1 2009-01-13 Jakub Jelinek <jakub@redhat.com>
3 PR rtl-optimization/38774
4 * combine.c (simplify_set): When undoing cc_use change, don't do
5 PUT_CODE on the newly created comparison, but instead put back the
8 2009-01-13 Joseph Myers <joseph@codesourcery.com>
10 * doc/invoke.texi (ARM Options): Update lists of -mcpu and -march
11 values. Remove duplicate arm8 entry.
13 2009-01-13 Sebastian Pop <sebastian.pop@amd.com>
15 PR tree-optimization/38786
16 * graphite.c (expand_scalar_variables_ssa_name): New, outlined from
17 the SSA_NAME case of expand_scalar_variables_expr.
18 Set the type of an expression to the type of its assign statement.
19 (expand_scalar_variables_expr): Also gather the scalar computation
20 used to index the memory access. Do not pass loop_p.
21 Fix comment. Stop recursion on tcc_constant or tcc_declaration.
22 (expand_scalar_variables_stmt): Pass to expand_scalar_variables_expr
23 the gimple_stmt_iterator where it inserts new code. Do not pass loop_p.
24 (copy_bb_and_scalar_dependences): Do not pass loop_p.
25 (translate_clast): Update call to copy_bb_and_scalar_dependences.
27 2009-01-13 Sebastian Pop <sebastian.pop@amd.com>
29 * graphite.h (debug_value): Removed.
30 * graphite.c (debug_value): Removed.
32 2009-01-13 Richard Earnshaw <rearnsha@arm.com>
34 * arm.c (output_move_double): Don't synthesize thumb-2 ldrd/strd with
35 two 32-bit instructions.
37 2009-01-13 Richard Earnshaw <rearnsha@arm.com>
39 * arm.c (struct processors): Pass for speed down into cost helper
41 (const_ok_for_op): Handle COMPARE and inequality nodes.
42 (arm_rtx_costs_1): Rewrite.
43 (arm_size_rtx_costs): Update prototype.
44 (arm_rtx_costs): Pass speed down to helper functions.
45 (arm_slowmul_rtx_costs): Rework cost calculations.
46 (arm_fastmul_rtx_costs, arm_xscale_rtx_costs): Likewise.
47 (arm_9e_rtx_costs): Likewise.
49 2009-01-13 Uros Bizjak <ubizjak@gmail.com>
51 * config/alpha/alpha.c (alpha_legitimate_address_p): Explicit
52 relocations of local symbols wider than UNITS_PER_WORD are not valid.
53 (alpha_legitimize_address): Do not split local symbols wider than
54 UNITS_PER_WORD into HIGH/LO_SUM parts.
56 2009-01-13 Danny Smith <dannysmith@users.sourceforge.net>
59 * gcc.c (process_command): Replace call to execvp with calls
62 2009-01-03 Anatoly Sokolov <aesok@post.ru>
65 * config/avr/t-avr (LIB1ASMFUNCS): Add _tablejump_elpm.
66 * config/avr/libgcc.S (__do_global_ctors, __do_global_dtors): Add
67 variant for devices with 3-byte PC.
68 (__tablejump_elpm__) : New.
70 2009-01-12 Jakub Jelinek <jakub@redhat.com>
73 * c-parser.c (c_parser_postfix_expression): Allow `->' in
74 offsetof member-designator, handle it as `[0].'.
76 2009-01-12 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
78 * pa.c (pa_asm_output_mi_thunk): Use pc-relative branch to thunk
79 function when not using named sections on targets with named sections
80 if branch distance is less than 262132.
82 2009-01-12 Richard Earnshaw <rearnsha@arm.com>
84 * combine.c (combine_instructions): Recompute
85 optimize_this_for_speed_p for each BB in the main combine loop.
87 2009-01-12 Tomas Bily <tbily@suse.cz>
90 * tree-loop-distribution.c (prop_phis): New function.
91 (generate_builtin): Call prop_phis.
92 * testsuite/gcc.dg/tree-ssa/pr38385.c: New file.
