1 2009-08-31 DJ Delorie <dj@redhat.com>
3 * config/mep/mep.c (machine_function): Add frame_locked flag. Set
4 it once we start generating the prologue or epilogue.
5 (mep_call_saves_register): If the frame is locked, re-use cached values.
6 (mep_assign_save_slots): New, broken out from mep_expand_prologue.
7 (mep_expand_prologue): Call it.
8 (mep_expand_epilogue): Likewise.
9 (mep_start_function): Use the same logic as mep_expand_prologue.
10 (mep_pass_by_reference): Make logic more readable.
11 (mep_return_in_memory): Zero-sized objects are passed in memory.
12 (mep_reorg_noframe): Make sure we have accurate REG_DEAD notes.
14 2009-08-31 Richard Guenther <rguenther@suse.de>
16 * builtins.c (fold_builtin_memory_op): Use the alias oracle
17 to query if the memory regions for memmove overlap.
18 * tree-ssa-alias.c (ptr_deref_may_alias_decl_p): Relax the
19 asserts on pointers, instead deal with odd trees.
20 (ptr_derefs_may_alias_p): Likewise.
21 (refs_may_alias_p_1): Constructor bases also never alias.
23 2009-08-31 Gerald Pfeifer <gerald@pfeifer.com>
25 * doc/install.texi (Final install): Adjust reference on where to
26 order printed manuals.
28 2009-08-30 Olivier Hainque <hainque@adacore.com>
30 * dwarf2out.c (switch_to_frame_table_section): Move
31 definition prior to first use.
33 2009-08-30 Richard Guenther <rguenther@suse.de>
35 PR tree-optimization/41186
36 * tree-ssa-alias.c (ptr_deref_may_alias_ref_p): Remove.
37 (ao_ref_init_from_ptr_and_size): New function.
38 (ref_maybe_used_by_call_p_1): Be more precise tracking
39 used ranges for builtin functions.
40 (ref_maybe_used_by_call_p): Adjust.
41 (call_may_clobber_ref_p_1): Be more precise tracking clobbered
42 ranges for builtin functions.
43 * tree-ssa-alias.h (ao_ref_init_from_ptr_and_size): Declare.
45 2009-08-30 Alan Modra <amodra@bigpond.net.au>
48 * fwprop.c (get_reg_use_in): Delete.
49 (free_load_extend): New function.
50 (forward_propagate_subreg): Use it.
52 2009-08-29 Kaz Kojima <kkojima@gcc.gnu.org>
54 * config.gcc (sh*-*-elf): Add superh stuff only for sh*-superh-elf.
56 2009-08-29 Kai Tietz<kai.tietz@onevision.com>
59 * config/i386.c (ix86_expand_epilogue): Correct stack adjustment for
62 2009-08-29 Douglas B Rupp <rupp@gnat.com>
64 * crtstuff.c (__do_global_dtors_aux): Use atexit if no
65 fini or fini_array section.
67 2009-08-28 Sebastian Pop <sebastian.pop@amd.com>
69 * graphite-dependences.c (graphite_legal_transform_bb): Call
70 pbb_remove_duplicate_pdrs.
71 * graphite-poly.c (can_collapse_pdr): Removed.
72 (pdr_find_duplicate): Removed.
73 (can_collapse_pdrs): New.
74 (pbb_remove_duplicate_pdrs): New.
75 (new_poly_dr): Do not look for duplicates.
76 * graphite-poly.h (struct poly_bb): New field pdr_duplicates_removed.
77 (PBB_PDR_DUPLICATES_REMOVED): New.
78 (pbb_remove_duplicate_pdrs): Declared.
80 2009-08-28 Sebastian Pop <sebastian.pop@amd.com>
82 * graphite-interchange.c (pbb_interchange_profitable_p): Adjust
83 the strides by multiplying by PDR_NB_REFS.
84 * graphite-poly.c (can_collapse_pdr): New.
85 (pdr_find_duplicate): New.
86 (new_poly_dr): Call pdr_find_duplicate. Collapse duplicate PDRs.
87 Initialize PDR_NB_REFS.
88 * graphite-poly.h (struct poly_dr): Add field nb_refs.
90 (new_poly_dr): Number of subscripts is a graphite_dim_t.
92 2009-08-28 Sebastian Pop <sebastian.pop@amd.com>
95 * graphite-poly.c (apply_poly_transforms): Remove legality test before
98 2009-08-28 Sebastian Pop <sebastian.pop@amd.com>
100 * graphite-dependences.c (pddr_original_scattering): Return NULL
101 for read-read dependence relations.
102 * graphite-poly.h (enum poly_dr_type): Fix comment.
105 (pdr_may_write_p): New.
107 2009-08-28 Sebastian Pop <sebastian.pop@amd.com>
109 * graphite-poly.h (enum POLY_DR_TYPE): Renamed poly_dr_type.
110 (struct poly_dr): Same.
112 * graphite-poly.c (new_poly_dr): Same.
113 * graphite-dependences.c (dot_deps): Disable call to system.
115 2009-08-28 Cary Coutant <ccoutant@google.com>
118 * dwarf2out.c (gen_type_die_with_usage): Use proper context for
119 struct/union/enum types local to a function.
121 2009-08-28 Konrad Trifunovic <konrad.trifunovic@gmail.com>
122 Sebastian Pop <sebastian.pop@amd.com>
124 * graphite-blocking.c (pbb_strip_mine_loop_depth): Renamed
125 pbb_strip_mine_time_depth. Changed the implementation so that
126 transformation is expressed as a transformation on
127 time (scatttering) dimensions. Also, ensures that the 2d+1
128 scheduling format is preserved.
129 (pbb_strip_mine_profitable_p): Profitability is based on the
130 iteration number of a given time (scattering) dimension,
131 and not on a original loop depth dimension.
132 (pbb_strip_mine): Call pbb_number_of_iterations_at_time.
133 (pbb_do_strip_mine): Call psct_dynamic_dim.
134 * graphite-poly.c (pbb_number_of_iterations_at_time): New.
135 * graphite-poly.h (pbb_number_of_iterations_at_time): Declared.
136 (pbb_nb_dynamic_scattering_transform): New.
137 (psct_dynamic_dim): New.
139 2009-08-28 Konrad Trifunovic <konrad.trifunovic@gmail.com>
141 * graphite-ppl.c (ppl_max_for_le): Renamed ppl_max_for_le_pointset.
142 * graphite-ppl.h (ppl_max_for_le): Renamed ppl_max_for_le_pointset.
143 * graphite-poly.c (pbb_number_of_iterations): Rename ppl_max_for_le.
144 * graphite-interchange.c (build_linearized_memory_access): Same.
145 (memory_stride_in_loop): Same.
147 2009-08-28 Sebastian Pop <sebastian.pop@amd.com>
149 * graphite-dependences.c (pddr_original_scattering): New.
150 (graphite_legal_transform_dr): Call pddr_original_scattering.
153 * graphite-dependences.h (dot_deps): Declared.
154 * graphite-poly.c (new_poly_dr): Initialize PDR_ID.
155 (print_pdr): Print PDR_ID.
156 * graphite-poly.h (struct poly_dr): Add field id.
159 * graphite-scop-detection.c (dot_all_scops_1): Cleanup comment.
161 2009-08-28 Sebastian Pop <sebastian.pop@amd.com>
163 * graphite-dependences.c (graphite_carried_dependence_level_k): Do
164 not delete the original dependence relation.
166 2009-08-28 Sebastian Pop <sebastian.pop@amd.com>
168 * graphite-dependences.c (new_poly_dr_pair): Renamed new_poly_ddr.
169 (eq_poly_dr_pair_p): Renamed eq_poly_ddr_p.
170 (hash_poly_dr_pair_p): Renamed hash_poly_ddr_p.
171 (free_poly_ddr): New.
172 (pddr_is_empty): New.
173 (dependence_polyhedron_1): Now returns a poly_ddr_p.
174 (dependence_polyhedron): Same. Remove useless gcc_assert.
176 (graphite_legal_transform_dr): Call pddr_is_empty and free_poly_ddr.
177 (graphite_carried_dependence_level_k): Call pddr_is_empty.
178 * graphite-dependences.h (enum poly_dependence_kind): New.
179 (poly_dr_pair): Renamed poly_ddr. Added a field kind.
180 (PDRP_SOURCE): Renamed PDDR_SOURCE.
181 (PDRP_SINK): Renamed PDDR_SINK.
182 (PDRP_DDP): Renamed PDDR_DDP.
184 (free_poly_ddr): Declared.
185 * graphite-poly.c (new_scop): Use the new hash function names.
186 * graphite-poly.h (struct scop): Renamed field original_pdr_pairs
188 (SCOP_ORIGINAL_PDR_PAIRS): Renamed SCOP_ORIGINAL_PDDRS.
190 2009-08-28 Sebastian Pop <sebastian.pop@amd.com>
192 * cfgloopmanip.c (create_empty_loop_on_edge): Generate upper
193 bounds with LT_EXPR to make niter analysis more precise on code
194 generated by Graphite.
196 2009-08-28 Sebastian Pop <sebastian.pop@amd.com>
198 * graphite-dependences.c (graphite_legal_transform_dr): Fix formatting.
199 (graphite_legal_transform_bb): Same.
200 (poly_drs_may_alias_p): Same.
202 2009-08-28 Richard Guenther <rguenther@suse.de>
204 * tree.def: Remove note about obsolete TYPE_NONCOPIED_PARTS.
206 2009-08-28 Jan Beulich <jbeulich@novell.com>
208 * config/i386/netware.c: Include langhooks.h.
209 (i386_nlm_encode_section_info): Simplify.
210 (netware_override_options): Delete.
211 * config/i386/netware.h (netware_override_options): Delete
213 (OVERRIDE_OPTIONS): Delete definition.
214 (SUBTARGET_OVERRIDE_OPTIONS): Define.
215 (ASM_COMMENT_START): Define.
216 * config/i386/nwld.h (SUBSUBTARGET_OVERRIDE_OPTIONS): Define.
218 2009-08-28 Jan Beulich <jbeulich@novell.com>
220 * configure.ac: For in-tree ld, do a plain version check to
221 determine whether comdat groups are supported.
222 * configure: Regenerate.
224 2009-08-28 Olivier Hainque <hainque@adacore.com>
226 * collect2.c (DO_COLLECT_EXPORT_LIST): New internal macro,
227 always defined. Reflect definition or absence of such for
228 COLLECT_EXPORT_LIST. Readability helper.
229 (scanfilter): New enum, to help control what symbols
230 are to be considered or ignored by scan_prog_file.
231 (enum pass): Rename as "scanpass", moved together with scanfilter
232 prior to scan_prog_file's prototype.
233 (scan_prog_file): Accept and honor scanpass and scanfilter arguments.
234 Group prototype with the scanpass/scanfilter definitions, factorize
235 head comments for the several implementations at the prototype.
236 (main): Reorganize the first pass link control to let AIX
237 drag only the needed frame tables in executables. Prevent
238 frame tables collection during the scan aimed at static ctors.
239 Pre-link and scan for frame tables later to compensate.
240 * doc/tm.texi (ASM_OUTPUT_DWARF_TABLE_REF): New macro.
241 A C statement to issue assembly directives that create a reference
242 to the given DWARF table identifier label from the current function
244 * dwarf2out.c (switch_to_eh_frame_section): Add a BACK argument
245 to differentiate first time section entry. Only emit a .data
246 tables start identifier label the first time around.
247 (switch_to_frame_table_section): New function. Helper for
248 output_call_frame_info to switch possibly BACK into the eh_frame
249 or the debug_frame section depending on FOR_EH.
250 (output_call_frame_info): Use helper to first enter the proper
252 (output_fde): Use ASM_OUTPUT_DWARF_TABLE_REF when defined to
253 emit a link to the frame table start label from each function
255 * config/rs6000/rs6000.c (rs6000_aix_asm_output_dwarf_table_ref):
256 New function. Implementation of ASM_OUTPUT_DWARF_TABLE_REF.
257 * config/rs6000/rs6000-protos.h: Declare it.
258 * config/rs6000/aix.h (ASM_OUTPUT_DWARF_TABLE_REF): Define.
260 2009-08-27 Kaz Kojima <kkojima@gcc.gnu.org>
262 * config/sh/sh.c (split_branches): Check the result of
265 2009-08-27 Steve Ellcey <sje@cup.hp.com>
267 * config/ia64/hpux.h (LIB_SPEC): Add -lrt for when
268 using -pthread -fopenmp
270 2009-08-27 Gerald Pfeifer <gerald@pfeifer.com>
272 * doc/service.texi (service directory): Update URL.
274 2009-08-27 Uros Bizjak <ubizjak@gmail.com>
276 PR rtl-optimization/40861
277 * simplify-rtx.c (simplify_subreg): Do not call simplify_gen_subreg to
278 extract word from a multi-word subreg for negative byte positions.
280 2009-08-27 Tristan Gingold <gingold@adacore.com>
281 Douglas B Rupp <rupp@gnat.com>
283 * config/ia64/ia64.c (ia64_attribute_table): Add "common_object" entry.
284 (SECTION_VMS_OVERLAY): Define.
285 (ia64_vms_common_object_attribute): Added. Handle the "common_object"
287 (ia64_vms_elf_asm_named_section): Added. Generate .section pseudo-op
289 (ia64_vms_output_aligned_decl_common): Added. Generate pseudo-op for
290 common_object declarations.
291 (ia64_section_type_flags): Set section flag for common_object.
292 * config/ia64/ia64-protos.h
293 (ia64_vms_output_aligned_decl_common): Declare.
294 (ia64_vms_elf_asm_named_section): Declare.
296 2009-08-27 Michael Matz <matz@suse.de>
298 * expr.c (expand_expr_real_2): New function taking exploded
299 unary or binary expression, split out from ...
300 (expand_expr_real_1): ... here. Move over all unary/binary
301 switch parts to above function, in particular these codes:
302 PAREN_EXPR, NOP_EXPR, CONVERT_EXPR, POINTER_PLUS_EXPR, PLUS_EXPR,
303 MINUS_EXPR, MULT_EXPR, TRUNC_DIV_EXPR, FLOOR_DIV_EXPR, CEIL_DIV_EXPR,
304 ROUND_DIV_EXPR, EXACT_DIV_EXPR, RDIV_EXPR, TRUNC_MOD_EXPR,
305 FLOOR_MOD_EXPR, CEIL_MOD_EXPR, ROUND_MOD_EXPR, FIXED_CONVERT_EXPR,
306 FIX_TRUNC_EXPR, FLOAT_EXPR, NEGATE_EXPR, ABS_EXPR, MAX_EXPR, MIN_EXPR,
307 BIT_NOT_EXPR, TRUTH_AND_EXPR, BIT_AND_EXPR, TRUTH_OR_EXPR,
308 BIT_IOR_EXPR, TRUTH_XOR_EXPR, BIT_XOR_EXPR, LROTATE_EXPR, RROTATE_EXPR,
309 LSHIFT_EXPR, RSHIFT_EXPR, LT_EXPR, LE_EXPR, GT_EXPR, GE_EXPR, EQ_EXPR,
310 NE_EXPR, UNORDERED_EXPR, ORDERED_EXPR, UNLT_EXPR, UNLE_EXPR, UNGT_EXPR,
311 UNGE_EXPR, UNEQ_EXPR, LTGT_EXPR, TRUTH_NOT_EXPR, COMPLEX_EXPR,
312 WIDEN_SUM_EXPR, REDUC_MAX_EXPR, REDUC_MIN_EXPR, REDUC_PLUS_EXPR,
313 VEC_EXTRACT_EVEN_EXPR, VEC_EXTRACT_ODD_EXPR, VEC_INTERLEAVE_HIGH_EXPR,
314 VEC_INTERLEAVE_LOW_EXPR, VEC_LSHIFT_EXPR, VEC_RSHIFT_EXPR,
315 VEC_UNPACK_HI_EXPR, VEC_UNPACK_LO_EXPR, VEC_UNPACK_FLOAT_HI_EXPR,
316 VEC_UNPACK_FLOAT_LO_EXPR, VEC_WIDEN_MULT_HI_EXPR,
317 VEC_WIDEN_MULT_LO_EXPR, VEC_PACK_TRUNC_EXPR, VEC_PACK_SAT_EXPR,
318 VEC_PACK_FIX_TRUNC_EXPR.
319 (<case PAREN_EXPR>): Call set_mem_attributes() with type, not the
322 2009-08-27 Richard Guenther <rguenther@suse.de>
324 * gengtype.c (main): Handle uint64_t.
325 * ipa-utils.c (get_base_var): Indent properly.
326 * tree-ssa-live.c (debug_scope_block): New function.
327 * tree-flow.h (debug_scope_block): Declare.
328 * tree-ssa-copy.c (replace_exp_1): Add vertical space.
329 * basic-block.h (enum profile_status): Rename to
330 enum profile_status_d.
331 (x_profile_status): Adjust type.
333 2009-08-27 Dodji Seketeli <dodji@redhat.com>
336 * dwarf2out.c (get_context_die): Declare this static function.
337 (gen_type_die_with_usage): Make sure a DIE is a generated for
338 the context of a typedef.
340 2009-08-26 Anatoly Sokolov <aesok@post.ru>
342 * doc/invoke.texi (AVR Options): Remove documentation of -minit-stack
345 2009-08-26 Richard Sandiford <rdsandiford@googlemail.com>
347 * config/mips/mips-protos.h (mips_output_sync): Declare.
348 (mips_sync_loop_insns): Likewise.
349 (mips_output_sync_loop): Replace first two parameters with an rtx.
350 * config/mips/mips.c (mips_multi_member): New structure.
351 (mips_multi_members): New variable.
352 (mips_multi_start): New function.
353 (mips_multi_add): Likewise.
354 (mips_multi_add_insn): Likewise.
355 (mips_multi_add_label): Likewise.
356 (mips_multi_last_index): Likewise.
357 (mips_multi_copy_insn): Likewise.
358 (mips_multi_set_operand): Likewise.
359 (mips_multi_write): Likewise.
360 (mips_print_operand_punctuation): Remove '%|' and '%-'.
361 (mips_init_print_operand_punct): Update accordingly.
362 (mips_start_ll_sc_sync_block): New function.
363 (mips_end_ll_sc_sync_block): Likewise.
364 (mips_output_sync): Likewise.
365 (mips_sync_insn1_template): Likewise.
366 (mips_sync_insn2_template): Likewise.
367 (mips_get_sync_operand): Likewise.
368 (mips_process_sync_loop): Likewise.
369 (mips_output_sync_loop): Use mips_process_sync_loop.
370 (mips_sync_loop_insns): New function.
371 * config/mips/mips.h (MIPS_COMPARE_AND_SWAP): Delete.
372 (MIPS_COMPARE_AND_SWAP_12): Likewise.
373 (MIPS_COMPARE_AND_SWAP_12_ZERO_OP): Likewise.
374 (MIPS_COMPARE_AND_SWAP_12_NONZERO_OP): Likewise.
375 (MIPS_SYNC_OP, MIPS_SYNC_OP_12): Likewise.
376 (MIPS_SYNC_OP_12_AND, MIPS_SYNC_OP_12_XOR): Likewise.
377 (MIPS_SYNC_OLD_OP_12): Likewise.
378 (MIPS_SYNC_OLD_OP_12_AND, MIPS_SYNC_OLD_OP_12_XOR): Likewise.
379 (MIPS_SYNC_NEW_OP_12): Likewise.
380 (MIPS_SYNC_NEW_OP_12_AND, MIPS_SYNC_NEW_OP_12_XOR): Likewise.
381 (MIPS_SYNC_OLD_OP, MIPS_SYNC_NEW_OP): Likewise.
382 (MIPS_SYNC_NAND, MIPS_SYNC_OLD_NAND, MIPS_SYNC_NEW_NAND): Likewise.
383 (MIPS_SYNC_EXCHANGE, MIPS_SYNC_EXCHANGE_12): Likewise.
384 (MIPS_SYNC_EXCHANGE_12_ZERO_OP): Likewise.
385 (MIPS_SYNC_EXCHANGE_12_NONZER_OP): Likewise.
386 * config/mips/mips.md (sync_mem): New attribute.
387 (sync_oldval, sync_newval, sync_inclusive_mask): Likewise.
388 (sync_exclusive_mask, sync_required_oldval): Likewise.
389 (sync_insn1_op2, sync_insn1, sync_insn2): Likewise.
390 (sync_release_barrier): Likewise.
391 (length): Handle sync loops.
392 (sync): Use mips_output_sync.
393 * config/mips/sync.md (*memory_barrier): Use mips_output_sync.
394 (sync_compare_and_swap<mode>): Set the new sync_* attributes
395 and use mips_output_sync_loop.
396 (compare_and_swap_12, sync_add<mode>, sync_<optab>_12): Likewise.
397 (sync_old_<optab>_12, sync_new_<optab>_12, sync_nand_12): Likewise.
398 (sync_old_nand_12, sync_new_nand_12, sync_sub<mode>): Likewise.
399 (sync_old_add<mode>, sync_old_sub<mode>): Likewise.
400 (sync_new_add<mode>, sync_new_sub<mode>): Likewise.
401 (sync_<optab><mode>, sync_old_<optab><mode>): Likewise.
402 (sync_new_<optab><mode>, sync_nand<mode>): Likewise.
403 (sync_old_nand<mode>, sync_new_nand<mode>): Likewise.
404 (sync_lock_test_and_set<mode>, test_and_set_12): Likewise.
406 2009-08-26 Richard Guenther <rguenther@suse.de>
409 * gimplify.c (gimplify_addr_expr): Canonicalize ADDR_EXPRs if
410 the types to not match.
411 * tree-cfg.c (verify_gimple_assign_single): Adjust ADDR_EXPR
413 * tree-ssa.c (useless_type_conversion_p): Conversions to
414 pointers to unprototyped functions are useless.
416 2009-08-26 Richard Guenther <rguenther@suse.de>
418 * tree-ssa-structalias.c (create_variable_info_for): Remove
419 strange whole-program condition, prepare to be called for non-globals.
420 (intra_create_variable_infos): For restrict qualified DECL_BY_REFERENCE
421 params build a representative with known type and track its fields.
423 2009-08-26 Uros Bizjak <ubizjak@gmail.com>
425 * config/alpha/sync.md: Update comment about unpredictable LL/SC lock
426 clearing by a taken branch.
427 (sync_<fetchop_name><mode>): Split when epilogue_completed is set,
428 effectively after bbro pass.
429 (sync_nand<mode>): Ditto.
430 (sync_old_<fetchop_name><mode>): Ditto.
431 (sync_old_nand<mode>): Ditto.
432 (sync_new_<fetchop_name><mode>): Dito.
433 (sync_new_nand<mode>): Ditto.
434 (sync_compare_and_swap<mode>_1): Ditto.
435 (*sync_compare_and_swap<mode>): Ditto.
436 (sync_lock_test_and_set<mode>_1): Ditto.
437 ("sync_lock_test_and_set<mode>): Ditto.
439 2009-08-25 Douglas B Rupp <rupp@gnat.com>
441 * hwint.h (HOST_LONG_FORMAT): New macro
442 * bitmap.c, c-decl.c, mips-tfile.c, print-rtl.c, print-tree.c:
444 * system.h (HOST_PTR_PRINTF): Resurrect old macro
445 * doc/hostconfig.texi (HOST_LONG_FORMAT): Document.
446 (HOST_PTR_PRINTF): Document.
448 2009-08-25 Jan Hubicka <jh@suse.cz>
450 * config/i386/bmmintrin.h: Replace by #error.
453 Michael Meissner <michael.meissner@amd.com>
454 Dwarakanath Rajagopal <dwarak.rajagopal@amd.com>
455 Tony Linthicum <tony.linthicum@amd.com>
457 * config/i386/i386.h (TARGET_SSE5): New macro for SSE5.
458 (TARGET_ROUND): New macro for the round/ptest instructions which
459 are shared between SSE4.1 and SSE5.
460 (OPTION_MASK_ISA_ROUND): Ditto.
461 (OPTION_ISA_ROUND): Ditto.
462 (TARGET_FUSED_MADD): New macro for -mfused-madd swtich.
463 (TARGET_CPU_CPP_BUILTINS): Add SSE5 support.
465 * config/i386/i386.opt (-msse5): New switch for SSE5 support.
466 (-mfused-madd): New switch to give users control over whether the
467 compiler optimizes to use the multiply/add SSE5 instructions.
469 * config/i386/i386.c (enum pta_flags): Add PTA_SSE5.
470 (ix86_handle_option): Turn off 3dnow if -msse5.
471 (override_options): Add SSE5 support.
472 (print_operand): %Y prints comparison codes for SSE5 com/pcom
474 (ix86_expand_sse_movcc): Add SSE5 support.
475 (ix86_expand_sse5_unpack): New function to use pperm to unpack a
476 vector type to the next largest size.
477 (ix86_expand_sse5_pack): New function to use pperm to pack a
478 vector type to the next smallest size.
479 (IX86_BUILTIN_FMADDSS): New for SSE5 intrinsic.
480 (IX86_BUILTIN_FMADDSD): Ditto.
481 (IX86_BUILTIN_FMADDPS): Ditto.
482 (IX86_BUILTIN_FMADDPD): Ditto.
483 (IX86_BUILTIN_FMSUBSS): Ditto.
484 (IX86_BUILTIN_FMSUBSD): Ditto.
485 (IX86_BUILTIN_FMSUBPS): Ditto.
