1 2014-03-21 Richard Biener <rguenther@suse.de>
3 PR tree-optimization/60577
4 * tree-core.h (struct tree_base): Document nothrow_flag use
5 in VAR_DECL_NONALIASED.
6 * tree.h (VAR_DECL_NONALIASED): New.
7 (may_be_aliased): Adjust.
8 * coverage.c (build_var): Set VAR_DECL_NONALIASED.
10 2014-03-20 Eric Botcazou <ebotcazou@adacore.com>
12 * expr.c (expand_expr_real_1): Remove outdated comment.
14 2014-03-20 Jakub Jelinek <jakub@redhat.com>
17 * ira.c (adjust_cleared_regs): Call copy_rtx on
18 *reg_equiv[REGNO (loc)].src_p before passing it to
19 simplify_replace_fn_rtx.
22 * config/i386/i386.c (x86_output_mi_thunk): Surround UNSPEC_GOT
23 into CONST, put pic register as first operand of PLUS. Use
24 gen_const_mem for both 32-bit and 64-bit PIC got loads.
26 2014-03-20 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
28 * config/aarch64/aarch64.c (MEMORY_MOVE_COST): Delete.
30 2014-03-20 Eric Botcazou <ebotcazou@adacore.com>
32 * config/sparc/sparc.c (sparc_do_work_around_errata): Implement work
33 around for store forwarding issue in the FPU on the UT699.
34 * config/sparc/sparc.md (in_branch_delay): Return false for single FP
35 loads and operations if -mfix-ut699 is specified.
36 (divtf3_hq): Tweak attribute.
37 (sqrttf2_hq): Likewise.
39 2014-03-20 Eric Botcazou <ebotcazou@adacore.com>
41 * calls.c (store_one_arg): Remove incorrect const qualification on the
42 type of the temporary.
43 * cfgexpand.c (expand_return): Likewise.
44 * expr.c (expand_constructor): Likewise.
45 (expand_expr_real_1): Likewise.
47 2014-03-20 Zhenqiang Chen <zhenqiang.chen@linaro.org>
49 * config/arm/arm.c (arm_dwarf_register_span): Update the element number
52 2014-03-19 Kaz Kojima <kkojima@gcc.gnu.org>
55 * config/sh/sh.md (udivsi3_i1): Clobber R1 register.
57 2014-03-19 James Greenhalgh <james.greenhalgh@arm.com>
59 * config/arm/aarch-common-protos.h
60 (alu_cost_table): Fix spelling of "extend".
61 * config/arm/arm.c (arm_new_rtx_costs): Fix spelling of "extend".
63 2014-03-19 Richard Biener <rguenther@suse.de>
66 * tree-core.h (tree_type_common): Re-order pointer members
67 to reduce recursion depth during GC walks.
69 2014-03-19 Marek Polacek <polacek@redhat.com>
72 * ubsan.c (ubsan_type_descriptor): Check that DECL_NAME is nonnull
75 2014-03-19 Richard Biener <rguenther@suse.de>
78 * lto-streamer-in.c (input_function): In WPA stage do not drop
81 2014-03-19 Jakub Jelinek <jakub@redhat.com>
83 PR tree-optimization/60559
84 * vectorizable_mask_load_store): Replace scalar MASK_LOAD
85 with build_zero_cst assignment.
87 2014-03-18 Kai Tietz <ktietz@redhat.com>
89 PR rtl-optimization/56356
90 * sdbout.c (sdbout_parms): Verify that parms'
91 incoming argument is valid.
92 (sdbout_reg_parms): Likewise.
94 2014-03-18 Richard Henderson <rth@redhat.com>
97 * config/i386/i386.md (*float<SWI48x><MODEF>2_i387): Move down to
98 be shadowed by *float<SWI48><MODEF>2_sse. Test X87_ENABLE_FLOAT.
99 (*float<SWI48><MODEF>2_sse): Check X87_ENABLE_FLOAT for alternative 0.
101 2014-03-18 Basile Starynkevitch <basile@starynkevitch.net>
103 * plugin.def: Improve comment for PLUGIN_INCLUDE_FILE.
104 * doc/plugins.texi (Plugin callbacks): Mention
106 Italicize plugin event names in description. Explain that
107 PLUGIN_PRAGMAS has no sense for lto1. Explain
109 Remind that no GCC functions should be called after
111 Explain what pragmas with expansion are.
113 2014-03-18 Martin Liska <mliska@suse.cz>
115 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Added case when
116 gimple call statement is update.
117 * gimple-fold.c (gimple_fold_call): Changed order for GIMPLE_ASSIGN and
118 GIMPLE_CALL, where gsi iterator still points to GIMPLE CALL.
120 2014-03-18 Jakub Jelinek <jakub@redhat.com>
123 * ubsan.c (ubsan_instrument_unreachable): Call
124 initialize_sanitizer_builtins.
125 (ubsan_pass): Likewise.
128 * ubsan.c (ubsan_type_descriptor, ubsan_create_data): Call
129 varpool_finalize_decl instead of rest_of_decl_compilation.
131 2014-03-18 Richard Biener <rguenther@suse.de>
133 * df-problems.c (df_rd_confluence_n): Avoid bitmap_copy
134 by using bitmap_and_compl instead of bitmap_and_compl_into.
135 (df_rd_transfer_function): Likewise.
137 2014-03-18 Richard Biener <rguenther@suse.de>
139 * doc/lto.texi (fresolution): Fix typo.
141 2014-03-18 Richard Biener <rguenther@suse.de>
143 * doc/invoke.texi (flto): Update for changes in 4.9.
145 2014-03-18 Richard Biener <rguenther@suse.de>
147 * doc/loop.texi: Remove section on the removed lambda framework.
148 Update loop docs with recent changes in preserving loop structure.
150 2014-03-18 Richard Biener <rguenther@suse.de>
152 * doc/lto.texi (-fresolution): Document.
154 2014-03-18 Richard Biener <rguenther@suse.de>
156 * doc/contrib.texi: Adjust my name.
158 2014-03-18 Jakub Jelinek <jakub@redhat.com>
161 * internal-fn.c: Include diagnostic-core.h.
162 (expand_BUILTIN_EXPECT): New function.
163 * gimplify.c (gimplify_call_expr): Use false instead of FALSE.
164 (gimplify_modify_expr): Gimplify 3 argument __builtin_expect into
165 IFN_BUILTIN_EXPECT call instead of __builtin_expect builtin call.
166 * ipa-inline-analysis.c (find_foldable_builtin_expect): Handle
168 * predict.c (expr_expected_value_1): Handle IFN_BUILTIN_EXPECT.
169 Revert 3 argument __builtin_expect code.
170 (strip_predict_hints): Handle IFN_BUILTIN_EXPECT.
171 * gimple-fold.c (gimple_fold_call): Likewise.
172 * tree.h (fold_builtin_expect): New prototype.
173 * builtins.c (build_builtin_expect_predicate): Add predictor
174 argument, if non-NULL, create 3 argument __builtin_expect.
175 (fold_builtin_expect): No longer static. Add ARG2 argument,
176 pass it through to build_builtin_expect_predicate.
177 (fold_builtin_2): Adjust caller.
178 (fold_builtin_3): Handle BUILT_IN_EXPECT.
179 * internal-fn.def (BUILTIN_EXPECT): New.
181 2014-03-18 Tobias Burnus <burnus@net-b.de>
184 * predict.def (PRED_FORTRAN_OVERFLOW, PRED_FORTRAN_FAIL_ALLOC,
185 PRED_FORTRAN_FAIL_IO, PRED_FORTRAN_WARN_ONCE, PRED_FORTRAN_SIZE_ZERO,
186 PRED_FORTRAN_INVALID_BOUND, PRED_FORTRAN_ABSENT_DUMMY): Add.
188 2014-03-18 Jan Hubicka <hubicka@ucw.cz>
191 * predict.c (combine_predictions_for_bb): Fix up formatting.
192 (expr_expected_value_1, expr_expected_value): Add predictor argument,
193 fill what it points to if non-NULL.
194 (tree_predict_by_opcode): Adjust caller, use the predictor.
195 * predict.def (PRED_COMPARE_AND_SWAP): Add.
197 2014-03-18 Eric Botcazou <ebotcazou@adacore.com>
199 * config/sparc/sparc.c (sparc_do_work_around_errata): Speed up and use
200 proper constant for the store mode.
202 2014-03-18 Ilya Enkovich <ilya.enkovich@intel.com>
204 * symtab.c (change_decl_assembler_name): Fix transparent alias
207 2014-03-16 Renlin Li <Renlin.Li@arm.com>
209 * config/aarch64/aarch64.c: Correct the comments about the
210 aarch64 stack layout.
212 2014-03-18 Thomas Schwinge <thomas@codesourcery.com>
214 * omp-low.c (lower_rec_input_clauses) <build_omp_barrier>: Restore
215 check for GF_OMP_FOR_KIND_FOR.
217 2013-03-18 Kirill Yukhin <kirill.yukhin@intel.com>
219 * config/i386/i386.h (ADDITIONAL_REGISTER_NAMES): Add
220 ymm and zmm register names.
222 2014-03-17 Jakub Jelinek <jakub@redhat.com>
225 * config/i386/i386.c (ix86_expand_epilogue): Adjust REG_CFA_ADJUST_CFA
226 note creation for the 2010-08-31 changes.
228 2014-03-17 Marek Polacek <polacek@redhat.com>
231 * omp-low.c (omp_max_vf): Treat -fno-tree-loop-optimize the same
232 as -fno-tree-loop-vectorize.
233 (expand_omp_simd): Likewise.
235 2014-03-15 Eric Botcazou <ebotcazou@adacore.com>
237 * config/sparc/sparc-protos.h (tls_call_delay): Delete.
238 (eligible_for_call_delay): New prototype.
239 * config/sparc/sparc.c (tls_call_delay): Rename into...
240 (eligible_for_call_delay): ...this. Return false if the instruction
241 cannot be put in the delay slot of a branch.
242 (eligible_for_restore_insn): Simplify.
243 (eligible_for_return_delay): Return false if the instruction cannot be
244 put in the delay slot of a branch and simplify.
245 (eligible_for_sibcall_delay): Return false if the instruction cannot be
246 put in the delay slot of a branch.
247 * config/sparc/sparc.md (fix_ut699): New attribute.
248 (tls_call_delay): Delete.
249 (in_call_delay): Reimplement.
250 (eligible_for_sibcall_delay): Rename into...
251 (in_sibcall_delay): ...this.
252 (eligible_for_return_delay): Rename into...
253 (in_return_delay): ...this.
254 (in_branch_delay): Reimplement.
255 (in_uncond_branch_delay): Delete.
256 (in_annul_branch_delay): Delete.
258 2014-03-14 Richard Henderson <rth@redhat.com>
261 * config/i386/i386.md (floathi<X87MODEF>2): Delete expander; rename
262 define_insn from *floathi<X87MODEF>2_i387; allow nonimmediate_operand.
263 (*floathi<X87MODEF>2_i387_with_temp): Remove.
264 (floathi splitters): Remove.
265 (float<SWI48x>xf2): New pattern.
266 (float<SWI48><MODEF>2): Rename from float<SWI48x><X87MODEF>2. Drop
267 code that tried to handle DImode for 32-bit, but which was excluded
268 by the pattern's condition. Drop allocation of stack temporary.
269 (*floatsi<MODEF>2_vector_mixed_with_temp): Remove.
270 (*float<SWI48><MODEF>2_mixed_with_temp): Remove.
271 (*float<SWI48><MODEF>2_mixed_interunit): Remove.
272 (*float<SWI48><MODEF>2_mixed_nointerunit): Remove.
273 (*floatsi<MODEF>2_vector_sse_with_temp): Remove.
274 (*float<SWI48><MODEF>2_sse_with_temp): Remove.
275 (*float<SWI48><MODEF>2_sse_interunit): Remove.
276 (*float<SWI48><MODEF>2_sse_nointerunit): Remove.
277 (*float<SWI48x><X87MODEF>2_i387_with_temp): Remove.
278 (*float<SWI48x><X87MODEF>2_i387): Remove.
279 (all float _with_temp splitters): Remove.
280 (*float<SWI48x><MODEF>2_i387): New pattern.
281 (*float<SWI48><MODEF>2_sse): New pattern.
282 (float TARGET_USE_VECTOR_CONVERTS splitters): Merge them.
283 (float TARGET_SSE_PARTIAL_REG_DEPENDENCY splitters): Merge them.
285 2014-03-14 Jakub Jelinek <jakub@redhat.com>
286 Marek Polacek <polacek@redhat.com>
289 * common.opt (dump_base_name_prefixed): New Variable.
290 * opts.c (finish_options): Don't prepend directory to x_dump_base_name
291 if x_dump_base_name_prefixed is already set, set it at the end.
293 2014-03-14 Vladimir Makarov <vmakarov@redhat.com>
295 PR rtl-optimization/60508
296 * lra-constraints.c (get_reload_reg): Add new parameter
298 (process_addr_reg, simplify_operand_subreg, curr_insn_transform):
299 Pass the new parameter values.
301 2014-03-14 Richard Biener <rguenther@suse.de>
303 * common.opt: Revert unintented changes from r205065.
306 2014-03-14 Richard Biener <rguenther@suse.de>
309 * cfghooks.c (split_block): Properly adjust all loops the
310 block was a latch of.
312 2014-03-14 Martin Jambor <mjambor@suse.cz>
315 * ipa-prop.c (ipa_modify_call_arguments): Fix iteration condition
318 2014-03-14 Georg-Johann Lay <avr@gjlay.de>
321 * config/avr/avr.c (avr_set_current_function): Pass function name
322 through default_strip_name_encoding before sanity checking instead
323 of skipping the first char of the assembler name.
325 2014-03-13 Richard Henderson <rth@redhat.com>
328 * config/i386/i386.c (ix86_split_fp_branch): Remove pushed argument.
329 (ix86_force_to_memory, ix86_free_from_memory): Remove.
330 * config/i386/i386-protos.h: Likewise.
331 * config/i386/i386.md (floathi<X87MODEF>2): Use assign_386_stack_local
332 in the expander instead of a splitter.
333 (float<SWI48x><X87MODEF>2): Use assign_386_stack_local if there is
334 any possibility of requiring a memory.
335 (*floatsi<MODEF>2_vector_mixed): Remove, and the splitters.
336 (*floatsi<MODEF>2_vector_sse): Remove, and the splitters.
337 (fp branch splitters): Update for ix86_split_fp_branch.
338 (*jcc<X87MODEF>_<SWI24>_i387): Remove r/f alternative.
339 (*jcc<X87MODEF>_<SWI24>_r_i387): Likewise.
340 (splitter for jcc<X87MODEF>_<SWI24>_i387 r/f): Remove.
341 (*fop_<MODEF>_2_i387): Remove f/r alternative.
342 (*fop_<MODEF>_3_i387): Likewise.
343 (*fop_xf_2_i387, *fop_xf_3_i387): Likewise.
344 (splitters for the fop_* register patterns): Remove.
345 (fscalexf4_i387): Rename from *fscalexf4_i387.
346 (ldexpxf3): Use gen_floatsixf2 and gen_fscalexf4_i387.
348 2014-03-13 Jakub Jelinek <jakub@redhat.com>
350 PR tree-optimization/59779
351 * tree-dfa.c (get_ref_base_and_extent): Use double_int
352 type for bitsize and maxsize instead of HOST_WIDE_INT.
354 2014-03-13 Steven Bosscher <steven@gcc.gnu.org>
356 PR rtl-optimization/57320
357 * function.c (rest_of_handle_thread_prologue_and_epilogue): Cleanup
358 the CFG after thread_prologue_and_epilogue_insns.
360 2014-03-13 Vladimir Makarov <vmakarov@redhat.com>
362 PR rtl-optimization/57189
363 * lra-constraints.c (process_alt_operands): Disfavor spilling
366 2014-03-13 Cesar Philippidis <cesar@codesourcery.com>
368 * lto-wrapper.c (maybe_unlink_file): Suppress diagnostic
371 2014-03-13 Jakub Jelinek <jakub@redhat.com>
373 PR tree-optimization/59025
375 * tree-ssa-reassoc.c (sort_by_operand_rank): For SSA_NAMEs with the
376 same rank, sort by bb_rank and gimple_uid of SSA_NAME_DEF_STMT first.
378 2014-03-13 Georg-Johann Lay <avr@gjlay.de>
381 * config/avr/avr.c (avr_out_plus): Swap cc_plus and cc_minus in
382 calls of avr_out_plus_1.
384 2014-03-13 Bin Cheng <bin.cheng@arm.com>
386 * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Record
387 BB's single pred and update the father loop's latch info later.
389 2014-03-12 Michael Meissner <meissner@linux.vnet.ibm.com>
391 * config/rs6000/vector.md (VEC_L): Add V1TI mode to vector types.
395 (VEC_base): Likewise.
396 (mov<MODE>, VEC_M modes): If we are loading TImode into VSX
397 registers, we need to swap double words in little endian mode.
399 * config/rs6000/rs6000-modes.def (V1TImode): Add new vector mode
400 to be a container mode for 128-bit integer operations added in ISA
401 2.07. Unlike TImode and PTImode, the preferred register set is
402 the Altivec/VMX registers for the 128-bit operations.
404 * config/rs6000/rs6000-protos.h (rs6000_move_128bit_ok_p): Add
406 (rs6000_split_128bit_ok_p): Likewise.
408 * config/rs6000/rs6000-builtin.def (BU_P8V_AV_3): Add new support
409 macros for creating ISA 2.07 normal and overloaded builtin
410 functions with 3 arguments.
411 (BU_P8V_OVERLOAD_3): Likewise.
412 (VPERM_1T): Add support for V1TImode in 128-bit vector operations
413 for use as overloaded functions.
414 (VPERM_1TI_UNS): Likewise.
415 (VSEL_1TI): Likewise.
416 (VSEL_1TI_UNS): Likewise.
417 (ST_INTERNAL_1ti): Likewise.
418 (LD_INTERNAL_1ti): Likewise.
419 (XXSEL_1TI): Likewise.
420 (XXSEL_1TI_UNS): Likewise.
421 (VPERM_1TI): Likewise.
422 (VPERM_1TI_UNS): Likewise.
423 (XXPERMDI_1TI): Likewise.
425 (LXVD2X_V1TI): Likewise.
426 (STXVD2X_V1TI): Likewise.
427 (VEC_INIT_V1TI): Likewise.
428 (VEC_SET_V1TI): Likewise.
429 (VEC_EXT_V1TI): Likewise.
430 (EQV_V1TI): Likewise.
431 (NAND_V1TI): Likewise.
432 (ORC_V1TI): Likewise.
433 (VADDCUQ): Add support for 128-bit integer arithmetic instructions
434 added in ISA 2.07. Add both normal 'altivec' builtins, and the
438 (VADDEUQM): Likewise.
439 (VADDECUQ): Likewise.
440 (VSUBEUQM): Likewise.
441 (VSUBECUQ): Likewise.
443 * config/rs6000/rs6000-c.c (__int128_type): New static to hold
444 __int128_t and __uint128_t types.
445 (__uint128_type): Likewise.
446 (altivec_categorize_keyword): Add support for vector __int128_t,
447 vector __uint128_t, vector __int128, and vector unsigned __int128
448 as a container type for TImode operations that need to be done in
449 VSX/Altivec registers.
450 (rs6000_macro_to_expand): Likewise.
451 (altivec_overloaded_builtins): Add ISA 2.07 overloaded functions
452 to support 128-bit integer instructions vaddcuq, vadduqm,
453 vaddecuq, vaddeuqm, vsubcuq, vsubuqm, vsubecuq, vsubeuqm.
454 (altivec_resolve_overloaded_builtin): Add support for V1TImode.
456 * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Add support
457 for V1TImode, and set up preferences to use VSX/Altivec registers.
458 Setup VSX reload handlers.
459 (rs6000_debug_reg_global): Likewise.
460 (rs6000_init_hard_regno_mode_ok): Likewise.
461 (rs6000_preferred_simd_mode): Likewise.
462 (vspltis_constant): Do not allow V1TImode as easy altivec constants.
463 (easy_altivec_constant): Likewise.
464 (output_vec_const_move): Likewise.
465 (rs6000_expand_vector_set): Convert V1TImode set and extract to
467 (rs6000_expand_vector_extract): Likewise.
468 (reg_offset_addressing_ok_p): Setup V1TImode to use VSX reg+reg
470 (rs6000_const_vec): Add support for V1TImode.
471 (rs6000_emit_le_vsx_load): Swap double words when loading or
472 storing TImode/V1TImode.
473 (rs6000_emit_le_vsx_store): Likewise.
474 (rs6000_emit_le_vsx_move): Likewise.
475 (rs6000_emit_move): Add support for V1TImode.
476 (altivec_expand_ld_builtin): Likewise.
477 (altivec_expand_st_builtin): Likewise.
478 (altivec_expand_vec_init_builtin): Likewise.
479 (altivec_expand_builtin): Likewise.
480 (rs6000_init_builtins): Add support for V1TImode type. Add
481 support for ISA 2.07 128-bit integer builtins. Define type names
482 for the VSX/Altivec vector types.
483 (altivec_init_builtins): Add support for overloaded vector
484 functions with V1TImode type.
485 (rs6000_preferred_reload_class): Prefer Altivec registers for V1TImode.
486 (rs6000_move_128bit_ok_p): Move 128-bit move/split validation to
488 (rs6000_split_128bit_ok_p): Likewise.
489 (rs6000_handle_altivec_attribute): Create V1TImode from vector
490 __int128_t and vector __uint128_t.
492 * config/rs6000/vsx.md (VSX_L): Add V1TImode to vector iterators
500 (VS_scalar): Likewise.
501 (VS_double): Likewise.
502 (vsx_set_v1ti): New builtin function to create V1TImode from TImode.
504 * config/rs6000/rs6000.h (TARGET_VADDUQM): New macro to say whether
505 we support the ISA 2.07 128-bit integer arithmetic instructions.
506 (ALTIVEC_OR_VSX_VECTOR_MODE): Add V1TImode.
507 (enum rs6000_builtin_type_index): Add fields to hold V1TImode
508 and TImode types for use with the builtin functions.
509 (V1TI_type_node): Likewise.
510 (unsigned_V1TI_type_node): Likewise.
511 (intTI_type_internal_node): Likewise.
512 (uintTI_type_internal_node): Likewise.
514 * config/rs6000/altivec.md (UNSPEC_VADDCUQ): New unspecs for ISA 2.07
515 128-bit builtin functions.
516 (UNSPEC_VADDEUQM): Likewise.
517 (UNSPEC_VADDECUQ): Likewise.
518 (UNSPEC_VSUBCUQ): Likewise.
519 (UNSPEC_VSUBEUQM): Likewise.
520 (UNSPEC_VSUBECUQ): Likewise.
521 (VM): Add V1TImode to vector mode iterators.
524 (altivec_vadduqm): Add ISA 2.07 128-bit binary builtins.
525 (altivec_vaddcuq): Likewise.
526 (altivec_vsubuqm): Likewise.
527 (altivec_vsubcuq): Likewise.
528 (altivec_vaddeuqm): Likewise.
529 (altivec_vaddecuq): Likewise.
530 (altivec_vsubeuqm): Likewise.
531 (altivec_vsubecuq): Likewise.
533 * config/rs6000/rs6000.md (FMOVE128_GPR): Add V1TImode to vector
535 (BOOL_128): Likewise.
536 (BOOL_REGS_OUTPUT): Likewise.
537 (BOOL_REGS_OP1): Likewise.
538 (BOOL_REGS_OP2): Likewise.
539 (BOOL_REGS_UNARY): Likewise.
540 (BOOL_REGS_AND_CR0): Likewise.
542 * config/rs6000/altivec.h (vec_vaddcuq): Add support for ISA 2.07
543 128-bit integer builtin support.
544 (vec_vadduqm): Likewise.
545 (vec_vaddecuq): Likewise.
546 (vec_vaddeuqm): Likewise.
547 (vec_vsubecuq): Likewise.
548 (vec_vsubeuqm): Likewise.
549 (vec_vsubcuq): Likewise.
550 (vec_vsubuqm): Likewise.
552 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions):
553 Document vec_vaddcuq, vec_vadduqm, vec_vaddecuq, vec_vaddeuqm,
554 vec_subecuq, vec_subeuqm, vec_vsubcuq, vec_vsubeqm builtins adding
555 128-bit integer add/subtract to ISA 2.07.
