1 2007-10-09 James E. Wilson <wilson@specifix.com>
3 PR tree-optimization/33655
5 * tree-sra.c (bitfield_overlaps_p): When fld->element is INTEGER_CST,
6 convert it to bitsizetype before size_binop call.
8 2007-10-09 Alexandre Oliva <aoliva@redhat.com>
10 PR tree-optimization/33572
11 * tree-inline.c (update_ssa_across_abnormal_edges): Tolerate
12 the absence of a corresponding edge from the exit block.
14 2007-10-09 Alexandre Oliva <aoliva@redhat.com>
17 * tree-sra.c (instantiate_element): Use BYTES_BIG_ENDIAN for
19 (sra_build_assignment): Likewise. Set up mask depending on
21 (sra_build_bf_assignment): Use BYTES_BIG_ENDIAN. Don't overflow
23 (sra_build_elt_assignment): Don't view-convert from signed to
25 (sra_explode_bitfield_assignment): Use bit-field type if
26 possible. Use BYTES_BIG_ENDIAN.
28 2007-10-08 Alexandre Oliva <aoliva@redhat.com>
31 * tree-sra.c (scalarize_lsdt): Fix thinko in testing whether
32 the original stmt can throw.
33 (sra_build_bf_assignment): Fix type mismatch when applying negated
36 2007-10-08 Geoffrey Keating <geoffk@apple.com>
38 * config/darwin.h (DARWIN_REGISTER_TARGET_PRAGMAS): Register
39 '#pragma mark' to be executed at preprocessing time.
41 2007-10-08 Ollie Wild <aaw@google.com>
43 * varasm.c (compare_constant): Removed call to
44 lang_hooks.expand_constant.
45 (copy_constants): Removed call to lang_hooks.expand_constant.
46 (compute_reloc_for_constant): Removed call to
47 lang_hooks.expand_constant.
48 (output_addressed_constants): Removed call to
49 lang_hooks.expand_constant.
50 (constructor_static_from_elts_p): Removed call to
51 lang_hooks.expand_constant.
52 (output_constant): Removed calls to lang_hooks.expand_constant.
53 * langhooks.h (struct lang_hooks): Removed field expand_constant.
54 * langhooks-def.h (lhd_return_tree): Removed.
55 (LANG_HOOKS_EXPAND_CONSTANT): Removed.
56 (LANG_HOOKS_INITIALIZER): Removed LANG_HOOKS_EXPAND_CONSTANT.
57 * langhooks.c (lhd_return_tree): Removed.
59 2007-10-08 Mark Shinwell <shinwell@codesourcery.com>
61 * combine.c (setup_incoming_promotions): Ensure that
62 arguments that have not undergone mode promotions do not
63 incorrectly get marked as being sign- or zero-extended.
65 2007-10-08 Richard Guenther <rguenther@suse.de>
70 * fold-const.c (fold_binary): Use correct types in folding
71 of a * (1 << b) to (a << b). Likewise for ~A & ~B to ~(A | B)
72 and building of RROTATE_EXPR.
74 2007-10-08 Richard Guenther <rguenther@suse.de>
79 * fold-const.c (fold_binary): Use the correct types when
80 folding (A | CST1) & CST2 to (A & CST2) | (CST1 & CST2).
81 (fold_binary): Use the correct types when folding
83 (fold_unary): Use the correct types when folding ~(X).
85 2007-10-08 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
87 * doc/invoke.texi (Wall): fix formatting issues.
89 2007-10-07 Richard Sandiford <rsandifo@nildram.co.uk>
91 * simplify-rtx.c (simplify_binary_operation_1): Canonicalize
92 truncated shift counts.
94 2007-10-07 Kazu Hirata <kazu@codesourcery.com>
96 * config/m68k/m68k.c, config/m68k/m68k.md: Use the assembly
97 syntax for ASSEMBLER_DIALECT.
98 * config/m68k/m68k.h (ASSEMBLER_DIALECT): New.
100 2007-10-06 Eric Botcazou <ebotcazou@adacore.com>
101 Nathan Froyd <froydnj@codesourcery.com>
103 * dwarf2out.c (dwarf2out_frame_init): Check for DWARF2_FRAME_INFO
104 when determining whether to record INCOMING_RETURN_ADDR_RTX.
106 2007-10-06 Eric Botcazou <ebotcazou@adacore.com>
109 2007-02-12 Eric Botcazou <ebotcazou@adacore.com>
111 * tree.h (DECL_IGNORED_P): Document further effect for FUNCTION_DECL.
112 * cgraphunit.c (cgraph_expand_function): If DECL_IGNORED_P is set on
113 the function, temporarily point the debug interface to the null one.
115 2007-10-06 Alexandre Oliva <aoliva@redhat.com>
117 PR tree-optimization/33655
119 * tree-sra.c (bitfield_overlaps_p): Handle array and complex
122 2007-10-06 Alexandre Oliva <aoliva@redhat.com>
124 PR tree-optimization/33572
125 * tree-cfg.c (verify_stmts): Check for missing PHI defs.
126 * tree-inline.c (update_ssa_across_eh_edges): Renamed to...
127 (update_ssa_across_abnormal_edges): ... this. Set slots in the
129 (copy_edges_for_bb): Handle nonlocal label edges.
130 (make_nonlocal_label_edges): Deleted.
131 (optimize_inline_calls): Don't call it.
133 2007-10-05 Hans-Peter Nilsson <hp@axis.com>
135 * gthr-single.h: Revert last change.
137 2007-10-05 Michael Matz <matz@suse.de>
140 * lower-subreg.c (decompose_multiword_subregs): Use
141 validate_unshare_change().
143 2007-10-05 Peter Bergner <bergner@vnet.ibm.com>
145 * ra-conflict.c: Include "sparseset.h".
146 (conflicts): Change to HOST_WIDEST_FAST_INT.
147 (allocnos_live): Redefine variable as a sparseset.
148 (SET_ALLOCNO_LIVE, CLEAR_ALLOCNO_LIVE, GET_ALLOCNO_LIVE): Delete macros.
149 (allocno_row_words): Removed global variable.
150 (partial_bitnum, max_bitnum, adjacency_pool, adjacency): New variables.
151 (CONFLICT_BITNUM, CONFLICT_BITNUM_FAST): New defines.
152 (conflict_p, set_conflict_p, set_conflicts_p): New functions.
153 (record_one_conflict_between_regnos): Cache allocno values and reuse.
155 (record_one_conflict): Update uses of allocnos_live to use
156 the sparseset routines. Use set_conflicts_p.
157 (mark_reg_store): Likewise.
158 (set_reg_in_live): Likewise.
159 (global_conflicts): Update uses of allocnos_live.
160 Use the new adjacency list to visit an allocno's neighbors
161 rather than iterating over all possible allocnos.
162 Call set_conflicts_p to setup conflicts rather than adding
164 * global.c: Comments updated.
165 (CONFLICTP): Delete define.
166 (regno_compare): New function. Add prototype.
167 (global_alloc): Sort the allocno to regno mapping according to
168 which basic blocks the regnos are referenced in. Modify the
169 conflict bit matrix to a compressed triangular bitmatrix.
170 Only allocate the conflict bit matrix and adjacency lists if
171 we are actually going to allocate something.
172 (expand_preferences): Use conflict_p. Update uses of allocnos_live.
173 (prune_preferences): Use the FOR_EACH_CONFLICT macro to visit an
174 allocno's neighbors rather than iterating over all possible allocnos.
175 (mirror_conflicts): Removed function.
176 (dump_conflicts): Iterate over regnos rather than allocnos so
177 that all dump output will be sorted by regno number.
178 Use the FOR_EACH_CONFLICT macro.
179 * ra.h: Comments updated.
180 (conflicts): Update prototype to HOST_WIDEST_FAST_INT.
181 (partial_bitnum, max_bitnum, adjacency, adjacency_pool): Add prototypes.
182 (ADJACENCY_VEC_LENGTH, FOR_EACH_CONFLICT): New defines.
183 (adjacency_list_d, adjacency_iterator_d): New types.
184 (add_neighbor, adjacency_iter_init, adjacency_iter_done,
185 adjacency_iter_next, regno_basic_block): New static inline functions.
186 (EXECUTE_IF_SET_IN_ALLOCNO_SET): Removed define.
187 (conflict_p): Add function prototype.
188 * sparseset.h, sparseset.c: New files.
189 * Makefile.in (OBJS-common): Add sparseset.o.
190 (sparseset.o): New rule.
192 2007-10-05 Richard Guenther <rguenther@suse.de>
195 * fold-const.c (fold_unary): Do not fold (long long)(int)ptr
198 2007-10-05 Michael Matz <matz@suse.de>
201 * function.c (match_asm_constraints_1): Check for input
202 being used in the outputs.
204 2007-10-05 Richard Guenther <rguenther@suse.de>
206 * tree-cfg.c (verify_gimple_expr): Accept OBJ_TYPE_REF.
208 2007-10-05 Richard Sandiford <rsandifo@nildram.co.uk>
211 * config/mips/mips.c (mips_register_move_cost): Rewrite to use
212 subset checks. Make the cost of FPR -> FPR moves depend on
213 mips_mode_ok_for_mov_fmt_p.
215 2007-10-04 Doug Kwan <dougkwan@google.com>
217 * gthr-posix.h (__gthread_cond_broadcast, __gthread_cond_wait,
218 __gthread_cond_wait_recursive): Add to extend interface for POSIX
219 conditional variables. (__GTHREAD_HAS_COND): Macro defined to signify
220 support of conditional variables.
221 * gthr-posix95.h (__gthread_cond_broadcast, __gthread_cond_wait,
222 __gthread_cond_wait_recursive): Add to extend interface for POSIX
223 conditional variables. (__GTHREAD_HAS_COND): Macro defined to signify
224 support of conditional variables.
225 * gthr-single.h (__gthread_cond_broadcast, __gthread_cond_wait,
226 __gthread_cond_wait_recursive): Add to extend interface for POSIX
227 conditional variables.
228 * gthr.h: Update comments to document new interface.
230 2007-10-04 Geoffrey Keating <geoffk@apple.com>
232 * cgraphunit.c (cgraph_build_static_cdtor): Don't set
235 2007-10-04 Anatoly Sokolov <aesok@post.ru>
237 * config/avr/avr.c (expand_epilogue): Don't set RTX_FRAME_RELATED_P.
239 2007-10-04 Richard Guenther <rguenther@suse.de>
242 * tree-cfg.c (verify_gimple_expr): Operand one of POINTER_PLUS_EXPR
243 does not need to be of INTEGER_TYPE.
244 (verify_gimple_2): New function split out from ...
245 (verify_gimple_1): ... here. ICE if there was an error during
248 2007-10-04 Michael Matz <matz@suse.de>
250 PR rtl-optimization/33653
251 * dce.c (deletable_insn_p_1): Use volatile_refs_p().
252 * dse.c (scan_insn): Same.
254 2007-10-04 Kazu Hirata <kazu@codesourcery.com>
256 * config.gcc: Remove USE_GAS for m68k targets.
258 2007-10-04 Richard Guenther <rguenther@suse.de>
260 PR tree-optimization/33627
261 * tree-gimple.h (canonicalize_cond_expr_cond): Declare.
262 * tree-gimple.c (canonicalize_cond_expr_cond): New function,
264 * tree-ssa-forwprop.c (combine_cond_expr_cond): ... here.
265 * tree-ssa-ifcombine.c (ifcombine_iforif): Use it.
267 2007-10-04 Anatoly Sokolov <aesok@post.ru>
269 * config/avr/avr.c (commands_in_file, commands_in_prologues,
270 commands_in_epilogues): Remove variables.
271 (avr_file_start): Remove unneded initializations of commands_in_file,
272 commands_in_prologues and commands_in_epilogues variables.
273 (avr_file_end): Remove dead code.
275 2007-10-04 Kazu Hirata <kazu@codesourcery.com>
277 * config/m68k/m68k.c (m68k_output_movem): Use the MOTOROLA if
280 2007-10-03 Richard Sandiford <rsandifo@nildram.co.uk>
283 * config/mips/mips-protos.h (mips_split_64bit_move): Rename to...
284 (mips_split_doubleword_move): ...this.
285 * config/mips/mips.c (mips_subword): Extend to handle 64-bit words;
286 use natural endianness for multi-format FPR values.
287 (mips_split_64bit_move): Rename to...
288 (mips_split_doubleword_move): ...this and extend to 64-bit words.
289 Use move_doubleword_fpr* patterns for moves involving FPRs.
290 (mips_save_reg): Update the call to mips_split_64bit_move.
291 (mips_secondary_reload_class): Return NO_REGS for any reload of a
292 nonzero constant into an FPR if the constant can be forced to memory.
293 * config/mips/mips.md: Update the splitter calls to
294 mips_split_64bit_move.
295 (UNSPEC_LOAD_DF_LOW): Rename to...
296 (UNSPEC_LOAD_LOW): ...this.
297 (UNSPEC_LOAD_DF_HIGH): Rename to...
298 (UNSPEC_LOAD_HIGH): ...this.
299 (UNSPEC_STORE_DF_HIGH): Rename to...
300 (UNSPEC_STORE_WORD): ...this.
301 (SPLITF): New mode iterator.
302 (HALFMODE): New mode attribute.
303 (movtf): New expander.
304 (*movtf_internal): New define_insn_and_split.
305 (move_doubleword_fpr<mode>): New expander.
306 (load_df_low, load_df_high, store_df_high, mthc1, mfhc1): Replace
308 (load_low<mode>, load_high<mode>, store_word<mode>, mthc1<mode>)
309 (mfhc1<mode>): ...these more general patterns.
311 2007-10-03 Alexandre Oliva <aoliva@redhat.com>
313 * cfgrtl.c (rtl_block_ends_with_call_p): Skip notes at the end.
315 2007-10-03 Alexandre Oliva <aoliva@redhat.com>
317 * gcse.c (hash_scan_set): Insert set in insn before note at
318 the end of basic block.
320 2007-10-03 Sebastian Pop <sebastian.pop@amd.com>
322 PR tree-optimization/33576
323 * testsuite/gcc.dg/tree-ssa/pr33576.c: New.
324 * tree-loop-linear.c (linear_transform_loops): Call remove_iv.
325 * lambda.h (lambda_loopnest_to_gcc_loopnest): New parameter.
326 (remove_iv): Declared.
327 * lambda-code.c (remove_iv): Not static.
328 (lambda_loopnest_to_gcc_loopnest): New parameter remove_ivs.
329 Don't remove ivs there, save ivs in the buffer.
331 2007-10-03 Jason Merrill <jason@redhat.com>
334 * tree-eh.c (same_handler_p): New fn.
335 (optimize_double_finally): New fn.
336 (refactor_eh_r): New fn.
337 (refactor_eh): New fn.
338 (pass_refactor_eh): New pass.
339 * tree-pass.h: Declare it.
340 * passes.c (init_optimization_passes): Add it.
342 2007-10-03 Doug Kwan <dougkwan@google.com>
343 Richard Guenther <rguenther@suse.de>
346 * dwarf2out.c (reference_to_unused): Disable sanity checking,
347 be conservative instead.
349 2007-10-03 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
352 * doc/invoke.texi (-fwhole-program): Document that Fortran
353 doesn't support this option.
355 2007-10-02 Richard Sandiford <rsandifo@nildram.co.uk>
358 * expr.c (expand_expr_addr_expr_1): Pass CONSTRUCTORs to
361 2007-10-02 David Daney <ddaney@avtrex.com>
363 * config/mips/mips.md (sync_compare_and_swap<mode>): Handle compare
364 against constant zero.
365 * config/mips/mips.h (MIPS_COMPARE_AND_SWAP): Handle constant zero
368 2007-09-02 Kenneth Zadeck <zadeck@naturalbridge.com>
370 * ra-conflict.c: New file.
372 * reload.c (push_reload, find_dummy_reload): Change DF_RA_LIVE
373 usage to DF_LIVE usage.
374 * rtlanal.c (subreg_nregs_with_regno): New function.
375 * df-scan.c (df_def_record_1, df_uses_record): Add code to set
376 DF_REF_EXTRACT, DF_REF_STRICT_LOWER_PART, and DF_REF_SUBREG flags.
377 (df_has_eh_preds): Removed.
378 (df_bb_refs_collect, df_bb_refs_collect, df_bb_refs_collect,
379 df_exit_block_uses_collect): Changed call from df_has_eh_preds to
381 * global.c (allocno, max_allocno, conflicts, allocno_row_words,
382 reg_allocno, EXECUTE_IF_SET_IN_ALLOCNO_SET): Moved to ra.h
383 (SET_ALLOCNO_LIVE, CLEAR_ALLOCNO_LIVE): Moved to ra-conflicts.c.
384 (regs_set, record_one_conflict, record_conflicts, mark_reg_store,
385 mark_reg_clobber, mark_reg_conflicts, mark_reg_death): Deleted.
386 (global_alloc): Turn off rescanning insns after call to
387 global_conflicts and added call to set_preferences.
388 (global_conflicts): Moved to ra-alloc.c.
389 (set_preferences_1, set_preferences): New function.
390 (mirror_conflicts): Changed types for various variables.
391 (mark_elimination): Change DF_RA_LIVE
392 usage to DF_LIVE usage.
393 (build_insn_chain): Rewritten from scratch and made local.
394 (print_insn_chain, print_insn_chains): New functions.
395 (dump_conflicts): Do not print conflicts for fixed_regs.
396 (rest_of_handle_global_alloc): Turn off insn rescanning.
397 * hard-reg-set.h: Fixed comment.
398 * local-alloc.c (update_equiv_regs): Change DF_RA_LIVE
399 usage to DF_LIVE usage and delete refs to TOP sets.
400 (block_alloc): Mark regs as live if they are in the artificial
401 defs at top of block.
402 (find_stack_regs): New function.
403 (rest_of_handle_local_alloc): Changed urec problem to live
404 problem and do not turn off df rescanning.
405 * df.h (DF_UREC, DF_UREC_BB_INFO, DF_LIVE_TOP, DF_RA_LIVE_IN,
406 DF_RA_LIVE_TOP, DF_RA_LIVE_OUT, df_urec_bb_info, df_urec,
407 df_urec_add_problem, df_urec_get_bb_info, df_has_eh_preds): Removed.
408 (DF_CHAIN, DF_NOTE, DF_CHAIN): Renumbered.
409 (DF_REF_EXTRACT, DF_REF_STRICT_LOWER_PART, DF_REF_SUBREG): New
410 fields in df_ref_flags. The rest have been renumbered.
411 * init-regs.c (initialize_uninitialized_regs): Enhanced debugging
413 * rtl.h (subreg_nregs_with_regno): New function.
414 * df-problems.c: (df_get_live_out, df_get_live_in,
415 df_get_live_top): Removed reference to DF_RA_LIVE.
416 (df_lr_reset, df_lr_transfer_function, df_live_free_bb_info,
417 df_live_alloc, df_live_reset, df_live_local_finalize,
418 df_live_free): Make top set only if different from in set.
419 (df_lr_top_dump, df_live_top_dump): Only print top set if
420 different from in set.
421 (df_lr_bb_local_compute): Removed unnecessary check.
422 (df_urec_problem_data, df_urec_set_bb_info, df_urec_free_bb_info,
423 df_urec_alloc, df_urec_mark_reg_change, earlyclobber_regclass,
424 df_urec_check_earlyclobber, df_urec_mark_reg_use_for_earlyclobber,
425 df_urec_mark_reg_use_for_earlyclobber_1, df_urec_bb_local_compute,
426 df_urec_local_compute, df_urec_init, df_urec_local_finalize,
427 df_urec_confluence_n, df_urec_transfer_function, df_urec_free,
428 df_urec_top_dump, df_urec_bottom_dump, problem_UREC,
429 df_urec_add_problem): Removed.
430 (df_simulate_fixup_sets): Changed call from df_has_eh_preds to
432 * Makefile.in (ra-conflict.o, ra.h): New dependencies.
433 * basic_block.h (bb_has_abnormal_pred): New function.
434 * reload1.c (compute_use_by_pseudos): Change DF_RA_LIVE
435 usage to DF_LIVE usage.
437 2007-10-02 Revital Eres <eres@il.ibm.com>
439 * config/rs6000/predicates.md (easy_vector_constant): Return false
440 for 750CL paired vectors.
441 * config/rs6000/paired.md (movv2sf_paired): Fix move of easy
443 (vec_initv2sf): Add new description.
444 (vconcatsf): Likewise.
445 * config/rs6000/rs6000-protos.h: Declare paired_expand_vector_init.
446 * config/rs6000/rs6000.c (paired_expand_vector_init): New function.
448 2007-10-01 Alexandre Oliva <aoliva@redhat.com>
450 * tree-ssa-sink.c (sink_code_in_bb): Don't stop sinking after
451 sinking the last stmt in a BB.
453 2007-10-01 Alexandre Oliva <aoliva@redhat.com>
456 * tree-sra.c (struct sra_elt): Add in_bitfld_block.
457 (sra_hash_tree): Handle BIT_FIELD_REFs.
458 (sra_elt_hash): Don't hash bitfld blocks.
459 (sra_elt_eq): Skip them in parent compares as well. Handle
461 (build_element_name_1): Handle BIT_FIELD_REFs.
462 (instantiate_element): Propagate nowarn from parents. Create
463 BIT_FIELD_REF for variables that are widened by scalarization.
464 Gimple-zero-initialize all bit-field variables that are not
465 part of parameters that are going to be scalarized on entry.
466 (instantiate_missing_elements_1): Return the sra_elt.
467 (canon_type_for_field): New.
468 (try_instantiate_multiple_fields): New. Infer widest possible
469 access mode from decl or member type, but clip it at word
470 size, and only widen it if a field crosses an alignment
472 (instantiate_missing_elements): Use them.
