute_frame_layout): If -fstack-clash-protection and
5647 the frame is larger than the probe interval, then use pushes
5648 to save registers rather than reg->mem moves.
5649 (ix86_expand_prologue): Remove conditional for int_registers_saved
5652 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
5655 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
5656 min/max for never referenced object.
5658 2018-01-24 Jakub Jelinek <jakub@redhat.com>
5661 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
5663 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
5664 attributes from DECL_ATTRIBUTES (decl) without affecting
5665 DECL_ATTRIBUTES (current_function_decl).
5666 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
5667 functions with non-NULL DECL_ABSTRACT_ORIGIN.
5669 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
5671 PR tree-optimization/83979
5672 * fold-const.c (fold_comparison): Use constant_boolean_node
5673 instead of boolean_{true,false}_node.
5675 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
5677 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
5680 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5682 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5683 Simplify the clause that sets the length attribute.
5684 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5685 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
5686 clause that sets the length attribute.
5687 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5689 2018-01-24 Tom de Vries <tom@codesourcery.com>
5692 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
5693 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
5694 Add strict parameter.
5695 (prevent_branch_around_nothing): Insert dummy insn between branch to
5696 label and label with no ptx insn inbetween.
5697 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
5699 2018-01-24 Tom de Vries <tom@codesourcery.com>
5702 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5703 for neutered threads in warp.
5704 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
5706 2018-01-24 Richard Biener <rguenther@suse.de>
5708 PR tree-optimization/83176
5709 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
5712 2018-01-24 Richard Biener <rguenther@suse.de>
5714 PR tree-optimization/82819
5715 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
5716 code generating pluses that are no-ops in the target precision.
5718 2018-01-24 Richard Biener <rguenther@suse.de>
5721 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
5723 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5725 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
5726 to merge probabilities.
5727 * predict.c (probably_never_executed): Also mark as cold functions
5728 with global 0 profile and guessed local profile.
5729 * profile-count.c (profile_probability::combine_with_count): New
5731 * profile-count.h (profile_probability::operator*,
5732 profile_probability::operator*=, profile_probability::operator/,
5733 profile_probability::operator/=): Reduce precision to adjusted
5734 and set value to guessed on contradictory divisions.
5735 (profile_probability::combine_with_freq): Remove.
5736 (profile_probability::combine_wiht_count): Declare.
5737 (profile_count::force_nonzero):: Set to adjusted.
5738 (profile_count::probability_in):: Set quality to adjusted.
5739 * tree-ssa-tail-merge.c (replace_block_by): Use
5742 2018-01-23 Andrew Waterman <andrew@sifive.com>
5743 Jim Wilson <jimw@sifive.com>
5745 * config/riscv/riscv.c (riscv_stack_boundary): New.
5746 (riscv_option_override): Set riscv_stack_boundary. Handle
5747 riscv_preferred_stack_boundary_arg.
5748 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
5749 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
5750 (STACK_BOUNDARY): Set to riscv_stack_boundary.
5751 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
5752 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
5753 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
5755 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
5758 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
5759 of struct ix86_frame.
5760 (ix86_expand_epilogue): Likewise. Add a local variable for
5761 the reg_save_offset field in struct ix86_frame.
5763 2018-01-23 Bin Cheng <bin.cheng@arm.com>
5765 PR tree-optimization/82604
5766 * tree-loop-distribution.c (enum partition_kind): New enum item
5767 PKIND_PARTIAL_MEMSET.
5768 (partition_builtin_p): Support above new enum item.
5769 (generate_code_for_partition): Ditto.
5770 (compute_access_range): Differentiate cases that equality can be
5771 proven at all loops, the innermost loops or no loops.
5772 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
5773 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
5774 (finalize_partitions, distribute_loop): Don't fuse partition of
5775 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
5776 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
5779 2018-01-23 Martin Liska <mliska@suse.cz>
5781 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
5782 order to ignore the predictor.
5783 (PRED_POLYMORPHIC_CALL): Likewise.
5784 (PRED_RECURSIVE_CALL): Likewise.
5786 2018-01-23 Martin Liska <mliska@suse.cz>
5788 * tree-profile.c (tree_profiling): Print function header to
5789 aware reader which function we are working on.
5790 * value-prof.c (gimple_find_values_to_profile): Do not print
5791 not interesting value histograms.
5793 2018-01-23 Martin Liska <mliska@suse.cz>
5795 * profile-count.h (enum profile_quality): Add
5796 profile_uninitialized as the first value. Do not number values
5797 as they are zero based.
5798 (profile_count::verify): Update sanity check.
5799 (profile_probability::verify): Likewise.
5801 2018-01-23 Nathan Sidwell <nathan@acm.org>
5803 * doc/invoke.texi (ffor-scope): Deprecate.
5805 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5807 PR tree-optimization/83510
5808 * domwalk.c (set_all_edges_as_executable): New function.
5809 (dom_walker::dom_walker): Convert bool param
5810 "skip_unreachable_blocks" to enum reachability. Move setup of
5811 edge flags to set_all_edges_as_executable and only do it when
5812 reachability is REACHABLE_BLOCKS.
5813 * domwalk.h (enum dom_walker::reachability): New enum.
5814 (dom_walker::dom_walker): Convert bool param
5815 "skip_unreachable_blocks" to enum reachability.
5816 (set_all_edges_as_executable): New decl.
5817 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
5818 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
5820 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
5821 but converting true to REACHABLE_BLOCKS.
5822 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
5824 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
5825 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
5826 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
5828 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
5829 if check_all_array_refs will be called.
5831 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5833 * tree.c (selftest::test_location_wrappers): Add more test
5836 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5838 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
5839 (selftest::test_bit_in_range): Likewise.
5841 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5844 * doc/sourcebuild.texi (vect_float): Say that the selector
5845 only describes the situation when -funsafe-math-optimizations is on.
5846 (vect_float_strict): Document.
5848 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5850 PR tree-optimization/83965
5851 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5852 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
5853 instead of checking only for a reduction.
5854 (vect_recog_widen_sum_pattern): Likewise.
5856 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5858 * predict.c (probably_never_executed): Only use precise profile info.
5859 (compute_function_frequency): Skip after inlining hack since we now
5860 have quality checking.
5862 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5864 * profile-count.h (profile_probability::very_unlikely,
5865 profile_probability::unlikely, profile_probability::even): Set
5866 precision to guessed.
5868 2018-01-23 Richard Biener <rguenther@suse.de>
5870 PR tree-optimization/83963
5871 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
5872 Properly terminate dominator walk when crossing the exit edge not
5873 when visiting its source block.
5875 2018-01-23 Jakub Jelinek <jakub@redhat.com>
5878 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
5879 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
5881 2018-01-22 Jakub Jelinek <jakub@redhat.com>
5883 PR tree-optimization/83957
5884 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
5885 semicolon after for body surrounded by braces.
5887 PR tree-optimization/83081
5888 * profile-count.h (profile_probability::split): New method.
5889 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
5890 Use profile_probability::split.
5891 (do_compare_rtx_and_jump): Fix adjustment of probabilities
5892 when splitting a single conditional jump into 2.
5894 2018-01-22 David Malcolm <dmalcolm@redhat.com>
5896 PR tree-optimization/69452
5897 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
5900 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5902 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
5903 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
5904 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
5906 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5908 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
5909 * config/rl78/rl78.md (movdi): New define_expand.
5910 * config/rl78/rl78.c (rl78_split_movdi): New function.
5912 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
5915 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
5917 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
5918 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
5919 128-bit to produce an UNSPEC move to get the double word with the
5920 signbit and then a shift directly to do signbit.
5921 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
5922 implementation with a new version that just does either a direct
5923 move or a regular move. Move memory interface to separate insns.
5924 Move insns so they are next to the expander.
5925 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
5926 with signbit move. Split big and little endian case.
5927 (signbit<mode>2_dm_mem_le): Likewise.
5928 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
5929 (signbit<mode>2_dm2): Likewise.
5931 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5933 * config/rl78/rl78.md (anddi3): New define_expand.
5935 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5937 * config/rl78/rl78.md (umindi3): New define_expand.
5939 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5941 * config/rl78/rl78.md (smindi3): New define_expand.
5943 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5945 * config/rl78/rl78.md (smaxdi3): New define_expand.
5947 2018-01-22 Carl Love <cel@us.ibm.com>
5949 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
5950 LVX_V1TI): Add macro expansion.
5951 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
5952 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
5953 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
5954 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
5955 Change check to determine if the instruction is a byte reversing
5956 entry. Fix typo in comment.
5957 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
5958 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
5959 Add def_builtin calls for new builtins.
5960 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
5961 Add define_insn expansion.
5963 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5965 * config/rl78/rl78.md (umaxdi3): New define_expand.
5967 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5969 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
5970 for non-QImode registers.
5972 2018-01-22 Richard Biener <rguenther@suse.de>
5974 PR tree-optimization/83963
5975 * graphite-scop-detection.c (scop_detection::get_sese): Delay
5976 including the loop exit block.
5977 (scop_detection::merge_sese): Likewise.
5978 (scop_detection::add_scop): Do it here instead.
5980 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5982 * doc/sourcebuild.texi (arm_softfloat): Document.
5984 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
5987 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
5988 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
5989 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
5991 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5992 David Edelsohn <dje.gcc@gmail.com>
5995 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5996 Change "crset eq" to "crset 2".
5997 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5998 (*call_indirect_aix<mode>_nospec): Likewise.
5999 (*call_value_indirect_aix<mode>_nospec): Likewise.
6000 (*call_indirect_elfv2<mode>_nospec): Likewise.
6001 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6002 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6003 change assembly output from . to $.
6004 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6005 (indirect_jump<mode>_nospec): Change assembly output from . to $.
6006 (*tablejump<mode>_internal1_nospec): Likewise.
6008 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
6011 * config/sh/sh_optimize_sett_clrt.cc:
6012 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6014 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
6016 PR tree-optimization/83940
6017 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6018 offset_dt to vect_constant_def rather than vect_unknown_def_type.
6019 (vect_check_load_store_mask): Add a mask_dt_out parameter and
6020 use it to pass back the definition type.
6021 (vect_check_store_rhs): Likewise rhs_dt_out.
6022 (vect_build_gather_load_calls): Add a mask_dt argument and use
6023 it instead of a call to vect_is_simple_use.
6024 (vectorizable_store): Update calls to vect_check_load_store_mask
6025 and vect_check_store_rhs. Use the dt returned by the latter instead
6026 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
6027 instead of calls to vect_is_simple_use. Pass the scalar rather
6028 than the vector operand to vect_is_simple_use when handling
6029 second and subsequent copies of an rhs value.
6030 (vectorizable_load): Update calls to vect_check_load_store_mask
6031 and vect_build_gather_load_calls. Use the cached mask_dt and
6032 gs_info.offset_dt instead of calls to vect_is_simple_use.
6034 2018-01-20 Jakub Jelinek <jakub@redhat.com>
6037 * tree-emutls.c: Include gimplify.h.
6038 (lower_emutls_2): New function.
6039 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6040 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6041 it before further processing.
6044 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6045 UINTVAL (trueop1) instead of INTVAL (op1).
6047 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6051 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6052 INCOMING_FRAME_SP_OFFSET if not defined.
6053 (scan_trace): Add ENTRY argument. If true and
6054 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6055 emit a note to adjust the CFA offset.
6056 (create_cfi_notes): Adjust scan_trace callers.
6057 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6058 INCOMING_FRAME_SP_OFFSET in the CIE.
6059 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6060 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6062 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6063 * doc/tm.texi: Regenerated.
6065 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6067 PR rtl-optimization/83147
6068 * lra-constraints.c (remove_inheritance_pseudos): Use
6069 lra_substitute_pseudo_within_insn.
6071 2018-01-19 Tom de Vries <tom@codesourcery.com>
6072 Cesar Philippidis <cesar@codesourcery.com>
6075 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6077 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6080 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6081 spaces for function labels.
6083 2018-01-19 Martin Liska <mliska@suse.cz>
6085 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6086 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6087 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6088 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6089 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6090 (PRED_CONST_RETURN): Change from 69 to 65.
6091 (PRED_NULL_RETURN): Change from 91 to 71.
6092 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6093 (PRED_LOOP_GUARD): Change from 66 to 73.
6095 2018-01-19 Martin Liska <mliska@suse.cz>
6097 * predict.c (predict_insn_def): Add new assert.
6098 (struct branch_predictor): Change type to signed integer.
6099 (test_prediction_value_range): Amend test to cover
6101 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6102 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6103 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6104 (PRED_LOOP_IV_COMPARE): Likewise.
6105 * predict.h (PROB_UNINITIALIZED): Define new constant.
6107 2018-01-19 Martin Liska <mliska@suse.cz>
6109 * predict.c (dump_prediction): Add new format for
6110 analyze_brprob.py script which is enabled with -details
6112 * profile-count.h (precise_p): New function.
6114 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6116 PR tree-optimization/83922
6117 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6118 there are no statements that need masking.
6119 (vect_active_double_reduction_p): New function.
6120 (vect_analyze_loop_operations): Use it when handling phis that
6121 are not in the loop header.
6123 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6125 PR tree-optimization/83914
6126 * tree-vect-loop.c (vectorizable_induction): Don't convert
6127 init_expr or apply the peeling adjustment for inductions
6128 that are nested within the vectorized loop.
6130 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6132 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6135 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6139 * function.h (gimplify_parameters): Add gimple_seq * argument.
6140 * function.c: Include gimple.h and options.h.
6141 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6142 for the added local temporaries if needed.
6143 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6144 if there are any parameter cleanups, wrap whole body into a
6145 try/finally with the cleanups.
6147 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6150 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6151 Use GET_MODE_CLASS for scalar floating point.
6153 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6157 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6158 Fix call of call_cgraph_insertion_hooks.
6160 2018-01-18 Martin Sebor <msebor@redhat.com>
6162 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6164 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6167 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6170 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
6173 * common.opt: (-ffile-prefix-map): New option.
6174 * opts.c (common_handle_option): Defer it.
6175 * opts-global.c (handle_common_deferred_options): Handle it.
6176 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6177 * file-prefix-map.h: New file.
6178 (remap_debug_filename, add_debug_prefix_map): ...here.
6179 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6180 * final.c (debug_prefix_map, add_debug_prefix_map
6181 remap_debug_filename): Move to...
6182 * file-prefix-map.c: New file.
6183 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6184 generalize, get rid of alloca(), use strrchr() instead of strchr().
6185 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6186 Implement in terms of add_prefix_map().
6187 (remap_macro_filename, remap_debug_filename): Implement in term of
6189 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6190 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6191 * dbxout.c: Include file-prefix-map.h.
6192 * varasm.c: Likewise.
6193 * vmsdbgout.c: Likewise.
6194 * xcoffout.c: Likewise.
6195 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6196 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6197 * doc/invoke.texi (-ffile-prefix-map): Document.
6198 (-fdebug-prefix-map): Update description.
6200 2018-01-18 Martin Liska <mliska@suse.cz>
6202 * config/i386/i386.c (indirect_thunk_name): Document that also
6204 (output_indirect_thunk): Document why both instructions
6205 (pause and lfence) are generated.
6207 2018-01-18 Richard Biener <rguenther@suse.de>
6209 PR tree-optimization/83887
6210 * graphite-scop-detection.c
6211 (scop_detection::get_nearest_dom_with_single_entry): Remove.
6212 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6213 (scop_detection::merge_sese): Re-implement with a flood-fill
6214 algorithm that properly finds a SESE region if it exists.
6216 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6219 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
6220 pointer_diff optimizations use view_convert instead of convert.
6222 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6224 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6225 Generate different code for -mno-speculate-indirect-jumps.
6226 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6227 (*call_indirect_aix<mode>): Disable for
6228 -mno-speculate-indirect-jumps.
6229 (*call_indirect_aix<mode>_nospec): New define_insn.
6230 (*call_value_indirect_aix<mode>): Disable for
6231 -mno-speculate-indirect-jumps.
6232 (*call_value_indirect_aix<mode>_nospec): New define_insn.
6233 (*sibcall_nonlocal_sysv<mode>): Generate different code for
6234 -mno-speculate-indirect-jumps.
6235 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6237 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6239 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
6240 long double type, set the flags for noting the default long double
6241 type, even if we don't pass or return a long double type.
6243 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
6246 * ipa-inline.c (flatten_function): Do not overwrite final inlining
6249 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
6251 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6252 support for merge[hl].
6253 (fold_mergehl_helper): New helper function.
6254 (tree-vector-builder.h): New #include for tree_vector_builder usage.
6255 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6256 (altivec_vmrglw_direct): Add xxmrglw insn.
6258 2018-01-17 Andrew Waterman <andrew@sifive.com>
6260 * config/riscv/riscv.c (riscv_conditional_register_usage): If
6261 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6263 2018-01-17 David Malcolm <dmalcolm@redhat.com>
6266 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6267 call the lto_location_cache before reading the
6268 DECL_SOURCE_LOCATION of the types.
6270 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
6271 Richard Sandiford <richard.sandiford@linaro.org>
6273 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
6274 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6275 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6276 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6277 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6279 * config/aarch64/constraints.md (aarch64_movti_operand):
6281 * config/aarch64/predicates.md (Uti): Add new constraint.
6283 2018-01-17 Carl Love <cel@us.ibm.com>
6285 * config/rs6000/vsx.md (define_expand xl_len_r,
6286 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6287 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6289 (define_expand, define_insn): Move the shift left from the
6290 define_insn to the define_expand for lxvl and stxvl instructions.
6291 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6292 and XL_LEN_R definitions to PURE.
6294 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
6296 * config/i386/i386.c (indirect_thunk_name): Declare regno
6297 as unsigned int. Compare regno with INVALID_REGNUM.
6298 (output_indirect_thunk): Ditto.
6299 (output_indirect_thunk_function): Ditto.
6300 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6301 in the call to output_indirect_thunk_function.
6303 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6306 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6307 rather than the size of inner_type to determine the stack slot size
6308 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6310 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6313 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6316 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6318 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6319 endian Linux systems to optionally enable multilibs for selecting
6320 the long double type if the user configured an explicit type.
6321 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6322 have no long double multilibs if not defined.
6323 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6324 warn if the user used -mabi={ieee,ibm}longdouble and we built
6325 multilibs for long double.
6326 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6327 appropriate multilib option.
6328 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6330 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6331 for building long double multilibs.
6332 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6334 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6336 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6339 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6341 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6344 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6347 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6350 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6352 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6353 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6355 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6357 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6358 different rtl trees depending on TARGET_64BIT.
6359 (rs6000_gen_lvx): Likewise.
6361 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6363 * config/visium/visium.md (nop): Tweak comment.
6364 (hazard_nop): Likewise.
6366 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6368 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6369 -mspeculate-indirect-jumps.
6370 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6371 for -mno-speculate-indirect-jumps.
6372 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6373 (*call_value_indirect_elfv2<mode>): Disable for
6374 -mno-speculate-indirect-jumps.
6375 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6376 (indirect_jump): Emit different RTL for
6377 -mno-speculate-indirect-jumps.
6378 (*indirect_jump<mode>): Disable for
6379 -mno-speculate-indirect-jumps.
6380 (*indirect_jump<mode>_nospec): New define_insn.
6381 (tablejump): Emit different RTL for
6382 -mno-speculate-indirect-jumps.
6383 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6384 (tablejumpsi_nospec): New define_expand.
6385 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6386 (tablejumpdi_nospec): New define_expand.
6387 (*tablejump<mode>_internal1): Disable for
6388 -mno-speculate-indirect-jumps.
6389 (*tablejump<mode>_internal1_nospec): New define_insn.
6390 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6393 2018-01-16 Artyom Skrobov tyomitch@gmail.com
6395 * caller-save.c (insert_save): Drop unnecessary parameter. All
6398 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6399 Richard Biener <rguenth@suse.de>
6402 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6403 return early, inline manually is_gimple_sizepos. Make sure if we
6404 call gimplify_expr we don't end up with a gimple constant.
6405 * tree.c (variably_modified_type_p): Don't return true for
6406 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
6407 * gimplify.h (is_gimple_sizepos): Remove.
6409 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6411 PR tree-optimization/83857
6412 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6413 vectorizable_live_operation for pure SLP statements.
6414 (vectorizable_live_operation): Handle PHIs.
6416 2018-01-16 Richard Biener <rguenther@suse.de>
6418 PR tree-optimization/83867
6419 * tree-vect-stmts.c (vect_transform_stmt): Precompute
6420 nested_in_vect_loop_p since the scalar stmt may get invalidated.
6422 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6425 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
6426 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
6427 If off is not INTEGER_CST, issue a may not be aligned warning
6428 rather than isn't aligned. Use isn%'t rather than isn't.
6429 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
6431 <case MULT_EXPR>: Improve the case when bottom and one of the
6432 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
6433 operand, in that case check if the other operand is multiple of
6434 bottom divided by the INTEGER_CST operand.