486 (IX86_BUILTIN_FMSUBPD): Ditto.
487 (IX86_BUILTIN_FNMADDSS): Ditto.
488 (IX86_BUILTIN_FNMADDSD): Ditto.
489 (IX86_BUILTIN_FNMADDPS): Ditto.
490 (IX86_BUILTIN_FNMADDPD): Ditto.
491 (IX86_BUILTIN_FNMSUBSS): Ditto.
492 (IX86_BUILTIN_FNMSUBSD): Ditto.
493 (IX86_BUILTIN_FNMSUBPS): Ditto.
494 (IX86_BUILTIN_FNMSUBPD): Ditto.
495 (IX86_BUILTIN_PCMOV_V2DI): Ditto.
496 (IX86_BUILTIN_PCMOV_V4SI): Ditto.
497 (IX86_BUILTIN_PCMOV_V8HI): Ditto.
498 (IX86_BUILTIN_PCMOV_V16QI): Ditto.
499 (IX86_BUILTIN_PCMOV_V4SF): Ditto.
500 (IX86_BUILTIN_PCMOV_V2DF): Ditto.
501 (IX86_BUILTIN_PPERM): Ditto.
502 (IX86_BUILTIN_PERMPS): Ditto.
503 (IX86_BUILTIN_PERMPD): Ditto.
504 (IX86_BUILTIN_PMACSSWW): Ditto.
505 (IX86_BUILTIN_PMACSWW): Ditto.
506 (IX86_BUILTIN_PMACSSWD): Ditto.
507 (IX86_BUILTIN_PMACSWD): Ditto.
508 (IX86_BUILTIN_PMACSSDD): Ditto.
509 (IX86_BUILTIN_PMACSDD): Ditto.
510 (IX86_BUILTIN_PMACSSDQL): Ditto.
511 (IX86_BUILTIN_PMACSSDQH): Ditto.
512 (IX86_BUILTIN_PMACSDQL): Ditto.
513 (IX86_BUILTIN_PMACSDQH): Ditto.
514 (IX86_BUILTIN_PMADCSSWD): Ditto.
515 (IX86_BUILTIN_PMADCSWD): Ditto.
516 (IX86_BUILTIN_PHADDBW): Ditto.
517 (IX86_BUILTIN_PHADDBD): Ditto.
518 (IX86_BUILTIN_PHADDBQ): Ditto.
519 (IX86_BUILTIN_PHADDWD): Ditto.
520 (IX86_BUILTIN_PHADDWQ): Ditto.
521 (IX86_BUILTIN_PHADDDQ): Ditto.
522 (IX86_BUILTIN_PHADDUBW): Ditto.
523 (IX86_BUILTIN_PHADDUBD): Ditto.
524 (IX86_BUILTIN_PHADDUBQ): Ditto.
525 (IX86_BUILTIN_PHADDUWD): Ditto.
526 (IX86_BUILTIN_PHADDUWQ): Ditto.
527 (IX86_BUILTIN_PHADDUDQ): Ditto.
528 (IX86_BUILTIN_PHSUBBW): Ditto.
529 (IX86_BUILTIN_PHSUBWD): Ditto.
530 (IX86_BUILTIN_PHSUBDQ): Ditto.
531 (IX86_BUILTIN_PROTB): Ditto.
532 (IX86_BUILTIN_PROTW): Ditto.
533 (IX86_BUILTIN_PROTD): Ditto.
534 (IX86_BUILTIN_PROTQ): Ditto.
535 (IX86_BUILTIN_PROTB_IMM): Ditto.
536 (IX86_BUILTIN_PROTW_IMM): Ditto.
537 (IX86_BUILTIN_PROTD_IMM): Ditto.
538 (IX86_BUILTIN_PROTQ_IMM): Ditto.
539 (IX86_BUILTIN_PSHLB): Ditto.
540 (IX86_BUILTIN_PSHLW): Ditto.
541 (IX86_BUILTIN_PSHLD): Ditto.
542 (IX86_BUILTIN_PSHLQ): Ditto.
543 (IX86_BUILTIN_PSHAB): Ditto.
544 (IX86_BUILTIN_PSHAW): Ditto.
545 (IX86_BUILTIN_PSHAD): Ditto.
546 (IX86_BUILTIN_PSHAQ): Ditto.
547 (IX86_BUILTIN_FRCZSS): Ditto.
548 (IX86_BUILTIN_FRCZSD): Ditto.
549 (IX86_BUILTIN_FRCZPS): Ditto.
550 (IX86_BUILTIN_FRCZPD): Ditto.
551 (IX86_BUILTIN_CVTPH2PS): Ditto.
552 (IX86_BUILTIN_CVTPS2PH): Ditto.
553 (IX86_BUILTIN_COMEQSS): Ditto.
554 (IX86_BUILTIN_COMNESS): Ditto.
555 (IX86_BUILTIN_COMLTSS): Ditto.
556 (IX86_BUILTIN_COMLESS): Ditto.
557 (IX86_BUILTIN_COMGTSS): Ditto.
558 (IX86_BUILTIN_COMGESS): Ditto.
559 (IX86_BUILTIN_COMUEQSS): Ditto.
560 (IX86_BUILTIN_COMUNESS): Ditto.
561 (IX86_BUILTIN_COMULTSS): Ditto.
562 (IX86_BUILTIN_COMULESS): Ditto.
563 (IX86_BUILTIN_COMUGTSS): Ditto.
564 (IX86_BUILTIN_COMUGESS): Ditto.
565 (IX86_BUILTIN_COMORDSS): Ditto.
566 (IX86_BUILTIN_COMUNORDSS): Ditto.
567 (IX86_BUILTIN_COMFALSESS): Ditto.
568 (IX86_BUILTIN_COMTRUESS): Ditto.
569 (IX86_BUILTIN_COMEQSD): Ditto.
570 (IX86_BUILTIN_COMNESD): Ditto.
571 (IX86_BUILTIN_COMLTSD): Ditto.
572 (IX86_BUILTIN_COMLESD): Ditto.
573 (IX86_BUILTIN_COMGTSD): Ditto.
574 (IX86_BUILTIN_COMGESD): Ditto.
575 (IX86_BUILTIN_COMUEQSD): Ditto.
576 (IX86_BUILTIN_COMUNESD): Ditto.
577 (IX86_BUILTIN_COMULTSD): Ditto.
578 (IX86_BUILTIN_COMULESD): Ditto.
579 (IX86_BUILTIN_COMUGTSD): Ditto.
580 (IX86_BUILTIN_COMUGESD): Ditto.
581 (IX86_BUILTIN_COMORDSD): Ditto.
582 (IX86_BUILTIN_COMUNORDSD): Ditto.
583 (IX86_BUILTIN_COMFALSESD): Ditto.
584 (IX86_BUILTIN_COMTRUESD): Ditto.
585 (IX86_BUILTIN_COMEQPS): Ditto.
586 (IX86_BUILTIN_COMNEPS): Ditto.
587 (IX86_BUILTIN_COMLTPS): Ditto.
588 (IX86_BUILTIN_COMLEPS): Ditto.
589 (IX86_BUILTIN_COMGTPS): Ditto.
590 (IX86_BUILTIN_COMGEPS): Ditto.
591 (IX86_BUILTIN_COMUEQPS): Ditto.
592 (IX86_BUILTIN_COMUNEPS): Ditto.
593 (IX86_BUILTIN_COMULTPS): Ditto.
594 (IX86_BUILTIN_COMULEPS): Ditto.
595 (IX86_BUILTIN_COMUGTPS): Ditto.
596 (IX86_BUILTIN_COMUGEPS): Ditto.
597 (IX86_BUILTIN_COMORDPS): Ditto.
598 (IX86_BUILTIN_COMUNORDPS): Ditto.
599 (IX86_BUILTIN_COMFALSEPS): Ditto.
600 (IX86_BUILTIN_COMTRUEPS): Ditto.
601 (IX86_BUILTIN_COMEQPD): Ditto.
602 (IX86_BUILTIN_COMNEPD): Ditto.
603 (IX86_BUILTIN_COMLTPD): Ditto.
604 (IX86_BUILTIN_COMLEPD): Ditto.
605 (IX86_BUILTIN_COMGTPD): Ditto.
606 (IX86_BUILTIN_COMGEPD): Ditto.
607 (IX86_BUILTIN_COMUEQPD): Ditto.
608 (IX86_BUILTIN_COMUNEPD): Ditto.
609 (IX86_BUILTIN_COMULTPD): Ditto.
610 (IX86_BUILTIN_COMULEPD): Ditto.
611 (IX86_BUILTIN_COMUGTPD): Ditto.
612 (IX86_BUILTIN_COMUGEPD): Ditto.
613 (IX86_BUILTIN_COMORDPD): Ditto.
614 (IX86_BUILTIN_COMUNORDPD): Ditto.
615 (IX86_BUILTIN_COMFALSEPD): Ditto.
616 (IX86_BUILTIN_COMTRUEPD): Ditto.
617 (IX86_BUILTIN_PCOMEQUB): Ditto.
618 (IX86_BUILTIN_PCOMNEUB): Ditto.
619 (IX86_BUILTIN_PCOMLTUB): Ditto.
620 (IX86_BUILTIN_PCOMLEUB): Ditto.
621 (IX86_BUILTIN_PCOMGTUB): Ditto.
622 (IX86_BUILTIN_PCOMGEUB): Ditto.
623 (IX86_BUILTIN_PCOMFALSEUB): Ditto.
624 (IX86_BUILTIN_PCOMTRUEUB): Ditto.
625 (IX86_BUILTIN_PCOMEQUW): Ditto.
626 (IX86_BUILTIN_PCOMNEUW): Ditto.
627 (IX86_BUILTIN_PCOMLTUW): Ditto.
628 (IX86_BUILTIN_PCOMLEUW): Ditto.
629 (IX86_BUILTIN_PCOMGTUW): Ditto.
630 (IX86_BUILTIN_PCOMGEUW): Ditto.
631 (IX86_BUILTIN_PCOMFALSEUW): Ditto.
632 (IX86_BUILTIN_PCOMTRUEUW): Ditto.
633 (IX86_BUILTIN_PCOMEQUD): Ditto.
634 (IX86_BUILTIN_PCOMNEUD): Ditto.
635 (IX86_BUILTIN_PCOMLTUD): Ditto.
636 (IX86_BUILTIN_PCOMLEUD): Ditto.
637 (IX86_BUILTIN_PCOMGTUD): Ditto.
638 (IX86_BUILTIN_PCOMGEUD): Ditto.
639 (IX86_BUILTIN_PCOMFALSEUD): Ditto.
640 (IX86_BUILTIN_PCOMTRUEUD): Ditto.
641 (IX86_BUILTIN_PCOMEQUQ): Ditto.
642 (IX86_BUILTIN_PCOMNEUQ): Ditto.
643 (IX86_BUILTIN_PCOMLTUQ): Ditto.
644 (IX86_BUILTIN_PCOMLEUQ): Ditto.
645 (IX86_BUILTIN_PCOMGTUQ): Ditto.
646 (IX86_BUILTIN_PCOMGEUQ): Ditto.
647 (IX86_BUILTIN_PCOMFALSEUQ): Ditto.
648 (IX86_BUILTIN_PCOMTRUEUQ): Ditto.
649 (IX86_BUILTIN_PCOMEQB): Ditto.
650 (IX86_BUILTIN_PCOMNEB): Ditto.
651 (IX86_BUILTIN_PCOMLTB): Ditto.
652 (IX86_BUILTIN_PCOMLEB): Ditto.
653 (IX86_BUILTIN_PCOMGTB): Ditto.
654 (IX86_BUILTIN_PCOMGEB): Ditto.
655 (IX86_BUILTIN_PCOMFALSEB): Ditto.
656 (IX86_BUILTIN_PCOMTRUEB): Ditto.
657 (IX86_BUILTIN_PCOMEQW): Ditto.
658 (IX86_BUILTIN_PCOMNEW): Ditto.
659 (IX86_BUILTIN_PCOMLTW): Ditto.
660 (IX86_BUILTIN_PCOMLEW): Ditto.
661 (IX86_BUILTIN_PCOMGTW): Ditto.
662 (IX86_BUILTIN_PCOMGEW): Ditto.
663 (IX86_BUILTIN_PCOMFALSEW): Ditto.
664 (IX86_BUILTIN_PCOMTRUEW): Ditto.
665 (IX86_BUILTIN_PCOMEQD): Ditto.
666 (IX86_BUILTIN_PCOMNED): Ditto.
667 (IX86_BUILTIN_PCOMLTD): Ditto.
668 (IX86_BUILTIN_PCOMLED): Ditto.
669 (IX86_BUILTIN_PCOMGTD): Ditto.
670 (IX86_BUILTIN_PCOMGED): Ditto.
671 (IX86_BUILTIN_PCOMFALSED): Ditto.
672 (IX86_BUILTIN_PCOMTRUED): Ditto.
673 (IX86_BUILTIN_PCOMEQQ): Ditto.
674 (IX86_BUILTIN_PCOMNEQ): Ditto.
675 (IX86_BUILTIN_PCOMLTQ): Ditto.
676 (IX86_BUILTIN_PCOMLEQ): Ditto.
677 (IX86_BUILTIN_PCOMGTQ): Ditto.
678 (IX86_BUILTIN_PCOMGEQ): Ditto.
679 (IX86_BUILTIN_PCOMFALSEQ): Ditto.
680 (IX86_BUILTIN_PCOMTRUEQ): Ditto.
681 (enum multi_arg_type): New enum for describing the various SSE5
682 intrinsic argument types.
683 (bdesc_multi_arg): New table for SSE5 intrinsics.
684 (ix86_init_mmx_sse_builtins): Add SSE5 intrinsic support.
685 (ix86_expand_multi_arg_builtin): New function for creating SSE5
687 (ix86_expand_builtin): Add SSE5 intrinsic support.
688 (ix86_sse5_valid_op_p): New function to validate SSE5 3 and 4
689 operand instructions.
690 (ix86_expand_sse5_multiple_memory): New function to split the
691 second memory reference from SSE5 instructions.
692 (type_has_variadic_args_p): Delete in favor of stdarg_p.
693 (ix86_return_pops_args): Use stdarg_p to determine if the function
694 has variable arguments.
695 (ix86_setup_incoming_varargs): Ditto.
696 (x86_this_parameter): Ditto.
698 * config/i386/i386-protos.h (ix86_expand_sse5_unpack): Add
700 (ix86_expand_sse5_pack): Ditto.
701 (ix86_sse5_valid_op_p): Ditto.
702 (ix86_expand_sse5_multiple_memory): Ditto.
704 * config/i386/i386.md (UNSPEC_SSE5_INTRINSIC): Add new UNSPEC
705 constant for SSE5 support.
706 (UNSPEC_SSE5_UNSIGNED_CMP): Ditto.
707 (UNSPEC_SSE5_TRUEFALSE): Ditto.
708 (UNSPEC_SSE5_PERMUTE): Ditto.
709 (UNSPEC_SSE5_ASHIFT): Ditto.
710 (UNSPEC_SSE5_LSHIFT): Ditto.
711 (UNSPEC_FRCZ): Ditto.
712 (UNSPEC_CVTPH2PS): Ditto.
713 (UNSPEC_CVTPS2PH): Ditto.
714 (PCOM_FALSE): Add new constant for true/false SSE5 comparisons.
716 (COM_FALSE_S): Ditto.
717 (COM_FALSE_P): Ditto.
720 (type attribute): Add ssemuladd, sseiadd1, ssecvt1, sse4arg types.
721 (unit attribute): Add support for ssemuladd, ssecvt1, sseiadd1 sse4arg
723 (memory attribute): Ditto.
724 (sse4_1_round<mode>2): Use TARGET_ROUND instead of TARGET_SSE4_1.
725 Use SSE4_1_ROUND_* constants instead of hard coded numbers.
726 (rint<mode>2): Use TARGET_ROUND instead of TARGET_SSE4_1.
727 (floor<mode>2): Ditto.
728 (ceil<mode>2): Ditto.
729 (btrunc<mode>2): Ditto.
730 (nearbyintdf2): Ditto.
731 (nearbyintsf2): Ditto.
732 (sse_setccsf): Disable if SSE5.
733 (sse_setccdf): Ditto.
734 (sse5_setcc<mode>): New support for SSE5 conditional move.
735 (sse5_pcmov_<mode>): Ditto.
737 * config/i386/sse.md (SSEMODE1248): New mode iterator for SSE5.
740 (ssemodesuffixf4): New mode attribute for SSE5.
741 (ssemodesuffixf2s): Ditto.
742 (ssemodesuffixf2c): Ditto.
743 (sserotatemax): Ditto.
744 (ssescalarmode): Ditto.
745 (sse_maskcmpv4sf3): Disable if SSE5.
746 (sse_maskcmpv2df3): Ditto.
747 (sse_vmmaskcmpv4sf3): Ditto.
748 (sse5_fmadd<mode>4): Add SSE5 floating point multiply/add instructions.
749 (sse5_vmfmadd<mode>4): Ditto.
750 (sse5_fmsub<mode>4): Ditto.
751 (sse5_vmfmsub<mode>4): Ditto.
752 (sse5_fnmadd<mode>4): Ditto.
753 (sse5_vmfnmadd<mode>4): Ditto.
754 (sse5_fnmsub<mode>4): Ditto.
755 (sse5_vmfnmsub<mode>4): Ditto.
756 (sse5i_fmadd<mode>4): Ditto.
757 (sse5i_fmsub<mode>4): Ditto.
758 (sse5i_fnmadd<mode>4): Ditto.
759 (sse5i_fnmsub<mode>4): Ditto.
760 (sse5i_vmfmadd<mode>4): Ditto.
761 (sse5i_vmfmsub<mode>4): Ditto.
762 (sse5i_vmfnmadd<mode>4): Ditto.
763 (sse5i_vmfnmsub<mode>4): Ditto.
764 (mulv16qi3): Add SSE5 support.
766 (sse5_mulv4si3): New insn for 32-bit multiply support on SSE5.
767 (sse2_mulv4si3): Disable if SSE5.
768 (sse4_1_roundpd): Use TARGET_ROUND instead of TARGET_SSE4_1.
769 (sse4_1_roundps): Ditto.
770 (sse4_1_roundsd): Ditto.
771 (sse4_1_roundss): Ditto.
772 (sse_maskcmpv4sf3): Disable if SSE5 so the SSE5 instruction will
774 (sse_maskcmpsf3): Ditto.
775 (sse_vmmaskcmpv4sf3): Ditto.
776 (sse2_maskcmpv2df3): Ditto.
777 (sse2_maskcmpdf3): Ditto.
778 (sse2_vmmaskcmpv2df3): Ditto.
779 (sse2_eq<mode>3): Ditto.
780 (sse2_gt<mode>3): Ditto.
781 (sse5_pcmov_<mode>): Add SSE5 support.
782 (vec_unpacku_hi_v16qi): Ditto.
783 (vec_unpacks_hi_v16qi): Ditto.
784 (vec_unpacku_lo_v16qi): Ditto.
785 (vec_unpacks_lo_v16qi): Ditto.
786 (vec_unpacku_hi_v8hi): Ditto.
787 (vec_unpacks_hi_v8hi): Ditto.
788 (vec_unpacku_lo_v8hi): Ditto.
789 (vec_unpacks_lo_v8hi): Ditto.
790 (vec_unpacku_hi_v4si): Ditto.
791 (vec_unpacks_hi_v4si): Ditto.
792 (vec_unpacku_lo_v4si): Ditto.
793 (vec_unpacks_lo_v4si): Ditto.
794 (sse5_pmacsww): New SSE5 intrinsic insn.
795 (sse5_pmacssww): Ditto.
796 (sse5_pmacsdd): Ditto.
797 (sse5_pmacssdd): Ditto.
798 (sse5_pmacssdql): Ditto.
799 (sse5_pmacssdqh): Ditto.
800 (sse5_pmacsdqh): Ditto.
801 (sse5_pmacsswd): Ditto.
802 (sse5_pmacswd): Ditto.
803 (sse5_pmadcsswd): Ditto.
804 (sse5_pmadcswd): Ditto.
805 (sse5_pcmov_<move>): Conditional move support on SSE5.
806 (sse5_phaddbw): New SSE5 intrinsic insn.
807 (sse5_phaddbd): Ditto.
808 (sse5_phaddbq): Ditto.
809 (sse5_phaddwd): Ditto.
810 (sse5_phaddwq): Ditto.
811 (sse5_phadddq): Ditto.
812 (sse5_phaddubw): Ditto.
813 (sse5_phaddubd): Ditto.
814 (sse5_phaddubq): Ditto.
815 (sse5_phadduwd): Ditto.
816 (sse5_phadduwq): Ditto.
817 (sse5_phaddudq): Ditto.
818 (sse5_phsubbw): Ditto.
819 (sse5_phsubwd): Ditto.
820 (sse5_phsubdq): Ditto.
822 (sse5_pperm_sign_v16qi_v8hi): New insns for pack/unpack with SSE5.
823 (sse5_pperm_zero_v16qi_v8hi): Ditto.
824 (sse5_pperm_sign_v8hi_v4si): Ditto.
825 (sse5_pperm_zero_v8hi_v4si): Ditto.
826 (sse5_pperm_sign_v4si_v2di): Ditto.
827 (sse5_pperm_sign_v4si_v2di): Ditto.
828 (sse5_pperm_pack_v2di_v4si): Ditto.
829 (sse5_pperm_pack_v4si_v8hi): Ditto.
830 (sse5_pperm_pack_v8hi_v16qi): Ditto.
831 (sse5_perm<mode>): New SSE5 intrinsic insn.
832 (rotl<mode>3): Ditto.
833 (sse5_rotl<mode>3): Ditto.
834 (sse5_ashl<mode>3): Ditto.
835 (sse5_lshl<mode>3): Ditto.
836 (sse5_frcz<mode>2): Ditto.
837 (sse5s_frcz<mode>2): Ditto.
838 (sse5_cvtph2ps): Ditto.
839 (sse5_cvtps2ph): Ditto.
840 (sse5_vmmaskcmp<mode>3): Ditto.
841 (sse5_com_tf<mode>3): Ditto.
842 (sse5_maskcmp<mode>3): Ditto.
843 (sse5_maskcmp_uns<mode>3): Ditto.
844 (sse5_maskcmp_uns2<mode>3): Ditto.
845 (sse5_pcom_tf<mode>3): Ditto.
847 * config/i386/predicates.md (sse5_comparison_float_operator):
848 New predicate to match the comparison operators supported by
849 the SSE5 com instruction.
850 (ix86_comparison_int_operator): New predicate to match just the
851 signed int comparisons.
852 (ix86_comparison_uns_operator): New predicate to match just the
853 unsigned int comparisons.
855 * doc/invoke.texi (-msse5): Add documentation.
856 (-mfused-madd): Ditto.
858 * doc/extend.texi (x86 intrinsics): Document new SSE5 intrinsics.
860 * config.gcc (i[34567]86-*-*): Include bmmintrin.h and
864 * config/i386/cpuid.h (bit_SSE5): Define SSE5 bit.
866 * config/i386/bmmintrin.h: New file, provide common x86 compiler
867 intrinisics for SSE5.
869 * config/i386/smmintrin.h: Move instructions shared with SSE5 to
872 * config/i386/mmintrin-common.h: New file, to contain common
873 instructions between SSE4.1 and SSE5.
875 * config/i386/netware.c (gen_stdcall_or_fastcall_decoration): Use
876 FOREACH_FUNCTION_ARGS to iterate over the argument list.
877 (gen_regparm_prefix): Ditto.
879 * config/i386/winnt.c (gen_stdcall_or_fastcall_suffix): Use
880 FOREACH_FUNCTION_ARGS to iterate over the argument list. Use
881 prototype_p to determine if a function is prototyped.
883 2009-08-25 Ville Voutilainen <ville.voutilainen@gmail.com>
885 * c-common.c (c_common_reswords) add the alignof keyword,
886 with same RID as __alignof and __alignof__
888 2009-08-25 Anatoly Sokolov <aesok@post.ru>
890 * hooks.h (hook_bool_const_int_const_int_true): Declare.
891 * hooks.c (hook_bool_const_int_const_int_true): New function.
892 * target.h (struct gcc_target): Add can_eliminate field.
893 * target-def.h (TARGET_CAN_ELIMINATE): Define.
894 (TARGET_INITIALIZER): Use TARGET_CAN_ELIMINATE.
895 * ira.c (setup_eliminable_regset): Use can_eliminate target hook.
896 * reload1.c (update_eliminables, init_elim_table): (Ditto.).
897 (elim_table): Revise comment.
898 * system.h (CAN_ELIMINATE): Poison.
899 * defaults.h (CAN_ELIMINATE): Remove.
900 * doc/tm.texi (CAN_ELIMINATE): Revise documentation.
902 * config/alpha/vms.h (CAN_ELIMINATE): Remove macro.
903 * config/alpha/alpha.c (TARGET_CAN_ELIMINATE) [TARGET_ABI_OPEN_VMS]:
905 (alpha_vms_can_eliminate): Declare as static, change return type to
907 * config/alpha/alpha-protos.h (alpha_vms_can_eliminate): Remove.
909 * config/arm/arm.h (CAN_ELIMINATE): Remove macro.
910 * config/arm/arm.c (TARGET_CAN_ELIMINATE): Define macro.
911 (arm_can_eliminate): New function.
913 * config/avr/avr.h (CAN_ELIMINATE): Remove macro.
