1 2007-10-03 Alexandre Oliva <aoliva@redhat.com>
3 * gcse.c (hash_scan_set): Insert set in insn before note at
4 the end of basic block.
6 2007-10-03 Sebastian Pop <sebastian.pop@amd.com>
8 PR tree-optimization/33576
9 * testsuite/gcc.dg/tree-ssa/pr33576.c: New.
10 * tree-loop-linear.c (linear_transform_loops): Call remove_iv.
11 * lambda.h (lambda_loopnest_to_gcc_loopnest): New parameter.
12 (remove_iv): Declared.
13 * lambda-code.c (remove_iv): Not static.
14 (lambda_loopnest_to_gcc_loopnest): New parameter remove_ivs.
15 Don't remove ivs there, save ivs in the buffer.
17 2007-10-03 Jason Merrill <jason@redhat.com>
20 * tree-eh.c (same_handler_p): New fn.
21 (optimize_double_finally): New fn.
22 (refactor_eh_r): New fn.
23 (refactor_eh): New fn.
24 (pass_refactor_eh): New pass.
25 * tree-pass.h: Declare it.
26 * passes.c (init_optimization_passes): Add it.
28 2007-10-03 Doug Kwan <dougkwan@google.com>
29 Richard Guenther <rguenther@suse.de>
32 * dwarf2out.c (reference_to_unused): Disable sanity checking,
33 be conservative instead.
35 2007-10-03 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
38 * doc/invoke.texi (-fwhole-program): Document that Fortran
39 doesn't support this option.
41 2007-10-02 Richard Sandiford <rsandifo@nildram.co.uk>
44 * expr.c (expand_expr_addr_expr_1): Pass CONSTRUCTORs to
47 2007-10-02 David Daney <ddaney@avtrex.com>
49 * config/mips/mips.md (sync_compare_and_swap<mode>): Handle compare
50 against constant zero.
51 * config/mips/mips.h (MIPS_COMPARE_AND_SWAP): Handle constant zero
54 2007-09-02 Kenneth Zadeck <zadeck@naturalbridge.com>
56 * ra-conflict.c: New file.
58 * reload.c (push_reload, find_dummy_reload): Change DF_RA_LIVE
59 usage to DF_LIVE usage.
60 * rtlanal.c (subreg_nregs_with_regno): New function.
61 * df-scan.c (df_def_record_1, df_uses_record): Add code to set
62 DF_REF_EXTRACT, DF_REF_STRICT_LOWER_PART, and DF_REF_SUBREG flags.
63 (df_has_eh_preds): Removed.
64 (df_bb_refs_collect, df_bb_refs_collect, df_bb_refs_collect,
65 df_exit_block_uses_collect): Changed call from df_has_eh_preds to
67 * global.c (allocno, max_allocno, conflicts, allocno_row_words,
68 reg_allocno, EXECUTE_IF_SET_IN_ALLOCNO_SET): Moved to ra.h
69 (SET_ALLOCNO_LIVE, CLEAR_ALLOCNO_LIVE): Moved to ra-conflicts.c.
70 (regs_set, record_one_conflict, record_conflicts, mark_reg_store,
71 mark_reg_clobber, mark_reg_conflicts, mark_reg_death): Deleted.
72 (global_alloc): Turn off rescanning insns after call to
73 global_conflicts and added call to set_preferences.
74 (global_conflicts): Moved to ra-alloc.c.
75 (set_preferences_1, set_preferences): New function.
76 (mirror_conflicts): Changed types for various variables.
77 (mark_elimination): Change DF_RA_LIVE
78 usage to DF_LIVE usage.
79 (build_insn_chain): Rewritten from scratch and made local.
80 (print_insn_chain, print_insn_chains): New functions.
81 (dump_conflicts): Do not print conflicts for fixed_regs.
82 (rest_of_handle_global_alloc): Turn off insn rescanning.
83 * hard-reg-set.h: Fixed comment.
84 * local-alloc.c (update_equiv_regs): Change DF_RA_LIVE
85 usage to DF_LIVE usage and delete refs to TOP sets.
86 (block_alloc): Mark regs as live if they are in the artificial
88 (find_stack_regs): New function.
89 (rest_of_handle_local_alloc): Changed urec problem to live
90 problem and do not turn off df rescanning.
91 * df.h (DF_UREC, DF_UREC_BB_INFO, DF_LIVE_TOP, DF_RA_LIVE_IN,
92 DF_RA_LIVE_TOP, DF_RA_LIVE_OUT, df_urec_bb_info, df_urec,
93 df_urec_add_problem, df_urec_get_bb_info, df_has_eh_preds): Removed.
94 (DF_CHAIN, DF_NOTE, DF_CHAIN): Renumbered.
95 (DF_REF_EXTRACT, DF_REF_STRICT_LOWER_PART, DF_REF_SUBREG): New
96 fields in df_ref_flags. The rest have been renumbered.
97 * init-regs.c (initialize_uninitialized_regs): Enhanced debugging
99 * rtl.h (subreg_nregs_with_regno): New function.
100 * df-problems.c: (df_get_live_out, df_get_live_in,
101 df_get_live_top): Removed reference to DF_RA_LIVE.
102 (df_lr_reset, df_lr_transfer_function, df_live_free_bb_info,
103 df_live_alloc, df_live_reset, df_live_local_finalize,
104 df_live_free): Make top set only if different from in set.
105 (df_lr_top_dump, df_live_top_dump): Only print top set if
106 different from in set.
107 (df_lr_bb_local_compute): Removed unnecessary check.
108 (df_urec_problem_data, df_urec_set_bb_info, df_urec_free_bb_info,
109 df_urec_alloc, df_urec_mark_reg_change, earlyclobber_regclass,
110 df_urec_check_earlyclobber, df_urec_mark_reg_use_for_earlyclobber,
111 df_urec_mark_reg_use_for_earlyclobber_1, df_urec_bb_local_compute,
112 df_urec_local_compute, df_urec_init, df_urec_local_finalize,
113 df_urec_confluence_n, df_urec_transfer_function, df_urec_free,
114 df_urec_top_dump, df_urec_bottom_dump, problem_UREC,
115 df_urec_add_problem): Removed.
116 (df_simulate_fixup_sets): Changed call from df_has_eh_preds to
118 * Makefile.in (ra-conflict.o, ra.h): New dependencies.
119 * basic_block.h (bb_has_abnormal_pred): New function.
120 * reload1.c (compute_use_by_pseudos): Change DF_RA_LIVE
121 usage to DF_LIVE usage.
123 2007-10-02 Revital Eres <eres@il.ibm.com>
125 * config/rs6000/predicates.md (easy_vector_constant): Return false
126 for 750CL paired vectors.
127 * config/rs6000/paired.md (movv2sf_paired): Fix move of easy
129 (vec_initv2sf): Add new description.
130 (vconcatsf): Likewise.
131 * config/rs6000/rs6000-protos.h: Declare paired_expand_vector_init.
132 * config/rs6000/rs6000.c (paired_expand_vector_init): New function.
134 2007-10-01 Alexandre Oliva <aoliva@redhat.com>
136 * tree-ssa-sink.c (sink_code_in_bb): Don't stop sinking after
137 sinking the last stmt in a BB.
139 2007-10-01 Alexandre Oliva <aoliva@redhat.com>
142 * tree-sra.c (struct sra_elt): Add in_bitfld_block.
143 (sra_hash_tree): Handle BIT_FIELD_REFs.
144 (sra_elt_hash): Don't hash bitfld blocks.
145 (sra_elt_eq): Skip them in parent compares as well. Handle
147 (build_element_name_1): Handle BIT_FIELD_REFs.
148 (instantiate_element): Propagate nowarn from parents. Create
149 BIT_FIELD_REF for variables that are widened by scalarization.
150 Gimple-zero-initialize all bit-field variables that are not
151 part of parameters that are going to be scalarized on entry.
152 (instantiate_missing_elements_1): Return the sra_elt.
153 (canon_type_for_field): New.
154 (try_instantiate_multiple_fields): New. Infer widest possible
155 access mode from decl or member type, but clip it at word
156 size, and only widen it if a field crosses an alignment
158 (instantiate_missing_elements): Use them.
159 (generate_one_element_ref): Handle BIT_FIELD_REFs.
160 (scalar_bitfield_p): New.
161 (sra_build_assignment): Optimize assignments from scalarizable
162 BIT_FIELD_REFs. Use BITS_BIG_ENDIAN to determine shift
165 (sra_build_bf_assignment): New. Optimize assignments to
166 scalarizable BIT_FIELD_REFs.
167 (sra_build_elt_assignment): New. Optimize BIT_FIELD_REF
168 assignments to full variables.
169 (generate_copy_inout): Use the new macros and functions.
170 (generate_element_copy): Likewise. Handle bitfld differences.
171 (generate_element_zero): Don't recurse for blocks. Use
172 sra_build_elt_assignment.
173 (generate_one_element_init): Take elt instead of var. Use
174 sra_build_elt_assignment.
175 (generate_element_init_1): Adjust.
176 (bitfield_overlap_info): New struct.
177 (bitfield_overlaps_p): New.
178 (sra_explode_bitfield_assignment): New. Adjust widened
179 variables to account for endianness.
180 (sra_sync_for_bitfield_assignment): New.
181 (scalarize_use): Re-expand assignment to/from scalarized
182 BIT_FIELD_REFs. Explode or sync needed members for
183 BIT_FIELD_REFs accesses or assignments. Use REPLDUP.
184 (scalarize_copy): Use REPLDUP.
185 (scalarize_ldst): Move assert before dereference. Adjust EH
187 (dump_sra_elt_name): Handle BIT_FIELD_REFs.
189 2007-10-01 Paolo Bonzini <bonzini@gnu.org>
191 * simplify-rtx.c (comparison_result, simplify_relational_operation_1):
192 Rename CR_* constants to CMP_*. Fix spacing.
194 2007-10-01 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
197 * Makefile.in (build_html_dir/gccinstall): gccinstall.texi needs
198 to be processed with the special script doc/install.texi2html.
200 2007-09-30 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
202 * doc/invoke.texi (Wall): List the options enabled by Wall.
203 (Wstrict-aliasing): Add missing @option.
205 2007-09-30 Richard Sandiford <rsandifo@nildram.co.uk>
207 * config/mips/mips.c (mips_split_64bit_move): Use gen_rtx_REG_offset
208 rather than gen_lowpart to change a register from DImode to DFmode.
209 (mips_cannot_change_mode_class): Only allow FPRs to change mode if
210 both FROM and TO are integer modes that are no bigger than 4 bytes.
211 (mips_mode_ok_for_mov_fmt_p): New function.
212 (mips_preferred_reload_class): Use it instead of FLOAT_MODE_P.
213 (mips_secondary_reload_class): Tweak formatting and comments.
214 Use reg_class_subset_p instead of direct comparisons with
215 classes. Only allow direct FPR<->FPR moves for modes that
216 satisfy mips_mode_ok_for_mov_fmt_p. Only allow loads and stores
217 for 4- and 8-byte types. Handle reloads in which X is an FPR.
218 * config/mips/mips.md (*movdi_gp32_fp64): Remove f<-f alternative.
219 (*movdi_64bit): Likewise.
220 (*movsi_internal): Likewise.
221 (*movhi_internal): Likewise.
222 (*movqi_internal): Likewise.
224 2007-09-30 Diego Novillo <dnovillo@google.com>
227 * tree-ssa-ter.c (is_replaceable_p): Return false if STMT may
230 2007-09-30 Uros Bizjak <ubizjak@gmail.com>
232 PR tree-optimization/33597
233 * tree-vect-analyze.c (vect_build_slp_tree): Check if optab handler
234 for LSHIFT_EXPR and RSHIFT_EXPR is available for vec_mode.
236 2007-09-28 Uros Bizjak <ubizjak@gmail.com>
238 * config/i386/i386.c (ix86_expand_move): Use can_create_pseudo_p ()
239 instead of variants of (!reload_in_progress && !reload_completed).
240 (x86_expand_vector_move): Ditto.
242 2007-09-28 Ollie Wild <aaw@google.com>
245 2007-09-27 Ollie Wild <aaw@google.com>
247 * varasm.c (compare_constant): Removed call to
248 lang_hooks.expand_constant.
249 (copy_constants): Removed call to lang_hooks.expand_constant.
250 (compute_reloc_for_constant): Removed call to
251 lang_hooks.expand_constant.
252 (output_addressed_constants): Removed call to
253 lang_hooks.expand_constant.
254 (constructor_static_from_elts_p): Removed call to
255 lang_hooks.expand_constant.
256 (output_constant): Removed calls to lang_hooks.expand_constant.
257 * langhooks.h (struct lang_hooks): Removed field expand_constant.
258 * langhooks-def.h (lhd_return_tree): Removed.
259 (LANG_HOOKS_EXPAND_CONSTANT): Removed.
260 (LANG_HOOKS_INITIALIZER): Removed LANG_HOOKS_EXPAND_CONSTANT.
261 * langhooks.c (lhd_return_tree): Removed.
263 2007-09-28 Andrew Pinski <andrew_pinski@playstation.sony.com>
266 * config/spu/spu.c (spu_expand_insv): Call copy_rtx on the second
267 argument to gen_selb.
269 2007-09-28 Chao-ying Fu <fu@mips.com>
271 * libgcc-std.ver: Add fixed-point routines to GCC_4.3.0 section.
272 * doc/libgcc.texi (Fixed-point fractional library routines):
273 Fix typos for neg and cmp functions.
275 2007-09-28 Michael Matz <matz@suse.de>
277 PR rtl-optimization/33552
278 * function.c (match_asm_constraints_1): Check for overlap in
279 inputs and replace all occurences.
281 2007-09-28 Richard Sandiford <rsandifo@nildram.co.uk>
283 * config/mips/mips.c (override_options): Fix comment typo.
285 2007-09-28 Jie Zhang <jie.zhang@analog.com>
287 * config.gcc (bfin*-linux-uclibc*): Set extra_parts
288 to "crtbegin.o crtbeginS.o crtend.o crtendS.o".
289 * config/bfin/t-bfin-linux (crti.o): Don't build.
291 (EXTRA_MULTILIB_PARTS): Remove crti.o and crtn.o.
292 * config/bfin/t-bfin-uclinux (crti.o): Don't build.
294 (EXTRA_MULTILIB_PARTS): Remove crti.o and crtn.o.
296 2007-09-27 Ollie Wild <aaw@google.com>
298 * varasm.c (compare_constant): Removed call to
299 lang_hooks.expand_constant.
300 (copy_constants): Removed call to lang_hooks.expand_constant.
301 (compute_reloc_for_constant): Removed call to
302 lang_hooks.expand_constant.
303 (output_addressed_constants): Removed call to
304 lang_hooks.expand_constant.
305 (constructor_static_from_elts_p): Removed call to
306 lang_hooks.expand_constant.
307 (output_constant): Removed calls to lang_hooks.expand_constant.
308 * langhooks.h (struct lang_hooks): Removed field expand_constant.
309 * langhooks-def.h (lhd_return_tree): Removed.
310 (LANG_HOOKS_EXPAND_CONSTANT): Removed.
311 (LANG_HOOKS_INITIALIZER): Removed LANG_HOOKS_EXPAND_CONSTANT.
312 * langhooks.c (lhd_return_tree): Removed.
314 2007-09-27 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
317 * expr.c (emit_group_load_1): Split constant double when destination
318 length is half source length.
320 2007-09-27 Richard Sandiford <rsandifo@nildram.co.uk>
322 * config/mips/mips.h (PTF_AVOID_BRANCHLIKELY): New macro.
323 (mips_cpu_info): Add tune_flags.
324 (GENERATE_BRANCHLIKELY): Remove TARGET_SR71K check.
325 * config/mips/mips.c (mips_cpu_info_table): Add tune_flags fields.
327 (override_options): Remove deprecation code. Use branch-likely
328 instructions for optimize_size or if the tuning flags do not
329 suggest otherwise. Tweak warning.
330 (mips_matching_cpu_name_p, mips_parse_cpu): Use ARRAY_SIZE.
332 2007-09-27 Matthias Klose <doko@ubuntu.com>
334 * config/i386/t-linux64 (MULTILIB_OSDIRNAMES): Use ../lib32 as the
335 multilib osdirname if it exists.
336 * config/rs6000/t-linux64 (MULTILIB_OSDIRNAMES): Likewise.
338 2007-09-27 Ian Lance Taylor <iant@google.com>
340 PR tree-optimization/33565
341 * tree-ssa-loop-ch.c (copy_loop_headers): Set TREE_NO_WARNING on
342 assignments of comparisons.
343 * tree-ssa-sccvn.c (simplify_binary_expression): Add stmt
344 parameter. Change caller. Defer overflow warnings around call to
346 * fold-const.c (fold_undefer_overflow_warnings): Don't warn if
347 TREE_NO_WARNING is set on the statement.
348 * tree-ssa-forwprop.c
349 (tree_ssa_forward_propagate_single_use_vars): Don't test
350 TREE_NO_WARNING when calling fold_undefer_overflow_warnings.
351 * tree-cfg.c (fold_cond_expr_cond): Likewise.
353 2007-09-27 Joseph Myers <joseph@codesourcery.com>
355 * config/rs6000/rs6000.c (rs6000_legitimize_address): Do not
356 reduce offset by units of 0x10000 for SPE vector modes or modes
357 used with E500 double instructions.
359 2007-09-04 Paolo Bonzini <bonzini@gnu.org>
361 * simplify-rtx.c (comparison_result): New.
362 (simplify_const_relational_operation): Use it instead of the five
363 "equal|op[01]ltu?" variables; consequently remove redundant "else"s.
364 Improve bounds-checking optimizations; remove subsumed POPCOUNT
365 optimizations. Extract nonzero_address_p optimizations into a
366 separate "if" together with optimizations where op1 is const0_rtx.
367 Optimize comparing an IOR with zero. Simplify op0 RELOP op0 for
368 floating-point arguments too when appropriate. Hoist test for ABS
369 outside the final switch statement.
370 * cse.c (fold_rtx): Don't look for an IOR equivalent of
371 folded_arg0 if we found a constant equivalent. Remove
372 transformations done in simplify-rtx.c for "op0 RELOP op0".
374 2007-09-27 Jakub Jelinek <jakub@redhat.com>
376 * builtins.c (expand_builtin, expand_builtin_object_size,
377 expand_builtin_memory_chk, maybe_emit_chk_warning,
378 maybe_emit_sprintf_chk_warning): Use new %K format string specifier
380 * expr.c (expand_expr_real_1): Likewise.
381 * langhooks-def.h (struct diagnostic_info): Add forward decl.
382 (lhd_print_error_function): Add third argument.
383 * langhooks.h (struct diagnostic_info): Add forward decl.
384 (struct lang_hooks): Add third argument to print_error_function.
385 * diagnostic.h (diagnostic_info): Add abstract_origin field.
386 (diagnostic_last_function_changed, diagnostic_set_last_function): Add
388 (diagnostic_report_current_function): Likewise.
389 * toplev.c (announce_function): Pass NULL as second argument to
390 diagnostic_set_last_function.
391 * diagnostic.c (diagnostic_report_current_function): Add second
392 argument, pass it as third argument to lang_hooks.print_error_function.
393 (default_diagnostic_starter): Pass DIAGNOSTIC as second argument
394 to diagnostic_report_current_function.
395 (diagnostic_report_diagnostic): Initialize diagnostic->abstract_origin
396 and message.abstract_origin.
397 (verbatim): Initialize abstract_origin.
398 * pretty-print.h (text_info): Add abstract_origin field.
399 * pretty-print.c (pp_base_format): Handle %K.
400 * langhooks.c (lhd_print_error_function): Add third argument. If
401 diagnostic->abstract_origin, print virtual backtrace.
402 * c-format.c (gcc_diag_char_table, gcc_tdiag_char_table,
403 gcc_cdiag_char_table, gcc_cxxdiag_char_table): Support %K.
404 (init_dynamic_diag_info): Likewise.
406 2007-09-26 David Daney <ddaney@avtrex.com>
409 * config/mips/mips.md (sync_compare_and_swap<mode>, sync_old_add<mode>,
410 sync_new_add<mode>, sync_old_<optab><mode>, sync_new_<optab><mode>,
411 sync_old_nand<mode>, sync_new_nand<mode>,
412 sync_lock_test_and_set<mode>): Fix '&' constraint modifiers.
413 Update length attributes.
414 (sync_add<mode>, sync_sub<mode>, sync_old_sub<mode>,
415 sync_new_sub<mode>, sync_<optab><mode>, sync_nand<mode>): Update
417 * config/mips/mips.h (MIPS_COMPARE_AND_SWAP, MIPS_SYNC_OP,
418 MIPS_SYNC_OLD_OP, MIPS_SYNC_NEW_OP, MIPS_SYNC_NAND,
419 MIPS_SYNC_OLD_NAND, MIPS_SYNC_NEW_NAND, MIPS_SYNC_EXCHANGE): Add
422 2007-09-26 Richard Guenther <rguenther@suse.de>
424 PR tree-optimization/33563
425 * tree-ssa-dse.c (get_use_of_stmt_lhs): Rename to ...
426 (get_kill_of_stmt_lhs): ... this. Re-structure. Handle
428 (dse_optimize_stmt): Call get_kill_of_stmt_lhs instead of
431 2007-09-26 Joseph Myers <joseph@codesourcery.com>
434 * c-common.c (complete_array_type): Diagnose too-large arrays and
435 set type to error_mark_node.
437 2007-09-26 Richard Guenther <rguenther@suse.de>
439 PR tree-optimization/30375
440 PR tree-optimization/33560
441 * tree-ssa-dse.c (get_use_of_stmt_lhs): Give up on uses
445 2006-05-22 Aldy Hernandez <aldyh@redhat.com>
447 * tree-ssa-dse.c (aggregate_vardecl_d): New.
448 (dse_global_data): Add aggregate_vardecl field.
449 (dse_possible_dead_store_p): New.
450 Add prev_defvar variable.
451 Allow immediate uses and previous immediate uses to differ
452 if they are setting different parts of the whole.
453 (get_aggregate_vardecl): New.
454 (dse_record_partial_aggregate_store): New.
455 (dse_whole_aggregate_clobbered_p): New.
456 (dse_partial_kill_p): New.
457 Call dse_maybe_record_aggregate_store().
458 When checking whether a STMT and its USE_STMT refer to the
459 same memory address, check also for partial kills that clobber
461 Move some variable definitions to the block where they are used.
462 (aggregate_vardecl_hash): New.
463 (aggregate_vardecl_eq): New.
464 (aggregate_vardecl_free): New.
465 (aggregate_whole_store_p): New.
466 (tree_ssa_dse): Initialize and free aggregate_vardecl.
467 Mark which aggregate stores we care about.
469 2007-09-25 DJ Delorie <dj@redhat.com>
472 * config/m32c/m32c.c (m32c_immd_dbl_mov): Use INTVAL instead of
475 2007-09-25 Michael Meissner <michael.meissner@amd.com>
478 * config/i386/i386.c (ix86_expand_sse5_unpack): Change to call
479 gen_sse5_pperm_sign_v4si_v2di and gen_sse5_pperm_zero_v4si_v2di
480 for vector int32 -> int64 conversions. Don't write beyond the end
481 of the allocated vector for int32 -> int64 conversions.
483 2007-09-25 Revital Eres <eres@il.ibm.com>
485 * config/rs6000/paired.h (paired_sel): New.
486 * config/rs6000/rs6000.c (bdesc_3arg): Add selv2sf4.
487 (rs6000_expand_ternop_builtin): Pass zero const_double operand
488 when expanding selv2sf.
489 * config/rs6000/rs6000.h (rs6000_builtins): Add
490 PAIRED_BUILTIN_SELV2SF4.
492 2007-09-25 Joseph Myers <joseph@codesourcery.com>
495 * c-typeck.c (default_conversion): Call require_complete_type
496 before perform_integral_promotions.
497 (build_unary_op): Call require_complete_type except for ADDR_EXPR.
498 (build_c_cast): Call require_complete_type except for casts to
500 (convert_for_assignment): Call require_complete_type.
502 2007-09-25 Revital Eres <eres@il.ibm.com>
504 * config/spu/spu.md: Fix doloop pattern.
506 2007-09-25 Bernd Schmidt <bernd.schmidt@analog.com>
508 * config/bfin/bfin.c (expand_prologue_reg_save,
509 expand_epilogue_reg_restore): Code to save and restore I/M/B/L regs and
511 (expand_interrupt_handler_prologue, expand_interrupt_handler_epilogue):
512 ... from here. New argument ALL; callers changed.
513 (n_regs_saved_by_prologue): Count ASTAT for plain saveall functions.
514 (bfin_expand_prologue, bfin_expand_epilogue): Deal with functions that
515 have the "saveall" attribute.
517 2007-09-25 Hans-Peter Nilsson <hp@bitrange.com>
519 * config/mmix/mmix.h (FUNCTION_INCOMING_ARG_REGNO_P): Don't define
521 (INCOMING_REGNO, OUTGOING_REGNO): Define.
522 * config/mmix/mmix.c (mmix_opposite_regno): New function.
523 * config/mmix/mmix-protos.h (mmix_opposite_regno): Prototype.
525 * config/mmix/mmix.md ("*cmpcc_folded"): In condition, gate
526 with REG_P tests before REGNO access.
528 2007-09-24 DJ Delorie <dj@redhat.com>
531 * config/m32c/cond.md (stzx_reversed_<mode>): Add an output
533 (movqicc_<code>_<mode>): Likewise.
534 (movhicc_<code>_<mode>): Likewise.
