1 2009-01-15 Richard Sandiford <rdsandiford@googlemail.com>
3 * caller-save.c (add_used_regs_1, add_used_regs): New functions.
4 (insert_one_insn): Use them instead of REG_DEAD and REG_INC notes.
5 Also use them when walking CALL_INSN_FUNCTION_USAGE.
7 2009-01-15 H.J. Lu <hongjiu.lu@intel.com>
8 Joey Ye <joey.ye@intel.com>
11 * cfgexpand.c (expand_stack_alignment): Don't update stack
12 boundary nor check incoming stack boundary here.
13 (gimple_expand_cfg): Update stack boundary and check incoming
16 2009-01-15 Kenneth Zadeck <zadeck@naturalbridge.com>
18 * dce.c (find_call_stack_args, delete_unmarked_insns): Fixed
21 2009-01-14 Jakub Jelinek <jakub@redhat.com>
23 PR rtl-optimization/38245
24 * calls.c (expand_call): Add stack arguments to
25 CALL_INSN_FUNCTION_USAGE even for pure calls (when
26 ACCUMULATE_OUTGOING_ARGS) and even for args partially passed
27 in regs and partially in memory or BLKmode arguments.
28 (emit_library_call_value_1): Add stack arguments to
29 CALL_INSN_FUNCTION_USAGE even for pure calls (when
30 ACCUMULATE_OUTGOING_ARGS).
31 * dce.c: Include tm_p.h.
32 (find_call_stack_args): New function.
33 (deletable_insn_p): Call it for CALL_P insns. Add ARG_STORES
35 (mark_insn): Call find_call_stack_args for CALL_Ps.
36 (prescan_insns_for_dce): Walk insns backwards in bb rather than
37 forwards. Allocate and free arg_stores bitmap if needed, pass it
38 down to deletable_insn_p, don't mark stores set in arg_stores
39 bitmap, clear the bitmap at the beginning of each bb.
40 * Makefile.in (dce.o): Depend on $(TM_P_H).
42 2009-01-14 Michael Meissner <gnu@the-meissners.org>
45 * i386.c (print_operand): Add tests for 'D', 'C', 'F', 'f' to make
46 sure the insn is a conditional test (bug 22599). Reformat a few long
49 2009-01-14 Sebastian Pop <sebastian.pop@amd.com>
52 * graphite.c (get_vdef_before_scop, scop_adjust_vphi): New.
53 (scop_adjust_phis_for_liveouts): Call scop_adjust_vphi.
54 (gloog): Do not call cleanup_tree_cfg.
55 (graphite_transform_loops): Call cleanup_tree_cfg after all
56 scops have been code generated.
58 2009-01-14 Vladimir Makarov <vmakarov@redhat.com>
60 * testsuite/g++.dg/torture/pr38811.C: New file.
62 2009-01-14 Basile Starynkevitch <basile@starynkevitch.net>
63 * doc/gty.texi (Invoking the garbage collector): Added new node
64 and section documenting ggc_collect.
66 2009-01-14 Richard Guenther <rguenther@suse.de>
68 PR tree-optimization/38826
70 * tree-ssa-structalias.c (emit_alias_warning): Emit the pointer
71 initialization notes only if we actually emitted a warning.
72 (intra_create_variable_infos): Add constraints for a result decl
73 that is passed by hidden reference.
74 (build_pred_graph): Mark all related variables non-direct on
77 2009-01-14 Nick Clifton <nickc@redhat.com>
79 * ira-conflicts.c: Include addresses.h for the definition of
81 (ira_build_conflicts): Use base_reg_class instead of
83 * Makefile.in: Add a dependency of ira-conflicts.o on
86 2009-01-13 Vladimir Makarov <vmakarov@redhat.com>
89 * Makefile.in (ira-lives.o): Add except.h.
91 * ira-lives.c: Include except.h.
92 (process_bb_node_lives): Process can_throw_internal.
94 2009-01-13 Jakub Jelinek <jakub@redhat.com>
96 PR rtl-optimization/38774
97 * combine.c (simplify_set): When undoing cc_use change, don't do
98 PUT_CODE on the newly created comparison, but instead put back the
101 2009-01-13 Joseph Myers <joseph@codesourcery.com>
103 * doc/invoke.texi (ARM Options): Update lists of -mcpu and -march
104 values. Remove duplicate arm8 entry.
106 2009-01-13 Sebastian Pop <sebastian.pop@amd.com>
108 PR tree-optimization/38786
109 * graphite.c (expand_scalar_variables_ssa_name): New, outlined from
110 the SSA_NAME case of expand_scalar_variables_expr.
