1 2014-02-25 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3 * config/rs6000/vector.md (*vector_ordered<mode>): Change split to
4 use canonical form for nor<mode>3.
5 (*vector_unordered<mode>): Likewise.
7 2014-02-25 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10 * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Allow 128 to 64-bit
13 2014-02-25 Ilya Tocar <ilya.tocar@intel.com>
15 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PREFETCHWT1_SET),
16 (OPTION_MASK_ISA_PREFETCHWT1_UNSET): New.
17 (ix86_handle_option): Handle OPT_mprefetchwt1.
18 * config/i386/cpuid.h (bit_PREFETCHWT1): New.
19 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
21 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
22 OPTION_MASK_ISA_PREFETCHWT1.
23 * config/i386/i386.c (ix86_target_string): Handle mprefetchwt1.
24 (PTA_PREFETCHWT1): New.
25 (ix86_option_override_internal): Handle PTA_PREFETCHWT1.
26 (ix86_valid_target_attribute_inner_p): Handle OPT_mprefetchwt1.
27 * config/i386/i386.h (TARGET_PREFETCHWT1), (TARGET_PREFETCHWT1_P):
29 * config/i386/i386.md (prefetch): Check TARGET_PREFETCHWT1
30 (*prefetch_avx512pf_<mode>_: Change into ...
31 (*prefetch_prefetchwt1_<mode>: This.
32 * config/i386/i386.opt (mprefetchwt1): New.
33 * config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET1.
34 (_mm_prefetch): Handle intent to write.
35 * doc/invoke.texi (mprefetchwt1), (mno-prefetchwt1): Doccument.
37 2014-02-25 Richard Biener <rguenther@suse.de>
40 * emit-rtl.c (mem_attrs_htab): Remove.
41 (mem_attrs_htab_hash): Likewise.
42 (mem_attrs_htab_eq): Likewise.
43 (set_mem_attrs): Always allocate new mem-attrs when something
45 (init_emit_once): Do not allocate mem_attrs_htab.
47 2014-02-25 Richard Biener <rguenther@suse.de>
50 * lto-opts.c (lto_write_options): Output non-explicit conservative
51 -fwrapv, -fno-trapv and -fno-strict-overflow.
52 * lto-wrapper.c (merge_and_complain): Handle merging those options.
53 (run_gcc): And pass them through.
55 2014-02-25 Andrey Belevantsev <abel@ispras.ru>
57 * sel-sched.c (calculate_new_fences): New parameter ptime.
58 Calculate it as a maximum over all fence cycles.
59 (sel_sched_region_2): Adjust the call to calculate_new_fences.
60 Print the final schedule timing when sched_verbose.
62 2014-02-25 Andrey Belevantsev <abel@ispras.ru>
64 PR rtl-optimization/60292
65 * sel-sched.c (fill_vec_av_set): Do not reset target availability
66 bit fot the fence instruction.
68 2014-02-24 Alangi Derick <alangiderick@gmail.com>
70 * calls.h: Fix typo in comment.
72 2014-02-24 John David Anglin <danglin@gcc.gnu.org>
74 * config/pa/pa.c (pa_output_move_double): Don't valididate when
75 adjusting offsetable addresses.
77 2014-02-24 Guozhi Wei <carrot@google.com>
79 * sparseset.h (sparseset_pop): Fix the wrong index.
81 2014-02-24 Walter Lee <walt@tilera.com>
83 * config.gcc (tilepro-*-*): Change to tilepro*-*-*.
84 (tilegx-*-linux*): Change to tilegx*-*-linux*; Support tilegxbe
86 * common/config/tilegx/tilegx-common.c
87 (TARGET_DEFAULT_TARGET_FLAGS): Define.
88 * config/tilegx/linux.h (ASM_SPEC): Add endian_spec.
90 * config/tilegx/sync.md (atomic_test_and_set): Handle big endian.
91 * config/tilegx/tilegx.c (tilegx_return_in_msb): New.
92 (tilegx_gimplify_va_arg_expr): Handle big endian.
93 (tilegx_expand_unaligned_load): Ditto.
94 (tilegx_expand_unaligned_store): Ditto.
95 (TARGET_RETURN_IN_MSB): New.
96 * config/tilegx/tilegx.h (TARGET_DEFAULT): New.
97 (TARGET_ENDIAN_DEFAULT): New.
98 (TARGET_BIG_ENDIAN): Handle big endian.
99 (BYTES_BIG_ENDIAN): Ditto.
100 (WORDS_BIG_ENDIAN): Ditto.
101 (FLOAT_WORDS_BIG_ENDIAN): Ditto.
104 * config/tilegx/tilegx.md (extv): Handle big endian.
107 (insn_st<n>_add<bitsuffix>): Ditto.
108 (insn_stnt<n>): Ditto.
109 (insn_stnt<n>_add<bitsuffix>):Ditto.
110 (vec_interleave_highv8qi): Handle big endian.
111 (vec_interleave_highv8qi_be): New.
112 (vec_interleave_highv8qi_le): New.
113 (insn_v1int_h): Handle big endian.
114 (vec_interleave_lowv8qi): Handle big endian.
115 (vec_interleave_lowv8qi_be): New.
116 (vec_interleave_lowv8qi_le): New.
117 (insn_v1int_l): Handle big endian.
118 (vec_interleave_highv4hi): Handle big endian.
119 (vec_interleave_highv4hi_be): New.
120 (vec_interleave_highv4hi_le): New.
121 (insn_v2int_h): Handle big endian.
122 (vec_interleave_lowv4hi): Handle big endian.
123 (vec_interleave_lowv4hi_be): New.
124 (vec_interleave_lowv4hi_le): New.
125 (insn_v2int_l): Handle big endian.
126 (vec_interleave_highv2si): Handle big endian.
127 (vec_interleave_highv2si_be): New.
128 (vec_interleave_highv2si_le): New.
129 (insn_v4int_h): Handle big endian.
130 (vec_interleave_lowv2si): Handle big endian.
131 (vec_interleave_lowv2si_be): New.
132 (vec_interleave_lowv2si_le): New.
133 (insn_v4int_l): Handle big endian.
134 * config/tilegx/tilegx.opt (mbig-endian): New option.
135 (mlittle-endian): New option.
136 * doc/install.texi: Document tilegxbe-linux.
137 * doc/invoke.texi: Document -mbig-endian and -mlittle-endian.
139 2014-02-24 Martin Jambor <mjambor@suse.cz>
142 * ipa-cp.c (propagate_constants_accross_call): Bail out early if
143 there are no parameter descriptors.
145 2014-02-24 Andrey Belevantsev <abel@ispras.ru>
147 PR rtl-optimization/60268
148 * sched-rgn.c (haifa_find_rgns): Move the nr_regions_initial variable
149 initialization to ...
150 (sched_rgn_init): ... here.
151 (schedule_region): Check for SCHED_PRESSURE_NONE earlier.
153 2014-02-23 David Holsgrove <david.holsgrove@xilinx.com>
155 * config/microblaze/microblaze.md: Correct ashrsi_reg / lshrsi_reg
158 2014-02-23 Edgar E. Iglesias <edgar.iglesias@xilinx.com>
160 * config/microblaze/microblaze.h: Remove SECONDARY_MEMORY_NEEDED
163 2014-02-23 David Holsgrove <david.holsgrove@xilinx.com>
165 * /config/microblaze/microblaze.c: Add microblaze_asm_output_mi_thunk
166 and define TARGET_ASM_OUTPUT_MI_THUNK and
167 TARGET_ASM_CAN_OUTPUT_MI_THUNK.
169 2014-02-23 David Holsgrove <david.holsgrove@xilinx.com>
171 * config/microblaze/predicates.md: Add cmp_op predicate.
172 * config/microblaze/microblaze.md: Add branch_compare instruction
173 which uses cmp_op predicate and emits cmp insn before branch.
174 * config/microblaze/microblaze.c (microblaze_emit_compare): Rename
175 to microblaze_expand_conditional_branch and consolidate logic.
176 (microblaze_expand_conditional_branch): emit branch_compare
177 insn instead of handling cmp op separate from branch insn.
179 2014-02-23 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
181 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Relax assert
184 2014-02-23 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
186 * config/rs6000/altivec.md (altivec_lve<VI_char>x): Replace
187 define_insn with define_expand and new define_insn
188 *altivec_lve<VI_char>x_internal.
189 (altivec_stve<VI_char>x): Replace define_insn with define_expand
190 and new define_insn *altivec_stve<VI_char>x_internal.
191 * config/rs6000/rs6000-protos.h (altivec_expand_stvex_be): New
193 * config/rs6000/rs6000.c (altivec_expand_lvx_be): Document use by
195 (altivec_expand_stvex_be): New function.
197 2014-02-22 Joern Rennecke <joern.rennecke@embecosm.com>
199 * config/avr/avr.c (avr_can_eliminate): Allow elimination from
201 to STACK_POINTER_REGNUM if !frame_pointer_needed.
202 * config/avr/avr.c (ELIMINABLE_REGS): Add elimination from
203 ARG_POINTER_REGNUM to STACK_POINTER_REGNUM.
205 2014-02-21 Vladimir Makarov <vmakarov@redhat.com>
208 * lra-constraints.c (inherit_reload_reg): Use lra_emit_move
209 instead of emit_move_insn.
211 2014-02-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
213 * config/rs6000/altivec.md (altivec_vsumsws): Replace second
215 (reduc_uplus_v16qi): Use gen_altivec_vsumsws_direct instead of
218 2014-02-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
220 * config/rs6000/altivec.md (altivec_lvxl): Rename as
221 *altivec_lvxl_<mode>_internal and use VM2 iterator instead of
223 (altivec_lvxl_<mode>): New define_expand incorporating
224 -maltivec=be semantics where needed.
225 (altivec_lvx): Rename as *altivec_lvx_<mode>_internal.
226 (altivec_lvx_<mode>): New define_expand incorporating -maltivec=be
227 semantics where needed.
228 (altivec_stvx): Rename as *altivec_stvx_<mode>_internal.
229 (altivec_stvx_<mode>): New define_expand incorporating
230 -maltivec=be semantics where needed.
231 (altivec_stvxl): Rename as *altivec_stvxl_<mode>_internal and use
232 VM2 iterator instead of V4SI.
233 (altivec_stvxl_<mode>): New define_expand incorporating
234 -maltivec=be semantics where needed.
235 * config/rs6000/rs6000-builtin.def: Add new built-in definitions
236 LVXL_V2DF, LVXL_V2DI, LVXL_V4SF, LVXL_V4SI, LVXL_V8HI, LVXL_V16QI,
237 LVX_V2DF, LVX_V2DI, LVX_V4SF, LVX_V4SI, LVX_V8HI, LVX_V16QI,
238 STVX_V2DF, STVX_V2DI, STVX_V4SF, STVX_V4SI, STVX_V8HI, STVX_V16QI,
239 STVXL_V2DF, STVXL_V2DI, STVXL_V4SF, STVXL_V4SI, STVXL_V8HI,
241 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Replace
242 ALTIVEC_BUILTIN_LVX with ALTIVEC_BUILTIN_LVX_<MODE> throughout;
243 similarly for ALTIVEC_BUILTIN_LVXL, ALTIVEC_BUILTIN_STVX, and
244 ALTIVEC_BUILTIN_STVXL.
245 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be): New
247 (altivec_expand_stvx_be): Likewise.
248 * config/rs6000/rs6000.c (swap_selector_for_mode): New function.
249 (altivec_expand_lvx_be): Likewise.
250 (altivec_expand_stvx_be): Likewise.
251 (altivec_expand_builtin): Add cases for
252 ALTIVEC_BUILTIN_STVX_<MODE>, ALTIVEC_BUILTIN_STVXL_<MODE>,
253 ALTIVEC_BUILTIN_LVXL_<MODE>, and ALTIVEC_BUILTIN_LVX_<MODE>.
254 (altivec_init_builtins): Add definitions for
255 __builtin_altivec_lvxl_<mode>, __builtin_altivec_lvx_<mode>,
256 __builtin_altivec_stvx_<mode>, and
257 __builtin_altivec_stvxl_<mode>.
259 2014-02-21 Catherine Moore <clm@codesourcery.com>
261 * doc/invoke.texi (mvirt, mno-virt): Document.
262 * config/mips/mips.opt (mvirt): New option.
263 * config/mips/mips.h (ASM_SPEC): Pass mvirt to the assembler.
265 2014-02-21 Richard Biener <rguenther@suse.de>
267 PR tree-optimization/60276
268 * tree-vectorizer.h (struct _stmt_vec_info): Add min_neg_dist field.
269 (STMT_VINFO_MIN_NEG_DIST): New macro.
270 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Record
271 STMT_VINFO_MIN_NEG_DIST.
272 * tree-vect-stmts.c (vectorizable_load): Verify if assumptions
273 made for negative dependence distances still hold.
275 2014-02-21 Richard Biener <rguenther@suse.de>
278 * tree-ssa-live.c (mark_all_vars_used_1): Do not walk
279 DECL_INITIAL for globals not in the current function context.
281 2014-02-21 Jakub Jelinek <jakub@redhat.com>
283 PR tree-optimization/56490
284 * params.def (PARAM_UNINIT_CONTROL_DEP_ATTEMPTS): New param.
285 * tree-ssa-uninit.c: Include params.h.
286 (compute_control_dep_chain): Add num_calls argument, return false
287 if it exceed PARAM_UNINIT_CONTROL_DEP_ATTEMPTS param, pass
288 num_calls to recursive call.
289 (find_predicates): Change dep_chain into normal array,
290 cur_chain into auto_vec<edge, MAX_CHAIN_LEN + 1>, add num_calls
291 variable and adjust compute_control_dep_chain caller.
292 (find_def_preds): Likewise.
294 2014-02-21 Thomas Schwinge <thomas@codesourcery.com>
296 * gimple-pretty-print.c (dump_gimple_omp_for) [flags & TDF_RAW]
297 <case GF_OMP_FOR_KIND_CILKSIMD>: Add missing break statement.
299 2014-02-21 Nick Clifton <nickc@redhat.com>
301 * config/stormy16/stormy16.md (pushdqi1): Add mode to post_inc.
303 (popqi1): Add mode to pre_dec.
306 2014-02-21 Jakub Jelinek <jakub@redhat.com>
308 * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
309 mode for mask of V8SFmode permutation.
311 2014-02-20 Richard Henderson <rth@redhat.com>
314 * builtins.c (expand_builtin_atomic_compare_exchange): Always make
315 a new pseudo for OLDVAL.
317 2014-02-20 Jakub Jelinek <jakub@redhat.com>
320 * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
321 gen_reg_rtx if d->testing_p.
322 (expand_vec_perm_pshufb2, expand_vec_perm_broadcast_1): Return early
323 if d->testing_p and we will certainly return true.
324 (expand_vec_perm_even_odd_1): Likewise. Don't call gen_reg_rtx
327 2014-02-20 Uros Bizjak <ubizjak@gmail.com>
329 * emit-rtl.c (gen_reg_rtx): Assert that
330 crtl->emit.regno_pointer_align_length is non-zero.
332 2014-02-20 Richard Henderson <rth@redhat.com>
335 * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
336 on failure the store back into EXPECT.
338 2014-02-20 Chung-Lin Tang <cltang@codesourcery.com>
339 Sandra Loosemore <sandra@codesourcery.com>
341 * config/nios2/nios2.md (unspec): Add UNSPEC_PIC_GOTOFF_SYM enum.
342 * config/nios2/nios2.c (nios2_function_profiler):
343 Add -fPIC (flag_pic == 2) support.
344 (nios2_handle_custom_fpu_cfg): Fix warning parameter.
345 (nios2_large_offset_p): New function.
346 (nios2_unspec_reloc_p): Move up position, update to use
347 nios2_large_offset_p.
348 (nios2_unspec_address): Remove function.
349 (nios2_unspec_offset): New function.
350 (nios2_large_got_address): New function.
351 (nios2_got_address): Add large offset support.
352 (nios2_legitimize_tls_address): Update usage of removed and new
354 (nios2_symbol_binds_local_p): New function.
355 (nios2_load_pic_address): Add -fPIC (flag_pic == 2) support.
356 (nios2_legitimize_address): Update to use nios2_large_offset_p.
357 (nios2_emit_move_sequence): Avoid legitimizing (const (unspec ...)).
358 (nios2_print_operand): Merge H/L processing, add hiadj/lo
359 processing for (const (unspec ...)).
360 (nios2_unspec_reloc_name): Add UNSPEC_PIC_GOTOFF_SYM case.
362 2014-02-20 Richard Biener <rguenther@suse.de>
364 * tree-cfg.c (replace_uses_by): Mark altered BBs before
365 doing the substitution.
366 (verify_gimple_assign_single): Also verify bare MEM_REFs on the lhs.
368 2014-02-20 Martin Jambor <mjambor@suse.cz>
371 * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
372 info when checking whether lattices are bottom.
374 2014-02-20 Richard Biener <rguenther@suse.de>
377 * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
380 2014-02-20 Jan Hubicka <hubicka@ucw.cz>
383 * ipa-inline-transform.c (clone_inlined_nodes): Add freq_scale
384 parameter specifying the scaling.
385 (inline_call): Update.
386 (want_inline_recursively): Guard division by zero.
387 (recursive_inlining): Update.
388 * ipa-inline.h (clone_inlined_nodes): Update.
390 2014-02-20 Ilya Tocar <ilya.tocar@intel.com>
393 * config/i386/i386.c (classify_argument): Pass structures of size
394 64 bytes or less in register.
396 2014-02-20 Ilya Tocar <ilya.tocar@intel.com>
397 Kirill Yukhin <kirill.yukhin@intel.com>
399 * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): Swap operands.
400 (_mm_rcp28_round_ss): Ditto.
401 (_mm_rsqrt28_round_sd): Ditto.
402 (_mm_rsqrt28_round_ss): Ditto.
403 * config/i386/avx512erintrin.h (_mm_rcp14_round_sd): Ditto.
404 (_mm_rcp14_round_ss): Ditto.
405 (_mm_rsqrt14_round_sd): Ditto.
406 (_mm_rsqrt14_round_ss): Ditto.