536 2007-09-24 Rask Ingemann Lambertsen <rask@sygehus.dk>
539 * config/m32c/m32c.c (m32c_eh_return_data_regno): Leave an address
542 2007-09-24 Danny Smith <dannysmith@user.sourceforge.net>
545 * config/i386/i386.c (ix86_comp_type_attributes): Check
548 2007-09-24 Roman Zippel <zippel@linux-m68k.org>
550 * config/m68k/m68k.h (ASM_OUTPUT_ALIGN_WITH_NOP): New, use
551 "move.l %a4,%a4" to produce nops.
552 * config/m68k/m68k.c (override_options): Reset align options,
553 if neccessary align macro isn't avaible.
555 2007-09-24 Roman Zippel <zippel@linux-m68k.org>
557 config/m68k/linux.h (PREFERRED_STACK_BOUNDARY): New.
559 2007-09-24 Roman Zippel <zippel@linux-m68k.org>
561 * config/m68k/m68k.c (strict_low_part_peephole_ok): Don't leave
563 * config/m68k/m68k.md (movsi_m68k): Allow certain constant when
565 (peephole pattern): Convert most of them to RTL peephole pattern.
567 2007-09-24 Roman Zippel <zippel@linux-m68k.org>
569 * config/m68k/m68k.c (notice_update_cc): Recognize fp compare
570 (moved from fp compare patterns).
571 * config/m68k/m68k.md (cmp<mode>, cmp<mode>_68881, cmp<mode>_cf):
572 Cleanup predicates to relieve reload.
573 (conditional_trap): Reject conditional trap with fp condition.
574 * gcc/config/m68k/predicates.md (fp_src_operand): New, reject
575 certain constants early.
577 2007-09-24 Roman Zippel <zippel@linux-m68k.org>
579 * gcc/final.c (final_scan_insn): Remove accidentally duplicated code.
581 2007-09-24 Andrew Pinski <andrew_pinski@playstation.sony.com>
583 * config.gcc (powerpc*-*-*): --with-cpu=cell is a 64bit CPU.
584 Allow --with-tune=cell and --with-cpu=cell.
586 2007-09-24 David Edelsohn <edelsohn@gnu.org>
588 * config/rs6000/rs6000-protos.h (rs6000_emit_swrsqrtsf): Declare.
589 * config/rs6000/rs6000.opt (swdiv): Change option to ...
591 * config/rs6000/rs6000.c (rs6000_builtin_reciprocal): New
593 (TARGET_BUILTIN_RECIPROCAL): Use it.
594 (rs6000_builtin_expand): Expand recip, recipf, and rsqrtf.
595 (rs6000_init_builtins): Initialize recip, recipf, and rsqrtf.
596 (rs6000_emit_swrsqrtsf): New.
597 * config/rs6000/rs6000.h (rs6000_builtins): Add recip, recipf, and
599 * config/rs6000/rs6000.md (UNSPEC_RSQRT): Define.
600 (divsf3): Remove swdiv support.
603 (rsqrt_internal1): New.
604 (divdf3): Remove swdiv support.
607 2007-09-24 Jakub Jelinek <jakub@redhat.com>
610 * langhooks.h (struct lang_hooks_for_types): Add type_hash_eq
612 * langhooks-def.h (LANG_HOOKS_TYPE_HASH_EQ): Define.
613 (LANG_HOOKS_FOR_TYPES_INITIALIZER): Add LANG_HOOKS_TYPE_HASH_EQ.
614 * tree.c (type_hash_eq): For FUNCTION_TYPE use
615 lang_hooks.type.type_hash_eq in addition to generic tests.
617 2007-09-24 Pranav Bhandarkar <pranav.bhandarkar@celunite.com>
618 Ramana Radhakrishnan <ramana@hercules.pun.celunite.com>
620 * tree-inline.h (eni_weights): Add field target_builtin_cost to
621 reflect the cost per call to a target specific builtin.
622 * tree-inline.c (estimate_num_insns_1): If it is a CALL_EXPR for a
623 call to a target specific builtin, then use target_builtin_call_cost.
624 (init_inline_once): Initialize target_builtin_call_cost field.
626 2007-09-24 Kai Tietz <kai.tietz@onevision.com>
629 * config/i386/i386.c (return_in_memory_ms_64): Handle return types for
632 2007-09-23 H.J. Lu <hongjiu.lu@intel.com>
634 * configure.ac (ld_vers): Support GNU linker version xx.xx.*
635 * configure: Regenerated.
637 2007-09-23 Ollie Wild <aaw@google.com>
639 * fold-const.c (fold_binary): Fold BIT_AND_EXPR's with a pointer
641 (get_pointer_modulus_and_residue): New function.
643 2007-09-23 Richard Sandiford <rsandifo@nildram.co.uk>
645 * config/mips/mips.c (build_mips16_call_stub): On 64-bit targets,
646 combine an SC return value into a single register.
648 2007-09-23 Richard Sandiford <rsandifo@nildram.co.uk>
650 * opth-gen.awk (target_flags_explicit): Declare.
651 * toplev.h (target_flags_explicit): Delete declaration.
652 * toplev.c (target_flags): Likewise.
653 * config/mips/mips.h (mips_llsc_setting, mips_llsc): Delete.
654 (GENERATE_SYNC, GENERATE_LL_SC): Redefine using target_explicit_flags.
655 Never return true for TARGET_MIPS16.
656 * config/mips/mips.c (mips_llsc): Delete.
657 (mips_handle_option): Remove -mllsc handling.
658 (mips_strip_unspec_address): Tweak comment.
659 * config/mips/mips.opt (mllsc): Use a target mask.
661 2007-09-23 Richard Sandiford <rsandifo@nildram.co.uk>
663 * doc/tm.texi (POINTERS_EXTEND_UNSIGNED): Document extensions to
664 word_mode as well as Pmode.
666 2007-09-23 Richard Sandiford <rsandifo@nildram.co.uk>
668 * function.c (assign_parm_setup_block): Explicitly convert BLKmode
669 parameters from word_mode to the subword type if such a truncation
672 2007-09-23 Jakub Jelinek <jakub@redhat.com>
674 * configure.ac (MAKEINFO): Handle makeinfo version 4.10 and above.
675 * configure: Regenerated.
677 2007-09-23 Jan Hubicka <jh@suse.cz>
679 * params.def (INLINE_CALL_COST): Set to 12.
680 * invoke.texi (inline-call-cost): Update default value.
682 2007-09-23 Eric Botcazou <ebotcazou@adacore.com>
684 * config/alpha/alpha.md (movti): Use operand_subword for the split.
686 2007-09-23 Ayal Zaks <zaks@il.ibm.com>
687 Revital Eres <eres@il.ibm.com>
689 * modulo-sched.c (doloop_register_get): Rewrite the loop which
690 checks whether the count_reg is found outside the control part.
692 2007-09-23 Jakub Jelinek <jakub@redhat.com>
694 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Use get_callee_fndecl
695 instead of checking CALL_EXPR_FN directly to test for builtins.
696 If error or warning attributes are present, print
698 * c-common.c (handle_error_attribute): New function.
699 (c_common_attribute_table): Add error and warning
701 * doc/extend.texi: Document error and warning attributes.
703 * tree.h (block_nonartificial_location): New prototype.
704 * tree.c (block_nonartificial_location): New function.
705 * dwarf2out.c (gen_subprogram_die): Add DW_AT_artificial
706 if artificial attribute is present on abstract inline decl.
707 * c-common.c (handle_artificial_attribute): New function.
708 (c_common_attribute_table): Add artificial attribute.
709 * final.c (override_filename, override_linenum): New variables.
710 (final_scan_insn): For DBX_DEBUG or SDB_DEBUG, set override_filename
711 and override_linenum if inside of a block inlined from
712 __attribute__((__artificial__)) function.
713 (notice_source_line): Honor override_filename and override_linenum.
714 * doc/extend.texi: Document __attribute__((__artificial__)).
715 * config/i386/emmintrin.h: Add __artificial__ attribute to
716 all __always_inline__ functions.
717 * config/i386/mmintrin.h: Likewise.
718 * config/i386/tmmintrin.h: Likewise.
719 * config/i386/mm3dnow.h: Likewise.
720 * config/i386/pmmintrin.h: Likewise.
721 * config/i386/ammintrin.h: Likewise.
722 * config/i386/xmmintrin.h: Likewise.
723 * config/i386/smmintrin.h: Likewise.
724 * config/i386/bmmintrin.h: Likewise.
725 * config/i386/mmintrin-common.h: Likewise.
728 * expr.c (expand_constructor): New function.
729 (expand_expr_real_1) <case CONSTRUCTOR>: Call it.
730 (expand_expr_real_1) <case ARRAY_REF>: Call it if VALUE is
733 2007-09-23 Richard Sandiford <rsandifo@nildram.co.uk>
735 * config/mips/mips.c (dump_constants_1): Generalize to include
736 fractional and accumulator modes.
738 2007-09-23 Richard Sandiford <rsandifo@nildram.co.uk>
740 * config/mips/mips.h (ISA_HAS_DSP, ISA_HAS_DSPR2): New macros.
741 * config/mips/mips.c (mips_set_mips16_mode): Don't clear the DSP
743 (override_options): Check TARGET_HARD_FLOAT_ABI instead of
744 TARGET_HARD_FLOAT when testing whether -mpaired-single is
746 (mips_conditional_register_usage): Check ISA_HAS_DSP instead of
748 * config/mips/constraints.md (ka): Check ISA_HAS_DSPR2 instead of
750 * config/mips/mips.md (ANYF): Require TARGET_HARD_FLOAT for V2SF.
751 (mulv2sf3, movv2sf, movv2sf_hardfloat_64bit): Require
753 (<u>mulsidi3_32bit_internal, <u>msubsidi4, <u>maddsidi4): Check
754 ISA_HAS_DSPR2 instead of TARGET_HAS_DSPR2.
755 * config/mips/mips-dsp.md: Use ISA_HAS_DSP instead of TARGET_HAS_DSP
757 * config/mips/mips-dspr2.md: Likewise ISA_HAS_DSPR2 and
759 * config/mips/mips-fixed.md: Use ISA_HAS_DSP and ISA_HAS_DSPR2
760 instead of TARGET_HAS_DSP and TARGET_HAS_DSPR2.
761 * config/mips/mips-ps-3d.md: Add TARGET_HARD_FLOAT to V2SF patterns.
763 2007-09-22 Jason Merrill <jason@redhat.com>
766 * attribs.c (lookup_attribute_spec): Split out...
767 (decl_attributes): From here.
768 * tree.h: Declare it.
770 2007-09-22 Richard Sandiford <rsandifo@nildram.co.uk>
772 * doc/sourcebuild.texi: Document dg-add-options mips16_attribute.
774 2007-09-22 Eric Botcazou <ebotcazou@adacore.com>
776 * tree-inline.c (remap_type_1): Correctly chain variants.
778 2007-09-22 Richard Guenther <rguenther@suse.de>
780 PR tree-optimization/33146
781 * fold-const.c (fold_binary): Use the original tree
783 * tree.h (STRIP_SIGN_NOPS): Converting from or to pointer
786 2007-09-22 Eric Botcazou <ebotcazou@adacore.com>
789 * except.c (finish_eh_generation): Call commit_edge_insertions if
790 there are insns queued on the entry edge.
791 * config/alpha/alpha.c (alpha_gp_save_rtx): Insert the insns on
794 2007-09-22 Richard Sandiford <rsandifo@nildram.co.uk>
796 * doc/sourcebuild.texi (dg-add-c99-runtime-options): Document.
798 2007-09-19 Michael Meissner <michael.meissner@amd.com>
800 * gcc/config/i386/i386.c: Delete trailing whitespace.
801 * gcc/config/i386/i386.h: Ditto.
802 * gcc/config/i386/bmmintrin.h: Ditto.
803 * gcc/config/i386/sync.md: Ditto.
804 * gcc/config/i386/ppro.md: Ditto.
805 * gcc/config/i386/mmx.md: Ditto.
806 * gcc/config/i386/constraints.md: Ditto.
807 * gcc/config/i386/sse.md: Ditto.
808 * gcc/config/i386/athlon.md: Ditto.
809 * gcc/config/i386/i386.md: Ditto.
811 2007-09-21 Richard Guenther <rguenther@suse.de>
813 PR tree-optimization/33508
814 * tree-ssa-alias.c (mark_aliases_call_clobbered): Avoid
815 quadratic loop by keeping a bitmap of variables we have
816 to clobber all subvariables for.
817 (set_initial_properties): Likewise.
819 2007-09-21 Richard Sandiford <rsandifo@nildram.co.uk>
821 * config/mips/t-sde (TARGET_LIBGCC2_CFLAGS): Delete.
823 2007-09-20 Nigel Stephens <nigel@mips.com>
824 Chao-ying Fu <fu@mips.com>
826 * c-decl.c (finish_declspecs): When _Sat is used without
827 _Fract or _Accum, set the default type to cts_fract.
828 This avoids a warning of "type defaults to int".
830 2007-09-20 Joseph Myers <joseph@codesourcery.com>
832 * c-decl.c (check_bitfield_type_and_width): Don't allow _Bool
833 bit-fields wider than one bit.
835 2007-09-20 Jakub Jelinek <jakub@redhat.com>
838 * dwarf2out.c (modified_type_die): Handle TYPE_DECL with NULL
840 * dbxout.c (dbxout_type): Likewise.
844 * gimplify.c (gimplify_vla_decl): New function.
845 (gimplify_decl_expr): Move VLA decl handling to gimplify_vla_decl.
847 (gimplify_target_expr): Handle variable length TARGET_EXPRs.
849 2007-09-20 Richard Sandiford <rsandifo@nildram.co.uk>
851 * doc/invoke.texi (-minterlink-mips16): Document.
852 * config/mips/mips.opt (minterlink-mips16): New option.
853 * config/mips/mips.c (mips_function_ok_for_sibcall): Handle
856 2007-09-20 Joseph Myers <joseph@codesourcery.com>
858 * doc/extend.texi (Attribute Syntax): Remove old speculative
861 2007-09-20 Mark Shinwell <shinwell@codesourcery.com>
863 * combine.c: Include cgraph.h.
864 (setup_incoming_promotions): Rework to allow more aggressive
865 elimination of sign extensions when all call sites of the
866 current function are known to lie within the current unit.
868 2007-09-20 Richard Sandiford <rsandifo@nildram.co.uk>
870 * dse.c (find_shift_sequence): No-op rework of control flow.
872 2007-09-19 Richard Sandiford <rsandifo@nildram.co.uk>
874 * config/mips/mips.c (build_mips16_call_stub): Tidy. Fix second
875 GPR for DCmode on 64-bit targets. Remove redundant fallback.
877 2007-09-19 Andrew Pinski <andrew_pinski@playstation.sony.com>
879 * cfgexpand.c (dump_stack_var_partition): Use the correct
880 index for the offset.
882 2007-09-19 Joseph Myers <joseph@codesourcery.com>
884 * config/mips/sde.h: Switch to GPLv3.
886 2007-09-19 Eric Botcazou <ebotcazou@adacore.com>
888 * tree-sra.c (decide_block_copy): Decide if there are groups.
890 2007-09-19 Roman Zippel <zippel@linux-m68k.org>
892 * config/m68k/m68k.c (output_move_himode): Remove jump table
894 config/m68k/m68k.md (lea): Likewise.
895 * config/m68k/m68k.c (print_operand_address): Use simple pc
898 2007-09-19 Bernd Schmidt <bernd.schmidt@analog.com>
900 * doc/tm.texi (IS_ASM_LOGICAL_LINE_SEPARATOR): Document new argument.
901 * final.c (IS_ASM_LOGICAL_LINE_SEPARATOR): Provide two-argument
903 (asm_insn_count): Pass template as second argument to it.
904 * config/avr/avr.h (IS_ASM_LOGICAL_LINE_SEPARATOR): Likewise.
905 * config/pa/pa.h (IS_ASM_LOGICAL_LINE_SEPARATOR): Likewise.
906 * config/stormy16/stormy16.h (IS_ASM_LOGICAL_LINE_SEPARATOR): Likewise.
907 * config/cris/cris.h (IS_ASM_LOGICAL_LINE_SEPARATOR): Likewise.
908 * config/sh/sh.c (IS_ASM_LOGICAL_LINE_SEPARATOR): Likewise.
909 (sh_insn_length_adjustment): Pass template as second argument to it.
910 * config/bfin/bfin.h (IS_ASM_LOGICAL_LINE_SEPARATOR): New macro.
912 * config/bfin/bfin.md (define_asm_attributes): New.
914 2007-09-19 Jie Zhang <jie.zhang@analog.com>
916 * config.gcc (bfin*-linux-uclibc*): Add ./linux-sysroot-suffix.h
918 * config/bfin/print-sysroot-suffix.sh: New.
919 * config/bfin/t-bfin-elf (EXTRA_PARTS): Remove.
920 (MULTILIB_OPTIONS, MULTILIB_DIRNAMES, MULTILIB_MATCHES,
921 MULTILIB_EXCEPTIONS): Redefine with new multilibs.
922 * config/bfin/t-bfin-uclinux (EXTRA_PARTS): Remove.
923 (MULTILIB_OPTIONS, MULTILIB_DIRNAMES, MULTILIB_MATCHES,
924 MULTILIB_EXCEPTIONS): Redefine with new multilibs.
925 * config/bfin/t-bfin-linux (EXTRA_PARTS): Remove.
926 (MULTILIB_OPTIONS, MULTILIB_DIRNAMES, MULTILIB_MATCHES,
927 MULTILIB_EXCEPTIONS): Redefine with new multilibs.
928 (linux-sysroot-suffix.h): New target.
929 * config/bfin/bfin.opt (mcsync-anomaly): Use Var instead of Mask.
930 (mspecld-anomaly): Likewise.
931 * config/bfin/bfin-protos.h (enum bfin_cpu_type): Renamed from
932 (enum bfin_cpu): ... this. Add BFIN_CPU_BF522, BFIN_CPU_BF525,
933 BFIN_CPU_BF527, BFIN_CPU_BF538, BFIN_CPU_BF539, BFIN_CPU_BF542,
934 BFIN_CPU_BF544, BFIN_CPU_BF548, and BFIN_CPU_BF549.
935 (bfin_si_revision): Declare.
936 (bfin_workarounds): Declare.
937 (WA_SPECULATIVE_LOADS): Define.
938 (ENABLE_WA_SPECULATIVE_LOADS): Define.
939 (WA_SPECULATIVE_SYNCS): Define.
940 (ENABLE_WA_SPECULATIVE_SYNCS): Define.
941 * config/bfin/elf.h (STARTFILE_SPEC): Rename crt532.o to basiccrt.o.
942 (LIB_SPEC): Add %s to the linker scripts.
943 Use proper linker script for bf522, bf525, bf527,
944 bf538, bf539, bf542, bf544, bf548, and bf549.
945 * config/bfin/bfin.c (bfin_si_revision): Define.
946 (bfin_workarounds): Define.
947 (struct bfin_cpu): New.
949 (bfin_handle_option): Handle silicon revision part of -mcpu option.
950 (override_options): Set bfin_workarounds.
951 (length_for_loop): Replace TARGET_CSYNC_ANOMALY with
952 ENABLE_WA_SPECULATIVE_SYNCS, TARGET_SPECLD_ANOMALY with
953 ENABLE_WA_SPECULATIVE_LOADS.
954 (bfin_reorg): Likewise.
955 * config/bfin/bfin.h (TARGET_CPU_CPP_BUILTINS): Define
956 macros for bf522, bf525, bf527, bf538, bf539,
957 bf542, bf544, bf548, and bf549.
958 Define __SILICON_REVISION__ and __WORKAROUND_* macros if needed.
959 Don't define __ID_SHARED_LIB__ when -msep-data.
960 (TARGET_DEFAULT): Define as 0.
961 (DRIVER_SELF_SPECS): Add -mcpu=bf532 if no -mcpu option.
962 * doc/invoke.texi (Blackfin Options): Document silicon
963 revision part of -mcpu option and it now accepts bf522, bf525,
964 bf527, bf538, bf539, bf542, bf544, bf548, and bf549.
966 2007-09-18 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
968 * alias.c (memory_modified_1): Deconstify.
969 (memory_modified_in_insn_p): Don't use const_note_stores.
970 * rtl.h (const_note_stores): Delete.
971 * rtlanal.c (const_note_stores): Likewise.
973 2007-09-18 Richard Sandiford <rsandifo@nildram.co.uk>
975 * dse.c (find_shift_sequence): Temporarily revert to forbidding
978 2007-09-18 Rask Ingemann Lambertsen <rask@sygehus.dk>
982 * config/fr30/fr30.md (addsi3): Check REG_P() before calling REGNO().
983 (addsi_small_int): Likewise.
984 (addsi_big_int): Use rtx_equal_p() instead of REGNO() comparison.
985 (one_cmplsi2): Likewise.
987 (enter_func): Expand insn using hard_frame_pointer_rtx and
990 * config/fr30/fr30.c (fr30_expand_prologue): Check for
991 hard_frame_pointer_rtx instead of using REGNO() check.
992 Properly sign extend GEN_INT() argument.
994 2007-09-18 Roman Zippel <zippel@linux-m68k.org>
996 * config/m68k/m68k.c (override_options): Remove USE_GAS,
998 (output_dbcc_and_branch, output_scc_di): Replace all jbcc
999 alternatives with just jcc.
1000 * config/m68k/m68k.md (addsi_lshrsi_31, beq0_di, bne0_di,
1001 bge0_di, blt0_di, bgtu, bltu, bgeu, bleu, bgtu_rev,
1002 bltu_rev, bgeu_rev, bleu_rev, jump, dbne_hi, dbne_si,
1003 dbge_hi, dbge_si): Likewise.
1005 2007-09-18 Roman Zippel <zippel@linux-m68k.org>
1007 * config/m68k/m68k.md (beq, bne, bgt, blt, bge, ble, bordered,
1008 bunordered, buneq, bunge, bungt, bunle, bunlt, bltgt, beq_rev,
1009 bne_rev, bgt_rev, blt_rev, bge_rev, ble_rev, bordered_rev,
1010 bunordered_rev, buneq_rev, bunge_rev, bungt_rev, bunle_rev,
1011 bunlt_rev, bltgt_rev): Replace all fbcc with fjcc.
1013 2007-09-18 Maxim Kuvyrkov <maxim@codesourcery.com>
1015 * config/m68k/m68k-devices.def (51qe): New device.
1016 * config/m68k/m68k.c (FL_FOR_isa_c): Remove division unit. Add it
1017 to all uses of FL_FOR_isa_c for compatibility.
1018 (all_microarchs): Add cfv1 microarchitecture.
1019 (m68k_handle_option): Handle m51qe option.
1020 * config/m68k/m68k.h (TARGET_CPU_CPP_BUILTINS): Handle 51QE.
1021 (TUNE_CFV1): New macro.
1022 (enum uarch_type): Add ucfv1.
1023 * doc/invoke.texi: Document 51qe device and cfv1 microarchitecture.
1025 2007-09-18 Richard Guenther <rguenther@suse.de>
1027 PR tree-optimization/31863
1028 * tree-ssa-structalias.c (create_variable_info_for): Always
1029 free the fieldstack.
1031 2007-09-18 Dorit Nuzman <dorit@il.ibm.com>
1033 * opts.c (decode_options): Enable vectorization under -O3.
1035 2007-09-18 Richard Guenther <rguenther@suse.de>
1037 PR tree-optimization/33340
1038 * tree-ssa-sccvn.c (set_ssa_val_to): Do not set values to
1039 SSA_NAMEs that occur in abnormal PHI nodes.
1041 2007-09-17 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1043 * tree-cfg.c (is_ctrl_altering_stmt, tree_block_ends_with_call_p):
1044 Don't use const_get_call_expr_in.
1045 * tree-gimple.c (const_get_call_expr_in): Delete.
1046 * tree-gimple.h (const_get_call_expr_in): Likewise.
1048 2007-09-18 Richard Sandiford <rsandifo@nildram.co.uk>
1050 * df-scan.c (df_notes_rescan): Do nothing if the instruction does
1051 not yet have a basic block.
1052 * dse.c (find_shift_sequence): Don't set DF_NO_INSN_RESCAN.
1054 2007-09-18 Richard Sandiford <rsandifo@nildram.co.uk>
1056 * config/mips/mips.c (mips_file_start): Add ".previous" directives
1057 to both ".section"s.
1059 2007-09-17 Richard Sandiford <rsandifo@nildram.co.uk>
1061 * config/mips/mips.c (mips_output_mi_thunk): Use
1062 mips_function_ok_for_sibcall and const_call_insn_operand
1063 to determine if a direct sibcall is allowed. Use
1064 mips_classify_symbol to determine a global pointer is needed.
1066 2007-09-17 Richard Sandiford <rsandifo@nildram.co.uk>
1068 * config/mips/mips.md (*clear_upper32): Use "W" as the memory operand.
1070 2007-09-17 Chao-ying Fu <fu@mips.com>
1071 Nigel Stephens <nigel@mips.com>
1073 * config/fixed-bit.h: New file.
1074 * config/fixed-bit.c: New file.
1075 * doc/libgcc.texi (Fixed-point fractional library routines): New node.
1077 2007-09-18 Jakub Jelinek <jakub@redhat.com>
1079 * c-format.h (format_kind_info): Add alloc_char field.
1080 * c-format.c (scanf_flag_specs): Add 'm'.
1081 (scanf_flag_pairs): Add 'a', 'm' pair.
1082 (scan_char_table): Allow 'm' modifier for c, s, [, C and S.
1083 (format_types_orig): Add alloc_char fields.
1084 (check_format_info_main): Rename aflag to alloc_flag.
1085 Handle fki->alloc_char. modifier after width and before length
1086 modifiers. Move FMT_FLAG_SCANF_A_KLUDGE handling before
1087 length modifiers as well.
1088 * config/sol2-c.c (solaris_format_types): Add alloc_char field.
1091 * builtins.c (expand_builtin_memory_chk): Handle COMPOUND_EXPRs
1092 returned by build_call_expr.
