1 2014-02-18 Jan Hubicka <hubicka@ucw.cz>
3 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Also remove
4 inline clones when edge disappears.
6 2014-02-18 Michael Meissner <meissner@linux.vnet.ibm.com>
9 * config/rs6000/rs6000.md (mov<mode>_64bit, TF/TDmode moves):
10 Split 64-bit moves into 2 patterns. Do not allow the use of
11 direct move for TDmode in little endian, since the decimal value
12 has little endian bytes within a word, but the 64-bit pieces are
13 ordered in a big endian fashion, and normal subreg's of TDmode are
15 (mov<mode>_64bit_dm): Likewise.
16 (movtd_64bit_nodm): Likewise.
18 2014-02-18 Eric Botcazou <ebotcazou@adacore.com>
20 PR tree-optimization/60174
21 * tree-ssa-reassoc.c (init_range_entry): Do not look into the defining
22 statement of an SSA_NAME that occurs in an abnormal PHI node.
24 2014-02-18 Jakub Jelinek <jakub@redhat.com>
27 * final.c (SEEN_BB): Remove.
28 (SEEN_NOTE, SEEN_EMITTED): Renumber.
29 (final_scan_insn): Don't force_source_line on second
30 NOTE_INSN_BASIC_BLOCK.
32 2014-02-18 Uros Bizjak <ubizjak@gmail.com>
35 * config/i386/i386.h (struct ix86_args): Add warn_avx512f.
36 * config/i386/i386.c (init_cumulative_args): Initialize warn_avx512f.
37 (type_natural_mode): Warn ABI change when %zmm register is not
38 available for AVX512F vector value passing.
40 2014-02-18 Kai Tietz <ktietz@redhat.com>
43 * config/i386/i386.c (ix86_expand_prologue): Use value in
44 rax register as displacement when restoring %r10 or %rax.
45 Fix wrong offset when restoring both registers.
47 2014-02-18 Eric Botcazou <ebotcazou@adacore.com>
49 * ipa-prop.c (compute_complex_ancestor_jump_func): Replace overzealous
50 assertion with conditional return.
52 2014-02-18 Jakub Jelinek <jakub@redhat.com>
53 Uros Bizjak <ubizjak@gmail.com>
56 * config/i386/driver-i386.c (host_detect_local_cpu): If
57 YMM state is not saved by the OS, also clear has_f16c. Move
58 CPUID 0x80000001 handling before YMM state saving checking.
60 2014-02-18 Andrey Belevantsev <abel@ispras.ru>
62 PR rtl-optimization/58960
63 * haifa-sched.c (alloc_global_sched_pressure_data): New,
65 (sched_init): ... here.
66 (free_global_sched_pressure_data): New, factored out from ...
67 (sched_finish): ... here.
68 * sched-int.h (free_global_sched_pressure_data): Declare.
69 * sched-rgn.c (nr_regions_initial): New static global.
70 (haifa_find_rgns): Initialize it.
71 (schedule_region): Disable sched-pressure for the newly
74 2014-02-17 Richard Biener <rguenther@suse.de>
76 * tree-vect-stmts.c (free_stmt_vec_info): Clear BB and
77 release SSA defs of pattern stmts.
79 2014-02-17 Richard Biener <rguenther@suse.de>
81 * tree-inline.c (expand_call_inline): Release the virtual
82 operand defined by the call we are about to inline.
84 2014-02-17 Richard Biener <rguenther@suse.de>
86 * tree-ssa.c (verify_ssa): If verify_def found an error, ICE.
88 2014-02-17 Kirill Yukhin <kirill.yukhin@intel.com>
89 Ilya Tocar <ilya.tocar@intel.com>
91 * config/i386/avx512fintrin.h (_mm512_maskz_permutexvar_epi64): Swap
92 arguments order in builtin.
93 (_mm512_permutexvar_epi64): Ditto.
94 (_mm512_mask_permutexvar_epi64): Ditto
95 (_mm512_maskz_permutexvar_epi32): Ditto
96 (_mm512_permutexvar_epi32): Ditto
97 (_mm512_mask_permutexvar_epi32): Ditto
99 2014-02-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
101 * config/rs6000/altivec.md (p8_vmrgew): Handle little endian targets.
102 (p8_vmrgow): Likewise.
104 2014-02-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
106 * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Handle little
109 2014-02-15 Michael Meissner <meissner@linux.vnet.ibm.com>
112 * config/rs6000/rs6000.md (rreg): Add TFmode, TDmode constraints.
113 (mov<mode>_internal, TFmode/TDmode): Split TFmode/TDmode moves
114 into 64-bit and 32-bit moves. On 64-bit moves, add support for
115 using direct move instructions on ISA 2.07. Also adjust
116 instruction length for 64-bit.
117 (mov<mode>_64bit, TFmode/TDmode): Likewise.
118 (mov<mode>_32bit, TFmode/TDmode): Likewise.
120 2014-02-15 Alan Modra <amodra@gmail.com>
124 * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Use
125 find_replacement on parts of insn rtl that might be reloaded.
127 2014-02-15 Richard Biener <rguenther@suse.de>
129 PR tree-optimization/60183
130 * tree-ssa-phiprop.c (propagate_with_phi): Avoid speculating loads.
131 (tree_ssa_phiprop): Calculate and free post-dominators.
133 2014-02-14 Jeff Law <law@redhat.com>
135 PR rtl-optimization/60131
136 * ree.c (get_extended_src_reg): New function.
137 (combine_reaching_defs): Use it rather than assuming location of REG.
138 (find_and_remove_re): Verify first operand of extension is
139 a REG before adding the insns to the copy list.
141 2014-02-14 Roland McGrath <mcgrathr@google.com>
143 * configure.ac (HAVE_AS_IX86_UD2): New test for 'ud2' mnemonic.
144 * configure: Regenerated.
145 * config.in: Regenerated.
146 * config/i386/i386.md (trap) [HAVE_AS_IX86_UD2]: Use the mnemonic
147 instead of ASM_SHORT.
149 2014-02-14 Vladimir Makarov <vmakarov@redhat.com>
150 Richard Earnshaw <rearnsha@arm.com>
152 PR rtl-optimization/59535
153 * lra-constraints.c (process_alt_operands): Encourage alternative
154 when unassigned pseudo class is superset of the alternative class.
155 (inherit_reload_reg): Don't inherit when optimizing for code size.
156 * config/arm/arm.h (MODE_BASE_REG_CLASS): Add version for LRA
157 returning CORE_REGS for anything but Thumb1 and BASE_REGS for
158 modes not less than 4 for Thumb1.
160 2014-02-14 Kyle McMartin <kyle@redhat.com>
163 * config/host-linux.c (TRY_EMPTY_VM_SPACE): Define for AArch64.
165 2014-02-14 Richard Biener <rguenther@suse.de>
167 * cilk-common.c (cilk_arrow): Build a MEM_REF, not an INDIRECT_REF.
168 (get_frame_arg): Drop the assert with langhook types_compatible_p.
169 Do not strip INDIRECT_REFs.
171 2014-02-14 Richard Biener <rguenther@suse.de>
174 * lto-streamer-out.c (DFS_write_tree_body): Do not follow
175 DECL_FUNCTION_SPECIFIC_TARGET.
176 (hash_tree): Do not hash DECL_FUNCTION_SPECIFIC_TARGET.
177 * tree-streamer-out.c (pack_ts_target_option): Remove.
178 (streamer_pack_tree_bitfields): Do not stream TS_TARGET_OPTION.
179 (write_ts_function_decl_tree_pointers): Do not stream
180 DECL_FUNCTION_SPECIFIC_TARGET.
181 * tree-streamer-in.c (unpack_ts_target_option): Remove.
182 (unpack_value_fields): Do not stream TS_TARGET_OPTION.
183 (lto_input_ts_function_decl_tree_pointers): Do not stream
184 DECL_FUNCTION_SPECIFIC_TARGET.
186 2014-02-14 Jakub Jelinek <jakub@redhat.com>
188 * tree-vect-loop.c (vect_is_slp_reduction): Don't set use_stmt twice.
189 (get_initial_def_for_induction, vectorizable_induction): Ignore
190 debug stmts when looking for exit_phi.
191 (vectorizable_live_operation): Fix up condition.
193 2014-02-14 Chung-Ju Wu <jasonwucj@gmail.com>
195 * config/nds32/nds32.c (nds32_asm_function_prologue): Do not use
196 nreverse() because it changes the content of original tree list.
198 2014-02-14 Chung-Ju Wu <jasonwucj@gmail.com>
200 * config/nds32/t-mlibs (MULTILIB_OPTIONS): Fix typo in comment.
201 * config/nds32/nds32.c (nds32_merge_decl_attributes): Likewise.
203 2014-02-14 Chung-Ju Wu <jasonwucj@gmail.com>
205 * config/nds32/nds32.c (nds32_naked_function_p): Follow the
206 GNU coding standards.
208 2014-02-13 Jakub Jelinek <jakub@redhat.com>
211 * dwarf2out.c (gen_subprogram_die): Don't call
212 add_calling_convention_attribute if subr_die is old_die.
214 2014-02-13 Sharad Singhai <singhai@google.com>
216 * doc/optinfo.texi: Fix order of nodes.
218 2014-02-13 Uros Bizjak <ubizjak@gmail.com>
220 * config/i386/sse.md (xop_vmfrcz<mode>2): Generate const0 in
221 operands[2], not operands[3].
223 2014-02-13 Richard Biener <rguenther@suse.de>
226 * doc/install.texi (ISL): Update recommended version to 0.12.2,
227 mention the possibility of an in-tree build.
228 (CLooG): Update recommended version to 0.18.1, mention the
229 possibility of an in-tree build and clarify that the ISL
230 bundled with CLooG does not work.
232 2014-02-13 Jakub Jelinek <jakub@redhat.com>
235 * expr.c (compress_float_constant): If x is a hard register,
236 extend into a pseudo and then move to x.
238 2014-02-13 Dominik Vogt <vogt@linux.vnet.ibm.com>
240 * config/s390/s390.c (s390_asm_output_function_label): Fix crash
241 caused by bad second argument to warning_at() with -mhotpatch and
242 nested functions (e.g. with gfortran).
244 2014-02-13 Richard Sandiford <rdsandiford@googlemail.com>
246 * opts.c (option_name): Remove "enabled by default" rider.
248 2014-02-12 John David Anglin <danglin@gcc.gnu.org>
250 * config/pa/pa.c (pa_option_override): Remove auto increment FIXME.
252 2014-02-12 H.J. Lu <hongjiu.lu@intel.com>
253 Uros Bizjak <ubizjak@gmail.com>
256 * configure.ac (HAVE_AS_GOTOFF_IN_DATA): Pass --32 to GNU assembler.
257 * configure: Regenerated.
259 2014-02-12 Richard Biener <rguenther@suse.de>
261 * vec.c (vec_prefix::calculate_allocation): Move as
262 inline variant to vec.h.
263 (vec_prefix::calculate_allocation_1): New out-of-line version.
264 * vec.h (vec_prefix::calculate_allocation_1): Declare.
265 (vec_prefix::m_has_auto_buf): Rename to ...
266 (vec_prefix::m_using_auto_storage): ... this.
267 (vec_prefix::calculate_allocation): Inline the easy cases
268 and dispatch to calculate_allocation_1 which doesn't need the
270 (va_heap::reserve): Use gcc_checking_assert.
271 (vec<T, A, vl_embed>::embedded_init): Add argument to initialize
272 m_using_auto_storage.
273 (auto_vec): Change m_vecpfx member to a vec<T, va_heap, vl_embed>
275 (vec<T, va_heap, vl_ptr>::reserve): Remove redundant check.
276 (vec<T, va_heap, vl_ptr>::release): Avoid casting.
277 (vec<T, va_heap, vl_ptr>::using_auto_storage): Simplify.
279 2014-02-12 Richard Biener <rguenther@suse.de>
281 * gcse.c (compute_transp): break from loop over canon_modify_mem_list
282 when we found a dependence.
284 2014-02-12 Thomas Schwinge <thomas@codesourcery.com>
286 * gimplify.c (gimplify_call_expr, gimplify_modify_expr): Move
288 (maybe_fold_stmt): ... into this new function.
289 * omp-low.c (lower_omp): Update comment.
291 * omp-low.c (lower_omp_target): Add clobber for sizes array, after
294 * omp-low.c (diagnose_sb_0): Make sure label_ctx is valid to
297 2014-02-12 James Greenhalgh <james.greenhalgh@arm.com>
299 * config/arm/aarch-cost-tables.h (generic_extra_costs): Fix
300 identifiers in comments.
301 (cortexa53_extra_costs): Likewise.
302 * config/arm/arm.c (cortexa9_extra_costs): Fix identifiers in comments.
303 (cortexa7_extra_costs): Likewise.
304 (cortexa12_extra_costs): Likewise.
305 (cortexa15_extra_costs): Likewise.
306 (v7m_extra_costs): Likewise.
308 2014-02-12 Richard Biener <rguenther@suse.de>
311 * gimple-low.c (lower_builtin_posix_memalign): Lower conditional
312 of posix_memalign being successful.
313 (lower_stmt): Restrict lowering of posix_memalign to when
314 -ftree-bit-ccp is enabled.
316 2014-02-12 Senthil Kumar Selvaraj <senthil_kumar.selvaraj@atmel.com>
318 * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Pass vNULL for
320 * config/spu/spu-c.c (spu_resolve_overloaded_builtin): Likewise.
322 2014-02-12 Eric Botcazou <ebotcazou@adacore.com>
324 PR rtl-optimization/60116
325 * combine.c (try_combine): Also remove dangling REG_DEAD notes on the
326 other_insn once the combination has been validated.
328 2014-02-11 Jan Hubicka <hubicka@ucw.cz>
331 * ipa-utils.h (possible_polymorphic_call_targets): Update prototype
333 * ipa-devirt.c: Include demangle.h
334 (odr_violation_reported): New static variable.
335 (add_type_duplicate): Update odr_violations.
336 (maybe_record_node): Add completep parameter; update it.
337 (record_target_from_binfo): Add COMPLETEP parameter;
339 (possible_polymorphic_call_targets_1): Likewise.
340 (struct polymorphic_call_target_d): Add nonconstruction_targets;
341 rename FINAL to COMPLETE.
342 (record_targets_from_bases): Sanity check we found the binfo;
343 fix COMPLETEP updating.
344 (possible_polymorphic_call_targets): Add NONCONSTRUTION_TARGETSP
345 parameter, fix computing of COMPLETEP.
346 (dump_possible_polymorphic_call_targets): Imrove readability of dump;
347 at LTO time do demangling.
348 (ipa_devirt): Use nonconstruction_targets; Improve dumps.
349 * gimple-fold.c (gimple_get_virt_method_for_vtable): Add can_refer
351 (gimple_get_virt_method_for_binfo): Likewise.
352 * gimple-fold.h (gimple_get_virt_method_for_binfo,
353 gimple_get_virt_method_for_vtable): Update prototypes.
355 2014-02-11 Vladimir Makarov <vmakarov@redhat.com>
358 * genautomata.c (add_presence_absence): Fix typo with
359 {final_}presence_list.
361 2014-02-11 Michael Meissner <meissner@linux.vnet.ibm.com>
364 * config/rs6000/rs6000.md (128-bit GPR splitter): Add a splitter
365 for VSX/Altivec vectors that land in GPR registers.
367 2014-02-11 Richard Henderson <rth@redhat.com>
368 Jakub Jelinek <jakub@redhat.com>
371 * tree-sra.c (load_assign_lhs_subreplacements): Add VIEW_CONVERT_EXPR
372 around drhs if type conversion to lacc->type is not useless.
374 2014-02-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
376 * config/aarch64/aarch64-cores.def (cortex-a57): Use cortexa57
378 (cortex-a57.cortex-a53): Likewise.
379 * config/aarch64/aarch64.c (cortexa57_tunings): New tuning struct.
381 2014-02-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
383 * config/arm/thumb2.md (*thumb2_movhi_insn): Add alternatives for
386 2014-02-11 Renlin Li <Renlin.Li@arm.com>
388 * doc/sourcebuild.texi: Document check_effective_target_arm_vfp3_ok and
389 add_options_for_arm_vfp3.
391 2014-02-11 Jeff Law <law@redhat.com>
394 * expr.c (expand_expr_addr_expr_1): Handle expand_expr returning an
395 object with an undesirable mode.
397 2014-02-11 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
400 * config/i386/sol2-9.h: New file.
401 * config.gcc (i[34567]86-*-solaris2* | x86_64-*-solaris2.1[0-9]*,
402 *-*-solaris2.9*): Use it.
404 2014-02-10 Nagaraju Mekala <nagaraju.mekala@xilinx.com>
406 * config/microblaze/microblaze.md: Add movsi4_rev insn pattern.
407 * config/microblaze/predicates.md: Add reg_or_mem_operand predicate.
409 2014-02-10 Nagaraju Mekala <nagaraju.mekala@xilinx.com>
411 * config/microblaze/microblaze.c: Extend mcpu version format
413 2014-02-10 David Holsgrove <david.holsgrove@xilinx.com>
415 * config/microblaze/microblaze.h: Define SIZE_TYPE and PTRDIFF_TYPE.
417 2014-02-10 Richard Henderson <rth@redhat.com>
420 * calls.c (expand_call): Don't double-push for reg_parm_stack_space.
421 * config/i386/i386.c (init_cumulative_args): Remove sorry for 64-bit
422 ms-abi vs -mno-accumulate-outgoing-args.
423 (ix86_expand_prologue): Unconditionally call ix86_eax_live_at_start_p.
424 * config/i386/i386.h (ACCUMULATE_OUTGOING_ARGS): Fix comment with
427 2014-02-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
430 * cfgexpand.c (expand_asm_operands): Attach source location to
431 ASM_INPUT rtx objects.
432 * print-rtl.c (print_rtx): Check for UNKNOWN_LOCATION.
434 2014-02-10 Nick Clifton <nickc@redhat.com>
436 * config/mn10300/mn10300.c (popcount): New function.
437 (mn10300_expand_prologue): Include saved registers in stack usage
440 2014-02-10 Jeff Law <law@redhat.com>
443 * reload1.c (emit_input_reload_insns): Do not create invalid RTL
444 when changing the SET_DEST of a prior insn to avoid an input reload.
446 2014-02-10 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
448 * config/rs6000/sysv4.h (ENDIAN_SELECT): Do not attempt to enforce
449 big-endian mode for -mcall-aixdesc, -mcall-freebsd, -mcall-netbsd,
450 -mcall-openbsd, or -mcall-linux.
451 (CC1_ENDIAN_BIG_SPEC): Remove.
452 (CC1_ENDIAN_LITTLE_SPEC): Remove.
453 (CC1_ENDIAN_DEFAULT_SPEC): Remove.
454 (CC1_SPEC): Remove (always empty) %cc1_endian_... spec.
455 (SUBTARGET_EXTRA_SPECS): Remove %cc1_endian_big, %cc1_endian_little,
456 and %cc1_endian_default.
457 * config/rs6000/sysv4le.h (CC1_ENDIAN_DEFAULT_SPEC): Remove.
459 2014-02-10 Richard Biener <rguenther@suse.de>
461 PR tree-optimization/60115
462 * tree-eh.c (tree_could_trap_p): Unify TARGET_MEM_REF and
463 MEM_REF handling. Properly verify that the accesses are not
464 out of the objects bound.
466 2014-02-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
468 * config/aarch64/aarch64.c (aarch64_override_options): Fix typo from
471 2014-02-10 Eric Botcazou <ebotcazou@adacore.com>
473 * ipa-devirt.c (get_polymorphic_call_info_from_invariant): Return
474 proper constants and fix formatting.
475 (possible_polymorphic_call_targets): Fix formatting.
477 2014-02-10 Kirill Yukhin <kirill.yukhin@intel.com>
478 Ilya Tocar <ilya.tocar@intel.com>
480 * config/i386/avx512fintrin.h (_mm512_storeu_epi64): Removed.
481 (_mm512_loadu_epi32): Renamed into...
482 (_mm512_loadu_si512): This.
483 (_mm512_storeu_epi32): Renamed into...
484 (_mm512_storeu_si512): This.
485 (_mm512_maskz_ceil_ps): Removed.
486 (_mm512_maskz_ceil_pd): Ditto.
487 (_mm512_maskz_floor_ps): Ditto.
488 (_mm512_maskz_floor_pd): Ditto.
489 (_mm512_floor_round_ps): Ditto.
490 (_mm512_floor_round_pd): Ditto.
491 (_mm512_ceil_round_ps): Ditto.
492 (_mm512_ceil_round_pd): Ditto.
493 (_mm512_mask_floor_round_ps): Ditto.
494 (_mm512_mask_floor_round_pd): Ditto.
495 (_mm512_mask_ceil_round_ps): Ditto.
496 (_mm512_mask_ceil_round_pd): Ditto.
497 (_mm512_maskz_floor_round_ps): Ditto.
498 (_mm512_maskz_floor_round_pd): Ditto.