407 * config/i386/sse.md (rsqrt14<mode>): Put nonimmediate operand as
408 the first input operand, get rid of match_dup.
409 (avx512er_exp2<mode><mask_name><round_saeonly_name>): Set type
411 (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
413 (avx512er_vmrcp28<mode><round_saeonly_name>): Put nonimmediate
414 operand as the first input operand, set type attribute.
415 (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
417 (avx512er_vmrsqrt28<mode><round_saeonly_name>): Put nonimmediate
418 operand as the first input operand, set type attribute.
420 2014-02-19 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
422 * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
425 2014-02-19 H.J. Lu <hongjiu.lu@intel.com>
428 * config/i386/i386.c (construct_container): Remove TFmode check
429 for X86_64_INTEGER_CLASS.
431 2014-02-19 Uros Bizjak <ubizjak@gmail.com>
434 * config/i386/i386.c (type_natural_mode): Warn for ABI changes
435 only when -Wpsabi is enabled.
437 2014-02-19 Michael Hudson-Doyle <michael.hudson@linaro.org>
440 * config/aarch64/aarch64.c (aarch64_pass_by_reference): The rules for
441 passing arrays in registers are the same as for structs, so remove the
442 special case for them.
444 2014-02-19 Eric Botcazou <ebotcazou@adacore.com>
446 * expr.c (expand_expr_real_1) <case VIEW_CONVERT_EXPR>: For a bit-field
447 destination type, extract only the valid bits if the source type is not
448 integral and has a different mode.
450 2014-02-19 Richard Biener <rguenther@suse.de>
453 * tree-inline.c (estimate_num_insns): Avoid calling cgraph_get_node
456 2014-02-19 Richard Biener <rguenther@suse.de>
459 * ipa-prop.c: Include stringpool.h and tree-ssanames.h.
460 (ipa_modify_call_arguments): Emit an argument load explicitely and
461 preserve virtual SSA form there and for the replacement call.
462 Do not update SSA form nor free dominance info.
464 2014-02-18 Jan Hubicka <hubicka@ucw.cz>
466 * ipa.c (function_and_variable_visibility): Also clear WEAK
467 flag when disolving COMDAT_GROUP.
469 2014-02-18 Jan Hubicka <hubicka@ucw.cz>
471 * ipa-prop.h (ipa_ancestor_jf_data): Update ocmment.
472 * ipa-prop.c (ipa_set_jf_known_type): Return early when
474 (ipa_set_ancestor_jf): Set type to NULL hwen it is not preserved;
475 do more sanity checks.
476 (detect_type_change): Return true when giving up early.
477 (compute_complex_assign_jump_func): Fix type parameter of
479 (compute_complex_ancestor_jump_func): Likewise.
480 (update_jump_functions_after_inlining): Fix updating of
482 * ipa-cp.c (ipa_get_jf_ancestor_result): Be ready for type to be NULL.
484 2014-02-18 Jan Hubicka <hubicka@ucw.cz>
486 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
487 inline clones when edge disappears.
489 2014-02-18 Michael Meissner <meissner@linux.vnet.ibm.com>
492 * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
493 Split 64-bit moves into 2 patterns. Do not allow the use of
494 direct move for TDmode in little endian, since the decimal value
495 has little endian bytes within a word, but the 64-bit pieces are
496 ordered in a big endian fashion, and normal subreg's of TDmode are
498 (mov<mode>_64bit_dm): Likewise.
499 (movtd_64bit_nodm): Likewise.
501 2014-02-18 Eric Botcazou <ebotcazou@adacore.com>
503 PR tree-optimization/60174
504 * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
505 statement of an SSA_NAME that occurs in an abnormal PHI node.
507 2014-02-18 Jakub Jelinek <jakub@redhat.com>
510 * final.c (SEEN_BB): Remove.
511 (SEEN_NOTE, SEEN_EMITTED): Renumber.
512 (final_scan_insn): Don't force_source_line on second
513 NOTE_INSN_BASIC_BLOCK.
515 2014-02-18 Uros Bizjak <ubizjak@gmail.com>
518 * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
519 * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
520 (type_natural_mode): Warn ABI change when %zmm register is not
521 available for AVX512F vector value passing.
523 2014-02-18 Kai Tietz <ktietz@redhat.com>
526 * config/i386/i386.c (ix86_expand_prologue): Use value in
527 rax register as displacement when restoring %r10 or %rax.
528 Fix wrong offset when restoring both registers.
530 2014-02-18 Eric Botcazou <ebotcazou@adacore.com>
532 * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
533 assertion with conditional return.
535 2014-02-18 Jakub Jelinek <jakub@redhat.com>
536 Uros Bizjak <ubizjak@gmail.com>
539 * config/i386/driver-i386.c (host_detect_local_cpu): If
540 YMM state is not saved by the OS, also clear has_f16c. Move
541 CPUID 0x80000001 handling before YMM state saving checking.
543 2014-02-18 Andrey Belevantsev <abel@ispras.ru>
545 PR rtl-optimization/58960
546 * haifa-sched.c (alloc_global_sched_pressure_data): New,
547 factored out from ...
548 (sched_init): ... here.
549 (free_global_sched_pressure_data): New, factored out from ...
550 (sched_finish): ... here.
551 * sched-int.h (free_global_sched_pressure_data): Declare.
552 * sched-rgn.c (nr_regions_initial): New static global.
553 (haifa_find_rgns): Initialize it.
554 (schedule_region): Disable sched-pressure for the newly
557 2014-02-17 Richard Biener <rguenther@suse.de>
559 * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
560 release SSA defs of pattern stmts.
562 2014-02-17 Richard Biener <rguenther@suse.de>
564 * tree-inline.c (expand_call_inline): Release the virtual
565 operand defined by the call we are about to inline.
567 2014-02-17 Richard Biener <rguenther@suse.de>
569 * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
571 2014-02-17 Kirill Yukhin <kirill.yukhin@intel.com>
572 Ilya Tocar <ilya.tocar@intel.com>
574 * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
575 arguments order in builtin.
576 (_mm512_permutexvar_epi64): Ditto.
577 (_mm512_mask_permutexvar_epi64): Ditto
578 (_mm512_maskz_permutexvar_epi32): Ditto
579 (_mm512_permutexvar_epi32): Ditto
580 (_mm512_mask_permutexvar_epi32): Ditto
582 2014-02-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
584 * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
585 (p8_vmrgow): Likewise.
587 2014-02-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
589 * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
592 2014-02-15 Michael Meissner <meissner@linux.vnet.ibm.com>
595 * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
596 (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
597 into 64-bit and 32-bit moves. On 64-bit moves, add support for
598 using direct move instructions on ISA 2.07. Also adjust
599 instruction length for 64-bit.
600 (mov<mode>_64bit, TFmode/TDmode): Likewise.
601 (mov<mode>_32bit, TFmode/TDmode): Likewise.
603 2014-02-15 Alan Modra <amodra@gmail.com>
607 * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
608 find_replacement on parts of insn rtl that might be reloaded.
610 2014-02-15 Richard Biener <rguenther@suse.de>
612 PR tree-optimization/60183
613 * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
614 (tree_ssa_phiprop): Calculate and free post-dominators.
616 2014-02-14 Jeff Law <law@redhat.com>
618 PR rtl-optimization/60131
619 * ree.c (get_extended_src_reg): New function.
620 (combine_reaching_defs): Use it rather than assuming location of REG.
621 (find_and_remove_re): Verify first operand of extension is
622 a REG before adding the insns to the copy list.
624 2014-02-14 Roland McGrath <mcgrathr@google.com>
626 * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
627 * configure: Regenerated.
628 * config.in: Regenerated.
629 * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
630 instead of ASM_SHORT.
632 2014-02-14 Vladimir Makarov <vmakarov@redhat.com>
633 Richard Earnshaw <rearnsha@arm.com>
635 PR rtl-optimization/59535
636 * lra-constraints.c (process_alt_operands): Encourage alternative
637 when unassigned pseudo class is superset of the alternative class.
638 (inherit_reload_reg): Don't inherit when optimizing for code size.
639 * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
640 returning CORE_REGS for anything but Thumb1 and BASE_REGS for
641 modes not less than 4 for Thumb1.
643 2014-02-14 Kyle McMartin <kyle@redhat.com>
646 * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
648 2014-02-14 Richard Biener <rguenther@suse.de>
650 * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
651 (get_frame_arg): Drop the assert with langhook types_compatible_p.
652 Do not strip INDIRECT_REFs.
654 2014-02-14 Richard Biener <rguenther@suse.de>
657 * lto-streamer-out.c (DFS_write_tree_body): Do not follow
658 DECL_FUNCTION_SPECIFIC_TARGET.
659 (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
660 * tree-streamer-out.c (pack_ts_target_option): Remove.
661 (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
662 (write_ts_function_decl_tree_pointers): Do not stream
663 DECL_FUNCTION_SPECIFIC_TARGET.
664 * tree-streamer-in.c (unpack_ts_target_option): Remove.
665 (unpack_value_fields): Do not stream TS_TARGET_OPTION.
666 (lto_input_ts_function_decl_tree_pointers): Do not stream
667 DECL_FUNCTION_SPECIFIC_TARGET.
669 2014-02-14 Jakub Jelinek <jakub@redhat.com>
671 * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
672 (get_initial_def_for_induction, vectorizable_induction): Ignore
673 debug stmts when looking for exit_phi.
674 (vectorizable_live_operation): Fix up condition.
676 2014-02-14 Chung-Ju Wu <jasonwucj@gmail.com>
678 * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
679 nreverse() because it changes the content of original tree list.
681 2014-02-14 Chung-Ju Wu <jasonwucj@gmail.com>
683 * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
684 * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
686 2014-02-14 Chung-Ju Wu <jasonwucj@gmail.com>
688 * config/nds32/nds32.c (nds32_naked_function_p): Follow the
689 GNU coding standards.
691 2014-02-13 Jakub Jelinek <jakub@redhat.com>
694 * dwarf2out.c (gen_subprogram_die): Don't call
695 add_calling_convention_attribute if subr_die is old_die.
697 2014-02-13 Sharad Singhai <singhai@google.com>
699 * doc/optinfo.texi: Fix order of nodes.
701 2014-02-13 Uros Bizjak <ubizjak@gmail.com>
703 * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
704 operands[2], not operands[3].
706 2014-02-13 Richard Biener <rguenther@suse.de>
709 * doc/install.texi (ISL): Update recommended version to 0.12.2,
710 mention the possibility of an in-tree build.
711 (CLooG): Update recommended version to 0.18.1, mention the
712 possibility of an in-tree build and clarify that the ISL
713 bundled with CLooG does not work.
715 2014-02-13 Jakub Jelinek <jakub@redhat.com>
718 * expr.c (compress_float_constant): If x is a hard register,
719 extend into a pseudo and then move to x.
721 2014-02-13 Dominik Vogt <vogt@linux.vnet.ibm.com>
723 * config/s390/s390.c (s390_asm_output_function_label): Fix crash
724 caused by bad second argument to warning_at() with -mhotpatch and
725 nested functions (e.g. with gfortran).
727 2014-02-13 Richard Sandiford <rdsandiford@googlemail.com>
729 * opts.c (option_name): Remove "enabled by default" rider.
731 2014-02-12 John David Anglin <danglin@gcc.gnu.org>
733 * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
735 2014-02-12 H.J. Lu <hongjiu.lu@intel.com>
736 Uros Bizjak <ubizjak@gmail.com>
739 * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
740 * configure: Regenerated.
742 2014-02-12 Richard Biener <rguenther@suse.de>
744 * vec.c (vec_prefix::calculate_allocation): Move as
745 inline variant to vec.h.
746 (vec_prefix::calculate_allocation_1): New out-of-line version.
747 * vec.h (vec_prefix::calculate_allocation_1): Declare.
748 (vec_prefix::m_has_auto_buf): Rename to ...
749 (vec_prefix::m_using_auto_storage): ... this.
750 (vec_prefix::calculate_allocation): Inline the easy cases
751 and dispatch to calculate_allocation_1 which doesn't need the
753 (va_heap::reserve): Use gcc_checking_assert.
754 (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
755 m_using_auto_storage.
756 (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
758 (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
759 (vec<T, va_heap, vl_ptr>::release): Avoid casting.
760 (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
762 2014-02-12 Richard Biener <rguenther@suse.de>
764 * gcse.c (compute_transp): break from loop over canon_modify_mem_list
765 when we found a dependence.
767 2014-02-12 Thomas Schwinge <thomas@codesourcery.com>
769 * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
771 (maybe_fold_stmt): ... into this new function.
772 * omp-low.c (lower_omp): Update comment.
774 * omp-low.c (lower_omp_target): Add clobber for sizes array, after
777 * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
780 2014-02-12 James Greenhalgh <james.greenhalgh@arm.com>
782 * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
783 identifiers in comments.
784 (cortexa53_extra_costs): Likewise.
785 * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
786 (cortexa7_extra_costs): Likewise.
787 (cortexa12_extra_costs): Likewise.
788 (cortexa15_extra_costs): Likewise.
789 (v7m_extra_costs): Likewise.
791 2014-02-12 Richard Biener <rguenther@suse.de>
794 * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
795 of posix_memalign being successful.
796 (lower_stmt): Restrict lowering of posix_memalign to when
797 -ftree-bit-ccp is enabled.
799 2014-02-12 Senthil Kumar Selvaraj <senthil_kumar.selvaraj@atmel.com>
801 * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
803 * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
805 2014-02-12 Eric Botcazou <ebotcazou@adacore.com>
807 PR rtl-optimization/60116
808 * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
809 other_insn once the combination has been validated.
811 2014-02-11 Jan Hubicka <hubicka@ucw.cz>
814 * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
816 * ipa-devirt.c: Include demangle.h
817 (odr_violation_reported): New static variable.
818 (add_type_duplicate): Update odr_violations.
819 (maybe_record_node): Add completep parameter; update it.
820 (record_target_from_binfo): Add COMPLETEP parameter;
822 (possible_polymorphic_call_targets_1): Likewise.
823 (struct polymorphic_call_target_d): Add nonconstruction_targets;
824 rename FINAL to COMPLETE.
825 (record_targets_from_bases): Sanity check we found the binfo;
826 fix COMPLETEP updating.
827 (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
828 parameter, fix computing of COMPLETEP.
829 (dump_possible_polymorphic_call_targets): Imrove readability of dump;
830 at LTO time do demangling.
831 (ipa_devirt): Use nonconstruction_targets; Improve dumps.
832 * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
834 (gimple_get_virt_method_for_binfo): Likewise.
835 * gimple-fold.h (gimple_get_virt_method_for_binfo,
836 gimple_get_virt_method_for_vtable): Update prototypes.
838 2014-02-11 Vladimir Makarov <vmakarov@redhat.com>
841 * genautomata.c (add_presence_absence): Fix typo with
842 {final_}presence_list.
844 2014-02-11 Michael Meissner <meissner@linux.vnet.ibm.com>
847 * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
848 for VSX/Altivec vectors that land in GPR registers.
850 2014-02-11 Richard Henderson <rth@redhat.com>
851 Jakub Jelinek <jakub@redhat.com>
854 * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
855 around drhs if type conversion to lacc->type is not useless.
857 2014-02-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
859 * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
861 (cortex-a57.cortex-a53): Likewise.
862 * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
864 2014-02-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
866 * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
869 2014-02-11 Renlin Li <Renlin.Li@arm.com>
871 * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
872 add_options_for_arm_vfp3.
874 2014-02-11 Jeff Law <law@redhat.com>
877 * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
878 object with an undesirable mode.
880 2014-02-11 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
883 * config/i386/sol2-9.h: New file.
884 * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
885 *-*-solaris2.9*): Use it.
887 2014-02-10 Nagaraju Mekala <nagaraju.mekala@xilinx.com>
889 * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
890 * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
892 2014-02-10 Nagaraju Mekala <nagaraju.mekala@xilinx.com>
894 * config/microblaze/microblaze.c: Extend mcpu version format
896 2014-02-10 David Holsgrove <david.holsgrove@xilinx.com>
898 * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
900 2014-02-10 Richard Henderson <rth@redhat.com>
903 * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
904 * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
905 ms-abi vs -mno-accumulate-outgoing-args.
906 (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
907 * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
910 2014-02-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
913 * cfgexpand.c (expand_asm_operands): Attach source location to
914 ASM_INPUT rtx objects.
915 * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
917 2014-02-10 Nick Clifton <nickc@redhat.com>
919 * config/mn10300/mn10300.c (popcount): New function.
920 (mn10300_expand_prologue): Include saved registers in stack usage
923 2014-02-10 Jeff Law <law@redhat.com>
926 * reload1.c (emit_input_reload_insns): Do not create invalid RTL
927 when changing the SET_DEST of a prior insn to avoid an input reload.
929 2014-02-10 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
931 * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
932 big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
933 -mcall-openbsd, or -mcall-linux.
934 (CC1_ENDIAN_BIG_SPEC): Remove.
935 (CC1_ENDIAN_LITTLE_SPEC): Remove.
936 (CC1_ENDIAN_DEFAULT_SPEC): Remove.
937 (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
938 (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
939 and %cc1_endian_default.
940 * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
942 2014-02-10 Richard Biener <rguenther@suse.de>
944 PR tree-optimization/60115
945 * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
946 MEM_REF handling. Properly verify that the accesses are not
947 out of the objects bound.
949 2014-02-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
951 * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
954 2014-02-10 Eric Botcazou <ebotcazou@adacore.com>
956 * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
957 proper constants and fix formatting.
958 (possible_polymorphic_call_targets): Fix formatting.
960 2014-02-10 Kirill Yukhin <kirill.yukhin@intel.com>
961 Ilya Tocar <ilya.tocar@intel.com>
963 * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
964 (_mm512_loadu_epi32): Renamed into...
965 (_mm512_loadu_si512): This.
966 (_mm512_storeu_epi32): Renamed into...
967 (_mm512_storeu_si512): This.
968 (_mm512_maskz_ceil_ps): Removed.
969 (_mm512_maskz_ceil_pd): Ditto.
970 (_mm512_maskz_floor_ps): Ditto.
971 (_mm512_maskz_floor_pd): Ditto.