111 Set the type of an expression to the type of its assign statement.
112 (expand_scalar_variables_expr): Also gather the scalar computation
113 used to index the memory access. Do not pass loop_p.
114 Fix comment. Stop recursion on tcc_constant or tcc_declaration.
115 (expand_scalar_variables_stmt): Pass to expand_scalar_variables_expr
116 the gimple_stmt_iterator where it inserts new code. Do not pass loop_p.
117 (copy_bb_and_scalar_dependences): Do not pass loop_p.
118 (translate_clast): Update call to copy_bb_and_scalar_dependences.
120 2009-01-13 Sebastian Pop <sebastian.pop@amd.com>
122 * graphite.h (debug_value): Removed.
123 * graphite.c (debug_value): Removed.
125 2009-01-13 Richard Earnshaw <rearnsha@arm.com>
127 * arm.c (output_move_double): Don't synthesize thumb-2 ldrd/strd with
128 two 32-bit instructions.
130 2009-01-13 Richard Earnshaw <rearnsha@arm.com>
132 * arm.c (struct processors): Pass for speed down into cost helper
134 (const_ok_for_op): Handle COMPARE and inequality nodes.
135 (arm_rtx_costs_1): Rewrite.
136 (arm_size_rtx_costs): Update prototype.
137 (arm_rtx_costs): Pass speed down to helper functions.
138 (arm_slowmul_rtx_costs): Rework cost calculations.
139 (arm_fastmul_rtx_costs, arm_xscale_rtx_costs): Likewise.
140 (arm_9e_rtx_costs): Likewise.
142 2009-01-13 Uros Bizjak <ubizjak@gmail.com>
144 * config/alpha/alpha.c (alpha_legitimate_address_p): Explicit
145 relocations of local symbols wider than UNITS_PER_WORD are not valid.
146 (alpha_legitimize_address): Do not split local symbols wider than
147 UNITS_PER_WORD into HIGH/LO_SUM parts.
149 2009-01-13 Danny Smith <dannysmith@users.sourceforge.net>
152 * gcc.c (process_command): Replace call to execvp with calls
155 2009-01-03 Anatoly Sokolov <aesok@post.ru>
158 * config/avr/t-avr (LIB1ASMFUNCS): Add _tablejump_elpm.
159 * config/avr/libgcc.S (__do_global_ctors, __do_global_dtors): Add
160 variant for devices with 3-byte PC.
161 (__tablejump_elpm__) : New.
163 2009-01-12 Jakub Jelinek <jakub@redhat.com>
166 * c-parser.c (c_parser_postfix_expression): Allow `->' in
167 offsetof member-designator, handle it as `[0].'.
169 2009-01-12 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
171 * pa.c (pa_asm_output_mi_thunk): Use pc-relative branch to thunk
172 function when not using named sections on targets with named sections
173 if branch distance is less than 262132.
175 2009-01-12 Richard Earnshaw <rearnsha@arm.com>
177 * combine.c (combine_instructions): Recompute
178 optimize_this_for_speed_p for each BB in the main combine loop.
180 2009-01-12 Tomas Bily <tbily@suse.cz>
183 * tree-loop-distribution.c (prop_phis): New function.
184 (generate_builtin): Call prop_phis.
185 * testsuite/gcc.dg/tree-ssa/pr38385.c: New file.
187 2009-01-12 Jakub Jelinek <jakub@redhat.com>
189 PR tree-optimization/38807
190 * tree-ssa-reassoc.c (remove_visited_stmt_chain): Don't look at
191 gimple_visited_p unless stmt is GIMPLE_ASSIGN.
193 2009-01-11 Adam Nemet <anemet@caviumnetworks.com>
195 * expmed.c (store_bit_field_1): Properly truncate the paradoxical
196 subreg of op0 to the original op0.
198 2009-01-11 Laurent GUERBY <laurent@guerby.net>
200 * doc/sourcebuild.texi (Source Tree): Move up intl and fixinc.
202 2009-01-11 Markus Schoepflin <markus.schoepflin@comsoft.de>
205 * gcc/mips-tfile.c (parse_def): Fix parsing of def strings
206 starting with digits.
208 2009-01-10 Jakub Jelinek <jakub@redhat.com>
211 * config/arm/arm.c (arm_is_long_call_p): Don't call
212 arm_function_in_section_p if decl isn't a FUNCTION_DECL.