499 (_mm512_maskz_ceil_round_ps): Ditto.
500 (_mm512_maskz_ceil_round_pd): Ditto.
501 (_mm512_expand_pd): Ditto.
502 (_mm512_expand_ps): Ditto.
503 * config/i386/i386.c (ix86_builtins): Remove
504 IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK.
506 * config/i386/predicates.md (const1256_operand): New.
507 (const_1_to_2_operand): Ditto.
508 * config/i386/sse.md (avx512pf_gatherpf<mode>sf): Change hint value.
509 (*avx512pf_gatherpf<mode>sf_mask): Ditto.
510 (*avx512pf_gatherpf<mode>sf): Ditto.
511 (avx512pf_gatherpf<mode>df): Ditto.
512 (*avx512pf_gatherpf<mode>df_mask): Ditto.
513 (*avx512pf_gatherpf<mode>df): Ditto.
514 (avx512pf_scatterpf<mode>sf): Ditto.
515 (*avx512pf_scatterpf<mode>sf_mask): Ditto.
516 (*avx512pf_scatterpf<mode>sf): Ditto.
517 (avx512pf_scatterpf<mode>df): Ditto.
518 (*avx512pf_scatterpf<mode>df_mask): Ditto.
519 (*avx512pf_scatterpf<mode>df): Ditto.
520 (avx512f_expand<mode>): Removed.
521 (<shift_insn><mode>3<mask_name>): Change predicate type.
523 2014-02-08 Jakub Jelinek <jakub@redhat.com>
525 * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers
526 not at the end of datarefs vector use ordered_remove to avoid
527 reordering datarefs vector.
530 * gimplify.c (gimplify_bind_expr): In ORT_SIMD region
531 mark local addressable non-static vars as GOVD_PRIVATE
532 instead of GOVD_LOCAL.
533 * omp-low.c (lower_omp_for): Move gimple_bind_vars
534 and BLOCK_VARS of gimple_bind_block to new_stmt rather
538 * tree-ssa-ccp.c (surely_varying_stmt_p): Don't return true
539 if TYPE_ATTRIBUTES (gimple_call_fntype ()) contain
540 assume_aligned or alloc_align attributes.
541 (bit_value_assume_aligned): Add ATTR, PTRVAL and ALLOC_ALIGN
542 arguments. Handle also assume_aligned and alloc_align attributes.
543 (evaluate_stmt): Adjust bit_value_assume_aligned caller. Handle
544 calls to functions with assume_aligned or alloc_align attributes.
545 * doc/extend.texi: Document assume_aligned and alloc_align attributes.
547 2014-02-08 Terry Guo <terry.guo@arm.com>
549 * doc/invoke.texi: Document ARM -march=armv7e-m.
551 2014-02-08 Jakub Jelinek <jakub@redhat.com>
553 * cilk-common.c (cilk_init_builtins): Clear TREE_NOTHROW
554 flag on __cilkrts_rethrow builtin.
557 * ipa-cp.c (determine_versionability): Fail at -O0
558 or __attribute__((optimize (0))) or -fno-ipa-cp functions.
559 * tree-sra.c (ipa_sra_preliminary_function_checks): Similarly.
562 2014-02-04 Jakub Jelinek <jakub@redhat.com>
565 * tree-inline.c (copy_forbidden): Fail for
566 __attribute__((optimize (0))) functions.
568 2014-02-07 Jan Hubicka <hubicka@ucw.cz>
570 * varpool.c: Include pointer-set.h.
571 (varpool_remove_unreferenced_decls): Variables in other partitions
572 will not be output; be however careful to not lose information
575 2014-02-07 Jan Hubicka <hubicka@ucw.cz>
577 * gimple-fold.c (gimple_get_virt_method_for_vtable): Do O(1)
578 lookup in the vtable constructor.
580 2014-02-07 Jeff Law <law@redhat.com>
583 * config/m68k/m68k.md (ashldi_extsi): Turn into a
584 define_insn_and_split.
586 * ipa-inline.c (inline_small_functions): Fix typos.
588 2014-02-07 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
590 * config/s390/s390-protos.h (s390_can_use_simple_return_insn)
591 (s390_can_use_return_insn): Declare.
592 * config/s390/s390.h (EPILOGUE_USES): Define.
593 * config/s390/s390.c (s390_mainpool_start): Allow two main_pool
595 (s390_chunkify_start): Handle return JUMP_LABELs.
596 (s390_early_mach): Emit a main_pool instruction on the entry edge.
597 (s300_set_up_by_prologue, s390_can_use_simple_return_insn)
598 (s390_can_use_return_insn): New functions.
599 (s390_fix_long_loop_prediction): Handle conditional returns.
600 (TARGET_SET_UP_BY_PROLOGUE): Define.
601 * config/s390/s390.md (ANY_RETURN): New code iterator.
602 (*creturn, *csimple_return, return, simple_return): New patterns.
604 2014-02-07 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
606 * config/s390/s390.c (s390_restore_gprs_from_fprs): Add REG_CFA_RESTORE
607 notes to each restore. Also add REG_CFA_DEF_CFA when restoring %r15.
608 (s390_optimize_prologue): Don't clear RTX_FRAME_RELATED_P. Update the
609 REG_CFA_RESTORE list when deciding not to restore a register.
611 2014-02-07 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
613 * config/s390/s390.c: Include tree-pass.h and context.h.
614 (s390_early_mach): New function, split out from...
615 (s390_emit_prologue): ...here.
616 (pass_data_s390_early_mach): New pass structure.
617 (pass_s390_early_mach): New class.
618 (s390_option_override): Create and register early_mach pass.
621 2014-02-07 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
623 * var-tracking.c (vt_stack_adjustments): Don't require stack_adjusts
624 to match for the exit block.
626 2014-02-07 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
628 * config/s390/s390.md ("atomic_load<mode>", "atomic_store<mode>")
629 ("atomic_compare_and_swap<mode>", "atomic_fetch_<atomic><mode>"):
630 Reject misaligned operands.
632 2014-02-07 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
634 * optabs.c (expand_atomic_compare_and_swap): Allow expander to fail.
636 2014-02-07 Richard Biener <rguenther@suse.de>
639 * gimple-low.c (lower_builtin_posix_memalign): New function.
640 (lower_stmt): Call it to lower posix_memalign in a way
641 to make alignment info accessible.
643 2014-02-07 Jakub Jelinek <jakub@redhat.com>
646 * tree.c (build_common_builtin_nodes): Set ECF_LEAF for
647 __builtin_setjmp_receiver.
649 2014-02-07 Richard Biener <rguenther@suse.de>
652 * builtin-types.def (BT_FN_INT_PTRPTR_SIZE_SIZE): Add.
653 * builtins.def (BUILT_IN_POSIX_MEMALIGN): Likewise.
654 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
655 Handle BUILT_IN_POSIX_MEMALIGN.
656 (find_func_clobbers): Likewise.
657 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
658 (call_may_clobber_ref_p_1): Likewise.
660 2014-02-06 Jan Hubicka <hubicka@ucw.cz>
663 * ipa-devirt.c (record_target_from_binfo): Remove overactive
666 2014-02-06 Jan Hubicka <hubicka@ucw.cz>
669 * lto-cgraph.c (lto_output_node): Use
670 symtab_get_symbol_partitioning_class.
671 (lto_output_varpool_node): likewise.
672 (symtab_get_symbol_partitioning_class): Move here from
674 * cgraph.h (symbol_partitioning_class): Likewise.
675 (symtab_get_symbol_partitioning_class): Declare.
677 2014-02-06 Jan Hubicka <hubicka@ucw.cz>
679 * ggc.h (ggc_internal_cleared_alloc): New macro.
680 * vec.h (vec_safe_copy): Handle memory stats.
681 * omp-low.c (simd_clone_struct_alloc): Use ggc_internal_cleared_alloc.
682 * target-globals.c (save_target_globals): Likewise.
684 2014-02-06 Jan Hubicka <hubicka@ucw.cz>
687 * expr.c (emit_move_resolve_push): Export; be bit more selective
688 on when to clear alias set.
689 * expr.h (emit_move_resolve_push): Declare.
690 * function.h (struct function): Add tail_call_marked.
691 * tree-tailcall.c (optimize_tail_call): Set tail_call_marked.
692 * config/i386/i386-protos.h (ix86_expand_push): Remove.
693 * config/i386/i386.md (TImode move expander): De not call
695 (FP push expanders): Preserve memory attributes.
696 * config/i386/sse.md (push<mode>1): Remove.
697 * config/i386/i386.c (ix86_expand_vector_move): Handle push operation.
698 (ix86_expand_push): Remove.
699 * config/i386/mmx.md (push<mode>1): Remove.
701 2014-02-06 Jakub Jelinek <jakub@redhat.com>
703 PR rtl-optimization/60030
704 * internal-fn.c (ubsan_expand_si_overflow_mul_check): Surround
705 lopart with paradoxical subreg before shifting it up by hprec.
707 2014-02-06 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
709 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): New table.
710 Remove extra newline at end of file.
711 * config/arm/arm.c (arm_cortex_a57_tune): New tuning struct.
712 (arm_issue_rate): Handle cortexa57.
713 * config/arm/arm-cores.def (cortex-a57): Use cortex_a57 tuning.
714 (cortex-a57.cortex-a53): Likewise.
716 2014-02-06 Jakub Jelinek <jakub@redhat.com>
719 * config/arm/arm.c (emit_multi_reg_push): Add dwarf_regs_mask argument,
720 don't record in REG_FRAME_RELATED_EXPR registers not set in that
722 (arm_expand_prologue): Adjust all callers.
723 (arm_unwind_emit_sequence): Allow saved, but not important for unwind
724 info, registers also at the lowest numbered registers side. Use
725 gcc_assert instead of abort, and SET_SRC/SET_DEST macros instead of
729 * var-tracking.c (adjust_mems): Before adding a SET to
730 amd->side_effects, adjust it's SET_SRC using simplify_replace_fn_rtx.
732 2014-02-06 Alan Modra <amodra@gmail.com>
735 * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_mode): Only
736 change SDmode to DDmode when lra_in_progress.
738 2014-02-06 Jakub Jelinek <jakub@redhat.com>
741 * tree-vect-data-refs.c (vect_analyze_data_refs): For clobbers, call
742 free_data_ref on the dr first, and before goto again also set dr
743 to the next dr. For simd_lane_access, free old datarefs[i] before
744 overwriting it. For get_vectype_for_scalar_type failure, don't
745 free_data_ref if simd_lane_access.
747 * Makefile.in (prefix.o, cppbuiltin.o): Depend on $(BASEVER).
750 * tree.h (opts_for_fn): New inline function.
751 (opt_for_fn): Define.
752 * config/i386/i386.c (ix86_function_regparm): Use
753 opt_for_fn (decl, optimize) instead of optimize.
755 2014-02-06 Marcus Shawcroft <marcus.shawcroft@arm.com>
757 * config/aarch64/aarch64.c (aarch64_classify_symbol): Fix logic
758 for SYMBOL_REF in large memory model.
760 2014-02-06 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
762 * config/aarch64/aarch64-cores.def (cortex-a53): Specify CRC32
764 (cortex-a57): Likewise.
765 (cortex-a57.cortex-a53): Likewise.
767 2014-02-06 Yury Gribov <y.gribov@samsung.com>
768 Kugan Vivekanandarajah <kuganv@linaro.org>
770 * config/arm/arm.c (arm_vector_alignment_reachable): Check
772 * config/arm/arm.c (arm_builtin_support_vector_misalignment): Likewise.
774 2014-02-06 Richard Biener <rguenther@suse.de>
776 * tree-cfg.c (gimple_duplicate_sese_region): Fix ordering of
777 set_loop_copy and initialize_original_copy_tables.
779 2014-02-06 Alex Velenko <Alex.Velenko@arm.com>
781 * config/aarch64/aarch64-simd.md
782 (aarch64_ashr_simddi): Change QI to SI.
784 2014-02-05 Jan Hubicka <hubicka@ucw.cz>
785 Jakub Jelinek <jakub@redhat.com>
788 * ipa-inline-analysis.c (compute_bb_predicates): Ensure monotonicity
791 2014-02-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
793 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Change
794 CODE_FOR_altivec_vpku[hw]um to
795 CODE_FOR_altivec_vpku[hw]um_direct.
796 * config/rs6000/altivec.md (vec_unpacks_hi_<VP_small_lc>): Change
797 UNSPEC_VUNPACK_HI_SIGN to UNSPEC_VUNPACK_HI_SIGN_DIRECT.
798 (vec_unpacks_lo_<VP_small_lc>): Change UNSPEC_VUNPACK_LO_SIGN to
799 UNSPEC_VUNPACK_LO_SIGN_DIRECT.
801 2014-02-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
803 * config/rs6000/altivec.md (altivec_vsum2sws): Adjust code
804 generation for -maltivec=be.
805 (altivec_vsumsws): Simplify redundant test.
807 2014-02-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
809 * altivec.md (UNSPEC_VPACK_UNS_UNS_MOD_DIRECT): New unspec.
810 (UNSPEC_VUNPACK_HI_SIGN_DIRECT): Likewise.
811 (UNSPEC_VUNPACK_LO_SIGN_DIRECT): Likewise.
812 (mulv8hi3): Use gen_altivec_vpkuwum_direct instead of
814 (altivec_vpkpx): Test for VECTOR_ELT_ORDER_BIG instead of for
816 (altivec_vpks<VI_char>ss): Likewise.
817 (altivec_vpks<VI_char>us): Likewise.
818 (altivec_vpku<VI_char>us): Likewise.
819 (altivec_vpku<VI_char>um): Likewise.
820 (altivec_vpku<VI_char>um_direct): New (copy of
821 altivec_vpku<VI_char>um that still relies on BYTES_BIG_ENDIAN, for
823 (altivec_vupkhs<VU_char>): Emit vupkls* instead of vupkhs* when
824 target is little endian and -maltivec=be is not specified.
825 (*altivec_vupkhs<VU_char>_direct): New (copy of
826 altivec_vupkhs<VU_char> that always emits vupkhs*, for internal use).
827 (altivec_vupkls<VU_char>): Emit vupkhs* instead of vupkls* when
828 target is little endian and -maltivec=be is not specified.
829 (*altivec_vupkls<VU_char>_direct): New (copy of
830 altivec_vupkls<VU_char> that always emits vupkls*, for internal use).
831 (altivec_vupkhpx): Emit vupklpx instead of vupkhpx when target is
832 little endian and -maltivec=be is not specified.
833 (altivec_vupklpx): Emit vupkhpx instead of vupklpx when target is
834 little endian and -maltivec=be is not specified.
836 2014-02-05 Richard Henderson <rth@redhat.com>
839 * combine-stack-adj.c: Revert r206943.
840 * sched-int.h (struct deps_desc): Add last_args_size.
841 * sched-deps.c (init_deps): Initialize it.
842 (sched_analyze_insn): Add OUTPUT dependencies between insns that
843 contain REG_ARGS_SIZE notes.
845 2014-02-05 Jan Hubicka <hubicka@ucw.cz>
847 * lto-cgraph.c (asm_nodes_output): Make global.
848 * lto-wrapper.c (run_gcc): Pass down paralelizm to WPA.
849 * gcc.c (AS_NEEDS_DASH_FOR_PIPED_INPUT): Allow WPA parameter
850 (driver_handle_option): Handle OPT_fwpa.
852 2014-02-05 Jakub Jelinek <jakub@redhat.com>
855 * ipa-devirt.c (possible_polymorphic_call_targets): Fix
856 a comment typo and formatting issue. If odr_hash hasn't been
857 created, return vNULL and set *completep to false.
860 * tree-eh.c (cleanup_empty_eh): Bail out on totally empty
861 bb with no successors.
863 2014-02-05 James Greenhalgh <james.greenhalgh@arm.com>
866 * doc/invoke.texi (-march): Clarify documentation for ARM.
870 2014-02-05 Richard Biener <rguenther@suse.de>
872 * tree-vect-loop.c (vect_analyze_loop_2): Be more informative
873 when not vectorizing because of too many alias checks.
874 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
875 Add more verboseness, avoid duplicate MSG_MISSED_OPTIMIZATION.
877 2014-02-05 Nick Clifton <nickc@redhat.com>
879 * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Do not
880 accept extended registers in any mode when compiling for the MN10300.
882 2014-02-05 Yury Gribov <y.gribov@samsung.com>
884 * cif-code.def (ATTRIBUTE_MISMATCH): New CIF code.
885 * ipa-inline.c (report_inline_failed_reason): Handle mismatched
886 sanitization attributes.
887 (can_inline_edge_p): Likewise.
888 (sanitize_attrs_match_for_inline_p): New function.
890 2014-02-04 Jan Hubicka <hubicka@ucw.cz>
892 * ipa-prop.c (detect_type_change): Shor circuit testing of
893 type changes on THIS pointer.
895 2014-02-04 John David Anglin <danglin@gcc.gnu.org>
898 * config/pa/pa.c (legitimize_tls_address): Return original address
899 if not passed a SYMBOL_REF rtx.
900 (hppa_legitimize_address): Call legitimize_tls_address for all TLS
902 (pa_emit_move_sequence): Simplify TLS source operands.
903 (pa_legitimate_constant_p): Reject all TLS constants.
904 * config/pa/pa.h (PA_SYMBOL_REF_TLS_P): Correct comment.
905 (CONSTANT_ADDRESS_P): Reject TLS CONST addresses.
907 2014-02-04 Jan Hubicka <hubicka@ucw.cz>
909 * ipa.c (function_and_variable_visibility): Decompose DECL_ONE_ONLY
910 groups when we know they are controlled by LTO.
911 * varasm.c (default_binds_local_p_1): If object is in other partition,
912 it will be resolved locally.
914 2014-02-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
916 * config/host-linux.c (linux_gt_pch_use_address): Don't
917 use SSIZE_MAX because it is not always defined.
919 2014-02-04 Vladimir Makarov <vmakarov@redhat.com>
922 * lra-constraints.c (need_for_split_p): Use more 3 reloads as
923 threshold for pseudo splitting.
924 (update_ebb_live_info): Process call argument hard registers and
925 hard registers from insn definition too.
926 (max_small_class_regs_num): New constant.
927 (inherit_in_ebb): Update live hard regs through EBBs. Update
928 reloads_num only for small register classes. Don't split for
931 2014-02-04 Markus Trippelsdorf <markus@trippelsdorf.de>
934 * ipa-cp.c (ipa_get_indirect_edge_target_1): Check that target
937 2014-02-04 Jan Hubicka <hubicka@ucw.cz>
939 * gimple-fold.c (can_refer_decl_in_current_unit_p): Default
942 2014-02-04 Marek Polacek <polacek@redhat.com>
944 * gdbinit.in (pel): Define.
946 2014-02-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
948 * doc/invoke.texi (fstrict-volatile-bitfields): Clarify current
951 2014-02-04 Richard Biener <rguenther@suse.de>
954 * lto-streamer-out.c (tree_is_indexable): Force NAMELIST_DECLs
955 in function context local.
956 (lto_output_tree_ref): Do not write trees from lto_output_tree_ref.
957 * lto-streamer-in.c (lto_input_tree_ref): Handle LTO_namelist_decl_ref
958 similar to LTO_imported_decl_ref.
960 2014-02-04 Jakub Jelinek <jakub@redhat.com>
962 PR tree-optimization/60002
963 * cgraphclones.c (build_function_decl_skip_args): Clear
966 PR tree-optimization/60023
967 * tree-if-conv.c (predicate_mem_writes): Pass true instead of
968 false to gsi_replace.
969 * tree-vect-stmts.c (vect_finish_stmt_generation): If stmt
970 has been in some EH region and vec_stmt could throw, add
971 vec_stmt into the same EH region.
972 * tree-data-ref.c (get_references_in_stmt): If IFN_MASK_LOAD
973 has no lhs, ignore it.
974 * internal-fn.c (expand_MASK_LOAD): Likewise.
977 * tree-inline.c (copy_forbidden): Fail for
978 __attribute__((optimize (0))) functions.
981 * omp-low.c (simd_clone_struct_copy): If from->inbranch
982 is set, copy one less argument.
983 (expand_simd_clones): Don't subtract clone_info->inbranch
984 from simd_clone_struct_alloc argument.
986 PR rtl-optimization/57915
987 * recog.c (simplify_while_replacing): If all unary/binary/relational
988 operation arguments are constant, attempt to simplify those.
991 * expmed.c (expand_mult): For MODE_VECTOR_INT multiplication
992 if there is no vashl<mode>3 or ashl<mode>3 insn, skip_synth.
994 2014-02-04 Richard Biener <rguenther@suse.de>
996 PR tree-optimization/60012
997 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Apply
998 TBAA disambiguation to all DDRs.
1000 2014-02-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1003 * config/sol2.h (LINK_LIBGCC_MAPFILE_SPEC): Define.
1004 (LINK_SPEC): Use it for -shared, -shared-libgcc.