557 2014-03-12 Joern Rennecke <joern.rennecke@embecosm.com>
559 * config/arc/arc.c (arc_predicate_delay_insns):
560 Fix third argument passed to conditionalize_nonjump.
562 2014-03-12 Yufeng Zhang <yufeng.zhang@arm.com>
564 * config/aarch64/aarch64-builtins.c
565 (aarch64_builtin_vectorized_function): Add BUILT_IN_LFLOORF,
566 BUILT_IN_LLFLOOR, BUILT_IN_LCEILF and BUILT_IN_LLCEIL.
567 * config/aarch64/arm_neon.h (vcvtaq_u64_f64): Call __builtin_llfloor
568 instead of __builtin_lfloor.
569 (vcvtnq_u64_f64): Call __builtin_llceil instead of __builtin_lceil.
571 2014-03-12 Jakub Jelinek <jakub@redhat.com>
573 * tree-ssa-ifcombine.c (forwarder_block_to): New function.
574 (tree_ssa_ifcombine_bb_1): New function.
575 (tree_ssa_ifcombine_bb): Use it. Handle also cases where else_bb
576 is an empty forwarder block to then_bb or vice versa and then_bb
577 and else_bb are effectively swapped.
579 2014-03-12 Christian Bruel <christian.bruel@st.com>
582 * config/arm/arm.c (arm_emit_vfp_multi_reg_pop): Emit a
583 REG_CFA_DEF_CFA note.
584 (arm_expand_epilogue_apcs_frame): call arm_add_cfa_adjust_cfa_note.
585 (arm_unwind_emit): Allow REG_CFA_DEF_CFA.
587 2014-03-12 Thomas Preud'homme <thomas.preudhomme@arm.com>
589 PR tree-optimization/60454
590 * tree-ssa-math-opts.c (find_bswap_1): Fix bswap detection.
592 2014-03-12 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
594 * config.gcc (aarch64*-*-*): Use ISA flags from aarch64-arches.def.
595 Do not define target_cpu_default2 to generic.
596 * config/aarch64/aarch64.h (TARGET_CPU_DEFAULT): Use generic cpu.
597 * config/aarch64/aarch64.c (aarch64_override_options): Update comment.
598 * config/aarch64/aarch64-arches.def (armv8-a): Use generic cpu.
600 2014-03-12 Jakub Jelinek <jakub@redhat.com>
601 Marc Glisse <marc.glisse@inria.fr>
603 PR tree-optimization/60502
604 * tree-ssa-reassoc.c (eliminate_not_pairs): Use build_all_ones_cst
605 instead of build_low_bits_mask.
607 2014-03-12 Jakub Jelinek <jakub@redhat.com>
610 * tree-vrp.c (register_edge_assert_for_1): Don't add assert
611 if there are multiple uses, but op doesn't live on E edge.
612 * tree-cfg.c (assert_unreachable_fallthru_edge_p): Also ignore
613 clobber stmts before __builtin_unreachable.
615 2014-03-11 Richard Sandiford <rdsandiford@googlemail.com>
617 * builtins.c (expand_builtin_setjmp_receiver): Use and clobber
618 hard_frame_pointer_rtx.
619 * cse.c (cse_insn): Remove volatile check.
620 * cselib.c (cselib_process_insn): Likewise.
621 * dse.c (scan_insn): Likewise.
623 2014-03-11 Joern Rennecke <joern.rennecke@embecosm.com>
625 * config/arc/arc.c (conditionalize_nonjump): New function,
627 (arc_ifcvt): ... this.
628 (arc_predicate_delay_insns): Use it.
630 2014-03-11 Joern Rennecke <joern.rennecke@embecosm.com>
632 * config/arc/predicates.md (extend_operand): During/after reload,
633 allow const_int_operand.
634 * config/arc/arc.md (mulsidi3_700): Use extend_operand predicate.
635 (umulsidi3_700): Likewise. Change operand 2 constraint back to "cL".
636 (mulsi3_highpart): Change operand 2 constraint alternatives 2 and 3
638 (umulsi3_highpart_i): Likewise.
640 2014-03-11 Richard Biener <rguenther@suse.de>
642 * tree-ssa-structalias.c (get_constraint_for_ptr_offset):
643 Add asserts to guard possible wrong-code bugs.
645 2014-03-11 Richard Biener <rguenther@suse.de>
647 PR tree-optimization/60429
648 PR tree-optimization/60485
649 * tree-ssa-structalias.c (set_union_with_increment): Properly
650 take into account all fields that overlap the shifted vars.
651 (do_sd_constraint): Likewise.
652 (do_ds_constraint): Likewise.
653 (get_constraint_for_ptr_offset): Likewise.
655 2014-03-11 Chung-Lin Tang <cltang@codesourcery.com>
657 * config/nios2/nios2.c (machine_function): Add fp_save_offset field.
658 (nios2_compute_frame_layout):
659 Add calculation of cfun->machine->fp_save_offset.
660 (nios2_expand_prologue): Correct setting of frame pointer register
662 (nios2_expand_epilogue): Update recovery of stack pointer from
663 frame pointer accordingly.
664 (nios2_initial_elimination_offset): Update calculation of offset
665 for eliminating to HARD_FRAME_POINTER_REGNUM.
667 2014-03-10 Jakub Jelinek <jakub@redhat.com>
670 * ipa.c (symtab_remove_unreachable_nodes): Don't call
671 cgraph_get_create_node on VAR_DECLs.
673 2014-03-10 Richard Biener <rguenther@suse.de>
676 * tree.c (signed_or_unsigned_type_for): Handle OFFSET_TYPEs.
678 2014-03-08 Douglas B Rupp <rupp@gnat.com>
680 * config/vms/vms.opt (vms_float_format): New variable.
682 2014-03-08 Tobias Burnus <burnus@net-b.de>
684 * doc/invoke.texi (-fcilkplus): Update implementation status.
686 2014-03-08 Paulo Matos <paulo@matos-sorge.com>
687 Richard Biener <rguenther@suse.de>
689 * lto-wrapper.c (merge_and_complain): Ensure -fshort-double is used
690 consistently accross all TUs.
691 (run_gcc): Enable -fshort-double automatically at link at link-time
692 and disallow override.
694 2014-03-08 Richard Sandiford <rdsandiford@googlemail.com>
697 * config/mips/mips.c (mips_option_override): Promote -mpaired-single
698 warning to an error. Disable TARGET_PAIRED_SINGLE and TARGET_MIPS3D
699 if they can't be used.
701 2014-03-07 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
703 * configure.ac (HAVE_AS_IX86_TLSLDMPLT): Improve test
704 for Solaris 11/x86 ld.
705 * configure: Regenerate.
707 2014-03-07 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
709 * configure.ac (TLS_SECTION_ASM_FLAG): Save as tls_section_flag.
710 (LIB_TLS_SPEC): Save as ld_tls_libs.
711 (HAVE_AS_IX86_TLSLDMPLT): Define as 1/0.
712 (HAVE_AS_IX86_TLSLDM): New test.
713 * configure, config.in: Regenerate.
714 * config/i386/i386.c (legitimize_tls_address): Fall back to
715 TLS_MODEL_GLOBAL_DYNAMIC on 32-bit Solaris/x86 if tool chain
716 cannot support TLS_MODEL_LOCAL_DYNAMIC.
717 * config/i386/i386.md (*tls_local_dynamic_base_32_gnu): Use if
718 instead of #ifdef in HAVE_AS_IX86_TLSLDMPLT test.
720 2014-03-07 Paulo Matos <paulo@matos-sorge.com>
722 * common.opt (fira-loop-pressure): Mark as optimization.
724 2014-03-07 Thomas Schwinge <thomas@codesourcery.com>
726 * langhooks.c (lhd_omp_mappable_type): The error_mark_node is not
727 an OpenMP mappable type.
729 2014-03-06 Matthias Klose <doko@ubuntu.com>
731 * Makefile.in (s-mlib): Only pass MULTIARCH_DIRNAME if
732 MULTILIB_OSDIRNAMES is not defined.
734 2014-03-06 Jakub Jelinek <jakub@redhat.com>
735 Meador Inge <meadori@codesourcery.com>
738 * config/arm/arm.c (arm_tls_symbol_p): Remove.
739 (arm_legitimize_address): Call legitimize_tls_address for any
740 arm_tls_referenced_p expression, handle constant addend. Call it
741 before testing for !TARGET_ARM.
742 (thumb_legitimize_address): Don't handle arm_tls_symbol_p here.
744 2014-03-06 Richard Biener <rguenther@suse.de>
750 2014-03-04 Paulo Matos <paulo@matos-sorge.com>
752 * tree-streamer.c (record_common_node): Assert we don't record
753 nodes with type double.
754 (preload_common_node): Skip type double, complex double and double
755 pointer since it is now frontend dependent due to fshort-double option.
757 2014-03-06 Richard Biener <rguenther@suse.de>
759 * gcc.c (PLUGIN_COND): Always enable unless -fno-use-linker-plugin
760 or -fno-lto is specified and the linker has full plugin support.
761 * collect2.c (lto_mode): Default to LTO_MODE_WHOPR if LTO is enabled.
762 (main): Remove -flto processing, adjust lto_mode using use_plugin late.
763 * lto-wrapper.c (merge_and_complain): Merge compile-time
765 (run_gcc): And pass it through to the link options.
767 2014-03-06 Alexandre Oliva <aoliva@redhat.com>
771 2014-02-28 Alexandre Oliva <aoliva@redhat.com>
773 * cselib.c (remove_useless_values): Skip to avoid quadratic
774 behavior if the condition moved from...
775 (cselib_process_insn): ... here holds.
777 2014-03-05 Jakub Jelinek <jakub@redhat.com>
780 * Makefile.in (PLUGIN_HEADERS): Add tree-phinodes.h, stor-layout.h,
781 ssa-iterators.h, $(RESOURCE_H) and tree-cfgcleanup.h.
784 * config/i386/t-i386 (OPTIONS_H_EXTRA): Add stringop.def.
785 (TM_H): Add x86-tune.def.
787 2014-03-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
789 * config/aarch64/aarch64.c (generic_tunings):
790 Use cortexa57_extra_costs.
792 2014-03-05 Jakub Jelinek <jakub@redhat.com>
795 * cfgexpand.c (expand_used_vars): Do not assume all SSA_NAMEs
796 of PARM/RESULT_DECLs must be coalesced with optimize && in_lto_p.
797 * tree-ssa-coalesce.c (coalesce_ssa_name): Use MUST_COALESCE_COST - 1
800 2014-03-04 Heiher <r@hev.cc>
802 * config/mips/mips-cpus.def (loongson3a): Mark as a MIPS64r2 processor.
803 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Adjust accordingly.
805 2014-03-04 Uros Bizjak <ubizjak@gmail.com>
807 * config/i386/predicates.md (const2356_operand): Change to ...
808 (const2367_operand): ... this.
809 * config/i386/sse.md (avx512pf_scatterpf<mode>sf): Use
811 (*avx512pf_scatterpf<mode>sf_mask): Ditto.
812 (*avx512pf_scatterpf<mode>sf): Ditto.
813 (avx512pf_scatterpf<mode>df): Ditto.
814 (*avx512pf_scatterpf<mode>df_mask): Ditto.
815 (*avx512pf_scatterpf<mode>df): Ditto.
816 * config/i386/i386.c (ix86_expand_builtin): Update
817 incorrect hint operand error message.
819 2014-03-04 Richard Biener <rguenther@suse.de>
821 * lto-section-in.c (lto_get_section_data): Fix const cast.
823 2014-03-04 Paulo Matos <paulo@matos-sorge.com>
825 * tree-streamer.c (record_common_node): Assert we don't record
826 nodes with type double.
827 (preload_common_node): Skip type double, complex double and double
828 pointer since it is now frontend dependent due to fshort-double option.
830 2014-03-04 Richard Biener <rguenther@suse.de>
833 * lto-streamer-in.c (lto_read_body): Remove LTO bytecode version check.
834 (lto_input_toplevel_asms): Likewise.
835 * lto-section-in.c (lto_get_section_data): Instead do it here
838 2014-03-04 Richard Biener <rguenther@suse.de>
840 PR tree-optimization/60382
841 * tree-vect-loop.c (vect_is_simple_reduction_1): Do not consider
842 dead PHIs a reduction.
844 2014-03-03 Uros Bizjak <ubizjak@gmail.com>
846 * config/i386/xmmintrin.h (enum _mm_hint) <_MM_HINT_ET0>: Correct
848 (_mm_prefetch): Move out of GCC target("sse") pragma.
849 * config/i386/prfchwintrin.h (_m_prefetchw): Move out of
850 GCC target("prfchw") pragma.
851 * config/i386/i386.md (prefetch): Emit prefetchwt1 only
853 * config/i386/i386.c (ix86_option_override_internal): Enable
854 -mprfchw with -mprefetchwt1.
856 2014-03-03 Joern Rennecke <joern.rennecke@embecosm.com>
858 * config/arc/arc.md (casesi_load) <length attribute alternative 0>:
861 2014-03-03 Joern Rennecke <joern.rennecke@embecosm.com>
863 * opts.h (CL_PCH_IGNORE): Define.
864 * targhooks.c (option_affects_pch_p):
865 Return false for options that have CL_PCH_IGNORE set.
866 * opt-functions.awk: Process PchIgnore.
867 * doc/options.texi: Document PchIgnore.
869 * config/arc/arc.opt (misize): Add PchIgnore property.
871 2014-03-03 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
873 * config/rs6000/rs6000.c (rs6000_preferred_reload_class): Disallow
874 reload of PLUS rtx's outside of GENERAL_REGS or BASE_REGS; relax
875 constraint on constants to permit them being loaded into
876 GENERAL_REGS or BASE_REGS.
878 2014-03-03 Nick Clifton <nickc@redhat.com>
880 * config/rl78/rl78-real.md (cbranchsi4_real_signed): Add
881 anti-cacnonical alternatives.
882 (negandhi3_real): New pattern.
883 * config/rl78/rl78-virt.md (negandhi3_virt): New pattern.
885 2014-03-03 Senthil Kumar Selvaraj <senthil_kumar.selvaraj@atmel.com>
887 * config/avr/avr-mcus.def: Remove atxmega16x1.
888 * config/avr/avr-tables.opt: Regenerate.
889 * config/avr/t-multilib: Regenerate.
890 * doc/avr-mmcu.texi: Regenerate.
892 2014-03-03 Tobias Grosser <tobias@grosser.es>
893 Mircea Namolaru <mircea.namolaru@inria.fr>
895 PR tree-optimization/58028
896 * graphite-clast-to-gimple.c (set_cloog_options): Don't remove
899 2014-03-03 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
901 * config/arm/neon.md (*movmisalign<mode>): Legitimize addresses
902 not handled by recognizers.
904 2014-03-03 Jakub Jelinek <jakub@redhat.com>
907 * function.c (expand_function_end): Don't emit
908 clobber_return_register sequence if clobber_after is a BARRIER.
909 * cfgexpand.c (construct_exit_block): Append instructions before
910 return_label to prev_bb.
912 2014-03-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
914 * config/rs6000/constraints.md: Document reserved use of "wc".
916 2014-03-02 Jan Hubicka <hubicka@ucw.cz>
919 * ipa.c (function_and_variable_visibility): When dissolving comdat
920 group, also set all symbols to local.
922 2014-03-02 Jan Hubicka <hubicka@ucw.cz>
927 2013-12-14 Jan Hubicka <jh@suse.cz>
929 * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Skip clobbers.
931 2014-03-02 Jon Beniston <jon@beniston.com>
937 * config/lm32/lm32.c (lm32_legitimate_constant_p): Remove, as incorrect.
938 (TARGET_LEGITIMATE_CONSTANT_P): Undefine, as not needed.
939 * config/lm32/lm32.md (movsi_insn): Add 32-bit immediate support.
940 (simple_return, *simple_return): New patterns
941 * config/lm32/predicates.md (movsi_rhs_operand): Remove as obsolete.
942 * configure.ac (force_sjlj_exceptions): Force sjlj exceptions for lm32.
944 2014-03-01 Paolo Carlini <paolo.carlini@oracle.com>
946 * dwarf2out.c (gen_subprogram_die): Tidy.
948 2014-03-01 Oleg Endo <olegendo@gcc.gnu.org>
951 * config/sh/sh.md (*mov_t_msb_neg): Split into ...
952 (*mov_t_msb_neg_negc): ... this new insn.
954 2014-02-28 Jason Merrill <jason@redhat.com>
957 * ipa-devirt.c (ipa_devirt): Don't choose an implicitly-declared
960 2014-02-28 Paolo Carlini <paolo.carlini@oracle.com>
963 * dwarf2out.c (decltype_auto_die): New static.
964 (gen_subprogram_die): Handle 'decltype(auto)' like 'auto'.
965 (gen_type_die_with_usage): Handle 'decltype(auto)'.
966 (is_cxx_auto): Likewise.
968 2014-02-28 Ian Bolton <ian.bolton@arm.com>
970 * config/aarch64/aarch64.h: Define __ARM_NEON by default if
971 we are not using general regs only.
973 2014-02-28 Richard Biener <rguenther@suse.de>
976 * tree-cfgcleanup.c (tree_forwarder_block_p): Restrict
977 previous fix and only allow to remove trivial pre-headers
978 and latches. Also honor LOOPS_MAY_HAVE_MULTIPLE_LATCHES.
979 (remove_forwarder_block): Properly update the latch of a loop.
981 2014-02-28 Alexandre Oliva <aoliva@redhat.com>
984 * cselib.c (cselib_hasher::equal): Special-case VALUE lookup.
985 (cselib_preserved_hash_table): New.
986 (preserve_constants_and_equivs): Move preserved vals to it.
987 (cselib_find_slot): Look it up first.
988 (cselib_init): Initialize it.
989 (cselib_finish): Release it.
990 (dump_cselib_table): Dump it.
992 2014-02-28 Alexandre Oliva <aoliva@redhat.com>
995 * cselib.c (remove_useless_values): Skip to avoid quadratic
996 behavior if the condition moved from...
997 (cselib_process_insn): ... here holds.
999 2014-02-28 Alexandre Oliva <aoliva@redhat.com>
1002 * var-tracking.c (vt_initialize): Apply the same condition to
1003 preserve the CFA base value.
1005 2014-02-28 Joey Ye <joey.ye@arm.com>
1008 * config/arm/arm.c (thumb_far_jump_used_p): Don't change
1009 if reload in progress or completed.
1011 2014-02-28 Tobias Burnus <burnus@net-b.de>
1014 * tree-pretty-print.c (dump_generic_node, print_declaration): Handle
1017 2014-02-27 H.J. Lu <hongjiu.lu@intel.com>
1019 * doc/tm.texi.in (Condition Code Status): Update documention for
1020 relative locations of cc0-setter and cc0-user.
1022 2014-02-27 Jeff Law <law@redhat.com>
1024 PR rtl-optimization/52714
1025 * combine.c (try_combine): When splitting an unrecognized PARALLEL
1026 into two independent simple sets, if I3 is a jump, ensure the
1027 pattern we place into I3 is a (set (pc) ...).
1029 2014-02-27 Mikael Pettersson <mikpe@it.uu.se>
1030 Jeff Law <law@redhat.com>
1032 PR rtl-optimization/49847
1033 * cse.c (fold_rtx) Handle case where cc0 setter and cc0 user
1034 are in different blocks.
1035 * doc/tm.texi (Condition Code Status): Update documention for
1036 relative locations of cc0-setter and cc0-user.
1038 2014-02-27 Vladimir Makarov <vmakarov@redhat.com>
1041 * lra.c (lra_emit_add): Check SUBREG too.
1043 2014-02-27 Andreas Schwab <schwab@suse.de>
1045 * config/m68k/m68k.c (m68k_option_override): Disable
1046 -flive-range-shrinkage for classic m68k.
1047 (m68k_override_options_after_change): Likewise.
1049 2014-02-27 Marek Polacek <polacek@redhat.com>
1052 * tree-ssa-uninit.c (gate_warn_uninitialized): Run the pass even for
1053 -Wmaybe-uninitialized.
1055 2014-02-27 Alan Modra <amodra@gmail.com>
1058 * reload1.c (emit_input_reload_insns): When reload_override_in,
1059 set old to rl->in_reg when rl->in_reg is a subreg.
1061 2014-02-26 Richard Biener <rguenther@suse.de>
1064 * lra-assigns.c (spill_for): Avoid mixed-sign comparison.
1066 2014-02-25 Ilya Tocar <ilya.tocar@intel.com>
1068 * common/config/i386/predicates.md (const1256_operand): Remove.
1069 (const2356_operand): New.
1070 (const_1_to_2_operand): Remove.
1071 * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
1072 (*avx512pf_gatherpf<mode>sf_mask): Ditto.
1073 (*avx512pf_gatherpf<mode>sf): Ditto.
1074 (avx512pf_gatherpf<mode>df): Ditto.
1075 (*avx512pf_gatherpf<mode>df_mask): Ditto.
1076 (*avx512pf_gatherpf<mode>df): Ditto.
1077 (avx512pf_scatterpf<mode>sf): Ditto.
1078 (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1079 (*avx512pf_scatterpf<mode>sf): Ditto.
1080 (avx512pf_scatterpf<mode>df): Ditto.
1081 (*avx512pf_scatterpf<mode>df_mask): Ditto.
1082 (*avx512pf_scatterpf<mode>df): Ditto.
1083 * common/config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET0.
1085 2014-02-26 Ilya Tocar <ilya.tocar@intel.com>
1087 * config/i386/avx512fintrin.h (_mm512_testn_epi32_mask),
1088 (_mm512_mask_testn_epi32_mask), (_mm512_testn_epi64_mask),
1089 (_mm512_mask_testn_epi64_mask): Move to ...
1090 * config/i386/avx512cdintrin.h: Here.
1091 * config/i386/i386.c (bdesc_args): Change MASK_ISA for testnm.
1092 * config/i386/sse.md (avx512f_vmscalef<mode><round_name>): Remove %.
1093 (avx512f_scalef<mode><mask_name><round_name>): Ditto.
1094 (avx512f_testnm<mode>3<mask_scalar_merge_name>): Change conditon to
1095 TARGET_AVX512F from TARGET_AVX512CD.
1097 2014-02-26 Richard Biener <rguenther@suse.de>
1100 * ipa.c (walk_polymorphic_call_targets): Properly guard
1101 call to inline_update_overall_summary.
1103 2014-02-26 Bin Cheng <bin.cheng@arm.com>
1106 * tree-cfgcleanup.c (tree_forwarder_block_p): Protect loop preheaders
1107 and latches only if requested. Fix latch if it is removed.
1108 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Set
1109 LOOPS_HAVE_PREHEADERS.
1111 2014-02-25 Andrew Pinski <apinski@cavium.com>
1113 * builtins.c (expand_builtin_thread_pointer): Create a new target
1114 when the target is NULL.
1116 2014-02-25 Vladimir Makarov <vmakarov@redhat.com>
1118 PR rtl-optimization/60317
1119 * params.def (PARAM_LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
1120 * params.h (LRA_MAX_CONSIDERED_RELOAD_PSEUDOS): New.
1121 * lra-assigns.c: Include params.h.
1122 (spill_for): Use LRA_MAX_CONSIDERED_RELOAD_PSEUDOS as guard for
1123 other reload pseudos considerations.
1125 2014-02-25 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1127 * config/rs6000/vector.md (*vector_unordered<mode>): Change split
1128 to use canonical form for nor<mode>3.
1130 2014-02-25 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1133 * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Allow 128 to 64-bit
1136 2014-02-25 Ilya Tocar <ilya.tocar@intel.com>
1138 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PREFETCHWT1_SET),
1139 (OPTION_MASK_ISA_PREFETCHWT1_UNSET): New.
1140 (ix86_handle_option): Handle OPT_mprefetchwt1.
1141 * config/i386/cpuid.h (bit_PREFETCHWT1): New.
1142 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1144 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1145 OPTION_MASK_ISA_PREFETCHWT1.
1146 * config/i386/i386.c (ix86_target_string): Handle mprefetchwt1.
1147 (PTA_PREFETCHWT1): New.
1148 (ix86_option_override_internal): Handle PTA_PREFETCHWT1.