214 2009-01-09 Steven Bosscher <steven@gcc.gnu.org>
216 * regrename.c (regrename_optimize): Fix dumping.
217 (find_oldest_value_reg): Preserve REG_POINTER.
218 (copy_hardreg_forward_1): Likewise.
220 2009-01-09 Diego Novillo <dnovillo@google.com>
222 * gimple.h (struct gimple_statement_base)<uid>: Document
223 the restrictions on its use.
225 (gimple_set_uid): Tidy.
227 2009-01-09 Jakub Jelinek <jakub@redhat.com>
229 * config/i386/i386.c (ix86_expand_movmem, ix86_expand_setmem): Add
230 zero guard even if align_bytes != 0 and count is smaller than
233 2008-01-09 Vladimir Makarov <vmakarov@redhat.com>
235 PR rtl-optimization/38495
236 * ira-emit.c (print_move_list, ira_debug_move_list): New functions.
237 (add_range_and_copies_from_move_list): Print all added ranges.
238 Add ranges to memory optimized destination.
240 2009-01-09 Jakub Jelinek <jakub@redhat.com>
244 * config/i386/i386.c (override_options): Reject
245 -mstringop-strategy=rep_8byte with -m32.
246 (ix86_expand_movmem): For size_needed == 1 set epilogue_size_needed
247 to 1. Do count comparison against epilogue_size_needed at compile
248 time even when count_exp was constant forced into register. For
249 size_needed don't jump to epilogue, instead just avoid aligning
250 and invoke the body algorithm. If need_zero_guard, add zero guard
251 even if count is non-zero, but smaller than size_needed + number of
252 bytes that could be stored for alignment.
253 (ix86_expand_setmem): For size_needed == 1 set epilogue_size_needed
254 to 1. If need_zero_guard, add zero guard even if count is non-zero,
255 but smaller than size_needed + number of bytes that could be stored
256 for alignment. Compare size_needed with epilogue_size_needed instead
257 of desired_align - align, don't adjust size_needed, pass
258 epilogue_size_needed to the epilogue expanders.
261 * c-pretty-print.c (pp_c_expression): Handle GOTO_EXPR like BIND_EXPR.
263 2009-01-09 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
265 * pa.c (last_address): Change to unsigned.
266 (update_total_code_bytes): Change argument to unsigned. Don't
267 check if insn addresses are set.
268 (pa_output_function_epilogue): Set last_address to UINT_MAX if insn
269 addresses are not set.
270 (pa_asm_output_mi_thunk): Handle wrap when updating last_address.
272 2009-01-09 Nick Clifton <nickc@redhat.com>
274 * config/sh/symbian.c: Replace uses of DECL_INLINE with
275 DECL_DECLARED_INLINE_P.
277 2009-01-09 Jakub Jelinek <jakub@redhat.com>
280 * dojump.c (do_jump_by_parts_zero_rtx): Use mode instead of
281 GET_MODE (op0) in operand_subword_force calls.
284 * fold-const.c (fold_unary): For COMPOUND_EXPR and COND_EXPR,
285 fold_convert arg0 operands to TREE_TYPE (op0) first.
287 2009-01-08 Vladimir Makarov <vmakarov@redhat.com>
289 * params.def (ira-max-conflict-table-size): Decrease default value
292 2009-01-08 Jakub Jelinek <jakub@redhat.com>
294 PR tree-optimization/37031
295 * lambda-code.c (lambda_collect_parameters): Call pointer_set_destroy
297 (build_access_matrix): Reserve correct size for AM_MATRIX vector,
298 allocate it using gc instead of heap, use VEC_quick_push instead of
300 * graphite.c (build_access_matrix): Allocate AM_MATRIX vector using gc
301 instead of heap, use VEC_quick_push instead of VEC_safe_push.
302 * tree-data-ref.h (struct access_matrix): Change matrix to gc
303 allocated vector from heap allocated.
304 * lambda.h: Add DEF_VEC_ALLOC_P for gc allocated lambda_vector.
305 * tree-loop-linear.c (linear_transform_loops): Allocate nest
306 vector only after perfect_loop_nest_depth call.
308 2009-01-08 Sebastian Pop <sebastian.pop@amd.com>
309 Jan Sjodin <jan.sjodin@amd.com>
311 PR tree-optimization/38559
312 * graphite.c (debug_value, copy_constraint,
313 swap_constraint_variables, scale_constraint_variable, ): New.
314 (get_lower_bound, get_upper_bound): Removed.
