Just enumerate all GF_OMP_FOR_KIND_* and GF_OMP_TARGET_KIND_*.
[official-gcc.git] / gcc / ChangeLog
blobb71499da442ef1ccea3a6d600345de8f464e1b7f
1 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
3         PR plugins/59335
4         * Makefile.in (PLUGIN_HEADERS): Add tree-phinodes.h, stor-layout.h,
5         ssa-iterators.h, $(RESOURCE_H) and tree-cfgcleanup.h.
7         PR plugins/59335
8         * config/i386/t-i386 (OPTIONS_H_EXTRA): Add stringop.def.
9         (TM_H): Add x86-tune.def.
11 2014-03-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13         * config/aarch64/aarch64.c (generic_tunings):
14         Use cortexa57_extra_costs.
16 2014-03-05  Jakub Jelinek  <jakub@redhat.com>
18         PR lto/60404
19         * cfgexpand.c (expand_used_vars): Do not assume all SSA_NAMEs
20         of PARM/RESULT_DECLs must be coalesced with optimize && in_lto_p.
21         * tree-ssa-coalesce.c (coalesce_ssa_name): Use MUST_COALESCE_COST - 1
22         cost for in_lto_p.
24 2014-03-04  Heiher  <r@hev.cc>
26         * config/mips/mips-cpus.def (loongson3a): Mark as a MIPS64r2 processor.
27         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Adjust accordingly.
29 2014-03-04  Uros Bizjak  <ubizjak@gmail.com>
31         * config/i386/predicates.md (const2356_operand): Change to ...
32         (const2367_operand): ... this.
33         * config/i386/sse.md (avx512pf_scatterpf<mode>sf): Use
34         const2367_operand.
35         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
36         (*avx512pf_scatterpf<mode>sf): Ditto.
37         (avx512pf_scatterpf<mode>df): Ditto.
38         (*avx512pf_scatterpf<mode>df_mask): Ditto.
39         (*avx512pf_scatterpf<mode>df): Ditto.
40         * config/i386/i386.c (ix86_expand_builtin): Update
41         incorrect hint operand error message.
43 2014-03-04  Richard Biener  <rguenther@suse.de>
45         * lto-section-in.c (lto_get_section_data): Fix const cast.
47 2014-03-04  Paulo Matos  <paulo@matos-sorge.com>
49         * tree-streamer.c (record_common_node): Assert we don't record
50         nodes with type double.
51         (preload_common_node): Skip type double, complex double and
52         double pointer since it is now frontend dependent due to
53         fshort-double option.
55 2014-03-04  Richard Biener  <rguenther@suse.de>
57         PR lto/60405
58         * lto-streamer-in.c (lto_read_body): Remove LTO bytecode version check.
59         (lto_input_toplevel_asms): Likewise.
60         * lto-section-in.c (lto_get_section_data): Instead do it here
61         for every section.
63 2014-03-04  Richard Biener  <rguenther@suse.de>
65         PR tree-optimization/60382
66         * tree-vect-loop.c (vect_is_simple_reduction_1): Do not consider
67         dead PHIs a reduction.
69 2014-03-03  Uros Bizjak  <ubizjak@gmail.com>
71         * config/i386/xmmintrin.h (enum _mm_hint) <_MM_HINT_ET0>: Correct
72         hint value.
73         (_mm_prefetch): Move out of GCC target("sse") pragma.
74         * config/i386/prfchwintrin.h (_m_prefetchw): Move out of
75         GCC target("prfchw") pragma.
76         * config/i386/i386.md (prefetch): Emit prefetchwt1 only
77         for locality <= 2.
78         * config/i386/i386.c (ix86_option_override_internal): Enable
79         -mprfchw with -mprefetchwt1.
81 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
83         * config/arc/arc.md (casesi_load) <length attribute alternative 0>:
84         Mark as varying.
86 2014-03-03  Joern Rennecke  <joern.rennecke@embecosm.com>
88         * opts.h (CL_PCH_IGNORE): Define.
89         * targhooks.c (option_affects_pch_p):
90         Return false for options that have CL_PCH_IGNORE set.
91         * opt-functions.awk: Process PchIgnore.
92         * doc/options.texi: Document PchIgnore.
94         * config/arc/arc.opt (misize): Add PchIgnore property.
96 2014-03-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
98         * config/rs6000/rs6000.c (rs6000_preferred_reload_class): Disallow
99         reload of PLUS rtx's outside of GENERAL_REGS or BASE_REGS; relax
100         constraint on constants to permit them being loaded into
101         GENERAL_REGS or BASE_REGS.
103 2014-03-03  Nick Clifton  <nickc@redhat.com>
105         * config/rl78/rl78-real.md (cbranchsi4_real_signed): Add
106         anti-cacnonical alternatives.
107         (negandhi3_real): New pattern.
108         * config/rl78/rl78-virt.md (negandhi3_virt): New pattern.
110 2014-03-03  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
112         * config/avr/avr-mcus.def: Remove atxmega16x1.
113         * config/avr/avr-tables.opt: Regenerate.
114         * config/avr/t-multilib: Regenerate.
115         * doc/avr-mmcu.texi: Regenerate.
117 2014-03-03  Tobias Grosser  <tobias@grosser.es>
118             Mircea Namolaru  <mircea.namolaru@inria.fr>
120         PR tree-optimization/58028
121         * graphite-clast-to-gimple.c (set_cloog_options): Don't remove
122         scalar dimensions.
124 2014-03-03  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
126         * config/arm/neon.md (*movmisalign<mode>): Legitimize addresses
127         not handled by recognizers.
129 2014-03-03  Jakub Jelinek  <jakub@redhat.com>
131         PR middle-end/60175
132         * function.c (expand_function_end): Don't emit
133         clobber_return_register sequence if clobber_after is a BARRIER.
134         * cfgexpand.c (construct_exit_block): Append instructions before
135         return_label to prev_bb.
137 2014-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
139         * config/rs6000/constraints.md: Document reserved use of "wc".
141 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
143         PR ipa/60150
144         * ipa.c (function_and_variable_visibility): When dissolving comdat
145         group, also set all symbols to local.
147 2014-03-02  Jan Hubicka  <hubicka@ucw.cz>
149         PR ipa/60306
151         Revert:
152         2013-12-14   Jan Hubicka  <jh@suse.cz>
153         PR middle-end/58477
154         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Skip clobbers.
156 2014-03-02  Jon Beniston  <jon@beniston.com>
158         PR bootstrap/48230
159         PR bootstrap/50927
160         PR bootstrap/52466
161         PR target/46898
162         * config/lm32/lm32.c (lm32_legitimate_constant_p): Remove, as incorrect.
163           (TARGET_LEGITIMATE_CONSTANT_P): Undefine, as not needed.
164         * config/lm32/lm32.md (movsi_insn): Add 32-bit immediate support.
165         (simple_return, *simple_return): New patterns
166         * config/lm32/predicates.md (movsi_rhs_operand): Remove as obsolete.
167         * configure.ac (force_sjlj_exceptions): Force sjlj exceptions for lm32.
169 2014-03-01  Paolo Carlini  <paolo.carlini@oracle.com>
171         * dwarf2out.c (gen_subprogram_die): Tidy.
173 2014-03-01  Oleg Endo  <olegendo@gcc.gnu.org>
175         PR target/60071
176         * config/sh/sh.md (*mov_t_msb_neg): Split into ...
177         (*mov_t_msb_neg_negc): ... this new insn.
179 2014-02-28  Jason Merrill  <jason@redhat.com>
181         PR c++/58678
182         * ipa-devirt.c (ipa_devirt): Don't choose an implicitly-declared
183         function.
185 2014-02-28  Paolo Carlini  <paolo.carlini@oracle.com>
187         PR c++/60314
188         * dwarf2out.c (decltype_auto_die): New static.
189         (gen_subprogram_die): Handle 'decltype(auto)' like 'auto'.
190         (gen_type_die_with_usage): Handle 'decltype(auto)'.
191         (is_cxx_auto): Likewise.
193 2014-02-28  Ian Bolton  <ian.bolton@arm.com>
195         * config/aarch64/aarch64.h: Define __ARM_NEON by default if
196         we are not using general regs only.
198 2014-02-28  Richard Biener  <rguenther@suse.de>
200         PR target/60280
201         * tree-cfgcleanup.c (tree_forwarder_block_p): Restrict
202         previous fix and only allow to remove trivial pre-headers
203         and latches.  Also honor LOOPS_MAY_HAVE_MULTIPLE_LATCHES.
204         (remove_forwarder_block): Properly update the latch of a loop.
206 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
208         PR debug/59992
209         * cselib.c (cselib_hasher::equal): Special-case VALUE lookup.
210         (cselib_preserved_hash_table): New.
211         (preserve_constants_and_equivs): Move preserved vals to it.
212         (cselib_find_slot): Look it up first.
213         (cselib_init): Initialize it.
214         (cselib_finish): Release it.
215         (dump_cselib_table): Dump it.
217 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
219         PR debug/59992
220         * cselib.c (remove_useless_values): Skip to avoid quadratic
221         behavior if the condition moved from...
222         (cselib_process_insn): ... here holds.
224 2014-02-28  Alexandre Oliva <aoliva@redhat.com>
226         PR debug/57232
227         * var-tracking.c (vt_initialize): Apply the same condition to
228         preserve the CFA base value.
230 2014-02-28  Joey Ye  <joey.ye@arm.com>
232         PR target/PR60169
233         * config/arm/arm.c (thumb_far_jump_used_p): Don't change
234         if reload in progress or completed.
236 2014-02-28  Tobias Burnus  <burnus@net-b.de>
238         PR middle-end/60147
239         * tree-pretty-print.c (dump_generic_node, print_declaration): Handle
240         NAMELIST_DECL.
242 2014-02-27  H.J. Lu  <hongjiu.lu@intel.com>
244         * doc/tm.texi.in (Condition Code Status): Update documention for
245         relative locations of cc0-setter and cc0-user.
247 2014-02-27  Jeff Law  <law@redhat.com>
249         PR rtl-optimization/52714
250         * combine.c (try_combine): When splitting an unrecognized PARALLEL
251         into two independent simple sets, if I3 is a jump, ensure the
252         pattern we place into I3 is a (set (pc) ...).
254 2014-02-27  Mikael Pettersson  <mikpe@it.uu.se>
255             Jeff Law  <law@redhat.com>
257         PR rtl-optimization/49847
258         * cse.c (fold_rtx) Handle case where cc0 setter and cc0 user
259         are in different blocks.
260         * doc/tm.texi (Condition Code Status): Update documention for
261         relative locations of cc0-setter and cc0-user.
263 2014-02-27  Vladimir Makarov  <vmakarov@redhat.com>
265         PR target/59222
266         * lra.c (lra_emit_add): Check SUBREG too.
268 2014-02-27  Andreas Schwab  <schwab@suse.de>
270         * config/m68k/m68k.c (m68k_option_override): Disable
271         -flive-range-shrinkage for classic m68k.
272         (m68k_override_options_after_change): Likewise.
274 2014-02-27  Marek Polacek  <polacek@redhat.com>
276         PR middle-end/59223
277         * tree-ssa-uninit.c (gate_warn_uninitialized): Run the pass even for
278         -Wmaybe-uninitialized.
280 2014-02-27  Alan Modra  <amodra@gmail.com>
282         PR target/57936
283         * reload1.c (emit_input_reload_insns): When reload_override_in,
284         set old to rl->in_reg when rl->in_reg is a subreg.
286 2014-02-26  Richard Biener  <rguenther@suse.de>
288         PR bootstrap/60343
289         * lra-assigns.c (spill_for): Avoid mixed-sign comparison.
291 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
293         * common/config/i386/predicates.md (const1256_operand): Remove.
294         (const2356_operand): New.
295         (const_1_to_2_operand): Remove.
296         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
297         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
298         (*avx512pf_gatherpf<mode>sf): Ditto.
299         (avx512pf_gatherpf<mode>df): Ditto.
300         (*avx512pf_gatherpf<mode>df_mask): Ditto.
301         (*avx512pf_gatherpf<mode>df): Ditto.
302         (avx512pf_scatterpf<mode>sf): Ditto.
303         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
304         (*avx512pf_scatterpf<mode>sf): Ditto.
305         (avx512pf_scatterpf<mode>df): Ditto.
306         (*avx512pf_scatterpf<mode>df_mask): Ditto.
307         (*avx512pf_scatterpf<mode>df): Ditto.
308         * common/config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET0.
310 2014-02-26  Ilya Tocar  <ilya.tocar@intel.com>
312         * config/i386/avx512fintrin.h (_mm512_testn_epi32_mask),
313         (_mm512_mask_testn_epi32_mask), (_mm512_testn_epi64_mask),
314         (_mm512_mask_testn_epi64_mask): Move to ...
315         * config/i386/avx512cdintrin.h: Here.
316         * config/i386/i386.c (bdesc_args): Change MASK_ISA for testnm.
317         * config/i386/sse.md (avx512f_vmscalef<mode><round_name>): Remove %.
318         (avx512f_scalef<mode><mask_name><round_name>): Ditto.
319         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Change conditon to
320         TARGET_AVX512F from TARGET_AVX512CD.
322 2014-02-26  Richard Biener  <rguenther@suse.de>
324         PR ipa/60327
325         * ipa.c (walk_polymorphic_call_targets): Properly guard
326         call to inline_update_overall_summary.
328 2014-02-26  Bin Cheng  <bin.cheng@arm.com>
330         PR target/60280
331         * tree-cfgcleanup.c (tree_forwarder_block_p): Protect loop preheaders
332         and latches only if requested.  Fix latch if it is removed.
333         * tree-ssa-dom.c (tree_ssa_dominator_optimize): Set
334         LOOPS_HAVE_PREHEADERS.
336 2014-02-25  Andrew Pinski  <apinski@cavium.com>
338         * builtins.c (expand_builtin_thread_pointer): Create a new target
339         when the target is NULL.
341 2014-02-25  Vladimir Makarov  <vmakarov@redhat.com>
343         PR rtl-optimization/60317
344         * params.def (PARAM_LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
345         * params.h (LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
346         * lra-assigns.c: Include params.h.
347         (spill_for): Use LRA_MAX_CONSIDERED_RELOAD_PSEUDOS as guard for
348         other reload pseudos considerations.
350 2014-02-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
352         * config/rs6000/vector.md (*vector_unordered<mode>): Change split
353         to use canonical form for nor<mode>3.
355 2014-02-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
357         PR target/55426
358         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Allow 128 to 64-bit
359         conversions.
361 2014-02-25  Ilya Tocar  <ilya.tocar@intel.com>
363         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PREFETCHWT1_SET),
364         (OPTION_MASK_ISA_PREFETCHWT1_UNSET): New.
365         (ix86_handle_option): Handle OPT_mprefetchwt1.
366         * config/i386/cpuid.h (bit_PREFETCHWT1): New.
367         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
368         PREFETCHWT1 CPUID.
369         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
370         OPTION_MASK_ISA_PREFETCHWT1.
371         * config/i386/i386.c (ix86_target_string): Handle mprefetchwt1.
372         (PTA_PREFETCHWT1): New.
373         (ix86_option_override_internal): Handle PTA_PREFETCHWT1.
374         (ix86_valid_target_attribute_inner_p): Handle OPT_mprefetchwt1.
375         * config/i386/i386.h (TARGET_PREFETCHWT1, TARGET_PREFETCHWT1_P): New.
376         * config/i386/i386.md (prefetch): Check TARGET_PREFETCHWT1
377         (*prefetch_avx512pf_<mode>_: Change into ...
378         (*prefetch_prefetchwt1_<mode>: This.
379         * config/i386/i386.opt (mprefetchwt1): New.
380         * config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET1.
381         (_mm_prefetch): Handle intent to write.
382         * doc/invoke.texi (mprefetchwt1), (mno-prefetchwt1): Doccument.
384 2014-02-25  Richard Biener  <rguenther@suse.de>
386         PR middle-end/60291
387         * emit-rtl.c (mem_attrs_htab): Remove.
388         (mem_attrs_htab_hash): Likewise.
389         (mem_attrs_htab_eq): Likewise.
390         (set_mem_attrs): Always allocate new mem-attrs when something changed.
391         (init_emit_once): Do not allocate mem_attrs_htab.
393 2014-02-25  Richard Biener  <rguenther@suse.de>
395         PR lto/60319
396         * lto-opts.c (lto_write_options): Output non-explicit conservative
397         -fwrapv, -fno-trapv and -fno-strict-overflow.
398         * lto-wrapper.c (merge_and_complain): Handle merging those options.
399         (run_gcc): And pass them through.
401 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
403         * sel-sched.c (calculate_new_fences): New parameter ptime.
404         Calculate it as a maximum over all fence cycles.
405         (sel_sched_region_2): Adjust the call to calculate_new_fences.
406         Print the final schedule timing when sched_verbose.
408 2014-02-25  Andrey Belevantsev  <abel@ispras.ru>
410         PR rtl-optimization/60292
411         * sel-sched.c (fill_vec_av_set): Do not reset target availability
412         bit fot the fence instruction.
414 2014-02-24  Alangi Derick  <alangiderick@gmail.com>
416         * calls.h: Fix typo in comment.
418 2014-02-24  John David Anglin  <danglin@gcc.gnu.org>
420         * config/pa/pa.c (pa_output_move_double): Don't valididate when
421         adjusting offsetable addresses.
423 2014-02-24  Guozhi Wei  <carrot@google.com>
425         * sparseset.h (sparseset_pop): Fix the wrong index.
427 2014-02-24  Walter Lee  <walt@tilera.com>
429         * config.gcc (tilepro-*-*): Change to tilepro*-*-*.
430         (tilegx-*-linux*): Change to tilegx*-*-linux*; Support tilegxbe
431         triplet.
432         * common/config/tilegx/tilegx-common.c
433         (TARGET_DEFAULT_TARGET_FLAGS): Define.
434         * config/tilegx/linux.h (ASM_SPEC): Add endian_spec.
435         (LINK_SPEC): Ditto.
436         * config/tilegx/sync.md (atomic_test_and_set): Handle big endian.
437         * config/tilegx/tilegx.c (tilegx_return_in_msb): New.
438         (tilegx_gimplify_va_arg_expr): Handle big endian.
439         (tilegx_expand_unaligned_load): Ditto.
440         (tilegx_expand_unaligned_store): Ditto.
441         (TARGET_RETURN_IN_MSB): New.
442         * config/tilegx/tilegx.h (TARGET_DEFAULT): New.
443         (TARGET_ENDIAN_DEFAULT): New.
444         (TARGET_BIG_ENDIAN): Handle big endian.
445         (BYTES_BIG_ENDIAN): Ditto.
446         (WORDS_BIG_ENDIAN): Ditto.
447         (FLOAT_WORDS_BIG_ENDIAN): Ditto.
448         (ENDIAN_SPEC): New.
449         (EXTRA_SPECS): New.
450         * config/tilegx/tilegx.md (extv): Handle big endian.
451         (extzv): Ditto.
452         (insn_st<n>): Ditto.