6436 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6439 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
6440 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
6441 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
6442 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
6443 * config/pa/pa.c (pa_function_arg_advance): Likewise.
6444 (pa_function_arg, pa_arg_partial_bytes): Likewise.
6445 (pa_function_arg_size): New function.
6447 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6449 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
6450 in a separate statement.
6452 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6454 PR tree-optimization/83847
6455 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
6456 group gathers and scatters.
6458 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6460 PR rtl-optimization/86620
6461 * params.def (max-sched-ready-insns): Bump minimum value to 1.
6463 PR rtl-optimization/83213
6464 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
6465 to last if both are JUMP_INSNs.
6467 PR tree-optimization/83843
6468 * gimple-ssa-store-merging.c
6469 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
6470 store_immediate_info for bswap/nop orig_stores.
6472 2018-01-15 Andrew Waterman <andrew@sifive.com>
6474 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
6476 <UDIV>: Increase cost if !TARGET_DIV.
6478 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
6480 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
6481 (define_attr "cr_logical_3op"): New.
6482 (cceq_ior_compare): Adjust.
6483 (cceq_ior_compare_complement): Adjust.
6484 (*cceq_rev_compare): Adjust.
6485 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
6486 (is_cracked_insn): Adjust.
6487 (insn_must_be_first_in_group): Adjust.
6488 * config/rs6000/40x.md: Adjust.
6489 * config/rs6000/440.md: Adjust.
6490 * config/rs6000/476.md: Adjust.
6491 * config/rs6000/601.md: Adjust.
6492 * config/rs6000/603.md: Adjust.
6493 * config/rs6000/6xx.md: Adjust.
6494 * config/rs6000/7450.md: Adjust.
6495 * config/rs6000/7xx.md: Adjust.
6496 * config/rs6000/8540.md: Adjust.
6497 * config/rs6000/cell.md: Adjust.
6498 * config/rs6000/e300c2c3.md: Adjust.
6499 * config/rs6000/e500mc.md: Adjust.
6500 * config/rs6000/e500mc64.md: Adjust.
6501 * config/rs6000/e5500.md: Adjust.
6502 * config/rs6000/e6500.md: Adjust.
6503 * config/rs6000/mpc.md: Adjust.
6504 * config/rs6000/power4.md: Adjust.
6505 * config/rs6000/power5.md: Adjust.
6506 * config/rs6000/power6.md: Adjust.
6507 * config/rs6000/power7.md: Adjust.
6508 * config/rs6000/power8.md: Adjust.
6509 * config/rs6000/power9.md: Adjust.
6510 * config/rs6000/rs64.md: Adjust.
6511 * config/rs6000/titan.md: Adjust.
6513 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6515 * config/i386/predicates.md (indirect_branch_operand): Rewrite
6516 ix86_indirect_branch_register logic.
6518 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6520 * config/i386/constraints.md (Bs): Update
6521 ix86_indirect_branch_register check. Don't check
6522 ix86_indirect_branch_register with GOT_memory_operand.
6524 * config/i386/predicates.md (GOT_memory_operand): Don't check
6525 ix86_indirect_branch_register here.
6526 (GOT32_symbol_operand): Likewise.
6528 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6530 * config/i386/predicates.md (constant_call_address_operand):
6531 Rewrite ix86_indirect_branch_register logic.
6532 (sibcall_insn_operand): Likewise.
6534 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6536 * config/i386/constraints.md (Bs): Replace
6537 ix86_indirect_branch_thunk_register with
6538 ix86_indirect_branch_register.
6540 * config/i386/i386.md (indirect_jump): Likewise.
6541 (tablejump): Likewise.
6542 (*sibcall_memory): Likewise.
6543 (*sibcall_value_memory): Likewise.
6544 Peepholes of indirect call and jump via memory: Likewise.
6545 * config/i386/i386.opt: Likewise.
6546 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6547 (GOT_memory_operand): Likewise.
6548 (call_insn_operand): Likewise.
6549 (sibcall_insn_operand): Likewise.
6550 (GOT32_symbol_operand): Likewise.
6552 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6555 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
6556 type rather than type addr's type points to.
6557 (expand_omp_atomic_mutex): Likewise.
6558 (expand_omp_atomic): Likewise.
6560 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6563 * config/i386/i386.c (output_indirect_thunk_function): Use
6564 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
6565 for __x86_return_thunk.
6567 2018-01-15 Richard Biener <rguenther@suse.de>
6570 * expmed.c (extract_bit_field_1): Fix typo.
6572 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6575 * config/arm/iterators.md (VF): New mode iterator.
6576 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
6577 Remove integer-related logic from pattern.
6578 (neon_vabd<mode>_3): Likewise.
6580 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6583 * common.opt (fstrict-overflow): No longer an alias.
6584 (fwrapv-pointer): New option.
6585 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
6586 also for pointer types based on flag_wrapv_pointer.
6587 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
6588 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
6589 opts->x_flag_wrapv got set.
6590 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
6591 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
6592 POINTER_TYPE_OVERFLOW_UNDEFINED.
6593 * match.pd: Likewise in address comparison pattern.
6594 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
6596 2018-01-15 Richard Biener <rguenther@suse.de>
6599 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
6600 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
6601 Reset type names to their identifier if their TYPE_DECL doesn't
6602 have linkage (and thus is used for ODR and devirt).
6603 (save_debug_info_for_decl): Remove.
6604 (save_debug_info_for_type): Likewise.
6605 (add_tree_to_fld_list): Adjust.
6606 * tree-pretty-print.c (dump_generic_node): Make dumping of
6607 type names more robust.
6609 2018-01-15 Richard Biener <rguenther@suse.de>
6611 * BASE-VER: Bump to 8.0.1.
6613 2018-01-14 Martin Sebor <msebor@redhat.com>
6616 * builtins.c (check_access): Avoid warning when the no-warning bit
6619 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
6621 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
6622 * ira-color (allocno_hard_regs_compare): Likewise.
6624 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
6627 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
6628 Use .pushsection/.popsection.
6630 2018-01-14 Martin Sebor <msebor@redhat.com>
6633 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
6635 2018-01-14 Jakub Jelinek <jakub@redhat.com>
6637 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
6638 entry from extra_headers.
6639 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
6640 extra_headers, make the list bitwise identical to the i?86-*-* one.
6642 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6644 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
6645 -mcmodel=large with -mindirect-branch=thunk,
6646 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
6647 -mfunction-return=thunk-extern.
6648 * doc/invoke.texi: Document -mcmodel=large is incompatible with
6649 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
6650 -mfunction-return=thunk and -mfunction-return=thunk-extern.
6652 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6654 * config/i386/i386.c (print_reg): Print the name of the full
6655 integer register without '%'.
6656 (ix86_print_operand): Handle 'V'.
6657 * doc/extend.texi: Document 'V' modifier.
6659 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6661 * config/i386/constraints.md (Bs): Disallow memory operand for
6662 -mindirect-branch-register.
6664 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6665 (GOT_memory_operand): Likewise.
6666 (call_insn_operand): Likewise.
6667 (sibcall_insn_operand): Likewise.
6668 (GOT32_symbol_operand): Likewise.
6669 * config/i386/i386.md (indirect_jump): Call convert_memory_address
6670 for -mindirect-branch-register.
6671 (tablejump): Likewise.
6672 (*sibcall_memory): Likewise.
6673 (*sibcall_value_memory): Likewise.
6674 Disallow peepholes of indirect call and jump via memory for
6675 -mindirect-branch-register.
6676 (*call_pop): Replace m with Bw.
6677 (*call_value_pop): Likewise.
6678 (*sibcall_pop_memory): Replace m with Bs.
6679 * config/i386/i386.opt (mindirect-branch-register): New option.
6680 * doc/invoke.texi: Document -mindirect-branch-register option.
6682 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6684 * config/i386/i386-protos.h (ix86_output_function_return): New.
6685 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
6686 set function_return_type.
6687 (indirect_thunk_name): Add ret_p to indicate thunk for function
6689 (output_indirect_thunk_function): Pass false to
6690 indirect_thunk_name.
6691 (ix86_output_indirect_branch_via_reg): Likewise.
6692 (ix86_output_indirect_branch_via_push): Likewise.
6693 (output_indirect_thunk_function): Create alias for function
6694 return thunk if regno < 0.
6695 (ix86_output_function_return): New function.
6696 (ix86_handle_fndecl_attribute): Handle function_return.
6697 (ix86_attribute_table): Add function_return.
6698 * config/i386/i386.h (machine_function): Add
6699 function_return_type.
6700 * config/i386/i386.md (simple_return_internal): Use
6701 ix86_output_function_return.
6702 (simple_return_internal_long): Likewise.
6703 * config/i386/i386.opt (mfunction-return=): New option.
6704 (indirect_branch): Mention -mfunction-return=.
6705 * doc/extend.texi: Document function_return function attribute.
6706 * doc/invoke.texi: Document -mfunction-return= option.
6708 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6710 * config/i386/i386-opts.h (indirect_branch): New.
6711 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
6712 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
6713 with local indirect jump when converting indirect call and jump.
6714 (ix86_set_indirect_branch_type): New.
6715 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
6716 (indirectlabelno): New.
6717 (indirect_thunk_needed): Likewise.
6718 (indirect_thunk_bnd_needed): Likewise.
6719 (indirect_thunks_used): Likewise.
6720 (indirect_thunks_bnd_used): Likewise.
6721 (INDIRECT_LABEL): Likewise.
6722 (indirect_thunk_name): Likewise.
6723 (output_indirect_thunk): Likewise.
6724 (output_indirect_thunk_function): Likewise.
6725 (ix86_output_indirect_branch_via_reg): Likewise.
6726 (ix86_output_indirect_branch_via_push): Likewise.
6727 (ix86_output_indirect_branch): Likewise.
6728 (ix86_output_indirect_jmp): Likewise.
6729 (ix86_code_end): Call output_indirect_thunk_function if needed.
6730 (ix86_output_call_insn): Call ix86_output_indirect_branch if
6732 (ix86_handle_fndecl_attribute): Handle indirect_branch.
6733 (ix86_attribute_table): Add indirect_branch.
6734 * config/i386/i386.h (machine_function): Add indirect_branch_type
6735 and has_local_indirect_jump.
6736 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
6738 (tablejump): Likewise.
6739 (*indirect_jump): Use ix86_output_indirect_jmp.
6740 (*tablejump_1): Likewise.
6741 (simple_return_indirect_internal): Likewise.
6742 * config/i386/i386.opt (mindirect-branch=): New option.
6743 (indirect_branch): New.
6746 (thunk-inline): Likewise.
6747 (thunk-extern): Likewise.
6748 * doc/extend.texi: Document indirect_branch function attribute.
6749 * doc/invoke.texi: Document -mindirect-branch= option.
6751 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
6754 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
6756 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
6758 * ipa-inline.c (want_inline_small_function_p): Return false if
6759 inlining has already failed with CIF_FINAL_ERROR.
6760 (update_caller_keys): Call want_inline_small_function_p before
6762 (update_callee_keys): Likewise.
6764 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6766 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
6768 (rs6000_quadword_masked_address_p): Likewise.
6769 (quad_aligned_load_p): Likewise.
6770 (quad_aligned_store_p): Likewise.
6771 (const_load_sequence_p): Add comment to describe the outer-most loop.
6772 (mimic_memory_attributes_and_flags): New function.
6773 (rs6000_gen_stvx): Likewise.
6774 (replace_swapped_aligned_store): Likewise.
6775 (rs6000_gen_lvx): Likewise.
6776 (replace_swapped_aligned_load): Likewise.
6777 (replace_swapped_load_constant): Capitalize argument name in
6778 comment describing this function.
6779 (rs6000_analyze_swaps): Add a third pass to search for vector loads
6780 and stores that access quad-word aligned addresses and replace
6781 with stvx or lvx instructions when appropriate.
6782 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
6783 New function prototype.
6784 (rs6000_quadword_masked_address_p): Likewise.
6785 (rs6000_gen_lvx): Likewise.
6786 (rs6000_gen_stvx): Likewise.
6787 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
6788 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
6789 when memory address is aligned.
6790 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
6791 this split to select lvx instruction when memory address is aligned.
6792 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
6793 instruction when memory address is aligned.
6794 (*vsx_le_perm_load_v16qi): Likewise.
6795 (four unnamed splitters): Modify to select the stvx instruction
6796 when memory is aligned.
6798 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
6800 * predict.c (determine_unlikely_bbs): Handle correctly BBs
6801 which appears in the queue multiple times.
6803 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6804 Alan Hayward <alan.hayward@arm.com>
6805 David Sherwood <david.sherwood@arm.com>
6807 * tree-vectorizer.h (vec_lower_bound): New structure.
6808 (_loop_vec_info): Add check_nonzero and lower_bounds.
6809 (LOOP_VINFO_CHECK_NONZERO): New macro.
6810 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
6811 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
6812 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
6813 fields. Make seg_len the distance travelled, not including the
6815 (dr_direction_indicator): Declare.
6816 (dr_zero_step_indicator): Likewise.
6817 (dr_known_forward_stride_p): Likewise.
6818 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
6820 (runtime_alias_check_p): Allow runtime alias checks with
6822 (operator ==): Compare access_size and align.
6823 (prune_runtime_alias_test_list): Rework for new distinction between
6824 the access_size and seg_len.
6825 (create_intersect_range_checks_index): Likewise. Cope with polynomial
6827 (get_segment_min_max): New function.
6828 (create_intersect_range_checks): Use it.
6829 (dr_step_indicator): New function.
6830 (dr_direction_indicator): Likewise.
6831 (dr_zero_step_indicator): Likewise.
6832 (dr_known_forward_stride_p): Likewise.
6833 * tree-loop-distribution.c (data_ref_segment_size): Return
6834 DR_STEP * (niters - 1).
6835 (compute_alias_check_pairs): Update call to the dr_with_seg_len
6837 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
6838 (vect_preserves_scalar_order_p): New function, split out from...
6839 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
6840 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
6841 (vect_vfa_access_size): New function.
6842 (vect_vfa_align): Likewise.
6843 (vect_compile_time_alias): Take access_size_a and access_b arguments.
6844 (dump_lower_bound): New function.
6845 (vect_check_lower_bound): Likewise.
6846 (vect_small_gap_p): Likewise.
6847 (vectorizable_with_step_bound_p): Likewise.
6848 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
6849 depencies if the vectorization factor is 1. Convert the checks
6850 for nonzero steps into checks on the bounds of DR_STEP. Try using
6851 a bunds check for variable steps if the minimum required step is
6852 relatively small. Update calls to the dr_with_seg_len
6853 constructor and to vect_compile_time_alias.
6854 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
6856 (vect_loop_versioning): Call it.
6857 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
6859 (vect_estimate_min_profitable_iters): Account for any bounds checks.
6861 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6862 Alan Hayward <alan.hayward@arm.com>
6863 David Sherwood <david.sherwood@arm.com>
6865 * doc/sourcebuild.texi (vect_scatter_store): Document.
6866 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
6868 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
6870 * genopinit.c (main): Add supports_vec_scatter_store and
6871 supports_vec_scatter_store_cached to target_optabs.
6872 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
6873 IFN_MASK_SCATTER_STORE.
6874 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
6876 * internal-fn.h (internal_store_fn_p): Declare.
6877 (internal_fn_stored_value_index): Likewise.
6878 * internal-fn.c (scatter_store_direct): New macro.
6879 (expand_scatter_store_optab_fn): New function.
6880 (direct_scatter_store_optab_supported_p): New macro.
6881 (internal_store_fn_p): New function.
6882 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
6883 IFN_MASK_SCATTER_STORE.
6884 (internal_fn_mask_index): Likewise.
6885 (internal_fn_stored_value_index): New function.
6886 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
6888 * optabs-query.h (supports_vec_scatter_store_p): Declare.
6889 * optabs-query.c (supports_vec_scatter_store_p): New function.
6890 * tree-vectorizer.h (vect_get_store_rhs): Declare.
6891 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
6892 true for scatter stores.
6893 (vect_gather_scatter_fn_p): Handle scatter stores too.
6894 (vect_check_gather_scatter): Consider using scatter stores if
6895 supports_vec_scatter_store_p.
6896 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
6898 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6899 internal_fn_stored_value_index.
6900 (check_load_store_masking): Handle scatter stores too.
6901 (vect_get_store_rhs): Make public.
6902 (vectorizable_call): Use internal_store_fn_p.
6903 (vectorizable_store): Handle scatter store internal functions.
6904 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
6905 when deciding whether the end of the group has been reached.
6906 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
6907 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
6908 (mask_scatter_store<mode>): New insns.
6910 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6911 Alan Hayward <alan.hayward@arm.com>
6912 David Sherwood <david.sherwood@arm.com>
6914 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
6915 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
6916 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
6918 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
6919 Use vect_truncate_gather_scatter_offset if we can't treat the
6920 operation as a normal gather load or scatter store.
6921 (get_group_load_store_type): Take the gather_scatter_info
6922 as argument. Try using a gather load or scatter store for
6923 single-element groups.
6924 (get_load_store_type): Update calls to get_group_load_store_type
6925 and vect_use_strided_gather_scatters_p.
6927 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6928 Alan Hayward <alan.hayward@arm.com>
6929 David Sherwood <david.sherwood@arm.com>
6931 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
6932 optional tree argument.
6933 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
6935 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
6936 but continue to use the current value as a fallback.
6937 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
6938 to compare the updates.
6939 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
6940 (get_load_store_type): Use it when handling a strided access.
6941 (vect_get_strided_load_store_ops): New function.
6942 (vect_get_data_ptr_increment): Likewise.
6943 (vectorizable_load): Handle strided gather loads. Always pass
6944 a step to vect_create_data_ref_ptr and bump_vector_ptr.
6946 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6947 Alan Hayward <alan.hayward@arm.com>
6948 David Sherwood <david.sherwood@arm.com>
6950 * doc/md.texi (gather_load@var{m}): Document.
6951 (mask_gather_load@var{m}): Likewise.
6952 * genopinit.c (main): Add supports_vec_gather_load and
6953 supports_vec_gather_load_cached to target_optabs.
6954 * optabs-tree.c (init_tree_optimization_optabs): Use
6955 ggc_cleared_alloc to allocate target_optabs.
6956 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
6957 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
6959 * internal-fn.h (internal_load_fn_p): Declare.
6960 (internal_gather_scatter_fn_p): Likewise.
6961 (internal_fn_mask_index): Likewise.
6962 (internal_gather_scatter_fn_supported_p): Likewise.
6963 * internal-fn.c (gather_load_direct): New macro.
6964 (expand_gather_load_optab_fn): New function.
6965 (direct_gather_load_optab_supported_p): New macro.
6966 (direct_internal_fn_optab): New function.
6967 (internal_load_fn_p): Likewise.
6968 (internal_gather_scatter_fn_p): Likewise.
6969 (internal_fn_mask_index): Likewise.
6970 (internal_gather_scatter_fn_supported_p): Likewise.
6971 * optabs-query.c (supports_at_least_one_mode_p): New function.
6972 (supports_vec_gather_load_p): Likewise.
6973 * optabs-query.h (supports_vec_gather_load_p): Declare.
6974 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
6975 and memory_type field.
6976 (NUM_PATTERNS): Bump to 15.
6977 * tree-vect-data-refs.c: Include internal-fn.h.
6978 (vect_gather_scatter_fn_p): New function.
6979 (vect_describe_gather_scatter_call): Likewise.
6980 (vect_check_gather_scatter): Try using internal functions for
6981 gather loads. Recognize existing calls to a gather load function.
6982 (vect_analyze_data_refs): Consider using gather loads if
6983 supports_vec_gather_load_p.
6984 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
6985 (vect_get_gather_scatter_offset_type): Likewise.
6986 (vect_convert_mask_for_vectype): Likewise.
6987 (vect_add_conversion_to_patterm): Likewise.
6988 (vect_try_gather_scatter_pattern): Likewise.
6989 (vect_recog_gather_scatter_pattern): New pattern recognizer.
6990 (vect_vect_recog_func_ptrs): Add it.
6991 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6992 internal_fn_mask_index and internal_gather_scatter_fn_p.
6993 (check_load_store_masking): Take the gather_scatter_info as an
6994 argument and handle gather loads.
6995 (vect_get_gather_scatter_ops): New function.
6996 (vectorizable_call): Check internal_load_fn_p.
6997 (vectorizable_load): Likewise. Handle gather load internal
6999 (vectorizable_store): Update call to check_load_store_masking.
7000 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7001 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7002 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7003 (aarch64_gather_scale_operand_d): New predicates.
7004 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7005 (mask_gather_load<mode>): New insns.
7007 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7008 Alan Hayward <alan.hayward@arm.com>
7009 David Sherwood <david.sherwood@arm.com>
7011 * optabs.def (fold_left_plus_optab): New optab.
7012 * doc/md.texi (fold_left_plus_@var{m}): Document.