1149 (ix86_valid_target_attribute_inner_p): Handle OPT_mprefetchwt1.
1150 * config/i386/i386.h (TARGET_PREFETCHWT1, TARGET_PREFETCHWT1_P): New.
1151 * config/i386/i386.md (prefetch): Check TARGET_PREFETCHWT1
1152 (*prefetch_avx512pf_<mode>_: Change into ...
1153 (*prefetch_prefetchwt1_<mode>: This.
1154 * config/i386/i386.opt (mprefetchwt1): New.
1155 * config/i386/xmmintrin.h (_mm_hint): Add _MM_HINT_ET1.
1156 (_mm_prefetch): Handle intent to write.
1157 * doc/invoke.texi (mprefetchwt1), (mno-prefetchwt1): Doccument.
1159 2014-02-25 Richard Biener <rguenther@suse.de>
1162 * emit-rtl.c (mem_attrs_htab): Remove.
1163 (mem_attrs_htab_hash): Likewise.
1164 (mem_attrs_htab_eq): Likewise.
1165 (set_mem_attrs): Always allocate new mem-attrs when something changed.
1166 (init_emit_once): Do not allocate mem_attrs_htab.
1168 2014-02-25 Richard Biener <rguenther@suse.de>
1171 * lto-opts.c (lto_write_options): Output non-explicit conservative
1172 -fwrapv, -fno-trapv and -fno-strict-overflow.
1173 * lto-wrapper.c (merge_and_complain): Handle merging those options.
1174 (run_gcc): And pass them through.
1176 2014-02-25 Andrey Belevantsev <abel@ispras.ru>
1178 * sel-sched.c (calculate_new_fences): New parameter ptime.
1179 Calculate it as a maximum over all fence cycles.
1180 (sel_sched_region_2): Adjust the call to calculate_new_fences.
1181 Print the final schedule timing when sched_verbose.
1183 2014-02-25 Andrey Belevantsev <abel@ispras.ru>
1185 PR rtl-optimization/60292
1186 * sel-sched.c (fill_vec_av_set): Do not reset target availability
1187 bit fot the fence instruction.
1189 2014-02-24 Alangi Derick <alangiderick@gmail.com>
1191 * calls.h: Fix typo in comment.
1193 2014-02-24 John David Anglin <danglin@gcc.gnu.org>
1195 * config/pa/pa.c (pa_output_move_double): Don't valididate when
1196 adjusting offsetable addresses.
1198 2014-02-24 Guozhi Wei <carrot@google.com>
1200 * sparseset.h (sparseset_pop): Fix the wrong index.
1202 2014-02-24 Walter Lee <walt@tilera.com>
1204 * config.gcc (tilepro-*-*): Change to tilepro*-*-*.
1205 (tilegx-*-linux*): Change to tilegx*-*-linux*; Support tilegxbe
1207 * common/config/tilegx/tilegx-common.c
1208 (TARGET_DEFAULT_TARGET_FLAGS): Define.
1209 * config/tilegx/linux.h (ASM_SPEC): Add endian_spec.
1211 * config/tilegx/sync.md (atomic_test_and_set): Handle big endian.
1212 * config/tilegx/tilegx.c (tilegx_return_in_msb): New.
1213 (tilegx_gimplify_va_arg_expr): Handle big endian.
1214 (tilegx_expand_unaligned_load): Ditto.
1215 (tilegx_expand_unaligned_store): Ditto.
1216 (TARGET_RETURN_IN_MSB): New.
1217 * config/tilegx/tilegx.h (TARGET_DEFAULT): New.
1218 (TARGET_ENDIAN_DEFAULT): New.
1219 (TARGET_BIG_ENDIAN): Handle big endian.
1220 (BYTES_BIG_ENDIAN): Ditto.
1221 (WORDS_BIG_ENDIAN): Ditto.
1222 (FLOAT_WORDS_BIG_ENDIAN): Ditto.
1225 * config/tilegx/tilegx.md (extv): Handle big endian.
1227 (insn_st<n>): Ditto.
1228 (insn_st<n>_add<bitsuffix>): Ditto.
1229 (insn_stnt<n>): Ditto.
1230 (insn_stnt<n>_add<bitsuffix>):Ditto.
1231 (vec_interleave_highv8qi): Handle big endian.
1232 (vec_interleave_highv8qi_be): New.
1233 (vec_interleave_highv8qi_le): New.
1234 (insn_v1int_h): Handle big endian.
1235 (vec_interleave_lowv8qi): Handle big endian.
1236 (vec_interleave_lowv8qi_be): New.
1237 (vec_interleave_lowv8qi_le): New.
1238 (insn_v1int_l): Handle big endian.
1239 (vec_interleave_highv4hi): Handle big endian.
1240 (vec_interleave_highv4hi_be): New.
1241 (vec_interleave_highv4hi_le): New.
1242 (insn_v2int_h): Handle big endian.
1243 (vec_interleave_lowv4hi): Handle big endian.
1244 (vec_interleave_lowv4hi_be): New.
1245 (vec_interleave_lowv4hi_le): New.
1246 (insn_v2int_l): Handle big endian.
1247 (vec_interleave_highv2si): Handle big endian.
1248 (vec_interleave_highv2si_be): New.
1249 (vec_interleave_highv2si_le): New.
1250 (insn_v4int_h): Handle big endian.
1251 (vec_interleave_lowv2si): Handle big endian.
1252 (vec_interleave_lowv2si_be): New.
1253 (vec_interleave_lowv2si_le): New.
1254 (insn_v4int_l): Handle big endian.
1255 * config/tilegx/tilegx.opt (mbig-endian): New option.
1256 (mlittle-endian): New option.
1257 * doc/install.texi: Document tilegxbe-linux.
1258 * doc/invoke.texi: Document -mbig-endian and -mlittle-endian.
1260 2014-02-24 Martin Jambor <mjambor@suse.cz>
1263 * ipa-cp.c (propagate_constants_accross_call): Bail out early if
1264 there are no parameter descriptors.
1266 2014-02-24 Andrey Belevantsev <abel@ispras.ru>
1268 PR rtl-optimization/60268
1269 * sched-rgn.c (haifa_find_rgns): Move the nr_regions_initial variable
1270 initialization to ...
1271 (sched_rgn_init): ... here.
1272 (schedule_region): Check for SCHED_PRESSURE_NONE earlier.
1274 2014-02-23 David Holsgrove <david.holsgrove@xilinx.com>
1276 * config/microblaze/microblaze.md: Correct ashrsi_reg / lshrsi_reg
1279 2014-02-23 Edgar E. Iglesias <edgar.iglesias@xilinx.com>
1281 * config/microblaze/microblaze.h: Remove SECONDARY_MEMORY_NEEDED
1284 2014-02-23 David Holsgrove <david.holsgrove@xilinx.com>
1286 * /config/microblaze/microblaze.c: Add microblaze_asm_output_mi_thunk,
1287 define TARGET_ASM_OUTPUT_MI_THUNK and TARGET_ASM_CAN_OUTPUT_MI_THUNK.
1289 2014-02-23 David Holsgrove <david.holsgrove@xilinx.com>
1291 * config/microblaze/predicates.md: Add cmp_op predicate.
1292 * config/microblaze/microblaze.md: Add branch_compare instruction
1293 which uses cmp_op predicate and emits cmp insn before branch.
1294 * config/microblaze/microblaze.c (microblaze_emit_compare): Rename
1295 to microblaze_expand_conditional_branch and consolidate logic.
1296 (microblaze_expand_conditional_branch): emit branch_compare
1297 insn instead of handling cmp op separate from branch insn.
1299 2014-02-23 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1301 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_move): Relax assert
1304 2014-02-23 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1306 * config/rs6000/altivec.md (altivec_lve<VI_char>x): Replace
1307 define_insn with define_expand and new define_insn
1308 *altivec_lve<VI_char>x_internal.
1309 (altivec_stve<VI_char>x): Replace define_insn with define_expand
1310 and new define_insn *altivec_stve<VI_char>x_internal.
1311 * config/rs6000/rs6000-protos.h (altivec_expand_stvex_be): New
1313 * config/rs6000/rs6000.c (altivec_expand_lvx_be): Document use by
1315 (altivec_expand_stvex_be): New function.
1317 2014-02-22 Joern Rennecke <joern.rennecke@embecosm.com>
1319 * config/avr/avr.c (avr_can_eliminate): Allow elimination from
1320 ARG_POINTER_REGNUM to STACK_POINTER_REGNUM if !frame_pointer_needed.
1321 * config/avr/avr.c (ELIMINABLE_REGS): Add elimination from
1322 ARG_POINTER_REGNUM to STACK_POINTER_REGNUM.
1324 2014-02-21 Vladimir Makarov <vmakarov@redhat.com>
1327 * lra-constraints.c (inherit_reload_reg): Use lra_emit_move
1328 instead of emit_move_insn.
1330 2014-02-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1332 * config/rs6000/altivec.md (altivec_vsumsws): Replace second
1334 (reduc_uplus_v16qi): Use gen_altivec_vsumsws_direct instead of
1335 gen_altivec_vsumsws.
1337 2014-02-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1339 * config/rs6000/altivec.md (altivec_lvxl): Rename as
1340 *altivec_lvxl_<mode>_internal and use VM2 iterator instead of V4SI.
1341 (altivec_lvxl_<mode>): New define_expand incorporating
1342 -maltivec=be semantics where needed.
1343 (altivec_lvx): Rename as *altivec_lvx_<mode>_internal.
1344 (altivec_lvx_<mode>): New define_expand incorporating -maltivec=be
1345 semantics where needed.
1346 (altivec_stvx): Rename as *altivec_stvx_<mode>_internal.
1347 (altivec_stvx_<mode>): New define_expand incorporating
1348 -maltivec=be semantics where needed.
1349 (altivec_stvxl): Rename as *altivec_stvxl_<mode>_internal and use
1350 VM2 iterator instead of V4SI.
1351 (altivec_stvxl_<mode>): New define_expand incorporating
1352 -maltivec=be semantics where needed.
1353 * config/rs6000/rs6000-builtin.def: Add new built-in definitions
1354 LVXL_V2DF, LVXL_V2DI, LVXL_V4SF, LVXL_V4SI, LVXL_V8HI, LVXL_V16QI,
1355 LVX_V2DF, LVX_V2DI, LVX_V4SF, LVX_V4SI, LVX_V8HI, LVX_V16QI, STVX_V2DF,
1356 STVX_V2DI, STVX_V4SF, STVX_V4SI, STVX_V8HI, STVX_V16QI, STVXL_V2DF,
1357 STVXL_V2DI, STVXL_V4SF, STVXL_V4SI, STVXL_V8HI, STVXL_V16QI.
1358 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Replace
1359 ALTIVEC_BUILTIN_LVX with ALTIVEC_BUILTIN_LVX_<MODE> throughout;
1360 similarly for ALTIVEC_BUILTIN_LVXL, ALTIVEC_BUILTIN_STVX, and
1361 ALTIVEC_BUILTIN_STVXL.
1362 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be): New prototype.
1363 (altivec_expand_stvx_be): Likewise.
1364 * config/rs6000/rs6000.c (swap_selector_for_mode): New function.
1365 (altivec_expand_lvx_be): Likewise.
1366 (altivec_expand_stvx_be): Likewise.
1367 (altivec_expand_builtin): Add cases for
1368 ALTIVEC_BUILTIN_STVX_<MODE>, ALTIVEC_BUILTIN_STVXL_<MODE>,
1369 ALTIVEC_BUILTIN_LVXL_<MODE>, and ALTIVEC_BUILTIN_LVX_<MODE>.
1370 (altivec_init_builtins): Add definitions for
1371 __builtin_altivec_lvxl_<mode>, __builtin_altivec_lvx_<mode>,
1372 __builtin_altivec_stvx_<mode>, and __builtin_altivec_stvxl_<mode>.
1374 2014-02-21 Catherine Moore <clm@codesourcery.com>
1376 * doc/invoke.texi (mvirt, mno-virt): Document.
1377 * config/mips/mips.opt (mvirt): New option.
1378 * config/mips/mips.h (ASM_SPEC): Pass mvirt to the assembler.
1380 2014-02-21 Richard Biener <rguenther@suse.de>
1382 PR tree-optimization/60276
1383 * tree-vectorizer.h (struct _stmt_vec_info): Add min_neg_dist field.
1384 (STMT_VINFO_MIN_NEG_DIST): New macro.
1385 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Record
1386 STMT_VINFO_MIN_NEG_DIST.
1387 * tree-vect-stmts.c (vectorizable_load): Verify if assumptions
1388 made for negative dependence distances still hold.
1390 2014-02-21 Richard Biener <rguenther@suse.de>
1393 * tree-ssa-live.c (mark_all_vars_used_1): Do not walk
1394 DECL_INITIAL for globals not in the current function context.
1396 2014-02-21 Jakub Jelinek <jakub@redhat.com>
1398 PR tree-optimization/56490
1399 * params.def (PARAM_UNINIT_CONTROL_DEP_ATTEMPTS): New param.
1400 * tree-ssa-uninit.c: Include params.h.
1401 (compute_control_dep_chain): Add num_calls argument, return false
1402 if it exceed PARAM_UNINIT_CONTROL_DEP_ATTEMPTS param, pass
1403 num_calls to recursive call.
1404 (find_predicates): Change dep_chain into normal array,
1405 cur_chain into auto_vec<edge, MAX_CHAIN_LEN + 1>, add num_calls
1406 variable and adjust compute_control_dep_chain caller.
1407 (find_def_preds): Likewise.
1409 2014-02-21 Thomas Schwinge <thomas@codesourcery.com>
1411 * gimple-pretty-print.c (dump_gimple_omp_for) [flags & TDF_RAW]
1412 <case GF_OMP_FOR_KIND_CILKSIMD>: Add missing break statement.
1414 2014-02-21 Nick Clifton <nickc@redhat.com>
1416 * config/stormy16/stormy16.md (pushdqi1): Add mode to post_inc.
1417 (pushhi1): Likewise.
1418 (popqi1): Add mode to pre_dec.
1421 2014-02-21 Jakub Jelinek <jakub@redhat.com>
1423 * config/i386/i386.c (ix86_expand_vec_perm): Use V8SImode
1424 mode for mask of V8SFmode permutation.
1426 2014-02-20 Richard Henderson <rth@redhat.com>
1429 * builtins.c (expand_builtin_atomic_compare_exchange): Always make
1430 a new pseudo for OLDVAL.
1432 2014-02-20 Jakub Jelinek <jakub@redhat.com>
1435 * config/i386/i386.c (expand_vec_perm_interleave2): Don't call
1436 gen_reg_rtx if d->testing_p.
1437 (expand_vec_perm_pshufb2, expand_vec_perm_broadcast_1): Return early
1438 if d->testing_p and we will certainly return true.
1439 (expand_vec_perm_even_odd_1): Likewise. Don't call gen_reg_rtx
1442 2014-02-20 Uros Bizjak <ubizjak@gmail.com>
1444 * emit-rtl.c (gen_reg_rtx): Assert that
1445 crtl->emit.regno_pointer_align_length is non-zero.
1447 2014-02-20 Richard Henderson <rth@redhat.com>
1450 * builtins.c (expand_builtin_atomic_compare_exchange): Conditionalize
1451 on failure the store back into EXPECT.
1453 2014-02-20 Chung-Lin Tang <cltang@codesourcery.com>
1454 Sandra Loosemore <sandra@codesourcery.com>
1456 * config/nios2/nios2.md (unspec): Add UNSPEC_PIC_GOTOFF_SYM enum.
1457 * config/nios2/nios2.c (nios2_function_profiler): Add
1458 -fPIC (flag_pic == 2) support.
1459 (nios2_handle_custom_fpu_cfg): Fix warning parameter.
1460 (nios2_large_offset_p): New function.
1461 (nios2_unspec_reloc_p): Move up position, update to use
1462 nios2_large_offset_p.
1463 (nios2_unspec_address): Remove function.
1464 (nios2_unspec_offset): New function.
1465 (nios2_large_got_address): New function.
1466 (nios2_got_address): Add large offset support.
1467 (nios2_legitimize_tls_address): Update usage of removed and new
1469 (nios2_symbol_binds_local_p): New function.
1470 (nios2_load_pic_address): Add -fPIC (flag_pic == 2) support.
1471 (nios2_legitimize_address): Update to use nios2_large_offset_p.
1472 (nios2_emit_move_sequence): Avoid legitimizing (const (unspec ...)).
1473 (nios2_print_operand): Merge H/L processing, add hiadj/lo
1474 processing for (const (unspec ...)).
1475 (nios2_unspec_reloc_name): Add UNSPEC_PIC_GOTOFF_SYM case.
1477 2014-02-20 Richard Biener <rguenther@suse.de>
1479 * tree-cfg.c (replace_uses_by): Mark altered BBs before
1480 doing the substitution.
1481 (verify_gimple_assign_single): Also verify bare MEM_REFs on the lhs.
1483 2014-02-20 Martin Jambor <mjambor@suse.cz>
1486 * ipa-cp.c (cgraph_edge_brings_all_agg_vals_for_node): Uce correct
1487 info when checking whether lattices are bottom.
1489 2014-02-20 Richard Biener <rguenther@suse.de>
1492 * tree-eh.c (execute_cleanup_eh_1): Also cleanup empty EH
1495 2014-02-20 Jan Hubicka <hubicka@ucw.cz>
1498 * ipa-inline-transform.c (clone_inlined_nodes): Add freq_scale
1499 parameter specifying the scaling.
1500 (inline_call): Update.
1501 (want_inline_recursively): Guard division by zero.
1502 (recursive_inlining): Update.
1503 * ipa-inline.h (clone_inlined_nodes): Update.
1505 2014-02-20 Ilya Tocar <ilya.tocar@intel.com>
1508 * config/i386/i386.c (classify_argument): Pass structures of size
1509 64 bytes or less in register.
1511 2014-02-20 Ilya Tocar <ilya.tocar@intel.com>
1512 Kirill Yukhin <kirill.yukhin@intel.com>
1514 * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): Swap operands.
1515 (_mm_rcp28_round_ss): Ditto.
1516 (_mm_rsqrt28_round_sd): Ditto.
1517 (_mm_rsqrt28_round_ss): Ditto.
1518 * config/i386/avx512erintrin.h (_mm_rcp14_round_sd): Ditto.
1519 (_mm_rcp14_round_ss): Ditto.
1520 (_mm_rsqrt14_round_sd): Ditto.
1521 (_mm_rsqrt14_round_ss): Ditto.
1522 * config/i386/sse.md (rsqrt14<mode>): Put nonimmediate operand as
1523 the first input operand, get rid of match_dup.
1524 (avx512er_exp2<mode><mask_name><round_saeonly_name>): Set type
1526 (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
1528 (avx512er_vmrcp28<mode><round_saeonly_name>): Put nonimmediate
1529 operand as the first input operand, set type attribute.
1530 (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
1532 (avx512er_vmrsqrt28<mode><round_saeonly_name>): Put nonimmediate
1533 operand as the first input operand, set type attribute.
1535 2014-02-19 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1537 * config/rs6000/rs6000.c (vspltis_constant): Fix most significant
1540 2014-02-19 H.J. Lu <hongjiu.lu@intel.com>
1543 * config/i386/i386.c (construct_container): Remove TFmode check
1544 for X86_64_INTEGER_CLASS.
1546 2014-02-19 Uros Bizjak <ubizjak@gmail.com>
1549 * config/i386/i386.c (type_natural_mode): Warn for ABI changes
1550 only when -Wpsabi is enabled.
1552 2014-02-19 Michael Hudson-Doyle <michael.hudson@linaro.org>
1555 * config/aarch64/aarch64.c (aarch64_pass_by_reference): The rules for
1556 passing arrays in registers are the same as for structs, so remove the
1557 special case for them.
1559 2014-02-19 Eric Botcazou <ebotcazou@adacore.com>
1561 * expr.c (expand_expr_real_1) <case VIEW_CONVERT_EXPR>: For a bit-field
1562 destination type, extract only the valid bits if the source type is not
1563 integral and has a different mode.
1565 2014-02-19 Richard Biener <rguenther@suse.de>
1568 * tree-inline.c (estimate_num_insns): Avoid calling cgraph_get_node
1571 2014-02-19 Richard Biener <rguenther@suse.de>
1574 * ipa-prop.c: Include stringpool.h and tree-ssanames.h.
1575 (ipa_modify_call_arguments): Emit an argument load explicitely and
1576 preserve virtual SSA form there and for the replacement call.
1577 Do not update SSA form nor free dominance info.
1579 2014-02-18 Jan Hubicka <hubicka@ucw.cz>
1581 * ipa.c (function_and_variable_visibility): Also clear WEAK
1582 flag when disolving COMDAT_GROUP.
1584 2014-02-18 Jan Hubicka <hubicka@ucw.cz>
1586 * ipa-prop.h (ipa_ancestor_jf_data): Update ocmment.
1587 * ipa-prop.c (ipa_set_jf_known_type): Return early when
1589 (ipa_set_ancestor_jf): Set type to NULL hwen it is not preserved;
1590 do more sanity checks.
1591 (detect_type_change): Return true when giving up early.
1592 (compute_complex_assign_jump_func): Fix type parameter of
1593 ipa_set_ancestor_jf.
1594 (compute_complex_ancestor_jump_func): Likewise.
1595 (update_jump_functions_after_inlining): Fix updating of
1597 * ipa-cp.c (ipa_get_jf_ancestor_result): Be ready for type to be NULL.
1599 2014-02-18 Jan Hubicka <hubicka@ucw.cz>
1601 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
1602 inline clones when edge disappears.
1604 2014-02-18 Michael Meissner <meissner@linux.vnet.ibm.com>
1607 * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
1608 Split 64-bit moves into 2 patterns. Do not allow the use of
1609 direct move for TDmode in little endian, since the decimal value
1610 has little endian bytes within a word, but the 64-bit pieces are
1611 ordered in a big endian fashion, and normal subreg's of TDmode are
1613 (mov<mode>_64bit_dm): Likewise.
1614 (movtd_64bit_nodm): Likewise.
1616 2014-02-18 Eric Botcazou <ebotcazou@adacore.com>
1618 PR tree-optimization/60174
1619 * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
1620 statement of an SSA_NAME that occurs in an abnormal PHI node.
1622 2014-02-18 Jakub Jelinek <jakub@redhat.com>
1625 * final.c (SEEN_BB): Remove.
1626 (SEEN_NOTE, SEEN_EMITTED): Renumber.
1627 (final_scan_insn): Don't force_source_line on second
1628 NOTE_INSN_BASIC_BLOCK.
1630 2014-02-18 Uros Bizjak <ubizjak@gmail.com>
1633 * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
1634 * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
1635 (type_natural_mode): Warn ABI change when %zmm register is not
1636 available for AVX512F vector value passing.
1638 2014-02-18 Kai Tietz <ktietz@redhat.com>
1641 * config/i386/i386.c (ix86_expand_prologue): Use value in
1642 rax register as displacement when restoring %r10 or %rax.
1643 Fix wrong offset when restoring both registers.
1645 2014-02-18 Eric Botcazou <ebotcazou@adacore.com>
1647 * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
1648 assertion with conditional return.
1650 2014-02-18 Jakub Jelinek <jakub@redhat.com>
1651 Uros Bizjak <ubizjak@gmail.com>
1654 * config/i386/driver-i386.c (host_detect_local_cpu): If
1655 YMM state is not saved by the OS, also clear has_f16c. Move
1656 CPUID 0x80000001 handling before YMM state saving checking.
1658 2014-02-18 Andrey Belevantsev <abel@ispras.ru>
1660 PR rtl-optimization/58960
1661 * haifa-sched.c (alloc_global_sched_pressure_data): New,
1662 factored out from ...
1663 (sched_init): ... here.
1664 (free_global_sched_pressure_data): New, factored out from ...
1665 (sched_finish): ... here.
1666 * sched-int.h (free_global_sched_pressure_data): Declare.
1667 * sched-rgn.c (nr_regions_initial): New static global.
1668 (haifa_find_rgns): Initialize it.
1669 (schedule_region): Disable sched-pressure for the newly
1672 2014-02-17 Richard Biener <rguenther@suse.de>
1674 * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
1675 release SSA defs of pattern stmts.
1677 2014-02-17 Richard Biener <rguenther@suse.de>
1679 * tree-inline.c (expand_call_inline): Release the virtual
1680 operand defined by the call we are about to inline.
