gcc/
[official-gcc.git] / gcc / ChangeLog
blobda99a66b0fa372a2484e576732798b26647be84b
1 2014-01-27  Kirill Yukhin  <kirill.yukhin@intel.com>
3         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd): New.
4         (_mm512_mask_prefetch_i64gather_pd): Ditto.
5         (_mm512_prefetch_i32scatter_pd): Ditto.
6         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
7         (_mm512_prefetch_i64scatter_pd): Ditto.
8         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
9         (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
10         (_mm512_mask_prefetch_i64gather_ps): Ditto.
11         (_mm512_prefetch_i32scatter_ps): Ditto.
12         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
13         (_mm512_prefetch_i64scatter_ps): Ditto.
14         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
15         * config/i386/i386-builtin-types.def: Define
16         VOID_FTYPE_QI_V8SI_PCINT64_INT_INT and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
17         * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
18         IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
19         IX86_BUILTIN_SCATTERPFQPD.
20         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
21         __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
22         __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
23         __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
24         __builtin_ia32_scatterpfqps.
25         (ix86_expand_builtin): Expand new built-ins.
26         * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
27         fix memory access data type.
28         (*avx512pf_gatherpf<mode>_mask): Ditto.
29         (*avx512pf_gatherpf<mode>): Ditto.
30         (avx512pf_scatterpf<mode>): Ditto.
31         (*avx512pf_scatterpf<mode>_mask): Ditto.
32         (*avx512pf_scatterpf<mode>): Ditto.
33         (GATHER_SCATTER_SF_MEM_MODE): New.
34         (avx512pf_gatherpf<mode>df): Ditto.
35         (*avx512pf_gatherpf<mode>df_mask): Ditto.
36         (*avx512pf_scatterpf<mode>df): Ditto.
38 2014-01-27  Jakub Jelinek  <jakub@redhat.com>
40         PR bootstrap/59934
41         * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
42         know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
43         reached.
45 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
47         * common/config/arm/arm-common.c
48         (arm_rewrite_mcpu): Handle multiple names.
49         * config/arm/arm.h
50         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
52 2014-01-27  James Greenhalgh  <james.greenhalgh@arm.com>
54         * gimple-builder.h (create_gimple_tmp): Delete.
56 2014-01-27  Christian Bruel  <christian.bruel@st.com>
58         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
59         words comparisons.
61 2014-01-26  John David Anglin  <danglin@gcc.gnu.org>
63         * config/pa/pa.md (call): Generate indirect long calls to non-local
64         functions when outputing 32-bit code.
65         (call_value): Likewise except for special call to buggy powf function.
67         * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
68         portable runtime and PIC indirect calls.
69         (pa_output_indirect_call): Remove unnecessary nop from portable runtime
70         and PIC call sequences.  Use ldo instead of blr to set return register
71         in PIC call sequence.
73 2014-01-25  Walter Lee  <walt@tilera.com>
75         * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
76         avoid clobbering a live register.
78 2014-01-25  Walter Lee  <walt@tilera.com>
80         * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
81         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
82         * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
83         Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
85 2014-01-25  Walter Lee  <walt@tilera.com>
87         * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
88         arguments on even registers.
89         (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
90         STACK_BOUNDARY.
91         * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
92         (BIGGEST_ALIGNMENT): Ditto.
93         (BIGGEST_FIELD_ALIGNMENT): Ditto.
95 2014-01-25  Walter Lee  <walt@tilera.com>
97         * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
98         insns before bundling.
99         * config/tilegx/tilegx.md (tile_network_barrier): Update
100         comment.
102 2014-01-25  Walter Lee  <walt@tilera.com>
104         * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
105         PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
106         * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
108 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
110         * config/mips/constraints.md (kl): Delete.
111         * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
112         define expands, using...
113         (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
114         instructions for MIPS16.
115         (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
116         non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
118 2014-01-25  Walter Lee  <walt@tilera.com>
120         * config/tilepro/tilepro.md (ctzdi2): Use register_operand
121         predicate.
122         (clzdi2): Ditto.
123         (ffsdi2): Ditto.
125 2014-01-25  Walter Lee  <walt@tilera.com>
127         * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
128         (TARGET_EXPAND_TO_RTL_HOOK): Define.
130 2014-01-25  Richard Sandiford  <rdsandiford@googlemail.com>
132         * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
133         Handle XOR.
135 2014-01-25  Jakub Jelinek  <jakub@redhat.com>
137         * print-rtl.c (in_call_function_usage): New var.
138         (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
139         EXPR_LIST mode as mode and not as reg note name.
141         PR middle-end/59561
142         * cfgloopmanip.c (copy_loop_info): If
143         loop->warned_aggressive_loop_optimizations, make sure
144         the flag is set in target loop too.
146 2014-01-24  Balaji V. Iyer  <balaji.v.iyer@intel.com>
148         * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
149         flag_cilkplus.
150         * builtins.def: Likewise.
151         * cilk.h (fn_contains_cilk_spawn_p): Likewise.
152         * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
153         * ira.c (ira_setup_eliminable_regset): Likewise.
154         * omp-low.c (gate_expand_omp): Likewise.
155         (execute_lower_omp): Likewise.
156         (diagnose_sb_0): Likewise.
157         (gate_diagnose_omp_blocks): Likewise.
158         (simd_clone_clauses_extract): Likewise.
159         (gate): Likewise.
161 2014-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
163         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
164         correction for little endian...
165         * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
166         here.
168 2014-01-24  Jeff Law  <law@redhat.com>
170         PR tree-optimization/59919
171         * tree-vrp.c (find_assert_locations_1): Do not register asserts
172         for non-returning calls.
174 2014-01-24  James Greenhalgh  <james.greenhalgh@arm.com>
176         * common/config/aarch64/aarch64-common.c
177         (aarch64_rewrite_mcpu): Handle multiple names.
178         * config/aarch64/aarch64.h
179         (BIG_LITTLE_SPEC): Do not discard mcpu switches.
181 2014-01-24  Dodji Seketeli  <dodji@redhat.com>
183         * input.c (add_file_to_cache_tab): Handle the case where fopen
184         returns NULL.
186 2014-01-23  H.J. Lu  <hongjiu.lu@intel.com>
188         PR target/59929
189         * config/i386/i386.md (pushsf splitter): Get stack adjustment
190         from push operand if code of push isn't PRE_DEC.
192 2014-01-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
194         PR target/59909
195         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
196         -mquad-memory-atomic.  Update -mquad-memory documentation to say
197         it is only used for non-atomic loads/stores.
199         * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
200         -mquad-memory or -mquad-memory-atomic switches.
202         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
203         -mquad-memory-atomic to ISA 2.07 support.
205         * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
206         to separate support of normal quad word memory operations (ldq,
207         stq) from the atomic quad word memory operations.
209         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
210         support to separate non-atomic quad word operations from atomic
211         quad word operations.  Disable non-atomic quad word operations in
212         little endian mode so that we don't have to swap words after the
213         load and before the store.
214         (quad_load_store_p): Add comment about atomic quad word support.
215         (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
216         options printed with -mdebug=reg.
218         * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
219         -mquad-memory-atomic as the test for whether we have quad word
220         atomic instructions.
221         (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic,
222         -mquad-memory, or -mp8-vector are used, allow byte/half-word
223         atomic operations.
225         * config/rs6000/sync.md (load_lockedti): Insure that the address
226         is a proper indexed or indirect address for the lqarx instruction.
227         On little endian systems, swap the hi/lo registers after the lqarx
228         instruction.
229         (load_lockedpti): Use indexed_or_indirect_operand predicate to
230         insure the address is valid for the lqarx instruction.
231         (store_conditionalti): Insure that the address is a proper indexed
232         or indirect address for the stqcrx. instruction.  On little endian
233         systems, swap the hi/lo registers before doing the stqcrx.
234         instruction.
235         (store_conditionalpti): Use indexed_or_indirect_operand predicate to
236         insure the address is valid for the stqcrx. instruction.
238         * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
239         Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
240         type of quad memory support is available.
242 2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
244         PR regression/59915
245         * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
246         there is a danger of looping.
248 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
250         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
251         force flag_ira_loop_pressure if set via command line.
