1 2014-02-25 Richard Biener <rguenther@suse.de>
4 * lto-opts.c (lto_write_options): Output non-explicit conservative
5 -fwrapv, -fno-trapv and -fno-strict-overflow.
6 * lto-wrapper.c (merge_and_complain): Handle merging those options.
7 (run_gcc): And pass them through.
9 2014-02-25 Andrey Belevantsev <abel@ispras.ru>
11 * sel-sched.c (calculate_new_fences): New parameter ptime.
12 Calculate it as a maximum over all fence cycles.
13 (sel_sched_region_2): Adjust the call to calculate_new_fences.
14 Print the final schedule timing when sched_verbose.
16 2014-02-25 Andrey Belevantsev <abel@ispras.ru>
18 PR rtl-optimization/60292
19 * sel-sched.c (fill_vec_av_set): Do not reset target availability
20 bit fot the fence instruction.
22 2014-02-24 Alangi Derick <alangiderick@gmail.com>
24 * calls.h: Fix typo in comment.
26 2014-02-24 John David Anglin <danglin@gcc.gnu.org>
28 * config/pa/pa.c (pa_output_move_double): Don't valididate when
29 adjusting offsetable addresses.
31 2014-02-24 Guozhi Wei <carrot@google.com>
33 * sparseset.h (sparseset_pop): Fix the wrong index.
35 2014-02-24 Walter Lee <walt@tilera.com>
37 * config.gcc (tilepro-*-*): Change to tilepro*-*-*.
38 (tilegx-*-linux*): Change to tilegx*-*-linux*; Support tilegxbe
40 * common/config/tilegx/tilegx-common.c
41 (TARGET_DEFAULT_TARGET_FLAGS): Define.
42 * config/tilegx/linux.h (ASM_SPEC): Add endian_spec.
44 * config/tilegx/sync.md (atomic_test_and_set): Handle big endian.
45 * config/tilegx/tilegx.c (tilegx_return_in_msb): New.
46 (tilegx_gimplify_va_arg_expr): Handle big endian.
47 (tilegx_expand_unaligned_load): Ditto.
48 (tilegx_expand_unaligned_store): Ditto.
49 (TARGET_RETURN_IN_MSB): New.
50 * config/tilegx/tilegx.h (TARGET_DEFAULT): New.
51 (TARGET_ENDIAN_DEFAULT): New.
52 (TARGET_BIG_ENDIAN): Handle big endian.
53 (BYTES_BIG_ENDIAN): Ditto.
54 (WORDS_BIG_ENDIAN): Ditto.
55 (FLOAT_WORDS_BIG_ENDIAN): Ditto.
58 * config/tilegx/tilegx.md (extv): Handle big endian.
61 (insn_st<n>_add<bitsuffix>): Ditto.
62 (insn_stnt<n>): Ditto.
63 (insn_stnt<n>_add<bitsuffix>):Ditto.
64 (vec_interleave_highv8qi): Handle big endian.
65 (vec_interleave_highv8qi_be): New.
66 (vec_interleave_highv8qi_le): New.
67 (insn_v1int_h): Handle big endian.
68 (vec_interleave_lowv8qi): Handle big endian.
69 (vec_interleave_lowv8qi_be): New.
70 (vec_interleave_lowv8qi_le): New.
71 (insn_v1int_l): Handle big endian.
72 (vec_interleave_highv4hi): Handle big endian.
73 (vec_interleave_highv4hi_be): New.
74 (vec_interleave_highv4hi_le): New.
75 (insn_v2int_h): Handle big endian.
76 (vec_interleave_lowv4hi): Handle big endian.
77 (vec_interleave_lowv4hi_be): New.
78 (vec_interleave_lowv4hi_le): New.
79 (insn_v2int_l): Handle big endian.
80 (vec_interleave_highv2si): Handle big endian.
81 (vec_interleave_highv2si_be): New.
82 (vec_interleave_highv2si_le): New.
83 (insn_v4int_h): Handle big endian.
84 (vec_interleave_lowv2si): Handle big endian.
85 (vec_interleave_lowv2si_be): New.
86 (vec_interleave_lowv2si_le): New.
87 (insn_v4int_l): Handle big endian.
88 * config/tilegx/tilegx.opt (mbig-endian): New option.
89 (mlittle-endian): New option.
90 * doc/install.texi: Document tilegxbe-linux.
91 * doc/invoke.texi: Document -mbig-endian and -mlittle-endian.
93 2014-02-24 Martin Jambor <mjambor@suse.cz>
96 * ipa-cp.c (propagate_constants_accross_call): Bail out early if
97 there are no parameter descriptors.
99 2014-02-24 Andrey Belevantsev <abel@ispras.ru>
101 PR rtl-optimization/60268
102 * sched-rgn.c (haifa_find_rgns): Move the nr_regions_initial variable
103 initialization to ...
104 (sched_rgn_init): ... here.
105 (schedule_region): Check for SCHED_PRESSURE_NONE earlier.
107 2014-02-23 David Holsgrove <david.holsgrove@xilinx.com>
109 * config/microblaze/microblaze.md: Correct ashrsi_reg / lshrsi_reg
112 2014-02-23 Edgar E. Iglesias <edgar.iglesias@xilinx.com>
114 * config/microblaze/microblaze.h: Remove SECONDARY_MEMORY_NEEDED
117 2014-02-23 David Holsgrove <david.holsgrove@xilinx.com>
119 * /config/microblaze/microblaze.c: Add microblaze_asm_output_mi_thunk
120 and define TARGET_ASM_OUTPUT_MI_THUNK and
121 TARGET_ASM_CAN_OUTPUT_MI_THUNK.
123 2014-02-23 David Holsgrove <david.holsgrove@xilinx.com>
125 * config/microblaze/predicates.md: Add cmp_op predicate.
126 * config/microblaze/microblaze.md: Add branch_compare instruction
127 which uses cmp_op predicate and emits cmp insn before branch.
128 * config/microblaze/microblaze.c (microblaze_emit_compare): Rename
129 to microblaze_expand_conditional_branch and consolidate logic.
130 (microblaze_expand_conditional_branch): emit branch_compare
131 insn instead of handling cmp op separate from branch insn.
133 2014-02-23 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
135 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Relax assert
138 2014-02-23 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
140 * config/rs6000/altivec.md (altivec_lve<VI_char>x): Replace
141 define_insn with define_expand and new define_insn
142 *altivec_lve<VI_char>x_internal.
143 (altivec_stve<VI_char>x): Replace define_insn with define_expand
144 and new define_insn *altivec_stve<VI_char>x_internal.
145 * config/rs6000/rs6000-protos.h (altivec_expand_stvex_be): New
147 * config/rs6000/rs6000.c (altivec_expand_lvx_be): Document use by
149 (altivec_expand_stvex_be): New function.
151 2014-02-22 Joern Rennecke <joern.rennecke@embecosm.com>
153 * config/avr/avr.c (avr_can_eliminate): Allow elimination from
155 to STACK_POINTER_REGNUM if !frame_pointer_needed.
156 * config/avr/avr.c (ELIMINABLE_REGS): Add elimination from
157 ARG_POINTER_REGNUM to STACK_POINTER_REGNUM.
159 2014-02-21 Vladimir Makarov <vmakarov@redhat.com>
162 * lra-constraints.c (inherit_reload_reg): Use lra_emit_move
163 instead of emit_move_insn.
165 2014-02-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
167 * config/rs6000/altivec.md (altivec_vsumsws): Replace second
169 (reduc_uplus_v16qi): Use gen_altivec_vsumsws_direct instead of
172 2014-02-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
174 * config/rs6000/altivec.md (altivec_lvxl): Rename as
175 *altivec_lvxl_<mode>_internal and use VM2 iterator instead of
177 (altivec_lvxl_<mode>): New define_expand incorporating
178 -maltivec=be semantics where needed.
179 (altivec_lvx): Rename as *altivec_lvx_<mode>_internal.
180 (altivec_lvx_<mode>): New define_expand incorporating -maltivec=be
181 semantics where needed.
182 (altivec_stvx): Rename as *altivec_stvx_<mode>_internal.
183 (altivec_stvx_<mode>): New define_expand incorporating
184 -maltivec=be semantics where needed.
185 (altivec_stvxl): Rename as *altivec_stvxl_<mode>_internal and use
186 VM2 iterator instead of V4SI.
187 (altivec_stvxl_<mode>): New define_expand incorporating
188 -maltivec=be semantics where needed.
189 * config/rs6000/rs6000-builtin.def: Add new built-in definitions
190 LVXL_V2DF, LVXL_V2DI, LVXL_V4SF, LVXL_V4SI, LVXL_V8HI, LVXL_V16QI,
191 LVX_V2DF, LVX_V2DI, LVX_V4SF, LVX_V4SI, LVX_V8HI, LVX_V16QI,
192 STVX_V2DF, STVX_V2DI, STVX_V4SF, STVX_V4SI, STVX_V8HI, STVX_V16QI,
193 STVXL_V2DF, STVXL_V2DI, STVXL_V4SF, STVXL_V4SI, STVXL_V8HI,
195 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Replace
196 ALTIVEC_BUILTIN_LVX with ALTIVEC_BUILTIN_LVX_<MODE> throughout;
197 similarly for ALTIVEC_BUILTIN_LVXL, ALTIVEC_BUILTIN_STVX, and
198 ALTIVEC_BUILTIN_STVXL.
199 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be): New
201 (altivec_expand_stvx_be): Likewise.
202 * config/rs6000/rs6000.c (swap_selector_for_mode): New function.
203 (altivec_expand_lvx_be): Likewise.
204 (altivec_expand_stvx_be): Likewise.
205 (altivec_expand_builtin): Add cases for
206 ALTIVEC_BUILTIN_STVX_<MODE>, ALTIVEC_BUILTIN_STVXL_<MODE>,
207 ALTIVEC_BUILTIN_LVXL_<MODE>, and ALTIVEC_BUILTIN_LVX_<MODE>.
208 (altivec_init_builtins): Add definitions for
209 __builtin_altivec_lvxl_<mode>, __builtin_altivec_lvx_<mode>,
210 __builtin_altivec_stvx_<mode>, and
211 __builtin_altivec_stvxl_<mode>.
213 2014-02-21 Catherine Moore <clm@codesourcery.com>
215 * doc/invoke.texi (mvirt, mno-virt): Document.
216 * config/mips/mips.opt (mvirt): New option.
217 * config/mips/mips.h (ASM_SPEC): Pass mvirt to the assembler.
219 2014-02-21 Richard Biener <rguenther@suse.de>
221 PR tree-optimization/60276
222 * tree-vectorizer.h (struct _stmt_vec_info): Add min_neg_dist field.
223 (STMT_VINFO_MIN_NEG_DIST): New macro.
224 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Record
225 STMT_VINFO_MIN_NEG_DIST.
226 * tree-vect-stmts.c (vectorizable_load): Verify if assumptions
227 made for negative dependence distances still hold.
229 2014-02-21 Richard Biener <rguenther@suse.de>
232 * tree-ssa-live.c (mark_all_vars_used_1): Do not walk
233 DECL_INITIAL for globals not in the current function context.
235 2014-02-21 Jakub Jelinek <jakub@redhat.com>
237 PR tree-optimization/56490
238 * params.def (PARAM_UNINIT_CONTROL_DEP_ATTEMPTS): New param.
239 * tree-ssa-uninit.c: Include params.h.
240 (compute_control_dep_chain): Add num_calls argument, return false
241 if it exceed PARAM_UNINIT_CONTROL_DEP_ATTEMPTS param, pass
242 num_calls to recursive call.
243 (find_predicates): Change dep_chain into normal array,
244 cur_chain into auto_vec<edge, MAX_CHAIN_LEN + 1>, add num_calls
245 variable and adjust compute_control_dep_chain caller.
246 (find_def_preds): Likewise.
248 2014-02-21 Thomas Schwinge <thomas@codesourcery.com>
250 * gimple-pretty-print.c (dump_gimple_omp_for) [flags & TDF_RAW]
251 <case GF_OMP_FOR_KIND_CILKSIMD>: Add missing break statement.
253 2014-02-21 Nick Clifton <nickc@redhat.com>
255 * config/stormy16/stormy16.md (pushdqi1): Add mode to post_inc.
257 (popqi1): Add mode to pre_dec.
260 2014-02-21 Jakub Jelinek <jakub@redhat.com>
262 * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
263 mode for mask of V8SFmode permutation.
265 2014-02-20 Richard Henderson <rth@redhat.com>
268 * builtins.c (expand_builtin_atomic_compare_exchange): Always make
269 a new pseudo for OLDVAL.
271 2014-02-20 Jakub Jelinek <jakub@redhat.com>
274 * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
275 gen_reg_rtx if d->testing_p.
276 (expand_vec_perm_pshufb2, expand_vec_perm_broadcast_1): Return early
277 if d->testing_p and we will certainly return true.
278 (expand_vec_perm_even_odd_1): Likewise. Don't call gen_reg_rtx
281 2014-02-20 Uros Bizjak <ubizjak@gmail.com>
283 * emit-rtl.c (gen_reg_rtx): Assert that
284 crtl->emit.regno_pointer_align_length is non-zero.
286 2014-02-20 Richard Henderson <rth@redhat.com>
289 * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
290 on failure the store back into EXPECT.
292 2014-02-20 Chung-Lin Tang <cltang@codesourcery.com>
293 Sandra Loosemore <sandra@codesourcery.com>
295 * config/nios2/nios2.md (unspec): Add UNSPEC_PIC_GOTOFF_SYM enum.
296 * config/nios2/nios2.c (nios2_function_profiler):
297 Add -fPIC (flag_pic == 2) support.
298 (nios2_handle_custom_fpu_cfg): Fix warning parameter.
299 (nios2_large_offset_p): New function.
300 (nios2_unspec_reloc_p): Move up position, update to use
301 nios2_large_offset_p.
302 (nios2_unspec_address): Remove function.
303 (nios2_unspec_offset): New function.
304 (nios2_large_got_address): New function.
305 (nios2_got_address): Add large offset support.
306 (nios2_legitimize_tls_address): Update usage of removed and new
308 (nios2_symbol_binds_local_p): New function.
309 (nios2_load_pic_address): Add -fPIC (flag_pic == 2) support.
310 (nios2_legitimize_address): Update to use nios2_large_offset_p.
311 (nios2_emit_move_sequence): Avoid legitimizing (const (unspec ...)).
312 (nios2_print_operand): Merge H/L processing, add hiadj/lo
313 processing for (const (unspec ...)).
314 (nios2_unspec_reloc_name): Add UNSPEC_PIC_GOTOFF_SYM case.
316 2014-02-20 Richard Biener <rguenther@suse.de>
318 * tree-cfg.c (replace_uses_by): Mark altered BBs before
319 doing the substitution.
320 (verify_gimple_assign_single): Also verify bare MEM_REFs on the lhs.
322 2014-02-20 Martin Jambor <mjambor@suse.cz>
325 * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
326 info when checking whether lattices are bottom.
328 2014-02-20 Richard Biener <rguenther@suse.de>
331 * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
334 2014-02-20 Jan Hubicka <hubicka@ucw.cz>
337 * ipa-inline-transform.c (clone_inlined_nodes): Add freq_scale
338 parameter specifying the scaling.
339 (inline_call): Update.
340 (want_inline_recursively): Guard division by zero.
341 (recursive_inlining): Update.
342 * ipa-inline.h (clone_inlined_nodes): Update.
344 2014-02-20 Ilya Tocar <ilya.tocar@intel.com>
347 * config/i386/i386.c (classify_argument): Pass structures of size
348 64 bytes or less in register.
350 2014-02-20 Ilya Tocar <ilya.tocar@intel.com>
351 Kirill Yukhin <kirill.yukhin@intel.com>
353 * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): Swap operands.
354 (_mm_rcp28_round_ss): Ditto.
355 (_mm_rsqrt28_round_sd): Ditto.
356 (_mm_rsqrt28_round_ss): Ditto.
357 * config/i386/avx512erintrin.h (_mm_rcp14_round_sd): Ditto.
358 (_mm_rcp14_round_ss): Ditto.
359 (_mm_rsqrt14_round_sd): Ditto.
360 (_mm_rsqrt14_round_ss): Ditto.
361 * config/i386/sse.md (rsqrt14<mode>): Put nonimmediate operand as
362 the first input operand, get rid of match_dup.
363 (avx512er_exp2<mode><mask_name><round_saeonly_name>): Set type
365 (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
367 (avx512er_vmrcp28<mode><round_saeonly_name>): Put nonimmediate
368 operand as the first input operand, set type attribute.
369 (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
371 (avx512er_vmrsqrt28<mode><round_saeonly_name>): Put nonimmediate
372 operand as the first input operand, set type attribute.
374 2014-02-19 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
376 * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
379 2014-02-19 H.J. Lu <hongjiu.lu@intel.com>
382 * config/i386/i386.c (construct_container): Remove TFmode check
383 for X86_64_INTEGER_CLASS.
385 2014-02-19 Uros Bizjak <ubizjak@gmail.com>
388 * config/i386/i386.c (type_natural_mode): Warn for ABI changes
389 only when -Wpsabi is enabled.
391 2014-02-19 Michael Hudson-Doyle <michael.hudson@linaro.org>
394 * config/aarch64/aarch64.c (aarch64_pass_by_reference): The rules for
395 passing arrays in registers are the same as for structs, so remove the
396 special case for them.
398 2014-02-19 Eric Botcazou <ebotcazou@adacore.com>
400 * expr.c (expand_expr_real_1) <case VIEW_CONVERT_EXPR>: For a bit-field
401 destination type, extract only the valid bits if the source type is not
402 integral and has a different mode.
404 2014-02-19 Richard Biener <rguenther@suse.de>
407 * tree-inline.c (estimate_num_insns): Avoid calling cgraph_get_node
410 2014-02-19 Richard Biener <rguenther@suse.de>
413 * ipa-prop.c: Include stringpool.h and tree-ssanames.h.
