1 2018-04-18 Martin Liska <mliska@suse.cz>
4 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
7 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
10 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
13 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
14 ENDBR after calling __morestack.
16 2018-04-18 David Malcolm <dmalcolm@redhat.com>
19 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
20 by using gcc_base_ver to generate a gcc_driver_version, and use
21 it when generating GCC_DRIVER_NAME.
22 * configure: Regenerate.
24 2018-04-18 Jakub Jelinek <jakub@redhat.com>
27 * config.gcc: Obsolete powerpc*-*-*spe*.
29 2018-04-17 Jakub Jelinek <jakub@redhat.com>
32 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
33 (stabstr_D): Change type of unum from unsigned int to
34 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
37 2018-04-17 Jim Wilson <jimw@sifive.com>
40 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
41 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
42 Set arg_pointer_offset after using pretend_args_size.
44 2018-04-17 Jakub Jelinek <jakub@redhat.com>
46 PR rtl-optimization/85431
47 * dse.c (record_store): Ignore zero width stores.
50 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
51 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
52 __builtin_stack_restore rather than after it.
53 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
54 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
55 argument instead of virtual_dynamic_stack_rtx.
57 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
59 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
61 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
62 Add note to error message to explain internal mapping of overloaded
63 built-in function name to non-overloaded built-in function name.
64 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
67 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
70 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
71 where the inputs overlap with the output.
73 2018-04-17 Jakub Jelinek <jakub@redhat.com>
75 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
76 (=v, v) alternative and explicit "memory" attribute.
77 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
78 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
80 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
81 "sselog1" type instead of "sselog".
82 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
83 "sselog". Remove explicit "memory" attribute.
84 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
85 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
87 (vec_extract_hi_v32hi): Merge all alternatives into one, use
88 "sselog1" type instead of "sselog". Remove explicit "memory"
90 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
91 use "sselog1" type instead of "sselog". Remove explicit "memory"
93 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
94 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
96 (vec_extract_hi_v64qi): Merge all alternatives into one, use
97 "sselog1" type instead of "sselog". Remove explicit "memory"
99 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
100 use "sselog1" type instead of "sselog". Remove explicit "memory"
104 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
107 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
110 2018-04-17 Martin Jambor <mjambor@suse.cz>
113 * ipa-cp.c (create_specialized_node): Call
114 expand_all_artificial_thunks if necessary.
116 2018-04-17 Martin Liska <mliska@suse.cz>
119 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
120 in message, remote space in between '_G' and '('.
122 2018-04-17 Jakub Jelinek <jakub@redhat.com>
125 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
126 avx512f_vmcmp<mode>3<round_saeonly_name>,
127 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
128 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
129 avx512f_rndscale<mode><round_saeonly_name>,
130 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
131 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
132 Use %<iptr>2 instead of %2 for -masm=intel.
133 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
134 avx512f_vcvttss2usi<round_saeonly_name>,
135 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
137 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
138 avx512f_vcvttsd2usi<round_saeonly_name>,
139 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
140 Use %q1 instead of %1 for -masm=intel.
141 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
142 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
143 of %3 for -masm=intel.
144 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
146 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
148 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
150 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
151 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
153 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
155 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
156 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
157 %g1 and one with %0 and %1.
158 (avx512er_vmrcp28<mode><round_saeonly_name>,
159 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
161 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
162 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
163 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
164 of %0 and %{%4%} for -masm=intel.
165 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
166 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
167 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
168 order of %0 and %{%5%}%{z%} for -masm=intel.
170 2018-04-17 Jan Hubicka <jh@suse.cz>
173 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
175 2018-04-17 Martin Liska <mliska@suse.cz>
178 * multiple_target.c (create_dispatcher_calls): Set apostrophes
179 for target_clone error message. Make default implementation
180 clone to be a local declaration.
181 (separate_attrs): Add new argument and check for an empty
183 (expand_target_clones): Handle it.
184 (ipa_target_clone): Make redirection just for target_clones
187 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
188 Tom de Vries <tom@codesourcery.com>
191 * omp-expand.c (expand_oacc_for): Add dummy false branch for
192 tiled basic blocks without omp continue statements.
194 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
197 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
198 vec_extract expression as having side effects to make sure it gets
201 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
204 * config/i386/i386.c (get_builtin_code_for_version): Check
207 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
210 * gcc/config.gcc: Support "skylake".
211 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
213 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
214 (processor_target_table): Add "skylake".
215 (ix86_option_override_internal): Add "skylake".
216 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
217 PROCESSOR_CANNONLAKE.
218 (get_builtin_code_for_version): Fix priority for
219 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
220 PROCESSOR_SKYLAKE-AVX512.
221 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
222 (processor_type): Add PROCESSOR_SKYLAKE.
224 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
225 Jason Merrill <jason@redhat.com>
228 * convert.c (convert_to_integer_1): Use direct recursion for
229 enumeral types and types with a precision less than the number
230 of bits in their mode.
232 2018-04-16 Julia Koval <julia.koval@intel.com>
235 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
236 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
238 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
241 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
242 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
243 and -mno-direct-move.
245 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
248 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
249 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
250 Ensure negative shifts result in {0}.
252 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
254 PR rtl-optimization/79916
255 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
256 regs (if any) to define how to gnerate SD moves when LRA is in
259 2018-04-13 Jakub Jelinek <jakub@redhat.com>
261 PR rtl-optimization/85393
262 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
263 * except.c (expand_dw2_landing_pad_for_region): Make static.
264 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
265 a label and unconditional jump to old_bb, rather than
266 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
269 PR rtl-optimization/85376
270 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
271 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
272 instead of a specific value.
274 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
275 Bin Cheng <bin.cheng@arm.com>
277 PR tree-optimization/82965
278 PR tree-optimization/83991
279 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
280 by_profile_only parameter.
281 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
282 information if the loop was predicted to iterate too many times.
283 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
285 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
288 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
291 2018-04-13 Martin Liska <mliska@suse.cz>
292 Jakub Jelinek <jakub@redhat.com>
295 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
296 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
297 * builtins.c (expand_builtin_memory_copy_args): Use
298 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
299 handle dest_addr == pc_rtx.
301 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
304 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
305 asked to not generate direct moves.
306 (fix_trunc<mode>si2_stfiwx): Similar.
307 (fix_trunc<mode>si2_internal): Similar.
309 2018-04-12 Jakub Jelinek <jakub@redhat.com>
312 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
313 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
314 lookup if dest in some wider mode is known to be const0_rtx and
315 if so, record permanent equivalence for it to be ZERO_EXTEND of
316 the narrower mode destination.
318 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
320 * lto-streamer-out.c (output_function): Revert 259346.
321 * omp-expand.c (expand_oacc_for): Likewise.
323 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
325 PR rtl-optimization/85354
326 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
327 * sel-sched.c (sel_global_init): ... here.
329 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
332 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
333 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
334 mode for PE-COFF targets.
335 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
336 (i386_pe_asm_lto_end): Likewise.
337 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
338 (TARGET_ASM_LTO_END): Likewise.
339 * config/i386/winnt.c (saved_debug_info_level): New static variable.
340 (i386_pe_asm_lto_start): New function.
341 (i386_pe_asm_lto_end): Likewise.
343 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
344 Richard Biener <rguenther@suse.de>
347 * lto-streamer-out.c (output_function): Fix CFG loop state before
349 * omp-expand.c (expand_oacc_for): Handle calls to internal
350 functions like regular functions.
352 2018-04-12 Richard Biener <rguenther@suse.de>
355 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
356 for the early LTO debug to properly generate references to it
357 during DIE emission. Do not re-use that for the skeleton for
359 (dwarf2out_early_finish): Likewise.
361 2018-04-12 Jakub Jelinek <jakub@redhat.com>
365 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
366 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
367 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
368 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
369 and output is a reg, avoid creating invalid lowpart subreg, but
370 instead split into a 512-bit move. Don't split if not AVX512VL,
371 input is xmm16+ reg and output is a mem.
372 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
373 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
374 xmm16+ reg and output is a mem.
376 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
378 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
379 also for flag_dwarf2_cfi_asm.
381 2018-04-12 Jakub Jelinek <jakub@redhat.com>
383 PR rtl-optimization/85342
384 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
385 a bool scalar var inside of the loop instead. Don't try to update
386 recog_data.operand after failed apply_change_group.
388 2018-04-12 Tom de Vries <tom@codesourcery.com>
391 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
392 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
393 array with flexible array member as array without given dimension.
394 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
395 argument for undefined param to true.
397 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
400 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
401 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
402 from PowerPC section.
403 * config/rs6000/sysv4.opt (mcall-): Improve help text.
404 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
405 help text that is too long.
406 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
407 help text that is too long.
408 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
409 help text that is too long.
411 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
413 * config/alpha/alpha.md (stack_probe_internal): Rename
414 from "probe_stack". Update all callers.
416 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
418 PR rtl-optimization/84566
419 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
420 sched_macro_fuse_insns.
422 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
425 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
426 (compute_block_dependences): ... from here.
428 2018-04-11 Jakub Jelinek <jakub@redhat.com>
430 PR tree-optimization/85331
431 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
432 from int to HOST_WIDE_INT.
434 2018-04-11 Martin Jambor <mjambor@suse.cz>
437 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
438 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
439 not the same as the source val.
440 (cgraph_edge_brings_value_p): New parameter.
441 (gather_edges_for_value): Pass destination value to
442 cgraph_edge_brings_value_p.
443 (perhaps_add_new_callers): Likewise.
444 (get_info_about_necessary_edges): Likewise and exclude values brought
445 only by self-recursive edges.
446 (create_specialized_node): Redirect only clones of self-calling edges.
447 (+self_recursive_pass_through_p): New function.
448 (find_more_scalar_values_for_callers_subset): Use it.
449 (find_aggregate_values_for_callers_subset): Likewise.
450 (known_aggs_to_agg_replacement_list): Removed.
451 (decide_whether_version_node): Re-calculate known constants for all
452 remaining context clones.
454 2018-04-11 Richard Biener <rguenther@suse.de>
457 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
458 from early DWARF output.
459 (dwarf2out_early_finish): Output line info unconditionally into
460 early DWARF and add reference to it.
462 2018-04-11 Jakub Jelinek <jakub@redhat.com>
465 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
466 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
467 other than V2DFmode using iptr mode attribute.
468 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
470 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
472 PR rtl-optimization/84659
473 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
475 2018-04-11 Jakub Jelinek <jakub@redhat.com>
478 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
480 (output_loc_list): Pass address of a dummy size variable even in the
481 locview handling loop.
482 (index_location_lists): Add comment on why skip_loc_list_entry can't
485 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
488 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
491 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
494 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
495 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
496 and -mstring-compare-inline-limit.
498 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
501 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
502 for stack clash protection in a register whenever we need it to be in
505 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
507 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
508 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
510 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
513 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
515 (mlong-double-): Ditto.
516 * config/rs6000/sysv4.opt (msdata=): Ditto.
519 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
521 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
522 erroneous entries for
523 "vector int vec_ldl (int, long int *)", and
524 "vector unsigned int vec_ldl (int, unsigned long int *)".
525 Add comments and entries for
526 "vector bool char vec_ldl (int, bool char *)",
527 "vector bool short vec_ldl (int, bool short *)",
528 "vector bool int vec_ldl (int, bool int *)",
529 "vector bool long long vec_ldl (int, bool long long *)",
530 "vector pixel vec_ldl (int, pixel *)",
531 "vector long long vec_ldl (int, long long *)",
532 "vector unsigned long long vec_ldl (int, unsigned long long *)".
533 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
534 type tree bool_long_long_type_node and correct definition of
535 bool_V2DI_type_node to make reference to this new type tree.
536 (rs6000_mangle_type): Replace erroneous reference to
537 bool_long_type_node with bool_long_long_type_node.
538 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
539 comments to emphasize sign distinctions for char and int types and
540 replace RS6000_BTI_bool_long constant with
541 RS6000_BTI_bool_long_long constant. Also add comment to restrict
542 use of RS6000_BTI_pixel.
543 (bool_long_type_node): Remove this macro definition.
544 (bool_long_long_type_node): New macro definition
546 2018-04-10 Jakub Jelinek <jakub@redhat.com>
548 PR rtl-optimization/85300
549 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
550 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
551 simplify_unary_operation fails.
553 2018-04-10 Martin Liska <mliska@suse.cz>
555 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
556 cgraph_edge and ipa_ref.
558 2018-04-10 Jakub Jelinek <jakub@redhat.com>
563 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
564 computation of the VEC_MERGE selector from mask.
565 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
566 Fix decoding of the VEC_MERGE selector into mask.
568 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
570 PR tree-optimization/85286
571 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
573 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
575 * final.c (final_1): Set insn_last_address as well as
576 insn_current_address.
578 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
581 * explow.c (emit_stack_probe): Call validize_mem on memory location
582 before passing it to gen_probe_stack. Create address operand and
583 legitimize it for the probe_stack_address case.
585 2018-04-09 Jan Hubicka <jh@suse.cz>
588 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
589 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
590 * tree.c (free_lang_data_in_type): Fix handling of binfos;
592 (free_lang_data): Rebuild type inheritance graph.
594 2018-04-09 Martin Sebor <msebor@redhat.com>
596 * invoke.texi (-finline-small-functions): Mention other optimization
598 (-findirect-inlining, -fpartial-inlining): Same.
599 (-finline-functions-called-once): Same.
600 (-freorder-blocks-and-partition): Same.
602 2018-04-09 Jan Hubicka <jh@suse.cz>
605 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
606 jumps; choose last target that matches the criteria (i.e.
607 no partition changes for non-crossing jumps).
608 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
609 support for redirecting crossing jumps to non-crossing.
611 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
613 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
614 also for naked functions.
616 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
618 * config/arc/arc.md (add_shift): New pattern.
619 (add_shift2): Likewise.
620 (sub_shift): Likewise.
621 (sub_shift_cmp0_noout): Likewise.
622 (compare_si_ashiftsi): Likewise.
623 (xbfu_cmp0_noout): New combine pattern.
624 (xbfu_cmp0"): Likewise.
625 (movsi_set_cc_insn): Place the predicable variant first.
626 (commutative_binary_cmp0_noout): Remove clobber.
627 (commutative_binary_cmp0): New pattern.
628 (noncommutative_binary_cmp0): Likewise.
629 (noncommutative_binary_cmp0_noout): Likewise.
630 (noncommutative_binary_comparison_result_used): Removed.
631 (rsub_cmp0): New pattern.
632 (rsub_cmp0_noout): Likewise.
633 (extzvsi): Changed, keep only meaningful variants.
634 (SQH, SEZ): New iterators.
635 (SQH_postfix): New mode attribute.
636 (SEZ_prefix): New code attribute.
637 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
638 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
639 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
641 (noncommutative_operator): Check the availability of barrel
644 2018-04-09 Richard Biener <rguenther@suse.de>
646 PR tree-optimization/85284
647 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
648 Only use the niter constraining form of simple_iv when the exit
651 2018-04-09 Tom de Vries <tom@codesourcery.com>
654 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
655 (define_expand "*memory_barrier"): New define_expand.
656 (define_insn "memory_barrier"): New insn.
658 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
660 PR rtl-optimization/80463
661 PR rtl-optimization/83972
662 PR rtl-optimization/83480
664 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
665 correct producer for the insn.
666 (tidy_control_flow): Fixup seqnos in case of debug insns.
668 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
670 PR rtl-optimization/83913
672 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
673 different sched-times when merging exprs.
675 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
677 PR rtl-optimization/83962
679 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
680 tidy_fallthru_edge and tidy_control_flow.
682 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
684 PR rtl-optimization/83530
686 * sel-sched.c (force_next_insn): New global variable.
687 (remove_insn_for_debug): When force_next_insn is true, also leave only
688 next insn in the ready list.
689 (sel_sched_region): When the region wasn't scheduled, make another pass
690 over it with force_next_insn set to 1.
692 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
694 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
696 * config/nds32/constants.md (unspec_volatile_element): Add enum values
697 for interrupt control.
698 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
699 functions for interrupt control.
700 * config/nds32/nds32-intrinsic.md: Likewise.
701 * config/nds32/nds32_intrinsic.h: Likewise.
702 * config/nds32/nds32.h (nds32_builtins): Likewise.
704 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
706 * config/nds32/nds32.c (nds32_init_machine_status,
707 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
708 strict_aligned_p field.
709 (nds32_expand_to_rtl_hook): New function.
710 (TARGET_EXPAND_TO_RTL_HOOK): Define.
711 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
713 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
714 Chung-Ju Wu <jasonwucj@gmail.com>
716 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
717 * config/nds32/nds32-n7.md: New file.
718 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
719 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
721 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
722 * config/nds32/nds32.md (pipeline_model): Add n7.
723 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
724 * config/nds32/pipelines.md: Include n7 settings.
726 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
727 Chung-Ju Wu <jasonwucj@gmail.com>
729 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
730 * config/nds32/nds32-e8.md: New file.
731 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
732 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
734 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
735 * config/nds32/nds32.md (pipeline_model): Add e8.
736 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
737 * config/nds32/pipelines.md: Include e8 settings.
739 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
740 Chung-Ju Wu <jasonwucj@gmail.com>
742 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
743 * config/nds32/nds32-n8.md: New file.
744 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
745 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
747 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
748 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
749 * config/nds32/nds32.md (pipeline_model): Add n8.
750 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
751 * config/nds32/pipelines.md: Include n8 settings.
753 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
754 Chung-Ju Wu <jasonwucj@gmail.com>
756 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
757 * config/nds32/nds32-n9-2r1w.md: New file.
758 * config/nds32/nds32-n9-3r2w.md: New file.
759 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
760 nds32_register_ports): New or modify for cpu n9.
761 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
763 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
764 * config/nds32/nds32-utils.c: New file.
765 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
766 TARGET_MUL_SLOW): Define.
767 * config/nds32/nds32.md (pipeline_model): New attribute.
768 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
769 New options that support cpu n9.
770 * config/nds32/pipelines.md: Include n9 settings.
771 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
773 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
775 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
776 information if necessary.
777 (output_cond_branch_compare_zero): Likewise.
778 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
779 (nds32_target_alignment): Refine for alignment.
780 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
781 (FUNCTION_BOUNDARY): Modify.
782 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
784 * config/nds32/nds32.opt (malways-align, malign-functions): New.
786 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
788 * config/nds32/constants.md (unspec_volatile_element): Add values for
789 TLB operation and data prefetch.
790 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
791 functions for TLB operation and data prefetch.
792 * config/nds32/nds32-intrinsic.md: Likewise.
793 * config/nds32/nds32_intrinsic.h: Likewise.
794 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
795 (nds32_print_operand): Likewise.
796 * config/nds32/nds32.h (nds32_builtins): Likewise.
798 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
799 Andrew Pinski <pinsika@gcc.gnu.org>
802 * match.pd: Use constant_boolean_node of correct type instead of
803 boolean_true_node or boolean_false_node for simplifying
804 pointer comparisons to zero.
806 2018-04-07 Jakub Jelinek <jakub@redhat.com>
808 PR tree-optimization/80021
809 * tree.c (verify_type_variant): Make error call in verify_variant_match
810 translatable and remove final full stop.
812 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
814 * config/nds32/constants.md (unspec_volatile_element): Add
815 UNSPEC_VOLATILE_EH_RETURN.
816 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
817 nds32_output_stack_pop): Support dwarf exception handling process.
818 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
819 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
820 exception handling process.
821 (nds32_compute_stack_frame): Likewise.
822 (nds32_return_addr_rtx): Likewise.
823 (nds32_initial_elimination_offset): Likewise.
824 (nds32_expand_prologue): Likewise.
825 (nds32_expand_epilogue): Likewise.
826 (nds32_dynamic_chain_address): New function.
827 * config/nds32/nds32.h (machine_function): Add fields for dwarf
829 (DYNAMIC_CHAIN_ADDRESS): Define.
830 (EH_RETURN_DATA_REGNO): Define.
831 (EH_RETURN_STACKADJ_RTX): Define.
832 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
833 patterns for dwarf exception handling.
835 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
837 * config/nds32/nds32.h: Clean up obsolete macros.
839 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
841 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
842 Add enum values for particular instructions.
843 * config/nds32/nds32-intrinsic.c: Implementation of expanding
844 particular intrinsic functions.
845 * config/nds32/nds32-intrinsic.md: Likewise.
846 * config/nds32/nds32_intrinsic.h: Likewise.
847 * config/nds32/nds32.h (nds32_builtins): Likewise.
848 * config/nds32/nds32.md (type): Add pbsad and pbsada.
849 (btst, ave): New patterns for particular instructions.
851 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
853 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
854 Add enum values for atomic load/store and memory sync.
855 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
857 * config/nds32/nds32-intrinsic.md: Likewise.
858 * config/nds32/nds32_intrinsic.h: Likewise.
859 * config/nds32/nds32.h (nds32_builtins): Likewise.
861 2018-04-07 Jakub Jelinek <jakub@redhat.com>
863 PR tree-optimization/85257
864 * fold-const.c (native_encode_vector): If not all elts could fit
865 and off is -1, return 0 rather than offset.
866 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
867 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
868 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
869 adjust buffer in native_interpret_expr call.
871 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
873 * config/nds32/constants.md (unspec_volatile_element): Add cache
875 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
876 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
877 * config/nds32/nds32.c (nds32_cctl_names): New.
878 (nds32_print_operand): Handle cache control register names.
879 * config/nds32/nds32.h (nds32_builtins): New enum values.
880 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
882 * config/nds32/nds32.md (type): Add mmu.
883 * config/nds32/pipelines.md (simple_insn): Add mmu.
885 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
887 * config/nds32/nds32.md (type): Remove call.
888 * config/nds32/pipelines.md (simple_insn): Likewise.
890 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
892 * config/nds32/constants.md (unspec_volatile_element): Add
893 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
894 UNSPEC_VOLATILE_FMFCFG.
895 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
896 description for fmfcfg and fmfcsr.
897 (bdesc_1arg): Add fmtcsr.
898 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
899 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
900 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
901 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
902 unspec_fmfcfg): New patterns.
903 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
904 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
905 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
906 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
907 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
908 __nds32__fmfcfg): Define.
910 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
912 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
913 intrinsic register names.
914 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
915 intrinsic register enum values and macros.
917 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
919 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
920 for load/store addressing form.
921 (nds32_print_operand_address): Likewise.
923 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
926 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
927 based on LABEL_REF. Remove useless assertion.
928 (pic_address_needs_scratch): Fix formatting.
929 (sparc_legitimize_pic_address): Minor tweaks.
930 (sparc_delegitimize_address): Adjust assertion accordingly.
931 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
932 into symbolic_operand.
933 (movsi_high_pic_label_ref): Likewise.
934 (movsi_lo_sum_pic_label_ref): Likewise.
935 (movdi_pic_label_ref): Likewise.
936 (movdi_high_pic_label_ref): Likewise.
937 (movdi_lo_sum_pic_label_ref): Likewise.
939 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
941 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
942 custom LIB_SPEC setup.
944 2018-04-06 Ruslan Bukin <br@bsdpad.com>
945 Kito Cheng <kito.cheng@gmail.com>
947 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
948 * config/riscv/freebsd.h: New.
950 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
952 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
953 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
956 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
957 Kito Cheng <kito.cheng@gmail.com>
959 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
960 nds32_output_call, nds32_symbol_binds_local_p): New functions.
961 * config/nds32/nds32-protos.h (nds32_output_call,
962 nds32_output_return): Declare.
963 * config/nds32/nds32.md: Refine all the call and return patterns.
965 2018-04-06 Jakub Jelinek <jakub@redhat.com>
968 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
969 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
971 PR rtl-optimization/84872
972 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
973 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
976 2018-04-06 Tamar Christina <tamar.christina@arm.com>
978 * expr.c (copy_blkmode_to_reg): Revert 254862.
979 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
981 2018-04-06 Richard Biener <rguenther@suse.de>
984 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
985 after seeing a component reference with an adjacent field. Treat
986 refs to arrays at struct end of external decls similar to
987 refs to unconstrained commons.
989 2018-04-06 Jakub Jelinek <jakub@redhat.com>
992 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
993 look through SAVE_EXPRs with non-side-effects argument. Adjust
995 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
998 2018-04-06 Richard Biener <rguenther@suse.de>
1001 * alias.c (find_base_term): New wrapper around find_base_term
1002 unwinding CSELIB_VAL_PTR changes.
1003 (find_base_term): Do not restore CSELIB_VAL_PTR during the
1006 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1008 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1010 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1011 constant definitions.
1012 ("nop"): lr 0,0 -> nopr r0
1013 ("nop_lr0", "nop_lr1"): New insn definitions.
1015 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1017 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1018 NDS32_V3PUSH_AVAILABLE_P macro.
1020 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
1021 Chung-Ju Wu <jasonwucj@gmail.com>
1023 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1024 (nds32*-*-*): Add float and fpu_config into supported_defaults.
1025 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1026 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1027 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1028 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1029 * config/nds32/constraints.md: New constraints and checking for hard
1030 float configuration.
1031 * config/nds32/iterators.md: New mode iterator and attribute for hard
1032 float configuration.
1033 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1035 * config/nds32/nds32-fpu.md: New file.
1036 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1037 deal with hard float code generation.
1038 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1040 (abi_type, float_reg_number): New enum type.
1041 * config/nds32/nds32-predicates.c: New predicates for hard float.
1042 * config/nds32/nds32-protos.h: Declare functions for hard float.
1043 * config/nds32/nds32.c: Implementation for hard float configuration.
1044 * config/nds32/nds32.h: Definitions for hard float configuration.
1045 * config/nds32/nds32.md: Include hard float machine description and
1046 modify patterns for hard float configuration.
1047 * config/nds32/nds32.opt: New options for hard float configuration.
1048 * config/nds32/predicates.md: New predicates for hard float
1051 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1053 * common/config/nds32/nds32-common.c
1054 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1056 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1059 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1060 CONSTRUCTOR_ELT (ctor, ...)->value.
1062 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1065 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1067 2018-04-05 Tom de Vries <tom@codesourcery.com>
1070 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1073 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1074 Kito Cheng <kito.cheng@gmail.com>
1076 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1077 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1078 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1079 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1081 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1082 Kito Cheng <kito.cheng@gmail.com>
1084 * config/nds32/constraints.md (Ufe): New memory constraint.
1085 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1086 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1087 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1089 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1090 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1092 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1094 * config/nds32/nds32.md: Use optimize_size in the condition for
1095 alu-shift instructions.
1097 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1099 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1101 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1103 * config/nds32/nds32.md (negsi2): Refine pattern.
1105 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1106 Chung-Ju Wu <jasonwucj@gmail.com>
1108 * config/nds32/iterators.md (shift_rotate): New code iterator.
1109 (shift): New code attribute.
1110 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1111 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1112 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1113 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1114 bit-wise operations.
1115 (andsi3, *andsi3): Ditto.
1116 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1117 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1118 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1119 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1120 nds32_ior_operand, nds32_xor_operand): New predicates.
1122 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1124 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1125 (addsi3, subsi3): ... this.
1127 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1129 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1131 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1133 * config/nds32/nds32.md: Adjust indention.
1135 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1137 * config/nds32/nds32.md (feature): New attribute.
1139 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1141 * config/nds32/nds32.md (subtype): New attribute.
1143 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1146 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1147 expansion to perform a bitwise AND of the argument followed by a
1148 boolean negation of the result.
1150 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1152 PR rtl-optimization/84878
1153 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1154 the basic block. Assert the use reference is not artificial and that
1155 it has an associated insn.
1157 2018-04-04 Michael Matz <matz@suse.de>
1159 * builtins.c (compute_objsize): Pass correct operand
1160 to array_at_struct_end_p.
1162 2018-04-04 Richard Biener <rguenther@suse.de>
1165 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1166 from contexts for DINFO_LEVEL_TERSE and below.
1168 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1170 * config/nds32/nds32-doubleword.md (move_<mode>): Require
1171 resiter_operand condition.
1172 * config/nds32/nds32.md (*move<mode>): Ditto.
1174 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1175 Monk Chiang <sh.chiang04@gmail.com>
1177 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1179 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1181 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1183 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1184 Kito Cheng <kito.cheng@gmail.com>
1186 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1187 nds32_cond_code_str, output_cond_branch,
1188 output_cond_branch_compare_zero, nds32_expand_cbranch,
1189 nds32_expand_cstore, nds32_expand_movcc,
1190 nds32_output_cbranchsi4_equality_zero,
1191 nds32_output_cbranchsi4_equality_reg,
1192 nds32_output_cbranchsi4_equality_reg_or_const_int,
1193 nds32_output_cbranchsi4_greater_less_zero: New functions.
1194 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1195 nds32_expand_cstore, nds32_expand_movcc,
1196 nds32_output_cbranchsi4_equality_zero,
1197 nds32_output_cbranchsi4_equality_reg,
1198 nds32_output_cbranchsi4_equality_reg_or_const_int,
1199 nds32_output_cbranchsi4_greater_less_zero): Declare.
1200 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1201 nds32_rimm11s_operand): New predicates.
1202 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1203 * config/nds32/nds32.md: Rewrite all the branch and conditional move
1206 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1208 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
1209 * config/nds32/nds32.md: Ditto.
1210 * config/nds32/pipelines.md: Ditto.
1212 2018-04-04 Richard Biener <rguenther@suse.de>
1214 PR tree-optimization/85168
1215 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
1216 propagating abnormals.
1218 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1220 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
1222 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1223 Kito Cheng <kito.cheng@gmail.com>
1225 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
1226 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
1227 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
1228 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
1229 * config/nds32/nds32.md (sibcall_internal): New.
1230 (sibcall_register): Remove.
1231 (sibcall_immediate): Remove.
1232 (sibcall_value_internal): New.
1233 (sibcall_value_register): Remove.
1234 (sibcall_value_immediate): Remove.
1235 * config/nds32/predicates.md (nds32_general_register_operand): New.
1236 (nds32_call_address_operand): New.
1238 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1240 PR rtl-optimization/85167
1241 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1242 bb_defs if *split_p, instead preinitialize it to NULL.
1244 PR tree-optimization/85156
1245 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1246 evaluating the argument multiple times.
1248 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
1250 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1252 (_mm_cvtpd_ps): Likewise.
1253 (_mm_cvttpd_epi32): Likewise.
1254 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1255 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1256 vector, pixel, and bool following altivec.h include.
1258 2018-04-03 Martin Sebor <msebor@redhat.com>
1260 * doc/extend.texi (Common Function Attributes): Clarify.
1261 (const attribute): Likewise.
1262 (pure attribute): Likewise.
1264 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1267 * config/i386/i386.c (ix86_expand_vector_set): Use
1268 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
1270 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
1272 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1273 instructions when changing rounding bits to preserve precision bits
1274 in the x87 control word.
1276 2018-04-03 Martin Liska <mliska@suse.cz>
1278 PR tree-optimization/82491
1279 * rtl.h (strip_offset_and_add): Replace += suboffset with
1280 poly_uint64 () + suboffset.
1282 2018-03-29 Martin Liska <mliska@suse.cz>
1283 Martin Jambor <mjambor@suse.cz>
1286 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1287 param_type is not an integral or pointer type.
1289 2018-04-03 Richard Biener <rguenther@suse.de>
1291 * sese.h (recompute_all_dominators): Remove.
1293 2018-04-02 Martin Sebor <msebor@redhat.com>
1295 * doc/invoke.texi (-Wrestrict): Fix typos.
1297 2018-04-02 Jim Wilson <jimw@sifive.com>
1299 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1300 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1301 (<optab>di3, <optab>si3_extend): Likewise.
1302 (<optab>si3_mask, <optab>si3_mask_1): New.
1303 (<optab>di3_mask, <optab>di3_mask_1): New.
1304 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1305 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1306 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1308 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1310 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1313 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1315 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1316 (nds32_canonicalize_comparison): New function.
1318 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1319 Kito Cheng <kito.cheng@gmail.com>
1320 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1322 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1323 * config/nds32/constants.md (unspec_volatile_element): Add
1324 UNSPEC_VOLATILE_RELAX_GROUP.
1325 * config/nds32/nds32-relax-opt.c: New file.
1326 * config/nds32/nds32-predicates.c
1327 (nds32_symbol_load_store_p): New function.
1328 * config/nds32/nds32-protos.h
1329 (nds32_symbol_load_store_p): Declare function.
1330 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1331 * config/nds32/nds32.c
1332 (nds32_register_pass): New function to register pass.
1333 (nds32_register_passes): New function to register passes.
1334 * config/nds32/nds32.md (relax_group): New pattern.
1335 * config/nds32/nds32.opt (mrelax-hint): New option.
1336 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1338 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
1340 * config/nds32/t-nds32: Modify files dependency.
1342 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1344 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1345 (PROFILE_HOOK): Define its implementation.
1347 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1349 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1350 type and 32-bit size.
1352 2018-04-01 Jakub Jelinek <jakub@redhat.com>
1355 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1356 (V_128_256): New mode iterator.
1357 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1358 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1359 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1361 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1362 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1364 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1367 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1368 NaN inputs correctly.
1370 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1373 * config/rs6000/vsx.md (??r): New mode attribute.
1374 (*vsx_mov<mode>_64bit): Use it.
1375 (*vsx_mov<mode>_32bit): Likewise.
1377 2018-03-30 Martin Sebor <msebor@redhat.com>
1379 PR tree-optimization/84818
1380 * builtins.c (check_access): Use warning_n.
1382 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1385 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1387 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1390 2018-03-30 Julia Koval <julia.koval@intel.com>
1393 * x86-tune.def (movx, partial_reg_dependency): Enable for
1396 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
1399 * lra-constraints.c (process_alt_operands): Move setting
1400 this_alternative_matches below.
1402 2018-03-29 Martin Liska <mliska@suse.cz>
1405 * doc/invoke.texi: Document how LTO works with debug info.
1406 Describe auto-load support of binutils. Mention 'x86-64'
1407 as valid option value of -march option.
1409 2018-03-29 Jakub Jelinek <jakub@redhat.com>
1411 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
1414 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
1415 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
1416 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
1419 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1422 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
1423 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
1424 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
1425 for RS6000_BTM_POWERPC64.
1426 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
1427 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
1428 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
1433 2018-03-28 Carl Love <cel@us.ibm.com>
1436 2017-09-27 Carl Love <cel@us.ibm.com>
1438 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
1439 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
1440 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
1443 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1445 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
1446 instead of __vector bool.
1447 (_mm_max_pu8): Likewise.
1448 (_mm_min_pi16): Likewise.
1450 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1453 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
1454 (DIVWEUO): Likewise.
1456 (DIVDEUO): Likewise.
1457 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
1458 DIVWEUO and DIVDEUO.
1459 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
1460 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
1461 (div_extend): Likewise.
1462 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
1464 (__builtin_divweuo): Likewise.
1465 (__builtin_divdeo): Likewise.
1466 (__builtin_divdeuo): Likewise.
1468 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1471 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
1472 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
1474 PR tree-optimization/82004
1475 * gimple-match-head.c (optimize_pow_to_exp): New function.
1476 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
1477 Don't fold to exp if optimize_pow_to_exp is false.
1479 2018-03-28 Martin Liska <mliska@suse.cz>
1482 * calls.c (initialize_argument_information): Fix trailing space.
1483 * common.opt: Fix typo and provide better explanation for
1484 -fsanitize-coverage option.
1485 * config/i386/i386.opt: Fix typo.
1487 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1488 Martin Liska <mliska@suse.cz>
1491 * gimplify.c (asan_poison_variable): Don't do the check for
1492 gimplify_omp_ctxp here.
1493 (gimplify_decl_expr): Do it here.
1494 (gimplify_target_expr): Likewise.
1496 2018-03-28 Martin Liska <mliska@suse.cz>
1499 * config/i386/i386.c (ix86_function_arg_advance): Do not call
1500 chkp_type_bounds_count if MPX is not enabled.
1502 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
1504 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
1506 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
1509 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
1510 function to create the function decl for complex long double
1511 multiply and divide for -mabi=ieeelongdouble.
1512 (init_float128_ieee): Call it.
1514 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
1517 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
1518 -fcf-protection=branch -mibt.
1519 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
1521 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1524 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
1526 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
1529 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
1530 extern array declarations.
1532 2018-03-27 Richard Biener <rguenther@suse.de>
1535 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
1536 explicit single_use checks.
1538 2018-03-27 Richard Biener <rguenther@suse.de>
1540 PR tree-optimization/85082
1541 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
1544 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1546 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
1547 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
1548 Turn on fasynchronous-unwind-tables and funwind-tables.
1550 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
1553 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
1554 (*bmi_blsr_<mode>_ccz): Ditto.
1556 2018-03-26 Tom de Vries <tom@codesourcery.com>
1558 PR tree-optimization/85063
1559 * omp-general.c (offloading_function_p): New function. Factor out
1561 * omp-offload.c (pass_omp_target_link::gate): ... here.
1562 * omp-general.h (offloading_function_p): Declare.
1563 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
1564 with attribute omp declare target for offloading functions.
1566 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
1568 PR tree-optimization/84005
1569 * tree-data-ref.h (get_base_for_alignment): Declare.
1570 * tree-data-ref.c (get_base_for_alignment_1): New function.
1571 (get_base_for_alignment): Likewise.
1572 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
1573 get_base_for_alignment to find a suitable base object, instead
1574 of always using drb->base_address.
1576 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1579 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
1580 known size by default.
1582 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
1585 * lra-constraints.c (process_alt_operands): Don't match BLKmode
1586 and non BLKmode operands.
1588 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1591 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
1592 Clean up attributes.
1594 2018-03-23 Richard Biener <rguenther@suse.de>
1597 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
1598 we are going to emit early debug for LTO.
1600 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1603 * function.c (match_asm_constraints_1): Don't optimize if input
1604 doesn't satisfy general_operand predicate for output's mode.
1607 * alias.c (write_dependence_p): Don't require for x_canonicalized
1608 non-VOIDmode if x has VOIDmode.
1611 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
1612 just don't try to optimize it rather than assert it never happens.
1614 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1616 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
1617 macro expansions for definition of ST_INTERNAL_<mode> and
1618 LD_INTERNAL_<mode> builtins.
1619 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
1621 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
1623 (altivec_expand_st_builtin): Likewise.
1624 (altivec_expand_builtin): Remove calls to deleted functions.
1625 (rs6000_address_for_altivec): Delete this function.
1626 * config/rs6000/vector.md: Remove expands for
1627 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
1629 2018-03-22 Sudakshina Das <sudi.das@arm.com>
1632 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
1633 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
1634 re-computing once computed.
1635 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
1636 (arm_init_machine_status): Initialize
1637 machine->static_chain_stack_bytes.
1639 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1642 * doc/extend.texi: Add four new prototypes for vec_ld.
1643 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
1644 definitions for more logical presentation.
1645 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
1646 entries for V1TI variants of __builtin_altivec_ld builtin.
1647 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
1648 handling of V1TI variant of LVX icode pattern.
