* config/msp430/msp430.opt (-minrt): New.
[official-gcc.git] / gcc / ChangeLog
blobec72eb2014963aac5e5d23a5efa1b5c4f44b933e
1 2014-01-29  DJ Delorie  <dj@redhat.com>
3         * config/msp430/msp430.opt (-minrt): New.
4         * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
5         if -minrt given.
6         (ENDFILE_SPEC): Likewise.
8 2014-01-29  Jan Hubicka  <jh@suse.cz>
10         * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
11         (estimate_function_body_sizes): Use it.
13 2014-01-29  Paolo Carlini  <paolo.carlini@oracle.com>
15         PR c++/58561
16         * dwarf2out.c (is_cxx_auto): New.
17         (is_base_type): Use it.
18         (gen_type_die_with_usage): Likewise.
20 2014-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):  Use
23         CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
24         * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
25         -maltivec=be with LE targets.
26         (vsx_mergeh_<mode>): Likewise.
27         * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New
28         unspecs.
29         (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
30         (altivec_vmrghb): Replace with define_expand and new
31         *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE
32         targets.
33         (altivec_vmrghb_direct): New define_insn.
34         (altivec_vmrghh): Replace with define_expand and new
35         *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE
36         targets.
37         (altivec_vmrghh_direct): New define_insn.
38         (altivec_vmrghw): Replace with define_expand and new
39         *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE
40         targets.
41         (altivec_vmrghw_direct): New define_insn.
42         (*altivec_vmrghsf): Adjust for endianness.
43         (altivec_vmrglb): Replace with define_expand and new
44         *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE
45         targets.
46         (altivec_vmrglb_direct): New define_insn.
47         (altivec_vmrglh): Replace with define_expand and new
48         *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE
49         targets.
50         (altivec_vmrglh_direct): New define_insn.
51         (altivec_vmrglw): Replace with define_expand and new
52         *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE
53         targets.
54         (altivec_vmrglw_direct): New define_insn.
55         (*altivec_vmrglsf): Adjust for endianness.
56         (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
57         (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
58         (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
59         (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
60         (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
61         (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
62         (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
63         (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
65 2014-01-29  Marcus Shawcroft  <marcus.shawcroft@arm.com>
67         * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
68         (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
69         whitespace.
71 2014-01-29  Richard Biener  <rguenther@suse.de>
73         PR tree-optimization/58742
74         * tree-ssa-forwprop.c (associate_pointerplus): Rename to
75         associate_pointerplus_align.
76         (associate_pointerplus_diff): New function.
77         (associate_pointerplus): Likewise.  Call associate_pointerplus_align
78         and associate_pointerplus_diff.
80 2014-01-29  Richard Biener  <rguenther@suse.de>
82         * lto-streamer.h (LTO_major_version): Bump to 3.
83         (LTO_minor_version): Reset to 0.
85 2014-01-29  Renlin Li  <Renlin.Li@arm.com>
87         * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
88         * config/arm/arm.c (FL_FOR_ARCH7VE): New.
89         (arm_file_start): Generate correct asm header for armv7ve.
90         * config/arm/bpabi.h: Add multilib support for armv7ve.
91         * config/arm/driver-arm.c: Change the architectures of cortex-a7
92         and cortex-a15 to armv7ve.
93         * config/arm/t-aprofile: Add multilib support for armv7ve.
94         * doc/invoke.texi: Document -march=armv7ve.
96 2014-01-29  Richard Biener  <rguenther@suse.de>
98         PR tree-optimization/58742
99         * tree-ssa-forwprop.c (associate_plusminus): Return true
100         if we changed sth, defer EH cleanup to ...
101         (ssa_forward_propagate_and_combine): ... here.  Call simplify_mult.
102         (simplify_mult): New function.
104 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
106         PR middle-end/59917
107         PR tree-optimization/59920
108         * tree.c (build_common_builtin_nodes): Remove
109         __builtin_setjmp_dispatcher initialization.
110         * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
111         * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
112         instead of gsi_after_labels + manually skipping debug stmts.
113         Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
114         ignore bbs with IFN_ABNORMAL_DISPATCHER.
115         * tree-inline.c (copy_edges_for_bb): Remove
116         can_make_abnormal_goto argument, instead add abnormal_goto_dest
117         argument.  Ignore computed_goto_p stmts.  Don't call
118         make_abnormal_goto_edges.  If a call might need abnormal edges
119         for non-local gotos, see if it already has an edge to
120         IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
121         with true argument, don't do anything then, otherwise add
122         EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
123         (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
124         caller.
125         * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
126         (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
127         (lower_stmt): Don't set data->calls_builtin_setjmp.
128         (lower_builtin_setjmp): Adjust comment.
129         * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
130         * tree-cfg.c (found_computed_goto): Remove.
131         (factor_computed_gotos): Remove.
132         (make_goto_expr_edges): Return bool, true for computed gotos.
133         Don't call make_abnormal_goto_edges.
134         (build_gimple_cfg): Don't set found_computed_goto, don't call
135         factor_computed_gotos.
136         (computed_goto_p): No longer static.
137         (make_blocks): Don't set found_computed_goto.
138         (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
139         (make_edges): If make_goto_expr_edges returns true, push bb
140         into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
141         instead of calling make_abnormal_goto_edges push bb into ab_edge_call
142         vector.  Record mapping between bbs and OpenMP regions if there
143         are any, adjust make_gimple_omp_edges caller.  Call
144         handle_abnormal_edges.
145         (make_abnormal_goto_edges): Remove.
146         * tree-cfg.h (make_abnormal_goto_edges): Remove.
147         (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
148         * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
149         * builtins.c (expand_builtin): Don't handle
150         BUILT_IN_SETJMP_DISPATCHER.
151         * internal-fn.def (ABNORMAL_DISPATCHER): New.
152         * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
153         filling *region also set *region_idx to (*region)->entry->index.
155         PR other/58712
156         * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
157         For REGs set ORIGINAL_REGNO.
159 2014-01-29  Bingfeng Mei  <bmei@broadcom.com>
161         * doc/md.texi: Mention that a target shouldn't implement 
162         vec_widen_(s|u)mul_even/odd pair if it is less efficient
163         than hi/lo pair.
165 2014-01-29  Jakub Jelinek  <jakub@redhat.com>
167         PR tree-optimization/59594
168         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
169         a copy of the datarefs vector rather than the vector itself.
171 2014-01-28  Jason Merrill  <jason@redhat.com>
173         PR c++/53756
174         * dwarf2out.c (auto_die): New static.
175         (gen_type_die_with_usage): Handle C++1y 'auto'.
176         (gen_subprogram_die): If in-class DIE had 'auto', emit type again
177         on definition.
179 2014-01-28  H.J. Lu  <hongjiu.lu@intel.com>
181         PR target/59672
182         * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
183         (SPEC_X32): Likewise.
184         (SPEC_64): Likewise.
185         * config/i386/i386.c (ix86_option_override_internal): Turn off
186         OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
187         for TARGET_16BIT.
188         (x86_file_start): Output .code16gcc for TARGET_16BIT.
189         * config/i386/i386.h (TARGET_16BIT): New macro.
190         (TARGET_16BIT_P): Likewise.
191         * config/i386/i386.opt: Add m16.
192         * doc/invoke.texi: Document -m16.
194 2014-01-28  Jakub Jelinek  <jakub@redhat.com>
196         PR preprocessor/59935
197         * input.c (location_get_source_line): Bail out on when line number
198         is zero, and test the return value of
199         lookup_or_add_file_to_cache_tab.
201 2014-01-28  Richard Biener  <rguenther@suse.de>
203         PR tree-optimization/58742
204         * tree-ssa-forwprop.c (associate_plusminus): Handle
205         pointer subtraction of the form (T)(P + A) - (T)P.
207 2014-01-28  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
209         * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
210         at const_int_cost.
212 2014-01-28  Richard Biener  <rguenther@suse.de>
214         Revert
215         2014-01-28  Richard Biener  <rguenther@suse.de>
217         PR rtl-optimization/45364
218         PR rtl-optimization/59890
219         * var-tracking.c (local_get_addr_clear_given_value): Handle
220         already cleared slot.
221         (val_reset): Handle not allocated local_get_addr_cache.
222         (vt_find_locations): Use post-order on the inverted CFG.
