1 2014-03-11 Joern Rennecke <joern.rennecke@embecosm.com>
3 * config/arc/predicates.md (extend_operand): During/after reload,
4 allow const_int_operand.
5 * config/arc/arc.md (mulsidi3_700): Use extend_operand predicate.
6 (umulsidi3_700): Likewise. Change operand 2 constraint back to "cL".
7 (mulsi3_highpart): Change operand 2 constraint alternatives 2 and 3
9 (umulsi3_highpart_i): Likewise.
11 2014-03-11 Richard Biener <rguenther@suse.de>
13 * tree-ssa-structalias.c (get_constraint_for_ptr_offset):
14 Add asserts to guard possible wrong-code bugs.
16 2014-03-11 Richard Biener <rguenther@suse.de>
18 PR tree-optimization/60429
19 PR tree-optimization/60485
20 * tree-ssa-structalias.c (set_union_with_increment): Properly
21 take into account all fields that overlap the shifted vars.
22 (do_sd_constraint): Likewise.
23 (do_ds_constraint): Likewise.
24 (get_constraint_for_ptr_offset): Likewise.
26 2014-03-11 Chung-Lin Tang <cltang@codesourcery.com>
28 * config/nios2/nios2.c (machine_function): Add fp_save_offset field.
29 (nios2_compute_frame_layout):
30 Add calculation of cfun->machine->fp_save_offset.
31 (nios2_expand_prologue): Correct setting of frame pointer register
33 (nios2_expand_epilogue): Update recovery of stack pointer from
34 frame pointer accordingly.
35 (nios2_initial_elimination_offset): Update calculation of offset
36 for eliminating to HARD_FRAME_POINTER_REGNUM.
38 2014-03-10 Jakub Jelinek <jakub@redhat.com>
41 * ipa.c (symtab_remove_unreachable_nodes): Don't call
42 cgraph_get_create_node on VAR_DECLs.
44 2014-03-10 Richard Biener <rguenther@suse.de>
47 * tree.c (signed_or_unsigned_type_for): Handle OFFSET_TYPEs.
49 2014-03-08 Douglas B Rupp <rupp@gnat.com>
51 * config/vms/vms.opt (vms_float_format): New variable.
53 2014-03-08 Tobias Burnus <burnus@net-b.de>
55 * doc/invoke.texi (-fcilkplus): Update implementation
58 2014-03-08 Paulo Matos <paulo@matos-sorge.com>
59 Richard Biener <rguenther@suse.de>
61 * lto-wrapper.c (merge_and_complain): Ensure -fshort-double is used
62 consistently accross all TUs.
63 (run_gcc): Enable -fshort-double automatically at link at link-time
64 and disallow override.
66 2014-03-08 Richard Sandiford <rdsandiford@googlemail.com>
69 * config/mips/mips.c (mips_option_override): Promote -mpaired-single
70 warning to an error. Disable TARGET_PAIRED_SINGLE and TARGET_MIPS3D
71 if they can't be used.
73 2014-03-07 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
75 * configure.ac (HAVE_AS_IX86_TLSLDMPLT): Improve test for Solaris
77 * configure: Regenerate.
79 2014-03-07 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
81 * configure.ac (TLS_SECTION_ASM_FLAG): Save as tls_section_flag.
82 (LIB_TLS_SPEC): Save as ld_tls_libs.
83 (HAVE_AS_IX86_TLSLDMPLT): Define as 1/0.
84 (HAVE_AS_IX86_TLSLDM): New test.
85 * configure, config.in: Regenerate.
86 * config/i386/i386.c (legitimize_tls_address): Fall back to
87 TLS_MODEL_GLOBAL_DYNAMIC on 32-bit Solaris/x86 if tool chain
88 cannot support TLS_MODEL_LOCAL_DYNAMIC.
89 * config/i386/i386.md (*tls_local_dynamic_base_32_gnu): Use if
90 instead of #ifdef in HAVE_AS_IX86_TLSLDMPLT test.
92 2014-03-07 Paulo Matos <paulo@matos-sorge.com>
94 * common.opt (fira-loop-pressure): Mark as optimization.
96 2014-03-07 Thomas Schwinge <thomas@codesourcery.com>
98 * langhooks.c (lhd_omp_mappable_type): The error_mark_node is not
99 an OpenMP mappable type.
101 2014-03-06 Matthias Klose <doko@ubuntu.com>
103 * Makefile.in (s-mlib): Only pass MULTIARCH_DIRNAME if
104 MULTILIB_OSDIRNAMES is not defined.
106 2014-03-06 Jakub Jelinek <jakub@redhat.com>
107 Meador Inge <meadori@codesourcery.com>
110 * config/arm/arm.c (arm_tls_symbol_p): Remove.
111 (arm_legitimize_address): Call legitimize_tls_address for any
112 arm_tls_referenced_p expression, handle constant addend. Call it
113 before testing for !TARGET_ARM.
114 (thumb_legitimize_address): Don't handle arm_tls_symbol_p here.
116 2014-03-06 Richard Biener <rguenther@suse.de>
122 2014-03-04 Paulo Matos <paulo@matos-sorge.com>
124 * tree-streamer.c (record_common_node): Assert we don't record
125 nodes with type double.
126 (preload_common_node): Skip type double, complex double and
127 double pointer since it is now frontend dependent due to
128 fshort-double option.
130 2014-03-06 Richard Biener <rguenther@suse.de>
132 * gcc.c (PLUGIN_COND): Always enable unless -fno-use-linker-plugin
133 or -fno-lto is specified and the linker has full plugin support.
134 * collect2.c (lto_mode): Default to LTO_MODE_WHOPR if LTO is
136 (main): Remove -flto processing, adjust lto_mode using
138 * lto-wrapper.c (merge_and_complain): Merge compile-time
140 (run_gcc): And pass it through to the link options.
142 2014-03-06 Alexandre Oliva <aoliva@redhat.com>
146 2014-02-28 Alexandre Oliva <aoliva@redhat.com>
148 * cselib.c (remove_useless_values): Skip to avoid quadratic
149 behavior if the condition moved from...
150 (cselib_process_insn): ... here holds.
152 2014-03-05 Jakub Jelinek <jakub@redhat.com>
155 * Makefile.in (PLUGIN_HEADERS): Add tree-phinodes.h, stor-layout.h,
156 ssa-iterators.h, $(RESOURCE_H) and tree-cfgcleanup.h.
159 * config/i386/t-i386 (OPTIONS_H_EXTRA): Add stringop.def.
160 (TM_H): Add x86-tune.def.
162 2014-03-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
164 * config/aarch64/aarch64.c (generic_tunings):
165 Use cortexa57_extra_costs.
167 2014-03-05 Jakub Jelinek <jakub@redhat.com>
170 * cfgexpand.c (expand_used_vars): Do not assume all SSA_NAMEs
171 of PARM/RESULT_DECLs must be coalesced with optimize && in_lto_p.
172 * tree-ssa-coalesce.c (coalesce_ssa_name): Use MUST_COALESCE_COST - 1
175 2014-03-04 Heiher <r@hev.cc>
177 * config/mips/mips-cpus.def (loongson3a): Mark as a MIPS64r2 processor.
178 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Adjust accordingly.
180 2014-03-04 Uros Bizjak <ubizjak@gmail.com>
182 * config/i386/predicates.md (const2356_operand): Change to ...
183 (const2367_operand): ... this.
184 * config/i386/sse.md (avx512pf_scatterpf<mode>sf): Use
186 (*avx512pf_scatterpf<mode>sf_mask): Ditto.
187 (*avx512pf_scatterpf<mode>sf): Ditto.
188 (avx512pf_scatterpf<mode>df): Ditto.
189 (*avx512pf_scatterpf<mode>df_mask): Ditto.
190 (*avx512pf_scatterpf<mode>df): Ditto.
191 * config/i386/i386.c (ix86_expand_builtin): Update
192 incorrect hint operand error message.
194 2014-03-04 Richard Biener <rguenther@suse.de>
196 * lto-section-in.c (lto_get_section_data): Fix const cast.
198 2014-03-04 Paulo Matos <paulo@matos-sorge.com>
200 * tree-streamer.c (record_common_node): Assert we don't record
201 nodes with type double.
202 (preload_common_node): Skip type double, complex double and
203 double pointer since it is now frontend dependent due to
204 fshort-double option.
206 2014-03-04 Richard Biener <rguenther@suse.de>
209 * lto-streamer-in.c (lto_read_body): Remove LTO bytecode version check.
210 (lto_input_toplevel_asms): Likewise.
211 * lto-section-in.c (lto_get_section_data): Instead do it here
214 2014-03-04 Richard Biener <rguenther@suse.de>
216 PR tree-optimization/60382
217 * tree-vect-loop.c (vect_is_simple_reduction_1): Do not consider
218 dead PHIs a reduction.
220 2014-03-03 Uros Bizjak <ubizjak@gmail.com>
222 * config/i386/xmmintrin.h (enum _mm_hint) <_MM_HINT_ET0>: Correct
224 (_mm_prefetch): Move out of GCC target("sse") pragma.
225 * config/i386/prfchwintrin.h (_m_prefetchw): Move out of
226 GCC target("prfchw") pragma.
227 * config/i386/i386.md (prefetch): Emit prefetchwt1 only
229 * config/i386/i386.c (ix86_option_override_internal): Enable
230 -mprfchw with -mprefetchwt1.
232 2014-03-03 Joern Rennecke <joern.rennecke@embecosm.com>
234 * config/arc/arc.md (casesi_load) <length attribute alternative 0>:
237 2014-03-03 Joern Rennecke <joern.rennecke@embecosm.com>
239 * opts.h (CL_PCH_IGNORE): Define.
240 * targhooks.c (option_affects_pch_p):
241 Return false for options that have CL_PCH_IGNORE set.
242 * opt-functions.awk: Process PchIgnore.
243 * doc/options.texi: Document PchIgnore.
245 * config/arc/arc.opt (misize): Add PchIgnore property.
247 2014-03-03 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
249 * config/rs6000/rs6000.c (rs6000_preferred_reload_class): Disallow
250 reload of PLUS rtx's outside of GENERAL_REGS or BASE_REGS; relax
251 constraint on constants to permit them being loaded into
252 GENERAL_REGS or BASE_REGS.
254 2014-03-03 Nick Clifton <nickc@redhat.com>
256 * config/rl78/rl78-real.md (cbranchsi4_real_signed): Add
257 anti-cacnonical alternatives.
258 (negandhi3_real): New pattern.
259 * config/rl78/rl78-virt.md (negandhi3_virt): New pattern.
261 2014-03-03 Senthil Kumar Selvaraj <senthil_kumar.selvaraj@atmel.com>
263 * config/avr/avr-mcus.def: Remove atxmega16x1.
264 * config/avr/avr-tables.opt: Regenerate.
265 * config/avr/t-multilib: Regenerate.
266 * doc/avr-mmcu.texi: Regenerate.
268 2014-03-03 Tobias Grosser <tobias@grosser.es>
269 Mircea Namolaru <mircea.namolaru@inria.fr>
271 PR tree-optimization/58028
272 * graphite-clast-to-gimple.c (set_cloog_options): Don't remove
275 2014-03-03 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
277 * config/arm/neon.md (*movmisalign<mode>): Legitimize addresses
278 not handled by recognizers.
280 2014-03-03 Jakub Jelinek <jakub@redhat.com>
283 * function.c (expand_function_end): Don't emit
284 clobber_return_register sequence if clobber_after is a BARRIER.
285 * cfgexpand.c (construct_exit_block): Append instructions before
286 return_label to prev_bb.
288 2014-03-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
290 * config/rs6000/constraints.md: Document reserved use of "wc".
292 2014-03-02 Jan Hubicka <hubicka@ucw.cz>
295 * ipa.c (function_and_variable_visibility): When dissolving comdat
296 group, also set all symbols to local.
298 2014-03-02 Jan Hubicka <hubicka@ucw.cz>
303 2013-12-14 Jan Hubicka <jh@suse.cz>
305 * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Skip clobbers.
307 2014-03-02 Jon Beniston <jon@beniston.com>
313 * config/lm32/lm32.c (lm32_legitimate_constant_p): Remove, as incorrect.
314 (TARGET_LEGITIMATE_CONSTANT_P): Undefine, as not needed.
315 * config/lm32/lm32.md (movsi_insn): Add 32-bit immediate support.
316 (simple_return, *simple_return): New patterns
317 * config/lm32/predicates.md (movsi_rhs_operand): Remove as obsolete.
318 * configure.ac (force_sjlj_exceptions): Force sjlj exceptions for lm32.
320 2014-03-01 Paolo Carlini <paolo.carlini@oracle.com>
322 * dwarf2out.c (gen_subprogram_die): Tidy.
324 2014-03-01 Oleg Endo <olegendo@gcc.gnu.org>
327 * config/sh/sh.md (*mov_t_msb_neg): Split into ...
328 (*mov_t_msb_neg_negc): ... this new insn.
330 2014-02-28 Jason Merrill <jason@redhat.com>
333 * ipa-devirt.c (ipa_devirt): Don't choose an implicitly-declared
336 2014-02-28 Paolo Carlini <paolo.carlini@oracle.com>
339 * dwarf2out.c (decltype_auto_die): New static.
340 (gen_subprogram_die): Handle 'decltype(auto)' like 'auto'.
341 (gen_type_die_with_usage): Handle 'decltype(auto)'.
342 (is_cxx_auto): Likewise.
344 2014-02-28 Ian Bolton <ian.bolton@arm.com>
346 * config/aarch64/aarch64.h: Define __ARM_NEON by default if
347 we are not using general regs only.
349 2014-02-28 Richard Biener <rguenther@suse.de>
352 * tree-cfgcleanup.c (tree_forwarder_block_p): Restrict
353 previous fix and only allow to remove trivial pre-headers
354 and latches. Also honor LOOPS_MAY_HAVE_MULTIPLE_LATCHES.
355 (remove_forwarder_block): Properly update the latch of a loop.
357 2014-02-28 Alexandre Oliva <aoliva@redhat.com>
360 * cselib.c (cselib_hasher::equal): Special-case VALUE lookup.
361 (cselib_preserved_hash_table): New.
362 (preserve_constants_and_equivs): Move preserved vals to it.
363 (cselib_find_slot): Look it up first.
364 (cselib_init): Initialize it.
365 (cselib_finish): Release it.
366 (dump_cselib_table): Dump it.
368 2014-02-28 Alexandre Oliva <aoliva@redhat.com>
371 * cselib.c (remove_useless_values): Skip to avoid quadratic
372 behavior if the condition moved from...
373 (cselib_process_insn): ... here holds.
375 2014-02-28 Alexandre Oliva <aoliva@redhat.com>
378 * var-tracking.c (vt_initialize): Apply the same condition to
379 preserve the CFA base value.
381 2014-02-28 Joey Ye <joey.ye@arm.com>
384 * config/arm/arm.c (thumb_far_jump_used_p): Don't change
385 if reload in progress or completed.
387 2014-02-28 Tobias Burnus <burnus@net-b.de>
390 * tree-pretty-print.c (dump_generic_node, print_declaration): Handle
393 2014-02-27 H.J. Lu <hongjiu.lu@intel.com>
395 * doc/tm.texi.in (Condition Code Status): Update documention for
396 relative locations of cc0-setter and cc0-user.
398 2014-02-27 Jeff Law <law@redhat.com>
400 PR rtl-optimization/52714
401 * combine.c (try_combine): When splitting an unrecognized PARALLEL
402 into two independent simple sets, if I3 is a jump, ensure the
403 pattern we place into I3 is a (set (pc) ...).
405 2014-02-27 Mikael Pettersson <mikpe@it.uu.se>
406 Jeff Law <law@redhat.com>
408 PR rtl-optimization/49847
409 * cse.c (fold_rtx) Handle case where cc0 setter and cc0 user
410 are in different blocks.
411 * doc/tm.texi (Condition Code Status): Update documention for
412 relative locations of cc0-setter and cc0-user.
414 2014-02-27 Vladimir Makarov <vmakarov@redhat.com>
417 * lra.c (lra_emit_add): Check SUBREG too.
419 2014-02-27 Andreas Schwab <schwab@suse.de>
421 * config/m68k/m68k.c (m68k_option_override): Disable
422 -flive-range-shrinkage for classic m68k.
423 (m68k_override_options_after_change): Likewise.
425 2014-02-27 Marek Polacek <polacek@redhat.com>
428 * tree-ssa-uninit.c (gate_warn_uninitialized): Run the pass even for
429 -Wmaybe-uninitialized.
431 2014-02-27 Alan Modra <amodra@gmail.com>
434 * reload1.c (emit_input_reload_insns): When reload_override_in,
435 set old to rl->in_reg when rl->in_reg is a subreg.
437 2014-02-26 Richard Biener <rguenther@suse.de>
440 * lra-assigns.c (spill_for): Avoid mixed-sign comparison.
442 2014-02-25 Ilya Tocar <ilya.tocar@intel.com>
444 * common/config/i386/predicates.md (const1256_operand): Remove.
445 (const2356_operand): New.
446 (const_1_to_2_operand): Remove.
447 * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
448 (*avx512pf_gatherpf<mode>sf_mask): Ditto.
449 (*avx512pf_gatherpf<mode>sf): Ditto.
450 (avx512pf_gatherpf<mode>df): Ditto.
451 (*avx512pf_gatherpf<mode>df_mask): Ditto.
452 (*avx512pf_gatherpf<mode>df): Ditto.
453 (avx512pf_scatterpf<mode>sf): Ditto.
454 (*avx512pf_scatterpf<mode>sf_mask): Ditto.
455 (*avx512pf_scatterpf<mode>sf): Ditto.
456 (avx512pf_scatterpf<mode>df): Ditto.
457 (*avx512pf_scatterpf<mode>df_mask): Ditto.
458 (*avx512pf_scatterpf<mode>df): Ditto.
459 * common/config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET0.
461 2014-02-26 Ilya Tocar <ilya.tocar@intel.com>
463 * config/i386/avx512fintrin.h (_mm512_testn_epi32_mask),
464 (_mm512_mask_testn_epi32_mask), (_mm512_testn_epi64_mask),
465 (_mm512_mask_testn_epi64_mask): Move to ...
466 * config/i386/avx512cdintrin.h: Here.
467 * config/i386/i386.c (bdesc_args): Change MASK_ISA for testnm.
468 * config/i386/sse.md (avx512f_vmscalef<mode><round_name>): Remove %.
469 (avx512f_scalef<mode><mask_name><round_name>): Ditto.
470 (avx512f_testnm<mode>3<mask_scalar_merge_name>): Change conditon to
471 TARGET_AVX512F from TARGET_AVX512CD.
473 2014-02-26 Richard Biener <rguenther@suse.de>
476 * ipa.c (walk_polymorphic_call_targets): Properly guard
477 call to inline_update_overall_summary.
479 2014-02-26 Bin Cheng <bin.cheng@arm.com>
482 * tree-cfgcleanup.c (tree_forwarder_block_p): Protect loop preheaders
483 and latches only if requested. Fix latch if it is removed.
484 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Set
485 LOOPS_HAVE_PREHEADERS.
487 2014-02-25 Andrew Pinski <apinski@cavium.com>
489 * builtins.c (expand_builtin_thread_pointer): Create a new target
490 when the target is NULL.
492 2014-02-25 Vladimir Makarov <vmakarov@redhat.com>
494 PR rtl-optimization/60317
495 * params.def (PARAM_LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
496 * params.h (LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
497 * lra-assigns.c: Include params.h.
498 (spill_for): Use LRA_MAX_CONSIDERED_RELOAD_PSEUDOS as guard for
499 other reload pseudos considerations.
501 2014-02-25 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
503 * config/rs6000/vector.md (*vector_unordered<mode>): Change split
504 to use canonical form for nor<mode>3.
506 2014-02-25 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
509 * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Allow 128 to 64-bit
512 2014-02-25 Ilya Tocar <ilya.tocar@intel.com>
514 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PREFETCHWT1_SET),
515 (OPTION_MASK_ISA_PREFETCHWT1_UNSET): New.
516 (ix86_handle_option): Handle OPT_mprefetchwt1.
517 * config/i386/cpuid.h (bit_PREFETCHWT1): New.
518 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
520 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
521 OPTION_MASK_ISA_PREFETCHWT1.
522 * config/i386/i386.c (ix86_target_string): Handle mprefetchwt1.
523 (PTA_PREFETCHWT1): New.
524 (ix86_option_override_internal): Handle PTA_PREFETCHWT1.
525 (ix86_valid_target_attribute_inner_p): Handle OPT_mprefetchwt1.
526 * config/i386/i386.h (TARGET_PREFETCHWT1, TARGET_PREFETCHWT1_P): New.