473 (generate_one_element_ref): Handle BIT_FIELD_REFs.
474 (scalar_bitfield_p): New.
475 (sra_build_assignment): Optimize assignments from scalarizable
476 BIT_FIELD_REFs. Use BITS_BIG_ENDIAN to determine shift
479 (sra_build_bf_assignment): New. Optimize assignments to
480 scalarizable BIT_FIELD_REFs.
481 (sra_build_elt_assignment): New. Optimize BIT_FIELD_REF
482 assignments to full variables.
483 (generate_copy_inout): Use the new macros and functions.
484 (generate_element_copy): Likewise. Handle bitfld differences.
485 (generate_element_zero): Don't recurse for blocks. Use
486 sra_build_elt_assignment.
487 (generate_one_element_init): Take elt instead of var. Use
488 sra_build_elt_assignment.
489 (generate_element_init_1): Adjust.
490 (bitfield_overlap_info): New struct.
491 (bitfield_overlaps_p): New.
492 (sra_explode_bitfield_assignment): New. Adjust widened
493 variables to account for endianness.
494 (sra_sync_for_bitfield_assignment): New.
495 (scalarize_use): Re-expand assignment to/from scalarized
496 BIT_FIELD_REFs. Explode or sync needed members for
497 BIT_FIELD_REFs accesses or assignments. Use REPLDUP.
498 (scalarize_copy): Use REPLDUP.
499 (scalarize_ldst): Move assert before dereference. Adjust EH
501 (dump_sra_elt_name): Handle BIT_FIELD_REFs.
503 2007-10-01 Paolo Bonzini <bonzini@gnu.org>
505 * simplify-rtx.c (comparison_result, simplify_relational_operation_1):
506 Rename CR_* constants to CMP_*. Fix spacing.
508 2007-10-01 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
511 * Makefile.in (build_html_dir/gccinstall): gccinstall.texi needs
512 to be processed with the special script doc/install.texi2html.
514 2007-09-30 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
516 * doc/invoke.texi (Wall): List the options enabled by Wall.
517 (Wstrict-aliasing): Add missing @option.
519 2007-09-30 Richard Sandiford <rsandifo@nildram.co.uk>
521 * config/mips/mips.c (mips_split_64bit_move): Use gen_rtx_REG_offset
522 rather than gen_lowpart to change a register from DImode to DFmode.
523 (mips_cannot_change_mode_class): Only allow FPRs to change mode if
524 both FROM and TO are integer modes that are no bigger than 4 bytes.
525 (mips_mode_ok_for_mov_fmt_p): New function.
526 (mips_preferred_reload_class): Use it instead of FLOAT_MODE_P.
527 (mips_secondary_reload_class): Tweak formatting and comments.
528 Use reg_class_subset_p instead of direct comparisons with
529 classes. Only allow direct FPR<->FPR moves for modes that
530 satisfy mips_mode_ok_for_mov_fmt_p. Only allow loads and stores
531 for 4- and 8-byte types. Handle reloads in which X is an FPR.
532 * config/mips/mips.md (*movdi_gp32_fp64): Remove f<-f alternative.
533 (*movdi_64bit): Likewise.
534 (*movsi_internal): Likewise.
535 (*movhi_internal): Likewise.
536 (*movqi_internal): Likewise.
538 2007-09-30 Diego Novillo <dnovillo@google.com>
541 * tree-ssa-ter.c (is_replaceable_p): Return false if STMT may
544 2007-09-30 Uros Bizjak <ubizjak@gmail.com>
546 PR tree-optimization/33597
547 * tree-vect-analyze.c (vect_build_slp_tree): Check if optab handler
548 for LSHIFT_EXPR and RSHIFT_EXPR is available for vec_mode.
550 2007-09-28 Uros Bizjak <ubizjak@gmail.com>
552 * config/i386/i386.c (ix86_expand_move): Use can_create_pseudo_p ()
553 instead of variants of (!reload_in_progress && !reload_completed).
554 (x86_expand_vector_move): Ditto.
556 2007-09-28 Ollie Wild <aaw@google.com>
559 2007-09-27 Ollie Wild <aaw@google.com>
561 * varasm.c (compare_constant): Removed call to
562 lang_hooks.expand_constant.
563 (copy_constants): Removed call to lang_hooks.expand_constant.
564 (compute_reloc_for_constant): Removed call to
565 lang_hooks.expand_constant.
566 (output_addressed_constants): Removed call to
567 lang_hooks.expand_constant.
568 (constructor_static_from_elts_p): Removed call to
569 lang_hooks.expand_constant.
570 (output_constant): Removed calls to lang_hooks.expand_constant.
571 * langhooks.h (struct lang_hooks): Removed field expand_constant.
572 * langhooks-def.h (lhd_return_tree): Removed.
573 (LANG_HOOKS_EXPAND_CONSTANT): Removed.
574 (LANG_HOOKS_INITIALIZER): Removed LANG_HOOKS_EXPAND_CONSTANT.
575 * langhooks.c (lhd_return_tree): Removed.
577 2007-09-28 Andrew Pinski <andrew_pinski@playstation.sony.com>
580 * config/spu/spu.c (spu_expand_insv): Call copy_rtx on the second
581 argument to gen_selb.
583 2007-09-28 Chao-ying Fu <fu@mips.com>
585 * libgcc-std.ver: Add fixed-point routines to GCC_4.3.0 section.
586 * doc/libgcc.texi (Fixed-point fractional library routines):
587 Fix typos for neg and cmp functions.
589 2007-09-28 Michael Matz <matz@suse.de>
591 PR rtl-optimization/33552
592 * function.c (match_asm_constraints_1): Check for overlap in
593 inputs and replace all occurences.
595 2007-09-28 Richard Sandiford <rsandifo@nildram.co.uk>
597 * config/mips/mips.c (override_options): Fix comment typo.
599 2007-09-28 Jie Zhang <jie.zhang@analog.com>
601 * config.gcc (bfin*-linux-uclibc*): Set extra_parts
602 to "crtbegin.o crtbeginS.o crtend.o crtendS.o".
603 * config/bfin/t-bfin-linux (crti.o): Don't build.
605 (EXTRA_MULTILIB_PARTS): Remove crti.o and crtn.o.
606 * config/bfin/t-bfin-uclinux (crti.o): Don't build.
608 (EXTRA_MULTILIB_PARTS): Remove crti.o and crtn.o.
610 2007-09-27 Ollie Wild <aaw@google.com>
612 * varasm.c (compare_constant): Removed call to
613 lang_hooks.expand_constant.
614 (copy_constants): Removed call to lang_hooks.expand_constant.
615 (compute_reloc_for_constant): Removed call to
616 lang_hooks.expand_constant.
617 (output_addressed_constants): Removed call to
618 lang_hooks.expand_constant.
619 (constructor_static_from_elts_p): Removed call to
620 lang_hooks.expand_constant.
621 (output_constant): Removed calls to lang_hooks.expand_constant.
622 * langhooks.h (struct lang_hooks): Removed field expand_constant.
623 * langhooks-def.h (lhd_return_tree): Removed.
624 (LANG_HOOKS_EXPAND_CONSTANT): Removed.
625 (LANG_HOOKS_INITIALIZER): Removed LANG_HOOKS_EXPAND_CONSTANT.
626 * langhooks.c (lhd_return_tree): Removed.
628 2007-09-27 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
631 * expr.c (emit_group_load_1): Split constant double when destination
632 length is half source length.
634 2007-09-27 Richard Sandiford <rsandifo@nildram.co.uk>
636 * config/mips/mips.h (PTF_AVOID_BRANCHLIKELY): New macro.
637 (mips_cpu_info): Add tune_flags.
638 (GENERATE_BRANCHLIKELY): Remove TARGET_SR71K check.
639 * config/mips/mips.c (mips_cpu_info_table): Add tune_flags fields.
641 (override_options): Remove deprecation code. Use branch-likely
642 instructions for optimize_size or if the tuning flags do not
643 suggest otherwise. Tweak warning.
644 (mips_matching_cpu_name_p, mips_parse_cpu): Use ARRAY_SIZE.
646 2007-09-27 Matthias Klose <doko@ubuntu.com>
648 * config/i386/t-linux64 (MULTILIB_OSDIRNAMES): Use ../lib32 as the
649 multilib osdirname if it exists.
650 * config/rs6000/t-linux64 (MULTILIB_OSDIRNAMES): Likewise.
652 2007-09-27 Ian Lance Taylor <iant@google.com>
654 PR tree-optimization/33565
655 * tree-ssa-loop-ch.c (copy_loop_headers): Set TREE_NO_WARNING on
656 assignments of comparisons.
657 * tree-ssa-sccvn.c (simplify_binary_expression): Add stmt
658 parameter. Change caller. Defer overflow warnings around call to
660 * fold-const.c (fold_undefer_overflow_warnings): Don't warn if
661 TREE_NO_WARNING is set on the statement.
662 * tree-ssa-forwprop.c
663 (tree_ssa_forward_propagate_single_use_vars): Don't test
664 TREE_NO_WARNING when calling fold_undefer_overflow_warnings.
665 * tree-cfg.c (fold_cond_expr_cond): Likewise.
667 2007-09-27 Joseph Myers <joseph@codesourcery.com>
669 * config/rs6000/rs6000.c (rs6000_legitimize_address): Do not
670 reduce offset by units of 0x10000 for SPE vector modes or modes
671 used with E500 double instructions.
673 2007-09-04 Paolo Bonzini <bonzini@gnu.org>
675 * simplify-rtx.c (comparison_result): New.
676 (simplify_const_relational_operation): Use it instead of the five
677 "equal|op[01]ltu?" variables; consequently remove redundant "else"s.
678 Improve bounds-checking optimizations; remove subsumed POPCOUNT
679 optimizations. Extract nonzero_address_p optimizations into a
680 separate "if" together with optimizations where op1 is const0_rtx.
681 Optimize comparing an IOR with zero. Simplify op0 RELOP op0 for
682 floating-point arguments too when appropriate. Hoist test for ABS
683 outside the final switch statement.
684 * cse.c (fold_rtx): Don't look for an IOR equivalent of
685 folded_arg0 if we found a constant equivalent. Remove
686 transformations done in simplify-rtx.c for "op0 RELOP op0".
688 2007-09-27 Jakub Jelinek <jakub@redhat.com>
690 * builtins.c (expand_builtin, expand_builtin_object_size,
691 expand_builtin_memory_chk, maybe_emit_chk_warning,
692 maybe_emit_sprintf_chk_warning): Use new %K format string specifier
694 * expr.c (expand_expr_real_1): Likewise.
695 * langhooks-def.h (struct diagnostic_info): Add forward decl.
696 (lhd_print_error_function): Add third argument.
697 * langhooks.h (struct diagnostic_info): Add forward decl.
698 (struct lang_hooks): Add third argument to print_error_function.
699 * diagnostic.h (diagnostic_info): Add abstract_origin field.
700 (diagnostic_last_function_changed, diagnostic_set_last_function): Add
702 (diagnostic_report_current_function): Likewise.
703 * toplev.c (announce_function): Pass NULL as second argument to
704 diagnostic_set_last_function.
705 * diagnostic.c (diagnostic_report_current_function): Add second
706 argument, pass it as third argument to lang_hooks.print_error_function.
707 (default_diagnostic_starter): Pass DIAGNOSTIC as second argument
708 to diagnostic_report_current_function.
709 (diagnostic_report_diagnostic): Initialize diagnostic->abstract_origin
710 and message.abstract_origin.
711 (verbatim): Initialize abstract_origin.
712 * pretty-print.h (text_info): Add abstract_origin field.
713 * pretty-print.c (pp_base_format): Handle %K.
714 * langhooks.c (lhd_print_error_function): Add third argument. If
715 diagnostic->abstract_origin, print virtual backtrace.
716 * c-format.c (gcc_diag_char_table, gcc_tdiag_char_table,
717 gcc_cdiag_char_table, gcc_cxxdiag_char_table): Support %K.
718 (init_dynamic_diag_info): Likewise.
720 2007-09-26 David Daney <ddaney@avtrex.com>
723 * config/mips/mips.md (sync_compare_and_swap<mode>, sync_old_add<mode>,
724 sync_new_add<mode>, sync_old_<optab><mode>, sync_new_<optab><mode>,
725 sync_old_nand<mode>, sync_new_nand<mode>,
726 sync_lock_test_and_set<mode>): Fix '&' constraint modifiers.
727 Update length attributes.
728 (sync_add<mode>, sync_sub<mode>, sync_old_sub<mode>,
729 sync_new_sub<mode>, sync_<optab><mode>, sync_nand<mode>): Update
731 * config/mips/mips.h (MIPS_COMPARE_AND_SWAP, MIPS_SYNC_OP,
732 MIPS_SYNC_OLD_OP, MIPS_SYNC_NEW_OP, MIPS_SYNC_NAND,
733 MIPS_SYNC_OLD_NAND, MIPS_SYNC_NEW_NAND, MIPS_SYNC_EXCHANGE): Add
736 2007-09-26 Richard Guenther <rguenther@suse.de>
738 PR tree-optimization/33563
739 * tree-ssa-dse.c (get_use_of_stmt_lhs): Rename to ...
740 (get_kill_of_stmt_lhs): ... this. Re-structure. Handle
742 (dse_optimize_stmt): Call get_kill_of_stmt_lhs instead of
745 2007-09-26 Joseph Myers <joseph@codesourcery.com>
748 * c-common.c (complete_array_type): Diagnose too-large arrays and
749 set type to error_mark_node.
751 2007-09-26 Richard Guenther <rguenther@suse.de>
753 PR tree-optimization/30375
754 PR tree-optimization/33560
755 * tree-ssa-dse.c (get_use_of_stmt_lhs): Give up on uses
759 2006-05-22 Aldy Hernandez <aldyh@redhat.com>
761 * tree-ssa-dse.c (aggregate_vardecl_d): New.
762 (dse_global_data): Add aggregate_vardecl field.
763 (dse_possible_dead_store_p): New.
764 Add prev_defvar variable.
765 Allow immediate uses and previous immediate uses to differ
766 if they are setting different parts of the whole.
767 (get_aggregate_vardecl): New.
768 (dse_record_partial_aggregate_store): New.
769 (dse_whole_aggregate_clobbered_p): New.
770 (dse_partial_kill_p): New.
771 Call dse_maybe_record_aggregate_store().
772 When checking whether a STMT and its USE_STMT refer to the
773 same memory address, check also for partial kills that clobber
775 Move some variable definitions to the block where they are used.
776 (aggregate_vardecl_hash): New.
777 (aggregate_vardecl_eq): New.
778 (aggregate_vardecl_free): New.
779 (aggregate_whole_store_p): New.
780 (tree_ssa_dse): Initialize and free aggregate_vardecl.
781 Mark which aggregate stores we care about.
783 2007-09-25 DJ Delorie <dj@redhat.com>
786 * config/m32c/m32c.c (m32c_immd_dbl_mov): Use INTVAL instead of
789 2007-09-25 Michael Meissner <michael.meissner@amd.com>
792 * config/i386/i386.c (ix86_expand_sse5_unpack): Change to call
793 gen_sse5_pperm_sign_v4si_v2di and gen_sse5_pperm_zero_v4si_v2di
794 for vector int32 -> int64 conversions. Don't write beyond the end
795 of the allocated vector for int32 -> int64 conversions.
797 2007-09-25 Revital Eres <eres@il.ibm.com>
799 * config/rs6000/paired.h (paired_sel): New.
800 * config/rs6000/rs6000.c (bdesc_3arg): Add selv2sf4.
801 (rs6000_expand_ternop_builtin): Pass zero const_double operand
802 when expanding selv2sf.
803 * config/rs6000/rs6000.h (rs6000_builtins): Add
804 PAIRED_BUILTIN_SELV2SF4.
806 2007-09-25 Joseph Myers <joseph@codesourcery.com>
809 * c-typeck.c (default_conversion): Call require_complete_type
810 before perform_integral_promotions.
811 (build_unary_op): Call require_complete_type except for ADDR_EXPR.
812 (build_c_cast): Call require_complete_type except for casts to
814 (convert_for_assignment): Call require_complete_type.
816 2007-09-25 Revital Eres <eres@il.ibm.com>
818 * config/spu/spu.md: Fix doloop pattern.
820 2007-09-25 Bernd Schmidt <bernd.schmidt@analog.com>
822 * config/bfin/bfin.c (expand_prologue_reg_save,
823 expand_epilogue_reg_restore): Code to save and restore I/M/B/L regs and
825 (expand_interrupt_handler_prologue, expand_interrupt_handler_epilogue):
826 ... from here. New argument ALL; callers changed.
827 (n_regs_saved_by_prologue): Count ASTAT for plain saveall functions.
828 (bfin_expand_prologue, bfin_expand_epilogue): Deal with functions that
829 have the "saveall" attribute.
831 2007-09-25 Hans-Peter Nilsson <hp@bitrange.com>
833 * config/mmix/mmix.h (FUNCTION_INCOMING_ARG_REGNO_P): Don't define
835 (INCOMING_REGNO, OUTGOING_REGNO): Define.
836 * config/mmix/mmix.c (mmix_opposite_regno): New function.
837 * config/mmix/mmix-protos.h (mmix_opposite_regno): Prototype.
839 * config/mmix/mmix.md ("*cmpcc_folded"): In condition, gate
840 with REG_P tests before REGNO access.
842 2007-09-24 DJ Delorie <dj@redhat.com>
845 * config/m32c/cond.md (stzx_reversed_<mode>): Add an output
847 (movqicc_<code>_<mode>): Likewise.
848 (movhicc_<code>_<mode>): Likewise.
850 2007-09-24 Rask Ingemann Lambertsen <rask@sygehus.dk>
853 * config/m32c/m32c.c (m32c_eh_return_data_regno): Leave an address
856 2007-09-24 Danny Smith <dannysmith@user.sourceforge.net>
859 * config/i386/i386.c (ix86_comp_type_attributes): Check
862 2007-09-24 Roman Zippel <zippel@linux-m68k.org>
864 * config/m68k/m68k.h (ASM_OUTPUT_ALIGN_WITH_NOP): New, use
865 "move.l %a4,%a4" to produce nops.
866 * config/m68k/m68k.c (override_options): Reset align options,
867 if neccessary align macro isn't avaible.
869 2007-09-24 Roman Zippel <zippel@linux-m68k.org>
871 config/m68k/linux.h (PREFERRED_STACK_BOUNDARY): New.
873 2007-09-24 Roman Zippel <zippel@linux-m68k.org>
875 * config/m68k/m68k.c (strict_low_part_peephole_ok): Don't leave
877 * config/m68k/m68k.md (movsi_m68k): Allow certain constant when
879 (peephole pattern): Convert most of them to RTL peephole pattern.
881 2007-09-24 Roman Zippel <zippel@linux-m68k.org>
883 * config/m68k/m68k.c (notice_update_cc): Recognize fp compare
884 (moved from fp compare patterns).
885 * config/m68k/m68k.md (cmp<mode>, cmp<mode>_68881, cmp<mode>_cf):
886 Cleanup predicates to relieve reload.
887 (conditional_trap): Reject conditional trap with fp condition.
888 * gcc/config/m68k/predicates.md (fp_src_operand): New, reject
889 certain constants early.
891 2007-09-24 Roman Zippel <zippel@linux-m68k.org>
893 * gcc/final.c (final_scan_insn): Remove accidentally duplicated code.
895 2007-09-24 Andrew Pinski <andrew_pinski@playstation.sony.com>
897 * config.gcc (powerpc*-*-*): --with-cpu=cell is a 64bit CPU.
898 Allow --with-tune=cell and --with-cpu=cell.
900 2007-09-24 David Edelsohn <edelsohn@gnu.org>
902 * config/rs6000/rs6000-protos.h (rs6000_emit_swrsqrtsf): Declare.
903 * config/rs6000/rs6000.opt (swdiv): Change option to ...
905 * config/rs6000/rs6000.c (rs6000_builtin_reciprocal): New
907 (TARGET_BUILTIN_RECIPROCAL): Use it.
908 (rs6000_builtin_expand): Expand recip, recipf, and rsqrtf.
909 (rs6000_init_builtins): Initialize recip, recipf, and rsqrtf.
910 (rs6000_emit_swrsqrtsf): New.
911 * config/rs6000/rs6000.h (rs6000_builtins): Add recip, recipf, and
913 * config/rs6000/rs6000.md (UNSPEC_RSQRT): Define.
914 (divsf3): Remove swdiv support.
917 (rsqrt_internal1): New.
918 (divdf3): Remove swdiv support.
921 2007-09-24 Jakub Jelinek <jakub@redhat.com>
924 * langhooks.h (struct lang_hooks_for_types): Add type_hash_eq
926 * langhooks-def.h (LANG_HOOKS_TYPE_HASH_EQ): Define.
927 (LANG_HOOKS_FOR_TYPES_INITIALIZER): Add LANG_HOOKS_TYPE_HASH_EQ.
928 * tree.c (type_hash_eq): For FUNCTION_TYPE use
929 lang_hooks.type.type_hash_eq in addition to generic tests.
931 2007-09-24 Pranav Bhandarkar <pranav.bhandarkar@celunite.com>
932 Ramana Radhakrishnan <ramana@hercules.pun.celunite.com>
934 * tree-inline.h (eni_weights): Add field target_builtin_cost to
935 reflect the cost per call to a target specific builtin.
936 * tree-inline.c (estimate_num_insns_1): If it is a CALL_EXPR for a
937 call to a target specific builtin, then use target_builtin_call_cost.
938 (init_inline_once): Initialize target_builtin_call_cost field.