1006 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
1009 * tree.c (get_binfo_at_offset): Do not get confused by empty classes;
1011 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
1013 * gimple-fold.c (gimple_extract_devirt_binfo_from_cst): Remove.
1014 * gimple-fold.h (gimple_extract_devirt_binfo_from_cst): Remove.
1016 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
1019 * ipa-cp.c (ipa_get_indirect_edge_target_1): Use ipa-devirt
1020 to figure out targets of polymorphic calls with known decl.
1021 * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
1022 * ipa-utils.h (get_polymorphic_call_info_from_invariant): Declare.
1023 * ipa-devirt.c (get_polymorphic_call_info_for_decl): Break out from ...
1024 (get_polymorphic_call_info): ... here.
1025 (get_polymorphic_call_info_from_invariant): New function.
1027 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
1029 * ipa-cp.c (ipa_get_indirect_edge_target_1): Do direct
1030 lookup via vtable pointer; check for type consistency
1031 and turn inconsitent facts into UNREACHABLE.
1032 * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
1033 * gimple-fold.c (gimple_get_virt_method_for_vtable): Do not ICE on
1034 type inconsistent querries; return UNREACHABLE instead.
1036 2014-02-03 Richard Henderson <rth@twiddle.net>
1039 * tree-ssa-uninit.c (push_to_worklist): Don't re-push if we've
1040 already processed this node.
1041 (normalize_one_pred_1): Pass along mark_set.
1042 (normalize_one_pred): Create and destroy a pointer_set_t.
1043 (normalize_one_pred_chain): Likewise.
1045 2014-02-03 Laurent Aflonsi <laurent.alfonsi@st.com>
1047 PR gcov-profile/58602
1048 * gcc/gcov-io.c (gcov_open): Open with truncation when mode < 0.
1050 2014-02-03 Jan Hubicka <hubicka@ucw.cz>
1053 * ipa-cp.c (ipa_get_indirect_edge_target_1): Give up on
1054 -fno-devirtualize; try to devirtualize by the knowledge of
1055 virtual table pointer given by aggregate propagation.
1056 * ipa-prop.c (try_make_edge_direct_virtual_call): Likewise.
1057 (ipa_print_node_jump_functions): Dump also offset that
1058 is relevant for polymorphic calls.
1059 (determine_known_aggregate_parts): Add arg_type parameter; use it
1060 instead of determining the type from pointer type.
1061 (ipa_compute_jump_functions_for_edge): Update call of
1062 determine_known_aggregate_parts.
1063 * gimple-fold.c (gimple_get_virt_method_for_vtable): Break out from ...
1064 (gimple_get_virt_method_for_binfo): ... here; simplify using
1065 vtable_pointer_value_to_vtable.
1066 * gimple-fold.h (gimple_get_virt_method_for_vtable): Declare.
1067 * ipa-devirt.c (subbinfo_with_vtable_at_offset): Turn OFFSET parameter
1068 to unsigned HOST_WIDE_INT; use vtable_pointer_value_to_vtable.
1069 (vtable_pointer_value_to_vtable): Break out from ...; handle also
1071 (vtable_pointer_value_to_binfo): ... here.
1072 * ipa-utils.h (vtable_pointer_value_to_vtable): Declare.
1074 2014-02-03 Teresa Johnson <tejohnson@google.com>
1076 * tree-vect-slp.c (vect_supported_load_permutation_p): Avoid
1077 redef of outer loop index variable.
1079 2014-02-03 Marc Glisse <marc.glisse@inria.fr>
1083 * doc/extend.texi (Function Attributes): Typo.
1085 2014-02-03 Cong Hou <congh@google.com>
1087 PR tree-optimization/60000
1088 * tree-vect-loop.c (vect_transform_loop): Set pattern_def_seq to NULL
1089 if the vectorized statement is a store. A store statement can only
1090 appear at the end of pattern statements.
1092 2014-02-03 H.J. Lu <hongjiu.lu@intel.com>
1094 * config/i386/i386.c (flag_opts): Add -mlong-double-128.
1095 (ix86_option_override_internal): Default long double to 64-bit for
1096 32-bit Bionic and to 128-bit for 64-bit Bionic.
1098 * config/i386/i386.h (LONG_DOUBLE_TYPE_SIZE): Use 128 if
1099 TARGET_LONG_DOUBLE_128 is true.
1100 (LIBGCC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1102 * config/i386/i386.opt (mlong-double-80): Negate -mlong-double-64.
1103 (mlong-double-64): Negate -mlong-double-128.
1104 (mlong-double-128): New option.
1106 * config/i386/i386-c.c (ix86_target_macros): Define
1107 __LONG_DOUBLE_128__ for TARGET_LONG_DOUBLE_128.
1109 * doc/invoke.texi: Document -mlong-double-128.
1111 2014-02-03 H.J. Lu <hongjiu.lu@intel.com>
1113 PR rtl-optimization/60024
1114 * sel-sched.c (init_regs_for_mode): Check if mode is OK first.
1116 2014-02-03 Markus Trippelsdorf <markus@trippelsdorf.de>
1118 * doc/invoke.texi (fprofile-reorder-functions): Fix typo.
1120 2014-02-03 Andrey Belevantsev <abel@ispras.ru>
1122 PR rtl-optimization/57662
1123 * sel-sched.c (code_motion_path_driver): Do not mark already not
1124 existing blocks in the visiting bitmap.
1126 2014-02-03 Andrey Belevantsev <abel@ispras.ru>
1128 * sel-sched-ir.c (sel_gen_insn_from_expr_after): Reset INSN_DELETED_P
1129 on the insn being emitted.
1131 2014-02-03 James Greenhalgh <james.greenhalgh@arm.com>
1132 Will Deacon <will.deacon@arm.com>
1134 * doc/gimple.texi (gimple_asm_clear_volatile): Remove.
1136 2014-02-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1138 * config/arm/arm-tables.opt: Regenerate.
1140 2014-02-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1142 * config/rs6000/rs6000.c (altivec_expand_vec_perm_le): Generalize
1143 for vector types other than V16QImode.
1144 * config/rs6000/altivec.md (altivec_vperm_<mode>): Change to a
1145 define_expand, and call altivec_expand_vec_perm_le when producing
1146 code with little endian element order.
1147 (*altivec_vperm_<mode>_internal): New insn having previous
1148 behavior of altivec_vperm_<mode>.
1149 (altivec_vperm_<mode>_uns): Change to a define_expand, and call
1150 altivec_expand_vec_perm_le when producing code with little endian
1152 (*altivec_vperm_<mode>_uns_internal): New insn having previous
1153 behavior of altivec_vperm_<mode>_uns.
1155 2014-02-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1157 * config/rs6000/altivec.md (UNSPEC_VSUMSWS_DIRECT): New unspec.
1158 (altivec_vsumsws): Add handling for -maltivec=be with a little
1160 (altivec_vsumsws_direct): New.
1161 (reduc_splus_<mode>): Call gen_altivec_vsumsws_direct instead of
1162 gen_altivec_vsumsws.
1164 2014-02-02 Jan Hubicka <hubicka@ucw.cz>
1166 * ipa-devirt.c (subbinfo_with_vtable_at_offset,
1167 vtable_pointer_value_to_binfo): New functions.
1168 * ipa-utils.h (vtable_pointer_value_to_binfo): Declare.
1169 * ipa-prop.c (extr_type_from_vtbl_ptr_store): Use it.
1171 2014-02-02 Sandra Loosemore <sandra@codesourcery.com>
1173 * config/nios2/nios2.md (load_got_register): Initialize GOT
1174 pointer from _gp_got instead of _GLOBAL_OFFSET_TABLE_.
1175 * config/nios2/nios2.c (nios2_function_profiler): Likewise.
1177 2014-02-02 Jan Hubicka <hubicka@ucw.cz>
1179 * ipa-prop.c (update_jump_functions_after_inlining): When type is not
1180 preserverd by passthrough, do not propagate the type.
1182 2014-02-02 Richard Sandiford <rdsandiford@googlemail.com>
1184 * config/mips/mips.c (MIPS_GET_FCSR, MIPS_SET_FCSR): New macros.
1185 (mips_atomic_assign_expand_fenv): New function.
1186 (TARGET_ATOMIC_ASSIGN_EXPAND_FENV): Define.
1188 2014-02-02 Richard Sandiford <rdsandiford@googlemail.com>
1190 * doc/extend.texi (__builtin_mips_get_fcsr): Document.
1191 (__builtin_mips_set_fcsr): Likewise.
1192 * config/mips/mips-ftypes.def: Add MIPS_VOID_FTYPE_USI and
1193 MIPS_USI_FTYPE_VOID.
1194 * config/mips/mips-protos.h (mips16_expand_get_fcsr): Declare
1195 (mips16_expand_set_fcsr): Likewise.
1196 * config/mips/mips.c (mips16_get_fcsr_stub): New variable.
1197 (mips16_set_fcsr_stub): Likewise.
1198 (mips16_get_fcsr_one_only_stub): New class.
1199 (mips16_set_fcsr_one_only_stub): Likewise.
1200 (mips16_expand_get_fcsr, mips16_expand_set_fcsr): New functions.
1201 (mips_code_end): Output the get_fcsr and set_fcsr stubs, if needed.
1202 (BUILTIN_AVAIL_MIPS16, AVAIL_ALL): New macros.
1203 (hard_float): New availability predicate.
1204 (mips_builtins): Add get_fcsr and set_fcsr.
1205 (mips_expand_builtin): Check BUILTIN_AVAIL_MIPS16.
1206 * config/mips/mips.md (UNSPEC_GET_FCSR, UNSPEC_SET_FCSR): New unspecs.
1207 (GET_FCSR_REGNUM, SET_FCSR_REGNUM): New constants.
1208 (mips_get_fcsr, *mips_get_fcsr, mips_get_fcsr_mips16_<mode>)
1209 (mips_set_fcsr, *mips_set_fcsr, mips_set_fcsr_mips16_<mode>): New
1212 2014-02-02 Richard Sandiford <rdsandiford@googlemail.com>
1214 * config/mips/mips.c (mips_one_only_stub): New class.
1215 (mips_need_mips16_rdhwr_p): Replace with...
1216 (mips16_rdhwr_stub): ...this new variable.
1217 (mips16_stub_call_address): New function.
1218 (mips16_rdhwr_one_only_stub): New class.
1219 (mips_expand_thread_pointer): Use mips16_stub_call_address.
1220 (mips_output_mips16_rdhwr): Delete.
1221 (mips_finish_stub): New function.
1222 (mips_code_end): Use it to handle rdhwr stubs.
1224 2014-02-02 Uros Bizjak <ubizjak@gmail.com>
1227 * config/i386/i386.c (classify_argument): Fix handling of bit_offset
1228 when calculating size of integer atomic types.
1230 2014-02-02 H.J. Lu <hongjiu.lu@intel.com>
1232 * ipa-inline-analysis.c (true_predicate_p): Fix a typo in comments.
1234 2014-02-01 Jakub Jelinek <jakub@redhat.com>
1236 PR tree-optimization/60003
1237 * gimple-low.c (lower_builtin_setjmp): Set cfun->has_nonlocal_label.
1238 * profile.c (branch_prob): Use gimple_call_builtin_p
1239 to check for BUILT_IN_SETJMP_RECEIVER.
1240 * tree-inline.c (copy_bb): Call notice_special_calls.
1242 2014-01-31 Vladimir Makarov <vmakarov@redhat.com>
1245 * lra-constraints.c (process_alt_operands): Update reload_sum only
1248 2014-01-31 Richard Henderson <rth@redhat.com>
1251 * tree-eh.c (lower_try_finally_switch): Delay lowering finally block
1252 until after else_eh is processed.
1254 2014-01-31 Ilya Tocar <ilya.tocar@intel.com>
1256 * config/i386/avx512fintrin.h (_MM_FROUND_TO_NEAREST_INT),
1257 (_MM_FROUND_TO_NEG_INF), (_MM_FROUND_TO_POS_INF),
1258 (_MM_FROUND_TO_ZERO), (_MM_FROUND_CUR_DIRECTION): Are already defined
1259 in smmintrin.h, remove them.
1260 (_MM_FROUND_NO_EXC): Same as above, bit also wrong value.
1261 * config/i386/i386.c (ix86_print_operand): Split sae and rounding.
1262 * config/i386/i386.md (ROUND_SAE): Fix value.
1263 * config/i386/predicates.md (const_4_or_8_to_11_operand): New.
1264 (const48_operand): New.
1265 * config/i386/subst.md (round), (round_expand): Use
1266 const_4_or_8_to_11_operand.
1267 (round_saeonly), (round_saeonly_expand): Use const48_operand.
1269 2014-01-31 Ilya Tocar <ilya.tocar@intel.com>
1271 * config/i386/constraints.md (Yk): Swap meaning with k.
1272 * config/i386/i386.md (movhi_internal): Change Yk to k.
1273 (movqi_internal): Ditto.
1274 (*k<logic><mode>): Ditto.
1277 (kandn<mode>): Ditto.
1278 (*<code>hi_1): Ditto.
1279 (*<code>qi_1): Ditto.
1280 (kxnor<mode>): Ditto.
1281 (kortestzhi): Ditto.
1282 (kortestchi): Ditto.
1284 (*one_cmplhi2_1): Ditto.
1285 (*one_cmplqi2_1): Ditto.
1286 * config/i386/sse.md (): Change k to Yk.
1287 (avx512f_load<mode>_mask): Ditto.
1288 (avx512f_blendm<mode>): Ditto.
1289 (avx512f_store<mode>_mask): Ditto.
1290 (avx512f_storeu<ssemodesuffix>512_mask): Ditto.
1291 (avx512f_storedqu<mode>_mask): Ditto.
1292 (avx512f_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
1294 (avx512f_ucmp<mode>3<mask_scalar_merge_name>): Ditto.
1295 (avx512f_vmcmp<mode>3<round_saeonly_name>): Ditto.
1296 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Ditto.
1297 (avx512f_maskcmp<mode>3): Ditto.
1298 (avx512f_fmadd_<mode>_mask<round_name>): Ditto.
1299 (avx512f_fmadd_<mode>_mask3<round_name>): Ditto.
1300 (avx512f_fmsub_<mode>_mask<round_name>): Ditto.
1301 (avx512f_fmsub_<mode>_mask3<round_name>): Ditto.
1302 (avx512f_fnmadd_<mode>_mask<round_name>): Ditto.
1303 (avx512f_fnmadd_<mode>_mask3<round_name>): Ditto.
1304 (avx512f_fnmsub_<mode>_mask<round_name>): Ditto.
1305 (avx512f_fnmsub_<mode>_mask3<round_name>): Ditto.
1306 (avx512f_fmaddsub_<mode>_mask<round_name>): Ditto.
1307 (avx512f_fmaddsub_<mode>_mask3<round_name>): Ditto.
1308 (avx512f_fmsubadd_<mode>_mask<round_name>): Ditto.
1309 (avx512f_fmsubadd_<mode>_mask3<round_name>): Ditto.
1310 (avx512f_vextract<shuffletype>32x4_1_maskm): Ditto.
1311 (vec_extract_lo_<mode>_maskm): Ditto.
1312 (vec_extract_hi_<mode>_maskm): Ditto.
1313 (avx512f_vternlog<mode>_mask): Ditto.
1314 (avx512f_fixupimm<mode>_mask<round_saeonly_name>): Ditto.
1315 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Ditto.
1316 (avx512f_<code><pmov_src_lower><mode>2_mask): Ditto.
1317 (avx512f_<code>v8div16qi2_mask): Ditto.
1318 (avx512f_<code>v8div16qi2_mask_store): Ditto.
1319 (avx512f_eq<mode>3<mask_scalar_merge_name>_1): Ditto.
1320 (avx512f_gt<mode>3<mask_scalar_merge_name>): Ditto.
1321 (avx512f_testm<mode>3<mask_scalar_merge_name>): Ditto.
1322 (avx512f_testnm<mode>3<mask_scalar_merge_name>): Ditto.
1323 (*avx512pf_gatherpf<mode>sf_mask): Ditto.
1324 (*avx512pf_gatherpf<mode>df_mask): Ditto.
1325 (*avx512pf_scatterpf<mode>sf_mask): Ditto.
1326 (*avx512pf_scatterpf<mode>df_mask): Ditto.
1327 (avx512cd_maskb_vec_dupv8di): Ditto.
1328 (avx512cd_maskw_vec_dupv16si): Ditto.
1329 (avx512f_vpermi2var<mode>3_maskz): Ditto.
1330 (avx512f_vpermi2var<mode>3_mask): Ditto.
1331 (avx512f_vpermi2var<mode>3_mask): Ditto.
1332 (avx512f_vpermt2var<mode>3_maskz): Ditto.
1333 (*avx512f_gathersi<mode>): Ditto.
1334 (*avx512f_gathersi<mode>_2): Ditto.
1335 (*avx512f_gatherdi<mode>): Ditto.
1336 (*avx512f_gatherdi<mode>_2): Ditto.
1337 (*avx512f_scattersi<mode>): Ditto.
1338 (*avx512f_scatterdi<mode>): Ditto.
1339 (avx512f_compress<mode>_mask): Ditto.
1340 (avx512f_compressstore<mode>_mask): Ditto.
1341 (avx512f_expand<mode>_mask): Ditto.
1342 * config/i386/subst.md (mask): Change k to Yk.
1343 (mask_scalar_merge): Ditto.
1346 2014-01-31 Marc Glisse <marc.glisse@inria.fr>
1348 * doc/extend.texi (Vector Extensions): Document ?: in C++.
1350 2014-01-31 Richard Biener <rguenther@suse.de>
1353 * builtins.c (fold_builtin_memory_op): Make sure to not
1354 use a floating-point mode or a boolean or enumeral type for
1357 2014-01-30 DJ Delorie <dj@redhat.com>
1359 * config/msp430/msp430.h (LIB_SPEC): Add -lcrt
1360 * config/msp430/msp430.md (msp430_refsym_need_exit): New.
1361 * config/msp430/msp430.c (msp430_expand_epilogue): Call it
1362 whenever main() has an epilogue.
1364 2014-01-30 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1366 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Remove
1367 unused variable "field".
1368 * config/rs6000/vsx.md (vsx_mergel_<mode>): Add missing DONE.
1369 (vsx_mergeh_<mode>): Likewise.
1370 * config/rs6000/altivec.md (altivec_vmrghb): Likewise.
1371 (altivec_vmrghh): Likewise.
1372 (altivec_vmrghw): Likewise.
1373 (altivec_vmrglb): Likewise.
1374 (altivec_vmrglh): Likewise.
1375 (altivec_vmrglw): Likewise.
1376 (altivec_vspltb): Add missing uses.
1377 (altivec_vsplth): Likewise.
1378 (altivec_vspltw): Likewise.
1379 (altivec_vspltsf): Likewise.
1381 2014-01-30 Jakub Jelinek <jakub@redhat.com>
1384 * ifcvt.c (cond_exec_process_insns): Don't conditionalize
1385 frame related instructions.
1387 2014-01-30 Vladimir Makarov <vmakarov@redhat.com>
1389 PR rtl-optimization/59959
1390 * lra-constrains.c (simplify_operand_subreg): Assign NO_REGS to
1391 any reload of register whose subreg is invalid.
1393 2014-01-30 Jakub Jelinek <jakub@redhat.com>
1395 * config/i386/f16cintrin.h (_cvtsh_ss): Avoid -Wnarrowing warning.
1396 * config/i386/avx512fintrin.h (_mm512_mask_cvtusepi64_storeu_epi32):
1397 Add missing return type - void.
1399 2014-01-30 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1401 * gcc/config/rs6000/rs6000.c (rs6000_expand_vector_init): Use
1402 gen_vsx_xxspltw_v4sf_direct instead of gen_vsx_xxspltw_v4sf;
1403 remove element index adjustment for endian (now handled in vsx.md
1405 (altivec_expand_vec_perm_const): Use
1406 gen_altivec_vsplt[bhw]_direct instead of gen_altivec_vsplt[bhw].
1407 * gcc/config/rs6000/vsx.md (UNSPEC_VSX_XXSPLTW): New unspec.
1408 (vsx_xxspltw_<mode>): Adjust element index for little endian.
1409 * gcc/config/rs6000/altivec.md (altivec_vspltb): Divide into a
1410 define_expand and a new define_insn *altivec_vspltb_internal;
1411 adjust for -maltivec=be on a little endian target.
1412 (altivec_vspltb_direct): New.
1413 (altivec_vsplth): Divide into a define_expand and a new
1414 define_insn *altivec_vsplth_internal; adjust for -maltivec=be on a
1415 little endian target.
1416 (altivec_vsplth_direct): New.
1417 (altivec_vspltw): Divide into a define_expand and a new
1418 define_insn *altivec_vspltw_internal; adjust for -maltivec=be on a
1419 little endian target.
1420 (altivec_vspltw_direct): New.
1421 (altivec_vspltsf): Divide into a define_expand and a new
1422 define_insn *altivec_vspltsf_internal; adjust for -maltivec=be on
1423 a little endian target.