972 (_mm512_floor_round_ps): Ditto.
973 (_mm512_floor_round_pd): Ditto.
974 (_mm512_ceil_round_ps): Ditto.
975 (_mm512_ceil_round_pd): Ditto.
976 (_mm512_mask_floor_round_ps): Ditto.
977 (_mm512_mask_floor_round_pd): Ditto.
978 (_mm512_mask_ceil_round_ps): Ditto.
979 (_mm512_mask_ceil_round_pd): Ditto.
980 (_mm512_maskz_floor_round_ps): Ditto.
981 (_mm512_maskz_floor_round_pd): Ditto.
982 (_mm512_maskz_ceil_round_ps): Ditto.
983 (_mm512_maskz_ceil_round_pd): Ditto.
984 (_mm512_expand_pd): Ditto.
985 (_mm512_expand_ps): Ditto.
986 * config/i386/i386.c (ix86_builtins): Remove
987 IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
989 * config/i386/predicates.md (const1256_operand): New.
990 (const_1_to_2_operand): Ditto.
991 * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
992 (*avx512pf_gatherpf<mode>sf_mask): Ditto.
993 (*avx512pf_gatherpf<mode>sf): Ditto.
994 (avx512pf_gatherpf<mode>df): Ditto.
995 (*avx512pf_gatherpf<mode>df_mask): Ditto.
996 (*avx512pf_gatherpf<mode>df): Ditto.
997 (avx512pf_scatterpf<mode>sf): Ditto.
998 (*avx512pf_scatterpf<mode>sf_mask): Ditto.
999 (*avx512pf_scatterpf<mode>sf): Ditto.
1000 (avx512pf_scatterpf<mode>df): Ditto.
1001 (*avx512pf_scatterpf<mode>df_mask): Ditto.
1002 (*avx512pf_scatterpf<mode>df): Ditto.
1003 (avx512f_expand<mode>): Removed.
1004 (<shift_insn><mode>3<mask_name>): Change predicate type.
1006 2014-02-08 Jakub Jelinek <jakub@redhat.com>
1008 * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
1009 not at the end of datarefs vector use ordered_remove to avoid
1010 reordering datarefs vector.
1013 * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
1014 mark local addressable non-static vars as GOVD_PRIVATE
1015 instead of GOVD_LOCAL.
1016 * omp-low.c (lower_omp_for): Move gimple_bind_vars
1017 and BLOCK_VARS of gimple_bind_block to new_stmt rather
1021 * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
1022 if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
1023 assume_aligned or alloc_align attributes.
1024 (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
1025 arguments. Handle also assume_aligned and alloc_align attributes.
1026 (evaluate_stmt): Adjust bit_value_assume_aligned caller. Handle
1027 calls to functions with assume_aligned or alloc_align attributes.
1028 * doc/extend.texi: Document assume_aligned and alloc_align attributes.
1030 2014-02-08 Terry Guo <terry.guo@arm.com>
1032 * doc/invoke.texi: Document ARM -march=armv7e-m.
1034 2014-02-08 Jakub Jelinek <jakub@redhat.com>
1036 * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
1037 flag on __cilkrts_rethrow builtin.
1040 * ipa-cp.c (determine_versionability): Fail at -O0
1041 or __attribute__((optimize (0))) or -fno-ipa-cp functions.
1042 * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
1045 2014-02-04 Jakub Jelinek <jakub@redhat.com>
1048 * tree-inline.c (copy_forbidden): Fail for
1049 __attribute__((optimize (0))) functions.
1051 2014-02-07 Jan Hubicka <hubicka@ucw.cz>
1053 * varpool.c: Include pointer-set.h.
1054 (varpool_remove_unreferenced_decls): Variables in other partitions
1055 will not be output; be however careful to not lose information
1058 2014-02-07 Jan Hubicka <hubicka@ucw.cz>
1060 * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
1061 lookup in the vtable constructor.
1063 2014-02-07 Jeff Law <law@redhat.com>
1066 * config/m68k/m68k.md (ashldi_extsi): Turn into a
1067 define_insn_and_split.
1069 * ipa-inline.c (inline_small_functions): Fix typos.
1071 2014-02-07 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
1073 * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
1074 (s390_can_use_return_insn): Declare.
1075 * config/s390/s390.h (EPILOGUE_USES): Define.
1076 * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
1078 (s390_chunkify_start): Handle return JUMP_LABELs.
1079 (s390_early_mach): Emit a main_pool instruction on the entry edge.
1080 (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
1081 (s390_can_use_return_insn): New functions.
1082 (s390_fix_long_loop_prediction): Handle conditional returns.
1083 (TARGET_SET_UP_BY_PROLOGUE): Define.
1084 * config/s390/s390.md (ANY_RETURN): New code iterator.
1085 (*creturn, *csimple_return, return, simple_return): New patterns.
1087 2014-02-07 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
1089 * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
1090 notes to each restore. Also add REG_CFA_DEF_CFA when restoring %r15.
1091 (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P. Update the
1092 REG_CFA_RESTORE list when deciding not to restore a register.
1094 2014-02-07 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
1096 * config/s390/s390.c: Include tree-pass.h and context.h.
1097 (s390_early_mach): New function, split out from...
1098 (s390_emit_prologue): ...here.
1099 (pass_data_s390_early_mach): New pass structure.
1100 (pass_s390_early_mach): New class.
1101 (s390_option_override): Create and register early_mach pass.
1102 Move to end of file.
1104 2014-02-07 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
1106 * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
1107 to match for the exit block.
1109 2014-02-07 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
1111 * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
1112 ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
1113 Reject misaligned operands.
1115 2014-02-07 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
1117 * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
1119 2014-02-07 Richard Biener <rguenther@suse.de>
1122 * gimple-low.c (lower_builtin_posix_memalign): New function.
1123 (lower_stmt): Call it to lower posix_memalign in a way
1124 to make alignment info accessible.
1126 2014-02-07 Jakub Jelinek <jakub@redhat.com>
1129 * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
1130 __builtin_setjmp_receiver.
1132 2014-02-07 Richard Biener <rguenther@suse.de>
1135 * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
1136 * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
1137 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
1138 Handle BUILT_IN_POSIX_MEMALIGN.
1139 (find_func_clobbers): Likewise.
1140 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
1141 (call_may_clobber_ref_p_1): Likewise.
1143 2014-02-06 Jan Hubicka <hubicka@ucw.cz>
1146 * ipa-devirt.c (record_target_from_binfo): Remove overactive
1149 2014-02-06 Jan Hubicka <hubicka@ucw.cz>
1152 * lto-cgraph.c (lto_output_node): Use
1153 symtab_get_symbol_partitioning_class.
1154 (lto_output_varpool_node): likewise.
1155 (symtab_get_symbol_partitioning_class): Move here from
1157 * cgraph.h (symbol_partitioning_class): Likewise.
1158 (symtab_get_symbol_partitioning_class): Declare.
1160 2014-02-06 Jan Hubicka <hubicka@ucw.cz>
1162 * ggc.h (ggc_internal_cleared_alloc): New macro.
1163 * vec.h (vec_safe_copy): Handle memory stats.
1164 * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
1165 * target-globals.c (save_target_globals): Likewise.
1167 2014-02-06 Jan Hubicka <hubicka@ucw.cz>
1170 * expr.c (emit_move_resolve_push): Export; be bit more selective
1171 on when to clear alias set.
1172 * expr.h (emit_move_resolve_push): Declare.
1173 * function.h (struct function): Add tail_call_marked.
1174 * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
1175 * config/i386/i386-protos.h (ix86_expand_push): Remove.
1176 * config/i386/i386.md (TImode move expander): De not call
1178 (FP push expanders): Preserve memory attributes.
1179 * config/i386/sse.md (push<mode>1): Remove.
1180 * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
1181 (ix86_expand_push): Remove.
1182 * config/i386/mmx.md (push<mode>1): Remove.
1184 2014-02-06 Jakub Jelinek <jakub@redhat.com>
1186 PR rtl-optimization/60030
1187 * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
1188 lopart with paradoxical subreg before shifting it up by hprec.
1190 2014-02-06 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1192 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
1193 Remove extra newline at end of file.
1194 * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
1195 (arm_issue_rate): Handle cortexa57.
1196 * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
1197 (cortex-a57.cortex-a53): Likewise.
1199 2014-02-06 Jakub Jelinek <jakub@redhat.com>
1202 * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
1203 don't record in REG_FRAME_RELATED_EXPR registers not set in that
1205 (arm_expand_prologue): Adjust all callers.
1206 (arm_unwind_emit_sequence): Allow saved, but not important for unwind
1207 info, registers also at the lowest numbered registers side. Use
1208 gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
1212 * var-tracking.c (adjust_mems): Before adding a SET to
1213 amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
1215 2014-02-06 Alan Modra <amodra@gmail.com>
1218 * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
1219 change SDmode to DDmode when lra_in_progress.
1221 2014-02-06 Jakub Jelinek <jakub@redhat.com>
1224 * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
1225 free_data_ref on the dr first, and before goto again also set dr
1226 to the next dr. For simd_lane_access, free old datarefs[i] before
1227 overwriting it. For get_vectype_for_scalar_type failure, don't
1228 free_data_ref if simd_lane_access.
1230 * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
1233 * tree.h (opts_for_fn): New inline function.
1234 (opt_for_fn): Define.
1235 * config/i386/i386.c (ix86_function_regparm): Use
1236 opt_for_fn (decl, optimize) instead of optimize.
1238 2014-02-06 Marcus Shawcroft <marcus.shawcroft@arm.com>
1240 * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
1241 for SYMBOL_REF in large memory model.
1243 2014-02-06 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1245 * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
1247 (cortex-a57): Likewise.
1248 (cortex-a57.cortex-a53): Likewise.
1250 2014-02-06 Yury Gribov <y.gribov@samsung.com>
1251 Kugan Vivekanandarajah <kuganv@linaro.org>
1253 * config/arm/arm.c (arm_vector_alignment_reachable): Check
1255 * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
1257 2014-02-06 Richard Biener <rguenther@suse.de>
1259 * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
1260 set_loop_copy and initialize_original_copy_tables.
1262 2014-02-06 Alex Velenko <Alex.Velenko@arm.com>
1264 * config/aarch64/aarch64-simd.md
1265 (aarch64_ashr_simddi): Change QI to SI.
1267 2014-02-05 Jan Hubicka <hubicka@ucw.cz>
1268 Jakub Jelinek <jakub@redhat.com>
1271 * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
1274 2014-02-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1276 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
1277 CODE_FOR_altivec_vpku[hw]um to
1278 CODE_FOR_altivec_vpku[hw]um_direct.
1279 * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
1280 UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
1281 (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
1282 UNSPEC_VUNPACK_LO_SIGN_DIRECT.
1284 2014-02-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1286 * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
1287 generation for -maltivec=be.
1288 (altivec_vsumsws): Simplify redundant test.
1290 2014-02-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1292 * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
1293 (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
1294 (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
1295 (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
1296 gen_altivec_vpkuwum.
1297 (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
1299 (altivec_vpks<VI_char>ss): Likewise.
1300 (altivec_vpks<VI_char>us): Likewise.
1301 (altivec_vpku<VI_char>us): Likewise.
1302 (altivec_vpku<VI_char>um): Likewise.
1303 (altivec_vpku<VI_char>um_direct): New (copy of
1304 altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
1306 (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
1307 target is little endian and -maltivec=be is not specified.
1308 (*altivec_vupkhs<VU_char>_direct): New (copy of
1309 altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
1310 (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
1311 target is little endian and -maltivec=be is not specified.
1312 (*altivec_vupkls<VU_char>_direct): New (copy of
1313 altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
1314 (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
1315 little endian and -maltivec=be is not specified.
1316 (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
1317 little endian and -maltivec=be is not specified.
1319 2014-02-05 Richard Henderson <rth@redhat.com>
1322 * combine-stack-adj.c: Revert r206943.
1323 * sched-int.h (struct deps_desc): Add last_args_size.
1324 * sched-deps.c (init_deps): Initialize it.
1325 (sched_analyze_insn): Add OUTPUT dependencies between insns that
1326 contain REG_ARGS_SIZE notes.
1328 2014-02-05 Jan Hubicka <hubicka@ucw.cz>
1330 * lto-cgraph.c (asm_nodes_output): Make global.
1331 * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
1332 * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
1333 (driver_handle_option): Handle OPT_fwpa.
1335 2014-02-05 Jakub Jelinek <jakub@redhat.com>
1338 * ipa-devirt.c (possible_polymorphic_call_targets): Fix
1339 a comment typo and formatting issue. If odr_hash hasn't been
1340 created, return vNULL and set *completep to false.
1343 * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
1344 bb with no successors.
1346 2014-02-05 James Greenhalgh <james.greenhalgh@arm.com>
1349 * doc/invoke.texi (-march): Clarify documentation for ARM.
1353 2014-02-05 Richard Biener <rguenther@suse.de>
1355 * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
1356 when not vectorizing because of too many alias checks.
1357 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
1358 Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
1360 2014-02-05 Nick Clifton <nickc@redhat.com>
1362 * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
1363 accept extended registers in any mode when compiling for the MN10300.
1365 2014-02-05 Yury Gribov <y.gribov@samsung.com>
1367 * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
1368 * ipa-inline.c (report_inline_failed_reason): Handle mismatched
1369 sanitization attributes.
1370 (can_inline_edge_p): Likewise.
1371 (sanitize_attrs_match_for_inline_p): New function.
1373 2014-02-04 Jan Hubicka <hubicka@ucw.cz>
1375 * ipa-prop.c (detect_type_change): Shor circuit testing of
1376 type changes on THIS pointer.
1378 2014-02-04 John David Anglin <danglin@gcc.gnu.org>
1381 * config/pa/pa.c (legitimize_tls_address): Return original address
1382 if not passed a SYMBOL_REF rtx.
1383 (hppa_legitimize_address): Call legitimize_tls_address for all TLS
1385 (pa_emit_move_sequence): Simplify TLS source operands.
1386 (pa_legitimate_constant_p): Reject all TLS constants.
1387 * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
1388 (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
1390 2014-02-04 Jan Hubicka <hubicka@ucw.cz>
1392 * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
1393 groups when we know they are controlled by LTO.
1394 * varasm.c (default_binds_local_p_1): If object is in other partition,
1395 it will be resolved locally.
1397 2014-02-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
1399 * config/host-linux.c (linux_gt_pch_use_address): Don't
1400 use SSIZE_MAX because it is not always defined.
1402 2014-02-04 Vladimir Makarov <vmakarov@redhat.com>
1405 * lra-constraints.c (need_for_split_p): Use more 3 reloads as
1406 threshold for pseudo splitting.
1407 (update_ebb_live_info): Process call argument hard registers and
1408 hard registers from insn definition too.
1409 (max_small_class_regs_num): New constant.
1410 (inherit_in_ebb): Update live hard regs through EBBs. Update
1411 reloads_num only for small register classes. Don't split for
1414 2014-02-04 Markus Trippelsdorf <markus@trippelsdorf.de>
1417 * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
1420 2014-02-04 Jan Hubicka <hubicka@ucw.cz>
1422 * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
1425 2014-02-04 Marek Polacek <polacek@redhat.com>
1427 * gdbinit.in (pel): Define.
1429 2014-02-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
1431 * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
1434 2014-02-04 Richard Biener <rguenther@suse.de>
1437 * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
1438 in function context local.
1439 (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
1440 * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
1441 similar to LTO_imported_decl_ref.
1443 2014-02-04 Jakub Jelinek <jakub@redhat.com>
1445 PR tree-optimization/60002
1446 * cgraphclones.c (build_function_decl_skip_args): Clear
1449 PR tree-optimization/60023
1450 * tree-if-conv.c (predicate_mem_writes): Pass true instead of
1451 false to gsi_replace.
1452 * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
1453 has been in some EH region and vec_stmt could throw, add
1454 vec_stmt into the same EH region.
1455 * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
1456 has no lhs, ignore it.
1457 * internal-fn.c (expand_MASK_LOAD): Likewise.
1460 * tree-inline.c (copy_forbidden): Fail for
1461 __attribute__((optimize (0))) functions.
1464 * omp-low.c (simd_clone_struct_copy): If from->inbranch
1465 is set, copy one less argument.
1466 (expand_simd_clones): Don't subtract clone_info->inbranch
1467 from simd_clone_struct_alloc argument.
1469 PR rtl-optimization/57915
1470 * recog.c (simplify_while_replacing): If all unary/binary/relational
1471 operation arguments are constant, attempt to simplify those.
1474 * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
1475 if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
1477 2014-02-04 Richard Biener <rguenther@suse.de>
1479 PR tree-optimization/60012
1480 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
1481 TBAA disambiguation to all DDRs.
1483 2014-02-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1486 * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
1487 (LINK_SPEC): Use it for -shared, -shared-libgcc.
1489 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
1492 * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
1494 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
1496 * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
1497 * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
1499 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
1502 * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
1503 to figure out targets of polymorphic calls with known decl.
1504 * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
1505 * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
1506 * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
1507 (get_polymorphic_call_info): ... here.
1508 (get_polymorphic_call_info_from_invariant): New function.
1510 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
1512 * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
1513 lookup via vtable pointer; check for type consistency
1514 and turn inconsitent facts into UNREACHABLE.
1515 * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
1516 * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
1517 type inconsistent querries; return UNREACHABLE instead.
1519 2014-02-03 Richard Henderson <rth@twiddle.net>
1522 * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
1523 already processed this node.
1524 (normalize_one_pred_1): Pass along mark_set.
1525 (normalize_one_pred): Create and destroy a pointer_set_t.
1526 (normalize_one_pred_chain): Likewise.
1528 2014-02-03 Laurent Aflonsi <laurent.alfonsi@st.com>
1530 PR gcov-profile/58602
1531 * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
1533 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
1536 * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
1537 -fno-devirtualize; try to devirtualize by the knowledge of
1538 virtual table pointer given by aggregate propagation.
1539 * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
1540 (ipa_print_node_jump_functions): Dump also offset that
1541 is relevant for polymorphic calls.
1542 (determine_known_aggregate_parts): Add arg_type parameter; use it
1543 instead of determining the type from pointer type.