914 * config/avr/avr.c (TARGET_CAN_ELIMINATE): Define macro.
915 (avr_can_eliminate): Declare as static.
916 * config/avr/avr-protos.h (avr_can_eliminate): Remove.
918 * config/bfin/bfin.h (CAN_ELIMINATE): Remove macro.
919 * config/bfin/bfin.c (TARGET_CAN_ELIMINATE): Define macro.
920 (bfin_can_eliminate): New function.
922 * config/crx/crx.h (CAN_ELIMINATE): Remove macro.
923 * config/crx/crx.c (TARGET_CAN_ELIMINATE): Define macro.
924 (crx_can_eliminate): New function.
926 * config/fr30/fr30.h (CAN_ELIMINATE): Remove macro.
927 * config/fr30/fr30.c (TARGET_CAN_ELIMINATE): Define macro.
928 (fr30_can_eliminate): New function.
930 * config/frv/frv.h (CAN_ELIMINATE): Remove macro.
931 * config/frv/frv.c (TARGET_CAN_ELIMINATE): Define macro.
932 (frv_can_eliminate): New function.
934 * config/h8300/h8300.h (CAN_ELIMINATE): Remove macro.
935 * config/h8300/h8300.c (TARGET_CAN_ELIMINATE): Define macro.
936 (h8300_can_eliminate): New function.
938 * config/i386/i386.h (CAN_ELIMINATE): Remove macro.
939 * config/i386/i386.c (TARGET_CAN_ELIMINATE): Define macro.
940 (i386_can_eliminate): Declare as static, change return type to bool.
941 * config/i386/i386-protos.h (i386_can_eliminate): Remove.
943 * config/ia64/ia64.h (CAN_ELIMINATE): Remove macro.
944 * config/ia64/ia64.c (TARGET_CAN_ELIMINATE): Define macro.
945 (ia64_can_eliminate): New function.
947 * config/iq2000/iq2000.h (CAN_ELIMINATE): Remove macro.
948 * config/iq2000/iq2000.c (TARGET_CAN_ELIMINATE): Define macro.
949 (iq2000_can_eliminate): New function.
951 * config/m32r/m32r.h (CAN_ELIMINATE): Remove macro.
952 * config/m32r/m32r.c (TARGET_CAN_ELIMINATE): Define macro.
953 (m32r_can_eliminate): New function.
955 * config/m68hc11/m68hc11.h (CAN_ELIMINATE): Remove macro.
956 * config/m68hc11/m68hc11.c (TARGET_CAN_ELIMINATE): Define macro.
957 (m68hc11_can_eliminate): New function.
959 * config/m68k/m68k.h (CAN_ELIMINATE): Remove macro.
960 * config/m68k/m68k.c (TARGET_CAN_ELIMINATE): Define macro.
961 (m68k_can_eliminate): New function.
963 * config/mep/mep.h (CAN_ELIMINATE): Remove macro.
964 * config/mep/mep.c (TARGET_CAN_ELIMINATE): Define macro.
965 (mep_can_eliminate): New function.
967 * config/mips/mips.h (CAN_ELIMINATE): Remove macro.
968 * config/mips/mips.c (TARGET_CAN_ELIMINATE): Define macro.
969 (mips_can_eliminate): New function.
971 * config/rs6000/rs6000.h (CAN_ELIMINATE): Remove macro.
972 * config/rs6000/rs6000.c (TARGET_CAN_ELIMINATE): Define macro.
973 (rs6000_can_eliminate): New function.
975 * config/s390/s390.h (CAN_ELIMINATE): Remove macro.
976 * config/s390/s390.c (TARGET_CAN_ELIMINATE): Define macro.
977 (s390_can_eliminate): Declare as static.
978 * config/s390/s390-protos.h (sparc_can_eliminate): Remove.
980 * config/score/score.h (CAN_ELIMINATE): Remove macro.
981 * config/score/score.c (TARGET_CAN_ELIMINATE): Define macro.
982 (score_can_eliminate): New function.
984 * config/sparc/sparc.h (CAN_ELIMINATE): Remove macro.
985 * config/sparc/sparc.c (TARGET_CAN_ELIMINATE): Define macro.
986 (sparc_can_eliminate): Declare as static.
987 * config/sparc/sparc-protos.h (sparc_can_eliminate): Remove.
989 * config/stormy16/stormy16.h (CAN_ELIMINATE): Remove macro.
990 * config/stormy16/stormy16.c (TARGET_CAN_ELIMINATE): Define macro.
991 (xstormy16_can_eliminate): New function.
993 * config/v850/v850.h (CAN_ELIMINATE): Remove macro.
994 * config/v850/v850.c (TARGET_CAN_ELIMINATE): Define macro.
995 (v850_can_eliminate): New function.
997 2009-08-25 Uros Bizjak <ubizjak@gmail.com>
999 * config/alpha/alpha.md (*cmpdf_ieee_ext[123]): Remove.
1000 (*cmpdf_internal): Enable for all ALPHA_FPTM levels.
1001 (*movdfcc_ext[1234]): Disable for IEEE mode.
1003 2009-08-25 Eric Botcazou <ebotcazou@adacore.com>
1005 * gimplify.c (prepare_gimple_addressable): New static function.
1006 (gimplify_modify_expr_to_memcpy): Invoke it on the RHS before marking
1008 (gimplify_addr_expr): Invoke it similarly on the operand instead of
1009 manually fiddling with it.
1011 2009-08-25 Michael Matz <matz@suse.de>
1013 * expr.h (jumpifnot_1, jumpif_1, do_jump_1): Declare.
1014 * dojump.c (do_jump_by_parts_greater): Take two operands instead of
1016 (do_jump_by_parts_equality, do_compare_and_jump): Ditto.
1017 (jumpifnot_1, jumpif_1): New wrappers for do_jump_1.
1018 (do_jump): Split out code for simple binary comparisons into ...
1019 (do_jump_1): ... this, taking the individual operands and code.
1020 Change callers to helper function above accordingly.
1021 * expr.c (expand_expr_real_1): Use jumpifnot_1 for simple binary
1024 2009-08-25 Michael Matz <matz@suse.de>
1026 * expr.h (struct separate_ops, sepops): New type for passing
1027 around an exploded simple expression.
1028 * optabs.c (expand_widen_pattern_expr, expand_vec_shift_expr):
1029 Use this structure instead of expression tree.
1030 (get_vcond_icode, expand_vec_cond_expr_p): Don't take whole
1031 expression, only its type.
1032 (expand_vec_cond_expr): Take type and individual operands instead
1034 * optabs.h (expand_widen_pattern_expr, expand_vec_cond_expr,
1035 expand_vec_shift_expr): Change prototype accordingly.
1036 * tree-vect-stmts.c (vectorizable_condition): Change call of
1037 expand_vec_cond_expr_p to pass only type.
1038 * expr.c (do_store_flags): Change prototype and implementation
1039 to take an exploded expression.
1040 (expand_expr_real_1): New local ops initialized with details
1041 of the full expression. Use it instead of full
1042 expression in calls to do_store_flags, expand_vec_cond_expr,
1043 expand_widen_pattern_expr and expand_vec_shift_expr.
1045 2009-08-25 Michael Matz <matz@suse.de>
1047 * expr.c (expand_expr_real_1): New local treeop0, treeop1,
1048 treeop2 initialized with first three operands of the full expression.
1049 Substitute all TREE_OPERAND (exp, [012]) calls with them.
1051 2009-08-25 Kai Tietz <kai.tietz@onevision.com>
1053 * gcc/gthr-win32.h (__UNUSED_PARAM): Define, if not already present.
1054 (__gthread_objc_condition_allocate): Mark arguments as unused.
1055 (__gthread_objc_condition_deallocate): Likewise.
1056 (__gthread_objc_condition_wait): Likewise.
1057 (__gthread_objc_condition_broadcast): Likewise.
1058 (__gthread_objc_condition_signal): Likewise.
1059 (__gthread_objc_thread_detach): Cast via INT_PTR to pointer.
1060 (__gthread_objc_thread_id): Likewise.
1062 2009-08-25 Janus Weil <janus@gcc.gnu.org>
1065 * tree-pretty-print.c (print_call_name): Print the correct call name
1066 for procedure pointer components.
1068 2009-08-24 Steve Ellcey <sje@cup.hp.com>
1070 * config/ia64/ia64.c (ia64_promote_function_mode): Call
1071 default_promote_function_mode when not VMS.
1073 2009-08-24 Olivier Hainque <hainque@adacore.com>
1075 * convert.c (convert_to_integer): Don't assume an input pointer is
1076 POINTER_SIZE wide. Fetch from the type instead.
1078 2009-08-24 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
1080 * configure.ac (AC_PREREQ): Bump to 2.64.
1082 2009-08-24 Rafael Avila de Espindola <espindola@google.com>
1084 * gcc.c (standard_exec_prefix_1,standard_exec_prefix_2): Remove.
1085 (process_command): Don't search standard_exec_prefix_1 and
1086 standard_exec_prefix_2.
1088 2009-08-24 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1090 * config/arm/arm.c (output_return_instruction): Handle for
1093 2009-08-24 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1095 * config/arm/arm.c (arm_select_cc_mode): Handle subreg.
1097 2009-08-24 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1099 * config/arm/vfp.md (*arm_movdi_vfp): Mark as predicable.
1100 (*arm_movdf_vfp): Likewise.
1102 2009-08-24 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1104 * config/arm/neon.md (vashl<mode>3): Rename from ashl<mode>3.
1105 (vashr<mode>3): Rename from ashr<mode>3.
1106 (vlshr<mode>3): Rename from lshr<mode>3.
1108 2009-08-24 Kai Tietz <kai.tietz@onevision.com>
1111 * c-format.c (format_wanted_type): Add new member scalar_identity_flag.
1112 (check_format_info_main): Use scalar_identify_flag.
1113 (check_format_types): Check for scalar size identity if
1114 scalar_identify_flag is set.
1115 (printf_length_specs): Extend by new field.
1116 (asm_fprintf_length_specs): Likewise.
1117 (gcc_diag_length_specs): Likewise.
1118 (scanf_length_specs): Likewise.
1119 (strfmon_length_specs): Likewise.
1120 (gcc_gfc_length_specs): Likewise.
1121 * config/i386/msformat-c.c (ms_printf_length_specs): Likewise.
1122 (ms_printf_flag_specs): Likewise.
1123 * c-format.h (format_length_info): Add new member scalar_identity_flag.
1125 2009-08-23 Uros Bizjak <ubizjak@gmail.com>
1128 * config/i386/i386.c (*call_pop_1): Disable for sibling calls.
1129 (*call_value_pop_1): Ditto.
1130 (*sibcall_pop_1): New insn pattern.
1131 (*sibcall_value_pop_1): Ditto.
1133 2009-08-23 Alan Modra <amodra@bigpond.net.au>
1136 * config/rs6000/rs6000.md (rotlsi3_64, ashlsi3_64, lshrsi3_64,
1139 2009-08-23 Alan Modra <amodra@bigpond.net.au>
1142 * fwprop.c (try_fwprop_subst): Allow multiple sets.
1143 (get_reg_use_in): New function.
1144 (forward_propagate_subreg): Propagate through subreg of zero_extend
1147 2009-08-22 Kaz Kojima <kkojima@gcc.gnu.org>
1149 * config/sh/t-sh (TARGET_LIBGCC2_CFLAGS): Define.
1150 * config/sh/t-netbsd (TARGET_LIBGCC2_CFLAGS): Add -mieee.
1152 2009-08-22 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
1154 * configure.ac: Remove --with-datarootdir, --with-docdir,
1155 --with-htmldir switches. No need to call AC_SUBST for
1156 datarootdir, docdir, htmldir any more.
1157 * configure: Regenerate.
1158 * doc/install.texi (Configuration): Document --datarootdir,
1159 --docdir, --htmldir, --pdfdir; update documentation for
1160 --infodir, --mandir.
1161 (Prerequisites): Bump Autoconf version to 2.64, Automake to 1.11,
1164 * aclocal.m4: Regenerate.
1165 * config.in: Regenerate.
1166 * configure: Regenerate.
1168 2009-08-21 Douglas B Rupp <rupp@gnat.com>
1169 Olivier Hainque <hainque@adacore.com>
1171 * config/ia64/ia64.c: Include libfuncs.h.
1172 (TARGET_PROMOTE_FUNCITON_MODE): Define target macro.
1173 (ia64_expand_call): Use reg 25 on VMS.
1174 (ia64_initialize_trampoline): Fix for VMS ABI.
1175 (ia64_function_arg_offset): Always returns 0 when TARGET_ABI_OPEN_VMS.
1176 (ia64_function_arg): Initialize reg 25 on VMS.
1177 Fix OpenVMS ABI issues for varargs.
1178 For OpenVMS, emit the Argument Information register set in the
1179 incoming/sibcall case as well.
1180 (ia64_arg_type): New function.
1181 (ia64_function_arg_advance): Keep track of cum->words.
1182 Fix OpenVMS ABI issues for varargs.
1183 (ia64_function_value): On VMS, promote mode of non-aggregate types.
1184 (ia64_override_options): Set flag_no_common on VMS.
1185 (ia64_init_builtins): Disable FWRITE builtin.
1186 (ia64_asm_output_external): Call DO_CRTL_NAMES.
1187 (ia64_vms_init_libfuncs): Add decc$ routines.
1188 (ia64_vms_valid_pointer_mode): New function.
1189 (ia64_struct_value_rtx): Allways NULL_RTX on VMS.
1190 (ia64_promote_function_mode): New function
1191 * config/ia64/ia64.h (TARGET_ABI_OPEN_VMS): Define as 0 for default.
1192 (LONG_DOUBLE_TYPE_SIZE): Force to 64 on VMS.
1193 (LIBCGC2_LONG_DOUBLE_TYPE_SIZE): Likewise.
1194 (INIT_CUMULATIVE_ARGS): Add atypes for VMS.
1195 (INIT_CUMULATIVE_INCOMING_ARGS): Likewise.
1196 (ASM_OUTPUT_DEF): Use ISDIGIT instead of isdigit.
1197 Suppress trailing '#' if VALUE is numeric.
1198 * config/ia64/vms.h (PROMOTE_FUNCTION_MODE): Remove, code moved to
1199 ia64_promote_function_mode.
1200 (TARGET_VALID_POINTER_MODE): Define.
1202 2009-08-21 Michael Meissner <meissner@linux.vnet.ibm.com>
1205 * config/rs6000/rs6000.c (rs6000_override_options): Use
1206 TARGET_64BIT instead of TARGET_POWERPC64 to set the size of pointers.
1209 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute): Fix
1210 reporting of vector + decimal/boolean/complex error.
1212 2009-08-21 Jakub Jelinek <jakub@redhat.com>
1214 * config/rs6000/rs6000.c (rs6000_init_builtins): Fix type of
1215 __vector double TYPE_DECL.
1217 2009-08-21 Richard Earnshaw <rearnsha@arm.com>
1219 * arm.h (MACHMODE): New define. Include insn-modes.h if available.
1220 (CUMULATIVE_ARGS): Use MACHMODE for declaration of aapcs_vfp_mode.
1221 * arm.c (aapcs_vfp_is_call_or_return_candidate): Change base_mode
1222 to pointer to enum machine_mode. Update all callers as needed.
1224 2009-08-21 Uros Bizjak <ubizjak@gmail.com>
1226 * config/alpha/alpha.md (exception_receiver): Emit alternative
1227 GP load sequence if flag_reorder_blocks_and_partition is set.
1228 (*exception_receiver_2): Also enable when
1229 flag_reorder_blocks_and_partition is set.
1231 2009-08-20 Matt Rice <ratmice@gmail.com>
1232 Diego Novillo <dnovillo@google.com>
1234 * Makefile.in (PLUGIN_HEADERS): Include incpath.h and
1237 2009-08-20 Richard Guenther <rguenther@suse.de>
1239 * c-objc-common.h (LANG_HOOKS_DUP_LANG_SPECIFIC_DECL): Do not define.
1240 * c-tree.h (c_dup_lang_specific_decl): Remove.
1241 (struct lang_decl, struct lang_type): Move definitions ...
1242 * c-lang.h: ... here. New file.
1243 * c-decl.c: Include c-lang.h.
1244 (c_dup_lang_specific_decl): Remove.
1245 * c-typeck.c: Include c-lang.h.
1246 * Makefile.in (c-decl.o): Add c-lang.h dependency.
1247 (c-typeck.o): Likewise.
1248 * c-config-lang.in (gtfiles): Add c-lang.h.
1249 * gengtype.c (get_output_file_with_visibility): Handle c-lang.h
1252 2009-08-20 Uros Bizjak <ubizjak@gmail.com>
1254 * config/alpha/alpha.c (alpha_end_function): Do not clear
1255 crtl->emit structure and free insn locators if cfun->is_thunk is true,
1256 this is now handled in generic code.
1258 2009-08-20 Andreas Krebbel <krebbel1@de.ibm.com>
1260 * config/s390/s390.c (Z10_PREDICT_DISTANCE): New macro.
1261 (s390_z10_fix_long_loop_prediction): New function.
1262 (s390_z10_optimize_cmp): INSN walk moved to callee - s390_reorg.
1263 (s390_reorg): Walk over the INSNs and invoke
1264 s390_z10_fix_long_loop_prediction and s390_z10_optimize_cmp.
1266 2009-08-20 Andreas Krebbel <krebbel1@de.ibm.com>
1268 * config/s390/s390.md ("*brx_stage1_<GPR:mode>", "*brxg_64bit",
1269 "*brx_64bit", "*brx_31bit"): New patterns.
1270 * config/s390/s390.c ('E'): New output modifier.
1272 2009-08-20 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1273 Richard Earnshaw <richard.earnshaw@arm.com>
1275 * config/arm/arm.c (arm_emit_movpair): Handle CONST_INT.
1276 * config/arm/arm.md (*arm_movtas_ze): New pattern for movt.
1278 2009-08-19 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1280 * pa.md (reload_inhi, reload_outhi, reload_inqi, reload_outqi): New
1282 * pa.c (emit_move_sequence): Check if address of operand1 is valid
1283 for mode mode of operand0 when doing secondary reload for SAR.
1285 2009-08-19 Jakub Jelinek <jakub@redhat.com>
1288 * expr.c (expand_expr_real_1) <normal_inner_ref>: Handle all kinds
1289 of CONCAT, not just bitpos 0 bitsize size of the whole CONCAT.
1291 2009-08-19 Jason Merrill <jason@redhat.com>
1293 * doc/invoke.texi (C++ Dialect Options): Note change of minimum
1294 supported template depth in C++0x.
1296 2009-08-19 Jakub Jelinek <jakub@redhat.com>
1298 * config/rs6000/rs6000.c (rs6000_output_mi_thunk): Don't call
1299 free_after_compilation.
1300 * config/score/score7.c (score7_output_mi_thunk): Likewise.
1301 * config/score/score3.c (score3_output_mi_thunk): Likewise.
1302 * config/ia64/ia64.c (ia64_output_mi_thunk): Likewise.
1303 * config/mips/mips.c (mips_output_mi_thunk): Likewise.
1304 * config/sh/sh.c (sh_output_mi_thunk): Likewise.
1305 * config/m68k/m68k.c (m68k_output_mi_thunk): Likewise.
1306 * config/sparc/sparc.c (sparc_output_mi_thunk): Likewise.
1308 2009-08-19 Ian Lance Taylor <iant@google.com>
1310 * doc/md.texi (Insn Canonicalizations): Correct canonicalization
1311 of (plus (mult (neg B) C) A).
1313 2009-08-18 Michael Matz <matz@suse.de>
1315 * omp-low.c (optimize_omp_library_calls): Use types_compatible_p
1316 instead of comparing TYPE_MAIN_VARIANT for equality.
1317 * tree-vect-patterns.c (vect_recog_dot_prod_pattern,
1318 vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern): Ditto.
1319 * tree-vect-loop.c (vect_is_simple_reduction): Ditto.
1320 * gimplify.c (goa_lhs_expr_p): Ditto and use
1321 STRIP_USELESS_TYPE_CONVERSION.
1323 2009-08-18 Michael Matz <matz@suse.de>
1325 * tree-ssa-structalias.c (create_variable_info_for): Also mark
1326 first field in a struct.
1327 (intra_create_variable_infos): Don't deal with flag_argument_noalias.
1329 2009-08-18 Uros Bizjak <ubizjak@gmail.com>
1331 * config/alpha/alpha.c (alpha_output_mi_thunk_osf): Allocate insn
1332 locators before emit_insn is called. Remove assert that
1334 (alpha_end_function): Clear crtl->emit structure and free insn
1335 locators if cfun->is_thunk is true.
1337 2009-08-18 Jason Merrill <jason@redhat.com>
1339 * config/elfos.h (ASM_DECLARE_OBJECT_NAME): Use gnu_unique_object
1341 * configure.ac: Test for it.
1342 * configure, config.in: Regenerate.
1343 * doc/install.texi: Document --enable-gnu-unique-object.
1345 2009-08-18 Richard Guenther <rguenther@suse.de>
1348 * builtins.c (fold_builtin_pow): Fold pow(pow(x,y),z) to
1349 pow(x,y*z) only if x is nonnegative.
1351 2009-08-18 Jakub Jelinek <jakub@redhat.com>
1353 * bb-reorder.c (fix_up_fall_thru_edges): Only call invert_jump
1357 * config/rs6000/rs6000.c (rs6000_legitimize_address): For
1358 [DT][FDI]mode ensure the offset isn't 4/8/12 bytes below 0x8000.
1360 2009-08-17 DJ Delorie <dj@redhat.com>
1362 * config/m32c/m32c.md (UNS_FSETB, UNS_FREIT): New.
1363 * config/m32c/prologue.md (epilogue_freit): New.
1365 * config/m32c/m32c.c (m32c_function_needs_enter): Add prototype.
1366 (bank_switch_p): Likewise.
1367 (fast_interrupt_p): Likewise.
1368 (interrupt_p): Likewise.
1369 (m32c_conditional_register_usage): Round memregs size up.
1370 (need_to_save): We only need to save $a0 when we use ENTER.
1371 (interrupt_p): Check for fast_interrupt too.
1372 (bank_switch_p): New.
1373 (fast_interrupt_p): New.
1374 (m32c_attribute_table): Add bank_switch and fast_interrupt.
1375 (m32c_emit_prolog): Support bank switching and fast interrupts.
1376 * doc/extend.texi (Function Attributes): Add bank_switch and
1379 2009-08-17 Douglas B Rupp <rupp@gnat.com>
1381 * config/alpha/alpha.c (vms_valid_pointer_mode): New function.
1382 * config/alpha/vms.h (TARGET_VALID_POINTER_MODE): Define.
1384 2009-08-16 Douglas B Rupp <rupp@gnat.com>
1386 * doc/invoke.texi (Target options): Add new option list for IA-64/VMS.
1387 (menu): Add IA-64/VMS Options.
1388 (IA-64/VMS Options): Likewise.
1390 2009-08-16 Richard Sandiford <rdsandiford@googlemail.com>
1393 * config/mips/mips.md (*lwxs): Use :P for pointer values.
1395 2009-08-16 Richard Sandiford <rdsandiford@googlemail.com>
1397 * config/mips/mips-protos.h (mips_push_asm_switch): New function.
1398 (mips_pop_asm_switch): Likewise.
1399 * config/mips/mips.c (set_noreorder, set_nomacro, set_noat): Replace
1401 (mips_noreorder, mips_nomacro, mips_noat): ...these new variables.
1402 (mips_push_asm_switch_1, mips_pop_asm_switch_1): New functions.
1403 (mips_push_asm_switch, mips_pop_asm_switch): Likewise.
1404 (mips_print_operand_punctuation): Use them. Check mips_noreorder
1405 instead of set_noreorder.
1406 (mips_output_function_prologue): Use the new functions.
1407 (mips_output_function_epilogue): Likewise.
1408 (mips_need_noat_wrapper_p): New function, split out from...
1409 (mips_final_prescan_insn, mips_final_postscan_insn): ...here.
1410 Use mips_push_asm_switch and mips_pop_asm_switch.
1411 * config/mips/mips.h (FUNCTION_PROFILER): Use mips_push_asm_switch
1412 and mips_pop_asm_switch.
1413 (ASM_OUTPUT_REG_POP): Likewise.
1414 (DBR_OUTPUT_SEQEND): Remove boilerplate comment.
1415 Use mips_pop_asm_switch.
1416 (mips_asm_switch): New structure.
1417 (set_noreorder, set_nomacro): Replace with...
1418 (mips_noreorder, mips_nomacro, mips_noat): ...these new variables.
1419 * config/mips/mips.md (fix_truncdfsi2_macro): Use mips_nomacro
1420 instead of set_nomacro.
1421 (fix_truncsfsi2_macro): Likewise.
1422 (cprestore): Likewise.
1423 (hazard): Use mips_noreorder instead of set_noreorder.
1424 * config/mips/sdemtk.h (FUNCTION_PROFILER): As for mips.h.
1426 2009-08-16 Uros Bizjak <ubizjak@gmail.com>
1428 * config/alpha/alpha.c (alpha_end_function): Handle NULL_RTX returned
1429 from prev_active_insn.
1431 2009-08-16 Anatoly Sokolov <aesok@post.ru>
1433 * config/avr/avr.h (AVR_HAVE_8BIT_SP): New macros.
1434 * config/avr/avr.c (avr_override_options): Initialize
1435 avr_current_arch variable.