94 2009-01-12 Jakub Jelinek <jakub@redhat.com>
96 PR tree-optimization/38807
97 * tree-ssa-reassoc.c (remove_visited_stmt_chain): Don't look at
98 gimple_visited_p unless stmt is GIMPLE_ASSIGN.
100 2009-01-11 Adam Nemet <anemet@caviumnetworks.com>
102 * expmed.c (store_bit_field_1): Properly truncate the paradoxical
103 subreg of op0 to the original op0.
105 2009-01-11 Laurent GUERBY <laurent@guerby.net>
107 * doc/sourcebuild.texi (Source Tree): Move up intl and fixinc.
109 2009-01-11 Markus Schoepflin <markus.schoepflin@comsoft.de>
112 * gcc/mips-tfile.c (parse_def): Fix parsing of def strings
113 starting with digits.
115 2009-01-10 Jakub Jelinek <jakub@redhat.com>
118 * config/arm/arm.c (arm_is_long_call_p): Don't call
119 arm_function_in_section_p if decl isn't a FUNCTION_DECL.
121 2009-01-09 Steven Bosscher <steven@gcc.gnu.org>
123 * regrename.c (regrename_optimize): Fix dumping.
124 (find_oldest_value_reg): Preserve REG_POINTER.
125 (copy_hardreg_forward_1): Likewise.
127 2009-01-09 Diego Novillo <dnovillo@google.com>
129 * gimple.h (struct gimple_statement_base)<uid>: Document
130 the restrictions on its use.
132 (gimple_set_uid): Tidy.
134 2009-01-09 Jakub Jelinek <jakub@redhat.com>
136 * config/i386/i386.c (ix86_expand_movmem, ix86_expand_setmem): Add
137 zero guard even if align_bytes != 0 and count is smaller than
140 2008-01-09 Vladimir Makarov <vmakarov@redhat.com>
142 PR rtl-optimization/38495
143 * ira-emit.c (print_move_list, ira_debug_move_list): New functions.
144 (add_range_and_copies_from_move_list): Print all added ranges.
145 Add ranges to memory optimized destination.
147 2009-01-09 Jakub Jelinek <jakub@redhat.com>
151 * config/i386/i386.c (override_options): Reject
152 -mstringop-strategy=rep_8byte with -m32.
153 (ix86_expand_movmem): For size_needed == 1 set epilogue_size_needed
154 to 1. Do count comparison against epilogue_size_needed at compile
155 time even when count_exp was constant forced into register. For
156 size_needed don't jump to epilogue, instead just avoid aligning
157 and invoke the body algorithm. If need_zero_guard, add zero guard
158 even if count is non-zero, but smaller than size_needed + number of
159 bytes that could be stored for alignment.
160 (ix86_expand_setmem): For size_needed == 1 set epilogue_size_needed
161 to 1. If need_zero_guard, add zero guard even if count is non-zero,
162 but smaller than size_needed + number of bytes that could be stored
163 for alignment. Compare size_needed with epilogue_size_needed instead
164 of desired_align - align, don't adjust size_needed, pass
165 epilogue_size_needed to the epilogue expanders.
168 * c-pretty-print.c (pp_c_expression): Handle GOTO_EXPR like BIND_EXPR.
170 2009-01-09 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
172 * pa.c (last_address): Change to unsigned.
173 (update_total_code_bytes): Change argument to unsigned. Don't
174 check if insn addresses are set.
175 (pa_output_function_epilogue): Set last_address to UINT_MAX if insn
176 addresses are not set.
177 (pa_asm_output_mi_thunk): Handle wrap when updating last_address.
179 2009-01-09 Nick Clifton <nickc@redhat.com>
181 * config/sh/symbian.c: Replace uses of DECL_INLINE with
182 DECL_DECLARED_INLINE_P.
184 2009-01-09 Jakub Jelinek <jakub@redhat.com>
187 * dojump.c (do_jump_by_parts_zero_rtx): Use mode instead of
188 GET_MODE (op0) in operand_subword_force calls.
191 * fold-const.c (fold_unary): For COMPOUND_EXPR and COND_EXPR,
192 fold_convert arg0 operands to TREE_TYPE (op0) first.