1094 2007-09-17 Eric Botcazou <ebotcazou@adacore.com>
1096 * tree-sra.c (maybe_lookup_element_for_expr) <COMPONENT_REF>: Return
1097 NULL for variable-sized records too.
1098 (sra_walk_expr) <COMPONENT_REF>: Stop at variable-sized records too.
1100 2007-09-17 Tom Tromey <tromey@redhat.com>
1102 * c-decl.c (pushdecl): Don't set DECL_LANG_SPECIFIC.
1103 (c_builtin_function): Likewise.
1104 (grokdeclarator): Likewise.
1106 2007-09-17 Zdenek Dvorak <ook@ucw.cz>
1108 PR rtl-optimization/26449
1109 * loop-invariant.c (move_invariant_reg): Do not use force_operand.
1110 (seq_insns_valid_p): Removed.
1112 2007-09-17 Eric Botcazou <ebotcazou@adacore.com>
1114 * tree-nomudflap.c (gate_mudflap): New static function.
1115 (pass_mudflap_1): Use it as gate function.
1116 (pass_mudflap_2): Likewise.
1118 2007-09-17 Jan Hubicka <jh@suse.cz>
1122 * loop-invariant.c (move_invariant_reg): Unshare sequence.
1124 2007-09-17 Victor Kaplansky <victork@il.ibm.com>
1126 PR tree-optimization/33319
1127 * tree-vect-analyze.c (vect_same_range_drs): New.
1128 (vect_vfa_range_equal): New.
1129 (vect_is_duplicate_ddr): Removed.
1130 (vect_mark_for_runtime_alias_test): Do not perform marking when
1131 optimizing for size or max_param for alias checking is zero.
1132 Move the function before vect_analyze_data_ref_dependence.
1133 (vect_analyze_data_ref_dependence): Add call to
1134 vect_mark_for_runtime_alias_test in two cases when dependence
1136 (vect_analyze_data_ref_dependences): Do not call to
1137 vect_mark_for_runtime_alias_test.
1138 (vect_prune_runtime_alias_test_list): New.
1139 (vect_analyze_loop): Add call to vect_prune_runtime_alias_test_list.
1140 * tree-vect-transform.c (vect_estimate_min_profitable_iters):
1141 Update vec_outside_cost.
1142 (vect_vfa_segment_size): More compact code, use TYPE_SIZE_UNIT.
1143 (vect_create_cond_for_alias_checks): Build the base address of data
1144 reference from DR_GROUP_FIRST_DR.
1145 (vect_loop_versioning): New.
1146 (vect_transform_loop): Add a call to vect_loop_versioning.
1147 Remove factored out code.
1149 2007-09-16 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1152 * expr.c (store_expr): Call adjust_address to change mode of dest_mem
1155 2007-09-16 Richard Sandiford <rsandifo@nildram.co.uk>
1157 * dse.c (find_shift_sequence): Allow word as well as subword shifts.
1158 Do the tentative shift expansion with the DF_NO_INSN_RESCAN flag set.
1159 Fix the call to insn_rtx_cost. Skip access sizes that require a
1160 real truncation of the store register. Use convert_move instead
1161 of gen_lowpart when narrowing the result.
1162 (replace_read): Use convert_move instead of gen_lowpart when
1163 narrowing the store rhs.
1165 2007-09-16 Richard Sandiford <rsandifo@nildram.co.uk>
1167 * config/mips/mips.md (SHORT): Fix long line.
1168 (SUBDI): New mode iterator. Extend the shift-and-truncate insns
1169 to QImode and HImode.
1171 2007-09-16 Richard Sandiford <rsandifo@nildram.co.uk>
1173 * config/mips/mips.h (POINTERS_EXTEND_UNSIGNED): Define.
1175 2007-09-15 Zdenek Dvorak <ook@ucw.cz>
1177 * tree-parloops.c: New file.
1178 * tree-ssa-operands.h (free_stmt_operands): Declare.
1179 * tree-ssa-loop-manip.c (split_loop_exit_edge): Return the new basic
1181 * tree-pass.h (pass_parallelize_loops): Declare.
1182 * omp-low.c (expand_omp_parallel, expand_omp_for): Update SSA form for
1184 (build_omp_regions_1): Allow analysing just a single OMP region and
1186 ( build_omp_regions_root, omp_expand_local): New functions.
1187 (build_omp_regions): Add argument to build_omp_regions_1 call.
1188 * builtins.def (DEF_GOMP_BUILTIN): Initialize OMP builtins when
1189 autoparallelization is run.
1190 * timevar.def (TV_TREE_PARALLELIZE_LOOPS): New.
1191 * tree-ssa-loop.c (gate_tree_parallelize_loops, tree_parallelize_loops,
1192 pass_parallelize_loops): New.
1193 * common.opt (ftree-parallelize-loops): New.
1194 * tree-flow.h (omp_expand_local, tree_duplicate_sese_tail,
1195 parallelize_loops): Declare.
1196 (add_phi_args_after_copy, split_loop_exit_edge): Declaration changed.
1197 * Makefile.in (tree-parloops.o): Added.
1198 * tree-cfg.c (add_phi_args_after_copy_edge, tree_duplicate_sese_tail):
1200 (add_phi_args_after_copy_bb): Use add_phi_args_after_copy_edge.
1201 (add_phi_args_after_copy): Call add_phi_args_after_copy_edge for
1202 one extra edge as well.
1203 (tree_duplicate_sese_region): Add argument to add_phi_args_after_copy.
1204 Use VEC_free to free doms vector.
1205 (move_block_to_fn): Update loop info. Remove phi nodes for virtual
1206 operands. Recompute operand caches in the new function.
1207 (move_sese_region_to_fn): Update loop info.
1208 * passes.c (init_optimization_passes): Add pass_parallelize_loops.
1209 * tree-ssa-operands.c (free_stmt_operands): New function.
1211 * doc/passes.texi: Document autoparallelization.
1212 * doc/invoke.texi (-ftree-parallelize-loops): New option.
1214 2007-09-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1217 * pa.c (function_value): Use GET_MODE_BITSIZE instead of
1220 2007-09-15 Dorit Nuzman <dorit@il.ibm.com>
1222 * tree-vect-transform.c (vect_get_vec_defs_for_stmt_copy): check if
1223 the VEC is not NULL.
1224 (vectorizable_type_demotion, vectorizable_type_promotion): Check that
1225 get_vectype_for_scalar_type succeeded.
1226 (vectorizable_conversion): Likewise.
1228 2007-09-14 Jan Hubicka <jh@suse.cz>
1230 * config/i386/i386.md (*floatdi<mode>2_i387): Guard against
1233 2007-09-14 Uros Bizjak <ubizjak@gmail.com>
1236 * config/i386/i386.md (fmodxf3): Copy operands[2] to temporary
1237 register when operands[2] equals operands[1].
1238 (remainderxf3): Ditto.
1240 2007-09-14 Sandra Loosemore <sandra@codesourcery.com>
1241 Nigel Stephens <nigel@mips.com>
1243 * doc/tm.texi (LIBGCC2_UNWIND_ATTRIBUTE): Document.
1244 * unwind-generic.h (LIBGCC2_UNWIND_ATTRIBUTE): Define.
1245 (_Unwind_RaiseException): Add LIBGCC2_UNWIND_ATTRIBUTE to
1247 (_Unwind_ForcedUnwind): Likewise.
1248 (_Unwind_Resume): Likewise.
1249 (_Unwind_Resume_or_Rethrow): Likewise.
1250 (_Unwind_Backtrace): Likewise.
1251 (_Unwind_SjLj_RaiseException): Likewise.
1252 (_Unwind_SjLj_ForcedUnwind): Likewise.
1253 (_Unwind_SjLj_Resume): Likewise.
1254 (_Unwind_SjLj_Resume_or_Rethrow): Likewise.
1255 * unwind.inc (_Unwind_RaiseException): Add LIBGCC2_UNWIND_ATTRIBUTE
1257 (_Unwind_ForcedUnwind): Likewise.
1258 (_Unwind_Resume): Likewise.
1259 (_Unwind_Resume_or_Rethrow): Likewise.
1260 (_Unwind_Backtrace): Likewise.
1261 * unwind-compat.c (_Unwind_Backtrace): Likewise.
1262 (_Unwind_ForcedUnwind): Likewise.
1263 (_Unwind_RaiseException): Likewise.
1264 (_Unwind_Resume): Likewise.
1265 (_Unwind_Resume_or_Rethrow): Likewise.
1267 * config/mips/mips.h (LIBGCC2_UNWIND_ATTRIBUTE): Define to force
1268 nomips16 mode when IN_LIBGCC2 with hard float.
1270 2007-09-14 Richard Sandiford <rsandifo@nildram.co.uk>
1272 * config/mips/sdemtk.opt: Update to GPLv3.
1273 * config/mips/sdemtk.h: Likewise.
1275 2007-09-14 Nigel Stephens <nigel@mips.com>
1277 * config.gcc (mips*-*-linux*): Recognise mipsisa32r2 and set
1278 MIPS_ISA_DEFAULT appropriately. Don't make soft-float the default
1279 for mipsisa32-*-linux*.
1281 2007-09-14 Nigel Stephens <nigel@mips.com>
1282 David Ung <davidu@mips.com>
1283 Thiemo Seufer <ths@mips.com>
1284 Richard Sandiford <richard@codesourcery.com>
1286 * config.gcc (mips*-sde-elf*): Add support for the SDE C libraries.
1287 * configure.ac: Add a mipssde threading type.
1288 * configure: Regenerate.
1289 * config/mips/sdemtk.h: New file.
1290 * config/mips/t-sdemtk: Likewise.
1291 * config/mips/sdemtk.opt: Likewise.
1292 * gthr-mipssde.h: Likewise.
1293 * config/mips/sde.h (FUNCTION_PROFILER): Move to config/mips/sdemtk.h.
1294 * config/mips/mips.h (MIPS_SAVE_REG_FOR_PROFILING_P): New macro.
1295 (MIPS_ICACHE_SYNC): New macro, split from ...
1296 * config/mips/mips.md (clear_cache): ...here.
1297 * config/mips/mips.c (mips_save_reg_p): Check
1298 MIPS_SAVE_REG_FOR_PROFILING_P on profiled functions.
1299 (build_mips16_function_stub): Use targetm.strip_name_encoding.
1300 (build_mips16_call_stub): Likewise.
1302 2007-09-14 Richard Sandiford <richard@codesourcery.com>
1304 * Makefile.in (stmp-int-hdrs): Depend on fixinc_list.
1306 2007-09-14 Jakub Jelinek <jakub@redhat.com>
1309 * config/ia64/ia64.c (find_gr_spill): Don't decrement
1310 current_frame_info.n_local_regs. Don't return emitted local
1312 (ia64_compute_frame_size): Improve unwind hack to put
1313 RP, PFS, FP in that order by allowing some of the registers
1314 been already emitted, as long as they are emitted to the
1317 2007-09-14 Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
1319 * config/spu/vmx2spu.h (vec_extract, vec_insert, vec_lvlx,
1320 vec_lvlxl, vec_lvrx, vec_lvrxl, vec_promote, vec_splats,
1321 vec_stvlx, vec_stvlxl, vec_stvrx, vec_stvrxl): New intrinsics.
1323 2007-09-13 Eric Christopher <echristo@apple.com>
1324 Kenneth Zadeck <zadeck@naturalbridge.com>
1326 * dse.c (find_shift_sequence): New function.
1327 (replace_read): Add case to remove read if it requires shift.
1328 * config/i386/i386.c (ix86_expand_prologue): Fixed typo in comment.
1330 2007-09-13 Tom Tromey <tromey@redhat.com>
1332 * c-common.c (fname_as_string): Update.
1333 * c-parser.c (c_parser) <lex_untranslated_string>: New field.
1334 (c_lex_one_token): Update. Add 'parser' argument.
1335 (c_parser_simple_asm_expr): Update.
1336 (c_parser_attributes): Update.
1337 (c_parser_asm_statement): Update.
1338 (c_parser_asm_operands): Update.
1339 (c_parser_peek_token): Update.
1340 (c_parser_peek_2nd_token): Update.
1341 * c-lex.c (c_lex_string_translate): Remove.
1342 (c_lex_return_raw_strings): Likewise.
1343 (c_lex_with_flags): Added 'lex_flags' argument.
1344 (lex_string): Added 'translate' argument.
1345 * c-pragma.h (c_lex_with_flags): Update.
1346 (c_lex_string_translate, c_lex_return_raw_strings): Remove.
1347 (C_LEX_STRING_NO_TRANSLATE): New define.
1348 (C_LEX_RAW_STRINGS): Likewise.
1350 2007-09-13 Bernd Schmidt <bernd.schmidt@analog.com>
1353 * config/bfin/bfin.c (enum bfin_builtins): Add BFIN_BUILTIN_ONES,
1354 BFIN_BUILTIN_CPLX_MUL_16_S40, BFIN_BUILTIN_CPLX_MAC_16_S40,
1355 BFIN_BUILTIN_CPLX_MSU_16_S40, and BFIN_BUILTIN_CPLX_SQU.
1356 (bfin_init_builtins): Initialize __builtin_bfin_ones,
1357 __builtin_bfin_min_fr1x16, __builtin_bfin_max_fr1x16,
1358 __builtin_bfin_min_fr1x32, __builtin_bfin_max_fr1x32,
1359 __builtin_bfin_cmplx_add, __builtin_bfin_cmplx_sub,
1360 __builtin_bfin_cmplx_mul_s40, __builtin_bfin_cmplx_mac_s40,
1361 __builtin_bfin_cmplx_msu_s40 and __builtin_bfin_csqu_fr16.
1362 (bdesc_1arg): Add __builtin_bfin_ones.
1363 (bfin_expand_builtin): Expand __builtin_bfin_cmplx_mul_s40,
1364 __builtin_bfin_cmplx_mac_s40, __builtin_bfin_cmplx_msu_s40,
1365 and __builtin_bfin_csqu_fr16.
1366 * config/bfin/bfin.md (UNSPEC_ONES): New constant.
1367 (ones): New define_insn.
1368 (ssaddhi3_parts): New define_insn.
1369 (sssubhi3_parts): New define_insn.
1370 (flag_mulhi_parts): New define_insn.
1372 2007-09-13 Seongbae Park <seongbae.park@gmail.com>
1374 * common.opt (femit-class-debug-always): Turn off by default.
1376 2007-09-13 Bernd Schmidt <bernd.schmidt@analog.com>
1378 * config/bfin/bfin.md (reload_outpdi, reload_inpdi): New patterns.
1379 * config/bfin/bfin.c (bfin_secondary_reload): Make sure we use them.
1381 2007-09-13 James E. Wilson <wilson@specifix.com>
1383 PR tree-optimization/33389
1384 * tree-ssa-operands.c (append_vuse): If ann->in_vdef_list true,
1385 then set build_loads before returning.
1387 2007-09-13 Sandra Loosemore <sandra@codesourcery.com>
1388 David Ung <davidu@mips.com>
1390 * config/mips/mips.h (ASM_OUTPUT_REG_PUSH): Replace {d}subu with
1391 {d}addiu and a negative immediate such that it works with MIPS16
1394 2007-09-13 H.J. Lu <hongjiu.lu@intel.com>
1397 * configure.ac (ld_vers): Support Linux linker.
1398 * configure: Regenerated.
1400 2007-09-13 Richard Sandiford <richard@codesourcery.com>
1401 Sandra Loosemore <sandra@codesourcery.com>
1403 * config/mips/mips.h (SYMBOL_FLAG_MIPS16_FUNC): Delete.
1404 (SYMBOL_REF_MIPS16_FUNC_P): Delete.
1405 * config/mips/mips.c (mips_attribute_table): Turn mips16 and
1406 nomips16 into decl attributes.
1407 (TARGET_INSERT_ATTRIBUTES): Override.
1408 (TARGET_MERGE_DECL_ATTRIBUTES): Likewise.
1409 (TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P): Always return true.
1410 (mips_mips16_type_p, mips_nomips16_type_p): Delete in favor of...
1411 (mips_mips16_decl_p, mips_nomips16_decl_p): ...these new functions.
1412 (mips_comp_type_attributes): Remove mips16 and nomips16 handling.
1413 (mips_use_mips16_mode_p): Reimplement as a function that takes
1414 a decl and considers only decl attributes. If the decl is nested
1415 function, use its parent attributes.
1416 (mips_function_ok_for_sibcall): Use mips_use_mips16_mode_p
1417 instead of SYMBOL_REF_MIPS16_FUNC_P.
1418 (mips_set_mips16_mode): Move call to sorry here from old
1419 mips_use_mips16_mode_p.
1420 (mflip_mips16_entry): New structure.
1421 (mflip_mips16_htab): New variable.
1422 (mflip_mips16_htab_hash, mflip_mips16_htab_eq): New functions.
1423 (mflip_mips16_use_mips16_p, mips_insert_attributes): Likewise.
1424 (mips_merge_decl_attributes): New function.
1425 (mips_set_current_function): Reinstate call to mips_set_mips16_mode.
1426 Use mips_use_mips16_mode_p.
1427 (mips_output_mi_thunk): Use mips_use_mips16_mode_p instead of
1428 SYMBOL_REF_MIPS16_FUNC_P.
1429 (mips_encode_section_info): Don't set SYMBOL_FLAG_MIPS16_FUNC.
1431 2007-09-13 Richard Sandiford <richard@codesourcery.com>
1433 * c-parser.c (c_parser_struct_declaration): Check for a null return.
1435 2007-09-13 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
1438 * gcc.c (xputenv): Make argument const, and use CONST_CAST.
1440 2007-09-12 Michael Meissner <michael.meissner@amd.com>
1441 Dwarakanath Rajagopal <dwarak.rajagopal@amd.com>
1442 Tony Linthicum <tony.linthicum@amd.com>
1444 * tree.h (function_args_iterator): New type to iterate over
1446 (FOREACH_FUNCTION_ARGS_PTR): Iterator macros for iterating over
1447 function arguments providing a pointer to the argument.
1448 (FOREACH_FUNCTION_ARGS): Iterator macros for iterating over
1449 function arguments providing the argument.
1450 (function_args_iter_init): Inline function to initialize
1451 function_args_iterator.
1452 (function_args_iter_cond_ptr): Inline function to return the next
1453 pointer to hold the argument.
1454 (function_args_iter_cond): Inline function to return the next
1456 (function_args_iter_cond_next): Advance the function args
1458 (stdarg_p): New function, return true if variable argument
1460 (prototype_p): New function, return true if function is
1462 (function_args_count): New function, count the number of arguments
1465 * tree.c (stdarg_p): New function, return true if variable
1467 (prototype_p): New function, return true if function is
1470 * config/i386/i386.h (TARGET_SSE5): New macro for SSE5.
1471 (TARGET_ROUND): New macro for the round/ptest instructions which
1472 are shared between SSE4.1 and SSE5.
1473 (OPTION_MASK_ISA_ROUND): Ditto.
1474 (OPTION_ISA_ROUND): Ditto.
1475 (TARGET_FUSED_MADD): New macro for -mfused-madd swtich.
1476 (TARGET_CPU_CPP_BUILTINS): Add SSE5 support.
1478 * config/i386/i386.opt (-msse5): New switch for SSE5 support.
1479 (-mfused-madd): New switch to give users control over whether the
1480 compiler optimizes to use the multiply/add SSE5 instructions.
1482 * config/i386/i386.c (m_AMD_MULTIPLE): Rename from
1483 m_ATHLON_K8_AMDFAM10, and change all uses.
1484 (enum pta_flags): Add PTA_SSE5.
1485 (ix86_handle_option): Turn off 3dnow if -msse5.
1486 (override_options): Add SSE5 support.
1487 (print_operand): %Y prints comparison codes for SSE5 com/pcom
1489 (ix86_expand_sse_movcc): Add SSE5 support.
1490 (ix86_expand_sse5_unpack): New function to use pperm to unpack a
1491 vector type to the next largest size.
1492 (ix86_expand_sse5_pack): New function to use pperm to pack a
1493 vector type to the next smallest size.
1494 (IX86_BUILTIN_FMADDSS): New for SSE5 intrinsic.
1495 (IX86_BUILTIN_FMADDSD): Ditto.
1496 (IX86_BUILTIN_FMADDPS): Ditto.
1497 (IX86_BUILTIN_FMADDPD): Ditto.
1498 (IX86_BUILTIN_FMSUBSS): Ditto.
1499 (IX86_BUILTIN_FMSUBSD): Ditto.
1500 (IX86_BUILTIN_FMSUBPS): Ditto.
1501 (IX86_BUILTIN_FMSUBPD): Ditto.
1502 (IX86_BUILTIN_FNMADDSS): Ditto.
1503 (IX86_BUILTIN_FNMADDSD): Ditto.
1504 (IX86_BUILTIN_FNMADDPS): Ditto.
1505 (IX86_BUILTIN_FNMADDPD): Ditto.
1506 (IX86_BUILTIN_FNMSUBSS): Ditto.
1507 (IX86_BUILTIN_FNMSUBSD): Ditto.
1508 (IX86_BUILTIN_FNMSUBPS): Ditto.
1509 (IX86_BUILTIN_FNMSUBPD): Ditto.
1510 (IX86_BUILTIN_PCMOV_V2DI): Ditto.
1511 (IX86_BUILTIN_PCMOV_V4SI): Ditto.
1512 (IX86_BUILTIN_PCMOV_V8HI): Ditto.
1513 (IX86_BUILTIN_PCMOV_V16QI): Ditto.
1514 (IX86_BUILTIN_PCMOV_V4SF): Ditto.
1515 (IX86_BUILTIN_PCMOV_V2DF): Ditto.
1516 (IX86_BUILTIN_PPERM): Ditto.
1517 (IX86_BUILTIN_PERMPS): Ditto.
1518 (IX86_BUILTIN_PERMPD): Ditto.
1519 (IX86_BUILTIN_PMACSSWW): Ditto.
1520 (IX86_BUILTIN_PMACSWW): Ditto.
1521 (IX86_BUILTIN_PMACSSWD): Ditto.
1522 (IX86_BUILTIN_PMACSWD): Ditto.
1523 (IX86_BUILTIN_PMACSSDD): Ditto.
1524 (IX86_BUILTIN_PMACSDD): Ditto.
1525 (IX86_BUILTIN_PMACSSDQL): Ditto.
1526 (IX86_BUILTIN_PMACSSDQH): Ditto.
1527 (IX86_BUILTIN_PMACSDQL): Ditto.
1528 (IX86_BUILTIN_PMACSDQH): Ditto.
1529 (IX86_BUILTIN_PMADCSSWD): Ditto.
1530 (IX86_BUILTIN_PMADCSWD): Ditto.
1531 (IX86_BUILTIN_PHADDBW): Ditto.
1532 (IX86_BUILTIN_PHADDBD): Ditto.
1533 (IX86_BUILTIN_PHADDBQ): Ditto.
1534 (IX86_BUILTIN_PHADDWD): Ditto.
1535 (IX86_BUILTIN_PHADDWQ): Ditto.
1536 (IX86_BUILTIN_PHADDDQ): Ditto.
1537 (IX86_BUILTIN_PHADDUBW): Ditto.
1538 (IX86_BUILTIN_PHADDUBD): Ditto.
1539 (IX86_BUILTIN_PHADDUBQ): Ditto.
1540 (IX86_BUILTIN_PHADDUWD): Ditto.
1541 (IX86_BUILTIN_PHADDUWQ): Ditto.
1542 (IX86_BUILTIN_PHADDUDQ): Ditto.
1543 (IX86_BUILTIN_PHSUBBW): Ditto.
1544 (IX86_BUILTIN_PHSUBWD): Ditto.
1545 (IX86_BUILTIN_PHSUBDQ): Ditto.
1546 (IX86_BUILTIN_PROTB): Ditto.
1547 (IX86_BUILTIN_PROTW): Ditto.
1548 (IX86_BUILTIN_PROTD): Ditto.
1549 (IX86_BUILTIN_PROTQ): Ditto.
1550 (IX86_BUILTIN_PROTB_IMM): Ditto.
1551 (IX86_BUILTIN_PROTW_IMM): Ditto.
1552 (IX86_BUILTIN_PROTD_IMM): Ditto.
1553 (IX86_BUILTIN_PROTQ_IMM): Ditto.
1554 (IX86_BUILTIN_PSHLB): Ditto.
1555 (IX86_BUILTIN_PSHLW): Ditto.
1556 (IX86_BUILTIN_PSHLD): Ditto.
1557 (IX86_BUILTIN_PSHLQ): Ditto.
1558 (IX86_BUILTIN_PSHAB): Ditto.
1559 (IX86_BUILTIN_PSHAW): Ditto.
1560 (IX86_BUILTIN_PSHAD): Ditto.
1561 (IX86_BUILTIN_PSHAQ): Ditto.
1562 (IX86_BUILTIN_FRCZSS): Ditto.
1563 (IX86_BUILTIN_FRCZSD): Ditto.
1564 (IX86_BUILTIN_FRCZPS): Ditto.
1565 (IX86_BUILTIN_FRCZPD): Ditto.
1566 (IX86_BUILTIN_CVTPH2PS): Ditto.
1567 (IX86_BUILTIN_CVTPS2PH): Ditto.
1568 (IX86_BUILTIN_COMEQSS): Ditto.
1569 (IX86_BUILTIN_COMNESS): Ditto.
1570 (IX86_BUILTIN_COMLTSS): Ditto.
1571 (IX86_BUILTIN_COMLESS): Ditto.
1572 (IX86_BUILTIN_COMGTSS): Ditto.
1573 (IX86_BUILTIN_COMGESS): Ditto.
1574 (IX86_BUILTIN_COMUEQSS): Ditto.
1575 (IX86_BUILTIN_COMUNESS): Ditto.
1576 (IX86_BUILTIN_COMULTSS): Ditto.
1577 (IX86_BUILTIN_COMULESS): Ditto.