1544 (ipa_compute_jump_functions_for_edge): Update call of
1545 determine_known_aggregate_parts.
1546 * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
1547 (gimple_get_virt_method_for_binfo): ... here; simplify using
1548 vtable_pointer_value_to_vtable.
1549 * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
1550 * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
1551 to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
1552 (vtable_pointer_value_to_vtable): Break out from ...; handle also
1554 (vtable_pointer_value_to_binfo): ... here.
1555 * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
1557 2014-02-03 Teresa Johnson <tejohnson@google.com>
1559 * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
1560 redef of outer loop index variable.
1562 2014-02-03 Marc Glisse <marc.glisse@inria.fr>
1566 * doc/extend.texi (Function Attributes): Typo.
1568 2014-02-03 Cong Hou <congh@google.com>
1570 PR tree-optimization/60000
1571 * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
1572 if the vectorized statement is a store. A store statement can only
1573 appear at the end of pattern statements.
1575 2014-02-03 H.J. Lu <hongjiu.lu@intel.com>
1577 * config/i386/i386.c (flag_opts): Add -mlong-double-128.
1578 (ix86_option_override_internal): Default long double to 64-bit for
1579 32-bit Bionic and to 128-bit for 64-bit Bionic.
1581 * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
1582 TARGET_LONG_DOUBLE_128 is true.
1583 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1585 * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
1586 (mlong-double-64): Negate -mlong-double-128.
1587 (mlong-double-128): New option.
1589 * config/i386/i386-c.c (ix86_target_macros): Define
1590 __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
1592 * doc/invoke.texi: Document -mlong-double-128.
1594 2014-02-03 H.J. Lu <hongjiu.lu@intel.com>
1596 PR rtl-optimization/60024
1597 * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
1599 2014-02-03 Markus Trippelsdorf <markus@trippelsdorf.de>
1601 * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
1603 2014-02-03 Andrey Belevantsev <abel@ispras.ru>
1605 PR rtl-optimization/57662
1606 * sel-sched.c (code_motion_path_driver): Do not mark already not
1607 existing blocks in the visiting bitmap.
1609 2014-02-03 Andrey Belevantsev <abel@ispras.ru>
1611 * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
1612 on the insn being emitted.
1614 2014-02-03 James Greenhalgh <james.greenhalgh@arm.com>
1615 Will Deacon <will.deacon@arm.com>
1617 * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
1619 2014-02-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1621 * config/arm/arm-tables.opt: Regenerate.
1623 2014-02-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1625 * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
1626 for vector types other than V16QImode.
1627 * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
1628 define_expand, and call altivec_expand_vec_perm_le when producing
1629 code with little endian element order.
1630 (*altivec_vperm_<mode>_internal): New insn having previous
1631 behavior of altivec_vperm_<mode>.
1632 (altivec_vperm_<mode>_uns): Change to a define_expand, and call
1633 altivec_expand_vec_perm_le when producing code with little endian
1635 (*altivec_vperm_<mode>_uns_internal): New insn having previous
1636 behavior of altivec_vperm_<mode>_uns.
1638 2014-02-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1640 * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
1641 (altivec_vsumsws): Add handling for -maltivec=be with a little
1643 (altivec_vsumsws_direct): New.
1644 (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
1645 gen_altivec_vsumsws.
1647 2014-02-02 Jan Hubicka <hubicka@ucw.cz>
1649 * ipa-devirt.c (subbinfo_with_vtable_at_offset,
1650 vtable_pointer_value_to_binfo): New functions.
1651 * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
1652 * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
1654 2014-02-02 Sandra Loosemore <sandra@codesourcery.com>
1656 * config/nios2/nios2.md (load_got_register): Initialize GOT
1657 pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
1658 * config/nios2/nios2.c (nios2_function_profiler): Likewise.
1660 2014-02-02 Jan Hubicka <hubicka@ucw.cz>
1662 * ipa-prop.c (update_jump_functions_after_inlining): When type is not
1663 preserverd by passthrough, do not propagate the type.
1665 2014-02-02 Richard Sandiford <rdsandiford@googlemail.com>
1667 * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
1668 (mips_atomic_assign_expand_fenv): New function.
1669 (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
1671 2014-02-02 Richard Sandiford <rdsandiford@googlemail.com>
1673 * doc/extend.texi (__builtin_mips_get_fcsr): Document.
1674 (__builtin_mips_set_fcsr): Likewise.
1675 * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
1676 MIPS_USI_FTYPE_VOID.
1677 * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
1678 (mips16_expand_set_fcsr): Likewise.
1679 * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
1680 (mips16_set_fcsr_stub): Likewise.
1681 (mips16_get_fcsr_one_only_stub): New class.
1682 (mips16_set_fcsr_one_only_stub): Likewise.
1683 (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
1684 (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
1685 (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
1686 (hard_float): New availability predicate.
1687 (mips_builtins): Add get_fcsr and set_fcsr.
1688 (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
1689 * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
1690 (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
1691 (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
1692 (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
1695 2014-02-02 Richard Sandiford <rdsandiford@googlemail.com>
1697 * config/mips/mips.c (mips_one_only_stub): New class.
1698 (mips_need_mips16_rdhwr_p): Replace with...
1699 (mips16_rdhwr_stub): ...this new variable.
1700 (mips16_stub_call_address): New function.
1701 (mips16_rdhwr_one_only_stub): New class.
1702 (mips_expand_thread_pointer): Use mips16_stub_call_address.
1703 (mips_output_mips16_rdhwr): Delete.
1704 (mips_finish_stub): New function.
1705 (mips_code_end): Use it to handle rdhwr stubs.
1707 2014-02-02 Uros Bizjak <ubizjak@gmail.com>
1710 * config/i386/i386.c (classify_argument): Fix handling of bit_offset
1711 when calculating size of integer atomic types.
1713 2014-02-02 H.J. Lu <hongjiu.lu@intel.com>
1715 * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
1717 2014-02-01 Jakub Jelinek <jakub@redhat.com>
1719 PR tree-optimization/60003
1720 * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
1721 * profile.c (branch_prob): Use gimple_call_builtin_p
1722 to check for BUILT_IN_SETJMP_RECEIVER.
1723 * tree-inline.c (copy_bb): Call notice_special_calls.
1725 2014-01-31 Vladimir Makarov <vmakarov@redhat.com>
1728 * lra-constraints.c (process_alt_operands): Update reload_sum only
1731 2014-01-31 Richard Henderson <rth@redhat.com>
1734 * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
1735 until after else_eh is processed.
1737 2014-01-31 Ilya Tocar <ilya.tocar@intel.com>
1739 * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
1740 (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
1741 (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
1742 in smmintrin.h, remove them.
1743 (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
1744 * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
1745 * config/i386/i386.md (ROUND_SAE): Fix value.
1746 * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
1747 (const48_operand): New.
1748 * config/i386/subst.md (round), (round_expand): Use
1749 const_4_or_8_to_11_operand.
1750 (round_saeonly), (round_saeonly_expand): Use const48_operand.
1752 2014-01-31 Ilya Tocar <ilya.tocar@intel.com>
1754 * config/i386/constraints.md (Yk): Swap meaning with k.
1755 * config/i386/i386.md (movhi_internal): Change Yk to k.
1756 (movqi_internal): Ditto.
1757 (*k<logic><mode>): Ditto.
1760 (kandn<mode>): Ditto.
1761 (*<code>hi_1): Ditto.
1762 (*<code>qi_1): Ditto.
1763 (kxnor<mode>): Ditto.
1764 (kortestzhi): Ditto.
1765 (kortestchi): Ditto.
1767 (*one_cmplhi2_1): Ditto.
1768 (*one_cmplqi2_1): Ditto.
1769 * config/i386/sse.md (): Change k to Yk.
1770 (avx512f_load<mode>_mask): Ditto.
1771 (avx512f_blendm<mode>): Ditto.
1772 (avx512f_store<mode>_mask): Ditto.
1773 (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
1774 (avx512f_storedqu<mode>_mask): Ditto.
1775 (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
1777 (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
1778 (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
1779 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
1780 (avx512f_maskcmp<mode>3): Ditto.
1781 (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
1782 (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
1783 (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
1784 (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
1785 (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
1786 (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
1787 (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
1788 (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
1789 (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
1790 (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
1791 (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
1792 (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
1793 (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
1794 (vec_extract_lo_<mode>_maskm): Ditto.
1795 (vec_extract_hi_<mode>_maskm): Ditto.
1796 (avx512f_vternlog<mode>_mask): Ditto.
1797 (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
1798 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
1799 (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
1800 (avx512f_<code>v8div16qi2_mask): Ditto.
1801 (avx512f_<code>v8div16qi2_mask_store): Ditto.
1802 (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
1803 (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
1804 (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
1805 (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
1806 (*avx512pf_gatherpf<mode>sf_mask): Ditto.
1807 (*avx512pf_gatherpf<mode>df_mask): Ditto.
1808 (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1809 (*avx512pf_scatterpf<mode>df_mask): Ditto.
1810 (avx512cd_maskb_vec_dupv8di): Ditto.
1811 (avx512cd_maskw_vec_dupv16si): Ditto.
1812 (avx512f_vpermi2var<mode>3_maskz): Ditto.
1813 (avx512f_vpermi2var<mode>3_mask): Ditto.
1814 (avx512f_vpermi2var<mode>3_mask): Ditto.
1815 (avx512f_vpermt2var<mode>3_maskz): Ditto.
1816 (*avx512f_gathersi<mode>): Ditto.
1817 (*avx512f_gathersi<mode>_2): Ditto.
1818 (*avx512f_gatherdi<mode>): Ditto.
1819 (*avx512f_gatherdi<mode>_2): Ditto.
1820 (*avx512f_scattersi<mode>): Ditto.
1821 (*avx512f_scatterdi<mode>): Ditto.
1822 (avx512f_compress<mode>_mask): Ditto.
1823 (avx512f_compressstore<mode>_mask): Ditto.
1824 (avx512f_expand<mode>_mask): Ditto.
1825 * config/i386/subst.md (mask): Change k to Yk.
1826 (mask_scalar_merge): Ditto.
1829 2014-01-31 Marc Glisse <marc.glisse@inria.fr>
1831 * doc/extend.texi (Vector Extensions): Document ?: in C++.
1833 2014-01-31 Richard Biener <rguenther@suse.de>
1836 * builtins.c (fold_builtin_memory_op): Make sure to not
1837 use a floating-point mode or a boolean or enumeral type for
1840 2014-01-30 DJ Delorie <dj@redhat.com>
1842 * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
1843 * config/msp430/msp430.md (msp430_refsym_need_exit): New.
1844 * config/msp430/msp430.c (msp430_expand_epilogue): Call it
1845 whenever main() has an epilogue.
1847 2014-01-30 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1849 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
1850 unused variable "field".
1851 * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
1852 (vsx_mergeh_<mode>): Likewise.
1853 * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
1854 (altivec_vmrghh): Likewise.
1855 (altivec_vmrghw): Likewise.
1856 (altivec_vmrglb): Likewise.
1857 (altivec_vmrglh): Likewise.
1858 (altivec_vmrglw): Likewise.
1859 (altivec_vspltb): Add missing uses.
1860 (altivec_vsplth): Likewise.
1861 (altivec_vspltw): Likewise.
1862 (altivec_vspltsf): Likewise.
1864 2014-01-30 Jakub Jelinek <jakub@redhat.com>
1867 * ifcvt.c (cond_exec_process_insns): Don't conditionalize
1868 frame related instructions.
1870 2014-01-30 Vladimir Makarov <vmakarov@redhat.com>
1872 PR rtl-optimization/59959
1873 * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
1874 any reload of register whose subreg is invalid.
1876 2014-01-30 Jakub Jelinek <jakub@redhat.com>
1878 * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
1879 * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
1880 Add missing return type - void.
1882 2014-01-30 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1884 * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
1885 gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
1886 remove element index adjustment for endian (now handled in vsx.md
1888 (altivec_expand_vec_perm_const): Use
1889 gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
1890 * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
1891 (vsx_xxspltw_<mode>): Adjust element index for little endian.
1892 * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
1893 define_expand and a new define_insn *altivec_vspltb_internal;
1894 adjust for -maltivec=be on a little endian target.
1895 (altivec_vspltb_direct): New.
1896 (altivec_vsplth): Divide into a define_expand and a new
1897 define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
1898 little endian target.
1899 (altivec_vsplth_direct): New.
1900 (altivec_vspltw): Divide into a define_expand and a new
1901 define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
1902 little endian target.
1903 (altivec_vspltw_direct): New.
1904 (altivec_vspltsf): Divide into a define_expand and a new
1905 define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
1906 a little endian target.
1908 2014-01-30 Richard Biener <rguenther@suse.de>
1910 PR tree-optimization/59993
1911 * tree-ssa-forwprop.c (associate_pointerplus): Check we
1912 can propagate form the earlier stmt and avoid the transform
1913 when the intermediate result is needed.
1915 2014-01-30 Alangi Derick <alangiderick@gmail.com>
1917 * README.Portability: Fix typo.
1919 2014-01-30 David Holsgrove <david.holsgrove@xilinx.com>
1921 * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
1922 comparison_operator with ordered_comparison_operator.
1924 2014-01-30 Nick Clifton <nickc@redhat.com>
1926 * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
1927 Rename to mn10300_store_multiple_regs.
1928 * config/mn10300/mn10300.c: Likewise.
1929 * config/mn10300/mn10300.md (store_movm): Fix typo: call
1930 store_multiple_regs.
1931 * config/mn10300/predicates.md (mn10300_store_multiple_operation):
1932 Call mn10300_store_multiple_regs.
1934 2014-01-30 Nick Clifton <nickc@redhat.com>
1935 DJ Delorie <dj@redhat.com>
1937 * config/rl78/rl78.c (register_sizes): Make the "upper half" of
1938 %fp 2 to keep registers after it properly word-aligned.
1939 (rl78_alloc_physical_registers_umul): Handle the case where both
1940 input operands are the same.
1942 2014-01-30 Richard Biener <rguenther@suse.de>
1944 PR tree-optimization/59903
1945 * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
1948 2014-01-30 Jason Merrill <jason@redhat.com>
1951 * tree.c (walk_type_fields): Handle VECTOR_TYPE.
1954 * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
1956 2014-01-30 Richard Biener <rguenther@suse.de>
1958 PR tree-optimization/59951
1959 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
1961 2014-01-30 Savin Zlobec <savin.zlobec@gmail.com>
1964 * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
1965 SFmode to DFmode case.
1967 2014-01-29 DJ Delorie <dj@redhat.com>
1969 * config/msp430/msp430.opt (-minrt): New.
1970 * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
1972 (ENDFILE_SPEC): Likewise.
1974 2014-01-29 Jan Hubicka <hubicka@ucw.cz>
1976 * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
1977 (estimate_function_body_sizes): Use it.
1979 2014-01-29 Paolo Carlini <paolo.carlini@oracle.com>
1982 * dwarf2out.c (is_cxx_auto): New.
1983 (is_base_type): Use it.
1984 (gen_type_die_with_usage): Likewise.
1986 2014-01-29 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1988 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Use
1989 CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
1990 * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
1991 -maltivec=be with LE targets.
1992 (vsx_mergeh_<mode>): Likewise.
1993 * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
1994 (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
1995 (altivec_vmrghb): Replace with define_expand and new
1996 *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
1997 (altivec_vmrghb_direct): New define_insn.
1998 (altivec_vmrghh): Replace with define_expand and new
1999 *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
2000 (altivec_vmrghh_direct): New define_insn.
2001 (altivec_vmrghw): Replace with define_expand and new
2002 *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
2003 (altivec_vmrghw_direct): New define_insn.
2004 (*altivec_vmrghsf): Adjust for endianness.
2005 (altivec_vmrglb): Replace with define_expand and new
2006 *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
2007 (altivec_vmrglb_direct): New define_insn.
2008 (altivec_vmrglh): Replace with define_expand and new
2009 *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
2010 (altivec_vmrglh_direct): New define_insn.
2011 (altivec_vmrglw): Replace with define_expand and new
2012 *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
2013 (altivec_vmrglw_direct): New define_insn.
2014 (*altivec_vmrglsf): Adjust for endianness.
2015 (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
2016 (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
2017 (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
2018 (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
2019 (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
2020 (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
2021 (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
2022 (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
2024 2014-01-29 Marcus Shawcroft <marcus.shawcroft@arm.com>
2026 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
2027 (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
2030 2014-01-29 Richard Biener <rguenther@suse.de>
2032 PR tree-optimization/58742
2033 * tree-ssa-forwprop.c (associate_pointerplus): Rename to
2034 associate_pointerplus_align.
2035 (associate_pointerplus_diff): New function.
2036 (associate_pointerplus): Likewise. Call associate_pointerplus_align
2037 and associate_pointerplus_diff.
2039 2014-01-29 Richard Biener <rguenther@suse.de>
2041 * lto-streamer.h (LTO_major_version): Bump to 3.
2042 (LTO_minor_version): Reset to 0.
2044 2014-01-29 Renlin Li <Renlin.Li@arm.com>
2046 * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
2047 * config/arm/arm.c (FL_FOR_ARCH7VE): New.
2048 (arm_file_start): Generate correct asm header for armv7ve.
2049 * config/arm/bpabi.h: Add multilib support for armv7ve.
2050 * config/arm/driver-arm.c: Change the architectures of cortex-a7
2051 and cortex-a15 to armv7ve.
2052 * config/arm/t-aprofile: Add multilib support for armv7ve.
2053 * doc/invoke.texi: Document -march=armv7ve.
2055 2014-01-29 Richard Biener <rguenther@suse.de>
2057 PR tree-optimization/58742
2058 * tree-ssa-forwprop.c (associate_plusminus): Return true
2059 if we changed sth, defer EH cleanup to ...
2060 (ssa_forward_propagate_and_combine): ... here. Call simplify_mult.
2061 (simplify_mult): New function.
2063 2014-01-29 Jakub Jelinek <jakub@redhat.com>
2066 PR tree-optimization/59920
2067 * tree.c (build_common_builtin_nodes): Remove
2068 __builtin_setjmp_dispatcher initialization.