1682 2014-02-17 Richard Biener <rguenther@suse.de>
1684 * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
1686 2014-02-17 Kirill Yukhin <kirill.yukhin@intel.com>
1687 Ilya Tocar <ilya.tocar@intel.com>
1689 * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
1690 arguments order in builtin.
1691 (_mm512_permutexvar_epi64): Ditto.
1692 (_mm512_mask_permutexvar_epi64): Ditto
1693 (_mm512_maskz_permutexvar_epi32): Ditto
1694 (_mm512_permutexvar_epi32): Ditto
1695 (_mm512_mask_permutexvar_epi32): Ditto
1697 2014-02-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1699 * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
1700 (p8_vmrgow): Likewise.
1702 2014-02-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1704 * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
1707 2014-02-15 Michael Meissner <meissner@linux.vnet.ibm.com>
1710 * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
1711 (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
1712 into 64-bit and 32-bit moves. On 64-bit moves, add support for
1713 using direct move instructions on ISA 2.07. Also adjust
1714 instruction length for 64-bit.
1715 (mov<mode>_64bit, TFmode/TDmode): Likewise.
1716 (mov<mode>_32bit, TFmode/TDmode): Likewise.
1718 2014-02-15 Alan Modra <amodra@gmail.com>
1722 * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
1723 find_replacement on parts of insn rtl that might be reloaded.
1725 2014-02-15 Richard Biener <rguenther@suse.de>
1727 PR tree-optimization/60183
1728 * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
1729 (tree_ssa_phiprop): Calculate and free post-dominators.
1731 2014-02-14 Jeff Law <law@redhat.com>
1733 PR rtl-optimization/60131
1734 * ree.c (get_extended_src_reg): New function.
1735 (combine_reaching_defs): Use it rather than assuming location of REG.
1736 (find_and_remove_re): Verify first operand of extension is
1737 a REG before adding the insns to the copy list.
1739 2014-02-14 Roland McGrath <mcgrathr@google.com>
1741 * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
1742 * configure: Regenerated.
1743 * config.in: Regenerated.
1744 * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
1745 instead of ASM_SHORT.
1747 2014-02-14 Vladimir Makarov <vmakarov@redhat.com>
1748 Richard Earnshaw <rearnsha@arm.com>
1750 PR rtl-optimization/59535
1751 * lra-constraints.c (process_alt_operands): Encourage alternative
1752 when unassigned pseudo class is superset of the alternative class.
1753 (inherit_reload_reg): Don't inherit when optimizing for code size.
1754 * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
1755 returning CORE_REGS for anything but Thumb1 and BASE_REGS for
1756 modes not less than 4 for Thumb1.
1758 2014-02-14 Kyle McMartin <kyle@redhat.com>
1761 * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
1763 2014-02-14 Richard Biener <rguenther@suse.de>
1765 * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
1766 (get_frame_arg): Drop the assert with langhook types_compatible_p.
1767 Do not strip INDIRECT_REFs.
1769 2014-02-14 Richard Biener <rguenther@suse.de>
1772 * lto-streamer-out.c (DFS_write_tree_body): Do not follow
1773 DECL_FUNCTION_SPECIFIC_TARGET.
1774 (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
1775 * tree-streamer-out.c (pack_ts_target_option): Remove.
1776 (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
1777 (write_ts_function_decl_tree_pointers): Do not stream
1778 DECL_FUNCTION_SPECIFIC_TARGET.
1779 * tree-streamer-in.c (unpack_ts_target_option): Remove.
1780 (unpack_value_fields): Do not stream TS_TARGET_OPTION.
1781 (lto_input_ts_function_decl_tree_pointers): Do not stream
1782 DECL_FUNCTION_SPECIFIC_TARGET.
1784 2014-02-14 Jakub Jelinek <jakub@redhat.com>
1786 * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
1787 (get_initial_def_for_induction, vectorizable_induction): Ignore
1788 debug stmts when looking for exit_phi.
1789 (vectorizable_live_operation): Fix up condition.
1791 2014-02-14 Chung-Ju Wu <jasonwucj@gmail.com>
1793 * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
1794 nreverse() because it changes the content of original tree list.
1796 2014-02-14 Chung-Ju Wu <jasonwucj@gmail.com>
1798 * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
1799 * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
1801 2014-02-14 Chung-Ju Wu <jasonwucj@gmail.com>
1803 * config/nds32/nds32.c (nds32_naked_function_p): Follow the
1804 GNU coding standards.
1806 2014-02-13 Jakub Jelinek <jakub@redhat.com>
1809 * dwarf2out.c (gen_subprogram_die): Don't call
1810 add_calling_convention_attribute if subr_die is old_die.
1812 2014-02-13 Sharad Singhai <singhai@google.com>
1814 * doc/optinfo.texi: Fix order of nodes.
1816 2014-02-13 Uros Bizjak <ubizjak@gmail.com>
1818 * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
1819 operands[2], not operands[3].
1821 2014-02-13 Richard Biener <rguenther@suse.de>
1824 * doc/install.texi (ISL): Update recommended version to 0.12.2,
1825 mention the possibility of an in-tree build.
1826 (CLooG): Update recommended version to 0.18.1, mention the
1827 possibility of an in-tree build and clarify that the ISL
1828 bundled with CLooG does not work.
1830 2014-02-13 Jakub Jelinek <jakub@redhat.com>
1833 * expr.c (compress_float_constant): If x is a hard register,
1834 extend into a pseudo and then move to x.
1836 2014-02-13 Dominik Vogt <vogt@linux.vnet.ibm.com>
1838 * config/s390/s390.c (s390_asm_output_function_label): Fix crash
1839 caused by bad second argument to warning_at() with -mhotpatch and
1840 nested functions (e.g. with gfortran).
1842 2014-02-13 Richard Sandiford <rdsandiford@googlemail.com>
1844 * opts.c (option_name): Remove "enabled by default" rider.
1846 2014-02-12 John David Anglin <danglin@gcc.gnu.org>
1848 * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
1850 2014-02-12 H.J. Lu <hongjiu.lu@intel.com>
1851 Uros Bizjak <ubizjak@gmail.com>
1854 * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
1855 * configure: Regenerated.
1857 2014-02-12 Richard Biener <rguenther@suse.de>
1859 * vec.c (vec_prefix::calculate_allocation): Move as
1860 inline variant to vec.h.
1861 (vec_prefix::calculate_allocation_1): New out-of-line version.
1862 * vec.h (vec_prefix::calculate_allocation_1): Declare.
1863 (vec_prefix::m_has_auto_buf): Rename to ...
1864 (vec_prefix::m_using_auto_storage): ... this.
1865 (vec_prefix::calculate_allocation): Inline the easy cases
1866 and dispatch to calculate_allocation_1 which doesn't need the
1868 (va_heap::reserve): Use gcc_checking_assert.
1869 (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
1870 m_using_auto_storage.
1871 (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
1873 (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
1874 (vec<T, va_heap, vl_ptr>::release): Avoid casting.
1875 (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
1877 2014-02-12 Richard Biener <rguenther@suse.de>
1879 * gcse.c (compute_transp): break from loop over canon_modify_mem_list
1880 when we found a dependence.
1882 2014-02-12 Thomas Schwinge <thomas@codesourcery.com>
1884 * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
1886 (maybe_fold_stmt): ... into this new function.
1887 * omp-low.c (lower_omp): Update comment.
1889 * omp-low.c (lower_omp_target): Add clobber for sizes array, after
1892 * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
1895 2014-02-12 James Greenhalgh <james.greenhalgh@arm.com>
1897 * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
1898 identifiers in comments.
1899 (cortexa53_extra_costs): Likewise.
1900 * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
1901 (cortexa7_extra_costs): Likewise.
1902 (cortexa12_extra_costs): Likewise.
1903 (cortexa15_extra_costs): Likewise.
1904 (v7m_extra_costs): Likewise.
1906 2014-02-12 Richard Biener <rguenther@suse.de>
1909 * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
1910 of posix_memalign being successful.
1911 (lower_stmt): Restrict lowering of posix_memalign to when
1912 -ftree-bit-ccp is enabled.
1914 2014-02-12 Senthil Kumar Selvaraj <senthil_kumar.selvaraj@atmel.com>
1916 * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
1918 * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
1920 2014-02-12 Eric Botcazou <ebotcazou@adacore.com>
1922 PR rtl-optimization/60116
1923 * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
1924 other_insn once the combination has been validated.
1926 2014-02-11 Jan Hubicka <hubicka@ucw.cz>
1929 * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
1931 * ipa-devirt.c: Include demangle.h
1932 (odr_violation_reported): New static variable.
1933 (add_type_duplicate): Update odr_violations.
1934 (maybe_record_node): Add completep parameter; update it.
1935 (record_target_from_binfo): Add COMPLETEP parameter;
1936 update it as needed.
1937 (possible_polymorphic_call_targets_1): Likewise.
1938 (struct polymorphic_call_target_d): Add nonconstruction_targets;
1939 rename FINAL to COMPLETE.
1940 (record_targets_from_bases): Sanity check we found the binfo;
1941 fix COMPLETEP updating.
1942 (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
1943 parameter, fix computing of COMPLETEP.
1944 (dump_possible_polymorphic_call_targets): Imrove readability of dump;
1945 at LTO time do demangling.
1946 (ipa_devirt): Use nonconstruction_targets; Improve dumps.
1947 * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
1949 (gimple_get_virt_method_for_binfo): Likewise.
1950 * gimple-fold.h (gimple_get_virt_method_for_binfo,
1951 gimple_get_virt_method_for_vtable): Update prototypes.
1953 2014-02-11 Vladimir Makarov <vmakarov@redhat.com>
1956 * genautomata.c (add_presence_absence): Fix typo with
1957 {final_}presence_list.
1959 2014-02-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1962 * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
1963 for VSX/Altivec vectors that land in GPR registers.
1965 2014-02-11 Richard Henderson <rth@redhat.com>
1966 Jakub Jelinek <jakub@redhat.com>
1969 * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
1970 around drhs if type conversion to lacc->type is not useless.
1972 2014-02-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1974 * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
1976 (cortex-a57.cortex-a53): Likewise.
1977 * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
1979 2014-02-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1981 * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
1984 2014-02-11 Renlin Li <Renlin.Li@arm.com>
1986 * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
1987 add_options_for_arm_vfp3.
1989 2014-02-11 Jeff Law <law@redhat.com>
1992 * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
1993 object with an undesirable mode.
1995 2014-02-11 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1998 * config/i386/sol2-9.h: New file.
1999 * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
2000 *-*-solaris2.9*): Use it.
2002 2014-02-10 Nagaraju Mekala <nagaraju.mekala@xilinx.com>
2004 * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
2005 * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
2007 2014-02-10 Nagaraju Mekala <nagaraju.mekala@xilinx.com>
2009 * config/microblaze/microblaze.c: Extend mcpu version format
2011 2014-02-10 David Holsgrove <david.holsgrove@xilinx.com>
2013 * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
2015 2014-02-10 Richard Henderson <rth@redhat.com>
2018 * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
2019 * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
2020 ms-abi vs -mno-accumulate-outgoing-args.
2021 (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
2022 * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
2025 2014-02-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
2028 * cfgexpand.c (expand_asm_operands): Attach source location to
2029 ASM_INPUT rtx objects.
2030 * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
2032 2014-02-10 Nick Clifton <nickc@redhat.com>
2034 * config/mn10300/mn10300.c (popcount): New function.
2035 (mn10300_expand_prologue): Include saved registers in stack usage
2038 2014-02-10 Jeff Law <law@redhat.com>
2041 * reload1.c (emit_input_reload_insns): Do not create invalid RTL
2042 when changing the SET_DEST of a prior insn to avoid an input reload.
2044 2014-02-10 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
2046 * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
2047 big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
2048 -mcall-openbsd, or -mcall-linux.
2049 (CC1_ENDIAN_BIG_SPEC): Remove.
2050 (CC1_ENDIAN_LITTLE_SPEC): Remove.
2051 (CC1_ENDIAN_DEFAULT_SPEC): Remove.
2052 (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
2053 (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
2054 and %cc1_endian_default.
2055 * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
2057 2014-02-10 Richard Biener <rguenther@suse.de>
2059 PR tree-optimization/60115
2060 * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
2061 MEM_REF handling. Properly verify that the accesses are not
2062 out of the objects bound.
2064 2014-02-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2066 * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
2069 2014-02-10 Eric Botcazou <ebotcazou@adacore.com>
2071 * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
2072 proper constants and fix formatting.
2073 (possible_polymorphic_call_targets): Fix formatting.
2075 2014-02-10 Kirill Yukhin <kirill.yukhin@intel.com>
2076 Ilya Tocar <ilya.tocar@intel.com>
2078 * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
2079 (_mm512_loadu_epi32): Renamed into...
2080 (_mm512_loadu_si512): This.
2081 (_mm512_storeu_epi32): Renamed into...
2082 (_mm512_storeu_si512): This.
2083 (_mm512_maskz_ceil_ps): Removed.
2084 (_mm512_maskz_ceil_pd): Ditto.
2085 (_mm512_maskz_floor_ps): Ditto.
2086 (_mm512_maskz_floor_pd): Ditto.
2087 (_mm512_floor_round_ps): Ditto.
2088 (_mm512_floor_round_pd): Ditto.
2089 (_mm512_ceil_round_ps): Ditto.
2090 (_mm512_ceil_round_pd): Ditto.
2091 (_mm512_mask_floor_round_ps): Ditto.
2092 (_mm512_mask_floor_round_pd): Ditto.
2093 (_mm512_mask_ceil_round_ps): Ditto.
2094 (_mm512_mask_ceil_round_pd): Ditto.
2095 (_mm512_maskz_floor_round_ps): Ditto.
2096 (_mm512_maskz_floor_round_pd): Ditto.
2097 (_mm512_maskz_ceil_round_ps): Ditto.
2098 (_mm512_maskz_ceil_round_pd): Ditto.
2099 (_mm512_expand_pd): Ditto.
2100 (_mm512_expand_ps): Ditto.
2101 * config/i386/i386.c (ix86_builtins): Remove
2102 IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
2103 (bdesc_args): Ditto.
2104 * config/i386/predicates.md (const1256_operand): New.
2105 (const_1_to_2_operand): Ditto.
2106 * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
2107 (*avx512pf_gatherpf<mode>sf_mask): Ditto.
2108 (*avx512pf_gatherpf<mode>sf): Ditto.
2109 (avx512pf_gatherpf<mode>df): Ditto.
2110 (*avx512pf_gatherpf<mode>df_mask): Ditto.
2111 (*avx512pf_gatherpf<mode>df): Ditto.
2112 (avx512pf_scatterpf<mode>sf): Ditto.
2113 (*avx512pf_scatterpf<mode>sf_mask): Ditto.
2114 (*avx512pf_scatterpf<mode>sf): Ditto.
2115 (avx512pf_scatterpf<mode>df): Ditto.
2116 (*avx512pf_scatterpf<mode>df_mask): Ditto.
2117 (*avx512pf_scatterpf<mode>df): Ditto.
2118 (avx512f_expand<mode>): Removed.
2119 (<shift_insn><mode>3<mask_name>): Change predicate type.
2121 2014-02-08 Jakub Jelinek <jakub@redhat.com>
2123 * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
2124 not at the end of datarefs vector use ordered_remove to avoid
2125 reordering datarefs vector.
2128 * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
2129 mark local addressable non-static vars as GOVD_PRIVATE
2130 instead of GOVD_LOCAL.
2131 * omp-low.c (lower_omp_for): Move gimple_bind_vars
2132 and BLOCK_VARS of gimple_bind_block to new_stmt rather
2136 * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
2137 if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
2138 assume_aligned or alloc_align attributes.
2139 (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
2140 arguments. Handle also assume_aligned and alloc_align attributes.
2141 (evaluate_stmt): Adjust bit_value_assume_aligned caller. Handle
2142 calls to functions with assume_aligned or alloc_align attributes.
2143 * doc/extend.texi: Document assume_aligned and alloc_align attributes.
2145 2014-02-08 Terry Guo <terry.guo@arm.com>
2147 * doc/invoke.texi: Document ARM -march=armv7e-m.
2149 2014-02-08 Jakub Jelinek <jakub@redhat.com>
2151 * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
2152 flag on __cilkrts_rethrow builtin.
2155 * ipa-cp.c (determine_versionability): Fail at -O0
2156 or __attribute__((optimize (0))) or -fno-ipa-cp functions.
2157 * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
2160 2014-02-04 Jakub Jelinek <jakub@redhat.com>
2163 * tree-inline.c (copy_forbidden): Fail for
2164 __attribute__((optimize (0))) functions.
2166 2014-02-07 Jan Hubicka <hubicka@ucw.cz>
2168 * varpool.c: Include pointer-set.h.
2169 (varpool_remove_unreferenced_decls): Variables in other partitions
2170 will not be output; be however careful to not lose information
2173 2014-02-07 Jan Hubicka <hubicka@ucw.cz>
2175 * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
2176 lookup in the vtable constructor.
2178 2014-02-07 Jeff Law <law@redhat.com>
2181 * config/m68k/m68k.md (ashldi_extsi): Turn into a
2182 define_insn_and_split.
2184 * ipa-inline.c (inline_small_functions): Fix typos.
2186 2014-02-07 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
2188 * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
2189 (s390_can_use_return_insn): Declare.
2190 * config/s390/s390.h (EPILOGUE_USES): Define.
2191 * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
2193 (s390_chunkify_start): Handle return JUMP_LABELs.
2194 (s390_early_mach): Emit a main_pool instruction on the entry edge.
2195 (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
2196 (s390_can_use_return_insn): New functions.
2197 (s390_fix_long_loop_prediction): Handle conditional returns.
2198 (TARGET_SET_UP_BY_PROLOGUE): Define.
2199 * config/s390/s390.md (ANY_RETURN): New code iterator.
2200 (*creturn, *csimple_return, return, simple_return): New patterns.
2202 2014-02-07 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
2204 * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
2205 notes to each restore. Also add REG_CFA_DEF_CFA when restoring %r15.
2206 (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P. Update the
2207 REG_CFA_RESTORE list when deciding not to restore a register.
2209 2014-02-07 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
2211 * config/s390/s390.c: Include tree-pass.h and context.h.
2212 (s390_early_mach): New function, split out from...
2213 (s390_emit_prologue): ...here.
2214 (pass_data_s390_early_mach): New pass structure.
2215 (pass_s390_early_mach): New class.
2216 (s390_option_override): Create and register early_mach pass.
2217 Move to end of file.
2219 2014-02-07 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
2221 * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
2222 to match for the exit block.
2224 2014-02-07 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
2226 * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
2227 ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
2228 Reject misaligned operands.
2230 2014-02-07 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
2232 * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
2234 2014-02-07 Richard Biener <rguenther@suse.de>
2237 * gimple-low.c (lower_builtin_posix_memalign): New function.
2238 (lower_stmt): Call it to lower posix_memalign in a way
2239 to make alignment info accessible.
2241 2014-02-07 Jakub Jelinek <jakub@redhat.com>
2244 * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
2245 __builtin_setjmp_receiver.
2247 2014-02-07 Richard Biener <rguenther@suse.de>
2250 * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
2251 * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
2252 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2253 Handle BUILT_IN_POSIX_MEMALIGN.
2254 (find_func_clobbers): Likewise.
2255 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
2256 (call_may_clobber_ref_p_1): Likewise.
2258 2014-02-06 Jan Hubicka <hubicka@ucw.cz>
2261 * ipa-devirt.c (record_target_from_binfo): Remove overactive
2264 2014-02-06 Jan Hubicka <hubicka@ucw.cz>
2267 * lto-cgraph.c (lto_output_node): Use
2268 symtab_get_symbol_partitioning_class.
2269 (lto_output_varpool_node): likewise.
2270 (symtab_get_symbol_partitioning_class): Move here from
2272 * cgraph.h (symbol_partitioning_class): Likewise.
2273 (symtab_get_symbol_partitioning_class): Declare.
2275 2014-02-06 Jan Hubicka <hubicka@ucw.cz>
2277 * ggc.h (ggc_internal_cleared_alloc): New macro.
2278 * vec.h (vec_safe_copy): Handle memory stats.
2279 * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
2280 * target-globals.c (save_target_globals): Likewise.
2282 2014-02-06 Jan Hubicka <hubicka@ucw.cz>
2285 * expr.c (emit_move_resolve_push): Export; be bit more selective
2286 on when to clear alias set.
2287 * expr.h (emit_move_resolve_push): Declare.
2288 * function.h (struct function): Add tail_call_marked.
2289 * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
2290 * config/i386/i386-protos.h (ix86_expand_push): Remove.
2291 * config/i386/i386.md (TImode move expander): De not call
2293 (FP push expanders): Preserve memory attributes.
2294 * config/i386/sse.md (push<mode>1): Remove.
2295 * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
2296 (ix86_expand_push): Remove.
2297 * config/i386/mmx.md (push<mode>1): Remove.
2299 2014-02-06 Jakub Jelinek <jakub@redhat.com>
2301 PR rtl-optimization/60030
2302 * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
2303 lopart with paradoxical subreg before shifting it up by hprec.
2305 2014-02-06 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2307 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
2308 Remove extra newline at end of file.
2309 * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
2310 (arm_issue_rate): Handle cortexa57.
2311 * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
2312 (cortex-a57.cortex-a53): Likewise.
2314 2014-02-06 Jakub Jelinek <jakub@redhat.com>
2317 * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
2318 don't record in REG_FRAME_RELATED_EXPR registers not set in that
2320 (arm_expand_prologue): Adjust all callers.
2321 (arm_unwind_emit_sequence): Allow saved, but not important for unwind
2322 info, registers also at the lowest numbered registers side. Use
2323 gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
2327 * var-tracking.c (adjust_mems): Before adding a SET to
2328 amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
2330 2014-02-06 Alan Modra <amodra@gmail.com>
2333 * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
2334 change SDmode to DDmode when lra_in_progress.
2336 2014-02-06 Jakub Jelinek <jakub@redhat.com>
2339 * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
2340 free_data_ref on the dr first, and before goto again also set dr
2341 to the next dr. For simd_lane_access, free old datarefs[i] before
2342 overwriting it. For get_vectype_for_scalar_type failure, don't
2343 free_data_ref if simd_lane_access.
2345 * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
2348 * tree.h (opts_for_fn): New inline function.
2349 (opt_for_fn): Define.
2350 * config/i386/i386.c (ix86_function_regparm): Use
2351 opt_for_fn (decl, optimize) instead of optimize.
2353 2014-02-06 Marcus Shawcroft <marcus.shawcroft@arm.com>
2355 * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
2356 for SYMBOL_REF in large memory model.
2358 2014-02-06 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2360 * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
2362 (cortex-a57): Likewise.
2363 (cortex-a57.cortex-a53): Likewise.
2365 2014-02-06 Yury Gribov <y.gribov@samsung.com>
2366 Kugan Vivekanandarajah <kuganv@linaro.org>
2368 * config/arm/arm.c (arm_vector_alignment_reachable): Check
2370 * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
2372 2014-02-06 Richard Biener <rguenther@suse.de>
2374 * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
2375 set_loop_copy and initialize_original_copy_tables.
2377 2014-02-06 Alex Velenko <Alex.Velenko@arm.com>
2379 * config/aarch64/aarch64-simd.md
2380 (aarch64_ashr_simddi): Change QI to SI.
2382 2014-02-05 Jan Hubicka <hubicka@ucw.cz>
2383 Jakub Jelinek <jakub@redhat.com>
2386 * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
2389 2014-02-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2391 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
2392 CODE_FOR_altivec_vpku[hw]um to
2393 CODE_FOR_altivec_vpku[hw]um_direct.
2394 * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
2395 UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
2396 (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
2397 UNSPEC_VUNPACK_LO_SIGN_DIRECT.
2399 2014-02-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2401 * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
2402 generation for -maltivec=be.
2403 (altivec_vsumsws): Simplify redundant test.
2405 2014-02-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2407 * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
2408 (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
2409 (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
2410 (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
2411 gen_altivec_vpkuwum.
2412 (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
2414 (altivec_vpks<VI_char>ss): Likewise.
2415 (altivec_vpks<VI_char>us): Likewise.
2416 (altivec_vpku<VI_char>us): Likewise.
2417 (altivec_vpku<VI_char>um): Likewise.