1425 2014-01-30 Richard Biener <rguenther@suse.de>
1427 PR tree-optimization/59993
1428 * tree-ssa-forwprop.c (associate_pointerplus): Check we
1429 can propagate form the earlier stmt and avoid the transform
1430 when the intermediate result is needed.
1432 2014-01-30 Alangi Derick <alangiderick@gmail.com>
1434 * README.Portability: Fix typo.
1436 2014-01-30 David Holsgrove <david.holsgrove@xilinx.com>
1438 * config/microblaze/microblaze.md(cstoresf4, cbranchsf4): Replace
1439 comparison_operator with ordered_comparison_operator.
1441 2014-01-30 Nick Clifton <nickc@redhat.com>
1443 * config/mn10300/mn10300-protos.h (mn10300_store_multiple_operation_p):
1444 Rename to mn10300_store_multiple_regs.
1445 * config/mn10300/mn10300.c: Likewise.
1446 * config/mn10300/mn10300.md (store_movm): Fix typo: call
1447 store_multiple_regs.
1448 * config/mn10300/predicates.md (mn10300_store_multiple_operation):
1449 Call mn10300_store_multiple_regs.
1451 2014-01-30 Nick Clifton <nickc@redhat.com>
1452 DJ Delorie <dj@redhat.com>
1454 * config/rl78/rl78.c (register_sizes): Make the "upper half" of
1455 %fp 2 to keep registers after it properly word-aligned.
1456 (rl78_alloc_physical_registers_umul): Handle the case where both
1457 input operands are the same.
1459 2014-01-30 Richard Biener <rguenther@suse.de>
1461 PR tree-optimization/59903
1462 * tree-vect-loop.c (vect_transform_loop): Guard multiple-types
1465 2014-01-30 Jason Merrill <jason@redhat.com>
1468 * tree.c (walk_type_fields): Handle VECTOR_TYPE.
1471 * cgraphunit.c (expand_thunk): Copy volatile arg to a temporary.
1473 2014-01-30 Richard Biener <rguenther@suse.de>
1475 PR tree-optimization/59951
1476 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Skip uses in debug insns.
1478 2014-01-30 Savin Zlobec <savin.zlobec@gmail.com>
1481 * config/nios2/nios2.c (nios2_fpu_insn_asm): Fix asm output of
1482 SFmode to DFmode case.
1484 2014-01-29 DJ Delorie <dj@redhat.com>
1486 * config/msp430/msp430.opt (-minrt): New.
1487 * config/msp430/msp430.h (STARTFILE_SPEC): Link alternate runtime
1489 (ENDFILE_SPEC): Likewise.
1491 2014-01-29 Jan Hubicka <hubicka@ucw.cz>
1493 * ipa-inline-analysis.c (clobber_only_eh_bb_p): New function.
1494 (estimate_function_body_sizes): Use it.
1496 2014-01-29 Paolo Carlini <paolo.carlini@oracle.com>
1499 * dwarf2out.c (is_cxx_auto): New.
1500 (is_base_type): Use it.
1501 (gen_type_die_with_usage): Likewise.
1503 2014-01-29 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1505 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Use
1506 CODE_FOR_altivec_vmrg*_direct rather than CODE_FOR_altivec_vmrg*.
1507 * config/rs6000/vsx.md (vsx_mergel_<mode>): Adjust for
1508 -maltivec=be with LE targets.
1509 (vsx_mergeh_<mode>): Likewise.
1510 * config/rs6000/altivec.md (UNSPEC_VMRG[HL]_DIRECT): New unspecs.
1511 (mulv8hi3): Use gen_altivec_vmrg[hl]w_direct.
1512 (altivec_vmrghb): Replace with define_expand and new
1513 *altivec_vmrghb_internal insn; adjust for -maltivec=be with LE targets.
1514 (altivec_vmrghb_direct): New define_insn.
1515 (altivec_vmrghh): Replace with define_expand and new
1516 *altivec_vmrghh_internal insn; adjust for -maltivec=be with LE targets.
1517 (altivec_vmrghh_direct): New define_insn.
1518 (altivec_vmrghw): Replace with define_expand and new
1519 *altivec_vmrghw_internal insn; adjust for -maltivec=be with LE targets.
1520 (altivec_vmrghw_direct): New define_insn.
1521 (*altivec_vmrghsf): Adjust for endianness.
1522 (altivec_vmrglb): Replace with define_expand and new
1523 *altivec_vmrglb_internal insn; adjust for -maltivec=be with LE targets.
1524 (altivec_vmrglb_direct): New define_insn.
1525 (altivec_vmrglh): Replace with define_expand and new
1526 *altivec_vmrglh_internal insn; adjust for -maltivec=be with LE targets.
1527 (altivec_vmrglh_direct): New define_insn.
1528 (altivec_vmrglw): Replace with define_expand and new
1529 *altivec_vmrglw_internal insn; adjust for -maltivec=be with LE targets.
1530 (altivec_vmrglw_direct): New define_insn.
1531 (*altivec_vmrglsf): Adjust for endianness.
1532 (vec_widen_umult_hi_v16qi): Use gen_altivec_vmrghh_direct.
1533 (vec_widen_umult_lo_v16qi): Use gen_altivec_vmrglh_direct.
1534 (vec_widen_smult_hi_v16qi): Use gen_altivec_vmrghh_direct.
1535 (vec_widen_smult_lo_v16qi): Use gen_altivec_vmrglh_direct.
1536 (vec_widen_umult_hi_v8hi): Use gen_altivec_vmrghw_direct.
1537 (vec_widen_umult_lo_v8hi): Use gen_altivec_vmrglw_direct.
1538 (vec_widen_smult_hi_v8hi): Use gen_altivec_vmrghw_direct.
1539 (vec_widen_smult_lo_v8hi): Use gen_altivec_vmrglw_direct.
1541 2014-01-29 Marcus Shawcroft <marcus.shawcroft@arm.com>
1543 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate)
1544 (aarch64_legitimate_address_p, aarch64_class_max_nregs): Adjust
1547 2014-01-29 Richard Biener <rguenther@suse.de>
1549 PR tree-optimization/58742
1550 * tree-ssa-forwprop.c (associate_pointerplus): Rename to
1551 associate_pointerplus_align.
1552 (associate_pointerplus_diff): New function.
1553 (associate_pointerplus): Likewise. Call associate_pointerplus_align
1554 and associate_pointerplus_diff.
1556 2014-01-29 Richard Biener <rguenther@suse.de>
1558 * lto-streamer.h (LTO_major_version): Bump to 3.
1559 (LTO_minor_version): Reset to 0.
1561 2014-01-29 Renlin Li <Renlin.Li@arm.com>
1563 * config/arm/arm-arches.def (ARM_ARCH): Add armv7ve arch.
1564 * config/arm/arm.c (FL_FOR_ARCH7VE): New.
1565 (arm_file_start): Generate correct asm header for armv7ve.
1566 * config/arm/bpabi.h: Add multilib support for armv7ve.
1567 * config/arm/driver-arm.c: Change the architectures of cortex-a7
1568 and cortex-a15 to armv7ve.
1569 * config/arm/t-aprofile: Add multilib support for armv7ve.
1570 * doc/invoke.texi: Document -march=armv7ve.
1572 2014-01-29 Richard Biener <rguenther@suse.de>
1574 PR tree-optimization/58742
1575 * tree-ssa-forwprop.c (associate_plusminus): Return true
1576 if we changed sth, defer EH cleanup to ...
1577 (ssa_forward_propagate_and_combine): ... here. Call simplify_mult.
1578 (simplify_mult): New function.
1580 2014-01-29 Jakub Jelinek <jakub@redhat.com>
1583 PR tree-optimization/59920
1584 * tree.c (build_common_builtin_nodes): Remove
1585 __builtin_setjmp_dispatcher initialization.
1586 * omp-low.h (make_gimple_omp_edges): Add a new int * argument.
1587 * profile.c (branch_prob): Use gsi_start_nondebug_after_labels_bb
1588 instead of gsi_after_labels + manually skipping debug stmts.
1589 Don't ignore bbs with BUILT_IN_SETJMP_DISPATCHER, instead
1590 ignore bbs with IFN_ABNORMAL_DISPATCHER.
1591 * tree-inline.c (copy_edges_for_bb): Remove
1592 can_make_abnormal_goto argument, instead add abnormal_goto_dest
1593 argument. Ignore computed_goto_p stmts. Don't call
1594 make_abnormal_goto_edges. If a call might need abnormal edges
1595 for non-local gotos, see if it already has an edge to
1596 IFN_ABNORMAL_DISPATCHER or if it is IFN_ABNORMAL_DISPATCHER
1597 with true argument, don't do anything then, otherwise add
1598 EDGE_ABNORMAL from the call's bb to abnormal_goto_dest.
1599 (copy_cfg_body): Compute abnormal_goto_dest, adjust copy_edges_for_bb
1601 * gimple-low.c (struct lower_data): Remove calls_builtin_setjmp.
1602 (lower_function_body): Don't emit __builtin_setjmp_dispatcher.
1603 (lower_stmt): Don't set data->calls_builtin_setjmp.
1604 (lower_builtin_setjmp): Adjust comment.
1605 * builtins.def (BUILT_IN_SETJMP_DISPATCHER): Remove.
1606 * tree-cfg.c (found_computed_goto): Remove.
1607 (factor_computed_gotos): Remove.
1608 (make_goto_expr_edges): Return bool, true for computed gotos.
1609 Don't call make_abnormal_goto_edges.
1610 (build_gimple_cfg): Don't set found_computed_goto, don't call
1611 factor_computed_gotos.
1612 (computed_goto_p): No longer static.
1613 (make_blocks): Don't set found_computed_goto.
1614 (get_abnormal_succ_dispatcher, handle_abnormal_edges): New functions.
1615 (make_edges): If make_goto_expr_edges returns true, push bb
1616 into ab_edge_goto vector, for stmt_can_make_abnormal_goto calls
1617 instead of calling make_abnormal_goto_edges push bb into ab_edge_call
1618 vector. Record mapping between bbs and OpenMP regions if there
1619 are any, adjust make_gimple_omp_edges caller. Call
1620 handle_abnormal_edges.
1621 (make_abnormal_goto_edges): Remove.
1622 * tree-cfg.h (make_abnormal_goto_edges): Remove.
1623 (computed_goto_p, get_abnormal_succ_dispatcher): New prototypes.
1624 * internal-fn.c (expand_ABNORMAL_DISPATCHER): New function.
1625 * builtins.c (expand_builtin): Don't handle BUILT_IN_SETJMP_DISPATCHER.
1626 * internal-fn.def (ABNORMAL_DISPATCHER): New.
1627 * omp-low.c (make_gimple_omp_edges): Add region_idx argument, when
1628 filling *region also set *region_idx to (*region)->entry->index.
1631 * read-rtl.c (read_rtx_code): Clear all of RTX_CODE_SIZE (code).
1632 For REGs set ORIGINAL_REGNO.
1634 2014-01-29 Bingfeng Mei <bmei@broadcom.com>
1636 * doc/md.texi: Mention that a target shouldn't implement
1637 vec_widen_(s|u)mul_even/odd pair if it is less efficient
1640 2014-01-29 Jakub Jelinek <jakub@redhat.com>
1642 PR tree-optimization/59594
1643 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Sort
1644 a copy of the datarefs vector rather than the vector itself.
1646 2014-01-28 Jason Merrill <jason@redhat.com>
1649 * dwarf2out.c (auto_die): New static.
1650 (gen_type_die_with_usage): Handle C++1y 'auto'.
1651 (gen_subprogram_die): If in-class DIE had 'auto', emit type again
1654 2014-01-28 H.J. Lu <hongjiu.lu@intel.com>
1657 * config/i386/gnu-user64.h (SPEC_32): Add "m16|" to "m32".
1658 (SPEC_X32): Likewise.
1659 (SPEC_64): Likewise.
1660 * config/i386/i386.c (ix86_option_override_internal): Turn off
1661 OPTION_MASK_ISA_64BIT, OPTION_MASK_ABI_X32 and OPTION_MASK_ABI_64
1663 (x86_file_start): Output .code16gcc for TARGET_16BIT.
1664 * config/i386/i386.h (TARGET_16BIT): New macro.
1665 (TARGET_16BIT_P): Likewise.
1666 * config/i386/i386.opt: Add m16.
1667 * doc/invoke.texi: Document -m16.
1669 2014-01-28 Jakub Jelinek <jakub@redhat.com>
1671 PR preprocessor/59935
1672 * input.c (location_get_source_line): Bail out on when line number
1673 is zero, and test the return value of lookup_or_add_file_to_cache_tab.
1675 2014-01-28 Richard Biener <rguenther@suse.de>
1677 PR tree-optimization/58742
1678 * tree-ssa-forwprop.c (associate_plusminus): Handle
1679 pointer subtraction of the form (T)(P + A) - (T)P.
1681 2014-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1683 * config/arm/arm.c (arm_new_rtx_costs): Remove useless statement
1686 2014-01-28 Richard Biener <rguenther@suse.de>
1689 2014-01-28 Richard Biener <rguenther@suse.de>
1691 PR rtl-optimization/45364
1692 PR rtl-optimization/59890
1693 * var-tracking.c (local_get_addr_clear_given_value): Handle
1694 already cleared slot.
1695 (val_reset): Handle not allocated local_get_addr_cache.
1696 (vt_find_locations): Use post-order on the inverted CFG.
1698 2014-01-28 Richard Biener <rguenther@suse.de>
1700 * tree-data-ref.h (ddr_is_anti_dependent, ddrs_have_anti_deps): Remove.
1702 2014-01-28 Richard Biener <rguenther@suse.de>
1704 PR rtl-optimization/45364
1705 PR rtl-optimization/59890
1706 * var-tracking.c (local_get_addr_clear_given_value): Handle
1707 already cleared slot.
1708 (val_reset): Handle not allocated local_get_addr_cache.
1709 (vt_find_locations): Use post-order on the inverted CFG.
1711 2014-01-28 Alan Modra <amodra@gmail.com>
1713 * Makefile.in (BUILD_CPPFLAGS): Do not use ALL_CPPFLAGS.
1714 * configure.ac <recursive call for build != host>: Define
1715 GENERATOR_FILE. Comment. Use CXX_FOR_BUILD, CXXFLAGS_FOR_BUILD
1716 and LD_FOR_BUILD too.
1717 * configure: Regenerate.
1719 2014-01-27 Allan Sandfeld Jensen <sandfeld@kde.org>
1721 * config/i386/i386.c (get_builtin_code_for_version): Separate
1722 Westmere from Nehalem, Ivy Bridge from Sandy Bridge and
1723 Broadwell from Haswell.
1725 2014-01-27 Steve Ellcey <sellcey@mips.com>
1727 * common/config/mips/mips-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1728 Remove TARGET_FP_EXCEPTIONS_DEFAULT and MASK_FUSED_MADD.
1729 * config/mips/mips.c (mips_option_override): Change setting
1731 * config/mips/mips.h (TARGET_FP_EXCEPTIONS_DEFAULT): Remove.
1732 * config/mips/mips.opt (DSP, DSPR2, FP_EXCEPTIONS, FUSED_MADD, MIPS3D):
1733 Change from Mask to Var.
1735 2014-01-27 Jeff Law <law@redhat.com>
1737 * ipa-inline.c (inline_small_functions): Fix typo.
1739 2014-01-27 Ilya Tocar <ilya.tocar@intel.com>
1741 * config/i386/avx512fintrin.h (_mm512_mask_cvtepi32_storeu_epi8): New.
1742 (_mm512_mask_cvtsepi32_storeu_epi8): Ditto.
1743 (_mm512_mask_cvtusepi32_storeu_epi8): Ditto.
1744 (_mm512_mask_cvtepi32_storeu_epi16): Ditto.
1745 (_mm512_mask_cvtsepi32_storeu_epi16): Ditto.
1746 (_mm512_mask_cvtusepi32_storeu_epi16): Ditto.
1747 (_mm512_mask_cvtepi64_storeu_epi32): Ditto.
1748 (_mm512_mask_cvtsepi64_storeu_epi32): Ditto.
1749 (_mm512_mask_cvtusepi64_storeu_epi32): Ditto.
1750 (_mm512_mask_cvtepi64_storeu_epi16): Ditto.
1751 (_mm512_mask_cvtsepi64_storeu_epi16): Ditto.
1752 (_mm512_mask_cvtusepi64_storeu_epi16): Ditto.
1753 (_mm512_mask_cvtepi64_storeu_epi8): Ditto.
1754 (_mm512_mask_cvtsepi64_storeu_epi8): Ditto.
1755 (_mm512_mask_cvtusepi64_storeu_epi8): Ditto.
1756 (_mm512_storeu_epi64): Ditto.
1757 (_mm512_cmpge_epi32_mask): Ditto.
1758 (_mm512_cmpge_epu32_mask): Ditto.
1759 (_mm512_cmpge_epi64_mask): Ditto.
1760 (_mm512_cmpge_epu64_mask): Ditto.
1761 (_mm512_cmple_epi32_mask): Ditto.
1762 (_mm512_cmple_epu32_mask): Ditto.
1763 (_mm512_cmple_epi64_mask): Ditto.
1764 (_mm512_cmple_epu64_mask): Ditto.
1765 (_mm512_cmplt_epi32_mask): Ditto.
1766 (_mm512_cmplt_epu32_mask): Ditto.
1767 (_mm512_cmplt_epi64_mask): Ditto.
1768 (_mm512_cmplt_epu64_mask): Ditto.
1769 (_mm512_cmpneq_epi32_mask): Ditto.
1770 (_mm512_cmpneq_epu32_mask): Ditto.
1771 (_mm512_cmpneq_epi64_mask): Ditto.
1772 (_mm512_cmpneq_epu64_mask): Ditto.
1773 (_mm512_expand_pd): Ditto.
1774 (_mm512_expand_ps): Ditto.
1775 * config/i386/i386-builtin-types.def: Add PV16QI, PV16QI, PV16HI,
1776 VOID_PV8SI_V8DI_QI, VOID_PV8HI_V8DI_QI, VOID_PV16QI_V8DI_QI,
1777 VOID_PV16QI_V16SI_HI, VOID_PV16HI_V16SI_HI.
1778 * config/i386/i386.c (ix86_builtins): Add
1779 IX86_BUILTIN_EXPANDPD512_NOMASK, IX86_BUILTIN_EXPANDPS512_NOMASK,
1780 IX86_BUILTIN_PMOVDB512_MEM, IX86_BUILTIN_PMOVDW512_MEM,
1781 IX86_BUILTIN_PMOVQB512_MEM, IX86_BUILTIN_PMOVQD512_MEM,
1782 IX86_BUILTIN_PMOVQW512_MEM, IX86_BUILTIN_PMOVSDB512_MEM,
1783 IX86_BUILTIN_PMOVSDW512_MEM, IX86_BUILTIN_PMOVSQB512_MEM,
1784 IX86_BUILTIN_PMOVSQD512_MEM, IX86_BUILTIN_PMOVSQW512_MEM,
1785 IX86_BUILTIN_PMOVUSDB512_MEM, IX86_BUILTIN_PMOVUSDW512_MEM,
1786 IX86_BUILTIN_PMOVUSQB512_MEM, IX86_BUILTIN_PMOVUSQD512_MEM,
1787 IX86_BUILTIN_PMOVUSQW512_MEM.
1788 (bdesc_special_args): Add __builtin_ia32_pmovusqd512mem_mask,
1789 __builtin_ia32_pmovsqd512mem_mask,
1790 __builtin_ia32_pmovqd512mem_mask,
1791 __builtin_ia32_pmovusqw512mem_mask,
1792 __builtin_ia32_pmovsqw512mem_mask,
1793 __builtin_ia32_pmovqw512mem_mask,
1794 __builtin_ia32_pmovusdw512mem_mask,
1795 __builtin_ia32_pmovsdw512mem_mask,
1796 __builtin_ia32_pmovdw512mem_mask,
1797 __builtin_ia32_pmovqb512mem_mask,
1798 __builtin_ia32_pmovusqb512mem_mask,
1799 __builtin_ia32_pmovsqb512mem_mask,
1800 __builtin_ia32_pmovusdb512mem_mask,
1801 __builtin_ia32_pmovsdb512mem_mask,
1802 __builtin_ia32_pmovdb512mem_mask.
1803 (bdesc_args): Add __builtin_ia32_expanddf512,
1804 __builtin_ia32_expandsf512.
1805 (ix86_expand_special_args_builtin): Handle VOID_FTYPE_PV8SI_V8DI_QI,
1806 VOID_FTYPE_PV8HI_V8DI_QI, VOID_FTYPE_PV16HI_V16SI_HI,
1807 VOID_FTYPE_PV16QI_V8DI_QI, VOID_FTYPE_PV16QI_V16SI_HI.
1808 * config/i386/sse.md (unspec): Add UNSPEC_EXPAND_NOMASK.
1809 (avx512f_<code><pmov_src_lower><mode>2_mask_store): New.
1810 (*avx512f_<code>v8div16qi2_store_mask): Renamed to ...