194 2009-01-08 Vladimir Makarov <vmakarov@redhat.com>
196 * params.def (ira-max-conflict-table-size): Decrease default value
199 2009-01-08 Jakub Jelinek <jakub@redhat.com>
201 PR tree-optimization/37031
202 * lambda-code.c (lambda_collect_parameters): Call pointer_set_destroy
204 (build_access_matrix): Reserve correct size for AM_MATRIX vector,
205 allocate it using gc instead of heap, use VEC_quick_push instead of
207 * graphite.c (build_access_matrix): Allocate AM_MATRIX vector using gc
208 instead of heap, use VEC_quick_push instead of VEC_safe_push.
209 * tree-data-ref.h (struct access_matrix): Change matrix to gc
210 allocated vector from heap allocated.
211 * lambda.h: Add DEF_VEC_ALLOC_P for gc allocated lambda_vector.
212 * tree-loop-linear.c (linear_transform_loops): Allocate nest
213 vector only after perfect_loop_nest_depth call.
215 2009-01-08 Sebastian Pop <sebastian.pop@amd.com>
216 Jan Sjodin <jan.sjodin@amd.com>
218 PR tree-optimization/38559
219 * graphite.c (debug_value, copy_constraint,
220 swap_constraint_variables, scale_constraint_variable, ): New.
221 (get_lower_bound, get_upper_bound): Removed.
222 (graphite_trans_bb_strip_mine): Clean up this code that works
223 only for constant number of iterations. Fully copy upper and
224 lower bound constraints, not only the constant part of them.
225 * graphite.h (debug_value): Declared.
227 2009-01-08 Ira Rosen <irar@il.ibm.com>
229 PR tree-optimization/37194
230 * tree-vect-transform.c (vect_estimate_min_profitable_iters):
231 Don't add the cost of cost model guard in prologue to scalar
232 outside cost in case of known number of iterations.
234 2009-01-07 Nathan Froyd <froydnj@codesourcery.com>
235 Alan Modra <amodra@bigpond.net.au>
237 * config/rs6000/rs6000.c (rs6000_legitimize_address): Check for
238 non-word-aligned REG+CONST addressing.
240 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
243 * config/alpha/alpha.c (alpha_end_function): For TARGET_ABI_OSF, call
244 free_after_compilation when outputting a thunk.
245 (alpha_output_mi_thunk_osf): Assert that we are processing a thunk.
246 Do not call free_after_compilation here.
248 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
250 * config/i386/i386.c (ix86_target_string): Use ARRAY_SIZE.
251 (ix86_valid_target_attribute_inner_p): Ditto.
253 2009-01-07 Jan Sjodin <jan.sjodin@amd.com>
255 PR tree-optimization/38492
256 PR tree-optimization/38498
257 * tree-check.c (operator_is_linear, scev_is_linear_expression): New.
258 * tree-chrec.h (scev_is_linear_expression): Declared.
259 * graphite.c (graphite_cannot_represent_loop_niter): New.
260 (scopdet_basic_block_info): Call graphite_cannot_represent_loop_niter.
261 (graphite_loop_normal_form): Use gcc_assert.
262 (scan_tree_for_params): Use CASE_CONVERT.
263 (phi_node_is_iv, bb_contains_non_iv_scalar_phi_nodes): New.
264 (build_scop_conditions_1): Call bb_contains_non_iv_scalar_phi_nodes.
265 Use gcc_assert. Discard scops that contain unhandled cases.
266 (build_scop_conditions): Return a boolean status for unhandled cases.
267 (strip_mine_profitable_p): Print the loop number, not its depth.
268 (is_interchange_valid): Pass the depth of the loop nest, don't
269 recompute it wrongly.
270 (graphite_trans_bb_block): Same.
271 (graphite_trans_bb_block): Print tentative of loop blocking.