414 (ipa_modify_call_arguments): Emit an argument load explicitely and
415 preserve virtual SSA form there and for the replacement call.
416 Do not update SSA form nor free dominance info.
418 2014-02-18 Jan Hubicka <hubicka@ucw.cz>
420 * ipa.c (function_and_variable_visibility): Also clear WEAK
421 flag when disolving COMDAT_GROUP.
423 2014-02-18 Jan Hubicka <hubicka@ucw.cz>
425 * ipa-prop.h (ipa_ancestor_jf_data): Update ocmment.
426 * ipa-prop.c (ipa_set_jf_known_type): Return early when
428 (ipa_set_ancestor_jf): Set type to NULL hwen it is not preserved;
429 do more sanity checks.
430 (detect_type_change): Return true when giving up early.
431 (compute_complex_assign_jump_func): Fix type parameter of
433 (compute_complex_ancestor_jump_func): Likewise.
434 (update_jump_functions_after_inlining): Fix updating of
436 * ipa-cp.c (ipa_get_jf_ancestor_result): Be ready for type to be NULL.
438 2014-02-18 Jan Hubicka <hubicka@ucw.cz>
440 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
441 inline clones when edge disappears.
443 2014-02-18 Michael Meissner <meissner@linux.vnet.ibm.com>
446 * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
447 Split 64-bit moves into 2 patterns. Do not allow the use of
448 direct move for TDmode in little endian, since the decimal value
449 has little endian bytes within a word, but the 64-bit pieces are
450 ordered in a big endian fashion, and normal subreg's of TDmode are
452 (mov<mode>_64bit_dm): Likewise.
453 (movtd_64bit_nodm): Likewise.
455 2014-02-18 Eric Botcazou <ebotcazou@adacore.com>
457 PR tree-optimization/60174
458 * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
459 statement of an SSA_NAME that occurs in an abnormal PHI node.
461 2014-02-18 Jakub Jelinek <jakub@redhat.com>
464 * final.c (SEEN_BB): Remove.
465 (SEEN_NOTE, SEEN_EMITTED): Renumber.
466 (final_scan_insn): Don't force_source_line on second
467 NOTE_INSN_BASIC_BLOCK.
469 2014-02-18 Uros Bizjak <ubizjak@gmail.com>
472 * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
473 * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
474 (type_natural_mode): Warn ABI change when %zmm register is not
475 available for AVX512F vector value passing.
477 2014-02-18 Kai Tietz <ktietz@redhat.com>
480 * config/i386/i386.c (ix86_expand_prologue): Use value in
481 rax register as displacement when restoring %r10 or %rax.
482 Fix wrong offset when restoring both registers.
484 2014-02-18 Eric Botcazou <ebotcazou@adacore.com>
486 * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
487 assertion with conditional return.
489 2014-02-18 Jakub Jelinek <jakub@redhat.com>
490 Uros Bizjak <ubizjak@gmail.com>
493 * config/i386/driver-i386.c (host_detect_local_cpu): If
494 YMM state is not saved by the OS, also clear has_f16c. Move
495 CPUID 0x80000001 handling before YMM state saving checking.
497 2014-02-18 Andrey Belevantsev <abel@ispras.ru>
499 PR rtl-optimization/58960
500 * haifa-sched.c (alloc_global_sched_pressure_data): New,
501 factored out from ...
502 (sched_init): ... here.
503 (free_global_sched_pressure_data): New, factored out from ...
504 (sched_finish): ... here.
505 * sched-int.h (free_global_sched_pressure_data): Declare.
506 * sched-rgn.c (nr_regions_initial): New static global.
507 (haifa_find_rgns): Initialize it.
508 (schedule_region): Disable sched-pressure for the newly
511 2014-02-17 Richard Biener <rguenther@suse.de>
513 * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
514 release SSA defs of pattern stmts.
516 2014-02-17 Richard Biener <rguenther@suse.de>
518 * tree-inline.c (expand_call_inline): Release the virtual
519 operand defined by the call we are about to inline.
521 2014-02-17 Richard Biener <rguenther@suse.de>
523 * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
525 2014-02-17 Kirill Yukhin <kirill.yukhin@intel.com>
526 Ilya Tocar <ilya.tocar@intel.com>
528 * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
529 arguments order in builtin.
530 (_mm512_permutexvar_epi64): Ditto.
531 (_mm512_mask_permutexvar_epi64): Ditto
532 (_mm512_maskz_permutexvar_epi32): Ditto
533 (_mm512_permutexvar_epi32): Ditto
534 (_mm512_mask_permutexvar_epi32): Ditto
536 2014-02-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
538 * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
539 (p8_vmrgow): Likewise.
541 2014-02-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
543 * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
546 2014-02-15 Michael Meissner <meissner@linux.vnet.ibm.com>
549 * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
550 (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
551 into 64-bit and 32-bit moves. On 64-bit moves, add support for
552 using direct move instructions on ISA 2.07. Also adjust
553 instruction length for 64-bit.
554 (mov<mode>_64bit, TFmode/TDmode): Likewise.
555 (mov<mode>_32bit, TFmode/TDmode): Likewise.
557 2014-02-15 Alan Modra <amodra@gmail.com>
561 * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
562 find_replacement on parts of insn rtl that might be reloaded.
564 2014-02-15 Richard Biener <rguenther@suse.de>
566 PR tree-optimization/60183
567 * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
568 (tree_ssa_phiprop): Calculate and free post-dominators.
570 2014-02-14 Jeff Law <law@redhat.com>
572 PR rtl-optimization/60131
573 * ree.c (get_extended_src_reg): New function.
574 (combine_reaching_defs): Use it rather than assuming location of REG.
575 (find_and_remove_re): Verify first operand of extension is
576 a REG before adding the insns to the copy list.
578 2014-02-14 Roland McGrath <mcgrathr@google.com>
580 * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
581 * configure: Regenerated.
582 * config.in: Regenerated.
583 * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
584 instead of ASM_SHORT.
586 2014-02-14 Vladimir Makarov <vmakarov@redhat.com>
587 Richard Earnshaw <rearnsha@arm.com>
589 PR rtl-optimization/59535
590 * lra-constraints.c (process_alt_operands): Encourage alternative
591 when unassigned pseudo class is superset of the alternative class.
592 (inherit_reload_reg): Don't inherit when optimizing for code size.
593 * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
594 returning CORE_REGS for anything but Thumb1 and BASE_REGS for
595 modes not less than 4 for Thumb1.
597 2014-02-14 Kyle McMartin <kyle@redhat.com>
600 * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
602 2014-02-14 Richard Biener <rguenther@suse.de>
604 * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
605 (get_frame_arg): Drop the assert with langhook types_compatible_p.
606 Do not strip INDIRECT_REFs.
608 2014-02-14 Richard Biener <rguenther@suse.de>
611 * lto-streamer-out.c (DFS_write_tree_body): Do not follow
612 DECL_FUNCTION_SPECIFIC_TARGET.
613 (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
614 * tree-streamer-out.c (pack_ts_target_option): Remove.
615 (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
616 (write_ts_function_decl_tree_pointers): Do not stream
617 DECL_FUNCTION_SPECIFIC_TARGET.
618 * tree-streamer-in.c (unpack_ts_target_option): Remove.
619 (unpack_value_fields): Do not stream TS_TARGET_OPTION.
620 (lto_input_ts_function_decl_tree_pointers): Do not stream
621 DECL_FUNCTION_SPECIFIC_TARGET.
623 2014-02-14 Jakub Jelinek <jakub@redhat.com>
625 * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
626 (get_initial_def_for_induction, vectorizable_induction): Ignore
627 debug stmts when looking for exit_phi.
628 (vectorizable_live_operation): Fix up condition.
630 2014-02-14 Chung-Ju Wu <jasonwucj@gmail.com>
632 * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
633 nreverse() because it changes the content of original tree list.
635 2014-02-14 Chung-Ju Wu <jasonwucj@gmail.com>
637 * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
638 * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
640 2014-02-14 Chung-Ju Wu <jasonwucj@gmail.com>
642 * config/nds32/nds32.c (nds32_naked_function_p): Follow the
643 GNU coding standards.
645 2014-02-13 Jakub Jelinek <jakub@redhat.com>
648 * dwarf2out.c (gen_subprogram_die): Don't call
649 add_calling_convention_attribute if subr_die is old_die.
651 2014-02-13 Sharad Singhai <singhai@google.com>
653 * doc/optinfo.texi: Fix order of nodes.
655 2014-02-13 Uros Bizjak <ubizjak@gmail.com>
657 * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
658 operands[2], not operands[3].
660 2014-02-13 Richard Biener <rguenther@suse.de>
663 * doc/install.texi (ISL): Update recommended version to 0.12.2,
664 mention the possibility of an in-tree build.
665 (CLooG): Update recommended version to 0.18.1, mention the
666 possibility of an in-tree build and clarify that the ISL
667 bundled with CLooG does not work.
669 2014-02-13 Jakub Jelinek <jakub@redhat.com>
672 * expr.c (compress_float_constant): If x is a hard register,
673 extend into a pseudo and then move to x.
675 2014-02-13 Dominik Vogt <vogt@linux.vnet.ibm.com>
677 * config/s390/s390.c (s390_asm_output_function_label): Fix crash
678 caused by bad second argument to warning_at() with -mhotpatch and
679 nested functions (e.g. with gfortran).
681 2014-02-13 Richard Sandiford <rdsandiford@googlemail.com>
683 * opts.c (option_name): Remove "enabled by default" rider.
685 2014-02-12 John David Anglin <danglin@gcc.gnu.org>
687 * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
689 2014-02-12 H.J. Lu <hongjiu.lu@intel.com>
690 Uros Bizjak <ubizjak@gmail.com>
693 * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
694 * configure: Regenerated.
696 2014-02-12 Richard Biener <rguenther@suse.de>
698 * vec.c (vec_prefix::calculate_allocation): Move as
699 inline variant to vec.h.
700 (vec_prefix::calculate_allocation_1): New out-of-line version.
701 * vec.h (vec_prefix::calculate_allocation_1): Declare.
702 (vec_prefix::m_has_auto_buf): Rename to ...
703 (vec_prefix::m_using_auto_storage): ... this.
704 (vec_prefix::calculate_allocation): Inline the easy cases
705 and dispatch to calculate_allocation_1 which doesn't need the
707 (va_heap::reserve): Use gcc_checking_assert.
708 (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
709 m_using_auto_storage.
710 (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
712 (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
713 (vec<T, va_heap, vl_ptr>::release): Avoid casting.
714 (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
716 2014-02-12 Richard Biener <rguenther@suse.de>
718 * gcse.c (compute_transp): break from loop over canon_modify_mem_list
719 when we found a dependence.
721 2014-02-12 Thomas Schwinge <thomas@codesourcery.com>
723 * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
725 (maybe_fold_stmt): ... into this new function.
726 * omp-low.c (lower_omp): Update comment.
728 * omp-low.c (lower_omp_target): Add clobber for sizes array, after
731 * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
734 2014-02-12 James Greenhalgh <james.greenhalgh@arm.com>
736 * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
737 identifiers in comments.
738 (cortexa53_extra_costs): Likewise.
739 * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
740 (cortexa7_extra_costs): Likewise.
741 (cortexa12_extra_costs): Likewise.
742 (cortexa15_extra_costs): Likewise.
743 (v7m_extra_costs): Likewise.
745 2014-02-12 Richard Biener <rguenther@suse.de>
748 * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
749 of posix_memalign being successful.
750 (lower_stmt): Restrict lowering of posix_memalign to when
751 -ftree-bit-ccp is enabled.
753 2014-02-12 Senthil Kumar Selvaraj <senthil_kumar.selvaraj@atmel.com>
755 * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
757 * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
759 2014-02-12 Eric Botcazou <ebotcazou@adacore.com>
761 PR rtl-optimization/60116
762 * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
763 other_insn once the combination has been validated.
765 2014-02-11 Jan Hubicka <hubicka@ucw.cz>
768 * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
770 * ipa-devirt.c: Include demangle.h
771 (odr_violation_reported): New static variable.
772 (add_type_duplicate): Update odr_violations.
773 (maybe_record_node): Add completep parameter; update it.
774 (record_target_from_binfo): Add COMPLETEP parameter;
776 (possible_polymorphic_call_targets_1): Likewise.
777 (struct polymorphic_call_target_d): Add nonconstruction_targets;
778 rename FINAL to COMPLETE.
779 (record_targets_from_bases): Sanity check we found the binfo;
780 fix COMPLETEP updating.
781 (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
782 parameter, fix computing of COMPLETEP.
783 (dump_possible_polymorphic_call_targets): Imrove readability of dump;
784 at LTO time do demangling.
785 (ipa_devirt): Use nonconstruction_targets; Improve dumps.
786 * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
788 (gimple_get_virt_method_for_binfo): Likewise.
789 * gimple-fold.h (gimple_get_virt_method_for_binfo,
790 gimple_get_virt_method_for_vtable): Update prototypes.
792 2014-02-11 Vladimir Makarov <vmakarov@redhat.com>
795 * genautomata.c (add_presence_absence): Fix typo with
796 {final_}presence_list.
798 2014-02-11 Michael Meissner <meissner@linux.vnet.ibm.com>
801 * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
802 for VSX/Altivec vectors that land in GPR registers.
804 2014-02-11 Richard Henderson <rth@redhat.com>
805 Jakub Jelinek <jakub@redhat.com>
808 * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
809 around drhs if type conversion to lacc->type is not useless.
811 2014-02-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
813 * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
815 (cortex-a57.cortex-a53): Likewise.
816 * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
818 2014-02-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
820 * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
823 2014-02-11 Renlin Li <Renlin.Li@arm.com>
825 * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
826 add_options_for_arm_vfp3.
828 2014-02-11 Jeff Law <law@redhat.com>
831 * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
832 object with an undesirable mode.
834 2014-02-11 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
837 * config/i386/sol2-9.h: New file.
838 * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
839 *-*-solaris2.9*): Use it.
841 2014-02-10 Nagaraju Mekala <nagaraju.mekala@xilinx.com>
843 * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
844 * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
846 2014-02-10 Nagaraju Mekala <nagaraju.mekala@xilinx.com>
848 * config/microblaze/microblaze.c: Extend mcpu version format
850 2014-02-10 David Holsgrove <david.holsgrove@xilinx.com>
852 * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
854 2014-02-10 Richard Henderson <rth@redhat.com>
857 * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
858 * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
859 ms-abi vs -mno-accumulate-outgoing-args.
860 (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
861 * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
864 2014-02-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
867 * cfgexpand.c (expand_asm_operands): Attach source location to
868 ASM_INPUT rtx objects.
869 * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
871 2014-02-10 Nick Clifton <nickc@redhat.com>
873 * config/mn10300/mn10300.c (popcount): New function.
874 (mn10300_expand_prologue): Include saved registers in stack usage
877 2014-02-10 Jeff Law <law@redhat.com>
880 * reload1.c (emit_input_reload_insns): Do not create invalid RTL
881 when changing the SET_DEST of a prior insn to avoid an input reload.
883 2014-02-10 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
885 * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
886 big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
887 -mcall-openbsd, or -mcall-linux.
888 (CC1_ENDIAN_BIG_SPEC): Remove.
889 (CC1_ENDIAN_LITTLE_SPEC): Remove.
890 (CC1_ENDIAN_DEFAULT_SPEC): Remove.
891 (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
892 (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
893 and %cc1_endian_default.
894 * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
896 2014-02-10 Richard Biener <rguenther@suse.de>
898 PR tree-optimization/60115
899 * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
900 MEM_REF handling. Properly verify that the accesses are not
901 out of the objects bound.
903 2014-02-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
905 * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
908 2014-02-10 Eric Botcazou <ebotcazou@adacore.com>
910 * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
911 proper constants and fix formatting.
912 (possible_polymorphic_call_targets): Fix formatting.
914 2014-02-10 Kirill Yukhin <kirill.yukhin@intel.com>
915 Ilya Tocar <ilya.tocar@intel.com>
917 * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
918 (_mm512_loadu_epi32): Renamed into...
919 (_mm512_loadu_si512): This.
920 (_mm512_storeu_epi32): Renamed into...
921 (_mm512_storeu_si512): This.
922 (_mm512_maskz_ceil_ps): Removed.
923 (_mm512_maskz_ceil_pd): Ditto.
924 (_mm512_maskz_floor_ps): Ditto.
925 (_mm512_maskz_floor_pd): Ditto.
926 (_mm512_floor_round_ps): Ditto.
927 (_mm512_floor_round_pd): Ditto.
928 (_mm512_ceil_round_ps): Ditto.
929 (_mm512_ceil_round_pd): Ditto.
930 (_mm512_mask_floor_round_ps): Ditto.
931 (_mm512_mask_floor_round_pd): Ditto.
932 (_mm512_mask_ceil_round_ps): Ditto.
933 (_mm512_mask_ceil_round_pd): Ditto.
934 (_mm512_maskz_floor_round_ps): Ditto.
935 (_mm512_maskz_floor_round_pd): Ditto.
936 (_mm512_maskz_ceil_round_ps): Ditto.
937 (_mm512_maskz_ceil_round_pd): Ditto.
938 (_mm512_expand_pd): Ditto.
939 (_mm512_expand_ps): Ditto.
940 * config/i386/i386.c (ix86_builtins): Remove
941 IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
943 * config/i386/predicates.md (const1256_operand): New.
944 (const_1_to_2_operand): Ditto.
945 * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
946 (*avx512pf_gatherpf<mode>sf_mask): Ditto.
947 (*avx512pf_gatherpf<mode>sf): Ditto.
948 (avx512pf_gatherpf<mode>df): Ditto.
949 (*avx512pf_gatherpf<mode>df_mask): Ditto.
950 (*avx512pf_gatherpf<mode>df): Ditto.
951 (avx512pf_scatterpf<mode>sf): Ditto.