1811 (avx512f_<code>v8div16qi2_mask_store): This.
1812 (avx512f_expand<mode>): New.
1814 2014-01-27 Kirill Yukhin <kirill.yukhin@intel.com>
1816 * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i32gather_pd):
1818 (_mm512_mask_prefetch_i64gather_pd): Ditto.
1819 (_mm512_prefetch_i32scatter_pd): Ditto.
1820 (_mm512_mask_prefetch_i32scatter_pd): Ditto.
1821 (_mm512_prefetch_i64scatter_pd): Ditto.
1822 (_mm512_mask_prefetch_i64scatter_pd): Ditto.
1823 (_mm512_mask_prefetch_i32gather_ps): Fix operand type.
1824 (_mm512_mask_prefetch_i64gather_ps): Ditto.
1825 (_mm512_prefetch_i32scatter_ps): Ditto.
1826 (_mm512_mask_prefetch_i32scatter_ps): Ditto.
1827 (_mm512_prefetch_i64scatter_ps): Ditto.
1828 (_mm512_mask_prefetch_i64scatter_ps): Ditto.
1829 * config/i386/i386-builtin-types.def: Define
1830 VOID_FTYPE_QI_V8SI_PCINT64_INT_INT
1831 and VOID_FTYPE_QI_V8DI_PCINT64_INT_INT.
1832 * config/i386/i386.c (ix86_builtins): Define IX86_BUILTIN_GATHERPFQPD,
1833 IX86_BUILTIN_GATHERPFDPD, IX86_BUILTIN_SCATTERPFDPD,
1834 IX86_BUILTIN_SCATTERPFQPD.
1835 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_gatherpfdpd,
1836 __builtin_ia32_gatherpfdps, __builtin_ia32_gatherpfqpd,
1837 __builtin_ia32_gatherpfqps, __builtin_ia32_scatterpfdpd,
1838 __builtin_ia32_scatterpfdps, __builtin_ia32_scatterpfqpd,
1839 __builtin_ia32_scatterpfqps.
1840 (ix86_expand_builtin): Expand new built-ins.
1841 * config/i386/sse.md (avx512pf_gatherpf<mode>): Add SF suffix,
1842 fix memory access data type.
1843 (*avx512pf_gatherpf<mode>_mask): Ditto.
1844 (*avx512pf_gatherpf<mode>): Ditto.
1845 (avx512pf_scatterpf<mode>): Ditto.
1846 (*avx512pf_scatterpf<mode>_mask): Ditto.
1847 (*avx512pf_scatterpf<mode>): Ditto.
1848 (GATHER_SCATTER_SF_MEM_MODE): New.
1849 (avx512pf_gatherpf<mode>df): Ditto.
1850 (*avx512pf_gatherpf<mode>df_mask): Ditto.
1851 (*avx512pf_scatterpf<mode>df): Ditto.
1853 2014-01-27 Jakub Jelinek <jakub@redhat.com>
1856 * expmed.h (expmed_mode_index): Rework so that analysis and optimziers
1857 know when the MODE_PARTIAL_INT and MODE_VECTOR_INT cases can never be
1860 2014-01-27 James Greenhalgh <james.greenhalgh@arm.com>
1862 * common/config/arm/arm-common.c
1863 (arm_rewrite_mcpu): Handle multiple names.
1865 (BIG_LITTLE_SPEC): Do not discard mcpu switches.
1867 2014-01-27 James Greenhalgh <james.greenhalgh@arm.com>
1869 * gimple-builder.h (create_gimple_tmp): Delete.
1871 2014-01-27 Christian Bruel <christian.bruel@st.com>
1873 * config/sh/sh-mem.cc (sh_expand_cmpnstr): Fix remaining bytes after
1876 2014-01-26 John David Anglin <danglin@gcc.gnu.org>
1878 * config/pa/pa.md (call): Generate indirect long calls to non-local
1879 functions when outputing 32-bit code.
1880 (call_value): Likewise except for special call to buggy powf function.
1882 * config/pa/pa.c (pa_attr_length_indirect_call): Adjust length of
1883 portable runtime and PIC indirect calls.
1884 (pa_output_indirect_call): Remove unnecessary nop from portable runtime
1885 and PIC call sequences. Use ldo instead of blr to set return register
1886 in PIC call sequence.
1888 2014-01-25 Walter Lee <walt@tilera.com>
1890 * config/tilegx/sync.md (atomic_fetch_sub): Fix negation and
1891 avoid clobbering a live register.
1893 2014-01-25 Walter Lee <walt@tilera.com>
1895 * config/tilegx/tilegx-c.c (tilegx_cpu_cpp_builtins):
1896 Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2}.
1897 * config/tilegx/tilepro-c.c (tilepro_cpu_cpp_builtins):
1898 Define __GCC_HAVE_SYNC_COMPARE_AND_SWAP_{1,2,4,8}.
1900 2014-01-25 Walter Lee <walt@tilera.com>
1902 * config/tilegx/tilegx.c (tilegx_function_arg): Start 16-byte
1903 arguments on even registers.
1904 (tilegx_gimplify_va_arg_expr): Align 16-byte var args to
1906 * config/tilegx/tilegx.h (STACK_BOUNDARY): Change to 16 bytes.
1907 (BIGGEST_ALIGNMENT): Ditto.
1908 (BIGGEST_FIELD_ALIGNMENT): Ditto.
1910 2014-01-25 Walter Lee <walt@tilera.com>
1912 * config/tilegx/tilegx.c (tilegx_gen_bundles): Delete barrier
1913 insns before bundling.
1914 * config/tilegx/tilegx.md (tile_network_barrier): Update comment.
1916 2014-01-25 Walter Lee <walt@tilera.com>
1918 * config/tilegx/tilegx.c (tilegx_expand_builtin): Set
1919 PREFETCH_SCHEDULE_BARRIER_P to true for prefetches.
1920 * config/tilepro/tilepro.c (tilepro_expand_builtin): Ditto.
1922 2014-01-25 Richard Sandiford <rdsandiford@googlemail.com>
1924 * config/mips/constraints.md (kl): Delete.
1925 * config/mips/mips.md (divmod<mode>4, udivmod<mode>4): Turn into
1926 define expands, using...
1927 (divmod<mode>4_mips16, udivmod<mode>4_mips16): ...these new
1928 instructions for MIPS16.
1929 (*divmod<mode>4, *udivmod<mode>4): New patterns, taken from the
1930 non-MIPS16 version of the old divmod<mode>4 and udivmod<mode>4.
1932 2014-01-25 Walter Lee <walt@tilera.com>
1934 * config/tilepro/tilepro.md (ctzdi2): Use register_operand predicate.
1938 2014-01-25 Walter Lee <walt@tilera.com>
1940 * config/tilegx/tilegx.c (tilegx_expand_to_rtl_hook): New.
1941 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1943 2014-01-25 Richard Sandiford <rdsandiford@googlemail.com>
1945 * rtlanal.c (canonicalize_condition): Split out duplicated mode check.
1948 2014-01-25 Jakub Jelinek <jakub@redhat.com>
1950 * print-rtl.c (in_call_function_usage): New var.
1951 (print_rtx): When in CALL_INSN_FUNCTION_USAGE, always print
1952 EXPR_LIST mode as mode and not as reg note name.
1955 * cfgloopmanip.c (copy_loop_info): If
1956 loop->warned_aggressive_loop_optimizations, make sure
1957 the flag is set in target loop too.
1959 2014-01-24 Balaji V. Iyer <balaji.v.iyer@intel.com>
1961 * builtins.c (is_builtin_name): Renamed flag_enable_cilkplus to
1963 * builtins.def: Likewise.
1964 * cilk.h (fn_contains_cilk_spawn_p): Likewise.
1965 * cppbuiltin.c (define_builtin_macros_for_compilation_flags): Likewise.
1966 * ira.c (ira_setup_eliminable_regset): Likewise.
1967 * omp-low.c (gate_expand_omp): Likewise.
1968 (execute_lower_omp): Likewise.
1969 (diagnose_sb_0): Likewise.
1970 (gate_diagnose_omp_blocks): Likewise.
1971 (simd_clone_clauses_extract): Likewise.
1974 2014-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1976 * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Remove
1977 correction for little endian...
1978 * config/rs6000/vsx.md (vsx_xxpermdi2_<mode>_1): ...and move it to
1981 2014-01-24 Jeff Law <law@redhat.com>
1983 PR tree-optimization/59919
1984 * tree-vrp.c (find_assert_locations_1): Do not register asserts
1985 for non-returning calls.
1987 2014-01-24 James Greenhalgh <james.greenhalgh@arm.com>
1989 * common/config/aarch64/aarch64-common.c
1990 (aarch64_rewrite_mcpu): Handle multiple names.
1991 * config/aarch64/aarch64.h
1992 (BIG_LITTLE_SPEC): Do not discard mcpu switches.
1994 2014-01-24 Dodji Seketeli <dodji@redhat.com>
1996 * input.c (add_file_to_cache_tab): Handle the case where fopen
1999 2014-01-23 H.J. Lu <hongjiu.lu@intel.com>
2002 * config/i386/i386.md (pushsf splitter): Get stack adjustment
2003 from push operand if code of push isn't PRE_DEC.
2005 2014-01-23 Michael Meissner <meissner@linux.vnet.ibm.com>
2008 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2009 -mquad-memory-atomic. Update -mquad-memory documentation to say
2010 it is only used for non-atomic loads/stores.
2012 * config/rs6000/predicates.md (quad_int_reg_operand): Allow either
2013 -mquad-memory or -mquad-memory-atomic switches.
2015 * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Add
2016 -mquad-memory-atomic to ISA 2.07 support.
2018 * config/rs6000/rs6000.opt (-mquad-memory-atomic): Add new switch
2019 to separate support of normal quad word memory operations (ldq, stq)
2020 from the atomic quad word memory operations.
2022 * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
2023 support to separate non-atomic quad word operations from atomic
2024 quad word operations. Disable non-atomic quad word operations in
2025 little endian mode so that we don't have to swap words after the
2026 load and before the store.
2027 (quad_load_store_p): Add comment about atomic quad word support.
2028 (rs6000_opt_masks): Add -mquad-memory-atomic to the list of
2029 options printed with -mdebug=reg.
2031 * config/rs6000/rs6000.h (TARGET_SYNC_TI): Use
2032 -mquad-memory-atomic as the test for whether we have quad word
2033 atomic instructions.
2034 (TARGET_SYNC_HI_QI): If either -mquad-memory-atomic, -mquad-memory,
2035 or -mp8-vector are used, allow byte/half-word atomic operations.
2037 * config/rs6000/sync.md (load_lockedti): Insure that the address
2038 is a proper indexed or indirect address for the lqarx instruction.
2039 On little endian systems, swap the hi/lo registers after the lqarx
2041 (load_lockedpti): Use indexed_or_indirect_operand predicate to
2042 insure the address is valid for the lqarx instruction.
2043 (store_conditionalti): Insure that the address is a proper indexed
2044 or indirect address for the stqcrx. instruction. On little endian
2045 systems, swap the hi/lo registers before doing the stqcrx.
2047 (store_conditionalpti): Use indexed_or_indirect_operand predicate to
2048 insure the address is valid for the stqcrx. instruction.
2050 * gcc/config/rs6000/rs6000-c.c (rs6000_target_modify_macros):
2051 Define __QUAD_MEMORY__ and __QUAD_MEMORY_ATOMIC__ based on what
2052 type of quad memory support is available.
2054 2014-01-23 Vladimir Makarov <vmakarov@redhat.com>
2057 * lra-constraints.c (simplify_operand_subreg): Spill pseudo if
2058 there is a danger of looping.
2060 2014-01-23 Pat Haugen <pthaugen@us.ibm.com>
2062 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
2063 force flag_ira_loop_pressure if set via command line.
2065 2014-01-23 Alex Velenko <Alex.Velenko@arm.com>
2067 * config/aarch64/aarch64-simd-builtins.def (ashr): DI mode removed.
2068 (ashr_simd): New builtin handling DI mode.
2069 * config/aarch64/aarch64-simd.md (aarch64_ashr_simddi): New pattern.
2070 (aarch64_sshr_simddi): New match pattern.
2071 * config/aarch64/arm_neon.h (vshr_n_s32): Builtin call modified.
2072 (vshrd_n_s64): Likewise.
2073 * config/aarch64/predicates.md (aarch64_shift_imm64_di): New predicate.
2075 2014-01-23 Nick Clifton <nickc@redhat.com>
2077 * config/msp430/msp430.h (ASM_SPEC): Pass the -mcpu as -mcpu.
2078 (LIB_SPEC): Drop use of memory.ld and peripherals.ld scripts in
2079 favour of mcu specific scripts.
2080 * config/msp430/t-msp430 (MULTILIB_MATCHES): Add more matches for
2083 2014-01-23 James Greenhalgh <james.greenhalgh@arm.com>
2084 Alex Velenko <Alex.Velenko@arm.com>
2086 * config/aarch64/arm_neon.h (vaddv_s8): __LANE0 cleanup.
2087 (vaddv_s16): Likewise.
2088 (vaddv_s32): Likewise.
2089 (vaddv_u8): Likewise.
2090 (vaddv_u16): Likewise.
2091 (vaddv_u32): Likewise.
2092 (vaddvq_s8): Likewise.
2093 (vaddvq_s16): Likewise.
2094 (vaddvq_s32): Likewise.
2095 (vaddvq_s64): Likewise.
2096 (vaddvq_u8): Likewise.
2097 (vaddvq_u16): Likewise.
2098 (vaddvq_u32): Likewise.
2099 (vaddvq_u64): Likewise.
2100 (vaddv_f32): Likewise.
2101 (vaddvq_f32): Likewise.
2102 (vaddvq_f64): Likewise.
2103 (vmaxv_f32): Likewise.
2104 (vmaxv_s8): Likewise.
2105 (vmaxv_s16): Likewise.
2106 (vmaxv_s32): Likewise.
2107 (vmaxv_u8): Likewise.
2108 (vmaxv_u16): Likewise.
2109 (vmaxv_u32): Likewise.
2110 (vmaxvq_f32): Likewise.
2111 (vmaxvq_f64): Likewise.
2112 (vmaxvq_s8): Likewise.
2113 (vmaxvq_s16): Likewise.
2114 (vmaxvq_s32): Likewise.
2115 (vmaxvq_u8): Likewise.
2116 (vmaxvq_u16): Likewise.
2117 (vmaxvq_u32): Likewise.
2118 (vmaxnmv_f32): Likewise.
2119 (vmaxnmvq_f32): Likewise.
2120 (vmaxnmvq_f64): Likewise.
2121 (vminv_f32): Likewise.
2122 (vminv_s8): Likewise.
2123 (vminv_s16): Likewise.
2124 (vminv_s32): Likewise.
2125 (vminv_u8): Likewise.
2126 (vminv_u16): Likewise.
2127 (vminv_u32): Likewise.
2128 (vminvq_f32): Likewise.
2129 (vminvq_f64): Likewise.
2130 (vminvq_s8): Likewise.
2131 (vminvq_s16): Likewise.
2132 (vminvq_s32): Likewise.
2133 (vminvq_u8): Likewise.
2134 (vminvq_u16): Likewise.
2135 (vminvq_u32): Likewise.
2136 (vminnmv_f32): Likewise.
2137 (vminnmvq_f32): Likewise.
2138 (vminnmvq_f64): Likewise.
2140 2014-01-23 James Greenhalgh <james.greenhalgh@arm.com>
2142 * config/aarch64/aarch64-simd.md
2143 (aarch64_dup_lane<mode>): Correct lane number on big-endian.
2144 (aarch64_dup_lane_<vswap_widthi_name><mode>): Likewise.
2145 (*aarch64_mul3_elt<mode>): Likewise.
2146 (*aarch64_mul3_elt<vswap_width_name><mode>): Likewise.
2147 (*aarch64_mul3_elt_to_64v2df): Likewise.
2148 (*aarch64_mla_elt<mode>): Likewise.
2149 (*aarch64_mla_elt_<vswap_width_name><mode>): Likewise.
2150 (*aarch64_mls_elt<mode>): Likewise.
2151 (*aarch64_mls_elt_<vswap_width_name><mode>): Likewise.
2152 (*aarch64_fma4_elt<mode>): Likewise.
2153 (*aarch64_fma4_elt_<vswap_width_name><mode>): Likewise.
2154 (*aarch64_fma4_elt_to_64v2df): Likewise.
2155 (*aarch64_fnma4_elt<mode>): Likewise.
2156 (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
2157 (*aarch64_fnma4_elt_to_64v2df): Likewise.
2158 (aarch64_sq<r>dmulh_lane<mode>): Likewise.
2159 (aarch64_sq<r>dmulh_laneq<mode>): Likewise.
2160 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
2161 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>_internal): Likewise.
2162 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
2163 (aarch64_sqdmull_lane<mode>_internal): Likewise.
2164 (aarch64_sqdmull2_lane<mode>_internal): Likewise.
2166 2013-01-23 Alex Velenko <Alex.Velenko@arm.com>
2168 * config/aarch64/aarch64-simd.md
2169 (aarch64_be_checked_get_lane<mode>): New define_expand.
2170 * config/aarch64/aarch64-simd-builtins.def
2171 (BUILTIN_VALL (GETLANE, be_checked_get_lane, 0)):
2172 New builtin definition.
2173 * config/aarch64/arm_neon.h: (__aarch64_vget_lane_any):
2174 Use new safe be builtin.
2176 2014-01-23 Alex Velenko <Alex.Velenko@arm.com>
2178 * config/aarch64/aarch64-simd.md (aarch64_be_ld1<mode>):
2180 (aarch64_be_st1<mode>): Likewise.
2181 (aarch_ld1<VALL:mode>): Define_expand modified.
2182 (aarch_st1<VALL:mode>): Likewise.
2183 * config/aarch64/aarch64.md (UNSPEC_LD1): New unspec definition.
2184 (UNSPEC_ST1): Likewise.
2186 2014-01-23 David Holsgrove <david.holsgrove@xilinx.com>
2188 * config/microblaze/microblaze.md: Add trap insn and attribute
2190 2014-01-23 Dodji Seketeli <dodji@redhat.com>
2192 PR preprocessor/58580
2193 * input.h (location_get_source_line): Take an additional line_size
2195 (void diagnostics_file_cache_fini): Declare new function.
2196 * input.c (struct fcache): New type.
2197 (fcache_tab_size, fcache_buffer_size, fcache_line_record_size):
2198 New static constants.
2199 (diagnostic_file_cache_init, total_lines_num)
2200 (lookup_file_in_cache_tab, evicted_cache_tab_entry)
2201 (add_file_to_cache_tab, lookup_or_add_file_to_cache_tab)
2202 (needs_read, needs_grow, maybe_grow, read_data, maybe_read_data)
2203 (get_next_line, read_next_line, goto_next_line, read_line_num):
2204 New static function definitions.
2205 (diagnostic_file_cache_fini): New function.
2206 (location_get_source_line): Take an additional output line_len
2207 parameter. Re-write using lookup_or_add_file_to_cache_tab and
2209 * diagnostic.c (diagnostic_finish): Call
2210 diagnostic_file_cache_fini.
2211 (adjust_line): Take an additional input parameter for the length
2212 of the line, rather than calculating it with strlen.
2213 (diagnostic_show_locus): Adjust the use of
2214 location_get_source_line and adjust_line with respect to their new
2215 signature. While displaying a line now, do not stop at the first
2216 null byte. Rather, display the zero byte as a space and keep
2217 going until we reach the size of the line.
2218 * Makefile.in: Add vec.o to OBJS-libcommon
2220 2014-01-23 Kirill Yukhin <kirill.yukhin@intel.com>
2221 Ilya Tocar <ilya.tocar@intel.com>
2223 * config/i386/avx512fintrin.h (_mm512_kmov): New.
2224 * config/i386/i386.c (IX86_BUILTIN_KMOV16): Ditto.
2225 (__builtin_ia32_kmov16): Ditto.
2226 * config/i386/i386.md (UNSPEC_KMOV): New.
2229 2014-01-23 Kirill Yukhin <kirill.yukhin@intel.com>
2231 * config/i386/avx512fintrin.h (_mm512_loadu_si512): Rename.
2232 (_mm512_storeu_si512): Ditto.
2234 2014-01-23 Richard Sandiford <rdsandiford@googlemail.com>
2237 * rtl.h (get_referenced_operands): Declare.
2238 * recog.c (get_referenced_operands): New function.
2239 * config/mips/mips.c (mips_reorg_process_insns): Check which asm
2240 operands have been referenced when recording LO_SUM references.
2242 2014-01-22 David Holsgrove <david.holsgrove@xilinx.com>
2244 * config/microblaze/microblaze.md: Correct bswaphi2 insn.