453         (insn_st<n>_add<bitsuffix>): Ditto.
454         (insn_stnt<n>): Ditto.
455         (insn_stnt<n>_add<bitsuffix>):Ditto.
456         (vec_interleave_highv8qi): Handle big endian.
457         (vec_interleave_highv8qi_be): New.
458         (vec_interleave_highv8qi_le): New.
459         (insn_v1int_h): Handle big endian.
460         (vec_interleave_lowv8qi): Handle big endian.
461         (vec_interleave_lowv8qi_be): New.
462         (vec_interleave_lowv8qi_le): New.
463         (insn_v1int_l): Handle big endian.
464         (vec_interleave_highv4hi): Handle big endian.
465         (vec_interleave_highv4hi_be): New.
466         (vec_interleave_highv4hi_le): New.
467         (insn_v2int_h): Handle big endian.
468         (vec_interleave_lowv4hi): Handle big endian.
469         (vec_interleave_lowv4hi_be): New.
470         (vec_interleave_lowv4hi_le): New.
471         (insn_v2int_l): Handle big endian.
472         (vec_interleave_highv2si): Handle big endian.
473         (vec_interleave_highv2si_be): New.
474         (vec_interleave_highv2si_le): New.
475         (insn_v4int_h): Handle big endian.
476         (vec_interleave_lowv2si): Handle big endian.
477         (vec_interleave_lowv2si_be): New.
478         (vec_interleave_lowv2si_le): New.
479         (insn_v4int_l): Handle big endian.
480         * config/tilegx/tilegx.opt (mbig-endian): New option.
481         (mlittle-endian): New option.
482         * doc/install.texi: Document tilegxbe-linux.
483         * doc/invoke.texi: Document -mbig-endian and -mlittle-endian.
485 2014-02-24  Martin Jambor  <mjambor@suse.cz>
487         PR ipa/60266
488         * ipa-cp.c (propagate_constants_accross_call): Bail out early if
489         there are no parameter descriptors.
491 2014-02-24  Andrey Belevantsev  <abel@ispras.ru>
493         PR rtl-optimization/60268
494         * sched-rgn.c (haifa_find_rgns): Move the nr_regions_initial variable
495         initialization to ...
496         (sched_rgn_init): ... here.
497         (schedule_region): Check for SCHED_PRESSURE_NONE earlier.
499 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
501         * config/microblaze/microblaze.md: Correct ashrsi_reg / lshrsi_reg
502         names.
504 2014-02-23  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
506         * config/microblaze/microblaze.h: Remove SECONDARY_MEMORY_NEEDED
507         definition.
509 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
511         * /config/microblaze/microblaze.c: Add microblaze_asm_output_mi_thunk,
512         define TARGET_ASM_OUTPUT_MI_THUNK and TARGET_ASM_CAN_OUTPUT_MI_THUNK.
514 2014-02-23  David Holsgrove <david.holsgrove@xilinx.com>
516         * config/microblaze/predicates.md: Add cmp_op predicate.
517         * config/microblaze/microblaze.md: Add branch_compare instruction
518         which uses cmp_op predicate and emits cmp insn before branch.
519         * config/microblaze/microblaze.c (microblaze_emit_compare): Rename
520         to microblaze_expand_conditional_branch and consolidate logic.
521         (microblaze_expand_conditional_branch): emit branch_compare
522         insn instead of handling cmp op separate from branch insn.
524 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
526         * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Relax assert
527         to permit subregs.
529 2014-02-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
531         * config/rs6000/altivec.md (altivec_lve<VI_char>x): Replace
532         define_insn with define_expand and new define_insn
533         *altivec_lve<VI_char>x_internal.
534         (altivec_stve<VI_char>x): Replace define_insn with define_expand
535         and new define_insn *altivec_stve<VI_char>x_internal.
536         * config/rs6000/rs6000-protos.h (altivec_expand_stvex_be): New
537         prototype.
538         * config/rs6000/rs6000.c (altivec_expand_lvx_be): Document use by
539         lve*x built-ins.
540         (altivec_expand_stvex_be): New function.
542 2014-02-22  Joern Rennecke  <joern.rennecke@embecosm.com>
544         * config/avr/avr.c (avr_can_eliminate): Allow elimination from
545         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM if !frame_pointer_needed.
546         * config/avr/avr.c (ELIMINABLE_REGS): Add elimination from
547         ARG_POINTER_REGNUM to STACK_POINTER_REGNUM.
549 2014-02-21  Vladimir Makarov  <vmakarov@redhat.com>
551         PR target/60298
552         * lra-constraints.c (inherit_reload_reg): Use lra_emit_move
553         instead of emit_move_insn.
555 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
557         * config/rs6000/altivec.md (altivec_vsumsws): Replace second
558         vspltw with vsldoi.
559         (reduc_uplus_v16qi): Use gen_altivec_vsumsws_direct instead of
560         gen_altivec_vsumsws.
562 2014-02-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
564         * config/rs6000/altivec.md (altivec_lvxl): Rename as
565         *altivec_lvxl_<mode>_internal and use VM2 iterator instead of V4SI.
566         (altivec_lvxl_<mode>): New define_expand incorporating
567         -maltivec=be semantics where needed.
568         (altivec_lvx): Rename as *altivec_lvx_<mode>_internal.
569         (altivec_lvx_<mode>): New define_expand incorporating -maltivec=be
570         semantics where needed.
571         (altivec_stvx): Rename as *altivec_stvx_<mode>_internal.
572         (altivec_stvx_<mode>): New define_expand incorporating
573         -maltivec=be semantics where needed.
574         (altivec_stvxl): Rename as *altivec_stvxl_<mode>_internal and use
575         VM2 iterator instead of V4SI.
576         (altivec_stvxl_<mode>): New define_expand incorporating
577         -maltivec=be semantics where needed.
578         * config/rs6000/rs6000-builtin.def: Add new built-in definitions
579         LVXL_V2DF, LVXL_V2DI, LVXL_V4SF, LVXL_V4SI, LVXL_V8HI, LVXL_V16QI,
580         LVX_V2DF, LVX_V2DI, LVX_V4SF, LVX_V4SI, LVX_V8HI, LVX_V16QI, STVX_V2DF,
581         STVX_V2DI, STVX_V4SF, STVX_V4SI, STVX_V8HI, STVX_V16QI, STVXL_V2DF,
582         STVXL_V2DI, STVXL_V4SF, STVXL_V4SI, STVXL_V8HI, STVXL_V16QI.
583         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Replace
584         ALTIVEC_BUILTIN_LVX with ALTIVEC_BUILTIN_LVX_<MODE> throughout;
585         similarly for ALTIVEC_BUILTIN_LVXL, ALTIVEC_BUILTIN_STVX, and
586         ALTIVEC_BUILTIN_STVXL.
587         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be): New prototype.
588         (altivec_expand_stvx_be): Likewise.
589         * config/rs6000/rs6000.c (swap_selector_for_mode): New function.
590         (altivec_expand_lvx_be): Likewise.
591         (altivec_expand_stvx_be): Likewise.
592         (altivec_expand_builtin): Add cases for
593         ALTIVEC_BUILTIN_STVX_<MODE>, ALTIVEC_BUILTIN_STVXL_<MODE>,
594         ALTIVEC_BUILTIN_LVXL_<MODE>, and ALTIVEC_BUILTIN_LVX_<MODE>.
595         (altivec_init_builtins): Add definitions for
596         __builtin_altivec_lvxl_<mode>, __builtin_altivec_lvx_<mode>,
597         __builtin_altivec_stvx_<mode>, and __builtin_altivec_stvxl_<mode>.
599 2014-02-21  Catherine Moore  <clm@codesourcery.com>
601         * doc/invoke.texi (mvirt, mno-virt): Document.
602         * config/mips/mips.opt (mvirt): New option.
603         * config/mips/mips.h (ASM_SPEC): Pass mvirt to the assembler.
605 2014-02-21  Richard Biener  <rguenther@suse.de>
607         PR tree-optimization/60276
608         * tree-vectorizer.h (struct _stmt_vec_info): Add min_neg_dist field.
609         (STMT_VINFO_MIN_NEG_DIST): New macro.
610         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Record
611         STMT_VINFO_MIN_NEG_DIST.
612         * tree-vect-stmts.c (vectorizable_load): Verify if assumptions
613         made for negative dependence distances still hold.
615 2014-02-21  Richard Biener  <rguenther@suse.de>
617         PR middle-end/60291
618         * tree-ssa-live.c (mark_all_vars_used_1): Do not walk
619         DECL_INITIAL for globals not in the current function context.
621 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
623         PR tree-optimization/56490
624         * params.def (PARAM_UNINIT_CONTROL_DEP_ATTEMPTS): New param.
625         * tree-ssa-uninit.c: Include params.h.
626         (compute_control_dep_chain): Add num_calls argument, return false
627         if it exceed PARAM_UNINIT_CONTROL_DEP_ATTEMPTS param, pass
628         num_calls to recursive call.
629         (find_predicates): Change dep_chain into normal array,
630         cur_chain into auto_vec<edge, MAX_CHAIN_LEN + 1>, add num_calls
631         variable and adjust compute_control_dep_chain caller.
632         (find_def_preds): Likewise.
634 2014-02-21  Thomas Schwinge  <thomas@codesourcery.com>
636         * gimple-pretty-print.c (dump_gimple_omp_for) [flags & TDF_RAW]
637         <case GF_OMP_FOR_KIND_CILKSIMD>: Add missing break statement.
639 2014-02-21  Nick Clifton  <nickc@redhat.com>
641         * config/stormy16/stormy16.md (pushdqi1): Add mode to post_inc.
642         (pushhi1): Likewise.
643         (popqi1): Add mode to pre_dec.
644         (pophi1): Likewise.
646 2014-02-21  Jakub Jelinek  <jakub@redhat.com>
648         * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
649         mode for mask of V8SFmode permutation.
651 2014-02-20  Richard Henderson  <rth@redhat.com>
653         PR c++/60272
654         * builtins.c (expand_builtin_atomic_compare_exchange): Always make
655         a new pseudo for OLDVAL.
657 2014-02-20  Jakub Jelinek  <jakub@redhat.com>
659         PR target/57896
660         * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
661         gen_reg_rtx if d->testing_p.
662         (expand_vec_perm_pshufb2, expand_vec_perm_broadcast_1): Return early
663         if d->testing_p and we will certainly return true.
664         (expand_vec_perm_even_odd_1): Likewise.  Don't call gen_reg_rtx
665         if d->testing_p.
667 2014-02-20  Uros Bizjak  <ubizjak@gmail.com>
669         * emit-rtl.c (gen_reg_rtx): Assert that
670         crtl->emit.regno_pointer_align_length is non-zero.
672 2014-02-20  Richard Henderson  <rth@redhat.com>
674         PR c++/60272
675         * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
676         on failure the store back into EXPECT.
678 2014-02-20  Chung-Lin Tang  <cltang@codesourcery.com>
679             Sandra Loosemore  <sandra@codesourcery.com>
681         * config/nios2/nios2.md (unspec): Add UNSPEC_PIC_GOTOFF_SYM enum.
682         * config/nios2/nios2.c (nios2_function_profiler): Add
683         -fPIC (flag_pic == 2) support.
684         (nios2_handle_custom_fpu_cfg): Fix warning parameter.
685         (nios2_large_offset_p): New function.
686         (nios2_unspec_reloc_p): Move up position, update to use
687         nios2_large_offset_p.
688         (nios2_unspec_address): Remove function.
689         (nios2_unspec_offset): New function.
690         (nios2_large_got_address): New function.
691         (nios2_got_address): Add large offset support.
692         (nios2_legitimize_tls_address): Update usage of removed and new
693         functions.
694         (nios2_symbol_binds_local_p): New function.
695         (nios2_load_pic_address): Add -fPIC (flag_pic == 2) support.
696         (nios2_legitimize_address): Update to use nios2_large_offset_p.
697         (nios2_emit_move_sequence): Avoid legitimizing (const (unspec ...)).
698         (nios2_print_operand): Merge H/L processing, add hiadj/lo
699         processing for (const (unspec ...)).
700         (nios2_unspec_reloc_name): Add UNSPEC_PIC_GOTOFF_SYM case.
702 2014-02-20  Richard Biener  <rguenther@suse.de>
704         * tree-cfg.c (replace_uses_by): Mark altered BBs before
705         doing the substitution.
706         (verify_gimple_assign_single): Also verify bare MEM_REFs on the lhs.
708 2014-02-20  Martin Jambor  <mjambor@suse.cz>
710         PR ipa/55260
711         * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
712         info when checking whether lattices are bottom.
714 2014-02-20  Richard Biener  <rguenther@suse.de>
716         PR middle-end/60221
717         * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
718         regions at -O0.
720 2014-02-20  Jan Hubicka  <hubicka@ucw.cz>
722         PR ipa/58555
723         * ipa-inline-transform.c (clone_inlined_nodes): Add freq_scale
724         parameter specifying the scaling.
725         (inline_call): Update.
726         (want_inline_recursively): Guard division by zero.
727         (recursive_inlining): Update.
728         * ipa-inline.h (clone_inlined_nodes): Update.
730 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
732         PR target/60204
733         * config/i386/i386.c (classify_argument): Pass structures of size
734         64 bytes or less in register.
736 2014-02-20  Ilya Tocar  <ilya.tocar@intel.com>
737             Kirill Yukhin  <kirill.yukhin@intel.com>
739         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): Swap operands.
740         (_mm_rcp28_round_ss): Ditto.
741         (_mm_rsqrt28_round_sd): Ditto.
742         (_mm_rsqrt28_round_ss): Ditto.
743         * config/i386/avx512erintrin.h (_mm_rcp14_round_sd): Ditto.
744         (_mm_rcp14_round_ss): Ditto.
745         (_mm_rsqrt14_round_sd): Ditto.
746         (_mm_rsqrt14_round_ss): Ditto.
747         * config/i386/sse.md (rsqrt14<mode>): Put nonimmediate operand as
748         the first input operand, get rid of match_dup.
749         (avx512er_exp2<mode><mask_name><round_saeonly_name>): Set type
750         attribute to sse.
751         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
752         Ditto.
753         (avx512er_vmrcp28<mode><round_saeonly_name>): Put nonimmediate
754         operand as the first input operand, set type attribute.
755         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
756         Set type attribute.
757         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Put nonimmediate
758         operand as the first input operand, set type attribute.
760 2014-02-19  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
762         * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
763         bit of zero.
765 2014-02-19  H.J. Lu  <hongjiu.lu@intel.com>
767         PR target/60207
768         * config/i386/i386.c (construct_container): Remove TFmode check
769         for X86_64_INTEGER_CLASS.
771 2014-02-19  Uros Bizjak  <ubizjak@gmail.com>
773         PR target/59794
774         * config/i386/i386.c (type_natural_mode): Warn for ABI changes
775         only when -Wpsabi is enabled.
777 2014-02-19  Michael Hudson-Doyle  <michael.hudson@linaro.org>
779          PR target/59799
780         * config/aarch64/aarch64.c (aarch64_pass_by_reference): The rules for
781         passing arrays in registers are the same as for structs, so remove the
782         special case for them.
784 2014-02-19  Eric Botcazou  <ebotcazou@adacore.com>
786         * expr.c (expand_expr_real_1) <case VIEW_CONVERT_EXPR>: For a bit-field
787         destination type, extract only the valid bits if the source type is not
788         integral and has a different mode.
790 2014-02-19  Richard Biener  <rguenther@suse.de>
792         PR ipa/60243
793         * tree-inline.c (estimate_num_insns): Avoid calling cgraph_get_node
794         for all calls.
796 2014-02-19  Richard Biener  <rguenther@suse.de>
798         PR ipa/60243
799         * ipa-prop.c: Include stringpool.h and tree-ssanames.h.
800         (ipa_modify_call_arguments): Emit an argument load explicitely and
801         preserve virtual SSA form there and for the replacement call.
802         Do not update SSA form nor free dominance info.
804 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
806         * ipa.c (function_and_variable_visibility): Also clear WEAK
807         flag when disolving COMDAT_GROUP.
809 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
811         * ipa-prop.h (ipa_ancestor_jf_data): Update ocmment.
812         * ipa-prop.c (ipa_set_jf_known_type): Return early when
813         not devirtualizing.
814         (ipa_set_ancestor_jf): Set type to NULL hwen it is not preserved;
815         do more sanity checks.
816         (detect_type_change): Return true when giving up early.
817         (compute_complex_assign_jump_func): Fix type parameter of
818         ipa_set_ancestor_jf.
819         (compute_complex_ancestor_jump_func): Likewise.
820         (update_jump_functions_after_inlining): Fix updating of
821         ancestor function.
822         * ipa-cp.c (ipa_get_jf_ancestor_result): Be ready for type to be NULL.
824 2014-02-18  Jan Hubicka  <hubicka@ucw.cz>
826         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
827         inline clones when edge disappears.
829 2014-02-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
831         PR target/60203
832         * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
833         Split 64-bit moves into 2 patterns.  Do not allow the use of
834         direct move for TDmode in little endian, since the decimal value
835         has little endian bytes within a word, but the 64-bit pieces are
836         ordered in a big endian fashion, and normal subreg's of TDmode are
837         not allowed.
838         (mov<mode>_64bit_dm): Likewise.
839         (movtd_64bit_nodm): Likewise.
841 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
843         PR tree-optimization/60174
844         * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
845         statement of an SSA_NAME that occurs in an abnormal PHI node.
847 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
849         PR sanitizer/60142
850         * final.c (SEEN_BB): Remove.
851         (SEEN_NOTE, SEEN_EMITTED): Renumber.
852         (final_scan_insn): Don't force_source_line on second
853         NOTE_INSN_BASIC_BLOCK.
855 2014-02-18  Uros Bizjak  <ubizjak@gmail.com>
857         PR target/60205
858         * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
859         * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
860         (type_natural_mode): Warn ABI change when %zmm register is not
861         available for AVX512F vector value passing.
863 2014-02-18  Kai Tietz  <ktietz@redhat.com>
865         PR target/60193
866         * config/i386/i386.c (ix86_expand_prologue): Use value in
867         rax register as displacement when restoring %r10 or %rax.
868         Fix wrong offset when restoring both registers.
870 2014-02-18  Eric Botcazou  <ebotcazou@adacore.com>
872         * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
873         assertion with conditional return.
875 2014-02-18  Jakub Jelinek  <jakub@redhat.com>
876             Uros Bizjak  <ubizjak@gmail.com>
878         PR driver/60233
879         * config/i386/driver-i386.c (host_detect_local_cpu): If
880         YMM state is not saved by the OS, also clear has_f16c.  Move
881         CPUID 0x80000001 handling before YMM state saving checking.
883 2014-02-18  Andrey Belevantsev  <abel@ispras.ru>
885         PR rtl-optimization/58960
886         * haifa-sched.c (alloc_global_sched_pressure_data): New,
887         factored out from ...
888         (sched_init): ... here.
889         (free_global_sched_pressure_data): New, factored out from ...
890         (sched_finish): ... here.