7013 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7014 * internal-fn.c (fold_left_direct): Define.
7015 (expand_fold_left_optab_fn): Likewise.
7016 (direct_fold_left_optab_supported_p): Likewise.
7017 * fold-const-call.c (fold_const_fold_left): New function.
7018 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7019 * tree-parloops.c (valid_reduction_p): New function.
7020 (gather_scalar_reductions): Use it.
7021 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7022 (vect_finish_replace_stmt): Declare.
7023 * tree-vect-loop.c (fold_left_reduction_fn): New function.
7024 (needs_fold_left_reduction_p): New function, split out from...
7025 (vect_is_simple_reduction): ...here. Accept reductions that
7026 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7027 (vect_force_simple_reduction): Also store the reduction type in
7028 the assignment's STMT_VINFO_REDUC_TYPE.
7029 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7030 (merge_with_identity): New function.
7031 (vect_expand_fold_left): Likewise.
7032 (vectorize_fold_left_reduction): Likewise.
7033 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
7034 scalar phi in place for it. Check for target support and reject
7035 cases that would reassociate the operation. Defer the transform
7036 phase to vectorize_fold_left_reduction.
7037 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7038 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7039 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7041 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7043 * tree-if-conv.c (predicate_mem_writes): Remove redundant
7044 call to ifc_temp_var.
7046 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7047 Alan Hayward <alan.hayward@arm.com>
7048 David Sherwood <david.sherwood@arm.com>
7050 * target.def (legitimize_address_displacement): Take the original
7051 offset as a poly_int.
7052 * targhooks.h (default_legitimize_address_displacement): Update
7054 * targhooks.c (default_legitimize_address_displacement): Likewise.
7055 * doc/tm.texi: Regenerate.
7056 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7057 as an argument, moving assert of ad->disp == ad->disp_term to...
7058 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
7059 Try calling targetm.legitimize_address_displacement before expanding
7060 the address rather than afterwards, and adjust for the new interface.
7061 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7062 Match the new hook interface. Handle SVE addresses.
7063 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7066 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7068 * Makefile.in (OBJS): Add early-remat.o.
7069 * target.def (select_early_remat_modes): New hook.
7070 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7071 * doc/tm.texi: Regenerate.
7072 * targhooks.h (default_select_early_remat_modes): Declare.
7073 * targhooks.c (default_select_early_remat_modes): New function.
7074 * timevar.def (TV_EARLY_REMAT): New timevar.
7075 * passes.def (pass_early_remat): New pass.
7076 * tree-pass.h (make_pass_early_remat): Declare.
7077 * early-remat.c: New file.
7078 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7080 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7082 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7083 Alan Hayward <alan.hayward@arm.com>
7084 David Sherwood <david.sherwood@arm.com>
7086 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7087 vfm1 with a bound_epilog parameter.
7088 (vect_do_peeling): Update calls accordingly, and move the prologue
7089 call earlier in the function. Treat the base bound_epilog as 0 for
7090 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7091 this base when peeling for gaps.
7092 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7093 with fully-masked loops.
7094 (vect_estimate_min_profitable_iters): Handle the single peeled
7095 iteration in that case.
7097 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7098 Alan Hayward <alan.hayward@arm.com>
7099 David Sherwood <david.sherwood@arm.com>
7101 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7102 single-element interleaving even if the size is not a power of 2.
7103 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7104 accesses for single-element interleaving if the group size is
7107 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7108 Alan Hayward <alan.hayward@arm.com>
7109 David Sherwood <david.sherwood@arm.com>
7111 * doc/md.texi (fold_extract_last_@var{m}): Document.
7112 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7113 * optabs.def (fold_extract_last_optab): New optab.
7114 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7115 * internal-fn.c (fold_extract_direct): New macro.
7116 (expand_fold_extract_optab_fn): Likewise.
7117 (direct_fold_extract_optab_supported_p): Likewise.
7118 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7119 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7120 EXTRACT_LAST_REDUCTION.
7121 (get_initial_def_for_reduction): Do not create an initial vector
7122 for EXTRACT_LAST_REDUCTION reductions.
7123 (vectorizable_reduction): Leave the scalar phi in place for
7124 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7125 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7126 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7127 transform phase to vectorizable_condition.
7128 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7130 (vect_finish_stmt_generation): ...here.
7131 (vect_finish_replace_stmt): New function.
7132 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7133 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7135 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7137 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7138 Alan Hayward <alan.hayward@arm.com>
7139 David Sherwood <david.sherwood@arm.com>
7141 * doc/md.texi (extract_last_@var{m}): Document.
7142 * optabs.def (extract_last_optab): New optab.
7143 * internal-fn.def (EXTRACT_LAST): New internal function.
7144 * internal-fn.c (cond_unary_direct): New macro.
7145 (expand_cond_unary_optab_fn): Likewise.
7146 (direct_cond_unary_optab_supported_p): Likewise.
7147 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7148 loops using EXTRACT_LAST.
7149 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7150 (extract_last_<mode>): ...this optab.
7151 (vec_extract<mode><Vel>): Update accordingly.
7153 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7154 Alan Hayward <alan.hayward@arm.com>
7155 David Sherwood <david.sherwood@arm.com>
7157 * target.def (empty_mask_is_expensive): New hook.
7158 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7159 * doc/tm.texi: Regenerate.
7160 * targhooks.h (default_empty_mask_is_expensive): Declare.
7161 * targhooks.c (default_empty_mask_is_expensive): New function.
7162 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7163 if the target says that empty masks are expensive.
7164 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7166 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
7168 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7169 Alan Hayward <alan.hayward@arm.com>
7170 David Sherwood <david.sherwood@arm.com>
7172 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7173 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7174 (vect_use_loop_mask_for_alignment_p): New function.
7175 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7176 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7177 niters_skip argument. Make sure that the first niters_skip elements
7178 of the first iteration are inactive.
7179 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7180 Update call to vect_set_loop_masks_directly.
7181 (get_misalign_in_elems): New function, split out from...
7182 (vect_gen_prolog_loop_niters): ...here.
7183 (vect_update_init_of_dr): Take a code argument that specifies whether
7184 the adjustment should be added or subtracted.
7185 (vect_update_init_of_drs): Likewise.
7186 (vect_prepare_for_masked_peels): New function.
7187 (vect_do_peeling): Skip prologue peeling if we're using a mask
7188 instead. Update call to vect_update_inits_of_drs.
7189 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7191 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7192 alignment. Do not include the number of peeled iterations in
7193 the minimum threshold in that case.
7194 (vectorizable_induction): Adjust the start value down by
7195 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7196 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7197 Take the number of skipped iterations into account when calculating
7199 * tree-vect-stmts.c (vect_gen_while_not): New function.
7201 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7202 Alan Hayward <alan.hayward@arm.com>
7203 David Sherwood <david.sherwood@arm.com>
7205 * doc/sourcebuild.texi (vect_fully_masked): Document.
7206 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7208 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7210 (vect_analyze_loop_2): ...here. Don't check the vectorization
7211 factor against the number of loop iterations if the loop is
7214 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7215 Alan Hayward <alan.hayward@arm.com>
7216 David Sherwood <david.sherwood@arm.com>
7218 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
7219 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
7220 (dump_groups): Update accordingly.
7221 (iv_use::mem_type): New member variable.
7222 (address_p): New function.
7223 (record_use): Add a mem_type argument and initialize the new
7225 (record_group_use): Add a mem_type argument. Use address_p.
7226 Remove obsolete null checks of base_object. Update call to record_use.
7227 (find_interesting_uses_op): Update call to record_group_use.
7228 (find_interesting_uses_cond): Likewise.
7229 (find_interesting_uses_address): Likewise.
7230 (get_mem_type_for_internal_fn): New function.
7231 (find_address_like_use): Likewise.
7232 (find_interesting_uses_stmt): Try find_address_like_use before
7233 calling find_interesting_uses_op.
7234 (addr_offset_valid_p): Use the iv mem_type field as the type
7235 of the addressed memory.
7236 (add_autoinc_candidates): Likewise.
7237 (get_address_cost): Likewise.
7238 (split_small_address_groups_p): Use address_p.
7239 (split_address_groups): Likewise.
7240 (add_iv_candidate_for_use): Likewise.
7241 (autoinc_possible_for_pair): Likewise.
7242 (rewrite_groups): Likewise.
7243 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
7244 (determine_group_iv_cost): Update after split of USE_ADDRESS.
7245 (get_alias_ptr_type_for_ptr_address): New function.
7246 (rewrite_use_address): Rewrite address uses in calls that were
7247 identified by find_address_like_use.
7249 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7250 Alan Hayward <alan.hayward@arm.com>
7251 David Sherwood <david.sherwood@arm.com>
7253 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7255 * gimple-expr.h (is_gimple_addressable: Likewise.
7256 * gimple-expr.c (is_gimple_address): Likewise.
7257 * internal-fn.c (expand_call_mem_ref): New function.
7258 (expand_mask_load_optab_fn): Use it.
7259 (expand_mask_store_optab_fn): Likewise.
7261 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7262 Alan Hayward <alan.hayward@arm.com>
7263 David Sherwood <david.sherwood@arm.com>
7265 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7266 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7267 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7268 (cond_umax@var{mode}): Document.
7269 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
7270 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
7271 (cond_umin_optab, cond_umax_optab): New optabs.
7272 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
7273 (COND_IOR, COND_XOR): New internal functions.
7274 * internal-fn.h (get_conditional_internal_fn): Declare.
7275 * internal-fn.c (cond_binary_direct): New macro.
7276 (expand_cond_binary_optab_fn): Likewise.
7277 (direct_cond_binary_optab_supported_p): Likewise.
7278 (get_conditional_internal_fn): New function.
7279 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7280 Cope with reduction statements that are vectorized as calls rather
7282 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7283 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7284 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7285 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7286 (UNSPEC_COND_EOR): New unspecs.
7287 (optab): Add mappings for them.
7288 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7289 (sve_int_op, sve_fp_op): New int attributes.
7291 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7292 Alan Hayward <alan.hayward@arm.com>
7293 David Sherwood <david.sherwood@arm.com>
7295 * optabs.def (while_ult_optab): New optab.
7296 * doc/md.texi (while_ult@var{m}@var{n}): Document.
7297 * internal-fn.def (WHILE_ULT): New internal function.
7298 * internal-fn.h (direct_internal_fn_supported_p): New override
7299 that takes two types as argument.
7300 * internal-fn.c (while_direct): New macro.
7301 (expand_while_optab_fn): New function.
7302 (convert_optab_supported_p): Likewise.
7303 (direct_while_optab_supported_p): New macro.
7304 * wide-int.h (wi::udiv_ceil): New function.
7305 * tree-vectorizer.h (rgroup_masks): New structure.
7306 (vec_loop_masks): New typedef.
7307 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7309 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7310 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7311 (vect_max_vf): New function.
7312 (slpeel_make_loop_iterate_ntimes): Delete.
7313 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7314 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7315 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7316 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7317 internal-fn.h, stor-layout.h and optabs-query.h.
7318 (vect_set_loop_mask): New function.
7319 (add_preheader_seq): Likewise.
7320 (add_header_seq): Likewise.
7321 (interleave_supported_p): Likewise.
7322 (vect_maybe_permute_loop_masks): Likewise.
7323 (vect_set_loop_masks_directly): Likewise.
7324 (vect_set_loop_condition_masked): Likewise.
7325 (vect_set_loop_condition_unmasked): New function, split out from
7326 slpeel_make_loop_iterate_ntimes.
7327 (slpeel_make_loop_iterate_ntimes): Rename to..
7328 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7329 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7330 (vect_do_peeling): Update call accordingly.
7331 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7333 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7334 mask_compare_type, can_fully_mask_p and fully_masked_p.
7335 (release_vec_loop_masks): New function.
7336 (_loop_vec_info): Use it to free the loop masks.
7337 (can_produce_all_loop_masks_p): New function.
7338 (vect_get_max_nscalars_per_iter): Likewise.
7339 (vect_verify_full_masking): Likewise.
7340 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7341 retries, and free the mask rgroups before retrying. Check loop-wide
7342 reasons for disallowing fully-masked loops. Make the final decision
7343 about whether use a fully-masked loop or not.
7344 (vect_estimate_min_profitable_iters): Do not assume that peeling
7345 for the number of iterations will be needed for fully-masked loops.
7346 (vectorizable_reduction): Disable fully-masked loops.
7347 (vectorizable_live_operation): Likewise.
7348 (vect_halve_mask_nunits): New function.
7349 (vect_double_mask_nunits): Likewise.
7350 (vect_record_loop_mask): Likewise.
7351 (vect_get_loop_mask): Likewise.
7352 (vect_transform_loop): Handle the case in which the final loop
7353 iteration might handle a partial vector. Call vect_set_loop_condition
7354 instead of slpeel_make_loop_iterate_ntimes.
7355 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7356 (check_load_store_masking): New function.
7357 (prepare_load_store_mask): Likewise.
7358 (vectorizable_store): Handle fully-masked loops.
7359 (vectorizable_load): Likewise.
7360 (supportable_widening_operation): Use vect_halve_mask_nunits for
7362 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7363 (vect_gen_while): New function.
7364 * config/aarch64/aarch64.md (umax<mode>3): New expander.
7365 (aarch64_uqdec<mode>): New insn.
7367 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7368 Alan Hayward <alan.hayward@arm.com>
7369 David Sherwood <david.sherwood@arm.com>
7371 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7372 (reduc_xor_scal_optab): New optabs.
7373 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7374 (reduc_xor_scal_@var{m}): Document.
7375 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7376 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7378 * fold-const-call.c (fold_const_call): Handle them.
7379 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7380 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7381 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7382 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7383 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7384 (UNSPEC_XORV): New unspecs.
7385 (optab): Add entries for them.
7386 (BITWISEV): New int iterator.
7387 (bit_reduc_op): New int attributes.
7389 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7390 Alan Hayward <alan.hayward@arm.com>
7391 David Sherwood <david.sherwood@arm.com>
7393 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7394 * internal-fn.def (VEC_SHL_INSERT): New internal function.
7395 * optabs.def (vec_shl_insert_optab): New optab.
7396 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7397 (duplicate_and_interleave): Likewise.
7398 * tree-vect-loop.c: Include internal-fn.h.
7399 (neutral_op_for_slp_reduction): New function, split out from
7400 get_initial_defs_for_reduction.
7401 (get_initial_def_for_reduction): Handle option 2 for variable-length
7402 vectors by loading the neutral value into a vector and then shifting
7403 the initial value into element 0.
7404 (get_initial_defs_for_reduction): Replace the code argument with
7405 the neutral value calculated by neutral_op_for_slp_reduction.
7406 Use gimple_build_vector for constant-length vectors.
7407 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7408 but the first group_size elements have a neutral value.
7409 Use duplicate_and_interleave otherwise.
7410 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7411 Update call to get_initial_defs_for_reduction. Handle SLP
7412 reductions for variable-length vectors by creating one vector
7413 result for each scalar result, with the elements associated
7414 with other scalar results stubbed out with the neutral value.
7415 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7416 Require IFN_VEC_SHL_INSERT for double reductions on
7417 variable-length vectors, or SLP reductions that have
7418 a neutral value. Require can_duplicate_and_interleave_p
7419 support for variable-length unchained SLP reductions if there
7420 is no neutral value, such as for MIN/MAX reductions. Also require
7421 the number of vector elements to be a multiple of the number of
7422 SLP statements when doing variable-length unchained SLP reductions.
7423 Update call to vect_create_epilog_for_reduction.
7424 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
7425 and remove initial values.
7426 (duplicate_and_interleave): Make public.
7427 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
7428 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
7430 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7431 Alan Hayward <alan.hayward@arm.com>
7432 David Sherwood <david.sherwood@arm.com>
7434 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
7435 (can_duplicate_and_interleave_p): New function.
7436 (vect_get_and_check_slp_defs): Take the vector of statements
7437 rather than just the current one. Remove excess parentheses.
7438 Restriction rejectinon of vect_constant_def and vect_external_def
7439 for variable-length vectors to boolean types, or types for which
7440 can_duplicate_and_interleave_p is false.
7441 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
7442 (duplicate_and_interleave): New function.
7443 (vect_get_constant_vectors): Use gimple_build_vector for
7444 constant-length vectors and suitable variable-length constant
7445 vectors. Use duplicate_and_interleave for other variable-length
7446 vectors. Don't defer the update when inserting new statements.
7448 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7449 Alan Hayward <alan.hayward@arm.com>
7450 David Sherwood <david.sherwood@arm.com>
7452 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
7453 min_profitable_iters doesn't go negative.
7455 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7456 Alan Hayward <alan.hayward@arm.com>
7457 David Sherwood <david.sherwood@arm.com>
7459 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
7460 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
7461 * optabs.def (vec_mask_load_lanes_optab): New optab.
7462 (vec_mask_store_lanes_optab): Likewise.
7463 * internal-fn.def (MASK_LOAD_LANES): New internal function.
7464 (MASK_STORE_LANES): Likewise.
7465 * internal-fn.c (mask_load_lanes_direct): New macro.
7466 (mask_store_lanes_direct): Likewise.
7467 (expand_mask_load_optab_fn): Handle masked operations.
7468 (expand_mask_load_lanes_optab_fn): New macro.
7469 (expand_mask_store_optab_fn): Handle masked operations.
7470 (expand_mask_store_lanes_optab_fn): New macro.
7471 (direct_mask_load_lanes_optab_supported_p): Likewise.
7472 (direct_mask_store_lanes_optab_supported_p): Likewise.
7473 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
7475 (vect_load_lanes_supported): Likewise.
7476 * tree-vect-data-refs.c (strip_conversion): New function.
7477 (can_group_stmts_p): Likewise.
7478 (vect_analyze_data_ref_accesses): Use it instead of checking
7479 for a pair of assignments.
7480 (vect_store_lanes_supported): Take a masked_p parameter.
7481 (vect_load_lanes_supported): Likewise.
7482 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
7483 vect_store_lanes_supported and vect_load_lanes_supported.
7484 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7485 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
7486 parameter. Don't allow gaps for masked accesses.
7487 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
7488 and vect_load_lanes_supported.
7489 (get_load_store_type): Take a masked_p parameter and update
7490 call to get_group_load_store_type.
7491 (vectorizable_store): Update call to get_load_store_type.
7492 Handle IFN_MASK_STORE_LANES.
7493 (vectorizable_load): Update call to get_load_store_type.
7494 Handle IFN_MASK_LOAD_LANES.
7496 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7497 Alan Hayward <alan.hayward@arm.com>
7498 David Sherwood <david.sherwood@arm.com>
7500 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
7502 * config/aarch64/aarch64-protos.h
7503 (aarch64_sve_struct_memory_operand_p): Declare.
7504 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
7505 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
7506 (VPRED, vpred): Handle SVE structure modes.
7507 * config/aarch64/constraints.md (Utx): New constraint.
7508 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
7509 (aarch64_sve_struct_nonimmediate_operand): New predicates.
7510 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
7511 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
7512 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
7513 structure modes. Split into pieces after RA.
7514 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
7515 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
7517 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
7518 SVE structure modes.
7519 (aarch64_classify_address): Likewise.
7520 (sizetochar): Move earlier in file.
7521 (aarch64_print_operand): Handle SVE register lists.
7522 (aarch64_array_mode): New function.
7523 (aarch64_sve_struct_memory_operand_p): Likewise.
7524 (TARGET_ARRAY_MODE): Redefine.
7526 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7527 Alan Hayward <alan.hayward@arm.com>
7528 David Sherwood <david.sherwood@arm.com>
7530 * target.def (array_mode): New target hook.
7531 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
7532 * doc/tm.texi: Regenerate.
7533 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
7534 * hooks.c (hook_optmode_mode_uhwi_none): New function.
7535 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
7537 * stor-layout.c (mode_for_array): Likewise. Support polynomial
7540 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7541 Alan Hayward <alan.hayward@arm.com>
7542 David Sherwood <david.sherwood@arm.com>
7544 * fold-const.c (fold_binary_loc): Check the argument types
7545 rather than the result type when testing for a vector operation.
7547 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7549 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
7550 * doc/tm.texi: Regenerate.
7552 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7553 Alan Hayward <alan.hayward@arm.com>
7554 David Sherwood <david.sherwood@arm.com>
7556 * doc/invoke.texi (-msve-vector-bits=): Document new option.
7557 (sve): Document new AArch64 extension.
7558 * doc/md.texi (w): Extend the description of the AArch64
7559 constraint to include SVE vectors.
7560 (Upl, Upa): Document new AArch64 predicate constraints.
7561 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
7563 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
7564 (msve-vector-bits=): New option.
7565 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
7566 SVE when these are disabled.
7567 (sve): New extension.
7568 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
7569 modes. Adjust their number of units based on aarch64_sve_vg.
7570 (MAX_BITSIZE_MODE_ANY_MODE): Define.
7571 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
7572 aarch64_addr_query_type.