224 2014-01-28  Richard Biener  <rguenther@suse.de>
226         * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps):
227         Remove.
229 2014-01-28  Richard Biener  <rguenther@suse.de>
231         PR rtl-optimization/45364
232         PR rtl-optimization/59890
233         * var-tracking.c (local_get_addr_clear_given_value): Handle
234         already cleared slot.
235         (val_reset): Handle not allocated local_get_addr_cache.
236         (vt_find_locations): Use post-order on the inverted CFG.
238 2014-01-28  Alan Modra  <amodra@gmail.com>
240         * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
241         * configure.ac <recursive call for build != host>: Define
242         GENERATOR_FILE.  Comment.  Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
243         and LD_FOR_BUILD too.
244         * configure: Regenerate.
246 2014-01-27  Allan Sandfeld Jensen  <sandfeld@kde.org>
248         * config/i386/i386.c (get_builtin_code_for_version): Separate
249         Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
250         Broadwell from Haswell.
252 2014-01-27  Steve Ellcey  <sellcey@mips.com>
254         * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
255         Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
256         * config/mips/mips.c (mips_option_override): Change setting
257         of TARGET_DSP.
258         * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
259         * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
260         Change from Mask to Var.
262 2014-01-27  Jeff Law  <law@redhat.com>
264         * ipa-inline.c (inline_small_functions): Fix typo.
266 2014-01-27  Ilya Tocar  <ilya.tocar@intel.com>
268         * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
269         (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
270         (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
271         (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
272         (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
273         (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
274         (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
275         (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
276         (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
277         (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
278         (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
279         (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
280         (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
281         (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
282         (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
283         (_mm512_storeu_epi64): Ditto.
284         (_mm512_cmpge_epi32_mask): Ditto.
285         (_mm512_cmpge_epu32_mask): Ditto.
286         (_mm512_cmpge_epi64_mask): Ditto.
287         (_mm512_cmpge_epu64_mask): Ditto.
288         (_mm512_cmple_epi32_mask): Ditto.
289         (_mm512_cmple_epu32_mask): Ditto.
290         (_mm512_cmple_epi64_mask): Ditto.
291         (_mm512_cmple_epu64_mask): Ditto.
292         (_mm512_cmplt_epi32_mask): Ditto.
293         (_mm512_cmplt_epu32_mask): Ditto.
294         (_mm512_cmplt_epi64_mask): Ditto.
295         (_mm512_cmplt_epu64_mask): Ditto.
296         (_mm512_cmpneq_epi32_mask): Ditto.
297         (_mm512_cmpneq_epu32_mask): Ditto.
298         (_mm512_cmpneq_epi64_mask): Ditto.
299         (_mm512_cmpneq_epu64_mask): Ditto.
300         (_mm512_expand_pd): Ditto.
301         (_mm512_expand_ps): Ditto.
302         * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
303         VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
304         VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
305         * config/i386/i386.c (ix86_builtins): Add
306         IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
307         IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
308         IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
309         IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
310         IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
311         IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
312         IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
313         IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
314         IX86_BUILTIN_PMOVUSQW512_MEM.
315         (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
316         __builtin_ia32_pmovsqd512mem_mask,
317         __builtin_ia32_pmovqd512mem_mask,
318         __builtin_ia32_pmovusqw512mem_mask,
319         __builtin_ia32_pmovsqw512mem_mask,
320         __builtin_ia32_pmovqw512mem_mask,
321         __builtin_ia32_pmovusdw512mem_mask,
322         __builtin_ia32_pmovsdw512mem_mask,
323         __builtin_ia32_pmovdw512mem_mask,
324         __builtin_ia32_pmovqb512mem_mask,
325         __builtin_ia32_pmovusqb512mem_mask,
326         __builtin_ia32_pmovsqb512mem_mask,
327         __builtin_ia32_pmovusdb512mem_mask,
328         __builtin_ia32_pmovsdb512mem_mask,
329         __builtin_ia32_pmovdb512mem_mask.
330         (bdesc_args): Add __builtin_ia32_expanddf512,
331         __builtin_ia32_expandsf512.
332         (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
333         VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
334         VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
335         * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
336         (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
337         (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
338         (avx512f_<code>v8div16qi2_mask_store): This.
339         (avx512f_expand<mode>): New.
341 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
343         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
344         New.
345         (_mm512_mask_prefetch_i64gather_pd): Ditto.
346         (_mm512_prefetch_i32scatter_pd): Ditto.
347         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
348         (_mm512_prefetch_i64scatter_pd): Ditto.
349         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
350         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
351         (_mm512_mask_prefetch_i64gather_ps): Ditto.
352         (_mm512_prefetch_i32scatter_ps): Ditto.
353         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
354         (_mm512_prefetch_i64scatter_ps): Ditto.
355         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
356         * config/i386/i386-builtin-types.def: Define
357         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
358         and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
359         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
360         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
361         IX86_BUILTIN_SCATTERPFQPD.
362         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
363         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
364         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
365         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
366         __builtin_ia32_scatterpfqps.
367         (ix86_expand_builtin): Expand new built-ins.
368         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
369         fix memory access data type.
370         (*avx512pf_gatherpf<mode>_mask): Ditto.
371         (*avx512pf_gatherpf<mode>): Ditto.
372         (avx512pf_scatterpf<mode>): Ditto.
373         (*avx512pf_scatterpf<mode>_mask): Ditto.
374         (*avx512pf_scatterpf<mode>): Ditto.
375         (GATHER_SCATTER_SF_MEM_MODE): New.
376         (avx512pf_gatherpf<mode>df): Ditto.
377         (*avx512pf_gatherpf<mode>df_mask): Ditto.
378         (*avx512pf_scatterpf<mode>df): Ditto.
380 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
382         PR bootstrap/59934
383         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
384         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
385         reached.
387 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
389         * common/config/arm/arm-common.c
390         (arm_rewrite_mcpu): Handle multiple names.
391         * config/arm/arm.h
392         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
394 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
396         * gimple-builder.h (create_gimple_tmp): Delete.
398 2014-01-27  Christian Bruel  <christian.bruel@st.com>
400         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
401         words comparisons.
403 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
405         * config/pa/pa.md (call): Generate indirect long calls to non-local
406         functions when outputing 32-bit code.
407         (call_value): Likewise except for special call to buggy powf function.
409         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
410         portable runtime and PIC indirect calls.
411         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
412         and PIC call sequences.  Use ldo instead of blr to set return register
413         in PIC call sequence.
415 2014-01-25  Walter Lee  <walt@tilera.com>
417         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
418         avoid clobbering a live register.
420 2014-01-25  Walter Lee  <walt@tilera.com>
422         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
423         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
424         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
425         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
427 2014-01-25  Walter Lee  <walt@tilera.com>
429         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
430         arguments on even registers.
431         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
432         STACK_BOUNDARY.
433         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
434         (BIGGEST_ALIGNMENT): Ditto.
435         (BIGGEST_FIELD_ALIGNMENT): Ditto.
437 2014-01-25  Walter Lee  <walt@tilera.com>
439         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
440         insns before bundling.
441         * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
443 2014-01-25  Walter Lee  <walt@tilera.com>
445         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
446         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
447         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
449 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
451         * config/mips/constraints.md (kl): Delete.
452         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
453         define expands, using...
454         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
455         instructions for MIPS16.
456         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
457         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
459 2014-01-25  Walter Lee  <walt@tilera.com>
461         * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
462         (clzdi2): Ditto.
463         (ffsdi2): Ditto.
465 2014-01-25  Walter Lee  <walt@tilera.com>
467         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
468         (TARGET_EXPAND_TO_RTL_HOOK): Define.
470 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
472         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
473         Handle XOR.
475 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
477         * print-rtl.c (in_call_function_usage): New var.
478         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
479         EXPR_LIST mode as mode and not as reg note name.
481         PR middle-end/59561
482         * cfgloopmanip.c (copy_loop_info): If
483         loop->warned_aggressive_loop_optimizations, make sure
484         the flag is set in target loop too.
486 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
488         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
489         flag_cilkplus.
490         * builtins.def: Likewise.
491         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
492         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
493         * ira.c (ira_setup_eliminable_regset): Likewise.
494         * omp-low.c (gate_expand_omp): Likewise.
495         (execute_lower_omp): Likewise.
496         (diagnose_sb_0): Likewise.
497         (gate_diagnose_omp_blocks): Likewise.
498         (simd_clone_clauses_extract): Likewise.
499         (gate): Likewise.
501 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
503         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
504         correction for little endian...
505         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
506         here.