940 2007-09-24 Kai Tietz <kai.tietz@onevision.com>
943 * config/i386/i386.c (return_in_memory_ms_64): Handle return types for
946 2007-09-23 H.J. Lu <hongjiu.lu@intel.com>
948 * configure.ac (ld_vers): Support GNU linker version xx.xx.*
949 * configure: Regenerated.
951 2007-09-23 Ollie Wild <aaw@google.com>
953 * fold-const.c (fold_binary): Fold BIT_AND_EXPR's with a pointer
955 (get_pointer_modulus_and_residue): New function.
957 2007-09-23 Richard Sandiford <rsandifo@nildram.co.uk>
959 * config/mips/mips.c (build_mips16_call_stub): On 64-bit targets,
960 combine an SC return value into a single register.
962 2007-09-23 Richard Sandiford <rsandifo@nildram.co.uk>
964 * opth-gen.awk (target_flags_explicit): Declare.
965 * toplev.h (target_flags_explicit): Delete declaration.
966 * toplev.c (target_flags): Likewise.
967 * config/mips/mips.h (mips_llsc_setting, mips_llsc): Delete.
968 (GENERATE_SYNC, GENERATE_LL_SC): Redefine using target_explicit_flags.
969 Never return true for TARGET_MIPS16.
970 * config/mips/mips.c (mips_llsc): Delete.
971 (mips_handle_option): Remove -mllsc handling.
972 (mips_strip_unspec_address): Tweak comment.
973 * config/mips/mips.opt (mllsc): Use a target mask.
975 2007-09-23 Richard Sandiford <rsandifo@nildram.co.uk>
977 * doc/tm.texi (POINTERS_EXTEND_UNSIGNED): Document extensions to
978 word_mode as well as Pmode.
980 2007-09-23 Richard Sandiford <rsandifo@nildram.co.uk>
982 * function.c (assign_parm_setup_block): Explicitly convert BLKmode
983 parameters from word_mode to the subword type if such a truncation
986 2007-09-23 Jakub Jelinek <jakub@redhat.com>
988 * configure.ac (MAKEINFO): Handle makeinfo version 4.10 and above.
989 * configure: Regenerated.
991 2007-09-23 Jan Hubicka <jh@suse.cz>
993 * params.def (INLINE_CALL_COST): Set to 12.
994 * invoke.texi (inline-call-cost): Update default value.
996 2007-09-23 Eric Botcazou <ebotcazou@adacore.com>
998 * config/alpha/alpha.md (movti): Use operand_subword for the split.
1000 2007-09-23 Ayal Zaks <zaks@il.ibm.com>
1001 Revital Eres <eres@il.ibm.com>
1003 * modulo-sched.c (doloop_register_get): Rewrite the loop which
1004 checks whether the count_reg is found outside the control part.
1006 2007-09-23 Jakub Jelinek <jakub@redhat.com>
1008 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Use get_callee_fndecl
1009 instead of checking CALL_EXPR_FN directly to test for builtins.
1010 If error or warning attributes are present, print
1011 error resp. warning.
1012 * c-common.c (handle_error_attribute): New function.
1013 (c_common_attribute_table): Add error and warning
1015 * doc/extend.texi: Document error and warning attributes.
1017 * tree.h (block_nonartificial_location): New prototype.
1018 * tree.c (block_nonartificial_location): New function.
1019 * dwarf2out.c (gen_subprogram_die): Add DW_AT_artificial
1020 if artificial attribute is present on abstract inline decl.
1021 * c-common.c (handle_artificial_attribute): New function.
1022 (c_common_attribute_table): Add artificial attribute.
1023 * final.c (override_filename, override_linenum): New variables.
1024 (final_scan_insn): For DBX_DEBUG or SDB_DEBUG, set override_filename
1025 and override_linenum if inside of a block inlined from
1026 __attribute__((__artificial__)) function.
1027 (notice_source_line): Honor override_filename and override_linenum.
1028 * doc/extend.texi: Document __attribute__((__artificial__)).
1029 * config/i386/emmintrin.h: Add __artificial__ attribute to
1030 all __always_inline__ functions.
1031 * config/i386/mmintrin.h: Likewise.
1032 * config/i386/tmmintrin.h: Likewise.
1033 * config/i386/mm3dnow.h: Likewise.
1034 * config/i386/pmmintrin.h: Likewise.
1035 * config/i386/ammintrin.h: Likewise.
1036 * config/i386/xmmintrin.h: Likewise.
1037 * config/i386/smmintrin.h: Likewise.
1038 * config/i386/bmmintrin.h: Likewise.
1039 * config/i386/mmintrin-common.h: Likewise.
1042 * expr.c (expand_constructor): New function.
1043 (expand_expr_real_1) <case CONSTRUCTOR>: Call it.
1044 (expand_expr_real_1) <case ARRAY_REF>: Call it if VALUE is
1047 2007-09-23 Richard Sandiford <rsandifo@nildram.co.uk>
1049 * config/mips/mips.c (dump_constants_1): Generalize to include
1050 fractional and accumulator modes.
1052 2007-09-23 Richard Sandiford <rsandifo@nildram.co.uk>
1054 * config/mips/mips.h (ISA_HAS_DSP, ISA_HAS_DSPR2): New macros.
1055 * config/mips/mips.c (mips_set_mips16_mode): Don't clear the DSP
1057 (override_options): Check TARGET_HARD_FLOAT_ABI instead of
1058 TARGET_HARD_FLOAT when testing whether -mpaired-single is
1060 (mips_conditional_register_usage): Check ISA_HAS_DSP instead of
1062 * config/mips/constraints.md (ka): Check ISA_HAS_DSPR2 instead of
1064 * config/mips/mips.md (ANYF): Require TARGET_HARD_FLOAT for V2SF.
1065 (mulv2sf3, movv2sf, movv2sf_hardfloat_64bit): Require
1067 (<u>mulsidi3_32bit_internal, <u>msubsidi4, <u>maddsidi4): Check
1068 ISA_HAS_DSPR2 instead of TARGET_HAS_DSPR2.
1069 * config/mips/mips-dsp.md: Use ISA_HAS_DSP instead of TARGET_HAS_DSP
1071 * config/mips/mips-dspr2.md: Likewise ISA_HAS_DSPR2 and
1073 * config/mips/mips-fixed.md: Use ISA_HAS_DSP and ISA_HAS_DSPR2
1074 instead of TARGET_HAS_DSP and TARGET_HAS_DSPR2.
1075 * config/mips/mips-ps-3d.md: Add TARGET_HARD_FLOAT to V2SF patterns.
1077 2007-09-22 Jason Merrill <jason@redhat.com>
1080 * attribs.c (lookup_attribute_spec): Split out...
1081 (decl_attributes): From here.
1082 * tree.h: Declare it.
1084 2007-09-22 Richard Sandiford <rsandifo@nildram.co.uk>
1086 * doc/sourcebuild.texi: Document dg-add-options mips16_attribute.
1088 2007-09-22 Eric Botcazou <ebotcazou@adacore.com>
1090 * tree-inline.c (remap_type_1): Correctly chain variants.
1092 2007-09-22 Richard Guenther <rguenther@suse.de>
1094 PR tree-optimization/33146
1095 * fold-const.c (fold_binary): Use the original tree
1097 * tree.h (STRIP_SIGN_NOPS): Converting from or to pointer
1098 also changes "sign".
1100 2007-09-22 Eric Botcazou <ebotcazou@adacore.com>
1103 * except.c (finish_eh_generation): Call commit_edge_insertions if
1104 there are insns queued on the entry edge.
1105 * config/alpha/alpha.c (alpha_gp_save_rtx): Insert the insns on
1108 2007-09-22 Richard Sandiford <rsandifo@nildram.co.uk>
1110 * doc/sourcebuild.texi (dg-add-c99-runtime-options): Document.
1112 2007-09-19 Michael Meissner <michael.meissner@amd.com>
1114 * gcc/config/i386/i386.c: Delete trailing whitespace.
1115 * gcc/config/i386/i386.h: Ditto.
1116 * gcc/config/i386/bmmintrin.h: Ditto.
1117 * gcc/config/i386/sync.md: Ditto.
1118 * gcc/config/i386/ppro.md: Ditto.
1119 * gcc/config/i386/mmx.md: Ditto.
1120 * gcc/config/i386/constraints.md: Ditto.
1121 * gcc/config/i386/sse.md: Ditto.
1122 * gcc/config/i386/athlon.md: Ditto.
1123 * gcc/config/i386/i386.md: Ditto.
1125 2007-09-21 Richard Guenther <rguenther@suse.de>
1127 PR tree-optimization/33508
1128 * tree-ssa-alias.c (mark_aliases_call_clobbered): Avoid
1129 quadratic loop by keeping a bitmap of variables we have
1130 to clobber all subvariables for.
1131 (set_initial_properties): Likewise.
1133 2007-09-21 Richard Sandiford <rsandifo@nildram.co.uk>
1135 * config/mips/t-sde (TARGET_LIBGCC2_CFLAGS): Delete.
1137 2007-09-20 Nigel Stephens <nigel@mips.com>
1138 Chao-ying Fu <fu@mips.com>
1140 * c-decl.c (finish_declspecs): When _Sat is used without
1141 _Fract or _Accum, set the default type to cts_fract.
1142 This avoids a warning of "type defaults to int".
1144 2007-09-20 Joseph Myers <joseph@codesourcery.com>
1146 * c-decl.c (check_bitfield_type_and_width): Don't allow _Bool
1147 bit-fields wider than one bit.
1149 2007-09-20 Jakub Jelinek <jakub@redhat.com>
1152 * dwarf2out.c (modified_type_die): Handle TYPE_DECL with NULL
1154 * dbxout.c (dbxout_type): Likewise.
1158 * gimplify.c (gimplify_vla_decl): New function.
1159 (gimplify_decl_expr): Move VLA decl handling to gimplify_vla_decl.
1161 (gimplify_target_expr): Handle variable length TARGET_EXPRs.
1163 2007-09-20 Richard Sandiford <rsandifo@nildram.co.uk>
1165 * doc/invoke.texi (-minterlink-mips16): Document.
1166 * config/mips/mips.opt (minterlink-mips16): New option.
1167 * config/mips/mips.c (mips_function_ok_for_sibcall): Handle
1170 2007-09-20 Joseph Myers <joseph@codesourcery.com>
1172 * doc/extend.texi (Attribute Syntax): Remove old speculative
1175 2007-09-20 Mark Shinwell <shinwell@codesourcery.com>
1177 * combine.c: Include cgraph.h.
1178 (setup_incoming_promotions): Rework to allow more aggressive
1179 elimination of sign extensions when all call sites of the
1180 current function are known to lie within the current unit.
1182 2007-09-20 Richard Sandiford <rsandifo@nildram.co.uk>
1184 * dse.c (find_shift_sequence): No-op rework of control flow.
1186 2007-09-19 Richard Sandiford <rsandifo@nildram.co.uk>
1188 * config/mips/mips.c (build_mips16_call_stub): Tidy. Fix second
1189 GPR for DCmode on 64-bit targets. Remove redundant fallback.
1191 2007-09-19 Andrew Pinski <andrew_pinski@playstation.sony.com>
1193 * cfgexpand.c (dump_stack_var_partition): Use the correct
1194 index for the offset.
1196 2007-09-19 Joseph Myers <joseph@codesourcery.com>
1198 * config/mips/sde.h: Switch to GPLv3.
1200 2007-09-19 Eric Botcazou <ebotcazou@adacore.com>
1202 * tree-sra.c (decide_block_copy): Decide if there are groups.
1204 2007-09-19 Roman Zippel <zippel@linux-m68k.org>
1206 * config/m68k/m68k.c (output_move_himode): Remove jump table
1208 config/m68k/m68k.md (lea): Likewise.
1209 * config/m68k/m68k.c (print_operand_address): Use simple pc
1210 relative addressing.
1212 2007-09-19 Bernd Schmidt <bernd.schmidt@analog.com>
1214 * doc/tm.texi (IS_ASM_LOGICAL_LINE_SEPARATOR): Document new argument.
1215 * final.c (IS_ASM_LOGICAL_LINE_SEPARATOR): Provide two-argument
1217 (asm_insn_count): Pass template as second argument to it.
1218 * config/avr/avr.h (IS_ASM_LOGICAL_LINE_SEPARATOR): Likewise.
1219 * config/pa/pa.h (IS_ASM_LOGICAL_LINE_SEPARATOR): Likewise.
1220 * config/stormy16/stormy16.h (IS_ASM_LOGICAL_LINE_SEPARATOR): Likewise.
1221 * config/cris/cris.h (IS_ASM_LOGICAL_LINE_SEPARATOR): Likewise.
1222 * config/sh/sh.c (IS_ASM_LOGICAL_LINE_SEPARATOR): Likewise.
1223 (sh_insn_length_adjustment): Pass template as second argument to it.
1224 * config/bfin/bfin.h (IS_ASM_LOGICAL_LINE_SEPARATOR): New macro.
1226 * config/bfin/bfin.md (define_asm_attributes): New.
1228 2007-09-19 Jie Zhang <jie.zhang@analog.com>
1230 * config.gcc (bfin*-linux-uclibc*): Add ./linux-sysroot-suffix.h
1232 * config/bfin/print-sysroot-suffix.sh: New.
1233 * config/bfin/t-bfin-elf (EXTRA_PARTS): Remove.
1234 (MULTILIB_OPTIONS, MULTILIB_DIRNAMES, MULTILIB_MATCHES,
1235 MULTILIB_EXCEPTIONS): Redefine with new multilibs.
1236 * config/bfin/t-bfin-uclinux (EXTRA_PARTS): Remove.
1237 (MULTILIB_OPTIONS, MULTILIB_DIRNAMES, MULTILIB_MATCHES,
1238 MULTILIB_EXCEPTIONS): Redefine with new multilibs.
1239 * config/bfin/t-bfin-linux (EXTRA_PARTS): Remove.
1240 (MULTILIB_OPTIONS, MULTILIB_DIRNAMES, MULTILIB_MATCHES,
1241 MULTILIB_EXCEPTIONS): Redefine with new multilibs.
1242 (linux-sysroot-suffix.h): New target.
1243 * config/bfin/bfin.opt (mcsync-anomaly): Use Var instead of Mask.
1244 (mspecld-anomaly): Likewise.
1245 * config/bfin/bfin-protos.h (enum bfin_cpu_type): Renamed from
1246 (enum bfin_cpu): ... this. Add BFIN_CPU_BF522, BFIN_CPU_BF525,
1247 BFIN_CPU_BF527, BFIN_CPU_BF538, BFIN_CPU_BF539, BFIN_CPU_BF542,
1248 BFIN_CPU_BF544, BFIN_CPU_BF548, and BFIN_CPU_BF549.
1249 (bfin_si_revision): Declare.
1250 (bfin_workarounds): Declare.
1251 (WA_SPECULATIVE_LOADS): Define.
1252 (ENABLE_WA_SPECULATIVE_LOADS): Define.
1253 (WA_SPECULATIVE_SYNCS): Define.
1254 (ENABLE_WA_SPECULATIVE_SYNCS): Define.
1255 * config/bfin/elf.h (STARTFILE_SPEC): Rename crt532.o to basiccrt.o.
1256 (LIB_SPEC): Add %s to the linker scripts.
1257 Use proper linker script for bf522, bf525, bf527,
1258 bf538, bf539, bf542, bf544, bf548, and bf549.
1259 * config/bfin/bfin.c (bfin_si_revision): Define.
1260 (bfin_workarounds): Define.
1261 (struct bfin_cpu): New.
1263 (bfin_handle_option): Handle silicon revision part of -mcpu option.
1264 (override_options): Set bfin_workarounds.
1265 (length_for_loop): Replace TARGET_CSYNC_ANOMALY with
1266 ENABLE_WA_SPECULATIVE_SYNCS, TARGET_SPECLD_ANOMALY with
1267 ENABLE_WA_SPECULATIVE_LOADS.
1268 (bfin_reorg): Likewise.
1269 * config/bfin/bfin.h (TARGET_CPU_CPP_BUILTINS): Define
1270 macros for bf522, bf525, bf527, bf538, bf539,
1271 bf542, bf544, bf548, and bf549.
1272 Define __SILICON_REVISION__ and __WORKAROUND_* macros if needed.
1273 Don't define __ID_SHARED_LIB__ when -msep-data.
1274 (TARGET_DEFAULT): Define as 0.
1275 (DRIVER_SELF_SPECS): Add -mcpu=bf532 if no -mcpu option.
1276 * doc/invoke.texi (Blackfin Options): Document silicon
1277 revision part of -mcpu option and it now accepts bf522, bf525,
1278 bf527, bf538, bf539, bf542, bf544, bf548, and bf549.
1280 2007-09-18 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1282 * alias.c (memory_modified_1): Deconstify.
1283 (memory_modified_in_insn_p): Don't use const_note_stores.
1284 * rtl.h (const_note_stores): Delete.
1285 * rtlanal.c (const_note_stores): Likewise.
1287 2007-09-18 Richard Sandiford <rsandifo@nildram.co.uk>
1289 * dse.c (find_shift_sequence): Temporarily revert to forbidding
1292 2007-09-18 Rask Ingemann Lambertsen <rask@sygehus.dk>
1296 * config/fr30/fr30.md (addsi3): Check REG_P() before calling REGNO().
1297 (addsi_small_int): Likewise.
1298 (addsi_big_int): Use rtx_equal_p() instead of REGNO() comparison.
1299 (one_cmplsi2): Likewise.
1301 (enter_func): Expand insn using hard_frame_pointer_rtx and
1304 * config/fr30/fr30.c (fr30_expand_prologue): Check for
1305 hard_frame_pointer_rtx instead of using REGNO() check.
1306 Properly sign extend GEN_INT() argument.
1308 2007-09-18 Roman Zippel <zippel@linux-m68k.org>
1310 * config/m68k/m68k.c (override_options): Remove USE_GAS,
1312 (output_dbcc_and_branch, output_scc_di): Replace all jbcc
1313 alternatives with just jcc.
1314 * config/m68k/m68k.md (addsi_lshrsi_31, beq0_di, bne0_di,
1315 bge0_di, blt0_di, bgtu, bltu, bgeu, bleu, bgtu_rev,
1316 bltu_rev, bgeu_rev, bleu_rev, jump, dbne_hi, dbne_si,
1317 dbge_hi, dbge_si): Likewise.
1319 2007-09-18 Roman Zippel <zippel@linux-m68k.org>
1321 * config/m68k/m68k.md (beq, bne, bgt, blt, bge, ble, bordered,
1322 bunordered, buneq, bunge, bungt, bunle, bunlt, bltgt, beq_rev,
1323 bne_rev, bgt_rev, blt_rev, bge_rev, ble_rev, bordered_rev,
1324 bunordered_rev, buneq_rev, bunge_rev, bungt_rev, bunle_rev,
1325 bunlt_rev, bltgt_rev): Replace all fbcc with fjcc.
1327 2007-09-18 Maxim Kuvyrkov <maxim@codesourcery.com>
1329 * config/m68k/m68k-devices.def (51qe): New device.
1330 * config/m68k/m68k.c (FL_FOR_isa_c): Remove division unit. Add it
1331 to all uses of FL_FOR_isa_c for compatibility.
1332 (all_microarchs): Add cfv1 microarchitecture.
1333 (m68k_handle_option): Handle m51qe option.
1334 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Handle 51QE.
1335 (TUNE_CFV1): New macro.
1336 (enum uarch_type): Add ucfv1.
1337 * doc/invoke.texi: Document 51qe device and cfv1 microarchitecture.
1339 2007-09-18 Richard Guenther <rguenther@suse.de>
1341 PR tree-optimization/31863
1342 * tree-ssa-structalias.c (create_variable_info_for): Always
1343 free the fieldstack.
1345 2007-09-18 Dorit Nuzman <dorit@il.ibm.com>
1347 * opts.c (decode_options): Enable vectorization under -O3.
1349 2007-09-18 Richard Guenther <rguenther@suse.de>
1351 PR tree-optimization/33340
1352 * tree-ssa-sccvn.c (set_ssa_val_to): Do not set values to
1353 SSA_NAMEs that occur in abnormal PHI nodes.
1355 2007-09-17 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1357 * tree-cfg.c (is_ctrl_altering_stmt, tree_block_ends_with_call_p):
1358 Don't use const_get_call_expr_in.
1359 * tree-gimple.c (const_get_call_expr_in): Delete.
1360 * tree-gimple.h (const_get_call_expr_in): Likewise.
1362 2007-09-18 Richard Sandiford <rsandifo@nildram.co.uk>
1364 * df-scan.c (df_notes_rescan): Do nothing if the instruction does
1365 not yet have a basic block.
1366 * dse.c (find_shift_sequence): Don't set DF_NO_INSN_RESCAN.
1368 2007-09-18 Richard Sandiford <rsandifo@nildram.co.uk>
1370 * config/mips/mips.c (mips_file_start): Add ".previous" directives
1371 to both ".section"s.
1373 2007-09-17 Richard Sandiford <rsandifo@nildram.co.uk>
1375 * config/mips/mips.c (mips_output_mi_thunk): Use
1376 mips_function_ok_for_sibcall and const_call_insn_operand
1377 to determine if a direct sibcall is allowed. Use
1378 mips_classify_symbol to determine a global pointer is needed.
1380 2007-09-17 Richard Sandiford <rsandifo@nildram.co.uk>
1382 * config/mips/mips.md (*clear_upper32): Use "W" as the memory operand.
1384 2007-09-17 Chao-ying Fu <fu@mips.com>
1385 Nigel Stephens <nigel@mips.com>
1387 * config/fixed-bit.h: New file.
1388 * config/fixed-bit.c: New file.
1389 * doc/libgcc.texi (Fixed-point fractional library routines): New node.
1391 2007-09-18 Jakub Jelinek <jakub@redhat.com>
1393 * c-format.h (format_kind_info): Add alloc_char field.