1578 (IX86_BUILTIN_COMUGTSS): Ditto.
1579 (IX86_BUILTIN_COMUGESS): Ditto.
1580 (IX86_BUILTIN_COMORDSS): Ditto.
1581 (IX86_BUILTIN_COMUNORDSS): Ditto.
1582 (IX86_BUILTIN_COMFALSESS): Ditto.
1583 (IX86_BUILTIN_COMTRUESS): Ditto.
1584 (IX86_BUILTIN_COMEQSD): Ditto.
1585 (IX86_BUILTIN_COMNESD): Ditto.
1586 (IX86_BUILTIN_COMLTSD): Ditto.
1587 (IX86_BUILTIN_COMLESD): Ditto.
1588 (IX86_BUILTIN_COMGTSD): Ditto.
1589 (IX86_BUILTIN_COMGESD): Ditto.
1590 (IX86_BUILTIN_COMUEQSD): Ditto.
1591 (IX86_BUILTIN_COMUNESD): Ditto.
1592 (IX86_BUILTIN_COMULTSD): Ditto.
1593 (IX86_BUILTIN_COMULESD): Ditto.
1594 (IX86_BUILTIN_COMUGTSD): Ditto.
1595 (IX86_BUILTIN_COMUGESD): Ditto.
1596 (IX86_BUILTIN_COMORDSD): Ditto.
1597 (IX86_BUILTIN_COMUNORDSD): Ditto.
1598 (IX86_BUILTIN_COMFALSESD): Ditto.
1599 (IX86_BUILTIN_COMTRUESD): Ditto.
1600 (IX86_BUILTIN_COMEQPS): Ditto.
1601 (IX86_BUILTIN_COMNEPS): Ditto.
1602 (IX86_BUILTIN_COMLTPS): Ditto.
1603 (IX86_BUILTIN_COMLEPS): Ditto.
1604 (IX86_BUILTIN_COMGTPS): Ditto.
1605 (IX86_BUILTIN_COMGEPS): Ditto.
1606 (IX86_BUILTIN_COMUEQPS): Ditto.
1607 (IX86_BUILTIN_COMUNEPS): Ditto.
1608 (IX86_BUILTIN_COMULTPS): Ditto.
1609 (IX86_BUILTIN_COMULEPS): Ditto.
1610 (IX86_BUILTIN_COMUGTPS): Ditto.
1611 (IX86_BUILTIN_COMUGEPS): Ditto.
1612 (IX86_BUILTIN_COMORDPS): Ditto.
1613 (IX86_BUILTIN_COMUNORDPS): Ditto.
1614 (IX86_BUILTIN_COMFALSEPS): Ditto.
1615 (IX86_BUILTIN_COMTRUEPS): Ditto.
1616 (IX86_BUILTIN_COMEQPD): Ditto.
1617 (IX86_BUILTIN_COMNEPD): Ditto.
1618 (IX86_BUILTIN_COMLTPD): Ditto.
1619 (IX86_BUILTIN_COMLEPD): Ditto.
1620 (IX86_BUILTIN_COMGTPD): Ditto.
1621 (IX86_BUILTIN_COMGEPD): Ditto.
1622 (IX86_BUILTIN_COMUEQPD): Ditto.
1623 (IX86_BUILTIN_COMUNEPD): Ditto.
1624 (IX86_BUILTIN_COMULTPD): Ditto.
1625 (IX86_BUILTIN_COMULEPD): Ditto.
1626 (IX86_BUILTIN_COMUGTPD): Ditto.
1627 (IX86_BUILTIN_COMUGEPD): Ditto.
1628 (IX86_BUILTIN_COMORDPD): Ditto.
1629 (IX86_BUILTIN_COMUNORDPD): Ditto.
1630 (IX86_BUILTIN_COMFALSEPD): Ditto.
1631 (IX86_BUILTIN_COMTRUEPD): Ditto.
1632 (IX86_BUILTIN_PCOMEQUB): Ditto.
1633 (IX86_BUILTIN_PCOMNEUB): Ditto.
1634 (IX86_BUILTIN_PCOMLTUB): Ditto.
1635 (IX86_BUILTIN_PCOMLEUB): Ditto.
1636 (IX86_BUILTIN_PCOMGTUB): Ditto.
1637 (IX86_BUILTIN_PCOMGEUB): Ditto.
1638 (IX86_BUILTIN_PCOMFALSEUB): Ditto.
1639 (IX86_BUILTIN_PCOMTRUEUB): Ditto.
1640 (IX86_BUILTIN_PCOMEQUW): Ditto.
1641 (IX86_BUILTIN_PCOMNEUW): Ditto.
1642 (IX86_BUILTIN_PCOMLTUW): Ditto.
1643 (IX86_BUILTIN_PCOMLEUW): Ditto.
1644 (IX86_BUILTIN_PCOMGTUW): Ditto.
1645 (IX86_BUILTIN_PCOMGEUW): Ditto.
1646 (IX86_BUILTIN_PCOMFALSEUW): Ditto.
1647 (IX86_BUILTIN_PCOMTRUEUW): Ditto.
1648 (IX86_BUILTIN_PCOMEQUD): Ditto.
1649 (IX86_BUILTIN_PCOMNEUD): Ditto.
1650 (IX86_BUILTIN_PCOMLTUD): Ditto.
1651 (IX86_BUILTIN_PCOMLEUD): Ditto.
1652 (IX86_BUILTIN_PCOMGTUD): Ditto.
1653 (IX86_BUILTIN_PCOMGEUD): Ditto.
1654 (IX86_BUILTIN_PCOMFALSEUD): Ditto.
1655 (IX86_BUILTIN_PCOMTRUEUD): Ditto.
1656 (IX86_BUILTIN_PCOMEQUQ): Ditto.
1657 (IX86_BUILTIN_PCOMNEUQ): Ditto.
1658 (IX86_BUILTIN_PCOMLTUQ): Ditto.
1659 (IX86_BUILTIN_PCOMLEUQ): Ditto.
1660 (IX86_BUILTIN_PCOMGTUQ): Ditto.
1661 (IX86_BUILTIN_PCOMGEUQ): Ditto.
1662 (IX86_BUILTIN_PCOMFALSEUQ): Ditto.
1663 (IX86_BUILTIN_PCOMTRUEUQ): Ditto.
1664 (IX86_BUILTIN_PCOMEQB): Ditto.
1665 (IX86_BUILTIN_PCOMNEB): Ditto.
1666 (IX86_BUILTIN_PCOMLTB): Ditto.
1667 (IX86_BUILTIN_PCOMLEB): Ditto.
1668 (IX86_BUILTIN_PCOMGTB): Ditto.
1669 (IX86_BUILTIN_PCOMGEB): Ditto.
1670 (IX86_BUILTIN_PCOMFALSEB): Ditto.
1671 (IX86_BUILTIN_PCOMTRUEB): Ditto.
1672 (IX86_BUILTIN_PCOMEQW): Ditto.
1673 (IX86_BUILTIN_PCOMNEW): Ditto.
1674 (IX86_BUILTIN_PCOMLTW): Ditto.
1675 (IX86_BUILTIN_PCOMLEW): Ditto.
1676 (IX86_BUILTIN_PCOMGTW): Ditto.
1677 (IX86_BUILTIN_PCOMGEW): Ditto.
1678 (IX86_BUILTIN_PCOMFALSEW): Ditto.
1679 (IX86_BUILTIN_PCOMTRUEW): Ditto.
1680 (IX86_BUILTIN_PCOMEQD): Ditto.
1681 (IX86_BUILTIN_PCOMNED): Ditto.
1682 (IX86_BUILTIN_PCOMLTD): Ditto.
1683 (IX86_BUILTIN_PCOMLED): Ditto.
1684 (IX86_BUILTIN_PCOMGTD): Ditto.
1685 (IX86_BUILTIN_PCOMGED): Ditto.
1686 (IX86_BUILTIN_PCOMFALSED): Ditto.
1687 (IX86_BUILTIN_PCOMTRUED): Ditto.
1688 (IX86_BUILTIN_PCOMEQQ): Ditto.
1689 (IX86_BUILTIN_PCOMNEQ): Ditto.
1690 (IX86_BUILTIN_PCOMLTQ): Ditto.
1691 (IX86_BUILTIN_PCOMLEQ): Ditto.
1692 (IX86_BUILTIN_PCOMGTQ): Ditto.
1693 (IX86_BUILTIN_PCOMGEQ): Ditto.
1694 (IX86_BUILTIN_PCOMFALSEQ): Ditto.
1695 (IX86_BUILTIN_PCOMTRUEQ): Ditto.
1696 (bdesc_ptest): Change OPTION_MASK_ISA_SSE4_1 to
1697 OPTION_MASK_ISA_ROUND for instructions that are shared between
1699 (bdesc_2arg): Ditto.
1700 (bdesc_sse_3arg): Ditto.
1701 (enum multi_arg_type): New enum for describing the various SSE5
1702 intrinsic argument types.
1703 (bdesc_multi_arg): New table for SSE5 intrinsics.
1704 (ix86_init_mmx_sse_builtins): Add SSE5 intrinsic support.
1705 (ix86_expand_multi_arg_builtin): New function for creating SSE5
1707 (ix86_expand_builtin): Add SSE5 intrinsic support.
1708 (ix86_sse5_valid_op_p): New function to validate SSE5 3 and 4
1709 operand instructions.
1710 (ix86_expand_sse5_multiple_memory): New function to split the
1711 second memory reference from SSE5 instructions.
1712 (type_has_variadic_args_p): Delete in favor of stdarg_p.
1713 (ix86_return_pops_args): Use stdarg_p to determine if the function
1714 has variable arguments.
1715 (ix86_setup_incoming_varargs): Ditto.
1716 (x86_this_parameter): Ditto.
1718 * config/i386/i386-protos.h (ix86_expand_sse5_unpack): Add
1720 (ix86_expand_sse5_pack): Ditto.
1721 (ix86_sse5_valid_op_p): Ditto.
1722 (ix86_expand_sse5_multiple_memory): Ditto.
1724 * config/i386/i386.md (UNSPEC_SSE5_INTRINSIC): Add new UNSPEC
1725 constant for SSE5 support.
1726 (UNSPEC_SSE5_UNSIGNED_CMP): Ditto.
1727 (UNSPEC_SSE5_TRUEFALSE): Ditto.
1728 (UNSPEC_SSE5_PERMUTE): Ditto.
1729 (UNSPEC_SSE5_ASHIFT): Ditto.
1730 (UNSPEC_SSE5_LSHIFT): Ditto.
1731 (UNSPEC_FRCZ): Ditto.
1732 (UNSPEC_CVTPH2PS): Ditto.
1733 (UNSPEC_CVTPS2PH): Ditto.
1734 (PCOM_FALSE): Add new constant for true/false SSE5 comparisons.
1736 (COM_FALSE_S): Ditto.
1737 (COM_FALSE_P): Ditto.
1738 (COM_TRUE_S): Ditto.
1739 (COM_TRUE_P): Ditto.
1740 (type attribute): Add ssemuladd, sseiadd1, ssecvt1, sse4arg types.
1741 (unit attribute): Add support for ssemuladd, ssecvt1, sseiadd1 sse4arg
1743 (memory attribute): Ditto.
1744 (sse4_1_round<mode>2): Use TARGET_ROUND instead of TARGET_SSE4_1.
1745 Use SSE4_1_ROUND_* constants instead of hard coded numbers.
1746 (rint<mode>2): Use TARGET_ROUND instead of TARGET_SSE4_1.
1747 (floor<mode>2): Ditto.
1748 (ceil<mode>2): Ditto.
1749 (btrunc<mode>2): Ditto.
1750 (nearbyintdf2): Ditto.
1751 (nearbyintsf2): Ditto.
1752 (sse_setccsf): Disable if SSE5.
1753 (sse_setccdf): Ditto.
1754 (sse5_setcc<mode>): New support for SSE5 conditional move.
1755 (sse5_pcmov_<mode>): Ditto.
1757 * config/i386/sse.md (SSEMODE1248): New mode iterator for SSE5.
1759 (SSEMODEF2P): Ditto.
1760 (ssemodesuffixf4): New mode attribute for SSE5.
1761 (ssemodesuffixf2s): Ditto.
1762 (ssemodesuffixf2c): Ditto.
1763 (sserotatemax): Ditto.
1764 (ssescalarmode): Ditto.
1765 (sse_maskcmpv4sf3): Disable if SSE5.
1766 (sse_maskcmpv2df3): Ditto.
1767 (sse_vmmaskcmpv4sf3): Ditto.
1768 (sse5_fmadd<mode>4): Add SSE5 floating point multiply/add
1770 (sse5_vmfmadd<mode>4): Ditto.
1771 (sse5_fmsub<mode>4): Ditto.
1772 (sse5_vmfmsub<mode>4): Ditto.
1773 (sse5_fnmadd<mode>4): Ditto.
1774 (sse5_vmfnmadd<mode>4): Ditto.
1775 (sse5_fnmsub<mode>4): Ditto.
1776 (sse5_vmfnmsub<mode>4): Ditto.
1777 (sse5i_fmadd<mode>4): Ditto.
1778 (sse5i_fmsub<mode>4): Ditto.
1779 (sse5i_fnmadd<mode>4): Ditto.
1780 (sse5i_fnmsub<mode>4): Ditto.
1781 (sse5i_vmfmadd<mode>4): Ditto.
1782 (sse5i_vmfmsub<mode>4): Ditto.
1783 (sse5i_vmfnmadd<mode>4): Ditto.
1784 (sse5i_vmfnmsub<mode>4): Ditto.
1785 (mulv16qi3): Add SSE5 support.
1787 (sse5_mulv4si3): New insn for 32-bit multiply support on SSE5.
1788 (sse2_mulv4si3): Disable if SSE5.
1789 (sse4_1_roundpd): Use TARGET_ROUND instead of TARGET_SSE4_1.
1790 (sse4_1_roundps): Ditto.
1791 (sse4_1_roundsd): Ditto.
1792 (sse4_1_roundss): Ditto.
1793 (sse_maskcmpv4sf3): Disable if SSE5 so the SSE5 instruction will
1795 (sse_maskcmpsf3): Ditto.
1796 (sse_vmmaskcmpv4sf3): Ditto.
1797 (sse2_maskcmpv2df3): Ditto.
1798 (sse2_maskcmpdf3): Ditto.
1799 (sse2_vmmaskcmpv2df3): Ditto.
1800 (sse2_eq<mode>3): Ditto.
1801 (sse2_gt<mode>3): Ditto.
1802 (sse5_pcmov_<mode>): Add SSE5 support.
1803 (vec_unpacku_hi_v16qi): Ditto.
1804 (vec_unpacks_hi_v16qi): Ditto.
1805 (vec_unpacku_lo_v16qi): Ditto.
1806 (vec_unpacks_lo_v16qi): Ditto.
1807 (vec_unpacku_hi_v8hi): Ditto.
1808 (vec_unpacks_hi_v8hi): Ditto.
1809 (vec_unpacku_lo_v8hi): Ditto.
1810 (vec_unpacks_lo_v8hi): Ditto.
1811 (vec_unpacku_hi_v4si): Ditto.
1812 (vec_unpacks_hi_v4si): Ditto.
1813 (vec_unpacku_lo_v4si): Ditto.
1814 (vec_unpacks_lo_v4si): Ditto.
1815 (sse5_pmacsww): New SSE5 intrinsic insn.
1816 (sse5_pmacssww): Ditto.
1817 (sse5_pmacsdd): Ditto.
1818 (sse5_pmacssdd): Ditto.
1819 (sse5_pmacssdql): Ditto.
1820 (sse5_pmacssdqh): Ditto.
1821 (sse5_pmacsdqh): Ditto.
1822 (sse5_pmacsswd): Ditto.
1823 (sse5_pmacswd): Ditto.
1824 (sse5_pmadcsswd): Ditto.
1825 (sse5_pmadcswd): Ditto.
1826 (sse5_pcmov_<move>): Conditional move support on SSE5.
1827 (sse5_phaddbw): New SSE5 intrinsic insn.
1828 (sse5_phaddbd): Ditto.
1829 (sse5_phaddbq): Ditto.
1830 (sse5_phaddwd): Ditto.
1831 (sse5_phaddwq): Ditto.
1832 (sse5_phadddq): Ditto.
1833 (sse5_phaddubw): Ditto.
1834 (sse5_phaddubd): Ditto.
1835 (sse5_phaddubq): Ditto.
1836 (sse5_phadduwd): Ditto.
1837 (sse5_phadduwq): Ditto.
1838 (sse5_phaddudq): Ditto.
1839 (sse5_phsubbw): Ditto.
1840 (sse5_phsubwd): Ditto.
1841 (sse5_phsubdq): Ditto.
1842 (sse5_pperm): Ditto.
1843 (sse5_pperm_sign_v16qi_v8hi): New insns for pack/unpack with SSE5.
1844 (sse5_pperm_zero_v16qi_v8hi): Ditto.
1845 (sse5_pperm_sign_v8hi_v4si): Ditto.
1846 (sse5_pperm_zero_v8hi_v4si): Ditto.
1847 (sse5_pperm_sign_v4si_v2di): Ditto.
1848 (sse5_pperm_sign_v4si_v2di): Ditto.
1849 (sse5_pperm_pack_v2di_v4si): Ditto.
1850 (sse5_pperm_pack_v4si_v8hi): Ditto.
1851 (sse5_pperm_pack_v8hi_v16qi): Ditto.
1852 (sse5_perm<mode>): New SSE5 intrinsic insn.
1853 (rotl<mode>3): Ditto.
1854 (sse5_rotl<mode>3): Ditto.
1855 (sse5_ashl<mode>3): Ditto.
1856 (sse5_lshl<mode>3): Ditto.
1857 (sse5_frcz<mode>2): Ditto.
1858 (sse5s_frcz<mode>2): Ditto.
1859 (sse5_cvtph2ps): Ditto.
1860 (sse5_cvtps2ph): Ditto.
1861 (sse5_vmmaskcmp<mode>3): Ditto.
1862 (sse5_com_tf<mode>3): Ditto.
1863 (sse5_maskcmp<mode>3): Ditto.
1864 (sse5_maskcmp_uns<mode>3): Ditto.
1865 (sse5_maskcmp_uns2<mode>3): Ditto.
1866 (sse5_pcom_tf<mode>3): Ditto.
1868 * config/i386/predicates.md (const_0_to_31_operand): New predicate
1870 (sse5_comparison_float_operator): New predicate to match the
1871 comparison operators supported by the SSE5 com instruction.
1872 (ix86_comparison_int_operator): New predicate to match just the
1873 signed int comparisons.
1874 (ix86_comparison_uns_operator): New predicate to match just the
1875 unsigned int comparisons.
1877 * doc/invoke.texi (-msse5): Add documentation.
1878 (-mfused-madd): Ditto.
1880 * doc/extend.texi (x86 intrinsics): Document new SSE5 intrinsics.
1882 * config.gcc (i[34567]86-*-*): Include bmmintrin.h and
1884 (x86_64-*-*): Ditto.
1886 * config/i386/cpuid.h (bit_SSE5): Define SSE5 bit.
1888 * config/i386/bmmintrin.h: New file, provide common x86 compiler
1889 intrinisics for SSE5.
1891 * config/i386/smmintrin.h: Move instructions shared with SSE5 to
1894 * config/i386/mmintrin-common.h: New file, to contain common
1895 instructions between SSE4.1 and SSE5.
1897 * config/i386/netware.c (gen_stdcall_or_fastcall_decoration): Use
1898 FOREACH_FUNCTION_ARGS to iterate over the argument list.
1899 (gen_regparm_prefix): Ditto.
1901 * config/i386/winnt.c (gen_stdcall_or_fastcall_suffix): Use
1902 FOREACH_FUNCTION_ARGS to iterate over the argument list. Use
1903 prototype_p to determine if a function is prototyped.
1905 2007-09-12 Janis Johnson <janis187@us.ibm.com>
1907 * config/dfp-bit.c (dfp_conversion_exception): New function.
1908 (DFP_TO_DFP) Add new variants to use direct conversions in decNumber.
1909 (DFP_TO_INT): Ditto.
1910 (INT_TO_DFP): Ditto.
1911 * config/dfp-bit.h (DEC_FLOAT_FROM_INT, DEC_FLOAT_TO_INT): New.
1913 2007-09-12 Jakub Jelinek <jakub@redhat.com>
1916 * config/ia64/ia64.c (ia64_expand_epilogue): Emit blockage
1917 before sp restoration even when total_size is 0, but
1918 frame_pointer_needed.
1920 2007-09-12 Bob Wilson <bob.wilson@acm.org>
1922 * config/xtensa/xtensa.c (machine_function): Add vararg_a7_copy.
1923 (xtensa_copy_incoming_a7): Use start_sequence instead of
1924 push_to_sequence. Stash insns in vararg_a7_copy for builtin_saveregs.
1925 (xtensa_builtin_saveregs): Place code from vararg_a7_copy at the start
1926 of the saveregs sequence.
1928 2007-09-12 Richard Sandiford <richard@codesourcery.com>
1930 * c-tree.h (grokfield): Add a "tree *" argument.
1931 * c-decl.c (grokdeclarator): Take a pointer to the decl's attributes.
1932 Chain nested decl attributes to it. Don't call decl_attributes here.
1933 (groktypename): Pass grokdeclarator a pointer to the attribute list.
1934 (start_decl, grokparm, push_parm_decl, start_function): Likewise.
1935 (grokfield): Take a pointer to the decl's attributes and pass
1936 it to grokdeclarator.
1937 * c-parser.c (c_parser_struct_declaration): Update the calls to
1938 grokfield. Call decl_attributes for anonymous struct and union
1941 2007-09-12 Jan Hubicka <jh@suse.cz>
1943 * c-objc-common.h (LANG_HOOKS_CALLGRAPH_EXPAND_FUNCTION): Kill.
1945 2007-09-12 Ira Rosen <irar@il.ibm.com>
1947 PR tree-optimization/32377
1948 * tree-vect-analyze.c (vect_analyze_data_ref_dependence): Distinguish
1949 between positive and negative dependence distance using DDR_REVERSED_P.
1951 2007-09-12 Dorit Nuzman <dorit@il.ibm.com>
1953 PR tree-optimization/33373
1954 * tree-vect-analyze (vect_determine_vectorization_factor): Call
1955 TREE_INT_CST_LOW when comparing TYPE_SIZE_UNIT.
1957 2007-09-12 Jan Hubicka <jh@suse.cz>
1960 * i386.md (floatsisf2_mixed_memory, floatsisf2_sse_memory): Disable
1963 2007-09-12 Christian Bruel <christian.bruel@st.com>
1965 * sh.h (SH_DBX_REGISTER_NUMBER): Added fpscr, fixed sr/gbr regs.
1966 * linux-unwind.h (SH_DWARF_FRAME_GBR): fixed.
1968 2007-09-12 Ira Rosen <irar@il.ibm.com>
1970 * tree-vect-transform.c (vect_get_slp_defs): Don't build a vector
1971 for oprnd1 if not required.
1972 (vectorizable_operation): Use scalar operand in SLP in case of
1973 shift with scalar argument.
1975 2007-09-12 Ira Rosen <irar@il.ibm.com>
1977 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change default and minimum
1980 2007-09-11 James E. Wilson <wilson@specifix.com>
1982 * defaults.h (DWARF2_UNWIND_INFO): Don't define if
1983 TARGET_UNWIND_INFO is defined.
1984 * config/ia64/ia64.h (INCOMING_RETURN_ADDR_RTX): Delete undef
1987 2007-09-12 Kaz Kojima <kkojima@gcc.gnu.org>
1989 * config/sh/sh.c (calc_live_regs): Use
1990 current_function_saves_all_registers instead of
1991 current_function_has_nonlocal_label.
1992 (sh_allocate_initial_value): Likewise.
1993 (sh_get_pr_initial_val): Likewise.
1994 * config/sh/sh.h (SHMEDIA_REGS_STACK_ADJUST): Likewise.
1995 * config/sh/sh.md (load_ra): Likewise.
1997 2007-09-12 Hans-Peter Nilsson <hp@axis.com>
1999 * config/cris/t-linux (LIMITS_H_TEST): Only define if not inhibit_libc.
2002 * config/cris/cris.c (cris_expand_pic_call_address): Fix typo in
2003 GET_CODE (x) == CONST_INT to CONST_INT_P (x) transformation.
2005 2007-09-12 Sa Liu <saliu@de.ibm.com>
2007 * config/spu/spu.c (spu_emit_branch_or_set): Handle NaN values as
2008 operands to DFmode GE or LE compares.
2010 2007-09-12 Bernd Schmidt <bernd.schmidt@analog.com>
2012 * config/bfin/bfin.h (enum reg_class, REG_CLASS_CONTENTS,
2013 REG_CLASS_NAMES): Add P0REGS.
2014 (REGNO_REG_CLASS): Return it where appropriate.
2015 (REG_CLASS_FROM_CONSTRAINT): Add 'qA'.
2016 (CLASS_LIKELY_SPILLED_P): P0REGS is likely_spilled.
2017 * doc/md.texi (Blackfin family): Document 'q' constraints.
2019 2007-09-11 Steve Kenton <skenton@ou.edu>
2021 * pa/linux-unwind.h: Guard with inhibit_libc.
2022 * pa/hpux-unwind.h: Likewise.
2024 2007-09-11 David Daney <ddaney@avtrex.com>
2026 * doc/invoke.texi: Document new MIPS -mllsc and -mno-llsc options.
2027 * doc/install.texi: Document new --with-llsc and --without-llsc
2029 * config.gcc: Handle --with-llsc and --without-llsc configure options.