527 * config/i386/i386.md (prefetch): Check TARGET_PREFETCHWT1
528 (*prefetch_avx512pf_<mode>_: Change into ...
529 (*prefetch_prefetchwt1_<mode>: This.
530 * config/i386/i386.opt (mprefetchwt1): New.
531 * config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET1.
532 (_mm_prefetch): Handle intent to write.
533 * doc/invoke.texi (mprefetchwt1), (mno-prefetchwt1): Doccument.
535 2014-02-25 Richard Biener <rguenther@suse.de>
538 * emit-rtl.c (mem_attrs_htab): Remove.
539 (mem_attrs_htab_hash): Likewise.
540 (mem_attrs_htab_eq): Likewise.
541 (set_mem_attrs): Always allocate new mem-attrs when something changed.
542 (init_emit_once): Do not allocate mem_attrs_htab.
544 2014-02-25 Richard Biener <rguenther@suse.de>
547 * lto-opts.c (lto_write_options): Output non-explicit conservative
548 -fwrapv, -fno-trapv and -fno-strict-overflow.
549 * lto-wrapper.c (merge_and_complain): Handle merging those options.
550 (run_gcc): And pass them through.
552 2014-02-25 Andrey Belevantsev <abel@ispras.ru>
554 * sel-sched.c (calculate_new_fences): New parameter ptime.
555 Calculate it as a maximum over all fence cycles.
556 (sel_sched_region_2): Adjust the call to calculate_new_fences.
557 Print the final schedule timing when sched_verbose.
559 2014-02-25 Andrey Belevantsev <abel@ispras.ru>
561 PR rtl-optimization/60292
562 * sel-sched.c (fill_vec_av_set): Do not reset target availability
563 bit fot the fence instruction.
565 2014-02-24 Alangi Derick <alangiderick@gmail.com>
567 * calls.h: Fix typo in comment.
569 2014-02-24 John David Anglin <danglin@gcc.gnu.org>
571 * config/pa/pa.c (pa_output_move_double): Don't valididate when
572 adjusting offsetable addresses.
574 2014-02-24 Guozhi Wei <carrot@google.com>
576 * sparseset.h (sparseset_pop): Fix the wrong index.
578 2014-02-24 Walter Lee <walt@tilera.com>
580 * config.gcc (tilepro-*-*): Change to tilepro*-*-*.
581 (tilegx-*-linux*): Change to tilegx*-*-linux*; Support tilegxbe
583 * common/config/tilegx/tilegx-common.c
584 (TARGET_DEFAULT_TARGET_FLAGS): Define.
585 * config/tilegx/linux.h (ASM_SPEC): Add endian_spec.
587 * config/tilegx/sync.md (atomic_test_and_set): Handle big endian.
588 * config/tilegx/tilegx.c (tilegx_return_in_msb): New.
589 (tilegx_gimplify_va_arg_expr): Handle big endian.
590 (tilegx_expand_unaligned_load): Ditto.
591 (tilegx_expand_unaligned_store): Ditto.
592 (TARGET_RETURN_IN_MSB): New.
593 * config/tilegx/tilegx.h (TARGET_DEFAULT): New.
594 (TARGET_ENDIAN_DEFAULT): New.
595 (TARGET_BIG_ENDIAN): Handle big endian.
596 (BYTES_BIG_ENDIAN): Ditto.
597 (WORDS_BIG_ENDIAN): Ditto.
598 (FLOAT_WORDS_BIG_ENDIAN): Ditto.
601 * config/tilegx/tilegx.md (extv): Handle big endian.
604 (insn_st<n>_add<bitsuffix>): Ditto.
605 (insn_stnt<n>): Ditto.
606 (insn_stnt<n>_add<bitsuffix>):Ditto.
607 (vec_interleave_highv8qi): Handle big endian.
608 (vec_interleave_highv8qi_be): New.
609 (vec_interleave_highv8qi_le): New.
610 (insn_v1int_h): Handle big endian.
611 (vec_interleave_lowv8qi): Handle big endian.
612 (vec_interleave_lowv8qi_be): New.
613 (vec_interleave_lowv8qi_le): New.
614 (insn_v1int_l): Handle big endian.
615 (vec_interleave_highv4hi): Handle big endian.
616 (vec_interleave_highv4hi_be): New.
617 (vec_interleave_highv4hi_le): New.
618 (insn_v2int_h): Handle big endian.
619 (vec_interleave_lowv4hi): Handle big endian.
620 (vec_interleave_lowv4hi_be): New.
621 (vec_interleave_lowv4hi_le): New.
622 (insn_v2int_l): Handle big endian.
623 (vec_interleave_highv2si): Handle big endian.
624 (vec_interleave_highv2si_be): New.
625 (vec_interleave_highv2si_le): New.
626 (insn_v4int_h): Handle big endian.
627 (vec_interleave_lowv2si): Handle big endian.
628 (vec_interleave_lowv2si_be): New.
629 (vec_interleave_lowv2si_le): New.
630 (insn_v4int_l): Handle big endian.
631 * config/tilegx/tilegx.opt (mbig-endian): New option.
632 (mlittle-endian): New option.
633 * doc/install.texi: Document tilegxbe-linux.
634 * doc/invoke.texi: Document -mbig-endian and -mlittle-endian.
636 2014-02-24 Martin Jambor <mjambor@suse.cz>
639 * ipa-cp.c (propagate_constants_accross_call): Bail out early if
640 there are no parameter descriptors.
642 2014-02-24 Andrey Belevantsev <abel@ispras.ru>
644 PR rtl-optimization/60268
645 * sched-rgn.c (haifa_find_rgns): Move the nr_regions_initial variable
646 initialization to ...
647 (sched_rgn_init): ... here.
648 (schedule_region): Check for SCHED_PRESSURE_NONE earlier.
650 2014-02-23 David Holsgrove <david.holsgrove@xilinx.com>
652 * config/microblaze/microblaze.md: Correct ashrsi_reg / lshrsi_reg
655 2014-02-23 Edgar E. Iglesias <edgar.iglesias@xilinx.com>
657 * config/microblaze/microblaze.h: Remove SECONDARY_MEMORY_NEEDED
660 2014-02-23 David Holsgrove <david.holsgrove@xilinx.com>
662 * /config/microblaze/microblaze.c: Add microblaze_asm_output_mi_thunk,
663 define TARGET_ASM_OUTPUT_MI_THUNK and TARGET_ASM_CAN_OUTPUT_MI_THUNK.
665 2014-02-23 David Holsgrove <david.holsgrove@xilinx.com>
667 * config/microblaze/predicates.md: Add cmp_op predicate.
668 * config/microblaze/microblaze.md: Add branch_compare instruction
669 which uses cmp_op predicate and emits cmp insn before branch.
670 * config/microblaze/microblaze.c (microblaze_emit_compare): Rename
671 to microblaze_expand_conditional_branch and consolidate logic.
672 (microblaze_expand_conditional_branch): emit branch_compare
673 insn instead of handling cmp op separate from branch insn.
675 2014-02-23 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
677 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Relax assert
680 2014-02-23 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
682 * config/rs6000/altivec.md (altivec_lve<VI_char>x): Replace
683 define_insn with define_expand and new define_insn
684 *altivec_lve<VI_char>x_internal.
685 (altivec_stve<VI_char>x): Replace define_insn with define_expand
686 and new define_insn *altivec_stve<VI_char>x_internal.
687 * config/rs6000/rs6000-protos.h (altivec_expand_stvex_be): New
689 * config/rs6000/rs6000.c (altivec_expand_lvx_be): Document use by
691 (altivec_expand_stvex_be): New function.
693 2014-02-22 Joern Rennecke <joern.rennecke@embecosm.com>
695 * config/avr/avr.c (avr_can_eliminate): Allow elimination from
696 ARG_POINTER_REGNUM to STACK_POINTER_REGNUM if !frame_pointer_needed.
697 * config/avr/avr.c (ELIMINABLE_REGS): Add elimination from
698 ARG_POINTER_REGNUM to STACK_POINTER_REGNUM.
700 2014-02-21 Vladimir Makarov <vmakarov@redhat.com>
703 * lra-constraints.c (inherit_reload_reg): Use lra_emit_move
704 instead of emit_move_insn.
706 2014-02-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
708 * config/rs6000/altivec.md (altivec_vsumsws): Replace second
710 (reduc_uplus_v16qi): Use gen_altivec_vsumsws_direct instead of
713 2014-02-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
715 * config/rs6000/altivec.md (altivec_lvxl): Rename as
716 *altivec_lvxl_<mode>_internal and use VM2 iterator instead of V4SI.
717 (altivec_lvxl_<mode>): New define_expand incorporating
718 -maltivec=be semantics where needed.
719 (altivec_lvx): Rename as *altivec_lvx_<mode>_internal.
720 (altivec_lvx_<mode>): New define_expand incorporating -maltivec=be
721 semantics where needed.
722 (altivec_stvx): Rename as *altivec_stvx_<mode>_internal.
723 (altivec_stvx_<mode>): New define_expand incorporating
724 -maltivec=be semantics where needed.
725 (altivec_stvxl): Rename as *altivec_stvxl_<mode>_internal and use
726 VM2 iterator instead of V4SI.
727 (altivec_stvxl_<mode>): New define_expand incorporating
728 -maltivec=be semantics where needed.
729 * config/rs6000/rs6000-builtin.def: Add new built-in definitions
730 LVXL_V2DF, LVXL_V2DI, LVXL_V4SF, LVXL_V4SI, LVXL_V8HI, LVXL_V16QI,
731 LVX_V2DF, LVX_V2DI, LVX_V4SF, LVX_V4SI, LVX_V8HI, LVX_V16QI, STVX_V2DF,
732 STVX_V2DI, STVX_V4SF, STVX_V4SI, STVX_V8HI, STVX_V16QI, STVXL_V2DF,
733 STVXL_V2DI, STVXL_V4SF, STVXL_V4SI, STVXL_V8HI, STVXL_V16QI.
734 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Replace
735 ALTIVEC_BUILTIN_LVX with ALTIVEC_BUILTIN_LVX_<MODE> throughout;
736 similarly for ALTIVEC_BUILTIN_LVXL, ALTIVEC_BUILTIN_STVX, and
737 ALTIVEC_BUILTIN_STVXL.
738 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be): New prototype.
739 (altivec_expand_stvx_be): Likewise.
740 * config/rs6000/rs6000.c (swap_selector_for_mode): New function.
741 (altivec_expand_lvx_be): Likewise.
742 (altivec_expand_stvx_be): Likewise.
743 (altivec_expand_builtin): Add cases for
744 ALTIVEC_BUILTIN_STVX_<MODE>, ALTIVEC_BUILTIN_STVXL_<MODE>,
745 ALTIVEC_BUILTIN_LVXL_<MODE>, and ALTIVEC_BUILTIN_LVX_<MODE>.
746 (altivec_init_builtins): Add definitions for
747 __builtin_altivec_lvxl_<mode>, __builtin_altivec_lvx_<mode>,
748 __builtin_altivec_stvx_<mode>, and __builtin_altivec_stvxl_<mode>.
750 2014-02-21 Catherine Moore <clm@codesourcery.com>
752 * doc/invoke.texi (mvirt, mno-virt): Document.
753 * config/mips/mips.opt (mvirt): New option.
754 * config/mips/mips.h (ASM_SPEC): Pass mvirt to the assembler.
756 2014-02-21 Richard Biener <rguenther@suse.de>
758 PR tree-optimization/60276
759 * tree-vectorizer.h (struct _stmt_vec_info): Add min_neg_dist field.
760 (STMT_VINFO_MIN_NEG_DIST): New macro.
761 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Record
762 STMT_VINFO_MIN_NEG_DIST.
763 * tree-vect-stmts.c (vectorizable_load): Verify if assumptions
764 made for negative dependence distances still hold.
766 2014-02-21 Richard Biener <rguenther@suse.de>
769 * tree-ssa-live.c (mark_all_vars_used_1): Do not walk
770 DECL_INITIAL for globals not in the current function context.
772 2014-02-21 Jakub Jelinek <jakub@redhat.com>
774 PR tree-optimization/56490
775 * params.def (PARAM_UNINIT_CONTROL_DEP_ATTEMPTS): New param.
776 * tree-ssa-uninit.c: Include params.h.
777 (compute_control_dep_chain): Add num_calls argument, return false
778 if it exceed PARAM_UNINIT_CONTROL_DEP_ATTEMPTS param, pass
779 num_calls to recursive call.
780 (find_predicates): Change dep_chain into normal array,
781 cur_chain into auto_vec<edge, MAX_CHAIN_LEN + 1>, add num_calls
782 variable and adjust compute_control_dep_chain caller.
783 (find_def_preds): Likewise.
785 2014-02-21 Thomas Schwinge <thomas@codesourcery.com>
787 * gimple-pretty-print.c (dump_gimple_omp_for) [flags & TDF_RAW]
788 <case GF_OMP_FOR_KIND_CILKSIMD>: Add missing break statement.
790 2014-02-21 Nick Clifton <nickc@redhat.com>
792 * config/stormy16/stormy16.md (pushdqi1): Add mode to post_inc.
794 (popqi1): Add mode to pre_dec.
797 2014-02-21 Jakub Jelinek <jakub@redhat.com>
799 * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
800 mode for mask of V8SFmode permutation.
802 2014-02-20 Richard Henderson <rth@redhat.com>
805 * builtins.c (expand_builtin_atomic_compare_exchange): Always make
806 a new pseudo for OLDVAL.
808 2014-02-20 Jakub Jelinek <jakub@redhat.com>
811 * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
812 gen_reg_rtx if d->testing_p.
813 (expand_vec_perm_pshufb2, expand_vec_perm_broadcast_1): Return early
814 if d->testing_p and we will certainly return true.
815 (expand_vec_perm_even_odd_1): Likewise. Don't call gen_reg_rtx
818 2014-02-20 Uros Bizjak <ubizjak@gmail.com>
820 * emit-rtl.c (gen_reg_rtx): Assert that
821 crtl->emit.regno_pointer_align_length is non-zero.
823 2014-02-20 Richard Henderson <rth@redhat.com>
826 * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
827 on failure the store back into EXPECT.
829 2014-02-20 Chung-Lin Tang <cltang@codesourcery.com>
830 Sandra Loosemore <sandra@codesourcery.com>
832 * config/nios2/nios2.md (unspec): Add UNSPEC_PIC_GOTOFF_SYM enum.
833 * config/nios2/nios2.c (nios2_function_profiler): Add
834 -fPIC (flag_pic == 2) support.
835 (nios2_handle_custom_fpu_cfg): Fix warning parameter.
836 (nios2_large_offset_p): New function.
837 (nios2_unspec_reloc_p): Move up position, update to use
838 nios2_large_offset_p.
839 (nios2_unspec_address): Remove function.
840 (nios2_unspec_offset): New function.
841 (nios2_large_got_address): New function.
842 (nios2_got_address): Add large offset support.
843 (nios2_legitimize_tls_address): Update usage of removed and new
845 (nios2_symbol_binds_local_p): New function.
846 (nios2_load_pic_address): Add -fPIC (flag_pic == 2) support.
847 (nios2_legitimize_address): Update to use nios2_large_offset_p.
848 (nios2_emit_move_sequence): Avoid legitimizing (const (unspec ...)).
849 (nios2_print_operand): Merge H/L processing, add hiadj/lo
850 processing for (const (unspec ...)).
851 (nios2_unspec_reloc_name): Add UNSPEC_PIC_GOTOFF_SYM case.
853 2014-02-20 Richard Biener <rguenther@suse.de>
855 * tree-cfg.c (replace_uses_by): Mark altered BBs before
856 doing the substitution.
857 (verify_gimple_assign_single): Also verify bare MEM_REFs on the lhs.
859 2014-02-20 Martin Jambor <mjambor@suse.cz>
862 * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
863 info when checking whether lattices are bottom.
865 2014-02-20 Richard Biener <rguenther@suse.de>
868 * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
871 2014-02-20 Jan Hubicka <hubicka@ucw.cz>
874 * ipa-inline-transform.c (clone_inlined_nodes): Add freq_scale
875 parameter specifying the scaling.
876 (inline_call): Update.
877 (want_inline_recursively): Guard division by zero.
878 (recursive_inlining): Update.
879 * ipa-inline.h (clone_inlined_nodes): Update.
881 2014-02-20 Ilya Tocar <ilya.tocar@intel.com>
884 * config/i386/i386.c (classify_argument): Pass structures of size
885 64 bytes or less in register.
887 2014-02-20 Ilya Tocar <ilya.tocar@intel.com>
888 Kirill Yukhin <kirill.yukhin@intel.com>
890 * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): Swap operands.
891 (_mm_rcp28_round_ss): Ditto.
892 (_mm_rsqrt28_round_sd): Ditto.
893 (_mm_rsqrt28_round_ss): Ditto.
894 * config/i386/avx512erintrin.h (_mm_rcp14_round_sd): Ditto.
895 (_mm_rcp14_round_ss): Ditto.
896 (_mm_rsqrt14_round_sd): Ditto.
897 (_mm_rsqrt14_round_ss): Ditto.
898 * config/i386/sse.md (rsqrt14<mode>): Put nonimmediate operand as
899 the first input operand, get rid of match_dup.
900 (avx512er_exp2<mode><mask_name><round_saeonly_name>): Set type
902 (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
904 (avx512er_vmrcp28<mode><round_saeonly_name>): Put nonimmediate
905 operand as the first input operand, set type attribute.
906 (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
908 (avx512er_vmrsqrt28<mode><round_saeonly_name>): Put nonimmediate
909 operand as the first input operand, set type attribute.
911 2014-02-19 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
913 * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
916 2014-02-19 H.J. Lu <hongjiu.lu@intel.com>
919 * config/i386/i386.c (construct_container): Remove TFmode check
920 for X86_64_INTEGER_CLASS.
922 2014-02-19 Uros Bizjak <ubizjak@gmail.com>
925 * config/i386/i386.c (type_natural_mode): Warn for ABI changes
926 only when -Wpsabi is enabled.
928 2014-02-19 Michael Hudson-Doyle <michael.hudson@linaro.org>
931 * config/aarch64/aarch64.c (aarch64_pass_by_reference): The rules for
932 passing arrays in registers are the same as for structs, so remove the
933 special case for them.
935 2014-02-19 Eric Botcazou <ebotcazou@adacore.com>
937 * expr.c (expand_expr_real_1) <case VIEW_CONVERT_EXPR>: For a bit-field
938 destination type, extract only the valid bits if the source type is not
939 integral and has a different mode.
941 2014-02-19 Richard Biener <rguenther@suse.de>
944 * tree-inline.c (estimate_num_insns): Avoid calling cgraph_get_node
947 2014-02-19 Richard Biener <rguenther@suse.de>
950 * ipa-prop.c: Include stringpool.h and tree-ssanames.h.
951 (ipa_modify_call_arguments): Emit an argument load explicitely and
952 preserve virtual SSA form there and for the replacement call.
953 Do not update SSA form nor free dominance info.
955 2014-02-18 Jan Hubicka <hubicka@ucw.cz>
957 * ipa.c (function_and_variable_visibility): Also clear WEAK
958 flag when disolving COMDAT_GROUP.
960 2014-02-18 Jan Hubicka <hubicka@ucw.cz>
962 * ipa-prop.h (ipa_ancestor_jf_data): Update ocmment.
963 * ipa-prop.c (ipa_set_jf_known_type): Return early when
965 (ipa_set_ancestor_jf): Set type to NULL hwen it is not preserved;
966 do more sanity checks.
967 (detect_type_change): Return true when giving up early.
968 (compute_complex_assign_jump_func): Fix type parameter of
970 (compute_complex_ancestor_jump_func): Likewise.
971 (update_jump_functions_after_inlining): Fix updating of
973 * ipa-cp.c (ipa_get_jf_ancestor_result): Be ready for type to be NULL.
975 2014-02-18 Jan Hubicka <hubicka@ucw.cz>
977 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
978 inline clones when edge disappears.
980 2014-02-18 Michael Meissner <meissner@linux.vnet.ibm.com>
983 * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
984 Split 64-bit moves into 2 patterns. Do not allow the use of
985 direct move for TDmode in little endian, since the decimal value
986 has little endian bytes within a word, but the 64-bit pieces are
987 ordered in a big endian fashion, and normal subreg's of TDmode are
989 (mov<mode>_64bit_dm): Likewise.
990 (movtd_64bit_nodm): Likewise.
992 2014-02-18 Eric Botcazou <ebotcazou@adacore.com>
994 PR tree-optimization/60174
995 * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
996 statement of an SSA_NAME that occurs in an abnormal PHI node.