1394 * c-format.c (scanf_flag_specs): Add 'm'.
1395 (scanf_flag_pairs): Add 'a', 'm' pair.
1396 (scan_char_table): Allow 'm' modifier for c, s, [, C and S.
1397 (format_types_orig): Add alloc_char fields.
1398 (check_format_info_main): Rename aflag to alloc_flag.
1399 Handle fki->alloc_char. modifier after width and before length
1400 modifiers. Move FMT_FLAG_SCANF_A_KLUDGE handling before
1401 length modifiers as well.
1402 * config/sol2-c.c (solaris_format_types): Add alloc_char field.
1405 * builtins.c (expand_builtin_memory_chk): Handle COMPOUND_EXPRs
1406 returned by build_call_expr.
1408 2007-09-17 Eric Botcazou <ebotcazou@adacore.com>
1410 * tree-sra.c (maybe_lookup_element_for_expr) <COMPONENT_REF>: Return
1411 NULL for variable-sized records too.
1412 (sra_walk_expr) <COMPONENT_REF>: Stop at variable-sized records too.
1414 2007-09-17 Tom Tromey <tromey@redhat.com>
1416 * c-decl.c (pushdecl): Don't set DECL_LANG_SPECIFIC.
1417 (c_builtin_function): Likewise.
1418 (grokdeclarator): Likewise.
1420 2007-09-17 Zdenek Dvorak <ook@ucw.cz>
1422 PR rtl-optimization/26449
1423 * loop-invariant.c (move_invariant_reg): Do not use force_operand.
1424 (seq_insns_valid_p): Removed.
1426 2007-09-17 Eric Botcazou <ebotcazou@adacore.com>
1428 * tree-nomudflap.c (gate_mudflap): New static function.
1429 (pass_mudflap_1): Use it as gate function.
1430 (pass_mudflap_2): Likewise.
1432 2007-09-17 Jan Hubicka <jh@suse.cz>
1436 * loop-invariant.c (move_invariant_reg): Unshare sequence.
1438 2007-09-17 Victor Kaplansky <victork@il.ibm.com>
1440 PR tree-optimization/33319
1441 * tree-vect-analyze.c (vect_same_range_drs): New.
1442 (vect_vfa_range_equal): New.
1443 (vect_is_duplicate_ddr): Removed.
1444 (vect_mark_for_runtime_alias_test): Do not perform marking when
1445 optimizing for size or max_param for alias checking is zero.
1446 Move the function before vect_analyze_data_ref_dependence.
1447 (vect_analyze_data_ref_dependence): Add call to
1448 vect_mark_for_runtime_alias_test in two cases when dependence
1450 (vect_analyze_data_ref_dependences): Do not call to
1451 vect_mark_for_runtime_alias_test.
1452 (vect_prune_runtime_alias_test_list): New.
1453 (vect_analyze_loop): Add call to vect_prune_runtime_alias_test_list.
1454 * tree-vect-transform.c (vect_estimate_min_profitable_iters):
1455 Update vec_outside_cost.
1456 (vect_vfa_segment_size): More compact code, use TYPE_SIZE_UNIT.
1457 (vect_create_cond_for_alias_checks): Build the base address of data
1458 reference from DR_GROUP_FIRST_DR.
1459 (vect_loop_versioning): New.
1460 (vect_transform_loop): Add a call to vect_loop_versioning.
1461 Remove factored out code.
1463 2007-09-16 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1466 * expr.c (store_expr): Call adjust_address to change mode of dest_mem
1469 2007-09-16 Richard Sandiford <rsandifo@nildram.co.uk>
1471 * dse.c (find_shift_sequence): Allow word as well as subword shifts.
1472 Do the tentative shift expansion with the DF_NO_INSN_RESCAN flag set.
1473 Fix the call to insn_rtx_cost. Skip access sizes that require a
1474 real truncation of the store register. Use convert_move instead
1475 of gen_lowpart when narrowing the result.
1476 (replace_read): Use convert_move instead of gen_lowpart when
1477 narrowing the store rhs.
1479 2007-09-16 Richard Sandiford <rsandifo@nildram.co.uk>
1481 * config/mips/mips.md (SHORT): Fix long line.
1482 (SUBDI): New mode iterator. Extend the shift-and-truncate insns
1483 to QImode and HImode.
1485 2007-09-16 Richard Sandiford <rsandifo@nildram.co.uk>
1487 * config/mips/mips.h (POINTERS_EXTEND_UNSIGNED): Define.
1489 2007-09-15 Zdenek Dvorak <ook@ucw.cz>
1491 * tree-parloops.c: New file.
1492 * tree-ssa-operands.h (free_stmt_operands): Declare.
1493 * tree-ssa-loop-manip.c (split_loop_exit_edge): Return the new basic
1495 * tree-pass.h (pass_parallelize_loops): Declare.
1496 * omp-low.c (expand_omp_parallel, expand_omp_for): Update SSA form for
1498 (build_omp_regions_1): Allow analysing just a single OMP region and
1500 ( build_omp_regions_root, omp_expand_local): New functions.
1501 (build_omp_regions): Add argument to build_omp_regions_1 call.
1502 * builtins.def (DEF_GOMP_BUILTIN): Initialize OMP builtins when
1503 autoparallelization is run.
1504 * timevar.def (TV_TREE_PARALLELIZE_LOOPS): New.
1505 * tree-ssa-loop.c (gate_tree_parallelize_loops, tree_parallelize_loops,
1506 pass_parallelize_loops): New.
1507 * common.opt (ftree-parallelize-loops): New.
1508 * tree-flow.h (omp_expand_local, tree_duplicate_sese_tail,
1509 parallelize_loops): Declare.
1510 (add_phi_args_after_copy, split_loop_exit_edge): Declaration changed.
1511 * Makefile.in (tree-parloops.o): Added.
1512 * tree-cfg.c (add_phi_args_after_copy_edge, tree_duplicate_sese_tail):
1514 (add_phi_args_after_copy_bb): Use add_phi_args_after_copy_edge.
1515 (add_phi_args_after_copy): Call add_phi_args_after_copy_edge for
1516 one extra edge as well.
1517 (tree_duplicate_sese_region): Add argument to add_phi_args_after_copy.
1518 Use VEC_free to free doms vector.
1519 (move_block_to_fn): Update loop info. Remove phi nodes for virtual
1520 operands. Recompute operand caches in the new function.
1521 (move_sese_region_to_fn): Update loop info.
1522 * passes.c (init_optimization_passes): Add pass_parallelize_loops.
1523 * tree-ssa-operands.c (free_stmt_operands): New function.
1525 * doc/passes.texi: Document autoparallelization.
1526 * doc/invoke.texi (-ftree-parallelize-loops): New option.
1528 2007-09-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1531 * pa.c (function_value): Use GET_MODE_BITSIZE instead of
1534 2007-09-15 Dorit Nuzman <dorit@il.ibm.com>
1536 * tree-vect-transform.c (vect_get_vec_defs_for_stmt_copy): check if
1537 the VEC is not NULL.
1538 (vectorizable_type_demotion, vectorizable_type_promotion): Check that
1539 get_vectype_for_scalar_type succeeded.
1540 (vectorizable_conversion): Likewise.
1542 2007-09-14 Jan Hubicka <jh@suse.cz>
1544 * config/i386/i386.md (*floatdi<mode>2_i387): Guard against
1547 2007-09-14 Uros Bizjak <ubizjak@gmail.com>
1550 * config/i386/i386.md (fmodxf3): Copy operands[2] to temporary
1551 register when operands[2] equals operands[1].
1552 (remainderxf3): Ditto.
1554 2007-09-14 Sandra Loosemore <sandra@codesourcery.com>
1555 Nigel Stephens <nigel@mips.com>
1557 * doc/tm.texi (LIBGCC2_UNWIND_ATTRIBUTE): Document.
1558 * unwind-generic.h (LIBGCC2_UNWIND_ATTRIBUTE): Define.
1559 (_Unwind_RaiseException): Add LIBGCC2_UNWIND_ATTRIBUTE to
1561 (_Unwind_ForcedUnwind): Likewise.
1562 (_Unwind_Resume): Likewise.
1563 (_Unwind_Resume_or_Rethrow): Likewise.
1564 (_Unwind_Backtrace): Likewise.
1565 (_Unwind_SjLj_RaiseException): Likewise.
1566 (_Unwind_SjLj_ForcedUnwind): Likewise.
1567 (_Unwind_SjLj_Resume): Likewise.
1568 (_Unwind_SjLj_Resume_or_Rethrow): Likewise.
1569 * unwind.inc (_Unwind_RaiseException): Add LIBGCC2_UNWIND_ATTRIBUTE
1571 (_Unwind_ForcedUnwind): Likewise.
1572 (_Unwind_Resume): Likewise.
1573 (_Unwind_Resume_or_Rethrow): Likewise.
1574 (_Unwind_Backtrace): Likewise.
1575 * unwind-compat.c (_Unwind_Backtrace): Likewise.
1576 (_Unwind_ForcedUnwind): Likewise.
1577 (_Unwind_RaiseException): Likewise.
1578 (_Unwind_Resume): Likewise.
1579 (_Unwind_Resume_or_Rethrow): Likewise.
1581 * config/mips/mips.h (LIBGCC2_UNWIND_ATTRIBUTE): Define to force
1582 nomips16 mode when IN_LIBGCC2 with hard float.
1584 2007-09-14 Richard Sandiford <rsandifo@nildram.co.uk>
1586 * config/mips/sdemtk.opt: Update to GPLv3.
1587 * config/mips/sdemtk.h: Likewise.
1589 2007-09-14 Nigel Stephens <nigel@mips.com>
1591 * config.gcc (mips*-*-linux*): Recognise mipsisa32r2 and set
1592 MIPS_ISA_DEFAULT appropriately. Don't make soft-float the default
1593 for mipsisa32-*-linux*.
1595 2007-09-14 Nigel Stephens <nigel@mips.com>
1596 David Ung <davidu@mips.com>
1597 Thiemo Seufer <ths@mips.com>
1598 Richard Sandiford <richard@codesourcery.com>
1600 * config.gcc (mips*-sde-elf*): Add support for the SDE C libraries.
1601 * configure.ac: Add a mipssde threading type.
1602 * configure: Regenerate.
1603 * config/mips/sdemtk.h: New file.
1604 * config/mips/t-sdemtk: Likewise.
1605 * config/mips/sdemtk.opt: Likewise.
1606 * gthr-mipssde.h: Likewise.
1607 * config/mips/sde.h (FUNCTION_PROFILER): Move to config/mips/sdemtk.h.
1608 * config/mips/mips.h (MIPS_SAVE_REG_FOR_PROFILING_P): New macro.
1609 (MIPS_ICACHE_SYNC): New macro, split from ...
1610 * config/mips/mips.md (clear_cache): ...here.
1611 * config/mips/mips.c (mips_save_reg_p): Check
1612 MIPS_SAVE_REG_FOR_PROFILING_P on profiled functions.
1613 (build_mips16_function_stub): Use targetm.strip_name_encoding.
1614 (build_mips16_call_stub): Likewise.
1616 2007-09-14 Richard Sandiford <richard@codesourcery.com>
1618 * Makefile.in (stmp-int-hdrs): Depend on fixinc_list.
1620 2007-09-14 Jakub Jelinek <jakub@redhat.com>
1623 * config/ia64/ia64.c (find_gr_spill): Don't decrement
1624 current_frame_info.n_local_regs. Don't return emitted local
1626 (ia64_compute_frame_size): Improve unwind hack to put
1627 RP, PFS, FP in that order by allowing some of the registers
1628 been already emitted, as long as they are emitted to the
1631 2007-09-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1633 * config/spu/vmx2spu.h (vec_extract, vec_insert, vec_lvlx,
1634 vec_lvlxl, vec_lvrx, vec_lvrxl, vec_promote, vec_splats,
1635 vec_stvlx, vec_stvlxl, vec_stvrx, vec_stvrxl): New intrinsics.
1637 2007-09-13 Eric Christopher <echristo@apple.com>
1638 Kenneth Zadeck <zadeck@naturalbridge.com>
1640 * dse.c (find_shift_sequence): New function.
1641 (replace_read): Add case to remove read if it requires shift.
1642 * config/i386/i386.c (ix86_expand_prologue): Fixed typo in comment.
1644 2007-09-13 Tom Tromey <tromey@redhat.com>
1646 * c-common.c (fname_as_string): Update.
1647 * c-parser.c (c_parser) <lex_untranslated_string>: New field.
1648 (c_lex_one_token): Update. Add 'parser' argument.
1649 (c_parser_simple_asm_expr): Update.
1650 (c_parser_attributes): Update.
1651 (c_parser_asm_statement): Update.
1652 (c_parser_asm_operands): Update.
1653 (c_parser_peek_token): Update.
1654 (c_parser_peek_2nd_token): Update.
1655 * c-lex.c (c_lex_string_translate): Remove.
1656 (c_lex_return_raw_strings): Likewise.
1657 (c_lex_with_flags): Added 'lex_flags' argument.
1658 (lex_string): Added 'translate' argument.
1659 * c-pragma.h (c_lex_with_flags): Update.
1660 (c_lex_string_translate, c_lex_return_raw_strings): Remove.
1661 (C_LEX_STRING_NO_TRANSLATE): New define.
1662 (C_LEX_RAW_STRINGS): Likewise.
1664 2007-09-13 Bernd Schmidt <bernd.schmidt@analog.com>
1667 * config/bfin/bfin.c (enum bfin_builtins): Add BFIN_BUILTIN_ONES,
1668 BFIN_BUILTIN_CPLX_MUL_16_S40, BFIN_BUILTIN_CPLX_MAC_16_S40,
1669 BFIN_BUILTIN_CPLX_MSU_16_S40, and BFIN_BUILTIN_CPLX_SQU.
1670 (bfin_init_builtins): Initialize __builtin_bfin_ones,
1671 __builtin_bfin_min_fr1x16, __builtin_bfin_max_fr1x16,
1672 __builtin_bfin_min_fr1x32, __builtin_bfin_max_fr1x32,
1673 __builtin_bfin_cmplx_add, __builtin_bfin_cmplx_sub,
1674 __builtin_bfin_cmplx_mul_s40, __builtin_bfin_cmplx_mac_s40,
1675 __builtin_bfin_cmplx_msu_s40 and __builtin_bfin_csqu_fr16.
1676 (bdesc_1arg): Add __builtin_bfin_ones.
1677 (bfin_expand_builtin): Expand __builtin_bfin_cmplx_mul_s40,
1678 __builtin_bfin_cmplx_mac_s40, __builtin_bfin_cmplx_msu_s40,
1679 and __builtin_bfin_csqu_fr16.
1680 * config/bfin/bfin.md (UNSPEC_ONES): New constant.
1681 (ones): New define_insn.
1682 (ssaddhi3_parts): New define_insn.
1683 (sssubhi3_parts): New define_insn.
1684 (flag_mulhi_parts): New define_insn.
1686 2007-09-13 Seongbae Park <seongbae.park@gmail.com>
1688 * common.opt (femit-class-debug-always): Turn off by default.
1690 2007-09-13 Bernd Schmidt <bernd.schmidt@analog.com>
1692 * config/bfin/bfin.md (reload_outpdi, reload_inpdi): New patterns.
1693 * config/bfin/bfin.c (bfin_secondary_reload): Make sure we use them.
1695 2007-09-13 James E. Wilson <wilson@specifix.com>
1697 PR tree-optimization/33389
1698 * tree-ssa-operands.c (append_vuse): If ann->in_vdef_list true,
1699 then set build_loads before returning.
1701 2007-09-13 Sandra Loosemore <sandra@codesourcery.com>
1702 David Ung <davidu@mips.com>
1704 * config/mips/mips.h (ASM_OUTPUT_REG_PUSH): Replace {d}subu with
1705 {d}addiu and a negative immediate such that it works with MIPS16
1708 2007-09-13 H.J. Lu <hongjiu.lu@intel.com>
1711 * configure.ac (ld_vers): Support Linux linker.
1712 * configure: Regenerated.
1714 2007-09-13 Richard Sandiford <richard@codesourcery.com>
1715 Sandra Loosemore <sandra@codesourcery.com>
1717 * config/mips/mips.h (SYMBOL_FLAG_MIPS16_FUNC): Delete.
1718 (SYMBOL_REF_MIPS16_FUNC_P): Delete.
1719 * config/mips/mips.c (mips_attribute_table): Turn mips16 and
1720 nomips16 into decl attributes.
1721 (TARGET_INSERT_ATTRIBUTES): Override.
1722 (TARGET_MERGE_DECL_ATTRIBUTES): Likewise.
1723 (TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P): Always return true.
1724 (mips_mips16_type_p, mips_nomips16_type_p): Delete in favor of...
1725 (mips_mips16_decl_p, mips_nomips16_decl_p): ...these new functions.
1726 (mips_comp_type_attributes): Remove mips16 and nomips16 handling.
1727 (mips_use_mips16_mode_p): Reimplement as a function that takes
1728 a decl and considers only decl attributes. If the decl is nested
1729 function, use its parent attributes.
1730 (mips_function_ok_for_sibcall): Use mips_use_mips16_mode_p
1731 instead of SYMBOL_REF_MIPS16_FUNC_P.
1732 (mips_set_mips16_mode): Move call to sorry here from old
1733 mips_use_mips16_mode_p.
1734 (mflip_mips16_entry): New structure.
1735 (mflip_mips16_htab): New variable.
1736 (mflip_mips16_htab_hash, mflip_mips16_htab_eq): New functions.
1737 (mflip_mips16_use_mips16_p, mips_insert_attributes): Likewise.
1738 (mips_merge_decl_attributes): New function.
1739 (mips_set_current_function): Reinstate call to mips_set_mips16_mode.
1740 Use mips_use_mips16_mode_p.
1741 (mips_output_mi_thunk): Use mips_use_mips16_mode_p instead of
1742 SYMBOL_REF_MIPS16_FUNC_P.
1743 (mips_encode_section_info): Don't set SYMBOL_FLAG_MIPS16_FUNC.
1745 2007-09-13 Richard Sandiford <richard@codesourcery.com>
1747 * c-parser.c (c_parser_struct_declaration): Check for a null return.
1749 2007-09-13 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
1752 * gcc.c (xputenv): Make argument const, and use CONST_CAST.
1754 2007-09-12 Michael Meissner <michael.meissner@amd.com>
1755 Dwarakanath Rajagopal <dwarak.rajagopal@amd.com>
1756 Tony Linthicum <tony.linthicum@amd.com>
1758 * tree.h (function_args_iterator): New type to iterate over
1760 (FOREACH_FUNCTION_ARGS_PTR): Iterator macros for iterating over
1761 function arguments providing a pointer to the argument.
1762 (FOREACH_FUNCTION_ARGS): Iterator macros for iterating over
1763 function arguments providing the argument.
1764 (function_args_iter_init): Inline function to initialize
1765 function_args_iterator.
1766 (function_args_iter_cond_ptr): Inline function to return the next
1767 pointer to hold the argument.
1768 (function_args_iter_cond): Inline function to return the next
1770 (function_args_iter_cond_next): Advance the function args
1772 (stdarg_p): New function, return true if variable argument
1774 (prototype_p): New function, return true if function is
1776 (function_args_count): New function, count the number of arguments
1779 * tree.c (stdarg_p): New function, return true if variable
1781 (prototype_p): New function, return true if function is
1784 * config/i386/i386.h (TARGET_SSE5): New macro for SSE5.
1785 (TARGET_ROUND): New macro for the round/ptest instructions which
1786 are shared between SSE4.1 and SSE5.
1787 (OPTION_MASK_ISA_ROUND): Ditto.
1788 (OPTION_ISA_ROUND): Ditto.
1789 (TARGET_FUSED_MADD): New macro for -mfused-madd swtich.
1790 (TARGET_CPU_CPP_BUILTINS): Add SSE5 support.
1792 * config/i386/i386.opt (-msse5): New switch for SSE5 support.
1793 (-mfused-madd): New switch to give users control over whether the
1794 compiler optimizes to use the multiply/add SSE5 instructions.
1796 * config/i386/i386.c (m_AMD_MULTIPLE): Rename from
1797 m_ATHLON_K8_AMDFAM10, and change all uses.
1798 (enum pta_flags): Add PTA_SSE5.
1799 (ix86_handle_option): Turn off 3dnow if -msse5.
1800 (override_options): Add SSE5 support.
1801 (print_operand): %Y prints comparison codes for SSE5 com/pcom
1803 (ix86_expand_sse_movcc): Add SSE5 support.
1804 (ix86_expand_sse5_unpack): New function to use pperm to unpack a
1805 vector type to the next largest size.
1806 (ix86_expand_sse5_pack): New function to use pperm to pack a
1807 vector type to the next smallest size.
1808 (IX86_BUILTIN_FMADDSS): New for SSE5 intrinsic.
1809 (IX86_BUILTIN_FMADDSD): Ditto.
1810 (IX86_BUILTIN_FMADDPS): Ditto.
1811 (IX86_BUILTIN_FMADDPD): Ditto.
1812 (IX86_BUILTIN_FMSUBSS): Ditto.
1813 (IX86_BUILTIN_FMSUBSD): Ditto.
1814 (IX86_BUILTIN_FMSUBPS): Ditto.
1815 (IX86_BUILTIN_FMSUBPD): Ditto.
1816 (IX86_BUILTIN_FNMADDSS): Ditto.
1817 (IX86_BUILTIN_FNMADDSD): Ditto.
1818 (IX86_BUILTIN_FNMADDPS): Ditto.
1819 (IX86_BUILTIN_FNMADDPD): Ditto.
1820 (IX86_BUILTIN_FNMSUBSS): Ditto.
1821 (IX86_BUILTIN_FNMSUBSD): Ditto.