891         * sched-int.h (free_global_sched_pressure_data): Declare.
892         * sched-rgn.c (nr_regions_initial): New static global.
893         (haifa_find_rgns): Initialize it.
894         (schedule_region): Disable sched-pressure for the newly
895         generated regions.
897 2014-02-17  Richard Biener  <rguenther@suse.de>
899         * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
900         release SSA defs of pattern stmts.
902 2014-02-17  Richard Biener  <rguenther@suse.de>
904         * tree-inline.c (expand_call_inline): Release the virtual
905         operand defined by the call we are about to inline.
907 2014-02-17  Richard Biener  <rguenther@suse.de>
909         * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
911 2014-02-17  Kirill Yukhin  <kirill.yukhin@intel.com>
912             Ilya Tocar  <ilya.tocar@intel.com>
914         * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
915         arguments order in builtin.
916         (_mm512_permutexvar_epi64): Ditto.
917         (_mm512_mask_permutexvar_epi64): Ditto
918         (_mm512_maskz_permutexvar_epi32): Ditto
919         (_mm512_permutexvar_epi32): Ditto
920         (_mm512_mask_permutexvar_epi32): Ditto
922 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
924         * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
925         (p8_vmrgow): Likewise.
927 2014-02-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
929         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
930         endian targets.
932 2014-02-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
934         PR target/60203
935         * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
936         (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
937         into 64-bit and 32-bit moves.  On 64-bit moves, add support for
938         using direct move instructions on ISA 2.07.  Also adjust
939         instruction length for 64-bit.
940         (mov<mode>_64bit, TFmode/TDmode): Likewise.
941         (mov<mode>_32bit, TFmode/TDmode): Likewise.
943 2014-02-15  Alan Modra  <amodra@gmail.com>
945         PR target/58675
946         PR target/57935
947         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
948         find_replacement on parts of insn rtl that might be reloaded.
950 2014-02-15  Richard Biener  <rguenther@suse.de>
952         PR tree-optimization/60183
953         * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
954         (tree_ssa_phiprop): Calculate and free post-dominators.
956 2014-02-14  Jeff Law  <law@redhat.com>
958         PR rtl-optimization/60131
959         * ree.c (get_extended_src_reg): New function.
960         (combine_reaching_defs): Use it rather than assuming location of REG.
961         (find_and_remove_re): Verify first operand of extension is
962         a REG before adding the insns to the copy list.
964 2014-02-14  Roland McGrath  <mcgrathr@google.com>
966         * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
967         * configure: Regenerated.
968         * config.in: Regenerated.
969         * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
970         instead of ASM_SHORT.
972 2014-02-14  Vladimir Makarov  <vmakarov@redhat.com>
973             Richard Earnshaw  <rearnsha@arm.com>
975         PR rtl-optimization/59535
976         * lra-constraints.c (process_alt_operands): Encourage alternative
977         when unassigned pseudo class is superset of the alternative class.
978         (inherit_reload_reg): Don't inherit when optimizing for code size.
979         * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
980         returning CORE_REGS for anything but Thumb1 and BASE_REGS for
981         modes not less than 4 for Thumb1.
983 2014-02-14  Kyle McMartin  <kyle@redhat.com>
985         PR pch/60010
986         * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
988 2014-02-14  Richard Biener  <rguenther@suse.de>
990         * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
991         (get_frame_arg): Drop the assert with langhook types_compatible_p.
992         Do not strip INDIRECT_REFs.
994 2014-02-14  Richard Biener  <rguenther@suse.de>
996         PR lto/60179
997         * lto-streamer-out.c (DFS_write_tree_body): Do not follow
998         DECL_FUNCTION_SPECIFIC_TARGET.
999         (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
1000         * tree-streamer-out.c (pack_ts_target_option): Remove.
1001         (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
1002         (write_ts_function_decl_tree_pointers): Do not stream
1003         DECL_FUNCTION_SPECIFIC_TARGET.
1004         * tree-streamer-in.c (unpack_ts_target_option): Remove.
1005         (unpack_value_fields): Do not stream TS_TARGET_OPTION.
1006         (lto_input_ts_function_decl_tree_pointers): Do not stream
1007         DECL_FUNCTION_SPECIFIC_TARGET.
1009 2014-02-14  Jakub Jelinek  <jakub@redhat.com>
1011         * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
1012         (get_initial_def_for_induction, vectorizable_induction): Ignore
1013         debug stmts when looking for exit_phi.
1014         (vectorizable_live_operation): Fix up condition.
1016 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
1018         * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
1019         nreverse() because it changes the content of original tree list.
1021 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
1023         * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
1024         * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
1026 2014-02-14  Chung-Ju Wu  <jasonwucj@gmail.com>
1028         * config/nds32/nds32.c (nds32_naked_function_p): Follow the
1029         GNU coding standards.
1031 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
1033         PR debug/60152
1034         * dwarf2out.c (gen_subprogram_die): Don't call
1035         add_calling_convention_attribute if subr_die is old_die.
1037 2014-02-13  Sharad Singhai  <singhai@google.com>
1039         * doc/optinfo.texi: Fix order of nodes.
1041 2014-02-13  Uros Bizjak  <ubizjak@gmail.com>
1043         * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
1044         operands[2], not operands[3].
1046 2014-02-13  Richard Biener  <rguenther@suse.de>
1048         PR bootstrap/59878
1049         * doc/install.texi (ISL): Update recommended version to 0.12.2,
1050         mention the possibility of an in-tree build.
1051         (CLooG): Update recommended version to 0.18.1, mention the
1052         possibility of an in-tree build and clarify that the ISL
1053         bundled with CLooG does not work.
1055 2014-02-13  Jakub Jelinek  <jakub@redhat.com>
1057         PR target/43546
1058         * expr.c (compress_float_constant): If x is a hard register,
1059         extend into a pseudo and then move to x.
1061 2014-02-13  Dominik Vogt  <vogt@linux.vnet.ibm.com>
1063         * config/s390/s390.c (s390_asm_output_function_label): Fix crash
1064         caused by bad second argument to warning_at() with -mhotpatch and
1065         nested functions (e.g. with gfortran).
1067 2014-02-13  Richard Sandiford  <rdsandiford@googlemail.com>
1069         * opts.c (option_name): Remove "enabled by default" rider.
1071 2014-02-12  John David Anglin  <danglin@gcc.gnu.org>
1073         * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
1075 2014-02-12  H.J. Lu  <hongjiu.lu@intel.com>
1076             Uros Bizjak  <ubizjak@gmail.com>
1078         PR target/60151
1079         * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
1080         * configure: Regenerated.
1082 2014-02-12  Richard Biener  <rguenther@suse.de>
1084         * vec.c (vec_prefix::calculate_allocation): Move as
1085         inline variant to vec.h.
1086         (vec_prefix::calculate_allocation_1): New out-of-line version.
1087         * vec.h (vec_prefix::calculate_allocation_1): Declare.
1088         (vec_prefix::m_has_auto_buf): Rename to ...
1089         (vec_prefix::m_using_auto_storage): ... this.
1090         (vec_prefix::calculate_allocation): Inline the easy cases
1091         and dispatch to calculate_allocation_1 which doesn't need the
1092         prefix address.
1093         (va_heap::reserve): Use gcc_checking_assert.
1094         (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
1095         m_using_auto_storage.
1096         (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
1097         member and adjust.
1098         (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
1099         (vec<T, va_heap, vl_ptr>::release): Avoid casting.
1100         (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
1102 2014-02-12  Richard Biener  <rguenther@suse.de>
1104         * gcse.c (compute_transp): break from loop over canon_modify_mem_list
1105         when we found a dependence.
1107 2014-02-12  Thomas Schwinge  <thomas@codesourcery.com>
1109         * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
1110         common code...
1111         (maybe_fold_stmt): ... into this new function.
1112         * omp-low.c (lower_omp): Update comment.
1114         * omp-low.c (lower_omp_target): Add clobber for sizes array, after
1115         last use.
1117         * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
1118         dereference.
1120 2014-02-12  James Greenhalgh  <james.greenhalgh@arm.com>
1122         * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
1123         identifiers in comments.
1124         (cortexa53_extra_costs): Likewise.
1125         * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
1126         (cortexa7_extra_costs): Likewise.
1127         (cortexa12_extra_costs): Likewise.
1128         (cortexa15_extra_costs): Likewise.
1129         (v7m_extra_costs): Likewise.
1131 2014-02-12  Richard Biener  <rguenther@suse.de>
1133         PR middle-end/60092
1134         * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
1135         of posix_memalign being successful.
1136         (lower_stmt): Restrict lowering of posix_memalign to when
1137         -ftree-bit-ccp is enabled.
1139 2014-02-12  Senthil Kumar Selvaraj  <senthil_kumar.selvaraj@atmel.com>
1141         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
1142         arg_loc.
1143         * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
1145 2014-02-12  Eric Botcazou  <ebotcazou@adacore.com>
1147         PR rtl-optimization/60116
1148         * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
1149         other_insn once the combination has been validated.
1151 2014-02-11  Jan Hubicka  <hubicka@ucw.cz>
1153         PR lto/59468
1154         * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
1155         and wrapper.
1156         * ipa-devirt.c: Include demangle.h
1157         (odr_violation_reported): New static variable.
1158         (add_type_duplicate): Update odr_violations.
1159         (maybe_record_node): Add completep parameter; update it.
1160         (record_target_from_binfo): Add COMPLETEP parameter;
1161         update it as needed.
1162         (possible_polymorphic_call_targets_1): Likewise.
1163         (struct polymorphic_call_target_d): Add nonconstruction_targets;
1164         rename FINAL to COMPLETE.
1165         (record_targets_from_bases): Sanity check we found the binfo;
1166         fix COMPLETEP updating.
1167         (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
1168         parameter, fix computing of COMPLETEP.
1169         (dump_possible_polymorphic_call_targets): Imrove readability of dump;
1170         at LTO time do demangling.
1171         (ipa_devirt): Use nonconstruction_targets; Improve dumps.
1172         * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
1173         parameter.
1174         (gimple_get_virt_method_for_binfo): Likewise.
1175         * gimple-fold.h (gimple_get_virt_method_for_binfo,
1176         gimple_get_virt_method_for_vtable): Update prototypes.
1178 2014-02-11  Vladimir Makarov  <vmakarov@redhat.com>
1180         PR target/49008
1181         * genautomata.c (add_presence_absence): Fix typo with
1182         {final_}presence_list.
1184 2014-02-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
1186         PR target/60137
1187         * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
1188         for VSX/Altivec vectors that land in GPR registers.
1190 2014-02-11  Richard Henderson  <rth@redhat.com>
1191             Jakub Jelinek  <jakub@redhat.com>
1193         PR debug/59776
1194         * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
1195         around drhs if type conversion to lacc->type is not useless.
1197 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1199         * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
1200         tuning struct.
1201         (cortex-a57.cortex-a53): Likewise.
1202         * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
1204 2014-02-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1206         * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
1207         arm_restrict_it.
1209 2014-02-11  Renlin Li  <Renlin.Li@arm.com>
1211         * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
1212         add_options_for_arm_vfp3.
1214 2014-02-11  Jeff Law  <law@redhat.com>
1216         PR middle-end/54041
1217         * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
1218         object with an undesirable mode.
1220 2014-02-11  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1222         PR libgomp/60107
1223         * config/i386/sol2-9.h: New file.
1224         * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
1225         *-*-solaris2.9*): Use it.
1227 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
1229         * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
1230         * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
1232 2014-02-10  Nagaraju Mekala <nagaraju.mekala@xilinx.com>
1234         * config/microblaze/microblaze.c: Extend mcpu version format
1236 2014-02-10  David Holsgrove <david.holsgrove@xilinx.com>
1238         * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
1240 2014-02-10  Richard Henderson  <rth@redhat.com>
1242         PR target/59927
1243         * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
1244         * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
1245         ms-abi vs -mno-accumulate-outgoing-args.
1246         (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
1247         * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
1248         respect to ms-abi.
1250 2014-02-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1252         PR middle-end/60080
1253         * cfgexpand.c (expand_asm_operands): Attach source location to
1254         ASM_INPUT rtx objects.
1255         * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
1257 2014-02-10  Nick Clifton  <nickc@redhat.com>
1259         * config/mn10300/mn10300.c (popcount): New function.
1260         (mn10300_expand_prologue): Include saved registers in stack usage
1261         count.
1263 2014-02-10  Jeff Law  <law@redhat.com>
1265         PR middle-end/52306
1266         * reload1.c (emit_input_reload_insns): Do not create invalid RTL
1267         when changing the SET_DEST of a prior insn to avoid an input reload.
1269 2014-02-10  Ulrich Weigand  <Ulrich.Weigand@de.ibm.com>
1271         * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
1272         big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
1273         -mcall-openbsd, or -mcall-linux.
1274         (CC1_ENDIAN_BIG_SPEC): Remove.
1275         (CC1_ENDIAN_LITTLE_SPEC): Remove.
1276         (CC1_ENDIAN_DEFAULT_SPEC): Remove.
1277         (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
1278         (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
1279         and %cc1_endian_default.
1280         * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
1282 2014-02-10  Richard Biener  <rguenther@suse.de>
1284         PR tree-optimization/60115
1285         * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
1286         MEM_REF handling.  Properly verify that the accesses are not
1287         out of the objects bound.
1289 2014-02-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1291         * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
1292         coretex to cortex.
1294 2014-02-10  Eric Botcazou  <ebotcazou@adacore.com>
1296         * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
1297         proper constants and fix formatting.
1298         (possible_polymorphic_call_targets): Fix formatting.
1300 2014-02-10  Kirill Yukhin  <kirill.yukhin@intel.com>
1301             Ilya Tocar  <ilya.tocar@intel.com>
1303         * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
1304         (_mm512_loadu_epi32): Renamed into...
1305         (_mm512_loadu_si512): This.
1306         (_mm512_storeu_epi32): Renamed into...
1307         (_mm512_storeu_si512): This.
1308         (_mm512_maskz_ceil_ps): Removed.
1309         (_mm512_maskz_ceil_pd): Ditto.
1310         (_mm512_maskz_floor_ps): Ditto.
1311         (_mm512_maskz_floor_pd): Ditto.
1312         (_mm512_floor_round_ps): Ditto.
1313         (_mm512_floor_round_pd): Ditto.
1314         (_mm512_ceil_round_ps): Ditto.
1315         (_mm512_ceil_round_pd): Ditto.
1316         (_mm512_mask_floor_round_ps): Ditto.
1317         (_mm512_mask_floor_round_pd): Ditto.
1318         (_mm512_mask_ceil_round_ps): Ditto.
1319         (_mm512_mask_ceil_round_pd): Ditto.
1320         (_mm512_maskz_floor_round_ps): Ditto.
1321         (_mm512_maskz_floor_round_pd): Ditto.
1322         (_mm512_maskz_ceil_round_ps): Ditto.
1323         (_mm512_maskz_ceil_round_pd): Ditto.
1324         (_mm512_expand_pd): Ditto.
1325         (_mm512_expand_ps): Ditto.
1326         * config/i386/i386.c (ix86_builtins): Remove
1327         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
1328         (bdesc_args): Ditto.
1329         * config/i386/predicates.md (const1256_operand): New.
1330         (const_1_to_2_operand): Ditto.
1331         * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
1332         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
1333         (*avx512pf_gatherpf<mode>sf): Ditto.
1334         (avx512pf_gatherpf<mode>df): Ditto.
1335         (*avx512pf_gatherpf<mode>df_mask): Ditto.
1336         (*avx512pf_gatherpf<mode>df): Ditto.
1337         (avx512pf_scatterpf<mode>sf): Ditto.
1338         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1339         (*avx512pf_scatterpf<mode>sf): Ditto.
1340         (avx512pf_scatterpf<mode>df): Ditto.
1341         (*avx512pf_scatterpf<mode>df_mask): Ditto.
1342         (*avx512pf_scatterpf<mode>df): Ditto.
1343         (avx512f_expand<mode>): Removed.
1344         (<shift_insn><mode>3<mask_name>): Change predicate type.
1346 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
1348         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
1349         not at the end of datarefs vector use ordered_remove to avoid
1350         reordering datarefs vector.
1352         PR c/59984
1353         * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
1354         mark local addressable non-static vars as GOVD_PRIVATE
1355         instead of GOVD_LOCAL.
1356         * omp-low.c (lower_omp_for): Move gimple_bind_vars
1357         and BLOCK_VARS of gimple_bind_block to new_stmt rather
1358         than copying them.
1360         PR middle-end/60092
1361         * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
1362         if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
1363         assume_aligned or alloc_align attributes.
1364         (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
1365         arguments.  Handle also assume_aligned and alloc_align attributes.
1366         (evaluate_stmt): Adjust bit_value_assume_aligned caller.  Handle
1367         calls to functions with assume_aligned or alloc_align attributes.
1368         * doc/extend.texi: Document assume_aligned and alloc_align attributes.
1370 2014-02-08  Terry Guo  <terry.guo@arm.com>
1372         * doc/invoke.texi: Document ARM -march=armv7e-m.
1374 2014-02-08  Jakub Jelinek  <jakub@redhat.com>
1376         * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
1377         flag on __cilkrts_rethrow builtin.
1379         PR ipa/60026
1380         * ipa-cp.c (determine_versionability): Fail at -O0
1381         or __attribute__((optimize (0))) or -fno-ipa-cp functions.
1382         * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
1384         Revert:
1385         2014-02-04  Jakub Jelinek  <jakub@redhat.com>
1387         PR ipa/60026
1388         * tree-inline.c (copy_forbidden): Fail for
1389         __attribute__((optimize (0))) functions.
1391 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
1393         * varpool.c: Include pointer-set.h.
1394         (varpool_remove_unreferenced_decls): Variables in other partitions
1395         will not be output; be however careful to not lose information
1396         about partitioning.
1398 2014-02-07  Jan Hubicka  <hubicka@ucw.cz>
1400         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
1401         lookup in the vtable constructor.
1403 2014-02-07  Jeff Law  <law@redhat.com>
1405         PR target/40977
1406         * config/m68k/m68k.md (ashldi_extsi): Turn into a
1407         define_insn_and_split.
1409         * ipa-inline.c (inline_small_functions): Fix typos.
1411 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
1413         * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
1414         (s390_can_use_return_insn): Declare.
1415         * config/s390/s390.h (EPILOGUE_USES): Define.
1416         * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
1417         instructions.
1418         (s390_chunkify_start): Handle return JUMP_LABELs.
1419         (s390_early_mach): Emit a main_pool instruction on the entry edge.
1420         (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
1421         (s390_can_use_return_insn): New functions.
1422         (s390_fix_long_loop_prediction): Handle conditional returns.
1423         (TARGET_SET_UP_BY_PROLOGUE): Define.
1424         * config/s390/s390.md (ANY_RETURN): New code iterator.
1425         (*creturn, *csimple_return, return, simple_return): New patterns.
1427 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
1429         * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
1430         notes to each restore.  Also add REG_CFA_DEF_CFA when restoring %r15.