253 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
255         * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
256         (ashr_simd): New builtin handling DI mode.
257         * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
258         (aarch64_sshr_simddi): New match pattern.
259         * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
260         (vshrd_n_s64): Likewise.
261         * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
263 2014-01-23  Nick Clifton  <nickc@redhat.com>
265         * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
266         (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
267         favour of mcu specific scripts.
268         * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
269         430x multilibs.
271 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
272             Alex Velenko  <Alex.Velenko@arm.com>
274         * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
275         (vaddv_s16): Likewise.
276         (vaddv_s32): Likewise.
277         (vaddv_u8): Likewise.
278         (vaddv_u16): Likewise.
279         (vaddv_u32): Likewise.
280         (vaddvq_s8): Likewise.
281         (vaddvq_s16): Likewise.
282         (vaddvq_s32): Likewise.
283         (vaddvq_s64): Likewise.
284         (vaddvq_u8): Likewise.
285         (vaddvq_u16): Likewise.
286         (vaddvq_u32): Likewise.
287         (vaddvq_u64): Likewise.
288         (vaddv_f32): Likewise.
289         (vaddvq_f32): Likewise.
290         (vaddvq_f64): Likewise.
291         (vmaxv_f32): Likewise.
292         (vmaxv_s8): Likewise.
293         (vmaxv_s16): Likewise.
294         (vmaxv_s32): Likewise.
295         (vmaxv_u8): Likewise.
296         (vmaxv_u16): Likewise.
297         (vmaxv_u32): Likewise.
298         (vmaxvq_f32): Likewise.
299         (vmaxvq_f64): Likewise.
300         (vmaxvq_s8): Likewise.
301         (vmaxvq_s16): Likewise.
302         (vmaxvq_s32): Likewise.
303         (vmaxvq_u8): Likewise.
304         (vmaxvq_u16): Likewise.
305         (vmaxvq_u32): Likewise.
306         (vmaxnmv_f32): Likewise.
307         (vmaxnmvq_f32): Likewise.
308         (vmaxnmvq_f64): Likewise.
309         (vminv_f32): Likewise.
310         (vminv_s8): Likewise.
311         (vminv_s16): Likewise.
312         (vminv_s32): Likewise.
313         (vminv_u8): Likewise.
314         (vminv_u16): Likewise.
315         (vminv_u32): Likewise.
316         (vminvq_f32): Likewise.
317         (vminvq_f64): Likewise.
318         (vminvq_s8): Likewise.
319         (vminvq_s16): Likewise.
320         (vminvq_s32): Likewise.
321         (vminvq_u8): Likewise.
322         (vminvq_u16): Likewise.
323         (vminvq_u32): Likewise.
324         (vminnmv_f32): Likewise.
325         (vminnmvq_f32): Likewise.
326         (vminnmvq_f64): Likewise.
328 2014-01-23  James Greenhalgh  <james.greenhalgh@arm.com>
330         * config/aarch64/aarch64-simd.md
331         (aarch64_dup_lane<mode>): Correct lane number on big-endian.
332         (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
333         (*aarch64_mul3_elt<mode>): Likewise.
334         (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
335         (*aarch64_mul3_elt_to_64v2df): Likewise.
336         (*aarch64_mla_elt<mode>): Likewise.
337         (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
338         (*aarch64_mls_elt<mode>): Likewise.
339         (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
340         (*aarch64_fma4_elt<mode>): Likewise.
341         (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
342         (*aarch64_fma4_elt_to_64v2df): Likewise.
343         (*aarch64_fnma4_elt<mode>): Likewise.
344         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
345         (*aarch64_fnma4_elt_to_64v2df): Likewise.
346         (aarch64_sq<r>dmulh_lane<mode>): Likewise.
347         (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
348         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
349         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
350         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
351         (aarch64_sqdmull_lane<mode>_internal): Likewise.
352         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
354 2013-01-23  Alex Velenko  <Alex.Velenko@arm.com>
356         * config/aarch64/aarch64-simd.md
357         (aarch64_be_checked_get_lane<mode>): New define_expand.
358         * config/aarch64/aarch64-simd-builtins.def
359         (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0):
360         New builtin definition.
361         * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
362         Use new safe be builtin.
364 2014-01-23  Alex Velenko  <Alex.Velenko@arm.com>
366         * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
367         New define_insn.
368         (aarch64_be_st1<mode>): Likewise.
369         (aarch_ld1<VALL:mode>): Define_expand modified.
370         (aarch_st1<VALL:mode>): Likewise.
371         * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
372         (UNSPEC_ST1): Likewise.
374 2014-01-23  David Holsgrove <david.holsgrove@xilinx.com>
376         * config/microblaze/microblaze.md: Add trap insn and attribute
378 2014-01-23  Dodji Seketeli  <dodji@redhat.com>
380         PR preprocessor/58580
381         * input.h (location_get_source_line): Take an additional line_size
382         parameter.
383         (void diagnostics_file_cache_fini): Declare new function.
384         * input.c (struct fcache): New type.
385         (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
386         New static constants.
387         (diagnostic_file_cache_init, total_lines_num)
388         (lookup_file_in_cache_tab, evicted_cache_tab_entry)
389         (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
390         (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
391         (get_next_line, read_next_line, goto_next_line, read_line_num):
392         New static function definitions.
393         (diagnostic_file_cache_fini): New function.
394         (location_get_source_line): Take an additional output line_len
395         parameter.  Re-write using lookup_or_add_file_to_cache_tab and
396         read_line_num.
397         * diagnostic.c (diagnostic_finish): Call
398         diagnostic_file_cache_fini.
399         (adjust_line): Take an additional input parameter for the length
400         of the line, rather than calculating it with strlen.
401         (diagnostic_show_locus): Adjust the use of
402         location_get_source_line and adjust_line with respect to their new
403         signature.  While displaying a line now, do not stop at the first
404         null byte.  Rather, display the zero byte as a space and keep
405         going until we reach the size of the line.
406         * Makefile.in: Add vec.o to OBJS-libcommon
408 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
409             Ilya Tocar     <ilya.tocar@intel.com>
411         * config/i386/avx512fintrin.h (_mm512_kmov): New.
412         * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
413         (__builtin_ia32_kmov16): Ditto.
414         * config/i386/i386.md (UNSPEC_KMOV): New.
415         (kmovw): Ditto.
417 2014-01-23  Kirill Yukhin  <kirill.yukhin@intel.com>
419         * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
420         (_mm512_storeu_si512): Ditto.
422 2014-01-23  Richard Sandiford  <rdsandiford@googlemail.com>
424         PR target/52125
425         * rtl.h (get_referenced_operands): Declare.
426         * recog.c (get_referenced_operands): New function.
427         * config/mips/mips.c (mips_reorg_process_insns): Check which asm
428         operands have been referenced when recording LO_SUM references.
430 2014-01-22  David Holsgrove <david.holsgrove@xilinx.com>
432         * config/microblaze/microblaze.md: Correct bswaphi2 insn.
434 2014-01-22  Jan Hubicka  <jh@suse.cz>
436         * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
437         Enable for generic and recent AMD targets.
439 2014-01-22  Jan Hubicka  <jh@suse.cz>
441         * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
442         ARG_SIZE note when adjustment was eliminated.
444 2014-01-22  Jeff Law  <law@redhat.com>
446         PR tree-optimization/59597
447         * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
448         in file.  Accept new argument REGISTERING and use it to modify
449         dump output appropriately.
450         (register_jump_thread): Corresponding changes.
451         (mark_threaded_blocks): Reinstate code to cancel unprofitable
452         thread paths involving joiner blocks.  Add code to dump cancelled
453         jump threading paths.
455 2014-01-22  Vladimir Makarov  <vmakarov@redhat.com>
457         PR rtl-optimization/59477
458         * lra-constraints.c (inherit_in_ebb): Process call for living hard
459         regs.  Update reloads_num and potential_reload_hard_regs for all
460         insns.
462 2014-01-22  Tom Tromey  <tromey@redhat.com>
464         * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
465         PARAMS.
466         * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
468 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
470         PR rtl-optimization/59896
471         * lra-constraints.c (process_alt_operands): Check unused note for
472         matched operands of insn with no output reloads.