1436 (avr_cpu_cpp_builtins): Define __AVR_HAVE_8BIT_SP__ or
1437 __AVR_HAVE_16BIT_SP__ according to the device type.
1438 (expand_prologue, output_movhi): Use AVR_HAVE_8BIT_SP instead of
1440 (expand_epilogue): Use correct QI mode frame pointer for tiny stack.
1441 Use AVR_HAVE_8BIT_SP instead of TARGET_TINY_STACK.
1443 2009-08-16 Dodji Seketeli <dodji@redhat.com>
1446 * gcc/dwarf2out.c (gen_inlined_subroutine_die): Concentrate on
1447 generating inlined subroutine die only. We shouldn't be
1448 called for anything else.
1449 (gen_block_die): Don't generate inline subroutine debug info for
1452 2009-08-15 Sebastian Pop <sebastian.pop@amd.com>
1454 * graphite-poly.c (print_pbb): Print PBB index.
1456 2009-08-15 Sebastian Pop <sebastian.pop@amd.com>
1459 * graphite-interchange.c (ppl_max_for_le): Moved...
1460 * graphite-poly.c (pbb_number_of_iterations): Call ppl_max_for_le.
1461 * graphite-ppl.c (ppl_max_for_le): ... here. Correct the use of
1462 ppl_Pointset_Powerset_C_Polyhedron_maximize.
1463 * graphite-ppl.h (ppl_max_for_le): Declared.
1465 2009-08-14 Olatunji Ruwase <tjruwase@google.com>
1467 * doc/extend.texi (Symbol-Renaming Pragmas): redefine_extname is
1468 supported on all platforms.
1469 * target.h (struct gcc_target): Remove handle_pragma_redefine_extname.
1470 * c-cppbuiltin.c: Remove use of targetm.handle_pragma_redefine_extname.
1471 * c-pragma.c: Likewise.
1472 * target-def.h (TARGET_INITIALIZER): Remove
1473 TARGET_HANDLE_PRAGMA_REDEFINE_EXTNAME.
1474 * config/sol2.h: Remove use of TARGET_HANDLE_PRAGMA_REDEFINE_EXTNAME.
1476 2009-08-14 Douglas B Rupp <rupp@gnat.com>
1478 * config/ia64/fde-vms.c: New file.
1479 * config/ia64/fde-glibc.c (_Unwind_FindTableEntry): Add dummy arg.
1480 * config/ia64/unwind-ia64.c (UNW_ accessors): Move to unwind-ia64.h
1481 (MD_UNW_COMPATIBLE_PERSONALITY_P): Provide default.
1482 (uw_frame_state_for): Only register a personality routine if it is
1483 known to be compatible with our expectations.
1484 (_Unwind_FindEnclosingFunction, uw_frame_state_for):
1485 Declare unw_table_entry stack variable and
1486 mod all calls to _Unwind_FindTableEntry to add arg.
1487 * config/ia64/unwind-ia64.h (UNW_ accessors): Move here.
1488 (_Unwind_FindTableEntry): Add arg to prototype.
1490 2009-08-14 Eric Botcazou <ebotcazou@adacore.com>
1492 * config/ia64/unwind-ia64.c (struct _Unwind_Context): Add new
1493 field 'signal_pfs_loc'.
1494 (uw_frame_state_for): Remove duplicate code dealing with leaf
1495 procedures without unwind info.
1496 If in the frame after unwinding through a signal handler, restore
1497 the AR.PFS register instead of the CFM if AR.PFS has not been saved.
1498 * config/ia64/linux-unwind.h (ia64_fallback_frame_state): Do not set
1499 'pfs_loc' to the AR.PFS location in the signal context; instead
1500 set 'signal_pfs_loc'.
1501 Manually generate the unwind info for the AR.PFS register.
1502 (ABI_MARKER_OLD_LINUX_SIGTRAMP, ABI_MARKER_OLD_LINUX_INTERRUPT,
1503 ABI_MARKER_LINUX_SIGTRAMP, ABI_MARKER_LINUX_INTERRUPT): Define.
1504 (ia64_handle_unwabi): Test 'fs->unwabi' against them.
1505 Do not set 'pfs_loc' to the AR.PFS location in the signal context;
1506 instead set 'signal_pfs_loc'.
1507 Remove code preventing the AR.PFS register from being restored
1508 from the signal context.
1510 2009-08-14 Douglas B Rupp <rupp@gnat.com>
1511 Tristan Gingold <gingold@adacore.com>
1513 * config.gcc (ia64-hp-*vms*): Insert ia64/t-ia64 in tmake_file.
1514 * config/ia64/t-vms: New file.
1515 * config/ia64/vms64.h: New file.
1516 * config/ia64/vms.h: New file.
1517 * config/ia64/vms-crtinit.asm: New file.
1518 * config/ia64/vms_symvec_libgcc_s.opt: New file.
1519 * config/ia64/vms-unwind.h: New file.
1521 2009-08-14 Uros Bizjak <ubizjak@gmail.com>
1523 * config/alpha/alpha.c (alpha_emit_conditional_move): Handle
1526 2009-08-14 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1529 * builtins.c (do_mpc_arg2): Make extern, define for any MPC version.
1531 * real.h (do_mpc_arg2): ... here.
1532 * fold-const.c (const_binop): Use MPC for complex MULT_EXPR
1535 2009-08-14 Rafael Avila de Espindola <espindola@google.com>
1537 * final.c (add_debug_prefix_map): Don't use GC memory for
1538 old_prefix and new_prefix.
1540 2009-08-14 Richard Guenther <rguenther@suse.de>
1542 * ipa-prop.c (compute_complex_pass_through): If we cannot
1543 compute a non-varying offset for IPA_JF_ANCESTOR punt.
1545 2009-08-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1547 * c-lex.c (c_lex_with_flags): Increase size of local variable
1548 to avoid memory clobber.
1550 2009-08-14 Paolo Bonzini <bonzini@gnu.org>
1553 * config/i386/i386.c (ix86_fp_comparison_strategy):
1554 Only enable/disable sahf at function granularity.
1556 2009-08-14 Hans-Peter Nilsson <hp@axis.com>
1558 PR rtl-optimization/41064
1559 * reload1.c (reload_as_needed): Don't call extract_insn
1560 for known invalid replacements after calling
1561 validate_replace_rtx_group and verify_changes.
1563 2009-08-14 Uros Bizjak <ubizjak@gmail.com>
1566 * config/i386/sse.md (SSEMODE124C8): New mode iterator.
1567 (vcond<SSEMODEF2P:mode>): Assert that operation is supported by
1568 ix86_expand_fp_vcond.
1569 (vcond<SSEMODE124C8:mode>): Use SSEMODE124C8 instead of SSEMODE124.
1570 Assert that operation is supported by ix86_expand_int_vcond.
1571 (vcondu<SSEMODE124C8:mode>): Ditto.
1573 2009-08-13 DJ Delorie <dj@redhat.com>
1575 * config/i386/djgpp-stdint.h: New.
1576 * config.gcc (djgpp): Use it.
1578 2009-08-13 Kaz Kojima <kkojima@gcc.gnu.org>
1580 * config/sh/sh.c (sh_override_options): When flag_exceptions or
1581 flag_unwind_tables is on, turn flag_reorder_blocks_and_partition off.
1583 2009-08-13 Ghassan Shobaki <ghassan.shobaki@amd.com>
1585 * tree-ssa-loop-prefetch.c
1586 (prune_ref_by_group_reuse): Enhance probabilistic analysis
1587 for long-stride pruning.
1588 (compute_miss_rate): New function to compute the probability
1589 that two memory references access different cache lines.
1591 2009-08-13 Dave Korn <dave.korn.cygwin@gmail.com>
1593 * gcc/config/i386/cygwin.h (LINK_SPEC): Add --enable-auto-image-base.
1595 2009-08-13 Richard Guenther <rguenther@suse.de>
1598 * tree-ssa-ccp.c (ccp_fold): When folding pointer additions
1599 use the constant pointer type.
1600 * gimplify.c (canonicalize_addr_expr): Canonicalize independent
1601 of CV qualifiers on the target pointer type.
1602 * tree-ssa.c (useless_type_conversion_p): Move incomplete pointer
1603 conversion check before restrict check.
1605 2009-08-12 Kaz Kojima <kkojima@gcc.gnu.org>
1608 * config/sh/sh.md (reload_outdf__RnFRm+4): Fix thinko.
1610 2009-08-12 Kaz Kojima <kkojima@gcc.gnu.org>
1612 * config/sh/sh.c (sh_promote_function_mode): Add ATTRIBUTE_UNUSED.
1614 2009-08-12 Richard Guenther <rguenther@suse.de>
1616 PR tree-optimization/41011
1617 * ipa-cp.c (ipcp_lattice_from_jfunc): Deal with failing fold
1618 and reference constructing.
1620 2009-08-12 Xinliang David Li <davidxl@google.com>
1622 PR tree-optimization/41012
1623 * tree-flow.h : New external interface.
1624 * gimple-low.c (check_call_arg): Change to public function.
1625 Remove argument mismatch check in lowering.
1626 * tree-inline.h (tree_can_inline_p): Interface change.
1627 * tree-inline.c (tree_can_inline_p): Fold argument mismatch check
1629 * ipa-inline.c (cgraph_decide_inlining_of_small_functions):
1630 Call change to tree_can_inline_p function.
1631 (cgraph_decide_inlining_incrementally): Ditto.
1633 2009-08-12 Richard Sandiford <rdsandiford@googlemail.com>
1635 PR tree-optimization/41031
1636 * tree-outof-ssa.c (insert_value_copy_on_edge): Use promote_decl_mode
1637 on the partition variable rather than promote_mode on the source
1638 type. Assert that the partition variable's type has the same
1639 mode as the source value's.
1641 2009-08-12 Paolo Bonzini <bonzini@gnu.org>
1643 * doc/tm.texi (TARGET_PROMOTE_FUNCTION_MODE): Add documentation
1644 for for_return == 2.
1645 * function.c (assign_parm_setup_reg): Use for_return == 2, improve
1647 * calls.c (expand_call): Fix typo.
1648 * explow.c (promote_decl_mode): Use for_return == 2 for RESULT_DECL
1650 * stmt.c (expand_value_return): Use promote_function_mode to copy out
1652 * targhooks.c (default_promote_function_mode): Handle for_return == 2.
1653 * config/cris/cris.c (cris_promote_function_mode): Likewise.
1654 * config/mmix/mmix.c (mmix_promote_function_mode): Likewise.
1655 * config/pa/pa.c (pa_promote_function_mode): Likewise.
1657 2009-08-12 Andrew Haley <aph@redhat.com>
1659 * config/arm/arm.c (arm_init_libfuncs): Add __sync_synchronize.
1661 2009-08-12 Sebastian Pop <sebastian.pop@amd.com>
1664 * graphite.c: Remove pragma GCC diagnostic warning "-Wc++-compat".
1666 2009-08-12 Richard Guenther <rguenther@suse.de>
1668 * alias.c (get_alias_set): Honor TYPE_STRUCTURAL_EQUALITY_P.
1669 * gimplify.c (gimplify_modify_expr): Do not use
1670 lang_hooks.types_compatible_p.
1671 * tree-ssa.c (useless_type_conversion_p): For aggregates
1672 just return false if the canonical types differ.
1674 2009-08-12 Sebastian Pop <sebastian.pop@amd.com>
1677 * sese.c (convert_for_phi_arg): New.
1678 (add_guard_exit_phis): Use convert_for_phi_arg.
1680 2009-08-12 Sebastian Pop <sebastian.pop@amd.com>
1682 * graphite-sese-to-poly.c (pdr_add_data_dimensions): Dont add
1683 unknown subscript upper bounds.
1685 2009-08-12 Sebastian Pop <sebastian.pop@amd.com>
1686 Pranav Garg <pranav.garg2107@gmail.com>
1688 * graphite-interchange.c (gather_access_strides): Removed.
1689 (ppl_max_for_le): New.
1690 (build_linearized_memory_access): New.
1691 (memory_stride_in_loop): New.
1692 (pbb_interchange_profitable_p): Reimplemented.
1693 * graphite-ppl.h (ppl_new_id_map): New.
1694 (ppl_interchange): New.
1696 2009-08-12 Sebastian Pop <sebastian.pop@amd.com>
1698 * graphite-interchange.c (compute_subscript): Removed.
1699 (compute_array_size_cstr): Removed.
1700 (compute_array_size_poly): Removed.
1701 (compute_array_size): Removed.
1702 (gather_access_strides_poly): Removed.
1703 (gather_access_strides): Empty.
1705 2009-08-12 Sebastian Pop <sebastian.pop@amd.com>
1707 * graphite-dependences.c (dependence_polyhedron_1): Replace
1708 pdr_nb_subscripts with PDR_NB_SUBSCRIPTS.
1709 (graphite_legal_transform_dr): Same.
1710 (graphite_carried_dependence_level_k): Same.
1711 * graphite-poly.c (new_poly_dr): Add a parameter nb_subscripts.
1712 Initialize PDR_NB_SUBSCRIPTS.
1713 (print_pdr_access_layout): Replace pdr_nb_subscripts with
1715 * graphite-poly.h (struct poly_dr): Add nb_subscripts field.
1716 (PDR_NB_SUBSCRIPTS): New.
1717 (pdr_nb_subscripts): Removed.
1718 (pdr_dim): Simplified.
1719 * graphite-sese-to-poly.c (build_poly_dr): Replace pdr_nb_subscripts
1720 with PDR_NB_SUBSCRIPTS.
1722 2009-08-12 Sebastian Pop <sebastian.pop@amd.com>
1724 * graphite-interchange.c (compute_array_size): Remove use of
1726 * graphite-poly.c (new_poly_dr): Remove argument data_container.
1727 Do not initialize PDR_DATA_CONTAINER.
1728 (print_pdr): Do not print PDR_DATA_CONTAINER.
1729 * graphite-poly.h (struct poly_dr): Remove data_container field.
1730 (PDR_DATA_CONTAINER): Removed.
1731 * graphite-sese-to-poly.c (pdr_add_data_dimensions): Remove use of
1733 (build_poly_dr): Same.
1735 2009-08-12 Konrad Trifunovic <konrad.trifunovic@gmail.com>
1736 Sebastian Pop <sebastian.pop@amd.com>
1738 * graphite-dependences.c (graphite_legal_transform_dr): Work on a
1739 copy of the dependence polyhedron. Free the temporary objects.
1740 (graphite_carried_dependence_level_k): Free unused objects before
1743 * testsuite/gcc.dg/graphite/interchange-1.c: XFAILed.
1744 * testsuite/gcc.dg/graphite/interchange-2.c: XFAILed.
1745 * testsuite/gcc.dg/graphite/interchange-3.c: XFAILed.
1746 * testsuite/gcc.dg/graphite/interchange-4.c: XFAILed.
1747 * testsuite/gcc.dg/graphite/interchange-7.c: XFAILed.
1749 2009-08-12 Sebastian Pop <sebastian.pop@amd.com>
1751 * graphite-blocking.c (scop_do_strip_mine): Call store_scattering.
1752 Early return without analyzing the data dependences if no
1753 transform has been done. Call restore_scattering if the transform
1755 (graphite-interchange.c): Same.
1756 * graphite-poly.c (print_scattering_function): Test for
1758 (graphite_read_transforms): Initialize PBB_TRANSFORMED.
1759 (apply_poly_transforms): Do not gcc_assert that
1760 the transform is legal.
1761 (new_poly_bb): Initialize PBB_TRANSFORMED, PBB_SAVED and PBB_ORIGINAL.
1762 Do not initialize PBB_NB_SCATTERING_TRANSFORM, PBB_NB_LOCAL_VARIABLES,
1763 PBB_TRANSFORMED_SCATTERING, and PBB_ORIGINAL_SCATTERING.
1764 (free_poly_dr): Free PBB_TRANSFORMED, PBB_SAVED, and PBB_ORIGINAL.
1765 * graphite-poly.h (struct poly_scattering): New.
1766 (struct poly_bb): Add original, transformed, and saved fields.
1767 Remove transformed_scattering, original_scattering,
1768 nb_local_variables and nb_scattering_transform fields.
1769 (PBB_ORIGINAL, PBB_TRANSFORMED, PBB_SAVED): New.
1770 (poly_scattering_new): New.
1771 (poly_scattering_free): New.
1772 (poly_scattering_copy): New.
1773 (store_scattering_pbb): New.
1774 (store_scattering): New.
1775 (restore_scattering_pbb): New.
1776 (restore_scattering): New.
1777 * graphite-sese-to-poly.c (build_pbb_scattering_polyhedrons):
1778 Initialize PBB_TRANSFORMED and PBB_ORIGINAL.
1780 2009-08-12 Sebastian Pop <sebastian.pop@amd.com>
1782 * graphite-poly.c (print_pbb): Add parentheses in the pretty print.
1785 2009-08-12 Sebastian Pop <sebastian.pop@amd.com>
1787 * Makefile.in (graphite.o): Depends on PREDICT_H.
1788 * graphite.c: Include predict.h.
1789 (graphite_finalize): Call tree_estimate_probability.
1790 * predict.c (predict_loops): Do not call scev_initialize and
1792 (tree_estimate_probability_bb): New.
1793 (tree_estimate_probability): Do not initialize loops: move that
1794 code to the driver. Call tree_estimate_probability_bb.
1795 (tree_estimate_probability_driver): New.
1796 (pass_profile): Use tree_estimate_probability_driver.
1797 * predict.h (tree_estimate_probability): Declared.
1799 2009-08-12 Sebastian Pop <sebastian.pop@amd.com>
1801 * graphite-clast-to-gimple.c (gloog): Add time to TV_GRAPHITE_CODE_GEN.
1802 * graphite-dependences.c (graphite_legal_transform): Add time to
1803 TV_GRAPHITE_DATA_DEPS.
1804 (dependency_between_pbbs_p): Same.
1805 * timevar.def (TV_GRAPHITE_DATA_DEPS, TV_GRAPHITE_CODE_GEN): New.
1807 2009-08-12 Andrey Belevantsev <abel@ispras.ru>
1809 PR rtl-optimization/41033
1810 * alias.c (nonoverlapping_component_refs_p): Punt when strict
1811 aliasing is disabled.
1813 2009-08-11 Adam Nemet <anemet@caviumnetworks.com>
1815 * config/mips/predicates.md (qi_mask_operand, hi_mask_operand,
1816 si_mask_operand, and_load_operand, low_bitmask_operand,
1817 and_reg_operand, and_operand): New predicates.
1818 * config/mips/constraints.md (Yb, Yh, Yw, Yz): New constraints.
1819 * config/mips/mips.c (and_operands_ok): New function.
1820 * config/mips/mips-protos.h (and_operands_ok): Declare it.
1821 * config/mips/mips.md (move_type): Add ext_ins and logical.
1822 (type): Handle them.
1823 (and<mode>3): Use and_reg_operand as the second operand's predicate.
1824 (*and<mode>3): Add alternatives for lbu, lhu, lwu, <d>ext and
1825 shift_shift. Remove commutative constraint modifier.
1826 (*and<mode>3_mips16): Add alternatives for lbu, lhu, lwu and
1828 (*clear_upper32_dext): Remove define_insn_and_split.
1829 (*clear_upper32): Turn this define_insn_and_split ...
1830 (splitter for ANDing register with 0xffff_ffff): .. into this.
1832 2009-08-11 Adam Nemet <anemet@caviumnetworks.com>
1834 * combine.c (try_widen_shift_mode): Factor out code to check if an
1835 integer constant is a low-order bitmask from here ...
1836 * rtlanal.c (low_bitmask_len): ... to here.
1837 * rtl.h (low_bitmask_len): Declare.
1839 2009-08-11 Uros Bizjak <ubizjak@gmail.com>
1842 * config/alpha/alpha.md (addsi3): Remove expander.
1843 (addsi3): Rename from *addsi3_internal insn pattern.
1844 (subsi3): Remove expander.
1845 (subsi3): Rename from *subsi3_internal insn pattern.
1847 2009-08-11 Douglas B Rupp <rupp@gnat.com>
1849 * config/alpha/alpha.c (alpha_init_builtins): Nullify FWRITE and
1852 2009-08-11 Vasiliy Fofanov <fofanov@adacore.com>
1853 Eric Botcazou <botcazou@adacore.com>
1854 Douglas B Rupp <rupp@gnat.com>
1856 * config/alpha/alpha.c (alpha_return_in_memory): On VMS, ensure
1857 that records that fit in 64 bits are returned by immediate value,
1858 as required by OpenVMS Calling Standard.
1859 (function_value): Adjust for above modification.
1860 (alpha_va_start) <TARGET_ABI_OPEN_VMS>: Use
1861 virtual_incoming_args_rtx as base object, not next_arg.
1862 * config/alpha/vms.h: (DEFAULT_PCC_STRUCT_RETURN): Define as 0.
1864 2009-08-11 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1866 * reload.c (find_reloads_subreg_address): Check the original
1867 req_equiv_mem address to detect the case where an address is
1868 not valid in the outer mode.
1870 2009-08-11 Richard Guenther <rguenther@suse.de>
1873 * builtins.c (gimplify_va_arg_expr): Do not call SET_EXPR_LOCATION.
1875 2009-08-10 Douglas B Rupp <rupp@gnat.com>
1877 * config/alpha/vms.h (OPTIMIZATION_OPTIONS): Remove
1878 (OVERRIDE_OPTIONS): Incorporate removed OPTIMIZATION_OPTIONS.
1880 2009-08-10 Olivier Hainque <hainqueu@adacore.com>
1881 Douglas B Rupp <rupp@gnat.com>
1883 * config/alpha/alpha.c (alpha_sa_size): Force procedure type to
1884 PT_STACK when frame_pointer_needed on OpenVMS.
1885 (alpha_pv_save_size, alpha_using_fp): Remove.
1886 (alpha_vms_can_eliminate): New function. Support for CAN_ELIMINATE
1887 with proper processing for PT_NULL.
1888 (alpha_vms_initial_elimination_offset): New function. Support for
1889 INITIAL_ELIMINATION_OFFSET with proper processing for PT_NULL.
1890 (alpha_sa_size): Force procedure type to PT_STACK when
1891 frame_pointer_needed on OpenVMS.
1892 * config/alpha/alpha-protos.h (alpha_pv_save_size): Remove prototype.
1893 (alpha_using_fp): Likewise.
1894 (alpha_vms_can_eliminate): Add prototype.
1895 (alpha_vms_initial_elimination_offset): Likewise.
1896 * config/alpha/vms.h (CAN_ELIMINATE, INITIAL_ELIMINATION_OFFSET):
1897 Call alpha_vms_can_eliminate and alpha_vms_initial_elimination_offset.
1899 2009-08-10 Eric Botcazou <botcazou@adacore.com>
1900 Douglas B Rupp <rupp@gnat.com>
1902 * config/alpha/alpha.c (common_object_handler): New function.
1903 (vms_attribute_table): Declare a single attribute "common_object".
1904 (vms_output_aligned_decl_common): New global function.
1905 (SECTION_VMS_OVERLAY): Delete.
1906 (SECTION_VMS_GLOBAL): Likewise.
1907 (SECTION_VMS_INITIALIZE): Likewise.
1908 (vms_asm_named_section): Remove support for above flags.
1909 (vms_section_type_flags): Delete.
1910 (TARGET_SECTION_TYPE_FLAGS): Likewise.
1911 * config/alpha/alpha-protos.h (vms_output_aligned_decl_common): New.
1912 * config/alpha/vms.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
1913 (ASM_OUTPUT_ALIGNED_DECL_COMMON): New macro.
1915 2009-08-10 SUGIOKA Toshinobu <sugioka@itonet.co.jp>
1918 * longlong.h [__sh__] (udiv_qrnnd): Add T register to clobber list.
1919 (sub_ddmmss): Likewise.
1921 2009-08-10 Andreas Tobler <a.tobler@schweiz.org>
1924 * config/rs6000/freebsd.h: Define SVR4_ASM_SPEC. Adjust copyright
1927 2009-08-10 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1930 * reload1.c (reload_as_needed): Use cancel_changes to completely
1931 undo a failed replacement attempt.
1933 2009-08-10 Richard Guenther <rguenther@suse.de>
1936 * tree-ssa.c (useless_type_conversion_p_1): Fold into ...
1937 (useless_type_conversion_p): ... here. Require pointer targets
1940 2009-08-10 Dodji Seketeli <dodji@redhat.com>
1943 * tree-inline.c (copy_statement_list): The resulting copy shouldn't
1944 loose the original type of the statement list.
1946 2009-08-09 Douglas B Rupp <rupp@gnat.com>
1948 * config/alpha/alpha.c: Include libfuncs.h
1949 (avms_asm_output_extern): New function.
1950 (alpha_init_libfuncs): Init some decc libfuncs.
1951 * config/alpha/alpha-protos.h (avms_asm_output_external): Declare.
1952 * config/alpha/vms.h (ASM_OUTPUT_EXTERNAL): Define.
1953 (DO_CRTL_NAMES): Define.
1955 * config/alpha/vms64.h: (POINTERS_EXTEND_UNSIGNED): Remove undef.
1956 (LONG_TYPE_SIZE): Define.
1957 (TARGET_OS_CPP_BUILTINS): Define with __LONG_POINTERS=1
1958 (SUBTARGET_SWITCHES): Define malloc64 switch.
1959 (TARGET_DEFAULT): Default MASK_MALLOC64 set.
1960 (MASK_RETURN_ADDR): Define.
1961 doc/invoke.texi (mmalloc64): Document switch.