508 2014-01-24  Jeff Law  <law@redhat.com>
510         PR tree-optimization/59919
511         * tree-vrp.c (find_assert_locations_1): Do not register asserts
512         for non-returning calls.
514 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
516         * common/config/aarch64/aarch64-common.c
517         (aarch64_rewrite_mcpu): Handle multiple names.
518         * config/aarch64/aarch64.h
519         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
521 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
523         * input.c (add_file_to_cache_tab): Handle the case where fopen
524         returns NULL.
526 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
528         PR target/59929
529         * config/i386/i386.md (pushsf splitter): Get stack adjustment
530         from push operand if code of push isn't PRE_DEC.
532 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
534         PR target/59909
535         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
536         -mquad-memory-atomic.  Update -mquad-memory documentation to say
537         it is only used for non-atomic loads/stores.
539         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
540         -mquad-memory or -mquad-memory-atomic switches.
542         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
543         -mquad-memory-atomic to ISA 2.07 support.
545         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
546         to separate support of normal quad word memory operations (ldq, stq)
547         from the atomic quad word memory operations.
549         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
550         support to separate non-atomic quad word operations from atomic
551         quad word operations.  Disable non-atomic quad word operations in
552         little endian mode so that we don't have to swap words after the
553         load and before the store.
554         (quad_load_store_p): Add comment about atomic quad word support.
555         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
556         options printed with -mdebug=reg.
558         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
559         -mquad-memory-atomic as the test for whether we have quad word
560         atomic instructions.
561         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
562         or -mp8-vector are used, allow byte/half-word atomic operations.
564         * config/rs6000/sync.md (load_lockedti): Insure that the address
565         is a proper indexed or indirect address for the lqarx instruction.
566         On little endian systems, swap the hi/lo registers after the lqarx
567         instruction.
568         (load_lockedpti): Use indexed_or_indirect_operand predicate to
569         insure the address is valid for the lqarx instruction.
570         (store_conditionalti): Insure that the address is a proper indexed
571         or indirect address for the stqcrx. instruction.  On little endian
572         systems, swap the hi/lo registers before doing the stqcrx.
573         instruction.
574         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
575         insure the address is valid for the stqcrx. instruction.
577         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
578         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
579         type of quad memory support is available.
581 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
583         PR regression/59915
584         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
585         there is a danger of looping.
587 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
589         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
590         force flag_ira_loop_pressure if set via command line.
592 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
594         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
595         (ashr_simd): New builtin handling DI mode.
596         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
597         (aarch64_sshr_simddi): New match pattern.
598         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
599         (vshrd_n_s64): Likewise.
600         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
602 2014-01-23  Nick Clifton  <nickc@redhat.com>
604         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
605         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
606         favour of mcu specific scripts.
607         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
608         430x multilibs.
610 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
611             Alex Velenko  <Alex.Velenko@arm.com>
613         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
614         (vaddv_s16): Likewise.
615         (vaddv_s32): Likewise.
616         (vaddv_u8): Likewise.
617         (vaddv_u16): Likewise.
618         (vaddv_u32): Likewise.
619         (vaddvq_s8): Likewise.
620         (vaddvq_s16): Likewise.
621         (vaddvq_s32): Likewise.
622         (vaddvq_s64): Likewise.
623         (vaddvq_u8): Likewise.
624         (vaddvq_u16): Likewise.
625         (vaddvq_u32): Likewise.
626         (vaddvq_u64): Likewise.
627         (vaddv_f32): Likewise.
628         (vaddvq_f32): Likewise.
629         (vaddvq_f64): Likewise.
630         (vmaxv_f32): Likewise.
631         (vmaxv_s8): Likewise.
632         (vmaxv_s16): Likewise.
633         (vmaxv_s32): Likewise.
634         (vmaxv_u8): Likewise.
635         (vmaxv_u16): Likewise.
636         (vmaxv_u32): Likewise.
637         (vmaxvq_f32): Likewise.
638         (vmaxvq_f64): Likewise.
639         (vmaxvq_s8): Likewise.
640         (vmaxvq_s16): Likewise.
641         (vmaxvq_s32): Likewise.
642         (vmaxvq_u8): Likewise.
643         (vmaxvq_u16): Likewise.
644         (vmaxvq_u32): Likewise.
645         (vmaxnmv_f32): Likewise.
646         (vmaxnmvq_f32): Likewise.
647         (vmaxnmvq_f64): Likewise.
648         (vminv_f32): Likewise.
649         (vminv_s8): Likewise.
650         (vminv_s16): Likewise.
651         (vminv_s32): Likewise.
652         (vminv_u8): Likewise.
653         (vminv_u16): Likewise.
654         (vminv_u32): Likewise.
655         (vminvq_f32): Likewise.
656         (vminvq_f64): Likewise.
657         (vminvq_s8): Likewise.
658         (vminvq_s16): Likewise.
659         (vminvq_s32): Likewise.
660         (vminvq_u8): Likewise.
661         (vminvq_u16): Likewise.
662         (vminvq_u32): Likewise.
663         (vminnmv_f32): Likewise.
664         (vminnmvq_f32): Likewise.
665         (vminnmvq_f64): Likewise.
667 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
669         * config/aarch64/aarch64-simd.md
670         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
671         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
672         (*aarch64_mul3_elt<mode>): Likewise.
673         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
674         (*aarch64_mul3_elt_to_64v2df): Likewise.
675         (*aarch64_mla_elt<mode>): Likewise.
676         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
677         (*aarch64_mls_elt<mode>): Likewise.
678         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
679         (*aarch64_fma4_elt<mode>): Likewise.
680         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
681         (*aarch64_fma4_elt_to_64v2df): Likewise.
682         (*aarch64_fnma4_elt<mode>): Likewise.
683         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
684         (*aarch64_fnma4_elt_to_64v2df): Likewise.
685         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
686         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
687         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
688         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
689         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
690         (aarch64_sqdmull_lane<mode>_internal): Likewise.
691         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
693 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
695         * config/aarch64/aarch64-simd.md
696         (aarch64_be_checked_get_lane<mode>): New define_expand.
697         * config/aarch64/aarch64-simd-builtins.def
698         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
699         New builtin definition.
700         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
701         Use new safe be builtin.
703 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
705         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
706         New define_insn.
707         (aarch64_be_st1<mode>): Likewise.
708         (aarch_ld1<VALL:mode>): Define_expand modified.
709         (aarch_st1<VALL:mode>): Likewise.
710         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
711         (UNSPEC_ST1): Likewise.
713 2014-01-23  David Holsgrove <david.holsgrove@xilinx.com>
715         * config/microblaze/microblaze.md: Add trap insn and attribute
717 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
719         PR preprocessor/58580
720         * input.h (location_get_source_line): Take an additional line_size
721         parameter.
722         (void diagnostics_file_cache_fini): Declare new function.
723         * input.c (struct fcache): New type.
724         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
725         New static constants.
726         (diagnostic_file_cache_init, total_lines_num)
727         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
728         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
729         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
730         (get_next_line, read_next_line, goto_next_line, read_line_num):
731         New static function definitions.
732         (diagnostic_file_cache_fini): New function.
733         (location_get_source_line): Take an additional output line_len
734         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
735         read_line_num.
736         * diagnostic.c (diagnostic_finish): Call
737         diagnostic_file_cache_fini.
738         (adjust_line): Take an additional input parameter for the length
739         of the line, rather than calculating it with strlen.
740         (diagnostic_show_locus): Adjust the use of
741         location_get_source_line and adjust_line with respect to their new
742         signature.  While displaying a line now, do not stop at the first
743         null byte.  Rather, display the zero byte as a space and keep
744         going until we reach the size of the line.
745         * Makefile.in: Add vec.o to OBJS-libcommon
747 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
748             Ilya Tocar     <ilya.tocar@intel.com>
750         * config/i386/avx512fintrin.h (_mm512_kmov): New.
751         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
752         (__builtin_ia32_kmov16): Ditto.
753         * config/i386/i386.md (UNSPEC_KMOV): New.
754         (kmovw): Ditto.
756 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
758         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
759         (_mm512_storeu_si512): Ditto.
761 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
763         PR target/52125
764         * rtl.h (get_referenced_operands): Declare.
765         * recog.c (get_referenced_operands): New function.
766         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
767         operands have been referenced when recording LO_SUM references.