1649 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
1650 (rs6000_gimple_fold_builtin): Likewise.
1651 (altivec_init_builtins): Add code to define
1652 __builtin_altivec_lvx_v1ti function.
1654 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1657 * function.c (match_asm_constraints_1): Don't do the optimization
1658 if input isn't a REG, SUBREG, MEM or constant.
1660 2018-03-22 Tom de Vries <tom@codesourcery.com>
1662 PR tree-optimization/84956
1663 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
1664 bb_has_abnormal_pred.
1666 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1669 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
1670 DECL_INITIAL (decl) to decl at the end.
1671 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
1674 2018-03-21 Joseph Myers <joseph@codesourcery.com>
1676 * doc/extend.texi (__builtin_tgmath): Document when complex
1677 integer types are treated as _Complex _Float64.
1679 2018-03-21 Tom de Vries <tom@codesourcery.com>
1681 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
1683 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1685 PR tree-optimization/84960
1686 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
1687 if it is ENTRY block, move them into single succ of ENTRY in that case.
1689 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
1691 PR tree-optimization/84811
1692 * poly-int.h (poly_span_traits): Remove the T3 parameter and
1693 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
1694 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
1695 (known_subrange_p): Update accordingly. Cast each value involved
1696 in the size comparison, rather than casting the result of the
1699 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1701 PR tree-optimization/84982
1702 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
1703 by flipping the least significant bit rather than all bits from
1704 bitpos to bitpos + bitsize - 1.
1706 2018-03-21 Nathan Sidwell <nathan@acm.org>
1708 * doc/extend.texi (Deprecated Features): Remove mention of
1709 long-deleted deprecations.
1711 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1714 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
1715 * configure: Regenerate.
1717 2018-03-21 Tom de Vries <tom@codesourcery.com>
1719 PR tree-optimization/83126
1720 * tree-parloops.c (num_phis): New function.
1721 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
1723 2018-03-21 Nathan Sidwell <nathan@acm.org>
1725 * doc/extend.texi (Deprecated Features): Update deprecated flags,
1726 mention anon-struct/union members and trailing attributes.
1728 2018-03-21 Bin Cheng <bin.cheng@arm.com>
1730 PR tree-optimization/84969
1731 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
1732 builtin memset partitions if they set different rhs values.
1734 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1736 PR rtl-optimization/84989
1737 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
1738 VEC_DUPLICATE with scalar result mode.
1740 2018-03-21 Martin Liska <mliska@suse.cz>
1743 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
1744 not intended return statement.
1746 2018-03-21 Martin Liska <mliska@suse.cz>
1749 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
1750 (chkp_find_bound_slots_1): Limit number of iterations.
1752 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
1755 * Minor grammar fixes for x86 options.
1757 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1760 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
1761 holding REG_CFA_RESTORE notes, instead turn them into a USE.
1763 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
1766 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
1767 (altivec_lvx_<mode>_1op): Likewise.
1768 (altivec_stvx_<mode>_2op): Likewise.
1769 (altivec_stvx_<mode>_1op): Likewise.
1770 (altivec_lvx_<VM2:mode>): New define_expand.
1771 (altivec_stvx_<VM2:mode>): Likewise.
1772 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
1773 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1774 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
1775 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1776 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
1777 (rs6000_gen_lvx): Likewise.
1778 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
1779 (altivec_expand_stv_builtin): Likewise.
1780 (altivec_expand_builtin): Likewise.
1781 * config/rs6000/vector.md: Likewise.
1783 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1786 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
1789 2018-03-20 Richard Biener <rguenther@suse.de>
1792 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
1793 sign-conversions as zero, fall back to standard scalar_stmt
1796 2018-03-20 Martin Liska <mliska@suse.cz>
1799 * predict.c (rebuild_frequencies): Handle case when we have
1800 PROFILE_ABSENT, but flag_guess_branch_prob is false.
1802 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1805 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
1806 flag_section_anchors.
1807 * varasm.c (use_blocks_for_decl_p): Remove hack for
1808 dw2_force_const_mem.
1811 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
1813 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
1814 be created, use lowpart_subreg of operands[0] rather than operands[0]
1816 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
1817 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
1818 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
1819 and n constraint instead of aarch64_shift_imm_di and Usd.
1820 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
1821 (*aarch64_<optab>_reg_minus<mode>3): ... this.
1823 2018-03-20 Sudakshina Das <sudi.das@arm.com>
1826 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
1827 to favor GPR over NEON registers.
1828 (<shift>di3_neon): Likewise.
1830 2018-03-20 Tom de Vries <tom@codesourcery.com>
1833 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
1834 (nvptx_process_pars): Emit bar.sync asap and alap.
1836 2018-03-20 Tom de Vries <tom@codesourcery.com>
1839 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
1840 seen_label if seen_label is already set.
1842 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1845 * config/i386/i386.c (fold_builtin_cpu): For features above 31
1846 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
1847 Use 1U instead of 1. Formatting fixes.
1850 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
1851 instead of TREE_TYPE (s1) for the return value.
1853 2018-03-19 Jakub Jelinek <jakub@redhat.com>
1855 PR tree-optimization/84946
1856 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
1857 bitsize + bitsize in poly_uint64 rather than poly_int64.
1860 * dwarf2asm.c: Include fold-const.c.
1861 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
1862 of decl rather than decl itself.
1864 PR rtl-optimization/84643
1865 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
1867 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
1870 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
1871 calling assemble_variable.
1873 2018-03-19 Sudakshina Das <sudi.das@arm.com>
1876 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
1877 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
1879 2018-03-19 Jim Wilson <jimw@sifive.com>
1882 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
1883 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
1884 (riscv_first_stack_step): Likewise.
1885 (riscv_option_override): Use STACK_BOUNDARY instead of
1887 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
1889 (BIGGEST_ALIGNMENT): Set to 128.
1890 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
1891 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
1894 2018-03-19 Richard Biener <rguenther@suse.de>
1896 PR tree-optimization/84933
1897 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
1898 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
1900 2018-03-19 Richard Biener <rguenther@suse.de>
1902 PR tree-optimization/84859
1903 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
1904 (cond_if_else_store_replacement): Perform sinking operation on
1905 single-store BBs regardless of MAX_STORES_TO_SINK setting.
1906 Generalize what a BB with a single eligible store is.
1908 2018-03-19 Richard Biener <rguenther@suse.de>
1910 PR tree-optimization/84929
1911 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
1912 chrec_is_positive against non-chrec arg.
1914 2018-03-19 Tamar Christina <tamar.christina@arm.com>
1917 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
1919 2018-03-18 Martin Liska <mliska@suse.cz>
1921 PR rtl-optimization/84635
1922 * regrename.c (build_def_use): Use matches_mode only when
1925 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
1927 PR tree-optimization/84913
1928 * tree-vect-loop.c (vectorizable_reduction): Don't try to
1929 vectorize chains of COND_EXPRs.
1931 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1933 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
1935 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1937 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
1939 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1941 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
1943 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1944 Kito Cheng <kito.cheng@gmail.com>
1946 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
1947 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
1948 (nds32_adjust_reg_alloc_order): New function.
1949 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
1951 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
1953 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
1954 nds32_print_operand, nds32_print_operand_address): Use
1955 HOST_WIDE_INT_PRINT_DEC instead.
1957 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1959 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
1961 2018-03-17 Jakub Jelinek <jakub@redhat.com>
1964 * config/i386/i386.c (initial_ix86_tune_features,
1965 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
1967 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
1968 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
1969 rather than 1u << ix86_tune. Formatting fix.
1970 (ix86_option_override_internal): Change ix86_arch_mask from
1971 unsigned int to unsigned HOST_WIDE_INT, initialize to
1972 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
1973 (ix86_function_specific_restore): Likewise.
1975 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1978 * postreload.c (reload_combine_recognize_pattern): Perform
1979 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
1980 truncate_int_for_mode the result for the destination's mode.
1983 * hsa-gen.c (mem_type_for_type): Fix comment typo.
1984 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
1986 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1989 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
1992 * lra-assigns.c (lra_split_hard_reg_for): Don't use
1993 regno_allocno_class_array and sorted_pseudos.
1994 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
1995 insns where regno is used.
1997 2018-03-16 Martin Liska <mliska@suse.cz>
2000 * multiple_target.c (create_dispatcher_calls): Redirect
2001 reference in the symbol table.
2003 2018-03-16 Martin Liska <mliska@suse.cz>
2006 * multiple_target.c (create_dispatcher_calls): Redirect also
2009 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2013 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2014 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2015 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2017 2018-03-16 Julia Koval <julia.koval@intel.com>
2019 * doc/invoke.texi (Skylake Server): Add CLWB.
2020 Cannonlake): Remove CLWB.
2022 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2024 PR tree-optimization/84841
2025 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2027 (FLOAT_ONE_CONST_TYPE): Define.
2028 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2029 (sort_by_operand_rank): Put entries with higher constant_type last
2030 rather than first to match comments.
2032 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
2034 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2037 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2040 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2041 check and use address of target always.
2043 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2046 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2047 (indirect_thunk_bnd_needed): Likewise.
2048 (indirect_thunks_used): Likewise.
2049 (indirect_thunks_bnd_used): Likewise.
2050 (indirect_return_needed): New.
2051 (indirect_return_bnd_needed): Likewise.
2052 (output_indirect_thunk_function): Add a bool argument for
2054 (output_indirect_thunk_function): Don't generate alias for
2055 function return thunk.
2056 (ix86_code_end): Call output_indirect_thunk_function to generate
2057 function return thunks.
2058 (ix86_output_function_return): Set indirect_return_bnd_needed
2059 and indirect_return_needed instead of indirect_thunk_bnd_needed
2060 and indirect_thunk_needed.
2062 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2064 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2065 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2066 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2068 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2069 Paul Hua <paul.hua.gm@gmail.com>
2072 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2074 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2076 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2077 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2080 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2083 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2084 instead of GET_MODE_SIZE when comparing Units.
2086 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2089 * varasm.c (hash_section): Return an unchangeble hash value
2090 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2091 Return !aarch64_can_use_per_function_literal_pools_p ().
2093 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2096 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2097 rather than address of cmode as last argument to prepare_cmp_insn.
2099 2018-03-15 Julia Koval <julia.koval@intel.com>
2101 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2102 F_AVX512VNNI, F_AVX512BITALG): New.
2104 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2107 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2108 insn for floating-point loads and stores.
2110 2018-03-14 Carl Love <cel@us.ibm.com>
2112 * config/rs6000/rs6000-c.c: Add macro definitions for
2113 ALTIVEC_BUILTIN_VEC_PERMXOR.
2114 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2115 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2116 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2117 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2119 * config/doc/extend.texi: Add prototypes for vec_permxor.
2121 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2124 * diagnostic-show-locus.c (class layout_point): Convert m_line
2125 from int to linenum_type.
2126 (line_span::comparator): Use linenum "compare" function when
2127 comparing line numbers.
2128 (test_line_span): New function.
2129 (layout_range::contains_point): Convert param "row" from int to
2131 (layout_range::intersects_line_p): Likewise.
2132 (layout::will_show_line_p): Likewise.
2133 (layout::print_source_line): Likewise.
2134 (layout::should_print_annotation_line_p): Likewise.
2135 (layout::print_annotation_line): Likewise.
2136 (layout::print_leading_fixits): Likewise.
2137 (layout::annotation_line_showed_range_p): Likewise.
2138 (struct line_corrections): Likewise for field m_row.
2139 (line_corrections::line_corrections): Likewise for param "row".
2140 (layout::print_trailing_fixits): Likewise.
2141 (layout::get_state_at_point): Likewise.
2142 (layout::get_x_bound_for_row): Likewise.
2143 (layout::print_line): Likewise.
2144 (diagnostic_show_locus): Likewise for locals "last_line" and
2146 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2147 * input.c (selftest::test_linenum_comparisons): New function.
2148 (selftest::input_c_tests): Call it.
2149 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2150 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2151 * selftest.h (ASSERT_GT): New macro.
2152 (ASSERT_GT_AT): New macro.
2153 (ASSERT_LT): New macro.
2154 (ASSERT_LT_AT): New macro.
2156 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
2158 PR rtl-optimization/84780
2159 * combine.c (distribute_links): Don't make a link based on pc_rtx.
2161 2018-03-14 Martin Liska <mliska@suse.cz>
2163 * tree.c (record_node_allocation_statistics): Use
2164 get_stats_node_kind.
2165 (get_stats_node_kind): New function extracted from
2166 record_node_allocation_statistics.
2167 (free_node): Use get_stats_node_kind.
2169 2018-03-14 Richard Biener <rguenther@suse.de>
2171 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2172 that the value-set of ANTIC_IN doesn't grow.
2175 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2177 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2178 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2180 2018-03-14 Julia Koval <julia.koval@intel.com>
2182 * config.gcc (icelake-client, icelake-server): New.
2184 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2185 (initial_ix86_arch_features): Ditto.
2186 (PTA_SKYLAKE): Add SGX.
2187 (PTA_ICELAKE): Remove.
2188 (PTA_ICELAKE_CLIENT): New.
2189 (PTA_ICELAKE_SERVER): New.
2190 (ix86_option_override_internal): Split up icelake on icelake client and
2192 (get_builtin_code_for_version): Ditto.
2193 (fold_builtin_cpu): Ditto.
2194 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2195 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2196 * config/i386/i386.h (processor_type): Ditto.
2197 * doc/invoke.texi: Ditto.
2199 2018-03-14 Jakub Jelinek <jakub@redhat.com>
2202 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2203 INTEGER_CST offset, add it together with bitpos / 8 and
2204 sign extend based on POINTER_SIZE.
2208 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
2211 * config/i386/constraints.md (Yc): New register constraint.
2212 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
2213 Use Yc constraint for alternative 2 of operand 0. Remove
2214 preferred_for_speed attribute.
2216 2018-03-14 Richard Biener <rguenther@suse.de>
2218 PR tree-optimization/84830
2219 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
2220 with the old one to avoid oscillations.
2222 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
2225 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
2227 (assign_by_spills): Return a flag of reload assignment failure.
2228 Do not process the reload assignment failures. Do not spill other
2229 reload pseudos if they has the same reg class. Update n if
2231 (lra_assign): Add a return arg. Set up from the result of
2232 assign_by_spills call.
2233 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2234 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2235 usage_insns if it is not NULL.
2236 (spill_hard_reg_in_range): New function.
2237 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2238 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2239 function prototypes.
2240 (lra_assign): Change prototype.
2241 * lra.c (lra): Add code to deal with fails by splitting hard reg
2244 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
2246 * config/riscv/riscv.opt (mrelax): New option.
2247 * config/riscv/riscv.c (riscv_file_start): Emit ".option
2248 "norelax" when riscv_mrelax is disabled.
2249 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2251 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2254 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2255 reassociation for int modes.
2257 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2259 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
2260 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
2262 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
2263 * config/aarch64/aarch64-sve.md
2264 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
2265 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2266 (*extend<mode><Vwide>2): Rename to...
2267 (aarch64_sve_extend<mode><Vwide>2): ...this.
2268 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2269 renaming the old pattern to...
2270 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
2272 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2273 define_expand, renaming the old pattern to...
2274 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2275 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2276 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2277 account when deciding which SVE instruction the optab should use.
2278 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2280 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2282 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2283 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2284 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2285 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2286 (tlsdesc_small_<mode>): Turn a define_expand and use
2287 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
2288 (tlsdesc_small_advsimd_<mode>): ...this.
2289 (tlsdesc_small_sve_<mode>): New pattern.
2291 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2293 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2294 (UNSPEC_UMUL_HIGHPART): New constants.
2295 (MUL_HIGHPART): New int iteraor.
2296 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2297 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2299 (*<su>mul<mode>3_highpart): New define_insn.
2301 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2304 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2307 2018-03-13 Martin Liska <mliska@suse.cz>
2310 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2312 (sem_item_optimizer::~sem_item_optimizer): Release it.
2313 (sem_item_optimizer::merge_classes): Register variable aliases.
2314 (sem_item_optimizer::fixup_pt_set): New function.
2315 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2316 * ipa-icf.h: Declare new variables and functions.
2318 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2321 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2322 integer_pow2p@2 and test integer_pow2p in condition.
2323 (A < 0 ? C : 0): Similarly for @1.
2326 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2327 characters starting at p contain '\0' character, don't look beyond
2331 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2332 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2335 * reg-stack.c (change_stack): Change update_end var from int to
2336 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2337 also call set_block_for_insn on the newly added insns and rescan.
2340 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2341 on the last operand.
2344 * tree.c (stabilize_reference_1): Return save_expr (e) for
2345 STATEMENT_LIST even if it doesn't have side-effects.
2347 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
2349 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2351 2018-03-12 Renlin Li <renlin.li@arm.com>
2353 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2354 aarch64_output_scalar_simd_mov_immediate.
2356 2018-03-12 Martin Sebor <msebor@redhat.com>
2358 PR tree-optimization/83456
2359 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2360 for perfectly overlapping calls to memcpy.
2361 (gimple_fold_builtin_memory_chk): Same.
2362 (gimple_fold_builtin_strcpy): Handle no-warning.
2363 (gimple_fold_builtin_stxcpy_chk): Same.
2364 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2366 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2368 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2369 parameter. Use it for SFmode.
2370 (rs6000_function_arg_advance_1): Adjust.
2371 (rs6000_function_arg): Adjust.
2372 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2374 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2376 PR rtl-optimization/84169
2377 PR rtl-optimization/84780
2378 * combine.c (can_combine_p): Check for a 2-insn combination whether
2379 the destination register is used between the two insns, too.
2381 2018-03-12 Richard Biener <rguenther@suse.de>
2383 PR tree-optimization/84803
2384 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2385 for refs DR analysis didn't process.
2387 2018-03-12 Richard Biener <rguenther@suse.de>
2389 PR tree-optimization/84777
2390 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2391 force-vectorize loops ignore whether we are optimizing for size.
2393 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
2395 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2396 (TARGET_MD_ASM_ADJUST): Define.
2398 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
2399 Kito Cheng <kito.cheng@gmail.com>
2400 Chung-Ju Wu <jasonwucj@gmail.com>
2402 * config/nds32/nds32.c (nds32_compute_stack_frame,
2403 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2404 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2405 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2406 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2407 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2408 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2409 * config/nds32/nds32.md (prologue, epilogue): Use macro
2410 NDS32_V3PUSH_AVAILABLE_P to do checking.
2412 2018-03-11 Jakub Jelinek <jakub@redhat.com>
2415 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
2416 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
2417 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
2418 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
2419 addition of most attributes on !orig_type_die or the attribute not
2420 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
2422 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2423 Chung-Ju Wu <jasonwucj@gmail.com>
2425 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
2427 * config/nds32/nds32.opt (mvh): New option.
2429 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2430 Chung-Ju Wu <jasonwucj@gmail.com>
2432 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
2434 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
2435 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
2438 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2439 Chung-Ju Wu <jasonwucj@gmail.com>
2441 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
2443 * config/nds32/nds32-multiple.md (strlensi): New pattern.
2444 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
2446 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
2447 Kito Cheng <kito.cheng@gmail.com>
2448 Chung-Ju Wu <jasonwucj@gmail.com>
2450 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
2451 UNSPEC_FFMISM and UNSPEC_FLMISM.
2452 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
2453 for ffb, ffmism and flmism.
2454 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
2455 (unspec_ffmism): Ditto.
2456 (unspec_flmism): Ditto.
2457 (nds32_expand_builtin_impl): Check if string extension is available.
2458 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
2459 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
2461 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
2464 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2467 * lra-assigns.c (assign_by_spills): Return a flag of reload
2468 assignment failure. Do not process the reload assignment
2469 failures. Do not spill other reload pseudos if they has the same
2471 (lra_assign): Add a return arg. Set up from the result of
2472 assign_by_spills call.
2473 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2474 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2475 usage_insns if it is not NULL.
2476 (spill_hard_reg_in_range): New function.
2477 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2478 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2479 function prototypes.
2480 (lra_assign): Change prototype.
2481 * lra.c (lra): Add code to deal with fails by splitting hard reg
2484 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
2487 * config/i386/i386.opt: Replace Enforcment with Enforcement.
2489 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
2492 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
2493 (dw_val_node): Add val_symbolic_view.
2494 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
2495 (symview_upper_bound): New.
2496 (new_line_info_table): Initialize symviews_since_reset.
2497 (dwarf2out_source_line): Count symviews_since_reset and set
2498 symview_upper_bound.
2499 (dw_val_equal_p): Handle symview.
2500 (add_AT_symview): New.
2501 (print_dw_val): Handle symview.
2502 (attr_checksum, attr_checksum_ordered): Likewise.
2503 (same_dw_val_p, size_of_die): Likewise.
2504 (value_format, output_die): Likewise.
2505 (add_high_low_attributes): Use add_AT_symview for entry_view.
2506 (dwarf2out_finish): Reset symview_upper_bound, clear
2509 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
2512 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
2513 Add strict argument and use it.
2514 (rs6000_split_multireg_move): Update for new strict argument.
2515 (mem_operand_gpr): Disallow all non-offsettable addresses.
2516 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
2518 2018-03-09 Jakub Jelinek <jakub@redhat.com>
2521 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
2522 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
2523 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
2526 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
2527 decl, use remap_type if we want to use the type.
2529 2018-03-09 Martin Sebor <msebor@redhat.com>
2531 PR tree-optimization/84526
2532 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2534 (builtin_access::generic_overlap): Be prepared to handle non-array
2537 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
2539 PR rtl-optimization/84682
2540 * lra-constraints.c (process_address_1): Check is_address flag
2541 for address constraints.
2542 (process_alt_operands): Likewise.
2543 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
2544 preprocess_constraints.
2545 * recog.h (preprocess_constraints): Add oploc parameter.
2547 * recog.c (preprocess_constraints): Test address_operand for
2548 CT_ADDRESS constraints.
2550 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2553 * lra-assigns.c (assign_by_spills): Return a flag of reload
2554 assignment failure. Do not process the reload assignment
2555 failures. Do not spill other reload pseudos if they has the same
2557 (lra_assign): Add a return arg. Set up from the result of
2558 assign_by_spills call.
2559 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2560 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2561 usage_insns if it is not NULL.
2562 (spill_hard_reg_in_range): New function.
2563 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2564 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2565 function prototypes.
2566 (lra_assign): Change prototype.
2567 * lra.c (lra): Add code to deal with fails by splitting hard reg
2570 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2573 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
2574 Accept complain bool parameter. Only emit errors if it is true.
2575 (arm_parse_cpu_option_name): Likewise.
2576 (arm_target_thumb_only): Adjust callers of the above.
2577 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
2578 prototype to take a default true bool parameter.
2579 (arm_parse_arch_option_name): Likewise.
2581 2018-03-09 David Malcolm <dmalcolm@redhat.com>
2582 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2586 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
2587 * configure: Regenerate.
2588 * configure.ac ("linker --version-script option"): New.
2589 ("linker soname option"): New.
2591 2018-03-09 Richard Biener <rguenther@suse.de>
2593 PR tree-optimization/84775
2594 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
2595 immediate uses of predicate stmts and mark them modified.
2598 PR tree-optimization/84178
2599 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2601 (version_loop_for_if_conversion): Delay update_ssa call.
2602 (tree_if_conversion): Delay update_ssa until after predicate
2605 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
2608 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
2609 when the function accesses prior frames.
2611 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2614 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
2615 gen_llsym, otherwise call maybe_gen_llsym.
2618 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
2619 has ',' character inside of it.
2621 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2624 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
2625 as clobbering CC_REGNUM.
2627 2018-03-08 Richard Biener <rguenther@suse.de>
2630 * tree-scalar-evolution.c: Include tree-into-ssa.h.
2631 (follow_copies_to_constant): Do not follow SSA names registered
2634 2018-03-08 Richard Biener <rguenther@suse.de>
2636 PR tree-optimization/84178
2637 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2639 (version_loop_for_if_conversion): Delay update_ssa call.
2640 (tree_if_conversion): Delay update_ssa until after predicate
2643 2018-03-08 David Malcolm <dmalcolm@redhat.com>
2645 PR tree-optimization/84178
2646 * tree-if-conv.c (release_bb_predicate): Remove the
2647 the assertion that the stmts have NULL use_ops.
2648 Discard the statements, asserting that they haven't
2649 yet been added to a BB.
2651 2018-03-08 Richard Biener <rguenther@suse.de>
2653 PR tree-optimization/84746
2654 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
2655 (phi_translate): Pass in destination ANTIC_OUT set.
2656 (phi_translate_1): Likewise. For a simplified result lookup
2657 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
2658 (phi_translate_set): Adjust.
2659 (do_pre_regular_insertion): Likewise.
2660 (do_pre_partial_partial_insertion): Likewise.
2662 2018-03-08 Martin Liska <mliska@suse.cz>
2664 PR gcov-profile/84735
2665 * doc/gcov.texi: Document usage of profile files.
2666 * gcov-io.h: Document changes in the format.
2668 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
2672 * dwarf2out.c (struct dw_line_info_table): Update comments for
2674 (FORCE_RESET_NEXT_VIEW): New.
2675 (FORCE_RESETTING_VIEW_P): New.
2676 (RESETTING_VIEW_P): Check for -1 too.
2677 (ZERO_VIEW_P): Likewise.
2678 (new_line_info_table): Force-reset next view.
2679 (dwarf2out_begin_function): Likewise.
2680 (dwarf2out_source_line): Simplify zero_view_p initialization.
2681 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
2682 view directly. Omit view when omitting .loc at line 0.
2684 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2686 PR tree-optimization/84740
2687 * tree-switch-conversion.c (process_switch): Call build_constructors
2688 only if info.phi_count is non-zero.
2690 PR tree-optimization/84739
2691 * tree-tailcall.c (find_tail_calls): Check call arguments against
2692 DECL_ARGUMENTS (current_function_decl) rather than
2693 DECL_ARGUMENTS (func) when checking for tail recursion.
2695 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2697 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
2698 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
2699 Volker Reichelt's entry and add entries for people that perform
2700 GCC fuzzy testing and report numerous bugs.
2702 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
2705 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
2706 readonly data in sdata, if that is disabled.
2707 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
2708 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2709 -mreadonly-in-sdata option.
2711 2018-03-07 Martin Sebor <msebor@redhat.com>
2713 PR tree-optimization/84468
2714 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
2715 basic block when looking for nul assignment.
2717 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
2720 * except.h (output_function_exception_table): Adjust prototype.
2721 * except.c (output_function_exception_table): Remove FNNAME parameter
2722 and add SECTION parameter. Ouput one part of the table at a time.
2723 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
2724 the first part of the exception table and emit unwind directives.
2725 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
2726 (i386_pe_seh_cold_init): Likewise.
2727 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
2728 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
2729 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
2730 (ix86_output_call_insn): Emit a nop in one more case for SEH.
2731 * config/i386/winnt.c: Include except.h.
2732 (struct seh_frame_state): Add reg_offset, after_prologue and
2733 in_cold_section fields.
2734 (i386_pe_seh_end_prologue): Set seh->after_prologue.
2735 (i386_pe_seh_cold_init): New function.
2736 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
2737 to seh->in_cold_section.
2738 (seh_emit_push): Record the offset of the push.
2739 (seh_emit_save): Record the offet of the save.
2740 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
2741 Test seh->after_prologue to disregard the epilogue.
2742 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
2743 (i386_pe_end_cold_function): New function.
2745 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2748 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
2749 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
2752 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
2754 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
2755 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
2758 * multiple_target.c: Include tree-inline.h and intl.h.
2759 (expand_target_clones): Diagnose and fail if node->definition and
2760 !tree_versionable_function_p (node->decl).
2762 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
2764 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
2766 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
2767 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
2768 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
2770 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2773 * combine.c (try_combine): Use reg_or_subregno instead of handling
2774 just paradoxical SUBREGs and REGs.
2776 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
2778 * config/arc/arc.c (arc_finalize_pic): Remove function.
2779 (arc_must_save_register): We use single base PIC register, remove
2780 checks to save/restore the PIC register.
2781 (arc_expand_prologue): Likewise.
2782 * config/arc/arc-protos.h (arc_set_default_type_attributes):
2784 (arc_verify_short): Likewise.
2785 (arc_attr_type): Likewise.
2786 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
2787 (walk_stores): Likewise.
2788 (arc_address_cost): Make it static.
2789 (arc_verify_short): Likewise.
2790 (branch_dest): Likewise.
2791 (arc_attr_type): Likewise.
2792 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
2793 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
2794 (arc_final_prescan_insn): Remove inserting the nops due to
2795 hardware hazards. It is done in reorg step.
2796 (insn_length_variant_t): Remove.
2797 (insn_length_parameters_t): Likewise.
2798 (arc_insn_length_parameters): Likewise.
2799 (arc_get_insn_variants): Likewise.
2800 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
2802 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2805 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
2808 PR tree-optimization/84687
2809 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
2811 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
2813 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2815 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
2816 Rename to ppc_speculation_barrier.
2817 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
2818 __builtin_ppc_speculation_barrier.
2820 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2823 * combine.c (combine_simplify_rtx): Don't try to simplify if
2824 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
2827 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
2829 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
2830 to 32 bytes when compiling for POWER9.
2832 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2835 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
2836 regparm >= 3 with no arg reg available also for calls with
2837 flag_force_indirect_call. Pass decl to ix86_function_regparm.
2840 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
2842 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
2844 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
2847 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
2849 2018-03-05 Richard Biener <rguenther@suse.de>
2851 PR tree-optimization/84486
2852 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
2853 When inserting a __builtin_assume_aligned call set the LHS
2854 SSA name alignment info accordingly.
2856 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
2858 PR tree-optimization/84114
2859 * config/aarch64/aarch64.c (aarch64_reassociation_width)
2860 Avoid reassociation of FLOAT_MODE addition.
2862 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
2864 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
2865 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
2866 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
2867 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
2868 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
2869 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
2870 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
2872 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
2873 __builtin_ia32_wbinvd): New builtins.
2874 (SPECIAL_ARGS2): New.
2875 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
2876 (SPECIAL_ARGS2): New.
2877 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
2878 (ix86_valid_target_attribute_inner_p): Ditto.
2879 (ix86_init_mmx_sse_builtins): Add special_args2.
2880 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
2881 TARGET_WBNOINVD_P): New.
2882 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
2883 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
2884 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
2885 * config/i386/immintrin.h (_wbinvd): New intrinsic.
2886 * config/i386/pconfigintrin.h: New file.
2887 * config/i386/wbnoinvdintrin.h: Ditto.
2888 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
2889 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
2891 2018-03-05 Richard Biener <rguenther@suse.de>
2893 PR tree-optimization/84670
2894 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2896 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2897 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2898 (compute_antic_aux): Only assert the number of values in ANTIC_IN
2899 doesn't grow if all successors (recursively) were visited at least
2902 2018-03-05 Richard Biener <rguenther@suse.de>
2904 PR tree-optimization/84650
2905 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
2906 if executed in the loop pipeline.
2908 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
2910 * doc/configfiles.texi (Configuration Files): Move info about
2911 conditionalizing $target-protos.h to...
2912 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
2913 differs from $target-protos.h.
2915 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
2916 Chung-Ju Wu <jasonwucj@gmail.com>
2918 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
2919 * config/nds32/nds32-multiple.md (setmemsi): Define.
2920 * config/nds32/nds32-memory-manipulation.c
2921 (nds32_gen_dup_4_byte_to_word_value): New.
2922 (emit_setmem_word_loop): New.
2923 (emit_setmem_byte_loop): New.
2924 (nds32_expand_setmem_loop): New.
2925 (nds32_expand_setmem_loop_v3m): New.
2926 (nds32_expand_setmem_unroll): New.
2927 (nds32_expand_setmem): New.
2929 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2930 Chung-Ju Wu <jasonwucj@gmail.com>
2932 * config/nds32/nds32-memory-manipulation.c
2933 (nds32_emit_load_store): New.
2934 (nds32_emit_post_inc_load_store): New.
2935 (nds32_emit_mem_move): New.
2936 (nds32_emit_mem_move_block): New.
2937 (nds32_expand_movmemsi_loop_unknown_size): New.
2938 (nds32_expand_movmemsi_loop_known_size): New.
2939 (nds32_expand_movmemsi_loop): New.
2940 (nds32_expand_movmemsi_unroll): New.
2941 (nds32_expand_movmemqi): Rename ...
2942 (nds32_expand_movmemsi): ... to this.
2943 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
2944 (movmemsi): ... to this.
2945 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
2946 (nds32_expand_movmemsi): ... to this.
2948 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2949 Monk Chiang <sh.chiang04@gmail.com>
2950 Chung-Ju Wu <jasonwucj@gmail.com>
2952 * config/nds32/nds32-protos.h
2953 (nds32_expand_load_multiple): New arguments.
2954 (nds32_expand_store_multiple): Ditto.
2955 (nds32_valid_multiple_load_store): Rename ...
2956 (nds32_valid_multiple_load_store_p): ... to this.
2957 * config/nds32/nds32-memory-manipulation.c
2958 (nds32_expand_load_multiple): Refine implementation.
2959 (nds32_expand_store_multiple): Ditto.
2960 * config/nds32/nds32-multiple.md
2961 (load_multiple): Update nds32_expand_load_multiple interface.
2962 (store_multiple): Update nds32_expand_store_multiple interface.
2963 * config/nds32/nds32-predicates.c
2964 (nds32_valid_multiple_load_store): Rename ...
2965 (nds32_valid_multiple_load_store_p): ... to this and refine
2967 * config/nds32/predicates.md
2968 (nds32_load_multiple_and_update_address_operation): New predicate.
2969 (nds32_store_multiple_and_update_address_operation): New predicate.
2971 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2972 Chung-Ju Wu <jasonwucj@gmail.com>
2974 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
2975 (combo): New attribute.
2976 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
2978 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
2980 * config/nds32/nds32.opt: Change -mcmodel= default value.
2982 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
2983 Monk Chiang <sh.chiang04@gmail.com>
2984 Chung-Ju Wu <jasonwucj@gmail.com>
2986 * config/nds32/constants.md (unspec_element): New enum.
2987 * config/nds32/constraints.md (Umw): New constraint.
2988 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
2989 * config/nds32/nds32-intrinsic.md: Likewise.
2990 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
2991 (nds32_valid_smw_lwm_base_p): New.
2992 (nds32_output_smw_single_word): New.
2993 (nds32_output_lmw_single_word): New.
2994 (nds32_expand_unaligned_load): New.
2995 (nds32_expand_unaligned_store): New.
2996 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
2997 (nds32_output_smw_single_word): Declare.
2998 (nds32_output_lmw_single_word): Declare.
2999 (nds32_expand_unaligned_load): Declare.
3000 (nds32_expand_unaligned_store): Declare.
3001 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3002 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3003 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3004 NDS32_BUILTIN_UASTORE_DW.
3005 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3008 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
3009 Kito Cheng <kito.cheng@gmail.com>
3010 Chung-Ju Wu <jasonwucj@gmail.com>
3012 * config/nds32/nds32-intrinsic.c
3013 (nds32_expand_builtin_null_ftype_reg): Delete.
3014 (nds32_expand_builtin_reg_ftype_imm): Ditto.
3015 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3016 (nds32_read_argument): New.
3017 (nds32_legitimize_target): Ditto.
3018 (nds32_legitimize_argument): Ditto.
3019 (nds32_check_constant_argument): Ditto.
3020 (nds32_expand_unop_builtin): Ditto.
3021 (nds32_expand_unopimm_builtin): Ditto.
3022 (nds32_expand_binop_builtin): Ditto.
3023 (nds32_builtin_decl_impl): Ditto.
3024 (builtin_description): Ditto.
3025 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3026 (nds32_init_builtins_impl): Ditto.
3027 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3028 (nds32_builtin_decl): New.
3029 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3030 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3032 2018-03-02 Jeff Law <law@redhat.com>
3034 * reorg.c (stop_search_p): Handle DEBUG_INSN.
3035 (redundant_insn, fill_simple_delay_slots): Likewise.
3036 (fill_slots_from_thread): Likewise.
3037 * resource.c (mark_referenced_resources): Likewise.
3038 (mark_set_resources, find_dead_or_set_registers): Likewise.
3040 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3042 * substring-locations.h (format_warning_va): Formatting fix for
3044 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3046 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3047 * substring-locations.c: Include intl.h.
3048 (format_warning_va): Turned into small wrapper around
3049 format_warning_n_va, renamed to ...
3050 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3051 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3053 (format_warning_at_substring_n): New function.
3054 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3055 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3056 format_warning_at_substring with just a shorter name instead of
3057 const function pointer.
3058 (fmtwarn_n): New function.
3059 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3060 appropriate, get rid of all the fmtstr temporaries, move conditionals
3061 with G_() wrapped string literals directly into fmtwarn arguments,
3062 cast dir.len to (int), formatting fixes.
3064 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3066 * doc/invoke.texi: Remove "Cilk Plus" references.
3068 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3069 Richard Biener <rguenther@suse.de>
3072 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3073 for error or warning attributes if CALL_FROM_THUNK_P is set.
3076 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3079 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3080 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3083 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3084 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3086 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3087 instead of -1U in last predictors element's probability member.
3089 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3092 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3093 arguments if they are comparable.
3095 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3097 PR tree-optimization/84634
3098 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3099 masks and masked_loop_p with a single loop_masks, making sure it's
3100 null for bb vectorization.
3102 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3104 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3105 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3106 loop->force_vectorize to check whether there is no alias.
3108 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3111 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3113 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3115 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3116 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3117 instead of a loop around prev_real_insn.
3118 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3122 * config/i386/i386.c (ix86_print_operand): Use conditional
3123 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3126 2018-03-02 Richard Biener <rguenther@suse.de>
3128 PR tree-optimization/84427
3129 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3130 (bitmap_set_subtract_values): Rewrite to handle multiple
3133 (prune_clobbered_mems): Likewise.
3134 (phi_translate): Take edge instead of pred/phiblock.
3135 (phi_translate_1): Likewise.
3136 (phi_translate_set): Likewise. Insert all translated
3137 exprs for a value into the set, keeping possibly multiple
3138 expressions per value.
3139 (compute_antic_aux): Adjust for phi_translate changes.
3140 When intersecting union the expressions and prune those
3141 not in the final value set, keeping possibly multiple
3142 expressions per value. Do not use value-insertion
3143 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3144 all expressions. Add verification that the value-sets
3145 only shrink during iteration.
3146 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3147 (do_pre_regular_insertion): Likewise.
3148 (do_pre_partial_partial_insertion): Likewise.
3150 2018-03-02 Richard Biener <rguenther@suse.de>
3153 * config/darwin.c (saved_debug_info_level): New static global.
3154 (darwin_asm_lto_start): Disable debug info generation for LTO out.
3155 (darwin_asm_lto_end): Restore debug info generation settings.