272 (graphite_trans_scop_block): Do not print that the loop has been
274 (graphite_transform_loops): Do not handle scops that contain condition
277 2009-01-07 H.J. Lu <hongjiu.lu@intel.com>
279 AVX Programming Reference (December, 2008)
280 * config/i386/avxintrin.h (_mm256_stream_si256): New.
281 (_mm256_stream_pd): Likewise.
282 (_mm256_stream_ps): Likewise.
284 * config/i386/i386.c (ix86_builtins): Add IX86_BUILTIN_MOVNTDQ256,
285 IX86_BUILTIN_MOVNTPD256 and IX86_BUILTIN_MOVNTPS256.
286 (ix86_special_builtin_type): Add VOID_FTYPE_PV4DI_V4DI.
287 (bdesc_special_args): Add __builtin_ia32_movntdq256,
288 __builtin_ia32_movntpd256 and __builtin_ia32_movntps256.
289 (ix86_init_mmx_sse_builtins): Handle VOID_FTYPE_PV4DI_V4DI.
290 (ix86_expand_special_args_builtin): Likewise.
292 * config/i386/sse.md (AVXMODEDI): New.
293 (avx_movnt<mode>): Likewise.
294 (avx_movnt<mode>): Likewise.
295 (<sse>_movnt<mode>): Remove AVX support.
296 (sse2_movntv2di): Likewise.
298 2009-01-07 Richard Guenther <rguenther@suse.de>
301 * fold-const.c (extract_muldiv): Remove obsolete comment.
302 (fold_plusminus_mult_expr): Undo MINUS_EXPR
303 to PLUS_EXPR canonicalization for the canonicalization.
305 2009-01-07 Gerald Pfeifer <gerald@pfeifer.com>
307 * doc/install.texi (alpha*-dec-osf*): Remove note on 32-bit
308 hosted cross-compilers generating less efficient code.
310 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
312 * function.h (rtl_data): Add a dbr_scheduled_p field.
313 * reorg.c (dbr_schedule): Set it.
314 (gate_handle_delay_slots): Check it.
315 * config/mips/mips.c (mips_base_delayed_branch): Delete.
316 (mips_reorg): Check flag_delayed_branch instead of
317 mips_base_delayed_branch.
318 (mips_override_options): Don't set mips_base_delayed_branch
319 or flag_delayed_branch.
321 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
323 PR rtl-optimization/38426.
324 * ira.c (ira): Set current_function_is_leaf earlier.
326 2009-01-06 Jakub Jelinek <jakub@redhat.com>
328 PR rtl-optimization/38722
329 * combine.c (try_combine): Don't modify PATTERN (i3) and notes
330 too early, only set a flag and modify after last possible
333 2009-01-06 Janis Johnson <janis187@us.ibm.com>
336 * ginclude/float.h: Rename DECnn_DEN to DECnn_SUBNORMAL_MIN.
337 * real.c (decimal_single_format): Correct values of emin and emax.
338 (decimal_double_format): Ditto.
339 (decimal_quad_format): Ditto.
340 * c-cppbuiltin.c (builtin_define_decimal_float_constants): Adjust
341 computation of DECnn_MIN and DECnn_MAX for corrected values of
342 emin and emax. Define __DECnn_SUBNORMAL_MIN__ instead of
343 __DECnn_MIN__, and adjust its computation for the corrected value
346 2009-01-06 Jan Hubicka <jh@suse.cz>
349 * i386.c (ix86_expand_call): Use ARRAY_SIZE.
351 2009-01-06 Gerald Pfeifer <gerald@pfeifer.com>
353 * doc/contrib.texi (Contributors): Slightly adjust the end note.
354 Add Robert Clark to the list of testers.
356 2009-01-06 Jan Hubicka <jh@suse.cz>
357 Kai Tietz <kai.tietz@onevision.com>
359 * i386.md (*msabi_syvabi): Add SSE regs clobbers.
360 * i386.c (ix86_expand_call): Add clobbers.