1822 (IX86_BUILTIN_FNMSUBPS): Ditto.
1823 (IX86_BUILTIN_FNMSUBPD): Ditto.
1824 (IX86_BUILTIN_PCMOV_V2DI): Ditto.
1825 (IX86_BUILTIN_PCMOV_V4SI): Ditto.
1826 (IX86_BUILTIN_PCMOV_V8HI): Ditto.
1827 (IX86_BUILTIN_PCMOV_V16QI): Ditto.
1828 (IX86_BUILTIN_PCMOV_V4SF): Ditto.
1829 (IX86_BUILTIN_PCMOV_V2DF): Ditto.
1830 (IX86_BUILTIN_PPERM): Ditto.
1831 (IX86_BUILTIN_PERMPS): Ditto.
1832 (IX86_BUILTIN_PERMPD): Ditto.
1833 (IX86_BUILTIN_PMACSSWW): Ditto.
1834 (IX86_BUILTIN_PMACSWW): Ditto.
1835 (IX86_BUILTIN_PMACSSWD): Ditto.
1836 (IX86_BUILTIN_PMACSWD): Ditto.
1837 (IX86_BUILTIN_PMACSSDD): Ditto.
1838 (IX86_BUILTIN_PMACSDD): Ditto.
1839 (IX86_BUILTIN_PMACSSDQL): Ditto.
1840 (IX86_BUILTIN_PMACSSDQH): Ditto.
1841 (IX86_BUILTIN_PMACSDQL): Ditto.
1842 (IX86_BUILTIN_PMACSDQH): Ditto.
1843 (IX86_BUILTIN_PMADCSSWD): Ditto.
1844 (IX86_BUILTIN_PMADCSWD): Ditto.
1845 (IX86_BUILTIN_PHADDBW): Ditto.
1846 (IX86_BUILTIN_PHADDBD): Ditto.
1847 (IX86_BUILTIN_PHADDBQ): Ditto.
1848 (IX86_BUILTIN_PHADDWD): Ditto.
1849 (IX86_BUILTIN_PHADDWQ): Ditto.
1850 (IX86_BUILTIN_PHADDDQ): Ditto.
1851 (IX86_BUILTIN_PHADDUBW): Ditto.
1852 (IX86_BUILTIN_PHADDUBD): Ditto.
1853 (IX86_BUILTIN_PHADDUBQ): Ditto.
1854 (IX86_BUILTIN_PHADDUWD): Ditto.
1855 (IX86_BUILTIN_PHADDUWQ): Ditto.
1856 (IX86_BUILTIN_PHADDUDQ): Ditto.
1857 (IX86_BUILTIN_PHSUBBW): Ditto.
1858 (IX86_BUILTIN_PHSUBWD): Ditto.
1859 (IX86_BUILTIN_PHSUBDQ): Ditto.
1860 (IX86_BUILTIN_PROTB): Ditto.
1861 (IX86_BUILTIN_PROTW): Ditto.
1862 (IX86_BUILTIN_PROTD): Ditto.
1863 (IX86_BUILTIN_PROTQ): Ditto.
1864 (IX86_BUILTIN_PROTB_IMM): Ditto.
1865 (IX86_BUILTIN_PROTW_IMM): Ditto.
1866 (IX86_BUILTIN_PROTD_IMM): Ditto.
1867 (IX86_BUILTIN_PROTQ_IMM): Ditto.
1868 (IX86_BUILTIN_PSHLB): Ditto.
1869 (IX86_BUILTIN_PSHLW): Ditto.
1870 (IX86_BUILTIN_PSHLD): Ditto.
1871 (IX86_BUILTIN_PSHLQ): Ditto.
1872 (IX86_BUILTIN_PSHAB): Ditto.
1873 (IX86_BUILTIN_PSHAW): Ditto.
1874 (IX86_BUILTIN_PSHAD): Ditto.
1875 (IX86_BUILTIN_PSHAQ): Ditto.
1876 (IX86_BUILTIN_FRCZSS): Ditto.
1877 (IX86_BUILTIN_FRCZSD): Ditto.
1878 (IX86_BUILTIN_FRCZPS): Ditto.
1879 (IX86_BUILTIN_FRCZPD): Ditto.
1880 (IX86_BUILTIN_CVTPH2PS): Ditto.
1881 (IX86_BUILTIN_CVTPS2PH): Ditto.
1882 (IX86_BUILTIN_COMEQSS): Ditto.
1883 (IX86_BUILTIN_COMNESS): Ditto.
1884 (IX86_BUILTIN_COMLTSS): Ditto.
1885 (IX86_BUILTIN_COMLESS): Ditto.
1886 (IX86_BUILTIN_COMGTSS): Ditto.
1887 (IX86_BUILTIN_COMGESS): Ditto.
1888 (IX86_BUILTIN_COMUEQSS): Ditto.
1889 (IX86_BUILTIN_COMUNESS): Ditto.
1890 (IX86_BUILTIN_COMULTSS): Ditto.
1891 (IX86_BUILTIN_COMULESS): Ditto.
1892 (IX86_BUILTIN_COMUGTSS): Ditto.
1893 (IX86_BUILTIN_COMUGESS): Ditto.
1894 (IX86_BUILTIN_COMORDSS): Ditto.
1895 (IX86_BUILTIN_COMUNORDSS): Ditto.
1896 (IX86_BUILTIN_COMFALSESS): Ditto.
1897 (IX86_BUILTIN_COMTRUESS): Ditto.
1898 (IX86_BUILTIN_COMEQSD): Ditto.
1899 (IX86_BUILTIN_COMNESD): Ditto.
1900 (IX86_BUILTIN_COMLTSD): Ditto.
1901 (IX86_BUILTIN_COMLESD): Ditto.
1902 (IX86_BUILTIN_COMGTSD): Ditto.
1903 (IX86_BUILTIN_COMGESD): Ditto.
1904 (IX86_BUILTIN_COMUEQSD): Ditto.
1905 (IX86_BUILTIN_COMUNESD): Ditto.
1906 (IX86_BUILTIN_COMULTSD): Ditto.
1907 (IX86_BUILTIN_COMULESD): Ditto.
1908 (IX86_BUILTIN_COMUGTSD): Ditto.
1909 (IX86_BUILTIN_COMUGESD): Ditto.
1910 (IX86_BUILTIN_COMORDSD): Ditto.
1911 (IX86_BUILTIN_COMUNORDSD): Ditto.
1912 (IX86_BUILTIN_COMFALSESD): Ditto.
1913 (IX86_BUILTIN_COMTRUESD): Ditto.
1914 (IX86_BUILTIN_COMEQPS): Ditto.
1915 (IX86_BUILTIN_COMNEPS): Ditto.
1916 (IX86_BUILTIN_COMLTPS): Ditto.
1917 (IX86_BUILTIN_COMLEPS): Ditto.
1918 (IX86_BUILTIN_COMGTPS): Ditto.
1919 (IX86_BUILTIN_COMGEPS): Ditto.
1920 (IX86_BUILTIN_COMUEQPS): Ditto.
1921 (IX86_BUILTIN_COMUNEPS): Ditto.
1922 (IX86_BUILTIN_COMULTPS): Ditto.
1923 (IX86_BUILTIN_COMULEPS): Ditto.
1924 (IX86_BUILTIN_COMUGTPS): Ditto.
1925 (IX86_BUILTIN_COMUGEPS): Ditto.
1926 (IX86_BUILTIN_COMORDPS): Ditto.
1927 (IX86_BUILTIN_COMUNORDPS): Ditto.
1928 (IX86_BUILTIN_COMFALSEPS): Ditto.
1929 (IX86_BUILTIN_COMTRUEPS): Ditto.
1930 (IX86_BUILTIN_COMEQPD): Ditto.
1931 (IX86_BUILTIN_COMNEPD): Ditto.
1932 (IX86_BUILTIN_COMLTPD): Ditto.
1933 (IX86_BUILTIN_COMLEPD): Ditto.
1934 (IX86_BUILTIN_COMGTPD): Ditto.
1935 (IX86_BUILTIN_COMGEPD): Ditto.
1936 (IX86_BUILTIN_COMUEQPD): Ditto.
1937 (IX86_BUILTIN_COMUNEPD): Ditto.
1938 (IX86_BUILTIN_COMULTPD): Ditto.
1939 (IX86_BUILTIN_COMULEPD): Ditto.
1940 (IX86_BUILTIN_COMUGTPD): Ditto.
1941 (IX86_BUILTIN_COMUGEPD): Ditto.
1942 (IX86_BUILTIN_COMORDPD): Ditto.
1943 (IX86_BUILTIN_COMUNORDPD): Ditto.
1944 (IX86_BUILTIN_COMFALSEPD): Ditto.
1945 (IX86_BUILTIN_COMTRUEPD): Ditto.
1946 (IX86_BUILTIN_PCOMEQUB): Ditto.
1947 (IX86_BUILTIN_PCOMNEUB): Ditto.
1948 (IX86_BUILTIN_PCOMLTUB): Ditto.
1949 (IX86_BUILTIN_PCOMLEUB): Ditto.
1950 (IX86_BUILTIN_PCOMGTUB): Ditto.
1951 (IX86_BUILTIN_PCOMGEUB): Ditto.
1952 (IX86_BUILTIN_PCOMFALSEUB): Ditto.
1953 (IX86_BUILTIN_PCOMTRUEUB): Ditto.
1954 (IX86_BUILTIN_PCOMEQUW): Ditto.
1955 (IX86_BUILTIN_PCOMNEUW): Ditto.
1956 (IX86_BUILTIN_PCOMLTUW): Ditto.
1957 (IX86_BUILTIN_PCOMLEUW): Ditto.
1958 (IX86_BUILTIN_PCOMGTUW): Ditto.
1959 (IX86_BUILTIN_PCOMGEUW): Ditto.
1960 (IX86_BUILTIN_PCOMFALSEUW): Ditto.
1961 (IX86_BUILTIN_PCOMTRUEUW): Ditto.
1962 (IX86_BUILTIN_PCOMEQUD): Ditto.
1963 (IX86_BUILTIN_PCOMNEUD): Ditto.
1964 (IX86_BUILTIN_PCOMLTUD): Ditto.
1965 (IX86_BUILTIN_PCOMLEUD): Ditto.
1966 (IX86_BUILTIN_PCOMGTUD): Ditto.
1967 (IX86_BUILTIN_PCOMGEUD): Ditto.
1968 (IX86_BUILTIN_PCOMFALSEUD): Ditto.
1969 (IX86_BUILTIN_PCOMTRUEUD): Ditto.
1970 (IX86_BUILTIN_PCOMEQUQ): Ditto.
1971 (IX86_BUILTIN_PCOMNEUQ): Ditto.
1972 (IX86_BUILTIN_PCOMLTUQ): Ditto.
1973 (IX86_BUILTIN_PCOMLEUQ): Ditto.
1974 (IX86_BUILTIN_PCOMGTUQ): Ditto.
1975 (IX86_BUILTIN_PCOMGEUQ): Ditto.
1976 (IX86_BUILTIN_PCOMFALSEUQ): Ditto.
1977 (IX86_BUILTIN_PCOMTRUEUQ): Ditto.
1978 (IX86_BUILTIN_PCOMEQB): Ditto.
1979 (IX86_BUILTIN_PCOMNEB): Ditto.
1980 (IX86_BUILTIN_PCOMLTB): Ditto.
1981 (IX86_BUILTIN_PCOMLEB): Ditto.
1982 (IX86_BUILTIN_PCOMGTB): Ditto.
1983 (IX86_BUILTIN_PCOMGEB): Ditto.
1984 (IX86_BUILTIN_PCOMFALSEB): Ditto.
1985 (IX86_BUILTIN_PCOMTRUEB): Ditto.
1986 (IX86_BUILTIN_PCOMEQW): Ditto.
1987 (IX86_BUILTIN_PCOMNEW): Ditto.
1988 (IX86_BUILTIN_PCOMLTW): Ditto.
1989 (IX86_BUILTIN_PCOMLEW): Ditto.
1990 (IX86_BUILTIN_PCOMGTW): Ditto.
1991 (IX86_BUILTIN_PCOMGEW): Ditto.
1992 (IX86_BUILTIN_PCOMFALSEW): Ditto.
1993 (IX86_BUILTIN_PCOMTRUEW): Ditto.
1994 (IX86_BUILTIN_PCOMEQD): Ditto.
1995 (IX86_BUILTIN_PCOMNED): Ditto.
1996 (IX86_BUILTIN_PCOMLTD): Ditto.
1997 (IX86_BUILTIN_PCOMLED): Ditto.
1998 (IX86_BUILTIN_PCOMGTD): Ditto.
1999 (IX86_BUILTIN_PCOMGED): Ditto.
2000 (IX86_BUILTIN_PCOMFALSED): Ditto.
2001 (IX86_BUILTIN_PCOMTRUED): Ditto.
2002 (IX86_BUILTIN_PCOMEQQ): Ditto.
2003 (IX86_BUILTIN_PCOMNEQ): Ditto.
2004 (IX86_BUILTIN_PCOMLTQ): Ditto.
2005 (IX86_BUILTIN_PCOMLEQ): Ditto.
2006 (IX86_BUILTIN_PCOMGTQ): Ditto.
2007 (IX86_BUILTIN_PCOMGEQ): Ditto.
2008 (IX86_BUILTIN_PCOMFALSEQ): Ditto.
2009 (IX86_BUILTIN_PCOMTRUEQ): Ditto.
2010 (bdesc_ptest): Change OPTION_MASK_ISA_SSE4_1 to
2011 OPTION_MASK_ISA_ROUND for instructions that are shared between
2013 (bdesc_2arg): Ditto.
2014 (bdesc_sse_3arg): Ditto.
2015 (enum multi_arg_type): New enum for describing the various SSE5
2016 intrinsic argument types.
2017 (bdesc_multi_arg): New table for SSE5 intrinsics.
2018 (ix86_init_mmx_sse_builtins): Add SSE5 intrinsic support.
2019 (ix86_expand_multi_arg_builtin): New function for creating SSE5
2021 (ix86_expand_builtin): Add SSE5 intrinsic support.
2022 (ix86_sse5_valid_op_p): New function to validate SSE5 3 and 4
2023 operand instructions.
2024 (ix86_expand_sse5_multiple_memory): New function to split the
2025 second memory reference from SSE5 instructions.
2026 (type_has_variadic_args_p): Delete in favor of stdarg_p.
2027 (ix86_return_pops_args): Use stdarg_p to determine if the function
2028 has variable arguments.
2029 (ix86_setup_incoming_varargs): Ditto.
2030 (x86_this_parameter): Ditto.
2032 * config/i386/i386-protos.h (ix86_expand_sse5_unpack): Add
2034 (ix86_expand_sse5_pack): Ditto.
2035 (ix86_sse5_valid_op_p): Ditto.
2036 (ix86_expand_sse5_multiple_memory): Ditto.
2038 * config/i386/i386.md (UNSPEC_SSE5_INTRINSIC): Add new UNSPEC
2039 constant for SSE5 support.
2040 (UNSPEC_SSE5_UNSIGNED_CMP): Ditto.
2041 (UNSPEC_SSE5_TRUEFALSE): Ditto.
2042 (UNSPEC_SSE5_PERMUTE): Ditto.
2043 (UNSPEC_SSE5_ASHIFT): Ditto.
2044 (UNSPEC_SSE5_LSHIFT): Ditto.
2045 (UNSPEC_FRCZ): Ditto.
2046 (UNSPEC_CVTPH2PS): Ditto.
2047 (UNSPEC_CVTPS2PH): Ditto.
2048 (PCOM_FALSE): Add new constant for true/false SSE5 comparisons.
2050 (COM_FALSE_S): Ditto.
2051 (COM_FALSE_P): Ditto.
2052 (COM_TRUE_S): Ditto.
2053 (COM_TRUE_P): Ditto.
2054 (type attribute): Add ssemuladd, sseiadd1, ssecvt1, sse4arg types.
2055 (unit attribute): Add support for ssemuladd, ssecvt1, sseiadd1 sse4arg
2057 (memory attribute): Ditto.
2058 (sse4_1_round<mode>2): Use TARGET_ROUND instead of TARGET_SSE4_1.
2059 Use SSE4_1_ROUND_* constants instead of hard coded numbers.
2060 (rint<mode>2): Use TARGET_ROUND instead of TARGET_SSE4_1.
2061 (floor<mode>2): Ditto.
2062 (ceil<mode>2): Ditto.
2063 (btrunc<mode>2): Ditto.
2064 (nearbyintdf2): Ditto.
2065 (nearbyintsf2): Ditto.
2066 (sse_setccsf): Disable if SSE5.
2067 (sse_setccdf): Ditto.
2068 (sse5_setcc<mode>): New support for SSE5 conditional move.
2069 (sse5_pcmov_<mode>): Ditto.
2071 * config/i386/sse.md (SSEMODE1248): New mode iterator for SSE5.
2073 (SSEMODEF2P): Ditto.
2074 (ssemodesuffixf4): New mode attribute for SSE5.
2075 (ssemodesuffixf2s): Ditto.
2076 (ssemodesuffixf2c): Ditto.
2077 (sserotatemax): Ditto.
2078 (ssescalarmode): Ditto.
2079 (sse_maskcmpv4sf3): Disable if SSE5.
2080 (sse_maskcmpv2df3): Ditto.
2081 (sse_vmmaskcmpv4sf3): Ditto.
2082 (sse5_fmadd<mode>4): Add SSE5 floating point multiply/add
2084 (sse5_vmfmadd<mode>4): Ditto.
2085 (sse5_fmsub<mode>4): Ditto.
2086 (sse5_vmfmsub<mode>4): Ditto.
2087 (sse5_fnmadd<mode>4): Ditto.
2088 (sse5_vmfnmadd<mode>4): Ditto.
2089 (sse5_fnmsub<mode>4): Ditto.
2090 (sse5_vmfnmsub<mode>4): Ditto.
2091 (sse5i_fmadd<mode>4): Ditto.
2092 (sse5i_fmsub<mode>4): Ditto.
2093 (sse5i_fnmadd<mode>4): Ditto.
2094 (sse5i_fnmsub<mode>4): Ditto.
2095 (sse5i_vmfmadd<mode>4): Ditto.
2096 (sse5i_vmfmsub<mode>4): Ditto.
2097 (sse5i_vmfnmadd<mode>4): Ditto.
2098 (sse5i_vmfnmsub<mode>4): Ditto.
2099 (mulv16qi3): Add SSE5 support.
2101 (sse5_mulv4si3): New insn for 32-bit multiply support on SSE5.
2102 (sse2_mulv4si3): Disable if SSE5.
2103 (sse4_1_roundpd): Use TARGET_ROUND instead of TARGET_SSE4_1.
2104 (sse4_1_roundps): Ditto.
2105 (sse4_1_roundsd): Ditto.
2106 (sse4_1_roundss): Ditto.
2107 (sse_maskcmpv4sf3): Disable if SSE5 so the SSE5 instruction will
2109 (sse_maskcmpsf3): Ditto.
2110 (sse_vmmaskcmpv4sf3): Ditto.
2111 (sse2_maskcmpv2df3): Ditto.
2112 (sse2_maskcmpdf3): Ditto.
2113 (sse2_vmmaskcmpv2df3): Ditto.
2114 (sse2_eq<mode>3): Ditto.
2115 (sse2_gt<mode>3): Ditto.
2116 (sse5_pcmov_<mode>): Add SSE5 support.
2117 (vec_unpacku_hi_v16qi): Ditto.
2118 (vec_unpacks_hi_v16qi): Ditto.
2119 (vec_unpacku_lo_v16qi): Ditto.
2120 (vec_unpacks_lo_v16qi): Ditto.
2121 (vec_unpacku_hi_v8hi): Ditto.
2122 (vec_unpacks_hi_v8hi): Ditto.
2123 (vec_unpacku_lo_v8hi): Ditto.
2124 (vec_unpacks_lo_v8hi): Ditto.
2125 (vec_unpacku_hi_v4si): Ditto.
2126 (vec_unpacks_hi_v4si): Ditto.
2127 (vec_unpacku_lo_v4si): Ditto.
2128 (vec_unpacks_lo_v4si): Ditto.
2129 (sse5_pmacsww): New SSE5 intrinsic insn.
2130 (sse5_pmacssww): Ditto.
2131 (sse5_pmacsdd): Ditto.
2132 (sse5_pmacssdd): Ditto.
2133 (sse5_pmacssdql): Ditto.
2134 (sse5_pmacssdqh): Ditto.
2135 (sse5_pmacsdqh): Ditto.
2136 (sse5_pmacsswd): Ditto.
2137 (sse5_pmacswd): Ditto.
2138 (sse5_pmadcsswd): Ditto.
2139 (sse5_pmadcswd): Ditto.
2140 (sse5_pcmov_<move>): Conditional move support on SSE5.
2141 (sse5_phaddbw): New SSE5 intrinsic insn.
2142 (sse5_phaddbd): Ditto.
2143 (sse5_phaddbq): Ditto.
2144 (sse5_phaddwd): Ditto.
2145 (sse5_phaddwq): Ditto.
2146 (sse5_phadddq): Ditto.
2147 (sse5_phaddubw): Ditto.
2148 (sse5_phaddubd): Ditto.
2149 (sse5_phaddubq): Ditto.
2150 (sse5_phadduwd): Ditto.
2151 (sse5_phadduwq): Ditto.
2152 (sse5_phaddudq): Ditto.
2153 (sse5_phsubbw): Ditto.
2154 (sse5_phsubwd): Ditto.
2155 (sse5_phsubdq): Ditto.
2156 (sse5_pperm): Ditto.
2157 (sse5_pperm_sign_v16qi_v8hi): New insns for pack/unpack with SSE5.