3157 2018-03-01 Martin Liska <mliska@suse.cz>
3160 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3163 2018-03-01 Richard Biener <rguenther@suse.de>
3166 * dwarf2out.c (gen_variable_die): Properly handle late VLA
3167 type annotation with LTO when debug was disabled at compile-time.
3169 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3171 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3173 (mips_final_postscan_insn): Likewise.
3175 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
3177 PR rtl-optimization/84528
3178 * alias.c (init_alias_target): Add commentary.
3179 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3180 a unique base value if the frame pointer is not eliminated
3181 to the stack pointer.
3183 2018-03-01 Tom de Vries <tom@codesourcery.com>
3185 PR rtl-optimization/83327
3186 * lra-int.h (hard_regs_spilled_into): Declare.
3187 * lra.c (hard_regs_spilled_into): Define.
3188 (init_reg_info): Init hard_regs_spilled_into.
3189 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3190 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3191 (process_bb_lives): Handle hard_regs_spilled_into.
3192 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3193 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3195 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3197 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3198 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3199 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3200 * config/rs6000/aix72.h: New file.
3202 2018-02-28 Jakub Jelinek <jakub@redhat.com>
3204 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3205 instead of warning_at with conditional singular and plural messages
3209 * stor-layout.c (update_alignment_for_field): For
3210 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
3211 && !DECL_PACKED (field), do the alignment update, just use
3212 only desired_align instead of MAX (type_align, desired_align)
3214 (place_field): Don't do known_align < desired_align handling
3215 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
3216 is non-NULL, instead do it after rli->prev_field handling and
3217 only if not within a bitfield word. For DECL_PACKED (field)
3218 use type_align of BITS_PER_UNIT.
3220 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
3222 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
3223 superfluous parentheses and trailing spaces.
3225 2018-02-28 Richard Biener <rguenther@suse.de>
3227 PR tree-optimization/84584
3228 * graphite-scop-detection.c (scop_detection::add_scop): Discard
3229 SCoPs with fake exit edge.
3231 2018-02-28 Martin Liska <mliska@suse.cz>
3234 * timevar.c (timer::print): Fix format to properly print 100%
3237 2018-02-28 Richard Biener <rguenther@suse.de>
3240 * genmatch.c (capture_info::walk_match): Do not mark
3241 captured expressions without operands as expr_p given
3242 they act more like predicates and should be subject to
3243 "lost tail" side-effect preserving.
3245 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
3247 PR rtl-optimization/81611
3248 * auto-inc-dec.c (attempt_change): Move dead note from
3249 mem_insn if it's the next use of regno
3250 (find_address): Take address use of reg holding
3251 non-incremented value. Add parm to limit search to the named
3253 (merge_in_block): Attempt to use a mem insn that is the next
3254 use of the original regno.
3256 2018-02-27 Martin Sebor <msebor@redhat.com>
3259 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
3260 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
3262 2018-02-27 Martin Sebor <msebor@redhat.com>
3264 PR translation/84207
3265 * diagnostic-core.h (warning_n, error_n, inform_n): Change
3266 n argument to unsigned HOST_WIDE_INT.
3267 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3268 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
3269 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3270 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3272 2018-02-27 Richard Biener <rguenther@suse.de>
3274 PR tree-optimization/84512
3275 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3276 Do not use the estimate returned from record_stmt_cost for
3277 the scalar iteration cost but sum properly using add_stmt_cost.
3279 2018-02-27 Richard Biener <rguenther@suse.de>
3281 PR tree-optimization/84466
3282 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3283 Adjust last change to less strictly validate use operands.
3285 2018-02-27 Martin Liska <mliska@suse.cz>
3287 PR gcov-profile/84548
3288 * gcov.c (process_file): Allow partial overlap and consider it
3289 also as group functions.
3290 (output_lines): Properly calculate range of lines for a group.
3292 2018-02-27 Martin Liska <mliska@suse.cz>
3294 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3295 'ggc' suffixes. Change first column width.
3296 (timer::print): Fix formatting of the column.
3298 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3300 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3301 preserve inline entry blocks for the sake of debug inline
3302 entry point markers alone.
3303 (remove_unused_locals): Suggest in comments a better place to
3304 force the preservation of inline entry blocks that are
3305 otherwise unused, but do not preserve them.
3307 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3309 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3311 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3314 * config/i386/constraints.md (Bs): Replace
3315 ix86_indirect_branch_register with
3316 TARGET_INDIRECT_BRANCH_REGISTER.
3318 * config/i386/i386.md (indirect_jump): Likewise.
3319 (tablejump): Likewise.
3320 (*sibcall_memory): Likewise.
3321 (*sibcall_value_memory): Likewise.
3322 Peepholes of indirect call and jump via memory: Likewise.
3323 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3324 (*sibcall_value_GOT_32): Likewise.
3325 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3326 (GOT_memory_operand): Likewise.
3327 (call_insn_operand): Likewise.
3328 (sibcall_insn_operand): Likewise.
3329 (GOT32_symbol_operand): Likewise.
3330 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3332 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3334 PR rtl-optimization/83496
3335 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3336 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3337 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3338 redundant insn, if any.
3339 (relax_delay_slots): Likewise.
3340 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3342 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
3344 PR tree-optimization/83965
3345 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3346 that grouped statements are part of a reduction chain. Return
3347 true if the statement is not marked as a reduction itself but
3349 (vect_recog_dot_prod_pattern): Don't check whether the statement
3350 is part of a group here.
3351 (vect_recog_sad_pattern): Likewise.
3352 (vect_recog_widen_sum_pattern): Likewise.
3354 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3357 * final.c (rest_of_clean_state): Also look for calls inside sequences.
3359 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3362 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3364 (ix86_output_indirect_function_return): New prototype.
3365 (ix86_split_simple_return_pop_internal): Likewise.
3366 * config/i386/i386.c (indirect_return_via_cx): New.
3367 (indirect_return_via_cx_bnd): Likewise.
3368 (indirect_thunk_name): Handle return va CX_REG.
3369 (output_indirect_thunk_function): Create alias for
3370 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3371 (ix86_output_indirect_jmp): Remove the bool argument.
3372 (ix86_output_indirect_function_return): New function.
3373 (ix86_split_simple_return_pop_internal): Likewise.
3374 * config/i386/i386.md (*indirect_jump): Don't pass false
3375 to ix86_output_indirect_jmp.
3376 (*tablejump_1): Likewise.
3377 (simple_return_pop_internal): Change it to define_insn_and_split.
3378 Call ix86_split_simple_return_pop_internal to split it for
3380 (simple_return_indirect_internal): Call
3381 ix86_output_indirect_function_return instead of
3382 ix86_output_indirect_jmp.
3384 2018-02-26 Jakub Jelinek <jakub@redhat.com>
3387 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3388 memset and value initialization afterwards.
3390 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
3392 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3394 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3397 * common/config/aarch64/aarch64-common.c
3398 (aarch_option_optimization_table[]): Switch
3399 off fomit-frame-pointer
3401 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3402 Chung-Ju Wu <jasonwucj@gmail.com>
3404 * config/nds32/nds32-multiple.md (load_multiple): Disallow
3406 (store_multiple): Ditto.
3408 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3410 * config.gcc: Add --with-cpu support for nds32 target.
3411 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
3412 * config/nds32/nds32.opt: Add -mcpu= option.
3414 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
3416 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
3417 isel=yes): Warn for these deprecated options.
3419 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
3421 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
3422 ISA_2_5_MASKS_EMBEDDED.
3424 2018-02-23 Jakub Jelinek <jakub@redhat.com>
3426 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
3427 p->max as pointers rather than using iterative_hash_expr.
3429 2018-02-23 Carl Love <cel@us.ibm.com>
3431 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
3432 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
3434 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
3435 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
3436 P8V_BUILTIN_VEC_VUNSIGNED2.
3438 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
3441 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
3442 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
3444 * lra-constraints.c (curr_insn_transform): Set up
3445 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
3447 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
3448 * lra-eliminations.c (spill_pseudos): Ditto.
3449 (process_insn_for_elimination): Ditto.
3450 * lra-lives.c (reg_early_clobber_p): Use the new macros.
3451 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
3452 LRA_NON_CLOBBERED_ALT.
3454 2018-02-22 Martin Sebor <msebor@redhat.com>
3456 PR tree-optimization/84480
3457 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
3458 to maybe_diag_stxncpy_trunc. Call it.
3459 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
3460 from gimple_fold_builtin_strcpy. Print inlining stack.
3461 (handle_builtin_stxncpy): Print inlining stack.
3462 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
3464 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
3467 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
3468 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
3469 and -fcheck-pointer-bounds are used together.
3470 (indirect_thunk_prefix): New enum.
3471 (indirect_thunk_need_prefix): New function.
3472 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
3473 "_nt" instead of "_bnd" for NOTRACK prefix.
3474 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
3475 (output_indirect_thunk_function): Likewise.
3477 (ix86_code_end): Update output_indirect_thunk_function calls.
3478 (ix86_output_indirect_branch_via_reg): Replace
3479 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
3480 (ix86_output_indirect_branch_via_push): Likewise.
3481 (ix86_output_function_return): Likewise.
3482 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
3483 incompatible with -fcf-protection=branch and
3484 -fcheck-pointer-bounds.
3486 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3489 * config/aarch64/aarch64.c (aarch64_print_address_internal):
3490 Change gcc_assert call to output_operand_lossage.
3492 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3494 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
3496 2018-02-22 DJ Delorie <dj@redhat.com>
3497 Sebastian Perta <sebastian.perta@renesas.com>
3498 Oleg Endo <olegendo@gcc.gnu.org>
3500 * config/rx/rx.c (rx_rtx_costs): New function.
3501 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
3503 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
3505 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
3507 2018-02-22 Martin Liska <mliska@suse.cz>
3510 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
3511 Add "native" as a possible value.
3513 2018-02-22 Martin Liska <mliska@suse.cz>
3516 * config/i386/i386.c (ix86_option_override_internal):
3517 Add "native" as a possible value for -march and -mtune.
3519 2018-02-22 Jakub Jelinek <jakub@redhat.com>
3522 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
3523 to all type variants.
3525 PR tree-optimization/84503
3526 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
3527 width as info->bitpos + info->bitsize - start.
3528 (merged_store_group::merge_overlapping): Simplify width computation.
3529 (check_no_overlap): New function.
3530 (imm_store_chain_info::try_coalesce_bswap): Compute expected
3531 start + width and last_order of the group, fail if check_no_overlap
3533 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
3534 to group if check_no_overlap fails.
3536 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3538 * config/rs6000/altivec.md: Delete contraint arguments to
3539 define_expand, define_split, and define_peephole2, and in
3540 define_insn_and_split if always unused.
3541 * config/rs6000/darwin.md: Ditto.
3542 * config/rs6000/dfp.md: Ditto.
3543 * config/rs6000/rs6000.md: Ditto.
3544 * config/rs6000/sync.md: Ditto.
3545 * config/rs6000/vector.md: Ditto.
3546 * config/rs6000/vsx.md: Ditto.
3548 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3550 * config/rs6000/altivec.md: Write output control strings as braced
3551 blocks instead of double-quoted strings.
3552 * config/rs6000/darwin.md: Ditto.
3553 * config/rs6000/rs6000.md: Ditto.
3554 * config/rs6000/vector.md: Ditto.
3555 * config/rs6000/vsx.md: Ditto.
3557 2018-02-21 Jason Merrill <jason@redhat.com>
3559 PR c++/84314 - ICE with templates and fastcall attribute.
3560 * attribs.c (build_type_attribute_qual_variant): Remove assert.
3562 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3564 * ipa-cp.c (determine_versionability): Fix comment typos.
3566 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3569 * ipa-cp.c (determine_versionability): Do not version functions caling
3572 2018-02-21 Martin Liska <mliska@suse.cz>
3575 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
3576 Add "native" as a possible value.
3577 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
3578 the macro when native cpu detection is available.
3580 2018-02-21 Martin Liska <mliska@suse.cz>
3583 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
3584 Add "native" as a possible value.
3585 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
3586 when native cpu detection is available.
3588 2018-02-21 Jakub Jelinek <jakub@redhat.com>
3589 Martin Sebor <msebor@redhat.com>
3591 PR tree-optimization/84478
3592 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
3594 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
3595 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
3596 support which is conservatively correct, for 2 only stay conservative
3597 for maxlen. Formatting and comment capitalization fixes. Add STRICT
3598 argument to the 2 argument get_range_strlen, adjust 6 arg
3599 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
3601 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
3602 (gimple_fold_builtin_strlen): Pass true as last argument to
3605 2018-02-20 Martin Sebor <msebor@redhat.com>
3608 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
3609 (builtin_memref::set_base_and_offset): Same. Handle inner references.
3610 (builtin_memref::builtin_memref): Factor out parts into
3611 set_base_and_offset and call it.
3613 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3616 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
3617 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
3618 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
3619 search at the associated MODE_INT.
3621 2018-02-20 Jeff Law <law@redhat.com>
3624 PR tree-optimization/81592
3626 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
3627 for range data rather than using global data.
3628 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
3629 range data rather than using global data.
3630 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
3631 pass it to children as needed.
3632 (struct directive::fmtresult): Similarly.
3633 (struct directive::set_width): Similarly.
3634 (struct directive::set_precision): Similarly.
3635 (format_integer, format_directive, parse_directive): Similarly.
3636 (format_none): Accept unnamed vr_values parameter.
3637 (format_percent, format_floating, format_character): Similarly.
3638 (format_string, format_plain): Similarly.
3639 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
3640 the EVRP range analyzer for range data rather than using global data.
3641 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
3642 gimple-ssa-evrp-analyze.h
3643 (class sprintf_dom_walker): Add after_dom_children member function.
3644 Add evrp_range_analyzer member.
3645 (sprintf_dom_walker::before_dom_children): Call into the EVRP
3646 range analyzer as needed.
3647 (sprintf_dom_walker::after_dom_children): New member function.
3648 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
3650 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
3651 (evrp_range_analyzer::pop_to_marker): Likewise.
3653 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3655 PR tree-optimization/84419
3656 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
3657 with the required type if its current type is compatible but
3660 2018-02-20 Jakub Jelinek <jakub@redhat.com>
3663 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
3664 after vectorization.
3666 2018-02-20 Martin Liska <mliska@suse.cz>
3669 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
3670 possible values if we don't have a hint.
3672 2018-02-20 Martin Liska <mliska@suse.cz>
3676 * final.c (shorten_branches): Build align_tab array with one
3678 * opts.c (finish_options): Add alignment option limit check.
3679 (MAX_CODE_ALIGN): Likewise.
3680 (MAX_CODE_ALIGN_VALUE): Likewise.
3681 * doc/invoke.texi: Document maximum allowed option value for
3682 all -falign-* options.
3684 2018-02-19 Jakub Jelinek <jakub@redhat.com>
3687 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
3688 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
3689 * var-tracking.c (emit_note_insn_var_location): Remove all references
3690 to NOTE_INSN_CALL_ARG_LOCATION.
3691 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
3692 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
3693 Use copy_rtx_if_shared.
3694 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
3695 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
3696 (dwarf2out_var_location): Remove handling of
3697 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
3699 * final.c (final_scan_insn): Remove all references to
3700 NOTE_INSN_CALL_ARG_LOCATION.
3701 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
3702 before dumping final insns.
3703 * except.c (emit_note_eh_region_end): Remove all references to
3704 NOTE_INSN_CALL_ARG_LOCATION.
3705 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
3706 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
3707 * config/arc/arc.c (hwloop_optimize): Likewise.
3708 * config/arm/arm.c (create_fix_barrier): Likewise.
3709 * config/s390/s390.c (s390_chunkify_start): Likewise.
3710 * config/sh/sh.c (find_barrier): Likewise.
3711 * config/i386/i386.c (rest_of_insert_endbranch,
3712 ix86_seh_fixup_eh_fallthru): Likewise.
3713 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
3714 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
3715 * config/frv/frv.c (frv_function_prologue): Likewise.
3716 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
3718 (note_outside_basic_block_p): Remove all references to
3719 NOTE_INSN_CALL_ARG_LOCATION.
3720 * gengtype.c (adjust_field_rtx_def): Likewise.
3721 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
3723 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
3724 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
3727 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
3730 PR tree-optimization/84452
3731 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
3732 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
3735 2018-02-19 Martin Liska <mliska@suse.cz>
3738 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
3740 2018-02-19 Martin Liska <mliska@suse.cz>
3741 Richard Sandiford <richard.sandiford@linaro.org>
3743 PR tree-optimization/82491
3744 * gimple-fold.c (get_base_constructor): Make earlier bail out
3747 2018-02-19 Carl Love <cel@us.ibm.com>
3749 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
3750 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
3752 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
3753 P8V_BUILTIN_VEC_NEG.
3755 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
3757 * config/rl78/rl78.md (movdf): New define expand.
3759 2018-02-19 Martin Liska <mliska@suse.cz>
3762 * doc/invoke.texi: Fix typo.
3763 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
3765 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
3767 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3768 handle rs6000_single_float and rs6000_double_float specially for
3771 2018-02-16 Jeff Law <law@redhat.com>
3773 * config/rx/rx.c (add_pop_cfi_notes): New function.;
3776 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3779 * ipa-inline.c (inline_small_functions): Fix a typo.
3781 2018-02-16 Nathan Sidwell <nathan@acm.org>
3783 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
3785 2018-02-16 Carl Love <cel@us.ibm.com>
3787 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
3788 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
3789 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
3790 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
3791 expansion to P8V_BUILTIN_VEC_FLOAT2.
3793 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
3795 PR rtl-optimization/70023
3796 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
3797 src_regno into account.
3799 2018-02-16 Carl Love <cel@us.ibm.com>
3801 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
3802 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
3803 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
3804 * config/rs6000/rs6000.c: Remove case statements for
3805 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
3806 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
3807 and P9V_BUILTIN_VEC_VINSERT4B.
3808 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
3809 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
3810 * config/rs6000/vsx.md:
3811 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
3814 2018-02-16 Carl Love <cel@us.ibm.com>
3816 * config/rs6000/altivec.h: Add builtin names vec_extract4b
3818 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
3820 * config/rs6000/rs6000-c.c: Add the definitions for
3821 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
3822 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
3823 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
3824 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
3825 definition for insert4b and define insn *insert3b_internal.
3826 * doc/extend.texi: Add documentation for vec_extract4b.
3828 2018-02-16 Nathan Sidwell <nathan@acm.org>
3830 * doc/extend.texi (Backwards Compatibility): Mention friend
3831 injection. Note for-scope is deprecated.
3832 * doc/invoke.texi (-ffriend-injection): Deprecate.
3834 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
3836 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
3837 that moved to I2, also allow destinations that are a paradoxical
3838 subreg (instead of a normal reg).
3840 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
3843 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
3846 2018-02-16 Richard Biener <rguenther@suse.de>
3848 PR tree-optimization/84037
3849 PR tree-optimization/84016
3851 * config/i386/i386.c (ix86_builtin_vectorization_cost):
3852 Adjust vec_construct for the fact we need additional higher latency
3853 128bit inserts for AVX256 and AVX512 vector builds.
3854 (ix86_add_stmt_cost): Scale vector construction cost for
3857 2018-02-16 Richard Biener <rguenther@suse.de>
3859 PR tree-optimization/84417
3860 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
3861 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
3862 (non_rewritable_lvalue_p): Likewise, use poly-ints.
3864 2018-02-16 Martin Liska <mliska@suse.cz>
3867 * internal-fn.def (ASAN_CHECK): Set proper flags.
3868 (ASAN_MARK): Likewise.
3870 2018-02-16 Julia Koval <julia.koval@intel.com>
3872 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
3873 from PTA_CANNONLAKE.
3875 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3878 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
3879 Use ++iter rather than iter++ for std::list iterators.
3880 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
3881 defer deleting them until all nodes in the forest are processed. Do
3882 free even leaf nodes. Change to_process into auto_vec.
3885 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
3886 * vec.h (vec_default_construct): Use memset instead of placement new
3887 if BROKEN_VALUE_INITIALIZATION is defined.
3888 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
3889 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
3892 PR rtl-optimization/83723
3893 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
3894 * lra.c (lra_substitute_pseudo): Likewise. If true, use
3895 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
3897 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
3899 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
3901 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
3903 PR rtl-optimization/81443
3904 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
3905 from inner REGs to paradoxical SUBREGs.
3907 2018-02-16 Richard Biener <rguenther@suse.de>
3909 PR tree-optimization/84399
3910 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3911 For operands we can analyze at their definition make sure we can
3912 analyze them at each use as well.
3914 2018-02-16 Richard Biener <rguenther@suse.de>
3916 PR tree-optimization/84190
3917 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
3918 volatile accesses if the decl isn't volatile.
3920 2018-02-15 Jason Merrill <jason@redhat.com>
3922 PR c++/84314 - ICE with templates and fastcall attribute.
3923 * attribs.c (build_type_attribute_qual_variant): Don't clobber
3924 TYPE_CANONICAL on an existing type.
3926 2018-02-15 Jakub Jelinek <jakub@redhat.com>
3928 PR tree-optimization/84383
3929 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
3930 dstoff nor call operand_equal_p if dstbase is NULL.
3932 PR tree-optimization/84334
3933 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
3934 also a CONSTANT_CLASS_P, punt.
3936 2018-02-14 Jim Wilson <jimw@sifive.com>
3938 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
3939 first SMALL_OPERAND check. New local min_second_step. Move assert
3940 to where locals are set. Add TARGET_RVC support.
3941 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
3943 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
3945 * doc/invoke.texi: Correct -Wformat-overflow code sample.
3947 2018-02-14 Martin Sebor <msebor@redhat.com>
3949 PR tree-optimization/83698
3950 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
3951 arrays constrain the offset range to their bounds.
3952 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
3953 (builtin_access::overlap): Avoid setting the size of overlap if it's
3955 (maybe_diag_overlap): Also consider arrays when deciding what values
3956 of offsets to include in diagnostics.
3958 2018-02-14 Martin Sebor <msebor@redhat.com>
3961 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
3962 that correspond to the kind of a declaration.
3964 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
3967 * config/pa/pa.md: Load address of PIC label using the linkage table
3968 if the label is nonlocal.
3970 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
3972 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
3973 warning message if user requests -maltivec=be.
3974 * doc/invoke.texi: Document deprecation of -maltivec=be.
3976 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
3979 * config/rs6000/rs6000-c.c: Update definitions for
3980 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
3981 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
3983 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3986 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
3987 add _get_ssp intrinsics. Remove argument from
3988 __builtin_ia32_rdssp[d|q].
3989 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
3990 * config/i386/i386-builtin.def: Remove argument from
3991 __builtin_ia32_rdssp[d|q].
3992 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
3993 ix86_expand_special_args_builtin for _rdssp[d|q].
3994 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
3995 Clear register before usage.
3996 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
3997 Add documentation for new _get_ssp and _inc_ssp intrinsics.
3999 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
4001 PR tree-optimization/84357
4002 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4003 operand 1 of an ARRAY_REF too.
4005 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
4008 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4009 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4011 (set_of_reg): New struct.
4012 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4013 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4014 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4016 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4017 Split into bitclr, bitset, bitinvert patterns if appropriate.
4018 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4019 use rx_fuse_in_memory_bitop.
4020 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4021 to named insn, correct maximum insn length.
4023 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
4026 * machmode.def: Define a complex mode for PARTIAL_INT.
4027 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4029 * doc/rtl.texi: Document CSPImode.
4030 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4032 (msp430_hard_regno_nregs_with_padding): Likewise.
4034 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4037 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4039 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4041 PR rtl-optimization/84169
4042 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4043 we generated a parallel as new i3 and we split that to new i2 and i3
4044 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4045 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4046 those to i2, not i1. Partially rewrite this scan code.
4048 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4051 * stor-layout.c (place_field): For variable length fields, adjust
4052 offset_align afterwards not just based on the field's alignment,
4053 but also on the size.
4056 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4057 of exps and logs in the use_exp2 case.
4059 2018-02-13 Jeff Law <law@redhat.com>
4061 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4064 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4067 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4071 * common.opt (gas-loc-support, gas-locview-support): New.
4072 (ginline-points, ginternal-reset-location-views): New.
4073 * doc/invoke.texi: Document them. Use @itemx where intended.
4074 (gvariable-location-views): Adjust.
4075 * target.def (reset_location_view): New.
4076 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4077 (TARGET_RESET_LOCATION_VIEW): New.
4078 * doc/tm.texi: Rebuilt.
4079 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4080 (dwarf2out_default_as_locview_support): New.
4081 (output_asm_line_debug_info): Use option variables.
4082 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4083 (output_loc_list): Likewise.
4084 (add_high_low_attributes): Check option variables.
4085 Don't output entry view attribute in strict mode.
4086 (gen_inlined_subroutine_die): Check option variables.
4087 (dwarf2out_inline_entry): Likewise.
4088 (init_sections_and_labels): Likewise.
4089 (dwarf2out_early_finish): Likewise.
4090 (maybe_reset_location_view): New, from...
4091 (dwarf2out_var_location): ... here. Call it.
4092 * debug.h (dwarf2out_default_as_loc_support): Declare.
4093 (dwarf2out_default_as_locview_support): Declare.
4094 * hooks.c (hook_int_rtx_insn_0): New.
4095 * hooks.h (hook_int_rtx_insn_0): Declare.
4096 * toplev.c (process_options): Take -gas-loc-support and
4097 -gas-locview-support from dwarf2out. Enable
4098 -gvariable-location-views by default only with locview
4099 assembler support. Enable -ginternal-reset-location-views by
4100 default only if the target defines the corresponding hook.
4101 Enable -ginline-points by default if location views are
4102 enabled; force it disabled if statement frontiers are
4104 * tree-inline.c (expand_call_inline): Check option variables.
4105 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4107 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4109 PR tree-optimization/84321
4110 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4111 handling. Also check whether the anti-range contains any values
4112 that satisfy the mask; switch to a VR_RANGE if not.
4114 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4117 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4119 2018-02-13 Martin Jambor <mjambor@suse.cz>
4122 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4123 of call statements, also set location of a load to a temporary.
4125 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4127 * config/rl78/rl78.c (add_vector_labels): New function.
4128 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4129 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4130 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4131 which checks that no arguments are passed.
4132 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4133 * doc/extend.texi: Documentation for the new attribute.
4135 2018-02-13 Andreas Schwab <schwab@suse.de>
4137 * config/riscv/linux.h (CPP_SPEC): Define.
4139 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4142 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4143 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4144 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4145 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4146 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4147 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4148 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4149 temporarily for AES and PCLMUL builtins.
4151 PR tree-optimization/84339
4152 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4153 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4157 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4158 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4159 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4161 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4163 * omp-simd-clone.h: New file.
4164 * omp-simd-clone.c: Include omp-simd-clone.h.
4165 (expand_simd_clones): No longer static.
4166 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4167 cgraph.h and omp-simd-clone.h.
4168 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4169 (vect_recog_widen_shift_pattern): Formatting fix.
4170 (vect_pattern_recog_1): Don't check optab for calls.
4173 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4174 operands[2] into a REG before using gen_lowpart on it.
4176 2018-02-12 Jeff Law <law@redhat.com>
4179 * config/sh/sh.c (find_barrier): Consider a sibling call
4182 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4183 successfully back substituting a reg.
4185 2018-02-12 Richard Biener <rguenther@suse.de>
4187 PR tree-optimization/84037
4188 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4189 parameter, move visited init to caller.
4190 (vect_slp_analyze_operations): Separate cost from validity
4191 check, initialize visited once for all instances.
4192 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4194 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4196 (vect_model_promotion_demotion_cost): Likewise.
4197 (vectorizable_bswap): Guard cost modeling with !slp_node
4198 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4200 (vectorizable_call): Likewise.
4201 (vectorizable_conversion): Likewise.
4202 (vectorizable_assignment): Likewise.
4203 (vectorizable_shift): Likewise.
4204 (vectorizable_operation): Likewise.
4205 (vectorizable_store): Likewise.
4206 (vectorizable_load): Likewise.
4207 (vectorizable_condition): Likewise.
4208 (vectorizable_comparison): Likewise.
4210 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
4213 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
4214 (ASAN_MARK): Fix fnspec to account for return value, change pointer
4215 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
4217 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4220 * params.def (inline-min-speedup): Increase from 8 to 15.
4221 (max-inline-insns-auto): Decrease from 40 to 30.
4222 * ipa-split.c (consider_split): Add some buffer for function to
4223 be considered inlining candidate.
4224 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
4227 2018-02-12 Richard Biener <rguenther@suse.de>
4229 PR tree-optimization/84037
4230 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
4231 matched stmts if we cannot swap the non-matched ones.
4233 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4235 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
4236 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4237 _mm_maskz_scalef_round_ss): New intrinsics.
4238 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4239 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4240 __builtin_ia32_scalefss_round): Remove.
4241 (__builtin_ia32_scalefsd_mask_round,
4242 __builtin_ia32_scalefss_mask_round): New intrinsics.
4243 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4244 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4245 ((match_operand:VF_128 2 "<round_nimm_predicate>"
4246 "<round_constraint>")): Changed to ...
4247 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4248 "<round_scalar_constraint>")): ... this.
4249 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4250 %0, %1, %2<round_op3>}"): Changed to ...
4251 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4252 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4253 %2<round_scalar_mask_op3>}"): ... this.
4254 * config/i386/subst.md (round_scalar_nimm_predicate): New.
4256 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4258 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4259 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
4260 (_mm_maskz_sqrt_round_ss): New intrinsics.
4261 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
4262 (__builtin_ia32_sqrtsd_mask_round)
4263 (__builtin_ia32_sqrtss_mask_round): New builtins.
4264 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
4265 (__builtin_ia32_sqrtss_round): Remove.
4266 (__builtin_ia32_sqrtsd_mask_round)
4267 (__builtin_ia32_sqrtss_mask_round): New builtins.
4268 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4269 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4270 ((match_operand:VF_128 1 "vector_operand"
4271 "xBm,<round_constraint>")): Changed to ...
4272 ((match_operand:VF_128 1 "vector_operand"
4273 "xBm,<round_scalar_constraint>")): ... this.
4274 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4275 %0, %2, %<iptr>1<round_op3>}): Changed to ...
4276 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4277 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4278 %<iptr>1<round_scalar_mask_op3>}): ... this.
4279 ((set_attr "prefix" "<round_prefix>")): Changed to ...
4280 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4282 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
4285 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4286 Cast vec_cmpeq result to correct type.
4287 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4288 Cast vec_cmpgt result to correct type.
4290 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4292 * final.c (final_scan_insn_1): Renamed from...
4293 (final_scan_insn): ... this. New wrapper, to recover
4294 seen from the outermost call in recursive ones.
4295 * config/sparc/sparc.c (output_return): Drop seen from call.
4296 (output_sibcall): Likewise.
4297 * config/visium/visium.c (output_branch): Likewise.
4299 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4301 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4304 2018-02-10 Alan Modra <amodra@gmail.com>
4307 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4308 Specify LR as an input.
4310 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4313 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4314 remove_member_access_dummy_vars): New functions.
4315 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4316 lower_omp_1, execute_lower_omp): Use them.
4318 PR rtl-optimization/84308
4319 * shrink-wrap.c (spread_components): Release todo vector.
4321 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4323 PR rtl-optimization/57193
4324 * ira-color.c (struct allocno_color_data): Add member
4325 conflict_allocno_hard_prefs.
4326 (update_conflict_allocno_hard_prefs): New.
4327 (bucket_allocno_compare_func): Add a preference based on
4328 conflict_allocno_hard_prefs.
4329 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4330 (color_allocnos): Remove a dead code. Initiate
4331 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4333 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4336 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4337 constraint from =wa to wa. Avoid a subreg on the output operand,
4338 instead use a pseudo and subreg it in a move.
4339 (p9_xxbrd_<mode>): Changed to ...
4340 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4341 (p9_xxbrd_v2df): New expander.
4342 (p9_xxbrw_<mode>): Changed to ...
4343 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4344 (p9_xxbrw_v4sf): New expander.
4346 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4348 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4350 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
4353 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4354 multiply in 32-bit mode.
4355 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4356 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4359 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4361 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4362 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4363 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4364 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4366 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4369 * dwarf2out.c (is_trivial_indirect_ref): New function.
4370 (dwarf2out_late_global_decl): Do not generate a location
4371 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4372 and that are not defined in the current unit.
4374 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4376 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4377 instead of a libcall for UNORDERED.
4379 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4382 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4383 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4385 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4388 * config/s390/s390.c (s390_set_current_function): Invoke
4389 s390_indirect_branch_settings also if fndecl didn't change.
4391 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4393 * config/rs6000/rs6000.md (blockage): Set length to zero.
4395 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4397 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4399 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4402 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4403 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4407 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4408 PARALLEL incoming that failed vt_get_decl_and_offset check.
4411 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
4412 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
4414 (get_variable_section): For decls in named .bss* sections pass true as
4415 second argument to bss_initializer_p.
4417 2018-02-09 Marek Polacek <polacek@redhat.com>
4418 Jakub Jelinek <jakub@redhat.com>
4421 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
4422 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
4423 Sync some changes from cxx_fold_indirect_ref.
4425 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4427 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
4429 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
4430 (BLOCK_INLINE_ENTRY_LABEL): New.
4431 (dwarf2out_var_location): Disregard inline entry markers.
4432 (inline_entry_data): New struct.
4433 (inline_entry_data_hasher): New hashtable type.
4434 (inline_entry_data_hasher::hash): New.
4435 (inline_entry_data_hasher::equal): New.
4436 (inline_entry_data_table): New variable.
4437 (add_high_low_attributes): Add DW_AT_entry_pc and
4438 DW_AT_GNU_entry_view attributes if a pending entry is found
4439 in inline_entry_data_table. Add old entry_pc attribute only
4440 if debug nonbinding markers are disabled.
4441 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
4442 markers are enabled.
4443 (block_within_block_p, dwarf2out_inline_entry): New.
4444 (dwarf2out_finish): Check that no entries remained in
4445 inline_entry_data_table.
4446 * final.c (reemit_insn_block_notes): Handle inline entry notes.
4447 (final_scan_insn, notice_source_line): Likewise.
4448 (rest_of_clean_state): Skip inline entry markers.
4449 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
4451 * gimple.c (gimple_build_debug_inline_entry): New.
4452 * gimple.h (enum gimple_debug_subcode): Add
4453 GIMPLE_DEBUG_INLINE_ENTRY.
4454 (gimple_build_debug_inline_entry): Declare.
4455 (gimple_debug_inline_entry_p): New.
4456 (gimple_debug_nonbind_marker_p): Adjust.
4457 * insn-notes.def (INLINE_ENTRY): New.
4458 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4459 inline entry marker notes.
4460 (print_insn): Likewise.
4461 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
4462 (INSN_DEBUG_MARKER_KIND): Likewise.
4463 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
4464 * tree-inline.c (expand_call_inline): Build and insert
4465 debug_inline_entry stmt.
4466 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
4467 inline entry blocks early, if nonbind markers are enabled.
4468 (dump_scope_block): Dump fragment info.
4469 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
4470 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
4471 (gimple_build_debug_inline_entry): New.
4472 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
4473 Enable/disable inline entry points too.
4474 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
4475 (DEBUG_INSN): Describe inline entry markers.
4477 * common.opt (gvariable-location-views): New.
4478 (gvariable-location-views=incompat5): New.
4479 * config.in: Rebuilt.
4480 * configure: Rebuilt.
4481 * configure.ac: Test assembler for view support.
4482 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
4483 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
4484 * dwarf2out.c (var_loc_view): New typedef.
4485 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
4486 (dwarf2out_locviews_in_attribute): New.
4487 (dwarf2out_locviews_in_loclist): New.
4488 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
4489 (enum dw_line_info_opcode): Add LI_adv_address.
4490 (struct dw_line_info_table): Add view.
4491 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
4492 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
4493 (zero_view_p): New variable.
4494 (ZERO_VIEW_P): New macro.
4495 (output_asm_line_debug_info): New.
4496 (struct var_loc_node): Add view.
4497 (add_AT_view_list, AT_loc_list): New.
4498 (add_var_loc_to_decl): Add view param. Test it against last.
4499 (new_loc_list): Add view params. Record them.
4500 (AT_loc_list_ptr): Handle loc and view lists.
4501 (view_list_to_loc_list_val_node): New.
4502 (print_dw_val): Handle dw_val_class_view_list.
4503 (size_of_die): Likewise.
4504 (value_format): Likewise.
4505 (loc_list_has_views): New.
4506 (gen_llsym): Set vl_symbol too.
4507 (maybe_gen_llsym, skip_loc_list_entry): New.
4508 (dwarf2out_maybe_output_loclist_view_pair): New.
4509 (output_loc_list): Output view list or entries too.
4510 (output_view_list_offset): New.
4511 (output_die): Handle dw_val_class_view_list.
4512 (output_dwarf_version): New.
4513 (output_compilation_unit_header): Use it.
4514 (output_skeleton_debug_sections): Likewise.
4515 (output_rnglists, output_line_info): Likewise.
4516 (output_pubnames, output_aranges): Update version comments.
4517 (output_one_line_info_table): Output view numbers in asm comments.
4518 (dw_loc_list): Determine current endview, pass it to new_loc_list.
4519 Call maybe_gen_llsym.
4520 (loc_list_from_tree_1): Adjust.
4521 (add_AT_location_description): Create view list attribute if
4522 needed, check it's absent otherwise.
4523 (convert_cfa_to_fb_loc_list): Adjust.
4524 (maybe_emit_file): Call output_asm_line_debug_info for test.
4525 (dwarf2out_var_location): Reset views as needed. Precompute
4526 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
4527 attribute. Set view.
4528 (new_line_info_table): Reset next view.
4529 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
4530 (dwarf2out_source_line): Likewise. Output view resets and labels to
4531 the assembler, or select appropriate line info opcodes.
4532 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
4533 (optimize_string_length): Catch it. Adjust.
4534 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
4535 dw_val_class_view_list, and remove it if no longer needed.
4536 (hash_loc_list): Hash view numbers.
4537 (loc_list_hasher::equal): Compare them.
4538 (optimize_location_lists): Check whether a view list symbol is
4539 needed, and whether the locview attribute is present, and
4540 whether they match. Remove the locview attribute if no longer
4542 (index_location_lists): Call skip_loc_list_entry for test.
4543 (dwarf2out_finish): Call output_asm_line_debug_info for test.
4544 Use output_dwarf_version.
4545 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
4546 (struct dw_val_node): Add val_view_list.
4547 * final.c (SEEN_NEXT_VIEW): New.
4548 (set_next_view_needed): New.
4549 (clear_next_view_needed): New.
4550 (maybe_output_next_view): New.
4551 (final_start_function): Rename to...
4552 (final_start_function_1): ... this. Take pointer to FIRST,
4553 add SEEN parameter. Emit param bindings in the initial view.
4554 (final_start_function): Reintroduce SEEN-less interface.
4555 (final): Rename to...
4556 (final_1): ... this. Take SEEN parameter. Output final pending
4557 next view at the end.