315 (graphite_trans_bb_strip_mine): Clean up this code that works
316 only for constant number of iterations. Fully copy upper and
317 lower bound constraints, not only the constant part of them.
318 * graphite.h (debug_value): Declared.
320 2009-01-08 Ira Rosen <irar@il.ibm.com>
322 PR tree-optimization/37194
323 * tree-vect-transform.c (vect_estimate_min_profitable_iters):
324 Don't add the cost of cost model guard in prologue to scalar
325 outside cost in case of known number of iterations.
327 2009-01-07 Nathan Froyd <froydnj@codesourcery.com>
328 Alan Modra <amodra@bigpond.net.au>
330 * config/rs6000/rs6000.c (rs6000_legitimize_address): Check for
331 non-word-aligned REG+CONST addressing.
333 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
336 * config/alpha/alpha.c (alpha_end_function): For TARGET_ABI_OSF, call
337 free_after_compilation when outputting a thunk.
338 (alpha_output_mi_thunk_osf): Assert that we are processing a thunk.
339 Do not call free_after_compilation here.
341 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
343 * config/i386/i386.c (ix86_target_string): Use ARRAY_SIZE.
344 (ix86_valid_target_attribute_inner_p): Ditto.
346 2009-01-07 Jan Sjodin <jan.sjodin@amd.com>
348 PR tree-optimization/38492
349 PR tree-optimization/38498
350 * tree-check.c (operator_is_linear, scev_is_linear_expression): New.
351 * tree-chrec.h (scev_is_linear_expression): Declared.
352 * graphite.c (graphite_cannot_represent_loop_niter): New.
353 (scopdet_basic_block_info): Call graphite_cannot_represent_loop_niter.
354 (graphite_loop_normal_form): Use gcc_assert.
355 (scan_tree_for_params): Use CASE_CONVERT.
356 (phi_node_is_iv, bb_contains_non_iv_scalar_phi_nodes): New.
357 (build_scop_conditions_1): Call bb_contains_non_iv_scalar_phi_nodes.
358 Use gcc_assert. Discard scops that contain unhandled cases.
359 (build_scop_conditions): Return a boolean status for unhandled cases.
360 (strip_mine_profitable_p): Print the loop number, not its depth.
361 (is_interchange_valid): Pass the depth of the loop nest, don't
362 recompute it wrongly.
363 (graphite_trans_bb_block): Same.
364 (graphite_trans_bb_block): Print tentative of loop blocking.
365 (graphite_trans_scop_block): Do not print that the loop has been
367 (graphite_transform_loops): Do not handle scops that contain condition
370 2009-01-07 H.J. Lu <hongjiu.lu@intel.com>
372 AVX Programming Reference (December, 2008)
373 * config/i386/avxintrin.h (_mm256_stream_si256): New.
374 (_mm256_stream_pd): Likewise.
375 (_mm256_stream_ps): Likewise.
377 * config/i386/i386.c (ix86_builtins): Add IX86_BUILTIN_MOVNTDQ256,
378 IX86_BUILTIN_MOVNTPD256 and IX86_BUILTIN_MOVNTPS256.
379 (ix86_special_builtin_type): Add VOID_FTYPE_PV4DI_V4DI.
380 (bdesc_special_args): Add __builtin_ia32_movntdq256,
381 __builtin_ia32_movntpd256 and __builtin_ia32_movntps256.
382 (ix86_init_mmx_sse_builtins): Handle VOID_FTYPE_PV4DI_V4DI.
383 (ix86_expand_special_args_builtin): Likewise.
385 * config/i386/sse.md (AVXMODEDI): New.
386 (avx_movnt<mode>): Likewise.
387 (avx_movnt<mode>): Likewise.
388 (<sse>_movnt<mode>): Remove AVX support.
389 (sse2_movntv2di): Likewise.
391 2009-01-07 Richard Guenther <rguenther@suse.de>
394 * fold-const.c (extract_muldiv): Remove obsolete comment.
395 (fold_plusminus_mult_expr): Undo MINUS_EXPR
396 to PLUS_EXPR canonicalization for the canonicalization.
398 2009-01-07 Gerald Pfeifer <gerald@pfeifer.com>
400 * doc/install.texi (alpha*-dec-osf*): Remove note on 32-bit
401 hosted cross-compilers generating less efficient code.
403 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
405 * function.h (rtl_data): Add a dbr_scheduled_p field.
406 * reorg.c (dbr_schedule): Set it.
407 (gate_handle_delay_slots): Check it.
408 * config/mips/mips.c (mips_base_delayed_branch): Delete.