2418 (altivec_vpku<VI_char>um_direct): New (copy of
2419 altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
2421 (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
2422 target is little endian and -maltivec=be is not specified.
2423 (*altivec_vupkhs<VU_char>_direct): New (copy of
2424 altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
2425 (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
2426 target is little endian and -maltivec=be is not specified.
2427 (*altivec_vupkls<VU_char>_direct): New (copy of
2428 altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
2429 (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
2430 little endian and -maltivec=be is not specified.
2431 (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
2432 little endian and -maltivec=be is not specified.
2434 2014-02-05 Richard Henderson <rth@redhat.com>
2437 * combine-stack-adj.c: Revert r206943.
2438 * sched-int.h (struct deps_desc): Add last_args_size.
2439 * sched-deps.c (init_deps): Initialize it.
2440 (sched_analyze_insn): Add OUTPUT dependencies between insns that
2441 contain REG_ARGS_SIZE notes.
2443 2014-02-05 Jan Hubicka <hubicka@ucw.cz>
2445 * lto-cgraph.c (asm_nodes_output): Make global.
2446 * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
2447 * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
2448 (driver_handle_option): Handle OPT_fwpa.
2450 2014-02-05 Jakub Jelinek <jakub@redhat.com>
2453 * ipa-devirt.c (possible_polymorphic_call_targets): Fix
2454 a comment typo and formatting issue. If odr_hash hasn't been
2455 created, return vNULL and set *completep to false.
2458 * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
2459 bb with no successors.
2461 2014-02-05 James Greenhalgh <james.greenhalgh@arm.com>
2464 * doc/invoke.texi (-march): Clarify documentation for ARM.
2468 2014-02-05 Richard Biener <rguenther@suse.de>
2470 * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
2471 when not vectorizing because of too many alias checks.
2472 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
2473 Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
2475 2014-02-05 Nick Clifton <nickc@redhat.com>
2477 * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
2478 accept extended registers in any mode when compiling for the MN10300.
2480 2014-02-05 Yury Gribov <y.gribov@samsung.com>
2482 * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
2483 * ipa-inline.c (report_inline_failed_reason): Handle mismatched
2484 sanitization attributes.
2485 (can_inline_edge_p): Likewise.
2486 (sanitize_attrs_match_for_inline_p): New function.
2488 2014-02-04 Jan Hubicka <hubicka@ucw.cz>
2490 * ipa-prop.c (detect_type_change): Shor circuit testing of
2491 type changes on THIS pointer.
2493 2014-02-04 John David Anglin <danglin@gcc.gnu.org>
2496 * config/pa/pa.c (legitimize_tls_address): Return original address
2497 if not passed a SYMBOL_REF rtx.
2498 (hppa_legitimize_address): Call legitimize_tls_address for all TLS
2500 (pa_emit_move_sequence): Simplify TLS source operands.
2501 (pa_legitimate_constant_p): Reject all TLS constants.
2502 * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
2503 (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
2505 2014-02-04 Jan Hubicka <hubicka@ucw.cz>
2507 * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
2508 groups when we know they are controlled by LTO.
2509 * varasm.c (default_binds_local_p_1): If object is in other partition,
2510 it will be resolved locally.
2512 2014-02-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
2514 * config/host-linux.c (linux_gt_pch_use_address): Don't
2515 use SSIZE_MAX because it is not always defined.
2517 2014-02-04 Vladimir Makarov <vmakarov@redhat.com>
2520 * lra-constraints.c (need_for_split_p): Use more 3 reloads as
2521 threshold for pseudo splitting.
2522 (update_ebb_live_info): Process call argument hard registers and
2523 hard registers from insn definition too.
2524 (max_small_class_regs_num): New constant.
2525 (inherit_in_ebb): Update live hard regs through EBBs. Update
2526 reloads_num only for small register classes. Don't split for
2529 2014-02-04 Markus Trippelsdorf <markus@trippelsdorf.de>
2532 * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
2535 2014-02-04 Jan Hubicka <hubicka@ucw.cz>
2537 * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
2540 2014-02-04 Marek Polacek <polacek@redhat.com>
2542 * gdbinit.in (pel): Define.
2544 2014-02-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
2546 * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
2549 2014-02-04 Richard Biener <rguenther@suse.de>
2552 * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
2553 in function context local.
2554 (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
2555 * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
2556 similar to LTO_imported_decl_ref.
2558 2014-02-04 Jakub Jelinek <jakub@redhat.com>
2560 PR tree-optimization/60002
2561 * cgraphclones.c (build_function_decl_skip_args): Clear
2564 PR tree-optimization/60023
2565 * tree-if-conv.c (predicate_mem_writes): Pass true instead of
2566 false to gsi_replace.
2567 * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
2568 has been in some EH region and vec_stmt could throw, add
2569 vec_stmt into the same EH region.
2570 * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
2571 has no lhs, ignore it.
2572 * internal-fn.c (expand_MASK_LOAD): Likewise.
2575 * tree-inline.c (copy_forbidden): Fail for
2576 __attribute__((optimize (0))) functions.
2579 * omp-low.c (simd_clone_struct_copy): If from->inbranch
2580 is set, copy one less argument.
2581 (expand_simd_clones): Don't subtract clone_info->inbranch
2582 from simd_clone_struct_alloc argument.
2584 PR rtl-optimization/57915
2585 * recog.c (simplify_while_replacing): If all unary/binary/relational
2586 operation arguments are constant, attempt to simplify those.
2589 * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
2590 if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
2592 2014-02-04 Richard Biener <rguenther@suse.de>
2594 PR tree-optimization/60012
2595 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
2596 TBAA disambiguation to all DDRs.
2598 2014-02-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2601 * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
2602 (LINK_SPEC): Use it for -shared, -shared-libgcc.
2604 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
2607 * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
2609 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
2611 * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
2612 * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
2614 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
2617 * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
2618 to figure out targets of polymorphic calls with known decl.
2619 * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2620 * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
2621 * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
2622 (get_polymorphic_call_info): ... here.
2623 (get_polymorphic_call_info_from_invariant): New function.
2625 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
2627 * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
2628 lookup via vtable pointer; check for type consistency
2629 and turn inconsitent facts into UNREACHABLE.
2630 * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2631 * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
2632 type inconsistent querries; return UNREACHABLE instead.
2634 2014-02-03 Richard Henderson <rth@twiddle.net>
2637 * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
2638 already processed this node.
2639 (normalize_one_pred_1): Pass along mark_set.
2640 (normalize_one_pred): Create and destroy a pointer_set_t.
2641 (normalize_one_pred_chain): Likewise.
2643 2014-02-03 Laurent Aflonsi <laurent.alfonsi@st.com>
2645 PR gcov-profile/58602
2646 * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
2648 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
2651 * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
2652 -fno-devirtualize; try to devirtualize by the knowledge of
2653 virtual table pointer given by aggregate propagation.
2654 * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
2655 (ipa_print_node_jump_functions): Dump also offset that
2656 is relevant for polymorphic calls.
2657 (determine_known_aggregate_parts): Add arg_type parameter; use it
2658 instead of determining the type from pointer type.
2659 (ipa_compute_jump_functions_for_edge): Update call of
2660 determine_known_aggregate_parts.
2661 * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
2662 (gimple_get_virt_method_for_binfo): ... here; simplify using
2663 vtable_pointer_value_to_vtable.
2664 * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
2665 * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
2666 to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
2667 (vtable_pointer_value_to_vtable): Break out from ...; handle also
2669 (vtable_pointer_value_to_binfo): ... here.
2670 * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
2672 2014-02-03 Teresa Johnson <tejohnson@google.com>
2674 * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
2675 redef of outer loop index variable.
2677 2014-02-03 Marc Glisse <marc.glisse@inria.fr>
2681 * doc/extend.texi (Function Attributes): Typo.
2683 2014-02-03 Cong Hou <congh@google.com>
2685 PR tree-optimization/60000
2686 * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
2687 if the vectorized statement is a store. A store statement can only
2688 appear at the end of pattern statements.
2690 2014-02-03 H.J. Lu <hongjiu.lu@intel.com>
2692 * config/i386/i386.c (flag_opts): Add -mlong-double-128.
2693 (ix86_option_override_internal): Default long double to 64-bit for
2694 32-bit Bionic and to 128-bit for 64-bit Bionic.
2696 * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
2697 TARGET_LONG_DOUBLE_128 is true.
2698 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
2700 * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
2701 (mlong-double-64): Negate -mlong-double-128.
2702 (mlong-double-128): New option.
2704 * config/i386/i386-c.c (ix86_target_macros): Define
2705 __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
2707 * doc/invoke.texi: Document -mlong-double-128.
2709 2014-02-03 H.J. Lu <hongjiu.lu@intel.com>
2711 PR rtl-optimization/60024
2712 * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
2714 2014-02-03 Markus Trippelsdorf <markus@trippelsdorf.de>
2716 * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
2718 2014-02-03 Andrey Belevantsev <abel@ispras.ru>
2720 PR rtl-optimization/57662
2721 * sel-sched.c (code_motion_path_driver): Do not mark already not
2722 existing blocks in the visiting bitmap.
2724 2014-02-03 Andrey Belevantsev <abel@ispras.ru>
2726 * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
2727 on the insn being emitted.
2729 2014-02-03 James Greenhalgh <james.greenhalgh@arm.com>
2730 Will Deacon <will.deacon@arm.com>
2732 * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
2734 2014-02-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2736 * config/arm/arm-tables.opt: Regenerate.
2738 2014-02-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2740 * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
2741 for vector types other than V16QImode.
2742 * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
2743 define_expand, and call altivec_expand_vec_perm_le when producing
2744 code with little endian element order.
2745 (*altivec_vperm_<mode>_internal): New insn having previous
2746 behavior of altivec_vperm_<mode>.
2747 (altivec_vperm_<mode>_uns): Change to a define_expand, and call
2748 altivec_expand_vec_perm_le when producing code with little endian
2750 (*altivec_vperm_<mode>_uns_internal): New insn having previous
2751 behavior of altivec_vperm_<mode>_uns.
2753 2014-02-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2755 * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
2756 (altivec_vsumsws): Add handling for -maltivec=be with a little
2758 (altivec_vsumsws_direct): New.
2759 (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
2760 gen_altivec_vsumsws.
2762 2014-02-02 Jan Hubicka <hubicka@ucw.cz>
2764 * ipa-devirt.c (subbinfo_with_vtable_at_offset,
2765 vtable_pointer_value_to_binfo): New functions.
2766 * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
2767 * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
2769 2014-02-02 Sandra Loosemore <sandra@codesourcery.com>
2771 * config/nios2/nios2.md (load_got_register): Initialize GOT
2772 pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
2773 * config/nios2/nios2.c (nios2_function_profiler): Likewise.
2775 2014-02-02 Jan Hubicka <hubicka@ucw.cz>
2777 * ipa-prop.c (update_jump_functions_after_inlining): When type is not
2778 preserverd by passthrough, do not propagate the type.
2780 2014-02-02 Richard Sandiford <rdsandiford@googlemail.com>
2782 * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
2783 (mips_atomic_assign_expand_fenv): New function.
2784 (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
2786 2014-02-02 Richard Sandiford <rdsandiford@googlemail.com>
2788 * doc/extend.texi (__builtin_mips_get_fcsr): Document.
2789 (__builtin_mips_set_fcsr): Likewise.
2790 * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
2791 MIPS_USI_FTYPE_VOID.
2792 * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
2793 (mips16_expand_set_fcsr): Likewise.
2794 * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
2795 (mips16_set_fcsr_stub): Likewise.
2796 (mips16_get_fcsr_one_only_stub): New class.
2797 (mips16_set_fcsr_one_only_stub): Likewise.
2798 (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
2799 (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
2800 (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
2801 (hard_float): New availability predicate.
2802 (mips_builtins): Add get_fcsr and set_fcsr.
2803 (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
2804 * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
2805 (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
2806 (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
2807 (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
2810 2014-02-02 Richard Sandiford <rdsandiford@googlemail.com>
2812 * config/mips/mips.c (mips_one_only_stub): New class.
2813 (mips_need_mips16_rdhwr_p): Replace with...
2814 (mips16_rdhwr_stub): ...this new variable.
2815 (mips16_stub_call_address): New function.
2816 (mips16_rdhwr_one_only_stub): New class.
2817 (mips_expand_thread_pointer): Use mips16_stub_call_address.
2818 (mips_output_mips16_rdhwr): Delete.
2819 (mips_finish_stub): New function.
2820 (mips_code_end): Use it to handle rdhwr stubs.
2822 2014-02-02 Uros Bizjak <ubizjak@gmail.com>
2825 * config/i386/i386.c (classify_argument): Fix handling of bit_offset
2826 when calculating size of integer atomic types.
2828 2014-02-02 H.J. Lu <hongjiu.lu@intel.com>
2830 * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
2832 2014-02-01 Jakub Jelinek <jakub@redhat.com>
2834 PR tree-optimization/60003
2835 * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
2836 * profile.c (branch_prob): Use gimple_call_builtin_p
2837 to check for BUILT_IN_SETJMP_RECEIVER.
2838 * tree-inline.c (copy_bb): Call notice_special_calls.
2840 2014-01-31 Vladimir Makarov <vmakarov@redhat.com>
2843 * lra-constraints.c (process_alt_operands): Update reload_sum only
2846 2014-01-31 Richard Henderson <rth@redhat.com>
2849 * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
2850 until after else_eh is processed.
2852 2014-01-31 Ilya Tocar <ilya.tocar@intel.com>
2854 * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
2855 (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
2856 (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
2857 in smmintrin.h, remove them.
2858 (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
2859 * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
2860 * config/i386/i386.md (ROUND_SAE): Fix value.
2861 * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
2862 (const48_operand): New.
2863 * config/i386/subst.md (round), (round_expand): Use
2864 const_4_or_8_to_11_operand.
2865 (round_saeonly), (round_saeonly_expand): Use const48_operand.
2867 2014-01-31 Ilya Tocar <ilya.tocar@intel.com>
2869 * config/i386/constraints.md (Yk): Swap meaning with k.
2870 * config/i386/i386.md (movhi_internal): Change Yk to k.
2871 (movqi_internal): Ditto.
2872 (*k<logic><mode>): Ditto.
2875 (kandn<mode>): Ditto.
2876 (*<code>hi_1): Ditto.
2877 (*<code>qi_1): Ditto.
2878 (kxnor<mode>): Ditto.
2879 (kortestzhi): Ditto.
2880 (kortestchi): Ditto.
2882 (*one_cmplhi2_1): Ditto.
2883 (*one_cmplqi2_1): Ditto.
2884 * config/i386/sse.md (): Change k to Yk.
2885 (avx512f_load<mode>_mask): Ditto.
2886 (avx512f_blendm<mode>): Ditto.
2887 (avx512f_store<mode>_mask): Ditto.
2888 (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
2889 (avx512f_storedqu<mode>_mask): Ditto.
2890 (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
2892 (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
2893 (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
2894 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
2895 (avx512f_maskcmp<mode>3): Ditto.
2896 (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
2897 (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
2898 (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
2899 (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
2900 (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
2901 (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
2902 (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
2903 (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
2904 (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
2905 (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
2906 (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
2907 (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
2908 (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
2909 (vec_extract_lo_<mode>_maskm): Ditto.
2910 (vec_extract_hi_<mode>_maskm): Ditto.
2911 (avx512f_vternlog<mode>_mask): Ditto.
2912 (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
2913 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
2914 (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
2915 (avx512f_<code>v8div16qi2_mask): Ditto.
2916 (avx512f_<code>v8div16qi2_mask_store): Ditto.
2917 (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
2918 (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
2919 (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
2920 (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
2921 (*avx512pf_gatherpf<mode>sf_mask): Ditto.
2922 (*avx512pf_gatherpf<mode>df_mask): Ditto.
2923 (*avx512pf_scatterpf<mode>sf_mask): Ditto.
2924 (*avx512pf_scatterpf<mode>df_mask): Ditto.
2925 (avx512cd_maskb_vec_dupv8di): Ditto.
2926 (avx512cd_maskw_vec_dupv16si): Ditto.
2927 (avx512f_vpermi2var<mode>3_maskz): Ditto.
2928 (avx512f_vpermi2var<mode>3_mask): Ditto.
2929 (avx512f_vpermi2var<mode>3_mask): Ditto.
2930 (avx512f_vpermt2var<mode>3_maskz): Ditto.
2931 (*avx512f_gathersi<mode>): Ditto.
2932 (*avx512f_gathersi<mode>_2): Ditto.
2933 (*avx512f_gatherdi<mode>): Ditto.
2934 (*avx512f_gatherdi<mode>_2): Ditto.
2935 (*avx512f_scattersi<mode>): Ditto.
2936 (*avx512f_scatterdi<mode>): Ditto.
2937 (avx512f_compress<mode>_mask): Ditto.
2938 (avx512f_compressstore<mode>_mask): Ditto.
2939 (avx512f_expand<mode>_mask): Ditto.
2940 * config/i386/subst.md (mask): Change k to Yk.
2941 (mask_scalar_merge): Ditto.
2944 2014-01-31 Marc Glisse <marc.glisse@inria.fr>
2946 * doc/extend.texi (Vector Extensions): Document ?: in C++.
2948 2014-01-31 Richard Biener <rguenther@suse.de>
2951 * builtins.c (fold_builtin_memory_op): Make sure to not
2952 use a floating-point mode or a boolean or enumeral type for
2955 2014-01-30 DJ Delorie <dj@redhat.com>
2957 * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
2958 * config/msp430/msp430.md (msp430_refsym_need_exit): New.
2959 * config/msp430/msp430.c (msp430_expand_epilogue): Call it
2960 whenever main() has an epilogue.
2962 2014-01-30 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2964 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
2965 unused variable "field".
2966 * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
2967 (vsx_mergeh_<mode>): Likewise.
2968 * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
2969 (altivec_vmrghh): Likewise.
2970 (altivec_vmrghw): Likewise.
2971 (altivec_vmrglb): Likewise.
2972 (altivec_vmrglh): Likewise.
2973 (altivec_vmrglw): Likewise.
2974 (altivec_vspltb): Add missing uses.
2975 (altivec_vsplth): Likewise.
2976 (altivec_vspltw): Likewise.
2977 (altivec_vspltsf): Likewise.
2979 2014-01-30 Jakub Jelinek <jakub@redhat.com>
2982 * ifcvt.c (cond_exec_process_insns): Don't conditionalize
2983 frame related instructions.
2985 2014-01-30 Vladimir Makarov <vmakarov@redhat.com>
2987 PR rtl-optimization/59959
2988 * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
2989 any reload of register whose subreg is invalid.
2991 2014-01-30 Jakub Jelinek <jakub@redhat.com>
2993 * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
2994 * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
2995 Add missing return type - void.
2997 2014-01-30 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2999 * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
3000 gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
3001 remove element index adjustment for endian (now handled in vsx.md
3003 (altivec_expand_vec_perm_const): Use
3004 gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
3005 * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
3006 (vsx_xxspltw_<mode>): Adjust element index for little endian.
3007 * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
3008 define_expand and a new define_insn *altivec_vspltb_internal;
3009 adjust for -maltivec=be on a little endian target.
3010 (altivec_vspltb_direct): New.
3011 (altivec_vsplth): Divide into a define_expand and a new
3012 define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
3013 little endian target.
3014 (altivec_vsplth_direct): New.
3015 (altivec_vspltw): Divide into a define_expand and a new
3016 define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
3017 little endian target.
3018 (altivec_vspltw_direct): New.
3019 (altivec_vspltsf): Divide into a define_expand and a new
3020 define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
3021 a little endian target.
3023 2014-01-30 Richard Biener <rguenther@suse.de>
3025 PR tree-optimization/59993
3026 * tree-ssa-forwprop.c (associate_pointerplus): Check we
3027 can propagate form the earlier stmt and avoid the transform
3028 when the intermediate result is needed.
3030 2014-01-30 Alangi Derick <alangiderick@gmail.com>
3032 * README.Portability: Fix typo.
3034 2014-01-30 David Holsgrove <david.holsgrove@xilinx.com>
3036 * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
3037 comparison_operator with ordered_comparison_operator.
3039 2014-01-30 Nick Clifton <nickc@redhat.com>
3041 * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
3042 Rename to mn10300_store_multiple_regs.
3043 * config/mn10300/mn10300.c: Likewise.
3044 * config/mn10300/mn10300.md (store_movm): Fix typo: call
3045 store_multiple_regs.
3046 * config/mn10300/predicates.md (mn10300_store_multiple_operation):
3047 Call mn10300_store_multiple_regs.
3049 2014-01-30 Nick Clifton <nickc@redhat.com>
3050 DJ Delorie <dj@redhat.com>
3052 * config/rl78/rl78.c (register_sizes): Make the "upper half" of
3053 %fp 2 to keep registers after it properly word-aligned.
3054 (rl78_alloc_physical_registers_umul): Handle the case where both
3055 input operands are the same.
3057 2014-01-30 Richard Biener <rguenther@suse.de>
3059 PR tree-optimization/59903
3060 * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
3063 2014-01-30 Jason Merrill <jason@redhat.com>
3066 * tree.c (walk_type_fields): Handle VECTOR_TYPE.
3069 * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
3071 2014-01-30 Richard Biener <rguenther@suse.de>
3073 PR tree-optimization/59951
3074 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
3076 2014-01-30 Savin Zlobec <savin.zlobec@gmail.com>
3079 * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
3080 SFmode to DFmode case.
3082 2014-01-29 DJ Delorie <dj@redhat.com>
3084 * config/msp430/msp430.opt (-minrt): New.
3085 * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
3087 (ENDFILE_SPEC): Likewise.
3089 2014-01-29 Jan Hubicka <hubicka@ucw.cz>
3091 * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
3092 (estimate_function_body_sizes): Use it.
3094 2014-01-29 Paolo Carlini <paolo.carlini@oracle.com>
3097 * dwarf2out.c (is_cxx_auto): New.
3098 (is_base_type): Use it.
3099 (gen_type_die_with_usage): Likewise.
3101 2014-01-29 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3103 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Use
3104 CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
3105 * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
3106 -maltivec=be with LE targets.
3107 (vsx_mergeh_<mode>): Likewise.
3108 * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
3109 (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
3110 (altivec_vmrghb): Replace with define_expand and new
3111 *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
3112 (altivec_vmrghb_direct): New define_insn.
3113 (altivec_vmrghh): Replace with define_expand and new
3114 *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
3115 (altivec_vmrghh_direct): New define_insn.
3116 (altivec_vmrghw): Replace with define_expand and new
3117 *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
3118 (altivec_vmrghw_direct): New define_insn.
3119 (*altivec_vmrghsf): Adjust for endianness.
3120 (altivec_vmrglb): Replace with define_expand and new
3121 *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
3122 (altivec_vmrglb_direct): New define_insn.
3123 (altivec_vmrglh): Replace with define_expand and new
3124 *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
3125 (altivec_vmrglh_direct): New define_insn.
3126 (altivec_vmrglw): Replace with define_expand and new
3127 *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
3128 (altivec_vmrglw_direct): New define_insn.
3129 (*altivec_vmrglsf): Adjust for endianness.
3130 (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
3131 (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
3132 (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
3133 (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
3134 (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
3135 (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
3136 (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
3137 (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
3139 2014-01-29 Marcus Shawcroft <marcus.shawcroft@arm.com>
3141 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
3142 (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
3145 2014-01-29 Richard Biener <rguenther@suse.de>
3147 PR tree-optimization/58742
3148 * tree-ssa-forwprop.c (associate_pointerplus): Rename to
3149 associate_pointerplus_align.
3150 (associate_pointerplus_diff): New function.
3151 (associate_pointerplus): Likewise. Call associate_pointerplus_align
3152 and associate_pointerplus_diff.
3154 2014-01-29 Richard Biener <rguenther@suse.de>
3156 * lto-streamer.h (LTO_major_version): Bump to 3.
3157 (LTO_minor_version): Reset to 0.
3159 2014-01-29 Renlin Li <Renlin.Li@arm.com>
3161 * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
3162 * config/arm/arm.c (FL_FOR_ARCH7VE): New.
3163 (arm_file_start): Generate correct asm header for armv7ve.
3164 * config/arm/bpabi.h: Add multilib support for armv7ve.
3165 * config/arm/driver-arm.c: Change the architectures of cortex-a7
3166 and cortex-a15 to armv7ve.