1431         (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P.  Update the
1432         REG_CFA_RESTORE list when deciding not to restore a register.
1434 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
1436         * config/s390/s390.c: Include tree-pass.h and context.h.
1437         (s390_early_mach): New function, split out from...
1438         (s390_emit_prologue): ...here.
1439         (pass_data_s390_early_mach): New pass structure.
1440         (pass_s390_early_mach): New class.
1441         (s390_option_override): Create and register early_mach pass.
1442         Move to end of file.
1444 2014-02-07  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
1446         * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
1447         to match for the exit block.
1449 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
1451         * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
1452         ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
1453         Reject misaligned operands.
1455 2014-02-07  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
1457         * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
1459 2014-02-07  Richard Biener  <rguenther@suse.de>
1461         PR middle-end/60092
1462         * gimple-low.c (lower_builtin_posix_memalign): New function.
1463         (lower_stmt): Call it to lower posix_memalign in a way
1464         to make alignment info accessible.
1466 2014-02-07  Jakub Jelinek  <jakub@redhat.com>
1468         PR c++/60082
1469         * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
1470         __builtin_setjmp_receiver.
1472 2014-02-07  Richard Biener  <rguenther@suse.de>
1474         PR middle-end/60092
1475         * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
1476         * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
1477         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
1478         Handle BUILT_IN_POSIX_MEMALIGN.
1479         (find_func_clobbers): Likewise.
1480         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
1481         (call_may_clobber_ref_p_1): Likewise.
1483 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
1485         PR ipa/59918
1486         * ipa-devirt.c (record_target_from_binfo): Remove overactive
1487         sanity check.
1489 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
1491         PR ipa/59469
1492         * lto-cgraph.c (lto_output_node): Use
1493         symtab_get_symbol_partitioning_class.
1494         (lto_output_varpool_node): likewise.
1495         (symtab_get_symbol_partitioning_class): Move here from
1496         lto/lto-partition.c
1497         * cgraph.h (symbol_partitioning_class): Likewise.
1498         (symtab_get_symbol_partitioning_class): Declare.
1500 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
1502         * ggc.h (ggc_internal_cleared_alloc): New macro.
1503         * vec.h (vec_safe_copy): Handle memory stats.
1504         * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
1505         * target-globals.c (save_target_globals): Likewise.
1507 2014-02-06  Jan Hubicka  <hubicka@ucw.cz>
1509         PR target/60077
1510         * expr.c (emit_move_resolve_push): Export; be bit more selective
1511         on when to clear alias set.
1512         * expr.h (emit_move_resolve_push): Declare.
1513         * function.h (struct function): Add tail_call_marked.
1514         * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
1515         * config/i386/i386-protos.h (ix86_expand_push): Remove.
1516         * config/i386/i386.md (TImode move expander): De not call
1517         ix86_expand_push.
1518         (FP push expanders): Preserve memory attributes.
1519         * config/i386/sse.md (push<mode>1): Remove.
1520         * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
1521         (ix86_expand_push): Remove.
1522         * config/i386/mmx.md (push<mode>1): Remove.
1524 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
1526         PR rtl-optimization/60030
1527         * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
1528         lopart with paradoxical subreg before shifting it up by hprec.
1530 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1532         * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
1533         Remove extra newline at end of file.
1534         * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
1535         (arm_issue_rate): Handle cortexa57.
1536         * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
1537         (cortex-a57.cortex-a53): Likewise.
1539 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
1541         PR target/59575
1542         * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
1543         don't record in REG_FRAME_RELATED_EXPR registers not set in that
1544         bitmask.
1545         (arm_expand_prologue): Adjust all callers.
1546         (arm_unwind_emit_sequence): Allow saved, but not important for unwind
1547         info, registers also at the lowest numbered registers side.  Use
1548         gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
1549         XEXP.
1551         PR debug/59992
1552         * var-tracking.c (adjust_mems): Before adding a SET to
1553         amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
1555 2014-02-06  Alan Modra  <amodra@gmail.com>
1557         PR target/60032
1558         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
1559         change SDmode to DDmode when lra_in_progress.
1561 2014-02-06  Jakub Jelinek  <jakub@redhat.com>
1563         PR middle-end/59150
1564         * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
1565         free_data_ref on the dr first, and before goto again also set dr
1566         to the next dr.  For simd_lane_access, free old datarefs[i] before
1567         overwriting it.  For get_vectype_for_scalar_type failure, don't
1568         free_data_ref if simd_lane_access.
1570         * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
1572         PR target/60062
1573         * tree.h (opts_for_fn): New inline function.
1574         (opt_for_fn): Define.
1575         * config/i386/i386.c (ix86_function_regparm): Use
1576         opt_for_fn (decl, optimize) instead of optimize.
1578 2014-02-06  Marcus Shawcroft  <marcus.shawcroft@arm.com>
1580         * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
1581         for SYMBOL_REF in large memory model.
1583 2014-02-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1585         * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
1586         and crypto support.
1587         (cortex-a57): Likewise.
1588         (cortex-a57.cortex-a53): Likewise.
1590 2014-02-06  Yury Gribov  <y.gribov@samsung.com>
1591             Kugan Vivekanandarajah  <kuganv@linaro.org>
1593         * config/arm/arm.c (arm_vector_alignment_reachable): Check
1594         unaligned_access.
1595         * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
1597 2014-02-06  Richard Biener  <rguenther@suse.de>
1599         * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
1600         set_loop_copy and initialize_original_copy_tables.
1602 2014-02-06  Alex Velenko  <Alex.Velenko@arm.com>
1604         * config/aarch64/aarch64-simd.md
1605         (aarch64_ashr_simddi): Change QI to SI.
1607 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
1608             Jakub Jelinek  <jakub@redhat.com>
1610         PR middle-end/60013
1611         * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
1612         of the dataflow.
1614 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1616         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
1617         CODE_FOR_altivec_vpku[hw]um to
1618         CODE_FOR_altivec_vpku[hw]um_direct.
1619         * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
1620         UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
1621         (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
1622         UNSPEC_VUNPACK_LO_SIGN_DIRECT.
1624 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1626         * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
1627         generation for -maltivec=be.
1628         (altivec_vsumsws): Simplify redundant test.
1630 2014-02-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1632         * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
1633         (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
1634         (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
1635         (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
1636         gen_altivec_vpkuwum.
1637         (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
1638         BYTES_BIG_ENDIAN.
1639         (altivec_vpks<VI_char>ss): Likewise.
1640         (altivec_vpks<VI_char>us): Likewise.
1641         (altivec_vpku<VI_char>us): Likewise.
1642         (altivec_vpku<VI_char>um): Likewise.
1643         (altivec_vpku<VI_char>um_direct): New (copy of
1644         altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
1645         internal use).
1646         (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
1647         target is little endian and -maltivec=be is not specified.
1648         (*altivec_vupkhs<VU_char>_direct): New (copy of
1649         altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
1650         (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
1651         target is little endian and -maltivec=be is not specified.
1652         (*altivec_vupkls<VU_char>_direct): New (copy of
1653         altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
1654         (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
1655         little endian and -maltivec=be is not specified.
1656         (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
1657         little endian and -maltivec=be is not specified.
1659 2014-02-05  Richard Henderson  <rth@redhat.com>
1661         PR debug/52727
1662         * combine-stack-adj.c: Revert r206943.
1663         * sched-int.h (struct deps_desc): Add last_args_size.
1664         * sched-deps.c (init_deps): Initialize it.
1665         (sched_analyze_insn): Add OUTPUT dependencies between insns that
1666         contain REG_ARGS_SIZE notes.
1668 2014-02-05  Jan Hubicka  <hubicka@ucw.cz>
1670         * lto-cgraph.c (asm_nodes_output): Make global.
1671         * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
1672         * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
1673         (driver_handle_option): Handle OPT_fwpa.
1675 2014-02-05  Jakub Jelinek  <jakub@redhat.com>
1677         PR ipa/59947
1678         * ipa-devirt.c (possible_polymorphic_call_targets): Fix
1679         a comment typo and formatting issue.  If odr_hash hasn't been
1680         created, return vNULL and set *completep to false.
1682         PR middle-end/57499
1683         * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
1684         bb with no successors.
1686 2014-02-05  James Greenhalgh  <james.greenhalgh@arm.com>
1688         PR target/59718
1689         * doc/invoke.texi (-march): Clarify documentation for ARM.
1690         (-mtune): Likewise.
1691         (-mcpu): Likewise.
1693 2014-02-05  Richard Biener  <rguenther@suse.de>
1695         * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
1696         when not vectorizing because of too many alias checks.
1697         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
1698         Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
1700 2014-02-05  Nick Clifton  <nickc@redhat.com>
1702         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
1703         accept extended registers in any mode when compiling for the MN10300.
1705 2014-02-05  Yury Gribov  <y.gribov@samsung.com>
1707         * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
1708         * ipa-inline.c (report_inline_failed_reason): Handle mismatched
1709         sanitization attributes.
1710         (can_inline_edge_p): Likewise.
1711         (sanitize_attrs_match_for_inline_p): New function.
1713 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
1715         * ipa-prop.c (detect_type_change): Shor circuit testing of
1716         type changes on THIS pointer.
1718 2014-02-04  John David Anglin  <danglin@gcc.gnu.org>
1720         PR target/59777
1721         * config/pa/pa.c (legitimize_tls_address): Return original address
1722         if not passed a SYMBOL_REF rtx.
1723         (hppa_legitimize_address): Call legitimize_tls_address for all TLS
1724         addresses.
1725         (pa_emit_move_sequence): Simplify TLS source operands.
1726         (pa_legitimate_constant_p): Reject all TLS constants.
1727         * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
1728         (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
1730 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
1732         * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
1733         groups when we know they are controlled by LTO.
1734         * varasm.c (default_binds_local_p_1): If object is in other partition,
1735         it will be resolved locally.
1737 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1739         * config/host-linux.c (linux_gt_pch_use_address): Don't
1740         use SSIZE_MAX because it is not always defined.
1742 2014-02-04  Vladimir Makarov  <vmakarov@redhat.com>
1744         PR bootstrap/59913
1745         * lra-constraints.c (need_for_split_p): Use more 3 reloads as
1746         threshold for pseudo splitting.
1747         (update_ebb_live_info): Process call argument hard registers and
1748         hard registers from insn definition too.
1749         (max_small_class_regs_num): New constant.
1750         (inherit_in_ebb): Update live hard regs through EBBs.  Update
1751         reloads_num only for small register classes.  Don't split for
1752         outputs of jumps.
1754 2014-02-04  Markus Trippelsdorf  <markus@trippelsdorf.de>
1756         PR ipa/60058
1757         * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
1758         is non-null.
1760 2014-02-04  Jan Hubicka  <hubicka@ucw.cz>
1762         * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
1763         visibility is safe.
1765 2014-02-04  Marek Polacek  <polacek@redhat.com>
1767         * gdbinit.in (pel): Define.
1769 2014-02-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1771         * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
1772         behavior.
1774 2014-02-04  Richard Biener  <rguenther@suse.de>
1776         PR lto/59723
1777         * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
1778         in function context local.
1779         (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
1780         * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
1781         similar to LTO_imported_decl_ref.
1783 2014-02-04  Jakub Jelinek  <jakub@redhat.com>
1785         PR tree-optimization/60002
1786         * cgraphclones.c (build_function_decl_skip_args): Clear
1787         DECL_LANG_SPECIFIC.
1789         PR tree-optimization/60023
1790         * tree-if-conv.c (predicate_mem_writes): Pass true instead of
1791         false to gsi_replace.
1792         * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
1793         has been in some EH region and vec_stmt could throw, add
1794         vec_stmt into the same EH region.
1795         * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
1796         has no lhs, ignore it.
1797         * internal-fn.c (expand_MASK_LOAD): Likewise.
1799         PR ipa/60026
1800         * tree-inline.c (copy_forbidden): Fail for
1801         __attribute__((optimize (0))) functions.
1803         PR other/58712
1804         * omp-low.c (simd_clone_struct_copy): If from->inbranch
1805         is set, copy one less argument.
1806         (expand_simd_clones): Don't subtract clone_info->inbranch
1807         from simd_clone_struct_alloc argument.
1809         PR rtl-optimization/57915
1810         * recog.c (simplify_while_replacing): If all unary/binary/relational
1811         operation arguments are constant, attempt to simplify those.
1813         PR middle-end/59261
1814         * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
1815         if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
1817 2014-02-04  Richard Biener  <rguenther@suse.de>
1819         PR tree-optimization/60012
1820         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
1821         TBAA disambiguation to all DDRs.
1823 2014-02-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
1825         PR target/59788
1826         * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
1827         (LINK_SPEC): Use it for -shared, -shared-libgcc.
1829 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
1831         PR ipa/59882
1832         * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
1834 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
1836         * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
1837         * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
1839 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
1841         PR ipa/59831
1842         * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
1843         to figure out targets of polymorphic calls with known decl.
1844         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
1845         * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
1846         * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
1847         (get_polymorphic_call_info): ... here.
1848         (get_polymorphic_call_info_from_invariant): New function.
1850 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
1852         * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
1853         lookup via vtable pointer; check for type consistency
1854         and turn inconsitent facts into UNREACHABLE.
1855         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
1856         * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
1857         type inconsistent querries; return UNREACHABLE instead.
1859 2014-02-03  Richard Henderson  <rth@twiddle.net>
1861         PR tree-opt/59924
1862         * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
1863         already processed this node.
1864         (normalize_one_pred_1): Pass along mark_set.
1865         (normalize_one_pred): Create and destroy a pointer_set_t.
1866         (normalize_one_pred_chain): Likewise.
1868 2014-02-03  Laurent Aflonsi  <laurent.alfonsi@st.com>
1870         PR gcov-profile/58602
1871         * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
1873 2014-02-03  Jan Hubicka  <hubicka@ucw.cz>
1875         PR ipa/59831
1876         * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
1877         -fno-devirtualize; try to devirtualize by the knowledge of
1878         virtual table pointer given by aggregate propagation.
1879         * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
1880         (ipa_print_node_jump_functions): Dump also offset that
1881         is relevant for polymorphic calls.
1882         (determine_known_aggregate_parts): Add arg_type parameter; use it
1883         instead of determining the type from pointer type.
1884         (ipa_compute_jump_functions_for_edge): Update call of
1885         determine_known_aggregate_parts.
1886         * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
1887         (gimple_get_virt_method_for_binfo): ... here; simplify using
1888         vtable_pointer_value_to_vtable.
1889         * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
1890         * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
1891         to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
1892         (vtable_pointer_value_to_vtable): Break out from ...; handle also
1893         POINTER_PLUS_EXPR.
1894         (vtable_pointer_value_to_binfo): ... here.
1895         * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
1897 2014-02-03  Teresa Johnson  <tejohnson@google.com>
1899         * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
1900         redef of outer loop index variable.
1902 2014-02-03  Marc Glisse  <marc.glisse@inria.fr>
1904         PR c++/53017
1905         PR c++/59211
1906         * doc/extend.texi (Function Attributes): Typo.
1908 2014-02-03  Cong Hou  <congh@google.com>
1910         PR tree-optimization/60000
1911         * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
1912         if the vectorized statement is a store.  A store statement can only
1913         appear at the end of pattern statements.
1915 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
1917         * config/i386/i386.c (flag_opts): Add -mlong-double-128.
1918         (ix86_option_override_internal): Default long double to 64-bit for
1919         32-bit Bionic and to 128-bit for 64-bit Bionic.
1921         * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
1922         TARGET_LONG_DOUBLE_128 is true.
1923         (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1925         * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
1926         (mlong-double-64): Negate -mlong-double-128.
1927         (mlong-double-128): New option.
1929         * config/i386/i386-c.c (ix86_target_macros): Define
1930         __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
1932         * doc/invoke.texi: Document -mlong-double-128.
1934 2014-02-03  H.J. Lu  <hongjiu.lu@intel.com>
1936         PR rtl-optimization/60024
1937         * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
1939 2014-02-03  Markus Trippelsdorf  <markus@trippelsdorf.de>
1941         * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
1943 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
1945         PR rtl-optimization/57662
1946         * sel-sched.c (code_motion_path_driver): Do not mark already not
1947         existing blocks in the visiting bitmap.
1949 2014-02-03  Andrey Belevantsev  <abel@ispras.ru>
1951         * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
1952         on the insn being emitted.
1954 2014-02-03  James Greenhalgh  <james.greenhalgh@arm.com>
1955             Will Deacon  <will.deacon@arm.com>
1957         * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
1959 2014-02-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1961         * config/arm/arm-tables.opt: Regenerate.
1963 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1965         * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
1966         for vector types other than V16QImode.
1967         * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
1968         define_expand, and call altivec_expand_vec_perm_le when producing
1969         code with little endian element order.
1970         (*altivec_vperm_<mode>_internal): New insn having previous
1971         behavior of altivec_vperm_<mode>.
1972         (altivec_vperm_<mode>_uns): Change to a define_expand, and call
1973         altivec_expand_vec_perm_le when producing code with little endian
1974         element order.
1975         (*altivec_vperm_<mode>_uns_internal): New insn having previous
1976         behavior of altivec_vperm_<mode>_uns.
1978 2014-02-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1980         * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
1981         (altivec_vsumsws): Add handling for -maltivec=be with a little
1982         endian target.
1983         (altivec_vsumsws_direct): New.
1984         (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
1985         gen_altivec_vsumsws.
1987 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
1989         * ipa-devirt.c (subbinfo_with_vtable_at_offset,
1990         vtable_pointer_value_to_binfo): New functions.
1991         * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
1992         * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
1994 2014-02-02  Sandra Loosemore  <sandra@codesourcery.com>
1996         * config/nios2/nios2.md (load_got_register): Initialize GOT
1997         pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
1998         * config/nios2/nios2.c (nios2_function_profiler): Likewise.
2000 2014-02-02  Jan Hubicka  <hubicka@ucw.cz>
2002         * ipa-prop.c (update_jump_functions_after_inlining): When type is not
2003         preserverd by passthrough, do not propagate the type.
2005 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
2007         * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
2008         (mips_atomic_assign_expand_fenv): New function.
2009         (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
2011 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
2013         * doc/extend.texi (__builtin_mips_get_fcsr): Document.
2014         (__builtin_mips_set_fcsr): Likewise.
2015         * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
2016         MIPS_USI_FTYPE_VOID.
2017         * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
2018         (mips16_expand_set_fcsr): Likewise.
2019         * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
2020         (mips16_set_fcsr_stub): Likewise.
2021         (mips16_get_fcsr_one_only_stub): New class.
2022         (mips16_set_fcsr_one_only_stub): Likewise.
2023         (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
2024         (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
2025         (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
2026         (hard_float): New availability predicate.
2027         (mips_builtins): Add get_fcsr and set_fcsr.
2028         (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
2029         * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
2030         (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
2031         (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
2032         (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
2033         patterns.
2035 2014-02-02  Richard Sandiford  <rdsandiford@googlemail.com>
2037         * config/mips/mips.c (mips_one_only_stub): New class.
2038         (mips_need_mips16_rdhwr_p): Replace with...