952 (*avx512pf_scatterpf<mode>sf_mask): Ditto.
953 (*avx512pf_scatterpf<mode>sf): Ditto.
954 (avx512pf_scatterpf<mode>df): Ditto.
955 (*avx512pf_scatterpf<mode>df_mask): Ditto.
956 (*avx512pf_scatterpf<mode>df): Ditto.
957 (avx512f_expand<mode>): Removed.
958 (<shift_insn><mode>3<mask_name>): Change predicate type.
960 2014-02-08 Jakub Jelinek <jakub@redhat.com>
962 * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
963 not at the end of datarefs vector use ordered_remove to avoid
964 reordering datarefs vector.
967 * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
968 mark local addressable non-static vars as GOVD_PRIVATE
969 instead of GOVD_LOCAL.
970 * omp-low.c (lower_omp_for): Move gimple_bind_vars
971 and BLOCK_VARS of gimple_bind_block to new_stmt rather
975 * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
976 if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
977 assume_aligned or alloc_align attributes.
978 (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
979 arguments. Handle also assume_aligned and alloc_align attributes.
980 (evaluate_stmt): Adjust bit_value_assume_aligned caller. Handle
981 calls to functions with assume_aligned or alloc_align attributes.
982 * doc/extend.texi: Document assume_aligned and alloc_align attributes.
984 2014-02-08 Terry Guo <terry.guo@arm.com>
986 * doc/invoke.texi: Document ARM -march=armv7e-m.
988 2014-02-08 Jakub Jelinek <jakub@redhat.com>
990 * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
991 flag on __cilkrts_rethrow builtin.
994 * ipa-cp.c (determine_versionability): Fail at -O0
995 or __attribute__((optimize (0))) or -fno-ipa-cp functions.
996 * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
999 2014-02-04 Jakub Jelinek <jakub@redhat.com>
1002 * tree-inline.c (copy_forbidden): Fail for
1003 __attribute__((optimize (0))) functions.
1005 2014-02-07 Jan Hubicka <hubicka@ucw.cz>
1007 * varpool.c: Include pointer-set.h.
1008 (varpool_remove_unreferenced_decls): Variables in other partitions
1009 will not be output; be however careful to not lose information
1012 2014-02-07 Jan Hubicka <hubicka@ucw.cz>
1014 * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
1015 lookup in the vtable constructor.
1017 2014-02-07 Jeff Law <law@redhat.com>
1020 * config/m68k/m68k.md (ashldi_extsi): Turn into a
1021 define_insn_and_split.
1023 * ipa-inline.c (inline_small_functions): Fix typos.
1025 2014-02-07 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
1027 * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
1028 (s390_can_use_return_insn): Declare.
1029 * config/s390/s390.h (EPILOGUE_USES): Define.
1030 * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
1032 (s390_chunkify_start): Handle return JUMP_LABELs.
1033 (s390_early_mach): Emit a main_pool instruction on the entry edge.
1034 (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
1035 (s390_can_use_return_insn): New functions.
1036 (s390_fix_long_loop_prediction): Handle conditional returns.
1037 (TARGET_SET_UP_BY_PROLOGUE): Define.
1038 * config/s390/s390.md (ANY_RETURN): New code iterator.
1039 (*creturn, *csimple_return, return, simple_return): New patterns.
1041 2014-02-07 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
1043 * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
1044 notes to each restore. Also add REG_CFA_DEF_CFA when restoring %r15.
1045 (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P. Update the
1046 REG_CFA_RESTORE list when deciding not to restore a register.
1048 2014-02-07 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
1050 * config/s390/s390.c: Include tree-pass.h and context.h.
1051 (s390_early_mach): New function, split out from...
1052 (s390_emit_prologue): ...here.
1053 (pass_data_s390_early_mach): New pass structure.
1054 (pass_s390_early_mach): New class.
1055 (s390_option_override): Create and register early_mach pass.
1056 Move to end of file.
1058 2014-02-07 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
1060 * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
1061 to match for the exit block.
1063 2014-02-07 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
1065 * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
1066 ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
1067 Reject misaligned operands.
1069 2014-02-07 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
1071 * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
1073 2014-02-07 Richard Biener <rguenther@suse.de>
1076 * gimple-low.c (lower_builtin_posix_memalign): New function.
1077 (lower_stmt): Call it to lower posix_memalign in a way
1078 to make alignment info accessible.
1080 2014-02-07 Jakub Jelinek <jakub@redhat.com>
1083 * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
1084 __builtin_setjmp_receiver.
1086 2014-02-07 Richard Biener <rguenther@suse.de>
1089 * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
1090 * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
1091 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
1092 Handle BUILT_IN_POSIX_MEMALIGN.
1093 (find_func_clobbers): Likewise.
1094 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
1095 (call_may_clobber_ref_p_1): Likewise.
1097 2014-02-06 Jan Hubicka <hubicka@ucw.cz>
1100 * ipa-devirt.c (record_target_from_binfo): Remove overactive
1103 2014-02-06 Jan Hubicka <hubicka@ucw.cz>
1106 * lto-cgraph.c (lto_output_node): Use
1107 symtab_get_symbol_partitioning_class.
1108 (lto_output_varpool_node): likewise.
1109 (symtab_get_symbol_partitioning_class): Move here from
1111 * cgraph.h (symbol_partitioning_class): Likewise.
1112 (symtab_get_symbol_partitioning_class): Declare.
1114 2014-02-06 Jan Hubicka <hubicka@ucw.cz>
1116 * ggc.h (ggc_internal_cleared_alloc): New macro.
1117 * vec.h (vec_safe_copy): Handle memory stats.
1118 * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
1119 * target-globals.c (save_target_globals): Likewise.
1121 2014-02-06 Jan Hubicka <hubicka@ucw.cz>
1124 * expr.c (emit_move_resolve_push): Export; be bit more selective
1125 on when to clear alias set.
1126 * expr.h (emit_move_resolve_push): Declare.
1127 * function.h (struct function): Add tail_call_marked.
1128 * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
1129 * config/i386/i386-protos.h (ix86_expand_push): Remove.
1130 * config/i386/i386.md (TImode move expander): De not call
1132 (FP push expanders): Preserve memory attributes.
1133 * config/i386/sse.md (push<mode>1): Remove.
1134 * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
1135 (ix86_expand_push): Remove.
1136 * config/i386/mmx.md (push<mode>1): Remove.
1138 2014-02-06 Jakub Jelinek <jakub@redhat.com>
1140 PR rtl-optimization/60030
1141 * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
1142 lopart with paradoxical subreg before shifting it up by hprec.
1144 2014-02-06 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1146 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
1147 Remove extra newline at end of file.
1148 * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
1149 (arm_issue_rate): Handle cortexa57.
1150 * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
1151 (cortex-a57.cortex-a53): Likewise.
1153 2014-02-06 Jakub Jelinek <jakub@redhat.com>
1156 * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
1157 don't record in REG_FRAME_RELATED_EXPR registers not set in that
1159 (arm_expand_prologue): Adjust all callers.
1160 (arm_unwind_emit_sequence): Allow saved, but not important for unwind
1161 info, registers also at the lowest numbered registers side. Use
1162 gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
1166 * var-tracking.c (adjust_mems): Before adding a SET to
1167 amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
1169 2014-02-06 Alan Modra <amodra@gmail.com>
1172 * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
1173 change SDmode to DDmode when lra_in_progress.
1175 2014-02-06 Jakub Jelinek <jakub@redhat.com>
1178 * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
1179 free_data_ref on the dr first, and before goto again also set dr
1180 to the next dr. For simd_lane_access, free old datarefs[i] before
1181 overwriting it. For get_vectype_for_scalar_type failure, don't
1182 free_data_ref if simd_lane_access.
1184 * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
1187 * tree.h (opts_for_fn): New inline function.
1188 (opt_for_fn): Define.
1189 * config/i386/i386.c (ix86_function_regparm): Use
1190 opt_for_fn (decl, optimize) instead of optimize.
1192 2014-02-06 Marcus Shawcroft <marcus.shawcroft@arm.com>
1194 * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
1195 for SYMBOL_REF in large memory model.
1197 2014-02-06 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1199 * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
1201 (cortex-a57): Likewise.
1202 (cortex-a57.cortex-a53): Likewise.
1204 2014-02-06 Yury Gribov <y.gribov@samsung.com>
1205 Kugan Vivekanandarajah <kuganv@linaro.org>
1207 * config/arm/arm.c (arm_vector_alignment_reachable): Check
1209 * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
1211 2014-02-06 Richard Biener <rguenther@suse.de>
1213 * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
1214 set_loop_copy and initialize_original_copy_tables.
1216 2014-02-06 Alex Velenko <Alex.Velenko@arm.com>
1218 * config/aarch64/aarch64-simd.md
1219 (aarch64_ashr_simddi): Change QI to SI.
1221 2014-02-05 Jan Hubicka <hubicka@ucw.cz>
1222 Jakub Jelinek <jakub@redhat.com>
1225 * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
1228 2014-02-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1230 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
1231 CODE_FOR_altivec_vpku[hw]um to
1232 CODE_FOR_altivec_vpku[hw]um_direct.
1233 * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
1234 UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
1235 (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
1236 UNSPEC_VUNPACK_LO_SIGN_DIRECT.
1238 2014-02-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1240 * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
1241 generation for -maltivec=be.
1242 (altivec_vsumsws): Simplify redundant test.
1244 2014-02-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1246 * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
1247 (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
1248 (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
1249 (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
1250 gen_altivec_vpkuwum.
1251 (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
1253 (altivec_vpks<VI_char>ss): Likewise.
1254 (altivec_vpks<VI_char>us): Likewise.
1255 (altivec_vpku<VI_char>us): Likewise.
1256 (altivec_vpku<VI_char>um): Likewise.
1257 (altivec_vpku<VI_char>um_direct): New (copy of
1258 altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
1260 (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
1261 target is little endian and -maltivec=be is not specified.
1262 (*altivec_vupkhs<VU_char>_direct): New (copy of
1263 altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
1264 (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
1265 target is little endian and -maltivec=be is not specified.
1266 (*altivec_vupkls<VU_char>_direct): New (copy of
1267 altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
1268 (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
1269 little endian and -maltivec=be is not specified.
1270 (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
1271 little endian and -maltivec=be is not specified.
1273 2014-02-05 Richard Henderson <rth@redhat.com>
1276 * combine-stack-adj.c: Revert r206943.
1277 * sched-int.h (struct deps_desc): Add last_args_size.
1278 * sched-deps.c (init_deps): Initialize it.
1279 (sched_analyze_insn): Add OUTPUT dependencies between insns that
1280 contain REG_ARGS_SIZE notes.
1282 2014-02-05 Jan Hubicka <hubicka@ucw.cz>
1284 * lto-cgraph.c (asm_nodes_output): Make global.
1285 * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
1286 * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
1287 (driver_handle_option): Handle OPT_fwpa.
1289 2014-02-05 Jakub Jelinek <jakub@redhat.com>
1292 * ipa-devirt.c (possible_polymorphic_call_targets): Fix
1293 a comment typo and formatting issue. If odr_hash hasn't been
1294 created, return vNULL and set *completep to false.
1297 * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
1298 bb with no successors.
1300 2014-02-05 James Greenhalgh <james.greenhalgh@arm.com>
1303 * doc/invoke.texi (-march): Clarify documentation for ARM.
1307 2014-02-05 Richard Biener <rguenther@suse.de>
1309 * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
1310 when not vectorizing because of too many alias checks.
1311 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
1312 Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
1314 2014-02-05 Nick Clifton <nickc@redhat.com>
1316 * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
1317 accept extended registers in any mode when compiling for the MN10300.
1319 2014-02-05 Yury Gribov <y.gribov@samsung.com>
1321 * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
1322 * ipa-inline.c (report_inline_failed_reason): Handle mismatched
1323 sanitization attributes.
1324 (can_inline_edge_p): Likewise.
1325 (sanitize_attrs_match_for_inline_p): New function.
1327 2014-02-04 Jan Hubicka <hubicka@ucw.cz>
1329 * ipa-prop.c (detect_type_change): Shor circuit testing of
1330 type changes on THIS pointer.
1332 2014-02-04 John David Anglin <danglin@gcc.gnu.org>
1335 * config/pa/pa.c (legitimize_tls_address): Return original address
1336 if not passed a SYMBOL_REF rtx.
1337 (hppa_legitimize_address): Call legitimize_tls_address for all TLS
1339 (pa_emit_move_sequence): Simplify TLS source operands.
1340 (pa_legitimate_constant_p): Reject all TLS constants.
1341 * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
1342 (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
1344 2014-02-04 Jan Hubicka <hubicka@ucw.cz>
1346 * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
1347 groups when we know they are controlled by LTO.
1348 * varasm.c (default_binds_local_p_1): If object is in other partition,
1349 it will be resolved locally.
1351 2014-02-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
1353 * config/host-linux.c (linux_gt_pch_use_address): Don't
1354 use SSIZE_MAX because it is not always defined.
1356 2014-02-04 Vladimir Makarov <vmakarov@redhat.com>
1359 * lra-constraints.c (need_for_split_p): Use more 3 reloads as
1360 threshold for pseudo splitting.
1361 (update_ebb_live_info): Process call argument hard registers and
1362 hard registers from insn definition too.
1363 (max_small_class_regs_num): New constant.
1364 (inherit_in_ebb): Update live hard regs through EBBs. Update
1365 reloads_num only for small register classes. Don't split for
1368 2014-02-04 Markus Trippelsdorf <markus@trippelsdorf.de>
1371 * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
1374 2014-02-04 Jan Hubicka <hubicka@ucw.cz>
1376 * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
1379 2014-02-04 Marek Polacek <polacek@redhat.com>
1381 * gdbinit.in (pel): Define.
1383 2014-02-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
1385 * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
1388 2014-02-04 Richard Biener <rguenther@suse.de>
1391 * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
1392 in function context local.
1393 (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
1394 * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
1395 similar to LTO_imported_decl_ref.
1397 2014-02-04 Jakub Jelinek <jakub@redhat.com>
1399 PR tree-optimization/60002
1400 * cgraphclones.c (build_function_decl_skip_args): Clear
1403 PR tree-optimization/60023
1404 * tree-if-conv.c (predicate_mem_writes): Pass true instead of
1405 false to gsi_replace.
1406 * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
1407 has been in some EH region and vec_stmt could throw, add
1408 vec_stmt into the same EH region.
1409 * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
1410 has no lhs, ignore it.
1411 * internal-fn.c (expand_MASK_LOAD): Likewise.
1414 * tree-inline.c (copy_forbidden): Fail for
1415 __attribute__((optimize (0))) functions.
1418 * omp-low.c (simd_clone_struct_copy): If from->inbranch
1419 is set, copy one less argument.
1420 (expand_simd_clones): Don't subtract clone_info->inbranch
1421 from simd_clone_struct_alloc argument.
1423 PR rtl-optimization/57915
1424 * recog.c (simplify_while_replacing): If all unary/binary/relational
1425 operation arguments are constant, attempt to simplify those.
1428 * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
1429 if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
1431 2014-02-04 Richard Biener <rguenther@suse.de>
1433 PR tree-optimization/60012
1434 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
1435 TBAA disambiguation to all DDRs.
1437 2014-02-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1440 * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
1441 (LINK_SPEC): Use it for -shared, -shared-libgcc.
1443 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
1446 * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
1448 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
1450 * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
1451 * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
1453 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
1456 * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
1457 to figure out targets of polymorphic calls with known decl.
1458 * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
1459 * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
1460 * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
1461 (get_polymorphic_call_info): ... here.
1462 (get_polymorphic_call_info_from_invariant): New function.
1464 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
1466 * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
1467 lookup via vtable pointer; check for type consistency
1468 and turn inconsitent facts into UNREACHABLE.
1469 * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
1470 * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
1471 type inconsistent querries; return UNREACHABLE instead.
1473 2014-02-03 Richard Henderson <rth@twiddle.net>
1476 * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
1477 already processed this node.
1478 (normalize_one_pred_1): Pass along mark_set.
1479 (normalize_one_pred): Create and destroy a pointer_set_t.
1480 (normalize_one_pred_chain): Likewise.
1482 2014-02-03 Laurent Aflonsi <laurent.alfonsi@st.com>
1484 PR gcov-profile/58602
1485 * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
1487 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
1490 * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
1491 -fno-devirtualize; try to devirtualize by the knowledge of
1492 virtual table pointer given by aggregate propagation.
1493 * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
1494 (ipa_print_node_jump_functions): Dump also offset that
1495 is relevant for polymorphic calls.
1496 (determine_known_aggregate_parts): Add arg_type parameter; use it
1497 instead of determining the type from pointer type.
1498 (ipa_compute_jump_functions_for_edge): Update call of
1499 determine_known_aggregate_parts.
1500 * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
1501 (gimple_get_virt_method_for_binfo): ... here; simplify using
1502 vtable_pointer_value_to_vtable.
1503 * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
1504 * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
1505 to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
1506 (vtable_pointer_value_to_vtable): Break out from ...; handle also
1508 (vtable_pointer_value_to_binfo): ... here.
1509 * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
1511 2014-02-03 Teresa Johnson <tejohnson@google.com>
1513 * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
1514 redef of outer loop index variable.
1516 2014-02-03 Marc Glisse <marc.glisse@inria.fr>
1520 * doc/extend.texi (Function Attributes): Typo.
1522 2014-02-03 Cong Hou <congh@google.com>
1524 PR tree-optimization/60000
1525 * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
1526 if the vectorized statement is a store. A store statement can only
1527 appear at the end of pattern statements.
1529 2014-02-03 H.J. Lu <hongjiu.lu@intel.com>
1531 * config/i386/i386.c (flag_opts): Add -mlong-double-128.
1532 (ix86_option_override_internal): Default long double to 64-bit for
1533 32-bit Bionic and to 128-bit for 64-bit Bionic.