474 2014-01-21  Richard Sandiford  <rdsandiford@googlemail.com>
476         * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
477         (mips_move_from_gpr_cost): Likewise.
479 2014-01-21  Vladimir Makarov  <vmakarov@redhat.com>
481         PR rtl-optimization/59858
482         * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
483         ira_class_hard_regs_num.
484         (process_alt_operands): Increase reject for dying matched operand.
486 2014-01-21  Jakub Jelinek  <jakub@redhat.com>
488         PR target/59003
489         * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
490         smaller than size, perform several stores or loads and stores
491         at dst + count - size to store or copy all of size bytes, rather
492         than just last modesize bytes.
494 2014-01-20  DJ Delorie  <dj@redhat.com>
496         * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
497         that CLOBBERs are REGs before propogating their values.
499 2014-01-20  H.J. Lu  <hongjiu.lu@intel.com>
501         PR middle-end/59789
502         * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
503         (cgraph_inline_failed_type): New function.
504         * cgraph.h (DEFCIFCODE): Add type.
505         (cgraph_inline_failed_type_t): New enum.
506         (cgraph_inline_failed_type): New prototype.
507         * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
508         FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
509         FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
510         LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
511         MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
512         RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
513         OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
514         INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL. 
515         Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
516         FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
517         EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
518         OPTIMIZATION_MISMATCH.
519         * tree-inline.c (expand_call_inline): Emit errors during
520         early_inlining if cgraph_inline_failed_type returns
521         CIF_FINAL_ERROR.
523 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
525         PR target/59685
526         * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
527         mode attribute in insn output.
529 2014-01-20  Eric Botcazou  <ebotcazou@adacore.com>
531         * output.h (output_constant): Delete.
532         * varasm.c (output_constant): Make private.
534 2014-01-20  Alex Velenko  <Alex.Velenko@arm.com>
536         * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
538 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
540         PR middle-end/59860
541         * tree.h (fold_builtin_strcat): New prototype.
542         * builtins.c (fold_builtin_strcat): No longer static.  Add len
543         argument, if non-NULL, don't call c_strlen.  Optimize
544         directly into __builtin_memcpy instead of __builtin_strcpy.
545         (fold_builtin_2): Adjust fold_builtin_strcat caller.
546         * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
548 2014-01-20  Uros Bizjak  <ubizjak@gmail.com>
550         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
551         for SImode_address_operand operands, having only a REG argument.
553 2014-01-20  Marcus Shawcroft  <marcus.shawcroft@arm.com>
555         * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
556         loader name using mbig-endian.
557         (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
559 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
561         * doc/invoke.texi (-march): Clarify documentation for AArch64.
562         (-mtune): Likewise.
563         (-mcpu): Likewise.
565 2014-01-20  Tejas Belagod  <tejas.belagod@arm.com>
567         * config/aarch64/aarch64-protos.h
568         (aarch64_cannot_change_mode_class_ptr): Declare.
569         * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
570         aarch64_cannot_change_mode_class_ptr): New.
571         * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
572         backend hook aarch64_cannot_change_mode_class.
574 2014-01-20  James Greenhalgh  <james.greenhalgh@arm.com>
576         * common/config/aarch64/aarch64-common.c
577         (aarch64_handle_option): Don't handle any option order logic here.
578         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
579         selected_cpu, warn on architecture version mismatch.
580         (aarch64_override_options): Fix parsing order for option strings.
582 2014-01-20  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
583             Iain Sandoe  <iain@codesourcery.com>
585         PR bootstrap/59496
586         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
587         warning.  Amend comment to reflect current functionality.
589 2014-01-20  Richard Biener  <rguenther@suse.de>
591         PR middle-end/59860
592         * builtins.c (fold_builtin_strcat): Remove case better handled
593         by tree-ssa-strlen.c.
595 2014-01-20  Alan Lawrence  <alan.lawrence@arm.com>
597         * config/aarch64/aarch64.opt
598         (mcpu, march, mtune): Make case-insensitive.
600 2014-01-20  Jakub Jelinek  <jakub@redhat.com>
602         PR target/59880
603         * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
604         if operands[1] is a REG or ZERO_EXTEND of a REG.
606 2014-01-19  Jan Hubicka  <jh@suse.cz>
608         * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
610 2014-01-19  John David Anglin  <danglin@gcc.gnu.org>
612         * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
613         long non-pic millicode calls.
615 2014-01-19  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
617         * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
619 2014-01-19  Kito Cheng  <kito@0xlab.org>
621         * builtins.c (expand_movstr): Check movstr expand done or fail.
623 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
624             H.J. Lu  <hongjiu.lu@intel.com>
626         PR target/59379
627         * config/i386/i386.md (*lea<mode>): Zero-extend return register
628         to DImode for zero-extended addresses.
630 2014-01-19  Jakub Jelinek  <jakub@redhat.com>
632         PR rtl-optimization/57763
633         * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
634         on the new indirect jump_insn and increment LABEL_NUSES (label).
636 2014-01-18  H.J. Lu  <hongjiu.lu@intel.com>
638         PR bootstrap/59580
639         PR bootstrap/59583
640         * config.gcc (x86_archs): New variable.
641         (x86_64_archs): Likewise.
642         (x86_cpus): Likewise.
643         Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
644         --with-arch/--with-cpu= options.
645         Support --with-arch=/--with-cpu={nehalem,westmere,
646         sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
648 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
650         * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
651         and PROCESSOR_ATHLON to simplify code.  Move "memory" calculation.
653 2014-01-18  Uros Bizjak  <ubizjak@gmail.com>
655         * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
657 2014-01-18  Jakub Jelinek  <jakub@redhat.com>
659         PR target/58944
660         * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
661         clear cpp_get_options (parse_in)->warn_unused_macros for
662         ix86_target_macros_internal with cpp_define.
664 2014-01-18  Richard Sandiford  <rdsandiford@googlemail.com>
666         * jump.c (delete_related_insns): Keep (use (insn))s.
667         * reorg.c (redundant_insn): Check for barriers too.
669 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
671         * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
673 2014-01-17  John David Anglin  <danglin@gcc.gnu.org>
675         * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
676         call to $$dyncall when TARGET_LONG_CALLS is true.
678 2014-01-17  Jeff Law  <law@redhat.com>
680         * ree.c (combine_set_extension): Temporarily disable test for
681         changing number of hard registers.
683 2014-01-17  Jan Hubicka  <jh@suse.cz>
685         PR middle-end/58125
686         * ipa-inline-analysis.c (inline_free_summary):
687         Do not free summary of aliases.
689 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
691         PR middle-end/59706
692         * gimplify.c (gimplify_expr): Use create_tmp_var
693         instead of create_tmp_var_raw.  If cond doesn't have
694         integral type, don't add the IFN_ANNOTATE builtin at all.
696 2014-01-17  Martin Jambor  <mjambor@suse.cz>
698         PR ipa/59736
699         * ipa-cp.c (prev_edge_clone): New variable.
700         (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
701         Also resize prev_edge_clone vector.
702         (ipcp_edge_duplication_hook): Also update prev_edge_clone.
703         (ipcp_edge_removal_hook): New function.
704         (ipcp_driver): Register ipcp_edge_removal_hook.
706 2014-01-17  Andrew Pinski <apinski@cavium.com>
707             Steve Ellcey  <sellcey@mips.com>
709         PR target/59462
710         * config/mips/mips.c (mips_print_operand): Check operand mode instead
711         of operator mode.
713 2014-01-17  Jeff Law  <law@redhat.com>
715         PR middle-end/57904
716         * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
717         so that pass_ccp runs first.
719 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
721         * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
722         (ix86_adjust_cost): Use !TARGET_XXX.
723         (do_reorder_for_imul): Likewise.
724         (swap_top_of_ready_list): Likewise.
725         (ix86_sched_reorder): Likewise.
727 2014-01-17  H.J. Lu  <hongjiu.lu@intel.com>
729         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
730         PROCESSOR_INTEL.  Treat like PROCESSOR_GENERIC.
731         * config/i386/i386.c (intel_memcpy): New.  Duplicate slm_memcpy.
