1 2009-01-14 Richard Guenther <rguenther@suse.de>
3 PR tree-optimization/38826
5 * tree-ssa-structalias.c (emit_alias_warning): Emit the pointer
6 initialization notes only if we actually emitted a warning.
7 (intra_create_variable_infos): Add constraints for a result decl
8 that is passed by hidden reference.
9 (build_pred_graph): Mark all related variables non-direct on
12 2009-01-14 Nick Clifton <nickc@redhat.com>
14 * ira-conflicts.c: Include addresses.h for the definition of
16 (ira_build_conflicts): Use base_reg_class instead of
18 * Makefile.in: Add a dependency of ira-conflicts.o on
21 2009-01-13 Vladimir Makarov <vmakarov@redhat.com>
24 * Makefile.in (ira-lives.o): Add except.h.
26 * ira-lives.c: Include except.h.
27 (process_bb_node_lives): Process can_throw_internal.
29 2009-01-13 Jakub Jelinek <jakub@redhat.com>
31 PR rtl-optimization/38774
32 * combine.c (simplify_set): When undoing cc_use change, don't do
33 PUT_CODE on the newly created comparison, but instead put back the
36 2009-01-13 Joseph Myers <joseph@codesourcery.com>
38 * doc/invoke.texi (ARM Options): Update lists of -mcpu and -march
39 values. Remove duplicate arm8 entry.
41 2009-01-13 Sebastian Pop <sebastian.pop@amd.com>
43 PR tree-optimization/38786
44 * graphite.c (expand_scalar_variables_ssa_name): New, outlined from
45 the SSA_NAME case of expand_scalar_variables_expr.
46 Set the type of an expression to the type of its assign statement.
47 (expand_scalar_variables_expr): Also gather the scalar computation
48 used to index the memory access. Do not pass loop_p.
49 Fix comment. Stop recursion on tcc_constant or tcc_declaration.
50 (expand_scalar_variables_stmt): Pass to expand_scalar_variables_expr
51 the gimple_stmt_iterator where it inserts new code. Do not pass loop_p.
52 (copy_bb_and_scalar_dependences): Do not pass loop_p.
53 (translate_clast): Update call to copy_bb_and_scalar_dependences.
55 2009-01-13 Sebastian Pop <sebastian.pop@amd.com>
57 * graphite.h (debug_value): Removed.
58 * graphite.c (debug_value): Removed.
60 2009-01-13 Richard Earnshaw <rearnsha@arm.com>
62 * arm.c (output_move_double): Don't synthesize thumb-2 ldrd/strd with
63 two 32-bit instructions.
65 2009-01-13 Richard Earnshaw <rearnsha@arm.com>
67 * arm.c (struct processors): Pass for speed down into cost helper
69 (const_ok_for_op): Handle COMPARE and inequality nodes.
70 (arm_rtx_costs_1): Rewrite.
71 (arm_size_rtx_costs): Update prototype.
72 (arm_rtx_costs): Pass speed down to helper functions.
73 (arm_slowmul_rtx_costs): Rework cost calculations.
74 (arm_fastmul_rtx_costs, arm_xscale_rtx_costs): Likewise.
75 (arm_9e_rtx_costs): Likewise.
77 2009-01-13 Uros Bizjak <ubizjak@gmail.com>
79 * config/alpha/alpha.c (alpha_legitimate_address_p): Explicit
80 relocations of local symbols wider than UNITS_PER_WORD are not valid.
81 (alpha_legitimize_address): Do not split local symbols wider than
82 UNITS_PER_WORD into HIGH/LO_SUM parts.
84 2009-01-13 Danny Smith <dannysmith@users.sourceforge.net>
87 * gcc.c (process_command): Replace call to execvp with calls
90 2009-01-03 Anatoly Sokolov <aesok@post.ru>
93 * config/avr/t-avr (LIB1ASMFUNCS): Add _tablejump_elpm.
94 * config/avr/libgcc.S (__do_global_ctors, __do_global_dtors): Add
95 variant for devices with 3-byte PC.
96 (__tablejump_elpm__) : New.
98 2009-01-12 Jakub Jelinek <jakub@redhat.com>
101 * c-parser.c (c_parser_postfix_expression): Allow `->' in
102 offsetof member-designator, handle it as `[0].'.
104 2009-01-12 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
106 * pa.c (pa_asm_output_mi_thunk): Use pc-relative branch to thunk
107 function when not using named sections on targets with named sections
108 if branch distance is less than 262132.
110 2009-01-12 Richard Earnshaw <rearnsha@arm.com>
112 * combine.c (combine_instructions): Recompute
113 optimize_this_for_speed_p for each BB in the main combine loop.
115 2009-01-12 Tomas Bily <tbily@suse.cz>
118 * tree-loop-distribution.c (prop_phis): New function.
119 (generate_builtin): Call prop_phis.
120 * testsuite/gcc.dg/tree-ssa/pr38385.c: New file.
122 2009-01-12 Jakub Jelinek <jakub@redhat.com>
124 PR tree-optimization/38807
125 * tree-ssa-reassoc.c (remove_visited_stmt_chain): Don't look at
126 gimple_visited_p unless stmt is GIMPLE_ASSIGN.