409 (mips_reorg): Check flag_delayed_branch instead of
410 mips_base_delayed_branch.
411 (mips_override_options): Don't set mips_base_delayed_branch
412 or flag_delayed_branch.
414 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
416 PR rtl-optimization/38426.
417 * ira.c (ira): Set current_function_is_leaf earlier.
419 2009-01-06 Jakub Jelinek <jakub@redhat.com>
421 PR rtl-optimization/38722
422 * combine.c (try_combine): Don't modify PATTERN (i3) and notes
423 too early, only set a flag and modify after last possible
426 2009-01-06 Janis Johnson <janis187@us.ibm.com>
429 * ginclude/float.h: Rename DECnn_DEN to DECnn_SUBNORMAL_MIN.
430 * real.c (decimal_single_format): Correct values of emin and emax.
431 (decimal_double_format): Ditto.
432 (decimal_quad_format): Ditto.
433 * c-cppbuiltin.c (builtin_define_decimal_float_constants): Adjust
434 computation of DECnn_MIN and DECnn_MAX for corrected values of
435 emin and emax. Define __DECnn_SUBNORMAL_MIN__ instead of
436 __DECnn_MIN__, and adjust its computation for the corrected value
439 2009-01-06 Jan Hubicka <jh@suse.cz>
442 * i386.c (ix86_expand_call): Use ARRAY_SIZE.
444 2009-01-06 Gerald Pfeifer <gerald@pfeifer.com>
446 * doc/contrib.texi (Contributors): Slightly adjust the end note.
447 Add Robert Clark to the list of testers.
449 2009-01-06 Jan Hubicka <jh@suse.cz>
450 Kai Tietz <kai.tietz@onevision.com>
452 * i386.md (*msabi_syvabi): Add SSE regs clobbers.
453 * i386.c (ix86_expand_call): Add clobbers.
455 2009-01-06 Jan Hubicka <jh@suse.cz>
456 Kai Tietz <kai.tietz@onevision.com>
458 * i386.h (CONDITIONAL_CALL_USAGE): SSE regs are not used for w64 ABI.
459 * i386.c (struct ix86_frame): Add padding0 and nsseregs.
460 (ix86_nsaved_regs): Count only general purpose regs.
461 (ix86_nsaved_sseregs): New.
462 (ix86_compute_frame_layout): Update nsseregs; set preferred alignment
463 to 16 for w64; compute padding and size of sse reg save area.
464 (ix86_emit_save_regs, ix86_emit_save_regs_using_mov): Save only
465 general purpose regs.
466 (ix86_emit_save_sse_regs_using_mov): New.
467 (ix86_expand_prologue): Save SSE regs if needed.
468 (ix86_emit_restore_regs_using_mov): Use only general purpose regs.
469 (ix86_emit_restore_sse_regs_using_mov): New.
470 (ix86_expand_epilogue): Save SSE regs if needed.
472 2009-01-06 Jan Hubicka <jh@suse.cz>
473 Kai Tietz <kai.tietz@onevision.com>
475 * i386.h (ACCUMULATE_OUTGOING_ARGS): Enable for MSABI
476 * i386.c (init_cumulative_args): Disallow calls of MSABI functions
477 when accumulate outgoing args is off.
479 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
482 * ira-color.c (ira_reuse_stack_slot): Check ENABLE_IRA_CHECKING
483 before using pseudos_have_intersected_live_ranges_p.
485 * ira-int.h (ira_assert): Always define.
487 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
489 AVX Programming Reference (December, 2008)
490 * config/i386/avxintrin.h (_mm_permute2_pd): Removed.
491 (_mm256_permute2_pd): Likewise.
492 (_mm_permute2_ps): Likewise.
493 (_mm256_permute2_ps): Likewise.
494 * config/i386/i386.md (UNSPEC_VPERMIL2): Likewise.
495 * config/i386/sse.md (avx_vpermil2<mode>3): Likewise.
497 * config/i386/i386.c (ix86_builtins): Remove
498 IX86_BUILTIN_VPERMIL2PD, IX86_BUILTIN_VPERMIL2PS,
499 IX86_BUILTIN_VPERMIL2PD256 and IX86_BUILTIN_VPERMIL2PS256.
500 (ix86_builtin_type): Remove V8SF_FTYPE_V8SF_V8SF_V8SI_INT,
501 V4DF_FTYPE_V4DF_V4DF_V4DI_INT, V4SF_FTYPE_V4SF_V4SF_V4SI_INT
502 and V2DF_FTYPE_V2DF_V2DF_V2DI_INT.