1963 2009-08-09 Olivier Hainque <hainque@adacore.com>
1964 Douglas B Rupp <rupp@gnat.com>
1966 * config/alpha/alpha.c (struct machine_function): New flag for VMS,
1967 uses_condition_handler.
1968 (alpha_expand_builtin_establish_vms_condition_handler): New expander.
1969 (alpha_expand_builtin_revert_vms_condition_handler): New expander.
1970 (enum alpha_builtin): New ALPHA_BUILTIN_REVERT_VMS_CONDITION_HANDLER
1971 and ALPHA_BUILTIN_ESTABLISH_VMS_CONDITION_HANDLER values.
1972 (code_for_builtin): New insn codes for the new alpha_builtins.
1973 (alpha_init_builtins): Register the new functions as BUILT_IN_MD.
1974 (alpha_sa_size): Account for uses_condition_handler.
1975 (alpha_expand_prologue): Likewise.
1976 (alpha_start_function): Likewise.
1977 (alpha_expand_epilogue): Likewise.
1978 * config/alpha/alpha-protos.h: Prototype the new alpha.c builtin
1979 establish/revert expanders.
1980 * config/alpha/alpha.h (DWARF_FRAME_REGNUM): Define.
1981 * config/alpha/alpha.md (builtin_establish_vms_condition_handler):
1982 New expander, resorting to the alpha.c associated function.
1983 (builtin_revert_vms_condition_handler): Likewise.
1984 * config/alpha/vms-gcc_shell_handler.c: New file. Implements
1985 __gcc_shell_handler, the static VMS condition handler used as
1986 an indirection wrapper to the current dynamically established
1988 * config/alpha/vms-unwind.h: Complete rewrite.
1989 * config/alpha/t-vms (LIB2FUNCS_EXTRA): Add vms-gcc_shell_handler.c
1990 * config/alpha/vms.h (MD_UNWIND_SUPPORT):
1992 2009-08-09 Eric Botcazou <botcazou@adacore.com>
1993 Douglas B Rupp <rupp@gnat.com>
1995 * config/alpha/alpha.c (alpha_links): Add 'target' field.
1996 (alpha_need_linkage): Handle aliases. Return function symbol.
1997 (alpha_use_linkage): Rename 'linkage' argument to 'func'.
1998 Use ultimate alias target for the linkage name.
1999 * config/alpha/alpha.md (movmemdi): Use the symbol returned
2000 by alpha_need_linkage for the function symbol.
2001 (setmemdi): Likewise.
2003 2009-08-09 Douglas B Rupp <rupp@gnat.com>
2005 * config/alpha/alpha.c (TARGET_ASM_UNALIGNED_*_OP): Define if on VMS.
2006 * config/alpha/vms.h (OBJECT_FORMAT_ELF): Define.
2007 (ASM_WEAKEN_LABEL): Define.
2008 (CRT_CALL_STATIC_FUNCTION): Define.
2009 (STARTFILE_SPEC): Add crtbegin.o crtbeginS.o.
2010 (ENDFILE_SPEC): Define.
2011 (INIT_SECTION_ASM_OP): Define.
2012 * config/alpha/vms-dwarf2eh.asm (__EH_FRAME_BEGIN__): Remove.
2013 * config/alpha/t-vms (EXTRA_PARTS): Add crtbegin.o crtbeginS.o
2015 (MULTILIB_OSDIRNAMES): Define.
2016 (shlib_version): Define.
2017 (SHLIB_EXT): Define.
2018 (SHLIB_OBJS): Define.
2019 (SHLIB_NAME): Define.
2020 (SHLIB_MULTILIB): Define.
2021 (SHLIB_INSTALL): Define.
2022 (SHLIB_SYMVEC): Define.
2023 (SHLIB_SYMVECX2): Define.
2024 (SHLIB_LINK): Define.
2026 2009-08-09 Douglas B Rupp <rupp@gnat.com>
2028 * config/alpha/alpha.c (alpha_initialize_trampoline):
2029 Initialize VMS trampoline IAW ABI for bounded procedure calls.
2030 (alpha_start_function): Emit transfer address on nested functions
2031 for VMS trampoline call.
2032 * config/alpha/t-vms (LIB2FUNCS_EXTRA): Remove vms_tramp.asm
2033 since no longer used.
2034 * config/alpha/vms-tramp.asm: Remove.
2035 * config/alpha/vms.h (TRAMPOLINE_TEMPLATE): Leave undefined
2036 since now only data initialized at runtime.
2038 2009-08-09 Douglas B Rupp <rupp@gnat.com>
2040 * config/alpha/vms.h (HANDLE_SYSV_PRAGMA): Define.
2041 (LINK_GCC_C_SEQUENCE_SPEC): Define.
2042 (MD_EXEC_PREFIX): Remove, no longer used.
2043 (MD_STARTFILE_PREFIX): Likewise.
2044 (INCLUDE_DEFAULTS): Likewise.
2045 * config/alpha/t-vms:
2046 (vms-dwarf2.o, vms-dwarf2eh.o): Use GCC_FOR_TARGET to compile.
2048 2009-08-09 Richard Guenther <rguenther@suse.de>
2050 PR tree-optimization/41016
2051 * tree-ssa-ifcombine.c (get_name_for_bit_test): Fix tuplification bug.
2052 (operand_precision): Remove.
2053 (integral_operand_p): Likewise.
2054 (recognize_single_bit_test): Adjust.
2056 2009-08-09 Richard Sandiford <rdsandiford@googlemail.com>
2058 * c-common.c (c_fully_fold_internal): Issue a warning if a binary
2059 operation overflows. Likewise non-cast unary arithmetic.
2060 If one arm of a conditional expression is always taken,
2061 inhibit evaluation warnings for the other arm. Likewise inhibit
2062 evaluation warnings for the second && or || operand if the first
2063 operand is enough to determine the result.
2064 * c-typeck.c (build_conditional_expr): Apply the same inhibition
2066 (build_binary_op): Prevent duplicate evaluation warnings.
2068 2009-08-09 Richard Sandiford <rdsandiford@googlemail.com>
2070 * tree-out-of-ssa.c (insert_value_copy_on_edge): If the source
2071 and destination have different modes, Use promote_mode to
2072 determine the signedness of the conversion. Assert that the
2073 promoted source mode matches the destination mode. Don't pass
2074 the destination and destination mode to expand_expr if the source
2075 mode is different. Simplify conversion logic.
2077 2009-08-09 Ira Rosen <irar@il.ibm.com>
2079 PR tree-optimization/41008
2080 * tree-vect-loop.c (vect_is_simple_reduction): Get operands
2081 from condition only in case it's a comparison. Adjust checks.
2083 2009-08-09 Bernd Schmidt <bernd.schmidt@analog.com>
2085 * tree-dfa.c (renumber_gimple_stmt_uids_in_blocks): New function.
2086 * tree-flow.h (renumber_gimple_stmt_uids_in_blocks): Declare it.
2087 * tree-ssa-loop-ivopts.c (comp_cost): Make COST an integer.
2088 (enum iv_position): Add IP_AFTER_USE and IP_BEFORE_USE.
2089 (dump_cand): Handle them.
2090 (struct iv_cand): New members COST_STEP and AINC_USE.
2091 (stmt_after_increment): Likewise.
2092 (stmt_after_inc_pos): Renamed from stmt_after_ip_original_pos. All
2093 callers changed. Use gimple_uid comparison instead of scanning.
2094 (add_candidate_1): When looking for identical candidates, take
2095 AINC_USE into account. Set it for new candidates.
2096 (force_expr_to_var_cost): Cast target_spill_cost to int.
2097 (get_address_cost): New arguments STMT_AFTER_INC and MAY_AUTOINC.
2098 All callers changed. Check for availability of autoinc addressing
2099 modes, both in general for a given mode, and in the specific use case.
2100 (get_computation_cost_at): New argument CAN_AUTOINC. All callers
2102 (get_computation_cost): Likewise.
2103 (autoinc_possible_for_pair, set_autoinc_for_original_candidates,
2104 add_autoinc_candidates): New static functions.
2105 (add_candidate): Call add_autoinc_candidates for candidates based on
2107 (find_iv_candidates): Call set_autoinc_for_original_candidates.
2108 (determine_use_iv_cost_address): If we have an autoinc candidate at
2109 the matching use, verify autoinc is possible and subtract the cost
2110 of the candidate's step from the cost.
2111 (determine_iv_cost): Record the cost of the increment in the COST_STEP
2112 member of the candidate.
2113 (tree_ssa_iv_optimize_loop): Swap the calls to determine_iv_costs and
2114 determine_use_iv_costs. Call renumber_gimple_stmt_uids_in_blocks.
2116 2009-08-09 Douglas B Rupp <rupp@gnat.com>
2118 * config.build (ia64-hp-*vms*): New target.
2119 (alpha64-dec-*vms*,alpha*-dec-*vms*): Fix for config/vms and unify
2121 * config.gcc (ia64-hp-*vms*): New target.
2122 (alpha64-dec-*vms*,alpha*-dec-*vms*): Fix for config/vms and unify
2124 * config.host (ia64-hp-*vms*): New target.
2125 (alpha64-dec-*vms*,alpha*-dec-*vms*): Fix for config/vms and unify
2128 2009-08-08 Richard Guenther <rguenther@suse.de>
2130 PR tree-optimization/40991
2131 * tree-ssa-pre.c (eliminate): Delay purging EH edges.
2133 2009-08-08 Richard Sandiford <rdsandiford@googlemail.com>
2135 * combine.c (gen_lowpart_or_truncate): Exclude CONST_INTs from
2136 mode check. Do truncations in an integer mode.
2137 (force_to_mode): Handle subregs for all mode types. Only do
2138 arithmetic simplifications on integer modes.
2140 2009-08-07 Richard Guenther <rguenther@suse.de>
2142 PR tree-optimization/40999
2143 * tree-ssa-ccp.c (get_symbol_constant_value): Handle CONST_DECLs.
2144 (maybe_fold_reference): Lookup constant initializers.
2145 (fold_gimple_assign): Likewise.
2147 2009-08-07 Richard Guenther <rguenther@suse.de>
2149 * tree-ssa.c (useless_type_conversion_p_1): Only for types
2150 that require structural equality defer to the langhook.
2152 2009-08-07 Martin Jambor <mjambor@suse.cz>
2154 * ipa-prop.h (enum jump_func_type): New value IPA_JF_ANCESTOR, changed
2156 (struct ipa_pass_through_data): New type.
2157 (struct ipa_ancestor_jf_data): New type.
2158 (union jump_func_value): Removed field formal_id, added fields
2159 pass_through and ancestor.
2160 (struct ipa_param_call_note): Changed type of formal_id to int from
2162 * ipa-prop.c (ipa_print_node_jump_functions): Print pass through with
2163 operations jump functions and ancestor jump functions.
2164 (compute_complex_pass_through): New function.
2165 (compute_scalar_jump_functions): Call compute_complex_pass_through,
2166 reflect changes in the jump function strucutre.
2167 (update_jump_functions_after_inlining): Ignore complex pass-through
2168 and ancestor jump functions.
2169 * ipa-cp.c (ipcp_lattice_from_jfunc): Added support for ancestor and
2170 polynomial pass-through with operation jump functions.
2172 2009-08-07 Jakub Jelinek <jakub@redhat.com>
2174 * dwarf2out.c (output_fde): When doing hot/cold partitioning, use
2175 fde->dw_fde_begin as begin label instead of hot/cold label.
2176 Use LLSDAC label instead of LLSDA for second section lsda.
2177 (dwarf2out_do_cfi_startproc): Add SECOND argument. Use LLSDAC
2178 label instead of LLSDA if it is true.
2179 (dwarf2out_begin_prologue, dwarf2out_switch_text_section): Adjust
2181 * except.c (add_call_site, dw2_size_of_call_site_table): Add
2182 SECTION argument. Use it as index into crtl->eh.call_site_record
2184 (dw2_output_call_site_table): Likewise. Add CS_FORMAT argument,
2185 use it to determine how to print table entries instead of using
2186 #ifdef HAVE_AS_LEB128. For SECTION > 0 use hot resp. cold
2187 label instead of normal begin label as base.
2188 (sjlj_assign_call_site_values): Adjust add_call_site caller.
2189 (convert_to_eh_region_ranges): When doing hot/cold partitioning,
2190 ensure no EH range spans between sections and that landing pads
2191 are always in the corresponding section.
2192 (sjlj_size_of_call_site_table, sjlj_output_call_site_table): Adjust
2193 for crtl->eh.call_site_record being an array rather than scalar.
2194 (output_one_function_exception_table): New function, copied
2195 from output_function_exception_table. Adjust
2196 dw2_size_of_call_site_table, dw2_output_call_site_table
2197 callers. For SECOND section use *C suffixed labels.
2198 (output_function_exception_table): Call
2199 output_one_function_exception_table and, when doing hot/cold
2200 partitioning, also another time for the second section.
2201 * opts.c: Include except.h.
2202 (decode_options): Allow -freorder-blocks-and-partition with
2203 exceptions, unless SJLJ or TARGET_UNWIND_INFO.
2204 * Makefile.in (opts.o): Depend on $(EXCEPT_H).
2205 * function.h (struct rtl_eh): Change call_site_record from
2206 scalar into array of 2 elements.
2208 2009-08-07 Martin Jambor <mjambor@suse.cz>
2210 * ipa-prop.c (count_formal_params_1): New function.
2211 (ipa_get_vector_of_formal_parms): New function.
2212 (get_vector_of_formal_parm_types): New function.
2213 (ipa_modify_formal_parameters): New function.
2214 (ipa_modify_call_arguments): New function.
2215 (index_in_adjustments_multiple_times_p): New function.
2216 (ipa_combine_adjustments): New function.
2217 (ipa_dump_param_adjustments): New function.
2218 * ipa-prop.h (struct ipa_parm_adjustment): New type.
2219 (ipa_get_vector_of_formal_parms): Declare.
2220 (ipa_modify_formal_parameters): Declare.
2221 (ipa_modify_call_arguments): Declare.
2222 (ipa_combine_adjustments): Declare.
2223 (ipa_dump_param_adjustments): Declare.
2224 (build_ref_for_offset): Declare.
2225 * Makefile.in (tree-sra.o): Add ipa-prop.h to dependencies.
2226 * tree-sra.c: Include ipa-prop.c.
2227 (build_ref_for_offset): Make public.
2229 2009-08-06 Neil Vachharajani <nvachhar@gmail.com>
2231 * value-prof.c (init_pid_map): Replace xmalloc with XCNEWVEC.
2233 2009-08-06 Thomas Schwinge <tschwinge@gnu.org>
2235 * gcc/doc/extend.texi (__builtin_extract_return_address)
2236 (__builtin_frob_return_address): Document.
2238 2009-08-06 Paul Brook <paul@codesourcery.com>
2240 * config/arm/lib1funcs.asm (ARM_DIV_BODY): Add Thumb-2 implementation.
2241 (udivsi3, aeabi_uidivmod, divsi3, aeabi_idivmod): Only use Thumb-1
2242 implementation on ARMv6-M.
2244 2009-08-06 Richard Earnshaw <rearnsha@arm.com>
2246 * doc/extend.texi (pcs): Document new attribute for ARM.
2248 2009-08-06 Richard Earnshaw <rearnsha@arm.com>
2250 * arm.c (pcs_attribute_args): Comment out unsupported attribute
2253 2009-08-06 Richard Earnshaw <rearnsha@arm.com>
2255 * arm.c (arm_handle_pcs_attribute): Pass the entire name object to
2258 2009-08-06 Richard Earnshaw <rearnsha@arm.com>
2260 * arm.c (arm_handle_pcs_attribute): Use %qE in warning.
2262 2009-08-06 Richard Earnshaw <rearnsha@arm.com>
2264 Merge ARM/hard_vfp_branch to trunk.
2266 2009-08-04 Richard Earnshaw <rearnsha@arm.com>
2268 * arm.c (libcall_eq): New function.
2269 (libcall_hash): New function.
2270 (add_libcall): New function.
2271 (arm_libcall_uses_aapcs_base): New function.
2272 (arm_libcall_value): Use arm_libcall_uses_aapcs_base to check for
2273 libcalls using the base PCS.
2274 (arm_init_cumulative_args): Likewise.
2276 2009-07-20 Joseph Myers <joseph@codesourcery.com>
2278 * config/arm/arm.c (arm_libcall_value, arm_init_cumulative_args):
2279 Use base ABI for conversion libfuncs between HFmode and SFmode.
2281 2009-05-12 Joseph Myers <joseph@codesourcery.com>
2283 * config/arm/arm.c (aapcs_vfp_sub_candidate): Use V2SImode and
2284 V4SImode as representatives of all 64-bit and 128-bit vector
2285 types. Allow vector types without vector modes.
2286 (aapcs_vfp_is_call_or_return_candidate): Handle vector types
2287 without vector modes like BLKmode.
2288 (aapcs_vfp_allocate): Handle TImode for non-TARGET_NEON like
2289 BLKmode. Avoid unsupported vector modes or TImode moves for
2291 (aapcs_vfp_allocate_return_reg): Likewise.
2292 (arm_vector_mode_supported_p): Only support V2SImode, V4HImode and
2293 V8QImode if TARGET_NEON || TARGET_IWMMXT.
2295 2009-05-12 Joseph Myers <joseph@codesourcery.com>
2297 * config/arm/arm.c (arm_handle_pcs_attribute): New.
2298 (arm_get_pcs_model): Pass attribute arguments to
2299 arm_pcs_from_attribute.
2300 (arm_init_cumulative_args): Use base AAPCS for conversions from
2301 floating-point types to DImode.
2302 (arm_attribute_table): Add pcs attribute.
2303 (arm_handle_pcs_attribute): New.
2304 * config/arm/bpabi.h (DECLARE_LIBRARY_RENAMES): When renaming
2305 conversions from floating-point types to DImode, also declare them
2306 to use base AAPCS and declare functions they call to use base
2307 AAPCS and their RTABI names.
2309 2009-05-12 Joseph Myers <joseph@codesourcery.com>
2311 * doc/invoke.texi (-mfloat-abi=@var{name}): Remove statement about
2312 -mfloat-abi=hard not being supported for VFP.
2314 2009-05-11 Kazu Hirata <kazu@codesourcery.com>
2316 * config/sparc/sparc.c (sparc_emit_float_lib_cmp): Pass a libcall
2317 SYMBOL_REF to hard_libcall_value.
2319 2009-03-05 Joseph Myers <joseph@codesourcery.com>
2320 Richard Earnshaw <rearnsha@arm.com>
2322 * config/arm/arm.c (aapcs_layout_arg): Once a co-processor argument
2323 has been put on the stack, all remaining co-processory arguments for
2324 that co-processor also go on the stack.
2326 2009-03-05 Joseph Myers <joseph@codesourcery.com>
2328 * config/arm/arm.c (arm_return_in_memory): Handle returning
2329 vectors of suitable size in registers also for AAPCS case.
2331 2009-01-13 Richard Earnshaw <rearnsha@arm.com>
2333 * doc/tm.texi (TARGET_LIBCALL_VALUE): Add missing end statement.
2335 2008-12-09 Richard Earnshaw <rearnsha@arm.com>
2337 ARM Hard-VFP calling convention
2338 * target-def.h (TARGET_LIBCALL_VALUE): New hook.
2339 * target.h (gcc_target): Add libcall_value to table of call hooks.
2340 * targhooks.h (default_libcall_value): Default implementation.
2341 * targhooks.c (default_libcall_value): Likewise.
2342 * doc/tm.texi (TARGET_LIBCALL_VALUE): Document it.
2343 * optabs.c (expand_unop): Use it.
2344 * expr.h (hard_libcall_value): Pass the function RTX through.
2345 * calls.c (emit_library_call_value_1): Update call to
2347 * explow.c (hard_libcall_value): Use new target hook.
2348 * testsuite/lib/target-supports.exp
2349 (check_effective_target_arm_hard_vfp_ok): New hook.
2350 (check_effective_target_arm_neon_ok): Improve test for neon
2352 * testsuite/gcc.target/arm/eabi1.c: Only run test in base variant.
2353 * config/arm/arm.c: Include cgraph.h
2354 (TARGET_FUNCTION_VALUE): Override default hook.
2355 (arm_pcs_default): New variable.
2356 (arm_override_options): Don't fault hard calling convention with VFP.
2357 Add support for AAPCS variants.
2358 (arm_function_value): Make static. Handle AAPCS variants.
2359 (arm_libcall_value): New function.
2360 (arm_apply_result_size): Handle VFP registers in results.
2361 (arm_return_in_memory): Rework all AAPCS variants; handle hard-vfp
2363 (pcs_attribute_args): New variable.
2364 (arm_pcs_from_attribute): New function.
2365 (arm_get_pcs_model): New function.
2366 (aapcs_vfp_cum_init): New function.
2367 (aapcs_vfp_sub_candidate): New function.
2368 (aapcs_vfp_is_return_candidate): New function.
2369 (aapcs_vfp_is_call_candidate): New function.
2370 (aapcs_vfp_allocate): New function.
2371 (aapcs_vfp_allocate_return_reg): New function.
2372 (aapcs_vfp_advance): New function.
2373 (aapcs_cp_arg_layout): New variable.
2374 (aapcs_select_call_coproc): New function.
2375 (aapcs_select_return_coproc): New function.
2376 (aapcs_allocate_return_reg): New function.
2377 (aapcs_libcall_value): New function.
2378 (aapcs_layout_arg): New function.
2379 (arm_init_cumulative_args): Initialize AAPCS args data.
2380 (arm_function_arg): Handle AAPCS variants using new interface.
2381 (arm_arg_parital_bytes): Likewise.
2382 (arm_function_arg_advance): New function.
2383 (arm_function_ok_for_sibcall): Ensure that sibling calls agree on
2384 calling conventions.
2385 (arm_setup_incoming_varargs): Handle new AAPCS args data.
2386 * arm.h (NUM_VFP_ARG_REGS): Define.
2387 (LIBCALL_VALUE): Update.
2388 (FUNCTION_VALUE): Delete.
2389 (FUNCTION_VALUE_REGNO_P): Add VFP regs.
2390 (arm_pcs): New enum.
2391 (CUMULATIVE_ARGS): New data to support AAPCS argument marshalling.
2392 (FUNCTION_ARG_ADVANCE): Call arm_function_arg_advance.
2393 (FUNCTION_ARG_REGNO_P): Add VFP regs.
2394 * arm-protos.h (arm_function_arg_advance): Add.
2395 (aapcs_libcall_value): Add.
2396 (arm_function_value): Delete.
2398 2009-08-06 Uros Bizjak <ubizjak@gmail.com>
2399 H.J. Lu <hongjiu.lu@intel.com>
2402 * config/i386/i386.c (standard_sse_mode_p): Remove.
2403 (standard_sse_constant_p): Return 2 for integer mode
2404 vector_all_ones_operand when SSE2 is enabled.
2405 (standard_sse_constant_opcode)<case 2>: Always return [v]pcmpeqd.
2406 (ix86_expand_vector_move): Do not check for negative values from
2407 standard_sse_constant_p.
2409 2009-08-06 Richard Guenther <rguenther@suse.de>
2411 * tree-ssa.c (useless_type_conversion_p_1): Make function and
2412 array type comparisons frontend independent.
2413 * Makefile.in (tree-ssa.o): Add $(TARGET_H) dependency.
2414 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Always fill
2415 out array reference lower bound and element size operands.
2416 (ao_ref_init_from_vn_reference): Properly compute the offset
2417 for ARRAY_RANGE_REF.
2418 (vn_reference_fold_indirect): Fill out array reference lower
2419 bound and element size operands.
2420 * tree-ssa-pre.c (phi_translate_1): Fail if we have to translate
2421 a non gimple valued reference operand which can happen for
2422 array reference lower bound or element size.
2423 (create_component_ref_by_pieces_1): Properly generate the
2424 element size operand for array references.
2426 2009-08-06 Richard Guenther <rguenther@suse.de>
2428 PR tree-optimization/40964
2429 * tree.c (iterative_hash_host_wide_int): Export.
2430 * tree.h (iterative_hash_host_wide_int): Declare.
2431 * tree-ssa-structalias.c (heapvar_map): New struct.
2432 (heapvar_map_eq): New function.
2433 (heapvar_map_hash): Likewise.
2434 (heapvar_lookup): Adjust.
2435 (heapvar_insert): Likewise.
2436 (make_constraint_from_heapvar): Allow multiple heap variables
2437 per decl at different offsets.
2438 (init_alias_heapvars): Adjust.
2440 2009-08-04 David Daney <ddaney@caviumnetworks.com>
2442 * config/mips/mips.h (TARGET_SYNC_AFTER_SC): New macro.
2443 * mips_output_sync_loop (mips_output_sync_loop): Only emit
2444 trailing sync if TARGET_SYNC_AFTER_SC.
2446 2009-08-05 David Daney <ddaney@caviumnetworks.com>
2448 * gcc/config/mips/sync.md (sync_compare_and_swap<mode>,
2449 compare_and_swap_12, sync_add<mode>, sync_<optab>_12,
2450 sync_old_<optab>_12, sync_new_<optab>_12, sync_nand_12,
2451 sync_old_nand_12, sync_new_nand_12, sync_sub<mode>,
2452 sync_old_add<mode>, sync_old_sub<mode>, sync_new_add<mode>,
2453 sync_new_sub<mode>, sync_<optab><mode>, sync_old_<optab><mode>,
2454 sync_new_<optab><mode>, sync_nand<mode>, sync_old_nand<mode>,
2455 sync_new_nand<mode>, sync_lock_test_and_set<mode>,
2456 test_and_set_12): Rewrite calls to mips_output_sync_loop.