998 2014-02-18 Jakub Jelinek <jakub@redhat.com>
1001 * final.c (SEEN_BB): Remove.
1002 (SEEN_NOTE, SEEN_EMITTED): Renumber.
1003 (final_scan_insn): Don't force_source_line on second
1004 NOTE_INSN_BASIC_BLOCK.
1006 2014-02-18 Uros Bizjak <ubizjak@gmail.com>
1009 * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
1010 * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
1011 (type_natural_mode): Warn ABI change when %zmm register is not
1012 available for AVX512F vector value passing.
1014 2014-02-18 Kai Tietz <ktietz@redhat.com>
1017 * config/i386/i386.c (ix86_expand_prologue): Use value in
1018 rax register as displacement when restoring %r10 or %rax.
1019 Fix wrong offset when restoring both registers.
1021 2014-02-18 Eric Botcazou <ebotcazou@adacore.com>
1023 * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
1024 assertion with conditional return.
1026 2014-02-18 Jakub Jelinek <jakub@redhat.com>
1027 Uros Bizjak <ubizjak@gmail.com>
1030 * config/i386/driver-i386.c (host_detect_local_cpu): If
1031 YMM state is not saved by the OS, also clear has_f16c. Move
1032 CPUID 0x80000001 handling before YMM state saving checking.
1034 2014-02-18 Andrey Belevantsev <abel@ispras.ru>
1036 PR rtl-optimization/58960
1037 * haifa-sched.c (alloc_global_sched_pressure_data): New,
1038 factored out from ...
1039 (sched_init): ... here.
1040 (free_global_sched_pressure_data): New, factored out from ...
1041 (sched_finish): ... here.
1042 * sched-int.h (free_global_sched_pressure_data): Declare.
1043 * sched-rgn.c (nr_regions_initial): New static global.
1044 (haifa_find_rgns): Initialize it.
1045 (schedule_region): Disable sched-pressure for the newly
1048 2014-02-17 Richard Biener <rguenther@suse.de>
1050 * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
1051 release SSA defs of pattern stmts.
1053 2014-02-17 Richard Biener <rguenther@suse.de>
1055 * tree-inline.c (expand_call_inline): Release the virtual
1056 operand defined by the call we are about to inline.
1058 2014-02-17 Richard Biener <rguenther@suse.de>
1060 * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
1062 2014-02-17 Kirill Yukhin <kirill.yukhin@intel.com>
1063 Ilya Tocar <ilya.tocar@intel.com>
1065 * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
1066 arguments order in builtin.
1067 (_mm512_permutexvar_epi64): Ditto.
1068 (_mm512_mask_permutexvar_epi64): Ditto
1069 (_mm512_maskz_permutexvar_epi32): Ditto
1070 (_mm512_permutexvar_epi32): Ditto
1071 (_mm512_mask_permutexvar_epi32): Ditto
1073 2014-02-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1075 * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
1076 (p8_vmrgow): Likewise.
1078 2014-02-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1080 * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
1083 2014-02-15 Michael Meissner <meissner@linux.vnet.ibm.com>
1086 * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
1087 (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
1088 into 64-bit and 32-bit moves. On 64-bit moves, add support for
1089 using direct move instructions on ISA 2.07. Also adjust
1090 instruction length for 64-bit.
1091 (mov<mode>_64bit, TFmode/TDmode): Likewise.
1092 (mov<mode>_32bit, TFmode/TDmode): Likewise.
1094 2014-02-15 Alan Modra <amodra@gmail.com>
1098 * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
1099 find_replacement on parts of insn rtl that might be reloaded.
1101 2014-02-15 Richard Biener <rguenther@suse.de>
1103 PR tree-optimization/60183
1104 * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
1105 (tree_ssa_phiprop): Calculate and free post-dominators.
1107 2014-02-14 Jeff Law <law@redhat.com>
1109 PR rtl-optimization/60131
1110 * ree.c (get_extended_src_reg): New function.
1111 (combine_reaching_defs): Use it rather than assuming location of REG.
1112 (find_and_remove_re): Verify first operand of extension is
1113 a REG before adding the insns to the copy list.
1115 2014-02-14 Roland McGrath <mcgrathr@google.com>
1117 * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
1118 * configure: Regenerated.
1119 * config.in: Regenerated.
1120 * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
1121 instead of ASM_SHORT.
1123 2014-02-14 Vladimir Makarov <vmakarov@redhat.com>
1124 Richard Earnshaw <rearnsha@arm.com>
1126 PR rtl-optimization/59535
1127 * lra-constraints.c (process_alt_operands): Encourage alternative
1128 when unassigned pseudo class is superset of the alternative class.
1129 (inherit_reload_reg): Don't inherit when optimizing for code size.
1130 * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
1131 returning CORE_REGS for anything but Thumb1 and BASE_REGS for
1132 modes not less than 4 for Thumb1.
1134 2014-02-14 Kyle McMartin <kyle@redhat.com>
1137 * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
1139 2014-02-14 Richard Biener <rguenther@suse.de>
1141 * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
1142 (get_frame_arg): Drop the assert with langhook types_compatible_p.
1143 Do not strip INDIRECT_REFs.
1145 2014-02-14 Richard Biener <rguenther@suse.de>
1148 * lto-streamer-out.c (DFS_write_tree_body): Do not follow
1149 DECL_FUNCTION_SPECIFIC_TARGET.
1150 (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
1151 * tree-streamer-out.c (pack_ts_target_option): Remove.
1152 (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
1153 (write_ts_function_decl_tree_pointers): Do not stream
1154 DECL_FUNCTION_SPECIFIC_TARGET.
1155 * tree-streamer-in.c (unpack_ts_target_option): Remove.
1156 (unpack_value_fields): Do not stream TS_TARGET_OPTION.
1157 (lto_input_ts_function_decl_tree_pointers): Do not stream
1158 DECL_FUNCTION_SPECIFIC_TARGET.
1160 2014-02-14 Jakub Jelinek <jakub@redhat.com>
1162 * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
1163 (get_initial_def_for_induction, vectorizable_induction): Ignore
1164 debug stmts when looking for exit_phi.
1165 (vectorizable_live_operation): Fix up condition.
1167 2014-02-14 Chung-Ju Wu <jasonwucj@gmail.com>
1169 * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
1170 nreverse() because it changes the content of original tree list.
1172 2014-02-14 Chung-Ju Wu <jasonwucj@gmail.com>
1174 * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
1175 * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
1177 2014-02-14 Chung-Ju Wu <jasonwucj@gmail.com>
1179 * config/nds32/nds32.c (nds32_naked_function_p): Follow the
1180 GNU coding standards.
1182 2014-02-13 Jakub Jelinek <jakub@redhat.com>
1185 * dwarf2out.c (gen_subprogram_die): Don't call
1186 add_calling_convention_attribute if subr_die is old_die.
1188 2014-02-13 Sharad Singhai <singhai@google.com>
1190 * doc/optinfo.texi: Fix order of nodes.
1192 2014-02-13 Uros Bizjak <ubizjak@gmail.com>
1194 * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
1195 operands[2], not operands[3].
1197 2014-02-13 Richard Biener <rguenther@suse.de>
1200 * doc/install.texi (ISL): Update recommended version to 0.12.2,
1201 mention the possibility of an in-tree build.
1202 (CLooG): Update recommended version to 0.18.1, mention the
1203 possibility of an in-tree build and clarify that the ISL
1204 bundled with CLooG does not work.
1206 2014-02-13 Jakub Jelinek <jakub@redhat.com>
1209 * expr.c (compress_float_constant): If x is a hard register,
1210 extend into a pseudo and then move to x.
1212 2014-02-13 Dominik Vogt <vogt@linux.vnet.ibm.com>
1214 * config/s390/s390.c (s390_asm_output_function_label): Fix crash
1215 caused by bad second argument to warning_at() with -mhotpatch and
1216 nested functions (e.g. with gfortran).
1218 2014-02-13 Richard Sandiford <rdsandiford@googlemail.com>
1220 * opts.c (option_name): Remove "enabled by default" rider.
1222 2014-02-12 John David Anglin <danglin@gcc.gnu.org>
1224 * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
1226 2014-02-12 H.J. Lu <hongjiu.lu@intel.com>
1227 Uros Bizjak <ubizjak@gmail.com>
1230 * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
1231 * configure: Regenerated.
1233 2014-02-12 Richard Biener <rguenther@suse.de>
1235 * vec.c (vec_prefix::calculate_allocation): Move as
1236 inline variant to vec.h.
1237 (vec_prefix::calculate_allocation_1): New out-of-line version.
1238 * vec.h (vec_prefix::calculate_allocation_1): Declare.
1239 (vec_prefix::m_has_auto_buf): Rename to ...
1240 (vec_prefix::m_using_auto_storage): ... this.
1241 (vec_prefix::calculate_allocation): Inline the easy cases
1242 and dispatch to calculate_allocation_1 which doesn't need the
1244 (va_heap::reserve): Use gcc_checking_assert.
1245 (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
1246 m_using_auto_storage.
1247 (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
1249 (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
1250 (vec<T, va_heap, vl_ptr>::release): Avoid casting.
1251 (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
1253 2014-02-12 Richard Biener <rguenther@suse.de>
1255 * gcse.c (compute_transp): break from loop over canon_modify_mem_list
1256 when we found a dependence.
1258 2014-02-12 Thomas Schwinge <thomas@codesourcery.com>
1260 * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
1262 (maybe_fold_stmt): ... into this new function.
1263 * omp-low.c (lower_omp): Update comment.
1265 * omp-low.c (lower_omp_target): Add clobber for sizes array, after
1268 * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
1271 2014-02-12 James Greenhalgh <james.greenhalgh@arm.com>
1273 * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
1274 identifiers in comments.
1275 (cortexa53_extra_costs): Likewise.
1276 * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
1277 (cortexa7_extra_costs): Likewise.
1278 (cortexa12_extra_costs): Likewise.
1279 (cortexa15_extra_costs): Likewise.
1280 (v7m_extra_costs): Likewise.
1282 2014-02-12 Richard Biener <rguenther@suse.de>
1285 * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
1286 of posix_memalign being successful.
1287 (lower_stmt): Restrict lowering of posix_memalign to when
1288 -ftree-bit-ccp is enabled.
1290 2014-02-12 Senthil Kumar Selvaraj <senthil_kumar.selvaraj@atmel.com>
1292 * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
1294 * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
1296 2014-02-12 Eric Botcazou <ebotcazou@adacore.com>
1298 PR rtl-optimization/60116
1299 * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
1300 other_insn once the combination has been validated.
1302 2014-02-11 Jan Hubicka <hubicka@ucw.cz>
1305 * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
1307 * ipa-devirt.c: Include demangle.h
1308 (odr_violation_reported): New static variable.
1309 (add_type_duplicate): Update odr_violations.
1310 (maybe_record_node): Add completep parameter; update it.
1311 (record_target_from_binfo): Add COMPLETEP parameter;
1312 update it as needed.
1313 (possible_polymorphic_call_targets_1): Likewise.
1314 (struct polymorphic_call_target_d): Add nonconstruction_targets;
1315 rename FINAL to COMPLETE.
1316 (record_targets_from_bases): Sanity check we found the binfo;
1317 fix COMPLETEP updating.
1318 (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
1319 parameter, fix computing of COMPLETEP.
1320 (dump_possible_polymorphic_call_targets): Imrove readability of dump;
1321 at LTO time do demangling.
1322 (ipa_devirt): Use nonconstruction_targets; Improve dumps.
1323 * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
1325 (gimple_get_virt_method_for_binfo): Likewise.
1326 * gimple-fold.h (gimple_get_virt_method_for_binfo,
1327 gimple_get_virt_method_for_vtable): Update prototypes.
1329 2014-02-11 Vladimir Makarov <vmakarov@redhat.com>
1332 * genautomata.c (add_presence_absence): Fix typo with
1333 {final_}presence_list.
1335 2014-02-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1338 * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
1339 for VSX/Altivec vectors that land in GPR registers.
1341 2014-02-11 Richard Henderson <rth@redhat.com>
1342 Jakub Jelinek <jakub@redhat.com>
1345 * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
1346 around drhs if type conversion to lacc->type is not useless.
1348 2014-02-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1350 * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
1352 (cortex-a57.cortex-a53): Likewise.
1353 * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
1355 2014-02-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1357 * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
1360 2014-02-11 Renlin Li <Renlin.Li@arm.com>
1362 * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
1363 add_options_for_arm_vfp3.
1365 2014-02-11 Jeff Law <law@redhat.com>
1368 * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
1369 object with an undesirable mode.
1371 2014-02-11 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1374 * config/i386/sol2-9.h: New file.
1375 * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
1376 *-*-solaris2.9*): Use it.
1378 2014-02-10 Nagaraju Mekala <nagaraju.mekala@xilinx.com>
1380 * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
1381 * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
1383 2014-02-10 Nagaraju Mekala <nagaraju.mekala@xilinx.com>
1385 * config/microblaze/microblaze.c: Extend mcpu version format
1387 2014-02-10 David Holsgrove <david.holsgrove@xilinx.com>
1389 * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
1391 2014-02-10 Richard Henderson <rth@redhat.com>
1394 * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
1395 * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
1396 ms-abi vs -mno-accumulate-outgoing-args.
1397 (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
1398 * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
1401 2014-02-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
1404 * cfgexpand.c (expand_asm_operands): Attach source location to
1405 ASM_INPUT rtx objects.
1406 * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
1408 2014-02-10 Nick Clifton <nickc@redhat.com>
1410 * config/mn10300/mn10300.c (popcount): New function.
1411 (mn10300_expand_prologue): Include saved registers in stack usage
1414 2014-02-10 Jeff Law <law@redhat.com>
1417 * reload1.c (emit_input_reload_insns): Do not create invalid RTL
1418 when changing the SET_DEST of a prior insn to avoid an input reload.
1420 2014-02-10 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1422 * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
1423 big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
1424 -mcall-openbsd, or -mcall-linux.
1425 (CC1_ENDIAN_BIG_SPEC): Remove.
1426 (CC1_ENDIAN_LITTLE_SPEC): Remove.
1427 (CC1_ENDIAN_DEFAULT_SPEC): Remove.
1428 (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
1429 (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
1430 and %cc1_endian_default.
1431 * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
1433 2014-02-10 Richard Biener <rguenther@suse.de>
1435 PR tree-optimization/60115
1436 * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
1437 MEM_REF handling. Properly verify that the accesses are not
1438 out of the objects bound.
1440 2014-02-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1442 * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
1445 2014-02-10 Eric Botcazou <ebotcazou@adacore.com>
1447 * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
1448 proper constants and fix formatting.
1449 (possible_polymorphic_call_targets): Fix formatting.
1451 2014-02-10 Kirill Yukhin <kirill.yukhin@intel.com>
1452 Ilya Tocar <ilya.tocar@intel.com>
1454 * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
1455 (_mm512_loadu_epi32): Renamed into...
1456 (_mm512_loadu_si512): This.
1457 (_mm512_storeu_epi32): Renamed into...
1458 (_mm512_storeu_si512): This.
1459 (_mm512_maskz_ceil_ps): Removed.
1460 (_mm512_maskz_ceil_pd): Ditto.
1461 (_mm512_maskz_floor_ps): Ditto.
1462 (_mm512_maskz_floor_pd): Ditto.
1463 (_mm512_floor_round_ps): Ditto.
1464 (_mm512_floor_round_pd): Ditto.
1465 (_mm512_ceil_round_ps): Ditto.
1466 (_mm512_ceil_round_pd): Ditto.
1467 (_mm512_mask_floor_round_ps): Ditto.
1468 (_mm512_mask_floor_round_pd): Ditto.
1469 (_mm512_mask_ceil_round_ps): Ditto.
1470 (_mm512_mask_ceil_round_pd): Ditto.
1471 (_mm512_maskz_floor_round_ps): Ditto.
1472 (_mm512_maskz_floor_round_pd): Ditto.
1473 (_mm512_maskz_ceil_round_ps): Ditto.
1474 (_mm512_maskz_ceil_round_pd): Ditto.
1475 (_mm512_expand_pd): Ditto.
1476 (_mm512_expand_ps): Ditto.
1477 * config/i386/i386.c (ix86_builtins): Remove
1478 IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
1479 (bdesc_args): Ditto.
1480 * config/i386/predicates.md (const1256_operand): New.
1481 (const_1_to_2_operand): Ditto.
1482 * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
1483 (*avx512pf_gatherpf<mode>sf_mask): Ditto.
1484 (*avx512pf_gatherpf<mode>sf): Ditto.
1485 (avx512pf_gatherpf<mode>df): Ditto.
1486 (*avx512pf_gatherpf<mode>df_mask): Ditto.
1487 (*avx512pf_gatherpf<mode>df): Ditto.
1488 (avx512pf_scatterpf<mode>sf): Ditto.
1489 (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1490 (*avx512pf_scatterpf<mode>sf): Ditto.
1491 (avx512pf_scatterpf<mode>df): Ditto.
1492 (*avx512pf_scatterpf<mode>df_mask): Ditto.
1493 (*avx512pf_scatterpf<mode>df): Ditto.
1494 (avx512f_expand<mode>): Removed.
1495 (<shift_insn><mode>3<mask_name>): Change predicate type.
1497 2014-02-08 Jakub Jelinek <jakub@redhat.com>
1499 * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
1500 not at the end of datarefs vector use ordered_remove to avoid
1501 reordering datarefs vector.
1504 * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
1505 mark local addressable non-static vars as GOVD_PRIVATE
1506 instead of GOVD_LOCAL.
1507 * omp-low.c (lower_omp_for): Move gimple_bind_vars
1508 and BLOCK_VARS of gimple_bind_block to new_stmt rather
1512 * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
1513 if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
1514 assume_aligned or alloc_align attributes.
1515 (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
1516 arguments. Handle also assume_aligned and alloc_align attributes.
1517 (evaluate_stmt): Adjust bit_value_assume_aligned caller. Handle
1518 calls to functions with assume_aligned or alloc_align attributes.
1519 * doc/extend.texi: Document assume_aligned and alloc_align attributes.
1521 2014-02-08 Terry Guo <terry.guo@arm.com>
1523 * doc/invoke.texi: Document ARM -march=armv7e-m.
1525 2014-02-08 Jakub Jelinek <jakub@redhat.com>
1527 * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
1528 flag on __cilkrts_rethrow builtin.
1531 * ipa-cp.c (determine_versionability): Fail at -O0
1532 or __attribute__((optimize (0))) or -fno-ipa-cp functions.
1533 * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
1536 2014-02-04 Jakub Jelinek <jakub@redhat.com>
1539 * tree-inline.c (copy_forbidden): Fail for
1540 __attribute__((optimize (0))) functions.
1542 2014-02-07 Jan Hubicka <hubicka@ucw.cz>
1544 * varpool.c: Include pointer-set.h.
1545 (varpool_remove_unreferenced_decls): Variables in other partitions
1546 will not be output; be however careful to not lose information
1549 2014-02-07 Jan Hubicka <hubicka@ucw.cz>
1551 * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
1552 lookup in the vtable constructor.
1554 2014-02-07 Jeff Law <law@redhat.com>
1557 * config/m68k/m68k.md (ashldi_extsi): Turn into a
1558 define_insn_and_split.
1560 * ipa-inline.c (inline_small_functions): Fix typos.
1562 2014-02-07 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
1564 * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
1565 (s390_can_use_return_insn): Declare.
1566 * config/s390/s390.h (EPILOGUE_USES): Define.
1567 * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
1569 (s390_chunkify_start): Handle return JUMP_LABELs.
1570 (s390_early_mach): Emit a main_pool instruction on the entry edge.
1571 (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
1572 (s390_can_use_return_insn): New functions.
1573 (s390_fix_long_loop_prediction): Handle conditional returns.
1574 (TARGET_SET_UP_BY_PROLOGUE): Define.
1575 * config/s390/s390.md (ANY_RETURN): New code iterator.
1576 (*creturn, *csimple_return, return, simple_return): New patterns.
1578 2014-02-07 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
1580 * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
1581 notes to each restore. Also add REG_CFA_DEF_CFA when restoring %r15.
1582 (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P. Update the
1583 REG_CFA_RESTORE list when deciding not to restore a register.
1585 2014-02-07 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
1587 * config/s390/s390.c: Include tree-pass.h and context.h.
1588 (s390_early_mach): New function, split out from...
1589 (s390_emit_prologue): ...here.
1590 (pass_data_s390_early_mach): New pass structure.
1591 (pass_s390_early_mach): New class.
1592 (s390_option_override): Create and register early_mach pass.
1593 Move to end of file.
1595 2014-02-07 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
1597 * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
1598 to match for the exit block.