503 (bdesc_args): Remove __builtin_ia32_vpermil2pd,
504 __builtin_ia32_vpermil2ps, __builtin_ia32_vpermil2pd256 and
505 __builtin_ia32_vpermil2ps256.
506 (ix86_init_mmx_sse_builtins): Updated.
507 (ix86_expand_args_builtin): Likewise.
509 2009-01-05 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
511 * pa.c (output_call): Relocate non-jump insns in the delay slot of
512 long absolute calls when generating PA 2.0 code.
514 2009-01-05 Vladimir Makarov <vmakarov@redhat.com>
516 PR rtl-optimization/38583
517 * params.h (IRA_MAX_CONFLICT_TABLE_SIZE): New macro.
519 * params.def (ira-max-conflict-table-size): New.
521 * doc/invoke.texi (ira-max-conflict-table-size): Decribe.
523 * ira.h (ira_conflicts_p): New external definition.
525 * ira-conflicts.c (build_conflict_bit_table): Do not build too big
526 table. Report this. Return result of building.
527 (ira_build_conflicts): Use ira_conflicts_p. Check result of
528 building conflict table.
530 * ira-color.c (fast_allocation): Use num instead of ira_allocnos_num.
531 (ira_color): Use ira_conflicts_p.
533 * global.c: Include ira.h.
534 (pseudo_for_reload_consideration_p, build_insn_chain): Use
537 * Makefile.in (global.o): Add ira.h.
539 * ira-build.c (mark_all_loops_for_removal,
540 propagate_some_info_from_allocno): New.
541 (remove_unnecessary_allocnos): Call
542 propagate_some_info_from_allocno.
543 (remove_low_level_allocnos): New.
544 (remove_unnecessary_regions): Add parameter. Call
545 mark_all_loops_for_removal and remove_low_level_allocnos. Pass
546 parameter to remove_unnecessary_regions.
547 (ira_build): Remove all regions but root if the conflict table was
548 not built. Update conflict hard regs for allocnos crossing calls.
550 * ira.c (ira_conflicts_p): New global.
551 (ira): Define and use ira_conflicts_p.
553 * reload1.c (compute_use_by_pseudos, reload, count_pseudo,
554 count_spilled_pseudo, find_reg, alter_reg, finish_spills,
555 emit_input_reload_insns, delete_output_reload): Use ira_conflicts_p.
557 2009-01-06 Ben Elliston <bje@au.ibm.com>
559 * gengtype-lex.l (YY_NO_INPUT): Define.
561 2009-01-05 Andrew Pinski <andrew_pinski@playstation.sony.com>
564 * c-common.c (handle_vector_size_attribute): Also reject
567 2009-01-05 Sebastian Pop <sebastian.pop@amd.com>
569 PR tree-optimization/38492
570 * graphite.c (rename_map_elt, debug_rename_elt,
571 debug_rename_map_1, debug_rename_map, new_rename_map_elt,
572 rename_map_elt_info, eq_rename_map_elts,
573 get_new_name_from_old_name, bb_in_sese_p): Moved around.
574 (sese_find_uses_to_rename_use): Renamed sese_build_livein_liveouts_use.
575 (sese_find_uses_to_rename_bb): Renamed sese_build_livein_liveouts_bb.
576 (sese_build_livein_liveouts): New.
577 (new_sese, free_sese): New.
578 (new_scop): Call new_sese.
579 (free_scop): Call free_sese.
580 (rename_variables_from_edge, rename_phis_end_scop): Removed.
581 (register_old_new_names): Renamed register_old_and_new_names.
582 (register_scop_liveout_renames, add_loop_exit_phis,
583 insert_loop_close_phis, struct igp,
584 default_liveout_before_guard, add_guard_exit_phis,
585 insert_guard_phis, copy_renames): New.
586 (translate_clast): Call insert_loop_close_phis and insert_guard_phis.
587 (sese_add_exit_phis_edge): Renamed scop_add_exit_phis_edge.
588 (rewrite_into_sese_closed_ssa): Renamed scop_insert_phis_for_liveouts.
589 (scop_adjust_phis_for_liveouts): New.
590 (gloog): Call scop_adjust_phis_for_liveouts.
592 * graphite.h (struct sese): Documented. Added fields liveout,
594 (SESE_LIVEOUT, SESE_LIVEIN, SESE_LIVEIN_VER, SESE_NUM_VER): New.
595 (new_sese, free_sese, sese_build_livein_liveouts): Declared.