362 2009-01-06 Jan Hubicka <jh@suse.cz>
363 Kai Tietz <kai.tietz@onevision.com>
365 * i386.h (CONDITIONAL_CALL_USAGE): SSE regs are not used for w64 ABI.
366 * i386.c (struct ix86_frame): Add padding0 and nsseregs.
367 (ix86_nsaved_regs): Count only general purpose regs.
368 (ix86_nsaved_sseregs): New.
369 (ix86_compute_frame_layout): Update nsseregs; set preferred alignment
370 to 16 for w64; compute padding and size of sse reg save area.
371 (ix86_emit_save_regs, ix86_emit_save_regs_using_mov): Save only
372 general purpose regs.
373 (ix86_emit_save_sse_regs_using_mov): New.
374 (ix86_expand_prologue): Save SSE regs if needed.
375 (ix86_emit_restore_regs_using_mov): Use only general purpose regs.
376 (ix86_emit_restore_sse_regs_using_mov): New.
377 (ix86_expand_epilogue): Save SSE regs if needed.
379 2009-01-06 Jan Hubicka <jh@suse.cz>
380 Kai Tietz <kai.tietz@onevision.com>
382 * i386.h (ACCUMULATE_OUTGOING_ARGS): Enable for MSABI
383 * i386.c (init_cumulative_args): Disallow calls of MSABI functions
384 when accumulate outgoing args is off.
386 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
389 * ira-color.c (ira_reuse_stack_slot): Check ENABLE_IRA_CHECKING
390 before using pseudos_have_intersected_live_ranges_p.
392 * ira-int.h (ira_assert): Always define.
394 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
396 AVX Programming Reference (December, 2008)
397 * config/i386/avxintrin.h (_mm_permute2_pd): Removed.
398 (_mm256_permute2_pd): Likewise.
399 (_mm_permute2_ps): Likewise.
400 (_mm256_permute2_ps): Likewise.
401 * config/i386/i386.md (UNSPEC_VPERMIL2): Likewise.
402 * config/i386/sse.md (avx_vpermil2<mode>3): Likewise.
404 * config/i386/i386.c (ix86_builtins): Remove
405 IX86_BUILTIN_VPERMIL2PD, IX86_BUILTIN_VPERMIL2PS,
406 IX86_BUILTIN_VPERMIL2PD256 and IX86_BUILTIN_VPERMIL2PS256.
407 (ix86_builtin_type): Remove V8SF_FTYPE_V8SF_V8SF_V8SI_INT,
408 V4DF_FTYPE_V4DF_V4DF_V4DI_INT, V4SF_FTYPE_V4SF_V4SF_V4SI_INT
409 and V2DF_FTYPE_V2DF_V2DF_V2DI_INT.
410 (bdesc_args): Remove __builtin_ia32_vpermil2pd,
411 __builtin_ia32_vpermil2ps, __builtin_ia32_vpermil2pd256 and
412 __builtin_ia32_vpermil2ps256.
413 (ix86_init_mmx_sse_builtins): Updated.
414 (ix86_expand_args_builtin): Likewise.
416 2009-01-05 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
418 * pa.c (output_call): Relocate non-jump insns in the delay slot of
419 long absolute calls when generating PA 2.0 code.
421 2009-01-05 Vladimir Makarov <vmakarov@redhat.com>
423 PR rtl-optimization/38583
424 * params.h (IRA_MAX_CONFLICT_TABLE_SIZE): New macro.
426 * params.def (ira-max-conflict-table-size): New.
428 * doc/invoke.texi (ira-max-conflict-table-size): Decribe.
430 * ira.h (ira_conflicts_p): New external definition.
432 * ira-conflicts.c (build_conflict_bit_table): Do not build too big
433 table. Report this. Return result of building.
434 (ira_build_conflicts): Use ira_conflicts_p. Check result of
435 building conflict table.
437 * ira-color.c (fast_allocation): Use num instead of ira_allocnos_num.
438 (ira_color): Use ira_conflicts_p.
440 * global.c: Include ira.h.
441 (pseudo_for_reload_consideration_p, build_insn_chain): Use
444 * Makefile.in (global.o): Add ira.h.