2030 * config/mips/mips.md (sync, memory_barrier): Wrap sync instrunction
2031 in %| and %- operand codes. Depend on GENERATE_SYNC instead of
2033 (sync_compare_and_swap<mode>, sync_add<mode>, sync_sub<mode>,
2034 sync_old_add<mode>, sync_old_sub<mode>, sync_new_add<mode>,
2035 sync_new_sub<mode>, sync_<optab><mode>, sync_old_<optab><mode>,
2036 sync_new_<optab><mode>, sync_nand<mode>, sync_old_nand<mode>,
2037 sync_new_nand<mode>, sync_lock_test_and_set<mode>): Depend on
2038 GENERATE_LL_SC instead of ISA_HAS_LL_SC.
2039 * config/mips/mips.opt (mllsc): New option.
2040 * config/mips/mips.c (mips_llsc): Define variable.
2041 (mips_handle_option): Handle mllsc option.
2042 (override_options): Set mips_print_operand_punct for '|' and '-'.
2043 (print_operand): Add new %| and %- operand codes.
2044 * config/mips/mips.h (mips_llsc_setting): New enum type.
2045 (mips_llsc): Declare.
2046 (OPTION_DEFAULT_SPECS): Add llsc handling.
2047 (GENERATE_SYNC): New macro.
2048 (GENERATE_LL_SC): New macro.
2049 (MIPS_COMPARE_AND_SWAP, MIPS_SYNC_OP, MIPS_SYNC_OLD_OP,
2050 MIPS_SYNC_NEW_OP, MIPS_SYNC_NAND, MIPS_SYNC_OLD_NAND,
2051 MIPS_SYNC_NEW_NAND, MIPS_SYNC_EXCHANGE): Wrap instructions
2052 in %| and %- operand codes.
2054 2007-09-11 Eric Botcazou <ebotcazou@adacore.com>
2056 * tree-ssa-structalias.c (push_fields_onto_fieldstack): Deal with
2057 TYPE_NONALIASED_COMPONENT like with DECL_NONADDRESSABLE_P.
2059 2007-09-11 Jason Merrill <jason@redhat.com>
2062 * stor-layout.c (layout_decl): Do pack variable size fields.
2064 2007-09-11 Maxim Kuvyrkov <maxim@codesourcery.com>
2066 * config/m68k/predicates.md (movsi_const0_operand,
2067 non_symbolic_call_operand): New predicates.
2069 * config/m68k/constraints.md (Cs, Ci, C0, Cj, CQ, CW, CZ, CS, Ap, Ac):
2071 * doc/md.texi (Constraints for Particular Machines: Motorola 680x0):
2072 Document constraints N, O, P, R, S, T, Q, U, W, Cs, Ci, C0, Cj, CQ,
2073 CW, CZ, CS, Ap and Ac.
2075 * config/m68k/m68k.md (UNSPEC_IB): New constant.
2076 (constraints.md): New include.
2077 (cpu, type, type1, opx, opy, opx_type, opy_type, size, opx_access,
2078 opx_mem, opy_mem, op_mem, guess, split): New attributes.
2079 (movdf_internal): Name pattern. Fix to use alternatives. Add split.
2081 (pushdi): Add split.
2082 (tstsi_internal): Name pattern. Fix to use alternatives. Specify
2083 attributes. Split tstsi_internal_68020_cf from it.
2084 (tstsi_internal_68020_cf): New pattern.
2085 (tsthi_internal, tstqi_internal): Name pattern. Specify attributes.
2086 (tst<mode>_cf): Specify attributea.
2087 (cmpsi_cf): Name pattern. Specify attributes.
2088 (cmp<mode>_68881, cmp<mode>_cf): Specify type attribute.
2089 (pushexthisi_const): Fix to use alternatives. Specify
2091 (movsi_const0): Split movsi_const0_68000_10 and movsi_const0_68040_60
2092 from it. Fix to use alternatives. Specify attributes.
2093 (movsi_const0_68040_10, movsi_const0_68040_60): New patterns.
2094 (movsi_cf, movstrictqi_cf): Fix to use alternatives. Specify
2096 (movsf_cf_soft): Specify attributes.
2097 (movdf_cf_soft): Add split.
2098 (pushasi, zero_extendhisi2_cf, zero_extendqisi2_cfv4,
2099 cfv4_extendhisi2, 68k_extendhisi2, extendqihi2, cfv4_extendqisi2,
2100 68k_extendqisi2, truncdfsf2_cf): Specify attributes.
2101 (truncdfsf2_68881): Name pattern. Specify attributes.
2102 (floatsi<mode>2_cf, floathi<mode>2_68881, floathi<mode>2_cf,
2103 floatqi<mode>2_68881, floatqi<mode>2_cf, ftrunc<mode>2_cf,
2104 fix<mode>qi2_cf, fix<mode>hi2_cf, fix<mode>si2_cf, adddi_dishl32):
2106 (addsi3_5200): Fix to use alternatives. Specify attributes.
2108 (add<mode>3_cf, subdi_dishl32): Specify attributes.
2109 (subsi3): Add alternative for subq.l. Specify attributes.
2110 (sub<mode>3_cf, mulhi3, mulhisi3): Specify attributes.
2111 (mulhisisi3_s, mulsi3_68020, mulsi3_cf): Name pattern. Specify
2113 (umulhisi3): Specify attributes.
2114 (mulhisisi3_z): Name pattern. Specify attributes.
2115 (fmul<mode>3_cf, div<mode>3_cf, negsi2_internal, negsi2_5200,
2116 sqrt<mode>2_68881, clzsi2, one_cmplsi2_5200, subreghi1ashrdi_const32,
2117 subregsi1ashrdi_const32, ashrsi3, subreg1lshrdi_const32, lshrsi3,
2118 bsetmemqi): Specify attributes.
2119 (bsetmemqi_ext): Name pattern. Specify attributes.
2120 (bclrmemqi): Specify attributes.
2121 (bclrmemqi_ext, scc, sls): Name pattern. Specify attributes.
2122 (beq, bne, bgt, bgtu, blt, bltu, bge, bgeu, ble, bleu): Specify
2124 (beq2, bne2, bgt2, bgtu2, blt2, bltu2, bge2, bgeu2, ble2, bleu2): Name
2125 pattern. Specify attributes.
2126 (jump): Specify attributes.
2127 (tablejump_internal): Name pattern. Specify attributes.
2128 (call_value): Split into non_symbolic_call_value,
2129 symbolic_call_value_jsr, symbolic_call_value_bsr. Fix to use
2130 alternatives. Specify attributes.
2131 (non_symbolic_call_value, symbolic_call_value_jsr,
2132 symbolic_call_value_bsr): New patterns.
2133 (nop, return, unlink, indirect_jump): Specify attributes.
2134 (trap): Fix condition. Specify attributes.
2137 * config/m68k/m68k.c (m68k_symbolic_call_var): New variable.
2138 (override_options): Initialize it. Initialize m68k_sched_cpu.
2139 (CONST_METHOD): Rename to M68K_CONST_METHOD, move to m68k.h.
2140 (const_method): Make global, rename to m68k_const_method.
2141 (const_int_cost, output_move_const_into_data_reg): Update.
2142 (output_move_double): Parametrize to emit rtl code, rename to
2144 (output_reg_adjust, emit_reg_adjust, output_compadr, output_movsi,
2145 emit_movsi): New static functions.
2146 (output_move_double): New function with semantics of old
2148 (m68k_emit_move_double): New function.
2149 (m68k_sched_cpu): New variable.
2150 (attr_op_type): New enum.
2151 (sched_guess_p): New variable.
2152 (sched_address_type, sched_operand_type, sched_attr_op_type):
2153 New static functions.
2154 (m68k_sched_attr_opx_type, m68k_sched_attr_opy_type,
2155 m68k_sched_attr_size, m68k_sched_attr_op_mem): New functions.
2156 (sched_branch_type): New static variable.
2157 (m68k_sched_branch_type): New function.
2158 * config/m68k/m68k.h (M68K_SYMBOLIC_CALL): New enum.
2159 (m68k_symbolic_call_var): Declare.
2160 (M68K_CONST_METHOD): Rename from CONST_METHOD. Move here from m68k.c.
2161 (m68k_const_method, m68k_emit_move_double, m68k_sched_cpu,
2162 m68k_sched_attr_opx_type, m68k_sched_attr_opy_type,
2163 m68k_sched_attr_size, m68k_sched_attr_op_mem, m68k_sched_branch_type):
2166 2007-09-11 Jakub Jelinek <jakub@redhat.com>
2168 * builtins.def (BUILT_IN_VA_ARG_PACK_LEN): New builtin.
2169 * builtins.c (expand_builtin) <case BUILT_IN_VA_ARG_PACK_LEN>: Issue
2170 error if __builtin_va_arg_pack_len () wasn't optimized out during
2172 * tree-inline.c (copy_bb): Replace __builtin_va_arg_pack_len ()
2173 with the number of inline's anonymous arguments.
2174 * doc/extend.texi: Document __builtin_va_arg_pack_len ().
2176 2007-09-11 Zdenek Dvorak <ook@ucw.cz>
2178 * fold-const.c (extract_muldiv_1): Do not simplify
2181 2007-09-11 Jan Hubicka <jh@suse.cz>
2183 * i386.h (ix86_tune_indices): Add X86_TUNE_INTER_UNIT_CONVERSIONS.
2184 (TARGET_INTER_UNIT_CONVERSIONS): New.
2185 * i386.md (floatsi expanders): Remove redundant check for SImode
2186 source; offload to memory when asked for.
2187 (floatsisf2_mixed, floatsisf2_sse, floatsidf2_mixed, floatsidf2_sse
2188 floatdisf2_mixed, floatsisf2_sse, floatsidf2_mixed, floatsidf2_sse):
2190 (floatsisf2_mixed_memory, floatsisf2_sse_memory,
2191 floatsidf2_mixed_memory, floatsidf2_sse_memory
2192 floatdisf2_mixed_memory, floatsisf2_sse_memory,
2193 floatsidf2_mixed_memory, floatsidf2_sse_memory): New.
2195 2007-09-11 Jan Hubicka <jh@suse.cz>
2197 * toplev.c (process_options): all frontends now do unit-at-a-time.
2198 * cgraphunit.c: update comments.
2199 (cgraph_expand_function): call passmanager dirrectly; emit thunks.
2200 * c-decl.c (finish_function): use cgraph_add_new_function.
2201 * function.c (expand_function_end): We are always unit-at-a-time.
2203 2007-09-11 Richard Sandiford <richard@codesourcery.com>
2205 * config/mips/mips.c (mips_set_mips16_mode): Use separate anchor
2206 settings for MIPS16.
2207 (mips_use_anchors_for_symbol_p): Use default_use_anchors_for_symbol_p.
2209 2007-09-11 Richard Sandiford <richard@codesourcery.com>
2211 * config/mips/mips.c (mips_symbol_insns_1): Allow LEAs of
2212 SYMBOL_FORCE_TO_MEM constants.
2213 (mips_rtx_costs): Give a cost of 1 to force_to_mem_operands.
2214 (mips16_rewrite_pool_refs_info): New structure.
2215 (mips16_rewrite_pool_constant): New function, split out from...
2216 (mips16_rewrite_pool_refs): ...here. Take a pointer to a
2217 mips16_rewrite_pool_refs_info structure rather than a pointer
2218 to a constant pool. Force force_to_mem_operands into memory.
2219 (mips16_lay_out_constants): Update call to mips16_rewrite_pool_refs.
2220 * config/mips/predicates.md (force_to_mem_operand): New predicate.
2221 * config/mips/constraints.md (kf): New constraint.
2222 * config/mips/mips.md (*movdi_64bit_mips16): Add a d <- kf alternative.
2223 (*movsi_mips16): Likewise.
2225 2007-09-11 Richard Sandiford <richard@codesourcery.com>
2226 Nigel Stephens <nigel@mips.com>
2227 David Ung <davidu@mips.com>
2229 * config/mips/mips.h (CONSTANT_POOL_COST): Move to...
2230 * config/mips/mips.c: ...here and set to 4 for TARGET_MIPS16.
2231 (mips16_constant_cost, mips_immediate_operand_p, mips_binary_cost)
2232 (mips_fp_mult_cost, mips_fp_div_cost, mips_sign_extend_cost)
2233 (mips_zero_extend_cost): New functions.
2234 (mips_rtx_costs): Treat COMPARE constants as having zero cost.
2235 Use the new functions. Tweak many cost estimates, both here
2236 and in the new subroutines. Return false when the cost of the
2237 operands has not been calculated. Check for *clear_upper32.
2238 Check for floating-point multiply-add, reciprocal and rsqrt
2239 patterns. Handle comparison and rotation codes.
2241 2007-09-11 Danny Smith <dannysmith@users.sourceforge.net>
2243 * config/i386/cygming.h (TARGET_STRIP_NAME_ENCODING): Don't
2245 * config/i386/i386.c (get_dllimport_decl): Don't strip
2248 2007-09-10 Janis Johnson <janis187@us.ibm.com>
2251 * config/dfp-bit.c: Don't skip TFmode conversions; move strto*
2252 declarations to top.
2253 (DFP_TO_BFP): Use for either XFmode or TFmode.
2254 (BFP_TO_DFP): Use for either XFmode or TFmode; always use cast
2256 * config/dfp-bit.h: Include float.h.
2257 (LONG_DOUBLE_HAS_XF_MODE, LONG_DOUBLE_HAS_TF_MODE): Define if long
2258 double is one of these modes, rather than using LIBGCC_HAS_*F_MODE
2259 which doesn't mean the same thing.
2260 (BFP_KIND): Use 4 to mean TFmode.
2261 (BFP_FMT): Specify the number of decimal digits based on the
2262 number of mantissa digits.
2263 (BFP_VIA_TYPE): Binary float type to use as cast for sprintf.
2264 (BFP_TO_DFP, DFP_TO_BFP): Define names for TFmode variants.
2265 (STR_TO_BFP): Use strtold for XFmode or TFmode.
2266 (TFtype): Define if TFmode is supported.
2267 * doc/libgcc.texi (Decimal float library routines): Document
2268 TF conversion functions.
2270 2007-09-10 Chao-ying Fu <fu@mips.com>
2272 * config/mips/mips.c (mips_scalar_mode_supported_p): Declare.
2273 (TARGET_SCALAR_MODE_SUPPORTED_P): Define.
2274 (mips_emit_compare): Process fixed-point modes.
2275 (mips_pad_arg_upward): Support fixed-point types.
2276 (override_options): Allow fixed-point modes in accumulators.
2277 (mips_pass_by_reference): Pass DQ, UDQ, DA, and UDA modes in registers.
2278 (mips_vector_mode_supported_p): Support V2HQmode, V2UHQmode, V2HAmode,
2279 V2UHAmode, V4QQmode, and V4UQQmode when TARGET_DSP.
2280 (mips_scalar_mode_supported_p): New function to accept fixed-point
2281 modes if the width is not greater than two BITS_PER_WORD.
2282 * config/mips/mips.h (SHORT_FRACT_TYPE_SIZE, FRACT_TYPE_SIZE,
2283 LONG_FRACT_TYPE_SIZE, LONG_LONG_FRACT_TYPE_SIZE,
2284 SHORT_ACCUM_TYPE_SIZE, ACCUM_TYPE_SIZE, LONG_ACCUM_TYPE_SIZE,
2285 LONG_LONG_ACCUM_TYPE_SIZE): Define.
2286 * config/mips/mips.md ("d"): Update mode attribute for fixed-point
2288 ("IMODE"): New mode attribute.
2289 (mips-fixed.md): Include.
2290 * config/mips/mips-modes.def: Create VECTOR_MODES for FRACT, UFRACT,
2292 * config/mips/mips-fixed.md: New file.
2294 2007-09-11 Ben Elliston <bje@au.ibm.com>
2296 * config/spu/spu.md: Formatting fixes.
2298 2007-09-10 Janis Johnson <janis187@us.ibm.com>
2300 * config/dfp-bit.c (dfp_unary_func): Delete.
2301 (dfp_unary_op): Delete.
2302 (dfp_binary_op): Use decFloat functions instead of decNumber
2303 functions for binary operations.
2304 (d32_binary_op): Convert 32-bit operands to 64 bits for evaluation.
2305 (dnn_binary_op): Call dfp_binary_op with decFloat rather than
2307 (dfp_compare_op): Use decFloat functions insteadof decNumber
2308 functions for comparisons.
2309 (d32_compare_op): Convert 32-bit operands to 64 bits for evaluation.
2310 (dnn_binary_op): Call dfp_compare_op with decFloat rather than
2312 (DFP_ADD, DFP_SUB, DFP_MULTIPLE, DFP_DIVIDE): Use macros for
2313 call to dxx_binary_op and decFloat function.
2314 (DFP_EQ, DFP_NE, DFP_LT, DFP_GT, DFP_LE, DFP_GE): Use macros for
2315 calls to dxx_binary_op and decFloat function.
2316 * config/dfp-bit.h: Include decFloat header files.
2317 (decFloat, DFP_BINARY_OP, DFP_COMPARE_OP, DEC_FLOAT_ADD,
2318 DEC_FLOAT_SUBTRACT, DEC_FLOAT_MULTIPLY, DEC_FLOAT_DIVIDE,
2319 DEC_FLOAT_COMPARE, DEC_FLOAT_IS_ZERO, DEC_FLOAT_IS_NAN,
2320 DEC_FLOAT_IS_SIGNED: Define for each of 3 operand widths.
2322 2007-09-10 Harsha Jagasia <harsha.jagasia@amd.com>
2323 Jan Sjodin <jan.sjodin@amd.com>
2325 * tree-vect-analyze.c (vect_analyze_operations): Change
2326 comparison of loop iterations with threshold to less than
2327 or equal to instead of less than. Reduce
2328 min_scalar_loop_bound by one.
2329 * tree-vect-transform.c (vect_estimate_min_profitable_iters):
2330 Change prologue and epilogue iterations estimate to vf/2,
2331 when unknown at compile-time. Change versioning guard
2332 cost to taken_branch_cost. If peeling for alignment is
2333 unknown at compile-time, change peel guard costs to one
2334 taken branch and one not-taken branch per peeled loop.
2335 If peeling for alignment is known but number of scalar loop
2336 iterations is unknown at compile-time, change peel guard
2337 costs to one taken branch per peeled loop. Change the cost
2338 model equation to consider vector iterations as the loop
2339 iterations less the prologue and epilogue iterations.
2340 Change outside vector cost check to less than or equal to
2341 zero instead of equal to zero.
2342 (vect_do_peeling_for_loop_bound): Reduce
2343 min_scalar_loop_bound by one.
2344 * tree-vectorizer.h: Add TARG_COND_TAKEN_BRANCH_COST and
2345 TARG_COND_NOT_TAKEN_BRANCH_COST.
2346 * config/i386/i386.h (processor_costs): Add
2347 scalar_stmt_cost, scalar_load_cost, scalar_store_cost,
2348 vec_stmt_cost, vec_to_scalar_cost, scalar_to_vec_cost,
2349 vec_align_load_cost, vect_unalign_load_cost,
2350 vec_store_cost, cond_taken_branch_cost,
2351 cond_not_taken_branch_cost.
2352 Define macros for x86 costs.
2353 * config/i386/i386.c:
2354 (size_cost): Set scalar_stmt_cost, scalar_load_cost,
2355 scalar_store_cost, vec_stmt_cost, vec_to_scalar_cost,
2356 scalar_to_vec_cost, vec_align_load_cost,
2357 vect_unalign_load_cost, vec_store_cost,
2358 cond_taken_branch_cost, cond_not_taken_branch_cost to one.
2359 (i386_cost, i486_cost, pentium_cost, pentiumpro_cost,
2360 geode_cost, k6_cost, athlon_cost, pentium4_cost, nocona_cost,
2361 core2_cost, generic64_cost, generic32_cost): Set to default
2363 (k8_cost, amdfam10_cost): Costs for vectorization tuned.
2364 (x86_builtin_vectorization_cost): New.
2366 2007-09-10 Janis Johnson <janis187@us.ibm.com>
2367 Ben Elliston <bje@au.ibm.com>
2369 * dfp.c: Include decimal128Local.h;
2370 (dfp_byte_swap): Remove.
2371 (encode_decimal32, decode_decimal32): Don't handle endianness.
2372 (encode_decimal64, decode_decimal64): Ditto.
2373 (encode_decimal128, decode_decimal128): Ditto.
2374 * config/dfp-bit.c (host_to_ieee32, ieee_to_host_32): Ditto.
2376 (host_to_ieee_64, ieee_to_host_64): Don't handle endianness.
2378 (host_to_ieee_128, ieee_to_host_128): Don't handle endianness.
2379 * Makefile.in (DECNUM_H): Add decimal128Local.h.
2381 2007-09-10 David Daney <ddaney@avtrex.com>
2383 * config/mips/mips.md (UNSPEC_MEMORY_BARRIER): New entry in
2385 (memory_barrier): Rewrote as an insn that clobbers memory.
2387 2007-09-10 Richard Sandiford <richard@codesourcery.com>
2389 * config/mips/mips.c (mips_global_pointer): Check
2390 call_really_used_regs instead of call_used_regs.
2391 (mips_save_reg_p): Likewise. Save all call-saved registers
2392 if current_function_saves_all_registers. Fix indentation.
2393 No longer treat $18 as a special case.
2394 (compute_frame_size): Guard FPR loop with TARGET_HARD_FLOAT.
2396 2007-09-10 Richard Sandiford <richard@codesourcery.com>
2398 * config/mips/mips.h (MIPS_ARCH_FLOAT_SPEC): New macro.
2399 * config/mips/mips.c (mips_cpu_info_table): Mention it in the
2400 the introductory comment.
2401 (MIPS_MARCH_CONTROLS_SOFT_FLOAT): Delete.
2402 (override_options): Don't test for it.
2403 * config/mips/sde.h (MIPS_MARCH_CONTROLS_SOFT_FLOAT): Delete.
2404 (DRIVER_SELF_SPECS): Add MIPS_ARCH_FLOAT_SPEC.
2405 * config/mips/vr.h: As for config/mips/sde.h.
2407 2007-09-10 Trevor Smigiel <trevor_smigiel@playstation.sony.com>
2408 Revital Eres <eres@il.ibm.com>
2410 * target.h (struct gcc_target.sched): New field: sms_res_mii.
2411 (struct ddg): Define.
2412 * target-def.h (TARGET_SCHED_SMS_RES_MII): Define.
2413 (TARGET_SCHED): Add TARGET_SCHED_SMS_RES_MII.
2414 * config/spu/spu.c: Include ddg.h.
2415 (TARGET_SCHED_SMS_RES_MII): Define.
2416 (spu_sms_res_mii): New function to calculate mii.
2417 * modulo-sched (res_MII): Use it.
2418 * doc/tm.texi: Document TARGET_SCHED_SMS_RES_MII.
2420 2007-09-10 Andreas Krebbel <krebbel1@de.ibm.com>
2422 * config/s390/s390.c (s390_dump_pool): Create copy of constant
2423 pool entries since they might hold values that must not be shared.
2425 2007-09-10 Uros Bizjak <ubizjak@gmail.com>
2428 * gcc/config/i386/sse.md (ashr<mode>3): Change op2 mode to SImode.
2429 Use 'N' operand constraint for op2.
2430 (lshr<mode>3): Ditto.
2431 (ashl<mode>3): Ditto.
2432 (vec_shl_<mode>): Use const_0_to_255_mul_8_operand predicate for op2.
2433 (vec_shr_<mode>): Ditto.
2434 * gcc/config/i386/i386.c (ix86_expand_builtin) [IX86_BUILTIN_PSLL?128,
2435 IX86_BUILTIN_PSRA*?128, IX86_BUILTIN_PSRL?128]: Convert op1 to SImode.
2437 2007-09-10 Andreas Krebbel <krebbel1@de.ibm.com>
2439 * config/s390/s390.md ("fixuns_trunc<BFP:mode><GPR:mode>2"):
2440 Change mode macro in the last real_2expN parameter to uppercase.
2442 2007-09-10 Michael Matz <matz@suse.de>
2444 * tree-pass.h (pass_cselim): Declare new pass.
2445 * passes.c (init_optimization_passes): Link in pass_cselim.
2446 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Renamed from
2447 tree_ssa_phiopt; add do_store_elim parameter, handle it by calling
2448 cond_store_replacement.
2449 (condstoretemp): New static variable.
2450 (cond_store_replacement): New function.
2451 (tree_ssa_phiopt, tree_ssa_cs_elim): New wrappers around
2452 tree_ssa_phiopt_worker.
2453 (struct name_to_bb): New.
2454 (get_non_trapping, name_to_bb_hash, name_to_bb_eq, add_or_mark_expr,
2455 nt_init_block, nt_fini_block): New static functions.
2456 (seen_ssa_names, nontrap_set): New static variables.
2457 (gate_cselim, pass_cselim): Define new pass.
2458 * common.opt (ftree-cselim): New flag.
2459 * toplev.c (process_options): Set flag_tree_cselim if required.
2461 2007-09-10 Hans-Peter Nilsson <hp@axis.com>
2463 * simplify-rtx.c (simplify_relational_operation_1): For recent
2464 canonicalization, don't recurse if op1 equals both PLUS arguments.
2466 2007-09-09 David Daney <ddaney@avtrex.com>
2468 * optabs.c (expand_sync_operation): Use plus insn if minus
2470 (expand_sync_fetch_operation): Ditto.
2472 2007-09-09 H.J. Lu <hongjiu.lu@intel.com>
2474 * i386.md (*floatsisf2_mixed_vector): Use cvtdq2ps instead
2476 (*floatsisf2_sse_vector): Likewise.
2478 2007-09-09 Krister Walfridsson <cato@df.lth.se>
2480 * config/netbsd.h (HANDLE_PRAGMA_PACK_PUSH_POP): Define to 1.
2482 2007-09-09 H.J. Lu <hongjiu.lu@intel.com>
2484 * config/i386/i386.h (ix86_tune_indices): Rename
2485 X86_USE_VECTOR_CONVERTS to X86_TUNE_USE_VECTOR_CONVERTS.
2486 (TARGET_USE_VECTOR_CONVERTS): Updated.