596 (struct scop): Added field liveout_renames.
597 (SCOP_LIVEOUT_RENAMES): New.
599 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
601 PR tree-optimization/38510
602 * graphite.c (recompute_all_dominators): Call mark_irreducible_loops.
603 (translate_clast): Call recompute_all_dominators before
605 (gloog): Call recompute_all_dominators before graphite_verify.
607 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
608 Jan Sjodin <jan.sjodin@amd.com>
610 PR tree-optimization/38500
611 * graphite.c (create_sese_edges): Call fix_loop_structure after
614 2009-01-05 Joel Sherrill <joel.sherrill@oarcorp.com>
616 * config.gcc: Add m32r*-*-rtems*.
617 * config/m32r/rtems.h: New file.
619 2009-01-05 Ben Elliston <bje@au.ibm.com>
621 * Makefile.in (.po.gmo): Use mkinstalldirs, not test -d || mkdir.
623 (po/gcc.pot): Likewise.
625 2009-01-04 David S. Miller <davem@davemloft.net>
627 * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
628 (STARTING_FRAME_OFFSET): Always set to zero.
630 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
632 * tree.def (LSHIFT_EXPR, RSHIFT_EXPR): Add commentary.
633 * tree-cfg.c (verify_gimple_assign_binary): Allow shifts of
634 fixed-point types, and vectors of the same.
636 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
638 * config/mips/sync.md (*mb_barrier): Rename to...
639 (*memory_barrier): ...this.
641 2009-01-04 Jonathan Wakely <jwakely.gcc@gmail.com>
643 * doc/extend.texi (Function Attributes): Move @cindex after @item
644 for 'artificial' and 'flatten'. Fix grammar for 'externally_visible'
645 and put in alphabetical order. Fix 'target' name and put in order.
646 * doc/invoke.texi (-Wstrict-null-sentinel, -fipa-matrix-reorg): Fix
649 2009-01-04 Uros Bizjak <ubizjak@gmail.com>
651 * config/s390/s390.md (UNSPEC_MB): Rename from UNSPECV_MB.
652 (memory_barrier): Expand as unspec instead of unspec_volatile.
653 Remove mem:BLK from insn operands. Use Pmode scratch register.
654 (*memory_barrier): Define as unspec instead of unspec_volatile.
655 Use (match_dup 0) as input operand.
657 * config/sparc/sparc.md (UNSPEC_MEMBAR): Rename from UNSPECV_MEMBAR.
658 * config/sparc/sync.md (memory_barrier): Expand as unspec instead of
659 unspec_volatile. Remove mem:BLK from insn operands. Use Pmode
660 scratch register. Remove operand 1.
661 (*stbar): Define as unspec instead of unspec_volatile.
662 Use (match_dup 0) as input operand, remove (const_int 8).
663 (*membar): Define as unspec instead of unspec_volatile.
664 Use (match_dup 0) as input operand, remove input operand 2.
666 * config/xtensa/xtensa.md (UNSPEC_MEMW): Rename from UNSPECV_MEMW.
667 (memory_barrier): Expand as unspec instead of unspec_volatile.
668 Remove mem:BLK from insn operands. Use Pmode scratch register.
669 (*memory_barrier): Define as unspec instead of unspec_volatile.
670 Use (match_dup 0) as input operand.
672 * config/ia64/sync.md (memory_barrier): Redefine as expander pattern.
673 Remove mem:BLK from insn operands. Use Pmode scratch register.
674 Set volatile flag on operand 0.
675 (*memory_barrier): New insn pattern.
677 * config/rs6000/sync.md (memory_barrier): Remove mem:BLK from
679 (*memory_barrier): Use (match_dup 0) as input operand.
681 * config/mips/sync.md (memory_barrier): Redefine as expander pattern.
682 Remove mem:BLK from insn operands. Use Pmode scratch register.
683 Set volatile flag on operand 0.
684 (*mb_internal): New insn pattern.
686 * config/alpha/sync.md (*memory_barrier): Rename from *mb_internal.
688 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
691 * function.c (struct temp_slot): Move to the section of the file
692 that deals with temp slots. Remove field 'address'.
693 (temp_slot_address_table): New hash table of address -> temp slot.
694 (struct temp_slot_address_entry): New struct, items for the table.
695 (temp_slot_address_compute_hash, temp_slot_address_hash,
696 temp_slot_address_eq, insert_temp_slot_address): Support functions
698 (find_temp_slot_from_address): Rewrite to use the new hash table.