2039         (mips16_rdhwr_stub): ...this new variable.
2040         (mips16_stub_call_address): New function.
2041         (mips16_rdhwr_one_only_stub): New class.
2042         (mips_expand_thread_pointer): Use mips16_stub_call_address.
2043         (mips_output_mips16_rdhwr): Delete.
2044         (mips_finish_stub): New function.
2045         (mips_code_end): Use it to handle rdhwr stubs.
2047 2014-02-02  Uros Bizjak  <ubizjak@gmail.com>
2049         PR target/60017
2050         * config/i386/i386.c (classify_argument): Fix handling of bit_offset
2051         when calculating size of integer atomic types.
2053 2014-02-02  H.J. Lu  <hongjiu.lu@intel.com>
2055         * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
2057 2014-02-01  Jakub Jelinek  <jakub@redhat.com>
2059         PR tree-optimization/60003
2060         * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
2061         * profile.c (branch_prob): Use gimple_call_builtin_p
2062         to check for BUILT_IN_SETJMP_RECEIVER.
2063         * tree-inline.c (copy_bb): Call notice_special_calls.
2065 2014-01-31  Vladimir Makarov  <vmakarov@redhat.com>
2067         PR bootstrap/59985
2068         * lra-constraints.c (process_alt_operands): Update reload_sum only
2069         on the first pass.
2071 2014-01-31  Richard Henderson  <rth@redhat.com>
2073         PR middle-end/60004
2074         * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
2075         until after else_eh is processed.
2077 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
2079         * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
2080         (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
2081         (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
2082         in smmintrin.h, remove them.
2083         (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
2084         * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
2085         * config/i386/i386.md (ROUND_SAE): Fix value.
2086         * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
2087         (const48_operand): New.
2088         * config/i386/subst.md (round), (round_expand): Use
2089         const_4_or_8_to_11_operand.
2090         (round_saeonly), (round_saeonly_expand): Use const48_operand.
2092 2014-01-31  Ilya Tocar  <ilya.tocar@intel.com>
2094         * config/i386/constraints.md (Yk): Swap meaning with k.
2095         * config/i386/i386.md (movhi_internal): Change Yk to k.
2096         (movqi_internal): Ditto.
2097         (*k<logic><mode>): Ditto.
2098         (*andhi_1): Ditto.
2099         (*andqi_1): Ditto.
2100         (kandn<mode>): Ditto.
2101         (*<code>hi_1): Ditto.
2102         (*<code>qi_1): Ditto.
2103         (kxnor<mode>): Ditto.
2104         (kortestzhi): Ditto.
2105         (kortestchi): Ditto.
2106         (kunpckhi): Ditto.
2107         (*one_cmplhi2_1): Ditto.
2108         (*one_cmplqi2_1): Ditto.
2109         * config/i386/sse.md (): Change k to Yk.
2110         (avx512f_load<mode>_mask): Ditto.
2111         (avx512f_blendm<mode>): Ditto.
2112         (avx512f_store<mode>_mask): Ditto.
2113         (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
2114         (avx512f_storedqu<mode>_mask): Ditto.
2115         (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
2116         Ditto.
2117         (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
2118         (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
2119         (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
2120         (avx512f_maskcmp<mode>3): Ditto.
2121         (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
2122         (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
2123         (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
2124         (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
2125         (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
2126         (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
2127         (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
2128         (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
2129         (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
2130         (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
2131         (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
2132         (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
2133         (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
2134         (vec_extract_lo_<mode>_maskm): Ditto.
2135         (vec_extract_hi_<mode>_maskm): Ditto.
2136         (avx512f_vternlog<mode>_mask): Ditto.
2137         (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
2138         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
2139         (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
2140         (avx512f_<code>v8div16qi2_mask): Ditto.
2141         (avx512f_<code>v8div16qi2_mask_store): Ditto.
2142         (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
2143         (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
2144         (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
2145         (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
2146         (*avx512pf_gatherpf<mode>sf_mask): Ditto.
2147         (*avx512pf_gatherpf<mode>df_mask): Ditto.
2148         (*avx512pf_scatterpf<mode>sf_mask): Ditto.
2149         (*avx512pf_scatterpf<mode>df_mask): Ditto.
2150         (avx512cd_maskb_vec_dupv8di): Ditto.
2151         (avx512cd_maskw_vec_dupv16si): Ditto.
2152         (avx512f_vpermi2var<mode>3_maskz): Ditto.
2153         (avx512f_vpermi2var<mode>3_mask): Ditto.
2154         (avx512f_vpermi2var<mode>3_mask): Ditto.
2155         (avx512f_vpermt2var<mode>3_maskz): Ditto.
2156         (*avx512f_gathersi<mode>): Ditto.
2157         (*avx512f_gathersi<mode>_2): Ditto.
2158         (*avx512f_gatherdi<mode>): Ditto.
2159         (*avx512f_gatherdi<mode>_2): Ditto.
2160         (*avx512f_scattersi<mode>): Ditto.
2161         (*avx512f_scatterdi<mode>): Ditto.
2162         (avx512f_compress<mode>_mask): Ditto.
2163         (avx512f_compressstore<mode>_mask): Ditto.
2164         (avx512f_expand<mode>_mask): Ditto.
2165         * config/i386/subst.md (mask): Change k to Yk.
2166         (mask_scalar_merge): Ditto.
2167         (sd): Ditto.
2169 2014-01-31  Marc Glisse  <marc.glisse@inria.fr>
2171         * doc/extend.texi (Vector Extensions): Document ?: in C++.
2173 2014-01-31  Richard Biener  <rguenther@suse.de>
2175         PR middle-end/59990
2176         * builtins.c (fold_builtin_memory_op): Make sure to not
2177         use a floating-point mode or a boolean or enumeral type for
2178         the copy operation.
2180 2014-01-30  DJ Delorie  <dj@redhat.com>
2182         * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
2183         * config/msp430/msp430.md (msp430_refsym_need_exit): New.
2184         * config/msp430/msp430.c (msp430_expand_epilogue): Call it
2185         whenever main() has an epilogue.
2187 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2189         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
2190         unused variable "field".
2191         * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
2192         (vsx_mergeh_<mode>): Likewise.
2193         * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
2194         (altivec_vmrghh): Likewise.
2195         (altivec_vmrghw): Likewise.
2196         (altivec_vmrglb): Likewise.
2197         (altivec_vmrglh): Likewise.
2198         (altivec_vmrglw): Likewise.
2199         (altivec_vspltb): Add missing uses.
2200         (altivec_vsplth): Likewise.
2201         (altivec_vspltw): Likewise.
2202         (altivec_vspltsf): Likewise.
2204 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
2206         PR target/59923
2207         * ifcvt.c (cond_exec_process_insns): Don't conditionalize
2208         frame related instructions.
2210 2014-01-30  Vladimir Makarov  <vmakarov@redhat.com>
2212         PR rtl-optimization/59959
2213         * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
2214         any reload of register whose subreg is invalid.
2216 2014-01-30  Jakub Jelinek  <jakub@redhat.com>
2218         * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
2219         * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
2220         Add missing return type - void.
2222 2014-01-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2224         * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
2225         gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
2226         remove element index adjustment for endian (now handled in vsx.md
2227         and altivec.md).
2228         (altivec_expand_vec_perm_const): Use
2229         gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
2230         * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
2231         (vsx_xxspltw_<mode>): Adjust element index for little endian.
2232         * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
2233         define_expand and a new define_insn *altivec_vspltb_internal;
2234         adjust for -maltivec=be on a little endian target.
2235         (altivec_vspltb_direct): New.
2236         (altivec_vsplth): Divide into a define_expand and a new
2237         define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
2238         little endian target.
2239         (altivec_vsplth_direct): New.
2240         (altivec_vspltw): Divide into a define_expand and a new
2241         define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
2242         little endian target.
2243         (altivec_vspltw_direct): New.
2244         (altivec_vspltsf): Divide into a define_expand and a new
2245         define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
2246         a little endian target.
2248 2014-01-30  Richard Biener  <rguenther@suse.de>
2250         PR tree-optimization/59993
2251         * tree-ssa-forwprop.c (associate_pointerplus): Check we
2252         can propagate form the earlier stmt and avoid the transform
2253         when the intermediate result is needed.
2255 2014-01-30  Alangi Derick  <alangiderick@gmail.com>
2257         * README.Portability: Fix typo.
2259 2014-01-30  David Holsgrove <david.holsgrove@xilinx.com>
2261         * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
2262         comparison_operator with ordered_comparison_operator.
2264 2014-01-30  Nick Clifton  <nickc@redhat.com>
2266         * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
2267         Rename to mn10300_store_multiple_regs.
2268         * config/mn10300/mn10300.c: Likewise.
2269         * config/mn10300/mn10300.md (store_movm): Fix typo: call
2270         store_multiple_regs.
2271         * config/mn10300/predicates.md (mn10300_store_multiple_operation):
2272         Call mn10300_store_multiple_regs.
2274 2014-01-30  Nick Clifton  <nickc@redhat.com>
2275             DJ Delorie  <dj@redhat.com>
2277         * config/rl78/rl78.c (register_sizes): Make the "upper half" of
2278         %fp 2 to keep registers after it properly word-aligned.
2279         (rl78_alloc_physical_registers_umul): Handle the case where both
2280         input operands are the same.
2282 2014-01-30  Richard Biener  <rguenther@suse.de>
2284         PR tree-optimization/59903
2285         * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
2286         check properly.
2288 2014-01-30  Jason Merrill  <jason@redhat.com>
2290         PR c++/59633
2291         * tree.c (walk_type_fields): Handle VECTOR_TYPE.
2293         PR c++/59645
2294         * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
2296 2014-01-30  Richard Biener  <rguenther@suse.de>
2298         PR tree-optimization/59951
2299         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
2301 2014-01-30  Savin Zlobec  <savin.zlobec@gmail.com>
2303         PR target/59784
2304         * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
2305         SFmode to DFmode case.
2307 2014-01-29  DJ Delorie  <dj@redhat.com>
2309         * config/msp430/msp430.opt (-minrt): New.
2310         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
2311         if -minrt given.
2312         (ENDFILE_SPEC): Likewise.
2314 2014-01-29  Jan Hubicka  <hubicka@ucw.cz>
2316         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
2317         (estimate_function_body_sizes): Use it.
2319 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
2321         PR c++/58561
2322         * dwarf2out.c (is_cxx_auto): New.
2323         (is_base_type): Use it.
2324         (gen_type_die_with_usage): Likewise.
2326 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2328         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
2329         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
2330         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
2331         -maltivec=be with LE targets.
2332         (vsx_mergeh_<mode>): Likewise.
2333         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
2334         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
2335         (altivec_vmrghb): Replace with define_expand and new
2336         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
2337         (altivec_vmrghb_direct): New define_insn.
2338         (altivec_vmrghh): Replace with define_expand and new
2339         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
2340         (altivec_vmrghh_direct): New define_insn.
2341         (altivec_vmrghw): Replace with define_expand and new
2342         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
2343         (altivec_vmrghw_direct): New define_insn.
2344         (*altivec_vmrghsf): Adjust for endianness.
2345         (altivec_vmrglb): Replace with define_expand and new
2346         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
2347         (altivec_vmrglb_direct): New define_insn.
2348         (altivec_vmrglh): Replace with define_expand and new
2349         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
2350         (altivec_vmrglh_direct): New define_insn.
2351         (altivec_vmrglw): Replace with define_expand and new
2352         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
2353         (altivec_vmrglw_direct): New define_insn.
2354         (*altivec_vmrglsf): Adjust for endianness.
2355         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
2356         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
2357         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
2358         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
2359         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
2360         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
2361         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
2362         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
2364 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
2366         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
2367         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
2368         whitespace.
2370 2014-01-29  Richard Biener  <rguenther@suse.de>
2372         PR tree-optimization/58742
2373         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
2374         associate_pointerplus_align.
2375         (associate_pointerplus_diff): New function.
2376         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
2377         and associate_pointerplus_diff.
2379 2014-01-29  Richard Biener  <rguenther@suse.de>
2381         * lto-streamer.h (LTO_major_version): Bump to 3.
2382         (LTO_minor_version): Reset to 0.
2384 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
2386         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
2387         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
2388         (arm_file_start): Generate correct asm header for armv7ve.
2389         * config/arm/bpabi.h: Add multilib support for armv7ve.
2390         * config/arm/driver-arm.c: Change the architectures of cortex-a7
2391         and cortex-a15 to armv7ve.
2392         * config/arm/t-aprofile: Add multilib support for armv7ve.
2393         * doc/invoke.texi: Document -march=armv7ve.
2395 2014-01-29  Richard Biener  <rguenther@suse.de>
2397         PR tree-optimization/58742
2398         * tree-ssa-forwprop.c (associate_plusminus): Return true
2399         if we changed sth, defer EH cleanup to ...
2400         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
2401         (simplify_mult): New function.
2403 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
2405         PR middle-end/59917
2406         PR tree-optimization/59920
2407         * tree.c (build_common_builtin_nodes): Remove
2408         __builtin_setjmp_dispatcher initialization.
2409         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
2410         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
2411         instead of gsi_after_labels + manually skipping debug stmts.
2412         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
2413         ignore bbs with IFN_ABNORMAL_DISPATCHER.
2414         * tree-inline.c (copy_edges_for_bb): Remove
2415         can_make_abnormal_goto argument, instead add abnormal_goto_dest
2416         argument.  Ignore computed_goto_p stmts.  Don't call
2417         make_abnormal_goto_edges.  If a call might need abnormal edges
2418         for non-local gotos, see if it already has an edge to
2419         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
2420         with true argument, don't do anything then, otherwise add
2421         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
2422         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
2423         caller.
2424         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
2425         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
2426         (lower_stmt): Don't set data->calls_builtin_setjmp.
2427         (lower_builtin_setjmp): Adjust comment.
2428         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
2429         * tree-cfg.c (found_computed_goto): Remove.
2430         (factor_computed_gotos): Remove.
2431         (make_goto_expr_edges): Return bool, true for computed gotos.
2432         Don't call make_abnormal_goto_edges.
2433         (build_gimple_cfg): Don't set found_computed_goto, don't call
2434         factor_computed_gotos.
2435         (computed_goto_p): No longer static.
2436         (make_blocks): Don't set found_computed_goto.
2437         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
2438         (make_edges): If make_goto_expr_edges returns true, push bb
2439         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
2440         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
2441         vector.  Record mapping between bbs and OpenMP regions if there
2442         are any, adjust make_gimple_omp_edges caller.  Call
2443         handle_abnormal_edges.
2444         (make_abnormal_goto_edges): Remove.
2445         * tree-cfg.h (make_abnormal_goto_edges): Remove.
2446         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
2447         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
2448         * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
2449         * internal-fn.def (ABNORMAL_DISPATCHER): New.
2450         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
2451         filling *region also set *region_idx to (*region)->entry->index.
2453         PR other/58712
2454         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
2455         For REGs set ORIGINAL_REGNO.
2457 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
2459         * doc/md.texi: Mention that a target shouldn't implement
2460         vec_widen_(s|u)mul_even/odd pair if it is less efficient
2461         than hi/lo pair.
2463 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
2465         PR tree-optimization/59594
2466         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
2467         a copy of the datarefs vector rather than the vector itself.
2469 2014-01-28  Jason Merrill  <jason@redhat.com>
2471         PR c++/53756
2472         * dwarf2out.c (auto_die): New static.
2473         (gen_type_die_with_usage): Handle C++1y 'auto'.
2474         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
2475         on definition.
2477 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
2479         PR target/59672
2480         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
2481         (SPEC_X32): Likewise.
2482         (SPEC_64): Likewise.
2483         * config/i386/i386.c (ix86_option_override_internal): Turn off
2484         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
2485         for TARGET_16BIT.
2486         (x86_file_start): Output .code16gcc for TARGET_16BIT.
2487         * config/i386/i386.h (TARGET_16BIT): New macro.
2488         (TARGET_16BIT_P): Likewise.
2489         * config/i386/i386.opt: Add m16.
2490         * doc/invoke.texi: Document -m16.
2492 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
2494         PR preprocessor/59935
2495         * input.c (location_get_source_line): Bail out on when line number
2496         is zero, and test the return value of lookup_or_add_file_to_cache_tab.
2498 2014-01-28  Richard Biener  <rguenther@suse.de>
2500         PR tree-optimization/58742
2501         * tree-ssa-forwprop.c (associate_plusminus): Handle
2502         pointer subtraction of the form (T)(P + A) - (T)P.
2504 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2506         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
2507         at const_int_cost.
2509 2014-01-28  Richard Biener  <rguenther@suse.de>
2511         Revert
2512         2014-01-28  Richard Biener  <rguenther@suse.de>
2514         PR rtl-optimization/45364
2515         PR rtl-optimization/59890
2516         * var-tracking.c (local_get_addr_clear_given_value): Handle
2517         already cleared slot.
2518         (val_reset): Handle not allocated local_get_addr_cache.
2519         (vt_find_locations): Use post-order on the inverted CFG.
2521 2014-01-28  Richard Biener  <rguenther@suse.de>
2523         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
2525 2014-01-28  Richard Biener  <rguenther@suse.de>
2527         PR rtl-optimization/45364
2528         PR rtl-optimization/59890
2529         * var-tracking.c (local_get_addr_clear_given_value): Handle
2530         already cleared slot.
2531         (val_reset): Handle not allocated local_get_addr_cache.
2532         (vt_find_locations): Use post-order on the inverted CFG.
2534 2014-01-28  Alan Modra  <amodra@gmail.com>
2536         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
2537         * configure.ac <recursive call for build != host>: Define
2538         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
2539         and LD_FOR_BUILD too.
2540         * configure: Regenerate.
2542 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
2544         * config/i386/i386.c (get_builtin_code_for_version): Separate
2545         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
2546         Broadwell from Haswell.
2548 2014-01-27  Steve Ellcey  <sellcey@mips.com>
2550         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2551         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
2552         * config/mips/mips.c (mips_option_override): Change setting
2553         of TARGET_DSP.
2554         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
2555         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
2556         Change from Mask to Var.
2558 2014-01-27  Jeff Law  <law@redhat.com>
2560         * ipa-inline.c (inline_small_functions): Fix typo.
2562 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
2564         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
2565         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
2566         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
2567         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
2568         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
2569         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
2570         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
2571         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
2572         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
2573         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
2574         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
2575         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
2576         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
2577         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
2578         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
2579         (_mm512_storeu_epi64): Ditto.
2580         (_mm512_cmpge_epi32_mask): Ditto.
2581         (_mm512_cmpge_epu32_mask): Ditto.
2582         (_mm512_cmpge_epi64_mask): Ditto.
2583         (_mm512_cmpge_epu64_mask): Ditto.
2584         (_mm512_cmple_epi32_mask): Ditto.
2585         (_mm512_cmple_epu32_mask): Ditto.
2586         (_mm512_cmple_epi64_mask): Ditto.
2587         (_mm512_cmple_epu64_mask): Ditto.
2588         (_mm512_cmplt_epi32_mask): Ditto.