2487 * config/i386/i386.c: Likewise.
2489 2007-09-09 Sandra Loosemore <sandra@codesourcery.com>
2490 Nigel Stephens <nigel@mips.com>
2492 * doc/invoke.texi (Overall Options): Add .sx file extension
2493 as a synonym for .S.
2494 * cppspec.c (known_suffixes): Likewise.
2495 * gcc.c (default_compilers): Likewise.
2497 2007-09-09 Rask Ingemann Lambertsen <rask@sygehus.dk>
2500 * config/i386/i386.h (CANONICALIZE_COMPARISON): Delete.
2501 * simplify-rtx.c (simplify_relational_operation_1): Add the
2502 canonicalization from i386.h.
2503 * doc/md.texi (Canonicalization of Instructions): Document it.
2505 2007-09-09 Jan Hubicka <jh@suse.cz>
2506 Dwarakanath Rajagopal <dwarak.rajagopal@amd.com>
2508 * i386.h (ix86_tune_indices): Add X86_USE_VECTOR_CONVERTS.
2509 (TARGET_USE_VECTOR_CONVERTS): New.
2510 * i386.md: New post-reload splitters for converting SF to DF and DF to
2512 (floatsi* expander): Special case vector conversions.
2513 (floatsisf2_mixed_vector, floatsisf2_sse_vector_nointernunit,
2514 floatsisf2_sse_vector_internunit, floatsisf2_sse_vector,
2515 floatsidf2_mixed_vector, floatsidf2_sse_vector): New.
2516 (floatsisf2_mixed, floatsisf2_sse, floatsidf2_mixed, floatsidf2_sse):
2517 Disable when doing vector converts.
2518 (floatsi<mode>2_i387): Disable when
2519 * sse.md (vec_dupv2df): Export.
2520 * i386.c (ix86_tune_features): Enable SSE conversions.
2522 2007-09-09 Richard Guenther <rguenther@suse.de>
2524 * tree-ssa-operands.c (add_virtual_operand): Only mark
2525 stores as has_volatile_ops if alias information is not available.
2527 2007-09-09 Revital Eres <eres@il.ibm.com>
2529 * doc/contrib.texi: Add myself.
2531 2007-09-09 Ira Rosen <irar@il.ibm.com>
2533 * tree-vectorizer.h (stmt_vinfo_set_inside_of_loop_cost,
2534 stmt_vinfo_set_outside_of_loop_cost): New functions.
2535 * tree-vect-transform.c (vect_get_cost_fields): Remove.
2536 (vect_model_simple_cost): Call
2537 stmt_vinfo_set_inside/outside_of_loop_cost to set the relevant cost
2538 field instead of calling vect_get_cost_fields.
2539 (vect_model_store_cost, vect_model_load_cost): Likewise.
2541 2007-09-09 Revital Eres <eres@il.ibm.com>
2543 * config/rs6000/rs6000.c (paired_init_builtins): Add const
2544 declaration to bdesc_paired_preds variable.
2545 (paired_expand_builtin): Likewise.
2547 2007-09-09 Revital Eres <eres@il.ibm.com>
2549 * dbgcnt.def (sms_sched_loop): New counter.
2550 * modulo-sched.c: Use sms_sched_loop instead of
2551 MAX_SMS_LOOP_NUMBER to determine the maximum number of loops to
2552 perform swing modulo scheduling on. Include dbgcnt.h.
2553 * Makefile.in: Add DBGCNT_H to modulo-sched.o.
2554 * params.def: Remove PARAM_MAX_SMS_LOOP_NUMBER.
2556 2007-09-09 Uros Bizjak <ubizjak@gmail.com>
2558 * config/i386/i386.md (X87MODEF12, SSEMODEF): Remove mode iterators.
2559 Substitute all uses with ...
2560 (MODEF): New mode iterator.
2562 (fix_trunc<mode>_fisttp_i387_1): Remove operand constraints
2563 from pre-regalloc define_insn_and_split splitter pattern.
2564 (*fix_trunc<mode>_i387_1): Ditto.
2565 (*fistdi2_1): Ditto.
2566 (*fist<mode>2_1): Ditto.
2567 (frndintxf2_floor): Ditto.
2568 (*fist<mode>2_floor_1): Ditto.
2569 (frndintxf2_ceil): Ditto.
2570 (*fist<mode>2_ceil_1): Ditto.
2571 (frndintxf2_trunc): Ditto.
2572 (frndintxf2_mask_pm): Ditto.
2574 (prologue): Use (const_int 0) as never generated filler insn.
2576 (sibcall_epilogue): Ditto.
2577 (eh_return_si): Ditto.
2578 (eh_return_di): Ditto.
2580 (add<mode>3): Rename from adddf3 and addsf3. Macroize expander
2581 using MODEF mode iterator.
2582 (sub<mode>3): Rename from subdf3 and subsf3. Macroize expander
2583 using MODEF mode iterator.
2584 (mul<mode>3): Rename from muldf3 and mulsf3. Macroize expander
2585 using MODEF mode iterator.
2586 (nearbyint<mode>2): Rename from nearbyintdf2 and nearbyintsf2.
2587 Macroize expander using MODEF mode iterator.
2589 (zero_extendsidi2): Remove operand constraints from expander.
2590 (smuldi3_highpart): Ditto.
2591 (indirect_jump): Ditto.
2594 * config/i386/sse.md (storentv4sf): Ditto.
2595 (storentv2df): Ditto.
2596 (storentv2di): Ditto.
2598 (sse2_cvtpd2ps): Ditto.
2599 (vec_interleave_highv16qi): Ditto.
2600 (vec_interleave_lowv16qi): Ditto.
2601 (vec_interleave_highv8hi): Ditto.
2602 (vec_interleave_lowv8hi): Ditto.
2603 (vec_interleave_highv4si): Ditto.
2604 (vec_interleave_lowv4si): Ditto.
2605 (vec_interleave_highv2di): Ditto.
2606 (vec_interleave_lowv2di): Ditto.
2607 (sse2_maskmovdqu): Ditto.
2608 * config/i386/mmx.md (mmx_maskmovq): Ditto.
2610 2007-09-09 Ira Rosen <irar@il.ibm.com>
2612 * tree-vectorizer.h (enum vect_def_type): Start enumeration from 1.
2613 (struct _slp_tree, struct _slp_instance): Define new data structures
2614 along macros for their access.
2615 (struct _loop_vec_info): Define new fields: strided_stores,
2616 slp_instances, and slp_unrolling_factor along macros for their access.
2617 (enum slp_vect_type): New.
2618 (struct _stmt_vec_info): Define new field, slp_type, and macros for
2620 (STMT_VINFO_STRIDED_ACCESS): New macro.
2621 (vect_free_slp_tree): Declare.
2622 (vectorizable_load): Add an argument of type slp_tree.
2623 (vectorizable_store, vectorizable_operation, vectorizable_conversion,
2624 vectorizable_assignment): Likewise.
2625 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2626 Declare (make extern).
2627 * tree-vectorizer.c (new_stmt_vec_info): Initialize the new field.
2628 (new_loop_vec_info): Likewise.
2629 (destroy_loop_vec_info): Free memory allocated for SLP structures.
2630 * tree-vect-analyze.c: Include recog.h.
2631 (vect_update_slp_costs_according_to_vf): New.
2632 (vect_analyze_operations): Add argument for calls to vectorizable_ ()
2633 functions. For not pure SLP stmts with strided access check that the
2634 group size is power of 2. Update the vectorization factor according to
2635 SLP. Call vect_update_slp_costs_according_to_vf.
2636 (vect_analyze_group_access): New.
2637 (vect_analyze_data_ref_access): Call vect_analyze_group_access.
2638 (vect_free_slp_tree): New functions.
2639 (vect_get_and_check_slp_defs, vect_build_slp_tree, vect_print_slp_tree,
2640 vect_mark_slp_stmts, vect_analyze_slp_instance, vect_analyze_slp,
2641 vect_make_slp_decision, vect_detect_hybrid_slp_stmts,
2642 vect_detect_hybrid_slp): Likewise.
2643 (vect_analyze_loop): Call vect_analyze_slp, vect_make_slp_decision
2644 and vect_detect_hybrid_slp.
2645 * tree-vect-transform.c (vect_estimate_min_profitable_iters): Take
2646 SLP costs into account.
2647 (vect_get_cost_fields): New function.
2648 (vect_model_simple_cost): Make extern, add SLP parameter and handle
2650 (vect_model_store_cost, vect_model_load_cost): Likewise.
2651 (vect_get_constant_vectors): New function.
2652 (vect_get_slp_vect_defs, vect_get_slp_defs,
2653 vect_get_vec_defs_for_stmt_copy, vect_get_vec_defs_for_stmt_copy,
2654 vect_get_vec_defs): Likewise.
2655 (vectorizable_reduction): Don't handle SLP for now.
2656 (vectorizable_call): Don't handle SLP for now. Add argument to
2657 vect_model_simple_cost.
2658 (vectorizable_conversion): Handle SLP (call vect_get_vec_defs to
2659 get SLPed and vectorized defs). Fix indentation and spacing.
2660 (vectorizable_assignment): Handle SLP.
2661 (vectorizable_induction): Don't handle SLP for now.
2662 (vectorizable_operation): Likewise.
2663 (vectorizable_type_demotion): Add argument to
2664 vect_model_simple_cost.
2665 (vectorizable_type_promotion): Likewise.
2666 (vectorizable_store, vectorizable_load): Handle SLP.
2667 (vectorizable_condition): Don't handle SLP for now.
2668 (vect_transform_stmt): Add a new argument for SLP. Check that there is
2669 no SLP transformation required for unsupported cases. Add SLP
2670 argument for supported cases.
2671 (vect_remove_stores): New function.
2672 (vect_schedule_slp_instance, vect_schedule_slp): Likewise.
2673 (vect_transform_loop): Schedule SLP instances.
2674 * Makefile.in: (tree-vect-analyze.o): Depend on recog.h.
2676 2007-09-09 Andrew Haley <aph@redhat.com>
2678 * optabs.c (sign_expand_binop): Set libcall_gen = NULL in the fake
2681 2007-09-09 Hans-Peter Nilsson <hp@axis.com>
2683 Divide REG_LABEL notes into REG_LABEL_OPERAND and REG_LABEL_TARGET.
2684 * doc/rtl.texi (Insns): Specify when a label_ref makes a jump_insn.
2685 Similar for what label_refs can go in the JUMP_TARGET field. Split
2686 REG_LABEL documentation into REG_LABEL_TARGET and REG_LABEL_OPERAND.
2687 * reload.c (find_reloads): Generate REG_LABEL_OPERAND, not
2688 REG_LABEL when replacing an operand with a LABEL_REF for a
2690 (subst_reloads): When replacing a LABEL_REG with a register,
2691 instead of generating a REG_LABEL note, assert that there already
2692 is one or that the label is a known target for the insn.
2693 * rtlanal.c (computed_jump_p): Instead of looking for a REG_LABEL
2694 note, check the JUMP_LABEL field. Remove "else" after return.
2695 * reorg.c (emit_delay_sequence): Replace case for REG_LABEL with
2696 cases for REG_LABEL_OPERAND and REG_LABEL_TARGET.
2697 (fill_slots_from_thread): Handle both REG_LABEL_OPERAND and
2698 REG_LABEL_TARGET notes, including the JUMP_TARGET field on JUMP_P
2699 insns. Iterate over all notes; don't assume there's only one.
2700 * cse.c (recorded_label_ref): Adjust comment to refer to
2702 (cse_extended_basic_block): Do LABEL_REF check for all INSN_P
2703 insns, not just NONJUMP_INSN_P.
2704 (check_for_label_ref): For JUMP_P insns, check that the LABEL_REF
2705 isn't a jump target.
2706 * jump.c (rebuild_jump_labels): Adjust head comment.
2707 (init_label_info): Ditto. Remove REG_LABEL_OPERAND notes only;
2708 don't reset REG_LABEL_TARGET notes, including the JUMP_LABEL field.
2709 (mark_all_labels): For JUMP_P insns without a target, check if the
2710 the target is noted on the previous nonjump insn.
2711 (mark_jump_label_1): New function, guts from mark_jump_label.
2712 <case IF_THEN_ELSE>: Handle first operand as a non-target when
2713 marking jump target labels.
2714 <case LABEL_REF>: Adjust for whether to generate a
2715 REG_LABEL_TARGET or a REG_LABEL_OPERAND note.
2716 For 'E' format rtl, iterate in descending element order.
2717 (delete_related_insns): Handle both REG_LABEL_TARGET and
2718 REG_LABEL_OPERAND notes. For JUMP_P insns with labels with zero
2719 reference count, delete and fallthrough. Move finding-next-
2720 non-deleted insn last in the function. Look at all INSN_P insns
2721 for REG_LABEL_OPERAND notes.
2722 (redirect_jump_2): Assert that OLABEL equals the old JUMP_LABEL of
2724 * print-rtl.c (print_rtx): For JUMP_P insns and a non-NULL
2725 JUMP_LABEL, output the INSN_UID of it.
2726 * gcse.c: Adjust comments as appropriate to say REG_LABEL_OPERAND
2727 and/or REG_LABEL_TARGET.
2728 (add_label_notes): Only add REG_LABEL_OPERAND notes. Put in line
2729 with jump.c copy by only adding notes for labels actually
2730 referenced in the insn.
2731 * emit-rtl.c (try_split): Don't assume only NONJUMP_INSN_P need
2732 usage count increment; handle all INSN_P trial insns.
2733 (emit_copy_of_insn_after): Change to not copy REG_LABEL_OPERAND
2735 * rtl.h (struct rtx_def) <volatil>: Adjust to mention
2736 REG_LABEL_TARGET and REG_LABEL_OPERAND.
2737 (LABEL_REF_NONLOCAL_P): Allow REG_LABEL_TARGET and
2739 * combine.c (distribute_notes): Adjust for REG_LABEL_TARGET on
2740 JUMP_P insns and REG_LABEL_OPERAND everywhere.
2741 * sched-rgn.c (is_cfg_nonregular): Check for REG_LABEL_OPERANDS
2742 on all INSN_P insns.
2743 * reg-notes.def (LABEL_TARGET, LABEL_OPERAND): Split from LABEL.
2744 * cfgrtl.c (delete_insn): Adjust to handle REG_LABEL_TARGET and
2745 REG_LABEL_OPERAND notes.
2746 * reload1.c (calculate_needs_all_insns): Adjust comments.
2747 (set_label_offsets): Adjust to look for REG_LABEL_OPERAND notes.
2748 * config/alpha/alpha.md (split for load of an address into a
2749 four-insn sequence on Unicos/Mk): Adjust to use
2751 * config/sh/sh.md (sh_reorg, final_prescan_insn): Ditto.
2753 2007-09-09 Laurynas Biveinis <laurynas.biveinis@gmail.com>
2756 2007-09-05 Laurynas Biveinis <laurynas.biveinis@gmail.com>
2757 * regrename.c (copyprop_hardreg_forward_1): New variable next. Use
2758 FOR_BB_INSNS_SAFE instead of for loop.
2759 * cse.c (cse_extended_basic_block): Likewise.
2760 * postreload.c (reload_cse_regs_1): New variable next. Make sure
2761 that the for loop does not invoke NEXT_INSN on a deleted insn.
2762 * function.c (instantiate_virtual_regs): Likewise.
2763 * lower-subreg.c (remove_retval_note): Likewise.
2764 (decompose_multiword_subregs): Use FOR_BB_INSNS_SAFE instead of
2766 * emit-rtl.c (remove_insn): Set NEXT_INSN and PREV_INSN to NULL on
2768 * cfgrtl.c (delete_insn): Set JUMP_LABEL to NULL on a deleted
2769 insn, if it's a jump.
2770 (try_redirect_by_replacing_jump): New variable jump_p. Call
2771 tablejump_p before delete_insn_chain.
2772 * reload1.c (reload): New variable next. Make sure that the for
2773 loop does not invoke NEXT_INSN on a deleted insn.
2774 (fixup_eh_region_note): Make the loop terminate if i becomes NULL.
2775 (delete_output_reload): New variable prev. Make sure the the for
2776 loops do not invoke PREV_INSN on a deleted insn.
2778 2007-09-08 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2780 * pa/constraints.md: Remove 'X' from unused letters comment.
2782 2007-09-08 Richard Guenther <rguenther@suse.de>
2784 * tree-tailcall.c (find_tail_calls): If we don't have aliases
2785 computed check stmt_ann->references_memory instead of counting
2788 2007-09-08 Segher Boessenkool <segher@kernel.crashing.org>
2790 * cse.c (fold_rtx): Use validate_unshare_change() instead of
2791 validate_change() in one more case.
2793 2007-09-08 Zdenek Dvorak <ook@ucw.cz>
2795 PR tree-optimization/32283
2796 * tree-ssa-loop-ivopts.c (may_eliminate_iv): Use
2797 estimated_loop_iterations.
2798 (determine_use_iv_cost_condition): Decrease cost of expressions
2799 used in iv elimination.
2801 2007-09-08 Richard Guenther <rguenther@suse.de>
2803 * tree-cfg.c (verify_gimple_expr): Avoid building new
2804 pointer types, use TYPE_POINTER_TO if available instead.
2806 2007-09-08 Uros Bizjak <ubizjak@gmail.com>
2809 PR rtl-optimization/26449
2810 * config/i386/sse.md (mulv4si3): Do not expand sse2 sequence.
2811 (*sse2_mulv4si3): New define_insn_and_split pattern. Split insn in
2813 (mulv16qi3): Implement as define_insn_and_split pattern instead of
2814 define_expand. Split insn in split1 pass.
2817 2007-09-08 Dorit Nuzman <dorit@il.ibm.com>
2819 PR tree-optimization/33301
2820 * tree-vect-analyze (analyze_operations): Look at the type of the rhs
2823 2007-09-07 Zdenek Dvorak <ook@ucw.cz>
2825 PR tree-optimization/32183
2826 * Makefile.in (tree-ssa-reassoc.o): Also depend on $(CFGLOOP_H).
2828 * tree-ssa-reassoc.c: Include cfgloop.h.
2829 (is_reassociable_op): Add a loop argument and return true only
2831 (linearize_expr): Updated.
2832 (should_break_up_subtract): Likewise.
2833 (linearize_expr_tree): Likewise.
2834 (init_reassoc): Call loop_optimizer_init with
2835 AVOID_CFG_MODIFICATIONS. Remove calculate_dominance_info call
2836 with CDI_DOMINATORS.
2837 (fini_reassoc): Call loop_optimizer_finalize.
2839 2007-09-07 Sterling Augustine <sterling@tensilica.com>
2841 * config/xtensa/lib2funcs.S (__xtensa_sync_caches): Use an ISYNC even
2842 if there is no i-cache.
2844 2007-09-07 Richard Guenther <rguenther@suse.de>
2847 2007-09-06 Richard Guenther <rguenther@suse.de>
2849 PR tree-optimization/32586
2850 * tree-ssa-sccvn.c (simplify_binary_expression): Avoid
2851 folding if nothing changed.
2852 (simplify_unary_expression): New function. Do tree combining
2853 on conversion like codes.
2854 (try_to_simplify): Call it.
2856 2007-09-07 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2859 * gthr-posix.h (__gthread_active_p): Add implementation for hppa-hpux.
2860 (__gthread_active,__gthread_start, __gthread_active_init): New.
2861 * gthr-posix95.h: Likewise.
2863 2007-09-07 Roman Zippel <zippel@linux-m68k.org>
2865 * function.h (struct function): Rename calls_unwind_init
2866 to saves_all_registers.
2867 (current_function_saves_all_registers): Renamed from
2868 current_function_calls_unwind_init.
2869 * reload1.c (reload): Set current_function_saves_all_registers.
2870 * except.c (expand_builtin_unwind_init): Likewise.
2871 * config/m68k/m68k.c (m68k_save_reg): Use
2872 current_function_saves_all_registers to save pic register.
2874 2007-09-07 Janis Johnson <janis187@us.ibm.com>
2876 config/m68k/m68k.c (floating_exact_log2): Update call to real_2expN.
2877 config/s390/s390.md (fixuns_trunc<BFP:mode><GPR:mode>2): Ditto.
2879 Reapply reverted changes:
2881 2007-09-06 Jan Hubicka <jh@suse.cz>
2883 * config/i386.c (ix86_expand_lround, ix86_expand_round): Update call
2886 2007-09-06 Richard Sandiford <richard@codesourcery.com>
2888 * config/mips/mips.md (fixuns_truncdfsi2, fixuns_truncdfdi2)
2889 (fixuns_truncsfsi2, fixuns_truncsfdi2): Update calls to real_2expN.
2891 2007-09-05 Janis Johnson <janis187@us.ibm.com>
2893 * optabs.c (expand_float): Convert unsigned integer as signed only
2894 if it provides sufficient accuracy; add mode argument to real_2expN.
2895 (expand_fix): Fix comment typos; extend binary float into mode
2896 wider than destination for converion to unsigned integer; add mode
2897 argument to real_2expN.
2898 * real.c (real_2expN): Add mode argument to special-case decimal
2900 * real.h (real_2expN): Ditto.
2901 * fixed-value.c (check_real_for_fixed_mode): Add mode argument to
2903 (fixed_from_string): Ditto.
2904 (fixed_to_decimal): Ditto.
2905 (fixed_convert_from_real): Ditto.
2906 (real_convert_from_fixed): Ditto.
2907 * config/rs6000/rs6000.md (FP): Include DD and TD modes.
2908 * config/rs6000/dfp.md (extendddtd2, adddd3, addtd3, subdd3, subtd3,
2909 muldd3, multd3, divdd3, divtd3, cmpdd_internal1, cmptd_internal1,
2910 floatditd2, ftruncdd2, fixdddi2, ftrunctd2, fixddi2): New.
2912 2007-09-07 Diego Novillo <dnovillo@google.com>
2914 * tree-flow.h (const_block_stmt_iterator): Remove.
2915 Update all users to use block_stmt_iterator.
2916 * tree-iterator.h (const_tree_stmt_iterator): Remove.
2917 Update all users to use tree_stmt_iterator.
2919 2007-09-07 Sandra Loosemore <sandra@codesourcery.com>
2921 * config/mips/mips.c (mips_set_current_function): Temporarily
2922 make this a no-op to fix bootstrap errors, pending rewrite.
2924 2007-09-07 Jan Hubicka <jh@suse.cz>
2926 * reorg.c (dbr_schedule): Move code removing placeholder USEs later
2929 2007-09-07 Dorit Nuzman <dorit@il.ibm.com>
2931 PR tree-optimization/33299
2932 * tree-vect-transform.c (vect_create_epilog_for_reduction): Update
2933 uses for all relevant loop-exit phis, not just the first.
2935 2007-09-07 Richard Guenther <rguenther@suse.de>
2938 * tree-ssa-operands.c (access_can_touch_variable): An access
2939 of the form (*p)[0] can touch a variable of same size.
2941 2007-09-07 Jan Hubicka <jh@suse.cz>
2943 * passes.c (init_optimization_passes): Add simple dce and addressable
2945 * tree-ssa.c (execute_update_addresses_taken): New function.
2946 (pass_update_address_taken): New.
2947 * tree-ssa-dse.c (execute_simple_dse): New function.
2948 (pass_simple_dse): New.
2949 * tree-pass.h (pass_simple_dse, pass_update_address_taken): Declare.
2951 2007-09-07 Tobias Burnus <burnus@net-b.de>
2954 * doc/invoke.texi: Fix -Wstrict-overflow= table.
2956 2007-09-07 Richard Guenther <rguenther@suse.de>
2959 2007-09-06 Richard Guenther <rguenther@suse.de>
2961 PR tree-optimization/32586
2962 * tree-ssa-sccvn.c (simplify_binary_expression): Avoid
2963 folding if nothing changed.
2964 (simplify_unary_expression): New function. Do tree combining
2965 on conversion like codes.
2966 (try_to_simplify): Call it.
2968 2007-09-07 Richard Guenther <rguenther@suse.de>
2969 Uros Bizjak <ubizjak@gmail.com>
2971 PR tree-optimization/32821
2972 * tree_if_conv.c (combine_blocks): Use alloc_stmt_list instead of
2973 NULL in the call to set_bb_stmt_list.
2975 2007-09-07 Richard Sandiford <richard@codesourcery.com>
2977 * config/mips/mips.c (build_mips16_call_stub): Emit all direct
2978 float calls here, rather than leaving some to the caller.
2979 Use call_internal_direct and call_value_internal_direct.
2980 * config/mips/mips.md (call_internal_direct): New pattern.
2981 (call_value_internal_direct): Likewise.
2983 2007-09-07 Richard Sandiford <richard@codesourcery.com>
2985 * config/mips/mips.c (mips_base_move_loop_invariants): New variable.
2986 (mips_set_mips16_mode): Restore flag_move_loop_invariants, then set
2988 (override_options): Set mips_base_move_loop_invariants.
2990 2007-09-07 Richard Sandiford <richard@codesourcery.com>
2992 * config/mips/mips.opt (mcode-readable=): Move to keep list
2995 2007-09-07 Richard Sandiford <richard@codesourcery.com>
2997 * doc/invoke.texi: Document the MIPS -mlocal-sdata, -mextern-sdata
2998 and -mgpopt options. Adjust the -G documentation to match.
2999 * config/mips/mips.opt (mextern-data, mgpopt, mlocal-data): New.
3000 * config/mips/mips.c (mips_rtx_constant_in_small_data_p): New
3001 function, split out from mips_classify_symbol. Return false for
3002 !TARGET_LOCAL_SDATA.
3003 (mips_classify_symbol): Call mips_rtx_constant_in_small_data_p.