4558 (final): Reintroduce seen-less interface.
4559 (final_scan_insn): Output pending next view before switching
4560 sections or ending a block. Mark the next view as needed when
4561 outputting variable locations. Notify debug backend of section
4562 changes, and of location view changes.
4563 (rest_of_handle_final): Adjust.
4564 * toplev.c (process_options): Autodetect value for debug variable
4565 location views option. Warn on incompat5 without -gdwarf-5.
4566 * doc/invoke.texi (gvariable-location-views): New.
4567 (gvariable-location-views=incompat5): New.
4568 (gno-variable-location-views): New.
4570 2018-02-08 David Malcolm <dmalcolm@redhat.com>
4572 PR tree-optimization/84136
4573 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
4574 that the result of find_edge is non-NULL.
4576 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4579 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
4580 storing integer register in SImode. Fix cost of 256 and 512
4581 byte aligned SSE register store.
4583 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4585 * config/i386/i386.c (ix86_multiplication_cost): Fix
4586 multiplication cost for TARGET_AVX512DQ.
4588 2018-02-08 Marek Polacek <polacek@redhat.com>
4590 PR tree-optimization/84238
4591 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
4594 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4596 PR tree-optimization/84265
4597 * tree-vect-stmts.c (vectorizable_store): Don't treat
4598 VMAT_CONTIGUOUS accesses as grouped.
4599 (vectorizable_load): Likewise.
4601 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4603 PR tree-optimization/81635
4604 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
4605 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
4606 (test_round_for_mask): New functions.
4607 (wide_int_cc_tests): Call test_round_for_mask.
4608 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
4609 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
4610 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
4611 range returned by get_range_info.
4613 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4616 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
4617 * symtab.c: Include builtins.h
4618 (symtab_node::output_to_lto_symbol_table_p): Move here
4619 from lto-streamer-out.c:output_symbol_p.
4620 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
4621 (output_symbol_p): Move all logic to symtab.c
4622 (produce_symtab): Update.
4624 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4626 * config/s390/s390-opts.h (enum indirect_branch): Define.
4627 * config/s390/s390-protos.h (s390_return_addr_from_memory)
4628 (s390_indirect_branch_via_thunk)
4629 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
4630 (enum s390_indirect_branch_type): Define.
4631 * config/s390/s390.c (struct s390_frame_layout, struct
4632 machine_function): Remove.
4633 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
4634 (indirect_branch_table_label_no, indirect_branch_table_name):
4636 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
4637 (enum s390_indirect_branch_option): Define.
4638 (s390_return_addr_from_memory): New function.
4639 (s390_handle_string_attribute): New function.
4640 (s390_attribute_table): Add new attribute handler.
4641 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
4642 (s390_indirect_branch_via_thunk): New function.
4643 (s390_indirect_branch_via_inline_thunk): New function.
4644 (s390_function_ok_for_sibcall): When jumping via thunk disallow
4645 sibling call optimization for non z10 compiles.
4646 (s390_emit_call): Force indirect branch target to be a single
4647 register. Add r1 clobber for non-z10 compiles.
4648 (s390_emit_epilogue): Emit return jump via return_use expander.
4649 (s390_reorg): Handle JUMP_INSNs as execute targets.
4650 (s390_option_override_internal): Perform validity checks for the
4651 new command line options.
4652 (s390_indirect_branch_attrvalue): New function.
4653 (s390_indirect_branch_settings): New function.
4654 (s390_set_current_function): Invoke s390_indirect_branch_settings.
4655 (s390_output_indirect_thunk_function): New function.
4656 (s390_code_end): Implement target hook.
4657 (s390_case_values_threshold): Implement target hook.
4658 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
4660 * config/s390/s390.h (struct s390_frame_layout)
4661 (struct machine_function): Move here from s390.c.
4662 (TARGET_INDIRECT_BRANCH_NOBP_RET)
4663 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
4664 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
4665 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
4666 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
4667 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
4668 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
4669 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
4670 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
4671 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
4672 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
4673 (mnemonic attribute): Add values which aren't recognized
4675 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
4676 pattern for branch conversion. Fix mnemonic attribute.
4677 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
4678 indirect branch via thunk if requested.
4679 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
4680 ("*indirect_jump"): Disable for branch conversion using out of
4682 ("indirect_jump_via_thunk<mode>_z10")
4683 ("indirect_jump_via_thunk<mode>")
4684 ("indirect_jump_via_inlinethunk<mode>_z10")
4685 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
4686 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
4687 ("casesi_jump_via_inlinethunk<mode>_z10")
4688 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
4689 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
4690 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
4691 ("*indirect2_jump"): Disable for branch conversion.
4692 ("casesi_jump"): Turn into expander and expand patterns for branch
4694 ("return_use"): New expander.
4695 ("*return"): Emit return via thunk and rename it to ...
4696 ("*return<mode>"): ... this one.
4697 * config/s390/s390.opt: Add new options and and enum for the
4700 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4702 * lra-constraints.c (match_reload): Unconditionally use
4703 gen_lowpart_SUBREG, rather than selecting between that
4704 and equivalent gen_rtx_SUBREG code.
4706 2018-02-08 Richard Biener <rguenther@suse.de>
4708 PR tree-optimization/84233
4709 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
4710 changed flag instead of boguously re-using phi_inserted.
4712 2018-02-08 Martin Jambor <mjambor@suse.cz>
4714 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
4715 static local variables.
4717 2018-02-08 Richard Biener <rguenther@suse.de>
4719 PR tree-optimization/84278
4720 * tree-vect-stmts.c (vectorizable_store): When looking for
4721 smaller vector types to perform grouped strided loads/stores
4722 make sure the mode is supported by the target.
4723 (vectorizable_load): Likewise.
4725 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4727 * config/aarch64/aarch64.c (aarch64_components_for_bb):
4728 Increase LDP/STP opportunities by adding adjacent callee-saves.
4730 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4732 PR rtl-optimization/84068
4733 PR rtl-optimization/83459
4734 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
4736 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
4738 PR tree-optimization/84224
4739 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
4740 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
4743 2018-02-07 Iain Sandoe <iain@codesourcery.com>
4746 * config/rs6000/altivec.md (*restore_world): Remove LR use.
4747 * config/rs6000/predicates.md (restore_world_operation): Adjust op
4748 count, remove one USE.
4750 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
4752 * doc/install.texi (Configuration): Document the
4753 --with-long-double-format={ibm,ieee} PowerPC configuration
4757 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
4758 Convert from define_expand to be define_insn_and_split. Rework
4759 float/double/_Float128 conversions to QI/HI/SImode to work with
4760 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
4761 conversions to QI/HImode types did a store and then a load to
4762 truncate the value. For conversions to VSX registers, don't split
4763 the insn, instead emit the code directly. Use the code iterator
4764 any_fix to combine signed and unsigned conversions.
4765 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
4766 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
4767 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4768 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
4769 (fix_<mode>di2_hw): Likewise.
4770 (fixuns_<mode>di2_hw): Likewise.
4771 (fix_<mode>si2_hw): Likewise.
4772 (fixuns_<mode>si2_hw): Likewise.
4773 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
4774 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4775 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
4776 fix<uns>_trunc<SFDF:mode>si2_p8.
4777 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
4779 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
4780 (fix<uns>_<mode>_mem): Likewise.
4781 (fctiw<u>z_<mode>_mem): Likewise.
4782 (fix<uns>_<mode>_mem): Likewise.
4783 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
4784 the register allocator from doing a direct move to the GPRs to do
4785 a store, and instead use the ISA 3.0 store byte/half-word from
4786 vector register instruction. For IEEE 128-bit floating point,
4787 also optimize stores of 32-bit ints.
4788 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
4790 2018-02-07 Alan Hayward <alan.hayward@arm.com>
4792 * genextract.c (push_pathstr_operand): New function to support
4794 (walk_rtx): Call push_pathstr_operand.
4795 (print_path): Support [a-zA-Z].
4797 2018-02-07 Richard Biener <rguenther@suse.de>
4799 PR tree-optimization/84037
4800 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
4801 (cse_and_gimplify_to_preheader): Declare.
4802 (vect_get_place_in_interleaving_chain): Likewise.
4803 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4805 (_loop_vec_info::~_loop_vec_info): Delete it.
4806 (cse_and_gimplify_to_preheader): New function.
4807 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
4808 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
4809 (vectorizable_load): Likewise. For grouped stores always base
4810 the IV on the first element.
4811 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
4812 condition before gimplifying.
4814 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4816 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
4817 *DIV_EXPR and *MOD_EXPR.
4819 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
4822 * config/i386/i386.c (ix86_option_override_internal): Mask out
4823 the CF_SET bit when checking -fcf-protection.
4825 2018-02-07 Tom de Vries <tom@codesourcery.com>
4828 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
4831 2018-02-07 Richard Biener <rguenther@suse.de>
4833 PR tree-optimization/84204
4834 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4837 PR tree-optimization/84205
4838 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
4839 special-case isl_ast_op_zdiv_r.
4841 PR tree-optimization/84223
4842 * graphite-scop-detection.c (gather_bbs::before_dom_children):
4843 Only add conditions from within the region.
4844 (gather_bbs::after_dom_children): Adjust.
4846 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
4849 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
4850 * config/avr/avr.md: Only post-reload split REG-REG moves if
4851 either register is GENERAL_REG_P.
4853 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4855 PR tree-optimization/84235
4856 * tree-ssa-scopedtables.c
4857 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
4858 if the subtraction is performed in floating point type where NaNs are
4859 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
4860 build 1. Formatting fix.
4862 2018-02-06 Jakub Jelinek <jakub@redhat.com>
4865 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
4866 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
4867 and skip it regardless of bb boundaries. Use CALL_P macro,
4868 don't test INSN_P (insn) together with CALL_P or JUMP_P check
4869 unnecessarily, formatting fix.
4871 2018-02-06 Michael Collison <michael.collison@arm.com>
4873 * config/arm/thumb2.md:
4874 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
4875 (*thumb_mov_notscc): Ditto.
4877 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
4880 * config/rs6000/rs6000.md (su code attribute): Use "u" for
4881 unsigned_fix, not "s".
4883 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4885 * configure.ac (gcc_fn_eh_frame_ro): New function.
4886 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
4887 correct .eh_frame permissions.
4888 * configure: Regenerate.
4890 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
4892 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
4895 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4897 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4898 Display warning message for -mno-speculate-indirect-jumps.
4900 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
4902 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
4904 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
4906 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
4908 PR tree-optimization/84225
4909 * tree-eh.c (find_trapping_overflow): Only call
4910 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
4912 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4915 * config/i386/i386.c: Reimplement the check of possible options
4916 -mibt/-mshstk conbination. Change error messages.
4917 * doc/invoke.texi: Fix a typo: remove extra '='.
4919 2018-02-06 Marek Polacek <polacek@redhat.com>
4921 PR tree-optimization/84228
4922 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
4924 2018-02-06 Tamar Christina <tamar.christina@arm.com>
4927 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
4928 emitted arch directives.
4929 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
4930 __ARM_FEATURE_COPROC before changing architectures.
4932 2018-02-06 Richard Biener <rguenther@suse.de>
4934 * config/i386/i386.c (print_reg): Fix typo.
4935 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
4937 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
4939 * configure: Regenerate.
4941 2018-02-05 Martin Sebor <msebor@redhat.com>
4943 PR tree-optimization/83369
4944 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
4947 2018-02-05 Martin Liska <mliska@suse.cz>
4949 * doc/invoke.texi: Cherry-pick upstream r323995.
4951 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
4953 * ira.c (ira_init_register_move_cost): Adjust comment.
4955 2018-02-05 Martin Liska <mliska@suse.cz>
4957 PR gcov-profile/84137
4958 * doc/gcov.texi: Fix typo in documentation.
4960 2018-02-05 Martin Liska <mliska@suse.cz>
4962 PR gcov-profile/83879
4963 * doc/gcov.texi: Document necessity of --dynamic-list-data when
4964 using dlopen functionality.
4966 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
4968 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
4969 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
4970 _mm_maskz_range_ss, _mm_mask_range_round_ss,
4971 _mm_maskz_range_round_ss): New intrinsics.
4972 (__builtin_ia32_rangesd128_round)
4973 (__builtin_ia32_rangess128_round): Remove.
4974 (__builtin_ia32_rangesd128_mask_round,
4975 __builtin_ia32_rangess128_mask_round): New builtins.
4976 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
4977 __builtin_ia32_rangess128_round): Remove.
4978 (__builtin_ia32_rangesd128_mask_round,
4979 __builtin_ia32_rangess128_mask_round): New builtins.
4980 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
4981 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
4982 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
4983 "<round_saeonly_constraint>")): Changed to ...
4984 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
4985 "<round_saeonly_scalar_constraint>")): ... this.
4986 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
4987 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
4988 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
4989 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
4990 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
4992 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
4994 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
4996 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
4997 Remove all values except native, 8540 and 8548.
4999 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
5001 * config/i386/i386.c (ix86_output_function_return): Pass
5002 INVALID_REGNUM, instead of -1, as invalid register number to
5003 indirect_thunk_name and output_indirect_thunk.
5005 2018-02-02 Julia Koval <julia.koval@intel.com>
5007 * config.gcc: Add -march=icelake.
5008 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5009 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5010 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5011 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5012 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5013 (processor_target_table): Add icelake.
5014 (ix86_option_override_internal): Handle new PTAs.
5015 (get_builtin_code_for_version): Handle icelake.
5016 (M_INTEL_COREI7_ICELAKE): New.
5017 (fold_builtin_cpu): Handle icelake.
5018 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5019 * doc/invoke.texi: Add -march=icelake.
5021 2018-02-02 Julia Koval <julia.koval@intel.com>
5023 * config/i386/i386.c (ix86_option_override_internal): Change flags type
5024 to wide_int_bitmask.
5025 * wide-int-bitmask.h: New.
5027 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5030 * config/i386/i386.md: Replace Pmode with word_mode in
5031 builtin_setjmp_setup and builtin_longjmp to support x32.
5033 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5037 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5039 (rs6000_supported_cpu_names): New static variable.
5040 (linux_cpu_translation_table): Likewise.
5041 (elf_platform) <cpu>: Define new static variable and use it.
5042 Translate kernel AT_PLATFORM name to canonical name if needed.
5043 Error if platform name is unknown.
5045 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
5048 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5050 2018-02-01 Jeff Law <law@redhat.com>
5053 * config/i386/i386.c (release_scratch_register_on_entry): Add new
5054 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
5055 the scratch if RELEASE_VIA_POP is false.
5056 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5057 If we have to save a temporary register, decrement SIZE appropriately.
5058 Pass new arguments to release_scratch_register_on_entry.
5059 (ix86_adjust_stack_and_probe): Likewise.
5060 (ix86_emit_probe_stack_range): Pass new arguments to
5061 release_scratch_register_on_entry.
5063 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5065 PR rtl-optimization/84157
5066 * combine.c (change_zero_ext): Use REG_P predicate in
5067 front of HARD_REGISTER_P predicate.
5069 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5071 * config/avr/avr.c (avr_option_override): Move disabling of
5072 -fdelete-null-pointer-checks to...
5073 * common/config/avr/avr-common.c (avr_option_optimization_table):
5076 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5078 PR tree-optimization/81635
5079 * tree-data-ref.c (split_constant_offset_1): For types that
5080 wrap on overflow, try to use range info to prove that wrapping
5083 2018-02-01 Renlin Li <renlin.li@arm.com>
5086 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5088 (aarch64_register_move_cost): Likewise.
5089 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5091 (REG_CLASS_NAMES): Likewise.
5092 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5093 TAILCALL_ADDR_REGS. Remove IP registers.
5094 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5096 2018-02-01 Richard Biener <rguenther@suse.de>
5098 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5099 for specifying RPO order and allow NULL for that.
5100 * domwalk.c (dom_walker::dom_walker): Likewise.
5101 (dom_walker::walk): Handle NULL RPO order.
5102 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5104 (rewrite_update_dom_walker): Likewise.
5105 (mark_def_dom_walker): Likewise.
5107 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5109 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5110 (aarch64_maybe_expand_sve_subreg_move): Declare.
5111 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5112 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5114 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5115 that are semantically a reverse operation.
5116 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5117 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5118 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5120 (aarch64_can_change_mode_class): For big-endian, forbid changes
5121 between two SVE modes if they have different element sizes.
5123 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5125 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5126 the TImode handling for big-endian targets.
5128 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5130 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5131 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5133 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5134 Remove BSWAP handing for big-endian targets and use the form of
5135 LD1RQ appropariate for the mode.
5137 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5139 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5140 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5143 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5146 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5147 check for operands that need to go through aarch64_sve_reload_be.
5149 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5151 PR tree-optimization/81661
5152 PR tree-optimization/84117
5153 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5154 * tree-eh.c: Include gimplify.h.
5155 (find_trapping_overflow, replace_trapping_overflow,
5156 rewrite_to_non_trapping_overflow): New functions.
5157 * tree-vect-loop.c: Include tree-eh.h.
5158 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5159 * tree-data-ref.c: Include tree-eh.h.
5160 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5162 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
5164 PR rtl-optimization/84123
5165 * combine.c (change_zero_ext): Check if hard register satisfies
5166 can_change_dest_mode before calling gen_lowpart_SUBREG.
5168 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
5171 * ira.c (ira_init_register_move_cost): Remove assert.
5173 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5175 PR rtl-optimization/84071
5176 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5177 * doc/tm.texi: Regenerate.
5179 2018-01-31 Richard Biener <rguenther@suse.de>
5181 PR tree-optimization/84132
5182 * tree-data-ref.c (analyze_miv_subscript): Properly
5183 check whether evolution_function_is_affine_multivariate_p
5184 before calling gcd_of_steps_may_divide_p.
5186 2018-01-31 Julia Koval <julia.koval@intel.com>
5189 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5190 * config/i386/i386.md (rdpid_rex64) New.
5191 (rdpid): Make 32bit only.
5193 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5196 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5197 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5199 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5202 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5204 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5206 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5208 PR rtl-optimization/84071
5209 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
5210 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
5212 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5214 * config/arc/arc.c (arc_handle_aux_attribute): New function.
5215 (arc_attribute_table): Add 'aux' attribute.
5216 (arc_in_small_data_p): Consider aux like variables.
5217 (arc_is_aux_reg_p): New function.
5218 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
5219 (arc_get_aux_arg): New function.
5220 (prepare_move_operands): Handle aux-register access.
5221 (arc_handle_aux_attribute): New function.
5222 * doc/extend.texi (ARC Variable attributes): Add subsection.
5224 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5226 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
5227 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
5228 (arc_attribute_table): Add 'uncached' attribute.
5229 (arc_print_operand): Print '.di' flag for uncached memory
5231 (arc_in_small_data_p): Do not consider for small data the uncached
5233 (arc_is_uncached_mem_p): New function.
5234 * config/arc/predicates.md (compact_store_memory_operand): Check
5235 for uncached memory accesses.
5236 (nonvol_nonimm_operand): Likewise.
5237 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5239 2018-01-31 Jakub Jelinek <jakub@redhat.com>
5242 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5243 falign-loops=): Add Optimization flag.
5245 2018-01-30 Jeff Law <law@redhat.com>
5248 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5249 INT_REGISTERS_SAVED. Check it prior to calling
5250 get_scratch_register_on_entry.
5251 (ix86_adjust_stack_and_probe): Similarly.
5252 (ix86_emit_probe_stack_range): Similarly.
5253 (ix86_expand_prologue): Corresponding changes.
5255 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5258 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5259 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
5261 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
5264 * lra-constraints.c (curr_insn_transform): Process AND in the
5267 2018-01-30 Jakub Jelinek <jakub@redhat.com>
5269 PR rtl-optimization/83986
5270 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5271 dependence against last_pending_memory_flush in addition to
5274 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
5276 PR tree-optimization/81611
5277 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5280 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5283 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5286 2018-01-30 Richard Biener <rguenther@suse.de>
5287 Jakub Jelinek <jakub@redhat.com>
5289 PR tree-optimization/84111
5290 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5291 inner loops added during recursion, as they don't have up-to-date
5294 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5297 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5298 (can_inline_edge_by_limits_p): ... here.
5299 (can_early_inline_edge_p, check_callers,
5300 update_caller_keys, update_callee_keys, recursive_inlining,
5301 add_new_edges_to_heap, speculation_useful_p,
5302 inline_small_functions,
5303 inline_small_functions, flatten_function,
5304 inline_to_all_callers_1): Update.
5306 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5308 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5311 2018-01-30 Richard Biener <rguenther@suse.de>
5313 PR tree-optimization/83008
5314 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5315 invariant and constant vector uses in stmts when they need
5318 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5321 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5322 * configure: Regenerate.
5324 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5326 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5328 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5329 Use gen_rtx_REG rather than gen_lowpart.
5331 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5333 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5334 rather than 0 when creating partial subregs.
5336 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5338 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5341 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
5344 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5345 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5346 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5347 flags. This restores the settings used before the 2017-07-24.
5348 Turning off pre increment/decrement/modify allows IVOPTS to
5349 optimize DF/SF loops where the index is an int.
5351 2018-01-29 Richard Biener <rguenther@suse.de>
5352 Kelvin Nilsen <kelvin@gcc.gnu.org>
5355 * tree-vect-stmts.c (vectorizable_call): Don't call
5356 targetm.vectorize_builtin_md_vectorized_function if callee is
5359 2018-01-22 Carl Love <cel@us.ibm.com>
5361 * doc/extend.tex: Fix typo in second arg in
5362 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5364 2018-01-29 Richard Biener <rguenther@suse.de>
5366 PR tree-optimization/84086
5367 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5368 (flush_ssaname_freelist): When SSA names were released reset
5369 the SCEV hash table.
5371 2018-01-29 Richard Biener <rguenther@suse.de>
5373 PR tree-optimization/84057
5374 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5375 removed paths when removing edges.
5377 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
5379 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5380 -mfunction-return=@var{choice}.
5382 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
5385 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5387 (layout::print_source_line): Likewise.
5388 (test_get_line_width_without_trailing_whitespace): Add test cases.
5390 2018-01-27 Jakub Jelinek <jakub@redhat.com>
5393 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5396 2018-01-26 Jim Wilson <jimw@sifive.com>
5398 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5400 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5403 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5405 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5406 and CMP + SUB-immediate -> SUBS.
5408 2018-01-26 Martin Sebor <msebor@redhat.com>
5410 PR tree-optimization/83896
5411 * tree-ssa-strlen.c (get_string_len): Rename...
5412 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
5413 Avoid assuming length is constant.
5414 (handle_char_store): Use HOST_WIDE_INT for string length.
5416 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
5419 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
5420 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
5422 2018-01-26 Richard Biener <rguenther@suse.de>
5424 PR rtl-optimization/84003
5425 * dse.c (record_store): Only record redundant stores when
5426 the earlier store aliases at least all accesses the later one does.
5428 2018-01-26 Jakub Jelinek <jakub@redhat.com>
5430 PR rtl-optimization/83985
5431 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
5432 REG_CFA_RESTORE insns.
5433 (delete_unmarked_insns): Don't ignore separate shrink wrapping
5434 REG_CFA_RESTORE insns here.
5437 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
5438 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
5440 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5442 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
5443 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
5444 (arc_init): Likewise.
5445 (arc_override_options): Likewise.
5446 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
5448 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
5450 * config/arc/arc.h (TARGET_DBNZ): Define.
5451 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
5452 properly set the tune attribute.
5453 (dbnz): Use TARGET_DBNZ guard.
5454 * config/arc/arc.opt (mtune): Add core3 option.
5456 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5458 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
5459 recognize new pic like addresses.
5460 (arc_delegitimize_address): Clean up.
5462 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5464 * config/arc/arc-arches.def: Option mrf16 valid for all
5466 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
5467 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
5468 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
5469 * config/arc/arc-tables.opt: Regenerate.
5470 * config/arc/arc.c (arc_conditional_register_usage): Handle
5471 reduced register file case.
5472 (arc_file_start): Set must have build attributes.
5473 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
5475 * config/arc/arc.opt (mrf16): Add new option.
5476 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
5477 * config/arc/genmultilib.awk: Handle new mrf16 option.
5478 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
5479 * config/arc/t-multilib: Regenerate.
5480 * doc/invoke.texi (ARC Options): Document mrf16 option.
5482 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5484 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
5485 * config/arc/arc.c (arc_handle_secure_attribute): New function.
5486 (arc_attribute_table): Add 'secure_call' attribute.
5487 (arc_print_operand): Print secure call operand.
5488 (arc_function_ok_for_sibcall): Don't optimize tail calls when
5490 (arc_is_secure_call_p): New function. * config/arc/arc.md
5491 (call_i): Add support for sjli instruction.
5492 (call_value_i): Likewise.
5493 * config/arc/constraints.md (Csc): New constraint.
5495 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5496 John Eric Martin <John.Martin@emmicro-us.com>
5498 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
5499 * config/arc/arc.c (_arc_jli_section): New struct.
5500 (arc_jli_section): New type.
5501 (rc_jli_sections): New static variable.
5502 (arc_handle_jli_attribute): New function.
5503 (arc_attribute_table): Add jli_always and jli_fixed attribute.
5504 (arc_file_end): New function.
5505 (TARGET_ASM_FILE_END): Define.
5506 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
5507 (arc_add_jli_section): New function.
5508 (jli_call_scan): Likewise.
5509 (arc_reorg): Call jli_call_scan.
5510 (arc_output_addsi): Remove 'S' from printing asm operand.
5511 (arc_is_jli_call_p): New function.
5512 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
5514 (movhi_insn): Likewise.
5515 (movsi_insn): Likewise.
5516 (movsi_set_cc_insn): Likewise.
5517 (loadqi_update): Likewise.
5518 (load_zeroextendqisi_update): Likewise.
5519 (load_signextendqisi_update): Likewise.
5520 (loadhi_update): Likewise.
5521 (load_zeroextendhisi_update): Likewise.
5522 (load_signextendhisi_update): Likewise.
5523 (loadsi_update): Likewise.
5524 (loadsf_update): Likewise.
5525 (movsicc_insn): Likewise.
5526 (bset_insn): Likewise.
5527 (bxor_insn): Likewise.
5528 (bclr_insn): Likewise.
5529 (bmsk_insn): Likewise.
5530 (bicsi3_insn): Likewise.
5531 (cmpsi_cc_c_insn): Likewise.
5532 (movsi_ne): Likewise.
5533 (movsi_cond_exec): Likewise.
5534 (clrsbsi2): Likewise.
5545 (call_i): Remove 'S' asm letter, add jli instruction.
5546 (call_value_i): Likewise.
5547 * config/arc/arc.op (mjli-always): New option.
5548 * config/arc/constraints.md (Cji): New constraint.
5549 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
5551 (subsf3_fpx): Likewise.
5552 (mulsf3_fpx): Likewise.
5553 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
5555 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
5557 * doc/invoke.texi (ARC): Document mjli-always option.
5559 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
5561 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
5562 avoid addition with 0 and use incw and decw where possible.
5564 2018-01-26 Richard Biener <rguenther@suse.de>
5566 PR tree-optimization/81082
5567 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
5568 association if it requires casting to unsigned.
5569 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
5570 from fold_plusminus_mult_expr to catch important cases late when
5571 range info is available.
5573 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5575 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
5576 * configure.ac (hidden_linkonce): New test.
5577 * configure: Regenerate.
5578 * config.in: Regenerate.
5580 2018-01-26 Julia Koval <julia.koval@intel.com>
5582 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
5583 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
5584 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
5585 _mm_mask_bitshuffle_epi64_mask): Fix type.
5586 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
5587 USI_FTYPE_V4DI_V4DI_USI): Remove.
5588 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
5589 __builtin_ia32_vpshufbitqmb256_mask,
5590 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
5591 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
5592 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
5594 2018-01-26 Alan Modra <amodra@gmail.com>
5597 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
5598 UNSPEC_VBPERMQ. Sort other unspecs.
5600 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
5602 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
5604 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
5607 * predict.c (drop_profile): Do not push/pop cfun; update also
5609 (handle_missing_profiles): Fix logic looking for zero profiles.
5611 2018-01-25 Jakub Jelinek <jakub@redhat.com>
5614 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
5615 on functions with #pragma omp declare simd or functions with simd
5617 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
5618 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5619 Remove trailing \n from warning_at calls.
5621 2018-01-25 Tom de Vries <tom@codesourcery.com>
5624 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5625 for neutered workers.
5627 2018-01-24 Joseph Myers <joseph@codesourcery.com>
5630 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
5631 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
5633 2018-01-24 Jeff Law <law@redhat.com>
5636 * i386.c (get_probe_interval): Move to earlier point.
5637 (ix86_compute_frame_layout): If -fstack-clash-protection and
5638 the frame is larger than the probe interval, then use pushes
5639 to save registers rather than reg->mem moves.
5640 (ix86_expand_prologue): Remove conditional for int_registers_saved
5643 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
5646 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
5647 min/max for never referenced object.
5649 2018-01-24 Jakub Jelinek <jakub@redhat.com>
5652 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
5654 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
5655 attributes from DECL_ATTRIBUTES (decl) without affecting
5656 DECL_ATTRIBUTES (current_function_decl).
5657 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
5658 functions with non-NULL DECL_ABSTRACT_ORIGIN.
5660 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
5662 PR tree-optimization/83979
5663 * fold-const.c (fold_comparison): Use constant_boolean_node
5664 instead of boolean_{true,false}_node.
5666 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
5668 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
5671 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5673 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5674 Simplify the clause that sets the length attribute.
5675 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5676 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
5677 clause that sets the length attribute.
5678 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5680 2018-01-24 Tom de Vries <tom@codesourcery.com>
5683 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
5684 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
5685 Add strict parameter.
5686 (prevent_branch_around_nothing): Insert dummy insn between branch to
5687 label and label with no ptx insn inbetween.
5688 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
5690 2018-01-24 Tom de Vries <tom@codesourcery.com>
5693 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5694 for neutered threads in warp.
5695 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
5697 2018-01-24 Richard Biener <rguenther@suse.de>
5699 PR tree-optimization/83176
5700 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
5703 2018-01-24 Richard Biener <rguenther@suse.de>
5705 PR tree-optimization/82819
5706 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
5707 code generating pluses that are no-ops in the target precision.
5709 2018-01-24 Richard Biener <rguenther@suse.de>
5712 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
5714 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5716 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
5717 to merge probabilities.
5718 * predict.c (probably_never_executed): Also mark as cold functions
5719 with global 0 profile and guessed local profile.
5720 * profile-count.c (profile_probability::combine_with_count): New
5722 * profile-count.h (profile_probability::operator*,
5723 profile_probability::operator*=, profile_probability::operator/,
5724 profile_probability::operator/=): Reduce precision to adjusted
5725 and set value to guessed on contradictory divisions.
5726 (profile_probability::combine_with_freq): Remove.
5727 (profile_probability::combine_wiht_count): Declare.
5728 (profile_count::force_nonzero):: Set to adjusted.
5729 (profile_count::probability_in):: Set quality to adjusted.
5730 * tree-ssa-tail-merge.c (replace_block_by): Use
5733 2018-01-23 Andrew Waterman <andrew@sifive.com>
5734 Jim Wilson <jimw@sifive.com>
5736 * config/riscv/riscv.c (riscv_stack_boundary): New.
5737 (riscv_option_override): Set riscv_stack_boundary. Handle
5738 riscv_preferred_stack_boundary_arg.
5739 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
5740 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
5741 (STACK_BOUNDARY): Set to riscv_stack_boundary.
5742 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
5743 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
5744 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
5746 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
5749 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
5750 of struct ix86_frame.
5751 (ix86_expand_epilogue): Likewise. Add a local variable for
5752 the reg_save_offset field in struct ix86_frame.
5754 2018-01-23 Bin Cheng <bin.cheng@arm.com>
5756 PR tree-optimization/82604
5757 * tree-loop-distribution.c (enum partition_kind): New enum item
5758 PKIND_PARTIAL_MEMSET.
5759 (partition_builtin_p): Support above new enum item.
5760 (generate_code_for_partition): Ditto.
5761 (compute_access_range): Differentiate cases that equality can be
5762 proven at all loops, the innermost loops or no loops.
5763 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
5764 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
5765 (finalize_partitions, distribute_loop): Don't fuse partition of
5766 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
5767 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
5770 2018-01-23 Martin Liska <mliska@suse.cz>
5772 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
5773 order to ignore the predictor.
5774 (PRED_POLYMORPHIC_CALL): Likewise.
5775 (PRED_RECURSIVE_CALL): Likewise.
5777 2018-01-23 Martin Liska <mliska@suse.cz>
5779 * tree-profile.c (tree_profiling): Print function header to
5780 aware reader which function we are working on.
5781 * value-prof.c (gimple_find_values_to_profile): Do not print
5782 not interesting value histograms.
5784 2018-01-23 Martin Liska <mliska@suse.cz>
5786 * profile-count.h (enum profile_quality): Add
5787 profile_uninitialized as the first value. Do not number values
5788 as they are zero based.
5789 (profile_count::verify): Update sanity check.
5790 (profile_probability::verify): Likewise.
5792 2018-01-23 Nathan Sidwell <nathan@acm.org>
5794 * doc/invoke.texi (ffor-scope): Deprecate.
5796 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5798 PR tree-optimization/83510
5799 * domwalk.c (set_all_edges_as_executable): New function.
5800 (dom_walker::dom_walker): Convert bool param
5801 "skip_unreachable_blocks" to enum reachability. Move setup of
5802 edge flags to set_all_edges_as_executable and only do it when
5803 reachability is REACHABLE_BLOCKS.
5804 * domwalk.h (enum dom_walker::reachability): New enum.
5805 (dom_walker::dom_walker): Convert bool param
5806 "skip_unreachable_blocks" to enum reachability.
5807 (set_all_edges_as_executable): New decl.
5808 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
5809 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
5811 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
5812 but converting true to REACHABLE_BLOCKS.
5813 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
5815 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
5816 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
5817 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
5819 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
5820 if check_all_array_refs will be called.
5822 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5824 * tree.c (selftest::test_location_wrappers): Add more test
5827 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5829 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
5830 (selftest::test_bit_in_range): Likewise.
5832 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5835 * doc/sourcebuild.texi (vect_float): Say that the selector
5836 only describes the situation when -funsafe-math-optimizations is on.
5837 (vect_float_strict): Document.
5839 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5841 PR tree-optimization/83965
5842 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5843 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
5844 instead of checking only for a reduction.
5845 (vect_recog_widen_sum_pattern): Likewise.
5847 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5849 * predict.c (probably_never_executed): Only use precise profile info.
5850 (compute_function_frequency): Skip after inlining hack since we now
5851 have quality checking.
5853 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5855 * profile-count.h (profile_probability::very_unlikely,
5856 profile_probability::unlikely, profile_probability::even): Set
5857 precision to guessed.
5859 2018-01-23 Richard Biener <rguenther@suse.de>
5861 PR tree-optimization/83963
5862 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
5863 Properly terminate dominator walk when crossing the exit edge not
5864 when visiting its source block.
5866 2018-01-23 Jakub Jelinek <jakub@redhat.com>
5869 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
5870 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
5872 2018-01-22 Jakub Jelinek <jakub@redhat.com>
5874 PR tree-optimization/83957
5875 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
5876 semicolon after for body surrounded by braces.
5878 PR tree-optimization/83081
5879 * profile-count.h (profile_probability::split): New method.
5880 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
5881 Use profile_probability::split.
5882 (do_compare_rtx_and_jump): Fix adjustment of probabilities
5883 when splitting a single conditional jump into 2.
5885 2018-01-22 David Malcolm <dmalcolm@redhat.com>
5887 PR tree-optimization/69452
5888 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
5891 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5893 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
5894 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
5895 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
5897 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5899 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
5900 * config/rl78/rl78.md (movdi): New define_expand.
5901 * config/rl78/rl78.c (rl78_split_movdi): New function.
5903 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
5906 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
5908 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
5909 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
5910 128-bit to produce an UNSPEC move to get the double word with the
5911 signbit and then a shift directly to do signbit.
5912 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
5913 implementation with a new version that just does either a direct
5914 move or a regular move. Move memory interface to separate insns.
5915 Move insns so they are next to the expander.
5916 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
5917 with signbit move. Split big and little endian case.
5918 (signbit<mode>2_dm_mem_le): Likewise.
5919 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
5920 (signbit<mode>2_dm2): Likewise.
5922 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5924 * config/rl78/rl78.md (anddi3): New define_expand.
5926 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5928 * config/rl78/rl78.md (umindi3): New define_expand.
5930 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5932 * config/rl78/rl78.md (smindi3): New define_expand.
5934 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5936 * config/rl78/rl78.md (smaxdi3): New define_expand.
5938 2018-01-22 Carl Love <cel@us.ibm.com>
5940 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
5941 LVX_V1TI): Add macro expansion.
5942 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
5943 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
5944 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
5945 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
5946 Change check to determine if the instruction is a byte reversing
5947 entry. Fix typo in comment.
5948 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
5949 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
5950 Add def_builtin calls for new builtins.
5951 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
5952 Add define_insn expansion.
5954 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5956 * config/rl78/rl78.md (umaxdi3): New define_expand.
5958 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5960 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
5961 for non-QImode registers.
5963 2018-01-22 Richard Biener <rguenther@suse.de>
5965 PR tree-optimization/83963
5966 * graphite-scop-detection.c (scop_detection::get_sese): Delay
5967 including the loop exit block.
5968 (scop_detection::merge_sese): Likewise.
5969 (scop_detection::add_scop): Do it here instead.
5971 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5973 * doc/sourcebuild.texi (arm_softfloat): Document.
5975 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
5978 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
5979 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
5980 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
5982 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5983 David Edelsohn <dje.gcc@gmail.com>
5986 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5987 Change "crset eq" to "crset 2".
5988 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5989 (*call_indirect_aix<mode>_nospec): Likewise.
5990 (*call_value_indirect_aix<mode>_nospec): Likewise.
5991 (*call_indirect_elfv2<mode>_nospec): Likewise.
5992 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
5993 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
5994 change assembly output from . to $.
5995 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5996 (indirect_jump<mode>_nospec): Change assembly output from . to $.
5997 (*tablejump<mode>_internal1_nospec): Likewise.
5999 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
6002 * config/sh/sh_optimize_sett_clrt.cc:
6003 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6005 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
6007 PR tree-optimization/83940
6008 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6009 offset_dt to vect_constant_def rather than vect_unknown_def_type.
6010 (vect_check_load_store_mask): Add a mask_dt_out parameter and
6011 use it to pass back the definition type.
6012 (vect_check_store_rhs): Likewise rhs_dt_out.
6013 (vect_build_gather_load_calls): Add a mask_dt argument and use
6014 it instead of a call to vect_is_simple_use.
6015 (vectorizable_store): Update calls to vect_check_load_store_mask
6016 and vect_check_store_rhs. Use the dt returned by the latter instead
6017 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
6018 instead of calls to vect_is_simple_use. Pass the scalar rather
6019 than the vector operand to vect_is_simple_use when handling
6020 second and subsequent copies of an rhs value.