2589         (_mm512_cmplt_epu32_mask): Ditto.
2590         (_mm512_cmplt_epi64_mask): Ditto.
2591         (_mm512_cmplt_epu64_mask): Ditto.
2592         (_mm512_cmpneq_epi32_mask): Ditto.
2593         (_mm512_cmpneq_epu32_mask): Ditto.
2594         (_mm512_cmpneq_epi64_mask): Ditto.
2595         (_mm512_cmpneq_epu64_mask): Ditto.
2596         (_mm512_expand_pd): Ditto.
2597         (_mm512_expand_ps): Ditto.
2598         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
2599         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
2600         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
2601         * config/i386/i386.c (ix86_builtins): Add
2602         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
2603         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
2604         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
2605         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
2606         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
2607         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
2608         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
2609         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
2610         IX86_BUILTIN_PMOVUSQW512_MEM.
2611         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
2612         __builtin_ia32_pmovsqd512mem_mask,
2613         __builtin_ia32_pmovqd512mem_mask,
2614         __builtin_ia32_pmovusqw512mem_mask,
2615         __builtin_ia32_pmovsqw512mem_mask,
2616         __builtin_ia32_pmovqw512mem_mask,
2617         __builtin_ia32_pmovusdw512mem_mask,
2618         __builtin_ia32_pmovsdw512mem_mask,
2619         __builtin_ia32_pmovdw512mem_mask,
2620         __builtin_ia32_pmovqb512mem_mask,
2621         __builtin_ia32_pmovusqb512mem_mask,
2622         __builtin_ia32_pmovsqb512mem_mask,
2623         __builtin_ia32_pmovusdb512mem_mask,
2624         __builtin_ia32_pmovsdb512mem_mask,
2625         __builtin_ia32_pmovdb512mem_mask.
2626         (bdesc_args): Add __builtin_ia32_expanddf512,
2627         __builtin_ia32_expandsf512.
2628         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
2629         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
2630         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
2631         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
2632         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
2633         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
2634         (avx512f_<code>v8div16qi2_mask_store): This.
2635         (avx512f_expand<mode>): New.
2637 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
2639         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
2640         New.
2641         (_mm512_mask_prefetch_i64gather_pd): Ditto.
2642         (_mm512_prefetch_i32scatter_pd): Ditto.
2643         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
2644         (_mm512_prefetch_i64scatter_pd): Ditto.
2645         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
2646         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
2647         (_mm512_mask_prefetch_i64gather_ps): Ditto.
2648         (_mm512_prefetch_i32scatter_ps): Ditto.
2649         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
2650         (_mm512_prefetch_i64scatter_ps): Ditto.
2651         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
2652         * config/i386/i386-builtin-types.def: Define
2653         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
2654         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
2655         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
2656         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
2657         IX86_BUILTIN_SCATTERPFQPD.
2658         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
2659         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
2660         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
2661         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
2662         __builtin_ia32_scatterpfqps.
2663         (ix86_expand_builtin): Expand new built-ins.
2664         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
2665         fix memory access data type.
2666         (*avx512pf_gatherpf<mode>_mask): Ditto.
2667         (*avx512pf_gatherpf<mode>): Ditto.
2668         (avx512pf_scatterpf<mode>): Ditto.
2669         (*avx512pf_scatterpf<mode>_mask): Ditto.
2670         (*avx512pf_scatterpf<mode>): Ditto.
2671         (GATHER_SCATTER_SF_MEM_MODE): New.
2672         (avx512pf_gatherpf<mode>df): Ditto.
2673         (*avx512pf_gatherpf<mode>df_mask): Ditto.
2674         (*avx512pf_scatterpf<mode>df): Ditto.
2676 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
2678         PR bootstrap/59934
2679         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
2680         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
2681         reached.
2683 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
2685         * common/config/arm/arm-common.c
2686         (arm_rewrite_mcpu): Handle multiple names.
2687         * config/arm/arm.h
2688         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
2690 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
2692         * gimple-builder.h (create_gimple_tmp): Delete.
2694 2014-01-27  Christian Bruel  <christian.bruel@st.com>
2696         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
2697         words comparisons.
2699 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
2701         * config/pa/pa.md (call): Generate indirect long calls to non-local
2702         functions when outputing 32-bit code.
2703         (call_value): Likewise except for special call to buggy powf function.
2705         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
2706         portable runtime and PIC indirect calls.
2707         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
2708         and PIC call sequences.  Use ldo instead of blr to set return register
2709         in PIC call sequence.
2711 2014-01-25  Walter Lee  <walt@tilera.com>
2713         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
2714         avoid clobbering a live register.
2716 2014-01-25  Walter Lee  <walt@tilera.com>
2718         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
2719         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
2720         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
2721         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
2723 2014-01-25  Walter Lee  <walt@tilera.com>
2725         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
2726         arguments on even registers.
2727         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
2728         STACK_BOUNDARY.
2729         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
2730         (BIGGEST_ALIGNMENT): Ditto.
2731         (BIGGEST_FIELD_ALIGNMENT): Ditto.
2733 2014-01-25  Walter Lee  <walt@tilera.com>
2735         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
2736         insns before bundling.
2737         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
2739 2014-01-25  Walter Lee  <walt@tilera.com>
2741         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
2742         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
2743         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
2745 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
2747         * config/mips/constraints.md (kl): Delete.
2748         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
2749         define expands, using...
2750         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
2751         instructions for MIPS16.
2752         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
2753         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
2755 2014-01-25  Walter Lee  <walt@tilera.com>
2757         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
2758         (clzdi2): Ditto.
2759         (ffsdi2): Ditto.
2761 2014-01-25  Walter Lee  <walt@tilera.com>
2763         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
2764         (TARGET_EXPAND_TO_RTL_HOOK): Define.
2766 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
2768         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
2769         Handle XOR.
2771 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
2773         * print-rtl.c (in_call_function_usage): New var.
2774         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
2775         EXPR_LIST mode as mode and not as reg note name.
2777         PR middle-end/59561
2778         * cfgloopmanip.c (copy_loop_info): If
2779         loop->warned_aggressive_loop_optimizations, make sure
2780         the flag is set in target loop too.
2782 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
2784         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
2785         flag_cilkplus.
2786         * builtins.def: Likewise.
2787         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
2788         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
2789         * ira.c (ira_setup_eliminable_regset): Likewise.
2790         * omp-low.c (gate_expand_omp): Likewise.
2791         (execute_lower_omp): Likewise.
2792         (diagnose_sb_0): Likewise.
2793         (gate_diagnose_omp_blocks): Likewise.
2794         (simd_clone_clauses_extract): Likewise.
2795         (gate): Likewise.
2797 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
2799         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
2800         correction for little endian...
2801         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
2802         here.
2804 2014-01-24  Jeff Law  <law@redhat.com>
2806         PR tree-optimization/59919
2807         * tree-vrp.c (find_assert_locations_1): Do not register asserts
2808         for non-returning calls.
2810 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
2812         * common/config/aarch64/aarch64-common.c
2813         (aarch64_rewrite_mcpu): Handle multiple names.
2814         * config/aarch64/aarch64.h
2815         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
2817 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
2819         * input.c (add_file_to_cache_tab): Handle the case where fopen
2820         returns NULL.
2822 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
2824         PR target/59929
2825         * config/i386/i386.md (pushsf splitter): Get stack adjustment
2826         from push operand if code of push isn't PRE_DEC.
2828 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
2830         PR target/59909
2831         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2832         -mquad-memory-atomic.  Update -mquad-memory documentation to say
2833         it is only used for non-atomic loads/stores.
2835         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
2836         -mquad-memory or -mquad-memory-atomic switches.
2838         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
2839         -mquad-memory-atomic to ISA 2.07 support.
2841         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
2842         to separate support of normal quad word memory operations (ldq, stq)
2843         from the atomic quad word memory operations.
2845         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
2846         support to separate non-atomic quad word operations from atomic
2847         quad word operations.  Disable non-atomic quad word operations in
2848         little endian mode so that we don't have to swap words after the
2849         load and before the store.
2850         (quad_load_store_p): Add comment about atomic quad word support.
2851         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
2852         options printed with -mdebug=reg.
2854         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
2855         -mquad-memory-atomic as the test for whether we have quad word
2856         atomic instructions.
2857         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
2858         or -mp8-vector are used, allow byte/half-word atomic operations.
2860         * config/rs6000/sync.md (load_lockedti): Insure that the address
2861         is a proper indexed or indirect address for the lqarx instruction.
2862         On little endian systems, swap the hi/lo registers after the lqarx
2863         instruction.
2864         (load_lockedpti): Use indexed_or_indirect_operand predicate to
2865         insure the address is valid for the lqarx instruction.
2866         (store_conditionalti): Insure that the address is a proper indexed
2867         or indirect address for the stqcrx. instruction.  On little endian
2868         systems, swap the hi/lo registers before doing the stqcrx.
2869         instruction.
2870         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
2871         insure the address is valid for the stqcrx. instruction.
2873         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
2874         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
2875         type of quad memory support is available.
2877 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
2879         PR regression/59915
2880         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
2881         there is a danger of looping.
2883 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
2885         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2886         force flag_ira_loop_pressure if set via command line.
2888 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
2890         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
2891         (ashr_simd): New builtin handling DI mode.
2892         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
2893         (aarch64_sshr_simddi): New match pattern.
2894         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
2895         (vshrd_n_s64): Likewise.
2896         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
2898 2014-01-23  Nick Clifton  <nickc@redhat.com>
2900         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
2901         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
2902         favour of mcu specific scripts.
2903         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
2904         430x multilibs.
2906 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
2907             Alex Velenko  <Alex.Velenko@arm.com>
2909         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
2910         (vaddv_s16): Likewise.
2911         (vaddv_s32): Likewise.
2912         (vaddv_u8): Likewise.
2913         (vaddv_u16): Likewise.
2914         (vaddv_u32): Likewise.
2915         (vaddvq_s8): Likewise.
2916         (vaddvq_s16): Likewise.
2917         (vaddvq_s32): Likewise.
2918         (vaddvq_s64): Likewise.
2919         (vaddvq_u8): Likewise.
2920         (vaddvq_u16): Likewise.
2921         (vaddvq_u32): Likewise.
2922         (vaddvq_u64): Likewise.
2923         (vaddv_f32): Likewise.
2924         (vaddvq_f32): Likewise.
2925         (vaddvq_f64): Likewise.
2926         (vmaxv_f32): Likewise.
2927         (vmaxv_s8): Likewise.
2928         (vmaxv_s16): Likewise.
2929         (vmaxv_s32): Likewise.
2930         (vmaxv_u8): Likewise.
2931         (vmaxv_u16): Likewise.
2932         (vmaxv_u32): Likewise.
2933         (vmaxvq_f32): Likewise.
2934         (vmaxvq_f64): Likewise.
2935         (vmaxvq_s8): Likewise.
2936         (vmaxvq_s16): Likewise.
2937         (vmaxvq_s32): Likewise.
2938         (vmaxvq_u8): Likewise.
2939         (vmaxvq_u16): Likewise.
2940         (vmaxvq_u32): Likewise.
2941         (vmaxnmv_f32): Likewise.
2942         (vmaxnmvq_f32): Likewise.
2943         (vmaxnmvq_f64): Likewise.
2944         (vminv_f32): Likewise.
2945         (vminv_s8): Likewise.
2946         (vminv_s16): Likewise.
2947         (vminv_s32): Likewise.
2948         (vminv_u8): Likewise.
2949         (vminv_u16): Likewise.
2950         (vminv_u32): Likewise.
2951         (vminvq_f32): Likewise.
2952         (vminvq_f64): Likewise.
2953         (vminvq_s8): Likewise.
2954         (vminvq_s16): Likewise.
2955         (vminvq_s32): Likewise.
2956         (vminvq_u8): Likewise.
2957         (vminvq_u16): Likewise.
2958         (vminvq_u32): Likewise.
2959         (vminnmv_f32): Likewise.
2960         (vminnmvq_f32): Likewise.
2961         (vminnmvq_f64): Likewise.
2963 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
2965         * config/aarch64/aarch64-simd.md
2966         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
2967         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
2968         (*aarch64_mul3_elt<mode>): Likewise.
2969         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
2970         (*aarch64_mul3_elt_to_64v2df): Likewise.
2971         (*aarch64_mla_elt<mode>): Likewise.
2972         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
2973         (*aarch64_mls_elt<mode>): Likewise.
2974         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
2975         (*aarch64_fma4_elt<mode>): Likewise.
2976         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
2977         (*aarch64_fma4_elt_to_64v2df): Likewise.
2978         (*aarch64_fnma4_elt<mode>): Likewise.
2979         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
2980         (*aarch64_fnma4_elt_to_64v2df): Likewise.
2981         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
2982         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
2983         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
2984         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
2985         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
2986         (aarch64_sqdmull_lane<mode>_internal): Likewise.
2987         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
2989 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
2991         * config/aarch64/aarch64-simd.md
2992         (aarch64_be_checked_get_lane<mode>): New define_expand.
2993         * config/aarch64/aarch64-simd-builtins.def
2994         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
2995         New builtin definition.
2996         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
2997         Use new safe be builtin.
2999 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
3001         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
3002         New define_insn.
3003         (aarch64_be_st1<mode>): Likewise.
3004         (aarch_ld1<VALL:mode>): Define_expand modified.
3005         (aarch_st1<VALL:mode>): Likewise.
3006         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
3007         (UNSPEC_ST1): Likewise.
3009 2014-01-23  David Holsgrove <david.holsgrove@xilinx.com>
3011         * config/microblaze/microblaze.md: Add trap insn and attribute
3013 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
3015         PR preprocessor/58580
3016         * input.h (location_get_source_line): Take an additional line_size
3017         parameter.
3018         (void diagnostics_file_cache_fini): Declare new function.
3019         * input.c (struct fcache): New type.
3020         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
3021         New static constants.
3022         (diagnostic_file_cache_init, total_lines_num)
3023         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
3024         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
3025         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
3026         (get_next_line, read_next_line, goto_next_line, read_line_num):
3027         New static function definitions.
3028         (diagnostic_file_cache_fini): New function.
3029         (location_get_source_line): Take an additional output line_len
3030         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
3031         read_line_num.
3032         * diagnostic.c (diagnostic_finish): Call
3033         diagnostic_file_cache_fini.
3034         (adjust_line): Take an additional input parameter for the length
3035         of the line, rather than calculating it with strlen.
3036         (diagnostic_show_locus): Adjust the use of
3037         location_get_source_line and adjust_line with respect to their new
3038         signature.  While displaying a line now, do not stop at the first
3039         null byte.  Rather, display the zero byte as a space and keep
3040         going until we reach the size of the line.
3041         * Makefile.in: Add vec.o to OBJS-libcommon
3043 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
3044             Ilya Tocar     <ilya.tocar@intel.com>
3046         * config/i386/avx512fintrin.h (_mm512_kmov): New.
3047         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
3048         (__builtin_ia32_kmov16): Ditto.
3049         * config/i386/i386.md (UNSPEC_KMOV): New.
3050         (kmovw): Ditto.
3052 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
3054         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
3055         (_mm512_storeu_si512): Ditto.
3057 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
3059         PR target/52125
3060         * rtl.h (get_referenced_operands): Declare.
3061         * recog.c (get_referenced_operands): New function.
3062         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
3063         operands have been referenced when recording LO_SUM references.
3065 2014-01-22  David Holsgrove <david.holsgrove@xilinx.com>
3067         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
3069 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
3071         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
3072         Enable for generic and recent AMD targets.
3074 2014-01-22  Jan Hubicka  <hubicka@ucw.cz>
3076         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
3077         ARG_SIZE note when adjustment was eliminated.
3079 2014-01-22  Jeff Law  <law@redhat.com>
3081         PR tree-optimization/59597
3082         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
3083         in file.  Accept new argument REGISTERING and use it to modify
3084         dump output appropriately.
3085         (register_jump_thread): Corresponding changes.
3086         (mark_threaded_blocks): Reinstate code to cancel unprofitable
3087         thread paths involving joiner blocks.  Add code to dump cancelled
3088         jump threading paths.
3090 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
3092         PR rtl-optimization/59477
3093         * lra-constraints.c (inherit_in_ebb): Process call for living hard
3094         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
3096 2014-01-22  Tom Tromey  <tromey@redhat.com>
3098         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
3099         PARAMS.
3100         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
3102 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
3104         PR rtl-optimization/59896
3105         * lra-constraints.c (process_alt_operands): Check unused note for
3106         matched operands of insn with no output reloads.
3108 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
3110         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
3111         (mips_move_from_gpr_cost): Likewise.
3113 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
3115         PR rtl-optimization/59858
3116         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
3117         ira_class_hard_regs_num.
3118         (process_alt_operands): Increase reject for dying matched operand.
3120 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
3122         PR target/59003
3123         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
3124         smaller than size, perform several stores or loads and stores
3125         at dst + count - size to store or copy all of size bytes, rather
3126         than just last modesize bytes.
3128 2014-01-20  DJ Delorie  <dj@redhat.com>
3130         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
3131         that CLOBBERs are REGs before propogating their values.
3133 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
3135         PR middle-end/59789
3136         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
3137         (cgraph_inline_failed_type): New function.
3138         * cgraph.h (DEFCIFCODE): Add type.
3139         (cgraph_inline_failed_type_t): New enum.
3140         (cgraph_inline_failed_type): New prototype.
3141         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
3142         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
3143         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
3144         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
3145         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
3146         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
3147         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
3148         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
3149         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
3150         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
3151         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
3152         OPTIMIZATION_MISMATCH.
3153         * tree-inline.c (expand_call_inline): Emit errors during
3154         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
3156 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
3158         PR target/59685
3159         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
3160         mode attribute in insn output.
3162 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
3164         * output.h (output_constant): Delete.
3165         * varasm.c (output_constant): Make private.
3167 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
3169         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
3171 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
3173         PR middle-end/59860
3174         * tree.h (fold_builtin_strcat): New prototype.
3175         * builtins.c (fold_builtin_strcat): No longer static.  Add len
3176         argument, if non-NULL, don't call c_strlen.  Optimize
3177         directly into __builtin_memcpy instead of __builtin_strcpy.
3178         (fold_builtin_2): Adjust fold_builtin_strcat caller.
3179         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
3181 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
3183         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
3184         for SImode_address_operand operands, having only a REG argument.
3186 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
3188         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
3189         loader name using mbig-endian.
3190         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
3192 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
3194         * doc/invoke.texi (-march): Clarify documentation for AArch64.
3195         (-mtune): Likewise.
3196         (-mcpu): Likewise.
3198 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
3200         * config/aarch64/aarch64-protos.h
3201         (aarch64_cannot_change_mode_class_ptr): Declare.
3202         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
3203         aarch64_cannot_change_mode_class_ptr): New.
3204         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
3205         backend hook aarch64_cannot_change_mode_class.
3207 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
3209         * common/config/aarch64/aarch64-common.c
3210         (aarch64_handle_option): Don't handle any option order logic here.