3167 * config/arm/t-aprofile: Add multilib support for armv7ve.
3168 * doc/invoke.texi: Document -march=armv7ve.
3170 2014-01-29 Richard Biener <rguenther@suse.de>
3172 PR tree-optimization/58742
3173 * tree-ssa-forwprop.c (associate_plusminus): Return true
3174 if we changed sth, defer EH cleanup to ...
3175 (ssa_forward_propagate_and_combine): ... here. Call simplify_mult.
3176 (simplify_mult): New function.
3178 2014-01-29 Jakub Jelinek <jakub@redhat.com>
3181 PR tree-optimization/59920
3182 * tree.c (build_common_builtin_nodes): Remove
3183 __builtin_setjmp_dispatcher initialization.
3184 * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
3185 * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
3186 instead of gsi_after_labels + manually skipping debug stmts.
3187 Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
3188 ignore bbs with IFN_ABNORMAL_DISPATCHER.
3189 * tree-inline.c (copy_edges_for_bb): Remove
3190 can_make_abnormal_goto argument, instead add abnormal_goto_dest
3191 argument. Ignore computed_goto_p stmts. Don't call
3192 make_abnormal_goto_edges. If a call might need abnormal edges
3193 for non-local gotos, see if it already has an edge to
3194 IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
3195 with true argument, don't do anything then, otherwise add
3196 EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
3197 (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
3199 * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
3200 (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
3201 (lower_stmt): Don't set data->calls_builtin_setjmp.
3202 (lower_builtin_setjmp): Adjust comment.
3203 * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
3204 * tree-cfg.c (found_computed_goto): Remove.
3205 (factor_computed_gotos): Remove.
3206 (make_goto_expr_edges): Return bool, true for computed gotos.
3207 Don't call make_abnormal_goto_edges.
3208 (build_gimple_cfg): Don't set found_computed_goto, don't call
3209 factor_computed_gotos.
3210 (computed_goto_p): No longer static.
3211 (make_blocks): Don't set found_computed_goto.
3212 (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
3213 (make_edges): If make_goto_expr_edges returns true, push bb
3214 into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
3215 instead of calling make_abnormal_goto_edges push bb into ab_edge_call
3216 vector. Record mapping between bbs and OpenMP regions if there
3217 are any, adjust make_gimple_omp_edges caller. Call
3218 handle_abnormal_edges.
3219 (make_abnormal_goto_edges): Remove.
3220 * tree-cfg.h (make_abnormal_goto_edges): Remove.
3221 (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
3222 * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
3223 * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
3224 * internal-fn.def (ABNORMAL_DISPATCHER): New.
3225 * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
3226 filling *region also set *region_idx to (*region)->entry->index.
3229 * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
3230 For REGs set ORIGINAL_REGNO.
3232 2014-01-29 Bingfeng Mei <bmei@broadcom.com>
3234 * doc/md.texi: Mention that a target shouldn't implement
3235 vec_widen_(s|u)mul_even/odd pair if it is less efficient
3238 2014-01-29 Jakub Jelinek <jakub@redhat.com>
3240 PR tree-optimization/59594
3241 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
3242 a copy of the datarefs vector rather than the vector itself.
3244 2014-01-28 Jason Merrill <jason@redhat.com>
3247 * dwarf2out.c (auto_die): New static.
3248 (gen_type_die_with_usage): Handle C++1y 'auto'.
3249 (gen_subprogram_die): If in-class DIE had 'auto', emit type again
3252 2014-01-28 H.J. Lu <hongjiu.lu@intel.com>
3255 * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
3256 (SPEC_X32): Likewise.
3257 (SPEC_64): Likewise.
3258 * config/i386/i386.c (ix86_option_override_internal): Turn off
3259 OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
3261 (x86_file_start): Output .code16gcc for TARGET_16BIT.
3262 * config/i386/i386.h (TARGET_16BIT): New macro.
3263 (TARGET_16BIT_P): Likewise.
3264 * config/i386/i386.opt: Add m16.
3265 * doc/invoke.texi: Document -m16.
3267 2014-01-28 Jakub Jelinek <jakub@redhat.com>
3269 PR preprocessor/59935
3270 * input.c (location_get_source_line): Bail out on when line number
3271 is zero, and test the return value of lookup_or_add_file_to_cache_tab.
3273 2014-01-28 Richard Biener <rguenther@suse.de>
3275 PR tree-optimization/58742
3276 * tree-ssa-forwprop.c (associate_plusminus): Handle
3277 pointer subtraction of the form (T)(P + A) - (T)P.
3279 2014-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3281 * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
3284 2014-01-28 Richard Biener <rguenther@suse.de>
3287 2014-01-28 Richard Biener <rguenther@suse.de>
3289 PR rtl-optimization/45364
3290 PR rtl-optimization/59890
3291 * var-tracking.c (local_get_addr_clear_given_value): Handle
3292 already cleared slot.
3293 (val_reset): Handle not allocated local_get_addr_cache.
3294 (vt_find_locations): Use post-order on the inverted CFG.
3296 2014-01-28 Richard Biener <rguenther@suse.de>
3298 * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
3300 2014-01-28 Richard Biener <rguenther@suse.de>
3302 PR rtl-optimization/45364
3303 PR rtl-optimization/59890
3304 * var-tracking.c (local_get_addr_clear_given_value): Handle
3305 already cleared slot.
3306 (val_reset): Handle not allocated local_get_addr_cache.
3307 (vt_find_locations): Use post-order on the inverted CFG.
3309 2014-01-28 Alan Modra <amodra@gmail.com>
3311 * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
3312 * configure.ac <recursive call for build != host>: Define
3313 GENERATOR_FILE. Comment. Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
3314 and LD_FOR_BUILD too.
3315 * configure: Regenerate.
3317 2014-01-27 Allan Sandfeld Jensen <sandfeld@kde.org>
3319 * config/i386/i386.c (get_builtin_code_for_version): Separate
3320 Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
3321 Broadwell from Haswell.
3323 2014-01-27 Steve Ellcey <sellcey@mips.com>
3325 * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3326 Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
3327 * config/mips/mips.c (mips_option_override): Change setting
3329 * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
3330 * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
3331 Change from Mask to Var.
3333 2014-01-27 Jeff Law <law@redhat.com>
3335 * ipa-inline.c (inline_small_functions): Fix typo.
3337 2014-01-27 Ilya Tocar <ilya.tocar@intel.com>
3339 * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
3340 (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
3341 (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
3342 (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
3343 (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
3344 (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
3345 (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
3346 (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
3347 (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
3348 (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
3349 (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
3350 (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
3351 (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
3352 (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
3353 (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
3354 (_mm512_storeu_epi64): Ditto.
3355 (_mm512_cmpge_epi32_mask): Ditto.
3356 (_mm512_cmpge_epu32_mask): Ditto.
3357 (_mm512_cmpge_epi64_mask): Ditto.
3358 (_mm512_cmpge_epu64_mask): Ditto.
3359 (_mm512_cmple_epi32_mask): Ditto.
3360 (_mm512_cmple_epu32_mask): Ditto.
3361 (_mm512_cmple_epi64_mask): Ditto.
3362 (_mm512_cmple_epu64_mask): Ditto.
3363 (_mm512_cmplt_epi32_mask): Ditto.
3364 (_mm512_cmplt_epu32_mask): Ditto.
3365 (_mm512_cmplt_epi64_mask): Ditto.
3366 (_mm512_cmplt_epu64_mask): Ditto.
3367 (_mm512_cmpneq_epi32_mask): Ditto.
3368 (_mm512_cmpneq_epu32_mask): Ditto.
3369 (_mm512_cmpneq_epi64_mask): Ditto.
3370 (_mm512_cmpneq_epu64_mask): Ditto.
3371 (_mm512_expand_pd): Ditto.
3372 (_mm512_expand_ps): Ditto.
3373 * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
3374 VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
3375 VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
3376 * config/i386/i386.c (ix86_builtins): Add
3377 IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
3378 IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
3379 IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
3380 IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
3381 IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
3382 IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
3383 IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
3384 IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
3385 IX86_BUILTIN_PMOVUSQW512_MEM.
3386 (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
3387 __builtin_ia32_pmovsqd512mem_mask,
3388 __builtin_ia32_pmovqd512mem_mask,
3389 __builtin_ia32_pmovusqw512mem_mask,
3390 __builtin_ia32_pmovsqw512mem_mask,
3391 __builtin_ia32_pmovqw512mem_mask,
3392 __builtin_ia32_pmovusdw512mem_mask,
3393 __builtin_ia32_pmovsdw512mem_mask,
3394 __builtin_ia32_pmovdw512mem_mask,
3395 __builtin_ia32_pmovqb512mem_mask,
3396 __builtin_ia32_pmovusqb512mem_mask,
3397 __builtin_ia32_pmovsqb512mem_mask,
3398 __builtin_ia32_pmovusdb512mem_mask,
3399 __builtin_ia32_pmovsdb512mem_mask,
3400 __builtin_ia32_pmovdb512mem_mask.
3401 (bdesc_args): Add __builtin_ia32_expanddf512,
3402 __builtin_ia32_expandsf512.
3403 (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
3404 VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
3405 VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
3406 * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
3407 (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
3408 (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
3409 (avx512f_<code>v8div16qi2_mask_store): This.
3410 (avx512f_expand<mode>): New.
3412 2014-01-27 Kirill Yukhin <kirill.yukhin@intel.com>
3414 * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
3416 (_mm512_mask_prefetch_i64gather_pd): Ditto.
3417 (_mm512_prefetch_i32scatter_pd): Ditto.
3418 (_mm512_mask_prefetch_i32scatter_pd): Ditto.
3419 (_mm512_prefetch_i64scatter_pd): Ditto.
3420 (_mm512_mask_prefetch_i64scatter_pd): Ditto.
3421 (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
3422 (_mm512_mask_prefetch_i64gather_ps): Ditto.
3423 (_mm512_prefetch_i32scatter_ps): Ditto.
3424 (_mm512_mask_prefetch_i32scatter_ps): Ditto.
3425 (_mm512_prefetch_i64scatter_ps): Ditto.
3426 (_mm512_mask_prefetch_i64scatter_ps): Ditto.
3427 * config/i386/i386-builtin-types.def: Define
3428 VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
3429 and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
3430 * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
3431 IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
3432 IX86_BUILTIN_SCATTERPFQPD.
3433 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
3434 __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
3435 __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
3436 __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
3437 __builtin_ia32_scatterpfqps.
3438 (ix86_expand_builtin): Expand new built-ins.
3439 * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
3440 fix memory access data type.
3441 (*avx512pf_gatherpf<mode>_mask): Ditto.
3442 (*avx512pf_gatherpf<mode>): Ditto.
3443 (avx512pf_scatterpf<mode>): Ditto.
3444 (*avx512pf_scatterpf<mode>_mask): Ditto.
3445 (*avx512pf_scatterpf<mode>): Ditto.
3446 (GATHER_SCATTER_SF_MEM_MODE): New.
3447 (avx512pf_gatherpf<mode>df): Ditto.
3448 (*avx512pf_gatherpf<mode>df_mask): Ditto.
3449 (*avx512pf_scatterpf<mode>df): Ditto.
3451 2014-01-27 Jakub Jelinek <jakub@redhat.com>
3454 * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
3455 know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
3458 2014-01-27 James Greenhalgh <james.greenhalgh@arm.com>
3460 * common/config/arm/arm-common.c
3461 (arm_rewrite_mcpu): Handle multiple names.
3463 (BIG_LITTLE_SPEC): Do not discard mcpu switches.
3465 2014-01-27 James Greenhalgh <james.greenhalgh@arm.com>
3467 * gimple-builder.h (create_gimple_tmp): Delete.
3469 2014-01-27 Christian Bruel <christian.bruel@st.com>
3471 * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
3474 2014-01-26 John David Anglin <danglin@gcc.gnu.org>
3476 * config/pa/pa.md (call): Generate indirect long calls to non-local
3477 functions when outputing 32-bit code.
3478 (call_value): Likewise except for special call to buggy powf function.
3480 * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
3481 portable runtime and PIC indirect calls.
3482 (pa_output_indirect_call): Remove unnecessary nop from portable runtime
3483 and PIC call sequences. Use ldo instead of blr to set return register
3484 in PIC call sequence.
3486 2014-01-25 Walter Lee <walt@tilera.com>
3488 * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
3489 avoid clobbering a live register.
3491 2014-01-25 Walter Lee <walt@tilera.com>
3493 * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
3494 Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
3495 * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
3496 Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
3498 2014-01-25 Walter Lee <walt@tilera.com>
3500 * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
3501 arguments on even registers.
3502 (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
3504 * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
3505 (BIGGEST_ALIGNMENT): Ditto.
3506 (BIGGEST_FIELD_ALIGNMENT): Ditto.
3508 2014-01-25 Walter Lee <walt@tilera.com>
3510 * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
3511 insns before bundling.
3512 * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
3514 2014-01-25 Walter Lee <walt@tilera.com>
3516 * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
3517 PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
3518 * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
3520 2014-01-25 Richard Sandiford <rdsandiford@googlemail.com>
3522 * config/mips/constraints.md (kl): Delete.
3523 * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
3524 define expands, using...
3525 (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
3526 instructions for MIPS16.
3527 (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
3528 non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
3530 2014-01-25 Walter Lee <walt@tilera.com>
3532 * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
3536 2014-01-25 Walter Lee <walt@tilera.com>
3538 * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
3539 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3541 2014-01-25 Richard Sandiford <rdsandiford@googlemail.com>
3543 * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
3546 2014-01-25 Jakub Jelinek <jakub@redhat.com>
3548 * print-rtl.c (in_call_function_usage): New var.
3549 (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
3550 EXPR_LIST mode as mode and not as reg note name.
3553 * cfgloopmanip.c (copy_loop_info): If
3554 loop->warned_aggressive_loop_optimizations, make sure
3555 the flag is set in target loop too.
3557 2014-01-24 Balaji V. Iyer <balaji.v.iyer@intel.com>
3559 * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
3561 * builtins.def: Likewise.
3562 * cilk.h (fn_contains_cilk_spawn_p): Likewise.
3563 * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
3564 * ira.c (ira_setup_eliminable_regset): Likewise.
3565 * omp-low.c (gate_expand_omp): Likewise.
3566 (execute_lower_omp): Likewise.
3567 (diagnose_sb_0): Likewise.
3568 (gate_diagnose_omp_blocks): Likewise.
3569 (simd_clone_clauses_extract): Likewise.
3572 2014-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3574 * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
3575 correction for little endian...
3576 * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
3579 2014-01-24 Jeff Law <law@redhat.com>
3581 PR tree-optimization/59919
3582 * tree-vrp.c (find_assert_locations_1): Do not register asserts
3583 for non-returning calls.
3585 2014-01-24 James Greenhalgh <james.greenhalgh@arm.com>
3587 * common/config/aarch64/aarch64-common.c
3588 (aarch64_rewrite_mcpu): Handle multiple names.
3589 * config/aarch64/aarch64.h
3590 (BIG_LITTLE_SPEC): Do not discard mcpu switches.
3592 2014-01-24 Dodji Seketeli <dodji@redhat.com>
3594 * input.c (add_file_to_cache_tab): Handle the case where fopen
3597 2014-01-23 H.J. Lu <hongjiu.lu@intel.com>
3600 * config/i386/i386.md (pushsf splitter): Get stack adjustment
3601 from push operand if code of push isn't PRE_DEC.
3603 2014-01-23 Michael Meissner <meissner@linux.vnet.ibm.com>
3606 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3607 -mquad-memory-atomic. Update -mquad-memory documentation to say
3608 it is only used for non-atomic loads/stores.
3610 * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
3611 -mquad-memory or -mquad-memory-atomic switches.
3613 * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
3614 -mquad-memory-atomic to ISA 2.07 support.
3616 * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
3617 to separate support of normal quad word memory operations (ldq, stq)
3618 from the atomic quad word memory operations.
3620 * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
3621 support to separate non-atomic quad word operations from atomic
3622 quad word operations. Disable non-atomic quad word operations in
3623 little endian mode so that we don't have to swap words after the
3624 load and before the store.
3625 (quad_load_store_p): Add comment about atomic quad word support.
3626 (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
3627 options printed with -mdebug=reg.
3629 * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
3630 -mquad-memory-atomic as the test for whether we have quad word
3631 atomic instructions.
3632 (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
3633 or -mp8-vector are used, allow byte/half-word atomic operations.
3635 * config/rs6000/sync.md (load_lockedti): Insure that the address
3636 is a proper indexed or indirect address for the lqarx instruction.
3637 On little endian systems, swap the hi/lo registers after the lqarx
3639 (load_lockedpti): Use indexed_or_indirect_operand predicate to
3640 insure the address is valid for the lqarx instruction.
3641 (store_conditionalti): Insure that the address is a proper indexed
3642 or indirect address for the stqcrx. instruction. On little endian
3643 systems, swap the hi/lo registers before doing the stqcrx.
3645 (store_conditionalpti): Use indexed_or_indirect_operand predicate to
3646 insure the address is valid for the stqcrx. instruction.
3648 * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
3649 Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
3650 type of quad memory support is available.
3652 2014-01-23 Vladimir Makarov <vmakarov@redhat.com>
3655 * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
3656 there is a danger of looping.
3658 2014-01-23 Pat Haugen <pthaugen@us.ibm.com>
3660 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3661 force flag_ira_loop_pressure if set via command line.
3663 2014-01-23 Alex Velenko <Alex.Velenko@arm.com>
3665 * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
3666 (ashr_simd): New builtin handling DI mode.
3667 * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
3668 (aarch64_sshr_simddi): New match pattern.
3669 * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
3670 (vshrd_n_s64): Likewise.
3671 * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
3673 2014-01-23 Nick Clifton <nickc@redhat.com>
3675 * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
3676 (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
3677 favour of mcu specific scripts.
3678 * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
3681 2014-01-23 James Greenhalgh <james.greenhalgh@arm.com>
3682 Alex Velenko <Alex.Velenko@arm.com>
3684 * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
3685 (vaddv_s16): Likewise.
3686 (vaddv_s32): Likewise.
3687 (vaddv_u8): Likewise.
3688 (vaddv_u16): Likewise.
3689 (vaddv_u32): Likewise.
3690 (vaddvq_s8): Likewise.
3691 (vaddvq_s16): Likewise.
3692 (vaddvq_s32): Likewise.
3693 (vaddvq_s64): Likewise.
3694 (vaddvq_u8): Likewise.
3695 (vaddvq_u16): Likewise.
3696 (vaddvq_u32): Likewise.
3697 (vaddvq_u64): Likewise.
3698 (vaddv_f32): Likewise.
3699 (vaddvq_f32): Likewise.
3700 (vaddvq_f64): Likewise.
3701 (vmaxv_f32): Likewise.
3702 (vmaxv_s8): Likewise.
3703 (vmaxv_s16): Likewise.
3704 (vmaxv_s32): Likewise.
3705 (vmaxv_u8): Likewise.
3706 (vmaxv_u16): Likewise.
3707 (vmaxv_u32): Likewise.
3708 (vmaxvq_f32): Likewise.
3709 (vmaxvq_f64): Likewise.
3710 (vmaxvq_s8): Likewise.
3711 (vmaxvq_s16): Likewise.
3712 (vmaxvq_s32): Likewise.
3713 (vmaxvq_u8): Likewise.
3714 (vmaxvq_u16): Likewise.
3715 (vmaxvq_u32): Likewise.
3716 (vmaxnmv_f32): Likewise.
3717 (vmaxnmvq_f32): Likewise.
3718 (vmaxnmvq_f64): Likewise.
3719 (vminv_f32): Likewise.
3720 (vminv_s8): Likewise.
3721 (vminv_s16): Likewise.
3722 (vminv_s32): Likewise.
3723 (vminv_u8): Likewise.
3724 (vminv_u16): Likewise.
3725 (vminv_u32): Likewise.
3726 (vminvq_f32): Likewise.
3727 (vminvq_f64): Likewise.
3728 (vminvq_s8): Likewise.
3729 (vminvq_s16): Likewise.
3730 (vminvq_s32): Likewise.
3731 (vminvq_u8): Likewise.
3732 (vminvq_u16): Likewise.
3733 (vminvq_u32): Likewise.
3734 (vminnmv_f32): Likewise.
3735 (vminnmvq_f32): Likewise.
3736 (vminnmvq_f64): Likewise.
3738 2014-01-23 James Greenhalgh <james.greenhalgh@arm.com>
3740 * config/aarch64/aarch64-simd.md
3741 (aarch64_dup_lane<mode>): Correct lane number on big-endian.
3742 (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
3743 (*aarch64_mul3_elt<mode>): Likewise.
3744 (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
3745 (*aarch64_mul3_elt_to_64v2df): Likewise.
3746 (*aarch64_mla_elt<mode>): Likewise.
3747 (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
3748 (*aarch64_mls_elt<mode>): Likewise.
3749 (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
3750 (*aarch64_fma4_elt<mode>): Likewise.
3751 (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
3752 (*aarch64_fma4_elt_to_64v2df): Likewise.
3753 (*aarch64_fnma4_elt<mode>): Likewise.
3754 (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
3755 (*aarch64_fnma4_elt_to_64v2df): Likewise.
3756 (aarch64_sq<r>dmulh_lane<mode>): Likewise.
3757 (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
3758 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
3759 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
3760 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
3761 (aarch64_sqdmull_lane<mode>_internal): Likewise.
3762 (aarch64_sqdmull2_lane<mode>_internal): Likewise.
3764 2013-01-23 Alex Velenko <Alex.Velenko@arm.com>
3766 * config/aarch64/aarch64-simd.md
3767 (aarch64_be_checked_get_lane<mode>): New define_expand.
3768 * config/aarch64/aarch64-simd-builtins.def
3769 (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
3770 New builtin definition.
3771 * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
3772 Use new safe be builtin.
3774 2014-01-23 Alex Velenko <Alex.Velenko@arm.com>
3776 * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
3778 (aarch64_be_st1<mode>): Likewise.
3779 (aarch_ld1<VALL:mode>): Define_expand modified.
3780 (aarch_st1<VALL:mode>): Likewise.
3781 * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
3782 (UNSPEC_ST1): Likewise.
3784 2014-01-23 David Holsgrove <david.holsgrove@xilinx.com>
3786 * config/microblaze/microblaze.md: Add trap insn and attribute
3788 2014-01-23 Dodji Seketeli <dodji@redhat.com>
3790 PR preprocessor/58580
3791 * input.h (location_get_source_line): Take an additional line_size
3793 (void diagnostics_file_cache_fini): Declare new function.
3794 * input.c (struct fcache): New type.
3795 (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
3796 New static constants.
3797 (diagnostic_file_cache_init, total_lines_num)
3798 (lookup_file_in_cache_tab, evicted_cache_tab_entry)
3799 (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
3800 (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
3801 (get_next_line, read_next_line, goto_next_line, read_line_num):
3802 New static function definitions.
3803 (diagnostic_file_cache_fini): New function.
3804 (location_get_source_line): Take an additional output line_len
3805 parameter. Re-write using lookup_or_add_file_to_cache_tab and
3807 * diagnostic.c (diagnostic_finish): Call
3808 diagnostic_file_cache_fini.
3809 (adjust_line): Take an additional input parameter for the length
3810 of the line, rather than calculating it with strlen.
3811 (diagnostic_show_locus): Adjust the use of
3812 location_get_source_line and adjust_line with respect to their new
3813 signature. While displaying a line now, do not stop at the first
3814 null byte. Rather, display the zero byte as a space and keep
3815 going until we reach the size of the line.
3816 * Makefile.in: Add vec.o to OBJS-libcommon
3818 2014-01-23 Kirill Yukhin <kirill.yukhin@intel.com>
3819 Ilya Tocar <ilya.tocar@intel.com>
3821 * config/i386/avx512fintrin.h (_mm512_kmov): New.
3822 * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
3823 (__builtin_ia32_kmov16): Ditto.
3824 * config/i386/i386.md (UNSPEC_KMOV): New.
3827 2014-01-23 Kirill Yukhin <kirill.yukhin@intel.com>
3829 * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
3830 (_mm512_storeu_si512): Ditto.
3832 2014-01-23 Richard Sandiford <rdsandiford@googlemail.com>
3835 * rtl.h (get_referenced_operands): Declare.