769 2014-01-22  David Holsgrove <david.holsgrove@xilinx.com>
771         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
773 2014-01-22  Jan Hubicka  <jh@suse.cz>
775         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
776         Enable for generic and recent AMD targets.
778 2014-01-22  Jan Hubicka  <jh@suse.cz>
780         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
781         ARG_SIZE note when adjustment was eliminated.
783 2014-01-22  Jeff Law  <law@redhat.com>
785         PR tree-optimization/59597
786         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
787         in file.  Accept new argument REGISTERING and use it to modify
788         dump output appropriately.
789         (register_jump_thread): Corresponding changes.
790         (mark_threaded_blocks): Reinstate code to cancel unprofitable
791         thread paths involving joiner blocks.  Add code to dump cancelled
792         jump threading paths.
794 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
796         PR rtl-optimization/59477
797         * lra-constraints.c (inherit_in_ebb): Process call for living hard
798         regs.  Update reloads_num and potential_reload_hard_regs for all insns.
800 2014-01-22  Tom Tromey  <tromey@redhat.com>
802         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
803         PARAMS.
804         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
806 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
808         PR rtl-optimization/59896
809         * lra-constraints.c (process_alt_operands): Check unused note for
810         matched operands of insn with no output reloads.
812 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
814         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
815         (mips_move_from_gpr_cost): Likewise.
817 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
819         PR rtl-optimization/59858
820         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
821         ira_class_hard_regs_num.
822         (process_alt_operands): Increase reject for dying matched operand.
824 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
826         PR target/59003
827         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
828         smaller than size, perform several stores or loads and stores
829         at dst + count - size to store or copy all of size bytes, rather
830         than just last modesize bytes.
832 2014-01-20  DJ Delorie  <dj@redhat.com>
834         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
835         that CLOBBERs are REGs before propogating their values.
837 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
839         PR middle-end/59789
840         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
841         (cgraph_inline_failed_type): New function.
842         * cgraph.h (DEFCIFCODE): Add type.
843         (cgraph_inline_failed_type_t): New enum.
844         (cgraph_inline_failed_type): New prototype.
845         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
846         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
847         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
848         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
849         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
850         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
851         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
852         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
853         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
854         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
855         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
856         OPTIMIZATION_MISMATCH.
857         * tree-inline.c (expand_call_inline): Emit errors during
858         early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
860 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
862         PR target/59685
863         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
864         mode attribute in insn output.
866 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
868         * output.h (output_constant): Delete.
869         * varasm.c (output_constant): Make private.
871 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
873         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
875 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
877         PR middle-end/59860
878         * tree.h (fold_builtin_strcat): New prototype.
879         * builtins.c (fold_builtin_strcat): No longer static.  Add len
880         argument, if non-NULL, don't call c_strlen.  Optimize
881         directly into __builtin_memcpy instead of __builtin_strcpy.
882         (fold_builtin_2): Adjust fold_builtin_strcat caller.
883         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
885 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
887         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
888         for SImode_address_operand operands, having only a REG argument.
890 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
892         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
893         loader name using mbig-endian.
894         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
896 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
898         * doc/invoke.texi (-march): Clarify documentation for AArch64.
899         (-mtune): Likewise.
900         (-mcpu): Likewise.
902 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
904         * config/aarch64/aarch64-protos.h
905         (aarch64_cannot_change_mode_class_ptr): Declare.
906         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
907         aarch64_cannot_change_mode_class_ptr): New.
908         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
909         backend hook aarch64_cannot_change_mode_class.
911 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
913         * common/config/aarch64/aarch64-common.c
914         (aarch64_handle_option): Don't handle any option order logic here.
915         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
916         selected_cpu, warn on architecture version mismatch.
917         (aarch64_override_options): Fix parsing order for option strings.
919 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
920             Iain Sandoe  <iain@codesourcery.com>
922         PR bootstrap/59496
923         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
924         warning.  Amend comment to reflect current functionality.
926 2014-01-20  Richard Biener  <rguenther@suse.de>
928         PR middle-end/59860
929         * builtins.c (fold_builtin_strcat): Remove case better handled
930         by tree-ssa-strlen.c.
932 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
934         * config/aarch64/aarch64.opt
935         (mcpu, march, mtune): Make case-insensitive.
937 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
939         PR target/59880
940         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
941         if operands[1] is a REG or ZERO_EXTEND of a REG.
943 2014-01-19  Jan Hubicka  <jh@suse.cz>
945         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
947 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
949         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
950         long non-pic millicode calls.
952 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
954         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
956 2014-01-19  Kito Cheng  <kito@0xlab.org>
958         * builtins.c (expand_movstr): Check movstr expand done or fail.
960 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
961             H.J. Lu  <hongjiu.lu@intel.com>
963         PR target/59379
964         * config/i386/i386.md (*lea<mode>): Zero-extend return register
965         to DImode for zero-extended addresses.
967 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
969         PR rtl-optimization/57763
970         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
971         on the new indirect jump_insn and increment LABEL_NUSES (label).
973 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
975         PR bootstrap/59580
976         PR bootstrap/59583
977         * config.gcc (x86_archs): New variable.
978         (x86_64_archs): Likewise.
979         (x86_cpus): Likewise.
980         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
981         --with-arch/--with-cpu= options.
982         Support --with-arch=/--with-cpu={nehalem,westmere,
983         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
985 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
987         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
988         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
990 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
992         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
994 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
996         PR target/58944
997         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
998         clear cpp_get_options (parse_in)->warn_unused_macros for
999         ix86_target_macros_internal with cpp_define.
1001 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
1003         * jump.c (delete_related_insns): Keep (use (insn))s.
1004         * reorg.c (redundant_insn): Check for barriers too.
1006 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
1008         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
1010 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
1012         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
1013         call to $$dyncall when TARGET_LONG_CALLS is true.
1015 2014-01-17  Jeff Law  <law@redhat.com>
1017         * ree.c (combine_set_extension): Temporarily disable test for
1018         changing number of hard registers.
1020 2014-01-17  Jan Hubicka  <jh@suse.cz>
1022         PR middle-end/58125
1023         * ipa-inline-analysis.c (inline_free_summary):
1024         Do not free summary of aliases.
1026 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
1028         PR middle-end/59706
1029         * gimplify.c (gimplify_expr): Use create_tmp_var
1030         instead of create_tmp_var_raw.  If cond doesn't have
1031         integral type, don't add the IFN_ANNOTATE builtin at all.
1033 2014-01-17  Martin Jambor  <mjambor@suse.cz>
1035         PR ipa/59736
1036         * ipa-cp.c (prev_edge_clone): New variable.
1037         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
1038         Also resize prev_edge_clone vector.
1039         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
1040         (ipcp_edge_removal_hook): New function.
1041         (ipcp_driver): Register ipcp_edge_removal_hook.
1043 2014-01-17  Andrew Pinski <apinski@cavium.com>
1044             Steve Ellcey  <sellcey@mips.com>
1046         PR target/59462
1047         * config/mips/mips.c (mips_print_operand): Check operand mode instead
1048         of operator mode.
1050 2014-01-17  Jeff Law  <law@redhat.com>
1052         PR middle-end/57904
1053         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
1054         so that pass_ccp runs first.
1056 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
1058         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
1059         (ix86_adjust_cost): Use !TARGET_XXX.
1060         (do_reorder_for_imul): Likewise.
1061         (swap_top_of_ready_list): Likewise.
1062         (ix86_sched_reorder): Likewise.
1064 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
1066         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1067         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
1068         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
1069         (intel_memset): New.  Duplicate slm_memset.
1070         (intel_cost): New.  Duplicate slm_cost.
1071         (m_INTEL): New macro.
1072         (processor_target_table): Add "intel".
1073         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
1074         with PROCESSOR_INTEL for "intel".
1075         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
1076         PROCESSOR_SILVERMONT.
1077         (ix86_issue_rate): Likewise.
1078         (ix86_adjust_cost): Likewise.
1079         (ia32_multipass_dfa_lookahead): Likewise.
1080         (swap_top_of_ready_list): Likewise.
1081         (ix86_sched_reorder): Likewise.
1082         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
1083         instead of TARGET_OPT_AGU.
1084         * config/i386/i386.h (TARGET_INTEL): New.
1085         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
1086         (processor_type): Add PROCESSOR_INTEL.
1087         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
1088         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
1090 2014-01-17  Marek Polacek  <polacek@redhat.com>
1092         PR c/58346
1093         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
1094         size is zero.