699 (remove_unused_temp_slot_addresses): Remove addresses of temp
700 slots that have been made available.
701 (remove_unused_temp_slot_addresses_1): Call-back for htab_traverse,
702 worker function for remove_unused_temp_slot_addresses.
703 (assign_stack_temp_for_type): Don't clear the temp slot address list.
704 Add the temp slot address to the address -> temp slot map.
705 (update_temp_slot_address): Update via insert_temp_slot_address.
706 (free_temp_slots): Call remove_unused_temp_slot_addresses.
707 (pop_temp_slots): Likewise.
708 (init_temp_slots): Allocate the address -> temp slot map, or empty
709 the map if it is already allocated.
710 (prepare_function_start): Initialize temp slot processing.
712 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
715 * cfgexpand.c (estimate_stack_frame_size): Simplify the estimate:
716 Calculate the size of all stack vars assuming no packing of stack
717 vars will happen, replacing a quadratic algorithm with a linear one.
719 2009-01-03 Jakub Jelinek <jakub@redhat.com>
722 * expmed.c (store_bit_field_1): Don't modify op0 if movstrict insn
725 2009-01-03 Diego Novillo <dnovillo@google.com>
727 * doc/contrib.texi: Update contributions.
729 2009-01-03 Jakub Jelinek <jakub@redhat.com>
732 * builtins.c (fold_builtin_memory_op): Give up if either operand
733 is volatile. Set srctype or desttype to non-qualified version
737 * builtins.c (fold_builtin_expect): Only check DECL_WEAK for VAR_DECLs
740 2009-01-02 Kenneth Zadeck <zadeck@naturalbridge.com>
742 PR rtl-optimization/35805
743 * df-problems.c (df_lr_finalize): Add recursive call to resolve lr
744 problem if fast dce is able to remove any instructions.
745 * dce.c (dce_process_block): Fix dump message.
747 2009-01-02 Mark Mitchell <mark@codesourcery.com>
750 * tree-ssa-pre.c (compute_antic): Correct loop bounds.
752 2009-01-02 Jakub Jelinek <jakub@redhat.com>
755 * tree-flow.h (op_code_prio, op_prio): New prototypes.
756 * tree-pretty-print.c (op_code_prio): New function.
757 (op_prio): No longer static. Use op_code_prio.
758 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs):
759 Use op_prio and op_code_prio to determine if () should be
760 printed around operand(s) or not.
762 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs,
763 dump_gimple_call, dump_gimple_switch, dump_gimple_cond,
764 dump_gimple_label, dump_gimple_try, dump_symbols, dump_gimple_phi,
765 dump_gimple_mem_ops, dump_bb_header, dump_bb_end, pp_cfg_jump): Use
766 pp_character instead of pp_string for single letter printing.
768 2009-01-02 Richard Sandiford <rdsandiford@googlemail.com>
770 * doc/extend.texi: Fix '#pragma GCC option' typo.
772 2009-01-02 Richard Guenther <rguenther@suse.de>
774 * doc/install.texi (--enable-checking): Mention different
776 (--enable-stage1-checking): Document.
778 2009-01-01 Andrew Pinski <pinskia@gmail.com>
781 * tree-cfg.c (verify_expr): Add INDIRECT_REF case. Change MODIFY_EXPR
784 2009-01-02 Ben Elliston <bje@au.ibm.com>
786 * config/fp-bit.h (pack_d): Constify argument.
787 * config/fp-bit.c (makenan): Constify return type. Remove casts.
788 (isnan): Constify argument.
792 (_fpadd_parts): Constify return type.
793 (_fpmul_parts): Likewise.
794 (_fpdiv_parts): Likewise.
796 2009-01-01 Jakub Jelinek <jakub@redhat.com>
799 * c-typeck.c (add_pending_init): Add IMPLICIT argument. Only
800 warn about overwriting initializer with side-effects or
801 -Woverride-init if !IMPLICIT.
802 (output_init_element): Likewise. Pass IMPLICIT down to
804 (process_init_element): Add IMPLICIT argument. Pass it down
805 to output_init_element.
806 (push_init_element, pop_init_level, set_designator): Adjust
807 process_init_element callers.
808 (set_nonincremental_init, set_nonincremental_init_from_string):
809 Adjust add_pending_init callers.
810 (output_pending_init_elements): Adjust output_init_element callers.
811 * c-tree.h (process_init_element): Adjust prototype.
812 * c-parser.c (c_parser_initelt, c_parser_initval): Adjust
813 process_init_element callers.