2457 * gcc/config/mips/mips-protos.h (mips_output_sync_loop): Make
2458 the prototype declaration match the definition.
2459 * gcc/config/mips/mips.c (mips_output_sync_loop): Emit sync
2460 instructions explicitly. Add barrier_before and operands
2462 * gcc/config/mips/mips.h (MIPS_COMPARE_AND_SWAP,
2463 MIPS_COMPARE_AND_SWAP_12, MIPS_SYNC_OP, MIPS_SYNC_OP_12,
2464 MIPS_SYNC_OLD_OP_12, MIPS_SYNC_NEW_OP_12, MIPS_SYNC_OLD_OP,
2465 MIPS_SYNC_NEW_OP, MIPS_SYNC_NAND, MIPS_SYNC_OLD_NAND,
2466 MIPS_SYNC_NEW_NAND, MIPS_SYNC_EXCHANGE,
2467 MIPS_SYNC_EXCHANGE_12): Remove sync instructions.
2469 2009-08-05 Andrew Pinski <pinskia@gmail.com>
2471 * tree-ssa-alias.c: Fix intervals to use [) syntax.
2473 2009-08-05 Uros Bizjak <ubizjak@gmail.com>
2474 Mikulas Patocka <mikulas@artax.karlin.mff.cuni.cz>
2477 * config/i386/i386.c (ix86_split_long_move): Fix push of multi-part
2480 2009-08-05 Jakub Jelinek <jakub@redhat.com>
2482 PR rtl-optimization/40924
2483 * dse.c (canon_address): Before calling cselib_expand_value_rtx
2484 make sure canon_rtx (mem_address) isn't simpler than
2485 canon_rtx (expanded_mem_address).
2487 2009-08-05 Li Feng <nemokingdom@gmail.com>
2489 * graphite-sese-to-poly.c (build_pbb_drs): Remove build alias set
2491 (build_scop_drs): Build alias set for each SCoP.
2493 2009-08-04 Sandra Loosemore <sandra@codesourcery.com>
2495 * doc/invoke.texi (MIPS Options): Document new 1004K -march options.
2496 * config/mips/mips.c (mips_cpu_info_table): Add 1004K cores.
2497 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Add pattern for 1004K.
2498 (MIPS_ARCH_FLOAT_SPEC): Likewise.
2499 (BASE_DRIVER_SELF_SPECS): Likewise.
2501 2009-08-04 Andrew Pinski <pinskia@gmail.com>
2503 * tree-ssa-alias.c: Fix some comment typos.
2505 2009-08-04 Kaz Kojima <kkojima@gcc.gnu.org>
2507 * config/sh/linux-atomic.asm (ATOMIC_BOOL_COMPARE_AND_SWAP,
2508 ATOMIC_OP_AND_FETCH, ATOMIC_COMBOP_AND_FETCH): Define.
2510 2009-08-03 Janis Johnson <janis187@us.ibm.com>
2513 * simplify-rtx.c (simplify_binary_operation_1): Disable
2514 simplifications for decimal float operations.
2516 2009-08-03 Jakub Jelinek <jakub@redhat.com>
2519 * tree-ssa.c (warn_uninitialized_var): Even on LHS warn for
2520 operand of INDIRECT_REF.
2522 2009-08-03 Uros Bizjak <ubizjak@gmail.com>
2524 * config/alpha/alpha.c (alpha_legitimate_constant_p): Reject CONST
2525 constants referencing TLS symbols.
2527 2009-08-03 SUGIOKA Toshinobu <sugioka@itonet.co.jp>
2529 * config/sh/linux-atomic.asm (ATOMIC_COMPARE_AND_SWAP): Rename
2530 __sync_compare_and_swap_* to __sync_val_compare_and_swap_*.
2532 2009-08-03 Richard Guenther <rguenther@suse.de>
2534 * tree.c (make_vector_type): Build a main variant first,
2535 get the canonical one and then build the variant.
2536 * tree-ssa.c (useless_type_conversion_p_1): Handle
2538 (useless_type_conversion_p): Conversions to pointers to
2539 incomplete record types are useless.
2541 2009-08-03 Richard Guenther <rguenther@suse.de>
2543 * tree-cfg.c (pass_warn_unused_result): Mark name that no dump
2544 file will be created.
2545 * omp-low.c (pass_diagnose_omp_blocks): Likewise.
2546 * toplev.c (compile_file): Adjust comment.
2548 2009-08-03 Kaz Kojima <kkojima@gcc.gnu.org>
2550 * config/sh/sh-protos.h (sh_promote_function_mode): Remove.
2551 * config/sh/sh.c (sh_promote_function_mode): Wrap long lines.
2552 (TARGET_PROMOTE_FUNCTION_MODE): Define.
2553 (TARGET_PROMOTE_FUNCTION_ARGS): Remove.
2554 (sh_promote_function_mode): Fix typo.
2556 2009-08-03 Andreas Krebbel <krebbel1@de.ibm.com>
2558 * explow.c (promote_mode): Mark TYPE and PUNSIGNEDP as possibly unused.
2560 2009-08-02 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2562 * pa.c (pa_promote_function_mode): Remove ATTRIBUTE_UNUSED from
2563 declaration arguments.
2565 2009-08-02 Uros Bizjak <ubizjak@gmail.com>
2567 * config/i386/i386.c (ix86_expand_fp_compare): Use const0_rtx instead
2568 of GEN_INT (0x00) and const1_rtx instead of GEN_INT (0x01).
2569 (ix86_split_ashl): Ditto.
2570 (ix86_expand_vector_init_one_nonzero): Ditto.
2571 (ix86_expand_vector_set): Ditto.
2572 (ix86_expand_reduc_v4sf): Ditto.
2574 2009-08-02 Paolo Bonzini <bonzini@gnu.org>
2576 * explow.c (promote_function_mode): Remove assert.
2577 * config/sh/sh.c (sh_promote_function_mode): Declare.
2579 2009-08-01 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2581 * config/pa/pa.c (pa_promote_function_mode): Declare.
2582 Change to static. Fix promote_mode call.
2584 * gthr-dce.h (CONST_CAST2): Define if not defined.
2585 (__gthread_setspecific): Use CONST_CAST2 to fix warning.
2587 * config.gcc (hppa[12]*-*-hpux10*): Add stdint support.
2589 2009-08-01 Paolo Bonzini <bonzini@gnu.org>
2591 * expr.c (store_constructor): Use promote_decl_mode. Remove
2592 now write-only variable unsignedp.
2593 (expand_expr_real_1): Use promote_decl_mode.
2594 * expr.h (promote_function_mode, promote_decl_mode): New.
2595 (promote_mode): Remove last argument.
2596 * function.c (assign_temp): Drop last argument of promote_mode.
2597 (assign_parm_find_data_types): Use promote_function_mode.
2598 (assign_parm_setup_reg): Likewise.
2599 (expand_function_end): Use promote_function_mode.
2600 * calls.c (initialize_argument_information): Use promote_function_mode.
2601 (precompute_arguments): Use promote_mode instead of checking if
2602 only PROMOTE_FUNCTION_MODE is defined.
2603 (expand_call): When making sibcall decisions, use promote_function_mode.
2604 Below, remove an if for targetm.calls.promote_function_return and
2605 and use promote_function_mode.
2606 (emit_library_call_value_1): Use promote_function_mode, fix bug
2607 where promote_mode was passed FOR_CALL == 0 for a return value in an
2609 * cfgexpand.c (expand_one_register_var): Use promote_decl_mode.
2610 * explow.c (promote_function_mode, promote_decl_mode): New.
2611 (promote_mode): Keep only the FOR_CALL == 0 case.
2612 * combine.c (setup_incoming_promotion): Remove test of
2613 promote_function_args. Use promote_function_mode.
2614 * stmt.c (expand_value_return): Use promote_decl_mode.
2615 (expand_decl): Use promote_decl_mode.
2617 * expr.c (store_constructor): Use promote_decl_mode. Remove
2618 now write-only variable unsignedp.
2619 (expand_expr_real_1): Use promote_decl_mode.
2620 * expr.h (promote_function_mode, promote_decl_mode): New.
2621 (promote_mode): Remove last argument.
2622 * function.c (assign_temp): Drop last argument of promote_mode.
2623 (assign_parm_find_data_types): Use promote_function_mode.
2624 (assign_parm_setup_reg): Likewise.
2625 (expand_function_end): Use promote_function_mode.
2626 * calls.c (initialize_argument_information): Use promote_function_mode.
2627 (precompute_arguments): Use promote_mode instead of checking if
2628 only PROMOTE_FUNCTION_MODE is defined.
2629 (expand_call): When making sibcall decisions, use promote_function_mode.
2630 Below, remove an if for targetm.calls.promote_function_return and
2631 and use promote_function_mode.
2632 (emit_library_call_value_1): Use promote_function_mode, fix bug
2633 where promote_mode was passed FOR_CALL == 0 for a return value in an
2635 * cfgexpand.c (expand_one_register_var): Use promote_decl_mode.
2636 * explow.c (promote_function_mode, promote_decl_mode): New.
2637 (promote_mode): Keep only the FOR_CALL == 0 case.
2638 * combine.c (setup_incoming_promotion): Remove test of
2639 promote_function_args. Use promote_function_mode.
2640 * stmt.c (expand_value_return): Use promote_decl_mode.
2641 (expand_decl): Use promote_decl_mode.
2643 * explow.c (promote_function_mode): Just call the target hook.
2644 * targhooks.c (default_promote_function_mode,
2645 default_promote_function_mode_always_promote): New.
2646 * targhooks.h (default_promote_function_mode,
2647 default_promote_function_mode_always_promote): Declare.
2648 * target.h (promote_function_args, promote_function_return): Remove.
2649 (promote_function_mode): New.
2650 * target-def.h (TARGET_PROMOTE_FUNCTION_ARGS,
2651 TARGET_PROMOTE_FUNCTION_RETURN): Remove.
2652 (TARGET_PROMOTE_FUNCTION_MODE): New.
2653 (TARGET_CALLS): Adjust.
2654 * system.h (TARGET_PROMOTE_FUNCTION_ARGS,
2655 TARGET_PROMOTE_FUNCTION_RETURN, PROMOTE_FUNCTION_MODE): Poison.
2657 * config/s390/s390.h (PROMOTE_FUNCTION_MODE): Move...
2658 * config/s390/s390.c (s390_promote_function_mode): ... here,
2659 with pointer handling.
2660 (TARGET_PROMOTE_FUNCTION_MODE): Define.
2661 (TARGET_PROMOTE_FUNCTION_ARGS, TARGET_PROMOTE_FUNCTION_RETURN): Remove.
2663 * config/sparc/sparc.h (PROMOTE_FUNCTION_MODE): Move...
2664 * config/sparc/sparc.c (sparc_promote_function_mode): ... here,
2665 with pointer handling.
2666 (TARGET_PROMOTE_FUNCTION_MODE): Define.
2667 (TARGET_PROMOTE_FUNCTION_ARGS, TARGET_PROMOTE_FUNCTION_RETURN): Remove.
2669 * config/sh/sh-protos.h (sh_promote_function_mode): New.
2670 * config/sh/sh.c (sh_promote_function_mode): New.
2671 (TARGET_PROMOTE_FUNCTION_MODE): Define.
2672 (TARGET_PROMOTE_FUNCTION_ARGS, TARGET_PROMOTE_FUNCTION_RETURN): Remove.
2674 * config/cris/cris.h (PROMOTE_FUNCTION_MODE): Move...
2675 * config/cris/cris.c (cris_promote_function_mode): ... here.
2676 (TARGET_PROMOTE_FUNCTION_MODE): Define.
2677 (TARGET_PROMOTE_FUNCTION_ARGS): Remove.
2679 * config/mmix/mmix.h (PROMOTE_FUNCTION_MODE): Move...
2680 * config/mmix/mmix.c (mmix_promote_function_mode): ... here.
2681 (TARGET_PROMOTE_FUNCTION_MODE): Define.
2682 (TARGET_PROMOTE_FUNCTION_ARGS): Remove.
2684 * config/arm/arm.h (PROMOTE_FUNCTION_MODE): Move...
2685 * config/arm/arm.c (arm_promote_function_mode): ... here, without
2686 complex type handling.
2687 (TARGET_PROMOTE_FUNCTION_MODE): Define.
2688 (TARGET_PROMOTE_FUNCTION_ARGS, TARGET_PROMOTE_FUNCTION_RETURN): Remove.
2690 * config/pa/pa.c (pa_promote_function_mode): New.
2691 (TARGET_PROMOTE_FUNCTION_MODE): Define.
2692 (TARGET_PROMOTE_FUNCTION_RETURN): Remove.
2694 * config/alpha/alpha.c (TARGET_PROMOTE_FUNCTION_ARGS,
2695 TARGET_PROMOTE_FUNCTION_RETURN): Remove.
2696 (TARGET_PROMOTE_FUNCTION_MODE): Define equivalently.
2697 * config/xtensa/xtensa.c: Likewise.
2698 * config/stormy16/stormy16.c: Likewise.
2699 * config/iq2000/iq2000.c: Likewise.
2700 * config/rs6000/rs6000.c: Likewise.
2701 * config/picochip/picochip.c: Likewise.
2702 * config/arc/arc.c: Likewise.
2703 * config/mcore/mcore.c: Likewise.
2704 * config/score/score.c: Likewise.
2705 * config/mips/mips.c: Likewise.
2706 * config/bfin/bfin.c: Likewise.
2707 * config/ia64/ia64.c: Likewise (disabled though).
2709 * config/frv/frv.h: Remove pointless remark.
2711 * doc/tm.texi (PROMOTE_FUNCTION_MODE,
2712 TARGET_PROMOTE_FUNCTION_ARGS,
2713 TARGET_PROMOTE_FUNCTION_RETURN): Consolidate into...
2714 (TARGET_PROMOTE_FUNCTION_MODE): ... this.
2716 2009-08-01 Sebastian Pop <sebastian.pop@amd.com>
2718 * doc/invoke.texi (-fgraphite-force-parallel): Renamed
2719 -floop-parallelize-all.
2720 * toplev.c (process_options): Rename flag_graphite_force_parallel to
2721 flag_loop_parallelize_all.
2722 * tree-ssa-loop.c (gate_graphite_transforms): Same.
2723 * graphite.c (graphite_transform_loops): Same.
2725 * graphite-poly.c (apply_poly_transforms): Same.
2727 2009-07-31 Richard Earnshaw <rearnsha@arm.com>
2729 PR tree-optimization/40914
2730 * ipa-prop.c (ipa_get_ptr_load_param): New argument use_delta,
2731 if set, then check the delta field of the PMF record.
2732 (ipa_get_stmt_member_ptr_load_param): Propagate new param use_delta.
2733 (ipa_analyze_call_uses): Handle machines where the vbit for a PMF
2734 call is stored in the delta.
2736 2009-07-31 Adam Nemet <anemet@caviumnetworks.com>
2738 * config/mips/mips.md (*clear_upper32_dext): New pattern.
2740 2009-07-31 Uros Bizjak <ubizjak@gmail.com>
2742 * config/i386/bsd.h (ASM_BYTE): New define.
2743 * config/i386/darwin.h (ASM_BYTE): Rename from ASM_BYTE_OP.
2744 * config/i386/att.h (ASM_BYTE): New define. Use ASM_BYTE instead of
2745 .byte. Use fputs or putc instead of fprintf where appropriate.
2746 * config/i386/i386-interix.h: Use ASM_BYTE instead of .byte. Use
2747 fputs or putc instead of fprintf where appropriate.
2748 * config/i386/i386elf.h: Ditto.
2749 * config/i386/sysv4.h: Ditto.
2751 * config/i386/i386.c (TARGET_ASM_BYTE_OP): New define.
2752 * config/i386/i386.md (x86_sahf_1): Use ASM_BYTE instead of .byte.
2753 (*tls_global_dynamic_64): Ditto.
2755 2009-07-31 Christian Bruel <christian.bruel@st.com>
2757 * gcc/config.gcc (sh*-*-elf): test with_libgloss.
2759 2009-07-31 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2761 * config/arm/arm.c (arm_arm_address_cost): Fix typo.
2762 Remove dead code for MINUS.
2764 2009-07-31 Anthony Green <green@moxielogic.com>
2766 * config/moxie/moxie.c (moxie_expand_prologue): Use $r5 instead of
2768 (moxie_expand_epilogue): Ditto for epilogue.
2769 (moxie_setup_incoming_varargs): ABI change. Use 5 registers for
2771 (moxie_function_arg): Ditto.
2772 (moxie_pass_by_reference): Ditto.
2773 (moxie_arg_partial_bytes): Ditto.
2774 * config/moxie/moxie.h (CALL_USED_REGISTERS): Ditto.
2775 (FUNCTION_ARG_ADVANCE) Ditto.
2776 (REG_PARM_STACK_SPACE) Ditto.
2777 (FUNCTION_ARG_REGNO_P) Dito.
2779 * config.gcc: Add moxie linux config support.
2780 * gcc/config/moxie/uclinux.h: New file.
2782 2009-07-31 DJ Delorie <dj@redhat.com>
2784 * config/sh/sh.md (UNSPECV_SP_SWITCH_B): New.
2785 (UNSPECV_SP_SWITCH_E): New.
2786 (sp_switch_1): Change to an unspec.
2787 (sp_switch_2): Change to an unspec. Don't use post-inc when we
2789 * config/sh/sh.c (sh_expand_prologue): Use the constant pool to
2790 reference the new stack's address
2792 2009-07-30 Sebastian Pop <sebastian.pop@amd.com>
2794 * Makefile.in (OBJS-common): Added dependence on graphite-blocking.o,
2795 graphite-clast-to-gimple.o, graphite-dependences.o,
2796 graphite-interchange.o, graphite-poly.o, graphite-ppl.o,
2797 graphite-scop-detection.o, graphite-sese-to-poly.o, and sese.o.
2798 (graphite-blocking.o,
2799 graphite-clast-to-gimple.o, graphite-dependences.o,
2800 graphite-interchange.o, graphite-poly.o, graphite-ppl.o,
2801 graphite-scop-detection.o, graphite-sese-to-poly.o, and sese.o): New.
2802 * cfgloop.c (alloc_loop): Set loop->can_be_parallel to false.
2803 * cfgloop.h (struct loop): Add can_be_parallel field.
2804 * common.opt (fgraphite-identity): Moved up.
2805 (fgraphite-force-parallel): New flag.
2806 * graphite.c: Rewrite.
2807 * graphite.h: Rewrite.
2808 * passes.c (init_optimization_passes): Schedule a pass of DCE and LIM
2810 * toplev.c (graphite_out_file): New file descriptor.
2811 (graphite_in_file): New.
2812 (process_options): flag_graphite_force_parallel cannot be used without
2814 * tree-ssa-loop.c: Include toplev.h.
2815 (gate_graphite_transforms): Enable flag_graphite for
2816 flag_graphite_force_parallel.
2818 2009-07-30 Sebastian Pop <sebastian.pop@amd.com>
2820 * ChangeLog.graphite: New.
2821 * graphite-blocking.c: New.
2822 * graphite-clast-to-gimple.c: New.
2823 * graphite-clast-to-gimple.h: New.
2824 * graphite-dependences.c: New.
2825 * graphite-dependences.h: New.
2826 * graphite-interchange.c: New.
2827 * graphite-poly.c: New.
2828 * graphite-poly.h: New.
2829 * graphite-ppl.c: New.
2830 * graphite-ppl.h: New.
2831 * graphite-scop-detection.c: New.
2832 * graphite-scop-detection.h: New.
2833 * graphite-sese-to-poly.c: New.
2834 * graphite-sese-to-poly.h: New.
2838 2009-07-30 Sebastian Pop <sebastian.pop@amd.com>
2840 * tree-chrec.c (evolution_function_right_is_integer_cst): New.
2841 * tree-chrec.h (evolution_function_right_is_integer_cst): Declared.
2843 2009-07-30 Sebastian Pop <sebastian.pop@amd.com>
2845 * tree-chrec.c (operator_is_linear): Handle BIT_NOT_EXPR.
2846 (scev_is_linear_expression): Return false if the evolution is not
2847 affine multivariate.
2849 2009-07-30 Sebastian Pop <sebastian.pop@amd.com>
2851 * tree-data-ref.c (graphite_find_data_references_in_stmt): New.
2852 * tree-data-ref.h (graphite_find_data_references_in_stmt): Declared.
2854 2009-07-30 Sebastian Pop <sebastian.pop@amd.com>
2856 * tree-data-ref.c (debug_data_references): New.
2857 (debug_data_reference): New.
2858 * tree-data-ref.h (debug_data_references): Declared.
2859 (debug_data_reference): Declared.
2861 2009-07-30 Sebastian Pop <sebastian.pop@amd.com>
2863 * tree-data-ref.c (stmt_simple_memref_p: Removed.
2864 * tree-data-ref.h (scop_p): Removed.
2865 (struct data_reference): Remove field scop.
2867 (stmt_simple_memref_p): Removed.
2869 2009-07-30 Sebastian Pop <sebastian.pop@amd.com>
2871 * cfgloop.h (create_empty_loop_on_edge): Pass an extra argument.
2872 * cfgloopmanip.c (create_empty_loop_on_edge): Leave the loop_latch
2875 2009-07-30 Sebastian Pop <sebastian.pop@amd.com>
2877 * doc/invoke.texi (-fgraphite-force-parallel): Documented.
2879 2009-07-30 Sebastian Pop <sebastian.pop@amd.com>
2881 * doc/invoke.texi (-fgraphite-identity): Documented.
2883 2009-07-30 Sebastian Pop <sebastian.pop@amd.com>
2885 * tree-scalar-evolution.c: Fix comment.
2886 (instantiate_scev_1): Return unknow from scev instantiation if the
2887 result is not above instantiate_below.
2889 2009-07-30 Sebastian Pop <sebastian.pop@amd.com>
2891 * tree-scalar-evolution.c (compute_overall_effect_of_inner_loop): Not
2892 static anymore. Instantiate the symbols that may have been introduced
2894 * tree-scalar-evolution.h (compute_overall_effect_of_inner_loop):
2897 2009-07-30 DJ Delorie <dj@redhat.com>
2899 * config/mep/mep.c (mep_asm_init_sections): Add section flags and
2900 .vliw directive to VLIW sections.
2902 2009-07-30 Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
2904 * Makefile.in (AUTOCONF, ACLOCAL, ACLOCAL_AMFLAGS, aclocal_deps):
2906 ($(srcdir)/configure, $(srcdir)/aclocal.m4): New rules.
2907 (AUTOHEADER): New variable.
2908 ($(srcdir)/cstamp-h.in): Use it.
2910 2009-07-30 Michael Meissner <meissner@linux.vnet.ibm.com>
2911 Pat Haugen <pthaugen@us.ibm.com>
2912 Revital Eres <ERES@il.ibm.com>
2914 * config/rs6000/vector.md (VEC_F): Add VSX support.
2918 (vector_load_<mode>): Ditto.
2919 (vector_store_<mode>): Ditto.
2920 (vector GPR move split): Ditto.
2921 (vec_reload_and_plus_<mptrsize>): Ditto.
2922 (vec_reload_and_reg_<mptrsize>): Ditto.
2923 (add<mode>3): Ditto.
2924 (sub<mode>3): Ditto.
2925 (mul<mode>3): Ditto.
2926 (neg<mode>2): Ditto.
2927 (abs<mode>2): Ditto.
2928 (smin<mode>3): Ditto.
2929 (smax<mode>3): Ditto.
2930 (vector_eq<mode>): Ditto.
2931 (vector_gt<mode>): Ditto.
2932 (vector_ge<mode>): Ditto.
2933 (vector_gtu<mode>): Ditto.
2934 (vector_select_<mode>_uns): Ditto.
2935 (vector_eq_<mode>_p): Ditto.
2936 (vector_gt_<mode>_p): Ditto.
2937 (vector_ge_<mode>_p): Ditto.
2938 (vector_gtu_<mode>_p): Ditto.
2939 (cr6_test_for_zero): Ditto.
2940 (cr6_test_for_zero_reverse): Ditto.
2941 (cr6_test_for_lt): Ditto.
2942 (cr6_test_for_lt_reverse): Ditto.
2943 (xor<mode>3): Ditto.
2944 (ior<mode>3): Ditto.
2945 (and<mode>3): Ditto.
2946 (one_cmpl<mode>2): Ditto.
2947 (nor<mode>2): Ditto.
2948 (andc<mode>2): Ditto.
2949 (float<VEC_int<mode>2): Ditto.
2950 (unsigned_float<VEC_int><mode>2): Ditto.
2951 (fix_trunc<mode><VEC_int>2): Ditto.
2952 (fixuns_trunc<mode><VEC_int>2): Ditto.
2954 (vec_set<mode>): Ditto.
2955 (vec_extract<mode>): Ditto.
2956 (vec_interleave_highv4sf): Ditto.
2957 (vec_interleave_lowv4sf): Ditto.
2958 (vec_realign_load_<mode>): Ditto.
2959 (vec_shl_<mode>): Ditto.
2960 (vec_shr_<mode>): Ditto.
2961 (div<mode>3): New patterns for VSX.
2962 (vec_interleave_highv2df): Ditto.
2963 (vec_interleave_lowv2df): Ditto.