446 * ira-build.c (mark_all_loops_for_removal,
447 propagate_some_info_from_allocno): New.
448 (remove_unnecessary_allocnos): Call
449 propagate_some_info_from_allocno.
450 (remove_low_level_allocnos): New.
451 (remove_unnecessary_regions): Add parameter. Call
452 mark_all_loops_for_removal and remove_low_level_allocnos. Pass
453 parameter to remove_unnecessary_regions.
454 (ira_build): Remove all regions but root if the conflict table was
455 not built. Update conflict hard regs for allocnos crossing calls.
457 * ira.c (ira_conflicts_p): New global.
458 (ira): Define and use ira_conflicts_p.
460 * reload1.c (compute_use_by_pseudos, reload, count_pseudo,
461 count_spilled_pseudo, find_reg, alter_reg, finish_spills,
462 emit_input_reload_insns, delete_output_reload): Use ira_conflicts_p.
464 2009-01-06 Ben Elliston <bje@au.ibm.com>
466 * gengtype-lex.l (YY_NO_INPUT): Define.
468 2009-01-05 Andrew Pinski <andrew_pinski@playstation.sony.com>
471 * c-common.c (handle_vector_size_attribute): Also reject
474 2009-01-05 Sebastian Pop <sebastian.pop@amd.com>
476 PR tree-optimization/38492
477 * graphite.c (rename_map_elt, debug_rename_elt,
478 debug_rename_map_1, debug_rename_map, new_rename_map_elt,
479 rename_map_elt_info, eq_rename_map_elts,
480 get_new_name_from_old_name, bb_in_sese_p): Moved around.
481 (sese_find_uses_to_rename_use): Renamed sese_build_livein_liveouts_use.
482 (sese_find_uses_to_rename_bb): Renamed sese_build_livein_liveouts_bb.
483 (sese_build_livein_liveouts): New.
484 (new_sese, free_sese): New.
485 (new_scop): Call new_sese.
486 (free_scop): Call free_sese.
487 (rename_variables_from_edge, rename_phis_end_scop): Removed.
488 (register_old_new_names): Renamed register_old_and_new_names.
489 (register_scop_liveout_renames, add_loop_exit_phis,
490 insert_loop_close_phis, struct igp,
491 default_liveout_before_guard, add_guard_exit_phis,
492 insert_guard_phis, copy_renames): New.
493 (translate_clast): Call insert_loop_close_phis and insert_guard_phis.
494 (sese_add_exit_phis_edge): Renamed scop_add_exit_phis_edge.
495 (rewrite_into_sese_closed_ssa): Renamed scop_insert_phis_for_liveouts.
496 (scop_adjust_phis_for_liveouts): New.
497 (gloog): Call scop_adjust_phis_for_liveouts.
499 * graphite.h (struct sese): Documented. Added fields liveout,
501 (SESE_LIVEOUT, SESE_LIVEIN, SESE_LIVEIN_VER, SESE_NUM_VER): New.
502 (new_sese, free_sese, sese_build_livein_liveouts): Declared.
503 (struct scop): Added field liveout_renames.
504 (SCOP_LIVEOUT_RENAMES): New.
506 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
508 PR tree-optimization/38510
509 * graphite.c (recompute_all_dominators): Call mark_irreducible_loops.
510 (translate_clast): Call recompute_all_dominators before
512 (gloog): Call recompute_all_dominators before graphite_verify.
514 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
515 Jan Sjodin <jan.sjodin@amd.com>
517 PR tree-optimization/38500
518 * graphite.c (create_sese_edges): Call fix_loop_structure after
521 2009-01-05 Joel Sherrill <joel.sherrill@oarcorp.com>
523 * config.gcc: Add m32r*-*-rtems*.
524 * config/m32r/rtems.h: New file.
526 2009-01-05 Ben Elliston <bje@au.ibm.com>
528 * Makefile.in (.po.gmo): Use mkinstalldirs, not test -d || mkdir.
530 (po/gcc.pot): Likewise.