1600 2014-02-07 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
1602 * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
1603 ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
1604 Reject misaligned operands.
1606 2014-02-07 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
1608 * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
1610 2014-02-07 Richard Biener <rguenther@suse.de>
1613 * gimple-low.c (lower_builtin_posix_memalign): New function.
1614 (lower_stmt): Call it to lower posix_memalign in a way
1615 to make alignment info accessible.
1617 2014-02-07 Jakub Jelinek <jakub@redhat.com>
1620 * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
1621 __builtin_setjmp_receiver.
1623 2014-02-07 Richard Biener <rguenther@suse.de>
1626 * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
1627 * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
1628 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
1629 Handle BUILT_IN_POSIX_MEMALIGN.
1630 (find_func_clobbers): Likewise.
1631 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
1632 (call_may_clobber_ref_p_1): Likewise.
1634 2014-02-06 Jan Hubicka <hubicka@ucw.cz>
1637 * ipa-devirt.c (record_target_from_binfo): Remove overactive
1640 2014-02-06 Jan Hubicka <hubicka@ucw.cz>
1643 * lto-cgraph.c (lto_output_node): Use
1644 symtab_get_symbol_partitioning_class.
1645 (lto_output_varpool_node): likewise.
1646 (symtab_get_symbol_partitioning_class): Move here from
1648 * cgraph.h (symbol_partitioning_class): Likewise.
1649 (symtab_get_symbol_partitioning_class): Declare.
1651 2014-02-06 Jan Hubicka <hubicka@ucw.cz>
1653 * ggc.h (ggc_internal_cleared_alloc): New macro.
1654 * vec.h (vec_safe_copy): Handle memory stats.
1655 * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
1656 * target-globals.c (save_target_globals): Likewise.
1658 2014-02-06 Jan Hubicka <hubicka@ucw.cz>
1661 * expr.c (emit_move_resolve_push): Export; be bit more selective
1662 on when to clear alias set.
1663 * expr.h (emit_move_resolve_push): Declare.
1664 * function.h (struct function): Add tail_call_marked.
1665 * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
1666 * config/i386/i386-protos.h (ix86_expand_push): Remove.
1667 * config/i386/i386.md (TImode move expander): De not call
1669 (FP push expanders): Preserve memory attributes.
1670 * config/i386/sse.md (push<mode>1): Remove.
1671 * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
1672 (ix86_expand_push): Remove.
1673 * config/i386/mmx.md (push<mode>1): Remove.
1675 2014-02-06 Jakub Jelinek <jakub@redhat.com>
1677 PR rtl-optimization/60030
1678 * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
1679 lopart with paradoxical subreg before shifting it up by hprec.
1681 2014-02-06 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1683 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
1684 Remove extra newline at end of file.
1685 * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
1686 (arm_issue_rate): Handle cortexa57.
1687 * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
1688 (cortex-a57.cortex-a53): Likewise.
1690 2014-02-06 Jakub Jelinek <jakub@redhat.com>
1693 * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
1694 don't record in REG_FRAME_RELATED_EXPR registers not set in that
1696 (arm_expand_prologue): Adjust all callers.
1697 (arm_unwind_emit_sequence): Allow saved, but not important for unwind
1698 info, registers also at the lowest numbered registers side. Use
1699 gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
1703 * var-tracking.c (adjust_mems): Before adding a SET to
1704 amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
1706 2014-02-06 Alan Modra <amodra@gmail.com>
1709 * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
1710 change SDmode to DDmode when lra_in_progress.
1712 2014-02-06 Jakub Jelinek <jakub@redhat.com>
1715 * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
1716 free_data_ref on the dr first, and before goto again also set dr
1717 to the next dr. For simd_lane_access, free old datarefs[i] before
1718 overwriting it. For get_vectype_for_scalar_type failure, don't
1719 free_data_ref if simd_lane_access.
1721 * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
1724 * tree.h (opts_for_fn): New inline function.
1725 (opt_for_fn): Define.
1726 * config/i386/i386.c (ix86_function_regparm): Use
1727 opt_for_fn (decl, optimize) instead of optimize.
1729 2014-02-06 Marcus Shawcroft <marcus.shawcroft@arm.com>
1731 * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
1732 for SYMBOL_REF in large memory model.
1734 2014-02-06 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1736 * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
1738 (cortex-a57): Likewise.
1739 (cortex-a57.cortex-a53): Likewise.
1741 2014-02-06 Yury Gribov <y.gribov@samsung.com>
1742 Kugan Vivekanandarajah <kuganv@linaro.org>
1744 * config/arm/arm.c (arm_vector_alignment_reachable): Check
1746 * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
1748 2014-02-06 Richard Biener <rguenther@suse.de>
1750 * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
1751 set_loop_copy and initialize_original_copy_tables.
1753 2014-02-06 Alex Velenko <Alex.Velenko@arm.com>
1755 * config/aarch64/aarch64-simd.md
1756 (aarch64_ashr_simddi): Change QI to SI.
1758 2014-02-05 Jan Hubicka <hubicka@ucw.cz>
1759 Jakub Jelinek <jakub@redhat.com>
1762 * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
1765 2014-02-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1767 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
1768 CODE_FOR_altivec_vpku[hw]um to
1769 CODE_FOR_altivec_vpku[hw]um_direct.
1770 * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
1771 UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
1772 (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
1773 UNSPEC_VUNPACK_LO_SIGN_DIRECT.
1775 2014-02-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1777 * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
1778 generation for -maltivec=be.
1779 (altivec_vsumsws): Simplify redundant test.
1781 2014-02-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1783 * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
1784 (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
1785 (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
1786 (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
1787 gen_altivec_vpkuwum.
1788 (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
1790 (altivec_vpks<VI_char>ss): Likewise.
1791 (altivec_vpks<VI_char>us): Likewise.
1792 (altivec_vpku<VI_char>us): Likewise.
1793 (altivec_vpku<VI_char>um): Likewise.
1794 (altivec_vpku<VI_char>um_direct): New (copy of
1795 altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
1797 (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
1798 target is little endian and -maltivec=be is not specified.
1799 (*altivec_vupkhs<VU_char>_direct): New (copy of
1800 altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
1801 (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
1802 target is little endian and -maltivec=be is not specified.
1803 (*altivec_vupkls<VU_char>_direct): New (copy of
1804 altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
1805 (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
1806 little endian and -maltivec=be is not specified.
1807 (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
1808 little endian and -maltivec=be is not specified.
1810 2014-02-05 Richard Henderson <rth@redhat.com>
1813 * combine-stack-adj.c: Revert r206943.
1814 * sched-int.h (struct deps_desc): Add last_args_size.
1815 * sched-deps.c (init_deps): Initialize it.
1816 (sched_analyze_insn): Add OUTPUT dependencies between insns that
1817 contain REG_ARGS_SIZE notes.
1819 2014-02-05 Jan Hubicka <hubicka@ucw.cz>
1821 * lto-cgraph.c (asm_nodes_output): Make global.
1822 * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
1823 * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
1824 (driver_handle_option): Handle OPT_fwpa.
1826 2014-02-05 Jakub Jelinek <jakub@redhat.com>
1829 * ipa-devirt.c (possible_polymorphic_call_targets): Fix
1830 a comment typo and formatting issue. If odr_hash hasn't been
1831 created, return vNULL and set *completep to false.
1834 * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
1835 bb with no successors.
1837 2014-02-05 James Greenhalgh <james.greenhalgh@arm.com>
1840 * doc/invoke.texi (-march): Clarify documentation for ARM.
1844 2014-02-05 Richard Biener <rguenther@suse.de>
1846 * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
1847 when not vectorizing because of too many alias checks.
1848 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
1849 Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
1851 2014-02-05 Nick Clifton <nickc@redhat.com>
1853 * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
1854 accept extended registers in any mode when compiling for the MN10300.
1856 2014-02-05 Yury Gribov <y.gribov@samsung.com>
1858 * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
1859 * ipa-inline.c (report_inline_failed_reason): Handle mismatched
1860 sanitization attributes.
1861 (can_inline_edge_p): Likewise.
1862 (sanitize_attrs_match_for_inline_p): New function.
1864 2014-02-04 Jan Hubicka <hubicka@ucw.cz>
1866 * ipa-prop.c (detect_type_change): Shor circuit testing of
1867 type changes on THIS pointer.
1869 2014-02-04 John David Anglin <danglin@gcc.gnu.org>
1872 * config/pa/pa.c (legitimize_tls_address): Return original address
1873 if not passed a SYMBOL_REF rtx.
1874 (hppa_legitimize_address): Call legitimize_tls_address for all TLS
1876 (pa_emit_move_sequence): Simplify TLS source operands.
1877 (pa_legitimate_constant_p): Reject all TLS constants.
1878 * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
1879 (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
1881 2014-02-04 Jan Hubicka <hubicka@ucw.cz>
1883 * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
1884 groups when we know they are controlled by LTO.
1885 * varasm.c (default_binds_local_p_1): If object is in other partition,
1886 it will be resolved locally.
1888 2014-02-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
1890 * config/host-linux.c (linux_gt_pch_use_address): Don't
1891 use SSIZE_MAX because it is not always defined.
1893 2014-02-04 Vladimir Makarov <vmakarov@redhat.com>
1896 * lra-constraints.c (need_for_split_p): Use more 3 reloads as
1897 threshold for pseudo splitting.
1898 (update_ebb_live_info): Process call argument hard registers and
1899 hard registers from insn definition too.
1900 (max_small_class_regs_num): New constant.
1901 (inherit_in_ebb): Update live hard regs through EBBs. Update
1902 reloads_num only for small register classes. Don't split for
1905 2014-02-04 Markus Trippelsdorf <markus@trippelsdorf.de>
1908 * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
1911 2014-02-04 Jan Hubicka <hubicka@ucw.cz>
1913 * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
1916 2014-02-04 Marek Polacek <polacek@redhat.com>
1918 * gdbinit.in (pel): Define.
1920 2014-02-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
1922 * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
1925 2014-02-04 Richard Biener <rguenther@suse.de>
1928 * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
1929 in function context local.
1930 (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
1931 * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
1932 similar to LTO_imported_decl_ref.
1934 2014-02-04 Jakub Jelinek <jakub@redhat.com>
1936 PR tree-optimization/60002
1937 * cgraphclones.c (build_function_decl_skip_args): Clear
1940 PR tree-optimization/60023
1941 * tree-if-conv.c (predicate_mem_writes): Pass true instead of
1942 false to gsi_replace.
1943 * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
1944 has been in some EH region and vec_stmt could throw, add
1945 vec_stmt into the same EH region.
1946 * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
1947 has no lhs, ignore it.
1948 * internal-fn.c (expand_MASK_LOAD): Likewise.
1951 * tree-inline.c (copy_forbidden): Fail for
1952 __attribute__((optimize (0))) functions.
1955 * omp-low.c (simd_clone_struct_copy): If from->inbranch
1956 is set, copy one less argument.
1957 (expand_simd_clones): Don't subtract clone_info->inbranch
1958 from simd_clone_struct_alloc argument.
1960 PR rtl-optimization/57915
1961 * recog.c (simplify_while_replacing): If all unary/binary/relational
1962 operation arguments are constant, attempt to simplify those.
1965 * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
1966 if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
1968 2014-02-04 Richard Biener <rguenther@suse.de>
1970 PR tree-optimization/60012
1971 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
1972 TBAA disambiguation to all DDRs.
1974 2014-02-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1977 * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
1978 (LINK_SPEC): Use it for -shared, -shared-libgcc.
1980 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
1983 * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
1985 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
1987 * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
1988 * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
1990 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
1993 * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
1994 to figure out targets of polymorphic calls with known decl.
1995 * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
1996 * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
1997 * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
1998 (get_polymorphic_call_info): ... here.
1999 (get_polymorphic_call_info_from_invariant): New function.
2001 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
2003 * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
2004 lookup via vtable pointer; check for type consistency
2005 and turn inconsitent facts into UNREACHABLE.
2006 * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2007 * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
2008 type inconsistent querries; return UNREACHABLE instead.
2010 2014-02-03 Richard Henderson <rth@twiddle.net>
2013 * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
2014 already processed this node.
2015 (normalize_one_pred_1): Pass along mark_set.
2016 (normalize_one_pred): Create and destroy a pointer_set_t.
2017 (normalize_one_pred_chain): Likewise.
2019 2014-02-03 Laurent Aflonsi <laurent.alfonsi@st.com>
2021 PR gcov-profile/58602
2022 * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
2024 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
2027 * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
2028 -fno-devirtualize; try to devirtualize by the knowledge of
2029 virtual table pointer given by aggregate propagation.
2030 * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2031 (ipa_print_node_jump_functions): Dump also offset that
2032 is relevant for polymorphic calls.
2033 (determine_known_aggregate_parts): Add arg_type parameter; use it
2034 instead of determining the type from pointer type.
2035 (ipa_compute_jump_functions_for_edge): Update call of
2036 determine_known_aggregate_parts.
2037 * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
2038 (gimple_get_virt_method_for_binfo): ... here; simplify using
2039 vtable_pointer_value_to_vtable.
2040 * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
2041 * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
2042 to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
2043 (vtable_pointer_value_to_vtable): Break out from ...; handle also
2045 (vtable_pointer_value_to_binfo): ... here.
2046 * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
2048 2014-02-03 Teresa Johnson <tejohnson@google.com>
2050 * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
2051 redef of outer loop index variable.
2053 2014-02-03 Marc Glisse <marc.glisse@inria.fr>
2057 * doc/extend.texi (Function Attributes): Typo.
2059 2014-02-03 Cong Hou <congh@google.com>
2061 PR tree-optimization/60000
2062 * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
2063 if the vectorized statement is a store. A store statement can only
2064 appear at the end of pattern statements.
2066 2014-02-03 H.J. Lu <hongjiu.lu@intel.com>
2068 * config/i386/i386.c (flag_opts): Add -mlong-double-128.
2069 (ix86_option_override_internal): Default long double to 64-bit for
2070 32-bit Bionic and to 128-bit for 64-bit Bionic.
2072 * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
2073 TARGET_LONG_DOUBLE_128 is true.
2074 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2076 * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
2077 (mlong-double-64): Negate -mlong-double-128.
2078 (mlong-double-128): New option.
2080 * config/i386/i386-c.c (ix86_target_macros): Define
2081 __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
2083 * doc/invoke.texi: Document -mlong-double-128.
2085 2014-02-03 H.J. Lu <hongjiu.lu@intel.com>
2087 PR rtl-optimization/60024
2088 * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
2090 2014-02-03 Markus Trippelsdorf <markus@trippelsdorf.de>
2092 * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
2094 2014-02-03 Andrey Belevantsev <abel@ispras.ru>
2096 PR rtl-optimization/57662
2097 * sel-sched.c (code_motion_path_driver): Do not mark already not
2098 existing blocks in the visiting bitmap.
2100 2014-02-03 Andrey Belevantsev <abel@ispras.ru>
2102 * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
2103 on the insn being emitted.
2105 2014-02-03 James Greenhalgh <james.greenhalgh@arm.com>
2106 Will Deacon <will.deacon@arm.com>
2108 * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
2110 2014-02-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2112 * config/arm/arm-tables.opt: Regenerate.
2114 2014-02-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2116 * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
2117 for vector types other than V16QImode.
2118 * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
2119 define_expand, and call altivec_expand_vec_perm_le when producing
2120 code with little endian element order.
2121 (*altivec_vperm_<mode>_internal): New insn having previous
2122 behavior of altivec_vperm_<mode>.
2123 (altivec_vperm_<mode>_uns): Change to a define_expand, and call
2124 altivec_expand_vec_perm_le when producing code with little endian
2126 (*altivec_vperm_<mode>_uns_internal): New insn having previous
2127 behavior of altivec_vperm_<mode>_uns.
2129 2014-02-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2131 * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
2132 (altivec_vsumsws): Add handling for -maltivec=be with a little
2134 (altivec_vsumsws_direct): New.
2135 (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
2136 gen_altivec_vsumsws.
2138 2014-02-02 Jan Hubicka <hubicka@ucw.cz>
2140 * ipa-devirt.c (subbinfo_with_vtable_at_offset,
2141 vtable_pointer_value_to_binfo): New functions.
2142 * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
2143 * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
2145 2014-02-02 Sandra Loosemore <sandra@codesourcery.com>
2147 * config/nios2/nios2.md (load_got_register): Initialize GOT
2148 pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
2149 * config/nios2/nios2.c (nios2_function_profiler): Likewise.
2151 2014-02-02 Jan Hubicka <hubicka@ucw.cz>
2153 * ipa-prop.c (update_jump_functions_after_inlining): When type is not
2154 preserverd by passthrough, do not propagate the type.
2156 2014-02-02 Richard Sandiford <rdsandiford@googlemail.com>
2158 * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
2159 (mips_atomic_assign_expand_fenv): New function.
2160 (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
2162 2014-02-02 Richard Sandiford <rdsandiford@googlemail.com>
2164 * doc/extend.texi (__builtin_mips_get_fcsr): Document.
2165 (__builtin_mips_set_fcsr): Likewise.
2166 * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
2167 MIPS_USI_FTYPE_VOID.
2168 * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
2169 (mips16_expand_set_fcsr): Likewise.
2170 * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
2171 (mips16_set_fcsr_stub): Likewise.
2172 (mips16_get_fcsr_one_only_stub): New class.
2173 (mips16_set_fcsr_one_only_stub): Likewise.
2174 (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
2175 (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
2176 (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
2177 (hard_float): New availability predicate.
2178 (mips_builtins): Add get_fcsr and set_fcsr.
2179 (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
2180 * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
2181 (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
2182 (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
2183 (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
2186 2014-02-02 Richard Sandiford <rdsandiford@googlemail.com>
2188 * config/mips/mips.c (mips_one_only_stub): New class.
2189 (mips_need_mips16_rdhwr_p): Replace with...
2190 (mips16_rdhwr_stub): ...this new variable.
2191 (mips16_stub_call_address): New function.
2192 (mips16_rdhwr_one_only_stub): New class.
2193 (mips_expand_thread_pointer): Use mips16_stub_call_address.
2194 (mips_output_mips16_rdhwr): Delete.
2195 (mips_finish_stub): New function.
2196 (mips_code_end): Use it to handle rdhwr stubs.
2198 2014-02-02 Uros Bizjak <ubizjak@gmail.com>
2201 * config/i386/i386.c (classify_argument): Fix handling of bit_offset
2202 when calculating size of integer atomic types.
2204 2014-02-02 H.J. Lu <hongjiu.lu@intel.com>
2206 * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
2208 2014-02-01 Jakub Jelinek <jakub@redhat.com>
2210 PR tree-optimization/60003
2211 * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
2212 * profile.c (branch_prob): Use gimple_call_builtin_p
2213 to check for BUILT_IN_SETJMP_RECEIVER.
2214 * tree-inline.c (copy_bb): Call notice_special_calls.
2216 2014-01-31 Vladimir Makarov <vmakarov@redhat.com>
2219 * lra-constraints.c (process_alt_operands): Update reload_sum only
2222 2014-01-31 Richard Henderson <rth@redhat.com>
2225 * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
2226 until after else_eh is processed.
2228 2014-01-31 Ilya Tocar <ilya.tocar@intel.com>
2230 * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
2231 (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
2232 (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
2233 in smmintrin.h, remove them.
2234 (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
2235 * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
2236 * config/i386/i386.md (ROUND_SAE): Fix value.
2237 * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
2238 (const48_operand): New.
2239 * config/i386/subst.md (round), (round_expand): Use
2240 const_4_or_8_to_11_operand.
2241 (round_saeonly), (round_saeonly_expand): Use const48_operand.
2243 2014-01-31 Ilya Tocar <ilya.tocar@intel.com>
2245 * config/i386/constraints.md (Yk): Swap meaning with k.
2246 * config/i386/i386.md (movhi_internal): Change Yk to k.
2247 (movqi_internal): Ditto.
2248 (*k<logic><mode>): Ditto.
2251 (kandn<mode>): Ditto.
2252 (*<code>hi_1): Ditto.
2253 (*<code>qi_1): Ditto.
2254 (kxnor<mode>): Ditto.
2255 (kortestzhi): Ditto.
2256 (kortestchi): Ditto.
2258 (*one_cmplhi2_1): Ditto.
2259 (*one_cmplqi2_1): Ditto.
2260 * config/i386/sse.md (): Change k to Yk.
2261 (avx512f_load<mode>_mask): Ditto.
2262 (avx512f_blendm<mode>): Ditto.
2263 (avx512f_store<mode>_mask): Ditto.
2264 (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
2265 (avx512f_storedqu<mode>_mask): Ditto.