2069 * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
2070 * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
2071 instead of gsi_after_labels + manually skipping debug stmts.
2072 Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
2073 ignore bbs with IFN_ABNORMAL_DISPATCHER.
2074 * tree-inline.c (copy_edges_for_bb): Remove
2075 can_make_abnormal_goto argument, instead add abnormal_goto_dest
2076 argument. Ignore computed_goto_p stmts. Don't call
2077 make_abnormal_goto_edges. If a call might need abnormal edges
2078 for non-local gotos, see if it already has an edge to
2079 IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
2080 with true argument, don't do anything then, otherwise add
2081 EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
2082 (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
2084 * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
2085 (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
2086 (lower_stmt): Don't set data->calls_builtin_setjmp.
2087 (lower_builtin_setjmp): Adjust comment.
2088 * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
2089 * tree-cfg.c (found_computed_goto): Remove.
2090 (factor_computed_gotos): Remove.
2091 (make_goto_expr_edges): Return bool, true for computed gotos.
2092 Don't call make_abnormal_goto_edges.
2093 (build_gimple_cfg): Don't set found_computed_goto, don't call
2094 factor_computed_gotos.
2095 (computed_goto_p): No longer static.
2096 (make_blocks): Don't set found_computed_goto.
2097 (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
2098 (make_edges): If make_goto_expr_edges returns true, push bb
2099 into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
2100 instead of calling make_abnormal_goto_edges push bb into ab_edge_call
2101 vector. Record mapping between bbs and OpenMP regions if there
2102 are any, adjust make_gimple_omp_edges caller. Call
2103 handle_abnormal_edges.
2104 (make_abnormal_goto_edges): Remove.
2105 * tree-cfg.h (make_abnormal_goto_edges): Remove.
2106 (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
2107 * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
2108 * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
2109 * internal-fn.def (ABNORMAL_DISPATCHER): New.
2110 * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
2111 filling *region also set *region_idx to (*region)->entry->index.
2114 * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
2115 For REGs set ORIGINAL_REGNO.
2117 2014-01-29 Bingfeng Mei <bmei@broadcom.com>
2119 * doc/md.texi: Mention that a target shouldn't implement
2120 vec_widen_(s|u)mul_even/odd pair if it is less efficient
2123 2014-01-29 Jakub Jelinek <jakub@redhat.com>
2125 PR tree-optimization/59594
2126 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
2127 a copy of the datarefs vector rather than the vector itself.
2129 2014-01-28 Jason Merrill <jason@redhat.com>
2132 * dwarf2out.c (auto_die): New static.
2133 (gen_type_die_with_usage): Handle C++1y 'auto'.
2134 (gen_subprogram_die): If in-class DIE had 'auto', emit type again
2137 2014-01-28 H.J. Lu <hongjiu.lu@intel.com>
2140 * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
2141 (SPEC_X32): Likewise.
2142 (SPEC_64): Likewise.
2143 * config/i386/i386.c (ix86_option_override_internal): Turn off
2144 OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
2146 (x86_file_start): Output .code16gcc for TARGET_16BIT.
2147 * config/i386/i386.h (TARGET_16BIT): New macro.
2148 (TARGET_16BIT_P): Likewise.
2149 * config/i386/i386.opt: Add m16.
2150 * doc/invoke.texi: Document -m16.
2152 2014-01-28 Jakub Jelinek <jakub@redhat.com>
2154 PR preprocessor/59935
2155 * input.c (location_get_source_line): Bail out on when line number
2156 is zero, and test the return value of lookup_or_add_file_to_cache_tab.
2158 2014-01-28 Richard Biener <rguenther@suse.de>
2160 PR tree-optimization/58742
2161 * tree-ssa-forwprop.c (associate_plusminus): Handle
2162 pointer subtraction of the form (T)(P + A) - (T)P.
2164 2014-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2166 * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
2169 2014-01-28 Richard Biener <rguenther@suse.de>
2172 2014-01-28 Richard Biener <rguenther@suse.de>
2174 PR rtl-optimization/45364
2175 PR rtl-optimization/59890
2176 * var-tracking.c (local_get_addr_clear_given_value): Handle
2177 already cleared slot.
2178 (val_reset): Handle not allocated local_get_addr_cache.
2179 (vt_find_locations): Use post-order on the inverted CFG.
2181 2014-01-28 Richard Biener <rguenther@suse.de>
2183 * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
2185 2014-01-28 Richard Biener <rguenther@suse.de>
2187 PR rtl-optimization/45364
2188 PR rtl-optimization/59890
2189 * var-tracking.c (local_get_addr_clear_given_value): Handle
2190 already cleared slot.
2191 (val_reset): Handle not allocated local_get_addr_cache.
2192 (vt_find_locations): Use post-order on the inverted CFG.
2194 2014-01-28 Alan Modra <amodra@gmail.com>
2196 * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
2197 * configure.ac <recursive call for build != host>: Define
2198 GENERATOR_FILE. Comment. Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
2199 and LD_FOR_BUILD too.
2200 * configure: Regenerate.
2202 2014-01-27 Allan Sandfeld Jensen <sandfeld@kde.org>
2204 * config/i386/i386.c (get_builtin_code_for_version): Separate
2205 Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
2206 Broadwell from Haswell.
2208 2014-01-27 Steve Ellcey <sellcey@mips.com>
2210 * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2211 Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
2212 * config/mips/mips.c (mips_option_override): Change setting
2214 * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
2215 * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
2216 Change from Mask to Var.
2218 2014-01-27 Jeff Law <law@redhat.com>
2220 * ipa-inline.c (inline_small_functions): Fix typo.
2222 2014-01-27 Ilya Tocar <ilya.tocar@intel.com>
2224 * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
2225 (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
2226 (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
2227 (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
2228 (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
2229 (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
2230 (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
2231 (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
2232 (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
2233 (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
2234 (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
2235 (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
2236 (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
2237 (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
2238 (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
2239 (_mm512_storeu_epi64): Ditto.
2240 (_mm512_cmpge_epi32_mask): Ditto.
2241 (_mm512_cmpge_epu32_mask): Ditto.
2242 (_mm512_cmpge_epi64_mask): Ditto.
2243 (_mm512_cmpge_epu64_mask): Ditto.
2244 (_mm512_cmple_epi32_mask): Ditto.
2245 (_mm512_cmple_epu32_mask): Ditto.
2246 (_mm512_cmple_epi64_mask): Ditto.
2247 (_mm512_cmple_epu64_mask): Ditto.
2248 (_mm512_cmplt_epi32_mask): Ditto.
2249 (_mm512_cmplt_epu32_mask): Ditto.
2250 (_mm512_cmplt_epi64_mask): Ditto.
2251 (_mm512_cmplt_epu64_mask): Ditto.
2252 (_mm512_cmpneq_epi32_mask): Ditto.
2253 (_mm512_cmpneq_epu32_mask): Ditto.
2254 (_mm512_cmpneq_epi64_mask): Ditto.
2255 (_mm512_cmpneq_epu64_mask): Ditto.
2256 (_mm512_expand_pd): Ditto.
2257 (_mm512_expand_ps): Ditto.
2258 * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
2259 VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
2260 VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
2261 * config/i386/i386.c (ix86_builtins): Add
2262 IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
2263 IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
2264 IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
2265 IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
2266 IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
2267 IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
2268 IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
2269 IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
2270 IX86_BUILTIN_PMOVUSQW512_MEM.
2271 (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
2272 __builtin_ia32_pmovsqd512mem_mask,
2273 __builtin_ia32_pmovqd512mem_mask,
2274 __builtin_ia32_pmovusqw512mem_mask,
2275 __builtin_ia32_pmovsqw512mem_mask,
2276 __builtin_ia32_pmovqw512mem_mask,
2277 __builtin_ia32_pmovusdw512mem_mask,
2278 __builtin_ia32_pmovsdw512mem_mask,
2279 __builtin_ia32_pmovdw512mem_mask,
2280 __builtin_ia32_pmovqb512mem_mask,
2281 __builtin_ia32_pmovusqb512mem_mask,
2282 __builtin_ia32_pmovsqb512mem_mask,
2283 __builtin_ia32_pmovusdb512mem_mask,
2284 __builtin_ia32_pmovsdb512mem_mask,
2285 __builtin_ia32_pmovdb512mem_mask.
2286 (bdesc_args): Add __builtin_ia32_expanddf512,
2287 __builtin_ia32_expandsf512.
2288 (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
2289 VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
2290 VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
2291 * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
2292 (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
2293 (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
2294 (avx512f_<code>v8div16qi2_mask_store): This.
2295 (avx512f_expand<mode>): New.
2297 2014-01-27 Kirill Yukhin <kirill.yukhin@intel.com>
2299 * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
2301 (_mm512_mask_prefetch_i64gather_pd): Ditto.
2302 (_mm512_prefetch_i32scatter_pd): Ditto.
2303 (_mm512_mask_prefetch_i32scatter_pd): Ditto.
2304 (_mm512_prefetch_i64scatter_pd): Ditto.
2305 (_mm512_mask_prefetch_i64scatter_pd): Ditto.
2306 (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
2307 (_mm512_mask_prefetch_i64gather_ps): Ditto.
2308 (_mm512_prefetch_i32scatter_ps): Ditto.
2309 (_mm512_mask_prefetch_i32scatter_ps): Ditto.
2310 (_mm512_prefetch_i64scatter_ps): Ditto.
2311 (_mm512_mask_prefetch_i64scatter_ps): Ditto.
2312 * config/i386/i386-builtin-types.def: Define
2313 VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
2314 and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
2315 * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
2316 IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
2317 IX86_BUILTIN_SCATTERPFQPD.
2318 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
2319 __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
2320 __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
2321 __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
2322 __builtin_ia32_scatterpfqps.
2323 (ix86_expand_builtin): Expand new built-ins.
2324 * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
2325 fix memory access data type.
2326 (*avx512pf_gatherpf<mode>_mask): Ditto.
2327 (*avx512pf_gatherpf<mode>): Ditto.
2328 (avx512pf_scatterpf<mode>): Ditto.
2329 (*avx512pf_scatterpf<mode>_mask): Ditto.
2330 (*avx512pf_scatterpf<mode>): Ditto.
2331 (GATHER_SCATTER_SF_MEM_MODE): New.
2332 (avx512pf_gatherpf<mode>df): Ditto.
2333 (*avx512pf_gatherpf<mode>df_mask): Ditto.
2334 (*avx512pf_scatterpf<mode>df): Ditto.
2336 2014-01-27 Jakub Jelinek <jakub@redhat.com>
2339 * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
2340 know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
2343 2014-01-27 James Greenhalgh <james.greenhalgh@arm.com>
2345 * common/config/arm/arm-common.c
2346 (arm_rewrite_mcpu): Handle multiple names.
2348 (BIG_LITTLE_SPEC): Do not discard mcpu switches.
2350 2014-01-27 James Greenhalgh <james.greenhalgh@arm.com>
2352 * gimple-builder.h (create_gimple_tmp): Delete.
2354 2014-01-27 Christian Bruel <christian.bruel@st.com>
2356 * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
2359 2014-01-26 John David Anglin <danglin@gcc.gnu.org>
2361 * config/pa/pa.md (call): Generate indirect long calls to non-local
2362 functions when outputing 32-bit code.
2363 (call_value): Likewise except for special call to buggy powf function.
2365 * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
2366 portable runtime and PIC indirect calls.
2367 (pa_output_indirect_call): Remove unnecessary nop from portable runtime
2368 and PIC call sequences. Use ldo instead of blr to set return register
2369 in PIC call sequence.
2371 2014-01-25 Walter Lee <walt@tilera.com>
2373 * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
2374 avoid clobbering a live register.
2376 2014-01-25 Walter Lee <walt@tilera.com>
2378 * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
2379 Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
2380 * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
2381 Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
2383 2014-01-25 Walter Lee <walt@tilera.com>
2385 * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
2386 arguments on even registers.
2387 (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
2389 * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
2390 (BIGGEST_ALIGNMENT): Ditto.
2391 (BIGGEST_FIELD_ALIGNMENT): Ditto.
2393 2014-01-25 Walter Lee <walt@tilera.com>
2395 * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
2396 insns before bundling.
2397 * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
2399 2014-01-25 Walter Lee <walt@tilera.com>
2401 * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
2402 PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
2403 * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
2405 2014-01-25 Richard Sandiford <rdsandiford@googlemail.com>
2407 * config/mips/constraints.md (kl): Delete.
2408 * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
2409 define expands, using...
2410 (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
2411 instructions for MIPS16.
2412 (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
2413 non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
2415 2014-01-25 Walter Lee <walt@tilera.com>
2417 * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
2421 2014-01-25 Walter Lee <walt@tilera.com>
2423 * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
2424 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2426 2014-01-25 Richard Sandiford <rdsandiford@googlemail.com>
2428 * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
2431 2014-01-25 Jakub Jelinek <jakub@redhat.com>
2433 * print-rtl.c (in_call_function_usage): New var.
2434 (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
2435 EXPR_LIST mode as mode and not as reg note name.
2438 * cfgloopmanip.c (copy_loop_info): If
2439 loop->warned_aggressive_loop_optimizations, make sure
2440 the flag is set in target loop too.
2442 2014-01-24 Balaji V. Iyer <balaji.v.iyer@intel.com>
2444 * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
2446 * builtins.def: Likewise.
2447 * cilk.h (fn_contains_cilk_spawn_p): Likewise.
2448 * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
2449 * ira.c (ira_setup_eliminable_regset): Likewise.
2450 * omp-low.c (gate_expand_omp): Likewise.
2451 (execute_lower_omp): Likewise.
2452 (diagnose_sb_0): Likewise.
2453 (gate_diagnose_omp_blocks): Likewise.
2454 (simd_clone_clauses_extract): Likewise.
2457 2014-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2459 * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
2460 correction for little endian...
2461 * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
2464 2014-01-24 Jeff Law <law@redhat.com>
2466 PR tree-optimization/59919
2467 * tree-vrp.c (find_assert_locations_1): Do not register asserts
2468 for non-returning calls.
2470 2014-01-24 James Greenhalgh <james.greenhalgh@arm.com>
2472 * common/config/aarch64/aarch64-common.c
2473 (aarch64_rewrite_mcpu): Handle multiple names.
2474 * config/aarch64/aarch64.h
2475 (BIG_LITTLE_SPEC): Do not discard mcpu switches.
2477 2014-01-24 Dodji Seketeli <dodji@redhat.com>
2479 * input.c (add_file_to_cache_tab): Handle the case where fopen
2482 2014-01-23 H.J. Lu <hongjiu.lu@intel.com>
2485 * config/i386/i386.md (pushsf splitter): Get stack adjustment
2486 from push operand if code of push isn't PRE_DEC.
2488 2014-01-23 Michael Meissner <meissner@linux.vnet.ibm.com>
2491 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2492 -mquad-memory-atomic. Update -mquad-memory documentation to say
2493 it is only used for non-atomic loads/stores.
2495 * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
2496 -mquad-memory or -mquad-memory-atomic switches.
2498 * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
2499 -mquad-memory-atomic to ISA 2.07 support.
2501 * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
2502 to separate support of normal quad word memory operations (ldq, stq)
2503 from the atomic quad word memory operations.
2505 * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
2506 support to separate non-atomic quad word operations from atomic
2507 quad word operations. Disable non-atomic quad word operations in
2508 little endian mode so that we don't have to swap words after the
2509 load and before the store.
2510 (quad_load_store_p): Add comment about atomic quad word support.
2511 (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
2512 options printed with -mdebug=reg.
2514 * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
2515 -mquad-memory-atomic as the test for whether we have quad word
2516 atomic instructions.
2517 (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
2518 or -mp8-vector are used, allow byte/half-word atomic operations.
2520 * config/rs6000/sync.md (load_lockedti): Insure that the address
2521 is a proper indexed or indirect address for the lqarx instruction.
2522 On little endian systems, swap the hi/lo registers after the lqarx
2524 (load_lockedpti): Use indexed_or_indirect_operand predicate to
2525 insure the address is valid for the lqarx instruction.
2526 (store_conditionalti): Insure that the address is a proper indexed
2527 or indirect address for the stqcrx. instruction. On little endian
2528 systems, swap the hi/lo registers before doing the stqcrx.
2530 (store_conditionalpti): Use indexed_or_indirect_operand predicate to
2531 insure the address is valid for the stqcrx. instruction.
2533 * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
2534 Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
2535 type of quad memory support is available.
2537 2014-01-23 Vladimir Makarov <vmakarov@redhat.com>
2540 * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
2541 there is a danger of looping.
2543 2014-01-23 Pat Haugen <pthaugen@us.ibm.com>
2545 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2546 force flag_ira_loop_pressure if set via command line.
2548 2014-01-23 Alex Velenko <Alex.Velenko@arm.com>
2550 * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
2551 (ashr_simd): New builtin handling DI mode.
2552 * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
2553 (aarch64_sshr_simddi): New match pattern.
2554 * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
2555 (vshrd_n_s64): Likewise.
2556 * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
2558 2014-01-23 Nick Clifton <nickc@redhat.com>
2560 * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
2561 (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
2562 favour of mcu specific scripts.
2563 * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
2566 2014-01-23 James Greenhalgh <james.greenhalgh@arm.com>
2567 Alex Velenko <Alex.Velenko@arm.com>
2569 * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
2570 (vaddv_s16): Likewise.
2571 (vaddv_s32): Likewise.
2572 (vaddv_u8): Likewise.
2573 (vaddv_u16): Likewise.
2574 (vaddv_u32): Likewise.
2575 (vaddvq_s8): Likewise.
2576 (vaddvq_s16): Likewise.
2577 (vaddvq_s32): Likewise.
2578 (vaddvq_s64): Likewise.
2579 (vaddvq_u8): Likewise.
2580 (vaddvq_u16): Likewise.
2581 (vaddvq_u32): Likewise.
2582 (vaddvq_u64): Likewise.
2583 (vaddv_f32): Likewise.
2584 (vaddvq_f32): Likewise.
2585 (vaddvq_f64): Likewise.
2586 (vmaxv_f32): Likewise.
2587 (vmaxv_s8): Likewise.