532 2009-01-04 David S. Miller <davem@davemloft.net>
534 * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
535 (STARTING_FRAME_OFFSET): Always set to zero.
537 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
539 * tree.def (LSHIFT_EXPR, RSHIFT_EXPR): Add commentary.
540 * tree-cfg.c (verify_gimple_assign_binary): Allow shifts of
541 fixed-point types, and vectors of the same.
543 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
545 * config/mips/sync.md (*mb_barrier): Rename to...
546 (*memory_barrier): ...this.
548 2009-01-04 Jonathan Wakely <jwakely.gcc@gmail.com>
550 * doc/extend.texi (Function Attributes): Move @cindex after @item
551 for 'artificial' and 'flatten'. Fix grammar for 'externally_visible'
552 and put in alphabetical order. Fix 'target' name and put in order.
553 * doc/invoke.texi (-Wstrict-null-sentinel, -fipa-matrix-reorg): Fix
556 2009-01-04 Uros Bizjak <ubizjak@gmail.com>
558 * config/s390/s390.md (UNSPEC_MB): Rename from UNSPECV_MB.
559 (memory_barrier): Expand as unspec instead of unspec_volatile.
560 Remove mem:BLK from insn operands. Use Pmode scratch register.
561 (*memory_barrier): Define as unspec instead of unspec_volatile.
562 Use (match_dup 0) as input operand.
564 * config/sparc/sparc.md (UNSPEC_MEMBAR): Rename from UNSPECV_MEMBAR.
565 * config/sparc/sync.md (memory_barrier): Expand as unspec instead of
566 unspec_volatile. Remove mem:BLK from insn operands. Use Pmode
567 scratch register. Remove operand 1.
568 (*stbar): Define as unspec instead of unspec_volatile.
569 Use (match_dup 0) as input operand, remove (const_int 8).
570 (*membar): Define as unspec instead of unspec_volatile.
571 Use (match_dup 0) as input operand, remove input operand 2.
573 * config/xtensa/xtensa.md (UNSPEC_MEMW): Rename from UNSPECV_MEMW.
574 (memory_barrier): Expand as unspec instead of unspec_volatile.
575 Remove mem:BLK from insn operands. Use Pmode scratch register.
576 (*memory_barrier): Define as unspec instead of unspec_volatile.
577 Use (match_dup 0) as input operand.
579 * config/ia64/sync.md (memory_barrier): Redefine as expander pattern.
580 Remove mem:BLK from insn operands. Use Pmode scratch register.
581 Set volatile flag on operand 0.
582 (*memory_barrier): New insn pattern.
584 * config/rs6000/sync.md (memory_barrier): Remove mem:BLK from
586 (*memory_barrier): Use (match_dup 0) as input operand.
588 * config/mips/sync.md (memory_barrier): Redefine as expander pattern.
589 Remove mem:BLK from insn operands. Use Pmode scratch register.
590 Set volatile flag on operand 0.
591 (*mb_internal): New insn pattern.
593 * config/alpha/sync.md (*memory_barrier): Rename from *mb_internal.
595 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
598 * function.c (struct temp_slot): Move to the section of the file
599 that deals with temp slots. Remove field 'address'.
600 (temp_slot_address_table): New hash table of address -> temp slot.
601 (struct temp_slot_address_entry): New struct, items for the table.
602 (temp_slot_address_compute_hash, temp_slot_address_hash,
603 temp_slot_address_eq, insert_temp_slot_address): Support functions
605 (find_temp_slot_from_address): Rewrite to use the new hash table.
606 (remove_unused_temp_slot_addresses): Remove addresses of temp
607 slots that have been made available.
608 (remove_unused_temp_slot_addresses_1): Call-back for htab_traverse,
609 worker function for remove_unused_temp_slot_addresses.
610 (assign_stack_temp_for_type): Don't clear the temp slot address list.
611 Add the temp slot address to the address -> temp slot map.
612 (update_temp_slot_address): Update via insert_temp_slot_address.
613 (free_temp_slots): Call remove_unused_temp_slot_addresses.