3004 Only use GP-relative accesses for non-pool symbols if TARGET_GPOPT.
3005 (override_options): Check whether the -mgpopt setting is consistent
3006 with the other flags.
3007 (symbolic_expression_p): Delete.
3008 (mips_select_rtx_section): Use mips_rtx_constant_in_small_data_p
3009 and default_elf_select_rtx_section.
3010 (mips_in_small_data_p): Honor section attributes for MIPS16 too.
3011 Return false for local data unless TARGET_LOCAL_SDATA. Likewise
3012 external data and TARGET_EXTERN_SDATA.
3014 2007-09-07 Tobias Burnus <burnus@net-b.de>
3017 * doc/cpp.texi (Common Predefined Macros): Add __GFORTRAN__.
3019 2007-09-05 Laurynas Biveinis <laurynas.biveinis@gmail.com>
3021 * regrename.c (copyprop_hardreg_forward_1): New variable next. Use
3022 FOR_BB_INSNS_SAFE instead of for loop.
3023 * cse.c (cse_extended_basic_block): Likewise.
3024 * postreload.c (reload_cse_regs_1): New variable next. Make sure
3025 that the for loop does not invoke NEXT_INSN on a deleted insn.
3026 * function.c (instantiate_virtual_regs): Likewise.
3027 * lower-subreg.c (remove_retval_note): Likewise.
3028 (decompose_multiword_subregs): Use FOR_BB_INSNS_SAFE instead of
3030 * emit-rtl.c (remove_insn): Set NEXT_INSN and PREV_INSN to NULL on
3032 * cfgrtl.c (delete_insn): Set JUMP_LABEL to NULL on a deleted
3033 insn, if it's a jump.
3034 (try_redirect_by_replacing_jump): New variable jump_p. Call
3035 tablejump_p before delete_insn_chain.
3036 * reload1.c (reload): New variable next. Make sure that the for
3037 loop does not invoke NEXT_INSN on a deleted insn.
3038 (fixup_eh_region_note): Make the loop terminate if i becomes NULL.
3039 (delete_output_reload): New variable prev. Make sure the the for
3040 loops do not invoke PREV_INSN on a deleted insn.
3042 2007-09-06 Zdenek Dvorak <ook@ucw.cz>
3044 * cgraphbuild.c (rebuild_cgraph_edges): Export.
3045 * cgraph.h (rebuild_cgraph_edges): Declare.
3046 * tree-pass.h (pass_expand_omp_ssa): New.
3047 * omp-low.c (find_omp_clause): Export.
3048 (copy_var_decl): Split from omp_copy_decl_2.
3049 (build_omp_barrier): Return the call to emit instead of emitting
3051 (lower_rec_input_clauses, expand_omp_single): Gimplify the result of
3053 (extract_omp_for_data, expand_parallel_call, expand_omp_parallel,
3054 expand_omp_for_generic, expand_omp_for_static_nochunk,
3055 expand_omp_for_static_chunk, expand_omp_for, expand_omp_sections):
3056 Adapted to work on SSA form.
3057 (execute_expand_omp): Do not invalidate dominance information.
3058 (gate_expand_omp): Do not run with -fopenmp-ssa flag.
3059 (gate_expand_omp_ssa, pass_expand_omp_ssa): New.
3060 * gimplify.c (gimplify_omp_for): Ensure that the control variable is
3062 (force_gimple_operand): Allow gimplifying code expressions without
3064 * tree-predcom.c (mark_virtual_ops_for_renaming): Handle phi nodes.
3065 * common.opt (fopenmp-ssa): New.
3066 * tree-flow.h (find_omp_clause, copy_var_decl): Declare.
3067 * Makefile.in (tree-cfg.o): Add TREE_INLINE_H dependency.
3068 * tree-cfg.c: Include tree-inline.h.
3069 (struct move_stmt_d): Replace vars_to_remove by vars_map field.
3070 (replace_by_duplicate_decl, replace_ssa_name,
3071 mark_virtual_ops_in_region): New functions.
3072 (move_stmt_r, move_block_to_fn, move_sese_region_to_fn): Adapted
3073 to work on SSA form.
3074 * passes.c (init_optimization_passes): Add pass_expand_omp_ssa pass.
3075 * tree-ssa-operands.c (get_expr_operands): Handle operands of OMP
3078 2007-09-06 Laurynas Biveinis <laurynas.biveinis@gmail.com>
3080 * tree-loop-linear.c: Include obstack.h.
3081 (linear_transform_loops): New obstack lambda_obstack.
3082 Initialize it, pass it to gcc_loopnest_to_lambda_loopnest,
3083 lambda_loopnest_transform, lambda_loopnest_to_gcc_loopnest calls
3084 and free afterwards.
3085 * lambda.h (struct obstack): New forward declaration.
3086 (lambda_linear_expression_new): New parameter of type struct
3088 (lambda_loopnest_new): Likewise.
3089 (lambda_loopnest_transform): Likewise.
3090 (lambda_body_vector_new): Likewise.
3091 (lambda_body_vector_compute_new): Likewise.
3092 (gcc_loopnest_to_lambda_loopnest): Likewise.
3093 (lambda_loopnest_to_gcc_loopnest): Likewise.
3094 * lambda-code.c: Include obstack.h.
3095 (lambda_lattice_new): New parameter lambda_obstack. Use it for
3097 (lambda_body_vector_new): Likewise.
3098 (lambda_linear_expression_new): Likewise.
3099 (lambda_lattice_new): Likewise.
3100 (lambda_loopnest_new): Likewise. Additionally use obstack to
3101 allocate LN_LOOPS(ret).
3102 (lambda_lattice_compute_base): New parameter lambda_obstack. Pass
3103 it to lambda_lattice_new.
3104 (lambda_body_vector_compute_new): New parameter lambda_obstack.
3105 Pass it to lambda_body_vector_new.
3106 (lambda_lattice_compute_base): New paramater lambda_obstack. Pass
3107 it to lambda_lattice_new.
3108 (compute_nest_using_fourier_motzkin): New parameter lambda_obstack.
3109 Pass it to lambda_loopnest_new, lambda_linear_expression_new.
3110 (lambda_compute_target_space): Likewise.
3111 (lambda_compute_auxillary_space): New parameter lambda_obstack.
3112 Pass it to lambda_lattice_compute_base and
3113 compute_nest_using_fourieer_motzkin.
3114 (lambda_loopnest_transform): New parameter lambda_obstack. Pass
3115 it to lambda_lattice_compute_base, lambda_lattice_auxillary_space
3116 and lambda_lattice_compute_target_space.
3117 (gcc_tree_to_linear_expression): Nex parameter lambda_obstack.
3118 Pass it to lambda_linear_expression_new.
3119 (gcc_loop_to_lambda_loop): New parameter lambda_obstack. Pass it
3120 to gcc_tree_to_linear_expression.
3121 (gcc_loopnest_to_lambda_loopnest): New parameter lambda_obstack.
3122 Pass it to gcc_loop_to_lambda_loop and lambda_loopnest_new.
3123 (lambda_loopnest_to_gcc_loopnest): New parameter lambda_obstack.
3124 Pass it to lambda_body_vector_new and
3125 lambda_body_vector_compute_new.
3126 * Makefile.in (tree-loop-linear.o): Add $(OBSTACK_H) dependency.
3127 (lambda-code.o): Likewise.
3129 2007-09-06 Chao-ying Fu <fu@mips.com>
3131 * ginclude/stdfix.h: New file.
3132 * Makefile.in (USER_H): Add $(srcdir)/ginclude/stdfix.h.
3133 (convert.o): Add dependence on fixed-value.h.
3134 * c-convert.c (convert): Support FIXED_POINT_TYPE.
3135 * c-cppbuiltin.c (builtin_define_fixed_point_constants): New function
3136 to define fixed-point constants.
3137 (c_cpp_builtins): Define fixed-point constants.
3138 * convert.c (fixed-value.h): New include.
3139 (convert_to_real): Update comment to include fixed-point.
3140 Support FIXED_POINT_TYPE.
3141 (convert_to_integer): Update comment to include fixed-point.
3142 Support FIXED_POINT_TYPE.
3143 (convert_to_complex): Support FIXED_POINT_TYPE.
3144 (convert_to_fixed): New function.
3145 * convert.h (convert_to_fixed): Declare.
3146 * genopinit.c: Add comment about $Q for only fixed-point modes.
3147 (optabs): Add fract_optab, fractuns_optab, satfract_optab,
3148 satfractuns_optab, add_optab, ssadd_optab, usadd_optab, sub_optab,
3149 sssub_optab, ussub_optab, smul_optab, ssmul_optab, usmul_optab,
3150 ssmadd_widen_optab, usmadd_widen_optab, ssdiv_optab, udiv_optab,
3151 usdiv_optab, ssashl_optab, usashl_optab, neg_optab, ssneg_optab,
3152 usneg_optab for fixed-point modes.
3153 (gen_insn): Add force_fixed to track the $Q format for all fixed-point
3155 * optabs.c (optab_for_tree_code): For *DIV_EXPR, LSHIFT_EXPR,
3156 PLUS_EXPR, MINUS_EXPR, MULT_EXPR, NEGATE_EXPR, return signed or
3157 unsigned saturation optabs, when type is saturating.
3158 (shift_optab_p): Return true for SS_ASHIFT or US_ASHIFT.
3159 (expand_fixed_convert): New function.
3160 (gen_fixed_libfunc, gen_signed_fixed_libfunc,
3161 gen_unsigned_fixed_libfunc, gen_int_fp_fixed_libfunc,
3162 gen_int_fp_signed_fixed_libfunc, gen_int_fixed_libfunc,
3163 gen_int_signed_fixed_libfunc, gen_int_unsigned_fixed_libfunc,
3164 gen_fract_conv_libfunc, gen_fractuns_conv_libfunc,
3165 gen_satfract_conv_libfunc, gen_satfractuns_conv_libfunc): New
3167 (init_optabs): Initialize ssadd_optab, usadd_optab, sssub_optab,
3168 ussub_optab, ssmul_optab, usmul_optab, ssmadd_widen_optab,
3169 usmadd_widen_optab, ssmsub_widen_optab, usmsub_widen_optab,
3170 ssdiv_optab, usdiv_optab, ssashl_optab, usashl_optab, ssneg_optab,
3171 usneg_optab, fract_optab, fractuns_optab, satfract_optab,
3173 Initialize fixed-point libraries, including add, ssadd, usadd, sub,
3174 sssub, ussub, mul, ssmul, usmul, div, ssdiv, udiv, usdiv, ashl,
3175 ssashl, usashl, ashr, lshr, neg, ssneg, usneg, cmp, fract, satfract,
3176 fractuns, satfractuns.
3177 * optabs.h (enum optab_index): Add OTI_ssadd, OTI_usadd, OTI_sssub,
3178 OTI_ussub, OTI_ssmul, OTI_usmul, OTI_ssdiv, OTI_usdiv, OTI_ssneg,
3179 OTI_usneg, OTI_ssashl, OTI_usashl, OTI_ssmadd_widen, OTI_usmadd_widen,
3180 OTI_ssmsub_widen, OTI_usmsub_widen.
3181 (ssadd_optab, usadd_optab, sssub_optab, ussub_optab, ssmul_optab,
3182 usmul_optab, ssdiv_optab, usdiv_optab, ssneg_optab, usneg_optab,
3183 ssashl_optab, usashl_optab, ssmadd_widen_optab, usmadd_widen_optab,
3184 umsub_widen_optab, usmsub_widen_optab): Define.
3185 (enum convert_optab_index): Add COI_fract, COI_fractuns, COI_satfract,
3187 (fract_optab, fractuns_optab, satfract_optab, satfractuns_optab):
3189 (expand_fixed_convert): Declare.
3190 * expr.c (convert_move): Support the move of fixed-point modes.
3191 (emit_move_insn_1): Handle fixed-point mode to move via integer.
3192 (categorize_ctor_elements_1): Handle FIXED_CST.
3193 (count_type_elements): Handle FIXED_POINT_TYPE.
3194 (expand_expr_real_1): For VECTOR_CST, check MODE_VECTOR_FRACT,
3195 MODE_VECTOR_UFRACT, MODE_VECTOR_ACCUM, MODE_VECTOR_UACCUM.
3197 For PLUS_EXPR and MINUS_EXPR, support saturating and non-saturating
3198 multiply and add/subtract for fixed-point types.
3199 For MULT_EXPR, *DIV_EXPR, *SHIFT_EXPR, if the mode if a fixed-point
3200 mode, we jump to binop directly.
3201 Support FIXED_CONVERT_EXPR.
3202 (do_store_flag): Check FIXED_CST to put a constant second.
3203 (vector_mode_valid_p): Handle MODE_VECTOR_FRACT,
3204 MODE_VECTOR_UFRACT, MODE_VECTOR_ACCUM, MODE_VECTOR_UACCUM.
3205 (const_vector_from_tree): Support FIXED_CST.
3206 * doc/extend.texi (Fixed-Point): New node.
3207 * doc/md.texi (ssadd, usadd, sssub, ussub, ssmul, usmul, ssdiv, usdiv,
3208 ssmadd, usmadd, ssmsub, usmsub, ssashl, usashl, ssneg, usneg, fract,
3209 satfract, fractuns, satfractuns): Document them.
3211 2007-09-07 Bernd Schmidt <bernd.schmidt@analog.com>
3213 * config/bfin/bfin.h (PREFERRED_RELOAD_CLASS): Don't reload autoinc
3214 addresses into I registers.
3216 2007-09-06 Jan Hubicka <jh@suse.cz>
3217 Andreas Tobler <a.tobler@schweiz.org>
3219 * config/darwin.c (machopic_indirect_data_reference): Avoid invalid
3221 (machopic_legitimize_pic_address): Likewise.
3223 2007-09-06 Andrew Pinski <andrew_pinski@playstation.sony.com>
3224 Jan Hubicka <jh@suse.cz>
3226 * config/spu/spu.md (floatsidf2): Use convert_optab_libfunc
3227 instead of ufloat_optab->handlers directly.
3228 (floatdidf2): Likewise.
3230 2007-09-06 Sandra Loosemore <sandra@codesourcery.com>
3232 * config/mips/mips.c: Include diagnostic.h.
3233 (mips_set_current_function): Check errorcount and sorrycount
3234 before generating RTL.
3236 2007-09-06 Richard Sandiford <richard@codesourcery.com>
3239 * config/mips/mips.c (mips_classify_symbolic_expression): New function.
3240 (mips_classify_address): Use it instead of mips_symbolic_constant_p.
3241 (print_operand_reloc): Likewise.
3243 2007-09-06 Janis Johnson <janis187@us.ibm.com>
3247 2007-09-06 Jan Hubicka <jh@suse.cz>
3249 * i386.c (ix86_expand_lround, ix86_expand_round): Update call of
3252 2007-09-06 Richard Sandiford <richard@codesourcery.com>
3254 * config/mips/mips.md (fixuns_truncdfsi2, fixuns_truncdfdi2)
3255 (fixuns_truncsfsi2, fixuns_truncsfdi2): Update calls to real_2expN.
3257 2007-09-05 Janis Johnson <janis187@us.ibm.com>
3259 * optabs.c (expand_float): Convert unsigned integer as signed only
3260 if it provides sufficient accuracy; add mode argument to real_2expN.
3261 (expand_fix): Fix comment typos; extend binary float into mode
3262 wider than destination for converion to unsigned integer; add mode
3263 argument to real_2expN.
3264 * real.c (real_2expN): Add mode argument to special-case decimal
3266 * real.h (real_2expN): Ditto.
3267 * fixed-value.c (check_real_for_fixed_mode): Add mode argument to
3269 (fixed_from_string): Ditto.
3270 (fixed_to_decimal): Ditto.
3271 (fixed_convert_from_real): Ditto.
3272 (real_convert_from_fixed): Ditto.
3273 * config/rs6000/rs6000.md (FP): Include DD and TD modes.
3274 * config/rs6000/dfp.md (extendddtd2, adddd3, addtd3, subdd3, subtd3,
3275 muldd3, multd3, divdd3, divtd3, cmpdd_internal1, cmptd_internal1,
3276 floatditd2, ftruncdd2, fixdddi2, ftrunctd2, fixddi2): New.
3278 2007-09-06 Tom Tromey <tromey@redhat.com>
3280 * tree-cfg.c (remove_bb): Only warn if line is non-zero.
3281 * c-pch.c (c_common_read_pch): Restore current location after
3283 * tree.c (expand_location): Update.
3284 (expr_filename): Changed return type. Unified the two cases.
3285 (expr_lineno): Likewise.
3286 (annotate_with_file_line): Don't use EXPR_LINENO and EXPR_FILENAME
3288 * toplev.c (line_table): Changed type.
3289 (general_init): Update.
3290 (realloc_for_line_map): New function.
3291 (general_init): Allocate line_table using GC.
3292 * fix-header.c (line_table): Changed type.
3293 (read_scan_file): Update.
3294 (read_scan_file): Update.
3295 * c-ppoutput.c (maybe_print_line): Update.
3296 (print_line): Update.
3297 (cb_line_change): Update.
3298 (cb_define): Update.
3299 (pp_file_change): Update.
3300 * c-opts.c (c_common_init_options): Update.
3301 (finish_options): Update.
3302 (push_command_line_include): Update.
3303 * c-lex.c (cb_line_change): Update.
3304 (cb_def_pragma): Update.
3305 (cb_define): Update.
3307 (c_lex_with_flags): Use cpp_get_token_with_location.
3308 * input.h (line_table): Changed type.
3309 (location_from_locus): New macro.
3310 * tree.h (EXPR_FILENAME): No longer an lvalue.
3311 (EXPR_LINENO): Likewise.
3312 (expr_locus, set_expr_locus): Declare separately for
3313 USE_MAPPED_LOCATION.
3314 (expr_filename, expr_lineno): Changed return type.
3315 * gimplify.c (tree_to_gimple_tuple): Use SET_EXPR_LOCUS.
3316 * cfgexpand.c (expand_gimple_cond_expr): Use location_from_locus.
3317 (expand_gimple_basic_block): Likewise.
3318 * final.c (final_scan_insn): Use expanded_location.
3320 2007-09-06 Richard Guenther <rguenther@suse.de>
3322 PR tree-optimization/32586
3323 * tree-ssa-sccvn.c (simplify_binary_expression): Avoid
3324 folding if nothing changed.
3325 (simplify_unary_expression): New function. Do tree combining
3326 on conversion like codes.
3327 (try_to_simplify): Call it.
3328 * builtins.c (fold_builtin_cexp): Fold the built expressions.
3329 * fold-const.c (fold_unary): Test result of get_callee_fndecl().
3331 2007-09-06 Jan Hubicka <jh@suse.cz>
3334 * cse.c (fold_rtx): Avoid invalid sharing.
3336 2007-09-06 Richard Sandiford <richard@codesourcery.com>
3338 * config/mips/mips.md (fixuns_truncdfsi2, fixuns_truncdfdi2)
3339 (fixuns_truncsfsi2, fixuns_truncsfdi2): Update calls to real_2expN.
3341 2007-09-06 Revital Eres <eres@il.ibm.com>
3343 * config/rs6000/rs6000.c (rs6000_legitimate_offset_address_p):
3344 Fix offsettable memory reference for 750CL.
3346 2007-09-06 Pat Haugen <pthaugen@us.ibm.com>
3348 * reload.c (find_reloads_address_1): Try to preserve original
3349 base/index regclass of operands.
3350 * regrename.c (scan_rtx_address): Likewise.
3351 (replace_oldest_value_addr): Likewise.
3353 2007-09-06 Richard Guenther <rguenther@suse.de>
3355 PR tree-optimization/33302
3356 * domwalk.c (walk_dominator_tree): The exit block is
3357 interesting even if it is not reachable.
3359 2007-09-06 Richard Sandiford <richard@codesourcery.com>
3362 * optabs.c (avoid_expensive_constant): Do nothing if MODE is VOIDmode.
3364 2007-09-06 Basile Starynkevitch <basile@starynkevitch.net>
3366 * cfg.c (dump_bb_info, dump_edge_info): Added cfun test for
3369 2007-09-05 Ian Lance Taylor <iant@google.com>
3371 * tree-pretty-print.c (dump_decl_name): Cast LABEL_DECL_UID to int
3373 (dump_generic_node): Likewise.
3374 * print-rtl.c (print_decl_name): Likewise.
3375 * print-tree.c (print_node_brief): Likewise.
3376 (print_node): Likewise.
3377 * Makefile.in (RTL_BASE_H): Add alias.h.
3380 2007-09-06 Jesper Nilsson <jesper.nilsson@axis.com>
3382 * longlong.h [__CRIS_arch_version >= 8] (count_trailing_zeros):
3384 * config/cris/cris.md (ctzsi2, cris_swap_bits): Implemented.
3385 * config/cris/cris.h (CTZ_DEFINED_VALUE_AT_ZERO): Defined.
3387 2007-09-06 Jie Zhang <jie.zhang@analog.com>
3389 * config.gcc (tm_file): Add linux.h for bfin*-uclinux*.
3390 (tm_defines): Define UCLIBC_DEFAULT to 1.
3391 (extra_options): Add linux.opt.
3392 * config/bfin/linux.h (CPLUSPLUS_CPP_SPEC): Don't define.
3393 (CRT_CALL_STATIC_FUNCTION): Likewise.
3394 (NO_IMPLICIT_EXTERN_C): Likewise.
3395 (TARGET_OS_CPP_BUILTINS): Define as LINUX_TARGET_OS_CPP_BUILTINS.
3396 * config/bfin/elf.h (OBJECT_FORMAT_ELF): Don't define.
3397 * config/bfin/uclinux.h (CPLUSPLUS_CPP_SPEC): Don't define.
3398 (ENDFILE_SPEC): Don't define.
3399 (LIB_SPEC): Likewise.
3400 (CRT_CALL_STATIC_FUNCTION): Likewise.
3401 (NO_IMPLICIT_EXTERN_C): Likewise.
3402 (LINUX_TARGET_OS_CPP_BUILTINS): Likewise.
3403 (TARGET_OS_CPP_BUILTINS): Define as LINUX_TARGET_OS_CPP_BUILTINS.
3405 2007-09-06 Jan Hubicka <jh@suse.cz>
3407 * i386.c (ix86_expand_lround, ix86_expand_round): Update call of
3410 2007-09-06 Jan Hubicka <jh@suse.cz>
3412 * opts.c (common_handle_option): Enable inlining functions for
3415 2007-09-06 Jakub Jelinek <jakub@redhat.com>
3417 * builtin-attrs.def (ATTR_NONNULL_3): New.
3418 (DEF_FORMAT_ATTRIBUTE): Use just ATTR_NONNULL_##FA instead of
3419 ATTR_NOTHROW_NONNULL_##FA.
3420 (DEF_FORMAT_ATTRIBUTE_NOTHROW, DEF_FORMAT_ATTRIBUTE_BOTH): New macros.
3421 (ATTR_FORMAT_PRINTF_NOTHROW_2_0, ATTR_PRINTF_NOTHROW_2_3,
3422 ATTR_FORMAT_PRINTF_NOTHROW_3_0, ATTR_FORMAT_PRINTF_NOTHROW_3_4,
3423 ATTR_FORMAT_PRINTF_NOTHROW_4_0, ATTR_PRINTF_NOTHROW_4_5,
3424 ATTR_FORMAT_PRINTF_NOTHROW_5_0, ATTR_FORMAT_PRINTF_NOTHROW_5_6,
3425 ATTR_FORMAT_SCANF_NOTHROW_2_0, ATTR_FORMAT_SCANF_NOTHROW_2_3,
3426 ATTR_FORMAT_STRFTIME_NOTHROW_3_0, ATTR_FORMAT_STRFMON_NOTHROW_3_4):
3428 (ATTR_FORMAT_PRINTF_4_0, ATTR_PRINTF_4_5, ATTR_FORMAT_PRINTF_5_0,
3429 ATTR_FORMAT_PRINTF_5_6, ATTR_FORMAT_STRFTIME_3_0,
3430 ATTR_FORMAT_NOTHROW_3_4): Remove.
3431 * builtins.def (snprintf, sprintf, sscanf, vsnprintf, vsprintf,
3432 vsscanf, strfmon, strftime, __snprintf_chk, __sprintf_chk,
3433 __vsnprintf_chk, __vsprintf_chk): Use ATTR_FORMAT_*_NOTHROW_*
3434 instead of ATTR_FORMAT_*_*.
3436 2007-09-06 Hans-Peter Nilsson <hp@axis.com>
3438 * config/cris/cris.c (cris_emit_movem_store): Call copy_rtx on
3439 parts re-used for REG_FRAME_RELATED_EXPR.
3441 2007-09-05 Janis Johnson <janis187@us.ibm.com>
3443 * optabs.c (expand_float): Convert unsigned integer as signed only
3444 if it provides sufficient accuracy; add mode argument to real_2expN.
3445 (expand_fix): Fix comment typos; extend binary float into mode
3446 wider than destination for converion to unsigned integer; add mode
3447 argument to real_2expN.
3448 * real.c (real_2expN): Add mode argument to special-case decimal
3450 * real.h (real_2expN): Ditto.
3451 * fixed-value.c (check_real_for_fixed_mode): Add mode argument to
3453 (fixed_from_string): Ditto.
3454 (fixed_to_decimal): Ditto.
3455 (fixed_convert_from_real): Ditto.
3456 (real_convert_from_fixed): Ditto.
3457 * config/rs6000/rs6000.md (FP): Include DD and TD modes.
3458 * config/rs6000/dfp.md (extendddtd2, adddd3, addtd3, subdd3, subtd3,
3459 muldd3, multd3, divdd3, divtd3, cmpdd_internal1, cmptd_internal1,
3460 floatditd2, ftruncdd2, fixdddi2, ftrunctd2, fixddi2): New.