2246 2014-01-22 Jan Hubicka <hubicka@ucw.cz>
2248 * config/i386/x86-tune.def (X86_TUNE_ACCUMULATE_OUTGOING_ARGS):
2249 Enable for generic and recent AMD targets.
2251 2014-01-22 Jan Hubicka <hubicka@ucw.cz>
2253 * combine-stack-adj.c (combine_stack_adjustments_for_block): Remove
2254 ARG_SIZE note when adjustment was eliminated.
2256 2014-01-22 Jeff Law <law@redhat.com>
2258 PR tree-optimization/59597
2259 * tree-ssa-threadupdate.c (dump_jump_thread_path): Move to earlier
2260 in file. Accept new argument REGISTERING and use it to modify
2261 dump output appropriately.
2262 (register_jump_thread): Corresponding changes.
2263 (mark_threaded_blocks): Reinstate code to cancel unprofitable
2264 thread paths involving joiner blocks. Add code to dump cancelled
2265 jump threading paths.
2267 2014-01-22 Vladimir Makarov <vmakarov@redhat.com>
2269 PR rtl-optimization/59477
2270 * lra-constraints.c (inherit_in_ebb): Process call for living hard
2271 regs. Update reloads_num and potential_reload_hard_regs for all insns.
2273 2014-01-22 Tom Tromey <tromey@redhat.com>
2275 * config/i386/i386-interix.h (i386_pe_unique_section): Don't use
2277 * config/cr16/cr16-protos.h (notice_update_cc): Don't use PARAMS.
2279 2014-01-21 Vladimir Makarov <vmakarov@redhat.com>
2281 PR rtl-optimization/59896
2282 * lra-constraints.c (process_alt_operands): Check unused note for
2283 matched operands of insn with no output reloads.
2285 2014-01-21 Richard Sandiford <rdsandiford@googlemail.com>
2287 * config/mips/mips.c (mips_move_to_gpr_cost): Add M16_REGS case.
2288 (mips_move_from_gpr_cost): Likewise.
2290 2014-01-21 Vladimir Makarov <vmakarov@redhat.com>
2292 PR rtl-optimization/59858
2293 * lra-constraints.c (SMALL_REGISTER_CLASS_P): Use
2294 ira_class_hard_regs_num.
2295 (process_alt_operands): Increase reject for dying matched operand.
2297 2014-01-21 Jakub Jelinek <jakub@redhat.com>
2300 * config/i386/i386.c (expand_small_movmem_or_setmem): If mode is
2301 smaller than size, perform several stores or loads and stores
2302 at dst + count - size to store or copy all of size bytes, rather
2303 than just last modesize bytes.
2305 2014-01-20 DJ Delorie <dj@redhat.com>
2307 * config/rl78/rl78.c (rl78_propogate_register_origins): Verify
2308 that CLOBBERs are REGs before propogating their values.
2310 2014-01-20 H.J. Lu <hongjiu.lu@intel.com>
2313 * cgraph.c (cgraph_inline_failed_string): Add type to DEFCIFCODE.
2314 (cgraph_inline_failed_type): New function.
2315 * cgraph.h (DEFCIFCODE): Add type.
2316 (cgraph_inline_failed_type_t): New enum.
2317 (cgraph_inline_failed_type): New prototype.
2318 * cif-code.def: Add CIF_FINAL_NORMAL to OK, FUNCTION_NOT_CONSIDERED,
2319 FUNCTION_NOT_OPTIMIZED, REDEFINED_EXTERN_INLINE,
2320 FUNCTION_NOT_INLINE_CANDIDATE, LARGE_FUNCTION_GROWTH_LIMIT,
2321 LARGE_STACK_FRAME_GROWTH_LIMIT, MAX_INLINE_INSNS_SINGLE_LIMIT,
2322 MAX_INLINE_INSNS_AUTO_LIMIT, INLINE_UNIT_GROWTH_LIMIT,
2323 RECURSIVE_INLINING, UNLIKELY_CALL, NOT_DECLARED_INLINED,
2324 OPTIMIZING_FOR_SIZE, ORIGINALLY_INDIRECT_CALL,
2325 INDIRECT_UNKNOWN_CALL, USES_COMDAT_LOCAL.
2326 Add CIF_FINAL_ERROR to UNSPECIFIED, BODY_NOT_AVAILABLE,
2327 FUNCTION_NOT_INLINABLE, OVERWRITABLE, MISMATCHED_ARGUMENTS,
2328 EH_PERSONALITY, NON_CALL_EXCEPTIONS, TARGET_OPTION_MISMATCH,
2329 OPTIMIZATION_MISMATCH.
2330 * tree-inline.c (expand_call_inline): Emit errors during
2331 early_inlining if cgraph_inline_failed_type returns CIF_FINAL_ERROR.
2333 2014-01-20 Uros Bizjak <ubizjak@gmail.com>
2336 * config/i386/sse.md (*andnot<mode>3<mask_name>): Handle MODE_V16SF
2337 mode attribute in insn output.
2339 2014-01-20 Eric Botcazou <ebotcazou@adacore.com>
2341 * output.h (output_constant): Delete.
2342 * varasm.c (output_constant): Make private.
2344 2014-01-20 Alex Velenko <Alex.Velenko@arm.com>
2346 * config/aarch64/aarch64-simd.md (vec_perm<mode>): Add BE check.
2348 2014-01-20 Jakub Jelinek <jakub@redhat.com>
2351 * tree.h (fold_builtin_strcat): New prototype.
2352 * builtins.c (fold_builtin_strcat): No longer static. Add len
2353 argument, if non-NULL, don't call c_strlen. Optimize
2354 directly into __builtin_memcpy instead of __builtin_strcpy.
2355 (fold_builtin_2): Adjust fold_builtin_strcat caller.
2356 * gimple-fold.c (gimple_fold_builtin): Handle BUILT_IN_STRCAT.
2358 2014-01-20 Uros Bizjak <ubizjak@gmail.com>
2360 * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
2361 for SImode_address_operand operands, having only a REG argument.
2363 2014-01-20 Marcus Shawcroft <marcus.shawcroft@arm.com>
2365 * config/aarch64/aarch64-linux.h (GLIBC_DYNAMIC_LINKER): Expand
2366 loader name using mbig-endian.
2367 (LINUX_TARGET_LINK_SPEC): Pass linker -m flag.
2369 2014-01-20 James Greenhalgh <james.greenhalgh@arm.com>
2371 * doc/invoke.texi (-march): Clarify documentation for AArch64.
2375 2014-01-20 Tejas Belagod <tejas.belagod@arm.com>
2377 * config/aarch64/aarch64-protos.h
2378 (aarch64_cannot_change_mode_class_ptr): Declare.
2379 * config/aarch64/aarch64.c (aarch64_cannot_change_mode_class,
2380 aarch64_cannot_change_mode_class_ptr): New.
2381 * config/aarch64/aarch64.h (CANNOT_CHANGE_MODE_CLASS): Change to call
2382 backend hook aarch64_cannot_change_mode_class.
2384 2014-01-20 James Greenhalgh <james.greenhalgh@arm.com>
2386 * common/config/aarch64/aarch64-common.c
2387 (aarch64_handle_option): Don't handle any option order logic here.
2388 * config/aarch64/aarch64.c (aarch64_parse_arch): Do not override
2389 selected_cpu, warn on architecture version mismatch.
2390 (aarch64_override_options): Fix parsing order for option strings.
2392 2014-01-20 Jan-Benedict Glaw <jbglaw@lug-owl.de>
2393 Iain Sandoe <iain@codesourcery.com>
2396 * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Fix unused variable
2397 warning. Amend comment to reflect current functionality.
2399 2014-01-20 Richard Biener <rguenther@suse.de>
2402 * builtins.c (fold_builtin_strcat): Remove case better handled
2403 by tree-ssa-strlen.c.
2405 2014-01-20 Alan Lawrence <alan.lawrence@arm.com>
2407 * config/aarch64/aarch64.opt
2408 (mcpu, march, mtune): Make case-insensitive.
2410 2014-01-20 Jakub Jelinek <jakub@redhat.com>
2413 * config/i386/i386.c (ix86_avoid_lea_for_addr): Return false
2414 if operands[1] is a REG or ZERO_EXTEND of a REG.
2416 2014-01-19 Jan Hubicka <hubicka@ucw.cz>
2418 * varasm.c (compute_reloc_for_constant): Use targetm.binds_local_p.
2420 2014-01-19 John David Anglin <danglin@gcc.gnu.org>
2422 * config/pa/pa.c (pa_attr_length_millicode_call): Correct length of
2423 long non-pic millicode calls.
2425 2014-01-19 Jan-Benedict Glaw <jbglaw@lug-owl.de>
2427 * config/vax/vax.h (FUNCTION_ARG_REGNO_P): Fix unused variable warning.
2429 2014-01-19 Kito Cheng <kito@0xlab.org>
2431 * builtins.c (expand_movstr): Check movstr expand done or fail.
2433 2014-01-18 Uros Bizjak <ubizjak@gmail.com>
2434 H.J. Lu <hongjiu.lu@intel.com>
2437 * config/i386/i386.md (*lea<mode>): Zero-extend return register
2438 to DImode for zero-extended addresses.
2440 2014-01-19 Jakub Jelinek <jakub@redhat.com>
2442 PR rtl-optimization/57763
2443 * bb-reorder.c (fix_crossing_unconditional_branches): Set JUMP_LABEL
2444 on the new indirect jump_insn and increment LABEL_NUSES (label).
2446 2014-01-18 H.J. Lu <hongjiu.lu@intel.com>
2450 * config.gcc (x86_archs): New variable.
2451 (x86_64_archs): Likewise.
2452 (x86_cpus): Likewise.
2453 Use $x86_archs, $x86_64_archs and $x86_cpus to check valid
2454 --with-arch/--with-cpu= options.
2455 Support --with-arch=/--with-cpu={nehalem,westmere,
2456 sandybridge,ivybridge,haswell,broadwell,bonnell,silvermont}.
2458 2014-01-18 Uros Bizjak <ubizjak@gmail.com>
2460 * config/i386/i386.c (ix86_adjust_cost): Reorder PROCESSOR_K8
2461 and PROCESSOR_ATHLON to simplify code. Move "memory" calculation.
2463 2014-01-18 Uros Bizjak <ubizjak@gmail.com>
2465 * config/i386/i386.md (*swap<mode>): Rename from swap<mode>.
2467 2014-01-18 Jakub Jelinek <jakub@redhat.com>
2470 * config/i386/i386-c.c (ix86_pragma_target_parse): Temporarily
2471 clear cpp_get_options (parse_in)->warn_unused_macros for
2472 ix86_target_macros_internal with cpp_define.
2474 2014-01-18 Richard Sandiford <rdsandiford@googlemail.com>
2476 * jump.c (delete_related_insns): Keep (use (insn))s.
2477 * reorg.c (redundant_insn): Check for barriers too.
2479 2014-01-17 H.J. Lu <hongjiu.lu@intel.com>
2481 * config/i386/i386.c (ix86_split_lea_for_addr): Fix a comment typo.
2483 2014-01-17 John David Anglin <danglin@gcc.gnu.org>
2485 * config/pa/pa.c (pa_attr_length_indirect_call): Don't output a short
2486 call to $$dyncall when TARGET_LONG_CALLS is true.
2488 2014-01-17 Jeff Law <law@redhat.com>
2490 * ree.c (combine_set_extension): Temporarily disable test for
2491 changing number of hard registers.
2493 2014-01-17 Jan Hubicka <hubicka@ucw.cz>
2496 * ipa-inline-analysis.c (inline_free_summary):
2497 Do not free summary of aliases.
2499 2014-01-17 Jakub Jelinek <jakub@redhat.com>
2502 * gimplify.c (gimplify_expr): Use create_tmp_var
2503 instead of create_tmp_var_raw. If cond doesn't have
2504 integral type, don't add the IFN_ANNOTATE builtin at all.
2506 2014-01-17 Martin Jambor <mjambor@suse.cz>
2509 * ipa-cp.c (prev_edge_clone): New variable.
2510 (grow_next_edge_clone_vector): Renamed to grow_edge_clone_vectors.
2511 Also resize prev_edge_clone vector.
2512 (ipcp_edge_duplication_hook): Also update prev_edge_clone.
2513 (ipcp_edge_removal_hook): New function.
2514 (ipcp_driver): Register ipcp_edge_removal_hook.
2516 2014-01-17 Andrew Pinski <apinski@cavium.com>
2517 Steve Ellcey <sellcey@mips.com>
2520 * config/mips/mips.c (mips_print_operand): Check operand mode instead
2523 2014-01-17 Jeff Law <law@redhat.com>
2526 * passes.def: Reorder pass_copy_prop, pass_unrolli, pass_ccp sequence
2527 so that pass_ccp runs first.
2529 2014-01-17 H.J. Lu <hongjiu.lu@intel.com>
2531 * config/i386/i386.c (ix86_lea_outperforms): Use TARGET_XXX.
2532 (ix86_adjust_cost): Use !TARGET_XXX.
2533 (do_reorder_for_imul): Likewise.
2534 (swap_top_of_ready_list): Likewise.
2535 (ix86_sched_reorder): Likewise.
2537 2014-01-17 H.J. Lu <hongjiu.lu@intel.com>
2539 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2540 PROCESSOR_INTEL. Treat like PROCESSOR_GENERIC.
2541 * config/i386/i386.c (intel_memcpy): New. Duplicate slm_memcpy.
2542 (intel_memset): New. Duplicate slm_memset.
2543 (intel_cost): New. Duplicate slm_cost.
2544 (m_INTEL): New macro.
2545 (processor_target_table): Add "intel".
2546 (ix86_option_override_internal): Replace PROCESSOR_SILVERMONT
2547 with PROCESSOR_INTEL for "intel".
2548 (ix86_lea_outperforms): Support PROCESSOR_INTEL. Duplicate
2549 PROCESSOR_SILVERMONT.
2550 (ix86_issue_rate): Likewise.
2551 (ix86_adjust_cost): Likewise.
2552 (ia32_multipass_dfa_lookahead): Likewise.
2553 (swap_top_of_ready_list): Likewise.
2554 (ix86_sched_reorder): Likewise.
2555 (ix86_avoid_lea_for_addr): Check TARGET_AVOID_LEA_FOR_ADDR
2556 instead of TARGET_OPT_AGU.
2557 * config/i386/i386.h (TARGET_INTEL): New.
2558 (TARGET_AVOID_LEA_FOR_ADDR): Likewise.
2559 (processor_type): Add PROCESSOR_INTEL.
2560 * config/i386/x86-tune.def: Support m_INTEL. Duplicate m_SILVERMONT.
2561 Add X86_TUNE_AVOID_LEA_FOR_ADDR.
2563 2014-01-17 Marek Polacek <polacek@redhat.com>
2566 * gimple-fold.c (fold_array_ctor_reference): Don't fold if element
2569 2014-01-17 Richard Biener <rguenther@suse.de>
2571 PR tree-optimization/46590
2572 * opts.c (default_options_table): Add entries for
2573 OPT_fbranch_count_reg, OPT_fmove_loop_invariants and OPT_ftree_pta,
2574 all enabled at -O1 but not for -Og.
2575 * common.opt (fbranch-count-reg): Remove Init(1).
2576 (fmove-loop-invariants): Likewise.
2577 (ftree-pta): Likewise.
2579 2014-01-17 Jakub Jelinek <jakub@redhat.com>
2581 * config/i386/i386.c (ix86_data_alignment): For compatibility with
2582 (incorrect) GCC 4.8 and earlier alignment assumptions ensure we align
2583 decls to at least the GCC 4.8 used alignments.
2586 * tree-nested.c (convert_nonlocal_reference_stmt,
2587 convert_local_reference_stmt): For NAMELIST_DECLs in gimple_bind_vars
2588 of GIMPLE_BIND stmts, adjust associated decls.
2590 2014-01-17 Richard Biener <rguenther@suse.de>
2592 PR tree-optimization/46590
2593 * vec.h (vec<>::bseach): New member function implementing
2594 binary search according to C89 bsearch.
2595 (vec<>::qsort): Avoid calling ::qsort for vectors with sizes 0 or 1.
2596 * tree-ssa-loop-im.c (struct mem_ref): Make stored member a
2597 bitmap pointer again. Make accesses_in_loop a flat array.
2598 (mem_ref_obstack): New global.
2599 (outermost_indep_loop): Adjust for mem_ref->stored changes.
2600 (mark_ref_stored): Likewise.
2601 (ref_indep_loop_p_2): Likewise.
2602 (set_ref_stored_in_loop): New helper function.
2603 (mem_ref_alloc): Allocate mem_refs on the mem_ref_obstack obstack.
2604 (memref_free): Adjust.
2605 (record_mem_ref_loc): Simplify.
2606 (gather_mem_refs_stmt): Adjust.
2607 (sort_locs_in_loop_postorder_cmp): New function.
2608 (analyze_memory_references): Sort accesses_in_loop after
2609 loop postorder number.
2610 (find_ref_loc_in_loop_cmp): New function.
2611 (for_all_locs_in_loop): Find relevant cluster of locs in
2612 accesses_in_loop and iterate without recursion.
2613 (execute_sm): Avoid uninit warning.
2614 (struct ref_always_accessed): Simplify.
2615 (ref_always_accessed::operator ()): Likewise.
2616 (ref_always_accessed_p): Likewise.
2617 (tree_ssa_lim_initialize): Initialize mem_ref_obstack, compute
2618 loop postorder numbers here.
2619 (tree_ssa_lim_finalize): Free mem_ref_obstack and loop postorder
2622 2014-01-17 Jan Hubicka <hubicka@ucw.cz>
2625 * passes.c (rest_of_decl_compilation): Don't call varpool_finalize_decl
2626 on decls for which assemble_alias has been called.
2628 2014-01-17 Nick Clifton <nickc@redhat.com>
2630 * config/msp430/msp430.opt: (mcpu): New option.
2631 * config/msp430/msp430.c (msp430_mcu_name): Use target_mcu.
2632 (msp430_option_override): Parse target_cpu. If the MCU name
2633 matches a generic string, clear target_mcu.
2634 (msp430_attr): Allow numeric interrupt values up to 63.
2635 (msp430_expand_epilogue): No longer invert operand 1 of gen_popm.
2636 * config/msp430/msp430.h (ASM_SPEC): Convert -mcpu into a -mmcu
2638 * config/msp430/t-msp430: (MULTILIB_MATCHES): Remove mcu matches.
2640 * config/msp430/msp430.md (popm): Use %J rather than %I.
2641 (addsi3): Use msp430_nonimmediate_operand for operand 2.
2642 (addhi_cy_i): Use immediate_operand for operand 2.
2643 * doc/invoke.texi: Document -mcpu option.
2645 2014-01-17 Richard Biener <rguenther@suse.de>
2647 PR rtl-optimization/38518
2648 * df.h (df_analyze_loop): Declare.
2649 * df-core.c: Include cfgloop.h.
2650 (df_analyze_1): Split out main part of df_analyze.
2651 (df_analyze): Adjust.
2652 (loop_inverted_post_order_compute): New function.
2653 (loop_post_order_compute): Likewise.
2654 (df_analyze_loop): New function avoiding whole-function
2656 * loop-invariant.c (find_defs): Use df_analyze_loop.
2657 (find_invariants): Adjust.
2658 * loop-iv.c (iv_analysis_loop_init): Use df_analyze_loop.
2660 2014-01-17 Zhenqiang Chen <zhenqiang.chen@arm.com>
2662 * config/arm/arm.c (arm_v7m_tune): Set max_insns_skipped to 2.
2663 (thumb2_final_prescan_insn): Set max to MAX_INSN_PER_IT_BLOCK.
2665 2014-01-16 Ilya Enkovich <ilya.enkovich@intel.com>
2667 * ipa-ref.c (ipa_remove_stmt_references): Fix references
2668 traversal when removing references.
2670 2014-01-16 Jan Hubicka <hubicka@ucw.cz>
2673 * tree.c (get_binfo_at_offset): Look harder for virtual bases.
2675 2014-01-16 Bernd Schmidt <bernds@codesourcery.com>
2678 * reload.c (find_reloads_address_1): Do not use RELOAD_OTHER when
2679 pushing a reload for an autoinc when we had previously reloaded an
2680 inner part of the address.
2682 2014-01-16 Jakub Jelinek <jakub@redhat.com>
2684 * tree-vectorizer.h (struct _loop_vec_info): Add no_data_dependencies
2686 (LOOP_VINFO_NO_DATA_DEPENDENCIES): Define.
2687 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Clear it
2688 when not giving up or versioning for alias only because of
2690 (vect_analyze_data_ref_dependences): Set to true.
2691 * tree-vect-stmts.c (hoist_defs_of_uses): Return false if def_stmt
2693 (vectorizable_load): Use LOOP_VINFO_NO_DATA_DEPENDENCIES instead of
2694 LOOP_REQUIRES_VERSIONING_FOR_ALIAS, add && !nested_in_vect_loop
2698 * expr.c (expand_expr_real_1): Handle init == NULL_TREE.
2701 * config/i386/i386.c (ix86_expand_builtin): If target doesn't satisfy
2702 operand 0 predicate for gathers, use a new pseudo as subtarget.