732         (intel_memset): New.  Duplicate slm_memset.
733         (intel_cost): New.  Duplicate slm_cost.
734         (m_INTEL): New macro.
735         (processor_target_table): Add "intel".
736         (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
737         with PROCESSOR_INTEL for "intel".
738         (ix86_lea_outperforms): Support PROCESSOR_INTEL.  Duplicate
739         PROCESSOR_SILVERMONT.
740         (ix86_issue_rate): Likewise.
741         (ix86_adjust_cost): Likewise.
742         (ia32_multipass_dfa_lookahead): Likewise.
743         (swap_top_of_ready_list): Likewise.
744         (ix86_sched_reorder): Likewise.
745         (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
746         instead of TARGET_OPT_AGU.
747         * config/i386/i386.h (TARGET_INTEL): New.
748         (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
749         (processor_type): Add PROCESSOR_INTEL.
750         * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
751         Add X86_TUNE_AVOID_LEA_FOR_ADDR.
753 2014-01-17  Marek Polacek  <polacek@redhat.com>
755         PR c/58346
756         * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
757         size is zero.
759 2014-01-17  Richard Biener  <rguenther@suse.de>
761         PR tree-optimization/46590
762         * opts.c (default_options_table): Add entries for
763         OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
764         all enabled at -O1 but not for -Og.
765         * common.opt (fbranch-count-reg): Remove Init(1).
766         (fmove-loop-invariants): Likewise.
767         (ftree-pta): Likewise.
769 2014-01-17  Jakub Jelinek  <jakub@redhat.com>
771         * config/i386/i386.c (ix86_data_alignment): For compatibility with
772         (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
773         decls to at least the GCC 4.8 used alignments.
775         PR fortran/59440
776         * tree-nested.c (convert_nonlocal_reference_stmt,
777         convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
778         of GIMPLE_BIND stmts, adjust associated decls.
780 2014-01-17  Richard Biener  <rguenther@suse.de>
782         PR tree-optimization/46590
783         * vec.h (vec<>::bseach): New member function implementing
784         binary search according to C89 bsearch.
785         (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
786         * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
787         bitmap pointer again.  Make accesses_in_loop a flat array.
788         (mem_ref_obstack): New global.
789         (outermost_indep_loop): Adjust for mem_ref->stored changes.
790         (mark_ref_stored): Likewise.
791         (ref_indep_loop_p_2): Likewise.
792         (set_ref_stored_in_loop): New helper function.
793         (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
794         (memref_free): Adjust.
795         (record_mem_ref_loc): Simplify.
796         (gather_mem_refs_stmt): Adjust.
797         (sort_locs_in_loop_postorder_cmp): New function.
798         (analyze_memory_references): Sort accesses_in_loop after
799         loop postorder number.
800         (find_ref_loc_in_loop_cmp): New function.
801         (for_all_locs_in_loop): Find relevant cluster of locs in
802         accesses_in_loop and iterate without recursion.
803         (execute_sm): Avoid uninit warning.
804         (struct ref_always_accessed): Simplify.
805         (ref_always_accessed::operator ()): Likewise.
806         (ref_always_accessed_p): Likewise.
807         (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
808         loop postorder numbers here.
809         (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
810         numbers.
812 2014-01-17  Jan Hubicka  <hubicka@ucw.cz>
814         PR c++/57945
815         * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
816         on decls for which assemble_alias has been called.
818 2014-01-17  Nick Clifton  <nickc@redhat.com>
820         * config/msp430/msp430.opt: (mcpu): New option.
821         * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
822         (msp430_option_override): Parse target_cpu.  If the MCU name
823         matches a generic string, clear target_mcu.
824         (msp430_attr): Allow numeric interrupt values up to 63.
825         (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
826         * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
827         option.
828         * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
829         Add mcpu matches.
830         * config/msp430/msp430.md (popm): Use %J rather than %I.
831         (addsi3): Use msp430_nonimmediate_operand for operand 2.
832         (addhi_cy_i): Use immediate_operand for operand 2.
833         * doc/invoke.texi: Document -mcpu option.
835 2014-01-17  Richard Biener  <rguenther@suse.de>
837         PR rtl-optimization/38518
838         * df.h (df_analyze_loop): Declare.
839         * df-core.c: Include cfgloop.h.
840         (df_analyze_1): Split out main part of df_analyze.
841         (df_analyze): Adjust.
842         (loop_inverted_post_order_compute): New function.
843         (loop_post_order_compute): Likewise.
844         (df_analyze_loop): New function avoiding whole-function
845         postorder computes.
846         * loop-invariant.c (find_defs): Use df_analyze_loop.
847         (find_invariants): Adjust.
848         * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
850 2014-01-17  Zhenqiang Chen  <zhenqiang.chen@arm.com>
852         * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
853         (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
855 2014-01-16  Ilya Enkovich  <ilya.enkovich@intel.com>
857         * ipa-ref.c (ipa_remove_stmt_references): Fix references
858         traversal when removing references.
860 2014-01-16  Jan Hubicka  <jh@suse.cz>
862         PR ipa/59775
863         * tree.c (get_binfo_at_offset): Look harder for virtual bases.
865 2014-01-16  Bernd Schmidt  <bernds@codesourcery.com>
867         PR middle-end/56791
868         * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
869         pushing a reload for an autoinc when we had previously reloaded an
870         inner part of the address.
872 2014-01-16  Jakub Jelinek  <jakub@redhat.com>
874         * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
875         field.
876         (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
877         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
878         when not giving up or versioning for alias only because of
879         loop->safelen.
880         (vect_analyze_data_ref_dependences): Set to true.
881         * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
882         is a GIMPLE_PHI.
883         (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
884         LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
885         to the condition.
887         PR middle-end/58344
888         * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
890         PR target/59839
891         * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
892         operand 0 predicate for gathers, use a new pseudo as subtarget.
894 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
896         PR middle-end/59609
897         * lra-constraints.c (process_alt_operands): Add printing debug info.
898         Check absence of input/output reloads for matched operands too.
900 2014-01-16  Vladimir Makarov  <vmakarov@redhat.com>
902         PR rtl-optimization/59835
903         * ira.c (ira_init_register_move_cost): Increase cost for
904         impossible modes.
906 2014-01-16  Alan Lawrence  <alan.lawrence@arm.com>
908         * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
910 2014-01-16  Richard Earnshaw  <rearnsha@arm.com>
912         PR target/59780
913         * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
914         non-register objects.  Use gen_(high/low)part more consistently.
915         Fix assertions.
917 2014-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
919         PR target/59844
920         * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
921         endian support, remove tests for WORDS_BIG_ENDIAN.
922         (p8_mfvsrd_3_<mode>): Likewise.
923         (reload_gpr_from_vsx<mode>): Likewise.
924         (reload_gpr_from_vsxsf): Likewise.
925         (p8_mfvsrd_4_disf): Likewise.
927 2014-01-16  Richard Biener  <rguenther@suse.de>
929         PR rtl-optimization/46590
930         * lcm.c (compute_antinout_edge): Use postorder iteration.
931         (compute_laterin): Use inverted postorder iteration.
933 2014-01-16  Nick Clifton  <nickc@redhat.com>
935         PR middle-end/28865
936         * varasm.c (output_constant): Return the number of bytes actually
937         emitted.
938         (output_constructor_array_range): Update the field size with the
939         number of bytes emitted by output_constant.
940         (output_constructor_regular_field): Likewise.  Also do not
941         complain if the total number of bytes emitted is now greater
942         than the expected fieldpos.
943         * output.h (output_constant): Update prototype and descriptive comment.
945 2014-01-16  Marek Polacek  <polacek@redhat.com>
947         PR middle-end/59827
948         * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
949         it is error_mark_node.
951 2014-01-15  Uros Bizjak  <ubizjak@gmail.com>
953         * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
954         VALID_AVX256_REG_OR_OI_MODE.
956 2014-01-15  Pat Haugen  <pthaugen@us.ibm.com>
958         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
959         current procedure should be profiled.
961 2014-01-15  Andrew Pinski  <apinski@cavium.com>
963         * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
964         of moving from/to the STACK_REG register class.