1535 * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
1536 TARGET_LONG_DOUBLE_128 is true.
1537 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1539 * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
1540 (mlong-double-64): Negate -mlong-double-128.
1541 (mlong-double-128): New option.
1543 * config/i386/i386-c.c (ix86_target_macros): Define
1544 __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
1546 * doc/invoke.texi: Document -mlong-double-128.
1548 2014-02-03 H.J. Lu <hongjiu.lu@intel.com>
1550 PR rtl-optimization/60024
1551 * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
1553 2014-02-03 Markus Trippelsdorf <markus@trippelsdorf.de>
1555 * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
1557 2014-02-03 Andrey Belevantsev <abel@ispras.ru>
1559 PR rtl-optimization/57662
1560 * sel-sched.c (code_motion_path_driver): Do not mark already not
1561 existing blocks in the visiting bitmap.
1563 2014-02-03 Andrey Belevantsev <abel@ispras.ru>
1565 * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
1566 on the insn being emitted.
1568 2014-02-03 James Greenhalgh <james.greenhalgh@arm.com>
1569 Will Deacon <will.deacon@arm.com>
1571 * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
1573 2014-02-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1575 * config/arm/arm-tables.opt: Regenerate.
1577 2014-02-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1579 * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
1580 for vector types other than V16QImode.
1581 * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
1582 define_expand, and call altivec_expand_vec_perm_le when producing
1583 code with little endian element order.
1584 (*altivec_vperm_<mode>_internal): New insn having previous
1585 behavior of altivec_vperm_<mode>.
1586 (altivec_vperm_<mode>_uns): Change to a define_expand, and call
1587 altivec_expand_vec_perm_le when producing code with little endian
1589 (*altivec_vperm_<mode>_uns_internal): New insn having previous
1590 behavior of altivec_vperm_<mode>_uns.
1592 2014-02-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1594 * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
1595 (altivec_vsumsws): Add handling for -maltivec=be with a little
1597 (altivec_vsumsws_direct): New.
1598 (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
1599 gen_altivec_vsumsws.
1601 2014-02-02 Jan Hubicka <hubicka@ucw.cz>
1603 * ipa-devirt.c (subbinfo_with_vtable_at_offset,
1604 vtable_pointer_value_to_binfo): New functions.
1605 * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
1606 * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
1608 2014-02-02 Sandra Loosemore <sandra@codesourcery.com>
1610 * config/nios2/nios2.md (load_got_register): Initialize GOT
1611 pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
1612 * config/nios2/nios2.c (nios2_function_profiler): Likewise.
1614 2014-02-02 Jan Hubicka <hubicka@ucw.cz>
1616 * ipa-prop.c (update_jump_functions_after_inlining): When type is not
1617 preserverd by passthrough, do not propagate the type.
1619 2014-02-02 Richard Sandiford <rdsandiford@googlemail.com>
1621 * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
1622 (mips_atomic_assign_expand_fenv): New function.
1623 (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
1625 2014-02-02 Richard Sandiford <rdsandiford@googlemail.com>
1627 * doc/extend.texi (__builtin_mips_get_fcsr): Document.
1628 (__builtin_mips_set_fcsr): Likewise.
1629 * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
1630 MIPS_USI_FTYPE_VOID.
1631 * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
1632 (mips16_expand_set_fcsr): Likewise.
1633 * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
1634 (mips16_set_fcsr_stub): Likewise.
1635 (mips16_get_fcsr_one_only_stub): New class.
1636 (mips16_set_fcsr_one_only_stub): Likewise.
1637 (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
1638 (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
1639 (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
1640 (hard_float): New availability predicate.
1641 (mips_builtins): Add get_fcsr and set_fcsr.
1642 (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
1643 * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
1644 (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
1645 (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
1646 (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
1649 2014-02-02 Richard Sandiford <rdsandiford@googlemail.com>
1651 * config/mips/mips.c (mips_one_only_stub): New class.
1652 (mips_need_mips16_rdhwr_p): Replace with...
1653 (mips16_rdhwr_stub): ...this new variable.
1654 (mips16_stub_call_address): New function.
1655 (mips16_rdhwr_one_only_stub): New class.
1656 (mips_expand_thread_pointer): Use mips16_stub_call_address.
1657 (mips_output_mips16_rdhwr): Delete.
1658 (mips_finish_stub): New function.
1659 (mips_code_end): Use it to handle rdhwr stubs.
1661 2014-02-02 Uros Bizjak <ubizjak@gmail.com>
1664 * config/i386/i386.c (classify_argument): Fix handling of bit_offset
1665 when calculating size of integer atomic types.
1667 2014-02-02 H.J. Lu <hongjiu.lu@intel.com>
1669 * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
1671 2014-02-01 Jakub Jelinek <jakub@redhat.com>
1673 PR tree-optimization/60003
1674 * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
1675 * profile.c (branch_prob): Use gimple_call_builtin_p
1676 to check for BUILT_IN_SETJMP_RECEIVER.
1677 * tree-inline.c (copy_bb): Call notice_special_calls.
1679 2014-01-31 Vladimir Makarov <vmakarov@redhat.com>
1682 * lra-constraints.c (process_alt_operands): Update reload_sum only
1685 2014-01-31 Richard Henderson <rth@redhat.com>
1688 * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
1689 until after else_eh is processed.
1691 2014-01-31 Ilya Tocar <ilya.tocar@intel.com>
1693 * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
1694 (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
1695 (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
1696 in smmintrin.h, remove them.
1697 (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
1698 * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
1699 * config/i386/i386.md (ROUND_SAE): Fix value.
1700 * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
1701 (const48_operand): New.
1702 * config/i386/subst.md (round), (round_expand): Use
1703 const_4_or_8_to_11_operand.
1704 (round_saeonly), (round_saeonly_expand): Use const48_operand.
1706 2014-01-31 Ilya Tocar <ilya.tocar@intel.com>
1708 * config/i386/constraints.md (Yk): Swap meaning with k.
1709 * config/i386/i386.md (movhi_internal): Change Yk to k.
1710 (movqi_internal): Ditto.
1711 (*k<logic><mode>): Ditto.
1714 (kandn<mode>): Ditto.
1715 (*<code>hi_1): Ditto.
1716 (*<code>qi_1): Ditto.
1717 (kxnor<mode>): Ditto.
1718 (kortestzhi): Ditto.
1719 (kortestchi): Ditto.
1721 (*one_cmplhi2_1): Ditto.
1722 (*one_cmplqi2_1): Ditto.
1723 * config/i386/sse.md (): Change k to Yk.
1724 (avx512f_load<mode>_mask): Ditto.
1725 (avx512f_blendm<mode>): Ditto.
1726 (avx512f_store<mode>_mask): Ditto.
1727 (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
1728 (avx512f_storedqu<mode>_mask): Ditto.
1729 (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
1731 (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
1732 (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
1733 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
1734 (avx512f_maskcmp<mode>3): Ditto.
1735 (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
1736 (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
1737 (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
1738 (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
1739 (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
1740 (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
1741 (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
1742 (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
1743 (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
1744 (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
1745 (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
1746 (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
1747 (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
1748 (vec_extract_lo_<mode>_maskm): Ditto.
1749 (vec_extract_hi_<mode>_maskm): Ditto.
1750 (avx512f_vternlog<mode>_mask): Ditto.
1751 (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
1752 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
1753 (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
1754 (avx512f_<code>v8div16qi2_mask): Ditto.
1755 (avx512f_<code>v8div16qi2_mask_store): Ditto.
1756 (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
1757 (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
1758 (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
1759 (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
1760 (*avx512pf_gatherpf<mode>sf_mask): Ditto.
1761 (*avx512pf_gatherpf<mode>df_mask): Ditto.
1762 (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1763 (*avx512pf_scatterpf<mode>df_mask): Ditto.
1764 (avx512cd_maskb_vec_dupv8di): Ditto.
1765 (avx512cd_maskw_vec_dupv16si): Ditto.
1766 (avx512f_vpermi2var<mode>3_maskz): Ditto.
1767 (avx512f_vpermi2var<mode>3_mask): Ditto.
1768 (avx512f_vpermi2var<mode>3_mask): Ditto.
1769 (avx512f_vpermt2var<mode>3_maskz): Ditto.
1770 (*avx512f_gathersi<mode>): Ditto.
1771 (*avx512f_gathersi<mode>_2): Ditto.
1772 (*avx512f_gatherdi<mode>): Ditto.
1773 (*avx512f_gatherdi<mode>_2): Ditto.
1774 (*avx512f_scattersi<mode>): Ditto.
1775 (*avx512f_scatterdi<mode>): Ditto.
1776 (avx512f_compress<mode>_mask): Ditto.
1777 (avx512f_compressstore<mode>_mask): Ditto.
1778 (avx512f_expand<mode>_mask): Ditto.
1779 * config/i386/subst.md (mask): Change k to Yk.
1780 (mask_scalar_merge): Ditto.
1783 2014-01-31 Marc Glisse <marc.glisse@inria.fr>
1785 * doc/extend.texi (Vector Extensions): Document ?: in C++.
1787 2014-01-31 Richard Biener <rguenther@suse.de>
1790 * builtins.c (fold_builtin_memory_op): Make sure to not
1791 use a floating-point mode or a boolean or enumeral type for
1794 2014-01-30 DJ Delorie <dj@redhat.com>
1796 * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
1797 * config/msp430/msp430.md (msp430_refsym_need_exit): New.
1798 * config/msp430/msp430.c (msp430_expand_epilogue): Call it
1799 whenever main() has an epilogue.
1801 2014-01-30 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1803 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
1804 unused variable "field".
1805 * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
1806 (vsx_mergeh_<mode>): Likewise.
1807 * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
1808 (altivec_vmrghh): Likewise.
1809 (altivec_vmrghw): Likewise.
1810 (altivec_vmrglb): Likewise.
1811 (altivec_vmrglh): Likewise.
1812 (altivec_vmrglw): Likewise.
1813 (altivec_vspltb): Add missing uses.
1814 (altivec_vsplth): Likewise.
1815 (altivec_vspltw): Likewise.
1816 (altivec_vspltsf): Likewise.
1818 2014-01-30 Jakub Jelinek <jakub@redhat.com>
1821 * ifcvt.c (cond_exec_process_insns): Don't conditionalize
1822 frame related instructions.
1824 2014-01-30 Vladimir Makarov <vmakarov@redhat.com>
1826 PR rtl-optimization/59959
1827 * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
1828 any reload of register whose subreg is invalid.
1830 2014-01-30 Jakub Jelinek <jakub@redhat.com>
1832 * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
1833 * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
1834 Add missing return type - void.
1836 2014-01-30 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1838 * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
1839 gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
1840 remove element index adjustment for endian (now handled in vsx.md
1842 (altivec_expand_vec_perm_const): Use
1843 gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
1844 * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
1845 (vsx_xxspltw_<mode>): Adjust element index for little endian.
1846 * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
1847 define_expand and a new define_insn *altivec_vspltb_internal;
1848 adjust for -maltivec=be on a little endian target.
1849 (altivec_vspltb_direct): New.
1850 (altivec_vsplth): Divide into a define_expand and a new
1851 define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
1852 little endian target.
1853 (altivec_vsplth_direct): New.
1854 (altivec_vspltw): Divide into a define_expand and a new
1855 define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
1856 little endian target.
1857 (altivec_vspltw_direct): New.
1858 (altivec_vspltsf): Divide into a define_expand and a new
1859 define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
1860 a little endian target.
1862 2014-01-30 Richard Biener <rguenther@suse.de>
1864 PR tree-optimization/59993
1865 * tree-ssa-forwprop.c (associate_pointerplus): Check we
1866 can propagate form the earlier stmt and avoid the transform
1867 when the intermediate result is needed.
1869 2014-01-30 Alangi Derick <alangiderick@gmail.com>
1871 * README.Portability: Fix typo.
1873 2014-01-30 David Holsgrove <david.holsgrove@xilinx.com>
1875 * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
1876 comparison_operator with ordered_comparison_operator.
1878 2014-01-30 Nick Clifton <nickc@redhat.com>
1880 * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
1881 Rename to mn10300_store_multiple_regs.
1882 * config/mn10300/mn10300.c: Likewise.
1883 * config/mn10300/mn10300.md (store_movm): Fix typo: call
1884 store_multiple_regs.
1885 * config/mn10300/predicates.md (mn10300_store_multiple_operation):
1886 Call mn10300_store_multiple_regs.
1888 2014-01-30 Nick Clifton <nickc@redhat.com>
1889 DJ Delorie <dj@redhat.com>
1891 * config/rl78/rl78.c (register_sizes): Make the "upper half" of
1892 %fp 2 to keep registers after it properly word-aligned.
1893 (rl78_alloc_physical_registers_umul): Handle the case where both
1894 input operands are the same.
1896 2014-01-30 Richard Biener <rguenther@suse.de>
1898 PR tree-optimization/59903
1899 * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
1902 2014-01-30 Jason Merrill <jason@redhat.com>
1905 * tree.c (walk_type_fields): Handle VECTOR_TYPE.
1908 * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
1910 2014-01-30 Richard Biener <rguenther@suse.de>
1912 PR tree-optimization/59951
1913 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
1915 2014-01-30 Savin Zlobec <savin.zlobec@gmail.com>
1918 * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
1919 SFmode to DFmode case.
1921 2014-01-29 DJ Delorie <dj@redhat.com>
1923 * config/msp430/msp430.opt (-minrt): New.
1924 * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
1926 (ENDFILE_SPEC): Likewise.
1928 2014-01-29 Jan Hubicka <hubicka@ucw.cz>
1930 * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
1931 (estimate_function_body_sizes): Use it.
1933 2014-01-29 Paolo Carlini <paolo.carlini@oracle.com>
1936 * dwarf2out.c (is_cxx_auto): New.
1937 (is_base_type): Use it.
1938 (gen_type_die_with_usage): Likewise.
1940 2014-01-29 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1942 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Use
1943 CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
1944 * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
1945 -maltivec=be with LE targets.
1946 (vsx_mergeh_<mode>): Likewise.
1947 * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
1948 (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
1949 (altivec_vmrghb): Replace with define_expand and new
1950 *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
1951 (altivec_vmrghb_direct): New define_insn.
1952 (altivec_vmrghh): Replace with define_expand and new
1953 *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
1954 (altivec_vmrghh_direct): New define_insn.
1955 (altivec_vmrghw): Replace with define_expand and new
1956 *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
1957 (altivec_vmrghw_direct): New define_insn.
1958 (*altivec_vmrghsf): Adjust for endianness.
1959 (altivec_vmrglb): Replace with define_expand and new
1960 *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
1961 (altivec_vmrglb_direct): New define_insn.
1962 (altivec_vmrglh): Replace with define_expand and new
1963 *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
1964 (altivec_vmrglh_direct): New define_insn.
1965 (altivec_vmrglw): Replace with define_expand and new
1966 *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
1967 (altivec_vmrglw_direct): New define_insn.
1968 (*altivec_vmrglsf): Adjust for endianness.
1969 (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
1970 (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
1971 (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
1972 (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
1973 (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
1974 (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
1975 (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
1976 (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
1978 2014-01-29 Marcus Shawcroft <marcus.shawcroft@arm.com>
1980 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
1981 (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
1984 2014-01-29 Richard Biener <rguenther@suse.de>
1986 PR tree-optimization/58742
1987 * tree-ssa-forwprop.c (associate_pointerplus): Rename to
1988 associate_pointerplus_align.
1989 (associate_pointerplus_diff): New function.
1990 (associate_pointerplus): Likewise. Call associate_pointerplus_align
1991 and associate_pointerplus_diff.
1993 2014-01-29 Richard Biener <rguenther@suse.de>
1995 * lto-streamer.h (LTO_major_version): Bump to 3.
1996 (LTO_minor_version): Reset to 0.
1998 2014-01-29 Renlin Li <Renlin.Li@arm.com>
2000 * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
2001 * config/arm/arm.c (FL_FOR_ARCH7VE): New.
2002 (arm_file_start): Generate correct asm header for armv7ve.
2003 * config/arm/bpabi.h: Add multilib support for armv7ve.
2004 * config/arm/driver-arm.c: Change the architectures of cortex-a7
2005 and cortex-a15 to armv7ve.
2006 * config/arm/t-aprofile: Add multilib support for armv7ve.
2007 * doc/invoke.texi: Document -march=armv7ve.
2009 2014-01-29 Richard Biener <rguenther@suse.de>
2011 PR tree-optimization/58742
2012 * tree-ssa-forwprop.c (associate_plusminus): Return true
2013 if we changed sth, defer EH cleanup to ...
2014 (ssa_forward_propagate_and_combine): ... here. Call simplify_mult.
2015 (simplify_mult): New function.
2017 2014-01-29 Jakub Jelinek <jakub@redhat.com>
2020 PR tree-optimization/59920
2021 * tree.c (build_common_builtin_nodes): Remove
2022 __builtin_setjmp_dispatcher initialization.
2023 * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
2024 * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
2025 instead of gsi_after_labels + manually skipping debug stmts.
2026 Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
2027 ignore bbs with IFN_ABNORMAL_DISPATCHER.
2028 * tree-inline.c (copy_edges_for_bb): Remove
2029 can_make_abnormal_goto argument, instead add abnormal_goto_dest
2030 argument. Ignore computed_goto_p stmts. Don't call
2031 make_abnormal_goto_edges. If a call might need abnormal edges
2032 for non-local gotos, see if it already has an edge to
2033 IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
2034 with true argument, don't do anything then, otherwise add
2035 EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
2036 (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
2038 * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
2039 (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
2040 (lower_stmt): Don't set data->calls_builtin_setjmp.
2041 (lower_builtin_setjmp): Adjust comment.
2042 * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
2043 * tree-cfg.c (found_computed_goto): Remove.
2044 (factor_computed_gotos): Remove.
2045 (make_goto_expr_edges): Return bool, true for computed gotos.