2588 (vmaxv_s16): Likewise.
2589 (vmaxv_s32): Likewise.
2590 (vmaxv_u8): Likewise.
2591 (vmaxv_u16): Likewise.
2592 (vmaxv_u32): Likewise.
2593 (vmaxvq_f32): Likewise.
2594 (vmaxvq_f64): Likewise.
2595 (vmaxvq_s8): Likewise.
2596 (vmaxvq_s16): Likewise.
2597 (vmaxvq_s32): Likewise.
2598 (vmaxvq_u8): Likewise.
2599 (vmaxvq_u16): Likewise.
2600 (vmaxvq_u32): Likewise.
2601 (vmaxnmv_f32): Likewise.
2602 (vmaxnmvq_f32): Likewise.
2603 (vmaxnmvq_f64): Likewise.
2604 (vminv_f32): Likewise.
2605 (vminv_s8): Likewise.
2606 (vminv_s16): Likewise.
2607 (vminv_s32): Likewise.
2608 (vminv_u8): Likewise.
2609 (vminv_u16): Likewise.
2610 (vminv_u32): Likewise.
2611 (vminvq_f32): Likewise.
2612 (vminvq_f64): Likewise.
2613 (vminvq_s8): Likewise.
2614 (vminvq_s16): Likewise.
2615 (vminvq_s32): Likewise.
2616 (vminvq_u8): Likewise.
2617 (vminvq_u16): Likewise.
2618 (vminvq_u32): Likewise.
2619 (vminnmv_f32): Likewise.
2620 (vminnmvq_f32): Likewise.
2621 (vminnmvq_f64): Likewise.
2623 2014-01-23 James Greenhalgh <james.greenhalgh@arm.com>
2625 * config/aarch64/aarch64-simd.md
2626 (aarch64_dup_lane<mode>): Correct lane number on big-endian.
2627 (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
2628 (*aarch64_mul3_elt<mode>): Likewise.
2629 (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
2630 (*aarch64_mul3_elt_to_64v2df): Likewise.
2631 (*aarch64_mla_elt<mode>): Likewise.
2632 (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
2633 (*aarch64_mls_elt<mode>): Likewise.
2634 (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
2635 (*aarch64_fma4_elt<mode>): Likewise.
2636 (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
2637 (*aarch64_fma4_elt_to_64v2df): Likewise.
2638 (*aarch64_fnma4_elt<mode>): Likewise.
2639 (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
2640 (*aarch64_fnma4_elt_to_64v2df): Likewise.
2641 (aarch64_sq<r>dmulh_lane<mode>): Likewise.
2642 (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
2643 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
2644 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
2645 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
2646 (aarch64_sqdmull_lane<mode>_internal): Likewise.
2647 (aarch64_sqdmull2_lane<mode>_internal): Likewise.
2649 2013-01-23 Alex Velenko <Alex.Velenko@arm.com>
2651 * config/aarch64/aarch64-simd.md
2652 (aarch64_be_checked_get_lane<mode>): New define_expand.
2653 * config/aarch64/aarch64-simd-builtins.def
2654 (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
2655 New builtin definition.
2656 * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
2657 Use new safe be builtin.
2659 2014-01-23 Alex Velenko <Alex.Velenko@arm.com>
2661 * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
2663 (aarch64_be_st1<mode>): Likewise.
2664 (aarch_ld1<VALL:mode>): Define_expand modified.
2665 (aarch_st1<VALL:mode>): Likewise.
2666 * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
2667 (UNSPEC_ST1): Likewise.
2669 2014-01-23 David Holsgrove <david.holsgrove@xilinx.com>
2671 * config/microblaze/microblaze.md: Add trap insn and attribute
2673 2014-01-23 Dodji Seketeli <dodji@redhat.com>
2675 PR preprocessor/58580
2676 * input.h (location_get_source_line): Take an additional line_size
2678 (void diagnostics_file_cache_fini): Declare new function.
2679 * input.c (struct fcache): New type.
2680 (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
2681 New static constants.
2682 (diagnostic_file_cache_init, total_lines_num)
2683 (lookup_file_in_cache_tab, evicted_cache_tab_entry)
2684 (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
2685 (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
2686 (get_next_line, read_next_line, goto_next_line, read_line_num):
2687 New static function definitions.
2688 (diagnostic_file_cache_fini): New function.
2689 (location_get_source_line): Take an additional output line_len
2690 parameter. Re-write using lookup_or_add_file_to_cache_tab and
2692 * diagnostic.c (diagnostic_finish): Call
2693 diagnostic_file_cache_fini.
2694 (adjust_line): Take an additional input parameter for the length
2695 of the line, rather than calculating it with strlen.
2696 (diagnostic_show_locus): Adjust the use of
2697 location_get_source_line and adjust_line with respect to their new
2698 signature. While displaying a line now, do not stop at the first
2699 null byte. Rather, display the zero byte as a space and keep
2700 going until we reach the size of the line.
2701 * Makefile.in: Add vec.o to OBJS-libcommon
2703 2014-01-23 Kirill Yukhin <kirill.yukhin@intel.com>
2704 Ilya Tocar <ilya.tocar@intel.com>
2706 * config/i386/avx512fintrin.h (_mm512_kmov): New.
2707 * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
2708 (__builtin_ia32_kmov16): Ditto.
2709 * config/i386/i386.md (UNSPEC_KMOV): New.
2712 2014-01-23 Kirill Yukhin <kirill.yukhin@intel.com>
2714 * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
2715 (_mm512_storeu_si512): Ditto.
2717 2014-01-23 Richard Sandiford <rdsandiford@googlemail.com>
2720 * rtl.h (get_referenced_operands): Declare.
2721 * recog.c (get_referenced_operands): New function.
2722 * config/mips/mips.c (mips_reorg_process_insns): Check which asm
2723 operands have been referenced when recording LO_SUM references.
2725 2014-01-22 David Holsgrove <david.holsgrove@xilinx.com>
2727 * config/microblaze/microblaze.md: Correct bswaphi2 insn.
2729 2014-01-22 Jan Hubicka <hubicka@ucw.cz>
2731 * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
2732 Enable for generic and recent AMD targets.
2734 2014-01-22 Jan Hubicka <hubicka@ucw.cz>
2736 * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
2737 ARG_SIZE note when adjustment was eliminated.
2739 2014-01-22 Jeff Law <law@redhat.com>
2741 PR tree-optimization/59597
2742 * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
2743 in file. Accept new argument REGISTERING and use it to modify
2744 dump output appropriately.
2745 (register_jump_thread): Corresponding changes.
2746 (mark_threaded_blocks): Reinstate code to cancel unprofitable
2747 thread paths involving joiner blocks. Add code to dump cancelled
2748 jump threading paths.
2750 2014-01-22 Vladimir Makarov <vmakarov@redhat.com>
2752 PR rtl-optimization/59477
2753 * lra-constraints.c (inherit_in_ebb): Process call for living hard
2754 regs. Update reloads_num and potential_reload_hard_regs for all insns.
2756 2014-01-22 Tom Tromey <tromey@redhat.com>
2758 * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
2760 * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
2762 2014-01-21 Vladimir Makarov <vmakarov@redhat.com>
2764 PR rtl-optimization/59896
2765 * lra-constraints.c (process_alt_operands): Check unused note for
2766 matched operands of insn with no output reloads.
2768 2014-01-21 Richard Sandiford <rdsandiford@googlemail.com>
2770 * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
2771 (mips_move_from_gpr_cost): Likewise.
2773 2014-01-21 Vladimir Makarov <vmakarov@redhat.com>
2775 PR rtl-optimization/59858
2776 * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
2777 ira_class_hard_regs_num.
2778 (process_alt_operands): Increase reject for dying matched operand.
2780 2014-01-21 Jakub Jelinek <jakub@redhat.com>
2783 * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
2784 smaller than size, perform several stores or loads and stores
2785 at dst + count - size to store or copy all of size bytes, rather
2786 than just last modesize bytes.
2788 2014-01-20 DJ Delorie <dj@redhat.com>
2790 * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
2791 that CLOBBERs are REGs before propogating their values.
2793 2014-01-20 H.J. Lu <hongjiu.lu@intel.com>
2796 * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
2797 (cgraph_inline_failed_type): New function.
2798 * cgraph.h (DEFCIFCODE): Add type.
2799 (cgraph_inline_failed_type_t): New enum.
2800 (cgraph_inline_failed_type): New prototype.
2801 * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
2802 FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
2803 FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
2804 LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
2805 MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
2806 RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
2807 OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
2808 INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
2809 Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
2810 FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
2811 EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
2812 OPTIMIZATION_MISMATCH.
2813 * tree-inline.c (expand_call_inline): Emit errors during
2814 early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
2816 2014-01-20 Uros Bizjak <ubizjak@gmail.com>
2819 * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
2820 mode attribute in insn output.
2822 2014-01-20 Eric Botcazou <ebotcazou@adacore.com>
2824 * output.h (output_constant): Delete.
2825 * varasm.c (output_constant): Make private.
2827 2014-01-20 Alex Velenko <Alex.Velenko@arm.com>
2829 * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
2831 2014-01-20 Jakub Jelinek <jakub@redhat.com>
2834 * tree.h (fold_builtin_strcat): New prototype.
2835 * builtins.c (fold_builtin_strcat): No longer static. Add len
2836 argument, if non-NULL, don't call c_strlen. Optimize
2837 directly into __builtin_memcpy instead of __builtin_strcpy.
2838 (fold_builtin_2): Adjust fold_builtin_strcat caller.
2839 * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
2841 2014-01-20 Uros Bizjak <ubizjak@gmail.com>
2843 * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
2844 for SImode_address_operand operands, having only a REG argument.
2846 2014-01-20 Marcus Shawcroft <marcus.shawcroft@arm.com>
2848 * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
2849 loader name using mbig-endian.
2850 (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
2852 2014-01-20 James Greenhalgh <james.greenhalgh@arm.com>
2854 * doc/invoke.texi (-march): Clarify documentation for AArch64.
2858 2014-01-20 Tejas Belagod <tejas.belagod@arm.com>
2860 * config/aarch64/aarch64-protos.h
2861 (aarch64_cannot_change_mode_class_ptr): Declare.
2862 * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
2863 aarch64_cannot_change_mode_class_ptr): New.
2864 * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
2865 backend hook aarch64_cannot_change_mode_class.
2867 2014-01-20 James Greenhalgh <james.greenhalgh@arm.com>
2869 * common/config/aarch64/aarch64-common.c
2870 (aarch64_handle_option): Don't handle any option order logic here.
2871 * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
2872 selected_cpu, warn on architecture version mismatch.
2873 (aarch64_override_options): Fix parsing order for option strings.
2875 2014-01-20 Jan-Benedict Glaw <jbglaw@lug-owl.de>
2876 Iain Sandoe <iain@codesourcery.com>
2879 * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
2880 warning. Amend comment to reflect current functionality.
2882 2014-01-20 Richard Biener <rguenther@suse.de>
2885 * builtins.c (fold_builtin_strcat): Remove case better handled
2886 by tree-ssa-strlen.c.
2888 2014-01-20 Alan Lawrence <alan.lawrence@arm.com>
2890 * config/aarch64/aarch64.opt
2891 (mcpu, march, mtune): Make case-insensitive.
2893 2014-01-20 Jakub Jelinek <jakub@redhat.com>
2896 * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
2897 if operands[1] is a REG or ZERO_EXTEND of a REG.
2899 2014-01-19 Jan Hubicka <hubicka@ucw.cz>
2901 * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
2903 2014-01-19 John David Anglin <danglin@gcc.gnu.org>
2905 * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
2906 long non-pic millicode calls.
2908 2014-01-19 Jan-Benedict Glaw <jbglaw@lug-owl.de>
2910 * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
2912 2014-01-19 Kito Cheng <kito@0xlab.org>
2914 * builtins.c (expand_movstr): Check movstr expand done or fail.
2916 2014-01-18 Uros Bizjak <ubizjak@gmail.com>
2917 H.J. Lu <hongjiu.lu@intel.com>
2920 * config/i386/i386.md (*lea<mode>): Zero-extend return register
2921 to DImode for zero-extended addresses.
2923 2014-01-19 Jakub Jelinek <jakub@redhat.com>
2925 PR rtl-optimization/57763
2926 * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
2927 on the new indirect jump_insn and increment LABEL_NUSES (label).
2929 2014-01-18 H.J. Lu <hongjiu.lu@intel.com>
2933 * config.gcc (x86_archs): New variable.
2934 (x86_64_archs): Likewise.
2935 (x86_cpus): Likewise.
2936 Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
2937 --with-arch/--with-cpu= options.
2938 Support --with-arch=/--with-cpu={nehalem,westmere,
2939 sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
2941 2014-01-18 Uros Bizjak <ubizjak@gmail.com>
2943 * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
2944 and PROCESSOR_ATHLON to simplify code. Move "memory" calculation.
2946 2014-01-18 Uros Bizjak <ubizjak@gmail.com>
2948 * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
2950 2014-01-18 Jakub Jelinek <jakub@redhat.com>
2953 * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
2954 clear cpp_get_options (parse_in)->warn_unused_macros for
2955 ix86_target_macros_internal with cpp_define.
2957 2014-01-18 Richard Sandiford <rdsandiford@googlemail.com>
2959 * jump.c (delete_related_insns): Keep (use (insn))s.
2960 * reorg.c (redundant_insn): Check for barriers too.
2962 2014-01-17 H.J. Lu <hongjiu.lu@intel.com>
2964 * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
2966 2014-01-17 John David Anglin <danglin@gcc.gnu.org>
2968 * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
2969 call to $$dyncall when TARGET_LONG_CALLS is true.
2971 2014-01-17 Jeff Law <law@redhat.com>
2973 * ree.c (combine_set_extension): Temporarily disable test for
2974 changing number of hard registers.
2976 2014-01-17 Jan Hubicka <hubicka@ucw.cz>
2979 * ipa-inline-analysis.c (inline_free_summary):
2980 Do not free summary of aliases.
2982 2014-01-17 Jakub Jelinek <jakub@redhat.com>
2985 * gimplify.c (gimplify_expr): Use create_tmp_var
2986 instead of create_tmp_var_raw. If cond doesn't have
2987 integral type, don't add the IFN_ANNOTATE builtin at all.
2989 2014-01-17 Martin Jambor <mjambor@suse.cz>
2992 * ipa-cp.c (prev_edge_clone): New variable.
2993 (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
2994 Also resize prev_edge_clone vector.
2995 (ipcp_edge_duplication_hook): Also update prev_edge_clone.
2996 (ipcp_edge_removal_hook): New function.
2997 (ipcp_driver): Register ipcp_edge_removal_hook.
2999 2014-01-17 Andrew Pinski <apinski@cavium.com>
3000 Steve Ellcey <sellcey@mips.com>
3003 * config/mips/mips.c (mips_print_operand): Check operand mode instead
3006 2014-01-17 Jeff Law <law@redhat.com>
3009 * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
3010 so that pass_ccp runs first.
3012 2014-01-17 H.J. Lu <hongjiu.lu@intel.com>
3014 * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
3015 (ix86_adjust_cost): Use !TARGET_XXX.
3016 (do_reorder_for_imul): Likewise.
3017 (swap_top_of_ready_list): Likewise.
3018 (ix86_sched_reorder): Likewise.
3020 2014-01-17 H.J. Lu <hongjiu.lu@intel.com>
3022 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3023 PROCESSOR_INTEL. Treat like PROCESSOR_GENERIC.
3024 * config/i386/i386.c (intel_memcpy): New. Duplicate slm_memcpy.
3025 (intel_memset): New. Duplicate slm_memset.
3026 (intel_cost): New. Duplicate slm_cost.
3027 (m_INTEL): New macro.
3028 (processor_target_table): Add "intel".
3029 (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
3030 with PROCESSOR_INTEL for "intel".
3031 (ix86_lea_outperforms): Support PROCESSOR_INTEL. Duplicate
3032 PROCESSOR_SILVERMONT.
3033 (ix86_issue_rate): Likewise.
3034 (ix86_adjust_cost): Likewise.
3035 (ia32_multipass_dfa_lookahead): Likewise.
3036 (swap_top_of_ready_list): Likewise.
3037 (ix86_sched_reorder): Likewise.
3038 (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
3039 instead of TARGET_OPT_AGU.
3040 * config/i386/i386.h (TARGET_INTEL): New.
3041 (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
3042 (processor_type): Add PROCESSOR_INTEL.
3043 * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
3044 Add X86_TUNE_AVOID_LEA_FOR_ADDR.
3046 2014-01-17 Marek Polacek <polacek@redhat.com>
3049 * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
3052 2014-01-17 Richard Biener <rguenther@suse.de>
3054 PR tree-optimization/46590
3055 * opts.c (default_options_table): Add entries for
3056 OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
3057 all enabled at -O1 but not for -Og.
3058 * common.opt (fbranch-count-reg): Remove Init(1).
3059 (fmove-loop-invariants): Likewise.
3060 (ftree-pta): Likewise.
3062 2014-01-17 Jakub Jelinek <jakub@redhat.com>
3064 * config/i386/i386.c (ix86_data_alignment): For compatibility with
3065 (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
3066 decls to at least the GCC 4.8 used alignments.
3069 * tree-nested.c (convert_nonlocal_reference_stmt,
3070 convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
3071 of GIMPLE_BIND stmts, adjust associated decls.
3073 2014-01-17 Richard Biener <rguenther@suse.de>
3075 PR tree-optimization/46590
3076 * vec.h (vec<>::bseach): New member function implementing
3077 binary search according to C89 bsearch.
3078 (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
3079 * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
3080 bitmap pointer again. Make accesses_in_loop a flat array.
3081 (mem_ref_obstack): New global.
3082 (outermost_indep_loop): Adjust for mem_ref->stored changes.
3083 (mark_ref_stored): Likewise.
3084 (ref_indep_loop_p_2): Likewise.
3085 (set_ref_stored_in_loop): New helper function.
3086 (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
3087 (memref_free): Adjust.
3088 (record_mem_ref_loc): Simplify.
3089 (gather_mem_refs_stmt): Adjust.