2266 (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
2268 (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
2269 (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
2270 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
2271 (avx512f_maskcmp<mode>3): Ditto.
2272 (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
2273 (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
2274 (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
2275 (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
2276 (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
2277 (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
2278 (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
2279 (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
2280 (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
2281 (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
2282 (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
2283 (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
2284 (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
2285 (vec_extract_lo_<mode>_maskm): Ditto.
2286 (vec_extract_hi_<mode>_maskm): Ditto.
2287 (avx512f_vternlog<mode>_mask): Ditto.
2288 (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
2289 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
2290 (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
2291 (avx512f_<code>v8div16qi2_mask): Ditto.
2292 (avx512f_<code>v8div16qi2_mask_store): Ditto.
2293 (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
2294 (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
2295 (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
2296 (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
2297 (*avx512pf_gatherpf<mode>sf_mask): Ditto.
2298 (*avx512pf_gatherpf<mode>df_mask): Ditto.
2299 (*avx512pf_scatterpf<mode>sf_mask): Ditto.
2300 (*avx512pf_scatterpf<mode>df_mask): Ditto.
2301 (avx512cd_maskb_vec_dupv8di): Ditto.
2302 (avx512cd_maskw_vec_dupv16si): Ditto.
2303 (avx512f_vpermi2var<mode>3_maskz): Ditto.
2304 (avx512f_vpermi2var<mode>3_mask): Ditto.
2305 (avx512f_vpermi2var<mode>3_mask): Ditto.
2306 (avx512f_vpermt2var<mode>3_maskz): Ditto.
2307 (*avx512f_gathersi<mode>): Ditto.
2308 (*avx512f_gathersi<mode>_2): Ditto.
2309 (*avx512f_gatherdi<mode>): Ditto.
2310 (*avx512f_gatherdi<mode>_2): Ditto.
2311 (*avx512f_scattersi<mode>): Ditto.
2312 (*avx512f_scatterdi<mode>): Ditto.
2313 (avx512f_compress<mode>_mask): Ditto.
2314 (avx512f_compressstore<mode>_mask): Ditto.
2315 (avx512f_expand<mode>_mask): Ditto.
2316 * config/i386/subst.md (mask): Change k to Yk.
2317 (mask_scalar_merge): Ditto.
2320 2014-01-31 Marc Glisse <marc.glisse@inria.fr>
2322 * doc/extend.texi (Vector Extensions): Document ?: in C++.
2324 2014-01-31 Richard Biener <rguenther@suse.de>
2327 * builtins.c (fold_builtin_memory_op): Make sure to not
2328 use a floating-point mode or a boolean or enumeral type for
2331 2014-01-30 DJ Delorie <dj@redhat.com>
2333 * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
2334 * config/msp430/msp430.md (msp430_refsym_need_exit): New.
2335 * config/msp430/msp430.c (msp430_expand_epilogue): Call it
2336 whenever main() has an epilogue.
2338 2014-01-30 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2340 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
2341 unused variable "field".
2342 * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
2343 (vsx_mergeh_<mode>): Likewise.
2344 * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
2345 (altivec_vmrghh): Likewise.
2346 (altivec_vmrghw): Likewise.
2347 (altivec_vmrglb): Likewise.
2348 (altivec_vmrglh): Likewise.
2349 (altivec_vmrglw): Likewise.
2350 (altivec_vspltb): Add missing uses.
2351 (altivec_vsplth): Likewise.
2352 (altivec_vspltw): Likewise.
2353 (altivec_vspltsf): Likewise.
2355 2014-01-30 Jakub Jelinek <jakub@redhat.com>
2358 * ifcvt.c (cond_exec_process_insns): Don't conditionalize
2359 frame related instructions.
2361 2014-01-30 Vladimir Makarov <vmakarov@redhat.com>
2363 PR rtl-optimization/59959
2364 * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
2365 any reload of register whose subreg is invalid.
2367 2014-01-30 Jakub Jelinek <jakub@redhat.com>
2369 * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
2370 * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
2371 Add missing return type - void.
2373 2014-01-30 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2375 * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
2376 gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
2377 remove element index adjustment for endian (now handled in vsx.md
2379 (altivec_expand_vec_perm_const): Use
2380 gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
2381 * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
2382 (vsx_xxspltw_<mode>): Adjust element index for little endian.
2383 * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
2384 define_expand and a new define_insn *altivec_vspltb_internal;
2385 adjust for -maltivec=be on a little endian target.
2386 (altivec_vspltb_direct): New.
2387 (altivec_vsplth): Divide into a define_expand and a new
2388 define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
2389 little endian target.
2390 (altivec_vsplth_direct): New.
2391 (altivec_vspltw): Divide into a define_expand and a new
2392 define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
2393 little endian target.
2394 (altivec_vspltw_direct): New.
2395 (altivec_vspltsf): Divide into a define_expand and a new
2396 define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
2397 a little endian target.
2399 2014-01-30 Richard Biener <rguenther@suse.de>
2401 PR tree-optimization/59993
2402 * tree-ssa-forwprop.c (associate_pointerplus): Check we
2403 can propagate form the earlier stmt and avoid the transform
2404 when the intermediate result is needed.
2406 2014-01-30 Alangi Derick <alangiderick@gmail.com>
2408 * README.Portability: Fix typo.
2410 2014-01-30 David Holsgrove <david.holsgrove@xilinx.com>
2412 * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
2413 comparison_operator with ordered_comparison_operator.
2415 2014-01-30 Nick Clifton <nickc@redhat.com>
2417 * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
2418 Rename to mn10300_store_multiple_regs.
2419 * config/mn10300/mn10300.c: Likewise.
2420 * config/mn10300/mn10300.md (store_movm): Fix typo: call
2421 store_multiple_regs.
2422 * config/mn10300/predicates.md (mn10300_store_multiple_operation):
2423 Call mn10300_store_multiple_regs.
2425 2014-01-30 Nick Clifton <nickc@redhat.com>
2426 DJ Delorie <dj@redhat.com>
2428 * config/rl78/rl78.c (register_sizes): Make the "upper half" of
2429 %fp 2 to keep registers after it properly word-aligned.
2430 (rl78_alloc_physical_registers_umul): Handle the case where both
2431 input operands are the same.
2433 2014-01-30 Richard Biener <rguenther@suse.de>
2435 PR tree-optimization/59903
2436 * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
2439 2014-01-30 Jason Merrill <jason@redhat.com>
2442 * tree.c (walk_type_fields): Handle VECTOR_TYPE.
2445 * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
2447 2014-01-30 Richard Biener <rguenther@suse.de>
2449 PR tree-optimization/59951
2450 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
2452 2014-01-30 Savin Zlobec <savin.zlobec@gmail.com>
2455 * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
2456 SFmode to DFmode case.
2458 2014-01-29 DJ Delorie <dj@redhat.com>
2460 * config/msp430/msp430.opt (-minrt): New.
2461 * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
2463 (ENDFILE_SPEC): Likewise.
2465 2014-01-29 Jan Hubicka <hubicka@ucw.cz>
2467 * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
2468 (estimate_function_body_sizes): Use it.
2470 2014-01-29 Paolo Carlini <paolo.carlini@oracle.com>
2473 * dwarf2out.c (is_cxx_auto): New.
2474 (is_base_type): Use it.
2475 (gen_type_die_with_usage): Likewise.
2477 2014-01-29 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2479 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Use
2480 CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
2481 * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
2482 -maltivec=be with LE targets.
2483 (vsx_mergeh_<mode>): Likewise.
2484 * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
2485 (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
2486 (altivec_vmrghb): Replace with define_expand and new
2487 *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
2488 (altivec_vmrghb_direct): New define_insn.
2489 (altivec_vmrghh): Replace with define_expand and new
2490 *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
2491 (altivec_vmrghh_direct): New define_insn.
2492 (altivec_vmrghw): Replace with define_expand and new
2493 *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
2494 (altivec_vmrghw_direct): New define_insn.
2495 (*altivec_vmrghsf): Adjust for endianness.
2496 (altivec_vmrglb): Replace with define_expand and new
2497 *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
2498 (altivec_vmrglb_direct): New define_insn.
2499 (altivec_vmrglh): Replace with define_expand and new
2500 *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
2501 (altivec_vmrglh_direct): New define_insn.
2502 (altivec_vmrglw): Replace with define_expand and new
2503 *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
2504 (altivec_vmrglw_direct): New define_insn.
2505 (*altivec_vmrglsf): Adjust for endianness.
2506 (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
2507 (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
2508 (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
2509 (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
2510 (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
2511 (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
2512 (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
2513 (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
2515 2014-01-29 Marcus Shawcroft <marcus.shawcroft@arm.com>
2517 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
2518 (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
2521 2014-01-29 Richard Biener <rguenther@suse.de>
2523 PR tree-optimization/58742
2524 * tree-ssa-forwprop.c (associate_pointerplus): Rename to
2525 associate_pointerplus_align.
2526 (associate_pointerplus_diff): New function.
2527 (associate_pointerplus): Likewise. Call associate_pointerplus_align
2528 and associate_pointerplus_diff.
2530 2014-01-29 Richard Biener <rguenther@suse.de>
2532 * lto-streamer.h (LTO_major_version): Bump to 3.
2533 (LTO_minor_version): Reset to 0.
2535 2014-01-29 Renlin Li <Renlin.Li@arm.com>
2537 * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
2538 * config/arm/arm.c (FL_FOR_ARCH7VE): New.
2539 (arm_file_start): Generate correct asm header for armv7ve.
2540 * config/arm/bpabi.h: Add multilib support for armv7ve.
2541 * config/arm/driver-arm.c: Change the architectures of cortex-a7
2542 and cortex-a15 to armv7ve.
2543 * config/arm/t-aprofile: Add multilib support for armv7ve.
2544 * doc/invoke.texi: Document -march=armv7ve.
2546 2014-01-29 Richard Biener <rguenther@suse.de>
2548 PR tree-optimization/58742
2549 * tree-ssa-forwprop.c (associate_plusminus): Return true
2550 if we changed sth, defer EH cleanup to ...
2551 (ssa_forward_propagate_and_combine): ... here. Call simplify_mult.
2552 (simplify_mult): New function.
2554 2014-01-29 Jakub Jelinek <jakub@redhat.com>
2557 PR tree-optimization/59920
2558 * tree.c (build_common_builtin_nodes): Remove
2559 __builtin_setjmp_dispatcher initialization.
2560 * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
2561 * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
2562 instead of gsi_after_labels + manually skipping debug stmts.
2563 Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
2564 ignore bbs with IFN_ABNORMAL_DISPATCHER.
2565 * tree-inline.c (copy_edges_for_bb): Remove
2566 can_make_abnormal_goto argument, instead add abnormal_goto_dest
2567 argument. Ignore computed_goto_p stmts. Don't call
2568 make_abnormal_goto_edges. If a call might need abnormal edges
2569 for non-local gotos, see if it already has an edge to
2570 IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
2571 with true argument, don't do anything then, otherwise add
2572 EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
2573 (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
2575 * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
2576 (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
2577 (lower_stmt): Don't set data->calls_builtin_setjmp.
2578 (lower_builtin_setjmp): Adjust comment.
2579 * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
2580 * tree-cfg.c (found_computed_goto): Remove.
2581 (factor_computed_gotos): Remove.
2582 (make_goto_expr_edges): Return bool, true for computed gotos.
2583 Don't call make_abnormal_goto_edges.
2584 (build_gimple_cfg): Don't set found_computed_goto, don't call
2585 factor_computed_gotos.
2586 (computed_goto_p): No longer static.
2587 (make_blocks): Don't set found_computed_goto.
2588 (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
2589 (make_edges): If make_goto_expr_edges returns true, push bb
2590 into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
2591 instead of calling make_abnormal_goto_edges push bb into ab_edge_call
2592 vector. Record mapping between bbs and OpenMP regions if there
2593 are any, adjust make_gimple_omp_edges caller. Call
2594 handle_abnormal_edges.
2595 (make_abnormal_goto_edges): Remove.
2596 * tree-cfg.h (make_abnormal_goto_edges): Remove.
2597 (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
2598 * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
2599 * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
2600 * internal-fn.def (ABNORMAL_DISPATCHER): New.
2601 * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
2602 filling *region also set *region_idx to (*region)->entry->index.
2605 * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
2606 For REGs set ORIGINAL_REGNO.
2608 2014-01-29 Bingfeng Mei <bmei@broadcom.com>
2610 * doc/md.texi: Mention that a target shouldn't implement
2611 vec_widen_(s|u)mul_even/odd pair if it is less efficient
2614 2014-01-29 Jakub Jelinek <jakub@redhat.com>
2616 PR tree-optimization/59594
2617 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
2618 a copy of the datarefs vector rather than the vector itself.
2620 2014-01-28 Jason Merrill <jason@redhat.com>
2623 * dwarf2out.c (auto_die): New static.
2624 (gen_type_die_with_usage): Handle C++1y 'auto'.
2625 (gen_subprogram_die): If in-class DIE had 'auto', emit type again
2628 2014-01-28 H.J. Lu <hongjiu.lu@intel.com>
2631 * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
2632 (SPEC_X32): Likewise.
2633 (SPEC_64): Likewise.
2634 * config/i386/i386.c (ix86_option_override_internal): Turn off
2635 OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
2637 (x86_file_start): Output .code16gcc for TARGET_16BIT.
2638 * config/i386/i386.h (TARGET_16BIT): New macro.
2639 (TARGET_16BIT_P): Likewise.
2640 * config/i386/i386.opt: Add m16.
2641 * doc/invoke.texi: Document -m16.
2643 2014-01-28 Jakub Jelinek <jakub@redhat.com>
2645 PR preprocessor/59935
2646 * input.c (location_get_source_line): Bail out on when line number
2647 is zero, and test the return value of lookup_or_add_file_to_cache_tab.
2649 2014-01-28 Richard Biener <rguenther@suse.de>
2651 PR tree-optimization/58742
2652 * tree-ssa-forwprop.c (associate_plusminus): Handle
2653 pointer subtraction of the form (T)(P + A) - (T)P.
2655 2014-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2657 * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
2660 2014-01-28 Richard Biener <rguenther@suse.de>
2663 2014-01-28 Richard Biener <rguenther@suse.de>
2665 PR rtl-optimization/45364
2666 PR rtl-optimization/59890
2667 * var-tracking.c (local_get_addr_clear_given_value): Handle
2668 already cleared slot.
2669 (val_reset): Handle not allocated local_get_addr_cache.
2670 (vt_find_locations): Use post-order on the inverted CFG.
2672 2014-01-28 Richard Biener <rguenther@suse.de>
2674 * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
2676 2014-01-28 Richard Biener <rguenther@suse.de>
2678 PR rtl-optimization/45364
2679 PR rtl-optimization/59890
2680 * var-tracking.c (local_get_addr_clear_given_value): Handle
2681 already cleared slot.
2682 (val_reset): Handle not allocated local_get_addr_cache.
2683 (vt_find_locations): Use post-order on the inverted CFG.
2685 2014-01-28 Alan Modra <amodra@gmail.com>
2687 * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
2688 * configure.ac <recursive call for build != host>: Define
2689 GENERATOR_FILE. Comment. Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
2690 and LD_FOR_BUILD too.
2691 * configure: Regenerate.
2693 2014-01-27 Allan Sandfeld Jensen <sandfeld@kde.org>
2695 * config/i386/i386.c (get_builtin_code_for_version): Separate
2696 Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
2697 Broadwell from Haswell.
2699 2014-01-27 Steve Ellcey <sellcey@mips.com>
2701 * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2702 Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
2703 * config/mips/mips.c (mips_option_override): Change setting
2705 * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
2706 * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
2707 Change from Mask to Var.
2709 2014-01-27 Jeff Law <law@redhat.com>
2711 * ipa-inline.c (inline_small_functions): Fix typo.
2713 2014-01-27 Ilya Tocar <ilya.tocar@intel.com>
2715 * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
2716 (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
2717 (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
2718 (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
2719 (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
2720 (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
2721 (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
2722 (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
2723 (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
2724 (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
2725 (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
2726 (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
2727 (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
2728 (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
2729 (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
2730 (_mm512_storeu_epi64): Ditto.
2731 (_mm512_cmpge_epi32_mask): Ditto.
2732 (_mm512_cmpge_epu32_mask): Ditto.
2733 (_mm512_cmpge_epi64_mask): Ditto.
2734 (_mm512_cmpge_epu64_mask): Ditto.
2735 (_mm512_cmple_epi32_mask): Ditto.
2736 (_mm512_cmple_epu32_mask): Ditto.
2737 (_mm512_cmple_epi64_mask): Ditto.
2738 (_mm512_cmple_epu64_mask): Ditto.
2739 (_mm512_cmplt_epi32_mask): Ditto.
2740 (_mm512_cmplt_epu32_mask): Ditto.
2741 (_mm512_cmplt_epi64_mask): Ditto.
2742 (_mm512_cmplt_epu64_mask): Ditto.
2743 (_mm512_cmpneq_epi32_mask): Ditto.
2744 (_mm512_cmpneq_epu32_mask): Ditto.
2745 (_mm512_cmpneq_epi64_mask): Ditto.
2746 (_mm512_cmpneq_epu64_mask): Ditto.
2747 (_mm512_expand_pd): Ditto.
2748 (_mm512_expand_ps): Ditto.
2749 * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
2750 VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
2751 VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
2752 * config/i386/i386.c (ix86_builtins): Add
2753 IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
2754 IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
2755 IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
2756 IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
2757 IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
2758 IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
2759 IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
2760 IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
2761 IX86_BUILTIN_PMOVUSQW512_MEM.
2762 (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
2763 __builtin_ia32_pmovsqd512mem_mask,
2764 __builtin_ia32_pmovqd512mem_mask,
2765 __builtin_ia32_pmovusqw512mem_mask,
2766 __builtin_ia32_pmovsqw512mem_mask,
2767 __builtin_ia32_pmovqw512mem_mask,
2768 __builtin_ia32_pmovusdw512mem_mask,
2769 __builtin_ia32_pmovsdw512mem_mask,
2770 __builtin_ia32_pmovdw512mem_mask,
2771 __builtin_ia32_pmovqb512mem_mask,
2772 __builtin_ia32_pmovusqb512mem_mask,
2773 __builtin_ia32_pmovsqb512mem_mask,
2774 __builtin_ia32_pmovusdb512mem_mask,
2775 __builtin_ia32_pmovsdb512mem_mask,
2776 __builtin_ia32_pmovdb512mem_mask.
2777 (bdesc_args): Add __builtin_ia32_expanddf512,
2778 __builtin_ia32_expandsf512.
2779 (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
2780 VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
2781 VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
2782 * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
2783 (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
2784 (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
2785 (avx512f_<code>v8div16qi2_mask_store): This.
2786 (avx512f_expand<mode>): New.
2788 2014-01-27 Kirill Yukhin <kirill.yukhin@intel.com>
2790 * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
2792 (_mm512_mask_prefetch_i64gather_pd): Ditto.
2793 (_mm512_prefetch_i32scatter_pd): Ditto.
2794 (_mm512_mask_prefetch_i32scatter_pd): Ditto.
2795 (_mm512_prefetch_i64scatter_pd): Ditto.
2796 (_mm512_mask_prefetch_i64scatter_pd): Ditto.
2797 (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
2798 (_mm512_mask_prefetch_i64gather_ps): Ditto.
2799 (_mm512_prefetch_i32scatter_ps): Ditto.
2800 (_mm512_mask_prefetch_i32scatter_ps): Ditto.
2801 (_mm512_prefetch_i64scatter_ps): Ditto.
2802 (_mm512_mask_prefetch_i64scatter_ps): Ditto.
2803 * config/i386/i386-builtin-types.def: Define
2804 VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
2805 and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
2806 * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
2807 IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
2808 IX86_BUILTIN_SCATTERPFQPD.
2809 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
2810 __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
2811 __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
2812 __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
2813 __builtin_ia32_scatterpfqps.
2814 (ix86_expand_builtin): Expand new built-ins.
2815 * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
2816 fix memory access data type.
2817 (*avx512pf_gatherpf<mode>_mask): Ditto.
2818 (*avx512pf_gatherpf<mode>): Ditto.
2819 (avx512pf_scatterpf<mode>): Ditto.
2820 (*avx512pf_scatterpf<mode>_mask): Ditto.
2821 (*avx512pf_scatterpf<mode>): Ditto.
2822 (GATHER_SCATTER_SF_MEM_MODE): New.
2823 (avx512pf_gatherpf<mode>df): Ditto.
2824 (*avx512pf_gatherpf<mode>df_mask): Ditto.
2825 (*avx512pf_scatterpf<mode>df): Ditto.