2046 Don't call make_abnormal_goto_edges.
2047 (build_gimple_cfg): Don't set found_computed_goto, don't call
2048 factor_computed_gotos.
2049 (computed_goto_p): No longer static.
2050 (make_blocks): Don't set found_computed_goto.
2051 (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
2052 (make_edges): If make_goto_expr_edges returns true, push bb
2053 into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
2054 instead of calling make_abnormal_goto_edges push bb into ab_edge_call
2055 vector. Record mapping between bbs and OpenMP regions if there
2056 are any, adjust make_gimple_omp_edges caller. Call
2057 handle_abnormal_edges.
2058 (make_abnormal_goto_edges): Remove.
2059 * tree-cfg.h (make_abnormal_goto_edges): Remove.
2060 (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
2061 * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
2062 * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
2063 * internal-fn.def (ABNORMAL_DISPATCHER): New.
2064 * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
2065 filling *region also set *region_idx to (*region)->entry->index.
2068 * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
2069 For REGs set ORIGINAL_REGNO.
2071 2014-01-29 Bingfeng Mei <bmei@broadcom.com>
2073 * doc/md.texi: Mention that a target shouldn't implement
2074 vec_widen_(s|u)mul_even/odd pair if it is less efficient
2077 2014-01-29 Jakub Jelinek <jakub@redhat.com>
2079 PR tree-optimization/59594
2080 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
2081 a copy of the datarefs vector rather than the vector itself.
2083 2014-01-28 Jason Merrill <jason@redhat.com>
2086 * dwarf2out.c (auto_die): New static.
2087 (gen_type_die_with_usage): Handle C++1y 'auto'.
2088 (gen_subprogram_die): If in-class DIE had 'auto', emit type again
2091 2014-01-28 H.J. Lu <hongjiu.lu@intel.com>
2094 * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
2095 (SPEC_X32): Likewise.
2096 (SPEC_64): Likewise.
2097 * config/i386/i386.c (ix86_option_override_internal): Turn off
2098 OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
2100 (x86_file_start): Output .code16gcc for TARGET_16BIT.
2101 * config/i386/i386.h (TARGET_16BIT): New macro.
2102 (TARGET_16BIT_P): Likewise.
2103 * config/i386/i386.opt: Add m16.
2104 * doc/invoke.texi: Document -m16.
2106 2014-01-28 Jakub Jelinek <jakub@redhat.com>
2108 PR preprocessor/59935
2109 * input.c (location_get_source_line): Bail out on when line number
2110 is zero, and test the return value of lookup_or_add_file_to_cache_tab.
2112 2014-01-28 Richard Biener <rguenther@suse.de>
2114 PR tree-optimization/58742
2115 * tree-ssa-forwprop.c (associate_plusminus): Handle
2116 pointer subtraction of the form (T)(P + A) - (T)P.
2118 2014-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2120 * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
2123 2014-01-28 Richard Biener <rguenther@suse.de>
2126 2014-01-28 Richard Biener <rguenther@suse.de>
2128 PR rtl-optimization/45364
2129 PR rtl-optimization/59890
2130 * var-tracking.c (local_get_addr_clear_given_value): Handle
2131 already cleared slot.
2132 (val_reset): Handle not allocated local_get_addr_cache.
2133 (vt_find_locations): Use post-order on the inverted CFG.
2135 2014-01-28 Richard Biener <rguenther@suse.de>
2137 * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
2139 2014-01-28 Richard Biener <rguenther@suse.de>
2141 PR rtl-optimization/45364
2142 PR rtl-optimization/59890
2143 * var-tracking.c (local_get_addr_clear_given_value): Handle
2144 already cleared slot.
2145 (val_reset): Handle not allocated local_get_addr_cache.
2146 (vt_find_locations): Use post-order on the inverted CFG.
2148 2014-01-28 Alan Modra <amodra@gmail.com>
2150 * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
2151 * configure.ac <recursive call for build != host>: Define
2152 GENERATOR_FILE. Comment. Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
2153 and LD_FOR_BUILD too.
2154 * configure: Regenerate.
2156 2014-01-27 Allan Sandfeld Jensen <sandfeld@kde.org>
2158 * config/i386/i386.c (get_builtin_code_for_version): Separate
2159 Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
2160 Broadwell from Haswell.
2162 2014-01-27 Steve Ellcey <sellcey@mips.com>
2164 * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2165 Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
2166 * config/mips/mips.c (mips_option_override): Change setting
2168 * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
2169 * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
2170 Change from Mask to Var.
2172 2014-01-27 Jeff Law <law@redhat.com>
2174 * ipa-inline.c (inline_small_functions): Fix typo.
2176 2014-01-27 Ilya Tocar <ilya.tocar@intel.com>
2178 * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
2179 (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
2180 (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
2181 (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
2182 (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
2183 (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
2184 (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
2185 (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
2186 (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
2187 (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
2188 (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
2189 (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
2190 (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
2191 (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
2192 (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
2193 (_mm512_storeu_epi64): Ditto.
2194 (_mm512_cmpge_epi32_mask): Ditto.
2195 (_mm512_cmpge_epu32_mask): Ditto.
2196 (_mm512_cmpge_epi64_mask): Ditto.
2197 (_mm512_cmpge_epu64_mask): Ditto.
2198 (_mm512_cmple_epi32_mask): Ditto.
2199 (_mm512_cmple_epu32_mask): Ditto.
2200 (_mm512_cmple_epi64_mask): Ditto.
2201 (_mm512_cmple_epu64_mask): Ditto.
2202 (_mm512_cmplt_epi32_mask): Ditto.
2203 (_mm512_cmplt_epu32_mask): Ditto.
2204 (_mm512_cmplt_epi64_mask): Ditto.
2205 (_mm512_cmplt_epu64_mask): Ditto.
2206 (_mm512_cmpneq_epi32_mask): Ditto.
2207 (_mm512_cmpneq_epu32_mask): Ditto.
2208 (_mm512_cmpneq_epi64_mask): Ditto.
2209 (_mm512_cmpneq_epu64_mask): Ditto.
2210 (_mm512_expand_pd): Ditto.
2211 (_mm512_expand_ps): Ditto.
2212 * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
2213 VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
2214 VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
2215 * config/i386/i386.c (ix86_builtins): Add
2216 IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
2217 IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
2218 IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
2219 IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
2220 IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
2221 IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
2222 IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
2223 IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
2224 IX86_BUILTIN_PMOVUSQW512_MEM.
2225 (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
2226 __builtin_ia32_pmovsqd512mem_mask,
2227 __builtin_ia32_pmovqd512mem_mask,
2228 __builtin_ia32_pmovusqw512mem_mask,
2229 __builtin_ia32_pmovsqw512mem_mask,
2230 __builtin_ia32_pmovqw512mem_mask,
2231 __builtin_ia32_pmovusdw512mem_mask,
2232 __builtin_ia32_pmovsdw512mem_mask,
2233 __builtin_ia32_pmovdw512mem_mask,
2234 __builtin_ia32_pmovqb512mem_mask,
2235 __builtin_ia32_pmovusqb512mem_mask,
2236 __builtin_ia32_pmovsqb512mem_mask,
2237 __builtin_ia32_pmovusdb512mem_mask,
2238 __builtin_ia32_pmovsdb512mem_mask,
2239 __builtin_ia32_pmovdb512mem_mask.
2240 (bdesc_args): Add __builtin_ia32_expanddf512,
2241 __builtin_ia32_expandsf512.
2242 (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
2243 VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
2244 VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
2245 * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
2246 (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
2247 (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
2248 (avx512f_<code>v8div16qi2_mask_store): This.
2249 (avx512f_expand<mode>): New.
2251 2014-01-27 Kirill Yukhin <kirill.yukhin@intel.com>
2253 * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
2255 (_mm512_mask_prefetch_i64gather_pd): Ditto.
2256 (_mm512_prefetch_i32scatter_pd): Ditto.
2257 (_mm512_mask_prefetch_i32scatter_pd): Ditto.
2258 (_mm512_prefetch_i64scatter_pd): Ditto.
2259 (_mm512_mask_prefetch_i64scatter_pd): Ditto.
2260 (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
2261 (_mm512_mask_prefetch_i64gather_ps): Ditto.
2262 (_mm512_prefetch_i32scatter_ps): Ditto.
2263 (_mm512_mask_prefetch_i32scatter_ps): Ditto.
2264 (_mm512_prefetch_i64scatter_ps): Ditto.
2265 (_mm512_mask_prefetch_i64scatter_ps): Ditto.
2266 * config/i386/i386-builtin-types.def: Define
2267 VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
2268 and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
2269 * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
2270 IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
2271 IX86_BUILTIN_SCATTERPFQPD.
2272 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
2273 __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
2274 __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
2275 __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
2276 __builtin_ia32_scatterpfqps.
2277 (ix86_expand_builtin): Expand new built-ins.
2278 * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
2279 fix memory access data type.
2280 (*avx512pf_gatherpf<mode>_mask): Ditto.
2281 (*avx512pf_gatherpf<mode>): Ditto.
2282 (avx512pf_scatterpf<mode>): Ditto.
2283 (*avx512pf_scatterpf<mode>_mask): Ditto.
2284 (*avx512pf_scatterpf<mode>): Ditto.
2285 (GATHER_SCATTER_SF_MEM_MODE): New.
2286 (avx512pf_gatherpf<mode>df): Ditto.
2287 (*avx512pf_gatherpf<mode>df_mask): Ditto.
2288 (*avx512pf_scatterpf<mode>df): Ditto.
2290 2014-01-27 Jakub Jelinek <jakub@redhat.com>
2293 * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
2294 know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
2297 2014-01-27 James Greenhalgh <james.greenhalgh@arm.com>
2299 * common/config/arm/arm-common.c
2300 (arm_rewrite_mcpu): Handle multiple names.
2302 (BIG_LITTLE_SPEC): Do not discard mcpu switches.
2304 2014-01-27 James Greenhalgh <james.greenhalgh@arm.com>
2306 * gimple-builder.h (create_gimple_tmp): Delete.
2308 2014-01-27 Christian Bruel <christian.bruel@st.com>
2310 * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
2313 2014-01-26 John David Anglin <danglin@gcc.gnu.org>
2315 * config/pa/pa.md (call): Generate indirect long calls to non-local
2316 functions when outputing 32-bit code.
2317 (call_value): Likewise except for special call to buggy powf function.
2319 * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
2320 portable runtime and PIC indirect calls.
2321 (pa_output_indirect_call): Remove unnecessary nop from portable runtime
2322 and PIC call sequences. Use ldo instead of blr to set return register
2323 in PIC call sequence.
2325 2014-01-25 Walter Lee <walt@tilera.com>
2327 * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
2328 avoid clobbering a live register.
2330 2014-01-25 Walter Lee <walt@tilera.com>
2332 * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
2333 Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
2334 * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
2335 Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
2337 2014-01-25 Walter Lee <walt@tilera.com>
2339 * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
2340 arguments on even registers.
2341 (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
2343 * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
2344 (BIGGEST_ALIGNMENT): Ditto.
2345 (BIGGEST_FIELD_ALIGNMENT): Ditto.
2347 2014-01-25 Walter Lee <walt@tilera.com>
2349 * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
2350 insns before bundling.
2351 * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
2353 2014-01-25 Walter Lee <walt@tilera.com>
2355 * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
2356 PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
2357 * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
2359 2014-01-25 Richard Sandiford <rdsandiford@googlemail.com>
2361 * config/mips/constraints.md (kl): Delete.
2362 * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
2363 define expands, using...
2364 (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
2365 instructions for MIPS16.
2366 (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
2367 non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
2369 2014-01-25 Walter Lee <walt@tilera.com>
2371 * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
2375 2014-01-25 Walter Lee <walt@tilera.com>
2377 * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
2378 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2380 2014-01-25 Richard Sandiford <rdsandiford@googlemail.com>
2382 * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
2385 2014-01-25 Jakub Jelinek <jakub@redhat.com>
2387 * print-rtl.c (in_call_function_usage): New var.
2388 (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
2389 EXPR_LIST mode as mode and not as reg note name.
2392 * cfgloopmanip.c (copy_loop_info): If
2393 loop->warned_aggressive_loop_optimizations, make sure
2394 the flag is set in target loop too.
2396 2014-01-24 Balaji V. Iyer <balaji.v.iyer@intel.com>
2398 * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
2400 * builtins.def: Likewise.
2401 * cilk.h (fn_contains_cilk_spawn_p): Likewise.
2402 * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
2403 * ira.c (ira_setup_eliminable_regset): Likewise.
2404 * omp-low.c (gate_expand_omp): Likewise.
2405 (execute_lower_omp): Likewise.
2406 (diagnose_sb_0): Likewise.
2407 (gate_diagnose_omp_blocks): Likewise.
2408 (simd_clone_clauses_extract): Likewise.
2411 2014-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2413 * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
2414 correction for little endian...
2415 * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
2418 2014-01-24 Jeff Law <law@redhat.com>
2420 PR tree-optimization/59919
2421 * tree-vrp.c (find_assert_locations_1): Do not register asserts
2422 for non-returning calls.
2424 2014-01-24 James Greenhalgh <james.greenhalgh@arm.com>
2426 * common/config/aarch64/aarch64-common.c
2427 (aarch64_rewrite_mcpu): Handle multiple names.
2428 * config/aarch64/aarch64.h
2429 (BIG_LITTLE_SPEC): Do not discard mcpu switches.
2431 2014-01-24 Dodji Seketeli <dodji@redhat.com>
2433 * input.c (add_file_to_cache_tab): Handle the case where fopen
2436 2014-01-23 H.J. Lu <hongjiu.lu@intel.com>
2439 * config/i386/i386.md (pushsf splitter): Get stack adjustment
2440 from push operand if code of push isn't PRE_DEC.
2442 2014-01-23 Michael Meissner <meissner@linux.vnet.ibm.com>
2445 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2446 -mquad-memory-atomic. Update -mquad-memory documentation to say
2447 it is only used for non-atomic loads/stores.
2449 * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
2450 -mquad-memory or -mquad-memory-atomic switches.
2452 * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
2453 -mquad-memory-atomic to ISA 2.07 support.
2455 * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
2456 to separate support of normal quad word memory operations (ldq, stq)
2457 from the atomic quad word memory operations.
2459 * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
2460 support to separate non-atomic quad word operations from atomic
2461 quad word operations. Disable non-atomic quad word operations in
2462 little endian mode so that we don't have to swap words after the
2463 load and before the store.
2464 (quad_load_store_p): Add comment about atomic quad word support.
2465 (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
2466 options printed with -mdebug=reg.
2468 * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
2469 -mquad-memory-atomic as the test for whether we have quad word
2470 atomic instructions.
2471 (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
2472 or -mp8-vector are used, allow byte/half-word atomic operations.
2474 * config/rs6000/sync.md (load_lockedti): Insure that the address
2475 is a proper indexed or indirect address for the lqarx instruction.
2476 On little endian systems, swap the hi/lo registers after the lqarx
2478 (load_lockedpti): Use indexed_or_indirect_operand predicate to
2479 insure the address is valid for the lqarx instruction.
2480 (store_conditionalti): Insure that the address is a proper indexed
2481 or indirect address for the stqcrx. instruction. On little endian
2482 systems, swap the hi/lo registers before doing the stqcrx.
2484 (store_conditionalpti): Use indexed_or_indirect_operand predicate to
2485 insure the address is valid for the stqcrx. instruction.
2487 * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
2488 Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
2489 type of quad memory support is available.
2491 2014-01-23 Vladimir Makarov <vmakarov@redhat.com>
2494 * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
2495 there is a danger of looping.
2497 2014-01-23 Pat Haugen <pthaugen@us.ibm.com>
2499 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2500 force flag_ira_loop_pressure if set via command line.
2502 2014-01-23 Alex Velenko <Alex.Velenko@arm.com>
2504 * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
2505 (ashr_simd): New builtin handling DI mode.
2506 * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
2507 (aarch64_sshr_simddi): New match pattern.
2508 * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
2509 (vshrd_n_s64): Likewise.
2510 * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
2512 2014-01-23 Nick Clifton <nickc@redhat.com>
2514 * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
2515 (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
2516 favour of mcu specific scripts.
2517 * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
2520 2014-01-23 James Greenhalgh <james.greenhalgh@arm.com>
2521 Alex Velenko <Alex.Velenko@arm.com>
2523 * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
2524 (vaddv_s16): Likewise.
2525 (vaddv_s32): Likewise.
2526 (vaddv_u8): Likewise.
2527 (vaddv_u16): Likewise.
2528 (vaddv_u32): Likewise.
2529 (vaddvq_s8): Likewise.
2530 (vaddvq_s16): Likewise.
2531 (vaddvq_s32): Likewise.
2532 (vaddvq_s64): Likewise.
2533 (vaddvq_u8): Likewise.
2534 (vaddvq_u16): Likewise.
2535 (vaddvq_u32): Likewise.
2536 (vaddvq_u64): Likewise.
2537 (vaddv_f32): Likewise.
2538 (vaddvq_f32): Likewise.
2539 (vaddvq_f64): Likewise.
2540 (vmaxv_f32): Likewise.
2541 (vmaxv_s8): Likewise.
2542 (vmaxv_s16): Likewise.
2543 (vmaxv_s32): Likewise.
2544 (vmaxv_u8): Likewise.
2545 (vmaxv_u16): Likewise.
2546 (vmaxv_u32): Likewise.
2547 (vmaxvq_f32): Likewise.
2548 (vmaxvq_f64): Likewise.
2549 (vmaxvq_s8): Likewise.
2550 (vmaxvq_s16): Likewise.
2551 (vmaxvq_s32): Likewise.
2552 (vmaxvq_u8): Likewise.
2553 (vmaxvq_u16): Likewise.
2554 (vmaxvq_u32): Likewise.
2555 (vmaxnmv_f32): Likewise.
2556 (vmaxnmvq_f32): Likewise.