7573 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
7574 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
7575 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
7576 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
7577 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
7578 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
7579 (aarch64_simd_imm_zero_p): Delete.
7580 (aarch64_check_zero_based_sve_index_immediate): Declare.
7581 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7582 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7583 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7584 (aarch64_sve_float_mul_immediate_p): Likewise.
7585 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7587 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
7588 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
7589 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
7590 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
7591 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
7592 (aarch64_regmode_natural_size): Likewise.
7593 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
7594 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
7596 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
7597 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
7598 for VG and the SVE predicate registers.
7599 (V_ALIASES): Add a "z"-prefixed alias.
7600 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
7601 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
7602 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
7603 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
7604 (REG_CLASS_NAMES): Add entries for them.
7605 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
7606 and the predicate registers.
7607 (aarch64_sve_vg): Declare.
7608 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
7609 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
7610 (REGMODE_NATURAL_SIZE): Define.
7611 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
7613 * config/aarch64/aarch64.c: Include cfgrtl.h.
7614 (simd_immediate_info): Add a constructor for series vectors,
7615 and an associated step field.
7616 (aarch64_sve_vg): New variable.
7617 (aarch64_dbx_register_number): Handle VG and the predicate registers.
7618 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
7619 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
7620 (VEC_ANY_DATA, VEC_STRUCT): New constants.
7621 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
7622 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
7623 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
7624 (aarch64_get_mask_mode): New functions.
7625 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
7626 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
7627 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
7628 predicate modes and predicate registers. Explicitly restrict
7629 GPRs to modes of 16 bytes or smaller. Only allow FP registers
7630 to store a vector mode if it is recognized by
7631 aarch64_classify_vector_mode.
7632 (aarch64_regmode_natural_size): New function.
7633 (aarch64_hard_regno_caller_save_mode): Return the original mode
7635 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
7636 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
7637 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
7638 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
7640 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
7641 does not overlap dest if the function is frame-related. Handle
7643 (aarch64_split_add_offset): New function.
7644 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
7645 them aarch64_add_offset.
7646 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
7647 and update call to aarch64_sub_sp.
7648 (aarch64_add_cfa_expression): New function.
7649 (aarch64_expand_prologue): Pass extra temporary registers to the
7650 functions above. Handle the case in which we need to emit new
7651 DW_CFA_expressions for registers that were originally saved
7652 relative to the stack pointer, but now have to be expressed
7653 relative to the frame pointer.
7654 (aarch64_output_mi_thunk): Pass extra temporary registers to the
7656 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
7657 IP0 and IP1 values for SVE frames.
7658 (aarch64_expand_vec_series): New function.
7659 (aarch64_expand_sve_widened_duplicate): Likewise.
7660 (aarch64_expand_sve_const_vector): Likewise.
7661 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
7662 Handle SVE constants. Use emit_move_insn to move a force_const_mem
7663 into the register, rather than emitting a SET directly.
7664 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
7665 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
7666 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
7667 (offset_9bit_signed_scaled_p): New functions.
7668 (aarch64_replicate_bitmask_imm): New function.
7669 (aarch64_bitmask_imm): Use it.
7670 (aarch64_cannot_force_const_mem): Reject expressions involving
7671 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
7672 (aarch64_classify_index): Handle SVE indices, by requiring
7673 a plain register index with a scale that matches the element size.
7674 (aarch64_classify_address): Handle SVE addresses. Assert that
7675 the mode of the address is VOIDmode or an integer mode.
7676 Update call to aarch64_classify_symbol.
7677 (aarch64_classify_symbolic_expression): Update call to
7678 aarch64_classify_symbol.
7679 (aarch64_const_vec_all_in_range_p): New function.
7680 (aarch64_print_vector_float_operand): Likewise.
7681 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
7682 "vN" for FP registers with SVE modes. Handle (const ...) vectors
7683 and the FP immediates 1.0 and 0.5.
7684 (aarch64_print_address_internal): Handle SVE addresses.
7685 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
7686 (aarch64_regno_regclass): Handle predicate registers.
7687 (aarch64_secondary_reload): Handle big-endian reloads of SVE
7689 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
7690 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
7691 (aarch64_convert_sve_vector_bits): New function.
7692 (aarch64_override_options): Use it to handle -msve-vector-bits=.
7693 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7695 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
7696 Handle SVE vector and predicate modes. Accept VL-based constants
7697 that need only one temporary register, and VL offsets that require
7698 no temporary registers.
7699 (aarch64_conditional_register_usage): Mark the predicate registers
7700 as fixed if SVE isn't available.
7701 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
7702 Return true for SVE vector and predicate modes.
7703 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
7704 rather than an unsigned int. Handle SVE modes.
7705 (aarch64_preferred_simd_mode): Update call accordingly. Handle
7707 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
7709 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7710 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7711 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7712 (aarch64_sve_float_mul_immediate_p): New functions.
7713 (aarch64_sve_valid_immediate): New function.
7714 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
7715 Explicitly reject structure modes. Check for INDEX constants.
7716 Handle PTRUE and PFALSE constants.
7717 (aarch64_check_zero_based_sve_index_immediate): New function.
7718 (aarch64_simd_imm_zero_p): Delete.
7719 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
7720 vector modes. Accept constants in the range of CNT[BHWD].
7721 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
7722 ask for an Advanced SIMD mode.
7723 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
7724 (aarch64_simd_vector_alignment): Handle SVE predicates.
7725 (aarch64_vectorize_preferred_vector_alignment): New function.
7726 (aarch64_simd_vector_alignment_reachable): Use it instead of
7728 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
7729 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
7731 (MAX_VECT_LEN): Delete.
7732 (expand_vec_perm_d): Add a vec_flags field.
7733 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
7734 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
7735 (aarch64_evpc_ext): Don't apply a big-endian lane correction
7737 (aarch64_evpc_rev): Rename to...
7738 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
7739 (aarch64_evpc_rev_global): New function.
7740 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
7741 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
7743 (aarch64_evpc_sve_tbl): New function.
7744 (aarch64_expand_vec_perm_const_1): Update after rename of
7745 aarch64_evpc_rev. Handle SVE permutes too, trying
7746 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
7747 than aarch64_evpc_tbl.
7748 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
7749 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
7750 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
7751 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
7752 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
7753 (aarch64_expand_sve_vcond): New functions.
7754 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
7755 of aarch64_vector_mode_p.
7756 (aarch64_dwarf_poly_indeterminate_value): New function.
7757 (aarch64_compute_pressure_classes): Likewise.
7758 (aarch64_can_change_mode_class): Likewise.
7759 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
7760 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
7761 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
7762 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
7763 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
7764 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
7765 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
7766 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
7768 (Dn, Dl, Dr): Accept const as well as const_vector.
7769 (Dz): Likewise. Compare against CONST0_RTX.
7770 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
7771 of "vector" where appropriate.
7772 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
7773 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
7774 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
7775 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
7776 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
7777 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
7778 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
7779 (v_int_equiv): Extend to SVE modes.
7780 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
7782 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
7783 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
7784 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
7785 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
7786 (SVE_COND_FP_CMP): New int iterators.
7787 (perm_hilo): Handle the new unpack unspecs.
7788 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
7790 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
7791 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
7792 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
7793 (aarch64_equality_operator, aarch64_constant_vector_operand)
7794 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
7795 (aarch64_sve_nonimmediate_operand): Likewise.
7796 (aarch64_sve_general_operand): Likewise.
7797 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
7798 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
7799 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
7800 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
7801 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
7802 (aarch64_sve_float_arith_immediate): Likewise.
7803 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
7804 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
7805 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
7806 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
7807 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
7808 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
7809 (aarch64_sve_float_arith_operand): Likewise.
7810 (aarch64_sve_float_arith_with_sub_operand): Likewise.
7811 (aarch64_sve_float_mul_operand): Likewise.
7812 (aarch64_sve_vec_perm_operand): Likewise.
7813 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
7814 (aarch64_mov_operand): Accept const_poly_int and const_vector.
7815 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
7816 as well as const_vector.
7817 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
7818 in file. Use CONST0_RTX and CONSTM1_RTX.
7819 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
7820 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
7821 Use aarch64_simd_imm_zero.
7822 * config/aarch64/aarch64-sve.md: New file.
7823 * config/aarch64/aarch64.md: Include it.
7824 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
7825 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
7826 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
7827 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
7828 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
7829 (sve): New attribute.
7830 (enabled): Disable instructions with the sve attribute unless
7832 (movqi, movhi): Pass CONST_POLY_INT operaneds through
7833 aarch64_expand_mov_immediate.
7834 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
7835 CNT[BHSD] immediates.
7836 (movti): Split CONST_POLY_INT moves into two halves.
7837 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
7838 Split additions that need a temporary here if the destination
7839 is the stack pointer.
7840 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
7841 (*add<mode>3_poly_1): New instruction.
7842 (set_clobber_cc): New expander.
7844 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7846 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
7847 parameter and use it instead of GET_MODE_SIZE (innermode). Use
7848 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
7849 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
7850 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
7851 Change innermode from fixed_mode_size to machine_mode.
7852 (simplify_subreg): Update call accordingly. Handle a constant-sized
7853 subreg of a variable-length CONST_VECTOR.
7855 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7856 Alan Hayward <alan.hayward@arm.com>
7857 David Sherwood <david.sherwood@arm.com>
7859 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
7860 (add_offset_to_base): New function, split out from...
7861 (create_mem_ref): ...here. When handling a scale other than 1,
7862 check first whether the address is valid without the offset.
7863 Add it into the base if so, leaving the index and scale as-is.
7865 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7868 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
7869 fold_for_warn before checking if arg2 is INTEGER_CST.
7871 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
7873 * config/rs6000/predicates.md (load_multiple_operation): Delete.
7874 (store_multiple_operation): Delete.
7875 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
7876 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
7877 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
7878 guarded by TARGET_STRING.
7879 (rs6000_output_load_multiple): Delete.
7880 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
7881 OPTION_MASK_STRING / TARGET_STRING handling.
7882 (print_operand) <'N', 'O'>: Add comment that these are unused now.
7883 (const rs6000_opt_masks) <"string">: Change mask to 0.
7884 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
7885 (MASK_STRING): Delete.
7886 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
7888 (load_multiple): Delete.
7895 (store_multiple): Delete.
7902 (movmemsi_8reg): Delete.
7903 (corresponding unnamed define_insn): Delete.
7904 (movmemsi_6reg): Delete.
7905 (corresponding unnamed define_insn): Delete.
7906 (movmemsi_4reg): Delete.
7907 (corresponding unnamed define_insn): Delete.
7908 (movmemsi_2reg): Delete.
7909 (corresponding unnamed define_insn): Delete.
7910 (movmemsi_1reg): Delete.
7911 (corresponding unnamed define_insn): Delete.
7912 * config/rs6000/rs6000.opt (mno-string): New.
7913 (mstring): Replace by deprecation warning stub.
7914 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
7916 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7918 * regrename.c (regrename_do_replace): If replacing the same
7919 reg multiple times, try to reuse last created gen_raw_REG.
7922 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
7923 main to workaround a bug in GDB.
7925 2018-01-12 Tom de Vries <tom@codesourcery.com>
7928 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
7930 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
7932 PR rtl-optimization/80481
7933 * ira-color.c (get_cap_member): New function.
7934 (allocnos_conflict_by_live_ranges_p): Use it.
7935 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
7936 (setup_slot_coalesced_allocno_live_ranges): Ditto.
7938 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
7941 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
7942 (*saddl_se_1): Ditto.
7944 (*ssubl_se_1): Ditto.
7946 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7948 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
7949 rather than wi::to_widest for DR_INITs.
7950 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
7951 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
7952 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
7954 (vect_analyze_group_access_1): Note that here.
7956 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7958 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
7959 polynomial type sizes.
7961 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7963 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
7964 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
7965 (gimple_add_tmp_var): Likewise.
7967 2018-01-12 Martin Liska <mliska@suse.cz>
7969 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
7970 (gimple_alloc_sizes): Likewise.
7971 (dump_gimple_statistics): Use PRIu64 in printf format.
7972 * gimple.h: Change uint64_t to int.
7974 2018-01-12 Martin Liska <mliska@suse.cz>
7976 * tree-core.h: Use uint64_t instead of int.
7977 * tree.c (tree_node_counts): Likewise.
7978 (tree_node_sizes): Likewise.
7979 (dump_tree_statistics): Use PRIu64 in printf format.
7981 2018-01-12 Martin Liska <mliska@suse.cz>
7983 * Makefile.in: As qsort_chk is implemented in vec.c, add
7984 vec.o to linkage of gencfn-macros.
7985 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
7986 passing the info to record_node_allocation_statistics.
7987 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
7989 * ggc-common.c (struct ggc_usage): Add operator== and use
7990 it in operator< and compare function.
7991 * mem-stats.h (struct mem_usage): Likewise.
7992 * vec.c (struct vec_usage): Remove operator< and compare
7993 function. Can be simply inherited.
7995 2018-01-12 Martin Jambor <mjambor@suse.cz>
7998 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
7999 * tree-ssa-math-opts.c: Include domwalk.h.
8000 (convert_mult_to_fma_1): New function.
8001 (fma_transformation_info): New type.
8002 (fma_deferring_state): Likewise.
8003 (cancel_fma_deferring): New function.
8004 (result_of_phi): Likewise.
8005 (last_fma_candidate_feeds_initial_phi): Likewise.
8006 (convert_mult_to_fma): Added deferring logic, split actual
8007 transformation to convert_mult_to_fma_1.
8008 (math_opts_dom_walker): New type.
8009 (math_opts_dom_walker::after_dom_children): New method, body moved
8010 here from pass_optimize_widening_mul::execute, added deferring logic
8012 (pass_optimize_widening_mul::execute): Moved most of code to
8013 math_opts_dom_walker::after_dom_children.
8014 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8015 * config/i386/i386.c (ix86_option_override_internal): Added
8016 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8018 2018-01-12 Richard Biener <rguenther@suse.de>
8021 * dwarf2out.c (gen_variable_die): Do not reset old_die for
8022 inline instance vars.
8024 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
8027 * config/rx/rx.c (rx_is_restricted_memory_address):
8030 2018-01-12 Richard Biener <rguenther@suse.de>
8032 PR tree-optimization/80846
8033 * target.def (split_reduction): New target hook.
8034 * targhooks.c (default_split_reduction): New function.
8035 * targhooks.h (default_split_reduction): Declare.
8036 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8037 target requests first reduce vectors by combining low and high
8039 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8040 (get_vectype_for_scalar_type_and_size): Export.
8041 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8042 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8043 * doc/tm.texi: Regenerate.
8044 * config/i386/i386.c (ix86_split_reduction): Implement
8045 TARGET_VECTORIZE_SPLIT_REDUCTION.
8047 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8050 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8051 in PIC mode except for TARGET_VXWORKS_RTP.
8052 * config/sparc/sparc.c: Include cfgrtl.h.
8053 (TARGET_INIT_PIC_REG): Define.
8054 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8055 (sparc_pic_register_p): New predicate.
8056 (sparc_legitimate_address_p): Use it.
8057 (sparc_legitimize_pic_address): Likewise.
8058 (sparc_delegitimize_address): Likewise.
8059 (sparc_mode_dependent_address_p): Likewise.
8060 (gen_load_pcrel_sym): Remove 4th parameter.
8061 (load_got_register): Adjust call to above. Remove obsolete stuff.
8062 (sparc_expand_prologue): Do not call load_got_register here.
8063 (sparc_flat_expand_prologue): Likewise.
8064 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8065 (sparc_use_pseudo_pic_reg): New function.
8066 (sparc_init_pic_reg): Likewise.
8067 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8068 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8070 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8072 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8073 Add item for branch_cost.
8075 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8077 PR rtl-optimization/83565
8078 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8079 not extend the result to a larger mode for rotate operations.
8080 (num_sign_bit_copies1): Likewise.
8082 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8085 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8087 Use values-Xc.o for -pedantic.
8088 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8090 2018-01-12 Martin Liska <mliska@suse.cz>
8093 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8095 (possible_polymorphic_call_targets): Use it.
8096 (ipa_devirt): Likewise.
8098 2018-01-12 Martin Liska <mliska@suse.cz>
8100 * profile-count.h (enum profile_quality): Use 0 as invalid
8101 enum value of profile_quality.
8103 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8105 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8106 -mext-string options.
8108 2018-01-12 Richard Biener <rguenther@suse.de>
8110 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8111 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8112 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8114 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8116 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8118 * configure.ac (--with-long-double-format): Add support for the
8119 configuration option to change the default long double format on
8121 * config.gcc (powerpc*-linux*-*): Likewise.
8122 * configure: Regenerate.
8123 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8124 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8125 used without modification.
8127 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8129 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8130 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8131 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8132 MISC_BUILTIN_SPEC_BARRIER.
8133 (rs6000_init_builtins): Likewise.
8134 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8136 (speculation_barrier): New define_insn.
8137 * doc/extend.texi: Document __builtin_speculation_barrier.
8139 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8142 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8143 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8144 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8146 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8147 integral modes instead of "ss" and "sd".
8148 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8149 vectors with 32-bit and 64-bit elements.
8150 (vecdupssescalarmodesuffix): New mode attribute.
8151 (vec_dup<mode>): Use it.
8153 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8156 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8157 frame if argument is passed on stack.
8159 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8162 * ree.c (combine_reaching_defs): Optimize also
8163 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8164 reg2=any_extend(exp); reg1=reg2;, formatting fix.
8166 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8169 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8171 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8174 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8175 after they are computed.
8177 2018-01-11 Bin Cheng <bin.cheng@arm.com>
8179 PR tree-optimization/83695
8180 * gimple-loop-linterchange.cc
8181 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8182 reset cached scev information after interchange.
8183 (pass_linterchange::execute): Remove call to scev_reset_htab.
8185 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8187 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8188 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8189 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8190 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8191 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8192 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8193 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8194 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8195 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8196 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8197 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8198 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8199 (V_lane_reg): Likewise.
8200 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8202 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8203 (vfmal_lane_low<mode>_intrinsic,
8204 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8205 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8206 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8207 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8208 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8209 vfmsl_lane_high<mode>_intrinsic): New define_insns.
8211 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8213 * config/arm/arm-cpus.in (fp16fml): New feature.
8214 (ALL_SIMD): Add fp16fml.
8215 (armv8.2-a): Add fp16fml as an option.
8216 (armv8.3-a): Likewise.
8217 (armv8.4-a): Add fp16fml as part of fp16.
8218 * config/arm/arm.h (TARGET_FP16FML): Define.
8219 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
8221 * config/arm/arm-modes.def (V2HF): Define.
8222 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
8223 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
8224 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
8225 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
8226 vfmsl_low, vfmsl_high): New set of builtins.
8227 * config/arm/iterators.md (PLUSMINUS): New code iterator.
8228 (vfml_op): New code attribute.
8229 (VFMLHALVES): New int iterator.
8230 (VFML, VFMLSEL): New mode attributes.
8231 (V_reg): Define mapping for V2HF.
8232 (V_hi, V_lo): New mode attributes.
8233 (VF_constraint): Likewise.
8234 (vfml_half, vfml_half_selector): New int attributes.
8235 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
8237 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
8238 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
8240 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
8241 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
8242 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
8243 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
8245 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8246 Document new effective target and option set.
8248 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8250 * config/arm/arm-cpus.in (armv8_4): New feature.
8251 (ARMv8_4a): New fgroup.
8252 (armv8.4-a): New arch.
8253 * config/arm/arm-tables.opt: Regenerate.
8254 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8255 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8256 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8257 Add matching rules for -march=armv8.4-a and extensions.
8258 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8260 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
8263 * config/rx/rx.md (BW): New mode attribute.
8264 (sync_lock_test_and_setsi): Add mode suffix to insn output.
8266 2018-01-11 Richard Biener <rguenther@suse.de>
8268 PR tree-optimization/83435
8269 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
8270 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
8271 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
8273 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8274 Alan Hayward <alan.hayward@arm.com>
8275 David Sherwood <david.sherwood@arm.com>
8277 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8279 (aarch64_classify_address): Initialize it. Track polynomial offsets.
8280 (aarch64_print_address_internal): Use it to check for a zero offset.
8282 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8283 Alan Hayward <alan.hayward@arm.com>
8284 David Sherwood <david.sherwood@arm.com>
8286 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8287 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8288 Return a poly_int64 rather than a HOST_WIDE_INT.
8289 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8290 rather than a HOST_WIDE_INT.
8291 * config/aarch64/aarch64.h (aarch64_frame): Protect with
8292 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
8293 hard_fp_offset, frame_size, initial_adjust, callee_offset and
8294 final_offset from HOST_WIDE_INT to poly_int64.
8295 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8296 to_constant when getting the number of units in an Advanced SIMD
8298 (aarch64_builtin_vectorized_function): Check for a constant number
8300 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8302 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8303 attribute instead of GET_MODE_NUNITS.