3211         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
3212         selected_cpu, warn on architecture version mismatch.
3213         (aarch64_override_options): Fix parsing order for option strings.
3215 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
3216             Iain Sandoe  <iain@codesourcery.com>
3218         PR bootstrap/59496
3219         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
3220         warning.  Amend comment to reflect current functionality.
3222 2014-01-20  Richard Biener  <rguenther@suse.de>
3224         PR middle-end/59860
3225         * builtins.c (fold_builtin_strcat): Remove case better handled
3226         by tree-ssa-strlen.c.
3228 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
3230         * config/aarch64/aarch64.opt
3231         (mcpu, march, mtune): Make case-insensitive.
3233 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
3235         PR target/59880
3236         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
3237         if operands[1] is a REG or ZERO_EXTEND of a REG.
3239 2014-01-19  Jan Hubicka  <hubicka@ucw.cz>
3241         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
3243 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
3245         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
3246         long non-pic millicode calls.
3248 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
3250         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
3252 2014-01-19  Kito Cheng  <kito@0xlab.org>
3254         * builtins.c (expand_movstr): Check movstr expand done or fail.
3256 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
3257             H.J. Lu  <hongjiu.lu@intel.com>
3259         PR target/59379
3260         * config/i386/i386.md (*lea<mode>): Zero-extend return register
3261         to DImode for zero-extended addresses.
3263 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
3265         PR rtl-optimization/57763
3266         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
3267         on the new indirect jump_insn and increment LABEL_NUSES (label).
3269 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
3271         PR bootstrap/59580
3272         PR bootstrap/59583
3273         * config.gcc (x86_archs): New variable.
3274         (x86_64_archs): Likewise.
3275         (x86_cpus): Likewise.
3276         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
3277         --with-arch/--with-cpu= options.
3278         Support --with-arch=/--with-cpu={nehalem,westmere,
3279         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
3281 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
3283         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
3284         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
3286 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
3288         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
3290 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
3292         PR target/58944
3293         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
3294         clear cpp_get_options (parse_in)->warn_unused_macros for
3295         ix86_target_macros_internal with cpp_define.
3297 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
3299         * jump.c (delete_related_insns): Keep (use (insn))s.
3300         * reorg.c (redundant_insn): Check for barriers too.
3302 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
3304         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
3306 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
3308         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
3309         call to $$dyncall when TARGET_LONG_CALLS is true.
3311 2014-01-17  Jeff Law  <law@redhat.com>
3313         * ree.c (combine_set_extension): Temporarily disable test for
3314         changing number of hard registers.
3316 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
3318         PR middle-end/58125
3319         * ipa-inline-analysis.c (inline_free_summary):
3320         Do not free summary of aliases.
3322 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
3324         PR middle-end/59706
3325         * gimplify.c (gimplify_expr): Use create_tmp_var
3326         instead of create_tmp_var_raw.  If cond doesn't have
3327         integral type, don't add the IFN_ANNOTATE builtin at all.
3329 2014-01-17  Martin Jambor  <mjambor@suse.cz>
3331         PR ipa/59736
3332         * ipa-cp.c (prev_edge_clone): New variable.
3333         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
3334         Also resize prev_edge_clone vector.
3335         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
3336         (ipcp_edge_removal_hook): New function.
3337         (ipcp_driver): Register ipcp_edge_removal_hook.
3339 2014-01-17  Andrew Pinski <apinski@cavium.com>
3340             Steve Ellcey  <sellcey@mips.com>
3342         PR target/59462
3343         * config/mips/mips.c (mips_print_operand): Check operand mode instead
3344         of operator mode.
3346 2014-01-17  Jeff Law  <law@redhat.com>
3348         PR middle-end/57904
3349         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
3350         so that pass_ccp runs first.
3352 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
3354         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
3355         (ix86_adjust_cost): Use !TARGET_XXX.
3356         (do_reorder_for_imul): Likewise.
3357         (swap_top_of_ready_list): Likewise.
3358         (ix86_sched_reorder): Likewise.
3360 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
3362         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3363         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
3364         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
3365         (intel_memset): New.  Duplicate slm_memset.
3366         (intel_cost): New.  Duplicate slm_cost.
3367         (m_INTEL): New macro.
3368         (processor_target_table): Add "intel".
3369         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
3370         with PROCESSOR_INTEL for "intel".
3371         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
3372         PROCESSOR_SILVERMONT.
3373         (ix86_issue_rate): Likewise.
3374         (ix86_adjust_cost): Likewise.
3375         (ia32_multipass_dfa_lookahead): Likewise.
3376         (swap_top_of_ready_list): Likewise.
3377         (ix86_sched_reorder): Likewise.
3378         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
3379         instead of TARGET_OPT_AGU.
3380         * config/i386/i386.h (TARGET_INTEL): New.
3381         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
3382         (processor_type): Add PROCESSOR_INTEL.
3383         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
3384         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
3386 2014-01-17  Marek Polacek  <polacek@redhat.com>
3388         PR c/58346
3389         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
3390         size is zero.
3392 2014-01-17  Richard Biener  <rguenther@suse.de>
3394         PR tree-optimization/46590
3395         * opts.c (default_options_table): Add entries for
3396         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
3397         all enabled at -O1 but not for -Og.
3398         * common.opt (fbranch-count-reg): Remove Init(1).
3399         (fmove-loop-invariants): Likewise.
3400         (ftree-pta): Likewise.
3402 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
3404         * config/i386/i386.c (ix86_data_alignment): For compatibility with
3405         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
3406         decls to at least the GCC 4.8 used alignments.
3408         PR fortran/59440
3409         * tree-nested.c (convert_nonlocal_reference_stmt,
3410         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
3411         of GIMPLE_BIND stmts, adjust associated decls.
3413 2014-01-17  Richard Biener  <rguenther@suse.de>
3415         PR tree-optimization/46590
3416         * vec.h (vec<>::bseach): New member function implementing
3417         binary search according to C89 bsearch.
3418         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
3419         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
3420         bitmap pointer again.  Make accesses_in_loop a flat array.
3421         (mem_ref_obstack): New global.
3422         (outermost_indep_loop): Adjust for mem_ref->stored changes.
3423         (mark_ref_stored): Likewise.
3424         (ref_indep_loop_p_2): Likewise.
3425         (set_ref_stored_in_loop): New helper function.
3426         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
3427         (memref_free): Adjust.
3428         (record_mem_ref_loc): Simplify.
3429         (gather_mem_refs_stmt): Adjust.
3430         (sort_locs_in_loop_postorder_cmp): New function.
3431         (analyze_memory_references): Sort accesses_in_loop after
3432         loop postorder number.
3433         (find_ref_loc_in_loop_cmp): New function.
3434         (for_all_locs_in_loop): Find relevant cluster of locs in
3435         accesses_in_loop and iterate without recursion.
3436         (execute_sm): Avoid uninit warning.
3437         (struct ref_always_accessed): Simplify.
3438         (ref_always_accessed::operator ()): Likewise.
3439         (ref_always_accessed_p): Likewise.
3440         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
3441         loop postorder numbers here.
3442         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
3443         numbers.
3445 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
3447         PR c++/57945
3448         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
3449         on decls for which assemble_alias has been called.
3451 2014-01-17  Nick Clifton  <nickc@redhat.com>
3453         * config/msp430/msp430.opt: (mcpu): New option.
3454         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
3455         (msp430_option_override): Parse target_cpu.  If the MCU name
3456         matches a generic string, clear target_mcu.
3457         (msp430_attr): Allow numeric interrupt values up to 63.
3458         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
3459         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
3460         option.
3461         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
3462         Add mcpu matches.
3463         * config/msp430/msp430.md (popm): Use %J rather than %I.
3464         (addsi3): Use msp430_nonimmediate_operand for operand 2.
3465         (addhi_cy_i): Use immediate_operand for operand 2.
3466         * doc/invoke.texi: Document -mcpu option.
3468 2014-01-17  Richard Biener  <rguenther@suse.de>
3470         PR rtl-optimization/38518
3471         * df.h (df_analyze_loop): Declare.
3472         * df-core.c: Include cfgloop.h.
3473         (df_analyze_1): Split out main part of df_analyze.
3474         (df_analyze): Adjust.
3475         (loop_inverted_post_order_compute): New function.
3476         (loop_post_order_compute): Likewise.
3477         (df_analyze_loop): New function avoiding whole-function
3478         postorder computes.
3479         * loop-invariant.c (find_defs): Use df_analyze_loop.
3480         (find_invariants): Adjust.
3481         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
3483 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
3485         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
3486         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
3488 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
3490         * ipa-ref.c (ipa_remove_stmt_references): Fix references
3491         traversal when removing references.
3493 2014-01-16  Jan Hubicka  <hubicka@ucw.cz>
3495         PR ipa/59775
3496         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
3498 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
3500         PR middle-end/56791
3501         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
3502         pushing a reload for an autoinc when we had previously reloaded an
3503         inner part of the address.
3505 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
3507         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
3508         field.
3509         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
3510         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
3511         when not giving up or versioning for alias only because of
3512         loop->safelen.
3513         (vect_analyze_data_ref_dependences): Set to true.
3514         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
3515         is a GIMPLE_PHI.
3516         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
3517         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
3518         to the condition.
3520         PR middle-end/58344
3521         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
3523         PR target/59839
3524         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
3525         operand 0 predicate for gathers, use a new pseudo as subtarget.
3527 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
3529         PR middle-end/59609
3530         * lra-constraints.c (process_alt_operands): Add printing debug info.
3531         Check absence of input/output reloads for matched operands too.
3533 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
3535         PR rtl-optimization/59835
3536         * ira.c (ira_init_register_move_cost): Increase cost for
3537         impossible modes.
3539 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
3541         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
3543 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
3545         PR target/59780
3546         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
3547         non-register objects.  Use gen_(high/low)part more consistently.
3548         Fix assertions.
3550 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
3552         PR target/59844
3553         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
3554         endian support, remove tests for WORDS_BIG_ENDIAN.
3555         (p8_mfvsrd_3_<mode>): Likewise.
3556         (reload_gpr_from_vsx<mode>): Likewise.
3557         (reload_gpr_from_vsxsf): Likewise.
3558         (p8_mfvsrd_4_disf): Likewise.
3560 2014-01-16  Richard Biener  <rguenther@suse.de>
3562         PR rtl-optimization/46590
3563         * lcm.c (compute_antinout_edge): Use postorder iteration.
3564         (compute_laterin): Use inverted postorder iteration.
3566 2014-01-16  Nick Clifton  <nickc@redhat.com>
3568         PR middle-end/28865
3569         * varasm.c (output_constant): Return the number of bytes actually
3570         emitted.
3571         (output_constructor_array_range): Update the field size with the
3572         number of bytes emitted by output_constant.
3573         (output_constructor_regular_field): Likewise.  Also do not
3574         complain if the total number of bytes emitted is now greater
3575         than the expected fieldpos.
3576         * output.h (output_constant): Update prototype and descriptive comment.
3578 2014-01-16  Marek Polacek  <polacek@redhat.com>
3580         PR middle-end/59827
3581         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
3582         it is error_mark_node.
3584 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
3586         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
3587         VALID_AVX256_REG_OR_OI_MODE.
3589 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
3591         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
3592         current procedure should be profiled.
3594 2014-01-15  Andrew Pinski  <apinski@cavium.com>
3596         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
3597         of moving from/to the STACK_REG register class.
3599 2014-01-15  Richard Henderson  <rth@redhat.com>
3601         PR debug/54694
3602         * reginfo.c (global_regs_decl): Globalize.
3603         * rtl.h (global_regs_decl): Declare.
3604         * ira.c (do_reload): Diagnose frame_pointer_needed and it
3605         reserved via global_regs.
3607 2014-01-15  Teresa Johnson  <tejohnson@google.com>
3609         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
3611 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
3613         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
3614         and vmulosh rather than call gen_vec_widen_smult_*.
3615         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
3616         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
3617         (vec_widen_smult_even_v16qi): Likewise.
3618         (vec_widen_umult_even_v8hi): Likewise.
3619         (vec_widen_smult_even_v8hi): Likewise.
3620         (vec_widen_umult_odd_v16qi): Likewise.
3621         (vec_widen_smult_odd_v16qi): Likewise.
3622         (vec_widen_umult_odd_v8hi): Likewise.
3623         (vec_widen_smult_odd_v8hi): Likewise.
3624         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
3625         vmuloub rather than call gen_vec_widen_umult_*.
3626         (vec_widen_umult_lo_v16qi): Likewise.
3627         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
3628         vmulosb rather than call gen_vec_widen_smult_*.
3629         (vec_widen_smult_lo_v16qi): Likewise.
3630         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
3631         rather than call gen_vec_widen_umult_*.
3632         (vec_widen_umult_lo_v8hi): Likewise.
3633         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
3634         rather than call gen_vec_widen_smult_*.
3635         (vec_widen_smult_lo_v8hi): Likewise.
3637 2014-01-15  Jeff Law  <law@redhat.com>
3639         PR tree-optimization/59747
3640         * ree.c (find_and_remove_re): Properly handle case where a second
3641         eliminated extension requires widening a copy created for elimination
3642         of a prior extension.
3643         (combine_set_extension): Ensure that the number of hard regs needed
3644         for a destination register does not change when we widen it.
3646 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
3648         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
3649         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
3650         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
3651         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
3652         (avr-*-rtems*): Likewise.
3653         (bfin*-rtems*): Likewise.
3654         (moxie-*-rtems*): Likewise.
3655         (h8300-*-rtems*): Likewise.
3656         (i[34567]86-*-rtems*): Likewise.
3657         (lm32-*-rtems*): Likewise.
3658         (m32r-*-rtems*): Likewise.
3659         (m68k-*-rtems*): Likewise.
3660         (microblaze*-*-rtems*): Likewise.
3661         (mips*-*-rtems*): Likewise.
3662         (powerpc-*-rtems*): Likewise.
3663         (sh-*-rtems*): Likewise.
3664         (sparc-*-rtems*): Likewise.
3665         (sparc64-*-rtems*): Likewise.
3666         (v850-*-rtems*): Likewise.
3667         (m32c-*-rtems*): Likewise.
3669 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
3671         PR rtl-optimization/59511
3672         * ira.c (ira_init_register_move_cost): Use memory costs for some
3673         cases of register move cost calculations.
3674         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
3675         instead of BB frequency.
3676         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
3677         * lra-assigns.c (find_hard_regno_for): Ditto.
3679 2014-01-15  Richard Biener  <rguenther@suse.de>
3681         PR tree-optimization/59822
3682         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
3683         (vectorizable_load): Use it to hoist defs of uses of invariant
3684         loads out of the loop.
3686 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
3687             Kugan Vivekanandarajah  <kuganv@linaro.org>
3689         PR target/59695
3690         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
3691         truncation.
3693 2014-01-15  Richard Biener  <rguenther@suse.de>
3695         PR rtl-optimization/59802
3696         * lcm.c (compute_available): Use inverted postorder to seed
3697         the initial worklist.
3699 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
3701         PR target/59803
3702         * config/s390/s390.c (s390_preferred_reload_class): Don't return
3703         ADDR_REGS for invalid symrefs in non-PIC code.
3705 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
3707         PR other/58712
3708         * builtins.c (determine_block_size): Initialize *probable_max_size
3709         even if len_rtx is CONST_INT.
3711 2014-01-14  Andrew Pinski  <apinski@cavium.com>
3713         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
3714         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
3715         (cortexa53_tunings): Likewise.
3716         (aarch64_sched_issue_rate): New function.
3717         (TARGET_SCHED_ISSUE_RATE): Define.
3719 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
3721         * ira-costs.c (find_costs_and_classes): Add missed
3722         ira_init_register_move_cost_if_necessary.
3724 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
3726         PR target/59787
3727         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
3729 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
3731         PR target/59794
3732         * config/i386/i386.c (type_natural_mode): Add a bool parameter
3733         to indicate if type is used for function return value.  Warn ABI
3734         change if the vector mode isn't available for function return value.
3735         (ix86_function_arg_advance): Pass false to type_natural_mode.
3736         (ix86_function_arg): Likewise.
3737         (ix86_gimplify_va_arg): Likewise.
3738         (function_arg_32): Don't warn ABI change.
3739         (ix86_function_value): Pass true to type_natural_mode.
3740         (ix86_return_in_memory): Likewise.
3741         (ix86_struct_value_rtx): Removed.
3742         (TARGET_STRUCT_VALUE_RTX): Likewise.
3744 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
3746         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
3747         converting a conditional jump into a conditional return.
3749 2014-01-14  Richard Biener  <rguenther@suse.de>
3751         PR tree-optimization/58921
3752         PR tree-optimization/59006
3753         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
3754         hoisting invariant stmts.
3755         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
3756         invariant loads on the preheader edge if possible.
3758 2014-01-14  Joey Ye  <joey.ye@arm.com>
3760         * doc/plugin.texi (Building GCC plugins): Update to C++.
3762 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
3764         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
3765         (_mm_rcp28_round_ss): Ditto.
3766         (_mm_rsqrt28_round_sd): Ditto.
3767         (_mm_rsqrt28_round_ss): Ditto.
3768         (_mm_rcp28_sd): Ditto.
3769         (_mm_rcp28_ss): Ditto.
3770         (_mm_rsqrt28_sd): Ditto.
3771         (_mm_rsqrt28_ss): Ditto.
3772         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
3773         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
3774         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
3775         (IX86_BUILTIN_RCP28SD): Ditto.
3776         (IX86_BUILTIN_RCP28SS): Ditto.
3777         (IX86_BUILTIN_RSQRT28SD): Ditto.
3778         (IX86_BUILTIN_RSQRT28SS): Ditto.
3779         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
3780         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
3781         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
3782         (ix86_expand_special_args_builtin): Expand new FTYPE.
3783         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
3784         (srcp14<mode>): Make insn unary.
3785         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
3786         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
3787         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
3788         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
3789         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
3790         Fix rounding: make it SAE only.
3791         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
3792         Ditto.
3793         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
3794         Ditto.
3795         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
3796         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
3797         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
3798         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
3799         (round_saeonly_mask_scalar_operand4): Ditto.
3800         (round_saeonly_mask_scalar_op3): Ditto.
3801         (round_saeonly_mask_scalar_op4): Ditto.
3803 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3805         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3806         Implement -maltivec=be for vec_insert and vec_extract.
3808 2014-01-10  DJ Delorie  <dj@redhat.com>
3810         * config/msp430/msp430.md (call_internal): Don't allow memory
3811         references with SP as the base register.
3812         (call_value_internal): Likewise.
3813         * config/msp430/constraints.md (Yc): New.  For memory references
3814         that don't use SP as a base register.
3816         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
3817         "an integer without a # prefix"
3818         * config/msp430/msp430.md (epilogue_helper): Use it.
3820 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
3822         PR target/59617
3823         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
3824         AVX512F gather builtins.
3825         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
3826         on gather decls with INTEGER_TYPE masktype.
3827         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
3828         directly into the builtin rather than hoisting it before loop.