1096 2014-01-17  Richard Biener  <rguenther@suse.de>
1098         PR tree-optimization/46590
1099         * opts.c (default_options_table): Add entries for
1100         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
1101         all enabled at -O1 but not for -Og.
1102         * common.opt (fbranch-count-reg): Remove Init(1).
1103         (fmove-loop-invariants): Likewise.
1104         (ftree-pta): Likewise.
1106 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
1108         * config/i386/i386.c (ix86_data_alignment): For compatibility with
1109         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
1110         decls to at least the GCC 4.8 used alignments.
1112         PR fortran/59440
1113         * tree-nested.c (convert_nonlocal_reference_stmt,
1114         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
1115         of GIMPLE_BIND stmts, adjust associated decls.
1117 2014-01-17  Richard Biener  <rguenther@suse.de>
1119         PR tree-optimization/46590
1120         * vec.h (vec<>::bseach): New member function implementing
1121         binary search according to C89 bsearch.
1122         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
1123         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
1124         bitmap pointer again.  Make accesses_in_loop a flat array.
1125         (mem_ref_obstack): New global.
1126         (outermost_indep_loop): Adjust for mem_ref->stored changes.
1127         (mark_ref_stored): Likewise.
1128         (ref_indep_loop_p_2): Likewise.
1129         (set_ref_stored_in_loop): New helper function.
1130         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
1131         (memref_free): Adjust.
1132         (record_mem_ref_loc): Simplify.
1133         (gather_mem_refs_stmt): Adjust.
1134         (sort_locs_in_loop_postorder_cmp): New function.
1135         (analyze_memory_references): Sort accesses_in_loop after
1136         loop postorder number.
1137         (find_ref_loc_in_loop_cmp): New function.
1138         (for_all_locs_in_loop): Find relevant cluster of locs in
1139         accesses_in_loop and iterate without recursion.
1140         (execute_sm): Avoid uninit warning.
1141         (struct ref_always_accessed): Simplify.
1142         (ref_always_accessed::operator ()): Likewise.
1143         (ref_always_accessed_p): Likewise.
1144         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
1145         loop postorder numbers here.
1146         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
1147         numbers.
1149 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
1151         PR c++/57945
1152         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
1153         on decls for which assemble_alias has been called.
1155 2014-01-17  Nick Clifton  <nickc@redhat.com>
1157         * config/msp430/msp430.opt: (mcpu): New option.
1158         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
1159         (msp430_option_override): Parse target_cpu.  If the MCU name
1160         matches a generic string, clear target_mcu.
1161         (msp430_attr): Allow numeric interrupt values up to 63.
1162         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
1163         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
1164         option.
1165         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
1166         Add mcpu matches.
1167         * config/msp430/msp430.md (popm): Use %J rather than %I.
1168         (addsi3): Use msp430_nonimmediate_operand for operand 2.
1169         (addhi_cy_i): Use immediate_operand for operand 2.
1170         * doc/invoke.texi: Document -mcpu option.
1172 2014-01-17  Richard Biener  <rguenther@suse.de>
1174         PR rtl-optimization/38518
1175         * df.h (df_analyze_loop): Declare.
1176         * df-core.c: Include cfgloop.h.
1177         (df_analyze_1): Split out main part of df_analyze.
1178         (df_analyze): Adjust.
1179         (loop_inverted_post_order_compute): New function.
1180         (loop_post_order_compute): Likewise.
1181         (df_analyze_loop): New function avoiding whole-function
1182         postorder computes.
1183         * loop-invariant.c (find_defs): Use df_analyze_loop.
1184         (find_invariants): Adjust.
1185         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
1187 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
1189         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
1190         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
1192 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
1194         * ipa-ref.c (ipa_remove_stmt_references): Fix references
1195         traversal when removing references.
1197 2014-01-16  Jan Hubicka  <jh@suse.cz>
1199         PR ipa/59775
1200         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
1202 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
1204         PR middle-end/56791
1205         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
1206         pushing a reload for an autoinc when we had previously reloaded an
1207         inner part of the address.
1209 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
1211         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
1212         field.
1213         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
1214         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
1215         when not giving up or versioning for alias only because of
1216         loop->safelen.
1217         (vect_analyze_data_ref_dependences): Set to true.
1218         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
1219         is a GIMPLE_PHI.
1220         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
1221         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
1222         to the condition.
1224         PR middle-end/58344
1225         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
1227         PR target/59839
1228         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
1229         operand 0 predicate for gathers, use a new pseudo as subtarget.
1231 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
1233         PR middle-end/59609
1234         * lra-constraints.c (process_alt_operands): Add printing debug info.
1235         Check absence of input/output reloads for matched operands too.
1237 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
1239         PR rtl-optimization/59835
1240         * ira.c (ira_init_register_move_cost): Increase cost for
1241         impossible modes.
1243 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
1245         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
1247 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
1249         PR target/59780
1250         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
1251         non-register objects.  Use gen_(high/low)part more consistently.
1252         Fix assertions.
1254 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
1256         PR target/59844
1257         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
1258         endian support, remove tests for WORDS_BIG_ENDIAN.
1259         (p8_mfvsrd_3_<mode>): Likewise.
1260         (reload_gpr_from_vsx<mode>): Likewise.
1261         (reload_gpr_from_vsxsf): Likewise.
1262         (p8_mfvsrd_4_disf): Likewise.
1264 2014-01-16  Richard Biener  <rguenther@suse.de>
1266         PR rtl-optimization/46590
1267         * lcm.c (compute_antinout_edge): Use postorder iteration.
1268         (compute_laterin): Use inverted postorder iteration.
1270 2014-01-16  Nick Clifton  <nickc@redhat.com>
1272         PR middle-end/28865
1273         * varasm.c (output_constant): Return the number of bytes actually
1274         emitted.
1275         (output_constructor_array_range): Update the field size with the
1276         number of bytes emitted by output_constant.
1277         (output_constructor_regular_field): Likewise.  Also do not
1278         complain if the total number of bytes emitted is now greater
1279         than the expected fieldpos.
1280         * output.h (output_constant): Update prototype and descriptive comment.
1282 2014-01-16  Marek Polacek  <polacek@redhat.com>
1284         PR middle-end/59827
1285         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
1286         it is error_mark_node.
1288 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
1290         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
1291         VALID_AVX256_REG_OR_OI_MODE.
1293 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
1295         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
1296         current procedure should be profiled.
1298 2014-01-15  Andrew Pinski  <apinski@cavium.com>
1300         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
1301         of moving from/to the STACK_REG register class.
1303 2014-01-15  Richard Henderson <rth@redhat.com>
1305         PR debug/54694
1306         * reginfo.c (global_regs_decl): Globalize.
1307         * rtl.h (global_regs_decl): Declare.
1308         * ira.c (do_reload): Diagnose frame_pointer_needed and it
1309         reserved via global_regs.
1311 2014-01-15  Teresa Johnson  <tejohnson@google.com>
1313         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
1315 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
1317         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
1318         and vmulosh rather than call gen_vec_widen_smult_*.
1319         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
1320         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
1321         (vec_widen_smult_even_v16qi): Likewise.
1322         (vec_widen_umult_even_v8hi): Likewise.
1323         (vec_widen_smult_even_v8hi): Likewise.
1324         (vec_widen_umult_odd_v16qi): Likewise.
1325         (vec_widen_smult_odd_v16qi): Likewise.
1326         (vec_widen_umult_odd_v8hi): Likewise.
1327         (vec_widen_smult_odd_v8hi): Likewise.
1328         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
1329         vmuloub rather than call gen_vec_widen_umult_*.
1330         (vec_widen_umult_lo_v16qi): Likewise.
1331         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
1332         vmulosb rather than call gen_vec_widen_smult_*.
1333         (vec_widen_smult_lo_v16qi): Likewise.
1334         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
1335         rather than call gen_vec_widen_umult_*.
1336         (vec_widen_umult_lo_v8hi): Likewise.
1337         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
1338         rather than call gen_vec_widen_smult_*.
1339         (vec_widen_smult_lo_v8hi): Likewise.
1341 2014-01-15  Jeff Law  <law@redhat.com>
1343         PR tree-optimization/59747
1344         * ree.c (find_and_remove_re): Properly handle case where a second
1345         eliminated extension requires widening a copy created for elimination
1346         of a prior extension.
1347         (combine_set_extension): Ensure that the number of hard regs needed
1348         for a destination register does not change when we widen it.
1350 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
1352         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
1353         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
1354         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
1355         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
1356         (avr-*-rtems*): Likewise.