966 2014-01-15  Richard Henderson <rth@redhat.com>
968         PR debug/54694
969         * reginfo.c (global_regs_decl): Globalize.
970         * rtl.h (global_regs_decl): Declare.
971         * ira.c (do_reload): Diagnose frame_pointer_needed and it
972         reserved via global_regs.
974 2014-01-15  Teresa Johnson  <tejohnson@google.com>
976         * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
978 2014-01-15  Bill Schmidt  <wschmidt@vnet.linux.ibm.com>
980         * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
981         and vmulosh rather than call gen_vec_widen_smult_*.
982         (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
983         than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
984         (vec_widen_smult_even_v16qi): Likewise.
985         (vec_widen_umult_even_v8hi): Likewise.
986         (vec_widen_smult_even_v8hi): Likewise.
987         (vec_widen_umult_odd_v16qi): Likewise.
988         (vec_widen_smult_odd_v16qi): Likewise.
989         (vec_widen_umult_odd_v8hi): Likewise.
990         (vec_widen_smult_odd_v8hi): Likewise.
991         (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
992         vmuloub rather than call gen_vec_widen_umult_*.
993         (vec_widen_umult_lo_v16qi): Likewise.
994         (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
995         vmulosb rather than call gen_vec_widen_smult_*.
996         (vec_widen_smult_lo_v16qi): Likewise.
997         (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
998         rather than call gen_vec_widen_umult_*.
999         (vec_widen_umult_lo_v8hi): Likewise.
1000         (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
1001         rather than call gen_vec_widen_smult_*.
1002         (vec_widen_smult_lo_v8hi): Likewise.
1004 2014-01-15  Jeff Law  <law@redhat.com>
1006         PR tree-optimization/59747
1007         * ree.c (find_and_remove_re): Properly handle case where a second
1008         eliminated extension requires widening a copy created for elimination
1009         of a prior extension.
1010         (combine_set_extension): Ensure that the number of hard regs needed
1011         for a destination register does not change when we widen it.
1013 2014-01-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
1015         * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
1016         (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
1017         (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
1018         (arm*-*-rtems*): Use t-rtems from existing tmake_file.
1019         (avr-*-rtems*): Likewise.
1020         (bfin*-rtems*): Likewise.
1021         (moxie-*-rtems*): Likewise.
1022         (h8300-*-rtems*): Likewise.
1023         (i[34567]86-*-rtems*): Likewise.
1024         (lm32-*-rtems*): Likewise.
1025         (m32r-*-rtems*): Likewise.
1026         (m68k-*-rtems*): Likewise.
1027         (microblaze*-*-rtems*): Likewise.
1028         (mips*-*-rtems*): Likewise.
1029         (powerpc-*-rtems*): Likewise.
1030         (sh-*-rtems*): Likewise.
1031         (sparc-*-rtems*): Likewise.
1032         (sparc64-*-rtems*): Likewise.
1033         (v850-*-rtems*): Likewise.
1034         (m32c-*-rtems*): Likewise.
1036 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
1038         PR rtl-optimization/59511
1039         * ira.c (ira_init_register_move_cost): Use memory costs for some
1040         cases of register move cost calculations.
1041         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
1042         instead of BB frequency.
1043         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
1044         * lra-assigns.c (find_hard_regno_for): Ditto.
1046 2014-01-15  Richard Biener  <rguenther@suse.de>
1048         PR tree-optimization/59822
1049         * tree-vect-stmts.c (hoist_defs_of_uses): New function.
1050         (vectorizable_load): Use it to hoist defs of uses of invariant
1051         loads out of the loop.
1053 2014-01-15  Matthew Gretton-Dann  <matthew.gretton-dann@linaro.org>
1054             Kugan Vivekanandarajah  <kuganv@linaro.org>
1056         PR target/59695
1057         * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
1058         truncation.
1060 2014-01-15  Richard Biener  <rguenther@suse.de>
1062         PR rtl-optimization/59802
1063         * lcm.c (compute_available): Use inverted postorder to seed
1064         the initial worklist.
1066 2014-01-15  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
1068         PR target/59803
1069         * config/s390/s390.c (s390_preferred_reload_class): Don't return
1070         ADDR_REGS for invalid symrefs in non-PIC code.
1072 2014-01-15  Jakub Jelinek  <jakub@redhat.com>
1074         PR other/58712
1075         * builtins.c (determine_block_size): Initialize *probable_max_size
1076         even if len_rtx is CONST_INT.
1078 2014-01-14  Andrew Pinski  <apinski@cavium.com>
1080         * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
1081         * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
1082         (cortexa53_tunings): Likewise.
1083         (aarch64_sched_issue_rate): New function.
1084         (TARGET_SCHED_ISSUE_RATE): Define.
1086 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
1088         * ira-costs.c (find_costs_and_classes): Add missed
1089         ira_init_register_move_cost_if_necessary.
1091 2014-01-14  Vladimir Makarov  <vmakarov@redhat.com>
1093         PR target/59787
1094         * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
1096 2014-01-14  H.J. Lu  <hongjiu.lu@intel.com>
1098         PR target/59794
1099         * config/i386/i386.c (type_natural_mode): Add a bool parameter
1100         to indicate if type is used for function return value.  Warn ABI
1101         change if the vector mode isn't available for function return value.
1102         (ix86_function_arg_advance): Pass false to type_natural_mode.
1103         (ix86_function_arg): Likewise.
1104         (ix86_gimplify_va_arg): Likewise.
1105         (function_arg_32): Don't warn ABI change.
1106         (ix86_function_value): Pass true to type_natural_mode.
1107         (ix86_return_in_memory): Likewise.
1108         (ix86_struct_value_rtx): Removed.
1109         (TARGET_STRUCT_VALUE_RTX): Likewise.
1111 2014-01-14  Richard Sandiford  <rsandifo@linux.vnet.ibm.com>
1113         * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
1114         converting a conditional jump into a conditional return.
1116 2014-01-14  Richard Biener  <rguenther@suse.de>
1118         PR tree-optimization/58921
1119         PR tree-optimization/59006
1120         * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
1121         hoisting invariant stmts.
1122         * tree-vect-stmts.c (vectorizable_load): Insert the splat of
1123         invariant loads on the preheader edge if possible.
1125 2014-01-14  Joey Ye  <joey.ye@arm.com>
1127         * doc/plugin.texi (Building GCC plugins): Update to C++.
1129 2014-01-14  Kirill Yukhin  <kirill.yukhin@intel.com>
1131         * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
1132         (_mm_rcp28_round_ss): Ditto.
1133         (_mm_rsqrt28_round_sd): Ditto.
1134         (_mm_rsqrt28_round_ss): Ditto.
1135         (_mm_rcp28_sd): Ditto.
1136         (_mm_rcp28_ss): Ditto.
1137         (_mm_rsqrt28_sd): Ditto.
1138         (_mm_rsqrt28_ss): Ditto.
1139         * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
1140         * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
1141         * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
1142         (IX86_BUILTIN_RCP28SD): Ditto.
1143         (IX86_BUILTIN_RCP28SS): Ditto.
1144         (IX86_BUILTIN_RSQRT28SD): Ditto.
1145         (IX86_BUILTIN_RSQRT28SS): Ditto.
1146         (bdesc_special_args): Define __builtin_ia32_movntdqa512,
1147         __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
1148         __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
1149         (ix86_expand_special_args_builtin): Expand new FTYPE.
1150         * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
1151         (srcp14<mode>): Make insn unary.
1152         (avx512f_vmscalef<mode><round_name>): Use substed predicate.
1153         (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
1154         (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
1155         (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
1156         (avx512er_exp2<mode><mask_name><round_saeonly_name>):
1157         Fix rounding: make it SAE only.
1158         (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
1159         Ditto.
1160         (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
1161         Ditto.
1162         (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
1163         (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
1164         (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
1165         * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
1166         (round_saeonly_mask_scalar_operand4): Ditto.
1167         (round_saeonly_mask_scalar_op3): Ditto.
1168         (round_saeonly_mask_scalar_op4): Ditto.
1170 2014-01-13  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1172         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1173         Implement -maltivec=be for vec_insert and vec_extract.