614 (pop_temp_slots): Likewise.
615 (init_temp_slots): Allocate the address -> temp slot map, or empty
616 the map if it is already allocated.
617 (prepare_function_start): Initialize temp slot processing.
619 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
622 * cfgexpand.c (estimate_stack_frame_size): Simplify the estimate:
623 Calculate the size of all stack vars assuming no packing of stack
624 vars will happen, replacing a quadratic algorithm with a linear one.
626 2009-01-03 Jakub Jelinek <jakub@redhat.com>
629 * expmed.c (store_bit_field_1): Don't modify op0 if movstrict insn
632 2009-01-03 Diego Novillo <dnovillo@google.com>
634 * doc/contrib.texi: Update contributions.
636 2009-01-03 Jakub Jelinek <jakub@redhat.com>
639 * builtins.c (fold_builtin_memory_op): Give up if either operand
640 is volatile. Set srctype or desttype to non-qualified version
644 * builtins.c (fold_builtin_expect): Only check DECL_WEAK for VAR_DECLs
647 2009-01-02 Kenneth Zadeck <zadeck@naturalbridge.com>
649 PR rtl-optimization/35805
650 * df-problems.c (df_lr_finalize): Add recursive call to resolve lr
651 problem if fast dce is able to remove any instructions.
652 * dce.c (dce_process_block): Fix dump message.
654 2009-01-02 Mark Mitchell <mark@codesourcery.com>
657 * tree-ssa-pre.c (compute_antic): Correct loop bounds.
659 2009-01-02 Jakub Jelinek <jakub@redhat.com>
662 * tree-flow.h (op_code_prio, op_prio): New prototypes.
663 * tree-pretty-print.c (op_code_prio): New function.
664 (op_prio): No longer static. Use op_code_prio.
665 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs):
666 Use op_prio and op_code_prio to determine if () should be
667 printed around operand(s) or not.
669 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs,
670 dump_gimple_call, dump_gimple_switch, dump_gimple_cond,
671 dump_gimple_label, dump_gimple_try, dump_symbols, dump_gimple_phi,
672 dump_gimple_mem_ops, dump_bb_header, dump_bb_end, pp_cfg_jump): Use
673 pp_character instead of pp_string for single letter printing.
675 2009-01-02 Richard Sandiford <rdsandiford@googlemail.com>
677 * doc/extend.texi: Fix '#pragma GCC option' typo.
679 2009-01-02 Richard Guenther <rguenther@suse.de>
681 * doc/install.texi (--enable-checking): Mention different
683 (--enable-stage1-checking): Document.
685 2009-01-01 Andrew Pinski <pinskia@gmail.com>
688 * tree-cfg.c (verify_expr): Add INDIRECT_REF case. Change MODIFY_EXPR
691 2009-01-02 Ben Elliston <bje@au.ibm.com>
693 * config/fp-bit.h (pack_d): Constify argument.
694 * config/fp-bit.c (makenan): Constify return type. Remove casts.
695 (isnan): Constify argument.
699 (_fpadd_parts): Constify return type.
700 (_fpmul_parts): Likewise.
701 (_fpdiv_parts): Likewise.
703 2009-01-01 Jakub Jelinek <jakub@redhat.com>
706 * c-typeck.c (add_pending_init): Add IMPLICIT argument. Only
707 warn about overwriting initializer with side-effects or
708 -Woverride-init if !IMPLICIT.
709 (output_init_element): Likewise. Pass IMPLICIT down to
711 (process_init_element): Add IMPLICIT argument. Pass it down
712 to output_init_element.
713 (push_init_element, pop_init_level, set_designator): Adjust
714 process_init_element callers.
715 (set_nonincremental_init, set_nonincremental_init_from_string):
716 Adjust add_pending_init callers.
717 (output_pending_init_elements): Adjust output_init_element callers.
718 * c-tree.h (process_init_element): Adjust prototype.
719 * c-parser.c (c_parser_initelt, c_parser_initval): Adjust
720 process_init_element callers.