1357         (bfin*-rtems*): Likewise.
1358         (moxie-*-rtems*): Likewise.
1359         (h8300-*-rtems*): Likewise.
1360         (i[34567]86-*-rtems*): Likewise.
1361         (lm32-*-rtems*): Likewise.
1362         (m32r-*-rtems*): Likewise.
1363         (m68k-*-rtems*): Likewise.
1364         (microblaze*-*-rtems*): Likewise.
1365         (mips*-*-rtems*): Likewise.
1366         (powerpc-*-rtems*): Likewise.
1367         (sh-*-rtems*): Likewise.
1368         (sparc-*-rtems*): Likewise.
1369         (sparc64-*-rtems*): Likewise.
1370         (v850-*-rtems*): Likewise.
1371         (m32c-*-rtems*): Likewise.
1373 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
1375         PR rtl-optimization/59511
1376         * ira.c (ira_init_register_move_cost): Use memory costs for some
1377         cases of register move cost calculations.
1378         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
1379         instead of BB frequency.
1380         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
1381         * lra-assigns.c (find_hard_regno_for): Ditto.
1383 2014-01-15  Richard Biener  <rguenther@suse.de>
1385         PR tree-optimization/59822
1386         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
1387         (vectorizable_load): Use it to hoist defs of uses of invariant
1388         loads out of the loop.
1390 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
1391             Kugan Vivekanandarajah  <kuganv@linaro.org>
1393         PR target/59695
1394         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
1395         truncation.
1397 2014-01-15  Richard Biener  <rguenther@suse.de>
1399         PR rtl-optimization/59802
1400         * lcm.c (compute_available): Use inverted postorder to seed
1401         the initial worklist.
1403 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
1405         PR target/59803
1406         * config/s390/s390.c (s390_preferred_reload_class): Don't return
1407         ADDR_REGS for invalid symrefs in non-PIC code.
1409 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
1411         PR other/58712
1412         * builtins.c (determine_block_size): Initialize *probable_max_size
1413         even if len_rtx is CONST_INT.
1415 2014-01-14  Andrew Pinski  <apinski@cavium.com>
1417         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
1418         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
1419         (cortexa53_tunings): Likewise.
1420         (aarch64_sched_issue_rate): New function.
1421         (TARGET_SCHED_ISSUE_RATE): Define.
1423 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
1425         * ira-costs.c (find_costs_and_classes): Add missed
1426         ira_init_register_move_cost_if_necessary.
1428 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
1430         PR target/59787
1431         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
1433 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
1435         PR target/59794
1436         * config/i386/i386.c (type_natural_mode): Add a bool parameter
1437         to indicate if type is used for function return value.  Warn ABI
1438         change if the vector mode isn't available for function return value.
1439         (ix86_function_arg_advance): Pass false to type_natural_mode.
1440         (ix86_function_arg): Likewise.
1441         (ix86_gimplify_va_arg): Likewise.
1442         (function_arg_32): Don't warn ABI change.
1443         (ix86_function_value): Pass true to type_natural_mode.
1444         (ix86_return_in_memory): Likewise.
1445         (ix86_struct_value_rtx): Removed.
1446         (TARGET_STRUCT_VALUE_RTX): Likewise.
1448 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
1450         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
1451         converting a conditional jump into a conditional return.
1453 2014-01-14  Richard Biener  <rguenther@suse.de>
1455         PR tree-optimization/58921
1456         PR tree-optimization/59006
1457         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
1458         hoisting invariant stmts.
1459         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
1460         invariant loads on the preheader edge if possible.
1462 2014-01-14  Joey Ye  <joey.ye@arm.com>
1464         * doc/plugin.texi (Building GCC plugins): Update to C++.
1466 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
1468         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
1469         (_mm_rcp28_round_ss): Ditto.
1470         (_mm_rsqrt28_round_sd): Ditto.
1471         (_mm_rsqrt28_round_ss): Ditto.
1472         (_mm_rcp28_sd): Ditto.
1473         (_mm_rcp28_ss): Ditto.
1474         (_mm_rsqrt28_sd): Ditto.
1475         (_mm_rsqrt28_ss): Ditto.
1476         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
1477         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
1478         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
1479         (IX86_BUILTIN_RCP28SD): Ditto.
1480         (IX86_BUILTIN_RCP28SS): Ditto.
1481         (IX86_BUILTIN_RSQRT28SD): Ditto.
1482         (IX86_BUILTIN_RSQRT28SS): Ditto.
1483         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
1484         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
1485         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
1486         (ix86_expand_special_args_builtin): Expand new FTYPE.
1487         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
1488         (srcp14<mode>): Make insn unary.
1489         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
1490         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
1491         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
1492         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
1493         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
1494         Fix rounding: make it SAE only.
1495         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
1496         Ditto.
1497         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
1498         Ditto.
1499         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
1500         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
1501         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
1502         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
1503         (round_saeonly_mask_scalar_operand4): Ditto.
1504         (round_saeonly_mask_scalar_op3): Ditto.
1505         (round_saeonly_mask_scalar_op4): Ditto.
1507 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1509         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1510         Implement -maltivec=be for vec_insert and vec_extract.
1512 2014-01-10  DJ Delorie  <dj@redhat.com>
1514         * config/msp430/msp430.md (call_internal): Don't allow memory
1515         references with SP as the base register.
1516         (call_value_internal): Likewise.
1517         * config/msp430/constraints.md (Yc): New.  For memory references
1518         that don't use SP as a base register.
1520         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
1521         "an integer without a # prefix"
1522         * config/msp430/msp430.md (epilogue_helper): Use it.
1524 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
1526         PR target/59617
1527         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
1528         AVX512F gather builtins.
1529         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
1530         on gather decls with INTEGER_TYPE masktype.
1531         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
1532         directly into the builtin rather than hoisting it before loop.
1534         PR tree-optimization/59387
1535         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
1536         (scev_const_prop): If folded_casts and type has undefined overflow,
1537         use force_gimple_operand instead of force_gimple_operand_gsi and
1538         for each added stmt if it is assign with
1539         arith_code_with_undefined_signed_overflow, call
1540         rewrite_to_defined_overflow.
1541         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
1542         gimple-fold.h instead.
1543         (arith_code_with_undefined_signed_overflow,
1544         rewrite_to_defined_overflow): Moved to ...
1545         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
1546         rewrite_to_defined_overflow): ... here.  No longer static.
1547         Include gimplify-me.h.
1548         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
1549         rewrite_to_defined_overflow): New prototypes.
1551 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1553         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
1555 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
1557         * builtins.c (get_object_alignment_2): Minor tweak.
1558         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
1560 2014-01-13  Christian Bruel  <christian.bruel@st.com>
1562         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
1563         optimized non constant lengths.
1565 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
1567         PR libgomp/59194
1568         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
1569         load as __atomic_load_N if possible.
1571 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
1573         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
1574         target parameter.
1575         (rs6000_expand_builtin): Adjust call.
1577 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
1579         PR target/58115
1580         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
1581         * config/rs6000/rs6000.c: Include target-globals.h.
1582         (rs6000_set_current_function): Instead of doing target_reinit
1583         unconditionally, use save_target_globals_default_opts and
1584         restore_target_globals.
1586         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
1587         FPSCR.
1588         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
1589         (rs6000_expand_builtin): Handle mffs and mtfsf.
1590         (rs6000_init_builtins): Define mffs and mtfsf.
1591         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
1592         (rs6000_mffs): New pattern.
1593         (rs6000_mtfsf): New pattern.
1595 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
1597         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
1598         Start narrowing with START.  Apply candidate-use pair
1599         and check overall cost in narrowing.
1600         (iv_ca_prune): Pass new argument.
1602 2014-01-10  Jeff Law  <law@redhat.com>
1604         PR middle-end/59743
1605         * ree.c (combine_reaching_defs): Ensure the defining statement
1606         occurs before the extension when optimizing extensions with
1607         different source and destination hard registers.
1609 2014-01-10  Jan Hubicka  <jh@suse.cz>
1611         PR ipa/58585
1612         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
1613         vtables into the type inheritance graph.
1615 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
1617         PR rtl-optimization/59754
1618         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
1619         modes in the REGNO != REGNO case.
1621 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1623         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
1625 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
1627         PR tree-optimization/59745
1628         * tree-predcom.c (tree_predictive_commoning_loop): Call
1629         free_affine_expand_cache if giving up because components is NULL.