6021 (vectorizable_load): Update calls to vect_check_load_store_mask
6022 and vect_build_gather_load_calls. Use the cached mask_dt and
6023 gs_info.offset_dt instead of calls to vect_is_simple_use.
6025 2018-01-20 Jakub Jelinek <jakub@redhat.com>
6028 * tree-emutls.c: Include gimplify.h.
6029 (lower_emutls_2): New function.
6030 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6031 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6032 it before further processing.
6035 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6036 UINTVAL (trueop1) instead of INTVAL (op1).
6038 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6042 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6043 INCOMING_FRAME_SP_OFFSET if not defined.
6044 (scan_trace): Add ENTRY argument. If true and
6045 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6046 emit a note to adjust the CFA offset.
6047 (create_cfi_notes): Adjust scan_trace callers.
6048 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6049 INCOMING_FRAME_SP_OFFSET in the CIE.
6050 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6051 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6053 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6054 * doc/tm.texi: Regenerated.
6056 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6058 PR rtl-optimization/83147
6059 * lra-constraints.c (remove_inheritance_pseudos): Use
6060 lra_substitute_pseudo_within_insn.
6062 2018-01-19 Tom de Vries <tom@codesourcery.com>
6063 Cesar Philippidis <cesar@codesourcery.com>
6066 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6068 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6071 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6072 spaces for function labels.
6074 2018-01-19 Martin Liska <mliska@suse.cz>
6076 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6077 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6078 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6079 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6080 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6081 (PRED_CONST_RETURN): Change from 69 to 65.
6082 (PRED_NULL_RETURN): Change from 91 to 71.
6083 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6084 (PRED_LOOP_GUARD): Change from 66 to 73.
6086 2018-01-19 Martin Liska <mliska@suse.cz>
6088 * predict.c (predict_insn_def): Add new assert.
6089 (struct branch_predictor): Change type to signed integer.
6090 (test_prediction_value_range): Amend test to cover
6092 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6093 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6094 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6095 (PRED_LOOP_IV_COMPARE): Likewise.
6096 * predict.h (PROB_UNINITIALIZED): Define new constant.
6098 2018-01-19 Martin Liska <mliska@suse.cz>
6100 * predict.c (dump_prediction): Add new format for
6101 analyze_brprob.py script which is enabled with -details
6103 * profile-count.h (precise_p): New function.
6105 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6107 PR tree-optimization/83922
6108 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6109 there are no statements that need masking.
6110 (vect_active_double_reduction_p): New function.
6111 (vect_analyze_loop_operations): Use it when handling phis that
6112 are not in the loop header.
6114 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6116 PR tree-optimization/83914
6117 * tree-vect-loop.c (vectorizable_induction): Don't convert
6118 init_expr or apply the peeling adjustment for inductions
6119 that are nested within the vectorized loop.
6121 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6123 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6126 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6130 * function.h (gimplify_parameters): Add gimple_seq * argument.
6131 * function.c: Include gimple.h and options.h.
6132 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6133 for the added local temporaries if needed.
6134 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6135 if there are any parameter cleanups, wrap whole body into a
6136 try/finally with the cleanups.
6138 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6141 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6142 Use GET_MODE_CLASS for scalar floating point.
6144 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6148 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6149 Fix call of call_cgraph_insertion_hooks.
6151 2018-01-18 Martin Sebor <msebor@redhat.com>
6153 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6155 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6158 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6161 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
6164 * common.opt: (-ffile-prefix-map): New option.
6165 * opts.c (common_handle_option): Defer it.
6166 * opts-global.c (handle_common_deferred_options): Handle it.
6167 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6168 * file-prefix-map.h: New file.
6169 (remap_debug_filename, add_debug_prefix_map): ...here.
6170 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6171 * final.c (debug_prefix_map, add_debug_prefix_map
6172 remap_debug_filename): Move to...
6173 * file-prefix-map.c: New file.
6174 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6175 generalize, get rid of alloca(), use strrchr() instead of strchr().
6176 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6177 Implement in terms of add_prefix_map().
6178 (remap_macro_filename, remap_debug_filename): Implement in term of
6180 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6181 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6182 * dbxout.c: Include file-prefix-map.h.
6183 * varasm.c: Likewise.
6184 * vmsdbgout.c: Likewise.
6185 * xcoffout.c: Likewise.
6186 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6187 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6188 * doc/invoke.texi (-ffile-prefix-map): Document.
6189 (-fdebug-prefix-map): Update description.
6191 2018-01-18 Martin Liska <mliska@suse.cz>
6193 * config/i386/i386.c (indirect_thunk_name): Document that also
6195 (output_indirect_thunk): Document why both instructions
6196 (pause and lfence) are generated.
6198 2018-01-18 Richard Biener <rguenther@suse.de>
6200 PR tree-optimization/83887
6201 * graphite-scop-detection.c
6202 (scop_detection::get_nearest_dom_with_single_entry): Remove.
6203 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6204 (scop_detection::merge_sese): Re-implement with a flood-fill
6205 algorithm that properly finds a SESE region if it exists.
6207 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6210 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
6211 pointer_diff optimizations use view_convert instead of convert.
6213 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6215 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6216 Generate different code for -mno-speculate-indirect-jumps.
6217 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6218 (*call_indirect_aix<mode>): Disable for
6219 -mno-speculate-indirect-jumps.
6220 (*call_indirect_aix<mode>_nospec): New define_insn.
6221 (*call_value_indirect_aix<mode>): Disable for
6222 -mno-speculate-indirect-jumps.
6223 (*call_value_indirect_aix<mode>_nospec): New define_insn.
6224 (*sibcall_nonlocal_sysv<mode>): Generate different code for
6225 -mno-speculate-indirect-jumps.
6226 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6228 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6230 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
6231 long double type, set the flags for noting the default long double
6232 type, even if we don't pass or return a long double type.
6234 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
6237 * ipa-inline.c (flatten_function): Do not overwrite final inlining
6240 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
6242 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6243 support for merge[hl].
6244 (fold_mergehl_helper): New helper function.
6245 (tree-vector-builder.h): New #include for tree_vector_builder usage.
6246 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6247 (altivec_vmrglw_direct): Add xxmrglw insn.
6249 2018-01-17 Andrew Waterman <andrew@sifive.com>
6251 * config/riscv/riscv.c (riscv_conditional_register_usage): If
6252 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6254 2018-01-17 David Malcolm <dmalcolm@redhat.com>
6257 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6258 call the lto_location_cache before reading the
6259 DECL_SOURCE_LOCATION of the types.
6261 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
6262 Richard Sandiford <richard.sandiford@linaro.org>
6264 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
6265 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6266 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6267 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6268 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6270 * config/aarch64/constraints.md (aarch64_movti_operand):
6272 * config/aarch64/predicates.md (Uti): Add new constraint.
6274 2018-01-17 Carl Love <cel@us.ibm.com>
6276 * config/rs6000/vsx.md (define_expand xl_len_r,
6277 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6278 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6280 (define_expand, define_insn): Move the shift left from the
6281 define_insn to the define_expand for lxvl and stxvl instructions.
6282 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6283 and XL_LEN_R definitions to PURE.
6285 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
6287 * config/i386/i386.c (indirect_thunk_name): Declare regno
6288 as unsigned int. Compare regno with INVALID_REGNUM.
6289 (output_indirect_thunk): Ditto.
6290 (output_indirect_thunk_function): Ditto.
6291 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6292 in the call to output_indirect_thunk_function.
6294 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6297 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6298 rather than the size of inner_type to determine the stack slot size
6299 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6301 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6304 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6307 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6309 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6310 endian Linux systems to optionally enable multilibs for selecting
6311 the long double type if the user configured an explicit type.
6312 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6313 have no long double multilibs if not defined.
6314 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6315 warn if the user used -mabi={ieee,ibm}longdouble and we built
6316 multilibs for long double.
6317 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6318 appropriate multilib option.
6319 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6321 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6322 for building long double multilibs.
6323 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6325 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6327 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6330 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6332 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6335 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6338 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6341 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6343 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6344 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6346 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6348 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6349 different rtl trees depending on TARGET_64BIT.
6350 (rs6000_gen_lvx): Likewise.
6352 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6354 * config/visium/visium.md (nop): Tweak comment.
6355 (hazard_nop): Likewise.
6357 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6359 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6360 -mspeculate-indirect-jumps.
6361 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6362 for -mno-speculate-indirect-jumps.
6363 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6364 (*call_value_indirect_elfv2<mode>): Disable for
6365 -mno-speculate-indirect-jumps.
6366 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6367 (indirect_jump): Emit different RTL for
6368 -mno-speculate-indirect-jumps.
6369 (*indirect_jump<mode>): Disable for
6370 -mno-speculate-indirect-jumps.
6371 (*indirect_jump<mode>_nospec): New define_insn.
6372 (tablejump): Emit different RTL for
6373 -mno-speculate-indirect-jumps.
6374 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6375 (tablejumpsi_nospec): New define_expand.
6376 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6377 (tablejumpdi_nospec): New define_expand.
6378 (*tablejump<mode>_internal1): Disable for
6379 -mno-speculate-indirect-jumps.
6380 (*tablejump<mode>_internal1_nospec): New define_insn.
6381 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6384 2018-01-16 Artyom Skrobov tyomitch@gmail.com
6386 * caller-save.c (insert_save): Drop unnecessary parameter. All
6389 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6390 Richard Biener <rguenth@suse.de>
6393 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6394 return early, inline manually is_gimple_sizepos. Make sure if we
6395 call gimplify_expr we don't end up with a gimple constant.
6396 * tree.c (variably_modified_type_p): Don't return true for
6397 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
6398 * gimplify.h (is_gimple_sizepos): Remove.
6400 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6402 PR tree-optimization/83857
6403 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6404 vectorizable_live_operation for pure SLP statements.
6405 (vectorizable_live_operation): Handle PHIs.
6407 2018-01-16 Richard Biener <rguenther@suse.de>
6409 PR tree-optimization/83867
6410 * tree-vect-stmts.c (vect_transform_stmt): Precompute
6411 nested_in_vect_loop_p since the scalar stmt may get invalidated.
6413 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6416 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
6417 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
6418 If off is not INTEGER_CST, issue a may not be aligned warning
6419 rather than isn't aligned. Use isn%'t rather than isn't.
6420 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
6422 <case MULT_EXPR>: Improve the case when bottom and one of the
6423 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
6424 operand, in that case check if the other operand is multiple of
6425 bottom divided by the INTEGER_CST operand.
6427 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6430 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
6431 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
6432 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
6433 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
6434 * config/pa/pa.c (pa_function_arg_advance): Likewise.
6435 (pa_function_arg, pa_arg_partial_bytes): Likewise.
6436 (pa_function_arg_size): New function.
6438 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6440 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
6441 in a separate statement.
6443 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6445 PR tree-optimization/83847
6446 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
6447 group gathers and scatters.
6449 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6451 PR rtl-optimization/86620
6452 * params.def (max-sched-ready-insns): Bump minimum value to 1.
6454 PR rtl-optimization/83213
6455 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
6456 to last if both are JUMP_INSNs.
6458 PR tree-optimization/83843
6459 * gimple-ssa-store-merging.c
6460 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
6461 store_immediate_info for bswap/nop orig_stores.
6463 2018-01-15 Andrew Waterman <andrew@sifive.com>
6465 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
6467 <UDIV>: Increase cost if !TARGET_DIV.
6469 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
6471 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
6472 (define_attr "cr_logical_3op"): New.
6473 (cceq_ior_compare): Adjust.
6474 (cceq_ior_compare_complement): Adjust.
6475 (*cceq_rev_compare): Adjust.
6476 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
6477 (is_cracked_insn): Adjust.
6478 (insn_must_be_first_in_group): Adjust.
6479 * config/rs6000/40x.md: Adjust.
6480 * config/rs6000/440.md: Adjust.
6481 * config/rs6000/476.md: Adjust.
6482 * config/rs6000/601.md: Adjust.
6483 * config/rs6000/603.md: Adjust.
6484 * config/rs6000/6xx.md: Adjust.
6485 * config/rs6000/7450.md: Adjust.
6486 * config/rs6000/7xx.md: Adjust.
6487 * config/rs6000/8540.md: Adjust.
6488 * config/rs6000/cell.md: Adjust.
6489 * config/rs6000/e300c2c3.md: Adjust.
6490 * config/rs6000/e500mc.md: Adjust.
6491 * config/rs6000/e500mc64.md: Adjust.
6492 * config/rs6000/e5500.md: Adjust.
6493 * config/rs6000/e6500.md: Adjust.
6494 * config/rs6000/mpc.md: Adjust.
6495 * config/rs6000/power4.md: Adjust.
6496 * config/rs6000/power5.md: Adjust.
6497 * config/rs6000/power6.md: Adjust.
6498 * config/rs6000/power7.md: Adjust.
6499 * config/rs6000/power8.md: Adjust.
6500 * config/rs6000/power9.md: Adjust.
6501 * config/rs6000/rs64.md: Adjust.
6502 * config/rs6000/titan.md: Adjust.
6504 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6506 * config/i386/predicates.md (indirect_branch_operand): Rewrite
6507 ix86_indirect_branch_register logic.
6509 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6511 * config/i386/constraints.md (Bs): Update
6512 ix86_indirect_branch_register check. Don't check
6513 ix86_indirect_branch_register with GOT_memory_operand.
6515 * config/i386/predicates.md (GOT_memory_operand): Don't check
6516 ix86_indirect_branch_register here.
6517 (GOT32_symbol_operand): Likewise.
6519 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6521 * config/i386/predicates.md (constant_call_address_operand):
6522 Rewrite ix86_indirect_branch_register logic.
6523 (sibcall_insn_operand): Likewise.
6525 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6527 * config/i386/constraints.md (Bs): Replace
6528 ix86_indirect_branch_thunk_register with
6529 ix86_indirect_branch_register.
6531 * config/i386/i386.md (indirect_jump): Likewise.
6532 (tablejump): Likewise.
6533 (*sibcall_memory): Likewise.
6534 (*sibcall_value_memory): Likewise.
6535 Peepholes of indirect call and jump via memory: Likewise.
6536 * config/i386/i386.opt: Likewise.
6537 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6538 (GOT_memory_operand): Likewise.
6539 (call_insn_operand): Likewise.
6540 (sibcall_insn_operand): Likewise.
6541 (GOT32_symbol_operand): Likewise.
6543 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6546 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
6547 type rather than type addr's type points to.
6548 (expand_omp_atomic_mutex): Likewise.
6549 (expand_omp_atomic): Likewise.
6551 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6554 * config/i386/i386.c (output_indirect_thunk_function): Use
6555 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
6556 for __x86_return_thunk.
6558 2018-01-15 Richard Biener <rguenther@suse.de>
6561 * expmed.c (extract_bit_field_1): Fix typo.
6563 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6566 * config/arm/iterators.md (VF): New mode iterator.
6567 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
6568 Remove integer-related logic from pattern.
6569 (neon_vabd<mode>_3): Likewise.
6571 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6574 * common.opt (fstrict-overflow): No longer an alias.
6575 (fwrapv-pointer): New option.
6576 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
6577 also for pointer types based on flag_wrapv_pointer.
6578 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
6579 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
6580 opts->x_flag_wrapv got set.
6581 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
6582 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
6583 POINTER_TYPE_OVERFLOW_UNDEFINED.
6584 * match.pd: Likewise in address comparison pattern.
6585 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
6587 2018-01-15 Richard Biener <rguenther@suse.de>
6590 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
6591 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
6592 Reset type names to their identifier if their TYPE_DECL doesn't
6593 have linkage (and thus is used for ODR and devirt).
6594 (save_debug_info_for_decl): Remove.
6595 (save_debug_info_for_type): Likewise.
6596 (add_tree_to_fld_list): Adjust.
6597 * tree-pretty-print.c (dump_generic_node): Make dumping of
6598 type names more robust.
6600 2018-01-15 Richard Biener <rguenther@suse.de>
6602 * BASE-VER: Bump to 8.0.1.
6604 2018-01-14 Martin Sebor <msebor@redhat.com>
6607 * builtins.c (check_access): Avoid warning when the no-warning bit
6610 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
6612 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
6613 * ira-color (allocno_hard_regs_compare): Likewise.
6615 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
6618 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
6619 Use .pushsection/.popsection.
6621 2018-01-14 Martin Sebor <msebor@redhat.com>
6624 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
6626 2018-01-14 Jakub Jelinek <jakub@redhat.com>
6628 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
6629 entry from extra_headers.
6630 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
6631 extra_headers, make the list bitwise identical to the i?86-*-* one.
6633 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6635 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
6636 -mcmodel=large with -mindirect-branch=thunk,
6637 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
6638 -mfunction-return=thunk-extern.
6639 * doc/invoke.texi: Document -mcmodel=large is incompatible with
6640 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
6641 -mfunction-return=thunk and -mfunction-return=thunk-extern.
6643 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6645 * config/i386/i386.c (print_reg): Print the name of the full
6646 integer register without '%'.
6647 (ix86_print_operand): Handle 'V'.
6648 * doc/extend.texi: Document 'V' modifier.
6650 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6652 * config/i386/constraints.md (Bs): Disallow memory operand for
6653 -mindirect-branch-register.
6655 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6656 (GOT_memory_operand): Likewise.
6657 (call_insn_operand): Likewise.
6658 (sibcall_insn_operand): Likewise.
6659 (GOT32_symbol_operand): Likewise.
6660 * config/i386/i386.md (indirect_jump): Call convert_memory_address
6661 for -mindirect-branch-register.
6662 (tablejump): Likewise.
6663 (*sibcall_memory): Likewise.
6664 (*sibcall_value_memory): Likewise.
6665 Disallow peepholes of indirect call and jump via memory for
6666 -mindirect-branch-register.
6667 (*call_pop): Replace m with Bw.
6668 (*call_value_pop): Likewise.
6669 (*sibcall_pop_memory): Replace m with Bs.
6670 * config/i386/i386.opt (mindirect-branch-register): New option.
6671 * doc/invoke.texi: Document -mindirect-branch-register option.
6673 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6675 * config/i386/i386-protos.h (ix86_output_function_return): New.
6676 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
6677 set function_return_type.
6678 (indirect_thunk_name): Add ret_p to indicate thunk for function
6680 (output_indirect_thunk_function): Pass false to
6681 indirect_thunk_name.
6682 (ix86_output_indirect_branch_via_reg): Likewise.
6683 (ix86_output_indirect_branch_via_push): Likewise.
6684 (output_indirect_thunk_function): Create alias for function
6685 return thunk if regno < 0.
6686 (ix86_output_function_return): New function.
6687 (ix86_handle_fndecl_attribute): Handle function_return.
6688 (ix86_attribute_table): Add function_return.
6689 * config/i386/i386.h (machine_function): Add
6690 function_return_type.
6691 * config/i386/i386.md (simple_return_internal): Use
6692 ix86_output_function_return.
6693 (simple_return_internal_long): Likewise.
6694 * config/i386/i386.opt (mfunction-return=): New option.
6695 (indirect_branch): Mention -mfunction-return=.
6696 * doc/extend.texi: Document function_return function attribute.
6697 * doc/invoke.texi: Document -mfunction-return= option.
6699 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6701 * config/i386/i386-opts.h (indirect_branch): New.
6702 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
6703 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
6704 with local indirect jump when converting indirect call and jump.
6705 (ix86_set_indirect_branch_type): New.
6706 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
6707 (indirectlabelno): New.
6708 (indirect_thunk_needed): Likewise.
6709 (indirect_thunk_bnd_needed): Likewise.
6710 (indirect_thunks_used): Likewise.
6711 (indirect_thunks_bnd_used): Likewise.
6712 (INDIRECT_LABEL): Likewise.
6713 (indirect_thunk_name): Likewise.
6714 (output_indirect_thunk): Likewise.
6715 (output_indirect_thunk_function): Likewise.
6716 (ix86_output_indirect_branch_via_reg): Likewise.
6717 (ix86_output_indirect_branch_via_push): Likewise.
6718 (ix86_output_indirect_branch): Likewise.
6719 (ix86_output_indirect_jmp): Likewise.
6720 (ix86_code_end): Call output_indirect_thunk_function if needed.
6721 (ix86_output_call_insn): Call ix86_output_indirect_branch if
6723 (ix86_handle_fndecl_attribute): Handle indirect_branch.
6724 (ix86_attribute_table): Add indirect_branch.
6725 * config/i386/i386.h (machine_function): Add indirect_branch_type
6726 and has_local_indirect_jump.
6727 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
6729 (tablejump): Likewise.
6730 (*indirect_jump): Use ix86_output_indirect_jmp.
6731 (*tablejump_1): Likewise.
6732 (simple_return_indirect_internal): Likewise.
6733 * config/i386/i386.opt (mindirect-branch=): New option.
6734 (indirect_branch): New.
6737 (thunk-inline): Likewise.
6738 (thunk-extern): Likewise.
6739 * doc/extend.texi: Document indirect_branch function attribute.
6740 * doc/invoke.texi: Document -mindirect-branch= option.
6742 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
6745 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
6747 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
6749 * ipa-inline.c (want_inline_small_function_p): Return false if
6750 inlining has already failed with CIF_FINAL_ERROR.
6751 (update_caller_keys): Call want_inline_small_function_p before
6753 (update_callee_keys): Likewise.
6755 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6757 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
6759 (rs6000_quadword_masked_address_p): Likewise.
6760 (quad_aligned_load_p): Likewise.
6761 (quad_aligned_store_p): Likewise.
6762 (const_load_sequence_p): Add comment to describe the outer-most loop.
6763 (mimic_memory_attributes_and_flags): New function.
6764 (rs6000_gen_stvx): Likewise.
6765 (replace_swapped_aligned_store): Likewise.
6766 (rs6000_gen_lvx): Likewise.
6767 (replace_swapped_aligned_load): Likewise.
6768 (replace_swapped_load_constant): Capitalize argument name in
6769 comment describing this function.
6770 (rs6000_analyze_swaps): Add a third pass to search for vector loads
6771 and stores that access quad-word aligned addresses and replace
6772 with stvx or lvx instructions when appropriate.
6773 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
6774 New function prototype.
6775 (rs6000_quadword_masked_address_p): Likewise.
6776 (rs6000_gen_lvx): Likewise.
6777 (rs6000_gen_stvx): Likewise.
6778 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
6779 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
6780 when memory address is aligned.
6781 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
6782 this split to select lvx instruction when memory address is aligned.
6783 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
6784 instruction when memory address is aligned.
6785 (*vsx_le_perm_load_v16qi): Likewise.
6786 (four unnamed splitters): Modify to select the stvx instruction
6787 when memory is aligned.
6789 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
6791 * predict.c (determine_unlikely_bbs): Handle correctly BBs
6792 which appears in the queue multiple times.
6794 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6795 Alan Hayward <alan.hayward@arm.com>
6796 David Sherwood <david.sherwood@arm.com>
6798 * tree-vectorizer.h (vec_lower_bound): New structure.
6799 (_loop_vec_info): Add check_nonzero and lower_bounds.
6800 (LOOP_VINFO_CHECK_NONZERO): New macro.
6801 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
6802 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
6803 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
6804 fields. Make seg_len the distance travelled, not including the
6806 (dr_direction_indicator): Declare.
6807 (dr_zero_step_indicator): Likewise.
6808 (dr_known_forward_stride_p): Likewise.
6809 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
6811 (runtime_alias_check_p): Allow runtime alias checks with
6813 (operator ==): Compare access_size and align.
6814 (prune_runtime_alias_test_list): Rework for new distinction between
6815 the access_size and seg_len.
6816 (create_intersect_range_checks_index): Likewise. Cope with polynomial
6818 (get_segment_min_max): New function.
6819 (create_intersect_range_checks): Use it.
6820 (dr_step_indicator): New function.
6821 (dr_direction_indicator): Likewise.
6822 (dr_zero_step_indicator): Likewise.
6823 (dr_known_forward_stride_p): Likewise.
6824 * tree-loop-distribution.c (data_ref_segment_size): Return
6825 DR_STEP * (niters - 1).
6826 (compute_alias_check_pairs): Update call to the dr_with_seg_len
6828 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
6829 (vect_preserves_scalar_order_p): New function, split out from...
6830 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
6831 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
6832 (vect_vfa_access_size): New function.
6833 (vect_vfa_align): Likewise.
6834 (vect_compile_time_alias): Take access_size_a and access_b arguments.
6835 (dump_lower_bound): New function.
6836 (vect_check_lower_bound): Likewise.
6837 (vect_small_gap_p): Likewise.
6838 (vectorizable_with_step_bound_p): Likewise.
6839 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
6840 depencies if the vectorization factor is 1. Convert the checks
6841 for nonzero steps into checks on the bounds of DR_STEP. Try using
6842 a bunds check for variable steps if the minimum required step is
6843 relatively small. Update calls to the dr_with_seg_len
6844 constructor and to vect_compile_time_alias.
6845 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
6847 (vect_loop_versioning): Call it.
6848 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
6850 (vect_estimate_min_profitable_iters): Account for any bounds checks.
6852 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6853 Alan Hayward <alan.hayward@arm.com>
6854 David Sherwood <david.sherwood@arm.com>
6856 * doc/sourcebuild.texi (vect_scatter_store): Document.
6857 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
6859 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
6861 * genopinit.c (main): Add supports_vec_scatter_store and
6862 supports_vec_scatter_store_cached to target_optabs.
6863 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
6864 IFN_MASK_SCATTER_STORE.
6865 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
6867 * internal-fn.h (internal_store_fn_p): Declare.
6868 (internal_fn_stored_value_index): Likewise.
6869 * internal-fn.c (scatter_store_direct): New macro.
6870 (expand_scatter_store_optab_fn): New function.
6871 (direct_scatter_store_optab_supported_p): New macro.
6872 (internal_store_fn_p): New function.
6873 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
6874 IFN_MASK_SCATTER_STORE.
6875 (internal_fn_mask_index): Likewise.
6876 (internal_fn_stored_value_index): New function.
6877 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
6879 * optabs-query.h (supports_vec_scatter_store_p): Declare.
6880 * optabs-query.c (supports_vec_scatter_store_p): New function.
6881 * tree-vectorizer.h (vect_get_store_rhs): Declare.
6882 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
6883 true for scatter stores.
6884 (vect_gather_scatter_fn_p): Handle scatter stores too.
6885 (vect_check_gather_scatter): Consider using scatter stores if
6886 supports_vec_scatter_store_p.
6887 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
6889 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6890 internal_fn_stored_value_index.
6891 (check_load_store_masking): Handle scatter stores too.
6892 (vect_get_store_rhs): Make public.
6893 (vectorizable_call): Use internal_store_fn_p.
6894 (vectorizable_store): Handle scatter store internal functions.
6895 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
6896 when deciding whether the end of the group has been reached.
6897 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
6898 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
6899 (mask_scatter_store<mode>): New insns.
6901 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6902 Alan Hayward <alan.hayward@arm.com>
6903 David Sherwood <david.sherwood@arm.com>
6905 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
6906 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
6907 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
6909 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
6910 Use vect_truncate_gather_scatter_offset if we can't treat the
6911 operation as a normal gather load or scatter store.
6912 (get_group_load_store_type): Take the gather_scatter_info
6913 as argument. Try using a gather load or scatter store for
6914 single-element groups.
6915 (get_load_store_type): Update calls to get_group_load_store_type
6916 and vect_use_strided_gather_scatters_p.
6918 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6919 Alan Hayward <alan.hayward@arm.com>
6920 David Sherwood <david.sherwood@arm.com>
6922 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
6923 optional tree argument.
6924 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
6926 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
6927 but continue to use the current value as a fallback.
6928 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
6929 to compare the updates.
6930 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
6931 (get_load_store_type): Use it when handling a strided access.
6932 (vect_get_strided_load_store_ops): New function.
6933 (vect_get_data_ptr_increment): Likewise.
6934 (vectorizable_load): Handle strided gather loads. Always pass
6935 a step to vect_create_data_ref_ptr and bump_vector_ptr.
6937 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6938 Alan Hayward <alan.hayward@arm.com>
6939 David Sherwood <david.sherwood@arm.com>
6941 * doc/md.texi (gather_load@var{m}): Document.
6942 (mask_gather_load@var{m}): Likewise.
6943 * genopinit.c (main): Add supports_vec_gather_load and
6944 supports_vec_gather_load_cached to target_optabs.
6945 * optabs-tree.c (init_tree_optimization_optabs): Use
6946 ggc_cleared_alloc to allocate target_optabs.
6947 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
6948 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
6950 * internal-fn.h (internal_load_fn_p): Declare.
6951 (internal_gather_scatter_fn_p): Likewise.
6952 (internal_fn_mask_index): Likewise.
6953 (internal_gather_scatter_fn_supported_p): Likewise.
6954 * internal-fn.c (gather_load_direct): New macro.
6955 (expand_gather_load_optab_fn): New function.
6956 (direct_gather_load_optab_supported_p): New macro.
6957 (direct_internal_fn_optab): New function.
6958 (internal_load_fn_p): Likewise.
6959 (internal_gather_scatter_fn_p): Likewise.
6960 (internal_fn_mask_index): Likewise.
6961 (internal_gather_scatter_fn_supported_p): Likewise.
6962 * optabs-query.c (supports_at_least_one_mode_p): New function.
6963 (supports_vec_gather_load_p): Likewise.
6964 * optabs-query.h (supports_vec_gather_load_p): Declare.
6965 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
6966 and memory_type field.
6967 (NUM_PATTERNS): Bump to 15.
6968 * tree-vect-data-refs.c: Include internal-fn.h.
6969 (vect_gather_scatter_fn_p): New function.
6970 (vect_describe_gather_scatter_call): Likewise.
6971 (vect_check_gather_scatter): Try using internal functions for
6972 gather loads. Recognize existing calls to a gather load function.
6973 (vect_analyze_data_refs): Consider using gather loads if
6974 supports_vec_gather_load_p.
6975 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
6976 (vect_get_gather_scatter_offset_type): Likewise.
6977 (vect_convert_mask_for_vectype): Likewise.
6978 (vect_add_conversion_to_patterm): Likewise.
6979 (vect_try_gather_scatter_pattern): Likewise.
6980 (vect_recog_gather_scatter_pattern): New pattern recognizer.
6981 (vect_vect_recog_func_ptrs): Add it.
6982 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6983 internal_fn_mask_index and internal_gather_scatter_fn_p.
6984 (check_load_store_masking): Take the gather_scatter_info as an
6985 argument and handle gather loads.
6986 (vect_get_gather_scatter_ops): New function.
6987 (vectorizable_call): Check internal_load_fn_p.
6988 (vectorizable_load): Likewise. Handle gather load internal
6990 (vectorizable_store): Update call to check_load_store_masking.
6991 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
6992 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
6993 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
6994 (aarch64_gather_scale_operand_d): New predicates.
6995 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
6996 (mask_gather_load<mode>): New insns.
6998 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6999 Alan Hayward <alan.hayward@arm.com>
7000 David Sherwood <david.sherwood@arm.com>
7002 * optabs.def (fold_left_plus_optab): New optab.
7003 * doc/md.texi (fold_left_plus_@var{m}): Document.
7004 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7005 * internal-fn.c (fold_left_direct): Define.
7006 (expand_fold_left_optab_fn): Likewise.
7007 (direct_fold_left_optab_supported_p): Likewise.
7008 * fold-const-call.c (fold_const_fold_left): New function.
7009 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7010 * tree-parloops.c (valid_reduction_p): New function.
7011 (gather_scalar_reductions): Use it.
7012 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7013 (vect_finish_replace_stmt): Declare.
7014 * tree-vect-loop.c (fold_left_reduction_fn): New function.
7015 (needs_fold_left_reduction_p): New function, split out from...
7016 (vect_is_simple_reduction): ...here. Accept reductions that
7017 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7018 (vect_force_simple_reduction): Also store the reduction type in
7019 the assignment's STMT_VINFO_REDUC_TYPE.
7020 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7021 (merge_with_identity): New function.
7022 (vect_expand_fold_left): Likewise.
7023 (vectorize_fold_left_reduction): Likewise.
7024 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
7025 scalar phi in place for it. Check for target support and reject
7026 cases that would reassociate the operation. Defer the transform
7027 phase to vectorize_fold_left_reduction.
7028 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7029 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7030 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7032 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7034 * tree-if-conv.c (predicate_mem_writes): Remove redundant
7035 call to ifc_temp_var.
7037 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7038 Alan Hayward <alan.hayward@arm.com>
7039 David Sherwood <david.sherwood@arm.com>
7041 * target.def (legitimize_address_displacement): Take the original
7042 offset as a poly_int.
7043 * targhooks.h (default_legitimize_address_displacement): Update
7045 * targhooks.c (default_legitimize_address_displacement): Likewise.
7046 * doc/tm.texi: Regenerate.
7047 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7048 as an argument, moving assert of ad->disp == ad->disp_term to...
7049 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
7050 Try calling targetm.legitimize_address_displacement before expanding
7051 the address rather than afterwards, and adjust for the new interface.
7052 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7053 Match the new hook interface. Handle SVE addresses.
7054 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7057 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7059 * Makefile.in (OBJS): Add early-remat.o.
7060 * target.def (select_early_remat_modes): New hook.
7061 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7062 * doc/tm.texi: Regenerate.
7063 * targhooks.h (default_select_early_remat_modes): Declare.
7064 * targhooks.c (default_select_early_remat_modes): New function.
7065 * timevar.def (TV_EARLY_REMAT): New timevar.
7066 * passes.def (pass_early_remat): New pass.
7067 * tree-pass.h (make_pass_early_remat): Declare.
7068 * early-remat.c: New file.
7069 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7071 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7073 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7074 Alan Hayward <alan.hayward@arm.com>
7075 David Sherwood <david.sherwood@arm.com>
7077 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7078 vfm1 with a bound_epilog parameter.
7079 (vect_do_peeling): Update calls accordingly, and move the prologue
7080 call earlier in the function. Treat the base bound_epilog as 0 for
7081 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7082 this base when peeling for gaps.
7083 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7084 with fully-masked loops.
7085 (vect_estimate_min_profitable_iters): Handle the single peeled
7086 iteration in that case.
7088 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7089 Alan Hayward <alan.hayward@arm.com>
7090 David Sherwood <david.sherwood@arm.com>
7092 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7093 single-element interleaving even if the size is not a power of 2.
7094 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7095 accesses for single-element interleaving if the group size is
7098 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7099 Alan Hayward <alan.hayward@arm.com>
7100 David Sherwood <david.sherwood@arm.com>
7102 * doc/md.texi (fold_extract_last_@var{m}): Document.
7103 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7104 * optabs.def (fold_extract_last_optab): New optab.
7105 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7106 * internal-fn.c (fold_extract_direct): New macro.
7107 (expand_fold_extract_optab_fn): Likewise.
7108 (direct_fold_extract_optab_supported_p): Likewise.
7109 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7110 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7111 EXTRACT_LAST_REDUCTION.
7112 (get_initial_def_for_reduction): Do not create an initial vector
7113 for EXTRACT_LAST_REDUCTION reductions.
7114 (vectorizable_reduction): Leave the scalar phi in place for
7115 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7116 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7117 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7118 transform phase to vectorizable_condition.
7119 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7121 (vect_finish_stmt_generation): ...here.
7122 (vect_finish_replace_stmt): New function.
7123 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7124 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7126 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7128 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7129 Alan Hayward <alan.hayward@arm.com>
7130 David Sherwood <david.sherwood@arm.com>
7132 * doc/md.texi (extract_last_@var{m}): Document.
7133 * optabs.def (extract_last_optab): New optab.
7134 * internal-fn.def (EXTRACT_LAST): New internal function.
7135 * internal-fn.c (cond_unary_direct): New macro.
7136 (expand_cond_unary_optab_fn): Likewise.
7137 (direct_cond_unary_optab_supported_p): Likewise.
7138 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7139 loops using EXTRACT_LAST.
7140 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7141 (extract_last_<mode>): ...this optab.
7142 (vec_extract<mode><Vel>): Update accordingly.
7144 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7145 Alan Hayward <alan.hayward@arm.com>
7146 David Sherwood <david.sherwood@arm.com>
7148 * target.def (empty_mask_is_expensive): New hook.
7149 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7150 * doc/tm.texi: Regenerate.
7151 * targhooks.h (default_empty_mask_is_expensive): Declare.
7152 * targhooks.c (default_empty_mask_is_expensive): New function.
7153 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7154 if the target says that empty masks are expensive.
7155 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7157 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
7159 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7160 Alan Hayward <alan.hayward@arm.com>
7161 David Sherwood <david.sherwood@arm.com>
7163 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7164 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7165 (vect_use_loop_mask_for_alignment_p): New function.
7166 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7167 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7168 niters_skip argument. Make sure that the first niters_skip elements
7169 of the first iteration are inactive.
7170 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7171 Update call to vect_set_loop_masks_directly.
7172 (get_misalign_in_elems): New function, split out from...
7173 (vect_gen_prolog_loop_niters): ...here.
7174 (vect_update_init_of_dr): Take a code argument that specifies whether
7175 the adjustment should be added or subtracted.
7176 (vect_update_init_of_drs): Likewise.
7177 (vect_prepare_for_masked_peels): New function.
7178 (vect_do_peeling): Skip prologue peeling if we're using a mask
7179 instead. Update call to vect_update_inits_of_drs.
7180 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7182 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7183 alignment. Do not include the number of peeled iterations in
7184 the minimum threshold in that case.
7185 (vectorizable_induction): Adjust the start value down by
7186 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7187 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7188 Take the number of skipped iterations into account when calculating
7190 * tree-vect-stmts.c (vect_gen_while_not): New function.
7192 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7193 Alan Hayward <alan.hayward@arm.com>
7194 David Sherwood <david.sherwood@arm.com>
7196 * doc/sourcebuild.texi (vect_fully_masked): Document.
7197 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7199 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7201 (vect_analyze_loop_2): ...here. Don't check the vectorization
7202 factor against the number of loop iterations if the loop is
7205 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7206 Alan Hayward <alan.hayward@arm.com>
7207 David Sherwood <david.sherwood@arm.com>
7209 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
7210 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
7211 (dump_groups): Update accordingly.
7212 (iv_use::mem_type): New member variable.
7213 (address_p): New function.
7214 (record_use): Add a mem_type argument and initialize the new
7216 (record_group_use): Add a mem_type argument. Use address_p.
7217 Remove obsolete null checks of base_object. Update call to record_use.
7218 (find_interesting_uses_op): Update call to record_group_use.
7219 (find_interesting_uses_cond): Likewise.
7220 (find_interesting_uses_address): Likewise.
7221 (get_mem_type_for_internal_fn): New function.
7222 (find_address_like_use): Likewise.
7223 (find_interesting_uses_stmt): Try find_address_like_use before
7224 calling find_interesting_uses_op.