2158 (sse5_pperm_zero_v16qi_v8hi): Ditto.
2159 (sse5_pperm_sign_v8hi_v4si): Ditto.
2160 (sse5_pperm_zero_v8hi_v4si): Ditto.
2161 (sse5_pperm_sign_v4si_v2di): Ditto.
2162 (sse5_pperm_sign_v4si_v2di): Ditto.
2163 (sse5_pperm_pack_v2di_v4si): Ditto.
2164 (sse5_pperm_pack_v4si_v8hi): Ditto.
2165 (sse5_pperm_pack_v8hi_v16qi): Ditto.
2166 (sse5_perm<mode>): New SSE5 intrinsic insn.
2167 (rotl<mode>3): Ditto.
2168 (sse5_rotl<mode>3): Ditto.
2169 (sse5_ashl<mode>3): Ditto.
2170 (sse5_lshl<mode>3): Ditto.
2171 (sse5_frcz<mode>2): Ditto.
2172 (sse5s_frcz<mode>2): Ditto.
2173 (sse5_cvtph2ps): Ditto.
2174 (sse5_cvtps2ph): Ditto.
2175 (sse5_vmmaskcmp<mode>3): Ditto.
2176 (sse5_com_tf<mode>3): Ditto.
2177 (sse5_maskcmp<mode>3): Ditto.
2178 (sse5_maskcmp_uns<mode>3): Ditto.
2179 (sse5_maskcmp_uns2<mode>3): Ditto.
2180 (sse5_pcom_tf<mode>3): Ditto.
2182 * config/i386/predicates.md (const_0_to_31_operand): New predicate
2184 (sse5_comparison_float_operator): New predicate to match the
2185 comparison operators supported by the SSE5 com instruction.
2186 (ix86_comparison_int_operator): New predicate to match just the
2187 signed int comparisons.
2188 (ix86_comparison_uns_operator): New predicate to match just the
2189 unsigned int comparisons.
2191 * doc/invoke.texi (-msse5): Add documentation.
2192 (-mfused-madd): Ditto.
2194 * doc/extend.texi (x86 intrinsics): Document new SSE5 intrinsics.
2196 * config.gcc (i[34567]86-*-*): Include bmmintrin.h and
2198 (x86_64-*-*): Ditto.
2200 * config/i386/cpuid.h (bit_SSE5): Define SSE5 bit.
2202 * config/i386/bmmintrin.h: New file, provide common x86 compiler
2203 intrinisics for SSE5.
2205 * config/i386/smmintrin.h: Move instructions shared with SSE5 to
2208 * config/i386/mmintrin-common.h: New file, to contain common
2209 instructions between SSE4.1 and SSE5.
2211 * config/i386/netware.c (gen_stdcall_or_fastcall_decoration): Use
2212 FOREACH_FUNCTION_ARGS to iterate over the argument list.
2213 (gen_regparm_prefix): Ditto.
2215 * config/i386/winnt.c (gen_stdcall_or_fastcall_suffix): Use
2216 FOREACH_FUNCTION_ARGS to iterate over the argument list. Use
2217 prototype_p to determine if a function is prototyped.
2219 2007-09-12 Janis Johnson <janis187@us.ibm.com>
2221 * config/dfp-bit.c (dfp_conversion_exception): New function.
2222 (DFP_TO_DFP) Add new variants to use direct conversions in decNumber.
2223 (DFP_TO_INT): Ditto.
2224 (INT_TO_DFP): Ditto.
2225 * config/dfp-bit.h (DEC_FLOAT_FROM_INT, DEC_FLOAT_TO_INT): New.
2227 2007-09-12 Jakub Jelinek <jakub@redhat.com>
2230 * config/ia64/ia64.c (ia64_expand_epilogue): Emit blockage
2231 before sp restoration even when total_size is 0, but
2232 frame_pointer_needed.
2234 2007-09-12 Bob Wilson <bob.wilson@acm.org>
2236 * config/xtensa/xtensa.c (machine_function): Add vararg_a7_copy.
2237 (xtensa_copy_incoming_a7): Use start_sequence instead of
2238 push_to_sequence. Stash insns in vararg_a7_copy for builtin_saveregs.
2239 (xtensa_builtin_saveregs): Place code from vararg_a7_copy at the start
2240 of the saveregs sequence.
2242 2007-09-12 Richard Sandiford <richard@codesourcery.com>
2244 * c-tree.h (grokfield): Add a "tree *" argument.
2245 * c-decl.c (grokdeclarator): Take a pointer to the decl's attributes.
2246 Chain nested decl attributes to it. Don't call decl_attributes here.
2247 (groktypename): Pass grokdeclarator a pointer to the attribute list.
2248 (start_decl, grokparm, push_parm_decl, start_function): Likewise.
2249 (grokfield): Take a pointer to the decl's attributes and pass
2250 it to grokdeclarator.
2251 * c-parser.c (c_parser_struct_declaration): Update the calls to
2252 grokfield. Call decl_attributes for anonymous struct and union
2255 2007-09-12 Jan Hubicka <jh@suse.cz>
2257 * c-objc-common.h (LANG_HOOKS_CALLGRAPH_EXPAND_FUNCTION): Kill.
2259 2007-09-12 Ira Rosen <irar@il.ibm.com>
2261 PR tree-optimization/32377
2262 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Distinguish
2263 between positive and negative dependence distance using DDR_REVERSED_P.
2265 2007-09-12 Dorit Nuzman <dorit@il.ibm.com>
2267 PR tree-optimization/33373
2268 * tree-vect-analyze (vect_determine_vectorization_factor): Call
2269 TREE_INT_CST_LOW when comparing TYPE_SIZE_UNIT.
2271 2007-09-12 Jan Hubicka <jh@suse.cz>
2274 * i386.md (floatsisf2_mixed_memory, floatsisf2_sse_memory): Disable
2277 2007-09-12 Christian Bruel <christian.bruel@st.com>
2279 * sh.h (SH_DBX_REGISTER_NUMBER): Added fpscr, fixed sr/gbr regs.
2280 * linux-unwind.h (SH_DWARF_FRAME_GBR): fixed.
2282 2007-09-12 Ira Rosen <irar@il.ibm.com>
2284 * tree-vect-transform.c (vect_get_slp_defs): Don't build a vector
2285 for oprnd1 if not required.
2286 (vectorizable_operation): Use scalar operand in SLP in case of
2287 shift with scalar argument.
2289 2007-09-12 Ira Rosen <irar@il.ibm.com>
2291 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change default and minimum
2294 2007-09-11 James E. Wilson <wilson@specifix.com>
2296 * defaults.h (DWARF2_UNWIND_INFO): Don't define if
2297 TARGET_UNWIND_INFO is defined.
2298 * config/ia64/ia64.h (INCOMING_RETURN_ADDR_RTX): Delete undef
2301 2007-09-12 Kaz Kojima <kkojima@gcc.gnu.org>
2303 * config/sh/sh.c (calc_live_regs): Use
2304 current_function_saves_all_registers instead of
2305 current_function_has_nonlocal_label.
2306 (sh_allocate_initial_value): Likewise.
2307 (sh_get_pr_initial_val): Likewise.
2308 * config/sh/sh.h (SHMEDIA_REGS_STACK_ADJUST): Likewise.
2309 * config/sh/sh.md (load_ra): Likewise.
2311 2007-09-12 Hans-Peter Nilsson <hp@axis.com>
2313 * config/cris/t-linux (LIMITS_H_TEST): Only define if not inhibit_libc.
2316 * config/cris/cris.c (cris_expand_pic_call_address): Fix typo in
2317 GET_CODE (x) == CONST_INT to CONST_INT_P (x) transformation.
2319 2007-09-12 Sa Liu <saliu@de.ibm.com>
2321 * config/spu/spu.c (spu_emit_branch_or_set): Handle NaN values as
2322 operands to DFmode GE or LE compares.
2324 2007-09-12 Bernd Schmidt <bernd.schmidt@analog.com>
2326 * config/bfin/bfin.h (enum reg_class, REG_CLASS_CONTENTS,
2327 REG_CLASS_NAMES): Add P0REGS.
2328 (REGNO_REG_CLASS): Return it where appropriate.
2329 (REG_CLASS_FROM_CONSTRAINT): Add 'qA'.
2330 (CLASS_LIKELY_SPILLED_P): P0REGS is likely_spilled.
2331 * doc/md.texi (Blackfin family): Document 'q' constraints.
2333 2007-09-11 Steve Kenton <skenton@ou.edu>
2335 * pa/linux-unwind.h: Guard with inhibit_libc.
2336 * pa/hpux-unwind.h: Likewise.
2338 2007-09-11 David Daney <ddaney@avtrex.com>
2340 * doc/invoke.texi: Document new MIPS -mllsc and -mno-llsc options.
2341 * doc/install.texi: Document new --with-llsc and --without-llsc
2343 * config.gcc: Handle --with-llsc and --without-llsc configure options.
2344 * config/mips/mips.md (sync, memory_barrier): Wrap sync instrunction
2345 in %| and %- operand codes. Depend on GENERATE_SYNC instead of
2347 (sync_compare_and_swap<mode>, sync_add<mode>, sync_sub<mode>,
2348 sync_old_add<mode>, sync_old_sub<mode>, sync_new_add<mode>,
2349 sync_new_sub<mode>, sync_<optab><mode>, sync_old_<optab><mode>,
2350 sync_new_<optab><mode>, sync_nand<mode>, sync_old_nand<mode>,
2351 sync_new_nand<mode>, sync_lock_test_and_set<mode>): Depend on
2352 GENERATE_LL_SC instead of ISA_HAS_LL_SC.
2353 * config/mips/mips.opt (mllsc): New option.
2354 * config/mips/mips.c (mips_llsc): Define variable.
2355 (mips_handle_option): Handle mllsc option.
2356 (override_options): Set mips_print_operand_punct for '|' and '-'.
2357 (print_operand): Add new %| and %- operand codes.
2358 * config/mips/mips.h (mips_llsc_setting): New enum type.
2359 (mips_llsc): Declare.
2360 (OPTION_DEFAULT_SPECS): Add llsc handling.
2361 (GENERATE_SYNC): New macro.
2362 (GENERATE_LL_SC): New macro.
2363 (MIPS_COMPARE_AND_SWAP, MIPS_SYNC_OP, MIPS_SYNC_OLD_OP,
2364 MIPS_SYNC_NEW_OP, MIPS_SYNC_NAND, MIPS_SYNC_OLD_NAND,
2365 MIPS_SYNC_NEW_NAND, MIPS_SYNC_EXCHANGE): Wrap instructions
2366 in %| and %- operand codes.
2368 2007-09-11 Eric Botcazou <ebotcazou@adacore.com>
2370 * tree-ssa-structalias.c (push_fields_onto_fieldstack): Deal with
2371 TYPE_NONALIASED_COMPONENT like with DECL_NONADDRESSABLE_P.
2373 2007-09-11 Jason Merrill <jason@redhat.com>
2376 * stor-layout.c (layout_decl): Do pack variable size fields.
2378 2007-09-11 Maxim Kuvyrkov <maxim@codesourcery.com>
2380 * config/m68k/predicates.md (movsi_const0_operand,
2381 non_symbolic_call_operand): New predicates.
2383 * config/m68k/constraints.md (Cs, Ci, C0, Cj, CQ, CW, CZ, CS, Ap, Ac):
2385 * doc/md.texi (Constraints for Particular Machines: Motorola 680x0):
2386 Document constraints N, O, P, R, S, T, Q, U, W, Cs, Ci, C0, Cj, CQ,
2387 CW, CZ, CS, Ap and Ac.
2389 * config/m68k/m68k.md (UNSPEC_IB): New constant.
2390 (constraints.md): New include.
2391 (cpu, type, type1, opx, opy, opx_type, opy_type, size, opx_access,
2392 opx_mem, opy_mem, op_mem, guess, split): New attributes.
2393 (movdf_internal): Name pattern. Fix to use alternatives. Add split.
2395 (pushdi): Add split.
2396 (tstsi_internal): Name pattern. Fix to use alternatives. Specify
2397 attributes. Split tstsi_internal_68020_cf from it.
2398 (tstsi_internal_68020_cf): New pattern.
2399 (tsthi_internal, tstqi_internal): Name pattern. Specify attributes.
2400 (tst<mode>_cf): Specify attributea.
2401 (cmpsi_cf): Name pattern. Specify attributes.
2402 (cmp<mode>_68881, cmp<mode>_cf): Specify type attribute.
2403 (pushexthisi_const): Fix to use alternatives. Specify
2405 (movsi_const0): Split movsi_const0_68000_10 and movsi_const0_68040_60
2406 from it. Fix to use alternatives. Specify attributes.
2407 (movsi_const0_68040_10, movsi_const0_68040_60): New patterns.
2408 (movsi_cf, movstrictqi_cf): Fix to use alternatives. Specify
2410 (movsf_cf_soft): Specify attributes.
2411 (movdf_cf_soft): Add split.
2412 (pushasi, zero_extendhisi2_cf, zero_extendqisi2_cfv4,
2413 cfv4_extendhisi2, 68k_extendhisi2, extendqihi2, cfv4_extendqisi2,
2414 68k_extendqisi2, truncdfsf2_cf): Specify attributes.
2415 (truncdfsf2_68881): Name pattern. Specify attributes.
2416 (floatsi<mode>2_cf, floathi<mode>2_68881, floathi<mode>2_cf,
2417 floatqi<mode>2_68881, floatqi<mode>2_cf, ftrunc<mode>2_cf,
2418 fix<mode>qi2_cf, fix<mode>hi2_cf, fix<mode>si2_cf, adddi_dishl32):
2420 (addsi3_5200): Fix to use alternatives. Specify attributes.
2422 (add<mode>3_cf, subdi_dishl32): Specify attributes.
2423 (subsi3): Add alternative for subq.l. Specify attributes.
2424 (sub<mode>3_cf, mulhi3, mulhisi3): Specify attributes.
2425 (mulhisisi3_s, mulsi3_68020, mulsi3_cf): Name pattern. Specify
2427 (umulhisi3): Specify attributes.
2428 (mulhisisi3_z): Name pattern. Specify attributes.
2429 (fmul<mode>3_cf, div<mode>3_cf, negsi2_internal, negsi2_5200,
2430 sqrt<mode>2_68881, clzsi2, one_cmplsi2_5200, subreghi1ashrdi_const32,
2431 subregsi1ashrdi_const32, ashrsi3, subreg1lshrdi_const32, lshrsi3,
2432 bsetmemqi): Specify attributes.
2433 (bsetmemqi_ext): Name pattern. Specify attributes.
2434 (bclrmemqi): Specify attributes.
2435 (bclrmemqi_ext, scc, sls): Name pattern. Specify attributes.
2436 (beq, bne, bgt, bgtu, blt, bltu, bge, bgeu, ble, bleu): Specify
2438 (beq2, bne2, bgt2, bgtu2, blt2, bltu2, bge2, bgeu2, ble2, bleu2): Name
2439 pattern. Specify attributes.
2440 (jump): Specify attributes.
2441 (tablejump_internal): Name pattern. Specify attributes.
2442 (call_value): Split into non_symbolic_call_value,
2443 symbolic_call_value_jsr, symbolic_call_value_bsr. Fix to use
2444 alternatives. Specify attributes.
2445 (non_symbolic_call_value, symbolic_call_value_jsr,
2446 symbolic_call_value_bsr): New patterns.
2447 (nop, return, unlink, indirect_jump): Specify attributes.
2448 (trap): Fix condition. Specify attributes.
2451 * config/m68k/m68k.c (m68k_symbolic_call_var): New variable.
2452 (override_options): Initialize it. Initialize m68k_sched_cpu.
2453 (CONST_METHOD): Rename to M68K_CONST_METHOD, move to m68k.h.
2454 (const_method): Make global, rename to m68k_const_method.
2455 (const_int_cost, output_move_const_into_data_reg): Update.
2456 (output_move_double): Parametrize to emit rtl code, rename to
2458 (output_reg_adjust, emit_reg_adjust, output_compadr, output_movsi,
2459 emit_movsi): New static functions.
2460 (output_move_double): New function with semantics of old
2462 (m68k_emit_move_double): New function.
2463 (m68k_sched_cpu): New variable.
2464 (attr_op_type): New enum.
2465 (sched_guess_p): New variable.
2466 (sched_address_type, sched_operand_type, sched_attr_op_type):
2467 New static functions.
2468 (m68k_sched_attr_opx_type, m68k_sched_attr_opy_type,
2469 m68k_sched_attr_size, m68k_sched_attr_op_mem): New functions.
2470 (sched_branch_type): New static variable.
2471 (m68k_sched_branch_type): New function.
2472 * config/m68k/m68k.h (M68K_SYMBOLIC_CALL): New enum.
2473 (m68k_symbolic_call_var): Declare.
2474 (M68K_CONST_METHOD): Rename from CONST_METHOD. Move here from m68k.c.
2475 (m68k_const_method, m68k_emit_move_double, m68k_sched_cpu,
2476 m68k_sched_attr_opx_type, m68k_sched_attr_opy_type,
2477 m68k_sched_attr_size, m68k_sched_attr_op_mem, m68k_sched_branch_type):
2480 2007-09-11 Jakub Jelinek <jakub@redhat.com>
2482 * builtins.def (BUILT_IN_VA_ARG_PACK_LEN): New builtin.
2483 * builtins.c (expand_builtin) <case BUILT_IN_VA_ARG_PACK_LEN>: Issue
2484 error if __builtin_va_arg_pack_len () wasn't optimized out during
2486 * tree-inline.c (copy_bb): Replace __builtin_va_arg_pack_len ()
2487 with the number of inline's anonymous arguments.
2488 * doc/extend.texi: Document __builtin_va_arg_pack_len ().
2490 2007-09-11 Zdenek Dvorak <ook@ucw.cz>
2492 * fold-const.c (extract_muldiv_1): Do not simplify
2495 2007-09-11 Jan Hubicka <jh@suse.cz>
2497 * i386.h (ix86_tune_indices): Add X86_TUNE_INTER_UNIT_CONVERSIONS.
2498 (TARGET_INTER_UNIT_CONVERSIONS): New.
2499 * i386.md (floatsi expanders): Remove redundant check for SImode
2500 source; offload to memory when asked for.
2501 (floatsisf2_mixed, floatsisf2_sse, floatsidf2_mixed, floatsidf2_sse
2502 floatdisf2_mixed, floatsisf2_sse, floatsidf2_mixed, floatsidf2_sse):
2504 (floatsisf2_mixed_memory, floatsisf2_sse_memory,
2505 floatsidf2_mixed_memory, floatsidf2_sse_memory
2506 floatdisf2_mixed_memory, floatsisf2_sse_memory,
2507 floatsidf2_mixed_memory, floatsidf2_sse_memory): New.
2509 2007-09-11 Jan Hubicka <jh@suse.cz>
2511 * toplev.c (process_options): all frontends now do unit-at-a-time.
2512 * cgraphunit.c: update comments.
2513 (cgraph_expand_function): call passmanager dirrectly; emit thunks.
2514 * c-decl.c (finish_function): use cgraph_add_new_function.
2515 * function.c (expand_function_end): We are always unit-at-a-time.
2517 2007-09-11 Richard Sandiford <richard@codesourcery.com>
2519 * config/mips/mips.c (mips_set_mips16_mode): Use separate anchor
2520 settings for MIPS16.
2521 (mips_use_anchors_for_symbol_p): Use default_use_anchors_for_symbol_p.
2523 2007-09-11 Richard Sandiford <richard@codesourcery.com>
2525 * config/mips/mips.c (mips_symbol_insns_1): Allow LEAs of
2526 SYMBOL_FORCE_TO_MEM constants.
2527 (mips_rtx_costs): Give a cost of 1 to force_to_mem_operands.
2528 (mips16_rewrite_pool_refs_info): New structure.
2529 (mips16_rewrite_pool_constant): New function, split out from...
2530 (mips16_rewrite_pool_refs): ...here. Take a pointer to a
2531 mips16_rewrite_pool_refs_info structure rather than a pointer
2532 to a constant pool. Force force_to_mem_operands into memory.
2533 (mips16_lay_out_constants): Update call to mips16_rewrite_pool_refs.
2534 * config/mips/predicates.md (force_to_mem_operand): New predicate.
2535 * config/mips/constraints.md (kf): New constraint.
2536 * config/mips/mips.md (*movdi_64bit_mips16): Add a d <- kf alternative.
2537 (*movsi_mips16): Likewise.
2539 2007-09-11 Richard Sandiford <richard@codesourcery.com>
2540 Nigel Stephens <nigel@mips.com>
2541 David Ung <davidu@mips.com>
2543 * config/mips/mips.h (CONSTANT_POOL_COST): Move to...
2544 * config/mips/mips.c: ...here and set to 4 for TARGET_MIPS16.
2545 (mips16_constant_cost, mips_immediate_operand_p, mips_binary_cost)
2546 (mips_fp_mult_cost, mips_fp_div_cost, mips_sign_extend_cost)
2547 (mips_zero_extend_cost): New functions.
2548 (mips_rtx_costs): Treat COMPARE constants as having zero cost.
2549 Use the new functions. Tweak many cost estimates, both here
2550 and in the new subroutines. Return false when the cost of the
2551 operands has not been calculated. Check for *clear_upper32.
2552 Check for floating-point multiply-add, reciprocal and rsqrt
2553 patterns. Handle comparison and rotation codes.
2555 2007-09-11 Danny Smith <dannysmith@users.sourceforge.net>
2557 * config/i386/cygming.h (TARGET_STRIP_NAME_ENCODING): Don't
2559 * config/i386/i386.c (get_dllimport_decl): Don't strip
2562 2007-09-10 Janis Johnson <janis187@us.ibm.com>
2565 * config/dfp-bit.c: Don't skip TFmode conversions; move strto*
2566 declarations to top.