2827 2014-01-27 Jakub Jelinek <jakub@redhat.com>
2830 * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
2831 know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
2834 2014-01-27 James Greenhalgh <james.greenhalgh@arm.com>
2836 * common/config/arm/arm-common.c
2837 (arm_rewrite_mcpu): Handle multiple names.
2839 (BIG_LITTLE_SPEC): Do not discard mcpu switches.
2841 2014-01-27 James Greenhalgh <james.greenhalgh@arm.com>
2843 * gimple-builder.h (create_gimple_tmp): Delete.
2845 2014-01-27 Christian Bruel <christian.bruel@st.com>
2847 * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
2850 2014-01-26 John David Anglin <danglin@gcc.gnu.org>
2852 * config/pa/pa.md (call): Generate indirect long calls to non-local
2853 functions when outputing 32-bit code.
2854 (call_value): Likewise except for special call to buggy powf function.
2856 * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
2857 portable runtime and PIC indirect calls.
2858 (pa_output_indirect_call): Remove unnecessary nop from portable runtime
2859 and PIC call sequences. Use ldo instead of blr to set return register
2860 in PIC call sequence.
2862 2014-01-25 Walter Lee <walt@tilera.com>
2864 * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
2865 avoid clobbering a live register.
2867 2014-01-25 Walter Lee <walt@tilera.com>
2869 * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
2870 Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
2871 * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
2872 Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
2874 2014-01-25 Walter Lee <walt@tilera.com>
2876 * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
2877 arguments on even registers.
2878 (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
2880 * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
2881 (BIGGEST_ALIGNMENT): Ditto.
2882 (BIGGEST_FIELD_ALIGNMENT): Ditto.
2884 2014-01-25 Walter Lee <walt@tilera.com>
2886 * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
2887 insns before bundling.
2888 * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
2890 2014-01-25 Walter Lee <walt@tilera.com>
2892 * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
2893 PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
2894 * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
2896 2014-01-25 Richard Sandiford <rdsandiford@googlemail.com>
2898 * config/mips/constraints.md (kl): Delete.
2899 * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
2900 define expands, using...
2901 (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
2902 instructions for MIPS16.
2903 (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
2904 non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
2906 2014-01-25 Walter Lee <walt@tilera.com>
2908 * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
2912 2014-01-25 Walter Lee <walt@tilera.com>
2914 * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
2915 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2917 2014-01-25 Richard Sandiford <rdsandiford@googlemail.com>
2919 * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
2922 2014-01-25 Jakub Jelinek <jakub@redhat.com>
2924 * print-rtl.c (in_call_function_usage): New var.
2925 (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
2926 EXPR_LIST mode as mode and not as reg note name.
2929 * cfgloopmanip.c (copy_loop_info): If
2930 loop->warned_aggressive_loop_optimizations, make sure
2931 the flag is set in target loop too.
2933 2014-01-24 Balaji V. Iyer <balaji.v.iyer@intel.com>
2935 * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
2937 * builtins.def: Likewise.
2938 * cilk.h (fn_contains_cilk_spawn_p): Likewise.
2939 * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
2940 * ira.c (ira_setup_eliminable_regset): Likewise.
2941 * omp-low.c (gate_expand_omp): Likewise.
2942 (execute_lower_omp): Likewise.
2943 (diagnose_sb_0): Likewise.
2944 (gate_diagnose_omp_blocks): Likewise.
2945 (simd_clone_clauses_extract): Likewise.
2948 2014-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2950 * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
2951 correction for little endian...
2952 * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
2955 2014-01-24 Jeff Law <law@redhat.com>
2957 PR tree-optimization/59919
2958 * tree-vrp.c (find_assert_locations_1): Do not register asserts
2959 for non-returning calls.
2961 2014-01-24 James Greenhalgh <james.greenhalgh@arm.com>
2963 * common/config/aarch64/aarch64-common.c
2964 (aarch64_rewrite_mcpu): Handle multiple names.
2965 * config/aarch64/aarch64.h
2966 (BIG_LITTLE_SPEC): Do not discard mcpu switches.
2968 2014-01-24 Dodji Seketeli <dodji@redhat.com>
2970 * input.c (add_file_to_cache_tab): Handle the case where fopen
2973 2014-01-23 H.J. Lu <hongjiu.lu@intel.com>
2976 * config/i386/i386.md (pushsf splitter): Get stack adjustment
2977 from push operand if code of push isn't PRE_DEC.
2979 2014-01-23 Michael Meissner <meissner@linux.vnet.ibm.com>
2982 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2983 -mquad-memory-atomic. Update -mquad-memory documentation to say
2984 it is only used for non-atomic loads/stores.
2986 * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
2987 -mquad-memory or -mquad-memory-atomic switches.
2989 * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
2990 -mquad-memory-atomic to ISA 2.07 support.
2992 * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
2993 to separate support of normal quad word memory operations (ldq, stq)
2994 from the atomic quad word memory operations.
2996 * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
2997 support to separate non-atomic quad word operations from atomic
2998 quad word operations. Disable non-atomic quad word operations in
2999 little endian mode so that we don't have to swap words after the
3000 load and before the store.
3001 (quad_load_store_p): Add comment about atomic quad word support.
3002 (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
3003 options printed with -mdebug=reg.
3005 * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
3006 -mquad-memory-atomic as the test for whether we have quad word
3007 atomic instructions.
3008 (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
3009 or -mp8-vector are used, allow byte/half-word atomic operations.
3011 * config/rs6000/sync.md (load_lockedti): Insure that the address
3012 is a proper indexed or indirect address for the lqarx instruction.
3013 On little endian systems, swap the hi/lo registers after the lqarx
3015 (load_lockedpti): Use indexed_or_indirect_operand predicate to
3016 insure the address is valid for the lqarx instruction.
3017 (store_conditionalti): Insure that the address is a proper indexed
3018 or indirect address for the stqcrx. instruction. On little endian
3019 systems, swap the hi/lo registers before doing the stqcrx.
3021 (store_conditionalpti): Use indexed_or_indirect_operand predicate to
3022 insure the address is valid for the stqcrx. instruction.
3024 * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
3025 Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
3026 type of quad memory support is available.
3028 2014-01-23 Vladimir Makarov <vmakarov@redhat.com>
3031 * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
3032 there is a danger of looping.
3034 2014-01-23 Pat Haugen <pthaugen@us.ibm.com>
3036 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3037 force flag_ira_loop_pressure if set via command line.
3039 2014-01-23 Alex Velenko <Alex.Velenko@arm.com>
3041 * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
3042 (ashr_simd): New builtin handling DI mode.
3043 * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
3044 (aarch64_sshr_simddi): New match pattern.
3045 * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
3046 (vshrd_n_s64): Likewise.
3047 * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
3049 2014-01-23 Nick Clifton <nickc@redhat.com>
3051 * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
3052 (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
3053 favour of mcu specific scripts.
3054 * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
3057 2014-01-23 James Greenhalgh <james.greenhalgh@arm.com>
3058 Alex Velenko <Alex.Velenko@arm.com>
3060 * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
3061 (vaddv_s16): Likewise.
3062 (vaddv_s32): Likewise.
3063 (vaddv_u8): Likewise.
3064 (vaddv_u16): Likewise.
3065 (vaddv_u32): Likewise.
3066 (vaddvq_s8): Likewise.
3067 (vaddvq_s16): Likewise.
3068 (vaddvq_s32): Likewise.
3069 (vaddvq_s64): Likewise.
3070 (vaddvq_u8): Likewise.
3071 (vaddvq_u16): Likewise.
3072 (vaddvq_u32): Likewise.
3073 (vaddvq_u64): Likewise.
3074 (vaddv_f32): Likewise.
3075 (vaddvq_f32): Likewise.
3076 (vaddvq_f64): Likewise.
3077 (vmaxv_f32): Likewise.
3078 (vmaxv_s8): Likewise.
3079 (vmaxv_s16): Likewise.
3080 (vmaxv_s32): Likewise.
3081 (vmaxv_u8): Likewise.
3082 (vmaxv_u16): Likewise.
3083 (vmaxv_u32): Likewise.
3084 (vmaxvq_f32): Likewise.
3085 (vmaxvq_f64): Likewise.
3086 (vmaxvq_s8): Likewise.
3087 (vmaxvq_s16): Likewise.
3088 (vmaxvq_s32): Likewise.
3089 (vmaxvq_u8): Likewise.
3090 (vmaxvq_u16): Likewise.
3091 (vmaxvq_u32): Likewise.
3092 (vmaxnmv_f32): Likewise.
3093 (vmaxnmvq_f32): Likewise.
3094 (vmaxnmvq_f64): Likewise.
3095 (vminv_f32): Likewise.
3096 (vminv_s8): Likewise.
3097 (vminv_s16): Likewise.
3098 (vminv_s32): Likewise.
3099 (vminv_u8): Likewise.
3100 (vminv_u16): Likewise.
3101 (vminv_u32): Likewise.
3102 (vminvq_f32): Likewise.
3103 (vminvq_f64): Likewise.
3104 (vminvq_s8): Likewise.
3105 (vminvq_s16): Likewise.
3106 (vminvq_s32): Likewise.
3107 (vminvq_u8): Likewise.
3108 (vminvq_u16): Likewise.
3109 (vminvq_u32): Likewise.
3110 (vminnmv_f32): Likewise.
3111 (vminnmvq_f32): Likewise.
3112 (vminnmvq_f64): Likewise.
3114 2014-01-23 James Greenhalgh <james.greenhalgh@arm.com>
3116 * config/aarch64/aarch64-simd.md
3117 (aarch64_dup_lane<mode>): Correct lane number on big-endian.
3118 (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
3119 (*aarch64_mul3_elt<mode>): Likewise.
3120 (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
3121 (*aarch64_mul3_elt_to_64v2df): Likewise.
3122 (*aarch64_mla_elt<mode>): Likewise.
3123 (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
3124 (*aarch64_mls_elt<mode>): Likewise.
3125 (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
3126 (*aarch64_fma4_elt<mode>): Likewise.
3127 (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
3128 (*aarch64_fma4_elt_to_64v2df): Likewise.
3129 (*aarch64_fnma4_elt<mode>): Likewise.
3130 (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
3131 (*aarch64_fnma4_elt_to_64v2df): Likewise.
3132 (aarch64_sq<r>dmulh_lane<mode>): Likewise.
3133 (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
3134 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
3135 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
3136 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
3137 (aarch64_sqdmull_lane<mode>_internal): Likewise.
3138 (aarch64_sqdmull2_lane<mode>_internal): Likewise.
3140 2013-01-23 Alex Velenko <Alex.Velenko@arm.com>
3142 * config/aarch64/aarch64-simd.md
3143 (aarch64_be_checked_get_lane<mode>): New define_expand.
3144 * config/aarch64/aarch64-simd-builtins.def
3145 (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
3146 New builtin definition.
3147 * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
3148 Use new safe be builtin.
3150 2014-01-23 Alex Velenko <Alex.Velenko@arm.com>
3152 * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
3154 (aarch64_be_st1<mode>): Likewise.
3155 (aarch_ld1<VALL:mode>): Define_expand modified.
3156 (aarch_st1<VALL:mode>): Likewise.
3157 * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
3158 (UNSPEC_ST1): Likewise.
3160 2014-01-23 David Holsgrove <david.holsgrove@xilinx.com>
3162 * config/microblaze/microblaze.md: Add trap insn and attribute
3164 2014-01-23 Dodji Seketeli <dodji@redhat.com>
3166 PR preprocessor/58580
3167 * input.h (location_get_source_line): Take an additional line_size
3169 (void diagnostics_file_cache_fini): Declare new function.
3170 * input.c (struct fcache): New type.
3171 (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
3172 New static constants.
3173 (diagnostic_file_cache_init, total_lines_num)
3174 (lookup_file_in_cache_tab, evicted_cache_tab_entry)
3175 (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
3176 (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
3177 (get_next_line, read_next_line, goto_next_line, read_line_num):
3178 New static function definitions.
3179 (diagnostic_file_cache_fini): New function.
3180 (location_get_source_line): Take an additional output line_len
3181 parameter. Re-write using lookup_or_add_file_to_cache_tab and
3183 * diagnostic.c (diagnostic_finish): Call
3184 diagnostic_file_cache_fini.
3185 (adjust_line): Take an additional input parameter for the length
3186 of the line, rather than calculating it with strlen.
3187 (diagnostic_show_locus): Adjust the use of
3188 location_get_source_line and adjust_line with respect to their new
3189 signature. While displaying a line now, do not stop at the first
3190 null byte. Rather, display the zero byte as a space and keep
3191 going until we reach the size of the line.
3192 * Makefile.in: Add vec.o to OBJS-libcommon
3194 2014-01-23 Kirill Yukhin <kirill.yukhin@intel.com>
3195 Ilya Tocar <ilya.tocar@intel.com>
3197 * config/i386/avx512fintrin.h (_mm512_kmov): New.
3198 * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
3199 (__builtin_ia32_kmov16): Ditto.
3200 * config/i386/i386.md (UNSPEC_KMOV): New.
3203 2014-01-23 Kirill Yukhin <kirill.yukhin@intel.com>
3205 * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
3206 (_mm512_storeu_si512): Ditto.
3208 2014-01-23 Richard Sandiford <rdsandiford@googlemail.com>
3211 * rtl.h (get_referenced_operands): Declare.
3212 * recog.c (get_referenced_operands): New function.
3213 * config/mips/mips.c (mips_reorg_process_insns): Check which asm
3214 operands have been referenced when recording LO_SUM references.
3216 2014-01-22 David Holsgrove <david.holsgrove@xilinx.com>
3218 * config/microblaze/microblaze.md: Correct bswaphi2 insn.
3220 2014-01-22 Jan Hubicka <hubicka@ucw.cz>
3222 * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
3223 Enable for generic and recent AMD targets.
3225 2014-01-22 Jan Hubicka <hubicka@ucw.cz>
3227 * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
3228 ARG_SIZE note when adjustment was eliminated.
3230 2014-01-22 Jeff Law <law@redhat.com>
3232 PR tree-optimization/59597
3233 * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
3234 in file. Accept new argument REGISTERING and use it to modify
3235 dump output appropriately.
3236 (register_jump_thread): Corresponding changes.
3237 (mark_threaded_blocks): Reinstate code to cancel unprofitable
3238 thread paths involving joiner blocks. Add code to dump cancelled
3239 jump threading paths.
3241 2014-01-22 Vladimir Makarov <vmakarov@redhat.com>
3243 PR rtl-optimization/59477
3244 * lra-constraints.c (inherit_in_ebb): Process call for living hard
3245 regs. Update reloads_num and potential_reload_hard_regs for all insns.
3247 2014-01-22 Tom Tromey <tromey@redhat.com>
3249 * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
3251 * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
3253 2014-01-21 Vladimir Makarov <vmakarov@redhat.com>
3255 PR rtl-optimization/59896
3256 * lra-constraints.c (process_alt_operands): Check unused note for
3257 matched operands of insn with no output reloads.
3259 2014-01-21 Richard Sandiford <rdsandiford@googlemail.com>
3261 * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
3262 (mips_move_from_gpr_cost): Likewise.
3264 2014-01-21 Vladimir Makarov <vmakarov@redhat.com>
3266 PR rtl-optimization/59858
3267 * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
3268 ira_class_hard_regs_num.
3269 (process_alt_operands): Increase reject for dying matched operand.
3271 2014-01-21 Jakub Jelinek <jakub@redhat.com>
3274 * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
3275 smaller than size, perform several stores or loads and stores
3276 at dst + count - size to store or copy all of size bytes, rather
3277 than just last modesize bytes.
3279 2014-01-20 DJ Delorie <dj@redhat.com>
3281 * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
3282 that CLOBBERs are REGs before propogating their values.
3284 2014-01-20 H.J. Lu <hongjiu.lu@intel.com>
3287 * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
3288 (cgraph_inline_failed_type): New function.
3289 * cgraph.h (DEFCIFCODE): Add type.
3290 (cgraph_inline_failed_type_t): New enum.
3291 (cgraph_inline_failed_type): New prototype.
3292 * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
3293 FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
3294 FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
3295 LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
3296 MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
3297 RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
3298 OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
3299 INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
3300 Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
3301 FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
3302 EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
3303 OPTIMIZATION_MISMATCH.
3304 * tree-inline.c (expand_call_inline): Emit errors during
3305 early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
3307 2014-01-20 Uros Bizjak <ubizjak@gmail.com>
3310 * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
3311 mode attribute in insn output.
3313 2014-01-20 Eric Botcazou <ebotcazou@adacore.com>
3315 * output.h (output_constant): Delete.
3316 * varasm.c (output_constant): Make private.
3318 2014-01-20 Alex Velenko <Alex.Velenko@arm.com>
3320 * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
3322 2014-01-20 Jakub Jelinek <jakub@redhat.com>
3325 * tree.h (fold_builtin_strcat): New prototype.
3326 * builtins.c (fold_builtin_strcat): No longer static. Add len
3327 argument, if non-NULL, don't call c_strlen. Optimize
3328 directly into __builtin_memcpy instead of __builtin_strcpy.
3329 (fold_builtin_2): Adjust fold_builtin_strcat caller.
3330 * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
3332 2014-01-20 Uros Bizjak <ubizjak@gmail.com>
3334 * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
3335 for SImode_address_operand operands, having only a REG argument.
3337 2014-01-20 Marcus Shawcroft <marcus.shawcroft@arm.com>
3339 * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
3340 loader name using mbig-endian.
3341 (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
3343 2014-01-20 James Greenhalgh <james.greenhalgh@arm.com>
3345 * doc/invoke.texi (-march): Clarify documentation for AArch64.
3349 2014-01-20 Tejas Belagod <tejas.belagod@arm.com>
3351 * config/aarch64/aarch64-protos.h
3352 (aarch64_cannot_change_mode_class_ptr): Declare.
3353 * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
3354 aarch64_cannot_change_mode_class_ptr): New.
3355 * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
3356 backend hook aarch64_cannot_change_mode_class.
3358 2014-01-20 James Greenhalgh <james.greenhalgh@arm.com>
3360 * common/config/aarch64/aarch64-common.c
3361 (aarch64_handle_option): Don't handle any option order logic here.
3362 * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
3363 selected_cpu, warn on architecture version mismatch.
3364 (aarch64_override_options): Fix parsing order for option strings.
3366 2014-01-20 Jan-Benedict Glaw <jbglaw@lug-owl.de>
3367 Iain Sandoe <iain@codesourcery.com>
3370 * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
3371 warning. Amend comment to reflect current functionality.
3373 2014-01-20 Richard Biener <rguenther@suse.de>
3376 * builtins.c (fold_builtin_strcat): Remove case better handled
3377 by tree-ssa-strlen.c.
3379 2014-01-20 Alan Lawrence <alan.lawrence@arm.com>
3381 * config/aarch64/aarch64.opt
3382 (mcpu, march, mtune): Make case-insensitive.
3384 2014-01-20 Jakub Jelinek <jakub@redhat.com>
3387 * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
3388 if operands[1] is a REG or ZERO_EXTEND of a REG.
3390 2014-01-19 Jan Hubicka <hubicka@ucw.cz>
3392 * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
3394 2014-01-19 John David Anglin <danglin@gcc.gnu.org>
3396 * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
3397 long non-pic millicode calls.
3399 2014-01-19 Jan-Benedict Glaw <jbglaw@lug-owl.de>
3401 * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
3403 2014-01-19 Kito Cheng <kito@0xlab.org>
3405 * builtins.c (expand_movstr): Check movstr expand done or fail.
3407 2014-01-18 Uros Bizjak <ubizjak@gmail.com>
3408 H.J. Lu <hongjiu.lu@intel.com>
3411 * config/i386/i386.md (*lea<mode>): Zero-extend return register
3412 to DImode for zero-extended addresses.
3414 2014-01-19 Jakub Jelinek <jakub@redhat.com>
3416 PR rtl-optimization/57763
3417 * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
3418 on the new indirect jump_insn and increment LABEL_NUSES (label).
3420 2014-01-18 H.J. Lu <hongjiu.lu@intel.com>
3424 * config.gcc (x86_archs): New variable.
3425 (x86_64_archs): Likewise.
3426 (x86_cpus): Likewise.
3427 Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
3428 --with-arch/--with-cpu= options.
3429 Support --with-arch=/--with-cpu={nehalem,westmere,
3430 sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
3432 2014-01-18 Uros Bizjak <ubizjak@gmail.com>
3434 * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
3435 and PROCESSOR_ATHLON to simplify code. Move "memory" calculation.
3437 2014-01-18 Uros Bizjak <ubizjak@gmail.com>
3439 * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
3441 2014-01-18 Jakub Jelinek <jakub@redhat.com>
3444 * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
3445 clear cpp_get_options (parse_in)->warn_unused_macros for
3446 ix86_target_macros_internal with cpp_define.