2964 (vec_pack_trunc_v2df): Ditto.
2965 (vec_pack_sfix_trunc_v2df): Ditto.
2966 (vec_pack_ufix_trunc_v2df): Ditto.
2967 (vec_unpacks_hi_v4sf): Ditto.
2968 (vec_unpacks_lo_v4sf): Ditto.
2969 (vec_unpacks_float_hi_v4si): Ditto.
2970 (vec_unpacks_float_lo_v4si): Ditto.
2971 (vec_unpacku_float_hi_v4si): Ditto.
2972 (vec_unpacku_float_lo_v4si): Ditto.
2973 (movmisalign<mode>): Ditto.
2974 (vector_ceil<mode>2): New patterns for vectorizing math library.
2975 (vector_floor<mode>2): Ditto.
2976 (vector_btrunc<mode>2): Ditto.
2977 (vector_copysign<mode>3): Ditto.
2979 * config/rs6000/predicates.md (easy_vector_constant_msb): New
2980 predicate for setting the high bit in each word, used for copysign.
2982 * config/rs6000/ppc-asm.h (f19): Whitespace.
2983 (f32-f63): Define if VSX.
2984 (v0-v31): Define if Altivec.
2985 (vs0-vs63): Define if VSX.
2987 * config/rs6000/t-rs6000 (MD_INCLUDES): Add power7.md and vsx.md.
2989 * config/rs6000/power7.md: New file, provide tuning parameters for
2992 * config/rs6000/rs6000-c.c (rs6000_macro_to_expand): Add VSX support.
2993 (rs6000_cpu_cpp_builtins): Ditto.
2994 (altivec_overloaded_builtins): Ditto.
2995 (altivec_resolve_overloaded_builtin): Ditto.
2997 * config/rs6000/rs6000.opt (-mno-vectorize-builtins): Add new
2998 debug switch to disable vectorizing simple math builtin
3001 * config/rs6000/rs6000.c (rs6000_builtin_vectorized_function):
3002 Vectorize simple math builtin functions.
3003 (TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION): Define target
3004 hook to vectorize math builtins.
3005 (rs6000_override_options): Enable -mvsx on -mcpu=power7.
3006 (rs6000_builtin_conversion): Add VSX/power7 support.
3007 (rs6000_builtin_vec_perm): Ditto.
3008 (vsplits_constant): Add support for loading up a vector constant
3009 with just the high bit set in each part.
3010 (rs6000_expand_vector_init): Add VSX/power7 support.
3011 (rs6000_expand_vector_set): Ditto.
3012 (rs6000_expand_vector_extract): Ditto.
3013 (rs6000_emit_move): Ditto.
3014 (bdesc_3arg): Ditto.
3015 (bdesc_2arg): Ditto.
3016 (bdesc_1arg): Ditto.
3017 (rs6000_expand_ternop_builtin): Ditto.
3018 (altivec_expand_builtin): Ditto.
3019 (rs6000_expand_unop_builtin): Ditto.
3020 (rs6000_init_builtins): Ditto.
3021 (altivec_init_builtins): Ditto.
3022 (builtin_function_type): Ditto.
3023 (rs6000_common_init_builtins): Ditto.
3024 (rs6000_handle_altivec_attribute); Ditto.
3025 (rs6000_mangle_type): Ditto.
3026 (rs6000_vector_mode_supported_p): Ditto.
3027 (rs6000_mode_dependent_address): Altivec addresses with AND -16
3030 * config/rs6000/vsx.md: New file for VSX support.
3032 * config/rs6000/rs6000.h (EASY_VECTOR_MSB): New macro for
3033 identifing values with just the most significant bit set.
3034 (enum rs6000_builtins): Add builtins for VSX. Add simple math
3035 vectorized builtins.
3037 * config/rs6000/altivec.md (UNSPEC_VRFIP): Delete.
3038 (UNSPEC_VRFIM): Delete.
3039 (splitter for loading up vector with most significant bit): New
3040 splitter for vectorizing copysign.
3041 (altivec_vrfiz): Rename from altivec_fturncv4sf2. Add support for
3042 vectorizing simple math functions.
3043 (altivec_vrfip): Add support for vectorizing simple math functions.
3044 (altivec_vrfim): Ditto.
3045 (altivec_copysign_v4sf3): New insn for Altivec copysign support.
3047 * config/rs6000/rs6000.md (UNSPEC_BPERM): New constant.
3048 (power7.md, vsx.md): Include for power7 support.
3049 (copysigndf3): Use VSX instructions if -mvsx.
3050 (negdf2_fpr): Ditto.
3051 (absdf2_fpr): Ditto.
3052 (nabsdf2_fpr): Ditto.
3053 (adddf3_fpr): Ditto.
3054 (subdf3_fpr): Ditto.
3055 (muldf3_fpr): Ditto.
3056 (divdf3_fpr): Ditto.
3057 (fix_truncdfdi2_fpr): Ditto.
3058 (cmpdf_internal1): Ditto.
3059 (fred, fred_fpr): Convert into expander/insn to add VSX support.
3060 (btruncdf2, btruncdf2_fpr): Ditto.
3061 (ceildf2, ceildf2_fpr): Ditto.
3062 (floordf2, floordf2_fpr): Ditto.
3063 (floatdidf2, floatdidf2_fpr): Ditto.
3064 (fmadddf4_fpr): Name insn. Use VSX instructions if -mvsx.
3065 (fmsubdf4_fpr): Ditto.
3066 (fnmadddf4_fpr_1): Ditto.
3067 (fnmadddf4_fpr_2): Ditto.
3068 (fnmsubdf4_fpr_1): Ditto.
3069 (fnmsubdf4_fpr_2): Ditto.
3070 (fixuns_truncdfdi2): Add expander for VSX support.
3071 (fix_truncdfdi2): Ditto.
3072 (fix_truncdfsi2): Ditto.
3074 (btruncsf2): Whitespace.
3075 (movdf_hardfloat32): Add support for VSX registers.
3076 (movdf_softfloat32): Ditto.
3077 (movdf_hardfloat64): Ditto.
3078 (movdf_hardfloat64_mfpgpr): Ditto.
3079 (movdf_softfloat64): Ditto.
3080 (movti splitters): Add check for vector registers supporting
3081 TImode in the future.
3082 (bpermd): Add power7 bpermd instruction.
3084 * config/rs6000/altivec.h (vec_div): Define if VSX.
3088 (vec_nearbyint): Ditto.
3091 (all predicates): Use the generic builtin function, and not the V4SF
3092 specific function so that the predicates will work with VSX's V2DF.
3096 * doc/extend.texi (PowerPC Altivec/VSX Built-in Functions):
3097 Document new VSX functions and types.
3099 * doc/invoke.texi (PowerPc options): Document -mpopcntd, -mvsx
3102 * doc/md.texi (PowerPC constraints): Document "wd", "wf", "ws",
3103 "wa", and "j" constraints. Modify "v" to talk about Altivec
3104 instead of just vector.
3106 2009-07-30 Andrew MacLeod <amacleod@redhat.com>
3109 * tree-into-ssa.c (insert_phi_nodes_for, rewrite_add_phi_arguments): Set
3110 location for phi arguments.
3111 (rewrite_update_phi_arguments): Find locations for reaching defs.
3112 * tree-ssa-threadupdate.c (create_edge_and_update_destination_phis):
3113 Add location to add_phi_arg calls.
3114 * tree-loop-districbution.c (update_phis_for_loop_copy): Add locations.
3115 * tree-ssa-loop-manip.c (create_iv, add_exit_phis_edge,
3116 split_loop_exit_edge, tree_transform_and_unroll_loop): Add locations.
3117 * tree-tailcall.c (add_successor_phi_arg, eliminate_tail_call,
3118 create_tailcall_accumulator, tree_optimize_tail_calls_1): Add locations.
3119 * tree.h (struct phi_arg_d): Add location_t to PHI arguments.
3120 * tree-phinodes.c (make_phi_node): Initialize location.
3121 (resize_phi_node): Initialize location to UNKNOWN_LOCATION.
3122 (add_phi_arg): Add location parameter.
3123 (remove_phi_arg_num): Move location when moving phi argument.
3124 * omp-low.c (expand_parallel_call, expand_omp_for_static_chunk): Set
3126 * tree-vect-loop-manip.c (slpeel_update_phis_for_duplicate_loop,
3127 slpeel_update_phi_nodes_for_guard1,
3128 slpeel_update_phi_nodes_for_guard2,
3129 slpeel_tree_duplicate_loop_to_edge_cfg, set_prologue_iterations,
3130 vect_loop_versioning): Set locations.
3131 * tree-parloops.c (create_phi_for_local_result,
3132 transform_to_exit_first_loop, create_parallel_loop): Add locations.
3133 * gimple-pretty-print.c (dump_gimple_phi): Dump lineno's if present.
3134 * tree-vect-loop.c (get_initial_def_for_induction,
3135 vect_create_epilog_for_reduction, vect_finalize_reduction): Add
3137 * tree-flow-inline.h (gimple_phi_arg_location): New. Return locus.
3138 (gimple_phi_arg_location_from_edge): New. Return locus from an edge.
3139 (gimple_phi_arg_set_location): New. Set locus.
3140 (gimple_phi_arg_has_location): New. Check for locus.
3141 (redirect_edge_var_map_location): New. Return locus from var_map.
3142 * tree-vect-data-refs.c (vect_setup_realignment): Set location.
3143 * tree-ssa-phiopt.c (conditional_replacement): Set locus when
3144 combining PHI arguments.
3145 (cond_store_replacement): Set location.
3146 * cfgexpand.c (gimple_assign_rhs_to_tree): Transfer locus if possible.
3147 * grpahite.c (add_loop_exit_phis, add_guard_exit_phis,
3148 scop_add_exit_phis_edge): Add locations.
3149 * tree-cfgcleanup.c (remove_forwarder_block,
3150 remove_forwarder_block_with_phi): Add locations.
3151 * tree-ssa-pre.c (insert_into_preds_of_block): Add locations.
3152 * tree-predcom.c (initialize_root_vars, initialize_root_vars_lm): Add
3154 * tree-ssa-dce.c (forward_edge_to_pdom): Add locations.
3155 * tree-ssa.c (redirect_edge_var_map_add, ssa_redirect_edge,
3156 flush_pending_stmts): Add source location.
3157 * lambda-code.c (perfect_nestify): Maintain location stack with argument
3158 stack to preserve locations.
3159 * tree-vect-stmts.c (vectorizable_load): Add location.
3160 * tree-inline.c (copy_phis_for_bb): Copy locus.
3161 (setup_one_parameter): Add call locus to inlined parameter stmts.
3162 (initialize_inlined_parameters): Pass in call location as parameter
3164 (tree_function_versioning): Pass location to setup_one_parameter.
3165 * tree-ssa-phiprop.c (phiprop_insert_phi): Set locations.
3166 * tree-outof-ssa.c (struct _elim_graph): Add source_location vecs for
3167 copy and edge lists.
3168 (insert_partition_copy_on_edge, insert_value_copy_on_edge,
3169 insert_rtx_to_part_on_edge, insert_part_to_rtx_on_edge): Provide a
3170 locus parameter and override the stmt default if provided.
3171 (new_elim_graph, clear_elim_graph, delete_elim_graph,
3172 elim_graph_add_edge, elim_graph_remove_succ_edge,
3173 FOR_EACH_ELIM_GRAPH_SUCC, FOR_EACH_ELIM_GRAPH_PRED, eliminate_build,
3174 elim_forward, elim_unvisited_predecessor, elim_backward, elim_create,
3175 eliminate_phi): Add locus info in elimination graph for each edge and
3177 (insert_backedge_copies): Copy locus if present.
3178 * tree-flow.h (struct _edge_var_map): Add locus field.
3179 * tree-switch_conversions.c (fix_phi_nodes): Add locations.
3180 * tree-cfg.c (reinstall_phi_args, gimple_make_forwarder_block,
3181 add_phi_args_after_copy_edge, gimple_lv_adjust_loop_header_phi): Add
3183 * ipa-struct-reorg.c (make_edge_and_fix_phis_of_dest): Add locations.
3185 2009-07-30 Martin Jambor <mjambor@suse.cz>
3187 PR tree-optimization/40570
3188 * ipa-inline.c (cgraph_decide_inlining): Watch out for dead single
3191 2009-07-30 Razya Ladelsky <razya@il.ibm.com>
3193 * ssa-loop-manip.c: Include langhooks.h.
3194 (rewrite_phi_with_iv): New.
3195 (rewrite_all_phi_nodes_with_iv): New.
3196 (canonicalize_loop_ivs): Move here from tree-parloops.c.
3197 Remove reduction_list argument. Use rewrite_all_phi_nodes_with_iv.
3198 * tree-parloops.c (loop_parallel_p): Move out all conditions
3199 except dependency check.
3200 (canonicalize_loop_ivs): Move to tree-ssa-loop-manip.c.
3201 (gen_parallel_loop): Call canonicalize_loop_ivs without
3202 reduction_list argument.
3203 (build_new_reduction): New.
3204 (gather_scalar_reductions): New.
3205 (try_get_loop_niter): New.
3206 (try_create_reduction_list): New.
3207 (parallleize_loops): Change the parallel conditions check.
3208 * tree-flow.h (canonicalize_loop_ivs): Remove one argument.
3209 * Makefile.in (tree-ssa-loop-manip.o): Add langhooks.h dependency.
3211 2009-07-30 Dave Korn <dave.korn.cygwin@gmail.com>
3213 * opt-functions.awk (opt_args): Allow argument to be enclosed in
3215 * doc/options.texi (Option properties): Mention new quoting syntax.
3217 2009-07-29 Douglas B Rupp <rupp@gnat.com>
3219 * config/alpha/alpha.c (alpha_start_function):
3220 Handle VMS_DEBUG_MAIN_POINTER
3221 * config/alpha/vms.h (VMS_DEBUG_MAIN_POINTER): Define new macro.
3222 * doc/invoke.texi: Document -mdebug-main switch.
3224 2009-07-29 Richard Henderson <rth@redhat.com>
3226 * cgraph.c (cgraph_set_call_stmt_including_clones): Tidy.
3227 (cgraph_create_edge_including_clones): Likewise.
3228 * tree-inline.c (copy_bb): Operate on the correct edges
3229 when updating the callgraph.
3231 2009-07-29 Douglas B Rupp <rupp@gnat.com>
3233 * config/alpha/vms-cc.c: Deleted.
3234 * config/alpha/vms-ld.c: Deleted.
3235 * config/alpha/t-vms64: Moved to config/vms
3236 * config/alpha/vms-crt0-64.c: Moved to config/vms
3237 * config/alpha/vms-crt0.c: Moved to config/vms
3238 * config/alpha/vms-psxcrt0-64.c: Moved to config/vms
3239 * config/alpha/vms-psxcrt0.c: Moved to config/vms
3240 * config/alpha/xm-vms.h: Moved to config/vms
3241 * config/alpha/x-vms: Moved to config/vms
3242 * config/alpha/t-vms (vcrt0.o, pcrt0.o): Move rules to new file
3244 * config/vms/t-vms: Moved here from config/alpha. Alpha specific
3245 parts removed. (STMP_FIXPROTO, STMP_FIXINC, LIMITS_H_TEST): Set.
3247 * config/vms/t-vms64: Moved here from config/alpha
3248 * config/vms/vms-crt0-64.c: Moved here from config/alpha.
3249 (argc,argv,envp): Enforce 32bit malloc'ing.
3250 * config/vms/vms-psxcrt0-64.c: Likewise.
3251 * config/vms/vms-crt0.c: Moved here from config/alpha.
3252 * config/vms/vms-psxcrt0.c: Likewise.
3253 * config/vms/vms-crtl-64.h: New file.
3254 * config/vms/vms-crtl.h: New file.
3255 * config/vms/vms.opt: New file.
3256 * config/vms/xm-vms64.h: New file.
3257 * config/vms/xm-vms.h: Moved here from config/alpha.
3258 (STANARD_EXEC_PREFIX, STANDARD_STARTFILE_PREFIX, STANDARD_INCLUDE_DIR):
3260 * config/vms/x-vms: Moved here from config/alpha.
3261 (version, VMS_EXTRA_PARTS): Moved to t-vms.
3262 (vms-ld.o, vms-cc.o): Removed.
3263 (LN, LN_S, USE_COLLECT2, POD2MAN): Set.
3265 2009-07-29 Douglas B Rupp <rupp@gnat.com>
3267 * dwarf2out.c (add_name_and_src_coords_attributes): Push on the
3268 correct stack (obvious VMS fix).
3270 2009-07-29 Douglas B Rupp <rupp@gnat.com>
3272 * dwarf2out.c (output_file_names): Output VMS style file name, size,
3273 date, version info if VMS_DEBUGGING_INFO defined.
3274 * vmsdgbout.c (vms_file_stats_name): New functon. VMS style file name,
3275 size, date calculating code moved here.
3277 2009-07-29 Paul Brook <paul@codesourcery.com>
3279 * config/arm/lib1funcs.asm (clear_cache): Use ARM_FUNC_START and
3282 2009-07-29 Uros Bizjak <ubizjak@gmail.com>
3285 * config/alpha/alpha.c (alpha_expand_unaligned_store): Convert src
3286 to DImode when generating insq_le insn.
3288 2009-07-28 Douglas B Rupp <rupp@gnat.com>
3290 * dwarf2out.c (DWARF2_INDIRECT_STRING_SUPPORT_MISSING_ON_TARGET):
3291 New macro set for VMS_DEBUGGGING_INFO.
3292 (AT_string_form): Use it.
3294 2009-07-28 DJ Delorie <dj@redhat.com>
3296 * config/mep/mep.c (vtext_section): New.
3297 (vftext_section): New.
3298 (ftext_section): New.
3299 (mep_select_section): Add support for functions.
3300 (mep_unique_section): Likewise.
3301 (mep_asm_init_sections): Likewise.
3302 (mep_encode_section_info): Remove it from here.
3304 * config/mep/mep.h (USE_SELECT_SECTION_FOR_FUNCTIONS): Define.
3306 2009-07-28 Paolo Bonzini <bonzinI@gnu.org>
3308 * tree.h (TREE_DEPRECATED): Document it is used for types too.
3309 (TYPE_VECTOR_OPAQUE): Use default_def_flag
3311 2009-07-28 Douglas B Rupp <rupp@gnat.com>
3313 * dwarf2out.c (output_file_names): Test new macro
3314 DWARF2_DIR_SHOULD_END_WITH_SEPARATOR.
3315 (add_comp_dir_attribute): Likewise.
3317 2009-07-28 Kai Tietz <kai.tietz@onevision.com>
3319 * config/i386/mingw-w64.h (LINK_SPEC): Add
3320 separating space between commands.
3322 2009-07-28 Jan Hubicka <jh@suse.cz>
3324 PR tree-optimization/40759
3325 * tree-ssa-dce.c (mark_virtual_phi_result_for_renaming): Mark all uses
3328 2009-07-27 DJ Delorie <dj@redhat.com>
3330 * config/mep/mep.c (mep_expand_builtin_saveregs): Make sure 64-bit
3331 types are dword-aligned.
3332 (mep_expand_va_start): Likewise.
3334 2009-07-27 Olivier Hainque <hainque@adacore.com>
3335 Douglas B Rupp <rupp@gnat.com>
3337 * convert.c (convert_to_pointer): Don't assume the target
3338 pointer type is POINTER_SIZE long. Fetch its precision instead.
3340 2009-07-27 Douglas B Rupp <rupp@gnat.com>
3342 * system.h (fopen): Undefine if macro.
3344 2009-07-27 Jakub Jelinek <jakub@redhat.com>
3346 * dwarf2out.c (output_cfi_p): Removed.
3347 (output_cfis): New function.
3348 (output_fde): New function, split from output_call_frame_info.
3349 (output_call_frame_info): Use it.
3350 (dwarf2out_switch_text_section): Use output_cfis.
3352 2009-07-24 Kai Tietz <kai.tietz@onevision.com>
3354 * config/i386/cygming.h (DWARF2_UNWIND_INFO): Error build when
3355 TARGET_BI_ARCH is specified without enabling SJLJ.
3356 * config/i386/mingw32.h (MD_UNWIND_SUPPORT): Define MD_UNWIND_SUPPORT,
3357 if TARGET_64BIT and TARGET_BI_ARCH aren't defined.
3359 2009-07-26 Mikael Pettersson <mikpe@it.uu.se>
3361 * arm.md (negdi2): Use DImode if forcing a value into a register.
3363 2009-07-26 Ira Rosen <irar@il.ibm.com>
3365 PR tree-optimization/40801
3366 * tree-vect-stmts.c (vectorizable_call): Get previous copy
3367 of vector operand from the previous copy of vector statement.
3368 Pass the correct definition type value to
3369 vect_get_vec_def_for_stmt_copy().
3371 2009-07-25 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
3373 * collect2.c (scan_libraries): Use CONST_CAST2 to perform char ** to
3374 const char ** conversion.
3376 2009-07-25 David Daney <ddaney@caviumnetworks.com>
3378 * system.h (gcc_assert): Invoke __builtin_unreachable() instead of
3379 fancy_abort() if !ENABLE_ASSERT_CHECKING.
3380 (gcc_unreachable): Invoke __builtin_unreachable() if
3381 !ENABLE_ASSERT_CHECKING.
3383 2009-07-25 David Daney <ddaney@caviumnetworks.com>
3385 PR rtl-optimization/40445
3386 * emit-rtl.c (next_nonnote_insn_bb): New function.
3387 * rtl.h (next_nonnote_insn_bb): Declare new function.
3388 * cfgcleanup.c (try_optimize_cfg): Don't remove an empty block
3389 with no successors that is the successor of the ENTRY_BLOCK.
3390 Continue from the top after removing an empty fallthrough block.
3391 * cfgrtl.c (get_last_bb_insn): Call next_nonnote_insn_bb instead
3392 of next_nonnote_insn.
3394 2009-07-25 David Daney <ddaney@caviumnetworks.com>
3396 * cfgcleanup.c (old_insns_match_p): Handle the case of empty blocks.
3398 2009-07-25 Martin Jambor <mjambor@suse.cz>
3400 * c-common.c (c_common_attribute_table): New element for noclone.
3401 (handle_noclone_attribute): New function. Forward-declare.
3402 * tree-inline.c (tree_versionable_function_p): Check for noclone
3404 * doc/extend.texi (Labels as Values): Document need for noclone.
3405 (Function Attributes): Document noclone attribute.
3407 2009-07-25 Jakub Jelinek <jakub@redhat.com>
3409 PR rtl-optimization/34999
3410 * dwarf2out.c (struct dw_fde_struct): Add dw_fde_switch_cfi
3411 and dw_fde_switched_cold_to_hot fields.
3412 (output_cfi_p): New function.
3413 (output_call_frame_info): If fde->dw_fde_switched_sections,
3414 output 2 FDEs instead of one with corrupted header.
3415 (dwarf2out_do_cfi_startproc): New function.
3416 (dwarf2out_begin_prologue): Use it. Initialize fde->dw_fde_switch_cfi
3417 and fde->dw_fde_switched_cold_to_hot.
3418 (dwarf2out_switch_text_section): Compute
3419 fde->dw_fde_switched_cold_to_hot. Switch to new text section here.
3420 If dwarf2out_do_cfi_asm, emit .cfi_endproc before it and call
3421 dwarf2out_do_cfi_startproc plus emit again currently active CFI insns.
3422 Otherwise, compute fde->dw_fde_switch_cfi.
3424 2009-07-24 Cary Coutant <ccoutant@google.com>
3426 * tree-cfg.c (assign_discriminator): Add explicit parentheses.
3428 2009-07-24 Cary Coutant <ccoutant@google.com>
3430 * cfghooks.c (split_block): Copy discriminator to new block.
3431 * tree-cfg.c (assign_discriminator): Check location of last
3432 instruction in block as well as first.
3434 2009-07-24 Uros Bizjak <ubizjak@gmail.com>
3436 * config/i386/linux.c: Use fputs or putc instead of fprintf
3438 * config/i386/gas.h: Ditto.
3439 * config/i386/x86-64.h: Ditto.
3440 * config/i386/att.h: Ditto.
3442 2009-07-24 Andreas Krebbel <Andreas.Krebbel@de.ibm.com>
3444 * expmed.c (emit_store_flag): Use a recursive call to optimize the
3447 2009-07-24 Martin Jambor <mjambor@suse.cz>
3449 * ipa-prop.h (struct ipa_node_params): New flag node_enqued.
3450 (ipa_push_func_to_list_1): Declare.
3451 (ipa_push_func_to_list): New function.
3453 * ipa-prop.c (ipa_push_func_to_list_1): New function.
3454 (ipa_init_func_list): Call ipa_push_func_to_list_1.
3455 (ipa_push_func_to_list): Removed.
3456 (ipa_pop_func_from_list): Clear node_enqueued flag.
3458 2009-07-24 Andreas Krebbel <krebbel1@de.ibm.com>
3460 * config/s390/s390.c (override_options): Default
3461 max-unrolled-insns to 100 for z10 tuning.