2567 (DFP_TO_BFP): Use for either XFmode or TFmode.
2568 (BFP_TO_DFP): Use for either XFmode or TFmode; always use cast
2570 * config/dfp-bit.h: Include float.h.
2571 (LONG_DOUBLE_HAS_XF_MODE, LONG_DOUBLE_HAS_TF_MODE): Define if long
2572 double is one of these modes, rather than using LIBGCC_HAS_*F_MODE
2573 which doesn't mean the same thing.
2574 (BFP_KIND): Use 4 to mean TFmode.
2575 (BFP_FMT): Specify the number of decimal digits based on the
2576 number of mantissa digits.
2577 (BFP_VIA_TYPE): Binary float type to use as cast for sprintf.
2578 (BFP_TO_DFP, DFP_TO_BFP): Define names for TFmode variants.
2579 (STR_TO_BFP): Use strtold for XFmode or TFmode.
2580 (TFtype): Define if TFmode is supported.
2581 * doc/libgcc.texi (Decimal float library routines): Document
2582 TF conversion functions.
2584 2007-09-10 Chao-ying Fu <fu@mips.com>
2586 * config/mips/mips.c (mips_scalar_mode_supported_p): Declare.
2587 (TARGET_SCALAR_MODE_SUPPORTED_P): Define.
2588 (mips_emit_compare): Process fixed-point modes.
2589 (mips_pad_arg_upward): Support fixed-point types.
2590 (override_options): Allow fixed-point modes in accumulators.
2591 (mips_pass_by_reference): Pass DQ, UDQ, DA, and UDA modes in registers.
2592 (mips_vector_mode_supported_p): Support V2HQmode, V2UHQmode, V2HAmode,
2593 V2UHAmode, V4QQmode, and V4UQQmode when TARGET_DSP.
2594 (mips_scalar_mode_supported_p): New function to accept fixed-point
2595 modes if the width is not greater than two BITS_PER_WORD.
2596 * config/mips/mips.h (SHORT_FRACT_TYPE_SIZE, FRACT_TYPE_SIZE,
2597 LONG_FRACT_TYPE_SIZE, LONG_LONG_FRACT_TYPE_SIZE,
2598 SHORT_ACCUM_TYPE_SIZE, ACCUM_TYPE_SIZE, LONG_ACCUM_TYPE_SIZE,
2599 LONG_LONG_ACCUM_TYPE_SIZE): Define.
2600 * config/mips/mips.md ("d"): Update mode attribute for fixed-point
2602 ("IMODE"): New mode attribute.
2603 (mips-fixed.md): Include.
2604 * config/mips/mips-modes.def: Create VECTOR_MODES for FRACT, UFRACT,
2606 * config/mips/mips-fixed.md: New file.
2608 2007-09-11 Ben Elliston <bje@au.ibm.com>
2610 * config/spu/spu.md: Formatting fixes.
2612 2007-09-10 Janis Johnson <janis187@us.ibm.com>
2614 * config/dfp-bit.c (dfp_unary_func): Delete.
2615 (dfp_unary_op): Delete.
2616 (dfp_binary_op): Use decFloat functions instead of decNumber
2617 functions for binary operations.
2618 (d32_binary_op): Convert 32-bit operands to 64 bits for evaluation.
2619 (dnn_binary_op): Call dfp_binary_op with decFloat rather than
2621 (dfp_compare_op): Use decFloat functions insteadof decNumber
2622 functions for comparisons.
2623 (d32_compare_op): Convert 32-bit operands to 64 bits for evaluation.
2624 (dnn_binary_op): Call dfp_compare_op with decFloat rather than
2626 (DFP_ADD, DFP_SUB, DFP_MULTIPLE, DFP_DIVIDE): Use macros for
2627 call to dxx_binary_op and decFloat function.
2628 (DFP_EQ, DFP_NE, DFP_LT, DFP_GT, DFP_LE, DFP_GE): Use macros for
2629 calls to dxx_binary_op and decFloat function.
2630 * config/dfp-bit.h: Include decFloat header files.
2631 (decFloat, DFP_BINARY_OP, DFP_COMPARE_OP, DEC_FLOAT_ADD,
2632 DEC_FLOAT_SUBTRACT, DEC_FLOAT_MULTIPLY, DEC_FLOAT_DIVIDE,
2633 DEC_FLOAT_COMPARE, DEC_FLOAT_IS_ZERO, DEC_FLOAT_IS_NAN,
2634 DEC_FLOAT_IS_SIGNED: Define for each of 3 operand widths.
2636 2007-09-10 Harsha Jagasia <harsha.jagasia@amd.com>
2637 Jan Sjodin <jan.sjodin@amd.com>
2639 * tree-vect-analyze.c (vect_analyze_operations): Change
2640 comparison of loop iterations with threshold to less than
2641 or equal to instead of less than. Reduce
2642 min_scalar_loop_bound by one.
2643 * tree-vect-transform.c (vect_estimate_min_profitable_iters):
2644 Change prologue and epilogue iterations estimate to vf/2,
2645 when unknown at compile-time. Change versioning guard
2646 cost to taken_branch_cost. If peeling for alignment is
2647 unknown at compile-time, change peel guard costs to one
2648 taken branch and one not-taken branch per peeled loop.
2649 If peeling for alignment is known but number of scalar loop
2650 iterations is unknown at compile-time, change peel guard
2651 costs to one taken branch per peeled loop. Change the cost
2652 model equation to consider vector iterations as the loop
2653 iterations less the prologue and epilogue iterations.
2654 Change outside vector cost check to less than or equal to
2655 zero instead of equal to zero.
2656 (vect_do_peeling_for_loop_bound): Reduce
2657 min_scalar_loop_bound by one.
2658 * tree-vectorizer.h: Add TARG_COND_TAKEN_BRANCH_COST and
2659 TARG_COND_NOT_TAKEN_BRANCH_COST.
2660 * config/i386/i386.h (processor_costs): Add
2661 scalar_stmt_cost, scalar_load_cost, scalar_store_cost,
2662 vec_stmt_cost, vec_to_scalar_cost, scalar_to_vec_cost,
2663 vec_align_load_cost, vect_unalign_load_cost,
2664 vec_store_cost, cond_taken_branch_cost,
2665 cond_not_taken_branch_cost.
2666 Define macros for x86 costs.
2667 * config/i386/i386.c:
2668 (size_cost): Set scalar_stmt_cost, scalar_load_cost,
2669 scalar_store_cost, vec_stmt_cost, vec_to_scalar_cost,
2670 scalar_to_vec_cost, vec_align_load_cost,
2671 vect_unalign_load_cost, vec_store_cost,
2672 cond_taken_branch_cost, cond_not_taken_branch_cost to one.
2673 (i386_cost, i486_cost, pentium_cost, pentiumpro_cost,
2674 geode_cost, k6_cost, athlon_cost, pentium4_cost, nocona_cost,
2675 core2_cost, generic64_cost, generic32_cost): Set to default
2677 (k8_cost, amdfam10_cost): Costs for vectorization tuned.
2678 (x86_builtin_vectorization_cost): New.
2680 2007-09-10 Janis Johnson <janis187@us.ibm.com>
2681 Ben Elliston <bje@au.ibm.com>
2683 * dfp.c: Include decimal128Local.h;
2684 (dfp_byte_swap): Remove.
2685 (encode_decimal32, decode_decimal32): Don't handle endianness.
2686 (encode_decimal64, decode_decimal64): Ditto.
2687 (encode_decimal128, decode_decimal128): Ditto.
2688 * config/dfp-bit.c (host_to_ieee32, ieee_to_host_32): Ditto.
2690 (host_to_ieee_64, ieee_to_host_64): Don't handle endianness.
2692 (host_to_ieee_128, ieee_to_host_128): Don't handle endianness.
2693 * Makefile.in (DECNUM_H): Add decimal128Local.h.
2695 2007-09-10 David Daney <ddaney@avtrex.com>
2697 * config/mips/mips.md (UNSPEC_MEMORY_BARRIER): New entry in
2699 (memory_barrier): Rewrote as an insn that clobbers memory.
2701 2007-09-10 Richard Sandiford <richard@codesourcery.com>
2703 * config/mips/mips.c (mips_global_pointer): Check
2704 call_really_used_regs instead of call_used_regs.
2705 (mips_save_reg_p): Likewise. Save all call-saved registers
2706 if current_function_saves_all_registers. Fix indentation.
2707 No longer treat $18 as a special case.
2708 (compute_frame_size): Guard FPR loop with TARGET_HARD_FLOAT.
2710 2007-09-10 Richard Sandiford <richard@codesourcery.com>
2712 * config/mips/mips.h (MIPS_ARCH_FLOAT_SPEC): New macro.
2713 * config/mips/mips.c (mips_cpu_info_table): Mention it in the
2714 the introductory comment.
2715 (MIPS_MARCH_CONTROLS_SOFT_FLOAT): Delete.
2716 (override_options): Don't test for it.
2717 * config/mips/sde.h (MIPS_MARCH_CONTROLS_SOFT_FLOAT): Delete.
2718 (DRIVER_SELF_SPECS): Add MIPS_ARCH_FLOAT_SPEC.
2719 * config/mips/vr.h: As for config/mips/sde.h.
2721 2007-09-10 Trevor Smigiel <trevor_smigiel@playstation.sony.com>
2722 Revital Eres <eres@il.ibm.com>
2724 * target.h (struct gcc_target.sched): New field: sms_res_mii.
2725 (struct ddg): Define.
2726 * target-def.h (TARGET_SCHED_SMS_RES_MII): Define.
2727 (TARGET_SCHED): Add TARGET_SCHED_SMS_RES_MII.
2728 * config/spu/spu.c: Include ddg.h.
2729 (TARGET_SCHED_SMS_RES_MII): Define.
2730 (spu_sms_res_mii): New function to calculate mii.
2731 * modulo-sched (res_MII): Use it.
2732 * doc/tm.texi: Document TARGET_SCHED_SMS_RES_MII.
2734 2007-09-10 Andreas Krebbel <krebbel1@de.ibm.com>
2736 * config/s390/s390.c (s390_dump_pool): Create copy of constant
2737 pool entries since they might hold values that must not be shared.
2739 2007-09-10 Uros Bizjak <ubizjak@gmail.com>
2742 * gcc/config/i386/sse.md (ashr<mode>3): Change op2 mode to SImode.
2743 Use 'N' operand constraint for op2.
2744 (lshr<mode>3): Ditto.
2745 (ashl<mode>3): Ditto.
2746 (vec_shl_<mode>): Use const_0_to_255_mul_8_operand predicate for op2.
2747 (vec_shr_<mode>): Ditto.
2748 * gcc/config/i386/i386.c (ix86_expand_builtin) [IX86_BUILTIN_PSLL?128,
2749 IX86_BUILTIN_PSRA*?128, IX86_BUILTIN_PSRL?128]: Convert op1 to SImode.
2751 2007-09-10 Andreas Krebbel <krebbel1@de.ibm.com>
2753 * config/s390/s390.md ("fixuns_trunc<BFP:mode><GPR:mode>2"):
2754 Change mode macro in the last real_2expN parameter to uppercase.
2756 2007-09-10 Michael Matz <matz@suse.de>
2758 * tree-pass.h (pass_cselim): Declare new pass.
2759 * passes.c (init_optimization_passes): Link in pass_cselim.
2760 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Renamed from
2761 tree_ssa_phiopt; add do_store_elim parameter, handle it by calling
2762 cond_store_replacement.
2763 (condstoretemp): New static variable.
2764 (cond_store_replacement): New function.
2765 (tree_ssa_phiopt, tree_ssa_cs_elim): New wrappers around
2766 tree_ssa_phiopt_worker.
2767 (struct name_to_bb): New.
2768 (get_non_trapping, name_to_bb_hash, name_to_bb_eq, add_or_mark_expr,
2769 nt_init_block, nt_fini_block): New static functions.
2770 (seen_ssa_names, nontrap_set): New static variables.
2771 (gate_cselim, pass_cselim): Define new pass.
2772 * common.opt (ftree-cselim): New flag.
2773 * toplev.c (process_options): Set flag_tree_cselim if required.
2775 2007-09-10 Hans-Peter Nilsson <hp@axis.com>
2777 * simplify-rtx.c (simplify_relational_operation_1): For recent
2778 canonicalization, don't recurse if op1 equals both PLUS arguments.
2780 2007-09-09 David Daney <ddaney@avtrex.com>
2782 * optabs.c (expand_sync_operation): Use plus insn if minus
2784 (expand_sync_fetch_operation): Ditto.
2786 2007-09-09 H.J. Lu <hongjiu.lu@intel.com>
2788 * i386.md (*floatsisf2_mixed_vector): Use cvtdq2ps instead
2790 (*floatsisf2_sse_vector): Likewise.
2792 2007-09-09 Krister Walfridsson <cato@df.lth.se>
2794 * config/netbsd.h (HANDLE_PRAGMA_PACK_PUSH_POP): Define to 1.
2796 2007-09-09 H.J. Lu <hongjiu.lu@intel.com>
2798 * config/i386/i386.h (ix86_tune_indices): Rename
2799 X86_USE_VECTOR_CONVERTS to X86_TUNE_USE_VECTOR_CONVERTS.
2800 (TARGET_USE_VECTOR_CONVERTS): Updated.
2801 * config/i386/i386.c: Likewise.
2803 2007-09-09 Sandra Loosemore <sandra@codesourcery.com>
2804 Nigel Stephens <nigel@mips.com>
2806 * doc/invoke.texi (Overall Options): Add .sx file extension
2807 as a synonym for .S.
2808 * cppspec.c (known_suffixes): Likewise.
2809 * gcc.c (default_compilers): Likewise.
2811 2007-09-09 Rask Ingemann Lambertsen <rask@sygehus.dk>
2814 * config/i386/i386.h (CANONICALIZE_COMPARISON): Delete.
2815 * simplify-rtx.c (simplify_relational_operation_1): Add the
2816 canonicalization from i386.h.
2817 * doc/md.texi (Canonicalization of Instructions): Document it.
2819 2007-09-09 Jan Hubicka <jh@suse.cz>
2820 Dwarakanath Rajagopal <dwarak.rajagopal@amd.com>
2822 * i386.h (ix86_tune_indices): Add X86_USE_VECTOR_CONVERTS.
2823 (TARGET_USE_VECTOR_CONVERTS): New.
2824 * i386.md: New post-reload splitters for converting SF to DF and DF to
2826 (floatsi* expander): Special case vector conversions.
2827 (floatsisf2_mixed_vector, floatsisf2_sse_vector_nointernunit,
2828 floatsisf2_sse_vector_internunit, floatsisf2_sse_vector,
2829 floatsidf2_mixed_vector, floatsidf2_sse_vector): New.
2830 (floatsisf2_mixed, floatsisf2_sse, floatsidf2_mixed, floatsidf2_sse):
2831 Disable when doing vector converts.
2832 (floatsi<mode>2_i387): Disable when
2833 * sse.md (vec_dupv2df): Export.
2834 * i386.c (ix86_tune_features): Enable SSE conversions.
2836 2007-09-09 Richard Guenther <rguenther@suse.de>
2838 * tree-ssa-operands.c (add_virtual_operand): Only mark
2839 stores as has_volatile_ops if alias information is not available.
2841 2007-09-09 Revital Eres <eres@il.ibm.com>
2843 * doc/contrib.texi: Add myself.
2845 2007-09-09 Ira Rosen <irar@il.ibm.com>
2847 * tree-vectorizer.h (stmt_vinfo_set_inside_of_loop_cost,
2848 stmt_vinfo_set_outside_of_loop_cost): New functions.
2849 * tree-vect-transform.c (vect_get_cost_fields): Remove.
2850 (vect_model_simple_cost): Call
2851 stmt_vinfo_set_inside/outside_of_loop_cost to set the relevant cost
2852 field instead of calling vect_get_cost_fields.
2853 (vect_model_store_cost, vect_model_load_cost): Likewise.
2855 2007-09-09 Revital Eres <eres@il.ibm.com>
2857 * config/rs6000/rs6000.c (paired_init_builtins): Add const
2858 declaration to bdesc_paired_preds variable.
2859 (paired_expand_builtin): Likewise.
2861 2007-09-09 Revital Eres <eres@il.ibm.com>
2863 * dbgcnt.def (sms_sched_loop): New counter.
2864 * modulo-sched.c: Use sms_sched_loop instead of
2865 MAX_SMS_LOOP_NUMBER to determine the maximum number of loops to
2866 perform swing modulo scheduling on. Include dbgcnt.h.
2867 * Makefile.in: Add DBGCNT_H to modulo-sched.o.
2868 * params.def: Remove PARAM_MAX_SMS_LOOP_NUMBER.
2870 2007-09-09 Uros Bizjak <ubizjak@gmail.com>
2872 * config/i386/i386.md (X87MODEF12, SSEMODEF): Remove mode iterators.
2873 Substitute all uses with ...
2874 (MODEF): New mode iterator.
2876 (fix_trunc<mode>_fisttp_i387_1): Remove operand constraints
2877 from pre-regalloc define_insn_and_split splitter pattern.
2878 (*fix_trunc<mode>_i387_1): Ditto.
2879 (*fistdi2_1): Ditto.
2880 (*fist<mode>2_1): Ditto.
2881 (frndintxf2_floor): Ditto.
2882 (*fist<mode>2_floor_1): Ditto.
2883 (frndintxf2_ceil): Ditto.
2884 (*fist<mode>2_ceil_1): Ditto.
2885 (frndintxf2_trunc): Ditto.
2886 (frndintxf2_mask_pm): Ditto.
2888 (prologue): Use (const_int 0) as never generated filler insn.
2890 (sibcall_epilogue): Ditto.
2891 (eh_return_si): Ditto.
2892 (eh_return_di): Ditto.
2894 (add<mode>3): Rename from adddf3 and addsf3. Macroize expander
2895 using MODEF mode iterator.
2896 (sub<mode>3): Rename from subdf3 and subsf3. Macroize expander
2897 using MODEF mode iterator.
2898 (mul<mode>3): Rename from muldf3 and mulsf3. Macroize expander
2899 using MODEF mode iterator.
2900 (nearbyint<mode>2): Rename from nearbyintdf2 and nearbyintsf2.
2901 Macroize expander using MODEF mode iterator.
2903 (zero_extendsidi2): Remove operand constraints from expander.
2904 (smuldi3_highpart): Ditto.
2905 (indirect_jump): Ditto.
2908 * config/i386/sse.md (storentv4sf): Ditto.
2909 (storentv2df): Ditto.
2910 (storentv2di): Ditto.
2912 (sse2_cvtpd2ps): Ditto.
2913 (vec_interleave_highv16qi): Ditto.
2914 (vec_interleave_lowv16qi): Ditto.
2915 (vec_interleave_highv8hi): Ditto.
2916 (vec_interleave_lowv8hi): Ditto.
2917 (vec_interleave_highv4si): Ditto.
2918 (vec_interleave_lowv4si): Ditto.
2919 (vec_interleave_highv2di): Ditto.
2920 (vec_interleave_lowv2di): Ditto.
2921 (sse2_maskmovdqu): Ditto.
2922 * config/i386/mmx.md (mmx_maskmovq): Ditto.
2924 2007-09-09 Ira Rosen <irar@il.ibm.com>
2926 * tree-vectorizer.h (enum vect_def_type): Start enumeration from 1.
2927 (struct _slp_tree, struct _slp_instance): Define new data structures
2928 along macros for their access.
2929 (struct _loop_vec_info): Define new fields: strided_stores,
2930 slp_instances, and slp_unrolling_factor along macros for their access.
2931 (enum slp_vect_type): New.
2932 (struct _stmt_vec_info): Define new field, slp_type, and macros for
2934 (STMT_VINFO_STRIDED_ACCESS): New macro.
2935 (vect_free_slp_tree): Declare.
2936 (vectorizable_load): Add an argument of type slp_tree.
2937 (vectorizable_store, vectorizable_operation, vectorizable_conversion,
2938 vectorizable_assignment): Likewise.
2939 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2940 Declare (make extern).
2941 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
2942 (new_loop_vec_info): Likewise.
2943 (destroy_loop_vec_info): Free memory allocated for SLP structures.
2944 * tree-vect-analyze.c: Include recog.h.
2945 (vect_update_slp_costs_according_to_vf): New.
2946 (vect_analyze_operations): Add argument for calls to vectorizable_ ()
2947 functions. For not pure SLP stmts with strided access check that the
2948 group size is power of 2. Update the vectorization factor according to
2949 SLP. Call vect_update_slp_costs_according_to_vf.
2950 (vect_analyze_group_access): New.
2951 (vect_analyze_data_ref_access): Call vect_analyze_group_access.
2952 (vect_free_slp_tree): New functions.
2953 (vect_get_and_check_slp_defs, vect_build_slp_tree, vect_print_slp_tree,
2954 vect_mark_slp_stmts, vect_analyze_slp_instance, vect_analyze_slp,
2955 vect_make_slp_decision, vect_detect_hybrid_slp_stmts,
2956 vect_detect_hybrid_slp): Likewise.
2957 (vect_analyze_loop): Call vect_analyze_slp, vect_make_slp_decision
2958 and vect_detect_hybrid_slp.
2959 * tree-vect-transform.c (vect_estimate_min_profitable_iters): Take
2960 SLP costs into account.
2961 (vect_get_cost_fields): New function.
2962 (vect_model_simple_cost): Make extern, add SLP parameter and handle
2964 (vect_model_store_cost, vect_model_load_cost): Likewise.
2965 (vect_get_constant_vectors): New function.
2966 (vect_get_slp_vect_defs, vect_get_slp_defs,
2967 vect_get_vec_defs_for_stmt_copy, vect_get_vec_defs_for_stmt_copy,
2968 vect_get_vec_defs): Likewise.