8304 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8305 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8306 GET_MODE_SIZE for fixed-size registers.
8307 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8308 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8309 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8310 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8311 (aarch64_print_operand, aarch64_print_address_internal)
8312 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8313 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8314 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8315 Handle polynomial GET_MODE_SIZE.
8316 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8317 wider than SImode without modification.
8318 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8319 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8320 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8321 passing and returning SVE modes.
8322 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8323 rather than GEN_INT.
8324 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8325 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8326 (aarch64_allocate_and_probe_stack_space): Likewise.
8327 (aarch64_layout_frame): Cope with polynomial offsets.
8328 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8329 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8331 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8332 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8333 poly_int64 rather than a HOST_WIDE_INT.
8334 (aarch64_get_separate_components, aarch64_process_components)
8335 (aarch64_expand_prologue, aarch64_expand_epilogue)
8336 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8337 (aarch64_anchor_offset): New function, split out from...
8338 (aarch64_legitimize_address): ...here.
8339 (aarch64_builtin_vectorization_cost): Handle polynomial
8340 TYPE_VECTOR_SUBPARTS.
8341 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8343 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8344 number of elements from the PARALLEL rather than the mode.
8345 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8346 rather than GET_MODE_BITSIZE.
8347 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8348 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8349 (aarch64_expand_vec_perm_const_1): Handle polynomial
8350 d->perm.length () and d->perm elements.
8351 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
8352 Apply to_constant to d->perm elements.
8353 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8354 polynomial CONST_VECTOR_NUNITS.
8355 (aarch64_move_pointer): Take amount as a poly_int64 rather
8357 (aarch64_progress_pointer): Avoid temporary variable.
8358 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8359 the mode attribute instead of GET_MODE.
8361 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8362 Alan Hayward <alan.hayward@arm.com>
8363 David Sherwood <david.sherwood@arm.com>
8365 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8366 x exists before using it.
8367 (aarch64_add_constant_internal): Rename to...
8368 (aarch64_add_offset_1): ...this. Replace regnum with separate
8369 src and dest rtxes. Handle the case in which they're different,
8370 including when the offset is zero. Replace scratchreg with an rtx.
8371 Use 2 additions if there is no spare register into which we can
8372 move a 16-bit constant.
8373 (aarch64_add_constant): Delete.
8374 (aarch64_add_offset): Replace reg with separate src and dest
8375 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8376 Use aarch64_add_offset_1.
8377 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8378 an rtx rather than an int. Take the delta as a poly_int64
8379 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8380 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8381 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8382 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8383 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8385 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8386 aarch64_add_constant.
8388 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8390 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8391 Use scalar_float_mode.
8393 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8395 * config/aarch64/aarch64-simd.md
8396 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8397 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8398 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8399 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8400 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8401 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8402 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8403 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8404 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8405 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8407 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8410 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8411 targ_options->x_arm_arch_string is non NULL.
8413 2018-01-11 Tamar Christina <tamar.christina@arm.com>
8415 * config/aarch64/aarch64.h
8416 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
8418 2018-01-11 Sudakshina Das <sudi.das@arm.com>
8421 * expmed.c (emit_store_flag_force): Swap if const op0
8422 and change VOIDmode to mode of op0.
8424 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8426 PR rtl-optimization/83761
8427 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
8428 than bytes to mode_for_size.
8430 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8433 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
8434 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
8437 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8440 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
8441 when in layout mode.
8442 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
8443 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
8446 2018-01-10 Michael Collison <michael.collison@arm.com>
8448 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
8449 * config/aarch64/aarch64-option-extension.def: Add
8450 AARCH64_OPT_EXTENSION of 'fp16fml'.
8451 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8452 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
8453 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
8454 * config/aarch64/constraints.md (Ui7): New constraint.
8455 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
8456 (VFMLA_SEL_W): Ditto.
8459 (VFMLA16_LOW): New int iterator.
8460 (VFMLA16_HIGH): Ditto.
8461 (UNSPEC_FMLAL): New unspec.
8462 (UNSPEC_FMLSL): Ditto.
8463 (UNSPEC_FMLAL2): Ditto.
8464 (UNSPEC_FMLSL2): Ditto.
8465 (f16mac): New code attribute.
8466 * config/aarch64/aarch64-simd-builtins.def
8467 (aarch64_fmlal_lowv2sf): Ditto.
8468 (aarch64_fmlsl_lowv2sf): Ditto.
8469 (aarch64_fmlalq_lowv4sf): Ditto.
8470 (aarch64_fmlslq_lowv4sf): Ditto.
8471 (aarch64_fmlal_highv2sf): Ditto.
8472 (aarch64_fmlsl_highv2sf): Ditto.
8473 (aarch64_fmlalq_highv4sf): Ditto.
8474 (aarch64_fmlslq_highv4sf): Ditto.
8475 (aarch64_fmlal_lane_lowv2sf): Ditto.
8476 (aarch64_fmlsl_lane_lowv2sf): Ditto.
8477 (aarch64_fmlal_laneq_lowv2sf): Ditto.
8478 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
8479 (aarch64_fmlalq_lane_lowv4sf): Ditto.
8480 (aarch64_fmlsl_lane_lowv4sf): Ditto.
8481 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
8482 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
8483 (aarch64_fmlal_lane_highv2sf): Ditto.
8484 (aarch64_fmlsl_lane_highv2sf): Ditto.
8485 (aarch64_fmlal_laneq_highv2sf): Ditto.
8486 (aarch64_fmlsl_laneq_highv2sf): Ditto.
8487 (aarch64_fmlalq_lane_highv4sf): Ditto.
8488 (aarch64_fmlsl_lane_highv4sf): Ditto.
8489 (aarch64_fmlalq_laneq_highv4sf): Ditto.
8490 (aarch64_fmlsl_laneq_highv4sf): Ditto.
8491 * config/aarch64/aarch64-simd.md:
8492 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
8493 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8494 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
8495 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8496 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
8497 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
8498 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
8499 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
8500 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
8501 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
8502 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
8503 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
8504 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
8505 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
8506 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
8507 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
8508 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
8509 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
8510 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
8511 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
8512 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
8513 (vfmlsl_low_u32): Ditto.
8514 (vfmlalq_low_u32): Ditto.
8515 (vfmlslq_low_u32): Ditto.
8516 (vfmlal_high_u32): Ditto.
8517 (vfmlsl_high_u32): Ditto.
8518 (vfmlalq_high_u32): Ditto.
8519 (vfmlslq_high_u32): Ditto.
8520 (vfmlal_lane_low_u32): Ditto.
8521 (vfmlsl_lane_low_u32): Ditto.
8522 (vfmlal_laneq_low_u32): Ditto.
8523 (vfmlsl_laneq_low_u32): Ditto.
8524 (vfmlalq_lane_low_u32): Ditto.
8525 (vfmlslq_lane_low_u32): Ditto.
8526 (vfmlalq_laneq_low_u32): Ditto.
8527 (vfmlslq_laneq_low_u32): Ditto.
8528 (vfmlal_lane_high_u32): Ditto.
8529 (vfmlsl_lane_high_u32): Ditto.
8530 (vfmlal_laneq_high_u32): Ditto.
8531 (vfmlsl_laneq_high_u32): Ditto.
8532 (vfmlalq_lane_high_u32): Ditto.
8533 (vfmlslq_lane_high_u32): Ditto.
8534 (vfmlalq_laneq_high_u32): Ditto.
8535 (vfmlslq_laneq_high_u32): Ditto.
8536 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
8537 (AARCH64_FL_FOR_ARCH8_4): New.
8538 (AARCH64_ISA_F16FML): New ISA flag.
8539 (TARGET_F16FML): New feature flag for fp16fml.
8540 (doc/invoke.texi): Document new fp16fml option.
8542 2018-01-10 Michael Collison <michael.collison@arm.com>
8544 * config/aarch64/aarch64-builtins.c:
8545 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
8546 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8547 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
8548 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
8549 (AARCH64_ISA_SHA3): New ISA flag.
8550 (TARGET_SHA3): New feature flag for sha3.
8551 * config/aarch64/iterators.md (sha512_op): New int attribute.
8552 (CRYPTO_SHA512): New int iterator.
8553 (UNSPEC_SHA512H): New unspec.
8554 (UNSPEC_SHA512H2): Ditto.
8555 (UNSPEC_SHA512SU0): Ditto.
8556 (UNSPEC_SHA512SU1): Ditto.
8557 * config/aarch64/aarch64-simd-builtins.def
8558 (aarch64_crypto_sha512hqv2di): New builtin.
8559 (aarch64_crypto_sha512h2qv2di): Ditto.
8560 (aarch64_crypto_sha512su0qv2di): Ditto.
8561 (aarch64_crypto_sha512su1qv2di): Ditto.
8562 (aarch64_eor3qv8hi): Ditto.
8563 (aarch64_rax1qv2di): Ditto.
8564 (aarch64_xarqv2di): Ditto.
8565 (aarch64_bcaxqv8hi): Ditto.
8566 * config/aarch64/aarch64-simd.md:
8567 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
8568 (aarch64_crypto_sha512su0qv2di): Ditto.
8569 (aarch64_crypto_sha512su1qv2di): Ditto.
8570 (aarch64_eor3qv8hi): Ditto.
8571 (aarch64_rax1qv2di): Ditto.
8572 (aarch64_xarqv2di): Ditto.
8573 (aarch64_bcaxqv8hi): Ditto.
8574 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
8575 (vsha512h2q_u64): Ditto.
8576 (vsha512su0q_u64): Ditto.
8577 (vsha512su1q_u64): Ditto.
8578 (veor3q_u16): Ditto.
8579 (vrax1q_u64): Ditto.
8581 (vbcaxq_u16): Ditto.
8582 * config/arm/types.md (crypto_sha512): New type attribute.
8583 (crypto_sha3): Ditto.
8584 (doc/invoke.texi): Document new sha3 option.
8586 2018-01-10 Michael Collison <michael.collison@arm.com>
8588 * config/aarch64/aarch64-builtins.c:
8589 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
8590 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8591 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
8592 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
8593 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
8594 (AARCH64_ISA_SM4): New ISA flag.
8595 (TARGET_SM4): New feature flag for sm4.
8596 * config/aarch64/aarch64-simd-builtins.def
8597 (aarch64_sm3ss1qv4si): Ditto.
8598 (aarch64_sm3tt1aq4si): Ditto.
8599 (aarch64_sm3tt1bq4si): Ditto.
8600 (aarch64_sm3tt2aq4si): Ditto.
8601 (aarch64_sm3tt2bq4si): Ditto.
8602 (aarch64_sm3partw1qv4si): Ditto.
8603 (aarch64_sm3partw2qv4si): Ditto.
8604 (aarch64_sm4eqv4si): Ditto.
8605 (aarch64_sm4ekeyqv4si): Ditto.
8606 * config/aarch64/aarch64-simd.md:
8607 (aarch64_sm3ss1qv4si): Ditto.
8608 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
8609 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
8610 (aarch64_sm4eqv4si): Ditto.
8611 (aarch64_sm4ekeyqv4si): Ditto.
8612 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
8613 (sm3part_op): Ditto.
8614 (CRYPTO_SM3TT): Ditto.
8615 (CRYPTO_SM3PART): Ditto.
8616 (UNSPEC_SM3SS1): New unspec.
8617 (UNSPEC_SM3TT1A): Ditto.
8618 (UNSPEC_SM3TT1B): Ditto.
8619 (UNSPEC_SM3TT2A): Ditto.
8620 (UNSPEC_SM3TT2B): Ditto.
8621 (UNSPEC_SM3PARTW1): Ditto.
8622 (UNSPEC_SM3PARTW2): Ditto.
8623 (UNSPEC_SM4E): Ditto.
8624 (UNSPEC_SM4EKEY): Ditto.
8625 * config/aarch64/constraints.md (Ui2): New constraint.
8626 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
8627 * config/arm/types.md (crypto_sm3): New type attribute.
8628 (crypto_sm4): Ditto.
8629 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
8630 (vsm3tt1aq_u32): Ditto.
8631 (vsm3tt1bq_u32): Ditto.
8632 (vsm3tt2aq_u32): Ditto.
8633 (vsm3tt2bq_u32): Ditto.
8634 (vsm3partw1q_u32): Ditto.
8635 (vsm3partw2q_u32): Ditto.
8636 (vsm4eq_u32): Ditto.
8637 (vsm4ekeyq_u32): Ditto.
8638 (doc/invoke.texi): Document new sm4 option.
8640 2018-01-10 Michael Collison <michael.collison@arm.com>
8642 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
8643 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
8644 (AARCH64_FL_FOR_ARCH8_4): New.
8645 (AARCH64_FL_V8_4): New flag.
8646 (doc/invoke.texi): Document new armv8.4-a option.
8648 2018-01-10 Michael Collison <michael.collison@arm.com>
8650 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8651 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
8652 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
8653 * config/aarch64/aarch64-option-extension.def: Add
8654 AARCH64_OPT_EXTENSION of 'sha2'.
8655 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
8656 (crypto): Disable sha2 and aes if crypto disabled.
8657 (crypto): Enable aes and sha2 if enabled.
8658 (simd): Disable sha2 and aes if simd disabled.
8659 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
8661 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
8662 (TARGET_SHA2): New feature flag for sha2.
8663 (TARGET_AES): New feature flag for aes.
8664 * config/aarch64/aarch64-simd.md:
8665 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
8666 conditional on TARGET_AES.
8667 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
8668 (aarch64_crypto_sha1hsi): Make pattern conditional
8670 (aarch64_crypto_sha1hv4si): Ditto.
8671 (aarch64_be_crypto_sha1hv4si): Ditto.
8672 (aarch64_crypto_sha1su1v4si): Ditto.
8673 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
8674 (aarch64_crypto_sha1su0v4si): Ditto.
8675 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
8676 (aarch64_crypto_sha256su0v4si): Ditto.
8677 (aarch64_crypto_sha256su1v4si): Ditto.
8678 (doc/invoke.texi): Document new aes and sha2 options.
8680 2018-01-10 Martin Sebor <msebor@redhat.com>
8682 PR tree-optimization/83781
8683 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
8686 2018-01-11 Martin Sebor <msebor@gmail.com>
8687 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8689 PR tree-optimization/83501
8690 PR tree-optimization/81703
8692 * tree-ssa-strlen.c (get_string_cst): Rename...
8693 (get_string_len): ...to this. Handle global constants.
8694 (handle_char_store): Adjust.
8696 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
8697 Jim Wilson <jimw@sifive.com>
8699 * config/riscv/riscv-protos.h (riscv_output_return): New.
8700 * config/riscv/riscv.c (struct machine_function): New naked_p field.
8701 (riscv_attribute_table, riscv_output_return),
8702 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
8703 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
8704 (riscv_compute_frame_info): Only compute frame->mask if not a naked
8706 (riscv_expand_prologue): Add early return for naked function.
8707 (riscv_expand_epilogue): Likewise.
8708 (riscv_function_ok_for_sibcall): Return false for naked function.
8709 (riscv_set_current_function): New.
8710 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
8711 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
8712 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
8713 * doc/extend.texi (RISC-V Function Attributes): New.
8715 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8717 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
8718 check for 128-bit long double before checking TCmode.
8719 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
8720 128-bit long doubles before checking TFmode or TCmode.
8721 (FLOAT128_IBM_P): Likewise.
8723 2018-01-10 Martin Sebor <msebor@redhat.com>
8725 PR tree-optimization/83671
8726 * builtins.c (c_strlen): Unconditionally return zero for the empty
8728 Use -Warray-bounds for warnings.
8729 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
8730 for non-constant array indices with COMPONENT_REF, arrays of
8731 arrays, and pointers to arrays.
8732 (gimple_fold_builtin_strlen): Determine and set length range for
8733 non-constant character arrays.
8735 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
8738 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
8741 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
8743 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
8745 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8748 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
8749 VECTOR_MEM_ALTIVEC_OR_VSX_P.
8750 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
8751 indexed_or_indirect_operand predicate.
8752 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
8753 (*vsx_le_perm_load_v8hi): Likewise.
8754 (*vsx_le_perm_load_v16qi): Likewise.
8755 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
8756 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
8757 (*vsx_le_perm_store_v8hi): Likewise.
8758 (*vsx_le_perm_store_v16qi): Likewise.
8759 (eight unnamed splitters): Likewise.
8761 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8763 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
8764 * config/rs6000/emmintrin.h: Likewise.
8765 * config/rs6000/mmintrin.h: Likewise.
8766 * config/rs6000/xmmintrin.h: Likewise.
8768 2018-01-10 David Malcolm <dmalcolm@redhat.com>
8771 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
8773 * tree.c (tree_nop_conversion): Return true for location wrapper
8775 (maybe_wrap_with_location): New function.
8776 (selftest::check_strip_nops): New function.
8777 (selftest::test_location_wrappers): New function.
8778 (selftest::tree_c_tests): Call it.
8779 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
8780 (maybe_wrap_with_location): New decl.
8781 (EXPR_LOCATION_WRAPPER_P): New macro.
8782 (location_wrapper_p): New inline function.
8783 (tree_strip_any_location_wrapper): New inline function.
8785 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
8788 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
8789 stack_realign_offset for the largest alignment of stack slot
8791 (ix86_find_max_used_stack_alignment): New function.
8792 (ix86_finalize_stack_frame_flags): Use it. Set
8793 max_used_stack_alignment if we don't realign stack.
8794 * config/i386/i386.h (machine_function): Add
8795 max_used_stack_alignment.
8797 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
8799 * config/arm/arm.opt (-mbranch-cost): New option.
8800 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
8803 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
8806 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
8807 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
8809 2018-01-10 Richard Biener <rguenther@suse.de>
8812 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
8813 early out so it also covers the case where we have a non-NULL
8816 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8818 PR tree-optimization/83753
8819 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
8820 for non-strided grouped accesses if the number of elements is 1.
8822 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8825 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
8826 * i386.h (TARGET_USE_GATHER): Define.
8827 * x86-tune.def (X86_TUNE_USE_GATHER): New.
8829 2018-01-10 Martin Liska <mliska@suse.cz>
8832 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
8833 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
8835 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
8836 CLEANUP_NO_PARTITIONING is not set.
8838 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8840 * doc/rtl.texi: Remove documentation of (const ...) wrappers
8841 for vectors, as a partial revert of r254296.
8842 * rtl.h (const_vec_p): Delete.
8843 (const_vec_duplicate_p): Don't test for vector CONSTs.
8844 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
8845 * expmed.c (make_tree): Likewise.
8848 * common.md (E, F): Use CONSTANT_P instead of checking for
8850 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
8851 checking for CONST_VECTOR.
8853 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8856 * predict.c (force_edge_cold): Handle in more sane way edges
8859 2018-01-09 Carl Love <cel@us.ibm.com>
8861 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
8863 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
8864 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
8865 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
8866 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
8867 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
8868 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
8869 * config/rs6000/rs6000-protos.h: Add extern defition for
8870 rs6000_generate_float2_double_code.
8871 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
8873 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
8874 (float2_v2df): Add define_expand.
8876 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
8879 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
8880 op_mode in the force_to_mode call.
8882 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8884 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
8885 instead of checking each element individually.
8886 (aarch64_evpc_uzp): Likewise.
8887 (aarch64_evpc_zip): Likewise.
8888 (aarch64_evpc_ext): Likewise.
8889 (aarch64_evpc_rev): Likewise.
8890 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
8891 instead of checking each element individually. Return true without
8893 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
8894 whether all selected elements come from the same input, instead of
8895 checking each element individually. Remove calls to gen_rtx_REG,
8896 start_sequence and end_sequence and instead assert that no rtl is
8899 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8901 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
8902 order of HIGH and CONST checks.
8904 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8906 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
8907 if the destination isn't an SSA_NAME.
8909 2018-01-09 Richard Biener <rguenther@suse.de>
8911 PR tree-optimization/83668
8912 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
8914 (canonicalize_loop_form): ... here, renamed from ...
8915 (canonicalize_loop_closed_ssa_form): ... this and amended to
8916 swap successor edges for loop exit blocks to make us use
8917 the RPO order we need for initial schedule generation.
8919 2018-01-09 Joseph Myers <joseph@codesourcery.com>
8921 PR tree-optimization/64811
8922 * match.pd: When optimizing comparisons with Inf, avoid
8923 introducing or losing exceptions from comparisons with NaN.
8925 2018-01-09 Martin Liska <mliska@suse.cz>
8928 * asan.c (shadow_mem_size): Add gcc_assert.
8930 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
8932 Don't save registers in main().
8935 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
8936 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
8937 * config/avr/avr.c (avr_set_current_function): Don't error if
8938 naked, OS_task or OS_main are specified at the same time.
8939 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
8941 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
8943 * common/config/avr/avr-common.c (avr_option_optimization_table):
8944 Switch on -mmain-is-OS_task for optimizing compilations.
8946 2018-01-09 Richard Biener <rguenther@suse.de>
8948 PR tree-optimization/83572
8949 * graphite.c: Include cfganal.h.