3830         PR tree-optimization/59387
3831         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
3832         (scev_const_prop): If folded_casts and type has undefined overflow,
3833         use force_gimple_operand instead of force_gimple_operand_gsi and
3834         for each added stmt if it is assign with
3835         arith_code_with_undefined_signed_overflow, call
3836         rewrite_to_defined_overflow.
3837         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
3838         gimple-fold.h instead.
3839         (arith_code_with_undefined_signed_overflow,
3840         rewrite_to_defined_overflow): Moved to ...
3841         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
3842         rewrite_to_defined_overflow): ... here.  No longer static.
3843         Include gimplify-me.h.
3844         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
3845         rewrite_to_defined_overflow): New prototypes.
3847 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3849         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
3851 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
3853         * builtins.c (get_object_alignment_2): Minor tweak.
3854         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
3856 2014-01-13  Christian Bruel  <christian.bruel@st.com>
3858         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
3859         optimized non constant lengths.
3861 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
3863         PR libgomp/59194
3864         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
3865         load as __atomic_load_N if possible.
3867 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
3869         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
3870         target parameter.
3871         (rs6000_expand_builtin): Adjust call.
3873 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
3875         PR target/58115
3876         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
3877         * config/rs6000/rs6000.c: Include target-globals.h.
3878         (rs6000_set_current_function): Instead of doing target_reinit
3879         unconditionally, use save_target_globals_default_opts and
3880         restore_target_globals.
3882         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
3883         FPSCR.
3884         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
3885         (rs6000_expand_builtin): Handle mffs and mtfsf.
3886         (rs6000_init_builtins): Define mffs and mtfsf.
3887         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
3888         (rs6000_mffs): New pattern.
3889         (rs6000_mtfsf): New pattern.
3891 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
3893         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
3894         Start narrowing with START.  Apply candidate-use pair
3895         and check overall cost in narrowing.
3896         (iv_ca_prune): Pass new argument.
3898 2014-01-10  Jeff Law  <law@redhat.com>
3900         PR middle-end/59743
3901         * ree.c (combine_reaching_defs): Ensure the defining statement
3902         occurs before the extension when optimizing extensions with
3903         different source and destination hard registers.
3905 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
3907         PR ipa/58585
3908         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
3909         vtables into the type inheritance graph.
3911 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
3913         PR rtl-optimization/59754
3914         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
3915         modes in the REGNO != REGNO case.
3917 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3919         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
3921 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
3923         PR tree-optimization/59745
3924         * tree-predcom.c (tree_predictive_commoning_loop): Call
3925         free_affine_expand_cache if giving up because components is NULL.
3927         * target-globals.c (save_target_globals): Allocate < 4KB structs using
3928         GC in payload of target_globals struct instead of allocating them on
3929         the heap and the larger structs separately using GC.
3930         * target-globals.h (struct target_globals): Make regs, hard_regs,
3931         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
3932         of GTY((skip)) and change type to void *.
3933         (reset_target_globals): Cast loads from those fields to corresponding
3934         types.
3936 2014-01-10  Steve Ellcey  <sellcey@mips.com>
3938         PR plugins/59335
3939         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
3940         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
3941         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
3943 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
3945         PR target/59744
3946         * aarch64-modes.def (CC_Zmode): New flags mode.
3947         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
3948         represents an equality.
3949         (aarch64_get_condition_code): Handle CC_Zmode.
3950         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
3952 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
3954         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
3955         extraction in good case.
3957 2014-01-10  Richard Biener  <rguenther@suse.de>
3959         PR tree-optimization/59374
3960         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
3961         checking after SLP discovery.  Mark stmts not participating
3962         in any SLP instance properly.
3964 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3966         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
3967         when handling a SET rtx.
3969 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3971         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
3972         (cortex-a57): Likewise.
3973         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
3975 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3977         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
3978         non-iwmmxt builtins.
3980 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
3982         PR ipa/58252
3983         PR ipa/59226
3984         * ipa-devirt.c record_target_from_binfo): Take as argument
3985         stack of binfos and lookup matching one for virtual inheritance.
3986         (possible_polymorphic_call_targets_1): Update.
3988 2014-01-10  Huacai Chen  <chenhc@lemote.com>
3990         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
3991         kernel strings for Loongson-2E/2F/3A.
3993 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
3995         PR middle-end/59670
3996         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
3997         is_gimple_call before calling gimple_call_internal_p.
3999 2014-01-09  Steve Ellcey  <sellcey@mips.com>
4001         * Makefile.in (TREE_FLOW_H): Remove.
4002         (TREE_SSA_H): Add file names from tree-flow.h.
4003         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
4004         * tree.h: Remove tree-flow.h reference.
4005         * hash-table.h: Remove tree-flow.h reference.
4006         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
4007         reference with tree-ssa-loop.h.
4009 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4011         * doc/invoke.texi: Add -maltivec={be,le} options, and document
4012         default element-order behavior for -maltivec.
4013         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
4014         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
4015         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
4016         when targeting big endian, at least for now.
4017         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
4019 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
4021         PR middle-end/47735
4022         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
4023         var satisfies use_register_for_decl, just take into account type
4024         alignment, rather than decl alignment.
4026         PR tree-optimization/59622
4027         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
4028         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
4029         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
4030         Don't devirtualize for inplace at all.  For targets.length () == 1,
4031         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
4033 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
4035         * config/i386/i386.md (cpu): Remove the unused btver1.
4037 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
4039         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
4041 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
4043         PR target/58115
4044         * tree-core.h (struct target_globals): New forward declaration.
4045         (struct tree_target_option): Add globals field.
4046         * tree.h (TREE_TARGET_GLOBALS): Define.
4047         (prepare_target_option_nodes_for_pch): New prototype.
4048         * target-globals.h (struct target_globals): Define even if
4049         !SWITCHABLE_TARGET.
4050         * tree.c (prepare_target_option_node_for_pch,
4051         prepare_target_option_nodes_for_pch): New functions.
4052         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
4053         * config/i386/i386.c: Include target-globals.h.
4054         (ix86_set_current_function): Instead of doing target_reinit
4055         unconditionally, use save_target_globals_default_opts and
4056         restore_target_globals.
4058 2014-01-09  Richard Biener  <rguenther@suse.de>
4060         PR tree-optimization/59715
4061         * tree-cfg.h (split_critical_edges): Declare.
4062         * tree-cfg.c (split_critical_edges): Export.
4063         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
4065 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
4067         * cfgexpand.c (expand_stack_vars): Optionally disable
4068         asan stack protection.
4069         (expand_used_vars): Likewise.
4070         (partition_stack_vars): Likewise.
4071         * asan.c (asan_emit_stack_protection): Optionally disable
4072         after return stack usage.
4073         (instrument_derefs): Optionally disable memory access instrumentation.
4074         (instrument_builtin_call): Likewise.
4075         (instrument_strlen_call): Likewise.
4076         (asan_protect_global): Optionally disable global variables protection.
4077         * doc/invoke.texi: Added doc for new options.
4078         * params.def: Added new options.
4079         * params.h: Likewise.
4081 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
4083         PR rtl-optimization/59724
4084         * ifcvt.c (cond_exec_process_if_block): Don't call
4085         flow_find_head_matching_sequence with 0 longest_match.
4086         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
4087         non-active insns if !stop_after.
4088         (try_head_merge_bb): Revert 2014-01-07 changes.
4090 2014-01-08  Jeff Law  <law@redhat.com>
4092         * ree.c (get_sub_rtx): New function, extracted from...
4093         (merge_def_and_ext): Here.
4094         (combine_reaching_defs): Use get_sub_rtx.
4096 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
4098         * cgraph.h (varpool_variable_node): Do not choke on null node.
4100 2014-01-08  Catherine Moore  <clm@codesourcery.com>
4102         * config/mips/mips.md (simple_return): Attempt to use JRC
4103         for microMIPS.
4104         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
4106 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
4108         PR rtl-optimization/59137
4109         * reorg.c (steal_delay_list_from_target): Call update_block for
4110         elided insns.
4111         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
4113 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
4115         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4116         two duplicate entries.
4118 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
4120         Revert:
4121         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
4123         * config/mips/mips.c (mips_truncated_op_cost): New function.
4124         (mips_rtx_costs): Adjust test for BADDU.
4125         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
4127         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
4129         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
4130         (*baddu_si): ...this new pattern.
4132 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
4134         PR ipa/59722
4135         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
4137 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4139         PR middle-end/57748
4140         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
4141         inner_reference_p.
4142         (expand_expr, expand_normal): Adjust.
4143         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
4144         inner_reference_p. Use inner_reference_p to expand inner references.
4145         (store_expr): Adjust.
4146         * cfgexpand.c (expand_call_stmt): Adjust.
4148 2014-01-08  Rong Xu  <xur@google.com>
4150         * gcov-io.c (gcov_var): Move from gcov-io.h.
4151         (gcov_position): Ditto.
4152         (gcov_is_error): Ditto.
4153         (gcov_rewrite): Ditto.
4154         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
4155         only part to libgcc/libgcov.h.
4157 2014-01-08  Marek Polacek  <polacek@redhat.com>
4159         PR middle-end/59669
4160         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
4162 2014-01-08  Marek Polacek  <polacek@redhat.com>
4164         PR sanitizer/59667
4165         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
4167 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
4169         PR rtl-optimization/59649
4170         * stor-layout.c (get_mode_bounds): For BImode return
4171         0 and STORE_FLAG_VALUE.
4173 2014-01-08  Richard Biener  <rguenther@suse.de>
4175         PR middle-end/59630
4176         * gimple.h (is_gimple_builtin_call): Remove.
4177         (gimple_builtin_call_types_compatible_p): New.
4178         (gimple_call_builtin_p): New overload.
4179         * gimple.c (is_gimple_builtin_call): Remove.
4180         (validate_call): Rename to ...
4181         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
4182         check return types.
4183         (validate_type): New static function.
4184         (gimple_call_builtin_p): New overload and adjust.
4185         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
4186         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
4187         (gimple_fold_stmt_to_constant_1): Likewise.
4188         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
4190 2014-01-08  Richard Biener  <rguenther@suse.de>
4192         PR middle-end/59471
4193         * gimplify.c (gimplify_expr): Gimplify register-register type
4194         VIEW_CONVERT_EXPRs to separate stmts.
4196 2014-01-07  Jeff Law  <law@redhat.com>
4198         PR middle-end/53623
4199         * ree.c (combine_set_extension): Handle case where source
4200         and destination registers in an extension insn are different.
4201         (combine_reaching_defs): Allow source and destination registers
4202         in extension to be different under limited circumstances.
4203         (add_removable_extension): Remove restriction that the
4204         source and destination registers in the extension are the same.
4205         (find_and_remove_re): Emit a copy from the extension's
4206         destination to its source after the defining insn if
4207         the source and destination registers are different.
4209         PR middle-end/59285
4210         * ifcvt.c (merge_if_block): If we are merging a block with more than
4211         one successor with a block with no successors, remove any BARRIER
4212         after the second block.
4214 2014-01-07  Dan Xio Qiang <ziyan01@163.com>
4216         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
4218 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
4220         PR target/59652
4221         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
4222         for 14-bit register offsets when INT14_OK_STRICT is false.
4224 2014-01-07  Roland Stigge  <stigge@antcom.de>
4225             Michael Meissner  <meissner@linux.vnet.ibm.com>
4227         PR 57386/target
4228         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
4229         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
4231 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
4233         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
4234         -mcpu.
4236 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
4238         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
4239         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
4240         rtx is const0_rtx or not.
4242 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
4244         PR target/58115
4245         * target-globals.c (save_target_globals): Remove this_fn_optab
4246         handling.
4247         * toplev.c: Include optabs.h.
4248         (target_reinit): Temporarily restore the global options if another
4249         set of options are in force.
4251 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
4253         PR rtl-optimization/58668
4254         * cfgcleanup.c (flow_find_cross_jump): Don't count
4255         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
4256         to determine what is counted.
4257         (flow_find_head_matching_sequence): Use active_insn_p to determine
4258         what is counted.
4259         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
4260         counting change.
4261         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
4262         determine what is counted.
4264         PR tree-optimization/59643
4265         * tree-predcom.c (split_data_refs_to_components): If one dr is
4266         read and one write, determine_offset fails and the write isn't
4267         in the bad component, just put the read into the bad component.
4269 2014-01-07  Mike Stump  <mikestump@comcast.net>
4270             Jakub Jelinek  <jakub@redhat.com>
4272         PR pch/59436
4273         * tree-core.h (struct tree_optimization_option): Change optabs
4274         type from unsigned char * to void *.
4275         * optabs.c (init_tree_optimization_optabs): Adjust
4276         TREE_OPTIMIZATION_OPTABS initialization.
4278 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
4280         PR target/59644
4281         * config/i386/i386.h (struct machine_function): Add
4282         no_drap_save_restore field.
4283         * config/i386/i386.c (ix86_save_reg): Use
4284         !cfun->machine->no_drap_save_restore instead of
4285         crtl->stack_realign_needed.
4286         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
4287         this function clears frame_pointer_needed.  Set
4288         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
4289         and DRAP reg is needed.
4291 2014-01-06  Marek Polacek  <polacek@redhat.com>
4293         PR c/57773
4294         * doc/implement-c.texi: Mention that other integer types are
4295         permitted as bit-field types in strictly conforming mode.
4297 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
4299         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
4300         is newly allocated.
4302 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
4304         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
4306 2014-01-06  Martin Jambor  <mjambor@suse.cz>
4308         PR ipa/59008
4309         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
4310         to int.
4311         * ipa-prop.c (ipa_print_node_params): Fix indentation.
4313 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
4315         PR debug/59350
4316         PR debug/59510
4317         * var-tracking.c (add_stores): Preserve the value of the source even if
4318         we don't record the store.
4320 2014-01-06  Terry Guo  <terry.guo@arm.com>
4322         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
4324 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
4326         PR bootstrap/59541
4327         * config/darwin.c (darwin_function_section): Adjust return values to
4328         correspond to optimisation changes made in r206070.
4330 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
4332         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
4333         from prefetch_block tune setting.
4334         (nocona_cost): Correct size of prefetch block to 64.
4336 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
4338         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
4339         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
4340         used to save the static chain register in the computation of the offset
4341         from which the FP registers need to be restored.
4343 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
4345         PR tree-optimization/59519
4346         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
4347         ICE if get_current_def (current_new_name) is already non-NULL, as long
4348         as it is a phi result of some other phi in *new_exit_bb that has
4349         the same argument.
4351         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
4352         or vmovdqu* for misaligned_operand.
4353         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
4354         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
4355         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
4356         aligned_mem for AVX512F masked aligned load and store builtins and for
4357         non-temporal moves.
4359 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
4361         PR tree-optimization/59651
4362         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
4363         Address range for negative step should be added by TYPE_SIZE_UNIT.
4365 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
4367         * config/m68k/m68k.c (handle_move_double): Handle pushes with
4368         overlapping registers also for registers other than the stack pointer.
4370 2014-01-03  Marek Polacek  <polacek@redhat.com>
4372         PR other/59661
4373         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
4374         __builtin_FILE.
4376 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
4378         PR target/59625
4379         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
4380         asm goto as jump.
4382         * config/i386/i386.md (MODE_SIZE): New mode attribute.
4383         (push splitter): Use <P:MODE_SIZE> instead of
4384         GET_MODE_SIZE (<P:MODE>mode).
4385         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
4386         (mov -1, reg peephole2): Likewise.
4387         * config/i386/sse.md (*mov<mode>_internal,
4388         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
4389         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
4390         *<code><mode>3, *andnot<mode>3<mask_name>,
4391         <mask_codefor><code><mode>3<mask_name>): Likewise.
4392         * config/i386/subst.md (mask_mode512bit_condition,
4393         sd_mask_mode512bit_condition): Likewise.
4395 2014-01-02  Xinliang David Li  <davidxl@google.com>
4397         PR tree-optimization/59303
4398         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
4399         (dump_predicates): Better output format.
4400         (pred_equal_p): New function.
4401         (is_neq_relop_p): Ditto.
4402         (is_neq_zero_form_p): Ditto.
4403         (pred_expr_equal_p): Ditto.
4404         (pred_neg_p): Ditto.
4405         (simplify_pred): Ditto.
4406         (simplify_preds_2): Ditto.
4407         (simplify_preds_3): Ditto.
4408         (simplify_preds_4): Ditto.
4409         (simplify_preds): Ditto.
4410         (push_pred): Ditto.
4411         (push_to_worklist): Ditto.
4412         (get_pred_info_from_cmp): Ditto.
4413         (is_degenerated_phi): Ditto.
4414         (normalize_one_pred_1): Ditto.
4415         (normalize_one_pred): Ditto.
4416         (normalize_one_pred_chain): Ditto.
4417         (normalize_preds): Ditto.
4418         (normalize_cond_1): Remove function.
4419         (normalize_cond): Ditto.
4420         (is_gcond_subset_of): Ditto.
4421         (is_subset_of_any): Ditto.
4422         (is_or_set_subset_of): Ditto.
4423         (is_and_set_subset_of): Ditto.
4424         (is_norm_cond_subset_of): Ditto.
4425         (pred_chain_length_cmp): Ditto.
4426         (convert_control_dep_chain_into_preds): Type change.
4427         (find_predicates): Ditto.
4428         (find_def_preds): Ditto.
4429         (destroy_predicates_vecs): Ditto.
4430         (find_matching_predicates_in_rest_chains): Ditto.
4431         (use_pred_not_overlap_with_undef_path_pred): Ditto.
4432         (is_pred_expr_subset): Ditto.
4433         (is_pred_chain_subset_of): Ditto.
4434         (is_included_in): Ditto.
4435         (is_superset_of): Ditto.
4437 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
4439         Update copyright years.
4441 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
4443         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
4444         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
4445         config/arc/arc.md, config/arc/arc.opt,
4446         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
4447         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
4448         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
4449         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
4450         config/linux-protos.h, config/linux.c, config/winnt-c.c,
4451         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
4452         vtable-verify.c, vtable-verify.h: Use the standard form for the
4453         copyright notice.
4455 2014-01-02  Tobias Burnus  <burnus@net-b.de>
4457         * gcc.c (process_command): Update copyright notice dates.
4458         * gcov-dump.c: Ditto.
4459         * gcov.c: Ditto.
4460         * doc/cpp.texi: Bump @copying's copyright year.
4461         * doc/cppinternals.texi: Ditto.
4462         * doc/gcc.texi: Ditto.
4463         * doc/gccint.texi: Ditto.
4464         * doc/gcov.texi: Ditto.
4465         * doc/install.texi: Ditto.
4466         * doc/invoke.texi: Ditto.
4468 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
4470         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
4472 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
4474         * config/i386/sse.md (*mov<mode>_internal): Guard
4475         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
4477         PR rtl-optimization/59647
4478         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
4479         new_rtx into UNSIGNED_FLOAT rtxes.
4481 Copyright (C) 2014 Free Software Foundation, Inc.
4483 Copying and distribution of this file, with or without modification,
4484 are permitted in any medium without royalty provided the copyright
4485 notice and this notice are preserved.