3463 2009-07-24 Tobias Grosser <grosser@fim.uni-passau.de>
3465 * Makefile.in (TREE_DATA_REF_H, tree-vrp.o, tree-cfg.o, tree-if-conv.o
3466 tree-ssa-loop.o, tree-ssa-loop-niter.o, tree-ssa-loop-ivcanon.o,
3467 tree-ssa-loop-prefetch.o, tree-predcom.o, tree-affine.o,
3468 tree-scalar-evolution.o, tree-data-ref.o, tree-vect-loop.o,
3469 tree-vect-data-refs.o, tree-loop-linear.o, tree-loop-distribution.o
3470 tree-parloops.o, tree-pretty-printer.o, fold-const.o, tree-ssa-dce.o,
3471 lambda-code.o, params.o): Cleanup use of SCEV_H and TREE_DATA_REF_H.
3473 2009-07-24 Kai Tietz <kai.tietz@onevision.com>
3475 * config/i386/mingw-w64.h (STANDARD_INCLUDE_DIR): Remove and
3476 use default set in mingw32.h header.
3477 * config/i386/mingw32.h (STANDARD_INCLUDE_DIR): Use for 32-bit and
3478 64-bit /mingw/include path.
3479 (STANDARD_STARTFILE_PREFIX_1): Use for 32-bit and 64-bit /mingw/lib
3482 2009-07-23 Neil Vachharajani <nvachhar@google.com>
3484 PR rtl-optimization/40209
3485 * loop-iv.c (iv_analysis_loop_init): Call df_note_add_problem.
3487 2009-07-23 Uros Bizjak <ubizjak@gmail.com>
3489 * config/i386/i386.c: Use ASM_LONG instead of .long. Concatenate
3490 ASM_LONG, LPREFIX, MCOUNT_NAME and PROFILE_COUNT_REGISTER strings
3491 with the rest of string where appropriate. Use fputs or putc
3492 instead of fprintf where appropriate.
3494 2009-07-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3495 Pat Haugen <pthaugen@us.ibm.com>
3496 Revital Eres <ERES@il.ibm.com>
3498 * config/rs6000/vector.md: New file. Move most of the vector
3499 expander support here from altivec.md to allow for the VSX vector
3500 unit in the future. Add support for secondary_reload patterns.
3501 Rewrite the patterns for vector comparison, and vector comparison
3502 predicate instructions so that the RTL expresses the desired
3503 behavior, instead of using unspec.
3505 * config/rs6000/constraints.md ("f" constraint): Use
3506 rs6000_constraints to hold the precalculated register class.
3507 ("d" constraint): Ditto.
3508 ("wd" constraint): New constraint for VSX.
3509 ("wf" constraint): Ditto.
3510 ("ws" constraint): Ditto.
3511 ("wa" constraint): Ditto.
3512 ("wZ" constraint): Ditto.
3513 ("j" constraint): Ditto.
3515 * config/rs6000/predicates.md (vsx_register_operand): New
3517 (vfloat_operand): New predicate for vector.md.
3518 (vint_operand): Ditto.
3519 (vlogical_operand): Ditto.
3520 (easy_fp_constant): If VSX, 0.0 is an easy constant.
3521 (easy_vector_constant): Add VSX support.
3522 (altivec_indexed_or_indirect_operand): New predicate for
3523 recognizing Altivec style memory references with AND -16.
3525 * config/rs6000/rs6000.c (rs6000_vector_reload): New static global
3526 for vector secondary reload support.
3527 (rs6000_vector_reg_class): Delete, replacing it with
3529 (rs6000_vsx_reg_class): Ditto.
3530 (rs6000_constraints): New array to hold the register classes of
3531 each of the register constraints that can vary at runtime.
3532 (builtin_mode_to_type): New static array for builtin function type
3534 (builtin_hash_table): New static hash table for builtin function
3536 (TARGET_SECONDARY_RELOAD): Define target hook.
3537 (TARGET_IRA_COVER_CLASSES): Ditto.
3538 (rs6000_hard_regno_nregs_internal): If -mvsx, floating point
3539 registers are 128 bits if VSX memory reference instructions are used.
3540 (rs6000_hard_regno_mode_ok): For VSX, only check if the VSX memory
3542 (rs6000_debug_vector_unit): Move into rs6000_debug_reg_global.
3543 (rs6000_debug_reg_global): Move -mdebug=reg statements here.
3544 Print several of the scheduling related parameters.
3545 (rs6000_init_hard_regno_mode_ok): Switch to putting constraints in
3546 rs6000_constraints instead of rs6000_vector_reg_class. Move
3547 -mdebug=reg code to rs6000_debug_reg_global. Add support for
3548 -mvsx-align-128 debug switch. Drop testing float_p if VSX or
3549 Altivec. Add VSX support. Setup for secondary reload support on
3550 Altivec/VSX registers.
3551 (rs6000_override_options): Make power7 set the scheduling groups
3552 like the power5. Add support for new debug switches to override
3553 the scheduling defaults. Temporarily disable -mcpu=power7 from
3554 setting -mvsx. Add support for debug switches -malways-hint,
3555 -msched-groups, and -malign-branch-targets.
3556 (rs6000_buitlin_conversion): Add support for returning unsigned
3557 vector conversion functions to fix regressions due to stricter
3559 (rs6000_builtin_mul_widen_even): Ditto.
3560 (rs6000_builtin_mul_widen_odd): Ditto.
3561 (rs6000_builtin_vec_perm): Ditto.
3562 (rs6000_vec_const_move): On VSX, use xxlxor to clear register.
3563 (rs6000_expand_vector_init): Initial VSX support for using xxlxor
3565 (rs6000_emit_move): Fixup invalid const symbol_ref+reg that is
3567 (bdesc_3arg): Add builtins for unsigned types. Add builtins for
3568 VSX types for bit operations. Changes to accomidate vector.md.
3569 (bdesc_2arg): Ditto.
3570 (bdesc_1arg): Ditto.
3571 (struct builtin_description_predicates): Rewrite predicate
3572 handling so that RTL describes the operation, instead of passing
3573 the instruction to be used as a string argument.
3574 (bdesc_altivec_preds): Ditto.
3575 (altivec_expand_predicate_builtin): Ditto.
3576 (altivec_expand_builtin): Ditto.
3577 (rs6000_expand_ternop_builtin): Use a switch instead of an if
3578 statement for vsldoi support.
3579 (altivec_expand_ld_builtin): Change to use new names from vector.md.
3580 (altivec_expand_st_builtin): Ditto.
3581 (paired_expand_builtin): Whitespace changes.
3582 (rs6000_init_builtins): Add V2DF/V2DI types. Initialize the
3583 builtin_mode_to_type table for secondary reload. Call
3584 builtin_function_type to build random builtin functions.
3585 (altivec_init_builtins): Change to use builtin_function_type to
3586 create builtin function types dynamically as we need them.
3587 (builtin_hash_function): New support for hashing the tree types
3588 for builtin function as we need it, rather than trying to build
3589 all of the trees that we need. Add initial preliminary VSX support.
3590 (builtin_function_type): Ditto.
3591 (builtin_function_eq): Ditto.
3592 (builtin_hash_struct): Ditto.
3593 (rs6000_init_builtins): Ditto.
3594 (rs6000_common_init_builtins): Ditto.
3595 (altivec_init_builtins): Ditto.
3596 (rs6000_common_init_builtins): Ditto.
3597 (enum reload_reg_type): New enum for simplifing reg classes.
3598 (rs6000_reload_register_type): Simplify register classes into GPR,
3599 Vector, and other registers. Altivec and VSX addresses in reload.
3600 (rs6000_secondary_reload_inner): Ditto.
3601 (rs6000_ira_cover_classes): New target hook, that returns the
3602 appropriate cover classes, based on -mvsx being used or not.
3603 (rs6000_secondary_reload_class): Add VSX support.
3604 (get_vec_cmp_insn): Delete, rewrite vector conditionals.
3605 (get_vsel_insn): Ditto.
3606 (rs6000_emit_vector_compare): Rewrite vector conditional support
3607 so that where we can, we use RTL operators, instead of blindly use
3609 (rs6000_emit_vector_select): Ditto.
3610 (rs6000_emit_vector_cond_expr): Ditto.
3611 (rs6000_emit_minmax): Directly generate min/max under altivec, vsx.
3612 (create_TOC_reference): Add -mdebug=addr support.
3613 (emit_frame_save): VSX loads/stores need register indexed addressing.
3615 * config/rs6000/rs6000.md: Include vector.md.
3617 * config/rs6000/t-rs6000 (MD_INCLUDES): Add vector.md.
3619 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
3620 support for V2DI, V2DF in logical, permute, select operations.
3622 * config/rs6000/rs6000.opt (-mvsx-scalar-double): Add new debug
3623 switch for vsx/power7.
3624 (-mvsx-scalar-memory): Ditto.
3625 (-mvsx-align-128): Ditto.
3626 (-mallow-movmisalign): Ditto.
3627 (-mallow-df-permute): Ditto.
3628 (-msched-groups): Ditto.
3629 (-malways-hint): Ditto.
3630 (-malign-branch-targets): Ditto.
3632 * config/rs6000/rs6000.h (IRA_COVER_CLASSES): Delete, use target
3634 (IRA_COVER_CLASSES_PRE_VSX): Cover classes if not -mvsx.
3635 (IRA_COVER_CLASSES_VSX): Cover classes if -mvsx.
3636 (rs6000_vector_reg_class): Delete.
3637 (rs6000_vsx_reg_class): Ditto.
3638 (enum rs6000_reg_class_enum): New enum for the constraints that
3639 vary based on target switches.
3640 (rs6000_constraints): New array to hold the register class for all
3641 of the register constraints that vary based on the switches used.
3642 (ALTIVEC_BUILTIN_*_UNS): Add unsigned builtin functions.
3643 (enum rs6000_builtins): Add unsigned varients for the builtin
3644 declarations returned by target hooks for expanding multiplies,
3645 select, and permute operations. Add VSX builtins.
3646 (enum rs6000_builtin_type_index): Add entries for VSX.
3647 (V2DI_type_node): Ditto.
3648 (V2DF_type_node): Ditto.
3649 (unsigned_V2DI_type_node): Ditto.
3650 (bool_long_type_node): Ditto.
3651 (intDI_type_internal_node): Ditto.
3652 (uintDI_type_internal_node): Ditto.
3653 (double_type_internal_node): Ditto.
3655 * config/rs6000/altivec.md (whole file): Move all expanders to
3656 vector.md from altivec.md. Rename insn matching functions to be
3658 (UNSPEC_VCMP*): Delete, rewrite vector comparisons.
3659 (altivec_vcmp*): Ditto.
3660 (UNSPEC_VPERM_UNS): New, add for unsigned types using vperm.
3661 (VM): New iterator for moves that includes the VSX types.
3662 (altivec_vperm_<mode>): Add VSX types. Add unsigned types.
3663 (altivec_vperm_<mode>_uns): New, for unsigned types.
3664 (altivec_vsel_*): Rewrite vector comparisons and predicate builtins.
3665 (altivec_eq<mode>): Ditto.
3666 (altivec_gt<mode>): Ditto.
3667 (altivec_gtu<mode>): Ditto.
3668 (altivec_eqv4sf): Ditto.
3669 (altivec_gev4sf): Ditto.
3670 (altivec_gtv4sf): Ditto.
3671 (altivec_vcmpbfp_p): Ditto.
3673 2009-07-23 Richard Earnshaw <rearnsha@arm.com>
3675 * arm.md (split for ior/xor with shift and zero-extend): Cast op3 to
3678 2009-07-23 Uros Bizjak <ubizjak@gmail.com>
3681 * config/i386/i386.c (output_387_ffreep): Rewrite to use
3682 ASM_SHORT instead of .word.
3683 * config/i386/i386.md (*tls_global_dynamic_64): Use ASM_SHORT
3684 instead of .word in asm template.
3686 2009-07-22 Vladimir Makarov <vmakarov@redhat.com>
3689 * ira-lives.c (bb_has_abnormal_call_pred): New function.
3690 (process_bb_node_lives): Use it.
3692 * ira.c (setup_cover_and_important_classes): Don't setup
3693 ira_important_class_nums. Add cover classes to the end of
3695 (cover_class_order, comp_reg_classes_func, reorder_important_classes):
3697 (find_reg_class_closure): Use reorder_important_classes.
3699 * config/i386/i386.h (IRA_COVER_CLASSES): Remove.
3701 * config/i386/i386.c (i386_ira_cover_classes): New function.
3702 (TARGET_IRA_COVER_CLASSES): Redefine.
3704 * doc/tm.texi (TARGET_IRA_COVER_CLASSES): Add a comment about
3705 importance of order of cover classes in the array.
3707 2009-07-22 Diego Novillo <dnovillo@google.com>
3709 * tree-pass.h (TDF_EH): Define.
3710 * gimple-pretty-print.c (dump_gimple_stmt): If FLAGS
3711 contains TDF_EH, print the EH region number holding GS.
3712 * tree-dump.c (dump_options): Add "eh".
3713 * doc/invoke.texi: Document it.
3715 2009-07-22 Doug Kwan <dougkwan@google.com>
3717 * config/arm/arm.md (subdi3) Copy non-reg values to DImode registers.
3719 2009-07-22 Michael Matz <matz@suse.de>
3721 PR tree-optimization/35229
3722 PR tree-optimization/39300
3724 * tree-ssa-pre.c (includes): Include tree-scalar-evolution.h.
3725 (inhibit_phi_insertion): New function.
3726 (insert_into_preds_of_block): Call it for REFERENCEs.
3727 (init_pre): Initialize and finalize scalar evolutions.
3728 * Makefile.in (tree-ssa-pre.o): Depend on tree-scalar-evolution.h .
3730 2009-07-22 Uros Bizjak <ubizjak@gmail.com>
3732 * config/i386/predicates.md (zero_extended_scalar_load_operand):
3733 Use CONST_VECTOR_NUNITS to determine number of elements.
3735 2009-07-22 Andreas Krebbel <krebbel1@de.ibm.com>
3737 * config/s390/constraints.md (ZQ, ZR, ZS, ZT): New constraints.
3738 (U, W): Constraints are now deprecated and will be removed if we
3740 * config/s390/s390.md (U, W): Replaced with ZQZR, ZSZT throughout
3742 ("prefetch"): Add the stcmh instruction for prefetching.
3743 * config/s390/s390.c (s390_symref_operand_p): Function moved. No
3745 (s390_short_displacement): Return always true if compiling for
3746 machines not providing the long displacement facility.
3747 (s390_mem_constraint): Support the new constraint letter Z.
3748 (s390_check_qrst_address): New function.
3750 2009-07-21 DJ Delorie <dj@redhat.com>
3752 * config/mep/mep.c (mep_legitimize_arg): Leave control registers
3755 2009-07-21 Jason Merrill <jason@redhat.com>
3757 * c-common.c (max_tinst_depth): Increase default to 1024.
3759 2009-07-21 Uros Bizjak <ubizjak@gmail.com>
3761 * config/i386/sse.md (vec_unpacku_float_hi_v4si): New expander.
3762 (vec_unpacku_float_lo_v4si): Ditto.
3764 2009-07-21 Uros Bizjak <ubizjak@gmail.com>
3767 * config/i386/sse.md (sse2_cvtudq2ps): New expander.
3768 (enum ix86_builtins): Add IX86_BUILTIN_CVTUDQ2PS.
3769 (builtin_description): Add __builtin_ia32_cvtudq2ps.
3770 (ix86_vectorize_builtin_conversion): Handle IX86_BUILTIN_CVTUDQ2PS.
3772 2009-07-21 Jakub Jelinek <jakub@redhat.com>
3774 PR tree-optimization/40813
3775 * tree-inline.c (copy_bb): Regimplify RHS after last stmt, not before
3778 2009-07-21 Kaz Kojima <kkojima@gcc.gnu.org>
3780 * config/sh/sh.c (sh_gimplify_va_arg_expr): Wrap the result
3781 with a NOP_EXPR if needed.
3783 2009-07-21 Paul Brook <paul@codesourcery.com>
3785 * tree-vectorizer.c (increase_alignment): Handle nested arrays.
3786 Terminate debug dump with newline.
3788 2009-07-20 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
3790 * pa.c (compute_zdepwi_operands): Limit deposit length to 32 - lsb.
3791 Cast "1" to unsigned HOST_WIDE_INT.
3792 (compute_zdepdi_operands): Limit maximum length to 64 bits. Limit
3793 deposit length to the maximum length - lsb. Extend length if
3794 HOST_BITS_PER_WIDE_INT is 32.
3796 2009-07-20 Olatunji Ruwase <tjruwase@google.com>
3798 * cgraph.h (constant_pool_htab): New function.
3799 (constant_descriptor_tree): Move from varasm.c.
3800 * varasm.c (constant_pool_htab): New function.
3801 (constant_descriptor_tree): Move to cgraph.h.
3803 2009-07-20 Olatunji Ruwase <tjruwase@google.com>
3805 * toplev.c: Invoke FINISH_UNIT callbacks before call to finalize().
3807 2009-07-20 Shujing Zhao <pearly.zhao@oracle.com>
3809 * Makefile.in (TREE_INLINE_H, tree-inline.o, cgraph.o): Remove
3812 2009-07-20 Xinliang David Li <davidxl@google.com>
3814 * dbgcnt.c (dbg_cnt_set_limit_by_name): Add length check.
3816 2009-07-20 Adam Nemet <anemet@caviumnetworks.com>
3818 * config/mips/mips.md (move_type): Add arith.
3819 (type): Handle arith.
3820 (zero_extendsidi2): Rename this into ...
3821 (*zero_extendsidi2): ... this. Don't match if ISA_HAS_EXT_INS.
3822 (zero_extendsidi2): New expander.
3823 (*zero_extendsidi2_dext): New pattern.
3825 2009-07-20 Nick Clifton <nickc@redhat.com>
3827 * config.gcc (mips64-*-*): Add definition of tm_defines in order
3828 to set MIPS_ABI_DEFAULT.
3829 * config/mips/vr.h (MIPS_ABI_DEFAULT): Remove definition.
3831 2009-07-20 Jakub Jelinek <jakub@redhat.com>
3833 * tree-object-size.c (addr_object_size): Handle unions with
3834 array in it as last field of structs in __bos (, 1) as __bos (, 0).
3836 PR tree-optimization/40792
3837 * tree.c (build_function_type_skip_args): Remove bogus assert.
3839 2009-07-20 Jan Hubicka <jh@suse.cz>
3840 Martin Jambor <mjambor@suse.cz>
3842 * cgraph.h (combined_args_to_skip): New field.
3843 * cgraph.c (cgraph_create_virtual_clone): Properly handle
3844 combined_args_to_skip and args_to_skip.
3845 * tree-inline.c (update_clone_info): New function.
3846 (tree_function_versioning): Call update_clone_info.
3847 * cgraphunit.c: (cgraph_materialize_clone): Dump materialized
3849 (cgraph_materialize_all_clones): More extensive dumping, working
3850 with combined_args_to_skip rather than args_to_skip.
3852 2009-07-20 Ira Rosen <irar@il.ibm.com>
3854 * tree-vectorizer.h (vectorizable_condition): Add parameters.
3855 * tree-vect-loop.c (vect_is_simple_reduction): Support COND_EXPR.
3856 (get_initial_def_for_reduction): Likewise.
3857 (vectorizable_reduction): Skip the check of first operand in case
3858 of COND_EXPR. Add check that it is outer loop vectorization if
3859 nested cycle was detected. Call vectorizable_condition() for
3860 COND_EXPR. If reduction epilogue cannot be created do not fail for
3861 nested cycles (if it is not double reduction). Assert that there
3862 is only one type in the loop in case of COND_EXPR. Call
3863 vectorizable_condition() to vectorize COND_EXPR.
3864 * tree-vect-stmts.c (vectorizable_condition): Update comment.
3865 Add parameters. Allow nested cycles if called from
3866 vectorizable_reduction(). Use reduction vector variable if provided.
3867 (vect_analyze_stmt): Call vectorizable_reduction() before
3868 vectorizable_condition().
3869 (vect_transform_stmt): Update call to vectorizable_condition().
3871 2009-07-20 Christian Bruel <christian.bruel@st.com>
3873 * config/sh/sh.opt (-mfmovd): Resurrect and document.
3874 * doc/invoke.texi (-mfmovd): Likewise.
3875 * config/sh/sh.h (TARGET_FMOVD, MASK_FMOVD): Remove default setting.
3877 2009-07-20 Jan Hubicka <jh@suse.cz>
3879 * tree-ssa-dce.c (remove_dead_phis): Only look for abnormal PHIs
3880 when handling SSA name.
3882 2009-07-19 Jan Hubicka <jh@suse.cz>
3884 PR tree-optimization/40676
3885 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Do renaming on all
3886 virtual PHIs in empty BBs.
3888 2009-07-18 Adam Nemet <anemet@caviumnetworks.com>
3890 * combine.c (make_compound_operation) <SUBREG>: If force_to_mode
3891 re-expanded the compound use gen_lowpart instead to convert to the
3894 2009-07-18 Adam Nemet <anemet@caviumnetworks.com>
3896 * combine.c (try_widen_shift_mode): Add COUNT, OUTER_CODE and
3897 OUTER_CONST arguments.
3898 <LSHIFTRT>: Use them to allow widening if the bits shifted in from
3899 the new wider mode will be masked off.
3900 (simplify_shift_const_1): Adjust calls to try_widen_shift_mode.
3902 2009-07-18 Adam Nemet <anemet@caviumnetworks.com>
3904 * combine.c (try_widen_shift_mode) <LSHIFTRT>: Allow widening if the
3905 high-order bits are zero.
3907 2009-07-18 Adam Nemet <anemet@caviumnetworks.com>
3909 * combine.c (simplify_shift_const_1): Split code to determine
3911 (try_widen_shift_mode): ... here. Allow widening for ASHIFTRT if the
3912 new bits shifted in are identical to the old sign bit.
3914 2009-07-18 Richard Guenther <rguenther@suse.de>
3917 * gimplify.c (gimplify_call_expr): Reject code using results from
3918 functions returning void.
3920 2009-07-18 Richard Sandiford <r.sandiford@uk.ibm.com>
3922 * doc/md.texi: Document the new PowerPC "es" constraint.
3923 Document that "m" can include automodified addresses on this target,
3924 and explain how %U must be used. Extend the "Q" and "Z" documentation
3925 to suggest "es" as well as "m".
3926 * config/rs6000/constraints.md (es): New memory constraint.
3927 (Q, Z): Update strings to match new documentation.
3929 2009-07-18 Richard Sandiford <r.sandiford@uk.ibm.com>
3931 * config/rs6000/rs6000.c (rs6000_mode_dependent_address): Allow any
3932 offset from virtual_stack_vars_rtx and arg_pointer_rtx.
3933 * config/rs6000/predicates.md (volatile_mem_operand): Use
3934 offsettable_nonstrict_memref_p.
3935 * config/rs6000/rs6000.md (*floatsidf2_internal): Remove split check.
3936 (*floatunssidf2_internal): Likewise.
3937 (*fix_truncdfsi2_internal): Likewise.
3938 (*fix_trunctfsi2_internal): Likewise.
3940 2009-07-17 Anatoly Sokolov <aesok@post.ru>
3942 * config/avr/avr-devices.c (avr_mcu_t): Add atmega8u2, atmega16u2 and
3944 * config/avr/t-avr (MULTILIB_MATCHES): (Ditto.).
3946 2009-07-17 Richard Guenther <rguenther@suse.de>
3949 * tree-pass.h (pass_diagnose_omp_blocks): Declare.
3950 (pass_warn_unused_result): Likewise.
3951 (TODO_set_props): Remove.
3952 * omp-low.c (diagnose_omp_structured_block_errors): Change to
3954 (pass_diagnose_omp_blocks): Define.
3955 * c-decl.c (pop_file_scope): Do not finalize the CU here.
3956 (c_gimple_diagnostics_recursively): Remove.
3957 (finish_function): Do not call it.
3958 (c_write_global_declarations): Continue after errors.
3959 Finalize the CU here.
3960 * c-gimplify.c (c_genericize): Do not gimplify here.
3961 * c-common.c (c_warn_unused_result): Move ...
3962 * tree-cfg.c (do_warn_unused_result): ... here.
3963 (run_warn_unused_result): New function.
3964 (gate_warn_unused_result): New function.
3965 (pass_warn_unused_result): New pass.
3966 * c-common.h (c_warn_unused_result): Remove.
3967 * flags.h (flag_warn_unused_result): Declare.
3968 * c-opts.c (c_common_init_options): Enable flag_warn_unused_result.
3969 * opts.c (flag_warn_unused_result): Initialize to false.
3970 * toplev.c (compile_file): Add comment.
3971 * omp-low.c (create_omp_child_function): Do not register
3972 the function with the frontend.
3973 (diagnose_omp_structured_block_errors): Prepare to be
3974 called as optimization pass.
3975 (gate_diagnose_omp_blocks): New function.
3976 (pass_diagnose_omp_blocks): New pass.
3977 * cgraph.h (cgraph_optimize): Remove.
3978 (cgraph_analyze_function): Likewise.
3979 * cgraph.c (cgraph_add_new_function): Gimplify C++ thunks.
3980 * cgraphunit.c (cgraph_lower_function): Lower nested functions
3981 before their parents here.
3982 (cgraph_finalize_function): Not here.
3983 (cgraph_analyze_function): Gimplify functions here.
3984 (cgraph_finalize_compilation_unit): Continue after errors.
3985 Optimize the callgraph from here.
3986 (cgraph_optimize): Make static.
3987 * langhooks.c (write_global_declarations): Finalize the CU.
3988 * gimplify.c (gimplify_asm_expr): Do not emit ASMs with errors.
3989 (gimplify_function_tree): Assert we gimplify only once.
OSDN Git Service