3448 2014-01-18 Richard Sandiford <rdsandiford@googlemail.com>
3450 * jump.c (delete_related_insns): Keep (use (insn))s.
3451 * reorg.c (redundant_insn): Check for barriers too.
3453 2014-01-17 H.J. Lu <hongjiu.lu@intel.com>
3455 * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
3457 2014-01-17 John David Anglin <danglin@gcc.gnu.org>
3459 * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
3460 call to $$dyncall when TARGET_LONG_CALLS is true.
3462 2014-01-17 Jeff Law <law@redhat.com>
3464 * ree.c (combine_set_extension): Temporarily disable test for
3465 changing number of hard registers.
3467 2014-01-17 Jan Hubicka <hubicka@ucw.cz>
3470 * ipa-inline-analysis.c (inline_free_summary):
3471 Do not free summary of aliases.
3473 2014-01-17 Jakub Jelinek <jakub@redhat.com>
3476 * gimplify.c (gimplify_expr): Use create_tmp_var
3477 instead of create_tmp_var_raw. If cond doesn't have
3478 integral type, don't add the IFN_ANNOTATE builtin at all.
3480 2014-01-17 Martin Jambor <mjambor@suse.cz>
3483 * ipa-cp.c (prev_edge_clone): New variable.
3484 (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
3485 Also resize prev_edge_clone vector.
3486 (ipcp_edge_duplication_hook): Also update prev_edge_clone.
3487 (ipcp_edge_removal_hook): New function.
3488 (ipcp_driver): Register ipcp_edge_removal_hook.
3490 2014-01-17 Andrew Pinski <apinski@cavium.com>
3491 Steve Ellcey <sellcey@mips.com>
3494 * config/mips/mips.c (mips_print_operand): Check operand mode instead
3497 2014-01-17 Jeff Law <law@redhat.com>
3500 * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
3501 so that pass_ccp runs first.
3503 2014-01-17 H.J. Lu <hongjiu.lu@intel.com>
3505 * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
3506 (ix86_adjust_cost): Use !TARGET_XXX.
3507 (do_reorder_for_imul): Likewise.
3508 (swap_top_of_ready_list): Likewise.
3509 (ix86_sched_reorder): Likewise.
3511 2014-01-17 H.J. Lu <hongjiu.lu@intel.com>
3513 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3514 PROCESSOR_INTEL. Treat like PROCESSOR_GENERIC.
3515 * config/i386/i386.c (intel_memcpy): New. Duplicate slm_memcpy.
3516 (intel_memset): New. Duplicate slm_memset.
3517 (intel_cost): New. Duplicate slm_cost.
3518 (m_INTEL): New macro.
3519 (processor_target_table): Add "intel".
3520 (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
3521 with PROCESSOR_INTEL for "intel".
3522 (ix86_lea_outperforms): Support PROCESSOR_INTEL. Duplicate
3523 PROCESSOR_SILVERMONT.
3524 (ix86_issue_rate): Likewise.
3525 (ix86_adjust_cost): Likewise.
3526 (ia32_multipass_dfa_lookahead): Likewise.
3527 (swap_top_of_ready_list): Likewise.
3528 (ix86_sched_reorder): Likewise.
3529 (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
3530 instead of TARGET_OPT_AGU.
3531 * config/i386/i386.h (TARGET_INTEL): New.
3532 (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
3533 (processor_type): Add PROCESSOR_INTEL.
3534 * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
3535 Add X86_TUNE_AVOID_LEA_FOR_ADDR.
3537 2014-01-17 Marek Polacek <polacek@redhat.com>
3540 * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
3543 2014-01-17 Richard Biener <rguenther@suse.de>
3545 PR tree-optimization/46590
3546 * opts.c (default_options_table): Add entries for
3547 OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
3548 all enabled at -O1 but not for -Og.
3549 * common.opt (fbranch-count-reg): Remove Init(1).
3550 (fmove-loop-invariants): Likewise.
3551 (ftree-pta): Likewise.
3553 2014-01-17 Jakub Jelinek <jakub@redhat.com>
3555 * config/i386/i386.c (ix86_data_alignment): For compatibility with
3556 (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
3557 decls to at least the GCC 4.8 used alignments.
3560 * tree-nested.c (convert_nonlocal_reference_stmt,
3561 convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
3562 of GIMPLE_BIND stmts, adjust associated decls.
3564 2014-01-17 Richard Biener <rguenther@suse.de>
3566 PR tree-optimization/46590
3567 * vec.h (vec<>::bseach): New member function implementing
3568 binary search according to C89 bsearch.
3569 (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
3570 * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
3571 bitmap pointer again. Make accesses_in_loop a flat array.
3572 (mem_ref_obstack): New global.
3573 (outermost_indep_loop): Adjust for mem_ref->stored changes.
3574 (mark_ref_stored): Likewise.
3575 (ref_indep_loop_p_2): Likewise.
3576 (set_ref_stored_in_loop): New helper function.
3577 (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
3578 (memref_free): Adjust.
3579 (record_mem_ref_loc): Simplify.
3580 (gather_mem_refs_stmt): Adjust.
3581 (sort_locs_in_loop_postorder_cmp): New function.
3582 (analyze_memory_references): Sort accesses_in_loop after
3583 loop postorder number.
3584 (find_ref_loc_in_loop_cmp): New function.
3585 (for_all_locs_in_loop): Find relevant cluster of locs in
3586 accesses_in_loop and iterate without recursion.
3587 (execute_sm): Avoid uninit warning.
3588 (struct ref_always_accessed): Simplify.
3589 (ref_always_accessed::operator ()): Likewise.
3590 (ref_always_accessed_p): Likewise.
3591 (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
3592 loop postorder numbers here.
3593 (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
3596 2014-01-17 Jan Hubicka <hubicka@ucw.cz>
3599 * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
3600 on decls for which assemble_alias has been called.
3602 2014-01-17 Nick Clifton <nickc@redhat.com>
3604 * config/msp430/msp430.opt: (mcpu): New option.
3605 * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
3606 (msp430_option_override): Parse target_cpu. If the MCU name
3607 matches a generic string, clear target_mcu.
3608 (msp430_attr): Allow numeric interrupt values up to 63.
3609 (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
3610 * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
3612 * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
3614 * config/msp430/msp430.md (popm): Use %J rather than %I.
3615 (addsi3): Use msp430_nonimmediate_operand for operand 2.
3616 (addhi_cy_i): Use immediate_operand for operand 2.
3617 * doc/invoke.texi: Document -mcpu option.
3619 2014-01-17 Richard Biener <rguenther@suse.de>
3621 PR rtl-optimization/38518
3622 * df.h (df_analyze_loop): Declare.
3623 * df-core.c: Include cfgloop.h.
3624 (df_analyze_1): Split out main part of df_analyze.
3625 (df_analyze): Adjust.
3626 (loop_inverted_post_order_compute): New function.
3627 (loop_post_order_compute): Likewise.
3628 (df_analyze_loop): New function avoiding whole-function
3630 * loop-invariant.c (find_defs): Use df_analyze_loop.
3631 (find_invariants): Adjust.
3632 * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
3634 2014-01-17 Zhenqiang Chen <zhenqiang.chen@arm.com>
3636 * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
3637 (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
3639 2014-01-16 Ilya Enkovich <ilya.enkovich@intel.com>
3641 * ipa-ref.c (ipa_remove_stmt_references): Fix references
3642 traversal when removing references.
3644 2014-01-16 Jan Hubicka <hubicka@ucw.cz>
3647 * tree.c (get_binfo_at_offset): Look harder for virtual bases.
3649 2014-01-16 Bernd Schmidt <bernds@codesourcery.com>
3652 * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
3653 pushing a reload for an autoinc when we had previously reloaded an
3654 inner part of the address.
3656 2014-01-16 Jakub Jelinek <jakub@redhat.com>
3658 * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
3660 (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
3661 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
3662 when not giving up or versioning for alias only because of
3664 (vect_analyze_data_ref_dependences): Set to true.
3665 * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
3667 (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
3668 LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
3672 * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
3675 * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
3676 operand 0 predicate for gathers, use a new pseudo as subtarget.
3678 2014-01-16 Vladimir Makarov <vmakarov@redhat.com>
3681 * lra-constraints.c (process_alt_operands): Add printing debug info.
3682 Check absence of input/output reloads for matched operands too.
3684 2014-01-16 Vladimir Makarov <vmakarov@redhat.com>
3686 PR rtl-optimization/59835
3687 * ira.c (ira_init_register_move_cost): Increase cost for
3690 2014-01-16 Alan Lawrence <alan.lawrence@arm.com>
3692 * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
3694 2014-01-16 Richard Earnshaw <rearnsha@arm.com>
3697 * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
3698 non-register objects. Use gen_(high/low)part more consistently.
3701 2014-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3704 * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
3705 endian support, remove tests for WORDS_BIG_ENDIAN.
3706 (p8_mfvsrd_3_<mode>): Likewise.
3707 (reload_gpr_from_vsx<mode>): Likewise.
3708 (reload_gpr_from_vsxsf): Likewise.
3709 (p8_mfvsrd_4_disf): Likewise.
3711 2014-01-16 Richard Biener <rguenther@suse.de>
3713 PR rtl-optimization/46590
3714 * lcm.c (compute_antinout_edge): Use postorder iteration.
3715 (compute_laterin): Use inverted postorder iteration.
3717 2014-01-16 Nick Clifton <nickc@redhat.com>
3720 * varasm.c (output_constant): Return the number of bytes actually
3722 (output_constructor_array_range): Update the field size with the
3723 number of bytes emitted by output_constant.
3724 (output_constructor_regular_field): Likewise. Also do not
3725 complain if the total number of bytes emitted is now greater
3726 than the expected fieldpos.
3727 * output.h (output_constant): Update prototype and descriptive comment.
3729 2014-01-16 Marek Polacek <polacek@redhat.com>
3732 * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
3733 it is error_mark_node.
3735 2014-01-15 Uros Bizjak <ubizjak@gmail.com>
3737 * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
3738 VALID_AVX256_REG_OR_OI_MODE.
3740 2014-01-15 Pat Haugen <pthaugen@us.ibm.com>
3742 * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
3743 current procedure should be profiled.
3745 2014-01-15 Andrew Pinski <apinski@cavium.com>
3747 * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
3748 of moving from/to the STACK_REG register class.
3750 2014-01-15 Richard Henderson <rth@redhat.com>
3753 * reginfo.c (global_regs_decl): Globalize.
3754 * rtl.h (global_regs_decl): Declare.
3755 * ira.c (do_reload): Diagnose frame_pointer_needed and it
3756 reserved via global_regs.
3758 2014-01-15 Teresa Johnson <tejohnson@google.com>
3760 * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
3762 2014-01-15 Bill Schmidt <wschmidt@vnet.linux.ibm.com>
3764 * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
3765 and vmulosh rather than call gen_vec_widen_smult_*.
3766 (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
3767 than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
3768 (vec_widen_smult_even_v16qi): Likewise.
3769 (vec_widen_umult_even_v8hi): Likewise.
3770 (vec_widen_smult_even_v8hi): Likewise.
3771 (vec_widen_umult_odd_v16qi): Likewise.
3772 (vec_widen_smult_odd_v16qi): Likewise.
3773 (vec_widen_umult_odd_v8hi): Likewise.
3774 (vec_widen_smult_odd_v8hi): Likewise.
3775 (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
3776 vmuloub rather than call gen_vec_widen_umult_*.
3777 (vec_widen_umult_lo_v16qi): Likewise.
3778 (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
3779 vmulosb rather than call gen_vec_widen_smult_*.
3780 (vec_widen_smult_lo_v16qi): Likewise.
3781 (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
3782 rather than call gen_vec_widen_umult_*.
3783 (vec_widen_umult_lo_v8hi): Likewise.
3784 (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
3785 rather than call gen_vec_widen_smult_*.
3786 (vec_widen_smult_lo_v8hi): Likewise.
3788 2014-01-15 Jeff Law <law@redhat.com>
3790 PR tree-optimization/59747
3791 * ree.c (find_and_remove_re): Properly handle case where a second
3792 eliminated extension requires widening a copy created for elimination
3793 of a prior extension.
3794 (combine_set_extension): Ensure that the number of hard regs needed
3795 for a destination register does not change when we widen it.
3797 2014-01-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
3799 * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
3800 (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
3801 (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
3802 (arm*-*-rtems*): Use t-rtems from existing tmake_file.
3803 (avr-*-rtems*): Likewise.
3804 (bfin*-rtems*): Likewise.
3805 (moxie-*-rtems*): Likewise.
3806 (h8300-*-rtems*): Likewise.
3807 (i[34567]86-*-rtems*): Likewise.
3808 (lm32-*-rtems*): Likewise.
3809 (m32r-*-rtems*): Likewise.
3810 (m68k-*-rtems*): Likewise.
3811 (microblaze*-*-rtems*): Likewise.
3812 (mips*-*-rtems*): Likewise.
3813 (powerpc-*-rtems*): Likewise.
3814 (sh-*-rtems*): Likewise.
3815 (sparc-*-rtems*): Likewise.
3816 (sparc64-*-rtems*): Likewise.
3817 (v850-*-rtems*): Likewise.
3818 (m32c-*-rtems*): Likewise.
3820 2014-01-15 Vladimir Makarov <vmakarov@redhat.com>
3822 PR rtl-optimization/59511
3823 * ira.c (ira_init_register_move_cost): Use memory costs for some
3824 cases of register move cost calculations.
3825 * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
3826 instead of BB frequency.
3827 * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
3828 * lra-assigns.c (find_hard_regno_for): Ditto.
3830 2014-01-15 Richard Biener <rguenther@suse.de>
3832 PR tree-optimization/59822
3833 * tree-vect-stmts.c (hoist_defs_of_uses): New function.
3834 (vectorizable_load): Use it to hoist defs of uses of invariant
3835 loads out of the loop.
3837 2014-01-15 Matthew Gretton-Dann <matthew.gretton-dann@linaro.org>
3838 Kugan Vivekanandarajah <kuganv@linaro.org>
3841 * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
3844 2014-01-15 Richard Biener <rguenther@suse.de>
3846 PR rtl-optimization/59802
3847 * lcm.c (compute_available): Use inverted postorder to seed
3848 the initial worklist.
3850 2014-01-15 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
3853 * config/s390/s390.c (s390_preferred_reload_class): Don't return
3854 ADDR_REGS for invalid symrefs in non-PIC code.
3856 2014-01-15 Jakub Jelinek <jakub@redhat.com>
3859 * builtins.c (determine_block_size): Initialize *probable_max_size
3860 even if len_rtx is CONST_INT.
3862 2014-01-14 Andrew Pinski <apinski@cavium.com>
3864 * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
3865 * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
3866 (cortexa53_tunings): Likewise.
3867 (aarch64_sched_issue_rate): New function.
3868 (TARGET_SCHED_ISSUE_RATE): Define.
3870 2014-01-14 Vladimir Makarov <vmakarov@redhat.com>
3872 * ira-costs.c (find_costs_and_classes): Add missed
3873 ira_init_register_move_cost_if_necessary.
3875 2014-01-14 Vladimir Makarov <vmakarov@redhat.com>
3878 * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
3880 2014-01-14 H.J. Lu <hongjiu.lu@intel.com>
3883 * config/i386/i386.c (type_natural_mode): Add a bool parameter
3884 to indicate if type is used for function return value. Warn ABI
3885 change if the vector mode isn't available for function return value.
3886 (ix86_function_arg_advance): Pass false to type_natural_mode.
3887 (ix86_function_arg): Likewise.
3888 (ix86_gimplify_va_arg): Likewise.
3889 (function_arg_32): Don't warn ABI change.
3890 (ix86_function_value): Pass true to type_natural_mode.
3891 (ix86_return_in_memory): Likewise.
3892 (ix86_struct_value_rtx): Removed.
3893 (TARGET_STRUCT_VALUE_RTX): Likewise.
3895 2014-01-14 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
3897 * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
3898 converting a conditional jump into a conditional return.
3900 2014-01-14 Richard Biener <rguenther@suse.de>
3902 PR tree-optimization/58921
3903 PR tree-optimization/59006
3904 * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
3905 hoisting invariant stmts.
3906 * tree-vect-stmts.c (vectorizable_load): Insert the splat of
3907 invariant loads on the preheader edge if possible.
3909 2014-01-14 Joey Ye <joey.ye@arm.com>
3911 * doc/plugin.texi (Building GCC plugins): Update to C++.
3913 2014-01-14 Kirill Yukhin <kirill.yukhin@intel.com>
3915 * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
3916 (_mm_rcp28_round_ss): Ditto.
3917 (_mm_rsqrt28_round_sd): Ditto.
3918 (_mm_rsqrt28_round_ss): Ditto.
3919 (_mm_rcp28_sd): Ditto.
3920 (_mm_rcp28_ss): Ditto.
3921 (_mm_rsqrt28_sd): Ditto.
3922 (_mm_rsqrt28_ss): Ditto.
3923 * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
3924 * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
3925 * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
3926 (IX86_BUILTIN_RCP28SD): Ditto.
3927 (IX86_BUILTIN_RCP28SS): Ditto.
3928 (IX86_BUILTIN_RSQRT28SD): Ditto.
3929 (IX86_BUILTIN_RSQRT28SS): Ditto.
3930 (bdesc_special_args): Define __builtin_ia32_movntdqa512,
3931 __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
3932 __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
3933 (ix86_expand_special_args_builtin): Expand new FTYPE.
3934 * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
3935 (srcp14<mode>): Make insn unary.
3936 (avx512f_vmscalef<mode><round_name>): Use substed predicate.
3937 (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
3938 (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
3939 (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
3940 (avx512er_exp2<mode><mask_name><round_saeonly_name>):
3941 Fix rounding: make it SAE only.
3942 (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
3944 (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
3946 (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
3947 (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
3948 (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
3949 * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
3950 (round_saeonly_mask_scalar_operand4): Ditto.
3951 (round_saeonly_mask_scalar_op3): Ditto.
3952 (round_saeonly_mask_scalar_op4): Ditto.
3954 2014-01-13 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3956 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3957 Implement -maltivec=be for vec_insert and vec_extract.
3959 2014-01-10 DJ Delorie <dj@redhat.com>
3961 * config/msp430/msp430.md (call_internal): Don't allow memory
3962 references with SP as the base register.
3963 (call_value_internal): Likewise.
3964 * config/msp430/constraints.md (Yc): New. For memory references
3965 that don't use SP as a base register.
3967 * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
3968 "an integer without a # prefix"
3969 * config/msp430/msp430.md (epilogue_helper): Use it.
3971 2014-01-13 Jakub Jelinek <jakub@redhat.com>
3974 * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
3975 AVX512F gather builtins.
3976 * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
3977 on gather decls with INTEGER_TYPE masktype.
3978 (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
3979 directly into the builtin rather than hoisting it before loop.
3981 PR tree-optimization/59387
3982 * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
3983 (scev_const_prop): If folded_casts and type has undefined overflow,
3984 use force_gimple_operand instead of force_gimple_operand_gsi and
3985 for each added stmt if it is assign with
3986 arith_code_with_undefined_signed_overflow, call
3987 rewrite_to_defined_overflow.
3988 * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
3989 gimple-fold.h instead.
3990 (arith_code_with_undefined_signed_overflow,
3991 rewrite_to_defined_overflow): Moved to ...
3992 * gimple-fold.c (arith_code_with_undefined_signed_overflow,
3993 rewrite_to_defined_overflow): ... here. No longer static.
3994 Include gimplify-me.h.
3995 * gimple-fold.h (arith_code_with_undefined_signed_overflow,
3996 rewrite_to_defined_overflow): New prototypes.
3998 2014-01-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4000 * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
4002 2014-01-13 Eric Botcazou <ebotcazou@adacore.com>
4004 * builtins.c (get_object_alignment_2): Minor tweak.
4005 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
4007 2014-01-13 Christian Bruel <christian.bruel@st.com>
4009 * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
4010 optimized non constant lengths.
4012 2014-01-13 Jakub Jelinek <jakub@redhat.com>
4015 * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
4016 load as __atomic_load_N if possible.
4018 2014-01-11 David Edelsohn <dje.gcc@gmail.com>
4020 * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
4022 (rs6000_expand_builtin): Adjust call.
4024 2014-01-11 David Edelsohn <dje.gcc@gmail.com>
4027 * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
4028 * config/rs6000/rs6000.c: Include target-globals.h.
4029 (rs6000_set_current_function): Instead of doing target_reinit
4030 unconditionally, use save_target_globals_default_opts and
4031 restore_target_globals.