2969 (vectorizable_reduction): Don't handle SLP for now.
2970 (vectorizable_call): Don't handle SLP for now. Add argument to
2971 vect_model_simple_cost.
2972 (vectorizable_conversion): Handle SLP (call vect_get_vec_defs to
2973 get SLPed and vectorized defs). Fix indentation and spacing.
2974 (vectorizable_assignment): Handle SLP.
2975 (vectorizable_induction): Don't handle SLP for now.
2976 (vectorizable_operation): Likewise.
2977 (vectorizable_type_demotion): Add argument to
2978 vect_model_simple_cost.
2979 (vectorizable_type_promotion): Likewise.
2980 (vectorizable_store, vectorizable_load): Handle SLP.
2981 (vectorizable_condition): Don't handle SLP for now.
2982 (vect_transform_stmt): Add a new argument for SLP. Check that there is
2983 no SLP transformation required for unsupported cases. Add SLP
2984 argument for supported cases.
2985 (vect_remove_stores): New function.
2986 (vect_schedule_slp_instance, vect_schedule_slp): Likewise.
2987 (vect_transform_loop): Schedule SLP instances.
2988 * Makefile.in: (tree-vect-analyze.o): Depend on recog.h.
2990 2007-09-09 Andrew Haley <aph@redhat.com>
2992 * optabs.c (sign_expand_binop): Set libcall_gen = NULL in the fake
2995 2007-09-09 Hans-Peter Nilsson <hp@axis.com>
2997 Divide REG_LABEL notes into REG_LABEL_OPERAND and REG_LABEL_TARGET.
2998 * doc/rtl.texi (Insns): Specify when a label_ref makes a jump_insn.
2999 Similar for what label_refs can go in the JUMP_TARGET field. Split
3000 REG_LABEL documentation into REG_LABEL_TARGET and REG_LABEL_OPERAND.
3001 * reload.c (find_reloads): Generate REG_LABEL_OPERAND, not
3002 REG_LABEL when replacing an operand with a LABEL_REF for a
3004 (subst_reloads): When replacing a LABEL_REG with a register,
3005 instead of generating a REG_LABEL note, assert that there already
3006 is one or that the label is a known target for the insn.
3007 * rtlanal.c (computed_jump_p): Instead of looking for a REG_LABEL
3008 note, check the JUMP_LABEL field. Remove "else" after return.
3009 * reorg.c (emit_delay_sequence): Replace case for REG_LABEL with
3010 cases for REG_LABEL_OPERAND and REG_LABEL_TARGET.
3011 (fill_slots_from_thread): Handle both REG_LABEL_OPERAND and
3012 REG_LABEL_TARGET notes, including the JUMP_TARGET field on JUMP_P
3013 insns. Iterate over all notes; don't assume there's only one.
3014 * cse.c (recorded_label_ref): Adjust comment to refer to
3016 (cse_extended_basic_block): Do LABEL_REF check for all INSN_P
3017 insns, not just NONJUMP_INSN_P.
3018 (check_for_label_ref): For JUMP_P insns, check that the LABEL_REF
3019 isn't a jump target.
3020 * jump.c (rebuild_jump_labels): Adjust head comment.
3021 (init_label_info): Ditto. Remove REG_LABEL_OPERAND notes only;
3022 don't reset REG_LABEL_TARGET notes, including the JUMP_LABEL field.
3023 (mark_all_labels): For JUMP_P insns without a target, check if the
3024 the target is noted on the previous nonjump insn.
3025 (mark_jump_label_1): New function, guts from mark_jump_label.
3026 <case IF_THEN_ELSE>: Handle first operand as a non-target when
3027 marking jump target labels.
3028 <case LABEL_REF>: Adjust for whether to generate a
3029 REG_LABEL_TARGET or a REG_LABEL_OPERAND note.
3030 For 'E' format rtl, iterate in descending element order.
3031 (delete_related_insns): Handle both REG_LABEL_TARGET and
3032 REG_LABEL_OPERAND notes. For JUMP_P insns with labels with zero
3033 reference count, delete and fallthrough. Move finding-next-
3034 non-deleted insn last in the function. Look at all INSN_P insns
3035 for REG_LABEL_OPERAND notes.
3036 (redirect_jump_2): Assert that OLABEL equals the old JUMP_LABEL of
3038 * print-rtl.c (print_rtx): For JUMP_P insns and a non-NULL
3039 JUMP_LABEL, output the INSN_UID of it.
3040 * gcse.c: Adjust comments as appropriate to say REG_LABEL_OPERAND
3041 and/or REG_LABEL_TARGET.
3042 (add_label_notes): Only add REG_LABEL_OPERAND notes. Put in line
3043 with jump.c copy by only adding notes for labels actually
3044 referenced in the insn.
3045 * emit-rtl.c (try_split): Don't assume only NONJUMP_INSN_P need
3046 usage count increment; handle all INSN_P trial insns.
3047 (emit_copy_of_insn_after): Change to not copy REG_LABEL_OPERAND
3049 * rtl.h (struct rtx_def) <volatil>: Adjust to mention
3050 REG_LABEL_TARGET and REG_LABEL_OPERAND.
3051 (LABEL_REF_NONLOCAL_P): Allow REG_LABEL_TARGET and
3053 * combine.c (distribute_notes): Adjust for REG_LABEL_TARGET on
3054 JUMP_P insns and REG_LABEL_OPERAND everywhere.
3055 * sched-rgn.c (is_cfg_nonregular): Check for REG_LABEL_OPERANDS
3056 on all INSN_P insns.
3057 * reg-notes.def (LABEL_TARGET, LABEL_OPERAND): Split from LABEL.
3058 * cfgrtl.c (delete_insn): Adjust to handle REG_LABEL_TARGET and
3059 REG_LABEL_OPERAND notes.
3060 * reload1.c (calculate_needs_all_insns): Adjust comments.
3061 (set_label_offsets): Adjust to look for REG_LABEL_OPERAND notes.
3062 * config/alpha/alpha.md (split for load of an address into a
3063 four-insn sequence on Unicos/Mk): Adjust to use
3065 * config/sh/sh.md (sh_reorg, final_prescan_insn): Ditto.
3067 2007-09-09 Laurynas Biveinis <laurynas.biveinis@gmail.com>
3070 2007-09-05 Laurynas Biveinis <laurynas.biveinis@gmail.com>
3071 * regrename.c (copyprop_hardreg_forward_1): New variable next. Use
3072 FOR_BB_INSNS_SAFE instead of for loop.
3073 * cse.c (cse_extended_basic_block): Likewise.
3074 * postreload.c (reload_cse_regs_1): New variable next. Make sure
3075 that the for loop does not invoke NEXT_INSN on a deleted insn.
3076 * function.c (instantiate_virtual_regs): Likewise.
3077 * lower-subreg.c (remove_retval_note): Likewise.
3078 (decompose_multiword_subregs): Use FOR_BB_INSNS_SAFE instead of
3080 * emit-rtl.c (remove_insn): Set NEXT_INSN and PREV_INSN to NULL on
3082 * cfgrtl.c (delete_insn): Set JUMP_LABEL to NULL on a deleted
3083 insn, if it's a jump.
3084 (try_redirect_by_replacing_jump): New variable jump_p. Call
3085 tablejump_p before delete_insn_chain.
3086 * reload1.c (reload): New variable next. Make sure that the for
3087 loop does not invoke NEXT_INSN on a deleted insn.
3088 (fixup_eh_region_note): Make the loop terminate if i becomes NULL.
3089 (delete_output_reload): New variable prev. Make sure the the for
3090 loops do not invoke PREV_INSN on a deleted insn.
3092 2007-09-08 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
3094 * pa/constraints.md: Remove 'X' from unused letters comment.
3096 2007-09-08 Richard Guenther <rguenther@suse.de>
3098 * tree-tailcall.c (find_tail_calls): If we don't have aliases
3099 computed check stmt_ann->references_memory instead of counting
3102 2007-09-08 Segher Boessenkool <segher@kernel.crashing.org>
3104 * cse.c (fold_rtx): Use validate_unshare_change() instead of
3105 validate_change() in one more case.
3107 2007-09-08 Zdenek Dvorak <ook@ucw.cz>
3109 PR tree-optimization/32283
3110 * tree-ssa-loop-ivopts.c (may_eliminate_iv): Use
3111 estimated_loop_iterations.
3112 (determine_use_iv_cost_condition): Decrease cost of expressions
3113 used in iv elimination.
3115 2007-09-08 Richard Guenther <rguenther@suse.de>
3117 * tree-cfg.c (verify_gimple_expr): Avoid building new
3118 pointer types, use TYPE_POINTER_TO if available instead.
3120 2007-09-08 Uros Bizjak <ubizjak@gmail.com>
3123 PR rtl-optimization/26449
3124 * config/i386/sse.md (mulv4si3): Do not expand sse2 sequence.
3125 (*sse2_mulv4si3): New define_insn_and_split pattern. Split insn in
3127 (mulv16qi3): Implement as define_insn_and_split pattern instead of
3128 define_expand. Split insn in split1 pass.
3131 2007-09-08 Dorit Nuzman <dorit@il.ibm.com>
3133 PR tree-optimization/33301
3134 * tree-vect-analyze (analyze_operations): Look at the type of the rhs
3137 2007-09-07 Zdenek Dvorak <ook@ucw.cz>
3139 PR tree-optimization/32183
3140 * Makefile.in (tree-ssa-reassoc.o): Also depend on $(CFGLOOP_H).
3142 * tree-ssa-reassoc.c: Include cfgloop.h.
3143 (is_reassociable_op): Add a loop argument and return true only
3145 (linearize_expr): Updated.
3146 (should_break_up_subtract): Likewise.
3147 (linearize_expr_tree): Likewise.
3148 (init_reassoc): Call loop_optimizer_init with
3149 AVOID_CFG_MODIFICATIONS. Remove calculate_dominance_info call
3150 with CDI_DOMINATORS.
3151 (fini_reassoc): Call loop_optimizer_finalize.
3153 2007-09-07 Sterling Augustine <sterling@tensilica.com>
3155 * config/xtensa/lib2funcs.S (__xtensa_sync_caches): Use an ISYNC even
3156 if there is no i-cache.
3158 2007-09-07 Richard Guenther <rguenther@suse.de>
3161 2007-09-06 Richard Guenther <rguenther@suse.de>
3163 PR tree-optimization/32586
3164 * tree-ssa-sccvn.c (simplify_binary_expression): Avoid
3165 folding if nothing changed.
3166 (simplify_unary_expression): New function. Do tree combining
3167 on conversion like codes.
3168 (try_to_simplify): Call it.
3170 2007-09-07 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
3173 * gthr-posix.h (__gthread_active_p): Add implementation for hppa-hpux.
3174 (__gthread_active,__gthread_start, __gthread_active_init): New.
3175 * gthr-posix95.h: Likewise.
3177 2007-09-07 Roman Zippel <zippel@linux-m68k.org>
3179 * function.h (struct function): Rename calls_unwind_init
3180 to saves_all_registers.
3181 (current_function_saves_all_registers): Renamed from
3182 current_function_calls_unwind_init.
3183 * reload1.c (reload): Set current_function_saves_all_registers.
3184 * except.c (expand_builtin_unwind_init): Likewise.
3185 * config/m68k/m68k.c (m68k_save_reg): Use
3186 current_function_saves_all_registers to save pic register.
3188 2007-09-07 Janis Johnson <janis187@us.ibm.com>
3190 config/m68k/m68k.c (floating_exact_log2): Update call to real_2expN.
3191 config/s390/s390.md (fixuns_trunc<BFP:mode><GPR:mode>2): Ditto.
3193 Reapply reverted changes:
3195 2007-09-06 Jan Hubicka <jh@suse.cz>
3197 * config/i386.c (ix86_expand_lround, ix86_expand_round): Update call
3200 2007-09-06 Richard Sandiford <richard@codesourcery.com>
3202 * config/mips/mips.md (fixuns_truncdfsi2, fixuns_truncdfdi2)
3203 (fixuns_truncsfsi2, fixuns_truncsfdi2): Update calls to real_2expN.
3205 2007-09-05 Janis Johnson <janis187@us.ibm.com>
3207 * optabs.c (expand_float): Convert unsigned integer as signed only
3208 if it provides sufficient accuracy; add mode argument to real_2expN.
3209 (expand_fix): Fix comment typos; extend binary float into mode
3210 wider than destination for converion to unsigned integer; add mode
3211 argument to real_2expN.
3212 * real.c (real_2expN): Add mode argument to special-case decimal
3214 * real.h (real_2expN): Ditto.
3215 * fixed-value.c (check_real_for_fixed_mode): Add mode argument to
3217 (fixed_from_string): Ditto.
3218 (fixed_to_decimal): Ditto.
3219 (fixed_convert_from_real): Ditto.
3220 (real_convert_from_fixed): Ditto.
3221 * config/rs6000/rs6000.md (FP): Include DD and TD modes.
3222 * config/rs6000/dfp.md (extendddtd2, adddd3, addtd3, subdd3, subtd3,
3223 muldd3, multd3, divdd3, divtd3, cmpdd_internal1, cmptd_internal1,
3224 floatditd2, ftruncdd2, fixdddi2, ftrunctd2, fixddi2): New.
3226 2007-09-07 Diego Novillo <dnovillo@google.com>
3228 * tree-flow.h (const_block_stmt_iterator): Remove.
3229 Update all users to use block_stmt_iterator.
3230 * tree-iterator.h (const_tree_stmt_iterator): Remove.
3231 Update all users to use tree_stmt_iterator.
3233 2007-09-07 Sandra Loosemore <sandra@codesourcery.com>
3235 * config/mips/mips.c (mips_set_current_function): Temporarily
3236 make this a no-op to fix bootstrap errors, pending rewrite.
3238 2007-09-07 Jan Hubicka <jh@suse.cz>
3240 * reorg.c (dbr_schedule): Move code removing placeholder USEs later
3243 2007-09-07 Dorit Nuzman <dorit@il.ibm.com>
3245 PR tree-optimization/33299
3246 * tree-vect-transform.c (vect_create_epilog_for_reduction): Update
3247 uses for all relevant loop-exit phis, not just the first.
3249 2007-09-07 Richard Guenther <rguenther@suse.de>
3252 * tree-ssa-operands.c (access_can_touch_variable): An access
3253 of the form (*p)[0] can touch a variable of same size.
3255 2007-09-07 Jan Hubicka <jh@suse.cz>
3257 * passes.c (init_optimization_passes): Add simple dce and addressable
3259 * tree-ssa.c (execute_update_addresses_taken): New function.
3260 (pass_update_address_taken): New.
3261 * tree-ssa-dse.c (execute_simple_dse): New function.
3262 (pass_simple_dse): New.
3263 * tree-pass.h (pass_simple_dse, pass_update_address_taken): Declare.
3265 2007-09-07 Tobias Burnus <burnus@net-b.de>
3268 * doc/invoke.texi: Fix -Wstrict-overflow= table.
3270 2007-09-07 Richard Guenther <rguenther@suse.de>
3273 2007-09-06 Richard Guenther <rguenther@suse.de>
3275 PR tree-optimization/32586
3276 * tree-ssa-sccvn.c (simplify_binary_expression): Avoid
3277 folding if nothing changed.
3278 (simplify_unary_expression): New function. Do tree combining
3279 on conversion like codes.
3280 (try_to_simplify): Call it.
3282 2007-09-07 Richard Guenther <rguenther@suse.de>
3283 Uros Bizjak <ubizjak@gmail.com>
3285 PR tree-optimization/32821
3286 * tree_if_conv.c (combine_blocks): Use alloc_stmt_list instead of
3287 NULL in the call to set_bb_stmt_list.
3289 2007-09-07 Richard Sandiford <richard@codesourcery.com>
3291 * config/mips/mips.c (build_mips16_call_stub): Emit all direct
3292 float calls here, rather than leaving some to the caller.
3293 Use call_internal_direct and call_value_internal_direct.
3294 * config/mips/mips.md (call_internal_direct): New pattern.
3295 (call_value_internal_direct): Likewise.
3297 2007-09-07 Richard Sandiford <richard@codesourcery.com>
3299 * config/mips/mips.c (mips_base_move_loop_invariants): New variable.
3300 (mips_set_mips16_mode): Restore flag_move_loop_invariants, then set
3302 (override_options): Set mips_base_move_loop_invariants.
3304 2007-09-07 Richard Sandiford <richard@codesourcery.com>
3306 * config/mips/mips.opt (mcode-readable=): Move to keep list
3309 2007-09-07 Richard Sandiford <richard@codesourcery.com>
3311 * doc/invoke.texi: Document the MIPS -mlocal-sdata, -mextern-sdata
3312 and -mgpopt options. Adjust the -G documentation to match.
3313 * config/mips/mips.opt (mextern-data, mgpopt, mlocal-data): New.
3314 * config/mips/mips.c (mips_rtx_constant_in_small_data_p): New
3315 function, split out from mips_classify_symbol. Return false for
3316 !TARGET_LOCAL_SDATA.
3317 (mips_classify_symbol): Call mips_rtx_constant_in_small_data_p.
3318 Only use GP-relative accesses for non-pool symbols if TARGET_GPOPT.
3319 (override_options): Check whether the -mgpopt setting is consistent
3320 with the other flags.
3321 (symbolic_expression_p): Delete.
3322 (mips_select_rtx_section): Use mips_rtx_constant_in_small_data_p
3323 and default_elf_select_rtx_section.
3324 (mips_in_small_data_p): Honor section attributes for MIPS16 too.
3325 Return false for local data unless TARGET_LOCAL_SDATA. Likewise
3326 external data and TARGET_EXTERN_SDATA.
3328 2007-09-07 Tobias Burnus <burnus@net-b.de>
3331 * doc/cpp.texi (Common Predefined Macros): Add __GFORTRAN__.
3333 2007-09-05 Laurynas Biveinis <laurynas.biveinis@gmail.com>
3335 * regrename.c (copyprop_hardreg_forward_1): New variable next. Use
3336 FOR_BB_INSNS_SAFE instead of for loop.
3337 * cse.c (cse_extended_basic_block): Likewise.
3338 * postreload.c (reload_cse_regs_1): New variable next. Make sure
3339 that the for loop does not invoke NEXT_INSN on a deleted insn.
3340 * function.c (instantiate_virtual_regs): Likewise.
3341 * lower-subreg.c (remove_retval_note): Likewise.
3342 (decompose_multiword_subregs): Use FOR_BB_INSNS_SAFE instead of
3344 * emit-rtl.c (remove_insn): Set NEXT_INSN and PREV_INSN to NULL on
3346 * cfgrtl.c (delete_insn): Set JUMP_LABEL to NULL on a deleted
3347 insn, if it's a jump.
3348 (try_redirect_by_replacing_jump): New variable jump_p. Call
3349 tablejump_p before delete_insn_chain.
3350 * reload1.c (reload): New variable next. Make sure that the for
3351 loop does not invoke NEXT_INSN on a deleted insn.
3352 (fixup_eh_region_note): Make the loop terminate if i becomes NULL.
3353 (delete_output_reload): New variable prev. Make sure the the for
3354 loops do not invoke PREV_INSN on a deleted insn.
3356 2007-09-06 Zdenek Dvorak <ook@ucw.cz>
3358 * cgraphbuild.c (rebuild_cgraph_edges): Export.
3359 * cgraph.h (rebuild_cgraph_edges): Declare.
3360 * tree-pass.h (pass_expand_omp_ssa): New.
3361 * omp-low.c (find_omp_clause): Export.
3362 (copy_var_decl): Split from omp_copy_decl_2.
3363 (build_omp_barrier): Return the call to emit instead of emitting
3365 (lower_rec_input_clauses, expand_omp_single): Gimplify the result of
3367 (extract_omp_for_data, expand_parallel_call, expand_omp_parallel,
3368 expand_omp_for_generic, expand_omp_for_static_nochunk,
3369 expand_omp_for_static_chunk, expand_omp_for, expand_omp_sections):
3370 Adapted to work on SSA form.
3371 (execute_expand_omp): Do not invalidate dominance information.
3372 (gate_expand_omp): Do not run with -fopenmp-ssa flag.
3373 (gate_expand_omp_ssa, pass_expand_omp_ssa): New.
3374 * gimplify.c (gimplify_omp_for): Ensure that the control variable is
3376 (force_gimple_operand): Allow gimplifying code expressions without
3378 * tree-predcom.c (mark_virtual_ops_for_renaming): Handle phi nodes.
3379 * common.opt (fopenmp-ssa): New.
3380 * tree-flow.h (find_omp_clause, copy_var_decl): Declare.
3381 * Makefile.in (tree-cfg.o): Add TREE_INLINE_H dependency.
3382 * tree-cfg.c: Include tree-inline.h.
3383 (struct move_stmt_d): Replace vars_to_remove by vars_map field.
3384 (replace_by_duplicate_decl, replace_ssa_name,
3385 mark_virtual_ops_in_region): New functions.
3386 (move_stmt_r, move_block_to_fn, move_sese_region_to_fn): Adapted
3387 to work on SSA form.
3388 * passes.c (init_optimization_passes): Add pass_expand_omp_ssa pass.
3389 * tree-ssa-operands.c (get_expr_operands): Handle operands of OMP
3392 2007-09-06 Laurynas Biveinis <laurynas.biveinis@gmail.com>
3394 * tree-loop-linear.c: Include obstack.h.
3395 (linear_transform_loops): New obstack lambda_obstack.
3396 Initialize it, pass it to gcc_loopnest_to_lambda_loopnest,
3397 lambda_loopnest_transform, lambda_loopnest_to_gcc_loopnest calls
3398 and free afterwards.
3399 * lambda.h (struct obstack): New forward declaration.<