3836 * recog.c (get_referenced_operands): New function.
3837 * config/mips/mips.c (mips_reorg_process_insns): Check which asm
3838 operands have been referenced when recording LO_SUM references.
3840 2014-01-22 David Holsgrove <david.holsgrove@xilinx.com>
3842 * config/microblaze/microblaze.md: Correct bswaphi2 insn.
3844 2014-01-22 Jan Hubicka <hubicka@ucw.cz>
3846 * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
3847 Enable for generic and recent AMD targets.
3849 2014-01-22 Jan Hubicka <hubicka@ucw.cz>
3851 * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
3852 ARG_SIZE note when adjustment was eliminated.
3854 2014-01-22 Jeff Law <law@redhat.com>
3856 PR tree-optimization/59597
3857 * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
3858 in file. Accept new argument REGISTERING and use it to modify
3859 dump output appropriately.
3860 (register_jump_thread): Corresponding changes.
3861 (mark_threaded_blocks): Reinstate code to cancel unprofitable
3862 thread paths involving joiner blocks. Add code to dump cancelled
3863 jump threading paths.
3865 2014-01-22 Vladimir Makarov <vmakarov@redhat.com>
3867 PR rtl-optimization/59477
3868 * lra-constraints.c (inherit_in_ebb): Process call for living hard
3869 regs. Update reloads_num and potential_reload_hard_regs for all insns.
3871 2014-01-22 Tom Tromey <tromey@redhat.com>
3873 * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
3875 * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
3877 2014-01-21 Vladimir Makarov <vmakarov@redhat.com>
3879 PR rtl-optimization/59896
3880 * lra-constraints.c (process_alt_operands): Check unused note for
3881 matched operands of insn with no output reloads.
3883 2014-01-21 Richard Sandiford <rdsandiford@googlemail.com>
3885 * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
3886 (mips_move_from_gpr_cost): Likewise.
3888 2014-01-21 Vladimir Makarov <vmakarov@redhat.com>
3890 PR rtl-optimization/59858
3891 * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
3892 ira_class_hard_regs_num.
3893 (process_alt_operands): Increase reject for dying matched operand.
3895 2014-01-21 Jakub Jelinek <jakub@redhat.com>
3898 * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
3899 smaller than size, perform several stores or loads and stores
3900 at dst + count - size to store or copy all of size bytes, rather
3901 than just last modesize bytes.
3903 2014-01-20 DJ Delorie <dj@redhat.com>
3905 * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
3906 that CLOBBERs are REGs before propogating their values.
3908 2014-01-20 H.J. Lu <hongjiu.lu@intel.com>
3911 * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
3912 (cgraph_inline_failed_type): New function.
3913 * cgraph.h (DEFCIFCODE): Add type.
3914 (cgraph_inline_failed_type_t): New enum.
3915 (cgraph_inline_failed_type): New prototype.
3916 * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
3917 FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
3918 FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
3919 LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
3920 MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
3921 RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
3922 OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
3923 INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
3924 Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
3925 FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
3926 EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
3927 OPTIMIZATION_MISMATCH.
3928 * tree-inline.c (expand_call_inline): Emit errors during
3929 early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
3931 2014-01-20 Uros Bizjak <ubizjak@gmail.com>
3934 * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
3935 mode attribute in insn output.
3937 2014-01-20 Eric Botcazou <ebotcazou@adacore.com>
3939 * output.h (output_constant): Delete.
3940 * varasm.c (output_constant): Make private.
3942 2014-01-20 Alex Velenko <Alex.Velenko@arm.com>
3944 * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
3946 2014-01-20 Jakub Jelinek <jakub@redhat.com>
3949 * tree.h (fold_builtin_strcat): New prototype.
3950 * builtins.c (fold_builtin_strcat): No longer static. Add len
3951 argument, if non-NULL, don't call c_strlen. Optimize
3952 directly into __builtin_memcpy instead of __builtin_strcpy.
3953 (fold_builtin_2): Adjust fold_builtin_strcat caller.
3954 * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
3956 2014-01-20 Uros Bizjak <ubizjak@gmail.com>
3958 * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
3959 for SImode_address_operand operands, having only a REG argument.
3961 2014-01-20 Marcus Shawcroft <marcus.shawcroft@arm.com>
3963 * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
3964 loader name using mbig-endian.
3965 (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
3967 2014-01-20 James Greenhalgh <james.greenhalgh@arm.com>
3969 * doc/invoke.texi (-march): Clarify documentation for AArch64.
3973 2014-01-20 Tejas Belagod <tejas.belagod@arm.com>
3975 * config/aarch64/aarch64-protos.h
3976 (aarch64_cannot_change_mode_class_ptr): Declare.
3977 * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
3978 aarch64_cannot_change_mode_class_ptr): New.
3979 * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
3980 backend hook aarch64_cannot_change_mode_class.
3982 2014-01-20 James Greenhalgh <james.greenhalgh@arm.com>
3984 * common/config/aarch64/aarch64-common.c
3985 (aarch64_handle_option): Don't handle any option order logic here.
3986 * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
3987 selected_cpu, warn on architecture version mismatch.
3988 (aarch64_override_options): Fix parsing order for option strings.
3990 2014-01-20 Jan-Benedict Glaw <jbglaw@lug-owl.de>
3991 Iain Sandoe <iain@codesourcery.com>
3994 * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
3995 warning. Amend comment to reflect current functionality.
3997 2014-01-20 Richard Biener <rguenther@suse.de>
4000 * builtins.c (fold_builtin_strcat): Remove case better handled
4001 by tree-ssa-strlen.c.
4003 2014-01-20 Alan Lawrence <alan.lawrence@arm.com>
4005 * config/aarch64/aarch64.opt
4006 (mcpu, march, mtune): Make case-insensitive.
4008 2014-01-20 Jakub Jelinek <jakub@redhat.com>
4011 * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
4012 if operands[1] is a REG or ZERO_EXTEND of a REG.
4014 2014-01-19 Jan Hubicka <hubicka@ucw.cz>
4016 * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
4018 2014-01-19 John David Anglin <danglin@gcc.gnu.org>
4020 * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
4021 long non-pic millicode calls.
4023 2014-01-19 Jan-Benedict Glaw <jbglaw@lug-owl.de>
4025 * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
4027 2014-01-19 Kito Cheng <kito@0xlab.org>
4029 * builtins.c (expand_movstr): Check movstr expand done or fail.
4031 2014-01-18 Uros Bizjak <ubizjak@gmail.com>
4032 H.J. Lu <hongjiu.lu@intel.com>
4035 * config/i386/i386.md (*lea<mode>): Zero-extend return register
4036 to DImode for zero-extended addresses.
4038 2014-01-19 Jakub Jelinek <jakub@redhat.com>
4040 PR rtl-optimization/57763
4041 * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
4042 on the new indirect jump_insn and increment LABEL_NUSES (label).
4044 2014-01-18 H.J. Lu <hongjiu.lu@intel.com>
4048 * config.gcc (x86_archs): New variable.
4049 (x86_64_archs): Likewise.
4050 (x86_cpus): Likewise.
4051 Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
4052 --with-arch/--with-cpu= options.
4053 Support --with-arch=/--with-cpu={nehalem,westmere,
4054 sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
4056 2014-01-18 Uros Bizjak <ubizjak@gmail.com>
4058 * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
4059 and PROCESSOR_ATHLON to simplify code. Move "memory" calculation.
4061 2014-01-18 Uros Bizjak <ubizjak@gmail.com>
4063 * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
4065 2014-01-18 Jakub Jelinek <jakub@redhat.com>
4068 * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
4069 clear cpp_get_options (parse_in)->warn_unused_macros for
4070 ix86_target_macros_internal with cpp_define.
4072 2014-01-18 Richard Sandiford <rdsandiford@googlemail.com>
4074 * jump.c (delete_related_insns): Keep (use (insn))s.
4075 * reorg.c (redundant_insn): Check for barriers too.
4077 2014-01-17 H.J. Lu <hongjiu.lu@intel.com>
4079 * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
4081 2014-01-17 John David Anglin <danglin@gcc.gnu.org>
4083 * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
4084 call to $$dyncall when TARGET_LONG_CALLS is true.
4086 2014-01-17 Jeff Law <law@redhat.com>
4088 * ree.c (combine_set_extension): Temporarily disable test for
4089 changing number of hard registers.
4091 2014-01-17 Jan Hubicka <hubicka@ucw.cz>
4094 * ipa-inline-analysis.c (inline_free_summary):
4095 Do not free summary of aliases.
4097 2014-01-17 Jakub Jelinek <jakub@redhat.com>
4100 * gimplify.c (gimplify_expr): Use create_tmp_var
4101 instead of create_tmp_var_raw. If cond doesn't have
4102 integral type, don't add the IFN_ANNOTATE builtin at all.
4104 2014-01-17 Martin Jambor <mjambor@suse.cz>
4107 * ipa-cp.c (prev_edge_clone): New variable.
4108 (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
4109 Also resize prev_edge_clone vector.
4110 (ipcp_edge_duplication_hook): Also update prev_edge_clone.
4111 (ipcp_edge_removal_hook): New function.
4112 (ipcp_driver): Register ipcp_edge_removal_hook.
4114 2014-01-17 Andrew Pinski <apinski@cavium.com>
4115 Steve Ellcey <sellcey@mips.com>
4118 * config/mips/mips.c (mips_print_operand): Check operand mode instead
4121 2014-01-17 Jeff Law <law@redhat.com>
4124 * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
4125 so that pass_ccp runs first.
4127 2014-01-17 H.J. Lu <hongjiu.lu@intel.com>
4129 * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
4130 (ix86_adjust_cost): Use !TARGET_XXX.
4131 (do_reorder_for_imul): Likewise.
4132 (swap_top_of_ready_list): Likewise.
4133 (ix86_sched_reorder): Likewise.
4135 2014-01-17 H.J. Lu <hongjiu.lu@intel.com>
4137 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4138 PROCESSOR_INTEL. Treat like PROCESSOR_GENERIC.
4139 * config/i386/i386.c (intel_memcpy): New. Duplicate slm_memcpy.
4140 (intel_memset): New. Duplicate slm_memset.
4141 (intel_cost): New. Duplicate slm_cost.
4142 (m_INTEL): New macro.
4143 (processor_target_table): Add "intel".
4144 (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
4145 with PROCESSOR_INTEL for "intel".
4146 (ix86_lea_outperforms): Support PROCESSOR_INTEL. Duplicate
4147 PROCESSOR_SILVERMONT.
4148 (ix86_issue_rate): Likewise.
4149 (ix86_adjust_cost): Likewise.
4150 (ia32_multipass_dfa_lookahead): Likewise.
4151 (swap_top_of_ready_list): Likewise.
4152 (ix86_sched_reorder): Likewise.
4153 (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
4154 instead of TARGET_OPT_AGU.
4155 * config/i386/i386.h (TARGET_INTEL): New.
4156 (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
4157 (processor_type): Add PROCESSOR_INTEL.
4158 * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
4159 Add X86_TUNE_AVOID_LEA_FOR_ADDR.
4161 2014-01-17 Marek Polacek <polacek@redhat.com>
4164 * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
4167 2014-01-17 Richard Biener <rguenther@suse.de>
4169 PR tree-optimization/46590
4170 * opts.c (default_options_table): Add entries for
4171 OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
4172 all enabled at -O1 but not for -Og.
4173 * common.opt (fbranch-count-reg): Remove Init(1).
4174 (fmove-loop-invariants): Likewise.
4175 (ftree-pta): Likewise.
4177 2014-01-17 Jakub Jelinek <jakub@redhat.com>
4179 * config/i386/i386.c (ix86_data_alignment): For compatibility with
4180 (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
4181 decls to at least the GCC 4.8 used alignments.
4184 * tree-nested.c (convert_nonlocal_reference_stmt,
4185 convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
4186 of GIMPLE_BIND stmts, adjust associated decls.
4188 2014-01-17 Richard Biener <rguenther@suse.de>
4190 PR tree-optimization/46590
4191 * vec.h (vec<>::bseach): New member function implementing
4192 binary search according to C89 bsearch.
4193 (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
4194 * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
4195 bitmap pointer again. Make accesses_in_loop a flat array.
4196 (mem_ref_obstack): New global.
4197 (outermost_indep_loop): Adjust for mem_ref->stored changes.
4198 (mark_ref_stored): Likewise.
4199 (ref_indep_loop_p_2): Likewise.
4200 (set_ref_stored_in_loop): New helper function.
4201 (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
4202 (memref_free): Adjust.
4203 (record_mem_ref_loc): Simplify.
4204 (gather_mem_refs_stmt): Adjust.
4205 (sort_locs_in_loop_postorder_cmp): New function.
4206 (analyze_memory_references): Sort accesses_in_loop after
4207 loop postorder number.
4208 (find_ref_loc_in_loop_cmp): New function.
4209 (for_all_locs_in_loop): Find relevant cluster of locs in
4210 accesses_in_loop and iterate without recursion.
4211 (execute_sm): Avoid uninit warning.
4212 (struct ref_always_accessed): Simplify.
4213 (ref_always_accessed::operator ()): Likewise.
4214 (ref_always_accessed_p): Likewise.
4215 (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
4216 loop postorder numbers here.
4217 (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
4220 2014-01-17 Jan Hubicka <hubicka@ucw.cz>
4223 * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
4224 on decls for which assemble_alias has been called.
4226 2014-01-17 Nick Clifton <nickc@redhat.com>
4228 * config/msp430/msp430.opt: (mcpu): New option.
4229 * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
4230 (msp430_option_override): Parse target_cpu. If the MCU name
4231 matches a generic string, clear target_mcu.
4232 (msp430_attr): Allow numeric interrupt values up to 63.
4233 (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
4234 * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
4236 * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
4238 * config/msp430/msp430.md (popm): Use %J rather than %I.
4239 (addsi3): Use msp430_nonimmediate_operand for operand 2.
4240 (addhi_cy_i): Use immediate_operand for operand 2.
4241 * doc/invoke.texi: Document -mcpu option.
4243 2014-01-17 Richard Biener <rguenther@suse.de>
4245 PR rtl-optimization/38518
4246 * df.h (df_analyze_loop): Declare.
4247 * df-core.c: Include cfgloop.h.
4248 (df_analyze_1): Split out main part of df_analyze.
4249 (df_analyze): Adjust.
4250 (loop_inverted_post_order_compute): New function.
4251 (loop_post_order_compute): Likewise.
4252 (df_analyze_loop): New function avoiding whole-function
4254 * loop-invariant.c (find_defs): Use df_analyze_loop.
4255 (find_invariants): Adjust.
4256 * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
4258 2014-01-17 Zhenqiang Chen <zhenqiang.chen@arm.com>
4260 * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
4261 (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
4263 2014-01-16 Ilya Enkovich <ilya.enkovich@intel.com>
4265 * ipa-ref.c (ipa_remove_stmt_references): Fix references
4266 traversal when removing references.
4268 2014-01-16 Jan Hubicka <hubicka@ucw.cz>
4271 * tree.c (get_binfo_at_offset): Look harder for virtual bases.
4273 2014-01-16 Bernd Schmidt <bernds@codesourcery.com>
4276 * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
4277 pushing a reload for an autoinc when we had previously reloaded an
4278 inner part of the address.
4280 2014-01-16 Jakub Jelinek <jakub@redhat.com>
4282 * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
4284 (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
4285 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
4286 when not giving up or versioning for alias only because of
4288 (vect_analyze_data_ref_dependences): Set to true.
4289 * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
4291 (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
4292 LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
4296 * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
4299 * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
4300 operand 0 predicate for gathers, use a new pseudo as subtarget.
4302 2014-01-16 Vladimir Makarov <vmakarov@redhat.com>
4305 * lra-constraints.c (process_alt_operands): Add printing debug info.
4306 Check absence of input/output reloads for matched operands too.
4308 2014-01-16 Vladimir Makarov <vmakarov@redhat.com>
4310 PR rtl-optimization/59835
4311 * ira.c (ira_init_register_move_cost): Increase cost for
4314 2014-01-16 Alan Lawrence <alan.lawrence@arm.com>
4316 * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
4318 2014-01-16 Richard Earnshaw <rearnsha@arm.com>
4321 * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
4322 non-register objects. Use gen_(high/low)part more consistently.
4325 2014-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
4328 * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
4329 endian support, remove tests for WORDS_BIG_ENDIAN.
4330 (p8_mfvsrd_3_<mode>): Likewise.
4331 (reload_gpr_from_vsx<mode>): Likewise.
4332 (reload_gpr_from_vsxsf): Likewise.
4333 (p8_mfvsrd_4_disf): Likewise.
4335 2014-01-16 Richard Biener <rguenther@suse.de>
4337 PR rtl-optimization/46590
4338 * lcm.c (compute_antinout_edge): Use postorder iteration.
4339 (compute_laterin): Use inverted postorder iteration.
4341 2014-01-16 Nick Clifton <nickc@redhat.com>
4344 * varasm.c (output_constant): Return the number of bytes actually
4346 (output_constructor_array_range): Update the field size with the
4347 number of bytes emitted by output_constant.
4348 (output_constructor_regular_field): Likewise. Also do not
4349 complain if the total number of bytes emitted is now greater
4350 than the expected fieldpos.
4351 * output.h (output_constant): Update prototype and descriptive comment.
4353 2014-01-16 Marek Polacek <polacek@redhat.com>
4356 * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
4357 it is error_mark_node.
4359 2014-01-15 Uros Bizjak <ubizjak@gmail.com>
4361 * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
4362 VALID_AVX256_REG_OR_OI_MODE.
4364 2014-01-15 Pat Haugen <pthaugen@us.ibm.com>
4366 * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
4367 current procedure should be profiled.
4369 2014-01-15 Andrew Pinski <apinski@cavium.com>
4371 * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
4372 of moving from/to the STACK_REG register class.
4374 2014-01-15 Richard Henderson <rth@redhat.com>
4377 * reginfo.c (global_regs_decl): Globalize.
4378 * rtl.h (global_regs_decl): Declare.
4379 * ira.c (do_reload): Diagnose frame_pointer_needed and it
4380 reserved via global_regs.
4382 2014-01-15 Teresa Johnson <tejohnson@google.com>
4384 * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
4386 2014-01-15 Bill Schmidt <wschmidt@vnet.linux.ibm.com>
4388 * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
4389 and vmulosh rather than call gen_vec_widen_smult_*.
4390 (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
4391 than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
4392 (vec_widen_smult_even_v16qi): Likewise.
4393 (vec_widen_umult_even_v8hi): Likewise.
4394 (vec_widen_smult_even_v8hi): Likewise.
4395 (vec_widen_umult_odd_v16qi): Likewise.
4396 (vec_widen_smult_odd_v16qi): Likewise.
4397 (vec_widen_umult_odd_v8hi): Likewise.
4398 (vec_widen_smult_odd_v8hi): Likewise.
4399 (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
4400 vmuloub rather than call gen_vec_widen_umult_*.
4401 (vec_widen_umult_lo_v16qi): Likewise.
4402 (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
4403 vmulosb rather than call gen_vec_widen_smult_*.
4404 (vec_widen_smult_lo_v16qi): Likewise.
4405 (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
4406 rather than call gen_vec_widen_umult_*.
4407 (vec_widen_umult_lo_v8hi): Likewise.
4408 (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
4409 rather than call gen_vec_widen_smult_*.
4410 (vec_widen_smult_lo_v8hi): Likewise.
4412 2014-01-15 Jeff Law <law@redhat.com>
4414 PR tree-optimization/59747
4415 * ree.c (find_and_remove_re): Properly handle case where a second
4416 eliminated extension requires widening a copy created for elimination
4417 of a prior extension.
4418 (combine_set_extension): Ensure that the number of hard regs needed
4419 for a destination register does not change when we widen it.
4421 2014-01-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
4423 * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
4424 (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
4425 (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
4426 (arm*-*-rtems*): Use t-rtems from existing tmake_file.
4427 (avr-*-rtems*): Likewise.
4428 (bfin*-rtems*): Likewise.
4429 (moxie-*-rtems*): Likewise.
4430 (h8300-*-rtems*): Likewise.
4431 (i[34567]86-*-rtems*): Likewise.
4432 (lm32-*-rtems*): Likewise.
4433 (m32r-*-rtems*): Likewise.
4434 (m68k-*-rtems*): Likewise.
4435 (microblaze*-*-rtems*): Likewise.
4436 (mips*-*-rtems*): Likewise.
4437 (powerpc-*-rtems*): Likewise.
4438 (sh-*-rtems*): Likewise.
4439 (sparc-*-rtems*): Likewise.
4440 (sparc64-*-rtems*): Likewise.
4441 (v850-*-rtems*): Likewise.
4442 (m32c-*-rtems*): Likewise.
4444 2014-01-15 Vladimir Makarov <vmakarov@redhat.com>
4446 PR rtl-optimization/59511
4447 * ira.c (ira_init_register_move_cost): Use memory costs for some
4448 cases of register move cost calculations.
4449 * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
4450 instead of BB frequency.
4451 * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
4452 * lra-assigns.c (find_hard_regno_for): Ditto.
4454 2014-01-15 Richard Biener <rguenther@suse.de>
4456 PR tree-optimization/59822
4457 * tree-vect-stmts.c (hoist_defs_of_uses): New function.
4458 (vectorizable_load): Use it to hoist defs of uses of invariant
4459 loads out of the loop.
4461 2014-01-15 Matthew Gretton-Dann <matthew.gretton-dann@linaro.org>
4462 Kugan Vivekanandarajah <kuganv@linaro.org>
4465 * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
4468 2014-01-15 Richard Biener <rguenther@suse.de>
4470 PR rtl-optimization/59802
4471 * lcm.c (compute_available): Use inverted postorder to seed
4472 the initial worklist.
4474 2014-01-15 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
4477 * config/s390/s390.c (s390_preferred_reload_class): Don't return
4478 ADDR_REGS for invalid symrefs in non-PIC code.
4480 2014-01-15 Jakub Jelinek <jakub@redhat.com>
4483 * builtins.c (determine_block_size): Initialize *probable_max_size
4484 even if len_rtx is CONST_INT.
4486 2014-01-14 Andrew Pinski <apinski@cavium.com>
4488 * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
4489 * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
4490 (cortexa53_tunings): Likewise.
4491 (aarch64_sched_issue_rate): New function.
4492 (TARGET_SCHED_ISSUE_RATE): Define.
4494 2014-01-14 Vladimir Makarov <vmakarov@redhat.com>
4496 * ira-costs.c (find_costs_and_classes): Add missed
4497 ira_init_register_move_cost_if_necessary.
4499 2014-01-14 Vladimir Makarov <vmakarov@redhat.com>
4502 * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
4504 2014-01-14 H.J. Lu <hongjiu.lu@intel.com>
4507 * config/i386/i386.c (type_natural_mode): Add a bool parameter
4508 to indicate if type is used for function return value. Warn ABI
4509 change if the vector mode isn't available for function return value.
4510 (ix86_function_arg_advance): Pass false to type_natural_mode.
4511 (ix86_function_arg): Likewise.
4512 (ix86_gimplify_va_arg): Likewise.
4513 (function_arg_32): Don't warn ABI change.
4514 (ix86_function_value): Pass true to type_natural_mode.
4515 (ix86_return_in_memory): Likewise.
4516 (ix86_struct_value_rtx): Removed.
4517 (TARGET_STRUCT_VALUE_RTX): Likewise.
4519 2014-01-14 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
4521 * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
4522 converting a conditional jump into a conditional return.
4524 2014-01-14 Richard Biener <rguenther@suse.de>
4526 PR tree-optimization/58921
4527 PR tree-optimization/59006
4528 * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
4529 hoisting invariant stmts.
4530 * tree-vect-stmts.c (vectorizable_load): Insert the splat of
4531 invariant loads on the preheader edge if possible.
4533 2014-01-14 Joey Ye <joey.ye@arm.com>
4535 * doc/plugin.texi (Building GCC plugins): Update to C++.
4537 2014-01-14 Kirill Yukhin <kirill.yukhin@intel.com>
4539 * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
4540 (_mm_rcp28_round_ss): Ditto.
4541 (_mm_rsqrt28_round_sd): Ditto.
4542 (_mm_rsqrt28_round_ss): Ditto.
4543 (_mm_rcp28_sd): Ditto.
4544 (_mm_rcp28_ss): Ditto.