8950 (graphite_transform_loops): Connect infinite loops to exit
8951 and remove fake edges at the end.
8953 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8955 * ipa-inline.c (edge_badness): Revert accidental checkin.
8957 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8960 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
8961 symbols; not inline clones.
8963 2018-01-09 Jakub Jelinek <jakub@redhat.com>
8966 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
8967 hard registers. Formatting fixes.
8969 PR preprocessor/83722
8970 * gcc.c (try_generate_repro): Pass
8971 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
8972 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
8975 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
8976 Kito Cheng <kito.cheng@gmail.com>
8978 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
8979 (riscv_leaf_function_p): Delete.
8980 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
8982 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8984 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
8986 (do_ifelse): New function.
8987 (do_isel): New function.
8988 (do_sub3): New function.
8989 (do_add3): New function.
8990 (do_load_mask_compare): New function.
8991 (do_overlap_load_compare): New function.
8992 (expand_compare_loop): New function.
8993 (expand_block_compare): Call expand_compare_loop() when appropriate.
8994 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
8996 (-mblock-compare-inline-loop-limit): New option.
8998 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9001 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9002 Reverse order of second and third operands in first alternative.
9003 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9004 of first and second elements in UNSPEC_VPERMR vector.
9005 (altivec_expand_vec_perm_le): Likewise.
9007 2018-01-08 Jeff Law <law@redhat.com>
9009 PR rtl-optimizatin/81308
9010 * tree-switch-conversion.c (cfg_altered): New file scoped static.
9011 (process_switch): If group_case_labels makes a change, then set
9013 (pass_convert_switch::execute): If a switch is converted, then
9014 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
9016 PR rtl-optimization/81308
9017 * recog.c (split_all_insns): Conditionally cleanup the CFG after
9020 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
9022 PR target/83663 - Revert r255946
9023 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9024 generation for cases where splatting a value is not useful.
9025 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9026 across a vec_duplicate and a paradoxical subreg forming a vector
9027 mode to a vec_concat.
9029 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9031 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9032 -march=armv8.3-a variants.
9033 * config/arm/t-multilib: Likewise.
9034 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
9036 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9038 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9040 (cceq_ior_compare_complement): Give it a name so I can use it, and
9041 change boolean_or_operator predicate to boolean_operator so it can
9042 be used to generate a crand.
9043 (eqne): New code iterator.
9044 (bd/bd_neg): New code_attrs.
9045 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9046 a single define_insn.
9047 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9048 decrement (bdnzt/bdnzf/bdzt/bdzf).
9049 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9050 with the new names of the branch decrement patterns, and added the
9051 names of the branch decrement conditional patterns.
9053 2018-01-08 Richard Biener <rguenther@suse.de>
9055 PR tree-optimization/83563
9056 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9059 2018-01-08 Richard Biener <rguenther@suse.de>
9062 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9064 2018-01-08 Richard Biener <rguenther@suse.de>
9066 PR tree-optimization/83685
9067 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9068 references to abnormals.
9070 2018-01-08 Richard Biener <rguenther@suse.de>
9073 * dwarf2out.c (output_indirect_strings): Handle empty
9074 skeleton_debug_str_hash.
9075 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9077 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9079 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9080 (emit_store_direct): Likewise.
9081 (arc_trampoline_adjust_address): Likewise.
9082 (arc_asm_trampoline_template): New function.
9083 (arc_initialize_trampoline): Use asm_trampoline_template.
9084 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9085 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9086 * config/arc/arc.md (flush_icache): Delete pattern.
9088 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9090 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9091 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9094 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9097 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9098 by not USED_FOR_TARGET.
9099 (make_pass_resolve_sw_modes): Likewise.
9101 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9103 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9106 2018-01-08 Richard Biener <rguenther@suse.de>
9109 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9111 2018-01-08 Richard Biener <rguenther@suse.de>
9114 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9116 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9119 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9120 basic blocks with a small number of successors.
9121 (convert_control_dep_chain_into_preds): Improve handling of
9123 (dump_predicates): Split apart into...
9124 (dump_pred_chain): ...here...
9125 (dump_pred_info): ...and here.
9126 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9127 (can_chain_union_be_invalidated_p): Improve check for invalidation
9129 (uninit_uses_cannot_happen): Avoid unnecessary if
9130 convert_control_dep_chain_into_preds yielded nothing.
9132 2018-01-06 Martin Sebor <msebor@redhat.com>
9134 PR tree-optimization/83640
9135 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9136 subtracting negative offset from size.
9137 (builtin_access::overlap): Adjust offset bounds of the access to fall
9138 within the size of the object if possible.
9140 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9142 PR rtl-optimization/83699
9143 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9144 extract_bit_field_as_subreg to cases in which the extracted
9145 value is also a vector.
9147 * lra-constraints.c (process_alt_operands): Test for the equivalence
9148 substitutions when detecting a possible reload cycle.
9150 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9153 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9154 by default if flag_selective_schedling{,2}. Formatting fixes.
9156 PR rtl-optimization/83682
9157 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9158 if it has non-VECTOR_MODE element mode.
9159 (vec_duplicate_p): Likewise.
9162 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9163 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9165 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9168 * config/i386/i386-builtin.def
9169 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9170 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9171 Require also OPTION_MASK_ISA_AVX512F in addition to
9172 OPTION_MASK_ISA_GFNI.
9173 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9174 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9175 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9176 to OPTION_MASK_ISA_GFNI.
9177 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9178 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9179 OPTION_MASK_ISA_AVX512BW.
9180 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9181 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9182 addition to OPTION_MASK_ISA_GFNI.
9183 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9184 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9185 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9186 to OPTION_MASK_ISA_GFNI.
9187 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9188 a requirement for all ISAs rather than any of them with a few
9190 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9192 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9193 bitmasks to be enabled with 3 exceptions, instead of requiring any
9194 enabled ISA with lots of exceptions.
9195 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9196 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9197 Change avx512bw in isa attribute to avx512f.
9198 * config/i386/sgxintrin.h: Add license boilerplate.
9199 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9200 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9201 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9202 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9204 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9205 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9206 temporarily sse2 rather than sse if not enabled already.
9209 * config/i386/sse.md (VI248_VLBW): Rename to ...
9210 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
9211 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9212 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9213 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9214 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9215 mode iterator instead of VI248_VLBW.
9217 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
9219 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
9220 (record_modified): Skip clobbers; add debug output.
9221 (param_change_prob): Use sreal frequencies.
9223 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9225 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
9226 punt for user-aligned variables.
9228 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9230 * tree-chrec.c (chrec_contains_symbols): Return true for
9233 2018-01-05 Sudakshina Das <sudi.das@arm.com>
9236 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
9237 of (x|y) == x for BICS pattern.
9239 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9241 PR tree-optimization/83605
9242 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
9243 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9246 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
9248 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9249 * config/epiphany/rtems.h: New file.
9251 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9252 Uros Bizjak <ubizjak@gmail.com>
9255 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9256 QIreg_operand instead of register_operand predicate.
9257 * config/i386/i386.c (ix86_rop_should_change_byte_p,
9258 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9259 comments instead of -fmitigate[-_]rop.
9261 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9264 * cgraphunit.c (symbol_table::compile): Switch to text_section
9265 before calling assembly_start debug hook.
9266 * run-rtl-passes.c (run_rtl_passes): Likewise.
9269 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9271 * tree-vrp.c (extract_range_from_binary_expr_1): Check
9272 range_int_cst_p rather than !symbolic_range_p before calling
9273 extract_range_from_multiplicative_op_1.
9275 2018-01-04 Jeff Law <law@redhat.com>
9277 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9278 redundant test in assertion.
9280 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9282 * doc/rtl.texi: Document machine_mode wrapper classes.
9284 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9286 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9289 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9291 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9292 the VEC_PERM_EXPR fold to fail.
9294 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9297 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9298 to switched_sections.
9300 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9303 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9306 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9309 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9310 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9312 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9315 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9316 is BLKmode and bitpos not zero or mode change is needed.
9318 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9321 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9324 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9327 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9328 instead of MULT rtx. Update all corresponding splitters.
9330 (*ssub<modesuffix>): Ditto.
9332 (*cmp_sadd_di): Update split patterns.
9333 (*cmp_sadd_si): Ditto.
9334 (*cmp_sadd_sidi): Ditto.
9335 (*cmp_ssub_di): Ditto.
9336 (*cmp_ssub_si): Ditto.
9337 (*cmp_ssub_sidi): Ditto.
9338 * config/alpha/predicates.md (const23_operand): New predicate.
9339 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9340 Look for ASHIFT, not MULT inner operand.
9341 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9343 2018-01-04 Martin Liska <mliska@suse.cz>
9345 PR gcov-profile/83669
9346 * gcov.c (output_intermediate_file): Add version to intermediate
9348 * doc/gcov.texi: Document new field 'version' in intermediate
9349 file format. Fix location of '-k' option of gcov command.
9351 2018-01-04 Martin Liska <mliska@suse.cz>
9354 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9356 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9358 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9360 2018-01-03 Martin Sebor <msebor@redhat.com>
9362 PR tree-optimization/83655
9363 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9364 checking calls with invalid arguments.
9366 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9368 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9369 (vectorizable_mask_load_store): Delete.
9370 (vectorizable_call): Return false for masked loads and stores.
9371 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9372 instead of gimple_assign_rhs1.
9373 (vectorizable_load): Handle IFN_MASK_LOAD.
9374 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9376 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9378 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9380 (vectorizable_mask_load_store): ...here.
9381 (vectorizable_load): ...and here.
9383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9385 * tree-vect-stmts.c (vect_build_all_ones_mask)
9386 (vect_build_zero_merge_argument): New functions, split out from...
9387 (vectorizable_load): ...here.
9389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9391 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9393 (vectorizable_mask_load_store): ...here.
9394 (vectorizable_store): ...and here.
9396 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9398 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9400 (vectorizable_mask_load_store): ...here.
9402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9404 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9405 (vect_model_store_cost): Take a vec_load_store_type instead of a
9407 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9408 (vect_model_store_cost): Take a vec_load_store_type instead of a
9410 (vectorizable_mask_load_store): Update accordingly.
9411 (vectorizable_store): Likewise.
9412 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9414 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9416 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9417 IFN_MASK_LOAD calls here rather than...
9418 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9420 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9421 Alan Hayward <alan.hayward@arm.com>
9422 David Sherwood <david.sherwood@arm.com>
9424 * expmed.c (extract_bit_field_1): For vector extracts,
9425 fall back to extract_bit_field_as_subreg if vec_extract
9428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9429 Alan Hayward <alan.hayward@arm.com>
9430 David Sherwood <david.sherwood@arm.com>
9432 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
9433 they are variable or constant sized.
9434 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
9435 slots for constant-sized data.
9437 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9438 Alan Hayward <alan.hayward@arm.com>
9439 David Sherwood <david.sherwood@arm.com>
9441 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
9442 handling COND_EXPRs with boolean comparisons, try to find a better
9443 basis for the mask type than the boolean itself.
9445 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9447 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
9448 is calculated and how it can be overridden.
9449 * genmodes.c (max_bitsize_mode_any_mode): New variable.
9450 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
9452 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
9455 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9456 Alan Hayward <alan.hayward@arm.com>
9457 David Sherwood <david.sherwood@arm.com>
9459 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
9460 Remove the mode argument.
9461 (aarch64_simd_valid_immediate): Remove the mode and inverse
9463 * config/aarch64/iterators.md (bitsize): New iterator.
9464 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
9465 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
9466 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
9467 aarch64_simd_valid_immediate.
9468 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
9469 (aarch64_reg_or_bic_imm): Likewise.
9470 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
9471 with an insn_type enum and msl with a modifier_type enum.
9472 Replace element_width with a scalar_mode. Change the shift
9473 to unsigned int. Add constructors for scalar_float_mode and
9474 scalar_int_mode elements.
9475 (aarch64_vect_float_const_representable_p): Delete.
9476 (aarch64_can_const_movi_rtx_p)
9477 (aarch64_simd_scalar_immediate_valid_for_move)
9478 (aarch64_simd_make_constant): Update call to
9479 aarch64_simd_valid_immediate.
9480 (aarch64_advsimd_valid_immediate_hs): New function.
9481 (aarch64_advsimd_valid_immediate): Likewise.
9482 (aarch64_simd_valid_immediate): Remove mode and inverse
9483 arguments. Rewrite to use the above. Use const_vec_duplicate_p
9484 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
9485 and aarch64_float_const_representable_p on the result.
9486 (aarch64_output_simd_mov_immediate): Remove mode argument.
9487 Update call to aarch64_simd_valid_immediate and use of
9488 simd_immediate_info.
9489 (aarch64_output_scalar_simd_mov_immediate): Update call
9492 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9493 Alan Hayward <alan.hayward@arm.com>
9494 David Sherwood <david.sherwood@arm.com>
9496 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
9497 (mode_nunits): Likewise CONST_MODE_NUNITS.
9498 * machmode.def (ADJUST_NUNITS): Document.
9499 * genmodes.c (mode_data::need_nunits_adj): New field.
9500 (blank_mode): Update accordingly.
9501 (adj_nunits): New variable.
9502 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
9504 (emit_mode_size_inline): Set need_bytesize_adj for all modes
9505 listed in adj_nunits.
9506 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
9507 listed in adj_nunits. Don't emit case statements for such modes.
9508 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
9509 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
9510 nothing if adj_nunits is nonnull.
9511 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
9512 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
9513 (emit_mode_fbit): Update use of print_maybe_const_decl.
9514 (emit_move_size): Likewise. Treat the array as non-const
9516 (emit_mode_adjustments): Handle adj_nunits.
9518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9520 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
9521 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
9522 (VECTOR_MODES): Use it.
9523 (make_vector_modes): Take the prefix as an argument.
9525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9526 Alan Hayward <alan.hayward@arm.com>
9527 David Sherwood <david.sherwood@arm.com>
9529 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
9530 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
9531 for MODE_VECTOR_BOOL.
9532 * machmode.def (VECTOR_BOOL_MODE): Document.
9533 * genmodes.c (VECTOR_BOOL_MODE): New macro.
9534 (make_vector_bool_mode): New function.
9535 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
9537 * lto-streamer-in.c (lto_input_mode_table): Likewise.
9538 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
9540 * stor-layout.c (int_mode_for_mode): Likewise.
9541 * tree.c (build_vector_type_for_mode): Likewise.
9542 * varasm.c (output_constant_pool_2): Likewise.
9543 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
9544 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
9545 for MODE_VECTOR_BOOL.
9546 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
9547 of mode class checks.
9548 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
9549 instead of a list of mode class checks.
9550 (expand_vector_scalar_condition): Likewise.
9551 (type_for_widest_vector_mode): Handle BImode as an inner mode.
9553 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9554 Alan Hayward <alan.hayward@arm.com>
9555 David Sherwood <david.sherwood@arm.com>
9557 * machmode.h (mode_size): Change from unsigned short to
9559 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
9560 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9561 or if measurement_type is not polynomial.
9562 (fixed_size_mode::includes_p): Check for constant-sized modes.
9563 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
9564 return a poly_uint16 rather than an unsigned short.
9565 (emit_mode_size): Change the type of mode_size from unsigned short
9566 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
9567 (emit_mode_adjustments): Cope with polynomial vector sizes.
9568 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9570 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9572 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
9573 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
9574 * caller-save.c (setup_save_areas): Likewise.
9575 (replace_reg_with_saved_mem): Likewise.
9576 * calls.c (emit_library_call_value_1): Likewise.
9577 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
9578 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
9579 (gen_lowpart_for_combine): Likewise.
9580 * convert.c (convert_to_integer_1): Likewise.
9581 * cse.c (equiv_constant, cse_insn): Likewise.
9582 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
9583 (cselib_subst_to_values): Likewise.
9584 * dce.c (word_dce_process_block): Likewise.
9585 * df-problems.c (df_word_lr_mark_ref): Likewise.
9586 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
9587 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
9588 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
9589 (rtl_for_decl_location): Likewise.
9590 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
9591 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
9592 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
9593 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
9594 (expand_expr_real_1): Likewise.
9595 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
9596 (pad_below): Likewise.
9597 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9598 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
9599 * ira.c (get_subreg_tracking_sizes): Likewise.
9600 * ira-build.c (ira_create_allocno_objects): Likewise.
9601 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
9602 (ira_sort_regnos_for_alter_reg): Likewise.
9603 * ira-costs.c (record_operand_costs): Likewise.
9604 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
9605 (resolve_simple_move): Likewise.
9606 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
9607 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
9608 (lra_constraints): Likewise.
9609 (CONST_POOL_OK_P): Reject variable-sized modes.
9610 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
9611 (add_pseudo_to_slot, lra_spill): Likewise.
9612 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9613 * optabs-query.c (get_best_extraction_insn): Likewise.
9614 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9615 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
9616 (expand_mult_highpart, valid_multiword_target_p): Likewise.
9617 * recog.c (offsettable_address_addr_space_p): Likewise.
9618 * regcprop.c (maybe_mode_change): Likewise.
9619 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
9620 * regrename.c (build_def_use): Likewise.
9621 * regstat.c (dump_reg_info): Likewise.
9622 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
9623 (find_reloads, find_reloads_subreg_address): Likewise.
9624 * reload1.c (eliminate_regs_1): Likewise.
9625 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
9626 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
9627 (simplify_binary_operation_1, simplify_subreg): Likewise.
9628 * targhooks.c (default_function_arg_padding): Likewise.
9629 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
9630 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
9631 (verify_gimple_assign_ternary): Likewise.
9632 * tree-inline.c (estimate_move_cost): Likewise.
9633 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9634 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
9635 (get_address_cost_ainc): Likewise.
9636 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9637 (vect_supportable_dr_alignment): Likewise.
9638 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9639 (vectorizable_reduction): Likewise.
9640 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
9641 (vectorizable_operation, vectorizable_load): Likewise.
9642 * tree.c (build_same_sized_truth_vector_type): Likewise.
9643 * valtrack.c (cleanup_auto_inc_dec): Likewise.
9644 * var-tracking.c (emit_note_insn_var_location): Likewise.
9645 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
9646 (ADDR_VEC_ALIGN): Likewise.
9648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9649 Alan Hayward <alan.hayward@arm.com>
9650 David Sherwood <david.sherwood@arm.com>
9652 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
9654 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9655 or if measurement_type is polynomial.
9656 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
9657 * combine.c (make_extraction): Likewise.
9658 * dse.c (find_shift_sequence): Likewise.
9659 * dwarf2out.c (mem_loc_descriptor): Likewise.
9660 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
9661 (extract_bit_field, extract_low_bits): Likewise.
9662 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
9663 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
9664 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
9665 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
9666 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9667 * reload.c (find_reloads): Likewise.
9668 * reload1.c (alter_reg): Likewise.
9669 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
9670 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
9671 * tree-if-conv.c (predicate_mem_writes): Likewise.
9672 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
9673 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
9674 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9675 * valtrack.c (dead_debug_insert_temp): Likewise.
9676 * varasm.c (mergeable_constant_section): Likewise.
9677 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
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 * expr.c (expand_assignment): Cope with polynomial mode sizes
9684 when assigning to a CONCAT.
9686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9687 Alan Hayward <alan.hayward@arm.com>
9688 David Sherwood <david.sherwood@arm.com>
9690 * machmode.h (mode_precision): Change from unsigned short to
9692 (mode_to_precision): Return a poly_uint16 rather than an unsigned
9694 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
9695 or if measurement_type is not polynomial.
9696 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
9697 in which the mode is already known to be a scalar_int_mode.
9698 * genmodes.c (emit_mode_precision): Change the type of mode_precision
9699 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
9701 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9702 for GET_MODE_PRECISION.
9703 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9704 for GET_MODE_PRECISION.
9705 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
9707 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
9708 (expand_field_assignment, make_extraction): Likewise.
9709 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
9710 (get_last_value): Likewise.
9711 * convert.c (convert_to_integer_1): Likewise.
9712 * cse.c (cse_insn): Likewise.
9713 * expr.c (expand_expr_real_1): Likewise.
9714 * lra-constraints.c (simplify_operand_subreg): Likewise.
9715 * optabs-query.c (can_atomic_load_p): Likewise.
9716 * optabs.c (expand_atomic_load): Likewise.
9717 (expand_atomic_store): Likewise.
9718 * ree.c (combine_reaching_defs): Likewise.
9719 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
9720 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
9721 * tree.h (type_has_mode_precision_p): Likewise.
9722 * ubsan.c (instrument_si_overflow): Likewise.
9724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9725 Alan Hayward <alan.hayward@arm.com>
9726 David Sherwood <david.sherwood@arm.com>
9728 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
9729 polynomial numbers of units.
9730 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
9731 (valid_vector_subparts_p): New function.
9732 (build_vector_type): Remove temporary shim and take the number
9733 of units as a poly_uint64 rather than an int.