1175 2014-01-10  DJ Delorie  <dj@redhat.com>
1177         * config/msp430/msp430.md (call_internal): Don't allow memory
1178         references with SP as the base register.
1179         (call_value_internal): Likewise.
1180         * config/msp430/constraints.md (Yc): New.  For memory references
1181         that don't use SP as a base register.
1183         * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
1184         "an integer without a # prefix"
1185         * config/msp430/msp430.md (epilogue_helper): Use it.
1187 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
1189         PR target/59617
1190         * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
1191         AVX512F gather builtins.
1192         * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
1193         on gather decls with INTEGER_TYPE masktype.
1194         (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
1195         directly into the builtin rather than hoisting it before loop.
1197         PR tree-optimization/59387
1198         * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
1199         (scev_const_prop): If folded_casts and type has undefined overflow,
1200         use force_gimple_operand instead of force_gimple_operand_gsi and
1201         for each added stmt if it is assign with
1202         arith_code_with_undefined_signed_overflow, call
1203         rewrite_to_defined_overflow.
1204         * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
1205         gimple-fold.h instead.
1206         (arith_code_with_undefined_signed_overflow,
1207         rewrite_to_defined_overflow): Moved to ...
1208         * gimple-fold.c (arith_code_with_undefined_signed_overflow,
1209         rewrite_to_defined_overflow): ... here.  No longer static.
1210         Include gimplify-me.h.
1211         * gimple-fold.h (arith_code_with_undefined_signed_overflow,
1212         rewrite_to_defined_overflow): New prototypes.
1214 2014-01-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1216         * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
1218 2014-01-13  Eric Botcazou  <ebotcazou@adacore.com>
1220         * builtins.c (get_object_alignment_2): Minor tweak.
1221         * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
1223 2014-01-13  Christian Bruel  <christian.bruel@st.com>
1225         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
1226         optimized non constant lengths.
1228 2014-01-13  Jakub Jelinek  <jakub@redhat.com>
1230         PR libgomp/59194
1231         * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
1232         load as __atomic_load_N if possible.
1234 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
1236         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
1237         target parameter.
1238         (rs6000_expand_builtin): Adjust call.
1240 2014-01-11  David Edelsohn  <dje.gcc@gmail.com>
1242         PR target/58115
1243         * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
1244         * config/rs6000/rs6000.c: Include target-globals.h.
1245         (rs6000_set_current_function): Instead of doing target_reinit
1246         unconditionally, use save_target_globals_default_opts and
1247         restore_target_globals.
1249         * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
1250         FPSCR.
1251         * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
1252         (rs6000_expand_builtin): Handle mffs and mtfsf.
1253         (rs6000_init_builtins): Define mffs and mtfsf.
1254         * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
1255         (rs6000_mffs): New pattern.
1256         (rs6000_mtfsf): New pattern.
1258 2014-01-11  Bin Cheng  <bin.cheng@arm.com>
1260         * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
1261         Start narrowing with START.  Apply candidate-use pair
1262         and check overall cost in narrowing.
1263         (iv_ca_prune): Pass new argument.
1265 2014-01-10  Jeff Law  <law@redhat.com>
1267         PR middle-end/59743
1268         * ree.c (combine_reaching_defs): Ensure the defining statement
1269         occurs before the extension when optimizing extensions with
1270         different source and destination hard registers.
1272 2014-01-10  Jan Hubicka  <jh@suse.cz>
1274         PR ipa/58585
1275         * ipa-devirt.c (build_type_inheritance_graph): Also add types of
1276         vtables into the type inheritance graph.
1278 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
1280         PR rtl-optimization/59754
1281         * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
1282         modes in the REGNO != REGNO case.
1284 2014-01-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1286         * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
1288 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
1290         PR tree-optimization/59745
1291         * tree-predcom.c (tree_predictive_commoning_loop): Call
1292         free_affine_expand_cache if giving up because components is NULL.
1294         * target-globals.c (save_target_globals): Allocate < 4KB structs using
1295         GC in payload of target_globals struct instead of allocating them on
1296         the heap and the larger structs separately using GC.
1297         * target-globals.h (struct target_globals): Make regs, hard_regs,
1298         reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
1299         of GTY((skip)) and change type to void *.
1300         (reset_target_globals): Cast loads from those fields to corresponding
1301         types.
1303 2014-01-10  Steve Ellcey  <sellcey@mips.com>
1305         PR plugins/59335
1306         * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
1307         gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
1308         tree-ssanames.h, print-tree.h, varasm.h, and context.h.
1310 2014-01-10  Richard Earnshaw  <rearnsha@arm.com>
1312         PR target/59744
1313         * aarch64-modes.def (CC_Zmode): New flags mode.
1314         * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
1315         represents an equality.
1316         (aarch64_get_condition_code): Handle CC_Zmode.
1317         * aarch64.md (compare_neg<mode>): Restrict to equality operations.
1319 2014-01-10  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
1321         * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
1322         extraction in good case.
1324 2014-01-10  Richard Biener  <rguenther@suse.de>
1326         PR tree-optimization/59374
1327         * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
1328         checking after SLP discovery.  Mark stmts not participating
1329         in any SLP instance properly.
1331 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1333         * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
1334         when handling a SET rtx.
1336 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1338         * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
1339         (cortex-a57): Likewise.
1340         (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
1342 2014-01-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1344         * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
1345         non-iwmmxt builtins.
1347 2014-01-10  Jan Hubicka  <hubicka@ucw.cz>
1349         PR ipa/58252
1350         PR ipa/59226
1351         * ipa-devirt.c record_target_from_binfo): Take as argument
1352         stack of binfos and lookup matching one for virtual inheritance.
1353         (possible_polymorphic_call_targets_1): Update.
1355 2014-01-10  Huacai Chen  <chenhc@lemote.com>
1357         * config/mips/driver-native.c (host_detect_local_cpu): Handle new
1358         kernel strings for Loongson-2E/2F/3A.
1360 2014-01-10  Jakub Jelinek  <jakub@redhat.com>
1362         PR middle-end/59670
1363         * tree-vect-data-refs.c (vect_analyze_data_refs): Check
1364         is_gimple_call before calling gimple_call_internal_p.
1366 2014-01-09  Steve Ellcey  <sellcey@mips.com>
1368         * Makefile.in (TREE_FLOW_H): Remove.
1369         (TREE_SSA_H): Add file names from tree-flow.h.
1370         * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
1371         * tree.h: Remove tree-flow.h reference.
1372         * hash-table.h: Remove tree-flow.h reference.
1373         * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
1374         reference with tree-ssa-loop.h.
1376 2014-01-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1378         * doc/invoke.texi: Add -maltivec={be,le} options, and document
1379         default element-order behavior for -maltivec.
1380         * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
1381         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
1382         that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
1383         when targeting big endian, at least for now.
1384         * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
1386 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
1388         PR middle-end/47735
1389         * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
1390         var satisfies use_register_for_decl, just take into account type
1391         alignment, rather than decl alignment.
1393         PR tree-optimization/59622
1394         * gimple-fold.c (gimple_fold_call): Fix a typo in message.  For
1395         __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
1396         __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
1397         Don't devirtualize for inplace at all.  For targets.length () == 1,
1398         if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
1400 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
1402         * config/i386/i386.md (cpu): Remove the unused btver1.
1404 2014-01-09  H.J. Lu  <hongjiu.lu@intel.com>
1406         * gdbasan.in: Put a breakpoint on __sanitizer::Report.
1408 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
1410         PR target/58115
1411         * tree-core.h (struct target_globals): New forward declaration.
1412         (struct tree_target_option): Add globals field.
1413         * tree.h (TREE_TARGET_GLOBALS): Define.
1414         (prepare_target_option_nodes_for_pch): New prototype.
1415         * target-globals.h (struct target_globals): Define even if
1416         !SWITCHABLE_TARGET.
1417         * tree.c (prepare_target_option_node_for_pch,
1418         prepare_target_option_nodes_for_pch): New functions.
1419         * config/i386/i386.h (SWITCHABLE_TARGET): Define.
1420         * config/i386/i386.c: Include target-globals.h.