2557 (vmaxnmvq_f64): Likewise.
2558 (vminv_f32): Likewise.
2559 (vminv_s8): Likewise.
2560 (vminv_s16): Likewise.
2561 (vminv_s32): Likewise.
2562 (vminv_u8): Likewise.
2563 (vminv_u16): Likewise.
2564 (vminv_u32): Likewise.
2565 (vminvq_f32): Likewise.
2566 (vminvq_f64): Likewise.
2567 (vminvq_s8): Likewise.
2568 (vminvq_s16): Likewise.
2569 (vminvq_s32): Likewise.
2570 (vminvq_u8): Likewise.
2571 (vminvq_u16): Likewise.
2572 (vminvq_u32): Likewise.
2573 (vminnmv_f32): Likewise.
2574 (vminnmvq_f32): Likewise.
2575 (vminnmvq_f64): Likewise.
2577 2014-01-23 James Greenhalgh <james.greenhalgh@arm.com>
2579 * config/aarch64/aarch64-simd.md
2580 (aarch64_dup_lane<mode>): Correct lane number on big-endian.
2581 (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
2582 (*aarch64_mul3_elt<mode>): Likewise.
2583 (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
2584 (*aarch64_mul3_elt_to_64v2df): Likewise.
2585 (*aarch64_mla_elt<mode>): Likewise.
2586 (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
2587 (*aarch64_mls_elt<mode>): Likewise.
2588 (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
2589 (*aarch64_fma4_elt<mode>): Likewise.
2590 (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
2591 (*aarch64_fma4_elt_to_64v2df): Likewise.
2592 (*aarch64_fnma4_elt<mode>): Likewise.
2593 (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
2594 (*aarch64_fnma4_elt_to_64v2df): Likewise.
2595 (aarch64_sq<r>dmulh_lane<mode>): Likewise.
2596 (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
2597 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
2598 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
2599 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
2600 (aarch64_sqdmull_lane<mode>_internal): Likewise.
2601 (aarch64_sqdmull2_lane<mode>_internal): Likewise.
2603 2013-01-23 Alex Velenko <Alex.Velenko@arm.com>
2605 * config/aarch64/aarch64-simd.md
2606 (aarch64_be_checked_get_lane<mode>): New define_expand.
2607 * config/aarch64/aarch64-simd-builtins.def
2608 (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
2609 New builtin definition.
2610 * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
2611 Use new safe be builtin.
2613 2014-01-23 Alex Velenko <Alex.Velenko@arm.com>
2615 * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
2617 (aarch64_be_st1<mode>): Likewise.
2618 (aarch_ld1<VALL:mode>): Define_expand modified.
2619 (aarch_st1<VALL:mode>): Likewise.
2620 * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
2621 (UNSPEC_ST1): Likewise.
2623 2014-01-23 David Holsgrove <david.holsgrove@xilinx.com>
2625 * config/microblaze/microblaze.md: Add trap insn and attribute
2627 2014-01-23 Dodji Seketeli <dodji@redhat.com>
2629 PR preprocessor/58580
2630 * input.h (location_get_source_line): Take an additional line_size
2632 (void diagnostics_file_cache_fini): Declare new function.
2633 * input.c (struct fcache): New type.
2634 (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
2635 New static constants.
2636 (diagnostic_file_cache_init, total_lines_num)
2637 (lookup_file_in_cache_tab, evicted_cache_tab_entry)
2638 (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
2639 (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
2640 (get_next_line, read_next_line, goto_next_line, read_line_num):
2641 New static function definitions.
2642 (diagnostic_file_cache_fini): New function.
2643 (location_get_source_line): Take an additional output line_len
2644 parameter. Re-write using lookup_or_add_file_to_cache_tab and
2646 * diagnostic.c (diagnostic_finish): Call
2647 diagnostic_file_cache_fini.
2648 (adjust_line): Take an additional input parameter for the length
2649 of the line, rather than calculating it with strlen.
2650 (diagnostic_show_locus): Adjust the use of
2651 location_get_source_line and adjust_line with respect to their new
2652 signature. While displaying a line now, do not stop at the first
2653 null byte. Rather, display the zero byte as a space and keep
2654 going until we reach the size of the line.
2655 * Makefile.in: Add vec.o to OBJS-libcommon
2657 2014-01-23 Kirill Yukhin <kirill.yukhin@intel.com>
2658 Ilya Tocar <ilya.tocar@intel.com>
2660 * config/i386/avx512fintrin.h (_mm512_kmov): New.
2661 * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
2662 (__builtin_ia32_kmov16): Ditto.
2663 * config/i386/i386.md (UNSPEC_KMOV): New.
2666 2014-01-23 Kirill Yukhin <kirill.yukhin@intel.com>
2668 * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
2669 (_mm512_storeu_si512): Ditto.
2671 2014-01-23 Richard Sandiford <rdsandiford@googlemail.com>
2674 * rtl.h (get_referenced_operands): Declare.
2675 * recog.c (get_referenced_operands): New function.
2676 * config/mips/mips.c (mips_reorg_process_insns): Check which asm
2677 operands have been referenced when recording LO_SUM references.
2679 2014-01-22 David Holsgrove <david.holsgrove@xilinx.com>
2681 * config/microblaze/microblaze.md: Correct bswaphi2 insn.
2683 2014-01-22 Jan Hubicka <hubicka@ucw.cz>
2685 * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
2686 Enable for generic and recent AMD targets.
2688 2014-01-22 Jan Hubicka <hubicka@ucw.cz>
2690 * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
2691 ARG_SIZE note when adjustment was eliminated.
2693 2014-01-22 Jeff Law <law@redhat.com>
2695 PR tree-optimization/59597
2696 * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
2697 in file. Accept new argument REGISTERING and use it to modify
2698 dump output appropriately.
2699 (register_jump_thread): Corresponding changes.
2700 (mark_threaded_blocks): Reinstate code to cancel unprofitable
2701 thread paths involving joiner blocks. Add code to dump cancelled
2702 jump threading paths.
2704 2014-01-22 Vladimir Makarov <vmakarov@redhat.com>
2706 PR rtl-optimization/59477
2707 * lra-constraints.c (inherit_in_ebb): Process call for living hard
2708 regs. Update reloads_num and potential_reload_hard_regs for all insns.
2710 2014-01-22 Tom Tromey <tromey@redhat.com>
2712 * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
2714 * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
2716 2014-01-21 Vladimir Makarov <vmakarov@redhat.com>
2718 PR rtl-optimization/59896
2719 * lra-constraints.c (process_alt_operands): Check unused note for
2720 matched operands of insn with no output reloads.
2722 2014-01-21 Richard Sandiford <rdsandiford@googlemail.com>
2724 * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
2725 (mips_move_from_gpr_cost): Likewise.
2727 2014-01-21 Vladimir Makarov <vmakarov@redhat.com>
2729 PR rtl-optimization/59858
2730 * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
2731 ira_class_hard_regs_num.
2732 (process_alt_operands): Increase reject for dying matched operand.
2734 2014-01-21 Jakub Jelinek <jakub@redhat.com>
2737 * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
2738 smaller than size, perform several stores or loads and stores
2739 at dst + count - size to store or copy all of size bytes, rather
2740 than just last modesize bytes.
2742 2014-01-20 DJ Delorie <dj@redhat.com>
2744 * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
2745 that CLOBBERs are REGs before propogating their values.
2747 2014-01-20 H.J. Lu <hongjiu.lu@intel.com>
2750 * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
2751 (cgraph_inline_failed_type): New function.
2752 * cgraph.h (DEFCIFCODE): Add type.
2753 (cgraph_inline_failed_type_t): New enum.
2754 (cgraph_inline_failed_type): New prototype.
2755 * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
2756 FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
2757 FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
2758 LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
2759 MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
2760 RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
2761 OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
2762 INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
2763 Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
2764 FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
2765 EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
2766 OPTIMIZATION_MISMATCH.
2767 * tree-inline.c (expand_call_inline): Emit errors during
2768 early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
2770 2014-01-20 Uros Bizjak <ubizjak@gmail.com>
2773 * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
2774 mode attribute in insn output.
2776 2014-01-20 Eric Botcazou <ebotcazou@adacore.com>
2778 * output.h (output_constant): Delete.
2779 * varasm.c (output_constant): Make private.
2781 2014-01-20 Alex Velenko <Alex.Velenko@arm.com>
2783 * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
2785 2014-01-20 Jakub Jelinek <jakub@redhat.com>
2788 * tree.h (fold_builtin_strcat): New prototype.
2789 * builtins.c (fold_builtin_strcat): No longer static. Add len
2790 argument, if non-NULL, don't call c_strlen. Optimize
2791 directly into __builtin_memcpy instead of __builtin_strcpy.
2792 (fold_builtin_2): Adjust fold_builtin_strcat caller.
2793 * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
2795 2014-01-20 Uros Bizjak <ubizjak@gmail.com>
2797 * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
2798 for SImode_address_operand operands, having only a REG argument.
2800 2014-01-20 Marcus Shawcroft <marcus.shawcroft@arm.com>
2802 * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
2803 loader name using mbig-endian.
2804 (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
2806 2014-01-20 James Greenhalgh <james.greenhalgh@arm.com>
2808 * doc/invoke.texi (-march): Clarify documentation for AArch64.
2812 2014-01-20 Tejas Belagod <tejas.belagod@arm.com>
2814 * config/aarch64/aarch64-protos.h
2815 (aarch64_cannot_change_mode_class_ptr): Declare.
2816 * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
2817 aarch64_cannot_change_mode_class_ptr): New.
2818 * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
2819 backend hook aarch64_cannot_change_mode_class.
2821 2014-01-20 James Greenhalgh <james.greenhalgh@arm.com>
2823 * common/config/aarch64/aarch64-common.c
2824 (aarch64_handle_option): Don't handle any option order logic here.
2825 * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
2826 selected_cpu, warn on architecture version mismatch.
2827 (aarch64_override_options): Fix parsing order for option strings.
2829 2014-01-20 Jan-Benedict Glaw <jbglaw@lug-owl.de>
2830 Iain Sandoe <iain@codesourcery.com>
2833 * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
2834 warning. Amend comment to reflect current functionality.
2836 2014-01-20 Richard Biener <rguenther@suse.de>
2839 * builtins.c (fold_builtin_strcat): Remove case better handled
2840 by tree-ssa-strlen.c.
2842 2014-01-20 Alan Lawrence <alan.lawrence@arm.com>
2844 * config/aarch64/aarch64.opt
2845 (mcpu, march, mtune): Make case-insensitive.
2847 2014-01-20 Jakub Jelinek <jakub@redhat.com>
2850 * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
2851 if operands[1] is a REG or ZERO_EXTEND of a REG.
2853 2014-01-19 Jan Hubicka <hubicka@ucw.cz>
2855 * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
2857 2014-01-19 John David Anglin <danglin@gcc.gnu.org>
2859 * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
2860 long non-pic millicode calls.
2862 2014-01-19 Jan-Benedict Glaw <jbglaw@lug-owl.de>
2864 * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
2866 2014-01-19 Kito Cheng <kito@0xlab.org>
2868 * builtins.c (expand_movstr): Check movstr expand done or fail.
2870 2014-01-18 Uros Bizjak <ubizjak@gmail.com>
2871 H.J. Lu <hongjiu.lu@intel.com>
2874 * config/i386/i386.md (*lea<mode>): Zero-extend return register
2875 to DImode for zero-extended addresses.
2877 2014-01-19 Jakub Jelinek <jakub@redhat.com>
2879 PR rtl-optimization/57763
2880 * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
2881 on the new indirect jump_insn and increment LABEL_NUSES (label).
2883 2014-01-18 H.J. Lu <hongjiu.lu@intel.com>
2887 * config.gcc (x86_archs): New variable.
2888 (x86_64_archs): Likewise.
2889 (x86_cpus): Likewise.
2890 Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
2891 --with-arch/--with-cpu= options.
2892 Support --with-arch=/--with-cpu={nehalem,westmere,
2893 sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
2895 2014-01-18 Uros Bizjak <ubizjak@gmail.com>
2897 * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
2898 and PROCESSOR_ATHLON to simplify code. Move "memory" calculation.
2900 2014-01-18 Uros Bizjak <ubizjak@gmail.com>
2902 * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
2904 2014-01-18 Jakub Jelinek <jakub@redhat.com>
2907 * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
2908 clear cpp_get_options (parse_in)->warn_unused_macros for
2909 ix86_target_macros_internal with cpp_define.
2911 2014-01-18 Richard Sandiford <rdsandiford@googlemail.com>
2913 * jump.c (delete_related_insns): Keep (use (insn))s.
2914 * reorg.c (redundant_insn): Check for barriers too.
2916 2014-01-17 H.J. Lu <hongjiu.lu@intel.com>
2918 * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
2920 2014-01-17 John David Anglin <danglin@gcc.gnu.org>
2922 * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
2923 call to $$dyncall when TARGET_LONG_CALLS is true.
2925 2014-01-17 Jeff Law <law@redhat.com>
2927 * ree.c (combine_set_extension): Temporarily disable test for
2928 changing number of hard registers.
2930 2014-01-17 Jan Hubicka <hubicka@ucw.cz>
2933 * ipa-inline-analysis.c (inline_free_summary):
2934 Do not free summary of aliases.
2936 2014-01-17 Jakub Jelinek <jakub@redhat.com>
2939 * gimplify.c (gimplify_expr): Use create_tmp_var
2940 instead of create_tmp_var_raw. If cond doesn't have
2941 integral type, don't add the IFN_ANNOTATE builtin at all.
2943 2014-01-17 Martin Jambor <mjambor@suse.cz>
2946 * ipa-cp.c (prev_edge_clone): New variable.
2947 (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
2948 Also resize prev_edge_clone vector.
2949 (ipcp_edge_duplication_hook): Also update prev_edge_clone.
2950 (ipcp_edge_removal_hook): New function.
2951 (ipcp_driver): Register ipcp_edge_removal_hook.
2953 2014-01-17 Andrew Pinski <apinski@cavium.com>
2954 Steve Ellcey <sellcey@mips.com>
2957 * config/mips/mips.c (mips_print_operand): Check operand mode instead
2960 2014-01-17 Jeff Law <law@redhat.com>
2963 * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
2964 so that pass_ccp runs first.
2966 2014-01-17 H.J. Lu <hongjiu.lu@intel.com>
2968 * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
2969 (ix86_adjust_cost): Use !TARGET_XXX.
2970 (do_reorder_for_imul): Likewise.
2971 (swap_top_of_ready_list): Likewise.
2972 (ix86_sched_reorder): Likewise.
2974 2014-01-17 H.J. Lu <hongjiu.lu@intel.com>
2976 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2977 PROCESSOR_INTEL. Treat like PROCESSOR_GENERIC.
2978 * config/i386/i386.c (intel_memcpy): New. Duplicate slm_memcpy.
2979 (intel_memset): New. Duplicate slm_memset.
2980 (intel_cost): New. Duplicate slm_cost.
2981 (m_INTEL): New macro.
2982 (processor_target_table): Add "intel".
2983 (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
2984 with PROCESSOR_INTEL for "intel".
2985 (ix86_lea_outperforms): Support PROCESSOR_INTEL. Duplicate
2986 PROCESSOR_SILVERMONT.
2987 (ix86_issue_rate): Likewise.
2988 (ix86_adjust_cost): Likewise.
2989 (ia32_multipass_dfa_lookahead): Likewise.
2990 (swap_top_of_ready_list): Likewise.
2991 (ix86_sched_reorder): Likewise.
2992 (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
2993 instead of TARGET_OPT_AGU.
2994 * config/i386/i386.h (TARGET_INTEL): New.
2995 (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
2996 (processor_type): Add PROCESSOR_INTEL.
2997 * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
2998 Add X86_TUNE_AVOID_LEA_FOR_ADDR.
3000 2014-01-17 Marek Polacek <polacek@redhat.com>
3003 * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
3006 2014-01-17 Richard Biener <rguenther@suse.de>
3008 PR tree-optimization/46590
3009 * opts.c (default_options_table): Add entries for
3010 OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
3011 all enabled at -O1 but not for -Og.
3012 * common.opt (fbranch-count-reg): Remove Init(1).
3013 (fmove-loop-invariants): Likewise.
3014 (ftree-pta): Likewise.
3016 2014-01-17 Jakub Jelinek <jakub@redhat.com>
3018 * config/i386/i386.c (ix86_data_alignment): For compatibility with
3019 (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
3020 decls to at least the GCC 4.8 used alignments.
3023 * tree-nested.c (convert_nonlocal_reference_stmt,
3024 convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
3025 of GIMPLE_BIND stmts, adjust associated decls.
3027 2014-01-17 Richard Biener <rguenther@suse.de>
3029 PR tree-optimization/46590
3030 * vec.h (vec<>::bseach): New member function implementing
3031 binary search according to C89 bsearch.
3032 (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
3033 * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
3034 bitmap pointer again. Make accesses_in_loop a flat array.
3035 (mem_ref_obstack): New global.
3036 (outermost_indep_loop): Adjust for mem_ref->stored changes.
3037 (mark_ref_stored): Likewise.
3038 (ref_indep_loop_p_2): Likewise.
3039 (set_ref_stored_in_loop): New helper function.
3040 (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
3041 (memref_free): Adjust.
3042 (record_mem_ref_loc): Simplify.
3043 (gather_mem_refs_stmt): Adjust.
3044 (sort_locs_in_loop_postorder_cmp): New function.
3045 (analyze_memory_references): Sort accesses_in_loop after
3046 loop postorder number.
3047 (find_ref_loc_in_loop_cmp): New function.
3048 (for_all_locs_in_loop): Find relevant cluster of locs in
3049 accesses_in_loop and iterate without recursion.
3050 (execute_sm): Avoid uninit warning.
3051 (struct ref_always_accessed): Simplify.
3052 (ref_always_accessed::operator ()): Likewise.
3053 (ref_always_accessed_p): Likewise.