1631         * target-globals.c (save_target_globals): Allocate < 4KB structs using
1632         GC in payload of target_globals struct instead of allocating them on
1633         the heap and the larger structs separately using GC.
1634         * target-globals.h (struct target_globals): Make regs, hard_regs,
1635         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
1636         of GTY((skip)) and change type to void *.
1637         (reset_target_globals): Cast loads from those fields to corresponding
1638         types.
1640 2014-01-10  Steve Ellcey  <sellcey@mips.com>
1642         PR plugins/59335
1643         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
1644         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
1645         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
1647 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
1649         PR target/59744
1650         * aarch64-modes.def (CC_Zmode): New flags mode.
1651         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
1652         represents an equality.
1653         (aarch64_get_condition_code): Handle CC_Zmode.
1654         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
1656 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
1658         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
1659         extraction in good case.
1661 2014-01-10  Richard Biener  <rguenther@suse.de>
1663         PR tree-optimization/59374
1664         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
1665         checking after SLP discovery.  Mark stmts not participating
1666         in any SLP instance properly.
1668 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1670         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
1671         when handling a SET rtx.
1673 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1675         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
1676         (cortex-a57): Likewise.
1677         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
1679 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1681         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
1682         non-iwmmxt builtins.
1684 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
1686         PR ipa/58252
1687         PR ipa/59226
1688         * ipa-devirt.c record_target_from_binfo): Take as argument
1689         stack of binfos and lookup matching one for virtual inheritance.
1690         (possible_polymorphic_call_targets_1): Update.
1692 2014-01-10  Huacai Chen  <chenhc@lemote.com>
1694         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
1695         kernel strings for Loongson-2E/2F/3A.
1697 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
1699         PR middle-end/59670
1700         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
1701         is_gimple_call before calling gimple_call_internal_p.
1703 2014-01-09  Steve Ellcey  <sellcey@mips.com>
1705         * Makefile.in (TREE_FLOW_H): Remove.
1706         (TREE_SSA_H): Add file names from tree-flow.h.
1707         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
1708         * tree.h: Remove tree-flow.h reference.
1709         * hash-table.h: Remove tree-flow.h reference.
1710         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
1711         reference with tree-ssa-loop.h.
1713 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1715         * doc/invoke.texi: Add -maltivec={be,le} options, and document
1716         default element-order behavior for -maltivec.
1717         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
1718         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
1719         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
1720         when targeting big endian, at least for now.
1721         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
1723 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
1725         PR middle-end/47735
1726         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
1727         var satisfies use_register_for_decl, just take into account type
1728         alignment, rather than decl alignment.
1730         PR tree-optimization/59622
1731         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
1732         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
1733         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
1734         Don't devirtualize for inplace at all.  For targets.length () == 1,
1735         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
1737 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
1739         * config/i386/i386.md (cpu): Remove the unused btver1.
1741 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
1743         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
1745 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
1747         PR target/58115
1748         * tree-core.h (struct target_globals): New forward declaration.
1749         (struct tree_target_option): Add globals field.
1750         * tree.h (TREE_TARGET_GLOBALS): Define.
1751         (prepare_target_option_nodes_for_pch): New prototype.
1752         * target-globals.h (struct target_globals): Define even if
1753         !SWITCHABLE_TARGET.
1754         * tree.c (prepare_target_option_node_for_pch,
1755         prepare_target_option_nodes_for_pch): New functions.
1756         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
1757         * config/i386/i386.c: Include target-globals.h.
1758         (ix86_set_current_function): Instead of doing target_reinit
1759         unconditionally, use save_target_globals_default_opts and
1760         restore_target_globals.
1762 2014-01-09  Richard Biener  <rguenther@suse.de>
1764         PR tree-optimization/59715
1765         * tree-cfg.h (split_critical_edges): Declare.
1766         * tree-cfg.c (split_critical_edges): Export.
1767         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
1769 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
1771         * cfgexpand.c (expand_stack_vars): Optionally disable
1772         asan stack protection.
1773         (expand_used_vars): Likewise.
1774         (partition_stack_vars): Likewise.
1775         * asan.c (asan_emit_stack_protection): Optionally disable
1776         after return stack usage.
1777         (instrument_derefs): Optionally disable memory access instrumentation.
1778         (instrument_builtin_call): Likewise.
1779         (instrument_strlen_call): Likewise.
1780         (asan_protect_global): Optionally disable global variables protection.
1781         * doc/invoke.texi: Added doc for new options.
1782         * params.def: Added new options.
1783         * params.h: Likewise.
1785 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
1787         PR rtl-optimization/59724
1788         * ifcvt.c (cond_exec_process_if_block): Don't call
1789         flow_find_head_matching_sequence with 0 longest_match.
1790         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
1791         non-active insns if !stop_after.
1792         (try_head_merge_bb): Revert 2014-01-07 changes.
1794 2014-01-08  Jeff Law  <law@redhat.com>
1796         * ree.c (get_sub_rtx): New function, extracted from...
1797         (merge_def_and_ext): Here.
1798         (combine_reaching_defs): Use get_sub_rtx.
1800 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
1802         * cgraph.h (varpool_variable_node): Do not choke on null node.
1804 2014-01-08  Catherine Moore  <clm@codesourcery.com>
1806         * config/mips/mips.md (simple_return): Attempt to use JRC
1807         for microMIPS.
1808         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
1810 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
1812         PR rtl-optimization/59137
1813         * reorg.c (steal_delay_list_from_target): Call update_block for
1814         elided insns.
1815         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
1817 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1819         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1820         two duplicate entries.
1822 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
1824         Revert:
1825         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
1827         * config/mips/mips.c (mips_truncated_op_cost): New function.
1828         (mips_rtx_costs): Adjust test for BADDU.
1829         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
1831         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
1833         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
1834         (*baddu_si): ...this new pattern.
1836 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
1838         PR ipa/59722
1839         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
1841 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1843         PR middle-end/57748
1844         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
1845         inner_reference_p.
1846         (expand_expr, expand_normal): Adjust.
1847         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
1848         inner_reference_p. Use inner_reference_p to expand inner references.
1849         (store_expr): Adjust.
1850         * cfgexpand.c (expand_call_stmt): Adjust.
1852 2014-01-08  Rong Xu  <xur@google.com>
1854         * gcov-io.c (gcov_var): Move from gcov-io.h.
1855         (gcov_position): Ditto.
1856         (gcov_is_error): Ditto.
1857         (gcov_rewrite): Ditto.
1858         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
1859         only part to libgcc/libgcov.h.
1861 2014-01-08  Marek Polacek  <polacek@redhat.com>
1863         PR middle-end/59669
1864         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
1866 2014-01-08  Marek Polacek  <polacek@redhat.com>
1868         PR sanitizer/59667
1869         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
1871 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
1873         PR rtl-optimization/59649
1874         * stor-layout.c (get_mode_bounds): For BImode return
1875         0 and STORE_FLAG_VALUE.
1877 2014-01-08  Richard Biener  <rguenther@suse.de>
1879         PR middle-end/59630
1880         * gimple.h (is_gimple_builtin_call): Remove.
1881         (gimple_builtin_call_types_compatible_p): New.
1882         (gimple_call_builtin_p): New overload.
1883         * gimple.c (is_gimple_builtin_call): Remove.
1884         (validate_call): Rename to ...
1885         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
1886         check return types.
1887         (validate_type): New static function.
1888         (gimple_call_builtin_p): New overload and adjust.
1889         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
1890         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
1891         (gimple_fold_stmt_to_constant_1): Likewise.
1892         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
1894 2014-01-08  Richard Biener  <rguenther@suse.de>
1896         PR middle-end/59471
1897         * gimplify.c (gimplify_expr): Gimplify register-register type
1898         VIEW_CONVERT_EXPRs to separate stmts.
1900 2014-01-07  Jeff Law  <law@redhat.com>
1902         PR middle-end/53623
1903         * ree.c (combine_set_extension): Handle case where source
1904         and destination registers in an extension insn are different.
1905         (combine_reaching_defs): Allow source and destination registers
1906         in extension to be different under limited circumstances.
1907         (add_removable_extension): Remove restriction that the
1908         source and destination registers in the extension are the same.
1909         (find_and_remove_re): Emit a copy from the extension's
1910         destination to its source after the defining insn if
1911         the source and destination registers are different.