3090 (sort_locs_in_loop_postorder_cmp): New function.
3091 (analyze_memory_references): Sort accesses_in_loop after
3092 loop postorder number.
3093 (find_ref_loc_in_loop_cmp): New function.
3094 (for_all_locs_in_loop): Find relevant cluster of locs in
3095 accesses_in_loop and iterate without recursion.
3096 (execute_sm): Avoid uninit warning.
3097 (struct ref_always_accessed): Simplify.
3098 (ref_always_accessed::operator ()): Likewise.
3099 (ref_always_accessed_p): Likewise.
3100 (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
3101 loop postorder numbers here.
3102 (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
3105 2014-01-17 Jan Hubicka <hubicka@ucw.cz>
3108 * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
3109 on decls for which assemble_alias has been called.
3111 2014-01-17 Nick Clifton <nickc@redhat.com>
3113 * config/msp430/msp430.opt: (mcpu): New option.
3114 * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
3115 (msp430_option_override): Parse target_cpu. If the MCU name
3116 matches a generic string, clear target_mcu.
3117 (msp430_attr): Allow numeric interrupt values up to 63.
3118 (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
3119 * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
3121 * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
3123 * config/msp430/msp430.md (popm): Use %J rather than %I.
3124 (addsi3): Use msp430_nonimmediate_operand for operand 2.
3125 (addhi_cy_i): Use immediate_operand for operand 2.
3126 * doc/invoke.texi: Document -mcpu option.
3128 2014-01-17 Richard Biener <rguenther@suse.de>
3130 PR rtl-optimization/38518
3131 * df.h (df_analyze_loop): Declare.
3132 * df-core.c: Include cfgloop.h.
3133 (df_analyze_1): Split out main part of df_analyze.
3134 (df_analyze): Adjust.
3135 (loop_inverted_post_order_compute): New function.
3136 (loop_post_order_compute): Likewise.
3137 (df_analyze_loop): New function avoiding whole-function
3139 * loop-invariant.c (find_defs): Use df_analyze_loop.
3140 (find_invariants): Adjust.
3141 * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
3143 2014-01-17 Zhenqiang Chen <zhenqiang.chen@arm.com>
3145 * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
3146 (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
3148 2014-01-16 Ilya Enkovich <ilya.enkovich@intel.com>
3150 * ipa-ref.c (ipa_remove_stmt_references): Fix references
3151 traversal when removing references.
3153 2014-01-16 Jan Hubicka <hubicka@ucw.cz>
3156 * tree.c (get_binfo_at_offset): Look harder for virtual bases.
3158 2014-01-16 Bernd Schmidt <bernds@codesourcery.com>
3161 * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
3162 pushing a reload for an autoinc when we had previously reloaded an
3163 inner part of the address.
3165 2014-01-16 Jakub Jelinek <jakub@redhat.com>
3167 * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
3169 (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
3170 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
3171 when not giving up or versioning for alias only because of
3173 (vect_analyze_data_ref_dependences): Set to true.
3174 * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
3176 (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
3177 LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
3181 * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
3184 * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
3185 operand 0 predicate for gathers, use a new pseudo as subtarget.
3187 2014-01-16 Vladimir Makarov <vmakarov@redhat.com>
3190 * lra-constraints.c (process_alt_operands): Add printing debug info.
3191 Check absence of input/output reloads for matched operands too.
3193 2014-01-16 Vladimir Makarov <vmakarov@redhat.com>
3195 PR rtl-optimization/59835
3196 * ira.c (ira_init_register_move_cost): Increase cost for
3199 2014-01-16 Alan Lawrence <alan.lawrence@arm.com>
3201 * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
3203 2014-01-16 Richard Earnshaw <rearnsha@arm.com>
3206 * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
3207 non-register objects. Use gen_(high/low)part more consistently.
3210 2014-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3213 * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
3214 endian support, remove tests for WORDS_BIG_ENDIAN.
3215 (p8_mfvsrd_3_<mode>): Likewise.
3216 (reload_gpr_from_vsx<mode>): Likewise.
3217 (reload_gpr_from_vsxsf): Likewise.
3218 (p8_mfvsrd_4_disf): Likewise.
3220 2014-01-16 Richard Biener <rguenther@suse.de>
3222 PR rtl-optimization/46590
3223 * lcm.c (compute_antinout_edge): Use postorder iteration.
3224 (compute_laterin): Use inverted postorder iteration.
3226 2014-01-16 Nick Clifton <nickc@redhat.com>
3229 * varasm.c (output_constant): Return the number of bytes actually
3231 (output_constructor_array_range): Update the field size with the
3232 number of bytes emitted by output_constant.
3233 (output_constructor_regular_field): Likewise. Also do not
3234 complain if the total number of bytes emitted is now greater
3235 than the expected fieldpos.
3236 * output.h (output_constant): Update prototype and descriptive comment.
3238 2014-01-16 Marek Polacek <polacek@redhat.com>
3241 * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
3242 it is error_mark_node.
3244 2014-01-15 Uros Bizjak <ubizjak@gmail.com>
3246 * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
3247 VALID_AVX256_REG_OR_OI_MODE.
3249 2014-01-15 Pat Haugen <pthaugen@us.ibm.com>
3251 * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
3252 current procedure should be profiled.
3254 2014-01-15 Andrew Pinski <apinski@cavium.com>
3256 * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
3257 of moving from/to the STACK_REG register class.
3259 2014-01-15 Richard Henderson <rth@redhat.com>
3262 * reginfo.c (global_regs_decl): Globalize.
3263 * rtl.h (global_regs_decl): Declare.
3264 * ira.c (do_reload): Diagnose frame_pointer_needed and it
3265 reserved via global_regs.
3267 2014-01-15 Teresa Johnson <tejohnson@google.com>
3269 * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
3271 2014-01-15 Bill Schmidt <wschmidt@vnet.linux.ibm.com>
3273 * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
3274 and vmulosh rather than call gen_vec_widen_smult_*.
3275 (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
3276 than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
3277 (vec_widen_smult_even_v16qi): Likewise.
3278 (vec_widen_umult_even_v8hi): Likewise.
3279 (vec_widen_smult_even_v8hi): Likewise.
3280 (vec_widen_umult_odd_v16qi): Likewise.
3281 (vec_widen_smult_odd_v16qi): Likewise.
3282 (vec_widen_umult_odd_v8hi): Likewise.
3283 (vec_widen_smult_odd_v8hi): Likewise.
3284 (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
3285 vmuloub rather than call gen_vec_widen_umult_*.
3286 (vec_widen_umult_lo_v16qi): Likewise.
3287 (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
3288 vmulosb rather than call gen_vec_widen_smult_*.
3289 (vec_widen_smult_lo_v16qi): Likewise.
3290 (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
3291 rather than call gen_vec_widen_umult_*.
3292 (vec_widen_umult_lo_v8hi): Likewise.
3293 (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
3294 rather than call gen_vec_widen_smult_*.
3295 (vec_widen_smult_lo_v8hi): Likewise.
3297 2014-01-15 Jeff Law <law@redhat.com>
3299 PR tree-optimization/59747
3300 * ree.c (find_and_remove_re): Properly handle case where a second
3301 eliminated extension requires widening a copy created for elimination
3302 of a prior extension.
3303 (combine_set_extension): Ensure that the number of hard regs needed
3304 for a destination register does not change when we widen it.
3306 2014-01-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
3308 * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
3309 (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
3310 (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
3311 (arm*-*-rtems*): Use t-rtems from existing tmake_file.
3312 (avr-*-rtems*): Likewise.
3313 (bfin*-rtems*): Likewise.
3314 (moxie-*-rtems*): Likewise.
3315 (h8300-*-rtems*): Likewise.
3316 (i[34567]86-*-rtems*): Likewise.
3317 (lm32-*-rtems*): Likewise.
3318 (m32r-*-rtems*): Likewise.
3319 (m68k-*-rtems*): Likewise.
3320 (microblaze*-*-rtems*): Likewise.
3321 (mips*-*-rtems*): Likewise.
3322 (powerpc-*-rtems*): Likewise.
3323 (sh-*-rtems*): Likewise.
3324 (sparc-*-rtems*): Likewise.
3325 (sparc64-*-rtems*): Likewise.
3326 (v850-*-rtems*): Likewise.
3327 (m32c-*-rtems*): Likewise.
3329 2014-01-15 Vladimir Makarov <vmakarov@redhat.com>
3331 PR rtl-optimization/59511
3332 * ira.c (ira_init_register_move_cost): Use memory costs for some
3333 cases of register move cost calculations.
3334 * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
3335 instead of BB frequency.
3336 * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
3337 * lra-assigns.c (find_hard_regno_for): Ditto.
3339 2014-01-15 Richard Biener <rguenther@suse.de>
3341 PR tree-optimization/59822
3342 * tree-vect-stmts.c (hoist_defs_of_uses): New function.
3343 (vectorizable_load): Use it to hoist defs of uses of invariant
3344 loads out of the loop.
3346 2014-01-15 Matthew Gretton-Dann <matthew.gretton-dann@linaro.org>
3347 Kugan Vivekanandarajah <kuganv@linaro.org>
3350 * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
3353 2014-01-15 Richard Biener <rguenther@suse.de>
3355 PR rtl-optimization/59802
3356 * lcm.c (compute_available): Use inverted postorder to seed
3357 the initial worklist.
3359 2014-01-15 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
3362 * config/s390/s390.c (s390_preferred_reload_class): Don't return
3363 ADDR_REGS for invalid symrefs in non-PIC code.
3365 2014-01-15 Jakub Jelinek <jakub@redhat.com>
3368 * builtins.c (determine_block_size): Initialize *probable_max_size
3369 even if len_rtx is CONST_INT.
3371 2014-01-14 Andrew Pinski <apinski@cavium.com>
3373 * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
3374 * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
3375 (cortexa53_tunings): Likewise.
3376 (aarch64_sched_issue_rate): New function.
3377 (TARGET_SCHED_ISSUE_RATE): Define.
3379 2014-01-14 Vladimir Makarov <vmakarov@redhat.com>
3381 * ira-costs.c (find_costs_and_classes): Add missed
3382 ira_init_register_move_cost_if_necessary.
3384 2014-01-14 Vladimir Makarov <vmakarov@redhat.com>
3387 * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
3389 2014-01-14 H.J. Lu <hongjiu.lu@intel.com>
3392 * config/i386/i386.c (type_natural_mode): Add a bool parameter
3393 to indicate if type is used for function return value. Warn ABI
3394 change if the vector mode isn't available for function return value.
3395 (ix86_function_arg_advance): Pass false to type_natural_mode.
3396 (ix86_function_arg): Likewise.
3397 (ix86_gimplify_va_arg): Likewise.
3398 (function_arg_32): Don't warn ABI change.
3399 (ix86_function_value): Pass true to type_natural_mode.
3400 (ix86_return_in_memory): Likewise.
3401 (ix86_struct_value_rtx): Removed.
3402 (TARGET_STRUCT_VALUE_RTX): Likewise.
3404 2014-01-14 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
3406 * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
3407 converting a conditional jump into a conditional return.
3409 2014-01-14 Richard Biener <rguenther@suse.de>
3411 PR tree-optimization/58921
3412 PR tree-optimization/59006
3413 * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
3414 hoisting invariant stmts.
3415 * tree-vect-stmts.c (vectorizable_load): Insert the splat of
3416 invariant loads on the preheader edge if possible.
3418 2014-01-14 Joey Ye <joey.ye@arm.com>
3420 * doc/plugin.texi (Building GCC plugins): Update to C++.
3422 2014-01-14 Kirill Yukhin <kirill.yukhin@intel.com>
3424 * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
3425 (_mm_rcp28_round_ss): Ditto.
3426 (_mm_rsqrt28_round_sd): Ditto.
3427 (_mm_rsqrt28_round_ss): Ditto.
3428 (_mm_rcp28_sd): Ditto.
3429 (_mm_rcp28_ss): Ditto.
3430 (_mm_rsqrt28_sd): Ditto.
3431 (_mm_rsqrt28_ss): Ditto.
3432 * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
3433 * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
3434 * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
3435 (IX86_BUILTIN_RCP28SD): Ditto.
3436 (IX86_BUILTIN_RCP28SS): Ditto.
3437 (IX86_BUILTIN_RSQRT28SD): Ditto.
3438 (IX86_BUILTIN_RSQRT28SS): Ditto.
3439 (bdesc_special_args): Define __builtin_ia32_movntdqa512,
3440 __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
3441 __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
3442 (ix86_expand_special_args_builtin): Expand new FTYPE.
3443 * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
3444 (srcp14<mode>): Make insn unary.
3445 (avx512f_vmscalef<mode><round_name>): Use substed predicate.
3446 (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
3447 (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
3448 (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
3449 (avx512er_exp2<mode><mask_name><round_saeonly_name>):
3450 Fix rounding: make it SAE only.
3451 (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
3453 (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
3455 (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
3456 (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
3457 (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
3458 * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
3459 (round_saeonly_mask_scalar_operand4): Ditto.
3460 (round_saeonly_mask_scalar_op3): Ditto.
3461 (round_saeonly_mask_scalar_op4): Ditto.
3463 2014-01-13 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3465 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3466 Implement -maltivec=be for vec_insert and vec_extract.
3468 2014-01-10 DJ Delorie <dj@redhat.com>
3470 * config/msp430/msp430.md (call_internal): Don't allow memory
3471 references with SP as the base register.
3472 (call_value_internal): Likewise.
3473 * config/msp430/constraints.md (Yc): New. For memory references
3474 that don't use SP as a base register.
3476 * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
3477 "an integer without a # prefix"
3478 * config/msp430/msp430.md (epilogue_helper): Use it.
3480 2014-01-13 Jakub Jelinek <jakub@redhat.com>
3483 * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
3484 AVX512F gather builtins.
3485 * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
3486 on gather decls with INTEGER_TYPE masktype.
3487 (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
3488 directly into the builtin rather than hoisting it before loop.
3490 PR tree-optimization/59387
3491 * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
3492 (scev_const_prop): If folded_casts and type has undefined overflow,
3493 use force_gimple_operand instead of force_gimple_operand_gsi and
3494 for each added stmt if it is assign with
3495 arith_code_with_undefined_signed_overflow, call
3496 rewrite_to_defined_overflow.
3497 * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
3498 gimple-fold.h instead.
3499 (arith_code_with_undefined_signed_overflow,
3500 rewrite_to_defined_overflow): Moved to ...
3501 * gimple-fold.c (arith_code_with_undefined_signed_overflow,
3502 rewrite_to_defined_overflow): ... here. No longer static.
3503 Include gimplify-me.h.
3504 * gimple-fold.h (arith_code_with_undefined_signed_overflow,
3505 rewrite_to_defined_overflow): New prototypes.
3507 2014-01-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3509 * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
3511 2014-01-13 Eric Botcazou <ebotcazou@adacore.com>
3513 * builtins.c (get_object_alignment_2): Minor tweak.
3514 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
3516 2014-01-13 Christian Bruel <christian.bruel@st.com>
3518 * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
3519 optimized non constant lengths.
3521 2014-01-13 Jakub Jelinek <jakub@redhat.com>
3524 * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
3525 load as __atomic_load_N if possible.
3527 2014-01-11 David Edelsohn <dje.gcc@gmail.com>
3529 * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
3531 (rs6000_expand_builtin): Adjust call.
3533 2014-01-11 David Edelsohn <dje.gcc@gmail.com>
3536 * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
3537 * config/rs6000/rs6000.c: Include target-globals.h.
3538 (rs6000_set_current_function): Instead of doing target_reinit
3539 unconditionally, use save_target_globals_default_opts and
3540 restore_target_globals.
3542 * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
3544 * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
3545 (rs6000_expand_builtin): Handle mffs and mtfsf.
3546 (rs6000_init_builtins): Define mffs and mtfsf.
3547 * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
3548 (rs6000_mffs): New pattern.
3549 (rs6000_mtfsf): New pattern.
3551 2014-01-11 Bin Cheng <bin.cheng@arm.com>
3553 * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
3554 Start narrowing with START. Apply candidate-use pair
3555 and check overall cost in narrowing.
3556 (iv_ca_prune): Pass new argument.
3558 2014-01-10 Jeff Law <law@redhat.com>
3561 * ree.c (combine_reaching_defs): Ensure the defining statement
3562 occurs before the extension when optimizing extensions with
3563 different source and destination hard registers.
3565 2014-01-10 Jan Hubicka <hubicka@ucw.cz>
3568 * ipa-devirt.c (build_type_inheritance_graph): Also add types of
3569 vtables into the type inheritance graph.
3571 2014-01-10 Jakub Jelinek <jakub@redhat.com>
3573 PR rtl-optimization/59754
3574 * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
3575 modes in the REGNO != REGNO case.
3577 2014-01-10 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3579 * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
3581 2014-01-10 Jakub Jelinek <jakub@redhat.com>
3583 PR tree-optimization/59745
3584 * tree-predcom.c (tree_predictive_commoning_loop): Call
3585 free_affine_expand_cache if giving up because components is NULL.
3587 * target-globals.c (save_target_globals): Allocate < 4KB structs using
3588 GC in payload of target_globals struct instead of allocating them on
3589 the heap and the larger structs separately using GC.
3590 * target-globals.h (struct target_globals): Make regs, hard_regs,
3591 reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
3592 of GTY((skip)) and change type to void *.
3593 (reset_target_globals): Cast loads from those fields to corresponding
3596 2014-01-10 Steve Ellcey <sellcey@mips.com>
3599 * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
3600 gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
3601 tree-ssanames.h, print-tree.h, varasm.h, and context.h.
3603 2014-01-10 Richard Earnshaw <rearnsha@arm.com>
3606 * aarch64-modes.def (CC_Zmode): New flags mode.
3607 * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
3608 represents an equality.
3609 (aarch64_get_condition_code): Handle CC_Zmode.
3610 * aarch64.md (compare_neg<mode>): Restrict to equality operations.
3612 2014-01-10 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
3614 * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
3615 extraction in good case.
3617 2014-01-10 Richard Biener <rguenther@suse.de>
3619 PR tree-optimization/59374
3620 * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
3621 checking after SLP discovery. Mark stmts not participating
3622 in any SLP instance properly.