9734 (build_opaque_vector_type): Take the number of units as a
9735 poly_uint64 rather than an int.
9736 * tree.c (build_vector_from_ctor): Handle polynomial
9737 TYPE_VECTOR_SUBPARTS.
9738 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
9739 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
9740 (build_vector_from_val): If the number of units is variable,
9741 use build_vec_duplicate_cst for constant operands and
9742 VEC_DUPLICATE_EXPR otherwise.
9743 (make_vector_type): Remove temporary is_constant ().
9744 (build_vector_type, build_opaque_vector_type): Take the number of
9745 units as a poly_uint64 rather than an int.
9746 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
9748 * cfgexpand.c (expand_debug_expr): Likewise.
9749 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
9750 (store_constructor, expand_expr_real_1): Likewise.
9751 (const_scalar_mask_from_tree): Likewise.
9752 * fold-const-call.c (fold_const_reduction): Likewise.
9753 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
9754 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
9755 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
9756 (fold_relational_const): Likewise.
9757 (native_interpret_vector): Likewise. Change the size from an
9758 int to an unsigned int.
9759 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
9760 TYPE_VECTOR_SUBPARTS.
9761 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
9762 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
9763 duplicating a non-constant operand into a variable-length vector.
9764 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
9765 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
9766 * ipa-icf.c (sem_variable::equals): Likewise.
9767 * match.pd: Likewise.
9768 * omp-simd-clone.c (simd_clone_subparts): Likewise.
9769 * print-tree.c (print_node): Likewise.
9770 * stor-layout.c (layout_type): Likewise.
9771 * targhooks.c (default_builtin_vectorization_cost): Likewise.
9772 * tree-cfg.c (verify_gimple_comparison): Likewise.
9773 (verify_gimple_assign_binary): Likewise.
9774 (verify_gimple_assign_ternary): Likewise.
9775 (verify_gimple_assign_single): Likewise.
9776 * tree-pretty-print.c (dump_generic_node): Likewise.
9777 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9778 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
9779 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
9780 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
9781 (vect_shift_permute_load_chain): Likewise.
9782 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
9783 (expand_vector_condition, optimize_vector_constructor): Likewise.
9784 (lower_vec_perm, get_compute_type): Likewise.
9785 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9786 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
9787 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
9788 (vect_recog_mask_conversion_pattern): Likewise.
9789 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
9790 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
9791 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9792 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
9793 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
9794 (vectorizable_shift, vectorizable_operation, vectorizable_store)
9795 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
9796 (supportable_widening_operation): Likewise.
9797 (supportable_narrowing_operation): Likewise.
9798 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
9800 * varasm.c (output_constant): Likewise.
9802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9803 Alan Hayward <alan.hayward@arm.com>
9804 David Sherwood <david.sherwood@arm.com>
9806 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
9807 so that both the length == 3 and length != 3 cases set up their
9808 own permute vectors. Add comments explaining why we know the
9809 number of elements is constant.
9810 (vect_permute_load_chain): Likewise.
9812 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9813 Alan Hayward <alan.hayward@arm.com>
9814 David Sherwood <david.sherwood@arm.com>
9816 * machmode.h (mode_nunits): Change from unsigned char to
9818 (ONLY_FIXED_SIZE_MODES): New macro.
9819 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
9820 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
9821 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
9823 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
9824 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
9825 or if measurement_type is not polynomial.
9826 * genmodes.c (ZERO_COEFFS): New macro.
9827 (emit_mode_nunits_inline): Make mode_nunits_inline return a
9829 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
9830 Use ZERO_COEFFS when emitting initializers.
9831 * data-streamer.h (bp_pack_poly_value): New function.
9832 (bp_unpack_poly_value): Likewise.
9833 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9834 for GET_MODE_NUNITS.
9835 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9836 for GET_MODE_NUNITS.
9837 * tree.c (make_vector_type): Remove temporary shim and make
9838 the real function take the number of units as a poly_uint64
9840 (build_vector_type_for_mode): Handle polynomial nunits.
9841 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
9842 * emit-rtl.c (const_vec_series_p_1): Likewise.
9843 (gen_rtx_CONST_VECTOR): Likewise.
9844 * fold-const.c (test_vec_duplicate_folding): Likewise.
9845 * genrecog.c (validate_pattern): Likewise.
9846 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
9847 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9848 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
9849 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
9850 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
9851 * rtlanal.c (subreg_get_info): Likewise.
9852 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9853 (vect_grouped_load_supported): Likewise.
9854 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
9855 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
9856 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
9857 (simplify_const_unary_operation, simplify_binary_operation_1)
9858 (simplify_const_binary_operation, simplify_ternary_operation)
9859 (test_vector_ops_duplicate, test_vector_ops): Likewise.
9860 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
9861 instead of CONST_VECTOR_NUNITS.
9862 * varasm.c (output_constant_pool_2): Likewise.
9863 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
9864 explicit-encoded elements in the XVEC for variable-length vectors.
9866 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9868 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
9870 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9871 Alan Hayward <alan.hayward@arm.com>
9872 David Sherwood <david.sherwood@arm.com>
9874 * coretypes.h (fixed_size_mode): Declare.
9875 (fixed_size_mode_pod): New typedef.
9876 * builtins.h (target_builtins::x_apply_args_mode)
9877 (target_builtins::x_apply_result_mode): Change type to
9878 fixed_size_mode_pod.
9879 * builtins.c (apply_args_size, apply_result_size, result_vector)
9880 (expand_builtin_apply_args_1, expand_builtin_apply)
9881 (expand_builtin_return): Update accordingly.
9883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9885 * cse.c (hash_rtx_cb): Hash only the encoded elements.
9886 * cselib.c (cselib_hash_rtx): Likewise.
9887 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
9888 CONST_VECTOR encoding.
9890 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9891 Jeff Law <law@redhat.com>
9894 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
9895 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
9896 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
9897 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
9900 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
9901 explicitly probe *sp in a noreturn function if there were any callee
9902 register saves or frame pointer is needed.
9904 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9907 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
9908 BLKmode for ternary, binary or unary expressions.
9911 * var-tracking.c (delete_vta_debug_insn): New inline function.
9912 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
9913 insns from get_insns () to NULL instead of each bb separately.
9914 Use delete_vta_debug_insn. No longer static.
9915 (vt_debug_insns_local, variable_tracking_main_1): Adjust
9916 delete_vta_debug_insns callers.
9917 * rtl.h (delete_vta_debug_insns): Declare.
9918 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
9919 instead of variable_tracking_main.
9921 2018-01-03 Martin Sebor <msebor@redhat.com>
9923 PR tree-optimization/83603
9924 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
9925 arguments past the endof the argument list in functions declared
9926 without a prototype.
9927 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
9928 Avoid checking when arguments are null.
9930 2018-01-03 Martin Sebor <msebor@redhat.com>
9933 * doc/extend.texi (attribute const): Fix a typo.
9934 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
9935 issuing -Wsuggest-attribute for void functions.
9937 2018-01-03 Martin Sebor <msebor@redhat.com>
9939 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
9940 offset_int::from instead of wide_int::to_shwi.
9941 (maybe_diag_overlap): Remove assertion.
9942 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
9943 * gimple-ssa-sprintf.c (format_directive): Same.
9944 (parse_directive): Same.
9945 (sprintf_dom_walker::compute_format_length): Same.
9946 (try_substitute_return_value): Same.
9948 2018-01-03 Jeff Law <law@redhat.com>
9951 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
9952 non-constant residual for zero at runtime and avoid probing in
9953 that case. Reorganize code for trailing problem to mirror handling
9956 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9958 PR tree-optimization/83501
9959 * tree-ssa-strlen.c (get_string_cst): New.
9960 (handle_char_store): Call get_string_cst.
9962 2018-01-03 Martin Liska <mliska@suse.cz>
9964 PR tree-optimization/83593
9965 * tree-ssa-strlen.c: Include tree-cfg.h.
9966 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
9967 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
9968 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
9970 (strlen_dom_walker::before_dom_children): Call
9971 gimple_purge_dead_eh_edges. Dump tranformation with details
9973 (strlen_dom_walker::before_dom_children): Update call by adding
9974 new argument cleanup_eh.
9975 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
9977 2018-01-03 Martin Liska <mliska@suse.cz>
9980 * cif-code.def (VARIADIC_THUNK): New enum value.
9981 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
9984 2018-01-03 Jan Beulich <jbeulich@suse.com>
9986 * sse.md (mov<mode>_internal): Tighten condition for when to use
9987 vmovdqu<ssescalarsize> for TI and OI modes.
9989 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9991 Update copyright years.
9993 2018-01-03 Martin Liska <mliska@suse.cz>
9996 * ipa-visibility.c (function_and_variable_visibility): Skip
9997 functions with noipa attribure.
9999 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10001 * gcc.c (process_command): Update copyright notice dates.
10002 * gcov-dump.c (print_version): Ditto.
10003 * gcov.c (print_version): Ditto.
10004 * gcov-tool.c (print_version): Ditto.
10005 * gengtype.c (create_file): Ditto.
10006 * doc/cpp.texi: Bump @copying's copyright year.
10007 * doc/cppinternals.texi: Ditto.
10008 * doc/gcc.texi: Ditto.
10009 * doc/gccint.texi: Ditto.
10010 * doc/gcov.texi: Ditto.
10011 * doc/install.texi: Ditto.
10012 * doc/invoke.texi: Ditto.
10014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10016 * vector-builder.h (vector_builder::m_full_nelts): Change from
10017 unsigned int to poly_uint64.
10018 (vector_builder::full_nelts): Update prototype accordingly.
10019 (vector_builder::new_vector): Likewise.
10020 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10021 (vector_builder::operator ==): Likewise.
10022 (vector_builder::finalize): Likewise.
10023 * int-vector-builder.h (int_vector_builder::int_vector_builder):
10024 Take the number of elements as a poly_uint64 rather than an
10026 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10027 from unsigned int to poly_uint64.
10028 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10029 (vec_perm_indices::new_vector): Likewise.
10030 (vec_perm_indices::length): Likewise.
10031 (vec_perm_indices::nelts_per_input): Likewise.
10032 (vec_perm_indices::input_nelts): Likewise.
10033 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10034 number of elements per input as a poly_uint64 rather than an
10035 unsigned int. Use the original encoding for variable-length
10036 vectors, rather than clamping each individual element.
10037 For the second and subsequent elements in each pattern,
10038 clamp the step and base before clamping their sum.
10039 (vec_perm_indices::series_p): Handle polynomial element counts.
10040 (vec_perm_indices::all_in_range_p): Likewise.
10041 (vec_perm_indices_to_tree): Likewise.
10042 (vec_perm_indices_to_rtx): Likewise.
10043 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10044 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10045 (tree_vector_builder::new_binary_operation): Handle polynomial
10046 element counts. Return false if we need to know the number
10047 of elements at compile time.
10048 * fold-const.c (fold_vec_perm): Punt if the number of elements
10049 isn't known at compile time.
10051 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10053 * vec-perm-indices.h (vec_perm_builder): Change element type
10054 from HOST_WIDE_INT to poly_int64.
10055 (vec_perm_indices::element_type): Update accordingly.
10056 (vec_perm_indices::clamp): Handle polynomial element_types.
10057 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10058 (vec_perm_indices::all_in_range_p): Likewise.
10059 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10061 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10062 polynomial vec_perm_indices element types.
10063 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10064 * fold-const.c (fold_vec_perm): Likewise.
10065 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10066 * tree-vect-generic.c (lower_vec_perm): Likewise.
10067 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10068 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10069 element type to HOST_WIDE_INT.
10071 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10072 Alan Hayward <alan.hayward@arm.com>
10073 David Sherwood <david.sherwood@arm.com>
10075 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10076 rather than an int. Use plus_constant.
10077 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10078 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10080 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10081 Alan Hayward <alan.hayward@arm.com>
10082 David Sherwood <david.sherwood@arm.com>
10084 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10085 a HOST_WIDE_INT to a poly_int64.
10087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10088 Alan Hayward <alan.hayward@arm.com>
10089 David Sherwood <david.sherwood@arm.com>
10091 * calls.c (load_register_parameters): Cope with polynomial
10092 mode sizes. Require a constant size for BLKmode parameters
10093 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10094 forces a parameter to be padded at the lsb end in order to
10095 fill a complete number of words, require the parameter size
10096 to be ordered wrt UNITS_PER_WORD.
10098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10099 Alan Hayward <alan.hayward@arm.com>
10100 David Sherwood <david.sherwood@arm.com>
10102 * reload1.c (spill_stack_slot_width): Change element type
10103 from unsigned int to poly_uint64_pod.
10104 (alter_reg): Treat mode sizes as polynomial.
10106 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10107 Alan Hayward <alan.hayward@arm.com>
10108 David Sherwood <david.sherwood@arm.com>
10110 * reload.c (complex_word_subreg_p): New function.
10111 (reload_inner_reg_of_subreg, push_reload): Use it.
10113 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10114 Alan Hayward <alan.hayward@arm.com>
10115 David Sherwood <david.sherwood@arm.com>
10117 * lra-constraints.c (process_alt_operands): Reject matched
10118 operands whose sizes aren't ordered.
10119 (match_reload): Refer to this check here.
10121 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10122 Alan Hayward <alan.hayward@arm.com>
10123 David Sherwood <david.sherwood@arm.com>
10125 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10126 that the mode size is in the set {1, 2, 4, 8, 16}.
10128 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10129 Alan Hayward <alan.hayward@arm.com>
10130 David Sherwood <david.sherwood@arm.com>
10132 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10133 Use plus_constant instead of gen_rtx_PLUS.
10135 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10136 Alan Hayward <alan.hayward@arm.com>
10137 David Sherwood <david.sherwood@arm.com>
10139 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10140 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10141 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10142 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10143 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10144 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10145 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10146 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10147 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10148 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10150 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10151 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10152 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10153 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10154 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10155 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10156 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10157 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10158 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10159 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10161 * expr.c (emit_move_resolve_push): Treat the input and result
10162 of PUSH_ROUNDING as a poly_int64.
10163 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10164 (emit_push_insn): Likewise.
10165 * lra-eliminations.c (mark_not_eliminable): Likewise.
10166 * recog.c (push_operand): Likewise.
10167 * reload1.c (elimination_effects): Likewise.
10168 * rtlanal.c (nonzero_bits1): Likewise.
10169 * calls.c (store_one_arg): Likewise. Require the padding to be
10170 known at compile time.
10172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10173 Alan Hayward <alan.hayward@arm.com>
10174 David Sherwood <david.sherwood@arm.com>
10176 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10177 Use plus_constant instead of gen_rtx_PLUS.
10179 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10180 Alan Hayward <alan.hayward@arm.com>
10181 David Sherwood <david.sherwood@arm.com>
10183 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10184 rather than an int.
10186 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10187 Alan Hayward <alan.hayward@arm.com>
10188 David Sherwood <david.sherwood@arm.com>
10190 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10191 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10192 via stack temporaries. Treat the mode size as polynomial too.
10194 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10195 Alan Hayward <alan.hayward@arm.com>
10196 David Sherwood <david.sherwood@arm.com>
10198 * expr.c (expand_expr_real_2): When handling conversions involving
10199 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10200 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10201 as a poly_uint64 too.
10203 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10204 Alan Hayward <alan.hayward@arm.com>
10205 David Sherwood <david.sherwood@arm.com>
10207 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
10209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10210 Alan Hayward <alan.hayward@arm.com>
10211 David Sherwood <david.sherwood@arm.com>
10213 * combine.c (can_change_dest_mode): Handle polynomial
10214 REGMODE_NATURAL_SIZE.
10215 * expmed.c (store_bit_field_1): Likewise.
10216 * expr.c (store_constructor): Likewise.
10217 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10218 and polynomial REGMODE_NATURAL_SIZE.
10219 (gen_lowpart_common): Likewise.
10220 * reginfo.c (record_subregs_of_mode): Likewise.
10221 * rtlanal.c (read_modify_subreg_p): Likewise.
10223 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10224 Alan Hayward <alan.hayward@arm.com>
10225 David Sherwood <david.sherwood@arm.com>
10227 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
10228 numbers of elements.
10230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10231 Alan Hayward <alan.hayward@arm.com>
10232 David Sherwood <david.sherwood@arm.com>
10234 * match.pd: Cope with polynomial numbers of vector elements.
10236 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10237 Alan Hayward <alan.hayward@arm.com>
10238 David Sherwood <david.sherwood@arm.com>
10240 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
10241 in a POINTER_PLUS_EXPR.
10243 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10244 Alan Hayward <alan.hayward@arm.com>
10245 David Sherwood <david.sherwood@arm.com>
10247 * omp-simd-clone.c (simd_clone_subparts): New function.
10248 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10249 (ipa_simd_modify_function_body): Likewise.
10251 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10252 Alan Hayward <alan.hayward@arm.com>
10253 David Sherwood <david.sherwood@arm.com>
10255 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
10256 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10257 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10258 (expand_vector_condition, vector_element): Likewise.
10259 (subparts_gt): New function.
10260 (get_compute_type): Use subparts_gt.
10261 (count_type_subparts): Delete.
10262 (expand_vector_operations_1): Use subparts_gt instead of
10263 count_type_subparts.
10265 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10266 Alan Hayward <alan.hayward@arm.com>
10267 David Sherwood <david.sherwood@arm.com>
10269 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
10270 (vect_compile_time_alias): ...this new function. Do the calculation
10271 on poly_ints rather than trees.
10272 (vect_prune_runtime_alias_test_list): Update call accordingly.
10274 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10275 Alan Hayward <alan.hayward@arm.com>
10276 David Sherwood <david.sherwood@arm.com>
10278 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
10280 (vect_schedule_slp_instance): Likewise.
10282 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10283 Alan Hayward <alan.hayward@arm.com>
10284 David Sherwood <david.sherwood@arm.com>
10286 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
10287 constant and extern definitions for variable-length vectors.
10288 (vect_get_constant_vectors): Note that the number of units
10289 is known to be constant.
10291 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10292 Alan Hayward <alan.hayward@arm.com>
10293 David Sherwood <david.sherwood@arm.com>
10295 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
10296 of units as polynomial. Choose between WIDE and NARROW based
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-vect-stmts.c (simd_clone_subparts): New function.
10304 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10306 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10307 Alan Hayward <alan.hayward@arm.com>
10308 David Sherwood <david.sherwood@arm.com>
10310 * tree-vect-stmts.c (vectorizable_call): Treat the number of
10311 vectors as polynomial. Use build_index_vector for
10312 IFN_GOMP_SIMD_LANE.
10314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10315 Alan Hayward <alan.hayward@arm.com>
10316 David Sherwood <david.sherwood@arm.com>
10318 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10319 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10320 for variable-length vectors.
10321 (vectorizable_mask_load_store): Treat the number of units as
10322 polynomial, asserting that it is constant if the condition has
10323 already been enforced.
10324 (vectorizable_store, vectorizable_load): Likewise.
10326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10327 Alan Hayward <alan.hayward@arm.com>
10328 David Sherwood <david.sherwood@arm.com>
10330 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10331 of units as polynomial. Punt if we can't tell at compile time
10332 which vector contains the final result.
10334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10335 Alan Hayward <alan.hayward@arm.com>
10336 David Sherwood <david.sherwood@arm.com>
10338 * tree-vect-loop.c (vectorizable_induction): Treat the number
10339 of units as polynomial. Punt on SLP inductions. Use an integer
10340 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10341 cast of such a series for variable-length floating-point
10344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10345 Alan Hayward <alan.hayward@arm.com>
10346 David Sherwood <david.sherwood@arm.com>
10348 * tree.h (build_index_vector): Declare.
10349 * tree.c (build_index_vector): New function.
10350 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10351 of units as polynomial, forcibly converting it to a constant if
10352 vectorizable_reduction has already enforced the condition.
10353 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
10354 to create a {1,2,3,...} vector.
10355 (vectorizable_reduction): Treat the number of units as polynomial.
10356 Choose vectype_in based on the largest scalar element size rather
10357 than the smallest number of units. Enforce the restrictions
10360 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10361 Alan Hayward <alan.hayward@arm.com>
10362 David Sherwood <david.sherwood@arm.com>
10364 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10365 number of units as polynomial.
10367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10368 Alan Hayward <alan.hayward@arm.com>
10369 David Sherwood <david.sherwood@arm.com>
10371 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10372 * target.def (autovectorize_vector_sizes): Return the vector sizes
10373 by pointer, using vector_sizes rather than a bitmask.
10374 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10375 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10376 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10378 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10379 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10380 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10381 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10382 * omp-general.c (omp_max_vf): Likewise.
10383 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10384 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10385 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10386 * tree-vect-slp.c (vect_slp_bb): Likewise.
10387 * doc/tm.texi: Regenerate.
10388 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10390 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10391 the vector size as a poly_uint64 rather than an unsigned int.
10392 (current_vector_size): Change from an unsigned int to a poly_uint64.