2704 2014-01-16 Vladimir Makarov <vmakarov@redhat.com>
2707 * lra-constraints.c (process_alt_operands): Add printing debug info.
2708 Check absence of input/output reloads for matched operands too.
2710 2014-01-16 Vladimir Makarov <vmakarov@redhat.com>
2712 PR rtl-optimization/59835
2713 * ira.c (ira_init_register_move_cost): Increase cost for
2716 2014-01-16 Alan Lawrence <alan.lawrence@arm.com>
2718 * config/arm/arm.opt (mcpu, march, mtune): Make case-insensitive.
2720 2014-01-16 Richard Earnshaw <rearnsha@arm.com>
2723 * aarch64.c (aarch64_split_128bit_move): Don't lookup REGNO on
2724 non-register objects. Use gen_(high/low)part more consistently.
2727 2014-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
2730 * config/rs6000/rs6000.md (reload_vsx_from_gprsf): Add little
2731 endian support, remove tests for WORDS_BIG_ENDIAN.
2732 (p8_mfvsrd_3_<mode>): Likewise.
2733 (reload_gpr_from_vsx<mode>): Likewise.
2734 (reload_gpr_from_vsxsf): Likewise.
2735 (p8_mfvsrd_4_disf): Likewise.
2737 2014-01-16 Richard Biener <rguenther@suse.de>
2739 PR rtl-optimization/46590
2740 * lcm.c (compute_antinout_edge): Use postorder iteration.
2741 (compute_laterin): Use inverted postorder iteration.
2743 2014-01-16 Nick Clifton <nickc@redhat.com>
2746 * varasm.c (output_constant): Return the number of bytes actually
2748 (output_constructor_array_range): Update the field size with the
2749 number of bytes emitted by output_constant.
2750 (output_constructor_regular_field): Likewise. Also do not
2751 complain if the total number of bytes emitted is now greater
2752 than the expected fieldpos.
2753 * output.h (output_constant): Update prototype and descriptive comment.
2755 2014-01-16 Marek Polacek <polacek@redhat.com>
2758 * cgraph.c (gimple_check_call_args): Don't use DECL_ARG_TYPE if
2759 it is error_mark_node.
2761 2014-01-15 Uros Bizjak <ubizjak@gmail.com>
2763 * config/i386/i386.c (ix86_hard_regno_mode_ok): Use
2764 VALID_AVX256_REG_OR_OI_MODE.
2766 2014-01-15 Pat Haugen <pthaugen@us.ibm.com>
2768 * config/rs6000/rs6000.c (rs6000_output_function_prologue): Check if
2769 current procedure should be profiled.
2771 2014-01-15 Andrew Pinski <apinski@cavium.com>
2773 * config/aarch64/aarch64.c (aarch64_register_move_cost): Correct cost
2774 of moving from/to the STACK_REG register class.
2776 2014-01-15 Richard Henderson <rth@redhat.com>
2779 * reginfo.c (global_regs_decl): Globalize.
2780 * rtl.h (global_regs_decl): Declare.
2781 * ira.c (do_reload): Diagnose frame_pointer_needed and it
2782 reserved via global_regs.
2784 2014-01-15 Teresa Johnson <tejohnson@google.com>
2786 * tree-ssa-sccvn.c (visit_reference_op_call): Handle NULL vdef.
2788 2014-01-15 Bill Schmidt <wschmidt@vnet.linux.ibm.com>
2790 * config/rs6000/altivec.md (mulv8hi3): Explicitly generate vmulesh
2791 and vmulosh rather than call gen_vec_widen_smult_*.
2792 (vec_widen_umult_even_v16qi): Test VECTOR_ELT_ORDER_BIG rather
2793 than BYTES_BIG_ENDIAN to determine use of even or odd instruction.
2794 (vec_widen_smult_even_v16qi): Likewise.
2795 (vec_widen_umult_even_v8hi): Likewise.
2796 (vec_widen_smult_even_v8hi): Likewise.
2797 (vec_widen_umult_odd_v16qi): Likewise.
2798 (vec_widen_smult_odd_v16qi): Likewise.
2799 (vec_widen_umult_odd_v8hi): Likewise.
2800 (vec_widen_smult_odd_v8hi): Likewise.
2801 (vec_widen_umult_hi_v16qi): Explicitly generate vmuleub and
2802 vmuloub rather than call gen_vec_widen_umult_*.
2803 (vec_widen_umult_lo_v16qi): Likewise.
2804 (vec_widen_smult_hi_v16qi): Explicitly generate vmulesb and
2805 vmulosb rather than call gen_vec_widen_smult_*.
2806 (vec_widen_smult_lo_v16qi): Likewise.
2807 (vec_widen_umult_hi_v8hi): Explicitly generate vmuleuh and vmulouh
2808 rather than call gen_vec_widen_umult_*.
2809 (vec_widen_umult_lo_v8hi): Likewise.
2810 (vec_widen_smult_hi_v8hi): Explicitly gnerate vmulesh and vmulosh
2811 rather than call gen_vec_widen_smult_*.
2812 (vec_widen_smult_lo_v8hi): Likewise.
2814 2014-01-15 Jeff Law <law@redhat.com>
2816 PR tree-optimization/59747
2817 * ree.c (find_and_remove_re): Properly handle case where a second
2818 eliminated extension requires widening a copy created for elimination
2819 of a prior extension.
2820 (combine_set_extension): Ensure that the number of hard regs needed
2821 for a destination register does not change when we widen it.
2823 2014-01-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
2825 * config.gcc (*-*-rtems*): Add t-rtems to tmake_file.
2826 (arm*-*-uclinux*eabi*): Do not override an existing tmake_file.
2827 (arm*-*-eabi* | arm*-*-symbianelf* | arm*-*-rtems*): Likwise.
2828 (arm*-*-rtems*): Use t-rtems from existing tmake_file.
2829 (avr-*-rtems*): Likewise.
2830 (bfin*-rtems*): Likewise.
2831 (moxie-*-rtems*): Likewise.
2832 (h8300-*-rtems*): Likewise.
2833 (i[34567]86-*-rtems*): Likewise.
2834 (lm32-*-rtems*): Likewise.
2835 (m32r-*-rtems*): Likewise.
2836 (m68k-*-rtems*): Likewise.
2837 (microblaze*-*-rtems*): Likewise.
2838 (mips*-*-rtems*): Likewise.
2839 (powerpc-*-rtems*): Likewise.
2840 (sh-*-rtems*): Likewise.
2841 (sparc-*-rtems*): Likewise.
2842 (sparc64-*-rtems*): Likewise.
2843 (v850-*-rtems*): Likewise.
2844 (m32c-*-rtems*): Likewise.
2846 2014-01-15 Vladimir Makarov <vmakarov@redhat.com>
2848 PR rtl-optimization/59511
2849 * ira.c (ira_init_register_move_cost): Use memory costs for some
2850 cases of register move cost calculations.
2851 * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
2852 instead of BB frequency.
2853 * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
2854 * lra-assigns.c (find_hard_regno_for): Ditto.
2856 2014-01-15 Richard Biener <rguenther@suse.de>
2858 PR tree-optimization/59822
2859 * tree-vect-stmts.c (hoist_defs_of_uses): New function.
2860 (vectorizable_load): Use it to hoist defs of uses of invariant
2861 loads out of the loop.
2863 2014-01-15 Matthew Gretton-Dann <matthew.gretton-dann@linaro.org>
2864 Kugan Vivekanandarajah <kuganv@linaro.org>
2867 * config/aarch64/aarch64.c (aarch64_build_constant): Fix incorrect
2870 2014-01-15 Richard Biener <rguenther@suse.de>
2872 PR rtl-optimization/59802
2873 * lcm.c (compute_available): Use inverted postorder to seed
2874 the initial worklist.
2876 2014-01-15 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
2879 * config/s390/s390.c (s390_preferred_reload_class): Don't return
2880 ADDR_REGS for invalid symrefs in non-PIC code.
2882 2014-01-15 Jakub Jelinek <jakub@redhat.com>
2885 * builtins.c (determine_block_size): Initialize *probable_max_size
2886 even if len_rtx is CONST_INT.
2888 2014-01-14 Andrew Pinski <apinski@cavium.com>
2890 * config/aarch64/aarch64-protos.h (tune_params): Add issue_rate.
2891 * config/aarch64/aarch64.c (generic_tunings): Add issue rate of 2.
2892 (cortexa53_tunings): Likewise.
2893 (aarch64_sched_issue_rate): New function.
2894 (TARGET_SCHED_ISSUE_RATE): Define.
2896 2014-01-14 Vladimir Makarov <vmakarov@redhat.com>
2898 * ira-costs.c (find_costs_and_classes): Add missed
2899 ira_init_register_move_cost_if_necessary.
2901 2014-01-14 Vladimir Makarov <vmakarov@redhat.com>
2904 * config/arm/arm.c (arm_coproc_mem_operand): Add lra_in_progress.
2906 2014-01-14 H.J. Lu <hongjiu.lu@intel.com>
2909 * config/i386/i386.c (type_natural_mode): Add a bool parameter
2910 to indicate if type is used for function return value. Warn ABI
2911 change if the vector mode isn't available for function return value.
2912 (ix86_function_arg_advance): Pass false to type_natural_mode.
2913 (ix86_function_arg): Likewise.
2914 (ix86_gimplify_va_arg): Likewise.
2915 (function_arg_32): Don't warn ABI change.
2916 (ix86_function_value): Pass true to type_natural_mode.
2917 (ix86_return_in_memory): Likewise.
2918 (ix86_struct_value_rtx): Removed.
2919 (TARGET_STRUCT_VALUE_RTX): Likewise.
2921 2014-01-14 Richard Sandiford <rsandifo@linux.vnet.ibm.com>
2923 * jump.c (redirect_jump_2): Remove REG_CROSSING_JUMP notes when
2924 converting a conditional jump into a conditional return.
2926 2014-01-14 Richard Biener <rguenther@suse.de>
2928 PR tree-optimization/58921
2929 PR tree-optimization/59006
2930 * tree-vect-loop-manip.c (vect_loop_versioning): Remove code
2931 hoisting invariant stmts.
2932 * tree-vect-stmts.c (vectorizable_load): Insert the splat of
2933 invariant loads on the preheader edge if possible.
2935 2014-01-14 Joey Ye <joey.ye@arm.com>
2937 * doc/plugin.texi (Building GCC plugins): Update to C++.
2939 2014-01-14 Kirill Yukhin <kirill.yukhin@intel.com>
2941 * config/i386/avx512erintrin.h (_mm_rcp28_round_sd): New.
2942 (_mm_rcp28_round_ss): Ditto.
2943 (_mm_rsqrt28_round_sd): Ditto.
2944 (_mm_rsqrt28_round_ss): Ditto.
2945 (_mm_rcp28_sd): Ditto.
2946 (_mm_rcp28_ss): Ditto.
2947 (_mm_rsqrt28_sd): Ditto.
2948 (_mm_rsqrt28_ss): Ditto.
2949 * config/i386/avx512fintrin.h (_mm512_stream_load_si512): Ditto.
2950 * config/i386/i386-builtin-types.def (V8DI_FTYPE_PV8DI): Ditto.
2951 * config/i386/i386.c (IX86_BUILTIN_MOVNTDQA512): Ditto.
2952 (IX86_BUILTIN_RCP28SD): Ditto.
2953 (IX86_BUILTIN_RCP28SS): Ditto.
2954 (IX86_BUILTIN_RSQRT28SD): Ditto.
2955 (IX86_BUILTIN_RSQRT28SS): Ditto.
2956 (bdesc_special_args): Define __builtin_ia32_movntdqa512,
2957 __builtin_ia32_rcp28sd_round, __builtin_ia32_rcp28ss_round,
2958 __builtin_ia32_rsqrt28sd_round, __builtin_ia32_rsqrt28ss_round.
2959 (ix86_expand_special_args_builtin): Expand new FTYPE.
2960 * config/i386/sse.md (define_mode_attr "sse4_1_avx2"): Expand to V8DI.
2961 (srcp14<mode>): Make insn unary.
2962 (avx512f_vmscalef<mode><round_name>): Use substed predicate.
2963 (avx512f_sgetexp<mode><round_saeonly_name>): Ditto.
2964 (avx512f_rndscale<mode><round_saeonly_name>): Ditto.
2965 (<sse4_1_avx2>_movntdqa): Extend to 512 bits.
2966 (avx512er_exp2<mode><mask_name><round_saeonly_name>):
2967 Fix rounding: make it SAE only.
2968 (<mask_codefor>avx512er_rcp28<mode><mask_name><round_saeonly_name>):
2970 (<mask_codefor>avx512er_rsqrt28<mode><mask_name><round_saeonly_name>):
2972 (avx512er_vmrcp28<mode><round_saeonly_name>): Ditto.
2973 (avx512er_vmrsqrt28<mode><round_saeonly_name>): Ditto.
2974 (avx512f_getmant<mode><mask_name><round_saeonly_name>): Ditto.
2975 * config/i386/subst.md (round_saeonly_mask_scalar_operand3): Remove.
2976 (round_saeonly_mask_scalar_operand4): Ditto.
2977 (round_saeonly_mask_scalar_op3): Ditto.
2978 (round_saeonly_mask_scalar_op4): Ditto.
2980 2014-01-13 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2982 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2983 Implement -maltivec=be for vec_insert and vec_extract.
2985 2014-01-10 DJ Delorie <dj@redhat.com>
2987 * config/msp430/msp430.md (call_internal): Don't allow memory
2988 references with SP as the base register.
2989 (call_value_internal): Likewise.
2990 * config/msp430/constraints.md (Yc): New. For memory references
2991 that don't use SP as a base register.
2993 * config/msp430/msp430.c (msp430_print_operand): Add 'J' to mean
2994 "an integer without a # prefix"
2995 * config/msp430/msp430.md (epilogue_helper): Use it.
2997 2014-01-13 Jakub Jelinek <jakub@redhat.com>
3000 * config/i386/i386.c (ix86_vectorize_builtin_gather): Uncomment
3001 AVX512F gather builtins.
3002 * tree-vect-stmts.c (vectorizable_mask_load_store): For now punt
3003 on gather decls with INTEGER_TYPE masktype.
3004 (vectorizable_load): For INTEGER_TYPE masktype, put the INTEGER_CST
3005 directly into the builtin rather than hoisting it before loop.
3007 PR tree-optimization/59387
3008 * tree-scalar-evolution.c: Include gimple-fold.h and gimplify-me.h.
3009 (scev_const_prop): If folded_casts and type has undefined overflow,
3010 use force_gimple_operand instead of force_gimple_operand_gsi and
3011 for each added stmt if it is assign with
3012 arith_code_with_undefined_signed_overflow, call
3013 rewrite_to_defined_overflow.
3014 * tree-ssa-loop-im.c: Don't include gimplify-me.h, include
3015 gimple-fold.h instead.
3016 (arith_code_with_undefined_signed_overflow,
3017 rewrite_to_defined_overflow): Moved to ...
3018 * gimple-fold.c (arith_code_with_undefined_signed_overflow,
3019 rewrite_to_defined_overflow): ... here. No longer static.
3020 Include gimplify-me.h.
3021 * gimple-fold.h (arith_code_with_undefined_signed_overflow,
3022 rewrite_to_defined_overflow): New prototypes.
3024 2014-01-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3026 * config/arm/arm.h (MAX_CONDITIONAL_EXECUTE): Fix typo in description.
3028 2014-01-13 Eric Botcazou <ebotcazou@adacore.com>
3030 * builtins.c (get_object_alignment_2): Minor tweak.
3031 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): Rewrite.
3033 2014-01-13 Christian Bruel <christian.bruel@st.com>
3035 * config/sh/sh-mem.cc (sh_expand_cmpnstr): Unroll small sizes and
3036 optimized non constant lengths.
3038 2014-01-13 Jakub Jelinek <jakub@redhat.com>
3041 * omp-low.c (expand_omp_atomic_pipeline): Expand the initial
3042 load as __atomic_load_N if possible.
3044 2014-01-11 David Edelsohn <dje.gcc@gmail.com>
3046 * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): Remove
3048 (rs6000_expand_builtin): Adjust call.
3050 2014-01-11 David Edelsohn <dje.gcc@gmail.com>
3053 * config/rs6000/rs6000.h (SWITCHABLE_TARGET): Define.
3054 * config/rs6000/rs6000.c: Include target-globals.h.
3055 (rs6000_set_current_function): Instead of doing target_reinit
3056 unconditionally, use save_target_globals_default_opts and
3057 restore_target_globals.
3059 * config/rs6000/rs6000-builtin.def (mffs, mtfsf): Add builtins for
3061 * config/rs6000/rs6000.c (rs6000_expand_mtfsf_builtin): New.
3062 (rs6000_expand_builtin): Handle mffs and mtfsf.
3063 (rs6000_init_builtins): Define mffs and mtfsf.
3064 * config/rs6000/rs6000.md (UNSPECV_MFFS, UNSPECV_MTFSF): New constants.
3065 (rs6000_mffs): New pattern.
3066 (rs6000_mtfsf): New pattern.
3068 2014-01-11 Bin Cheng <bin.cheng@arm.com>
3070 * tree-ssa-loop-ivopts.c (iv_ca_narrow): New parameter.
3071 Start narrowing with START. Apply candidate-use pair
3072 and check overall cost in narrowing.
3073 (iv_ca_prune): Pass new argument.
3075 2014-01-10 Jeff Law <law@redhat.com>
3078 * ree.c (combine_reaching_defs): Ensure the defining statement
3079 occurs before the extension when optimizing extensions with
3080 different source and destination hard registers.
3082 2014-01-10 Jan Hubicka <hubicka@ucw.cz>
3085 * ipa-devirt.c (build_type_inheritance_graph): Also add types of
3086 vtables into the type inheritance graph.
3088 2014-01-10 Jakub Jelinek <jakub@redhat.com>
3090 PR rtl-optimization/59754
3091 * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
3092 modes in the REGNO != REGNO case.
3094 2014-01-10 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3096 * config/rs6000/rs6000-builtin.def: Fix pasto for VPKSDUS.
3098 2014-01-10 Jakub Jelinek <jakub@redhat.com>
3100 PR tree-optimization/59745
3101 * tree-predcom.c (tree_predictive_commoning_loop): Call
3102 free_affine_expand_cache if giving up because components is NULL.
3104 * target-globals.c (save_target_globals): Allocate < 4KB structs using
3105 GC in payload of target_globals struct instead of allocating them on
3106 the heap and the larger structs separately using GC.
3107 * target-globals.h (struct target_globals): Make regs, hard_regs,
3108 reload, expmed, ira, ira_int and lra_fields GTY((atomic)) instead
3109 of GTY((skip)) and change type to void *.
3110 (reset_target_globals): Cast loads from those fields to corresponding
3113 2014-01-10 Steve Ellcey <sellcey@mips.com>
3116 * Makefile.in (PLUGIN_HEADERS): Add gimplify.h, gimple-iterator.h,
3117 gimple-ssa.h, fold-const.h, tree-cfg.h, tree-into-ssa.h,
3118 tree-ssanames.h, print-tree.h, varasm.h, and context.h.
3120 2014-01-10 Richard Earnshaw <rearnsha@arm.com>
3123 * aarch64-modes.def (CC_Zmode): New flags mode.
3124 * aarch64.c (aarch64_select_cc_mode): Only allow NEG when the condition
3125 represents an equality.
3126 (aarch64_get_condition_code): Handle CC_Zmode.
3127 * aarch64.md (compare_neg<mode>): Restrict to equality operations.
3129 2014-01-10 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
3131 * config/s390/s390.c (s390_expand_tbegin): Remove jump over CC
3132 extraction in good case.
3134 2014-01-10 Richard Biener <rguenther@suse.de>
3136 PR tree-optimization/59374
3137 * tree-vect-slp.c (vect_slp_analyze_bb_1): Move dependence
3138 checking after SLP discovery. Mark stmts not participating
3139 in any SLP instance properly.
3141 2014-01-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3143 * config/arm/arm.c (arm_new_rtx_costs): Use destination mode
3144 when handling a SET rtx.
3146 2014-01-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3148 * config/arm/arm-cores.def (cortex-a53): Specify FL_CRC32.
3149 (cortex-a57): Likewise.
3150 (cortex-a57.cortex-a53): Likewise. Remove redundant flags.
3152 2014-01-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3154 * config/arm/arm.c (arm_init_iwmmxt_builtins): Skip
3155 non-iwmmxt builtins.
3157 2014-01-10 Jan Hubicka <hubicka@ucw.cz>
3161 * ipa-devirt.c record_target_from_binfo): Take as argument
3162 stack of binfos and lookup matching one for virtual inheritance.
3163 (possible_polymorphic_call_targets_1): Update.
3165 2014-01-10 Huacai Chen <chenhc@lemote.com>
3167 * config/mips/driver-native.c (host_detect_local_cpu): Handle new
3168 kernel strings for Loongson-2E/2F/3A.