1421         (ix86_set_current_function): Instead of doing target_reinit
1422         unconditionally, use save_target_globals_default_opts and
1423         restore_target_globals.
1425 2014-01-09  Richard Biener  <rguenther@suse.de>
1427         PR tree-optimization/59715
1428         * tree-cfg.h (split_critical_edges): Declare.
1429         * tree-cfg.c (split_critical_edges): Export.
1430         * tree-ssa-sink.c (execute_sink_code): Split critical edges.
1432 2014-01-09  Max Ostapenko  <m.ostapenko@partner.samsung.com>
1434         * cfgexpand.c (expand_stack_vars): Optionally disable
1435         asan stack protection.
1436         (expand_used_vars): Likewise.
1437         (partition_stack_vars): Likewise.
1438         * asan.c (asan_emit_stack_protection): Optionally disable
1439         after return stack usage.
1440         (instrument_derefs): Optionally disable memory access instrumentation.
1441         (instrument_builtin_call): Likewise.
1442         (instrument_strlen_call): Likewise.
1443         (asan_protect_global): Optionally disable global variables protection.
1444         * doc/invoke.texi: Added doc for new options.
1445         * params.def: Added new options.
1446         * params.h: Likewise.
1448 2014-01-09  Jakub Jelinek  <jakub@redhat.com>
1450         PR rtl-optimization/59724
1451         * ifcvt.c (cond_exec_process_if_block): Don't call
1452         flow_find_head_matching_sequence with 0 longest_match.
1453         * cfgcleanup.c (flow_find_head_matching_sequence): Count even
1454         non-active insns if !stop_after.
1455         (try_head_merge_bb): Revert 2014-01-07 changes.
1457 2014-01-08  Jeff Law  <law@redhat.com>
1459         * ree.c (get_sub_rtx): New function, extracted from...
1460         (merge_def_and_ext): Here.
1461         (combine_reaching_defs): Use get_sub_rtx.
1463 2014-01-08  Eric Botcazou  <ebotcazou@adacore.com>
1465         * cgraph.h (varpool_variable_node): Do not choke on null node.
1467 2014-01-08  Catherine Moore  <clm@codesourcery.com>
1469         * config/mips/mips.md (simple_return): Attempt to use JRC
1470         for microMIPS.
1471         * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
1473 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
1475         PR rtl-optimization/59137
1476         * reorg.c (steal_delay_list_from_target): Call update_block for
1477         elided insns.
1478         (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
1480 2014-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
1482         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1483         two duplicate entries.
1485 2014-01-08  Richard Sandiford  <rdsandiford@googlemail.com>
1487         Revert:
1488         2012-10-07  Richard Sandiford  <rdsandiford@googlemail.com>
1490         * config/mips/mips.c (mips_truncated_op_cost): New function.
1491         (mips_rtx_costs): Adjust test for BADDU.
1492         * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
1494         2012-10-02  Richard Sandiford  <rdsandiford@googlemail.com>
1496         * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
1497         (*baddu_si): ...this new pattern.
1499 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
1501         PR ipa/59722
1502         * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
1504 2014-01-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1506         PR middle-end/57748
1507         * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
1508         inner_reference_p.
1509         (expand_expr, expand_normal): Adjust.
1510         * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
1511         inner_reference_p. Use inner_reference_p to expand inner references.
1512         (store_expr): Adjust.
1513         * cfgexpand.c (expand_call_stmt): Adjust.
1515 2014-01-08  Rong Xu  <xur@google.com>
1517         * gcov-io.c (gcov_var): Move from gcov-io.h.
1518         (gcov_position): Ditto.
1519         (gcov_is_error): Ditto.
1520         (gcov_rewrite): Ditto.
1521         * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
1522         only part to libgcc/libgcov.h.
1524 2014-01-08  Marek Polacek  <polacek@redhat.com>
1526         PR middle-end/59669
1527         * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
1529 2014-01-08  Marek Polacek  <polacek@redhat.com>
1531         PR sanitizer/59667
1532         * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
1534 2014-01-08  Jakub Jelinek  <jakub@redhat.com>
1536         PR rtl-optimization/59649
1537         * stor-layout.c (get_mode_bounds): For BImode return
1538         0 and STORE_FLAG_VALUE.
1540 2014-01-08  Richard Biener  <rguenther@suse.de>
1542         PR middle-end/59630
1543         * gimple.h (is_gimple_builtin_call): Remove.
1544         (gimple_builtin_call_types_compatible_p): New.
1545         (gimple_call_builtin_p): New overload.
1546         * gimple.c (is_gimple_builtin_call): Remove.
1547         (validate_call): Rename to ...
1548         (gimple_builtin_call_types_compatible_p): ... this and export.  Also
1549         check return types.
1550         (validate_type): New static function.
1551         (gimple_call_builtin_p): New overload and adjust.
1552         * gimple-fold.c (gimple_fold_builtin): Fold the return value.
1553         (gimple_fold_call): Likewise.  Use gimple_call_builtin_p.
1554         (gimple_fold_stmt_to_constant_1): Likewise.
1555         * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
1557 2014-01-08  Richard Biener  <rguenther@suse.de>
1559         PR middle-end/59471
1560         * gimplify.c (gimplify_expr): Gimplify register-register type
1561         VIEW_CONVERT_EXPRs to separate stmts.
1563 2014-01-07  Jeff Law  <law@redhat.com>
1565         PR middle-end/53623
1566         * ree.c (combine_set_extension): Handle case where source
1567         and destination registers in an extension insn are different.
1568         (combine_reaching_defs): Allow source and destination registers
1569         in extension to be different under limited circumstances.
1570         (add_removable_extension): Remove restriction that the
1571         source and destination registers in the extension are the same.
1572         (find_and_remove_re): Emit a copy from the extension's
1573         destination to its source after the defining insn if
1574         the source and destination registers are different.
1576         PR middle-end/59285
1577         * ifcvt.c (merge_if_block): If we are merging a block with more than
1578         one successor with a block with no successors, remove any BARRIER
1579         after the second block.
1581 2014-01-07  Dan Xio Qiang <ziyan01@163.com>
1583         * hw-doloop.c (reorg_loops): Release the bitmap obstack.
1585 2014-01-07  John David Anglin  <danglin@gcc.gnu.org>
1587         PR target/59652
1588         * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
1589         for 14-bit register offsets when INT14_OK_STRICT is false.
1591 2014-01-07  Roland Stigge  <stigge@antcom.de>
1592             Michael Meissner  <meissner@linux.vnet.ibm.com>
1594         PR 57386/target
1595         * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
1596         Only check TFmode for SPE constants.  Don't check TImode or TDmode.
1598 2014-01-07  James Greenhalgh  <james.greenhalgh@arm.com>
1600         * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
1601         -mcpu.
1603 2014-01-07  Yufeng Zhang  <yufeng.zhang@arm.com>
1605         * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
1606         with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
1607         rtx is const0_rtx or not.
1609 2014-01-07  Richard Sandiford  <rdsandiford@googlemail.com>
1611         PR target/58115
1612         * target-globals.c (save_target_globals): Remove this_fn_optab
1613         handling.
1614         * toplev.c: Include optabs.h.
1615         (target_reinit): Temporarily restore the global options if another
1616         set of options are in force.
1618 2014-01-07  Jakub Jelinek  <jakub@redhat.com>
1620         PR rtl-optimization/58668
1621         * cfgcleanup.c (flow_find_cross_jump): Don't count
1622         any jumps if dir_p is NULL.  Remove p1 variable, use active_insn_p
1623         to determine what is counted.
1624         (flow_find_head_matching_sequence): Use active_insn_p to determine
1625         what is counted.
1626         (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
1627         counting change.
1628         * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
1629         determine what is counted.
1631         PR tree-optimization/59643
1632         * tree-predcom.c (split_data_refs_to_components): If one dr is
1633         read and one write, determine_offset fails and the write isn't
1634         in the bad component, just put the read into the bad component.
1636 2014-01-07  Mike Stump  <mikestump@comcast.net>
1637             Jakub Jelinek  <jakub@redhat.com>
1639         PR pch/59436
1640         * tree-core.h (struct tree_optimization_option): Change optabs
1641         type from unsigned char * to void *.