10393 (get_vectype_for_scalar_type): Update accordingly.
10394 * tree.h (build_truth_vector_type): Take the size and number of
10395 units as a poly_uint64 rather than an unsigned int.
10396 (build_vector_type): Add a temporary overload that takes
10397 the number of units as a poly_uint64 rather than an unsigned int.
10398 * tree.c (make_vector_type): Likewise.
10399 (build_truth_vector_type): Take the number of units as a poly_uint64
10400 rather than an unsigned int.
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 * target.def (get_mask_mode): Take the number of units and length
10407 as poly_uint64s rather than unsigned ints.
10408 * targhooks.h (default_get_mask_mode): Update accordingly.
10409 * targhooks.c (default_get_mask_mode): Likewise.
10410 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10411 * doc/tm.texi: Regenerate.
10413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10414 Alan Hayward <alan.hayward@arm.com>
10415 David Sherwood <david.sherwood@arm.com>
10417 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10418 * omp-general.c (omp_max_vf): Likewise.
10419 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10420 (expand_omp_simd): Handle polynomial safelen.
10421 * omp-low.c (omplow_simd_context): Add a default constructor.
10422 (omplow_simd_context::max_vf): Change from int to poly_uint64.
10423 (lower_rec_simd_input_clauses): Update accordingly.
10424 (lower_rec_input_clauses): Likewise.
10426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10427 Alan Hayward <alan.hayward@arm.com>
10428 David Sherwood <david.sherwood@arm.com>
10430 * tree-vectorizer.h (vect_nunits_for_cost): New function.
10431 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
10432 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
10433 (vect_analyze_slp_cost): Likewise.
10434 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
10435 (vect_model_load_cost): Likewise.
10437 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10438 Alan Hayward <alan.hayward@arm.com>
10439 David Sherwood <david.sherwood@arm.com>
10441 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
10442 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
10443 from an unsigned int * to a poly_uint64_pod *.
10444 (calculate_unrolling_factor): New function.
10445 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
10447 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10448 Alan Hayward <alan.hayward@arm.com>
10449 David Sherwood <david.sherwood@arm.com>
10451 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
10452 from an unsigned int to a poly_uint64.
10453 (_loop_vec_info::slp_unrolling_factor): Likewise.
10454 (_loop_vec_info::vectorization_factor): Change from an int
10456 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
10457 (vect_get_num_vectors): New function.
10458 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
10459 (vect_get_num_copies): Use vect_get_num_vectors.
10460 (vect_analyze_data_ref_dependences): Change max_vf from an int *
10461 to an unsigned int *.
10462 (vect_analyze_data_refs): Change min_vf from an int * to a
10464 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10465 than an unsigned HOST_WIDE_INT.
10466 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10467 (vect_analyze_data_ref_dependence): Change max_vf from an int *
10468 to an unsigned int *.
10469 (vect_analyze_data_ref_dependences): Likewise.
10470 (vect_compute_data_ref_alignment): Handle polynomial vf.
10471 (vect_enhance_data_refs_alignment): Likewise.
10472 (vect_prune_runtime_alias_test_list): Likewise.
10473 (vect_shift_permute_load_chain): Likewise.
10474 (vect_supportable_dr_alignment): Likewise.
10475 (dependence_distance_ge_vf): Take the vectorization factor as a
10476 poly_uint64 rather than an unsigned HOST_WIDE_INT.
10477 (vect_analyze_data_refs): Change min_vf from an int * to a
10479 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
10480 vfm1 as a poly_uint64 rather than an int. Make the same change
10481 for the returned bound_scalar.
10482 (vect_gen_vector_loop_niters): Handle polynomial vf.
10483 (vect_do_peeling): Likewise. Update call to
10484 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
10485 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
10487 * tree-vect-loop.c (vect_determine_vectorization_factor)
10488 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
10489 (vect_get_known_peeling_cost): Likewise.
10490 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
10491 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
10492 (vect_transform_loop): Likewise. Use the lowest possible VF when
10493 updating the upper bounds of the loop.
10494 (vect_min_worthwhile_factor): Make static. Return an unsigned int
10495 rather than an int.
10496 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
10497 polynomial unroll factors.
10498 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
10499 (vect_make_slp_decision): Likewise.
10500 (vect_supported_load_permutation_p): Likewise, and polynomial
10502 (vect_analyze_slp_cost): Handle polynomial vf.
10503 (vect_slp_analyze_node_operations): Likewise.
10504 (vect_slp_analyze_bb_1): Likewise.
10505 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10506 than an unsigned HOST_WIDE_INT.
10507 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
10508 (vectorizable_load): Handle polynomial vf.
10509 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
10511 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
10513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10514 Alan Hayward <alan.hayward@arm.com>
10515 David Sherwood <david.sherwood@arm.com>
10517 * match.pd: Handle bit operations involving three constants
10518 and try to fold one pair.
10520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10522 * tree-vect-loop-manip.c: Include gimple-fold.h.
10523 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
10524 niters_maybe_zero parameters. Handle other cases besides a step of 1.
10525 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
10526 Add a path that uses a step of VF instead of 1, but disable it
10528 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
10529 and niters_no_overflow parameters. Update calls to
10530 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
10531 Create a new SSA name if the latter choses to use a ste other
10532 than zero, and return it via niters_vector_mult_vf_var.
10533 * tree-vect-loop.c (vect_transform_loop): Update calls to
10534 vect_do_peeling, vect_gen_vector_loop_niters and
10535 slpeel_make_loop_iterate_ntimes.
10536 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
10537 (vect_gen_vector_loop_niters): Update declarations after above changes.
10539 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
10541 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
10542 128-bit round to integer instructions.
10543 (ceil<mode>2): Likewise.
10544 (btrunc<mode>2): Likewise.
10545 (round<mode>2): Likewise.
10547 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10549 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
10550 unaligned VSX load/store on P8/P9.
10551 (expand_block_clear): Allow the use of unaligned VSX
10552 load/store on P8/P9.
10554 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10556 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
10558 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
10559 swap associated with both a load and a store.
10561 2018-01-02 Andrew Waterman <andrew@sifive.com>
10563 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
10564 * config/riscv/riscv.md (clear_cache): Use it.
10566 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
10568 * web.c: Remove out-of-date comment.
10570 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10572 * expr.c (fixup_args_size_notes): Check that any existing
10573 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
10574 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
10575 (emit_single_push_insn): ...here.
10577 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10579 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
10580 (const_vector_encoded_nelts): New function.
10581 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
10582 (const_vector_int_elt, const_vector_elt): Declare.
10583 * emit-rtl.c (const_vector_int_elt_1): New function.
10584 (const_vector_elt): Likewise.
10585 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
10586 of CONST_VECTOR_ELT.
10588 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10590 * expr.c: Include rtx-vector-builder.h.
10591 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
10592 directly on the tree encoding.
10593 (const_vector_from_tree): Likewise.
10594 * optabs.c: Include rtx-vector-builder.h.
10595 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
10596 sequence of "u" values.
10597 * vec-perm-indices.c: Include rtx-vector-builder.h.
10598 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
10599 directly on the vec_perm_indices encoding.
10601 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10603 * doc/rtl.texi (const_vector): Describe new encoding scheme.
10604 * Makefile.in (OBJS): Add rtx-vector-builder.o.
10605 * rtx-vector-builder.h: New file.
10606 * rtx-vector-builder.c: Likewise.
10607 * rtl.h (rtx_def::u2): Add a const_vector field.
10608 (CONST_VECTOR_NPATTERNS): New macro.
10609 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
10610 (CONST_VECTOR_DUPLICATE_P): Likewise.
10611 (CONST_VECTOR_STEPPED_P): Likewise.
10612 (CONST_VECTOR_ENCODED_ELT): Likewise.
10613 (const_vec_duplicate_p): Check for a duplicated vector encoding.
10614 (unwrap_const_vec_duplicate): Likewise.
10615 (const_vec_series_p): Check for a non-duplicated vector encoding.
10616 Say that the function only returns true for integer vectors.
10617 * emit-rtl.c: Include rtx-vector-builder.h.
10618 (gen_const_vec_duplicate_1): Delete.
10619 (gen_const_vector): Call gen_const_vec_duplicate instead of
10620 gen_const_vec_duplicate_1.
10621 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
10622 (gen_const_vec_duplicate): Use rtx_vector_builder.
10623 (gen_const_vec_series): Likewise.
10624 (gen_rtx_CONST_VECTOR): Likewise.
10625 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
10626 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10627 Build a new vector rather than modifying a CONST_VECTOR in-place.
10628 (handle_special_swappables): Update call accordingly.
10629 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
10630 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10631 Build a new vector rather than modifying a CONST_VECTOR in-place.
10632 (handle_special_swappables): Update call accordingly.
10634 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10636 * simplify-rtx.c (simplify_const_binary_operation): Use
10637 CONST_VECTOR_ELT instead of XVECEXP.
10639 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10641 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
10642 the selector elements to be different from the data elements
10643 if the selector is a VECTOR_CST.
10644 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
10645 ssizetype for the selector.
10647 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10649 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
10650 before testing each element individually.
10651 * tree-vect-generic.c (lower_vec_perm): Likewise.
10653 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10655 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
10656 * selftest-run-tests.c (selftest::run_tests): Call it.
10657 * vector-builder.h (vector_builder::operator ==): New function.
10658 (vector_builder::operator !=): Likewise.
10659 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
10660 (vec_perm_indices::all_from_input_p): New function.
10661 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10662 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
10663 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
10664 instead of reading the VECTOR_CST directly. Detect whether both
10665 vector inputs are the same before constructing the vec_perm_indices,
10666 and update the number of inputs argument accordingly. Use the
10667 utility functions added above. Only construct sel2 if we need to.
10669 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10671 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
10672 the broadcast of the low byte.
10673 (expand_mult_highpart): Use an explicit encoding for the permutes.
10674 * optabs-query.c (can_mult_highpart_p): Likewise.
10675 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
10676 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10677 (vectorizable_bswap): Likewise.
10678 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
10679 explicit encoding for the power-of-2 permutes.
10680 (vect_permute_store_chain): Likewise.
10681 (vect_grouped_load_supported): Likewise.
10682 (vect_permute_load_chain): Likewise.
10684 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10686 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
10687 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
10688 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
10689 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10690 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
10691 (vect_gen_perm_mask_any): Likewise.
10693 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10695 * int-vector-builder.h: New file.
10696 * vec-perm-indices.h: Include int-vector-builder.h.
10697 (vec_perm_indices): Redefine as an int_vector_builder.
10698 (auto_vec_perm_indices): Delete.
10699 (vec_perm_builder): Redefine as a stand-alone class.
10700 (vec_perm_indices::vec_perm_indices): New function.
10701 (vec_perm_indices::clamp): Likewise.
10702 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
10703 (vec_perm_indices::new_vector): New function.
10704 (vec_perm_indices::new_expanded_vector): Update for new
10705 vec_perm_indices class.
10706 (vec_perm_indices::rotate_inputs): New function.
10707 (vec_perm_indices::all_in_range_p): Operate directly on the
10708 encoded form, without computing elided elements.
10709 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
10710 encoding. Update for new vec_perm_indices class.
10711 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
10712 the given vec_perm_builder.
10713 (expand_vec_perm_var): Update vec_perm_builder constructor.
10714 (expand_mult_highpart): Use vec_perm_builder instead of
10715 auto_vec_perm_indices.
10716 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
10717 vec_perm_indices instead of auto_vec_perm_indices. Use a single
10718 or double series encoding as appropriate.
10719 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
10720 vec_perm_indices instead of auto_vec_perm_indices.
10721 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10722 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10723 (vect_permute_store_chain): Likewise.
10724 (vect_grouped_load_supported): Likewise.
10725 (vect_permute_load_chain): Likewise.
10726 (vect_shift_permute_load_chain): Likewise.
10727 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10728 (vect_transform_slp_perm_load): Likewise.
10729 (vect_schedule_slp_instance): Likewise.
10730 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10731 (vectorizable_mask_load_store): Likewise.
10732 (vectorizable_bswap): Likewise.
10733 (vectorizable_store): Likewise.
10734 (vectorizable_load): Likewise.
10735 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
10736 vec_perm_indices instead of auto_vec_perm_indices. Use
10737 tree_to_vec_perm_builder to read the vector from a tree.
10738 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
10739 vec_perm_builder instead of a vec_perm_indices.
10740 (have_whole_vector_shift): Use vec_perm_builder and
10741 vec_perm_indices instead of auto_vec_perm_indices. Leave the
10742 truncation to calc_vec_perm_mask_for_shift.
10743 (vect_create_epilog_for_reduction): Likewise.
10744 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
10745 from auto_vec_perm_indices to vec_perm_indices.
10746 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10747 instead of changing individual elements.
10748 (aarch64_vectorize_vec_perm_const): Use new_vector to install
10749 the vector in d.perm.
10750 * config/arm/arm.c (expand_vec_perm_d::perm): Change
10751 from auto_vec_perm_indices to vec_perm_indices.
10752 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10753 instead of changing individual elements.
10754 (arm_vectorize_vec_perm_const): Use new_vector to install
10755 the vector in d.perm.
10756 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
10757 Update vec_perm_builder constructor.
10758 (rs6000_expand_interleave): Likewise.
10759 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
10760 (rs6000_expand_interleave): Likewise.
10762 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10764 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
10765 to qimode could truncate the indices.
10766 * optabs.c (expand_vec_perm_var): Likewise.
10768 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10770 * Makefile.in (OBJS): Add vec-perm-indices.o.
10771 * vec-perm-indices.h: New file.
10772 * vec-perm-indices.c: Likewise.
10773 * target.h (vec_perm_indices): Replace with a forward class
10775 (auto_vec_perm_indices): Move to vec-perm-indices.h.
10776 * optabs.h: Include vec-perm-indices.h.
10777 (expand_vec_perm): Delete.
10778 (selector_fits_mode_p, expand_vec_perm_var): Declare.
10779 (expand_vec_perm_const): Declare.
10780 * target.def (vec_perm_const_ok): Replace with...
10781 (vec_perm_const): ...this new hook.
10782 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
10783 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
10784 * doc/tm.texi: Regenerate.
10785 * optabs.def (vec_perm_const): Delete.
10786 * doc/md.texi (vec_perm_const): Likewise.
10787 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
10788 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
10789 expand_vec_perm for constant permutation vectors. Assert that
10790 the mode of variable permutation vectors is the integer equivalent
10791 of the mode that is being permuted.
10792 * optabs-query.h (selector_fits_mode_p): Declare.
10793 * optabs-query.c: Include vec-perm-indices.h.
10794 (selector_fits_mode_p): New function.
10795 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
10796 is defined, instead of checking whether the vec_perm_const_optab
10797 exists. Use targetm.vectorize.vec_perm_const instead of
10798 targetm.vectorize.vec_perm_const_ok. Check whether the indices
10799 fit in the vector mode before using a variable permute.
10800 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
10801 vec_perm_indices instead of an rtx.
10802 (expand_vec_perm): Replace with...
10803 (expand_vec_perm_const): ...this new function. Take the selector
10804 as a vec_perm_indices rather than an rtx. Also take the mode of
10805 the selector. Update call to shift_amt_for_vec_perm_mask.
10806 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
10807 Use vec_perm_indices::new_expanded_vector to expand the original
10808 selector into bytes. Check whether the indices fit in the vector
10809 mode before using a variable permute.
10810 (expand_vec_perm_var): Make global.
10811 (expand_mult_highpart): Use expand_vec_perm_const.
10812 * fold-const.c: Includes vec-perm-indices.h.
10813 * tree-ssa-forwprop.c: Likewise.
10814 * tree-vect-data-refs.c: Likewise.
10815 * tree-vect-generic.c: Likewise.
10816 * tree-vect-loop.c: Likewise.
10817 * tree-vect-slp.c: Likewise.
10818 * tree-vect-stmts.c: Likewise.
10819 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
10821 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
10822 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
10823 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
10824 (aarch64_vectorize_vec_perm_const): ...this new function.
10825 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10826 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10827 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
10828 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
10829 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10830 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10831 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
10833 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
10834 check for NEON modes.
10835 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
10836 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
10837 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
10838 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
10840 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
10841 the old VEC_PERM_CONST conditions.
10842 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
10843 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
10844 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
10845 (ia64_vectorize_vec_perm_const_ok): Merge into...
10846 (ia64_vectorize_vec_perm_const): ...this new function.
10847 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
10848 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
10849 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
10850 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
10851 * config/mips/mips.c (mips_expand_vec_perm_const)
10852 (mips_vectorize_vec_perm_const_ok): Merge into...
10853 (mips_vectorize_vec_perm_const): ...this new function.
10854 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
10855 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
10856 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
10857 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
10858 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
10859 (rs6000_expand_vec_perm_const): Delete.
10860 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
10862 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10863 (altivec_expand_vec_perm_const_le): Take each operand individually.
10864 Operate on constant selectors rather than rtxes.
10865 (altivec_expand_vec_perm_const): Likewise. Update call to
10866 altivec_expand_vec_perm_const_le.
10867 (rs6000_expand_vec_perm_const): Delete.
10868 (rs6000_vectorize_vec_perm_const_ok): Delete.
10869 (rs6000_vectorize_vec_perm_const): New function.
10870 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10871 an element count and rtx array.
10872 (rs6000_expand_extract_even): Update call accordingly.
10873 (rs6000_expand_interleave): Likewise.
10874 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
10875 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
10876 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
10877 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
10878 (rs6000_expand_vec_perm_const): Delete.
10879 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10880 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10881 (altivec_expand_vec_perm_const_le): Take each operand individually.
10882 Operate on constant selectors rather than rtxes.
10883 (altivec_expand_vec_perm_const): Likewise. Update call to
10884 altivec_expand_vec_perm_const_le.
10885 (rs6000_expand_vec_perm_const): Delete.
10886 (rs6000_vectorize_vec_perm_const_ok): Delete.
10887 (rs6000_vectorize_vec_perm_const): New function. Remove stray
10888 reference to the SPE evmerge intructions.
10889 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10890 an element count and rtx array.
10891 (rs6000_expand_extract_even): Update call accordingly.
10892 (rs6000_expand_interleave): Likewise.
10893 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
10894 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
10896 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10898 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10900 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
10901 vector mode and that that mode matches the mode of the data
10903 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
10904 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
10905 directly using expand_vec_perm_1 when forcing selectors into
10907 (expand_vec_perm_var): New function, split out from expand_vec_perm.
10909 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10911 * optabs-query.h (can_vec_perm_p): Delete.
10912 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
10913 * optabs-query.c (can_vec_perm_p): Split into...
10914 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
10915 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
10916 particular selector is valid.
10917 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10918 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10919 (vect_grouped_load_supported): Likewise.
10920 (vect_shift_permute_load_chain): Likewise.
10921 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10922 (vect_transform_slp_perm_load): Likewise.
10923 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10924 (vectorizable_bswap): Likewise.
10925 (vect_gen_perm_mask_checked): Likewise.
10926 * fold-const.c (fold_ternary_loc): Likewise. Don't take
10927 implementations of variable permutation vectors into account
10928 when deciding which selector to use.
10929 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
10930 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
10931 with a false third argument.
10932 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
10933 to test whether the constant selector is valid and can_vec_perm_var_p
10934 to test whether a variable selector is valid.
10936 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10938 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
10939 * optabs-query.c (can_vec_perm_p): Likewise.
10940 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
10941 instead of vec_perm_indices.
10942 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
10943 (vect_gen_perm_mask_checked): Likewise,
10944 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
10945 (vect_gen_perm_mask_checked): Likewise,
10947 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10949 * optabs-query.h (qimode_for_vec_perm): Declare.
10950 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
10951 (qimode_for_vec_perm): ...this new function.
10952 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
10954 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10956 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
10957 does not have a conditional at the top.
10959 2018-01-02 Richard Biener <rguenther@suse.de>
10961 * ipa-inline.c (big_speedup_p): Fix expression.
10963 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10966 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
10969 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10973 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
10974 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
10975 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
10976 cond_taken_branch_cost 3->4.
10978 2018-01-01 Jakub Jelinek <jakub@redhat.com>
10980 PR tree-optimization/83581
10981 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
10982 TODO_cleanup_cfg if any changes have been made.
10984 PR middle-end/83608
10985 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
10986 convert_modes if target mode has the right side, but different mode
10989 PR middle-end/83609
10990 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
10991 last argument when extracting from CONCAT. If either from_real or
10992 from_imag is NULL, use expansion through memory. If result is not
10993 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
10994 the parts directly to inner mode, if even that fails, use expansion
10997 PR middle-end/83623
10998 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
10999 check for bswap in mode rather than HImode and use that in expand_unop
11002 Copyright (C) 2018 Free Software Foundation, Inc.
11004 Copying and distribution of this file, with or without modification,
11005 are permitted in any medium without royalty provided the copyright
11006 notice and this notice are preserved.