3462 2007-09-05 Ian Lance Taylor <iant@google.com>
3464 * init-regs.c (initialize_uninitialized_regs): Call
3465 emit_insn_before rather than add_insn_before.
3467 2007-09-05 Jakub Jelinek <jakub@redhat.com>
3469 * builtins.def (BUILT_IN_VA_ARG_PACK): New built-in.
3470 * tree.h (CALL_EXPR_VA_ARG_PACK): Define.
3471 * tree-inline.h (copy_body_data): Add call_expr field.
3472 * tree-inline.c (expand_call_inline): Initialize call_expr.
3473 (copy_bb): Append anonymous inline fn arguments to arguments
3474 when inlining a CALL_EXPR_VA_ARG_PACK call.
3475 * builtins.c (expand_builtin): Issue an error if
3476 BUILT_IN_VA_ARG_PACK is seen during expand.
3477 (fold_call_expr, fold_builtin_call_array): Don't fold
3478 CALL_EXPR_VA_ARG_PACK CALL_EXPRs or calls with
3479 __builtin_va_arg_pack () call as last argument.
3480 * gimplify.c (gimplify_call_expr): If last argument to a vararg
3481 function is __builtin_va_arg_pack (), decrease number of call
3482 arguments and instead set CALL_EXPR_VA_ARG_PACK on the CALL_EXPR.
3483 * expr.c (expand_expr_real_1): Issue an error if
3484 CALL_EXPR_VA_ARG_PACK CALL_EXPR is seen during expand.
3485 * tree-pretty-print.c (dump_generic_node): Handle printing
3486 CALL_EXPR_VA_ARG_PACK bit on CALL_EXPRs.
3487 * doc/extend.texi (__builtin_va_arg_pack): Document.
3489 2007-09-05 Adam Nemet <anemet@caviumnetworks.com>
3491 PR tree-optimization/21513
3492 * builtins.c (build_builtin_expect_predicate): New function.
3493 (fold_builtin_expect): Add argument for expected value.
3494 Distribute __builtin_expect over short-circuiting operations.
3495 Fold nested builtin_expects.
3496 (fold_builtin_2): Adjust call to fold_builtin_expect.
3498 2007-09-05 Rask Ingemann Lambertsen <rask@sygehus.dk>
3501 PR tree-optimization/13756
3502 * doc/invoke.texi (Options That Control Optimization): Document
3505 2007-09-05 Uros Bizjak <ubizjak@gmail.com>
3507 * gcc/config/i386/cpuid.h: New file.
3508 * gcc/config/i386/driver-i386.c: Include cpuid.h.
3509 (describe_cache): Shrink size and line strings to 100 bytes.
3510 (detect_caches_amd): Return "" for unsupported max_ext_level.
3511 Use __cpuid function.
3512 (detect_caches_intel): Return "" for unsupported max_level.
3513 Use __cpuid function.
3514 (host_detect_local_cpu): Change feature flag variables to
3515 unsigned int. Initialize only extended feature flag variables.
3516 Use __get_cpuid_max to determine max supported cpuid level.
3517 Use __cpuid function to determine supported features. Fix
3518 calculation of family id. Remove is_amd and check signature
3519 directly. Check for Geode signature. Handle family 4 id.
3520 [PROCESSOR_GENERIC32]: New default for unknown family id. Move
3521 cpu discovery code to lower part of the function.
3522 [PROCESSOR_PENTIUM, PROCESSOR_K6, PROCESSOR_ATHLON]: Do not tune
3523 for sub-architecture.
3524 [PROCESSOR_PENTIUMPRO]: Simplify cpu discovery code.
3525 [PROCESSOR_K8]: Add k8-sse3 architecture.
3526 [PROCESSOR_NOCONA]: Remove.
3527 [PROCESSOR_GENERIC64]: Ditto.
3528 * gcc/config/i386/x-i386 (driver-i386.o): Depend on cpuid.h.
3529 * gcc/config/i386/crtfastmath.c: Include cpuid.h. Use __get_cpuid
3530 to check for SSE and FXSAVE support.
3531 * gcc/config/i386/t-crtfm (crtfastmath.o): Depend on cpuid.h.
3532 Add -minline-all-stringops.
3533 * gcc/config.gcc (i[34567]86-*-*): Add cpuid.h to extra_headers.
3534 (x86_64-*-*): Ditto.
3536 2007-09-05 Jie Zhang <jie.zhang@analog.com>
3538 * config/bfin/linux-unwind.h: New file.
3539 * config/bfin/linux.h (MD_UNWIND_SUPPORT): Define.
3540 * config/bfin/uclinux.h (MD_UNWIND_SUPPORT): Define.
3542 2007-09-05 Laurynas Biveinis <laurynas.biveinis@gmail.com>
3544 * Makefile.in (stmt.o): Add alloc-pool.h dependency.
3545 * stmt.c: Include alloc-pool.h.
3546 (struct case_node): Remove GTY marker.
3547 (add_case_node): New parameter case_node_pool. Use it for
3549 (expand_case): New allocation pool case_node_pool. Initialize it,
3550 pass to add_case_node and free it.
3552 2007-09-05 Sandra Loosemore <sandra@codesourcery.com>
3553 David Ung <davidu@mips.com>
3554 Nigel Stephens <nigel@mips.com>
3556 Add mips16/nomips16 function attributes and -mflip-mips16 option
3557 for testing mixed-mode compilation.
3559 * config/mips/mips.opt (mflip-mips16): New.
3561 * config/mips/mips.h (SYMBOL_FLAG_MIPS16_FUNC): Define.
3562 (SYMBOL_FLAG_MIPS16_FUNC_P): Define.
3564 * config/mips/mips.c (mips_base_target_flags): New.
3565 (mips_base_mips16): New.
3566 (mips_base_schedule_insns): New.
3567 (mips_base_reorder_blocks_and_partition): New.
3568 (mips_base_align_loops): New.
3569 (mips_base_align_jumps): New.
3570 (mips_base_align_functions): New.
3571 (mips16_flipper): New.
3572 (mips_attribute_table): Add "mips16" and "nomips16" entries.
3573 (TARGET_SET_CURRENT_FUNCTION): Define.
3574 (mips_mips16_type_p, mips_nomips16_type_p): New.
3575 (mips_comp_type_attributes): Check mips16/nomips16 attributes.
3576 (mips_function_ok_for_sibcall): Make it deal with functions with
3578 (mips_init_split_addresses): New, split out from override_options.
3579 (mips_init_relocs): New, split out from override_options.
3580 (was_mips16_p): New.
3581 (mips_set_mips16_mode): New, split out from override_options.
3582 (mips_set_current_function): New.
3583 (override_options): Add sorry for unsupported mips16/pic
3584 combination. Remove error for mips16/dsp combination. Save
3585 base option settings.
3586 (mips_file_start): Move mips16 mode setting output from here....
3587 (mips_output_function_prologue): ....to here.
3588 (mips_output_mi_thunk): Check for mips16 function.
3589 (build_mips16_function_stub): Don't set .mips16 here.
3590 (build_mips16_call_stub): Likewise.
3591 (mips_expand_builtin): Error in mips16 mode.
3592 (mips_use_mips16_mode_p): New.
3593 (mips_encode_section_info): Check for mips16 function, and set
3594 SYMBOL_REF_FLAGS accordingly.
3596 * doc/extend.texi (Function Attributes): Document new
3597 mips16/nomips16 attributes.
3598 * doc/invoke.texi (Option Summary): Add -mflip-mips16.
3599 (MIPS Options): Document -mflip-mips16.
3601 2007-09-05 Sandra Loosemore <sandra@codesourcery.com>
3603 * config/mips/mips.c (mips_legitimize_tls_address): Call sorry
3604 if we encounter TLS address in MIPS16 mode.
3605 (mips_legitimize_const_move): Check cannot_force_const_mem for
3606 the (const (plus symbol offset)) case; this forces invalid TLS
3607 address in MIPS16 mode to be caught by the above call to sorry.
3608 (override_options): Don't reset targetm.have_tls in MIPS16 mode,
3609 because that now enables emutls, which is not ABI compatible
3610 with native TLS in non-MIPS16 mode.
3612 2007-09-05 Sandra Loosemore <sandra@codesourcery.com>
3614 Add target hook invoked when cfun changes.
3616 * doc/tm.texi (TARGET_SET_CURRENT_FUNCTION): Document.
3617 * target.h (struct gcc_target): Add set_current_function.
3618 * target-def.h (TARGET_SET_CURRENT_FUNCTION): Define.
3619 (TARGET_INITIALIZER): Add initializer for set_current_function.
3620 * tree.h (push_struct_function): New.
3621 * tree-inline.h (push_cfun, pop_cfun): Move declarations to...
3623 (set_cfun): Declare.
3624 * tree-inline.c (cfun_stack, push_cfun, pop_cfun): Moved to...
3626 (push_function_context_to): Use allocate_struct_function
3627 to create null context, not init_dummy_function_start. Use set_cfun.
3628 (pop_function_context_from): Use set_cfun.
3629 (in_dummy_function): New.
3630 (invoke_set_current_function_hook): New.
3632 (push_cfun, pop_cfun): Use set_cfun.
3633 (push_struct_function): New.
3634 (allocate_struct_function): Call invoke_set_current_function_hook
3636 (prepare_function_start): Don't set cfun here. Remove unused
3637 argument; fix all callers.
3638 (init_dummy_function_start): Fiddle with in_dummy_function. Call
3639 push_struct_function.
3640 (init_function_start): Set cfun here.
3641 (expand_dummy_function_end): Fiddle with in_dummy_function. Pop cfun.
3642 * omp-low.c (create_omp_child_function): Use push_struct_function
3643 and pop_cfun to save/restore state.
3644 (expand_omp_parallel): Remove unused saved_cfun variable.
3645 * cgraphunit.c (ipa_passes): Use set_cfun.
3646 * gimple-low.c (record_vars_into): Use push_cfun/pop_cfun here.
3647 * dwarf2out.c (dwarf2out_abstract_function): Likewise.
3648 * matrix-reorg.c (transform_allocation_sites): Likewise.
3649 (matrix_reorg): Use set_cfun.
3650 * gimplify.c (gimplify_function_tree): Use push_cfun/pop_cfun here.
3651 * tree-optimize.c (tree_rest_of_compilation): Remove one redundant
3652 assignment to cfun; use set_cfun for the other.
3653 * tree-cfg.c (move_sese_region_to_fn): Use set_cfun.
3654 (dump_function_to_file): Use push_cfun/pop_cfun here.
3655 * c-decl.c (finish_function): Use set_cfun.
3657 2007-09-05 Kenneth Zadeck <zadeck@naturalbridge.com>
3659 * regrename.c (rerename_optimize): Use deferred rescanning and
3660 insert explicit calls to rescan insns when changed.
3662 2007-09-05 Jan Hubicka <jh@suse.cz>
3664 * optabs.c (libfunc_entry): Change optab to integer.
3665 (hash_libfunc, convert_optab_libfunc, optab_libfunc,
3666 set_optab_libfunc, set_conv_libfunc): Use optable indexes instead
3667 of pointers for hashing.
3669 2007-09-05 Jan Hubicka <jh@suse.cz>
3671 * regrename.c (pass_regrename, pass_cprop_hardreg): Add RTL sharing
3673 * fwprop.c (pass_rtl_fwprop, pass_rtl_fwprop_add): Likewise.
3674 * see.c (pass_see): Likewise.
3675 * tracer.c (pass_tracer): Likewise.
3676 * postreload-gcse.c (pass_gcse2): Likewise.
3677 * postreload.c (pass_postreload_cse): Likewise.
3678 * mode-switching.c (pass_mode_switching): Likewise.
3679 * modulo-sched.c (pass_sms): Likewise.
3680 * cse.c (cse_insn): Likewise.
3681 * web.c (pass_web): Likweise.
3682 * combine-stack-adj.c (pass_stack_adjustments): Likewise.
3683 * dce.c (pass_ud_rtl_dce, pass_fast_rtl_dce): Likewise.
3684 * loop-init.c (pass_rtl_loop_init): Likewise.
3685 (pass_rtl_loop_done, pass_rtl_move_loop_, pass_rtl_unswitch,
3686 pass_rtl_unroll_and, pass_rtl_doloop): Likewise.
3687 * global.c (pass_global_alloc): Likewise.
3688 * ifcvt.c (pass_rtl_ifcvt, pass_if_after_combine,
3689 pass_if_after_reload): Likewise.
3690 * reload.c (pass_peephole2, pass_split_for_shoren_branches): Likewise.
3691 * dse.c (pass_rtl_dse1, pass_rtl_dse2): Likewise.
3692 * regmove.c (pass_regmove): Likewise.
3693 * function.c (pass_thread_prologugues_epilogues): Likewise.
3694 * gcse.c (pass_gcse): Likewise.
3695 * rtl-factoring.c (pass_rtl_seqabstr): Likewise.
3696 * lower-subreg.c (pass_lower_subreg2): Likewise.
3697 * bt-load.c (pass_branch_target_load): Likewise.
3698 * emit-rtl.c (pass_unshare_all_rtl): Likewise.
3699 * cfgcleanup.c (pass_jump, pass_jump2): Likewise.
3700 * combine.c (pass_combine): Likewise.
3701 * bb-reorder.c (pass_duplicate_comp, pass_reorder_blocks): Likewise.
3702 (pass_partition_blocks): Likewise.
3703 * var-tracking.c (pass_variable_track): Likewise.
3704 * reg-stack.c (pass_stack_regs_run): Likewise.
3705 * sched-rgn.c (pass_sched, pass_sched2): Likewise.
3706 * passes.c (pass_postreload): Likewise.
3707 (execute_function_todo): Add TODO_verify_rtl_sharing handling code.
3708 * tree-pass.h (TODO_verify_rtl_sharing): New.
3709 (TODO_update_ssa, TODO_update_ssa_no_phi, TODO_update_ssa_full_phi,
3710 TODO_update_ssa_only_virtuals, TODO_remove_unused_locals,
3711 TODO_set_props, TODO_df_finish, TODO_df_verify,
3712 TODO_mark_first_instance, TODO_rebuild_alias): Renumber.
3714 2007-09-05 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
3716 * fold-const.c (all_ones_mask_p, sign_bit_p, simple_operand_p,
3717 fold_real_zero_addition_p, reorder_operands_p,
3718 div_if_zero_remainder, fold_undefer_overflow_warnings,
3719 int_binop_types_match_p, fold_convert_const_int_from_int,
3720 fold_convert_const_int_from_real,
3721 fold_convert_const_int_from_fixed,
3722 fold_convert_const_real_from_real,
3723 fold_convert_const_real_from_fixed,
3724 fold_convert_const_fixed_from_fixed,
3725 fold_convert_const_fixed_from_int,
3726 fold_convert_const_fixed_from_real, maybe_lvalue_p,
3727 fold_checksum_tree, fold_check_failed, print_fold_checksum,
3728 debug_fold_checksum, multiple_of_p): Constify.
3729 * tree-flow-inline.h (get_lineno): Likewise.
3730 * tree-flow.h (get_lineno): Likewise.
3731 * tree-object-size.c (compute_object_offset, addr_object_size,
3732 alloc_object_size, pass_through_call): Likewise.
3733 * tree-pretty-print.c (op_symbol, print_call_name,
3734 print_struct_decl, do_niy): Likewise.
3735 * tree.h (fold_undefer_overflow_warnings, multiple_of_p,
3736 debug_fold_checksum): Likewise.
3738 2007-09-04 Steven Bosscher <steven@gcc.gnu.org>
3741 * lower-subreg.c (resolve_clobber): If we remove a REG_LIBCALL
3742 note, remove the associated REG_RETVAL note.
3744 2007-09-04 Andrew Pinski <andrew_pinski@playstation.sony.com>
3746 * config.gcc (powerpc*-*-*): Install
3747 spu2vmx.h, vec_types.h, and si2vmx.h headers.
3748 * config/rs6000/spu2vmx.h: New header.
3749 * config/rs6000/si2vmx.h: New header.
3750 * config/rs6000/vec_types.h: New header.
3752 2007-09-05 Ben Elliston <bje@au.ibm.com>
3754 * varasm.c (initializer_constant_valid_p): Fix comment typo.
3756 2007-09-05 Ben Elliston <bje@au.ibm.com>
3758 * config/rs6000/ppu_intrinsics.h (__ldarx): Use `Z' constraint,
3760 (__stdcx): Make asm volatile.
3761 (__stwcx, __stdcx): Return only 0 or 1.
3763 2007-09-05 Jan Hubicka <jh@suse.cz>
3767 2007-09-04 Jan Hubicka <jh@suse.cz>
3768 * tree-tailcall.c (eliminate_tail_call): Expect unrenamed return value.
3770 2007-09-04 Richard Guenther <rguenther@suse.de>
3771 * tree-ssa-operands.c (add_virtual_operand): Only mark
3772 stores as has_volatile_ops if alias information is not available.
3774 2007-09-05 Jakub Jelinek <jakub@redhat.com>
3776 * config/rs6000/tramp.asm: Include config.h.
3777 Check __PIC__ or __pic__ macro instead of SHARED.
3779 PR rtl-optimization/32300
3780 * see.c (see_copy_insn): New function.
3781 (see_def_extension_not_merged, see_merge_one_use_extension,
3782 see_merge_one_def_extension): Use it. Avoid changing
3783 PREV_INSN/NEXT_INSN chains directly, insted emit insns
3784 into sequences. Call df_insn_delete on temporary insns
3785 that won't be emitted into the insn stream.
3786 (rest_of_handle_see): Turn off DF_DEFER_INSN_RESCAN
3787 and run df_process_deferred_rescans () before run_fast_dce.
3789 PR tree-optimization/33017
3790 * tree-data-ref.c (split_constant_offset) <case SSA_NAME>: Don't
3791 recurse for pure or const function calls.
3793 2007-09-04 Laurynas Biveinis <laurynas.biveinis@gmail.com>
3795 * c-format.c: Include alloc-pool.h.
3796 (check_format_info_main): New argument fwt_alloc. Use allocation
3797 pool instead of GC. Remove GC deallocation code.
3798 (check_format_arg): Create allocation pool, pass it to
3799 check_format_info_main and free it afterwards.
3800 * Makefile.in (c-format.o): Add alloc-pool.h dependency.
3802 2007-09-05 Ben Elliston <bje@au.ibm.com>
3804 * config/rs6000/ppu_intrinsics.h (__protected_stream_count):
3805 Rename count macro arugment for clarity.
3806 (__protected_unlimited_stream_set): Pass UG=1 to __dcbt_TH1000.
3807 (__protected_stream_set): Likewise, pass 0.
3809 2007-09-04 Jan Hubicka <jh@suse.cz>
3811 * tree-tailcall.c (eliminate_tail_call): Expect unrenamed return value.
3813 2007-09-04 Jan Hubicka <jh@suse.cz>
3815 * invoke.texi (-finline-small-functions): Document.
3816 * ipa-inline.c (cgraph_default_inline_p): Do not use DECL_INLINE
3817 when deciding what is inlinable.
3818 (cgraph_decide_recursive_inlining): Handle flag_inline_functions.
3819 (cgraph_decide_inlining_of_small_function): Handle new flags.
3820 (cgraph_decide_inlining_incrementally): Likewise.
3821 * opts.c (decode_options): Enable flag_inline_small_functions at -O2
3822 * common.opt (finline-small-functions): New.
3823 * Makefile.in (build/gengtype.o-warn): Work around PR29478
3825 2007-09-04 Richard Guenther <rguenther@suse.de>
3827 * tree-ssa-operands.c (add_virtual_operand): Only mark
3828 stores as has_volatile_ops if alias information is not available.
3830 2007-09-04 Andrew Haley <aph@redhat.com>
3832 * config/arm/libgcc-bpabi.ver: Add _Unwind_Backtrace as GCC_4.3.0.
3834 2007-09-04 Andrew Haley <aph@redhat.com>
3836 * config/arm/libunwind.S (UNWIND_WRAPPER _Unwind_Backtrace): New.
3837 * config/arm/unwind-arm.h (__gnu_Unwind_Backtrace): New.
3838 * config/arm/unwind-arm.c (__gnu_Unwind_Backtrace): New.
3840 2007-09-04 Emmanuel Thome <Emmanuel.Thome@inria.fr>
3843 * common.opt: Mention ABI version 2 in comment.
3845 2007-09-04 Jan Hubicka <jh@suse.cz>
3847 * optabs.c (debug_optab_libfunc): Update; make available to gdb.
3848 (libfunc_entry): New structure.
3849 (libfunc_hash): New hashtable.
3850 (hash_libfunc): New function.
3851 (eq_libfunc): New function.
3852 (convert_optab_libfunc): New function.
3853 (optab_libfunc): New function.
3854 (expand_binop, sign_expand_binop, expand_twoval_binop_libfunc,
3855 expand_unop, prepare_cmp_insn, prepare_float_insn, gen_add2_insn,
3856 expand_float, expand_fix, new_optab, new_convert_optab):
3857 Update for new libfunc API.
3858 (init_libfunc, init_integral_libfuncs,
3859 init_floating_libfuncs, init_interclass_conv_libfuncs
3860 init_intraclass_conv_libfuncs): Remove; reorganize all logic to:
3861 (gen_libfunc, gen_int_libfunc, gen_fp_libfunc, gen_int_fp_libfunc,
3862 gen_intv_fp_libfunc, gen_interclass_conv_libfunc,
3863 gen_int_to_fp_conv_libfunc, gen_ufloat_conv_libfunc,
3864 gen_int_to_fp_nondecimal_conv_libfunc, gen_fp_to_int_conv_libfunc,
3865 gen_intraclass_conv_libfunc, gen_trunc_conv_libfunc,
3866 gen_extend_conv_libfunc): New.
3867 (init_one_libfunc): Revamp for hashtables.
3868 (set_conv_libfunc): Likewise.
3869 (init_optabs): Initialize hashtable; use lazy initialization where
3871 * optabs.h (optab_handlers): Move out of GGC.
3872 (optab, convert_optab): Move out of GGC; add lazy gen info.
3873 (code_to_optab, convert_optab_table, optab_table): Move out of GGC.
3874 (optab_libfunc, convert_optab_libfunc): New.
3875 * builtins.c (expand_builtin_powi): Update for new API.
3876 * expr.c (convert_move): Likewise.
3877 * expmed.c (expand_divmod): Likewise.
3879 2007-09-04 Daniel Jacobowitz <dan@codesourcery.com>
3881 * config/rs6000/rs6000.c (rs6000_stack_info): Allocate space for the
3882 GOT pointer only if there is a constant pool. Use the allocated space
3885 2007-09-04 Daniel Jacobowitz <dan@codesourcery.com>
3887 * config/rs6000/linux-unwind.h (ppc_fallback_frame_state): Correct
3890 2007-09-04 Revital Eres <eres@il.ibm.com>
3891 Richard Guenther <rguenther@suse.de>
3892 R. Clint Whaley <whaley@cs.utsa.edu>
3894 * doc/invoke.texi (-fassociative-math, -freciprocal-math):
3896 * tree-tailcall.c (process_assignment): Use -fassociative-math
3897 when reodering operands of floating-point type.
3898 * fold-const.c (fold_comparison, fold_binary): Use
3899 -fassociative-math and -freciprocal-math instead of
3900 -funsafe-math-optimization flag.
3901 * toplev.h (set_unsafe_math_optimizations_flags): Declare function.
3902 * tree-ssa-math-opts.c (gate_cse_reciprocals): Use
3903 -freciprocal-math instead of -funsafe-math-optimizations.
3904 * opts.c (set_fast_math_flags): Set -freciprocal-math and
3905 -fassociative-math when -ffast-math is set.
3906 (set_unsafe_math_optimizations_flags): New Function
3907 to set -freciprocal-math and -fassociative-math when
3908 -funsafe-math-optimizations is set.
3909 (common_handle_option): Call it.
3910 * tree-vectorizer.c (vect_is_simple_reduction): Use
3911 -fassociative-math when doing reduction on floats.
3912 * loop-unroll.c (analyze_insn_to_expand_var): Use
3913 -fassociative-math when expanding an accumulator of type float.
3914 * simplify-rtx.c (simplify_binary_operation_1): Use
3915 -fassociative-math and -freciprocal-math when reordeing operands
3916 of floating-point type.
3917 * combine.c (combine_simplify_rtx): Likewise.
3918 * tree-ssa-reassoc.c (break_up_subtract_bb, reassociate_bb):
3920 * common.opt (-fassociative-math, -freciprocal-math): New flags.
3922 2007-09-04 Paolo Carlini <pcarlini@suse.de>
3925 * doc/install.texi (--enable-__cxa_atexit): Fix typo.
3927 2007-09-04 Jan Hubicka <jh@suse.cz>
3929 * regrename.c(copyprop_hardreg_forward_1): Unshare when doing copyprop.
3931 2007-09-04 Jan Hubicka <jh@suse.cz>
3933 * regrename.c (build_def_use): Don't share RTL between MATCH_OPERATOR
3934 and corresponding MATCH_DUP.
3936 2007-09-04 Uros Bizjak <ubizjak@gmail.com>
3939 * combine.c (subst): Do not try to simplify X if it represents load
3940 of FP constant from the constant pool via float extension.
3942 2007-09-04 Ben Elliston <bje@au.ibm.com>
3944 * c-opts.c: Include "tm_p.h".
3945 * Makefile.in (c-opts.o): Add $(TM_P_H) as a dependency.
3947 2007-09-04 Richard Guenther <rguenther@suse.de>
3949 PR tree-optimization/33291
3950 * tree-pretty-print.c (dump_generic_node): Dump all
3951 qualifiers for pointer types, not only first. Dump
3952 qualifiers for aggregate types as well.
3953 * tree-ssa-ccp.c (maybe_fold_offset_to_array_ref): Always
3954 use the canonical type for building ARRAY_REFs.
3955 * gimplify.c (canonicalize_addr_expr): Clean up. The