3624 2014-01-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3626 * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
3627 when handling a SET rtx.
3629 2014-01-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3631 * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
3632 (cortex-a57): Likewise.
3633 (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
3635 2014-01-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3637 * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
3638 non-iwmmxt builtins.
3640 2014-01-10 Jan Hubicka <hubicka@ucw.cz>
3644 * ipa-devirt.c record_target_from_binfo): Take as argument
3645 stack of binfos and lookup matching one for virtual inheritance.
3646 (possible_polymorphic_call_targets_1): Update.
3648 2014-01-10 Huacai Chen <chenhc@lemote.com>
3650 * config/mips/driver-native.c (host_detect_local_cpu): Handle new
3651 kernel strings for Loongson-2E/2F/3A.
3653 2014-01-10 Jakub Jelinek <jakub@redhat.com>
3656 * tree-vect-data-refs.c (vect_analyze_data_refs): Check
3657 is_gimple_call before calling gimple_call_internal_p.
3659 2014-01-09 Steve Ellcey <sellcey@mips.com>
3661 * Makefile.in (TREE_FLOW_H): Remove.
3662 (TREE_SSA_H): Add file names from tree-flow.h.
3663 * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
3664 * tree.h: Remove tree-flow.h reference.
3665 * hash-table.h: Remove tree-flow.h reference.
3666 * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
3667 reference with tree-ssa-loop.h.
3669 2014-01-09 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3671 * doc/invoke.texi: Add -maltivec={be,le} options, and document
3672 default element-order behavior for -maltivec.
3673 * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
3674 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
3675 that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
3676 when targeting big endian, at least for now.
3677 * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
3679 2014-01-09 Jakub Jelinek <jakub@redhat.com>
3682 * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
3683 var satisfies use_register_for_decl, just take into account type
3684 alignment, rather than decl alignment.
3686 PR tree-optimization/59622
3687 * gimple-fold.c (gimple_fold_call): Fix a typo in message. For
3688 __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
3689 __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
3690 Don't devirtualize for inplace at all. For targets.length () == 1,
3691 if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
3693 2014-01-09 H.J. Lu <hongjiu.lu@intel.com>
3695 * config/i386/i386.md (cpu): Remove the unused btver1.
3697 2014-01-09 H.J. Lu <hongjiu.lu@intel.com>
3699 * gdbasan.in: Put a breakpoint on __sanitizer::Report.
3701 2014-01-09 Jakub Jelinek <jakub@redhat.com>
3704 * tree-core.h (struct target_globals): New forward declaration.
3705 (struct tree_target_option): Add globals field.
3706 * tree.h (TREE_TARGET_GLOBALS): Define.
3707 (prepare_target_option_nodes_for_pch): New prototype.
3708 * target-globals.h (struct target_globals): Define even if
3710 * tree.c (prepare_target_option_node_for_pch,
3711 prepare_target_option_nodes_for_pch): New functions.
3712 * config/i386/i386.h (SWITCHABLE_TARGET): Define.
3713 * config/i386/i386.c: Include target-globals.h.
3714 (ix86_set_current_function): Instead of doing target_reinit
3715 unconditionally, use save_target_globals_default_opts and
3716 restore_target_globals.
3718 2014-01-09 Richard Biener <rguenther@suse.de>
3720 PR tree-optimization/59715
3721 * tree-cfg.h (split_critical_edges): Declare.
3722 * tree-cfg.c (split_critical_edges): Export.
3723 * tree-ssa-sink.c (execute_sink_code): Split critical edges.
3725 2014-01-09 Max Ostapenko <m.ostapenko@partner.samsung.com>
3727 * cfgexpand.c (expand_stack_vars): Optionally disable
3728 asan stack protection.
3729 (expand_used_vars): Likewise.
3730 (partition_stack_vars): Likewise.
3731 * asan.c (asan_emit_stack_protection): Optionally disable
3732 after return stack usage.
3733 (instrument_derefs): Optionally disable memory access instrumentation.
3734 (instrument_builtin_call): Likewise.
3735 (instrument_strlen_call): Likewise.
3736 (asan_protect_global): Optionally disable global variables protection.
3737 * doc/invoke.texi: Added doc for new options.
3738 * params.def: Added new options.
3739 * params.h: Likewise.
3741 2014-01-09 Jakub Jelinek <jakub@redhat.com>
3743 PR rtl-optimization/59724
3744 * ifcvt.c (cond_exec_process_if_block): Don't call
3745 flow_find_head_matching_sequence with 0 longest_match.
3746 * cfgcleanup.c (flow_find_head_matching_sequence): Count even
3747 non-active insns if !stop_after.
3748 (try_head_merge_bb): Revert 2014-01-07 changes.
3750 2014-01-08 Jeff Law <law@redhat.com>
3752 * ree.c (get_sub_rtx): New function, extracted from...
3753 (merge_def_and_ext): Here.
3754 (combine_reaching_defs): Use get_sub_rtx.
3756 2014-01-08 Eric Botcazou <ebotcazou@adacore.com>
3758 * cgraph.h (varpool_variable_node): Do not choke on null node.
3760 2014-01-08 Catherine Moore <clm@codesourcery.com>
3762 * config/mips/mips.md (simple_return): Attempt to use JRC
3764 * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
3766 2014-01-08 Richard Sandiford <rdsandiford@googlemail.com>
3768 PR rtl-optimization/59137
3769 * reorg.c (steal_delay_list_from_target): Call update_block for
3771 (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
3773 2014-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3775 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3776 two duplicate entries.
3778 2014-01-08 Richard Sandiford <rdsandiford@googlemail.com>
3781 2012-10-07 Richard Sandiford <rdsandiford@googlemail.com>
3783 * config/mips/mips.c (mips_truncated_op_cost): New function.
3784 (mips_rtx_costs): Adjust test for BADDU.
3785 * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
3787 2012-10-02 Richard Sandiford <rdsandiford@googlemail.com>
3789 * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
3790 (*baddu_si): ...this new pattern.
3792 2014-01-08 Jakub Jelinek <jakub@redhat.com>
3795 * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
3797 2014-01-08 Bernd Edlinger <bernd.edlinger@hotmail.de>
3800 * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
3802 (expand_expr, expand_normal): Adjust.
3803 * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
3804 inner_reference_p. Use inner_reference_p to expand inner references.
3805 (store_expr): Adjust.
3806 * cfgexpand.c (expand_call_stmt): Adjust.
3808 2014-01-08 Rong Xu <xur@google.com>
3810 * gcov-io.c (gcov_var): Move from gcov-io.h.
3811 (gcov_position): Ditto.
3812 (gcov_is_error): Ditto.
3813 (gcov_rewrite): Ditto.
3814 * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
3815 only part to libgcc/libgcov.h.
3817 2014-01-08 Marek Polacek <polacek@redhat.com>
3820 * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
3822 2014-01-08 Marek Polacek <polacek@redhat.com>
3825 * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
3827 2014-01-08 Jakub Jelinek <jakub@redhat.com>
3829 PR rtl-optimization/59649
3830 * stor-layout.c (get_mode_bounds): For BImode return
3831 0 and STORE_FLAG_VALUE.
3833 2014-01-08 Richard Biener <rguenther@suse.de>
3836 * gimple.h (is_gimple_builtin_call): Remove.
3837 (gimple_builtin_call_types_compatible_p): New.
3838 (gimple_call_builtin_p): New overload.
3839 * gimple.c (is_gimple_builtin_call): Remove.
3840 (validate_call): Rename to ...
3841 (gimple_builtin_call_types_compatible_p): ... this and export. Also
3843 (validate_type): New static function.
3844 (gimple_call_builtin_p): New overload and adjust.
3845 * gimple-fold.c (gimple_fold_builtin): Fold the return value.
3846 (gimple_fold_call): Likewise. Use gimple_call_builtin_p.
3847 (gimple_fold_stmt_to_constant_1): Likewise.
3848 * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
3850 2014-01-08 Richard Biener <rguenther@suse.de>
3853 * gimplify.c (gimplify_expr): Gimplify register-register type
3854 VIEW_CONVERT_EXPRs to separate stmts.
3856 2014-01-07 Jeff Law <law@redhat.com>
3859 * ree.c (combine_set_extension): Handle case where source
3860 and destination registers in an extension insn are different.
3861 (combine_reaching_defs): Allow source and destination registers
3862 in extension to be different under limited circumstances.
3863 (add_removable_extension): Remove restriction that the
3864 source and destination registers in the extension are the same.
3865 (find_and_remove_re): Emit a copy from the extension's
3866 destination to its source after the defining insn if
3867 the source and destination registers are different.
3870 * ifcvt.c (merge_if_block): If we are merging a block with more than
3871 one successor with a block with no successors, remove any BARRIER
3872 after the second block.
3874 2014-01-07 Dan Xio Qiang <ziyan01@163.com>
3876 * hw-doloop.c (reorg_loops): Release the bitmap obstack.
3878 2014-01-07 John David Anglin <danglin@gcc.gnu.org>
3881 * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
3882 for 14-bit register offsets when INT14_OK_STRICT is false.
3884 2014-01-07 Roland Stigge <stigge@antcom.de>
3885 Michael Meissner <meissner@linux.vnet.ibm.com>
3888 * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
3889 Only check TFmode for SPE constants. Don't check TImode or TDmode.
3891 2014-01-07 James Greenhalgh <james.greenhalgh@arm.com>
3893 * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
3896 2014-01-07 Yufeng Zhang <yufeng.zhang@arm.com>
3898 * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
3899 with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
3900 rtx is const0_rtx or not.
3902 2014-01-07 Richard Sandiford <rdsandiford@googlemail.com>
3905 * target-globals.c (save_target_globals): Remove this_fn_optab
3907 * toplev.c: Include optabs.h.
3908 (target_reinit): Temporarily restore the global options if another
3909 set of options are in force.
3911 2014-01-07 Jakub Jelinek <jakub@redhat.com>
3913 PR rtl-optimization/58668
3914 * cfgcleanup.c (flow_find_cross_jump): Don't count
3915 any jumps if dir_p is NULL. Remove p1 variable, use active_insn_p
3916 to determine what is counted.
3917 (flow_find_head_matching_sequence): Use active_insn_p to determine
3919 (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
3921 * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
3922 determine what is counted.
3924 PR tree-optimization/59643
3925 * tree-predcom.c (split_data_refs_to_components): If one dr is
3926 read and one write, determine_offset fails and the write isn't
3927 in the bad component, just put the read into the bad component.
3929 2014-01-07 Mike Stump <mikestump@comcast.net>
3930 Jakub Jelinek <jakub@redhat.com>
3933 * tree-core.h (struct tree_optimization_option): Change optabs
3934 type from unsigned char * to void *.
3935 * optabs.c (init_tree_optimization_optabs): Adjust
3936 TREE_OPTIMIZATION_OPTABS initialization.
3938 2014-01-06 Jakub Jelinek <jakub@redhat.com>
3941 * config/i386/i386.h (struct machine_function): Add
3942 no_drap_save_restore field.
3943 * config/i386/i386.c (ix86_save_reg): Use
3944 !cfun->machine->no_drap_save_restore instead of
3945 crtl->stack_realign_needed.
3946 (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
3947 this function clears frame_pointer_needed. Set
3948 cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
3949 and DRAP reg is needed.
3951 2014-01-06 Marek Polacek <polacek@redhat.com>
3954 * doc/implement-c.texi: Mention that other integer types are
3955 permitted as bit-field types in strictly conforming mode.
3957 2014-01-06 Felix Yang <fei.yang0953@gmail.com>
3959 * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
3962 2014-01-06 Richard Earnshaw <rearnsha@arm.com>
3964 * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
3966 2014-01-06 Martin Jambor <mjambor@suse.cz>
3969 * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
3971 * ipa-prop.c (ipa_print_node_params): Fix indentation.
3973 2014-01-06 Eric Botcazou <ebotcazou@adacore.com>
3977 * var-tracking.c (add_stores): Preserve the value of the source even if
3978 we don't record the store.
3980 2014-01-06 Terry Guo <terry.guo@arm.com>
3982 * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
3984 2014-01-05 Iain Sandoe <iain@codesourcery.com>
3987 * config/darwin.c (darwin_function_section): Adjust return values to
3988 correspond to optimisation changes made in r206070.
3990 2014-01-05 Uros Bizjak <ubizjak@gmail.com>
3992 * config/i386/i386.c (ix86_data_alignment): Calculate max_align
3993 from prefetch_block tune setting.
3994 (nocona_cost): Correct size of prefetch block to 64.
3996 2014-01-04 Eric Botcazou <ebotcazou@adacore.com>
3998 * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
3999 (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
4000 used to save the static chain register in the computation of the offset
4001 from which the FP registers need to be restored.
4003 2014-01-04 Jakub Jelinek <jakub@redhat.com>
4005 PR tree-optimization/59519
4006 * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
4007 ICE if get_current_def (current_new_name) is already non-NULL, as long
4008 as it is a phi result of some other phi in *new_exit_bb that has
4011 * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
4012 or vmovdqu* for misaligned_operand.
4013 (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
4014 <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
4015 * config/i386/i386.c (ix86_expand_special_args_builtin): Set
4016 aligned_mem for AVX512F masked aligned load and store builtins and for
4019 2014-01-03 Bingfeng Mei <bmei@broadcom.com>
4021 PR tree-optimization/59651
4022 * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
4023 Address range for negative step should be added by TYPE_SIZE_UNIT.
4025 2014-01-03 Andreas Schwab <schwab@linux-m68k.org>
4027 * config/m68k/m68k.c (handle_move_double): Handle pushes with
4028 overlapping registers also for registers other than the stack pointer.
4030 2014-01-03 Marek Polacek <polacek@redhat.com>
4033 * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
4036 2014-01-03 Jakub Jelinek <jakub@redhat.com>
4039 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
4042 * config/i386/i386.md (MODE_SIZE): New mode attribute.
4043 (push splitter): Use <P:MODE_SIZE> instead of
4044 GET_MODE_SIZE (<P:MODE>mode).
4045 (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
4046 (mov -1, reg peephole2): Likewise.
4047 * config/i386/sse.md (*mov<mode>_internal,
4048 <sse>_storeu<ssemodesuffix><avxsizesuffix>,
4049 <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
4050 *<code><mode>3, *andnot<mode>3<mask_name>,
4051 <mask_codefor><code><mode>3<mask_name>): Likewise.
4052 * config/i386/subst.md (mask_mode512bit_condition,
4053 sd_mask_mode512bit_condition): Likewise.
4055 2014-01-02 Xinliang David Li <davidxl@google.com>
4057 PR tree-optimization/59303
4058 * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
4059 (dump_predicates): Better output format.
4060 (pred_equal_p): New function.
4061 (is_neq_relop_p): Ditto.
4062 (is_neq_zero_form_p): Ditto.
4063 (pred_expr_equal_p): Ditto.
4064 (pred_neg_p): Ditto.
4065 (simplify_pred): Ditto.
4066 (simplify_preds_2): Ditto.
4067 (simplify_preds_3): Ditto.
4068 (simplify_preds_4): Ditto.
4069 (simplify_preds): Ditto.
4071 (push_to_worklist): Ditto.
4072 (get_pred_info_from_cmp): Ditto.
4073 (is_degenerated_phi): Ditto.
4074 (normalize_one_pred_1): Ditto.
4075 (normalize_one_pred): Ditto.
4076 (normalize_one_pred_chain): Ditto.
4077 (normalize_preds): Ditto.
4078 (normalize_cond_1): Remove function.
4079 (normalize_cond): Ditto.
4080 (is_gcond_subset_of): Ditto.
4081 (is_subset_of_any): Ditto.
4082 (is_or_set_subset_of): Ditto.
4083 (is_and_set_subset_of): Ditto.
4084 (is_norm_cond_subset_of): Ditto.
4085 (pred_chain_length_cmp): Ditto.
4086 (convert_control_dep_chain_into_preds): Type change.
4087 (find_predicates): Ditto.
4088 (find_def_preds): Ditto.
4089 (destroy_predicates_vecs): Ditto.
4090 (find_matching_predicates_in_rest_chains): Ditto.
4091 (use_pred_not_overlap_with_undef_path_pred): Ditto.
4092 (is_pred_expr_subset): Ditto.
4093 (is_pred_chain_subset_of): Ditto.
4094 (is_included_in): Ditto.
4095 (is_superset_of): Ditto.
4097 2014-01-02 Richard Sandiford <rdsandiford@googlemail.com>
4099 Update copyright years.
4101 2014-01-02 Richard Sandiford <rdsandiford@googlemail.com>
4103 * common/config/arc/arc-common.c, config/arc/arc-modes.def,
4104 config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
4105 config/arc/arc.md, config/arc/arc.opt,
4106 config/arm/arm_neon_builtins.def, config/arm/crypto.def,
4107 config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
4108 config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
4109 config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
4110 config/linux-protos.h, config/linux.c, config/winnt-c.c,
4111 diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
4112 vtable-verify.c, vtable-verify.h: Use the standard form for the
4115 2014-01-02 Tobias Burnus <burnus@net-b.de>
4117 * gcc.c (process_command): Update copyright notice dates.
4118 * gcov-dump.c: Ditto.
4120 * doc/cpp.texi: Bump @copying's copyright year.
4121 * doc/cppinternals.texi: Ditto.
4122 * doc/gcc.texi: Ditto.
4123 * doc/gccint.texi: Ditto.
4124 * doc/gcov.texi: Ditto.
4125 * doc/install.texi: Ditto.
4126 * doc/invoke.texi: Ditto.
4128 2014-01-01 Jan-Benedict Glaw <jbglaw@lug-owl.de>
4130 * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
4132 2014-01-01 Jakub Jelinek <jakub@redhat.com>
4134 * config/i386/sse.md (*mov<mode>_internal): Guard
4135 EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
4137 PR rtl-optimization/59647
4138 * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
4139 new_rtx into UNSIGNED_FLOAT rtxes.
4141 Copyright (C) 2014 Free Software Foundation, Inc.
4143 Copying and distribution of this file, with or without modification,
4144 are permitted in any medium without royalty provided the copyright
4145 notice and this notice are preserved.