7225 (addr_offset_valid_p): Use the iv mem_type field as the type
7226 of the addressed memory.
7227 (add_autoinc_candidates): Likewise.
7228 (get_address_cost): Likewise.
7229 (split_small_address_groups_p): Use address_p.
7230 (split_address_groups): Likewise.
7231 (add_iv_candidate_for_use): Likewise.
7232 (autoinc_possible_for_pair): Likewise.
7233 (rewrite_groups): Likewise.
7234 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
7235 (determine_group_iv_cost): Update after split of USE_ADDRESS.
7236 (get_alias_ptr_type_for_ptr_address): New function.
7237 (rewrite_use_address): Rewrite address uses in calls that were
7238 identified by find_address_like_use.
7240 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7241 Alan Hayward <alan.hayward@arm.com>
7242 David Sherwood <david.sherwood@arm.com>
7244 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7246 * gimple-expr.h (is_gimple_addressable: Likewise.
7247 * gimple-expr.c (is_gimple_address): Likewise.
7248 * internal-fn.c (expand_call_mem_ref): New function.
7249 (expand_mask_load_optab_fn): Use it.
7250 (expand_mask_store_optab_fn): Likewise.
7252 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7253 Alan Hayward <alan.hayward@arm.com>
7254 David Sherwood <david.sherwood@arm.com>
7256 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7257 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7258 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7259 (cond_umax@var{mode}): Document.
7260 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
7261 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
7262 (cond_umin_optab, cond_umax_optab): New optabs.
7263 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
7264 (COND_IOR, COND_XOR): New internal functions.
7265 * internal-fn.h (get_conditional_internal_fn): Declare.
7266 * internal-fn.c (cond_binary_direct): New macro.
7267 (expand_cond_binary_optab_fn): Likewise.
7268 (direct_cond_binary_optab_supported_p): Likewise.
7269 (get_conditional_internal_fn): New function.
7270 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7271 Cope with reduction statements that are vectorized as calls rather
7273 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7274 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7275 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7276 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7277 (UNSPEC_COND_EOR): New unspecs.
7278 (optab): Add mappings for them.
7279 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7280 (sve_int_op, sve_fp_op): New int attributes.
7282 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7283 Alan Hayward <alan.hayward@arm.com>
7284 David Sherwood <david.sherwood@arm.com>
7286 * optabs.def (while_ult_optab): New optab.
7287 * doc/md.texi (while_ult@var{m}@var{n}): Document.
7288 * internal-fn.def (WHILE_ULT): New internal function.
7289 * internal-fn.h (direct_internal_fn_supported_p): New override
7290 that takes two types as argument.
7291 * internal-fn.c (while_direct): New macro.
7292 (expand_while_optab_fn): New function.
7293 (convert_optab_supported_p): Likewise.
7294 (direct_while_optab_supported_p): New macro.
7295 * wide-int.h (wi::udiv_ceil): New function.
7296 * tree-vectorizer.h (rgroup_masks): New structure.
7297 (vec_loop_masks): New typedef.
7298 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7300 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7301 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7302 (vect_max_vf): New function.
7303 (slpeel_make_loop_iterate_ntimes): Delete.
7304 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7305 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7306 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7307 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7308 internal-fn.h, stor-layout.h and optabs-query.h.
7309 (vect_set_loop_mask): New function.
7310 (add_preheader_seq): Likewise.
7311 (add_header_seq): Likewise.
7312 (interleave_supported_p): Likewise.
7313 (vect_maybe_permute_loop_masks): Likewise.
7314 (vect_set_loop_masks_directly): Likewise.
7315 (vect_set_loop_condition_masked): Likewise.
7316 (vect_set_loop_condition_unmasked): New function, split out from
7317 slpeel_make_loop_iterate_ntimes.
7318 (slpeel_make_loop_iterate_ntimes): Rename to..
7319 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7320 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7321 (vect_do_peeling): Update call accordingly.
7322 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7324 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7325 mask_compare_type, can_fully_mask_p and fully_masked_p.
7326 (release_vec_loop_masks): New function.
7327 (_loop_vec_info): Use it to free the loop masks.
7328 (can_produce_all_loop_masks_p): New function.
7329 (vect_get_max_nscalars_per_iter): Likewise.
7330 (vect_verify_full_masking): Likewise.
7331 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7332 retries, and free the mask rgroups before retrying. Check loop-wide
7333 reasons for disallowing fully-masked loops. Make the final decision
7334 about whether use a fully-masked loop or not.
7335 (vect_estimate_min_profitable_iters): Do not assume that peeling
7336 for the number of iterations will be needed for fully-masked loops.
7337 (vectorizable_reduction): Disable fully-masked loops.
7338 (vectorizable_live_operation): Likewise.
7339 (vect_halve_mask_nunits): New function.
7340 (vect_double_mask_nunits): Likewise.
7341 (vect_record_loop_mask): Likewise.
7342 (vect_get_loop_mask): Likewise.
7343 (vect_transform_loop): Handle the case in which the final loop
7344 iteration might handle a partial vector. Call vect_set_loop_condition
7345 instead of slpeel_make_loop_iterate_ntimes.
7346 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7347 (check_load_store_masking): New function.
7348 (prepare_load_store_mask): Likewise.
7349 (vectorizable_store): Handle fully-masked loops.
7350 (vectorizable_load): Likewise.
7351 (supportable_widening_operation): Use vect_halve_mask_nunits for
7353 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7354 (vect_gen_while): New function.
7355 * config/aarch64/aarch64.md (umax<mode>3): New expander.
7356 (aarch64_uqdec<mode>): New insn.
7358 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7359 Alan Hayward <alan.hayward@arm.com>
7360 David Sherwood <david.sherwood@arm.com>
7362 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7363 (reduc_xor_scal_optab): New optabs.
7364 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7365 (reduc_xor_scal_@var{m}): Document.
7366 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7367 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7369 * fold-const-call.c (fold_const_call): Handle them.
7370 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7371 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7372 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7373 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7374 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7375 (UNSPEC_XORV): New unspecs.
7376 (optab): Add entries for them.
7377 (BITWISEV): New int iterator.
7378 (bit_reduc_op): New int attributes.
7380 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7381 Alan Hayward <alan.hayward@arm.com>
7382 David Sherwood <david.sherwood@arm.com>
7384 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7385 * internal-fn.def (VEC_SHL_INSERT): New internal function.
7386 * optabs.def (vec_shl_insert_optab): New optab.
7387 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7388 (duplicate_and_interleave): Likewise.
7389 * tree-vect-loop.c: Include internal-fn.h.
7390 (neutral_op_for_slp_reduction): New function, split out from
7391 get_initial_defs_for_reduction.
7392 (get_initial_def_for_reduction): Handle option 2 for variable-length
7393 vectors by loading the neutral value into a vector and then shifting
7394 the initial value into element 0.
7395 (get_initial_defs_for_reduction): Replace the code argument with
7396 the neutral value calculated by neutral_op_for_slp_reduction.
7397 Use gimple_build_vector for constant-length vectors.
7398 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7399 but the first group_size elements have a neutral value.
7400 Use duplicate_and_interleave otherwise.
7401 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7402 Update call to get_initial_defs_for_reduction. Handle SLP
7403 reductions for variable-length vectors by creating one vector
7404 result for each scalar result, with the elements associated
7405 with other scalar results stubbed out with the neutral value.
7406 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7407 Require IFN_VEC_SHL_INSERT for double reductions on
7408 variable-length vectors, or SLP reductions that have
7409 a neutral value. Require can_duplicate_and_interleave_p
7410 support for variable-length unchained SLP reductions if there
7411 is no neutral value, such as for MIN/MAX reductions. Also require
7412 the number of vector elements to be a multiple of the number of
7413 SLP statements when doing variable-length unchained SLP reductions.
7414 Update call to vect_create_epilog_for_reduction.
7415 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
7416 and remove initial values.
7417 (duplicate_and_interleave): Make public.
7418 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
7419 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
7421 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7422 Alan Hayward <alan.hayward@arm.com>
7423 David Sherwood <david.sherwood@arm.com>
7425 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
7426 (can_duplicate_and_interleave_p): New function.
7427 (vect_get_and_check_slp_defs): Take the vector of statements
7428 rather than just the current one. Remove excess parentheses.
7429 Restriction rejectinon of vect_constant_def and vect_external_def
7430 for variable-length vectors to boolean types, or types for which
7431 can_duplicate_and_interleave_p is false.
7432 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
7433 (duplicate_and_interleave): New function.
7434 (vect_get_constant_vectors): Use gimple_build_vector for
7435 constant-length vectors and suitable variable-length constant
7436 vectors. Use duplicate_and_interleave for other variable-length
7437 vectors. Don't defer the update when inserting new statements.
7439 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7440 Alan Hayward <alan.hayward@arm.com>
7441 David Sherwood <david.sherwood@arm.com>
7443 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
7444 min_profitable_iters doesn't go negative.
7446 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7447 Alan Hayward <alan.hayward@arm.com>
7448 David Sherwood <david.sherwood@arm.com>
7450 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
7451 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
7452 * optabs.def (vec_mask_load_lanes_optab): New optab.
7453 (vec_mask_store_lanes_optab): Likewise.
7454 * internal-fn.def (MASK_LOAD_LANES): New internal function.
7455 (MASK_STORE_LANES): Likewise.
7456 * internal-fn.c (mask_load_lanes_direct): New macro.
7457 (mask_store_lanes_direct): Likewise.
7458 (expand_mask_load_optab_fn): Handle masked operations.
7459 (expand_mask_load_lanes_optab_fn): New macro.
7460 (expand_mask_store_optab_fn): Handle masked operations.
7461 (expand_mask_store_lanes_optab_fn): New macro.
7462 (direct_mask_load_lanes_optab_supported_p): Likewise.
7463 (direct_mask_store_lanes_optab_supported_p): Likewise.
7464 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
7466 (vect_load_lanes_supported): Likewise.
7467 * tree-vect-data-refs.c (strip_conversion): New function.
7468 (can_group_stmts_p): Likewise.
7469 (vect_analyze_data_ref_accesses): Use it instead of checking
7470 for a pair of assignments.
7471 (vect_store_lanes_supported): Take a masked_p parameter.
7472 (vect_load_lanes_supported): Likewise.
7473 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
7474 vect_store_lanes_supported and vect_load_lanes_supported.
7475 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7476 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
7477 parameter. Don't allow gaps for masked accesses.
7478 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
7479 and vect_load_lanes_supported.
7480 (get_load_store_type): Take a masked_p parameter and update
7481 call to get_group_load_store_type.
7482 (vectorizable_store): Update call to get_load_store_type.
7483 Handle IFN_MASK_STORE_LANES.
7484 (vectorizable_load): Update call to get_load_store_type.
7485 Handle IFN_MASK_LOAD_LANES.
7487 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7488 Alan Hayward <alan.hayward@arm.com>
7489 David Sherwood <david.sherwood@arm.com>
7491 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
7493 * config/aarch64/aarch64-protos.h
7494 (aarch64_sve_struct_memory_operand_p): Declare.
7495 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
7496 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
7497 (VPRED, vpred): Handle SVE structure modes.
7498 * config/aarch64/constraints.md (Utx): New constraint.
7499 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
7500 (aarch64_sve_struct_nonimmediate_operand): New predicates.
7501 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
7502 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
7503 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
7504 structure modes. Split into pieces after RA.
7505 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
7506 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
7508 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
7509 SVE structure modes.
7510 (aarch64_classify_address): Likewise.
7511 (sizetochar): Move earlier in file.
7512 (aarch64_print_operand): Handle SVE register lists.
7513 (aarch64_array_mode): New function.
7514 (aarch64_sve_struct_memory_operand_p): Likewise.
7515 (TARGET_ARRAY_MODE): Redefine.
7517 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7518 Alan Hayward <alan.hayward@arm.com>
7519 David Sherwood <david.sherwood@arm.com>
7521 * target.def (array_mode): New target hook.
7522 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
7523 * doc/tm.texi: Regenerate.
7524 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
7525 * hooks.c (hook_optmode_mode_uhwi_none): New function.
7526 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
7528 * stor-layout.c (mode_for_array): Likewise. Support polynomial
7531 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7532 Alan Hayward <alan.hayward@arm.com>
7533 David Sherwood <david.sherwood@arm.com>
7535 * fold-const.c (fold_binary_loc): Check the argument types
7536 rather than the result type when testing for a vector operation.
7538 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7540 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
7541 * doc/tm.texi: Regenerate.
7543 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7544 Alan Hayward <alan.hayward@arm.com>
7545 David Sherwood <david.sherwood@arm.com>
7547 * doc/invoke.texi (-msve-vector-bits=): Document new option.
7548 (sve): Document new AArch64 extension.
7549 * doc/md.texi (w): Extend the description of the AArch64
7550 constraint to include SVE vectors.
7551 (Upl, Upa): Document new AArch64 predicate constraints.
7552 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
7554 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
7555 (msve-vector-bits=): New option.
7556 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
7557 SVE when these are disabled.
7558 (sve): New extension.
7559 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
7560 modes. Adjust their number of units based on aarch64_sve_vg.
7561 (MAX_BITSIZE_MODE_ANY_MODE): Define.
7562 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
7563 aarch64_addr_query_type.
7564 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
7565 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
7566 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
7567 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
7568 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
7569 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
7570 (aarch64_simd_imm_zero_p): Delete.
7571 (aarch64_check_zero_based_sve_index_immediate): Declare.
7572 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7573 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7574 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7575 (aarch64_sve_float_mul_immediate_p): Likewise.
7576 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7578 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
7579 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
7580 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
7581 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
7582 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
7583 (aarch64_regmode_natural_size): Likewise.
7584 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
7585 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
7587 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
7588 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
7589 for VG and the SVE predicate registers.
7590 (V_ALIASES): Add a "z"-prefixed alias.
7591 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
7592 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
7593 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
7594 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
7595 (REG_CLASS_NAMES): Add entries for them.
7596 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
7597 and the predicate registers.
7598 (aarch64_sve_vg): Declare.
7599 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
7600 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
7601 (REGMODE_NATURAL_SIZE): Define.
7602 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
7604 * config/aarch64/aarch64.c: Include cfgrtl.h.
7605 (simd_immediate_info): Add a constructor for series vectors,
7606 and an associated step field.
7607 (aarch64_sve_vg): New variable.
7608 (aarch64_dbx_register_number): Handle VG and the predicate registers.
7609 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
7610 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
7611 (VEC_ANY_DATA, VEC_STRUCT): New constants.
7612 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
7613 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
7614 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
7615 (aarch64_get_mask_mode): New functions.
7616 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
7617 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
7618 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
7619 predicate modes and predicate registers. Explicitly restrict
7620 GPRs to modes of 16 bytes or smaller. Only allow FP registers
7621 to store a vector mode if it is recognized by
7622 aarch64_classify_vector_mode.
7623 (aarch64_regmode_natural_size): New function.
7624 (aarch64_hard_regno_caller_save_mode): Return the original mode
7626 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
7627 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
7628 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
7629 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
7631 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
7632 does not overlap dest if the function is frame-related. Handle
7634 (aarch64_split_add_offset): New function.
7635 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
7636 them aarch64_add_offset.
7637 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
7638 and update call to aarch64_sub_sp.
7639 (aarch64_add_cfa_expression): New function.
7640 (aarch64_expand_prologue): Pass extra temporary registers to the
7641 functions above. Handle the case in which we need to emit new
7642 DW_CFA_expressions for registers that were originally saved
7643 relative to the stack pointer, but now have to be expressed
7644 relative to the frame pointer.
7645 (aarch64_output_mi_thunk): Pass extra temporary registers to the
7647 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
7648 IP0 and IP1 values for SVE frames.
7649 (aarch64_expand_vec_series): New function.
7650 (aarch64_expand_sve_widened_duplicate): Likewise.
7651 (aarch64_expand_sve_const_vector): Likewise.
7652 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
7653 Handle SVE constants. Use emit_move_insn to move a force_const_mem
7654 into the register, rather than emitting a SET directly.
7655 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
7656 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
7657 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
7658 (offset_9bit_signed_scaled_p): New functions.
7659 (aarch64_replicate_bitmask_imm): New function.
7660 (aarch64_bitmask_imm): Use it.
7661 (aarch64_cannot_force_const_mem): Reject expressions involving
7662 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
7663 (aarch64_classify_index): Handle SVE indices, by requiring
7664 a plain register index with a scale that matches the element size.
7665 (aarch64_classify_address): Handle SVE addresses. Assert that
7666 the mode of the address is VOIDmode or an integer mode.
7667 Update call to aarch64_classify_symbol.
7668 (aarch64_classify_symbolic_expression): Update call to
7669 aarch64_classify_symbol.
7670 (aarch64_const_vec_all_in_range_p): New function.
7671 (aarch64_print_vector_float_operand): Likewise.
7672 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
7673 "vN" for FP registers with SVE modes. Handle (const ...) vectors
7674 and the FP immediates 1.0 and 0.5.
7675 (aarch64_print_address_internal): Handle SVE addresses.
7676 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
7677 (aarch64_regno_regclass): Handle predicate registers.
7678 (aarch64_secondary_reload): Handle big-endian reloads of SVE
7680 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
7681 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
7682 (aarch64_convert_sve_vector_bits): New function.
7683 (aarch64_override_options): Use it to handle -msve-vector-bits=.
7684 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7686 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
7687 Handle SVE vector and predicate modes. Accept VL-based constants
7688 that need only one temporary register, and VL offsets that require
7689 no temporary registers.
7690 (aarch64_conditional_register_usage): Mark the predicate registers
7691 as fixed if SVE isn't available.
7692 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
7693 Return true for SVE vector and predicate modes.
7694 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
7695 rather than an unsigned int. Handle SVE modes.
7696 (aarch64_preferred_simd_mode): Update call accordingly. Handle
7698 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
7700 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7701 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7702 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7703 (aarch64_sve_float_mul_immediate_p): New functions.
7704 (aarch64_sve_valid_immediate): New function.
7705 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
7706 Explicitly reject structure modes. Check for INDEX constants.
7707 Handle PTRUE and PFALSE constants.
7708 (aarch64_check_zero_based_sve_index_immediate): New function.
7709 (aarch64_simd_imm_zero_p): Delete.
7710 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
7711 vector modes. Accept constants in the range of CNT[BHWD].
7712 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
7713 ask for an Advanced SIMD mode.
7714 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
7715 (aarch64_simd_vector_alignment): Handle SVE predicates.
7716 (aarch64_vectorize_preferred_vector_alignment): New function.
7717 (aarch64_simd_vector_alignment_reachable): Use it instead of
7719 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
7720 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
7722 (MAX_VECT_LEN): Delete.
7723 (expand_vec_perm_d): Add a vec_flags field.
7724 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
7725 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
7726 (aarch64_evpc_ext): Don't apply a big-endian lane correction
7728 (aarch64_evpc_rev): Rename to...
7729 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
7730 (aarch64_evpc_rev_global): New function.
7731 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
7732 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
7734 (aarch64_evpc_sve_tbl): New function.
7735 (aarch64_expand_vec_perm_const_1): Update after rename of
7736 aarch64_evpc_rev. Handle SVE permutes too, trying
7737 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
7738 than aarch64_evpc_tbl.
7739 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
7740 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
7741 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
7742 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
7743 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
7744 (aarch64_expand_sve_vcond): New functions.
7745 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
7746 of aarch64_vector_mode_p.
7747 (aarch64_dwarf_poly_indeterminate_value): New function.
7748 (aarch64_compute_pressure_classes): Likewise.
7749 (aarch64_can_change_mode_class): Likewise.
7750 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
7751 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
7752 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
7753 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
7754 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
7755 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
7756 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
7757 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
7759 (Dn, Dl, Dr): Accept const as well as const_vector.
7760 (Dz): Likewise. Compare against CONST0_RTX.
7761 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
7762 of "vector" where appropriate.
7763 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
7764 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
7765 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
7766 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
7767 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
7768 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
7769 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
7770 (v_int_equiv): Extend to SVE modes.
7771 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
7773 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
7774 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
7775 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
7776 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
7777 (SVE_COND_FP_CMP): New int iterators.
7778 (perm_hilo): Handle the new unpack unspecs.
7779 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
7781 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
7782 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
7783 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
7784 (aarch64_equality_operator, aarch64_constant_vector_operand)
7785 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
7786 (aarch64_sve_nonimmediate_operand): Likewise.
7787 (aarch64_sve_general_operand): Likewise.
7788 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
7789 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
7790 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
7791 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
7792 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
7793 (aarch64_sve_float_arith_immediate): Likewise.
7794 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
7795 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
7796 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
7797 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
7798 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
7799 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
7800 (aarch64_sve_float_arith_operand): Likewise.
7801 (aarch64_sve_float_arith_with_sub_operand): Likewise.
7802 (aarch64_sve_float_mul_operand): Likewise.
7803 (aarch64_sve_vec_perm_operand): Likewise.
7804 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
7805 (aarch64_mov_operand): Accept const_poly_int and const_vector.
7806 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
7807 as well as const_vector.
7808 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
7809 in file. Use CONST0_RTX and CONSTM1_RTX.
7810 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
7811 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
7812 Use aarch64_simd_imm_zero.
7813 * config/aarch64/aarch64-sve.md: New file.
7814 * config/aarch64/aarch64.md: Include it.
7815 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
7816 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
7817 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
7818 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
7819 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
7820 (sve): New attribute.
7821 (enabled): Disable instructions with the sve attribute unless
7823 (movqi, movhi): Pass CONST_POLY_INT operaneds through
7824 aarch64_expand_mov_immediate.
7825 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
7826 CNT[BHSD] immediates.
7827 (movti): Split CONST_POLY_INT moves into two halves.
7828 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
7829 Split additions that need a temporary here if the destination
7830 is the stack pointer.
7831 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
7832 (*add<mode>3_poly_1): New instruction.
7833 (set_clobber_cc): New expander.
7835 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7837 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
7838 parameter and use it instead of GET_MODE_SIZE (innermode). Use
7839 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
7840 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
7841 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
7842 Change innermode from fixed_mode_size to machine_mode.
7843 (simplify_subreg): Update call accordingly. Handle a constant-sized
7844 subreg of a variable-length CONST_VECTOR.
7846 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7847 Alan Hayward <alan.hayward@arm.com>
7848 David Sherwood <david.sherwood@arm.com>
7850 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
7851 (add_offset_to_base): New function, split out from...
7852 (create_mem_ref): ...here. When handling a scale other than 1,
7853 check first whether the address is valid without the offset.
7854 Add it into the base if so, leaving the index and scale as-is.
7856 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7859 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
7860 fold_for_warn before checking if arg2 is INTEGER_CST.
7862 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
7864 * config/rs6000/predicates.md (load_multiple_operation): Delete.
7865 (store_multiple_operation): Delete.
7866 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
7867 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
7868 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
7869 guarded by TARGET_STRING.
7870 (rs6000_output_load_multiple): Delete.
7871 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
7872 OPTION_MASK_STRING / TARGET_STRING handling.
7873 (print_operand) <'N', 'O'>: Add comment that these are unused now.
7874 (const rs6000_opt_masks) <"string">: Change mask to 0.
7875 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
7876 (MASK_STRING): Delete.
7877 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
7879 (load_multiple): Delete.
7886 (store_multiple): Delete.
7893 (movmemsi_8reg): Delete.
7894 (corresponding unnamed define_insn): Delete.
7895 (movmemsi_6reg): Delete.
7896 (corresponding unnamed define_insn): Delete.
7897 (movmemsi_4reg): Delete.
7898 (corresponding unnamed define_insn): Delete.
7899 (movmemsi_2reg): Delete.
7900 (corresponding unnamed define_insn): Delete.
7901 (movmemsi_1reg): Delete.
7902 (corresponding unnamed define_insn): Delete.
7903 * config/rs6000/rs6000.opt (mno-string): New.
7904 (mstring): Replace by deprecation warning stub.
7905 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
7907 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7909 * regrename.c (regrename_do_replace): If replacing the same
7910 reg multiple times, try to reuse last created gen_raw_REG.
7913 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
7914 main to workaround a bug in GDB.
7916 2018-01-12 Tom de Vries <tom@codesourcery.com>
7919 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
7921 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
7923 PR rtl-optimization/80481
7924 * ira-color.c (get_cap_member): New function.
7925 (allocnos_conflict_by_live_ranges_p): Use it.
7926 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
7927 (setup_slot_coalesced_allocno_live_ranges): Ditto.
7929 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
7932 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
7933 (*saddl_se_1): Ditto.
7935 (*ssubl_se_1): Ditto.
7937 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7939 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
7940 rather than wi::to_widest for DR_INITs.
7941 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
7942 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
7943 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
7945 (vect_analyze_group_access_1): Note that here.
7947 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7949 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
7950 polynomial type sizes.
7952 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7954 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
7955 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
7956 (gimple_add_tmp_var): Likewise.
7958 2018-01-12 Martin Liska <mliska@suse.cz>
7960 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
7961 (gimple_alloc_sizes): Likewise.
7962 (dump_gimple_statistics): Use PRIu64 in printf format.
7963 * gimple.h: Change uint64_t to int.
7965 2018-01-12 Martin Liska <mliska@suse.cz>
7967 * tree-core.h: Use uint64_t instead of int.
7968 * tree.c (tree_node_counts): Likewise.
7969 (tree_node_sizes): Likewise.
7970 (dump_tree_statistics): Use PRIu64 in printf format.
7972 2018-01-12 Martin Liska <mliska@suse.cz>
7974 * Makefile.in: As qsort_chk is implemented in vec.c, add
7975 vec.o to linkage of gencfn-macros.
7976 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
7977 passing the info to record_node_allocation_statistics.
7978 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
7980 * ggc-common.c (struct ggc_usage): Add operator== and use
7981 it in operator< and compare function.
7982 * mem-stats.h (struct mem_usage): Likewise.
7983 * vec.c (struct vec_usage): Remove operator< and compare
7984 function. Can be simply inherited.
7986 2018-01-12 Martin Jambor <mjambor@suse.cz>
7989 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
7990 * tree-ssa-math-opts.c: Include domwalk.h.
7991 (convert_mult_to_fma_1): New function.
7992 (fma_transformation_info): New type.
7993 (fma_deferring_state): Likewise.
7994 (cancel_fma_deferring): New function.
7995 (result_of_phi): Likewise.
7996 (last_fma_candidate_feeds_initial_phi): Likewise.
7997 (convert_mult_to_fma): Added deferring logic, split actual
7998 transformation to convert_mult_to_fma_1.
7999 (math_opts_dom_walker): New type.
8000 (math_opts_dom_walker::after_dom_children): New method, body moved
8001 here from pass_optimize_widening_mul::execute, added deferring logic
8003 (pass_optimize_widening_mul::execute): Moved most of code to
8004 math_opts_dom_walker::after_dom_children.
8005 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8006 * config/i386/i386.c (ix86_option_override_internal): Added
8007 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8009 2018-01-12 Richard Biener <rguenther@suse.de>
8012 * dwarf2out.c (gen_variable_die): Do not reset old_die for
8013 inline instance vars.
8015 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
8018 * config/rx/rx.c (rx_is_restricted_memory_address):
8021 2018-01-12 Richard Biener <rguenther@suse.de>
8023 PR tree-optimization/80846
8024 * target.def (split_reduction): New target hook.
8025 * targhooks.c (default_split_reduction): New function.
8026 * targhooks.h (default_split_reduction): Declare.
8027 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8028 target requests first reduce vectors by combining low and high
8030 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8031 (get_vectype_for_scalar_type_and_size): Export.
8032 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8033 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8034 * doc/tm.texi: Regenerate.
8035 * config/i386/i386.c (ix86_split_reduction): Implement
8036 TARGET_VECTORIZE_SPLIT_REDUCTION.
8038 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8041 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8042 in PIC mode except for TARGET_VXWORKS_RTP.
8043 * config/sparc/sparc.c: Include cfgrtl.h.
8044 (TARGET_INIT_PIC_REG): Define.
8045 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8046 (sparc_pic_register_p): New predicate.
8047 (sparc_legitimate_address_p): Use it.
8048 (sparc_legitimize_pic_address): Likewise.
8049 (sparc_delegitimize_address): Likewise.
8050 (sparc_mode_dependent_address_p): Likewise.
8051 (gen_load_pcrel_sym): Remove 4th parameter.
8052 (load_got_register): Adjust call to above. Remove obsolete stuff.
8053 (sparc_expand_prologue): Do not call load_got_register here.
8054 (sparc_flat_expand_prologue): Likewise.
8055 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8056 (sparc_use_pseudo_pic_reg): New function.
8057 (sparc_init_pic_reg): Likewise.
8058 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8059 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8061 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8063 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8064 Add item for branch_cost.
8066 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8068 PR rtl-optimization/83565
8069 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8070 not extend the result to a larger mode for rotate operations.
8071 (num_sign_bit_copies1): Likewise.
8073 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8076 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8078 Use values-Xc.o for -pedantic.
8079 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8081 2018-01-12 Martin Liska <mliska@suse.cz>
8084 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8086 (possible_polymorphic_call_targets): Use it.
8087 (ipa_devirt): Likewise.
8089 2018-01-12 Martin Liska <mliska@suse.cz>
8091 * profile-count.h (enum profile_quality): Use 0 as invalid
8092 enum value of profile_quality.
8094 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8096 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8097 -mext-string options.
8099 2018-01-12 Richard Biener <rguenther@suse.de>
8101 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8102 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8103 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8105 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8107 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8109 * configure.ac (--with-long-double-format): Add support for the
8110 configuration option to change the default long double format on
8112 * config.gcc (powerpc*-linux*-*): Likewise.
8113 * configure: Regenerate.
8114 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8115 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8116 used without modification.
8118 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8120 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8121 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8122 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8123 MISC_BUILTIN_SPEC_BARRIER.
8124 (rs6000_init_builtins): Likewise.
8125 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8127 (speculation_barrier): New define_insn.
8128 * doc/extend.texi: Document __builtin_speculation_barrier.
8130 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8133 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8134 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8135 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8137 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8138 integral modes instead of "ss" and "sd".
8139 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8140 vectors with 32-bit and 64-bit elements.
8141 (vecdupssescalarmodesuffix): New mode attribute.
8142 (vec_dup<mode>): Use it.
8144 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8147 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8148 frame if argument is passed on stack.
8150 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8153 * ree.c (combine_reaching_defs): Optimize also
8154 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8155 reg2=any_extend(exp); reg1=reg2;, formatting fix.
8157 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8160 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8162 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8165 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8166 after they are computed.
8168 2018-01-11 Bin Cheng <bin.cheng@arm.com>
8170 PR tree-optimization/83695
8171 * gimple-loop-linterchange.cc
8172 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8173 reset cached scev information after interchange.
8174 (pass_linterchange::execute): Remove call to scev_reset_htab.
8176 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8178 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8179 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8180 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8181 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8182 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8183 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8184 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8185 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8186 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8187 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8188 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8189 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8190 (V_lane_reg): Likewise.
8191 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8193 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8194 (vfmal_lane_low<mode>_intrinsic,
8195 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8196 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8197 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8198 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8199 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8200 vfmsl_lane_high<mode>_intrinsic): New define_insns.
8202 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8204 * config/arm/arm-cpus.in (fp16fml): New feature.
8205 (ALL_SIMD): Add fp16fml.
8206 (armv8.2-a): Add fp16fml as an option.
8207 (armv8.3-a): Likewise.
8208 (armv8.4-a): Add fp16fml as part of fp16.
8209 * config/arm/arm.h (TARGET_FP16FML): Define.
8210 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
8212 * config/arm/arm-modes.def (V2HF): Define.
8213 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
8214 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
8215 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
8216 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
8217 vfmsl_low, vfmsl_high): New set of builtins.
8218 * config/arm/iterators.md (PLUSMINUS): New code iterator.
8219 (vfml_op): New code attribute.
8220 (VFMLHALVES): New int iterator.
8221 (VFML, VFMLSEL): New mode attributes.
8222 (V_reg): Define mapping for V2HF.
8223 (V_hi, V_lo): New mode attributes.
8224 (VF_constraint): Likewise.
8225 (vfml_half, vfml_half_selector): New int attributes.
8226 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
8228 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
8229 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
8231 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
8232 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
8233 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
8234 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
8236 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8237 Document new effective target and option set.
8239 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8241 * config/arm/arm-cpus.in (armv8_4): New feature.
8242 (ARMv8_4a): New fgroup.
8243 (armv8.4-a): New arch.
8244 * config/arm/arm-tables.opt: Regenerate.
8245 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8246 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8247 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8248 Add matching rules for -march=armv8.4-a and extensions.
8249 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8251 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
8254 * config/rx/rx.md (BW): New mode attribute.
8255 (sync_lock_test_and_setsi): Add mode suffix to insn output.
8257 2018-01-11 Richard Biener <rguenther@suse.de>
8259 PR tree-optimization/83435
8260 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
8261 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
8262 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
8264 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8265 Alan Hayward <alan.hayward@arm.com>
8266 David Sherwood <david.sherwood@arm.com>
8268 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8270 (aarch64_classify_address): Initialize it. Track polynomial offsets.
8271 (aarch64_print_address_internal): Use it to check for a zero offset.
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-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8278 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8279 Return a poly_int64 rather than a HOST_WIDE_INT.
8280 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8281 rather than a HOST_WIDE_INT.
8282 * config/aarch64/aarch64.h (aarch64_frame): Protect with
8283 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
8284 hard_fp_offset, frame_size, initial_adjust, callee_offset and
8285 final_offset from HOST_WIDE_INT to poly_int64.
8286 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8287 to_constant when getting the number of units in an Advanced SIMD
8289 (aarch64_builtin_vectorized_function): Check for a constant number
8291 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8293 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8294 attribute instead of GET_MODE_NUNITS.
8295 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8296 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8297 GET_MODE_SIZE for fixed-size registers.
8298 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8299 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8300 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8301 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8302 (aarch64_print_operand, aarch64_print_address_internal)
8303 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8304 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8305 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8306 Handle polynomial GET_MODE_SIZE.
8307 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8308 wider than SImode without modification.
8309 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8310 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8311 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8312 passing and returning SVE modes.
8313 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8314 rather than GEN_INT.
8315 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8316 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8317 (aarch64_allocate_and_probe_stack_space): Likewise.
8318 (aarch64_layout_frame): Cope with polynomial offsets.
8319 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8320 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8322 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8323 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8324 poly_int64 rather than a HOST_WIDE_INT.
8325 (aarch64_get_separate_components, aarch64_process_components)
8326 (aarch64_expand_prologue, aarch64_expand_epilogue)
8327 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8328 (aarch64_anchor_offset): New function, split out from...
8329 (aarch64_legitimize_address): ...here.
8330 (aarch64_builtin_vectorization_cost): Handle polynomial
8331 TYPE_VECTOR_SUBPARTS.
8332 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8334 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8335 number of elements from the PARALLEL rather than the mode.
8336 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8337 rather than GET_MODE_BITSIZE.
8338 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8339 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8340 (aarch64_expand_vec_perm_const_1): Handle polynomial
8341 d->perm.length () and d->perm elements.
8342 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
8343 Apply to_constant to d->perm elements.
8344 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8345 polynomial CONST_VECTOR_NUNITS.
8346 (aarch64_move_pointer): Take amount as a poly_int64 rather
8348 (aarch64_progress_pointer): Avoid temporary variable.
8349 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8350 the mode attribute instead of GET_MODE.
8352 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8353 Alan Hayward <alan.hayward@arm.com>
8354 David Sherwood <david.sherwood@arm.com>
8356 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8357 x exists before using it.
8358 (aarch64_add_constant_internal): Rename to...
8359 (aarch64_add_offset_1): ...this. Replace regnum with separate
8360 src and dest rtxes. Handle the case in which they're different,
8361 including when the offset is zero. Replace scratchreg with an rtx.
8362 Use 2 additions if there is no spare register into which we can
8363 move a 16-bit constant.
8364 (aarch64_add_constant): Delete.
8365 (aarch64_add_offset): Replace reg with separate src and dest
8366 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8367 Use aarch64_add_offset_1.
8368 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8369 an rtx rather than an int. Take the delta as a poly_int64
8370 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8371 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8372 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8373 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8374 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8376 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8377 aarch64_add_constant.
8379 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8381 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8382 Use scalar_float_mode.
8384 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8386 * config/aarch64/aarch64-simd.md
8387 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8388 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8389 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8390 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8391 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8392 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8393 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8394 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8395 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8396 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8398 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8401 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8402 targ_options->x_arm_arch_string is non NULL.
8404 2018-01-11 Tamar Christina <tamar.christina@arm.com>
8406 * config/aarch64/aarch64.h
8407 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
8409 2018-01-11 Sudakshina Das <sudi.das@arm.com>
8412 * expmed.c (emit_store_flag_force): Swap if const op0
8413 and change VOIDmode to mode of op0.
8415 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8417 PR rtl-optimization/83761
8418 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
8419 than bytes to mode_for_size.
8421 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8424 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
8425 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
8428 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8431 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
8432 when in layout mode.
8433 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
8434 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
8437 2018-01-10 Michael Collison <michael.collison@arm.com>
8439 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
8440 * config/aarch64/aarch64-option-extension.def: Add
8441 AARCH64_OPT_EXTENSION of 'fp16fml'.
8442 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8443 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
8444 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
8445 * config/aarch64/constraints.md (Ui7): New constraint.
8446 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
8447 (VFMLA_SEL_W): Ditto.
8450 (VFMLA16_LOW): New int iterator.
8451 (VFMLA16_HIGH): Ditto.
8452 (UNSPEC_FMLAL): New unspec.
8453 (UNSPEC_FMLSL): Ditto.
8454 (UNSPEC_FMLAL2): Ditto.
8455 (UNSPEC_FMLSL2): Ditto.
8456 (f16mac): New code attribute.
8457 * config/aarch64/aarch64-simd-builtins.def
8458 (aarch64_fmlal_lowv2sf): Ditto.
8459 (aarch64_fmlsl_lowv2sf): Ditto.
8460 (aarch64_fmlalq_lowv4sf): Ditto.
8461 (aarch64_fmlslq_lowv4sf): Ditto.
8462 (aarch64_fmlal_highv2sf): Ditto.
8463 (aarch64_fmlsl_highv2sf): Ditto.
8464 (aarch64_fmlalq_highv4sf): Ditto.
8465 (aarch64_fmlslq_highv4sf): Ditto.
8466 (aarch64_fmlal_lane_lowv2sf): Ditto.
8467 (aarch64_fmlsl_lane_lowv2sf): Ditto.
8468 (aarch64_fmlal_laneq_lowv2sf): Ditto.
8469 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
8470 (aarch64_fmlalq_lane_lowv4sf): Ditto.
8471 (aarch64_fmlsl_lane_lowv4sf): Ditto.
8472 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
8473 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
8474 (aarch64_fmlal_lane_highv2sf): Ditto.
8475 (aarch64_fmlsl_lane_highv2sf): Ditto.