4033 * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
4035 * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
4036 (rs6000_expand_builtin): Handle mffs and mtfsf.
4037 (rs6000_init_builtins): Define mffs and mtfsf.
4038 * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
4039 (rs6000_mffs): New pattern.
4040 (rs6000_mtfsf): New pattern.
4042 2014-01-11 Bin Cheng <bin.cheng@arm.com>
4044 * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
4045 Start narrowing with START. Apply candidate-use pair
4046 and check overall cost in narrowing.
4047 (iv_ca_prune): Pass new argument.
4049 2014-01-10 Jeff Law <law@redhat.com>
4052 * ree.c (combine_reaching_defs): Ensure the defining statement
4053 occurs before the extension when optimizing extensions with
4054 different source and destination hard registers.
4056 2014-01-10 Jan Hubicka <hubicka@ucw.cz>
4059 * ipa-devirt.c (build_type_inheritance_graph): Also add types of
4060 vtables into the type inheritance graph.
4062 2014-01-10 Jakub Jelinek <jakub@redhat.com>
4064 PR rtl-optimization/59754
4065 * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
4066 modes in the REGNO != REGNO case.
4068 2014-01-10 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4070 * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
4072 2014-01-10 Jakub Jelinek <jakub@redhat.com>
4074 PR tree-optimization/59745
4075 * tree-predcom.c (tree_predictive_commoning_loop): Call
4076 free_affine_expand_cache if giving up because components is NULL.
4078 * target-globals.c (save_target_globals): Allocate < 4KB structs using
4079 GC in payload of target_globals struct instead of allocating them on
4080 the heap and the larger structs separately using GC.
4081 * target-globals.h (struct target_globals): Make regs, hard_regs,
4082 reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
4083 of GTY((skip)) and change type to void *.
4084 (reset_target_globals): Cast loads from those fields to corresponding
4087 2014-01-10 Steve Ellcey <sellcey@mips.com>
4090 * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
4091 gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
4092 tree-ssanames.h, print-tree.h, varasm.h, and context.h.
4094 2014-01-10 Richard Earnshaw <rearnsha@arm.com>
4097 * aarch64-modes.def (CC_Zmode): New flags mode.
4098 * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
4099 represents an equality.
4100 (aarch64_get_condition_code): Handle CC_Zmode.
4101 * aarch64.md (compare_neg<mode>): Restrict to equality operations.
4103 2014-01-10 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
4105 * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
4106 extraction in good case.
4108 2014-01-10 Richard Biener <rguenther@suse.de>
4110 PR tree-optimization/59374
4111 * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
4112 checking after SLP discovery. Mark stmts not participating
4113 in any SLP instance properly.
4115 2014-01-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4117 * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
4118 when handling a SET rtx.
4120 2014-01-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4122 * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
4123 (cortex-a57): Likewise.
4124 (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
4126 2014-01-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4128 * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
4129 non-iwmmxt builtins.
4131 2014-01-10 Jan Hubicka <hubicka@ucw.cz>
4135 * ipa-devirt.c record_target_from_binfo): Take as argument
4136 stack of binfos and lookup matching one for virtual inheritance.
4137 (possible_polymorphic_call_targets_1): Update.
4139 2014-01-10 Huacai Chen <chenhc@lemote.com>
4141 * config/mips/driver-native.c (host_detect_local_cpu): Handle new
4142 kernel strings for Loongson-2E/2F/3A.
4144 2014-01-10 Jakub Jelinek <jakub@redhat.com>
4147 * tree-vect-data-refs.c (vect_analyze_data_refs): Check
4148 is_gimple_call before calling gimple_call_internal_p.
4150 2014-01-09 Steve Ellcey <sellcey@mips.com>
4152 * Makefile.in (TREE_FLOW_H): Remove.
4153 (TREE_SSA_H): Add file names from tree-flow.h.
4154 * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
4155 * tree.h: Remove tree-flow.h reference.
4156 * hash-table.h: Remove tree-flow.h reference.
4157 * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
4158 reference with tree-ssa-loop.h.
4160 2014-01-09 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4162 * doc/invoke.texi: Add -maltivec={be,le} options, and document
4163 default element-order behavior for -maltivec.
4164 * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
4165 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
4166 that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
4167 when targeting big endian, at least for now.
4168 * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
4170 2014-01-09 Jakub Jelinek <jakub@redhat.com>
4173 * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
4174 var satisfies use_register_for_decl, just take into account type
4175 alignment, rather than decl alignment.
4177 PR tree-optimization/59622
4178 * gimple-fold.c (gimple_fold_call): Fix a typo in message. For
4179 __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
4180 __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
4181 Don't devirtualize for inplace at all. For targets.length () == 1,
4182 if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
4184 2014-01-09 H.J. Lu <hongjiu.lu@intel.com>
4186 * config/i386/i386.md (cpu): Remove the unused btver1.
4188 2014-01-09 H.J. Lu <hongjiu.lu@intel.com>
4190 * gdbasan.in: Put a breakpoint on __sanitizer::Report.
4192 2014-01-09 Jakub Jelinek <jakub@redhat.com>
4195 * tree-core.h (struct target_globals): New forward declaration.
4196 (struct tree_target_option): Add globals field.
4197 * tree.h (TREE_TARGET_GLOBALS): Define.
4198 (prepare_target_option_nodes_for_pch): New prototype.
4199 * target-globals.h (struct target_globals): Define even if
4201 * tree.c (prepare_target_option_node_for_pch,
4202 prepare_target_option_nodes_for_pch): New functions.
4203 * config/i386/i386.h (SWITCHABLE_TARGET): Define.
4204 * config/i386/i386.c: Include target-globals.h.
4205 (ix86_set_current_function): Instead of doing target_reinit
4206 unconditionally, use save_target_globals_default_opts and
4207 restore_target_globals.
4209 2014-01-09 Richard Biener <rguenther@suse.de>
4211 PR tree-optimization/59715
4212 * tree-cfg.h (split_critical_edges): Declare.
4213 * tree-cfg.c (split_critical_edges): Export.
4214 * tree-ssa-sink.c (execute_sink_code): Split critical edges.
4216 2014-01-09 Max Ostapenko <m.ostapenko@partner.samsung.com>
4218 * cfgexpand.c (expand_stack_vars): Optionally disable
4219 asan stack protection.
4220 (expand_used_vars): Likewise.
4221 (partition_stack_vars): Likewise.
4222 * asan.c (asan_emit_stack_protection): Optionally disable
4223 after return stack usage.
4224 (instrument_derefs): Optionally disable memory access instrumentation.
4225 (instrument_builtin_call): Likewise.
4226 (instrument_strlen_call): Likewise.
4227 (asan_protect_global): Optionally disable global variables protection.
4228 * doc/invoke.texi: Added doc for new options.
4229 * params.def: Added new options.
4230 * params.h: Likewise.
4232 2014-01-09 Jakub Jelinek <jakub@redhat.com>
4234 PR rtl-optimization/59724
4235 * ifcvt.c (cond_exec_process_if_block): Don't call
4236 flow_find_head_matching_sequence with 0 longest_match.
4237 * cfgcleanup.c (flow_find_head_matching_sequence): Count even
4238 non-active insns if !stop_after.
4239 (try_head_merge_bb): Revert 2014-01-07 changes.
4241 2014-01-08 Jeff Law <law@redhat.com>
4243 * ree.c (get_sub_rtx): New function, extracted from...
4244 (merge_def_and_ext): Here.
4245 (combine_reaching_defs): Use get_sub_rtx.
4247 2014-01-08 Eric Botcazou <ebotcazou@adacore.com>
4249 * cgraph.h (varpool_variable_node): Do not choke on null node.
4251 2014-01-08 Catherine Moore <clm@codesourcery.com>
4253 * config/mips/mips.md (simple_return): Attempt to use JRC
4255 * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
4257 2014-01-08 Richard Sandiford <rdsandiford@googlemail.com>
4259 PR rtl-optimization/59137
4260 * reorg.c (steal_delay_list_from_target): Call update_block for
4262 (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
4264 2014-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4266 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4267 two duplicate entries.
4269 2014-01-08 Richard Sandiford <rdsandiford@googlemail.com>
4272 2012-10-07 Richard Sandiford <rdsandiford@googlemail.com>
4274 * config/mips/mips.c (mips_truncated_op_cost): New function.
4275 (mips_rtx_costs): Adjust test for BADDU.
4276 * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
4278 2012-10-02 Richard Sandiford <rdsandiford@googlemail.com>
4280 * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
4281 (*baddu_si): ...this new pattern.
4283 2014-01-08 Jakub Jelinek <jakub@redhat.com>
4286 * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
4288 2014-01-08 Bernd Edlinger <bernd.edlinger@hotmail.de>
4291 * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
4293 (expand_expr, expand_normal): Adjust.
4294 * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
4295 inner_reference_p. Use inner_reference_p to expand inner references.
4296 (store_expr): Adjust.
4297 * cfgexpand.c (expand_call_stmt): Adjust.
4299 2014-01-08 Rong Xu <xur@google.com>
4301 * gcov-io.c (gcov_var): Move from gcov-io.h.
4302 (gcov_position): Ditto.
4303 (gcov_is_error): Ditto.
4304 (gcov_rewrite): Ditto.
4305 * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
4306 only part to libgcc/libgcov.h.
4308 2014-01-08 Marek Polacek <polacek@redhat.com>
4311 * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
4313 2014-01-08 Marek Polacek <polacek@redhat.com>
4316 * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
4318 2014-01-08 Jakub Jelinek <jakub@redhat.com>
4320 PR rtl-optimization/59649
4321 * stor-layout.c (get_mode_bounds): For BImode return
4322 0 and STORE_FLAG_VALUE.
4324 2014-01-08 Richard Biener <rguenther@suse.de>
4327 * gimple.h (is_gimple_builtin_call): Remove.
4328 (gimple_builtin_call_types_compatible_p): New.
4329 (gimple_call_builtin_p): New overload.
4330 * gimple.c (is_gimple_builtin_call): Remove.
4331 (validate_call): Rename to ...
4332 (gimple_builtin_call_types_compatible_p): ... this and export. Also
4334 (validate_type): New static function.
4335 (gimple_call_builtin_p): New overload and adjust.
4336 * gimple-fold.c (gimple_fold_builtin): Fold the return value.
4337 (gimple_fold_call): Likewise. Use gimple_call_builtin_p.
4338 (gimple_fold_stmt_to_constant_1): Likewise.
4339 * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
4341 2014-01-08 Richard Biener <rguenther@suse.de>
4344 * gimplify.c (gimplify_expr): Gimplify register-register type
4345 VIEW_CONVERT_EXPRs to separate stmts.
4347 2014-01-07 Jeff Law <law@redhat.com>
4350 * ree.c (combine_set_extension): Handle case where source
4351 and destination registers in an extension insn are different.
4352 (combine_reaching_defs): Allow source and destination registers
4353 in extension to be different under limited circumstances.
4354 (add_removable_extension): Remove restriction that the
4355 source and destination registers in the extension are the same.
4356 (find_and_remove_re): Emit a copy from the extension's
4357 destination to its source after the defining insn if
4358 the source and destination registers are different.
4361 * ifcvt.c (merge_if_block): If we are merging a block with more than
4362 one successor with a block with no successors, remove any BARRIER
4363 after the second block.
4365 2014-01-07 Dan Xio Qiang <ziyan01@163.com>
4367 * hw-doloop.c (reorg_loops): Release the bitmap obstack.
4369 2014-01-07 John David Anglin <danglin@gcc.gnu.org>
4372 * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
4373 for 14-bit register offsets when INT14_OK_STRICT is false.
4375 2014-01-07 Roland Stigge <stigge@antcom.de>
4376 Michael Meissner <meissner@linux.vnet.ibm.com>
4379 * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
4380 Only check TFmode for SPE constants. Don't check TImode or TDmode.
4382 2014-01-07 James Greenhalgh <james.greenhalgh@arm.com>
4384 * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
4387 2014-01-07 Yufeng Zhang <yufeng.zhang@arm.com>
4389 * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
4390 with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
4391 rtx is const0_rtx or not.
4393 2014-01-07 Richard Sandiford <rdsandiford@googlemail.com>
4396 * target-globals.c (save_target_globals): Remove this_fn_optab
4398 * toplev.c: Include optabs.h.
4399 (target_reinit): Temporarily restore the global options if another
4400 set of options are in force.
4402 2014-01-07 Jakub Jelinek <jakub@redhat.com>
4404 PR rtl-optimization/58668
4405 * cfgcleanup.c (flow_find_cross_jump): Don't count
4406 any jumps if dir_p is NULL. Remove p1 variable, use active_insn_p
4407 to determine what is counted.
4408 (flow_find_head_matching_sequence): Use active_insn_p to determine
4410 (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
4412 * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
4413 determine what is counted.
4415 PR tree-optimization/59643
4416 * tree-predcom.c (split_data_refs_to_components): If one dr is
4417 read and one write, determine_offset fails and the write isn't
4418 in the bad component, just put the read into the bad component.
4420 2014-01-07 Mike Stump <mikestump@comcast.net>
4421 Jakub Jelinek <jakub@redhat.com>
4424 * tree-core.h (struct tree_optimization_option): Change optabs
4425 type from unsigned char * to void *.
4426 * optabs.c (init_tree_optimization_optabs): Adjust
4427 TREE_OPTIMIZATION_OPTABS initialization.
4429 2014-01-06 Jakub Jelinek <jakub@redhat.com>
4432 * config/i386/i386.h (struct machine_function): Add
4433 no_drap_save_restore field.
4434 * config/i386/i386.c (ix86_save_reg): Use
4435 !cfun->machine->no_drap_save_restore instead of
4436 crtl->stack_realign_needed.
4437 (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
4438 this function clears frame_pointer_needed. Set
4439 cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
4440 and DRAP reg is needed.
4442 2014-01-06 Marek Polacek <polacek@redhat.com>
4445 * doc/implement-c.texi: Mention that other integer types are
4446 permitted as bit-field types in strictly conforming mode.
4448 2014-01-06 Felix Yang <fei.yang0953@gmail.com>
4450 * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
4453 2014-01-06 Richard Earnshaw <rearnsha@arm.com>
4455 * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
4457 2014-01-06 Martin Jambor <mjambor@suse.cz>
4460 * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
4462 * ipa-prop.c (ipa_print_node_params): Fix indentation.
4464 2014-01-06 Eric Botcazou <ebotcazou@adacore.com>
4468 * var-tracking.c (add_stores): Preserve the value of the source even if
4469 we don't record the store.
4471 2014-01-06 Terry Guo <terry.guo@arm.com>
4473 * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
4475 2014-01-05 Iain Sandoe <iain@codesourcery.com>
4478 * config/darwin.c (darwin_function_section): Adjust return values to
4479 correspond to optimisation changes made in r206070.
4481 2014-01-05 Uros Bizjak <ubizjak@gmail.com>
4483 * config/i386/i386.c (ix86_data_alignment): Calculate max_align
4484 from prefetch_block tune setting.
4485 (nocona_cost): Correct size of prefetch block to 64.
4487 2014-01-04 Eric Botcazou <ebotcazou@adacore.com>
4489 * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
4490 (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
4491 used to save the static chain register in the computation of the offset
4492 from which the FP registers need to be restored.
4494 2014-01-04 Jakub Jelinek <jakub@redhat.com>
4496 PR tree-optimization/59519
4497 * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
4498 ICE if get_current_def (current_new_name) is already non-NULL, as long
4499 as it is a phi result of some other phi in *new_exit_bb that has
4502 * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
4503 or vmovdqu* for misaligned_operand.
4504 (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
4505 <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
4506 * config/i386/i386.c (ix86_expand_special_args_builtin): Set
4507 aligned_mem for AVX512F masked aligned load and store builtins and for
4510 2014-01-03 Bingfeng Mei <bmei@broadcom.com>
4512 PR tree-optimization/59651
4513 * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
4514 Address range for negative step should be added by TYPE_SIZE_UNIT.
4516 2014-01-03 Andreas Schwab <schwab@linux-m68k.org>
4518 * config/m68k/m68k.c (handle_move_double): Handle pushes with
4519 overlapping registers also for registers other than the stack pointer.
4521 2014-01-03 Marek Polacek <polacek@redhat.com>
4524 * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
4527 2014-01-03 Jakub Jelinek <jakub@redhat.com>
4530 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
4533 * config/i386/i386.md (MODE_SIZE): New mode attribute.
4534 (push splitter): Use <P:MODE_SIZE> instead of
4535 GET_MODE_SIZE (<P:MODE>mode).
4536 (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
4537 (mov -1, reg peephole2): Likewise.
4538 * config/i386/sse.md (*mov<mode>_internal,
4539 <sse>_storeu<ssemodesuffix><avxsizesuffix>,
4540 <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
4541 *<code><mode>3, *andnot<mode>3<mask_name>,
4542 <mask_codefor><code><mode>3<mask_name>): Likewise.
4543 * config/i386/subst.md (mask_mode512bit_condition,
4544 sd_mask_mode512bit_condition): Likewise.
4546 2014-01-02 Xinliang David Li <davidxl@google.com>
4548 PR tree-optimization/59303
4549 * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
4550 (dump_predicates): Better output format.
4551 (pred_equal_p): New function.
4552 (is_neq_relop_p): Ditto.
4553 (is_neq_zero_form_p): Ditto.
4554 (pred_expr_equal_p): Ditto.
4555 (pred_neg_p): Ditto.
4556 (simplify_pred): Ditto.
4557 (simplify_preds_2): Ditto.
4558 (simplify_preds_3): Ditto.
4559 (simplify_preds_4): Ditto.
4560 (simplify_preds): Ditto.
4562 (push_to_worklist): Ditto.
4563 (get_pred_info_from_cmp): Ditto.
4564 (is_degenerated_phi): Ditto.
4565 (normalize_one_pred_1): Ditto.
4566 (normalize_one_pred): Ditto.
4567 (normalize_one_pred_chain): Ditto.
4568 (normalize_preds): Ditto.
4569 (normalize_cond_1): Remove function.
4570 (normalize_cond): Ditto.
4571 (is_gcond_subset_of): Ditto.
4572 (is_subset_of_any): Ditto.
4573 (is_or_set_subset_of): Ditto.
4574 (is_and_set_subset_of): Ditto.
4575 (is_norm_cond_subset_of): Ditto.
4576 (pred_chain_length_cmp): Ditto.
4577 (convert_control_dep_chain_into_preds): Type change.
4578 (find_predicates): Ditto.
4579 (find_def_preds): Ditto.
4580 (destroy_predicates_vecs): Ditto.
4581 (find_matching_predicates_in_rest_chains): Ditto.
4582 (use_pred_not_overlap_with_undef_path_pred): Ditto.
4583 (is_pred_expr_subset): Ditto.
4584 (is_pred_chain_subset_of): Ditto.
4585 (is_included_in): Ditto.
4586 (is_superset_of): Ditto.
4588 2014-01-02 Richard Sandiford <rdsandiford@googlemail.com>
4590 Update copyright years.
4592 2014-01-02 Richard Sandiford <rdsandiford@googlemail.com>
4594 * common/config/arc/arc-common.c, config/arc/arc-modes.def,
4595 config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
4596 config/arc/arc.md, config/arc/arc.opt,
4597 config/arm/arm_neon_builtins.def, config/arm/crypto.def,
4598 config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
4599 config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
4600 config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
4601 config/linux-protos.h, config/linux.c, config/winnt-c.c,
4602 diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
4603 vtable-verify.c, vtable-verify.h: Use the standard form for the
4606 2014-01-02 Tobias Burnus <burnus@net-b.de>
4608 * gcc.c (process_command): Update copyright notice dates.
4609 * gcov-dump.c: Ditto.
4611 * doc/cpp.texi: Bump @copying's copyright year.
4612 * doc/cppinternals.texi: Ditto.
4613 * doc/gcc.texi: Ditto.
4614 * doc/gccint.texi: Ditto.
4615 * doc/gcov.texi: Ditto.
4616 * doc/install.texi: Ditto.
4617 * doc/invoke.texi: Ditto.
4619 2014-01-01 Jan-Benedict Glaw <jbglaw@lug-owl.de>
4621 * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
4623 2014-01-01 Jakub Jelinek <jakub@redhat.com>
4625 * config/i386/sse.md (*mov<mode>_internal): Guard
4626 EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
4628 PR rtl-optimization/59647
4629 * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
4630 new_rtx into UNSIGNED_FLOAT rtxes.
4632 Copyright (C) 2014 Free Software Foundation, Inc.
4634 Copying and distribution of this file, with or without modification,
4635 are permitted in any medium without royalty provided the copyright
4636 notice and this notice are preserved.