3054 (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
3055 loop postorder numbers here.
3056 (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
3059 2014-01-17 Jan Hubicka <hubicka@ucw.cz>
3062 * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
3063 on decls for which assemble_alias has been called.
3065 2014-01-17 Nick Clifton <nickc@redhat.com>
3067 * config/msp430/msp430.opt: (mcpu): New option.
3068 * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
3069 (msp430_option_override): Parse target_cpu. If the MCU name
3070 matches a generic string, clear target_mcu.
3071 (msp430_attr): Allow numeric interrupt values up to 63.
3072 (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
3073 * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
3075 * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
3077 * config/msp430/msp430.md (popm): Use %J rather than %I.
3078 (addsi3): Use msp430_nonimmediate_operand for operand 2.
3079 (addhi_cy_i): Use immediate_operand for operand 2.
3080 * doc/invoke.texi: Document -mcpu option.
3082 2014-01-17 Richard Biener <rguenther@suse.de>
3084 PR rtl-optimization/38518
3085 * df.h (df_analyze_loop): Declare.
3086 * df-core.c: Include cfgloop.h.
3087 (df_analyze_1): Split out main part of df_analyze.
3088 (df_analyze): Adjust.
3089 (loop_inverted_post_order_compute): New function.
3090 (loop_post_order_compute): Likewise.
3091 (df_analyze_loop): New function avoiding whole-function
3093 * loop-invariant.c (find_defs): Use df_analyze_loop.
3094 (find_invariants): Adjust.
3095 * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
3097 2014-01-17 Zhenqiang Chen <zhenqiang.chen@arm.com>
3099 * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
3100 (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
3102 2014-01-16 Ilya Enkovich <ilya.enkovich@intel.com>
3104 * ipa-ref.c (ipa_remove_stmt_references): Fix references
3105 traversal when removing references.
3107 2014-01-16 Jan Hubicka <hubicka@ucw.cz>
3110 * tree.c (get_binfo_at_offset): Look harder for virtual bases.
3112 2014-01-16 Bernd Schmidt <bernds@codesourcery.com>
3115 * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
3116 pushing a reload for an autoinc when we had previously reloaded an
3117 inner part of the address.
3119 2014-01-16 Jakub Jelinek <jakub@redhat.com>
3121 * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
3123 (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
3124 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
3125 when not giving up or versioning for alias only because of
3127 (vect_analyze_data_ref_dependences): Set to true.
3128 * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
3130 (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
3131 LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
3135 * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
3138 * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
3139 operand 0 predicate for gathers, use a new pseudo as subtarget.
3141 2014-01-16 Vladimir Makarov <vmakarov@redhat.com>
3144 * lra-constraints.c (process_alt_operands): Add printing debug info.
3145 Check absence of input/output reloads for matched operands too.
3147 2014-01-16 Vladimir Makarov <vmakarov@redhat.com>
3149 PR rtl-optimization/59835
3150 * ira.c (ira_init_register_move_cost): Increase cost for
3153 2014-01-16 Alan Lawrence <alan.lawrence@arm.com>
3155 * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
3157 2014-01-16 Richard Earnshaw <rearnsha@arm.com>
3160 * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
3161 non-register objects. Use gen_(high/low)part more consistently.
3164 2014-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3167 * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
3168 endian support, remove tests for WORDS_BIG_ENDIAN.
3169 (p8_mfvsrd_3_<mode>): Likewise.
3170 (reload_gpr_from_vsx<mode>): Likewise.
3171 (reload_gpr_from_vsxsf): Likewise.
3172 (p8_mfvsrd_4_disf): Likewise.
3174 2014-01-16 Richard Biener <rguenther@suse.de>
3176 PR rtl-optimization/46590
3177 * lcm.c (compute_antinout_edge): Use postorder iteration.
3178 (compute_laterin): Use inverted postorder iteration.
3180 2014-01-16 Nick Clifton <nickc@redhat.com>
3183 * varasm.c (output_constant): Return the number of bytes actually
3185 (output_constructor_array_range): Update the field size with the
3186 number of bytes emitted by output_constant.
3187 (output_constructor_regular_field): Likewise. Also do not
3188 complain if the total number of bytes emitted is now greater
3189 than the expected fieldpos.
3190 * output.h (output_constant): Update prototype and descriptive comment.
3192 2014-01-16 Marek Polacek <polacek@redhat.com>
3195 * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
3196 it is error_mark_node.
3198 2014-01-15 Uros Bizjak <ubizjak@gmail.com>
3200 * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
3201 VALID_AVX256_REG_OR_OI_MODE.
3203 2014-01-15 Pat Haugen <pthaugen@us.ibm.com>
3205 * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
3206 current procedure should be profiled.
3208 2014-01-15 Andrew Pinski <apinski@cavium.com>
3210 * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
3211 of moving from/to the STACK_REG register class.
3213 2014-01-15 Richard Henderson <rth@redhat.com>
3216 * reginfo.c (global_regs_decl): Globalize.
3217 * rtl.h (global_regs_decl): Declare.
3218 * ira.c (do_reload): Diagnose frame_pointer_needed and it
3219 reserved via global_regs.
3221 2014-01-15 Teresa Johnson <tejohnson@google.com>
3223 * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
3225 2014-01-15 Bill Schmidt <wschmidt@vnet.linux.ibm.com>
3227 * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
3228 and vmulosh rather than call gen_vec_widen_smult_*.
3229 (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
3230 than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
3231 (vec_widen_smult_even_v16qi): Likewise.
3232 (vec_widen_umult_even_v8hi): Likewise.
3233 (vec_widen_smult_even_v8hi): Likewise.
3234 (vec_widen_umult_odd_v16qi): Likewise.
3235 (vec_widen_smult_odd_v16qi): Likewise.
3236 (vec_widen_umult_odd_v8hi): Likewise.
3237 (vec_widen_smult_odd_v8hi): Likewise.
3238 (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
3239 vmuloub rather than call gen_vec_widen_umult_*.
3240 (vec_widen_umult_lo_v16qi): Likewise.
3241 (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
3242 vmulosb rather than call gen_vec_widen_smult_*.
3243 (vec_widen_smult_lo_v16qi): Likewise.
3244 (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
3245 rather than call gen_vec_widen_umult_*.
3246 (vec_widen_umult_lo_v8hi): Likewise.
3247 (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
3248 rather than call gen_vec_widen_smult_*.
3249 (vec_widen_smult_lo_v8hi): Likewise.
3251 2014-01-15 Jeff Law <law@redhat.com>
3253 PR tree-optimization/59747
3254 * ree.c (find_and_remove_re): Properly handle case where a second
3255 eliminated extension requires widening a copy created for elimination
3256 of a prior extension.
3257 (combine_set_extension): Ensure that the number of hard regs needed
3258 for a destination register does not change when we widen it.
3260 2014-01-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
3262 * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
3263 (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
3264 (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
3265 (arm*-*-rtems*): Use t-rtems from existing tmake_file.
3266 (avr-*-rtems*): Likewise.
3267 (bfin*-rtems*): Likewise.
3268 (moxie-*-rtems*): Likewise.
3269 (h8300-*-rtems*): Likewise.
3270 (i[34567]86-*-rtems*): Likewise.
3271 (lm32-*-rtems*): Likewise.
3272 (m32r-*-rtems*): Likewise.
3273 (m68k-*-rtems*): Likewise.
3274 (microblaze*-*-rtems*): Likewise.
3275 (mips*-*-rtems*): Likewise.
3276 (powerpc-*-rtems*): Likewise.
3277 (sh-*-rtems*): Likewise.
3278 (sparc-*-rtems*): Likewise.
3279 (sparc64-*-rtems*): Likewise.
3280 (v850-*-rtems*): Likewise.
3281 (m32c-*-rtems*): Likewise.
3283 2014-01-15 Vladimir Makarov <vmakarov@redhat.com>
3285 PR rtl-optimization/59511
3286 * ira.c (ira_init_register_move_cost): Use memory costs for some
3287 cases of register move cost calculations.
3288 * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
3289 instead of BB frequency.
3290 * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
3291 * lra-assigns.c (find_hard_regno_for): Ditto.
3293 2014-01-15 Richard Biener <rguenther@suse.de>
3295 PR tree-optimization/59822
3296 * tree-vect-stmts.c (hoist_defs_of_uses): New function.
3297 (vectorizable_load): Use it to hoist defs of uses of invariant
3298 loads out of the loop.
3300 2014-01-15 Matthew Gretton-Dann <matthew.gretton-dann@linaro.org>
3301 Kugan Vivekanandarajah <kuganv@linaro.org>
3304 * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
3307 2014-01-15 Richard Biener <rguenther@suse.de>
3309 PR rtl-optimization/59802
3310 * lcm.c (compute_available): Use inverted postorder to seed
3311 the initial worklist.
3313 2014-01-15 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
3316 * config/s390/s390.c (s390_preferred_reload_class): Don't return
3317 ADDR_REGS for invalid symrefs in non-PIC code.
3319 2014-01-15 Jakub Jelinek <jakub@redhat.com>
3322 * builtins.c (determine_block_size): Initialize *probable_max_size
3323 even if len_rtx is CONST_INT.
3325 2014-01-14 Andrew Pinski <apinski@cavium.com>
3327 * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
3328 * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
3329 (cortexa53_tunings): Likewise.
3330 (aarch64_sched_issue_rate): New function.
3331 (TARGET_SCHED_ISSUE_RATE): Define.
3333 2014-01-14 Vladimir Makarov <vmakarov@redhat.com>
3335 * ira-costs.c (find_costs_and_classes): Add missed
3336 ira_init_register_move_cost_if_necessary.
3338 2014-01-14 Vladimir Makarov <vmakarov@redhat.com>
3341 * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
3343 2014-01-14 H.J. Lu <hongjiu.lu@intel.com>
3346 * config/i386/i386.c (type_natural_mode): Add a bool parameter
3347 to indicate if type is used for function return value. Warn ABI
3348 change if the vector mode isn't available for function return value.
3349 (ix86_function_arg_advance): Pass false to type_natural_mode.
3350 (ix86_function_arg): Likewise.
3351 (ix86_gimplify_va_arg): Likewise.
3352 (function_arg_32): Don't warn ABI change.
3353 (ix86_function_value): Pass true to type_natural_mode.
3354 (ix86_return_in_memory): Likewise.
3355 (ix86_struct_value_rtx): Removed.
3356 (TARGET_STRUCT_VALUE_RTX): Likewise.
3358 2014-01-14 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
3360 * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
3361 converting a conditional jump into a conditional return.
3363 2014-01-14 Richard Biener <rguenther@suse.de>
3365 PR tree-optimization/58921
3366 PR tree-optimization/59006
3367 * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
3368 hoisting invariant stmts.
3369 * tree-vect-stmts.c (vectorizable_load): Insert the splat of
3370 invariant loads on the preheader edge if possible.
3372 2014-01-14 Joey Ye <joey.ye@arm.com>
3374 * doc/plugin.texi (Building GCC plugins): Update to C++.
3376 2014-01-14 Kirill Yukhin <kirill.yukhin@intel.com>
3378 * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
3379 (_mm_rcp28_round_ss): Ditto.
3380 (_mm_rsqrt28_round_sd): Ditto.
3381 (_mm_rsqrt28_round_ss): Ditto.
3382 (_mm_rcp28_sd): Ditto.
3383 (_mm_rcp28_ss): Ditto.
3384 (_mm_rsqrt28_sd): Ditto.
3385 (_mm_rsqrt28_ss): Ditto.
3386 * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
3387 * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
3388 * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
3389 (IX86_BUILTIN_RCP28SD): Ditto.
3390 (IX86_BUILTIN_RCP28SS): Ditto.
3391 (IX86_BUILTIN_RSQRT28SD): Ditto.
3392 (IX86_BUILTIN_RSQRT28SS): Ditto.
3393 (bdesc_special_args): Define __builtin_ia32_movntdqa512,
3394 __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
3395 __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
3396 (ix86_expand_special_args_builtin): Expand new FTYPE.
3397 * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
3398 (srcp14<mode>): Make insn unary.
3399 (avx512f_vmscalef<mode><round_name>): Use substed predicate.
3400 (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
3401 (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
3402 (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
3403 (avx512er_exp2<mode><mask_name><round_saeonly_name>):
3404 Fix rounding: make it SAE only.
3405 (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
3407 (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
3409 (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
3410 (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
3411 (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
3412 * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
3413 (round_saeonly_mask_scalar_operand4): Ditto.
3414 (round_saeonly_mask_scalar_op3): Ditto.
3415 (round_saeonly_mask_scalar_op4): Ditto.
3417 2014-01-13 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3419 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3420 Implement -maltivec=be for vec_insert and vec_extract.
3422 2014-01-10 DJ Delorie <dj@redhat.com>
3424 * config/msp430/msp430.md (call_internal): Don't allow memory
3425 references with SP as the base register.
3426 (call_value_internal): Likewise.
3427 * config/msp430/constraints.md (Yc): New. For memory references
3428 that don't use SP as a base register.
3430 * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
3431 "an integer without a # prefix"
3432 * config/msp430/msp430.md (epilogue_helper): Use it.
3434 2014-01-13 Jakub Jelinek <jakub@redhat.com>
3437 * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
3438 AVX512F gather builtins.
3439 * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
3440 on gather decls with INTEGER_TYPE masktype.
3441 (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
3442 directly into the builtin rather than hoisting it before loop.
3444 PR tree-optimization/59387
3445 * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
3446 (scev_const_prop): If folded_casts and type has undefined overflow,
3447 use force_gimple_operand instead of force_gimple_operand_gsi and
3448 for each added stmt if it is assign with
3449 arith_code_with_undefined_signed_overflow, call
3450 rewrite_to_defined_overflow.
3451 * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
3452 gimple-fold.h instead.
3453 (arith_code_with_undefined_signed_overflow,
3454 rewrite_to_defined_overflow): Moved to ...
3455 * gimple-fold.c (arith_code_with_undefined_signed_overflow,
3456 rewrite_to_defined_overflow): ... here. No longer static.
3457 Include gimplify-me.h.
3458 * gimple-fold.h (arith_code_with_undefined_signed_overflow,
3459 rewrite_to_defined_overflow): New prototypes.
3461 2014-01-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3463 * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
3465 2014-01-13 Eric Botcazou <ebotcazou@adacore.com>
3467 * builtins.c (get_object_alignment_2): Minor tweak.
3468 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
3470 2014-01-13 Christian Bruel <christian.bruel@st.com>
3472 * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
3473 optimized non constant lengths.
3475 2014-01-13 Jakub Jelinek <jakub@redhat.com>
3478 * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
3479 load as __atomic_load_N if possible.
3481 2014-01-11 David Edelsohn <dje.gcc@gmail.com>
3483 * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
3485 (rs6000_expand_builtin): Adjust call.
3487 2014-01-11 David Edelsohn <dje.gcc@gmail.com>
3490 * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
3491 * config/rs6000/rs6000.c: Include target-globals.h.
3492 (rs6000_set_current_function): Instead of doing target_reinit
3493 unconditionally, use save_target_globals_default_opts and
3494 restore_target_globals.
3496 * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
3498 * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
3499 (rs6000_expand_builtin): Handle mffs and mtfsf.
3500 (rs6000_init_builtins): Define mffs and mtfsf.
3501 * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
3502 (rs6000_mffs): New pattern.
3503 (rs6000_mtfsf): New pattern.
3505 2014-01-11 Bin Cheng <bin.cheng@arm.com>
3507 * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
3508 Start narrowing with START. Apply candidate-use pair
3509 and check overall cost in narrowing.
3510 (iv_ca_prune): Pass new argument.
3512 2014-01-10 Jeff Law <law@redhat.com>
3515 * ree.c (combine_reaching_defs): Ensure the defining statement
3516 occurs before the extension when optimizing extensions with
3517 different source and destination hard registers.
3519 2014-01-10 Jan Hubicka <hubicka@ucw.cz>
3522 * ipa-devirt.c (build_type_inheritance_graph): Also add types of
3523 vtables into the type inheritance graph.
3525 2014-01-10 Jakub Jelinek <jakub@redhat.com>
3527 PR rtl-optimization/59754
3528 * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
3529 modes in the REGNO != REGNO case.
3531 2014-01-10 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3533 * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
3535 2014-01-10 Jakub Jelinek <jakub@redhat.com>
3537 PR tree-optimization/59745
3538 * tree-predcom.c (tree_predictive_commoning_loop): Call
3539 free_affine_expand_cache if giving up because components is NULL.
3541 * target-globals.c (save_target_globals): Allocate < 4KB structs using
3542 GC in payload of target_globals struct instead of allocating them on
3543 the heap and the larger structs separately using GC.
3544 * target-globals.h (struct target_globals): Make regs, hard_regs,
3545 reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
3546 of GTY((skip)) and change type to void *.
3547 (reset_target_globals): Cast loads from those fields to corresponding
3550 2014-01-10 Steve Ellcey <sellcey@mips.com>
3553 * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
3554 gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
3555 tree-ssanames.h, print-tree.h, varasm.h, and context.h.
3557 2014-01-10 Richard Earnshaw <rearnsha@arm.com>
3560 * aarch64-modes.def (CC_Zmode): New flags mode.
3561 * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
3562 represents an equality.
3563 (aarch64_get_condition_code): Handle CC_Zmode.
3564 * aarch64.md (compare_neg<mode>): Restrict to equality operations.
3566 2014-01-10 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
3568 * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
3569 extraction in good case.
3571 2014-01-10 Richard Biener <rguenther@suse.de>
3573 PR tree-optimization/59374
3574 * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
3575 checking after SLP discovery. Mark stmts not participating
3576 in any SLP instance properly.
3578 2014-01-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3580 * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
3581 when handling a SET rtx.
3583 2014-01-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3585 * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
3586 (cortex-a57): Likewise.
3587 (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
3589 2014-01-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3591 * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
3592 non-iwmmxt builtins.