3170 2014-01-10 Jakub Jelinek <jakub@redhat.com>
3173 * tree-vect-data-refs.c (vect_analyze_data_refs): Check
3174 is_gimple_call before calling gimple_call_internal_p.
3176 2014-01-09 Steve Ellcey <sellcey@mips.com>
3178 * Makefile.in (TREE_FLOW_H): Remove.
3179 (TREE_SSA_H): Add file names from tree-flow.h.
3180 * doc/tree-ssa.texi (Annotations): Remove reference to tree-flow.h
3181 * tree.h: Remove tree-flow.h reference.
3182 * hash-table.h: Remove tree-flow.h reference.
3183 * tree-ssa-loop-niter.c (dump_affine_iv): Replace tree-flow.h
3184 reference with tree-ssa-loop.h.
3186 2014-01-09 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3188 * doc/invoke.texi: Add -maltivec={be,le} options, and document
3189 default element-order behavior for -maltivec.
3190 * config/rs6000/rs6000.opt: Add -maltivec={be,le} options.
3191 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ensure
3192 that -maltivec={le,be} implies -maltivec; disallow -maltivec=le
3193 when targeting big endian, at least for now.
3194 * config/rs6000/rs6000.h: Add #define of VECTOR_ELT_ORDER_BIG.
3196 2014-01-09 Jakub Jelinek <jakub@redhat.com>
3199 * cfgexpand.c (expand_one_var): For SSA_NAMEs, if the underlying
3200 var satisfies use_register_for_decl, just take into account type
3201 alignment, rather than decl alignment.
3203 PR tree-optimization/59622
3204 * gimple-fold.c (gimple_fold_call): Fix a typo in message. For
3205 __builtin_unreachable replace the OBJ_TYPE_REF call with a call to
3206 __builtin_unreachable and add if needed a setter of the lhs SSA_NAME.
3207 Don't devirtualize for inplace at all. For targets.length () == 1,
3208 if the call is noreturn and cfun isn't in SSA form yet, clear lhs.
3210 2014-01-09 H.J. Lu <hongjiu.lu@intel.com>
3212 * config/i386/i386.md (cpu): Remove the unused btver1.
3214 2014-01-09 H.J. Lu <hongjiu.lu@intel.com>
3216 * gdbasan.in: Put a breakpoint on __sanitizer::Report.
3218 2014-01-09 Jakub Jelinek <jakub@redhat.com>
3221 * tree-core.h (struct target_globals): New forward declaration.
3222 (struct tree_target_option): Add globals field.
3223 * tree.h (TREE_TARGET_GLOBALS): Define.
3224 (prepare_target_option_nodes_for_pch): New prototype.
3225 * target-globals.h (struct target_globals): Define even if
3227 * tree.c (prepare_target_option_node_for_pch,
3228 prepare_target_option_nodes_for_pch): New functions.
3229 * config/i386/i386.h (SWITCHABLE_TARGET): Define.
3230 * config/i386/i386.c: Include target-globals.h.
3231 (ix86_set_current_function): Instead of doing target_reinit
3232 unconditionally, use save_target_globals_default_opts and
3233 restore_target_globals.
3235 2014-01-09 Richard Biener <rguenther@suse.de>
3237 PR tree-optimization/59715
3238 * tree-cfg.h (split_critical_edges): Declare.
3239 * tree-cfg.c (split_critical_edges): Export.
3240 * tree-ssa-sink.c (execute_sink_code): Split critical edges.
3242 2014-01-09 Max Ostapenko <m.ostapenko@partner.samsung.com>
3244 * cfgexpand.c (expand_stack_vars): Optionally disable
3245 asan stack protection.
3246 (expand_used_vars): Likewise.
3247 (partition_stack_vars): Likewise.
3248 * asan.c (asan_emit_stack_protection): Optionally disable
3249 after return stack usage.
3250 (instrument_derefs): Optionally disable memory access instrumentation.
3251 (instrument_builtin_call): Likewise.
3252 (instrument_strlen_call): Likewise.
3253 (asan_protect_global): Optionally disable global variables protection.
3254 * doc/invoke.texi: Added doc for new options.
3255 * params.def: Added new options.
3256 * params.h: Likewise.
3258 2014-01-09 Jakub Jelinek <jakub@redhat.com>
3260 PR rtl-optimization/59724
3261 * ifcvt.c (cond_exec_process_if_block): Don't call
3262 flow_find_head_matching_sequence with 0 longest_match.
3263 * cfgcleanup.c (flow_find_head_matching_sequence): Count even
3264 non-active insns if !stop_after.
3265 (try_head_merge_bb): Revert 2014-01-07 changes.
3267 2014-01-08 Jeff Law <law@redhat.com>
3269 * ree.c (get_sub_rtx): New function, extracted from...
3270 (merge_def_and_ext): Here.
3271 (combine_reaching_defs): Use get_sub_rtx.
3273 2014-01-08 Eric Botcazou <ebotcazou@adacore.com>
3275 * cgraph.h (varpool_variable_node): Do not choke on null node.
3277 2014-01-08 Catherine Moore <clm@codesourcery.com>
3279 * config/mips/mips.md (simple_return): Attempt to use JRC
3281 * config/mips/mips.h (MIPS_CALL): Attempt to use JALS for microMIPS.
3283 2014-01-08 Richard Sandiford <rdsandiford@googlemail.com>
3285 PR rtl-optimization/59137
3286 * reorg.c (steal_delay_list_from_target): Call update_block for
3288 (steal_delay_list_from_fallthrough, relax_delay_slots): Likewise.
3290 2014-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3292 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
3293 two duplicate entries.
3295 2014-01-08 Richard Sandiford <rdsandiford@googlemail.com>
3298 2012-10-07 Richard Sandiford <rdsandiford@googlemail.com>
3300 * config/mips/mips.c (mips_truncated_op_cost): New function.
3301 (mips_rtx_costs): Adjust test for BADDU.
3302 * config/mips/mips.md (*baddu_di<mode>): Push truncates to operands.
3304 2012-10-02 Richard Sandiford <rdsandiford@googlemail.com>
3306 * config/mips/mips.md (*baddu_si_eb, *baddu_si_el): Merge into...
3307 (*baddu_si): ...this new pattern.
3309 2014-01-08 Jakub Jelinek <jakub@redhat.com>
3312 * ipa-prop.c (ipa_analyze_params_uses): Ignore uses in debug stmts.
3314 2014-01-08 Bernd Edlinger <bernd.edlinger@hotmail.de>
3317 * expr.h (expand_expr_real, expand_expr_real_1): Add new parameter
3319 (expand_expr, expand_normal): Adjust.
3320 * expr.c (expand_expr_real, expand_expr_real_1): Add new parameter
3321 inner_reference_p. Use inner_reference_p to expand inner references.
3322 (store_expr): Adjust.
3323 * cfgexpand.c (expand_call_stmt): Adjust.
3325 2014-01-08 Rong Xu <xur@google.com>
3327 * gcov-io.c (gcov_var): Move from gcov-io.h.
3328 (gcov_position): Ditto.
3329 (gcov_is_error): Ditto.
3330 (gcov_rewrite): Ditto.
3331 * gcov-io.h: Refactor. Move gcov_var to gcov-io.h, and libgcov
3332 only part to libgcc/libgcov.h.
3334 2014-01-08 Marek Polacek <polacek@redhat.com>
3337 * omp-low.c (simd_clone_adjust): Don't crash if def is NULL.
3339 2014-01-08 Marek Polacek <polacek@redhat.com>
3342 * ubsan.c (ubsan_type_descriptor): Call strip_array_types on type2.
3344 2014-01-08 Jakub Jelinek <jakub@redhat.com>
3346 PR rtl-optimization/59649
3347 * stor-layout.c (get_mode_bounds): For BImode return
3348 0 and STORE_FLAG_VALUE.
3350 2014-01-08 Richard Biener <rguenther@suse.de>
3353 * gimple.h (is_gimple_builtin_call): Remove.
3354 (gimple_builtin_call_types_compatible_p): New.
3355 (gimple_call_builtin_p): New overload.
3356 * gimple.c (is_gimple_builtin_call): Remove.
3357 (validate_call): Rename to ...
3358 (gimple_builtin_call_types_compatible_p): ... this and export. Also
3360 (validate_type): New static function.
3361 (gimple_call_builtin_p): New overload and adjust.
3362 * gimple-fold.c (gimple_fold_builtin): Fold the return value.
3363 (gimple_fold_call): Likewise. Use gimple_call_builtin_p.
3364 (gimple_fold_stmt_to_constant_1): Likewise.
3365 * tsan.c (instrument_gimple): Use gimple_call_builtin_p.
3367 2014-01-08 Richard Biener <rguenther@suse.de>
3370 * gimplify.c (gimplify_expr): Gimplify register-register type
3371 VIEW_CONVERT_EXPRs to separate stmts.
3373 2014-01-07 Jeff Law <law@redhat.com>
3376 * ree.c (combine_set_extension): Handle case where source
3377 and destination registers in an extension insn are different.
3378 (combine_reaching_defs): Allow source and destination registers
3379 in extension to be different under limited circumstances.
3380 (add_removable_extension): Remove restriction that the
3381 source and destination registers in the extension are the same.
3382 (find_and_remove_re): Emit a copy from the extension's
3383 destination to its source after the defining insn if
3384 the source and destination registers are different.
3387 * ifcvt.c (merge_if_block): If we are merging a block with more than
3388 one successor with a block with no successors, remove any BARRIER
3389 after the second block.
3391 2014-01-07 Dan Xio Qiang <ziyan01@163.com>
3393 * hw-doloop.c (reorg_loops): Release the bitmap obstack.
3395 2014-01-07 John David Anglin <danglin@gcc.gnu.org>
3398 * config/pa/pa.c (pa_legitimate_address_p): Return false before reload
3399 for 14-bit register offsets when INT14_OK_STRICT is false.
3401 2014-01-07 Roland Stigge <stigge@antcom.de>
3402 Michael Meissner <meissner@linux.vnet.ibm.com>
3405 * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
3406 Only check TFmode for SPE constants. Don't check TImode or TDmode.
3408 2014-01-07 James Greenhalgh <james.greenhalgh@arm.com>
3410 * config/aarch64/aarch64-elf.h (ASM_SPEC): Remove identity spec for
3413 2014-01-07 Yufeng Zhang <yufeng.zhang@arm.com>
3415 * config/arm/arm.c (arm_expand_neon_args): Call expand_expr
3416 with EXPAND_MEMORY for NEON_ARG_MEMORY; check if the returned
3417 rtx is const0_rtx or not.
3419 2014-01-07 Richard Sandiford <rdsandiford@googlemail.com>
3422 * target-globals.c (save_target_globals): Remove this_fn_optab
3424 * toplev.c: Include optabs.h.
3425 (target_reinit): Temporarily restore the global options if another
3426 set of options are in force.
3428 2014-01-07 Jakub Jelinek <jakub@redhat.com>
3430 PR rtl-optimization/58668
3431 * cfgcleanup.c (flow_find_cross_jump): Don't count
3432 any jumps if dir_p is NULL. Remove p1 variable, use active_insn_p
3433 to determine what is counted.
3434 (flow_find_head_matching_sequence): Use active_insn_p to determine
3436 (try_head_merge_bb): Adjust for the flow_find_head_matching_sequence
3438 * ifcvt.c (count_bb_insns): Use active_insn_p && !JUMP_P to
3439 determine what is counted.
3441 PR tree-optimization/59643
3442 * tree-predcom.c (split_data_refs_to_components): If one dr is
3443 read and one write, determine_offset fails and the write isn't
3444 in the bad component, just put the read into the bad component.
3446 2014-01-07 Mike Stump <mikestump@comcast.net>
3447 Jakub Jelinek <jakub@redhat.com>
3450 * tree-core.h (struct tree_optimization_option): Change optabs
3451 type from unsigned char * to void *.
3452 * optabs.c (init_tree_optimization_optabs): Adjust
3453 TREE_OPTIMIZATION_OPTABS initialization.
3455 2014-01-06 Jakub Jelinek <jakub@redhat.com>
3458 * config/i386/i386.h (struct machine_function): Add
3459 no_drap_save_restore field.
3460 * config/i386/i386.c (ix86_save_reg): Use
3461 !cfun->machine->no_drap_save_restore instead of
3462 crtl->stack_realign_needed.
3463 (ix86_finalize_stack_realign_flags): Don't clear drap_reg unless
3464 this function clears frame_pointer_needed. Set
3465 cfun->machine->no_drap_save_restore if clearing frame_pointer_needed
3466 and DRAP reg is needed.
3468 2014-01-06 Marek Polacek <polacek@redhat.com>
3471 * doc/implement-c.texi: Mention that other integer types are
3472 permitted as bit-field types in strictly conforming mode.
3474 2014-01-06 Felix Yang <fei.yang0953@gmail.com>
3476 * modulo-sched.c (schedule_reg_moves): Clear distance1_uses if it
3479 2014-01-06 Richard Earnshaw <rearnsha@arm.com>
3481 * aarch64.c (aarch64_rtx_costs): Fix cost calculation for MADD.
3483 2014-01-06 Martin Jambor <mjambor@suse.cz>
3486 * ipa-cp.c (ipcp_discover_new_direct_edges): Changed param_index type
3488 * ipa-prop.c (ipa_print_node_params): Fix indentation.
3490 2014-01-06 Eric Botcazou <ebotcazou@adacore.com>
3494 * var-tracking.c (add_stores): Preserve the value of the source even if
3495 we don't record the store.
3497 2014-01-06 Terry Guo <terry.guo@arm.com>
3499 * config.gcc (arm*-*-*): Check --with-arch against arm-arches.def.
3501 2014-01-05 Iain Sandoe <iain@codesourcery.com>
3504 * config/darwin.c (darwin_function_section): Adjust return values to
3505 correspond to optimisation changes made in r206070.
3507 2014-01-05 Uros Bizjak <ubizjak@gmail.com>
3509 * config/i386/i386.c (ix86_data_alignment): Calculate max_align
3510 from prefetch_block tune setting.
3511 (nocona_cost): Correct size of prefetch block to 64.
3513 2014-01-04 Eric Botcazou <ebotcazou@adacore.com>
3515 * config/arm/arm.c (arm_get_frame_offsets): Revamp long lines.
3516 (arm_expand_epilogue_apcs_frame): Take into account the number of bytes
3517 used to save the static chain register in the computation of the offset
3518 from which the FP registers need to be restored.
3520 2014-01-04 Jakub Jelinek <jakub@redhat.com>
3522 PR tree-optimization/59519
3523 * tree-vect-loop-manip.c (slpeel_update_phi_nodes_for_guard1): Don't
3524 ICE if get_current_def (current_new_name) is already non-NULL, as long
3525 as it is a phi result of some other phi in *new_exit_bb that has
3528 * config/i386/sse.md (avx512f_load<mode>_mask): Emit vmovup{s,d}
3529 or vmovdqu* for misaligned_operand.
3530 (<sse>_loadu<ssemodesuffix><avxsizesuffix><mask_name>,
3531 <sse2_avx_avx512f>_loaddqu<mode><mask_name>): Handle <mask_applied>.
3532 * config/i386/i386.c (ix86_expand_special_args_builtin): Set
3533 aligned_mem for AVX512F masked aligned load and store builtins and for
3536 2014-01-03 Bingfeng Mei <bmei@broadcom.com>
3538 PR tree-optimization/59651
3539 * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks):
3540 Address range for negative step should be added by TYPE_SIZE_UNIT.
3542 2014-01-03 Andreas Schwab <schwab@linux-m68k.org>
3544 * config/m68k/m68k.c (handle_move_double): Handle pushes with
3545 overlapping registers also for registers other than the stack pointer.
3547 2014-01-03 Marek Polacek <polacek@redhat.com>
3550 * doc/extend.texi: Fix the return value of __builtin_FUNCTION and
3553 2014-01-03 Jakub Jelinek <jakub@redhat.com>
3556 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Don't consider
3559 * config/i386/i386.md (MODE_SIZE): New mode attribute.
3560 (push splitter): Use <P:MODE_SIZE> instead of
3561 GET_MODE_SIZE (<P:MODE>mode).
3562 (lea splitter): Use <MODE_SIZE> instead of GET_MODE_SIZE (<MODE>mode).
3563 (mov -1, reg peephole2): Likewise.
3564 * config/i386/sse.md (*mov<mode>_internal,
3565 <sse>_storeu<ssemodesuffix><avxsizesuffix>,
3566 <sse2_avx_avx512f>_storedqu<mode>, <sse>_andnot<mode>3,
3567 *<code><mode>3, *andnot<mode>3<mask_name>,
3568 <mask_codefor><code><mode>3<mask_name>): Likewise.
3569 * config/i386/subst.md (mask_mode512bit_condition,
3570 sd_mask_mode512bit_condition): Likewise.
3572 2014-01-02 Xinliang David Li <davidxl@google.com>
3574 PR tree-optimization/59303
3575 * tree-ssa-uninit.c (is_use_properly_guarded): Main cleanup.
3576 (dump_predicates): Better output format.
3577 (pred_equal_p): New function.
3578 (is_neq_relop_p): Ditto.
3579 (is_neq_zero_form_p): Ditto.
3580 (pred_expr_equal_p): Ditto.
3581 (pred_neg_p): Ditto.
3582 (simplify_pred): Ditto.
3583 (simplify_preds_2): Ditto.
3584 (simplify_preds_3): Ditto.
3585 (simplify_preds_4): Ditto.
3586 (simplify_preds): Ditto.
3588 (push_to_worklist): Ditto.
3589 (get_pred_info_from_cmp): Ditto.
3590 (is_degenerated_phi): Ditto.
3591 (normalize_one_pred_1): Ditto.
3592 (normalize_one_pred): Ditto.
3593 (normalize_one_pred_chain): Ditto.
3594 (normalize_preds): Ditto.
3595 (normalize_cond_1): Remove function.
3596 (normalize_cond): Ditto.
3597 (is_gcond_subset_of): Ditto.
3598 (is_subset_of_any): Ditto.
3599 (is_or_set_subset_of): Ditto.
3600 (is_and_set_subset_of): Ditto.
3601 (is_norm_cond_subset_of): Ditto.
3602 (pred_chain_length_cmp): Ditto.
3603 (convert_control_dep_chain_into_preds): Type change.
3604 (find_predicates): Ditto.
3605 (find_def_preds): Ditto.
3606 (destroy_predicates_vecs): Ditto.
3607 (find_matching_predicates_in_rest_chains): Ditto.
3608 (use_pred_not_overlap_with_undef_path_pred): Ditto.
3609 (is_pred_expr_subset): Ditto.
3610 (is_pred_chain_subset_of): Ditto.
3611 (is_included_in): Ditto.
3612 (is_superset_of): Ditto.
3614 2014-01-02 Richard Sandiford <rdsandiford@googlemail.com>
3616 Update copyright years.
3618 2014-01-02 Richard Sandiford <rdsandiford@googlemail.com>
3620 * common/config/arc/arc-common.c, config/arc/arc-modes.def,
3621 config/arc/arc-protos.h, config/arc/arc.c, config/arc/arc.h,
3622 config/arc/arc.md, config/arc/arc.opt,
3623 config/arm/arm_neon_builtins.def, config/arm/crypto.def,
3624 config/i386/avx512cdintrin.h, config/i386/avx512erintrin.h,
3625 config/i386/avx512fintrin.h, config/i386/avx512pfintrin.h,
3626 config/i386/btver2.md, config/i386/shaintrin.h, config/i386/slm.md,
3627 config/linux-protos.h, config/linux.c, config/winnt-c.c,
3628 diagnostic-color.c, diagnostic-color.h, gimple-ssa-isolate-paths.c,
3629 vtable-verify.c, vtable-verify.h: Use the standard form for the
3632 2014-01-02 Tobias Burnus <burnus@net-b.de>
3634 * gcc.c (process_command): Update copyright notice dates.
3635 * gcov-dump.c: Ditto.
3637 * doc/cpp.texi: Bump @copying's copyright year.
3638 * doc/cppinternals.texi: Ditto.
3639 * doc/gcc.texi: Ditto.
3640 * doc/gccint.texi: Ditto.
3641 * doc/gcov.texi: Ditto.
3642 * doc/install.texi: Ditto.
3643 * doc/invoke.texi: Ditto.
3645 2014-01-01 Jan-Benedict Glaw <jbglaw@lug-owl.de>
3647 * config/nios2/nios2.h (BITS_PER_UNIT): Don't define it.
3649 2014-01-01 Jakub Jelinek <jakub@redhat.com>
3651 * config/i386/sse.md (*mov<mode>_internal): Guard
3652 EXT_REX_SSE_REGNO_P (REGNO ()) uses with REG_P.
3654 PR rtl-optimization/59647
3655 * cse.c (cse_process_notes_1): Don't substitute negative VOIDmode
3656 new_rtx into UNSIGNED_FLOAT rtxes.
3658 Copyright (C) 2014 Free Software Foundation, Inc.
3660 Copying and distribution of this file, with or without modification,
3661 are permitted in any medium without royalty provided the copyright
3662 notice and this notice are preserved.