8476 (aarch64_fmlal_laneq_highv2sf): Ditto.
8477 (aarch64_fmlsl_laneq_highv2sf): Ditto.
8478 (aarch64_fmlalq_lane_highv4sf): Ditto.
8479 (aarch64_fmlsl_lane_highv4sf): Ditto.
8480 (aarch64_fmlalq_laneq_highv4sf): Ditto.
8481 (aarch64_fmlsl_laneq_highv4sf): Ditto.
8482 * config/aarch64/aarch64-simd.md:
8483 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
8484 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8485 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
8486 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8487 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
8488 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
8489 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
8490 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
8491 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
8492 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
8493 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
8494 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
8495 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
8496 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
8497 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
8498 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
8499 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
8500 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
8501 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
8502 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
8503 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
8504 (vfmlsl_low_u32): Ditto.
8505 (vfmlalq_low_u32): Ditto.
8506 (vfmlslq_low_u32): Ditto.
8507 (vfmlal_high_u32): Ditto.
8508 (vfmlsl_high_u32): Ditto.
8509 (vfmlalq_high_u32): Ditto.
8510 (vfmlslq_high_u32): Ditto.
8511 (vfmlal_lane_low_u32): Ditto.
8512 (vfmlsl_lane_low_u32): Ditto.
8513 (vfmlal_laneq_low_u32): Ditto.
8514 (vfmlsl_laneq_low_u32): Ditto.
8515 (vfmlalq_lane_low_u32): Ditto.
8516 (vfmlslq_lane_low_u32): Ditto.
8517 (vfmlalq_laneq_low_u32): Ditto.
8518 (vfmlslq_laneq_low_u32): Ditto.
8519 (vfmlal_lane_high_u32): Ditto.
8520 (vfmlsl_lane_high_u32): Ditto.
8521 (vfmlal_laneq_high_u32): Ditto.
8522 (vfmlsl_laneq_high_u32): Ditto.
8523 (vfmlalq_lane_high_u32): Ditto.
8524 (vfmlslq_lane_high_u32): Ditto.
8525 (vfmlalq_laneq_high_u32): Ditto.
8526 (vfmlslq_laneq_high_u32): Ditto.
8527 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
8528 (AARCH64_FL_FOR_ARCH8_4): New.
8529 (AARCH64_ISA_F16FML): New ISA flag.
8530 (TARGET_F16FML): New feature flag for fp16fml.
8531 (doc/invoke.texi): Document new fp16fml option.
8533 2018-01-10 Michael Collison <michael.collison@arm.com>
8535 * config/aarch64/aarch64-builtins.c:
8536 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
8537 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8538 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
8539 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
8540 (AARCH64_ISA_SHA3): New ISA flag.
8541 (TARGET_SHA3): New feature flag for sha3.
8542 * config/aarch64/iterators.md (sha512_op): New int attribute.
8543 (CRYPTO_SHA512): New int iterator.
8544 (UNSPEC_SHA512H): New unspec.
8545 (UNSPEC_SHA512H2): Ditto.
8546 (UNSPEC_SHA512SU0): Ditto.
8547 (UNSPEC_SHA512SU1): Ditto.
8548 * config/aarch64/aarch64-simd-builtins.def
8549 (aarch64_crypto_sha512hqv2di): New builtin.
8550 (aarch64_crypto_sha512h2qv2di): Ditto.
8551 (aarch64_crypto_sha512su0qv2di): Ditto.
8552 (aarch64_crypto_sha512su1qv2di): Ditto.
8553 (aarch64_eor3qv8hi): Ditto.
8554 (aarch64_rax1qv2di): Ditto.
8555 (aarch64_xarqv2di): Ditto.
8556 (aarch64_bcaxqv8hi): Ditto.
8557 * config/aarch64/aarch64-simd.md:
8558 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
8559 (aarch64_crypto_sha512su0qv2di): Ditto.
8560 (aarch64_crypto_sha512su1qv2di): Ditto.
8561 (aarch64_eor3qv8hi): Ditto.
8562 (aarch64_rax1qv2di): Ditto.
8563 (aarch64_xarqv2di): Ditto.
8564 (aarch64_bcaxqv8hi): Ditto.
8565 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
8566 (vsha512h2q_u64): Ditto.
8567 (vsha512su0q_u64): Ditto.
8568 (vsha512su1q_u64): Ditto.
8569 (veor3q_u16): Ditto.
8570 (vrax1q_u64): Ditto.
8572 (vbcaxq_u16): Ditto.
8573 * config/arm/types.md (crypto_sha512): New type attribute.
8574 (crypto_sha3): Ditto.
8575 (doc/invoke.texi): Document new sha3 option.
8577 2018-01-10 Michael Collison <michael.collison@arm.com>
8579 * config/aarch64/aarch64-builtins.c:
8580 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
8581 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8582 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
8583 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
8584 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
8585 (AARCH64_ISA_SM4): New ISA flag.
8586 (TARGET_SM4): New feature flag for sm4.
8587 * config/aarch64/aarch64-simd-builtins.def
8588 (aarch64_sm3ss1qv4si): Ditto.
8589 (aarch64_sm3tt1aq4si): Ditto.
8590 (aarch64_sm3tt1bq4si): Ditto.
8591 (aarch64_sm3tt2aq4si): Ditto.
8592 (aarch64_sm3tt2bq4si): Ditto.
8593 (aarch64_sm3partw1qv4si): Ditto.
8594 (aarch64_sm3partw2qv4si): Ditto.
8595 (aarch64_sm4eqv4si): Ditto.
8596 (aarch64_sm4ekeyqv4si): Ditto.
8597 * config/aarch64/aarch64-simd.md:
8598 (aarch64_sm3ss1qv4si): Ditto.
8599 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
8600 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
8601 (aarch64_sm4eqv4si): Ditto.
8602 (aarch64_sm4ekeyqv4si): Ditto.
8603 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
8604 (sm3part_op): Ditto.
8605 (CRYPTO_SM3TT): Ditto.
8606 (CRYPTO_SM3PART): Ditto.
8607 (UNSPEC_SM3SS1): New unspec.
8608 (UNSPEC_SM3TT1A): Ditto.
8609 (UNSPEC_SM3TT1B): Ditto.
8610 (UNSPEC_SM3TT2A): Ditto.
8611 (UNSPEC_SM3TT2B): Ditto.
8612 (UNSPEC_SM3PARTW1): Ditto.
8613 (UNSPEC_SM3PARTW2): Ditto.
8614 (UNSPEC_SM4E): Ditto.
8615 (UNSPEC_SM4EKEY): Ditto.
8616 * config/aarch64/constraints.md (Ui2): New constraint.
8617 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
8618 * config/arm/types.md (crypto_sm3): New type attribute.
8619 (crypto_sm4): Ditto.
8620 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
8621 (vsm3tt1aq_u32): Ditto.
8622 (vsm3tt1bq_u32): Ditto.
8623 (vsm3tt2aq_u32): Ditto.
8624 (vsm3tt2bq_u32): Ditto.
8625 (vsm3partw1q_u32): Ditto.
8626 (vsm3partw2q_u32): Ditto.
8627 (vsm4eq_u32): Ditto.
8628 (vsm4ekeyq_u32): Ditto.
8629 (doc/invoke.texi): Document new sm4 option.
8631 2018-01-10 Michael Collison <michael.collison@arm.com>
8633 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
8634 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
8635 (AARCH64_FL_FOR_ARCH8_4): New.
8636 (AARCH64_FL_V8_4): New flag.
8637 (doc/invoke.texi): Document new armv8.4-a option.
8639 2018-01-10 Michael Collison <michael.collison@arm.com>
8641 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8642 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
8643 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
8644 * config/aarch64/aarch64-option-extension.def: Add
8645 AARCH64_OPT_EXTENSION of 'sha2'.
8646 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
8647 (crypto): Disable sha2 and aes if crypto disabled.
8648 (crypto): Enable aes and sha2 if enabled.
8649 (simd): Disable sha2 and aes if simd disabled.
8650 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
8652 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
8653 (TARGET_SHA2): New feature flag for sha2.
8654 (TARGET_AES): New feature flag for aes.
8655 * config/aarch64/aarch64-simd.md:
8656 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
8657 conditional on TARGET_AES.
8658 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
8659 (aarch64_crypto_sha1hsi): Make pattern conditional
8661 (aarch64_crypto_sha1hv4si): Ditto.
8662 (aarch64_be_crypto_sha1hv4si): Ditto.
8663 (aarch64_crypto_sha1su1v4si): Ditto.
8664 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
8665 (aarch64_crypto_sha1su0v4si): Ditto.
8666 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
8667 (aarch64_crypto_sha256su0v4si): Ditto.
8668 (aarch64_crypto_sha256su1v4si): Ditto.
8669 (doc/invoke.texi): Document new aes and sha2 options.
8671 2018-01-10 Martin Sebor <msebor@redhat.com>
8673 PR tree-optimization/83781
8674 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
8677 2018-01-11 Martin Sebor <msebor@gmail.com>
8678 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8680 PR tree-optimization/83501
8681 PR tree-optimization/81703
8683 * tree-ssa-strlen.c (get_string_cst): Rename...
8684 (get_string_len): ...to this. Handle global constants.
8685 (handle_char_store): Adjust.
8687 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
8688 Jim Wilson <jimw@sifive.com>
8690 * config/riscv/riscv-protos.h (riscv_output_return): New.
8691 * config/riscv/riscv.c (struct machine_function): New naked_p field.
8692 (riscv_attribute_table, riscv_output_return),
8693 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
8694 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
8695 (riscv_compute_frame_info): Only compute frame->mask if not a naked
8697 (riscv_expand_prologue): Add early return for naked function.
8698 (riscv_expand_epilogue): Likewise.
8699 (riscv_function_ok_for_sibcall): Return false for naked function.
8700 (riscv_set_current_function): New.
8701 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
8702 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
8703 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
8704 * doc/extend.texi (RISC-V Function Attributes): New.
8706 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8708 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
8709 check for 128-bit long double before checking TCmode.
8710 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
8711 128-bit long doubles before checking TFmode or TCmode.
8712 (FLOAT128_IBM_P): Likewise.
8714 2018-01-10 Martin Sebor <msebor@redhat.com>
8716 PR tree-optimization/83671
8717 * builtins.c (c_strlen): Unconditionally return zero for the empty
8719 Use -Warray-bounds for warnings.
8720 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
8721 for non-constant array indices with COMPONENT_REF, arrays of
8722 arrays, and pointers to arrays.
8723 (gimple_fold_builtin_strlen): Determine and set length range for
8724 non-constant character arrays.
8726 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
8729 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
8732 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
8734 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
8736 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8739 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
8740 VECTOR_MEM_ALTIVEC_OR_VSX_P.
8741 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
8742 indexed_or_indirect_operand predicate.
8743 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
8744 (*vsx_le_perm_load_v8hi): Likewise.
8745 (*vsx_le_perm_load_v16qi): Likewise.
8746 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
8747 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
8748 (*vsx_le_perm_store_v8hi): Likewise.
8749 (*vsx_le_perm_store_v16qi): Likewise.
8750 (eight unnamed splitters): Likewise.
8752 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8754 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
8755 * config/rs6000/emmintrin.h: Likewise.
8756 * config/rs6000/mmintrin.h: Likewise.
8757 * config/rs6000/xmmintrin.h: Likewise.
8759 2018-01-10 David Malcolm <dmalcolm@redhat.com>
8762 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
8764 * tree.c (tree_nop_conversion): Return true for location wrapper
8766 (maybe_wrap_with_location): New function.
8767 (selftest::check_strip_nops): New function.
8768 (selftest::test_location_wrappers): New function.
8769 (selftest::tree_c_tests): Call it.
8770 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
8771 (maybe_wrap_with_location): New decl.
8772 (EXPR_LOCATION_WRAPPER_P): New macro.
8773 (location_wrapper_p): New inline function.
8774 (tree_strip_any_location_wrapper): New inline function.
8776 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
8779 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
8780 stack_realign_offset for the largest alignment of stack slot
8782 (ix86_find_max_used_stack_alignment): New function.
8783 (ix86_finalize_stack_frame_flags): Use it. Set
8784 max_used_stack_alignment if we don't realign stack.
8785 * config/i386/i386.h (machine_function): Add
8786 max_used_stack_alignment.
8788 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
8790 * config/arm/arm.opt (-mbranch-cost): New option.
8791 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
8794 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
8797 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
8798 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
8800 2018-01-10 Richard Biener <rguenther@suse.de>
8803 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
8804 early out so it also covers the case where we have a non-NULL
8807 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8809 PR tree-optimization/83753
8810 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
8811 for non-strided grouped accesses if the number of elements is 1.
8813 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8816 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
8817 * i386.h (TARGET_USE_GATHER): Define.
8818 * x86-tune.def (X86_TUNE_USE_GATHER): New.
8820 2018-01-10 Martin Liska <mliska@suse.cz>
8823 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
8824 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
8826 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
8827 CLEANUP_NO_PARTITIONING is not set.
8829 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8831 * doc/rtl.texi: Remove documentation of (const ...) wrappers
8832 for vectors, as a partial revert of r254296.
8833 * rtl.h (const_vec_p): Delete.
8834 (const_vec_duplicate_p): Don't test for vector CONSTs.
8835 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
8836 * expmed.c (make_tree): Likewise.
8839 * common.md (E, F): Use CONSTANT_P instead of checking for
8841 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
8842 checking for CONST_VECTOR.
8844 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8847 * predict.c (force_edge_cold): Handle in more sane way edges
8850 2018-01-09 Carl Love <cel@us.ibm.com>
8852 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
8854 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
8855 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
8856 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
8857 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
8858 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
8859 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
8860 * config/rs6000/rs6000-protos.h: Add extern defition for
8861 rs6000_generate_float2_double_code.
8862 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
8864 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
8865 (float2_v2df): Add define_expand.
8867 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
8870 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
8871 op_mode in the force_to_mode call.
8873 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8875 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
8876 instead of checking each element individually.
8877 (aarch64_evpc_uzp): Likewise.
8878 (aarch64_evpc_zip): Likewise.
8879 (aarch64_evpc_ext): Likewise.
8880 (aarch64_evpc_rev): Likewise.
8881 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
8882 instead of checking each element individually. Return true without
8884 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
8885 whether all selected elements come from the same input, instead of
8886 checking each element individually. Remove calls to gen_rtx_REG,
8887 start_sequence and end_sequence and instead assert that no rtl is
8890 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8892 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
8893 order of HIGH and CONST checks.
8895 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8897 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
8898 if the destination isn't an SSA_NAME.
8900 2018-01-09 Richard Biener <rguenther@suse.de>
8902 PR tree-optimization/83668
8903 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
8905 (canonicalize_loop_form): ... here, renamed from ...
8906 (canonicalize_loop_closed_ssa_form): ... this and amended to
8907 swap successor edges for loop exit blocks to make us use
8908 the RPO order we need for initial schedule generation.
8910 2018-01-09 Joseph Myers <joseph@codesourcery.com>
8912 PR tree-optimization/64811
8913 * match.pd: When optimizing comparisons with Inf, avoid
8914 introducing or losing exceptions from comparisons with NaN.
8916 2018-01-09 Martin Liska <mliska@suse.cz>
8919 * asan.c (shadow_mem_size): Add gcc_assert.
8921 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
8923 Don't save registers in main().
8926 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
8927 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
8928 * config/avr/avr.c (avr_set_current_function): Don't error if
8929 naked, OS_task or OS_main are specified at the same time.
8930 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
8932 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
8934 * common/config/avr/avr-common.c (avr_option_optimization_table):
8935 Switch on -mmain-is-OS_task for optimizing compilations.
8937 2018-01-09 Richard Biener <rguenther@suse.de>
8939 PR tree-optimization/83572
8940 * graphite.c: Include cfganal.h.
8941 (graphite_transform_loops): Connect infinite loops to exit
8942 and remove fake edges at the end.
8944 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8946 * ipa-inline.c (edge_badness): Revert accidental checkin.
8948 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8951 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
8952 symbols; not inline clones.
8954 2018-01-09 Jakub Jelinek <jakub@redhat.com>
8957 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
8958 hard registers. Formatting fixes.
8960 PR preprocessor/83722
8961 * gcc.c (try_generate_repro): Pass
8962 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
8963 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
8966 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
8967 Kito Cheng <kito.cheng@gmail.com>
8969 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
8970 (riscv_leaf_function_p): Delete.
8971 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
8973 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8975 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
8977 (do_ifelse): New function.
8978 (do_isel): New function.
8979 (do_sub3): New function.
8980 (do_add3): New function.
8981 (do_load_mask_compare): New function.
8982 (do_overlap_load_compare): New function.
8983 (expand_compare_loop): New function.
8984 (expand_block_compare): Call expand_compare_loop() when appropriate.
8985 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
8987 (-mblock-compare-inline-loop-limit): New option.
8989 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8992 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
8993 Reverse order of second and third operands in first alternative.
8994 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
8995 of first and second elements in UNSPEC_VPERMR vector.
8996 (altivec_expand_vec_perm_le): Likewise.
8998 2018-01-08 Jeff Law <law@redhat.com>
9000 PR rtl-optimizatin/81308
9001 * tree-switch-conversion.c (cfg_altered): New file scoped static.
9002 (process_switch): If group_case_labels makes a change, then set
9004 (pass_convert_switch::execute): If a switch is converted, then
9005 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
9007 PR rtl-optimization/81308
9008 * recog.c (split_all_insns): Conditionally cleanup the CFG after
9011 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
9013 PR target/83663 - Revert r255946
9014 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9015 generation for cases where splatting a value is not useful.
9016 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9017 across a vec_duplicate and a paradoxical subreg forming a vector
9018 mode to a vec_concat.
9020 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9022 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9023 -march=armv8.3-a variants.
9024 * config/arm/t-multilib: Likewise.
9025 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
9027 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9029 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9031 (cceq_ior_compare_complement): Give it a name so I can use it, and
9032 change boolean_or_operator predicate to boolean_operator so it can
9033 be used to generate a crand.
9034 (eqne): New code iterator.
9035 (bd/bd_neg): New code_attrs.
9036 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9037 a single define_insn.
9038 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9039 decrement (bdnzt/bdnzf/bdzt/bdzf).
9040 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9041 with the new names of the branch decrement patterns, and added the
9042 names of the branch decrement conditional patterns.
9044 2018-01-08 Richard Biener <rguenther@suse.de>
9046 PR tree-optimization/83563
9047 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9050 2018-01-08 Richard Biener <rguenther@suse.de>
9053 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9055 2018-01-08 Richard Biener <rguenther@suse.de>
9057 PR tree-optimization/83685
9058 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9059 references to abnormals.
9061 2018-01-08 Richard Biener <rguenther@suse.de>
9064 * dwarf2out.c (output_indirect_strings): Handle empty
9065 skeleton_debug_str_hash.
9066 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9068 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9070 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9071 (emit_store_direct): Likewise.
9072 (arc_trampoline_adjust_address): Likewise.
9073 (arc_asm_trampoline_template): New function.
9074 (arc_initialize_trampoline): Use asm_trampoline_template.
9075 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9076 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9077 * config/arc/arc.md (flush_icache): Delete pattern.
9079 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9081 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9082 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9085 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9088 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9089 by not USED_FOR_TARGET.
9090 (make_pass_resolve_sw_modes): Likewise.
9092 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9094 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9097 2018-01-08 Richard Biener <rguenther@suse.de>
9100 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9102 2018-01-08 Richard Biener <rguenther@suse.de>
9105 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9107 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9110 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9111 basic blocks with a small number of successors.
9112 (convert_control_dep_chain_into_preds): Improve handling of
9114 (dump_predicates): Split apart into...
9115 (dump_pred_chain): ...here...
9116 (dump_pred_info): ...and here.
9117 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9118 (can_chain_union_be_invalidated_p): Improve check for invalidation
9120 (uninit_uses_cannot_happen): Avoid unnecessary if
9121 convert_control_dep_chain_into_preds yielded nothing.
9123 2018-01-06 Martin Sebor <msebor@redhat.com>
9125 PR tree-optimization/83640
9126 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9127 subtracting negative offset from size.
9128 (builtin_access::overlap): Adjust offset bounds of the access to fall
9129 within the size of the object if possible.
9131 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9133 PR rtl-optimization/83699
9134 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9135 extract_bit_field_as_subreg to cases in which the extracted
9136 value is also a vector.
9138 * lra-constraints.c (process_alt_operands): Test for the equivalence
9139 substitutions when detecting a possible reload cycle.
9141 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9144 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9145 by default if flag_selective_schedling{,2}. Formatting fixes.
9147 PR rtl-optimization/83682
9148 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9149 if it has non-VECTOR_MODE element mode.
9150 (vec_duplicate_p): Likewise.
9153 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9154 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9156 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9159 * config/i386/i386-builtin.def
9160 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9161 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9162 Require also OPTION_MASK_ISA_AVX512F in addition to
9163 OPTION_MASK_ISA_GFNI.
9164 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9165 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9166 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9167 to OPTION_MASK_ISA_GFNI.
9168 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9169 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9170 OPTION_MASK_ISA_AVX512BW.
9171 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9172 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9173 addition to OPTION_MASK_ISA_GFNI.
9174 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9175 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9176 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9177 to OPTION_MASK_ISA_GFNI.
9178 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9179 a requirement for all ISAs rather than any of them with a few
9181 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9183 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9184 bitmasks to be enabled with 3 exceptions, instead of requiring any
9185 enabled ISA with lots of exceptions.
9186 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9187 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9188 Change avx512bw in isa attribute to avx512f.
9189 * config/i386/sgxintrin.h: Add license boilerplate.
9190 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9191 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9192 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9193 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9195 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9196 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9197 temporarily sse2 rather than sse if not enabled already.
9200 * config/i386/sse.md (VI248_VLBW): Rename to ...
9201 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
9202 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9203 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9204 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9205 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9206 mode iterator instead of VI248_VLBW.
9208 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
9210 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
9211 (record_modified): Skip clobbers; add debug output.
9212 (param_change_prob): Use sreal frequencies.
9214 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9216 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
9217 punt for user-aligned variables.
9219 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9221 * tree-chrec.c (chrec_contains_symbols): Return true for
9224 2018-01-05 Sudakshina Das <sudi.das@arm.com>
9227 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
9228 of (x|y) == x for BICS pattern.
9230 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9232 PR tree-optimization/83605
9233 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
9234 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9237 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
9239 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9240 * config/epiphany/rtems.h: New file.
9242 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9243 Uros Bizjak <ubizjak@gmail.com>
9246 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9247 QIreg_operand instead of register_operand predicate.
9248 * config/i386/i386.c (ix86_rop_should_change_byte_p,
9249 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9250 comments instead of -fmitigate[-_]rop.
9252 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9255 * cgraphunit.c (symbol_table::compile): Switch to text_section
9256 before calling assembly_start debug hook.
9257 * run-rtl-passes.c (run_rtl_passes): Likewise.
9260 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9262 * tree-vrp.c (extract_range_from_binary_expr_1): Check
9263 range_int_cst_p rather than !symbolic_range_p before calling
9264 extract_range_from_multiplicative_op_1.
9266 2018-01-04 Jeff Law <law@redhat.com>
9268 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9269 redundant test in assertion.
9271 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9273 * doc/rtl.texi: Document machine_mode wrapper classes.
9275 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9277 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9280 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9282 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9283 the VEC_PERM_EXPR fold to fail.
9285 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9288 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9289 to switched_sections.
9291 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9294 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9297 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9300 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9301 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9303 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9306 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9307 is BLKmode and bitpos not zero or mode change is needed.
9309 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9312 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9315 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9318 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9319 instead of MULT rtx. Update all corresponding splitters.
9321 (*ssub<modesuffix>): Ditto.
9323 (*cmp_sadd_di): Update split patterns.
9324 (*cmp_sadd_si): Ditto.
9325 (*cmp_sadd_sidi): Ditto.
9326 (*cmp_ssub_di): Ditto.
9327 (*cmp_ssub_si): Ditto.
9328 (*cmp_ssub_sidi): Ditto.
9329 * config/alpha/predicates.md (const23_operand): New predicate.
9330 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9331 Look for ASHIFT, not MULT inner operand.
9332 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9334 2018-01-04 Martin Liska <mliska@suse.cz>
9336 PR gcov-profile/83669
9337 * gcov.c (output_intermediate_file): Add version to intermediate
9339 * doc/gcov.texi: Document new field 'version' in intermediate
9340 file format. Fix location of '-k' option of gcov command.
9342 2018-01-04 Martin Liska <mliska@suse.cz>
9345 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9347 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9349 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9351 2018-01-03 Martin Sebor <msebor@redhat.com>
9353 PR tree-optimization/83655
9354 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9355 checking calls with invalid arguments.
9357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9359 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9360 (vectorizable_mask_load_store): Delete.
9361 (vectorizable_call): Return false for masked loads and stores.
9362 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9363 instead of gimple_assign_rhs1.
9364 (vectorizable_load): Handle IFN_MASK_LOAD.
9365 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9369 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9371 (vectorizable_mask_load_store): ...here.
9372 (vectorizable_load): ...and here.
9374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9376 * tree-vect-stmts.c (vect_build_all_ones_mask)
9377 (vect_build_zero_merge_argument): New functions, split out from...
9378 (vectorizable_load): ...here.
9380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9382 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9384 (vectorizable_mask_load_store): ...here.
9385 (vectorizable_store): ...and here.
9387 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9389 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9391 (vectorizable_mask_load_store): ...here.
9393 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9395 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9396 (vect_model_store_cost): Take a vec_load_store_type instead of a
9398 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9399 (vect_model_store_cost): Take a vec_load_store_type instead of a
9401 (vectorizable_mask_load_store): Update accordingly.
9402 (vectorizable_store): Likewise.
9403 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9405 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9407 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9408 IFN_MASK_LOAD calls here rather than...
9409 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9412 Alan Hayward <alan.hayward@arm.com>
9413 David Sherwood <david.sherwood@arm.com>
9415 * expmed.c (extract_bit_field_1): For vector extracts,
9416 fall back to extract_bit_field_as_subreg if vec_extract
9419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9420 Alan Hayward <alan.hayward@arm.com>
9421 David Sherwood <david.sherwood@arm.com>
9423 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
9424 they are variable or constant sized.
9425 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
9426 slots for constant-sized data.
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 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
9433 handling COND_EXPRs with boolean comparisons, try to find a better
9434 basis for the mask type than the boolean itself.
9436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9438 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
9439 is calculated and how it can be overridden.
9440 * genmodes.c (max_bitsize_mode_any_mode): New variable.
9441 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
9443 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
9446 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9447 Alan Hayward <alan.hayward@arm.com>
9448 David Sherwood <david.sherwood@arm.com>
9450 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
9451 Remove the mode argument.
9452 (aarch64_simd_valid_immediate): Remove the mode and inverse
9454 * config/aarch64/iterators.md (bitsize): New iterator.
9455 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
9456 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
9457 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
9458 aarch64_simd_valid_immediate.
9459 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
9460 (aarch64_reg_or_bic_imm): Likewise.
9461 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
9462 with an insn_type enum and msl with a modifier_type enum.
9463 Replace element_width with a scalar_mode. Change the shift
9464 to unsigned int. Add constructors for scalar_float_mode and
9465 scalar_int_mode elements.
9466 (aarch64_vect_float_const_representable_p): Delete.
9467 (aarch64_can_const_movi_rtx_p)
9468 (aarch64_simd_scalar_immediate_valid_for_move)
9469 (aarch64_simd_make_constant): Update call to
9470 aarch64_simd_valid_immediate.
9471 (aarch64_advsimd_valid_immediate_hs): New function.
9472 (aarch64_advsimd_valid_immediate): Likewise.
9473 (aarch64_simd_valid_immediate): Remove mode and inverse
9474 arguments. Rewrite to use the above. Use const_vec_duplicate_p
9475 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
9476 and aarch64_float_const_representable_p on the result.
9477 (aarch64_output_simd_mov_immediate): Remove mode argument.
9478 Update call to aarch64_simd_valid_immediate and use of
9479 simd_immediate_info.
9480 (aarch64_output_scalar_simd_mov_immediate): Update call
9483 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9484 Alan Hayward <alan.hayward@arm.com>
9485 David Sherwood <david.sherwood@arm.com>
9487 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
9488 (mode_nunits): Likewise CONST_MODE_NUNITS.
9489 * machmode.def (ADJUST_NUNITS): Document.
9490 * genmodes.c (mode_data::need_nunits_adj): New field.
9491 (blank_mode): Update accordingly.
9492 (adj_nunits): New variable.
9493 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
9495 (emit_mode_size_inline): Set need_bytesize_adj for all modes
9496 listed in adj_nunits.
9497 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
9498 listed in adj_nunits. Don't emit case statements for such modes.
9499 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
9500 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
9501 nothing if adj_nunits is nonnull.
9502 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
9503 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
9504 (emit_mode_fbit): Update use of print_maybe_const_decl.
9505 (emit_move_size): Likewise. Treat the array as non-const
9507 (emit_mode_adjustments): Handle adj_nunits.
9509 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9511 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
9512 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
9513 (VECTOR_MODES): Use it.
9514 (make_vector_modes): Take the prefix as an argument.
9516 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9517 Alan Hayward <alan.hayward@arm.com>
9518 David Sherwood <david.sherwood@arm.com>
9520 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
9521 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
9522 for MODE_VECTOR_BOOL.
9523 * machmode.def (VECTOR_BOOL_MODE): Document.
9524 * genmodes.c (VECTOR_BOOL_MODE): New macro.
9525 (make_vector_bool_mode): New function.
9526 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
9528 * lto-streamer-in.c (lto_input_mode_table): Likewise.
9529 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
9531 * stor-layout.c (int_mode_for_mode): Likewise.
9532 * tree.c (build_vector_type_for_mode): Likewise.
9533 * varasm.c (output_constant_pool_2): Likewise.
9534 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
9535 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
9536 for MODE_VECTOR_BOOL.
9537 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
9538 of mode class checks.
9539 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
9540 instead of a list of mode class checks.
9541 (expand_vector_scalar_condition): Likewise.
9542 (type_for_widest_vector_mode): Handle BImode as an inner mode.
9544 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9545 Alan Hayward <alan.hayward@arm.com>
9546 David Sherwood <david.sherwood@arm.com>
9548 * machmode.h (mode_size): Change from unsigned short to
9550 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
9551 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9552 or if measurement_type is not polynomial.
9553 (fixed_size_mode::includes_p): Check for constant-sized modes.
9554 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
9555 return a poly_uint16 rather than an unsigned short.
9556 (emit_mode_size): Change the type of mode_size from unsigned short
9557 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
9558 (emit_mode_adjustments): Cope with polynomial vector sizes.
9559 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9561 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9563 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
9564 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
9565 * caller-save.c (setup_save_areas): Likewise.
9566 (replace_reg_with_saved_mem): Likewise.
9567 * calls.c (emit_library_call_value_1): Likewise.
9568 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
9569 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
9570 (gen_lowpart_for_combine): Likewise.
9571 * convert.c (convert_to_integer_1): Likewise.
9572 * cse.c (equiv_constant, cse_insn): Likewise.
9573 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
9574 (cselib_subst_to_values): Likewise.
9575 * dce.c (word_dce_process_block): Likewise.
9576 * df-problems.c (df_word_lr_mark_ref): Likewise.
9577 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
9578 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
9579 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
9580 (rtl_for_decl_location): Likewise.
9581 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
9582 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
9583 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
9584 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
9585 (expand_expr_real_1): Likewise.
9586 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
9587 (pad_below): Likewise.
9588 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9589 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
9590 * ira.c (get_subreg_tracking_sizes): Likewise.
9591 * ira-build.c (ira_create_allocno_objects): Likewise.
9592 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
9593 (ira_sort_regnos_for_alter_reg): Likewise.
9594 * ira-costs.c (record_operand_costs): Likewise.
9595 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
9596 (resolve_simple_move): Likewise.
9597 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
9598 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
9599 (lra_constraints): Likewise.
9600 (CONST_POOL_OK_P): Reject variable-sized modes.
9601 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
9602 (add_pseudo_to_slot, lra_spill): Likewise.
9603 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9604 * optabs-query.c (get_best_extraction_insn): Likewise.
9605 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9606 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
9607 (expand_mult_highpart, valid_multiword_target_p): Likewise.
9608 * recog.c (offsettable_address_addr_space_p): Likewise.
9609 * regcprop.c (maybe_mode_change): Likewise.
9610 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
9611 * regrename.c (build_def_use): Likewise.
9612 * regstat.c (dump_reg_info): Likewise.
9613 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
9614 (find_reloads, find_reloads_subreg_address): Likewise.
9615 * reload1.c (eliminate_regs_1): Likewise.
9616 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
9617 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
9618 (simplify_binary_operation_1, simplify_subreg): Likewise.
9619 * targhooks.c (default_function_arg_padding): Likewise.
9620 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
9621 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
9622 (verify_gimple_assign_ternary): Likewise.
9623 * tree-inline.c (estimate_move_cost): Likewise.
9624 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9625 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
9626 (get_address_cost_ainc): Likewise.
9627 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9628 (vect_supportable_dr_alignment): Likewise.
9629 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9630 (vectorizable_reduction): Likewise.
9631 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
9632 (vectorizable_operation, vectorizable_load): Likewise.
9633 * tree.c (build_same_sized_truth_vector_type): Likewise.
9634 * valtrack.c (cleanup_auto_inc_dec): Likewise.
9635 * var-tracking.c (emit_note_insn_var_location): Likewise.
9636 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
9637 (ADDR_VEC_ALIGN): Likewise.
9639 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9640 Alan Hayward <alan.hayward@arm.com>
9641 David Sherwood <david.sherwood@arm.com>
9643 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
9645 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9646 or if measurement_type is polynomial.
9647 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
9648 * combine.c (make_extraction): Likewise.
9649 * dse.c (find_shift_sequence): Likewise.
9650 * dwarf2out.c (mem_loc_descriptor): Likewise.
9651 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
9652 (extract_bit_field, extract_low_bits): Likewise.
9653 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
9654 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
9655 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
9656 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
9657 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9658 * reload.c (find_reloads): Likewise.
9659 * reload1.c (alter_reg): Likewise.
9660 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
9661 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
9662 * tree-if-conv.c (predicate_mem_writes): Likewise.
9663 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
9664 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
9665 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9666 * valtrack.c (dead_debug_insert_temp): Likewise.
9667 * varasm.c (mergeable_constant_section): Likewise.
9668 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
9670 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9671 Alan Hayward <alan.hayward@arm.com>
9672 David Sherwood <david.sherwood@arm.com>
9674 * expr.c (expand_assignment): Cope with polynomial mode sizes
9675 when assigning to a CONCAT.
9677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9678 Alan Hayward <alan.hayward@arm.com>
9679 David Sherwood <david.sherwood@arm.com>
9681 * machmode.h (mode_precision): Change from unsigned short to
9683 (mode_to_precision): Return a poly_uint16 rather than an unsigned
9685 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
9686 or if measurement_type is not polynomial.
9687 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
9688 in which the mode is already known to be a scalar_int_mode.
9689 * genmodes.c (emit_mode_precision): Change the type of mode_precision
9690 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
9692 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9693 for GET_MODE_PRECISION.
9694 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9695 for GET_MODE_PRECISION.
9696 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
9698 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
9699 (expand_field_assignment, make_extraction): Likewise.
9700 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
9701 (get_last_value): Likewise.
9702 * convert.c (convert_to_integer_1): Likewise.
9703 * cse.c (cse_insn): Likewise.
9704 * expr.c (expand_expr_real_1): Likewise.
9705 * lra-constraints.c (simplify_operand_subreg): Likewise.
9706 * optabs-query.c (can_atomic_load_p): Likewise.
9707 * optabs.c (expand_atomic_load): Likewise.
9708 (expand_atomic_store): Likewise.
9709 * ree.c (combine_reaching_defs): Likewise.
9710 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
9711 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
9712 * tree.h (type_has_mode_precision_p): Likewise.
9713 * ubsan.c (instrument_si_overflow): Likewise.
9715 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9716 Alan Hayward <alan.hayward@arm.com>
9717 David Sherwood <david.sherwood@arm.com>
9719 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
9720 polynomial numbers of units.
9721 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
9722 (valid_vector_subparts_p): New function.
9723 (build_vector_type): Remove temporary shim and take the number
9724 of units as a poly_uint64 rather than an int.
9725 (build_opaque_vector_type): Take the number of units as a
9726 poly_uint64 rather than an int.
9727 * tree.c (build_vector_from_ctor): Handle polynomial
9728 TYPE_VECTOR_SUBPARTS.
9729 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
9730 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
9731 (build_vector_from_val): If the number of units is variable,
9732 use build_vec_duplicate_cst for constant operands and
9733 VEC_DUPLICATE_EXPR otherwise.
9734 (make_vector_type): Remove temporary is_constant ().
9735 (build_vector_type, build_opaque_vector_type): Take the number of
9736 units as a poly_uint64 rather than an int.
9737 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
9739 * cfgexpand.c (expand_debug_expr): Likewise.
9740 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
9741 (store_constructor, expand_expr_real_1): Likewise.
9742 (const_scalar_mask_from_tree): Likewise.
9743 * fold-const-call.c (fold_const_reduction): Likewise.
9744 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
9745 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
9746 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
9747 (fold_relational_const): Likewise.
9748 (native_interpret_vector): Likewise. Change the size from an
9749 int to an unsigned int.
9750 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
9751 TYPE_VECTOR_SUBPARTS.
9752 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
9753 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
9754 duplicating a non-constant operand into a variable-length vector.
9755 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
9756 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
9757 * ipa-icf.c (sem_variable::equals): Likewise.
9758 * match.pd: Likewise.
9759 * omp-simd-clone.c (simd_clone_subparts): Likewise.
9760 * print-tree.c (print_node): Likewise.
9761 * stor-layout.c (layout_type): Likewise.
9762 * targhooks.c (default_builtin_vectorization_cost): Likewise.
9763 * tree-cfg.c (verify_gimple_comparison): Likewise.
9764 (verify_gimple_assign_binary): Likewise.
9765 (verify_gimple_assign_ternary): Likewise.
9766 (verify_gimple_assign_single): Likewise.
9767 * tree-pretty-print.c (dump_generic_node): Likewise.
9768 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9769 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
9770 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
9771 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
9772 (vect_shift_permute_load_chain): Likewise.
9773 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
9774 (expand_vector_condition, optimize_vector_constructor): Likewise.
9775 (lower_vec_perm, get_compute_type): Likewise.
9776 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9777 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
9778 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
9779 (vect_recog_mask_conversion_pattern): Likewise.
9780 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
9781 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
9782 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9783 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
9784 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
9785 (vectorizable_shift, vectorizable_operation, vectorizable_store)
9786 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
9787 (supportable_widening_operation): Likewise.
9788 (supportable_narrowing_operation): Likewise.