1913         PR middle-end/59285
1914         * ifcvt.c (merge_if_block): If we are merging a block with more than
1915         one successor with a block with no successors, remove any BARRIER
1916         after the second block.
1918 2014-01-07  Dan Xio Qiang <ziyan01@163.com>
1920         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
1922 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
1924         PR target/59652
1925         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
1926         for 14-bit register offsets when INT14_OK_STRICT is false.
1928 2014-01-07  Roland Stigge  <stigge@antcom.de>
1929             Michael Meissner  <meissner@linux.vnet.ibm.com>
1931         PR 57386/target
1932         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
1933         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
1935 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
1937         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
1938         -mcpu.
1940 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
1942         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
1943         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
1944         rtx is const0_rtx or not.
1946 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
1948         PR target/58115
1949         * target-globals.c (save_target_globals): Remove this_fn_optab
1950         handling.
1951         * toplev.c: Include optabs.h.
1952         (target_reinit): Temporarily restore the global options if another
1953         set of options are in force.
1955 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
1957         PR rtl-optimization/58668
1958         * cfgcleanup.c (flow_find_cross_jump): Don't count
1959         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
1960         to determine what is counted.
1961         (flow_find_head_matching_sequence): Use active_insn_p to determine
1962         what is counted.
1963         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
1964         counting change.
1965         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
1966         determine what is counted.
1968         PR tree-optimization/59643
1969         * tree-predcom.c (split_data_refs_to_components): If one dr is
1970         read and one write, determine_offset fails and the write isn't
1971         in the bad component, just put the read into the bad component.
1973 2014-01-07  Mike Stump  <mikestump@comcast.net>
1974             Jakub Jelinek  <jakub@redhat.com>
1976         PR pch/59436
1977         * tree-core.h (struct tree_optimization_option): Change optabs
1978         type from unsigned char * to void *.
1979         * optabs.c (init_tree_optimization_optabs): Adjust
1980         TREE_OPTIMIZATION_OPTABS initialization.
1982 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
1984         PR target/59644
1985         * config/i386/i386.h (struct machine_function): Add
1986         no_drap_save_restore field.
1987         * config/i386/i386.c (ix86_save_reg): Use
1988         !cfun->machine->no_drap_save_restore instead of
1989         crtl->stack_realign_needed.
1990         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
1991         this function clears frame_pointer_needed.  Set
1992         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
1993         and DRAP reg is needed.
1995 2014-01-06  Marek Polacek  <polacek@redhat.com>
1997         PR c/57773
1998         * doc/implement-c.texi: Mention that other integer types are
1999         permitted as bit-field types in strictly conforming mode.
2001 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
2003         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
2004         is newly allocated.
2006 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
2008         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
2010 2014-01-06  Martin Jambor  <mjambor@suse.cz>
2012         PR ipa/59008
2013         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
2014         to int.
2015         * ipa-prop.c (ipa_print_node_params): Fix indentation.
2017 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
2019         PR debug/59350
2020         PR debug/59510
2021         * var-tracking.c (add_stores): Preserve the value of the source even if
2022         we don't record the store.
2024 2014-01-06  Terry Guo  <terry.guo@arm.com>
2026         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
2028 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
2030         PR bootstrap/59541
2031         * config/darwin.c (darwin_function_section): Adjust return values to
2032         correspond to optimisation changes made in r206070.
2034 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
2036         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
2037         from prefetch_block tune setting.
2038         (nocona_cost): Correct size of prefetch block to 64.
2040 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
2042         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
2043         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
2044         used to save the static chain register in the computation of the offset
2045         from which the FP registers need to be restored.
2047 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
2049         PR tree-optimization/59519
2050         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
2051         ICE if get_current_def (current_new_name) is already non-NULL, as long
2052         as it is a phi result of some other phi in *new_exit_bb that has
2053         the same argument.
2055         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
2056         or vmovdqu* for misaligned_operand.
2057         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
2058         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
2059         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
2060         aligned_mem for AVX512F masked aligned load and store builtins and for
2061         non-temporal moves.
2063 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
2065         PR tree-optimization/59651
2066         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
2067         Address range for negative step should be added by TYPE_SIZE_UNIT.
2069 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
2071         * config/m68k/m68k.c (handle_move_double): Handle pushes with
2072         overlapping registers also for registers other than the stack pointer.
2074 2014-01-03  Marek Polacek  <polacek@redhat.com>
2076         PR other/59661
2077         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
2078         __builtin_FILE.
2080 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
2082         PR target/59625
2083         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
2084         asm goto as jump.
2086         * config/i386/i386.md (MODE_SIZE): New mode attribute.
2087         (push splitter): Use <P:MODE_SIZE> instead of
2088         GET_MODE_SIZE (<P:MODE>mode).
2089         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
2090         (mov -1, reg peephole2): Likewise.
2091         * config/i386/sse.md (*mov<mode>_internal,
2092         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
2093         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
2094         *<code><mode>3, *andnot<mode>3<mask_name>,
2095         <mask_codefor><code><mode>3<mask_name>): Likewise.
2096         * config/i386/subst.md (mask_mode512bit_condition,
2097         sd_mask_mode512bit_condition): Likewise.
2099 2014-01-02  Xinliang David Li  <davidxl@google.com>
2101         PR tree-optimization/59303
2102         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
2103         (dump_predicates): Better output format.
2104         (pred_equal_p): New function.
2105         (is_neq_relop_p): Ditto.
2106         (is_neq_zero_form_p): Ditto.
2107         (pred_expr_equal_p): Ditto.
2108         (pred_neg_p): Ditto.
2109         (simplify_pred): Ditto.
2110         (simplify_preds_2): Ditto.
2111         (simplify_preds_3): Ditto.
2112         (simplify_preds_4): Ditto.
2113         (simplify_preds): Ditto.
2114         (push_pred): Ditto.
2115         (push_to_worklist): Ditto.
2116         (get_pred_info_from_cmp): Ditto.
2117         (is_degenerated_phi): Ditto.
2118         (normalize_one_pred_1): Ditto.
2119         (normalize_one_pred): Ditto.
2120         (normalize_one_pred_chain): Ditto.
2121         (normalize_preds): Ditto.
2122         (normalize_cond_1): Remove function.
2123         (normalize_cond): Ditto.
2124         (is_gcond_subset_of): Ditto.
2125         (is_subset_of_any): Ditto.
2126         (is_or_set_subset_of): Ditto.
2127         (is_and_set_subset_of): Ditto.
2128         (is_norm_cond_subset_of): Ditto.
2129         (pred_chain_length_cmp): Ditto.
2130         (convert_control_dep_chain_into_preds): Type change.
2131         (find_predicates): Ditto.
2132         (find_def_preds): Ditto.
2133         (destroy_predicates_vecs): Ditto.
2134         (find_matching_predicates_in_rest_chains): Ditto.
2135         (use_pred_not_overlap_with_undef_path_pred): Ditto.
2136         (is_pred_expr_subset): Ditto.
2137         (is_pred_chain_subset_of): Ditto.
2138         (is_included_in): Ditto.
2139         (is_superset_of): Ditto.
2141 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
2143         Update copyright years.
2145 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
2147         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
2148         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
2149         config/arc/arc.md, config/arc/arc.opt,
2150         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
2151         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
2152         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
2153         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
2154         config/linux-protos.h, config/linux.c, config/winnt-c.c,
2155         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
2156         vtable-verify.c, vtable-verify.h: Use the standard form for the
2157         copyright notice.
2159 2014-01-02  Tobias Burnus  <burnus@net-b.de>
2161         * gcc.c (process_command): Update copyright notice dates.
2162         * gcov-dump.c: Ditto.
2163         * gcov.c: Ditto.
2164         * doc/cpp.texi: Bump @copying's copyright year.
2165         * doc/cppinternals.texi: Ditto.
2166         * doc/gcc.texi: Ditto.
2167         * doc/gccint.texi: Ditto.
2168         * doc/gcov.texi: Ditto.
2169         * doc/install.texi: Ditto.
2170         * doc/invoke.texi: Ditto.
2172 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
2174         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
2176 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
2178         * config/i386/sse.md (*mov<mode>_internal): Guard
2179         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
2181         PR rtl-optimization/59647
2182         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
2183         new_rtx into UNSIGNED_FLOAT rtxes.
2185 Copyright (C) 2014 Free Software Foundation, Inc.
2187 Copying and distribution of this file, with or without modification,
2188 are permitted in any medium without royalty provided the copyright
2189 notice and this notice are preserved.