1642         * optabs.c (init_tree_optimization_optabs): Adjust
1643         TREE_OPTIMIZATION_OPTABS initialization.
1645 2014-01-06  Jakub Jelinek  <jakub@redhat.com>
1647         PR target/59644
1648         * config/i386/i386.h (struct machine_function): Add
1649         no_drap_save_restore field.
1650         * config/i386/i386.c (ix86_save_reg): Use
1651         !cfun->machine->no_drap_save_restore instead of
1652         crtl->stack_realign_needed.
1653         (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
1654         this function clears frame_pointer_needed.  Set
1655         cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
1656         and DRAP reg is needed.
1658 2014-01-06  Marek Polacek  <polacek@redhat.com>
1660         PR c/57773
1661         * doc/implement-c.texi: Mention that other integer types are
1662         permitted as bit-field types in strictly conforming mode.
1664 2014-01-06  Felix Yang  <fei.yang0953@gmail.com>
1666         * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
1667         is newly allocated.
1669 2014-01-06  Richard Earnshaw  <rearnsha@arm.com>
1671         * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
1673 2014-01-06  Martin Jambor  <mjambor@suse.cz>
1675         PR ipa/59008
1676         * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
1677         to int.
1678         * ipa-prop.c (ipa_print_node_params): Fix indentation.
1680 2014-01-06  Eric Botcazou  <ebotcazou@adacore.com>
1682         PR debug/59350
1683         PR debug/59510
1684         * var-tracking.c (add_stores): Preserve the value of the source even if
1685         we don't record the store.
1687 2014-01-06  Terry Guo  <terry.guo@arm.com>
1689         * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
1691 2014-01-05  Iain Sandoe  <iain@codesourcery.com>
1693         PR bootstrap/59541
1694         * config/darwin.c (darwin_function_section): Adjust return values to
1695         correspond to optimisation changes made in r206070.
1697 2014-01-05  Uros Bizjak  <ubizjak@gmail.com>
1699         * config/i386/i386.c (ix86_data_alignment): Calculate max_align
1700         from prefetch_block tune setting.
1701         (nocona_cost): Correct size of prefetch block to 64.
1703 2014-01-04  Eric Botcazou  <ebotcazou@adacore.com>
1705         * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
1706         (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
1707         used to save the static chain register in the computation of the offset
1708         from which the FP registers need to be restored.
1710 2014-01-04  Jakub Jelinek  <jakub@redhat.com>
1712         PR tree-optimization/59519
1713         * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
1714         ICE if get_current_def (current_new_name) is already non-NULL, as long
1715         as it is a phi result of some other phi in *new_exit_bb that has
1716         the same argument.
1718         * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
1719         or vmovdqu* for misaligned_operand.
1720         (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
1721         <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
1722         * config/i386/i386.c (ix86_expand_special_args_builtin): Set
1723         aligned_mem for AVX512F masked aligned load and store builtins and for
1724         non-temporal moves.
1726 2014-01-03  Bingfeng Mei  <bmei@broadcom.com>
1728         PR tree-optimization/59651
1729         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
1730         Address range for negative step should be added by TYPE_SIZE_UNIT.
1732 2014-01-03  Andreas Schwab  <schwab@linux-m68k.org>
1734         * config/m68k/m68k.c (handle_move_double): Handle pushes with
1735         overlapping registers also for registers other than the stack pointer.
1737 2014-01-03  Marek Polacek  <polacek@redhat.com>
1739         PR other/59661
1740         * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
1741         __builtin_FILE.
1743 2014-01-03  Jakub Jelinek  <jakub@redhat.com>
1745         PR target/59625
1746         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
1747         asm goto as jump.
1749         * config/i386/i386.md (MODE_SIZE): New mode attribute.
1750         (push splitter): Use <P:MODE_SIZE> instead of
1751         GET_MODE_SIZE (<P:MODE>mode).
1752         (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
1753         (mov -1, reg peephole2): Likewise.
1754         * config/i386/sse.md (*mov<mode>_internal,
1755         <sse>_storeu<ssemodesuffix><avxsizesuffix>,
1756         <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
1757         *<code><mode>3, *andnot<mode>3<mask_name>,
1758         <mask_codefor><code><mode>3<mask_name>): Likewise.
1759         * config/i386/subst.md (mask_mode512bit_condition,
1760         sd_mask_mode512bit_condition): Likewise.
1762 2014-01-02  Xinliang David Li  <davidxl@google.com>
1764         PR tree-optimization/59303
1765         * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
1766         (dump_predicates): Better output format.
1767         (pred_equal_p): New function.
1768         (is_neq_relop_p): Ditto.
1769         (is_neq_zero_form_p): Ditto.
1770         (pred_expr_equal_p): Ditto.
1771         (pred_neg_p): Ditto.
1772         (simplify_pred): Ditto.
1773         (simplify_preds_2): Ditto.
1774         (simplify_preds_3): Ditto.
1775         (simplify_preds_4): Ditto.
1776         (simplify_preds): Ditto.
1777         (push_pred): Ditto.
1778         (push_to_worklist): Ditto.
1779         (get_pred_info_from_cmp): Ditto.
1780         (is_degenerated_phi): Ditto.
1781         (normalize_one_pred_1): Ditto.
1782         (normalize_one_pred): Ditto.
1783         (normalize_one_pred_chain): Ditto.
1784         (normalize_preds): Ditto.
1785         (normalize_cond_1): Remove function.
1786         (normalize_cond): Ditto.
1787         (is_gcond_subset_of): Ditto.
1788         (is_subset_of_any): Ditto.
1789         (is_or_set_subset_of): Ditto.
1790         (is_and_set_subset_of): Ditto.
1791         (is_norm_cond_subset_of): Ditto.
1792         (pred_chain_length_cmp): Ditto.
1793         (convert_control_dep_chain_into_preds): Type change.
1794         (find_predicates): Ditto.
1795         (find_def_preds): Ditto.
1796         (destroy_predicates_vecs): Ditto.
1797         (find_matching_predicates_in_rest_chains): Ditto.
1798         (use_pred_not_overlap_with_undef_path_pred): Ditto.
1799         (is_pred_expr_subset): Ditto.
1800         (is_pred_chain_subset_of): Ditto.
1801         (is_included_in): Ditto.
1802         (is_superset_of): Ditto.
1804 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
1806         Update copyright years.
1808 2014-01-02  Richard Sandiford  <rdsandiford@googlemail.com>
1810         * common/config/arc/arc-common.c, config/arc/arc-modes.def,
1811         config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
1812         config/arc/arc.md, config/arc/arc.opt,
1813         config/arm/arm_neon_builtins.def, config/arm/crypto.def,
1814         config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
1815         config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
1816         config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
1817         config/linux-protos.h, config/linux.c, config/winnt-c.c,
1818         diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
1819         vtable-verify.c, vtable-verify.h: Use the standard form for the
1820         copyright notice.
1822 2014-01-02  Tobias Burnus  <burnus@net-b.de>
1824         * gcc.c (process_command): Update copyright notice dates.
1825         * gcov-dump.c: Ditto.
1826         * gcov.c: Ditto.
1827         * doc/cpp.texi: Bump @copying's copyright year.
1828         * doc/cppinternals.texi: Ditto.
1829         * doc/gcc.texi: Ditto.
1830         * doc/gccint.texi: Ditto.
1831         * doc/gcov.texi: Ditto.
1832         * doc/install.texi: Ditto.
1833         * doc/invoke.texi: Ditto.
1835 2014-01-01  Jan-Benedict Glaw  <jbglaw@lug-owl.de>
1837         * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
1839 2014-01-01  Jakub Jelinek  <jakub@redhat.com>
1841         * config/i386/sse.md (*mov<mode>_internal): Guard
1842         EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
1844         PR rtl-optimization/59647
1845         * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
1846         new_rtx into UNSIGNED_FLOAT rtxes.
1848 Copyright (C) 2014 Free Software Foundation, Inc.
1850 Copying and distribution of this file, with or without modification,
1851 are permitted in any medium without royalty provided the copyright
1852 notice and this notice are preserved.