3594 2014-01-10 Jan Hubicka <hubicka@ucw.cz>
3598 * ipa-devirt.c record_target_from_binfo): Take as argument
3599 stack of binfos and lookup matching one for virtual inheritance.
3600 (possible_polymorphic_call_targets_1): Update.
3602 2014-01-10 Huacai Chen <chenhc@lemote.com>
3604 * config/mips/driver-native.c (host_detect_local_cpu): Handle new
3605 kernel strings for Loongson-2E/2F/3A.
3607 2014-01-10 Jakub Jelinek <jakub@redhat.com>
3610 * tree-vect-data-refs.c (vect_analyze_data_refs): Check
3611 is_gimple_call before calling gimple_call_internal_p.
3613 2014-01-09 Steve Ellcey <sellcey@mips.com>
3615 * Makefile.in (TREE_FLOW_H): Remove.
3616 (TREE_SSA_H): Add file names from tree-flow.h.
3617 * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
3618 * tree.h: Remove tree-flow.h reference.
3619 * hash-table.h: Remove tree-flow.h reference.
3620 * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
3621 reference with tree-ssa-loop.h.
3623 2014-01-09 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3625 * doc/invoke.texi: Add -maltivec={be,le} options, and document
3626 default element-order behavior for -maltivec.
3627 * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
3628 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
3629 that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
3630 when targeting big endian, at least for now.
3631 * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
3633 2014-01-09 Jakub Jelinek <jakub@redhat.com>
3636 * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
3637 var satisfies use_register_for_decl, just take into account type
3638 alignment, rather than decl alignment.
3640 PR tree-optimization/59622
3641 * gimple-fold.c (gimple_fold_call): Fix a typo in message. For
3642 __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
3643 __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
3644 Don't devirtualize for inplace at all. For targets.length () == 1,
3645 if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
3647 2014-01-09 H.J. Lu <hongjiu.lu@intel.com>
3649 * config/i386/i386.md (cpu): Remove the unused btver1.
3651 2014-01-09 H.J. Lu <hongjiu.lu@intel.com>
3653 * gdbasan.in: Put a breakpoint on __sanitizer::Report.
3655 2014-01-09 Jakub Jelinek <jakub@redhat.com>
3658 * tree-core.h (struct target_globals): New forward declaration.
3659 (struct tree_target_option): Add globals field.
3660 * tree.h (TREE_TARGET_GLOBALS): Define.
3661 (prepare_target_option_nodes_for_pch): New prototype.
3662 * target-globals.h (struct target_globals): Define even if
3664 * tree.c (prepare_target_option_node_for_pch,
3665 prepare_target_option_nodes_for_pch): New functions.
3666 * config/i386/i386.h (SWITCHABLE_TARGET): Define.
3667 * config/i386/i386.c: Include target-globals.h.
3668 (ix86_set_current_function): Instead of doing target_reinit
3669 unconditionally, use save_target_globals_default_opts and
3670 restore_target_globals.
3672 2014-01-09 Richard Biener <rguenther@suse.de>
3674 PR tree-optimization/59715
3675 * tree-cfg.h (split_critical_edges): Declare.
3676 * tree-cfg.c (split_critical_edges): Export.
3677 * tree-ssa-sink.c (execute_sink_code): Split critical edges.
3679 2014-01-09 Max Ostapenko <m.ostapenko@partner.samsung.com>
3681 * cfgexpand.c (expand_stack_vars): Optionally disable
3682 asan stack protection.
3683 (expand_used_vars): Likewise.
3684 (partition_stack_vars): Likewise.
3685 * asan.c (asan_emit_stack_protection): Optionally disable
3686 after return stack usage.
3687 (instrument_derefs): Optionally disable memory access instrumentation.
3688 (instrument_builtin_call): Likewise.
3689 (instrument_strlen_call): Likewise.
3690 (asan_protect_global): Optionally disable global variables protection.
3691 * doc/invoke.texi: Added doc for new options.
3692 * params.def: Added new options.
3693 * params.h: Likewise.
3695 2014-01-09 Jakub Jelinek <jakub@redhat.com>
3697 PR rtl-optimization/59724
3698 * ifcvt.c (cond_exec_process_if_block): Don't call
3699 flow_find_head_matching_sequence with 0 longest_match.
3700 * cfgcleanup.c (flow_find_head_matching_sequence): Count even
3701 non-active insns if !stop_after.
3702 (try_head_merge_bb): Revert 2014-01-07 changes.
3704 2014-01-08 Jeff Law <law@redhat.com>
3706 * ree.c (get_sub_rtx): New function, extracted from...
3707 (merge_def_and_ext): Here.
3708 (combine_reaching_defs): Use get_sub_rtx.
3710 2014-01-08 Eric Botcazou <ebotcazou@adacore.com>
3712 * cgraph.h (varpool_variable_node): Do not choke on null node.
3714 2014-01-08 Catherine Moore <clm@codesourcery.com>
3716 * config/mips/mips.md (simple_return): Attempt to use JRC
3718 * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
3720 2014-01-08 Richard Sandiford <rdsandiford@googlemail.com>
3722 PR rtl-optimization/59137
3723 * reorg.c (steal_delay_list_from_target): Call update_block for
3725 (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
3727 2014-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3729 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3730 two duplicate entries.
3732 2014-01-08 Richard Sandiford <rdsandiford@googlemail.com>
3735 2012-10-07 Richard Sandiford <rdsandiford@googlemail.com>
3737 * config/mips/mips.c (mips_truncated_op_cost): New function.
3738 (mips_rtx_costs): Adjust test for BADDU.
3739 * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
3741 2012-10-02 Richard Sandiford <rdsandiford@googlemail.com>
3743 * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
3744 (*baddu_si): ...this new pattern.
3746 2014-01-08 Jakub Jelinek <jakub@redhat.com>
3749 * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
3751 2014-01-08 Bernd Edlinger <bernd.edlinger@hotmail.de>
3754 * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
3756 (expand_expr, expand_normal): Adjust.
3757 * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
3758 inner_reference_p. Use inner_reference_p to expand inner references.
3759 (store_expr): Adjust.
3760 * cfgexpand.c (expand_call_stmt): Adjust.
3762 2014-01-08 Rong Xu <xur@google.com>
3764 * gcov-io.c (gcov_var): Move from gcov-io.h.
3765 (gcov_position): Ditto.
3766 (gcov_is_error): Ditto.
3767 (gcov_rewrite): Ditto.
3768 * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
3769 only part to libgcc/libgcov.h.
3771 2014-01-08 Marek Polacek <polacek@redhat.com>
3774 * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
3776 2014-01-08 Marek Polacek <polacek@redhat.com>
3779 * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
3781 2014-01-08 Jakub Jelinek <jakub@redhat.com>
3783 PR rtl-optimization/59649
3784 * stor-layout.c (get_mode_bounds): For BImode return
3785 0 and STORE_FLAG_VALUE.
3787 2014-01-08 Richard Biener <rguenther@suse.de>
3790 * gimple.h (is_gimple_builtin_call): Remove.
3791 (gimple_builtin_call_types_compatible_p): New.
3792 (gimple_call_builtin_p): New overload.
3793 * gimple.c (is_gimple_builtin_call): Remove.
3794 (validate_call): Rename to ...
3795 (gimple_builtin_call_types_compatible_p): ... this and export. Also
3797 (validate_type): New static function.
3798 (gimple_call_builtin_p): New overload and adjust.
3799 * gimple-fold.c (gimple_fold_builtin): Fold the return value.
3800 (gimple_fold_call): Likewise. Use gimple_call_builtin_p.
3801 (gimple_fold_stmt_to_constant_1): Likewise.
3802 * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
3804 2014-01-08 Richard Biener <rguenther@suse.de>
3807 * gimplify.c (gimplify_expr): Gimplify register-register type
3808 VIEW_CONVERT_EXPRs to separate stmts.
3810 2014-01-07 Jeff Law <law@redhat.com>
3813 * ree.c (combine_set_extension): Handle case where source
3814 and destination registers in an extension insn are different.
3815 (combine_reaching_defs): Allow source and destination registers
3816 in extension to be different under limited circumstances.
3817 (add_removable_extension): Remove restriction that the
3818 source and destination registers in the extension are the same.
3819 (find_and_remove_re): Emit a copy from the extension's
3820 destination to its source after the defining insn if
3821 the source and destination registers are different.
3824 * ifcvt.c (merge_if_block): If we are merging a block with more than
3825 one successor with a block with no successors, remove any BARRIER
3826 after the second block.
3828 2014-01-07 Dan Xio Qiang <ziyan01@163.com>
3830 * hw-doloop.c (reorg_loops): Release the bitmap obstack.
3832 2014-01-07 John David Anglin <danglin@gcc.gnu.org>
3835 * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
3836 for 14-bit register offsets when INT14_OK_STRICT is false.
3838 2014-01-07 Roland Stigge <stigge@antcom.de>
3839 Michael Meissner <meissner@linux.vnet.ibm.com>
3842 * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
3843 Only check TFmode for SPE constants. Don't check TImode or TDmode.
3845 2014-01-07 James Greenhalgh <james.greenhalgh@arm.com>
3847 * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
3850 2014-01-07 Yufeng Zhang <yufeng.zhang@arm.com>
3852 * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
3853 with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
3854 rtx is const0_rtx or not.
3856 2014-01-07 Richard Sandiford <rdsandiford@googlemail.com>
3859 * target-globals.c (save_target_globals): Remove this_fn_optab
3861 * toplev.c: Include optabs.h.
3862 (target_reinit): Temporarily restore the global options if another
3863 set of options are in force.
3865 2014-01-07 Jakub Jelinek <jakub@redhat.com>
3867 PR rtl-optimization/58668
3868 * cfgcleanup.c (flow_find_cross_jump): Don't count
3869 any jumps if dir_p is NULL. Remove p1 variable, use active_insn_p
3870 to determine what is counted.
3871 (flow_find_head_matching_sequence): Use active_insn_p to determine
3873 (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
3875 * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
3876 determine what is counted.
3878 PR tree-optimization/59643
3879 * tree-predcom.c (split_data_refs_to_components): If one dr is
3880 read and one write, determine_offset fails and the write isn't
3881 in the bad component, just put the read into the bad component.
3883 2014-01-07 Mike Stump <mikestump@comcast.net>
3884 Jakub Jelinek <jakub@redhat.com>
3887 * tree-core.h (struct tree_optimization_option): Change optabs
3888 type from unsigned char * to void *.
3889 * optabs.c (init_tree_optimization_optabs): Adjust
3890 TREE_OPTIMIZATION_OPTABS initialization.
3892 2014-01-06 Jakub Jelinek <jakub@redhat.com>
3895 * config/i386/i386.h (struct machine_function): Add
3896 no_drap_save_restore field.
3897 * config/i386/i386.c (ix86_save_reg): Use
3898 !cfun->machine->no_drap_save_restore instead of
3899 crtl->stack_realign_needed.
3900 (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
3901 this function clears frame_pointer_needed. Set
3902 cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
3903 and DRAP reg is needed.
3905 2014-01-06 Marek Polacek <polacek@redhat.com>
3908 * doc/implement-c.texi: Mention that other integer types are
3909 permitted as bit-field types in strictly conforming mode.
3911 2014-01-06 Felix Yang <fei.yang0953@gmail.com>
3913 * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
3916 2014-01-06 Richard Earnshaw <rearnsha@arm.com>
3918 * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
3920 2014-01-06 Martin Jambor <mjambor@suse.cz>
3923 * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
3925 * ipa-prop.c (ipa_print_node_params): Fix indentation.
3927 2014-01-06 Eric Botcazou <ebotcazou@adacore.com>
3931 * var-tracking.c (add_stores): Preserve the value of the source even if
3932 we don't record the store.
3934 2014-01-06 Terry Guo <terry.guo@arm.com>
3936 * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
3938 2014-01-05 Iain Sandoe <iain@codesourcery.com>
3941 * config/darwin.c (darwin_function_section): Adjust return values to
3942 correspond to optimisation changes made in r206070.
3944 2014-01-05 Uros Bizjak <ubizjak@gmail.com>
3946 * config/i386/i386.c (ix86_data_alignment): Calculate max_align
3947 from prefetch_block tune setting.
3948 (nocona_cost): Correct size of prefetch block to 64.
3950 2014-01-04 Eric Botcazou <ebotcazou@adacore.com>
3952 * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
3953 (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
3954 used to save the static chain register in the computation of the offset
3955 from which the FP registers need to be restored.
3957 2014-01-04 Jakub Jelinek <jakub@redhat.com>
3959 PR tree-optimization/59519
3960 * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
3961 ICE if get_current_def (current_new_name) is already non-NULL, as long
3962 as it is a phi result of some other phi in *new_exit_bb that has
3965 * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
3966 or vmovdqu* for misaligned_operand.
3967 (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
3968 <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
3969 * config/i386/i386.c (ix86_expand_special_args_builtin): Set
3970 aligned_mem for AVX512F masked aligned load and store builtins and for
3973 2014-01-03 Bingfeng Mei <bmei@broadcom.com>
3975 PR tree-optimization/59651
3976 * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
3977 Address range for negative step should be added by TYPE_SIZE_UNIT.
3979 2014-01-03 Andreas Schwab <schwab@linux-m68k.org>
3981 * config/m68k/m68k.c (handle_move_double): Handle pushes with
3982 overlapping registers also for registers other than the stack pointer.
3984 2014-01-03 Marek Polacek <polacek@redhat.com>
3987 * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
3990 2014-01-03 Jakub Jelinek <jakub@redhat.com>
3993 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
3996 * config/i386/i386.md (MODE_SIZE): New mode attribute.
3997 (push splitter): Use <P:MODE_SIZE> instead of
3998 GET_MODE_SIZE (<P:MODE>mode).
3999 (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
4000 (mov -1, reg peephole2): Likewise.
4001 * config/i386/sse.md (*mov<mode>_internal,
4002 <sse>_storeu<ssemodesuffix><avxsizesuffix>,
4003 <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
4004 *<code><mode>3, *andnot<mode>3<mask_name>,
4005 <mask_codefor><code><mode>3<mask_name>): Likewise.
4006 * config/i386/subst.md (mask_mode512bit_condition,
4007 sd_mask_mode512bit_condition): Likewise.
4009 2014-01-02 Xinliang David Li <davidxl@google.com>
4011 PR tree-optimization/59303
4012 * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
4013 (dump_predicates): Better output format.
4014 (pred_equal_p): New function.
4015 (is_neq_relop_p): Ditto.
4016 (is_neq_zero_form_p): Ditto.
4017 (pred_expr_equal_p): Ditto.
4018 (pred_neg_p): Ditto.
4019 (simplify_pred): Ditto.
4020 (simplify_preds_2): Ditto.
4021 (simplify_preds_3): Ditto.
4022 (simplify_preds_4): Ditto.
4023 (simplify_preds): Ditto.
4025 (push_to_worklist): Ditto.
4026 (get_pred_info_from_cmp): Ditto.
4027 (is_degenerated_phi): Ditto.
4028 (normalize_one_pred_1): Ditto.
4029 (normalize_one_pred): Ditto.
4030 (normalize_one_pred_chain): Ditto.
4031 (normalize_preds): Ditto.
4032 (normalize_cond_1): Remove function.
4033 (normalize_cond): Ditto.
4034 (is_gcond_subset_of): Ditto.
4035 (is_subset_of_any): Ditto.
4036 (is_or_set_subset_of): Ditto.
4037 (is_and_set_subset_of): Ditto.
4038 (is_norm_cond_subset_of): Ditto.
4039 (pred_chain_length_cmp): Ditto.
4040 (convert_control_dep_chain_into_preds): Type change.
4041 (find_predicates): Ditto.
4042 (find_def_preds): Ditto.
4043 (destroy_predicates_vecs): Ditto.
4044 (find_matching_predicates_in_rest_chains): Ditto.
4045 (use_pred_not_overlap_with_undef_path_pred): Ditto.
4046 (is_pred_expr_subset): Ditto.
4047 (is_pred_chain_subset_of): Ditto.
4048 (is_included_in): Ditto.
4049 (is_superset_of): Ditto.
4051 2014-01-02 Richard Sandiford <rdsandiford@googlemail.com>
4053 Update copyright years.
4055 2014-01-02 Richard Sandiford <rdsandiford@googlemail.com>
4057 * common/config/arc/arc-common.c, config/arc/arc-modes.def,
4058 config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
4059 config/arc/arc.md, config/arc/arc.opt,
4060 config/arm/arm_neon_builtins.def, config/arm/crypto.def,
4061 config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
4062 config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
4063 config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
4064 config/linux-protos.h, config/linux.c, config/winnt-c.c,
4065 diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
4066 vtable-verify.c, vtable-verify.h: Use the standard form for the
4069 2014-01-02 Tobias Burnus <burnus@net-b.de>
4071 * gcc.c (process_command): Update copyright notice dates.
4072 * gcov-dump.c: Ditto.
4074 * doc/cpp.texi: Bump @copying's copyright year.
4075 * doc/cppinternals.texi: Ditto.
4076 * doc/gcc.texi: Ditto.
4077 * doc/gccint.texi: Ditto.
4078 * doc/gcov.texi: Ditto.
4079 * doc/install.texi: Ditto.
4080 * doc/invoke.texi: Ditto.
4082 2014-01-01 Jan-Benedict Glaw <jbglaw@lug-owl.de>
4084 * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
4086 2014-01-01 Jakub Jelinek <jakub@redhat.com>
4088 * config/i386/sse.md (*mov<mode>_internal): Guard
4089 EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
4091 PR rtl-optimization/59647
4092 * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
4093 new_rtx into UNSIGNED_FLOAT rtxes.
4095 Copyright (C) 2014 Free Software Foundation, Inc.
4097 Copying and distribution of this file, with or without modification,
4098 are permitted in any medium without royalty provided the copyright
4099 notice and this notice are preserved.