9789 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
9791 * varasm.c (output_constant): Likewise.
9793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9794 Alan Hayward <alan.hayward@arm.com>
9795 David Sherwood <david.sherwood@arm.com>
9797 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
9798 so that both the length == 3 and length != 3 cases set up their
9799 own permute vectors. Add comments explaining why we know the
9800 number of elements is constant.
9801 (vect_permute_load_chain): Likewise.
9803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9804 Alan Hayward <alan.hayward@arm.com>
9805 David Sherwood <david.sherwood@arm.com>
9807 * machmode.h (mode_nunits): Change from unsigned char to
9809 (ONLY_FIXED_SIZE_MODES): New macro.
9810 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
9811 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
9812 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
9814 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
9815 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
9816 or if measurement_type is not polynomial.
9817 * genmodes.c (ZERO_COEFFS): New macro.
9818 (emit_mode_nunits_inline): Make mode_nunits_inline return a
9820 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
9821 Use ZERO_COEFFS when emitting initializers.
9822 * data-streamer.h (bp_pack_poly_value): New function.
9823 (bp_unpack_poly_value): Likewise.
9824 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9825 for GET_MODE_NUNITS.
9826 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9827 for GET_MODE_NUNITS.
9828 * tree.c (make_vector_type): Remove temporary shim and make
9829 the real function take the number of units as a poly_uint64
9831 (build_vector_type_for_mode): Handle polynomial nunits.
9832 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
9833 * emit-rtl.c (const_vec_series_p_1): Likewise.
9834 (gen_rtx_CONST_VECTOR): Likewise.
9835 * fold-const.c (test_vec_duplicate_folding): Likewise.
9836 * genrecog.c (validate_pattern): Likewise.
9837 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
9838 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9839 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
9840 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
9841 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
9842 * rtlanal.c (subreg_get_info): Likewise.
9843 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9844 (vect_grouped_load_supported): Likewise.
9845 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
9846 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
9847 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
9848 (simplify_const_unary_operation, simplify_binary_operation_1)
9849 (simplify_const_binary_operation, simplify_ternary_operation)
9850 (test_vector_ops_duplicate, test_vector_ops): Likewise.
9851 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
9852 instead of CONST_VECTOR_NUNITS.
9853 * varasm.c (output_constant_pool_2): Likewise.
9854 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
9855 explicit-encoded elements in the XVEC for variable-length vectors.
9857 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9859 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
9861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9862 Alan Hayward <alan.hayward@arm.com>
9863 David Sherwood <david.sherwood@arm.com>
9865 * coretypes.h (fixed_size_mode): Declare.
9866 (fixed_size_mode_pod): New typedef.
9867 * builtins.h (target_builtins::x_apply_args_mode)
9868 (target_builtins::x_apply_result_mode): Change type to
9869 fixed_size_mode_pod.
9870 * builtins.c (apply_args_size, apply_result_size, result_vector)
9871 (expand_builtin_apply_args_1, expand_builtin_apply)
9872 (expand_builtin_return): Update accordingly.
9874 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9876 * cse.c (hash_rtx_cb): Hash only the encoded elements.
9877 * cselib.c (cselib_hash_rtx): Likewise.
9878 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
9879 CONST_VECTOR encoding.
9881 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9882 Jeff Law <law@redhat.com>
9885 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
9886 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
9887 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
9888 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
9891 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
9892 explicitly probe *sp in a noreturn function if there were any callee
9893 register saves or frame pointer is needed.
9895 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9898 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
9899 BLKmode for ternary, binary or unary expressions.
9902 * var-tracking.c (delete_vta_debug_insn): New inline function.
9903 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
9904 insns from get_insns () to NULL instead of each bb separately.
9905 Use delete_vta_debug_insn. No longer static.
9906 (vt_debug_insns_local, variable_tracking_main_1): Adjust
9907 delete_vta_debug_insns callers.
9908 * rtl.h (delete_vta_debug_insns): Declare.
9909 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
9910 instead of variable_tracking_main.
9912 2018-01-03 Martin Sebor <msebor@redhat.com>
9914 PR tree-optimization/83603
9915 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
9916 arguments past the endof the argument list in functions declared
9917 without a prototype.
9918 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
9919 Avoid checking when arguments are null.
9921 2018-01-03 Martin Sebor <msebor@redhat.com>
9924 * doc/extend.texi (attribute const): Fix a typo.
9925 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
9926 issuing -Wsuggest-attribute for void functions.
9928 2018-01-03 Martin Sebor <msebor@redhat.com>
9930 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
9931 offset_int::from instead of wide_int::to_shwi.
9932 (maybe_diag_overlap): Remove assertion.
9933 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
9934 * gimple-ssa-sprintf.c (format_directive): Same.
9935 (parse_directive): Same.
9936 (sprintf_dom_walker::compute_format_length): Same.
9937 (try_substitute_return_value): Same.
9939 2018-01-03 Jeff Law <law@redhat.com>
9942 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
9943 non-constant residual for zero at runtime and avoid probing in
9944 that case. Reorganize code for trailing problem to mirror handling
9947 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9949 PR tree-optimization/83501
9950 * tree-ssa-strlen.c (get_string_cst): New.
9951 (handle_char_store): Call get_string_cst.
9953 2018-01-03 Martin Liska <mliska@suse.cz>
9955 PR tree-optimization/83593
9956 * tree-ssa-strlen.c: Include tree-cfg.h.
9957 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
9958 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
9959 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
9961 (strlen_dom_walker::before_dom_children): Call
9962 gimple_purge_dead_eh_edges. Dump tranformation with details
9964 (strlen_dom_walker::before_dom_children): Update call by adding
9965 new argument cleanup_eh.
9966 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
9968 2018-01-03 Martin Liska <mliska@suse.cz>
9971 * cif-code.def (VARIADIC_THUNK): New enum value.
9972 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
9975 2018-01-03 Jan Beulich <jbeulich@suse.com>
9977 * sse.md (mov<mode>_internal): Tighten condition for when to use
9978 vmovdqu<ssescalarsize> for TI and OI modes.
9980 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9982 Update copyright years.
9984 2018-01-03 Martin Liska <mliska@suse.cz>
9987 * ipa-visibility.c (function_and_variable_visibility): Skip
9988 functions with noipa attribure.
9990 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9992 * gcc.c (process_command): Update copyright notice dates.
9993 * gcov-dump.c (print_version): Ditto.
9994 * gcov.c (print_version): Ditto.
9995 * gcov-tool.c (print_version): Ditto.
9996 * gengtype.c (create_file): Ditto.
9997 * doc/cpp.texi: Bump @copying's copyright year.
9998 * doc/cppinternals.texi: Ditto.
9999 * doc/gcc.texi: Ditto.
10000 * doc/gccint.texi: Ditto.
10001 * doc/gcov.texi: Ditto.
10002 * doc/install.texi: Ditto.
10003 * doc/invoke.texi: Ditto.
10005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10007 * vector-builder.h (vector_builder::m_full_nelts): Change from
10008 unsigned int to poly_uint64.
10009 (vector_builder::full_nelts): Update prototype accordingly.
10010 (vector_builder::new_vector): Likewise.
10011 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10012 (vector_builder::operator ==): Likewise.
10013 (vector_builder::finalize): Likewise.
10014 * int-vector-builder.h (int_vector_builder::int_vector_builder):
10015 Take the number of elements as a poly_uint64 rather than an
10017 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10018 from unsigned int to poly_uint64.
10019 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10020 (vec_perm_indices::new_vector): Likewise.
10021 (vec_perm_indices::length): Likewise.
10022 (vec_perm_indices::nelts_per_input): Likewise.
10023 (vec_perm_indices::input_nelts): Likewise.
10024 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10025 number of elements per input as a poly_uint64 rather than an
10026 unsigned int. Use the original encoding for variable-length
10027 vectors, rather than clamping each individual element.
10028 For the second and subsequent elements in each pattern,
10029 clamp the step and base before clamping their sum.
10030 (vec_perm_indices::series_p): Handle polynomial element counts.
10031 (vec_perm_indices::all_in_range_p): Likewise.
10032 (vec_perm_indices_to_tree): Likewise.
10033 (vec_perm_indices_to_rtx): Likewise.
10034 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10035 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10036 (tree_vector_builder::new_binary_operation): Handle polynomial
10037 element counts. Return false if we need to know the number
10038 of elements at compile time.
10039 * fold-const.c (fold_vec_perm): Punt if the number of elements
10040 isn't known at compile time.
10042 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10044 * vec-perm-indices.h (vec_perm_builder): Change element type
10045 from HOST_WIDE_INT to poly_int64.
10046 (vec_perm_indices::element_type): Update accordingly.
10047 (vec_perm_indices::clamp): Handle polynomial element_types.
10048 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10049 (vec_perm_indices::all_in_range_p): Likewise.
10050 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10052 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10053 polynomial vec_perm_indices element types.
10054 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10055 * fold-const.c (fold_vec_perm): Likewise.
10056 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10057 * tree-vect-generic.c (lower_vec_perm): Likewise.
10058 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10059 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10060 element type to HOST_WIDE_INT.
10062 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10063 Alan Hayward <alan.hayward@arm.com>
10064 David Sherwood <david.sherwood@arm.com>
10066 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10067 rather than an int. Use plus_constant.
10068 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10069 Take the offset "c" as a poly_int64 rather than a 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 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10076 a HOST_WIDE_INT to a poly_int64.
10078 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10079 Alan Hayward <alan.hayward@arm.com>
10080 David Sherwood <david.sherwood@arm.com>
10082 * calls.c (load_register_parameters): Cope with polynomial
10083 mode sizes. Require a constant size for BLKmode parameters
10084 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10085 forces a parameter to be padded at the lsb end in order to
10086 fill a complete number of words, require the parameter size
10087 to be ordered wrt UNITS_PER_WORD.
10089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10090 Alan Hayward <alan.hayward@arm.com>
10091 David Sherwood <david.sherwood@arm.com>
10093 * reload1.c (spill_stack_slot_width): Change element type
10094 from unsigned int to poly_uint64_pod.
10095 (alter_reg): Treat mode sizes as polynomial.
10097 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10098 Alan Hayward <alan.hayward@arm.com>
10099 David Sherwood <david.sherwood@arm.com>
10101 * reload.c (complex_word_subreg_p): New function.
10102 (reload_inner_reg_of_subreg, push_reload): Use it.
10104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10105 Alan Hayward <alan.hayward@arm.com>
10106 David Sherwood <david.sherwood@arm.com>
10108 * lra-constraints.c (process_alt_operands): Reject matched
10109 operands whose sizes aren't ordered.
10110 (match_reload): Refer to this check here.
10112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10113 Alan Hayward <alan.hayward@arm.com>
10114 David Sherwood <david.sherwood@arm.com>
10116 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10117 that the mode size is in the set {1, 2, 4, 8, 16}.
10119 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10120 Alan Hayward <alan.hayward@arm.com>
10121 David Sherwood <david.sherwood@arm.com>
10123 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10124 Use plus_constant instead of gen_rtx_PLUS.
10126 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10127 Alan Hayward <alan.hayward@arm.com>
10128 David Sherwood <david.sherwood@arm.com>
10130 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10131 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10132 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10133 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10134 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10135 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10136 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10137 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10138 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10139 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10141 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10142 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10143 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10144 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10145 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10146 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10147 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10148 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10149 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10150 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10152 * expr.c (emit_move_resolve_push): Treat the input and result
10153 of PUSH_ROUNDING as a poly_int64.
10154 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10155 (emit_push_insn): Likewise.
10156 * lra-eliminations.c (mark_not_eliminable): Likewise.
10157 * recog.c (push_operand): Likewise.
10158 * reload1.c (elimination_effects): Likewise.
10159 * rtlanal.c (nonzero_bits1): Likewise.
10160 * calls.c (store_one_arg): Likewise. Require the padding to be
10161 known at compile time.
10163 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10164 Alan Hayward <alan.hayward@arm.com>
10165 David Sherwood <david.sherwood@arm.com>
10167 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10168 Use plus_constant instead of gen_rtx_PLUS.
10170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10171 Alan Hayward <alan.hayward@arm.com>
10172 David Sherwood <david.sherwood@arm.com>
10174 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10175 rather than an int.
10177 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10178 Alan Hayward <alan.hayward@arm.com>
10179 David Sherwood <david.sherwood@arm.com>
10181 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10182 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10183 via stack temporaries. Treat the mode size as polynomial too.
10185 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10186 Alan Hayward <alan.hayward@arm.com>
10187 David Sherwood <david.sherwood@arm.com>
10189 * expr.c (expand_expr_real_2): When handling conversions involving
10190 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10191 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10192 as a poly_uint64 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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
10200 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10201 Alan Hayward <alan.hayward@arm.com>
10202 David Sherwood <david.sherwood@arm.com>
10204 * combine.c (can_change_dest_mode): Handle polynomial
10205 REGMODE_NATURAL_SIZE.
10206 * expmed.c (store_bit_field_1): Likewise.
10207 * expr.c (store_constructor): Likewise.
10208 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10209 and polynomial REGMODE_NATURAL_SIZE.
10210 (gen_lowpart_common): Likewise.
10211 * reginfo.c (record_subregs_of_mode): Likewise.
10212 * rtlanal.c (read_modify_subreg_p): Likewise.
10214 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10215 Alan Hayward <alan.hayward@arm.com>
10216 David Sherwood <david.sherwood@arm.com>
10218 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
10219 numbers of elements.
10221 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10222 Alan Hayward <alan.hayward@arm.com>
10223 David Sherwood <david.sherwood@arm.com>
10225 * match.pd: Cope with polynomial numbers of vector elements.
10227 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10228 Alan Hayward <alan.hayward@arm.com>
10229 David Sherwood <david.sherwood@arm.com>
10231 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
10232 in a POINTER_PLUS_EXPR.
10234 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10235 Alan Hayward <alan.hayward@arm.com>
10236 David Sherwood <david.sherwood@arm.com>
10238 * omp-simd-clone.c (simd_clone_subparts): New function.
10239 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10240 (ipa_simd_modify_function_body): Likewise.
10242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10243 Alan Hayward <alan.hayward@arm.com>
10244 David Sherwood <david.sherwood@arm.com>
10246 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
10247 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10248 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10249 (expand_vector_condition, vector_element): Likewise.
10250 (subparts_gt): New function.
10251 (get_compute_type): Use subparts_gt.
10252 (count_type_subparts): Delete.
10253 (expand_vector_operations_1): Use subparts_gt instead of
10254 count_type_subparts.
10256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10257 Alan Hayward <alan.hayward@arm.com>
10258 David Sherwood <david.sherwood@arm.com>
10260 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
10261 (vect_compile_time_alias): ...this new function. Do the calculation
10262 on poly_ints rather than trees.
10263 (vect_prune_runtime_alias_test_list): Update call accordingly.
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-slp.c (vect_build_slp_tree_1): Handle polynomial
10271 (vect_schedule_slp_instance): Likewise.
10273 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10274 Alan Hayward <alan.hayward@arm.com>
10275 David Sherwood <david.sherwood@arm.com>
10277 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
10278 constant and extern definitions for variable-length vectors.
10279 (vect_get_constant_vectors): Note that the number of units
10280 is known to be constant.
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-stmts.c (vectorizable_conversion): Treat the number
10287 of units as polynomial. Choose between WIDE and NARROW based
10290 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10291 Alan Hayward <alan.hayward@arm.com>
10292 David Sherwood <david.sherwood@arm.com>
10294 * tree-vect-stmts.c (simd_clone_subparts): New function.
10295 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10297 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10298 Alan Hayward <alan.hayward@arm.com>
10299 David Sherwood <david.sherwood@arm.com>
10301 * tree-vect-stmts.c (vectorizable_call): Treat the number of
10302 vectors as polynomial. Use build_index_vector for
10303 IFN_GOMP_SIMD_LANE.
10305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10306 Alan Hayward <alan.hayward@arm.com>
10307 David Sherwood <david.sherwood@arm.com>
10309 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10310 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10311 for variable-length vectors.
10312 (vectorizable_mask_load_store): Treat the number of units as
10313 polynomial, asserting that it is constant if the condition has
10314 already been enforced.
10315 (vectorizable_store, vectorizable_load): Likewise.
10317 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10318 Alan Hayward <alan.hayward@arm.com>
10319 David Sherwood <david.sherwood@arm.com>
10321 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10322 of units as polynomial. Punt if we can't tell at compile time
10323 which vector contains the final result.
10325 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10326 Alan Hayward <alan.hayward@arm.com>
10327 David Sherwood <david.sherwood@arm.com>
10329 * tree-vect-loop.c (vectorizable_induction): Treat the number
10330 of units as polynomial. Punt on SLP inductions. Use an integer
10331 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10332 cast of such a series for variable-length floating-point
10335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10336 Alan Hayward <alan.hayward@arm.com>
10337 David Sherwood <david.sherwood@arm.com>
10339 * tree.h (build_index_vector): Declare.
10340 * tree.c (build_index_vector): New function.
10341 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10342 of units as polynomial, forcibly converting it to a constant if
10343 vectorizable_reduction has already enforced the condition.
10344 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
10345 to create a {1,2,3,...} vector.
10346 (vectorizable_reduction): Treat the number of units as polynomial.
10347 Choose vectype_in based on the largest scalar element size rather
10348 than the smallest number of units. Enforce the restrictions
10351 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10352 Alan Hayward <alan.hayward@arm.com>
10353 David Sherwood <david.sherwood@arm.com>
10355 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10356 number of units as polynomial.
10358 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10359 Alan Hayward <alan.hayward@arm.com>
10360 David Sherwood <david.sherwood@arm.com>
10362 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10363 * target.def (autovectorize_vector_sizes): Return the vector sizes
10364 by pointer, using vector_sizes rather than a bitmask.
10365 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10366 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10367 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10369 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10370 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10371 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10372 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10373 * omp-general.c (omp_max_vf): Likewise.
10374 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10375 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10376 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10377 * tree-vect-slp.c (vect_slp_bb): Likewise.
10378 * doc/tm.texi: Regenerate.
10379 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10381 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10382 the vector size as a poly_uint64 rather than an unsigned int.
10383 (current_vector_size): Change from an unsigned int to a poly_uint64.
10384 (get_vectype_for_scalar_type): Update accordingly.
10385 * tree.h (build_truth_vector_type): Take the size and number of
10386 units as a poly_uint64 rather than an unsigned int.
10387 (build_vector_type): Add a temporary overload that takes
10388 the number of units as a poly_uint64 rather than an unsigned int.
10389 * tree.c (make_vector_type): Likewise.
10390 (build_truth_vector_type): Take the number of units as a poly_uint64
10391 rather than an unsigned int.
10393 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10394 Alan Hayward <alan.hayward@arm.com>
10395 David Sherwood <david.sherwood@arm.com>
10397 * target.def (get_mask_mode): Take the number of units and length
10398 as poly_uint64s rather than unsigned ints.
10399 * targhooks.h (default_get_mask_mode): Update accordingly.
10400 * targhooks.c (default_get_mask_mode): Likewise.
10401 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10402 * doc/tm.texi: Regenerate.
10404 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10405 Alan Hayward <alan.hayward@arm.com>
10406 David Sherwood <david.sherwood@arm.com>
10408 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10409 * omp-general.c (omp_max_vf): Likewise.
10410 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10411 (expand_omp_simd): Handle polynomial safelen.
10412 * omp-low.c (omplow_simd_context): Add a default constructor.
10413 (omplow_simd_context::max_vf): Change from int to poly_uint64.
10414 (lower_rec_simd_input_clauses): Update accordingly.
10415 (lower_rec_input_clauses): Likewise.
10417 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10418 Alan Hayward <alan.hayward@arm.com>
10419 David Sherwood <david.sherwood@arm.com>
10421 * tree-vectorizer.h (vect_nunits_for_cost): New function.
10422 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
10423 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
10424 (vect_analyze_slp_cost): Likewise.
10425 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
10426 (vect_model_load_cost): Likewise.
10428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10429 Alan Hayward <alan.hayward@arm.com>
10430 David Sherwood <david.sherwood@arm.com>
10432 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
10433 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
10434 from an unsigned int * to a poly_uint64_pod *.
10435 (calculate_unrolling_factor): New function.
10436 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
10438 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10439 Alan Hayward <alan.hayward@arm.com>
10440 David Sherwood <david.sherwood@arm.com>
10442 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
10443 from an unsigned int to a poly_uint64.
10444 (_loop_vec_info::slp_unrolling_factor): Likewise.
10445 (_loop_vec_info::vectorization_factor): Change from an int
10447 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
10448 (vect_get_num_vectors): New function.
10449 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
10450 (vect_get_num_copies): Use vect_get_num_vectors.
10451 (vect_analyze_data_ref_dependences): Change max_vf from an int *
10452 to an unsigned int *.
10453 (vect_analyze_data_refs): Change min_vf from an int * to a
10455 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10456 than an unsigned HOST_WIDE_INT.
10457 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10458 (vect_analyze_data_ref_dependence): Change max_vf from an int *
10459 to an unsigned int *.
10460 (vect_analyze_data_ref_dependences): Likewise.
10461 (vect_compute_data_ref_alignment): Handle polynomial vf.
10462 (vect_enhance_data_refs_alignment): Likewise.
10463 (vect_prune_runtime_alias_test_list): Likewise.
10464 (vect_shift_permute_load_chain): Likewise.
10465 (vect_supportable_dr_alignment): Likewise.
10466 (dependence_distance_ge_vf): Take the vectorization factor as a
10467 poly_uint64 rather than an unsigned HOST_WIDE_INT.
10468 (vect_analyze_data_refs): Change min_vf from an int * to a
10470 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
10471 vfm1 as a poly_uint64 rather than an int. Make the same change
10472 for the returned bound_scalar.
10473 (vect_gen_vector_loop_niters): Handle polynomial vf.
10474 (vect_do_peeling): Likewise. Update call to
10475 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
10476 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
10478 * tree-vect-loop.c (vect_determine_vectorization_factor)
10479 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
10480 (vect_get_known_peeling_cost): Likewise.
10481 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
10482 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
10483 (vect_transform_loop): Likewise. Use the lowest possible VF when
10484 updating the upper bounds of the loop.
10485 (vect_min_worthwhile_factor): Make static. Return an unsigned int
10486 rather than an int.
10487 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
10488 polynomial unroll factors.
10489 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
10490 (vect_make_slp_decision): Likewise.
10491 (vect_supported_load_permutation_p): Likewise, and polynomial
10493 (vect_analyze_slp_cost): Handle polynomial vf.
10494 (vect_slp_analyze_node_operations): Likewise.
10495 (vect_slp_analyze_bb_1): Likewise.
10496 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10497 than an unsigned HOST_WIDE_INT.
10498 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
10499 (vectorizable_load): Handle polynomial vf.
10500 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
10502 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
10504 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10505 Alan Hayward <alan.hayward@arm.com>
10506 David Sherwood <david.sherwood@arm.com>
10508 * match.pd: Handle bit operations involving three constants
10509 and try to fold one pair.
10511 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10513 * tree-vect-loop-manip.c: Include gimple-fold.h.
10514 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
10515 niters_maybe_zero parameters. Handle other cases besides a step of 1.
10516 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
10517 Add a path that uses a step of VF instead of 1, but disable it
10519 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
10520 and niters_no_overflow parameters. Update calls to
10521 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
10522 Create a new SSA name if the latter choses to use a ste other
10523 than zero, and return it via niters_vector_mult_vf_var.
10524 * tree-vect-loop.c (vect_transform_loop): Update calls to
10525 vect_do_peeling, vect_gen_vector_loop_niters and
10526 slpeel_make_loop_iterate_ntimes.
10527 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
10528 (vect_gen_vector_loop_niters): Update declarations after above changes.
10530 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
10532 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
10533 128-bit round to integer instructions.
10534 (ceil<mode>2): Likewise.
10535 (btrunc<mode>2): Likewise.
10536 (round<mode>2): Likewise.
10538 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10540 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
10541 unaligned VSX load/store on P8/P9.
10542 (expand_block_clear): Allow the use of unaligned VSX
10543 load/store on P8/P9.
10545 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10547 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
10549 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
10550 swap associated with both a load and a store.
10552 2018-01-02 Andrew Waterman <andrew@sifive.com>
10554 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
10555 * config/riscv/riscv.md (clear_cache): Use it.
10557 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
10559 * web.c: Remove out-of-date comment.
10561 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10563 * expr.c (fixup_args_size_notes): Check that any existing
10564 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
10565 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
10566 (emit_single_push_insn): ...here.
10568 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10570 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
10571 (const_vector_encoded_nelts): New function.
10572 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
10573 (const_vector_int_elt, const_vector_elt): Declare.
10574 * emit-rtl.c (const_vector_int_elt_1): New function.
10575 (const_vector_elt): Likewise.
10576 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
10577 of CONST_VECTOR_ELT.
10579 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10581 * expr.c: Include rtx-vector-builder.h.
10582 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
10583 directly on the tree encoding.
10584 (const_vector_from_tree): Likewise.
10585 * optabs.c: Include rtx-vector-builder.h.
10586 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
10587 sequence of "u" values.
10588 * vec-perm-indices.c: Include rtx-vector-builder.h.
10589 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
10590 directly on the vec_perm_indices encoding.
10592 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10594 * doc/rtl.texi (const_vector): Describe new encoding scheme.
10595 * Makefile.in (OBJS): Add rtx-vector-builder.o.
10596 * rtx-vector-builder.h: New file.
10597 * rtx-vector-builder.c: Likewise.
10598 * rtl.h (rtx_def::u2): Add a const_vector field.
10599 (CONST_VECTOR_NPATTERNS): New macro.
10600 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
10601 (CONST_VECTOR_DUPLICATE_P): Likewise.
10602 (CONST_VECTOR_STEPPED_P): Likewise.
10603 (CONST_VECTOR_ENCODED_ELT): Likewise.
10604 (const_vec_duplicate_p): Check for a duplicated vector encoding.
10605 (unwrap_const_vec_duplicate): Likewise.
10606 (const_vec_series_p): Check for a non-duplicated vector encoding.
10607 Say that the function only returns true for integer vectors.
10608 * emit-rtl.c: Include rtx-vector-builder.h.
10609 (gen_const_vec_duplicate_1): Delete.
10610 (gen_const_vector): Call gen_const_vec_duplicate instead of
10611 gen_const_vec_duplicate_1.
10612 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
10613 (gen_const_vec_duplicate): Use rtx_vector_builder.
10614 (gen_const_vec_series): Likewise.
10615 (gen_rtx_CONST_VECTOR): Likewise.
10616 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
10617 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10618 Build a new vector rather than modifying a CONST_VECTOR in-place.
10619 (handle_special_swappables): Update call accordingly.
10620 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
10621 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10622 Build a new vector rather than modifying a CONST_VECTOR in-place.
10623 (handle_special_swappables): Update call accordingly.
10625 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10627 * simplify-rtx.c (simplify_const_binary_operation): Use
10628 CONST_VECTOR_ELT instead of XVECEXP.
10630 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10632 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
10633 the selector elements to be different from the data elements
10634 if the selector is a VECTOR_CST.
10635 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
10636 ssizetype for the selector.
10638 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10640 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
10641 before testing each element individually.
10642 * tree-vect-generic.c (lower_vec_perm): Likewise.
10644 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10646 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
10647 * selftest-run-tests.c (selftest::run_tests): Call it.
10648 * vector-builder.h (vector_builder::operator ==): New function.
10649 (vector_builder::operator !=): Likewise.
10650 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
10651 (vec_perm_indices::all_from_input_p): New function.
10652 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10653 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
10654 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
10655 instead of reading the VECTOR_CST directly. Detect whether both
10656 vector inputs are the same before constructing the vec_perm_indices,
10657 and update the number of inputs argument accordingly. Use the
10658 utility functions added above. Only construct sel2 if we need to.
10660 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10662 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
10663 the broadcast of the low byte.
10664 (expand_mult_highpart): Use an explicit encoding for the permutes.
10665 * optabs-query.c (can_mult_highpart_p): Likewise.
10666 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
10667 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10668 (vectorizable_bswap): Likewise.
10669 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
10670 explicit encoding for the power-of-2 permutes.
10671 (vect_permute_store_chain): Likewise.
10672 (vect_grouped_load_supported): Likewise.
10673 (vect_permute_load_chain): Likewise.
10675 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10677 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
10678 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
10679 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
10680 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10681 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
10682 (vect_gen_perm_mask_any): Likewise.
10684 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10686 * int-vector-builder.h: New file.
10687 * vec-perm-indices.h: Include int-vector-builder.h.
10688 (vec_perm_indices): Redefine as an int_vector_builder.
10689 (auto_vec_perm_indices): Delete.
10690 (vec_perm_builder): Redefine as a stand-alone class.
10691 (vec_perm_indices::vec_perm_indices): New function.
10692 (vec_perm_indices::clamp): Likewise.
10693 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
10694 (vec_perm_indices::new_vector): New function.
10695 (vec_perm_indices::new_expanded_vector): Update for new
10696 vec_perm_indices class.
10697 (vec_perm_indices::rotate_inputs): New function.
10698 (vec_perm_indices::all_in_range_p): Operate directly on the
10699 encoded form, without computing elided elements.
10700 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
10701 encoding. Update for new vec_perm_indices class.
10702 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
10703 the given vec_perm_builder.
10704 (expand_vec_perm_var): Update vec_perm_builder constructor.
10705 (expand_mult_highpart): Use vec_perm_builder instead of
10706 auto_vec_perm_indices.
10707 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
10708 vec_perm_indices instead of auto_vec_perm_indices. Use a single
10709 or double series encoding as appropriate.
10710 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
10711 vec_perm_indices instead of auto_vec_perm_indices.
10712 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10713 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10714 (vect_permute_store_chain): Likewise.
10715 (vect_grouped_load_supported): Likewise.
10716 (vect_permute_load_chain): Likewise.
10717 (vect_shift_permute_load_chain): Likewise.
10718 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10719 (vect_transform_slp_perm_load): Likewise.
10720 (vect_schedule_slp_instance): Likewise.
10721 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10722 (vectorizable_mask_load_store): Likewise.
10723 (vectorizable_bswap): Likewise.
10724 (vectorizable_store): Likewise.
10725 (vectorizable_load): Likewise.
10726 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
10727 vec_perm_indices instead of auto_vec_perm_indices. Use
10728 tree_to_vec_perm_builder to read the vector from a tree.
10729 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
10730 vec_perm_builder instead of a vec_perm_indices.
10731 (have_whole_vector_shift): Use vec_perm_builder and
10732 vec_perm_indices instead of auto_vec_perm_indices. Leave the
10733 truncation to calc_vec_perm_mask_for_shift.
10734 (vect_create_epilog_for_reduction): Likewise.
10735 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
10736 from auto_vec_perm_indices to vec_perm_indices.
10737 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10738 instead of changing individual elements.
10739 (aarch64_vectorize_vec_perm_const): Use new_vector to install
10740 the vector in d.perm.
10741 * config/arm/arm.c (expand_vec_perm_d::perm): Change
10742 from auto_vec_perm_indices to vec_perm_indices.
10743 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10744 instead of changing individual elements.
10745 (arm_vectorize_vec_perm_const): Use new_vector to install
10746 the vector in d.perm.
10747 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
10748 Update vec_perm_builder constructor.
10749 (rs6000_expand_interleave): Likewise.
10750 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
10751 (rs6000_expand_interleave): Likewise.
10753 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10755 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
10756 to qimode could truncate the indices.
10757 * optabs.c (expand_vec_perm_var): Likewise.
10759 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10761 * Makefile.in (OBJS): Add vec-perm-indices.o.
10762 * vec-perm-indices.h: New file.
10763 * vec-perm-indices.c: Likewise.
10764 * target.h (vec_perm_indices): Replace with a forward class
10766 (auto_vec_perm_indices): Move to vec-perm-indices.h.
10767 * optabs.h: Include vec-perm-indices.h.
10768 (expand_vec_perm): Delete.
10769 (selector_fits_mode_p, expand_vec_perm_var): Declare.
10770 (expand_vec_perm_const): Declare.
10771 * target.def (vec_perm_const_ok): Replace with...
10772 (vec_perm_const): ...this new hook.
10773 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
10774 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
10775 * doc/tm.texi: Regenerate.
10776 * optabs.def (vec_perm_const): Delete.
10777 * doc/md.texi (vec_perm_const): Likewise.
10778 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
10779 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
10780 expand_vec_perm for constant permutation vectors. Assert that
10781 the mode of variable permutation vectors is the integer equivalent
10782 of the mode that is being permuted.
10783 * optabs-query.h (selector_fits_mode_p): Declare.
10784 * optabs-query.c: Include vec-perm-indices.h.
10785 (selector_fits_mode_p): New function.
10786 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
10787 is defined, instead of checking whether the vec_perm_const_optab
10788 exists. Use targetm.vectorize.vec_perm_const instead of
10789 targetm.vectorize.vec_perm_const_ok. Check whether the indices
10790 fit in the vector mode before using a variable permute.
10791 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
10792 vec_perm_indices instead of an rtx.
10793 (expand_vec_perm): Replace with...
10794 (expand_vec_perm_const): ...this new function. Take the selector
10795 as a vec_perm_indices rather than an rtx. Also take the mode of
10796 the selector. Update call to shift_amt_for_vec_perm_mask.
10797 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
10798 Use vec_perm_indices::new_expanded_vector to expand the original
10799 selector into bytes. Check whether the indices fit in the vector
10800 mode before using a variable permute.
10801 (expand_vec_perm_var): Make global.
10802 (expand_mult_highpart): Use expand_vec_perm_const.
10803 * fold-const.c: Includes vec-perm-indices.h.
10804 * tree-ssa-forwprop.c: Likewise.
10805 * tree-vect-data-refs.c: Likewise.
10806 * tree-vect-generic.c: Likewise.
10807 * tree-vect-loop.c: Likewise.
10808 * tree-vect-slp.c: Likewise.
10809 * tree-vect-stmts.c: Likewise.
10810 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
10812 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
10813 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
10814 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
10815 (aarch64_vectorize_vec_perm_const): ...this new function.
10816 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10817 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10818 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
10819 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
10820 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10821 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10822 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
10824 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
10825 check for NEON modes.
10826 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
10827 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
10828 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
10829 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
10831 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
10832 the old VEC_PERM_CONST conditions.
10833 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
10834 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
10835 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
10836 (ia64_vectorize_vec_perm_const_ok): Merge into...
10837 (ia64_vectorize_vec_perm_const): ...this new function.
10838 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
10839 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
10840 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
10841 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
10842 * config/mips/mips.c (mips_expand_vec_perm_const)
10843 (mips_vectorize_vec_perm_const_ok): Merge into...
10844 (mips_vectorize_vec_perm_const): ...this new function.
10845 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
10846 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
10847 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
10848 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
10849 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
10850 (rs6000_expand_vec_perm_const): Delete.
10851 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
10853 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10854 (altivec_expand_vec_perm_const_le): Take each operand individually.
10855 Operate on constant selectors rather than rtxes.
10856 (altivec_expand_vec_perm_const): Likewise. Update call to
10857 altivec_expand_vec_perm_const_le.
10858 (rs6000_expand_vec_perm_const): Delete.
10859 (rs6000_vectorize_vec_perm_const_ok): Delete.
10860 (rs6000_vectorize_vec_perm_const): New function.
10861 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10862 an element count and rtx array.
10863 (rs6000_expand_extract_even): Update call accordingly.
10864 (rs6000_expand_interleave): Likewise.
10865 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
10866 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
10867 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
10868 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
10869 (rs6000_expand_vec_perm_const): Delete.
10870 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10871 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10872 (altivec_expand_vec_perm_const_le): Take each operand individually.
10873 Operate on constant selectors rather than rtxes.
10874 (altivec_expand_vec_perm_const): Likewise. Update call to
10875 altivec_expand_vec_perm_const_le.
10876 (rs6000_expand_vec_perm_const): Delete.
10877 (rs6000_vectorize_vec_perm_const_ok): Delete.
10878 (rs6000_vectorize_vec_perm_const): New function. Remove stray
10879 reference to the SPE evmerge intructions.
10880 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10881 an element count and rtx array.
10882 (rs6000_expand_extract_even): Update call accordingly.
10883 (rs6000_expand_interleave): Likewise.
10884 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
10885 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
10887 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10889 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10891 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
10892 vector mode and that that mode matches the mode of the data
10894 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
10895 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
10896 directly using expand_vec_perm_1 when forcing selectors into
10898 (expand_vec_perm_var): New function, split out from expand_vec_perm.
10900 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10902 * optabs-query.h (can_vec_perm_p): Delete.
10903 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
10904 * optabs-query.c (can_vec_perm_p): Split into...
10905 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
10906 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
10907 particular selector is valid.
10908 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10909 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10910 (vect_grouped_load_supported): Likewise.
10911 (vect_shift_permute_load_chain): Likewise.
10912 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10913 (vect_transform_slp_perm_load): Likewise.
10914 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10915 (vectorizable_bswap): Likewise.
10916 (vect_gen_perm_mask_checked): Likewise.
10917 * fold-const.c (fold_ternary_loc): Likewise. Don't take
10918 implementations of variable permutation vectors into account
10919 when deciding which selector to use.
10920 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
10921 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
10922 with a false third argument.
10923 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
10924 to test whether the constant selector is valid and can_vec_perm_var_p
10925 to test whether a variable selector is valid.
10927 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10929 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
10930 * optabs-query.c (can_vec_perm_p): Likewise.
10931 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
10932 instead of vec_perm_indices.
10933 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
10934 (vect_gen_perm_mask_checked): Likewise,
10935 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
10936 (vect_gen_perm_mask_checked): Likewise,
10938 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10940 * optabs-query.h (qimode_for_vec_perm): Declare.
10941 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
10942 (qimode_for_vec_perm): ...this new function.
10943 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
10945 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10947 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
10948 does not have a conditional at the top.
10950 2018-01-02 Richard Biener <rguenther@suse.de>
10952 * ipa-inline.c (big_speedup_p): Fix expression.
10954 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10957 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
10960 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10964 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
10965 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
10966 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
10967 cond_taken_branch_cost 3->4.
10969 2018-01-01 Jakub Jelinek <jakub@redhat.com>
10971 PR tree-optimization/83581
10972 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
10973 TODO_cleanup_cfg if any changes have been made.
10975 PR middle-end/83608
10976 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
10977 convert_modes if target mode has the right side, but different mode
10980 PR middle-end/83609
10981 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
10982 last argument when extracting from CONCAT. If either from_real or
10983 from_imag is NULL, use expansion through memory. If result is not
10984 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
10985 the parts directly to inner mode, if even that fails, use expansion
10988 PR middle-end/83623
10989 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
10990 check for bswap in mode rather than HImode and use that in expand_unop
10993 Copyright (C) 2018 Free Software Foundation, Inc.
10995 Copying and distribution of this file, with or without modification,
10996 are permitted in any medium without royalty provided the copyright
10997 notice and this notice are preserved.