4545 (_mm_rsqrt28_sd): Ditto.
4546 (_mm_rsqrt28_ss): Ditto.
4547 * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
4548 * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
4549 * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
4550 (IX86_BUILTIN_RCP28SD): Ditto.
4551 (IX86_BUILTIN_RCP28SS): Ditto.
4552 (IX86_BUILTIN_RSQRT28SD): Ditto.
4553 (IX86_BUILTIN_RSQRT28SS): Ditto.
4554 (bdesc_special_args): Define __builtin_ia32_movntdqa512,
4555 __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
4556 __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
4557 (ix86_expand_special_args_builtin): Expand new FTYPE.
4558 * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
4559 (srcp14<mode>): Make insn unary.
4560 (avx512f_vmscalef<mode><round_name>): Use substed predicate.
4561 (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
4562 (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
4563 (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
4564 (avx512er_exp2<mode><mask_name><round_saeonly_name>):
4565 Fix rounding: make it SAE only.
4566 (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
4568 (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
4570 (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
4571 (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
4572 (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
4573 * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
4574 (round_saeonly_mask_scalar_operand4): Ditto.
4575 (round_saeonly_mask_scalar_op3): Ditto.
4576 (round_saeonly_mask_scalar_op4): Ditto.
4578 2014-01-13 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4580 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4581 Implement -maltivec=be for vec_insert and vec_extract.
4583 2014-01-10 DJ Delorie <dj@redhat.com>
4585 * config/msp430/msp430.md (call_internal): Don't allow memory
4586 references with SP as the base register.
4587 (call_value_internal): Likewise.
4588 * config/msp430/constraints.md (Yc): New. For memory references
4589 that don't use SP as a base register.
4591 * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
4592 "an integer without a # prefix"
4593 * config/msp430/msp430.md (epilogue_helper): Use it.
4595 2014-01-13 Jakub Jelinek <jakub@redhat.com>
4598 * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
4599 AVX512F gather builtins.
4600 * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
4601 on gather decls with INTEGER_TYPE masktype.
4602 (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
4603 directly into the builtin rather than hoisting it before loop.
4605 PR tree-optimization/59387
4606 * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
4607 (scev_const_prop): If folded_casts and type has undefined overflow,
4608 use force_gimple_operand instead of force_gimple_operand_gsi and
4609 for each added stmt if it is assign with
4610 arith_code_with_undefined_signed_overflow, call
4611 rewrite_to_defined_overflow.
4612 * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
4613 gimple-fold.h instead.
4614 (arith_code_with_undefined_signed_overflow,
4615 rewrite_to_defined_overflow): Moved to ...
4616 * gimple-fold.c (arith_code_with_undefined_signed_overflow,
4617 rewrite_to_defined_overflow): ... here. No longer static.
4618 Include gimplify-me.h.
4619 * gimple-fold.h (arith_code_with_undefined_signed_overflow,
4620 rewrite_to_defined_overflow): New prototypes.
4622 2014-01-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4624 * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
4626 2014-01-13 Eric Botcazou <ebotcazou@adacore.com>
4628 * builtins.c (get_object_alignment_2): Minor tweak.
4629 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
4631 2014-01-13 Christian Bruel <christian.bruel@st.com>
4633 * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
4634 optimized non constant lengths.
4636 2014-01-13 Jakub Jelinek <jakub@redhat.com>
4639 * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
4640 load as __atomic_load_N if possible.
4642 2014-01-11 David Edelsohn <dje.gcc@gmail.com>
4644 * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
4646 (rs6000_expand_builtin): Adjust call.
4648 2014-01-11 David Edelsohn <dje.gcc@gmail.com>
4651 * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
4652 * config/rs6000/rs6000.c: Include target-globals.h.
4653 (rs6000_set_current_function): Instead of doing target_reinit
4654 unconditionally, use save_target_globals_default_opts and
4655 restore_target_globals.
4657 * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
4659 * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
4660 (rs6000_expand_builtin): Handle mffs and mtfsf.
4661 (rs6000_init_builtins): Define mffs and mtfsf.
4662 * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
4663 (rs6000_mffs): New pattern.
4664 (rs6000_mtfsf): New pattern.
4666 2014-01-11 Bin Cheng <bin.cheng@arm.com>
4668 * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
4669 Start narrowing with START. Apply candidate-use pair
4670 and check overall cost in narrowing.
4671 (iv_ca_prune): Pass new argument.
4673 2014-01-10 Jeff Law <law@redhat.com>
4676 * ree.c (combine_reaching_defs): Ensure the defining statement
4677 occurs before the extension when optimizing extensions with
4678 different source and destination hard registers.
4680 2014-01-10 Jan Hubicka <hubicka@ucw.cz>
4683 * ipa-devirt.c (build_type_inheritance_graph): Also add types of
4684 vtables into the type inheritance graph.
4686 2014-01-10 Jakub Jelinek <jakub@redhat.com>
4688 PR rtl-optimization/59754
4689 * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
4690 modes in the REGNO != REGNO case.
4692 2014-01-10 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4694 * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
4696 2014-01-10 Jakub Jelinek <jakub@redhat.com>
4698 PR tree-optimization/59745
4699 * tree-predcom.c (tree_predictive_commoning_loop): Call
4700 free_affine_expand_cache if giving up because components is NULL.
4702 * target-globals.c (save_target_globals): Allocate < 4KB structs using
4703 GC in payload of target_globals struct instead of allocating them on
4704 the heap and the larger structs separately using GC.
4705 * target-globals.h (struct target_globals): Make regs, hard_regs,
4706 reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
4707 of GTY((skip)) and change type to void *.
4708 (reset_target_globals): Cast loads from those fields to corresponding
4711 2014-01-10 Steve Ellcey <sellcey@mips.com>
4714 * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
4715 gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
4716 tree-ssanames.h, print-tree.h, varasm.h, and context.h.
4718 2014-01-10 Richard Earnshaw <rearnsha@arm.com>
4721 * aarch64-modes.def (CC_Zmode): New flags mode.
4722 * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
4723 represents an equality.
4724 (aarch64_get_condition_code): Handle CC_Zmode.
4725 * aarch64.md (compare_neg<mode>): Restrict to equality operations.
4727 2014-01-10 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
4729 * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
4730 extraction in good case.
4732 2014-01-10 Richard Biener <rguenther@suse.de>
4734 PR tree-optimization/59374
4735 * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
4736 checking after SLP discovery. Mark stmts not participating
4737 in any SLP instance properly.
4739 2014-01-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4741 * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
4742 when handling a SET rtx.
4744 2014-01-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4746 * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
4747 (cortex-a57): Likewise.
4748 (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
4750 2014-01-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4752 * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
4753 non-iwmmxt builtins.
4755 2014-01-10 Jan Hubicka <hubicka@ucw.cz>
4759 * ipa-devirt.c record_target_from_binfo): Take as argument
4760 stack of binfos and lookup matching one for virtual inheritance.
4761 (possible_polymorphic_call_targets_1): Update.
4763 2014-01-10 Huacai Chen <chenhc@lemote.com>
4765 * config/mips/driver-native.c (host_detect_local_cpu): Handle new
4766 kernel strings for Loongson-2E/2F/3A.
4768 2014-01-10 Jakub Jelinek <jakub@redhat.com>
4771 * tree-vect-data-refs.c (vect_analyze_data_refs): Check
4772 is_gimple_call before calling gimple_call_internal_p.
4774 2014-01-09 Steve Ellcey <sellcey@mips.com>
4776 * Makefile.in (TREE_FLOW_H): Remove.
4777 (TREE_SSA_H): Add file names from tree-flow.h.
4778 * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
4779 * tree.h: Remove tree-flow.h reference.
4780 * hash-table.h: Remove tree-flow.h reference.
4781 * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
4782 reference with tree-ssa-loop.h.
4784 2014-01-09 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4786 * doc/invoke.texi: Add -maltivec={be,le} options, and document
4787 default element-order behavior for -maltivec.
4788 * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
4789 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
4790 that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
4791 when targeting big endian, at least for now.
4792 * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
4794 2014-01-09 Jakub Jelinek <jakub@redhat.com>
4797 * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
4798 var satisfies use_register_for_decl, just take into account type
4799 alignment, rather than decl alignment.
4801 PR tree-optimization/59622
4802 * gimple-fold.c (gimple_fold_call): Fix a typo in message. For
4803 __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
4804 __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
4805 Don't devirtualize for inplace at all. For targets.length () == 1,
4806 if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
4808 2014-01-09 H.J. Lu <hongjiu.lu@intel.com>
4810 * config/i386/i386.md (cpu): Remove the unused btver1.
4812 2014-01-09 H.J. Lu <hongjiu.lu@intel.com>
4814 * gdbasan.in: Put a breakpoint on __sanitizer::Report.
4816 2014-01-09 Jakub Jelinek <jakub@redhat.com>
4819 * tree-core.h (struct target_globals): New forward declaration.
4820 (struct tree_target_option): Add globals field.
4821 * tree.h (TREE_TARGET_GLOBALS): Define.
4822 (prepare_target_option_nodes_for_pch): New prototype.
4823 * target-globals.h (struct target_globals): Define even if
4825 * tree.c (prepare_target_option_node_for_pch,
4826 prepare_target_option_nodes_for_pch): New functions.
4827 * config/i386/i386.h (SWITCHABLE_TARGET): Define.
4828 * config/i386/i386.c: Include target-globals.h.
4829 (ix86_set_current_function): Instead of doing target_reinit
4830 unconditionally, use save_target_globals_default_opts and
4831 restore_target_globals.
4833 2014-01-09 Richard Biener <rguenther@suse.de>
4835 PR tree-optimization/59715
4836 * tree-cfg.h (split_critical_edges): Declare.
4837 * tree-cfg.c (split_critical_edges): Export.
4838 * tree-ssa-sink.c (execute_sink_code): Split critical edges.
4840 2014-01-09 Max Ostapenko <m.ostapenko@partner.samsung.com>
4842 * cfgexpand.c (expand_stack_vars): Optionally disable
4843 asan stack protection.
4844 (expand_used_vars): Likewise.
4845 (partition_stack_vars): Likewise.
4846 * asan.c (asan_emit_stack_protection): Optionally disable
4847 after return stack usage.
4848 (instrument_derefs): Optionally disable memory access instrumentation.
4849 (instrument_builtin_call): Likewise.
4850 (instrument_strlen_call): Likewise.
4851 (asan_protect_global): Optionally disable global variables protection.
4852 * doc/invoke.texi: Added doc for new options.
4853 * params.def: Added new options.
4854 * params.h: Likewise.
4856 2014-01-09 Jakub Jelinek <jakub@redhat.com>
4858 PR rtl-optimization/59724
4859 * ifcvt.c (cond_exec_process_if_block): Don't call
4860 flow_find_head_matching_sequence with 0 longest_match.
4861 * cfgcleanup.c (flow_find_head_matching_sequence): Count even
4862 non-active insns if !stop_after.
4863 (try_head_merge_bb): Revert 2014-01-07 changes.
4865 2014-01-08 Jeff Law <law@redhat.com>
4867 * ree.c (get_sub_rtx): New function, extracted from...
4868 (merge_def_and_ext): Here.
4869 (combine_reaching_defs): Use get_sub_rtx.
4871 2014-01-08 Eric Botcazou <ebotcazou@adacore.com>
4873 * cgraph.h (varpool_variable_node): Do not choke on null node.
4875 2014-01-08 Catherine Moore <clm@codesourcery.com>
4877 * config/mips/mips.md (simple_return): Attempt to use JRC
4879 * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
4881 2014-01-08 Richard Sandiford <rdsandiford@googlemail.com>
4883 PR rtl-optimization/59137
4884 * reorg.c (steal_delay_list_from_target): Call update_block for
4886 (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
4888 2014-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4890 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4891 two duplicate entries.
4893 2014-01-08 Richard Sandiford <rdsandiford@googlemail.com>
4896 2012-10-07 Richard Sandiford <rdsandiford@googlemail.com>
4898 * config/mips/mips.c (mips_truncated_op_cost): New function.
4899 (mips_rtx_costs): Adjust test for BADDU.
4900 * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
4902 2012-10-02 Richard Sandiford <rdsandiford@googlemail.com>
4904 * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
4905 (*baddu_si): ...this new pattern.
4907 2014-01-08 Jakub Jelinek <jakub@redhat.com>
4910 * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
4912 2014-01-08 Bernd Edlinger <bernd.edlinger@hotmail.de>
4915 * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
4917 (expand_expr, expand_normal): Adjust.
4918 * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
4919 inner_reference_p. Use inner_reference_p to expand inner references.
4920 (store_expr): Adjust.
4921 * cfgexpand.c (expand_call_stmt): Adjust.
4923 2014-01-08 Rong Xu <xur@google.com>
4925 * gcov-io.c (gcov_var): Move from gcov-io.h.
4926 (gcov_position): Ditto.
4927 (gcov_is_error): Ditto.
4928 (gcov_rewrite): Ditto.
4929 * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
4930 only part to libgcc/libgcov.h.
4932 2014-01-08 Marek Polacek <polacek@redhat.com>
4935 * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
4937 2014-01-08 Marek Polacek <polacek@redhat.com>
4940 * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
4942 2014-01-08 Jakub Jelinek <jakub@redhat.com>
4944 PR rtl-optimization/59649
4945 * stor-layout.c (get_mode_bounds): For BImode return
4946 0 and STORE_FLAG_VALUE.
4948 2014-01-08 Richard Biener <rguenther@suse.de>
4951 * gimple.h (is_gimple_builtin_call): Remove.
4952 (gimple_builtin_call_types_compatible_p): New.
4953 (gimple_call_builtin_p): New overload.
4954 * gimple.c (is_gimple_builtin_call): Remove.
4955 (validate_call): Rename to ...
4956 (gimple_builtin_call_types_compatible_p): ... this and export. Also
4958 (validate_type): New static function.
4959 (gimple_call_builtin_p): New overload and adjust.
4960 * gimple-fold.c (gimple_fold_builtin): Fold the return value.
4961 (gimple_fold_call): Likewise. Use gimple_call_builtin_p.
4962 (gimple_fold_stmt_to_constant_1): Likewise.
4963 * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
4965 2014-01-08 Richard Biener <rguenther@suse.de>
4968 * gimplify.c (gimplify_expr): Gimplify register-register type
4969 VIEW_CONVERT_EXPRs to separate stmts.
4971 2014-01-07 Jeff Law <law@redhat.com>
4974 * ree.c (combine_set_extension): Handle case where source
4975 and destination registers in an extension insn are different.
4976 (combine_reaching_defs): Allow source and destination registers
4977 in extension to be different under limited circumstances.
4978 (add_removable_extension): Remove restriction that the
4979 source and destination registers in the extension are the same.
4980 (find_and_remove_re): Emit a copy from the extension's
4981 destination to its source after the defining insn if
4982 the source and destination registers are different.
4985 * ifcvt.c (merge_if_block): If we are merging a block with more than
4986 one successor with a block with no successors, remove any BARRIER
4987 after the second block.
4989 2014-01-07 Dan Xio Qiang <ziyan01@163.com>
4991 * hw-doloop.c (reorg_loops): Release the bitmap obstack.
4993 2014-01-07 John David Anglin <danglin@gcc.gnu.org>
4996 * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
4997 for 14-bit register offsets when INT14_OK_STRICT is false.
4999 2014-01-07 Roland Stigge <stigge@antcom.de>
5000 Michael Meissner <meissner@linux.vnet.ibm.com>
5003 * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
5004 Only check TFmode for SPE constants. Don't check TImode or TDmode.
5006 2014-01-07 James Greenhalgh <james.greenhalgh@arm.com>
5008 * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
5011 2014-01-07 Yufeng Zhang <yufeng.zhang@arm.com>
5013 * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
5014 with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
5015 rtx is const0_rtx or not.
5017 2014-01-07 Richard Sandiford <rdsandiford@googlemail.com>
5020 * target-globals.c (save_target_globals): Remove this_fn_optab
5022 * toplev.c: Include optabs.h.
5023 (target_reinit): Temporarily restore the global options if another
5024 set of options are in force.
5026 2014-01-07 Jakub Jelinek <jakub@redhat.com>
5028 PR rtl-optimization/58668
5029 * cfgcleanup.c (flow_find_cross_jump): Don't count
5030 any jumps if dir_p is NULL. Remove p1 variable, use active_insn_p
5031 to determine what is counted.
5032 (flow_find_head_matching_sequence): Use active_insn_p to determine
5034 (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
5036 * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
5037 determine what is counted.
5039 PR tree-optimization/59643
5040 * tree-predcom.c (split_data_refs_to_components): If one dr is
5041 read and one write, determine_offset fails and the write isn't
5042 in the bad component, just put the read into the bad component.
5044 2014-01-07 Mike Stump <mikestump@comcast.net>
5045 Jakub Jelinek <jakub@redhat.com>
5048 * tree-core.h (struct tree_optimization_option): Change optabs
5049 type from unsigned char * to void *.
5050 * optabs.c (init_tree_optimization_optabs): Adjust
5051 TREE_OPTIMIZATION_OPTABS initialization.
5053 2014-01-06 Jakub Jelinek <jakub@redhat.com>
5056 * config/i386/i386.h (struct machine_function): Add
5057 no_drap_save_restore field.
5058 * config/i386/i386.c (ix86_save_reg): Use
5059 !cfun->machine->no_drap_save_restore instead of
5060 crtl->stack_realign_needed.
5061 (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
5062 this function clears frame_pointer_needed. Set
5063 cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
5064 and DRAP reg is needed.
5066 2014-01-06 Marek Polacek <polacek@redhat.com>
5069 * doc/implement-c.texi: Mention that other integer types are
5070 permitted as bit-field types in strictly conforming mode.
5072 2014-01-06 Felix Yang <fei.yang0953@gmail.com>
5074 * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
5077 2014-01-06 Richard Earnshaw <rearnsha@arm.com>
5079 * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
5081 2014-01-06 Martin Jambor <mjambor@suse.cz>
5084 * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
5086 * ipa-prop.c (ipa_print_node_params): Fix indentation.
5088 2014-01-06 Eric Botcazou <ebotcazou@adacore.com>
5092 * var-tracking.c (add_stores): Preserve the value of the source even if
5093 we don't record the store.
5095 2014-01-06 Terry Guo <terry.guo@arm.com>
5097 * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
5099 2014-01-05 Iain Sandoe <iain@codesourcery.com>
5102 * config/darwin.c (darwin_function_section): Adjust return values to
5103 correspond to optimisation changes made in r206070.
5105 2014-01-05 Uros Bizjak <ubizjak@gmail.com>
5107 * config/i386/i386.c (ix86_data_alignment): Calculate max_align
5108 from prefetch_block tune setting.
5109 (nocona_cost): Correct size of prefetch block to 64.
5111 2014-01-04 Eric Botcazou <ebotcazou@adacore.com>
5113 * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
5114 (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
5115 used to save the static chain register in the computation of the offset
5116 from which the FP registers need to be restored.
5118 2014-01-04 Jakub Jelinek <jakub@redhat.com>
5120 PR tree-optimization/59519
5121 * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
5122 ICE if get_current_def (current_new_name) is already non-NULL, as long
5123 as it is a phi result of some other phi in *new_exit_bb that has
5126 * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
5127 or vmovdqu* for misaligned_operand.
5128 (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
5129 <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
5130 * config/i386/i386.c (ix86_expand_special_args_builtin): Set
5131 aligned_mem for AVX512F masked aligned load and store builtins and for
5134 2014-01-03 Bingfeng Mei <bmei@broadcom.com>
5136 PR tree-optimization/59651
5137 * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
5138 Address range for negative step should be added by TYPE_SIZE_UNIT.
5140 2014-01-03 Andreas Schwab <schwab@linux-m68k.org>
5142 * config/m68k/m68k.c (handle_move_double): Handle pushes with
5143 overlapping registers also for registers other than the stack pointer.
5145 2014-01-03 Marek Polacek <polacek@redhat.com>
5148 * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
5151 2014-01-03 Jakub Jelinek <jakub@redhat.com>
5154 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
5157 * config/i386/i386.md (MODE_SIZE): New mode attribute.
5158 (push splitter): Use <P:MODE_SIZE> instead of
5159 GET_MODE_SIZE (<P:MODE>mode).
5160 (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
5161 (mov -1, reg peephole2): Likewise.
5162 * config/i386/sse.md (*mov<mode>_internal,
5163 <sse>_storeu<ssemodesuffix><avxsizesuffix>,
5164 <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
5165 *<code><mode>3, *andnot<mode>3<mask_name>,
5166 <mask_codefor><code><mode>3<mask_name>): Likewise.
5167 * config/i386/subst.md (mask_mode512bit_condition,
5168 sd_mask_mode512bit_condition): Likewise.
5170 2014-01-02 Xinliang David Li <davidxl@google.com>
5172 PR tree-optimization/59303
5173 * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
5174 (dump_predicates): Better output format.
5175 (pred_equal_p): New function.
5176 (is_neq_relop_p): Ditto.
5177 (is_neq_zero_form_p): Ditto.
5178 (pred_expr_equal_p): Ditto.
5179 (pred_neg_p): Ditto.
5180 (simplify_pred): Ditto.
5181 (simplify_preds_2): Ditto.
5182 (simplify_preds_3): Ditto.
5183 (simplify_preds_4): Ditto.
5184 (simplify_preds): Ditto.
5186 (push_to_worklist): Ditto.
5187 (get_pred_info_from_cmp): Ditto.
5188 (is_degenerated_phi): Ditto.
5189 (normalize_one_pred_1): Ditto.
5190 (normalize_one_pred): Ditto.
5191 (normalize_one_pred_chain): Ditto.
5192 (normalize_preds): Ditto.
5193 (normalize_cond_1): Remove function.
5194 (normalize_cond): Ditto.
5195 (is_gcond_subset_of): Ditto.
5196 (is_subset_of_any): Ditto.
5197 (is_or_set_subset_of): Ditto.
5198 (is_and_set_subset_of): Ditto.
5199 (is_norm_cond_subset_of): Ditto.
5200 (pred_chain_length_cmp): Ditto.
5201 (convert_control_dep_chain_into_preds): Type change.
5202 (find_predicates): Ditto.
5203 (find_def_preds): Ditto.
5204 (destroy_predicates_vecs): Ditto.
5205 (find_matching_predicates_in_rest_chains): Ditto.
5206 (use_pred_not_overlap_with_undef_path_pred): Ditto.
5207 (is_pred_expr_subset): Ditto.
5208 (is_pred_chain_subset_of): Ditto.
5209 (is_included_in): Ditto.
5210 (is_superset_of): Ditto.
5212 2014-01-02 Richard Sandiford <rdsandiford@googlemail.com>
5214 Update copyright years.
5216 2014-01-02 Richard Sandiford <rdsandiford@googlemail.com>
5218 * common/config/arc/arc-common.c, config/arc/arc-modes.def,
5219 config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
5220 config/arc/arc.md, config/arc/arc.opt,
5221 config/arm/arm_neon_builtins.def, config/arm/crypto.def,
5222 config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
5223 config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
5224 config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
5225 config/linux-protos.h, config/linux.c, config/winnt-c.c,
5226 diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
5227 vtable-verify.c, vtable-verify.h: Use the standard form for the
5230 2014-01-02 Tobias Burnus <burnus@net-b.de>
5232 * gcc.c (process_command): Update copyright notice dates.
5233 * gcov-dump.c: Ditto.
5235 * doc/cpp.texi: Bump @copying's copyright year.
5236 * doc/cppinternals.texi: Ditto.
5237 * doc/gcc.texi: Ditto.
5238 * doc/gccint.texi: Ditto.
5239 * doc/gcov.texi: Ditto.
5240 * doc/install.texi: Ditto.
5241 * doc/invoke.texi: Ditto.
5243 2014-01-01 Jan-Benedict Glaw <jbglaw@lug-owl.de>
5245 * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
5247 2014-01-01 Jakub Jelinek <jakub@redhat.com>
5249 * config/i386/sse.md (*mov<mode>_internal): Guard
5250 EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
5252 PR rtl-optimization/59647
5253 * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
5254 new_rtx into UNSIGNED_FLOAT rtxes.
5256 Copyright (C) 2014 Free Software Foundation, Inc.
5258 Copying and distribution of this file, with or without modification,
5259 are permitted in any medium without royalty provided the copyright
5260 notice and this notice are preserved.