128 2009-01-11 Adam Nemet <anemet@caviumnetworks.com>
130 * expmed.c (store_bit_field_1): Properly truncate the paradoxical
131 subreg of op0 to the original op0.
133 2009-01-11 Laurent GUERBY <laurent@guerby.net>
135 * doc/sourcebuild.texi (Source Tree): Move up intl and fixinc.
137 2009-01-11 Markus Schoepflin <markus.schoepflin@comsoft.de>
140 * gcc/mips-tfile.c (parse_def): Fix parsing of def strings
141 starting with digits.
143 2009-01-10 Jakub Jelinek <jakub@redhat.com>
146 * config/arm/arm.c (arm_is_long_call_p): Don't call
147 arm_function_in_section_p if decl isn't a FUNCTION_DECL.
149 2009-01-09 Steven Bosscher <steven@gcc.gnu.org>
151 * regrename.c (regrename_optimize): Fix dumping.
152 (find_oldest_value_reg): Preserve REG_POINTER.
153 (copy_hardreg_forward_1): Likewise.
155 2009-01-09 Diego Novillo <dnovillo@google.com>
157 * gimple.h (struct gimple_statement_base)<uid>: Document
158 the restrictions on its use.
160 (gimple_set_uid): Tidy.
162 2009-01-09 Jakub Jelinek <jakub@redhat.com>
164 * config/i386/i386.c (ix86_expand_movmem, ix86_expand_setmem): Add
165 zero guard even if align_bytes != 0 and count is smaller than
168 2008-01-09 Vladimir Makarov <vmakarov@redhat.com>
170 PR rtl-optimization/38495
171 * ira-emit.c (print_move_list, ira_debug_move_list): New functions.
172 (add_range_and_copies_from_move_list): Print all added ranges.
173 Add ranges to memory optimized destination.
175 2009-01-09 Jakub Jelinek <jakub@redhat.com>
179 * config/i386/i386.c (override_options): Reject
180 -mstringop-strategy=rep_8byte with -m32.
181 (ix86_expand_movmem): For size_needed == 1 set epilogue_size_needed
182 to 1. Do count comparison against epilogue_size_needed at compile
183 time even when count_exp was constant forced into register. For
184 size_needed don't jump to epilogue, instead just avoid aligning
185 and invoke the body algorithm. If need_zero_guard, add zero guard
186 even if count is non-zero, but smaller than size_needed + number of
187 bytes that could be stored for alignment.
188 (ix86_expand_setmem): For size_needed == 1 set epilogue_size_needed
189 to 1. If need_zero_guard, add zero guard even if count is non-zero,
190 but smaller than size_needed + number of bytes that could be stored
191 for alignment. Compare size_needed with epilogue_size_needed instead
192 of desired_align - align, don't adjust size_needed, pass
193 epilogue_size_needed to the epilogue expanders.
196 * c-pretty-print.c (pp_c_expression): Handle GOTO_EXPR like BIND_EXPR.
198 2009-01-09 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
200 * pa.c (last_address): Change to unsigned.
201 (update_total_code_bytes): Change argument to unsigned. Don't
202 check if insn addresses are set.
203 (pa_output_function_epilogue): Set last_address to UINT_MAX if insn
204 addresses are not set.
205 (pa_asm_output_mi_thunk): Handle wrap when updating last_address.
207 2009-01-09 Nick Clifton <nickc@redhat.com>
209 * config/sh/symbian.c: Replace uses of DECL_INLINE with
210 DECL_DECLARED_INLINE_P.
212 2009-01-09 Jakub Jelinek <jakub@redhat.com>
215 * dojump.c (do_jump_by_parts_zero_rtx): Use mode instead of
216 GET_MODE (op0) in operand_subword_force calls.
219 * fold-const.c (fold_unary): For COMPOUND_EXPR and COND_EXPR,
220 fold_convert arg0 operands to TREE_TYPE (op0) first.
222 2009-01-08 Vladimir Makarov <vmakarov@redhat.com>
224 * params.def (ira-max-conflict-table-size): Decrease default value
227 2009-01-08 Jakub Jelinek <jakub@redhat.com>
229 PR tree-optimization/37031
230 * lambda-code.c (lambda_collect_parameters): Call pointer_set_destroy
232 (build_access_matrix): Reserve correct size for AM_MATRIX vector,
233 allocate it using gc instead of heap, use VEC_quick_push instead of
235 * graphite.c (build_access_matrix): Allocate AM_MATRIX vector using gc
236 instead of heap, use VEC_quick_push instead of VEC_safe_push.
237 * tree-data-ref.h (struct access_matrix): Change matrix to gc
238 allocated vector from heap allocated.
239 * lambda.h: Add DEF_VEC_ALLOC_P for gc allocated lambda_vector.
240 * tree-loop-linear.c (linear_transform_loops): Allocate nest
241 vector only after perfect_loop_nest_depth call.
243 2009-01-08 Sebastian Pop <sebastian.pop@amd.com>
244 Jan Sjodin <jan.sjodin@amd.com>
246 PR tree-optimization/38559
247 * graphite.c (debug_value, copy_constraint,
248 swap_constraint_variables, scale_constraint_variable, ): New.
249 (get_lower_bound, get_upper_bound): Removed.
250 (graphite_trans_bb_strip_mine): Clean up this code that works
251 only for constant number of iterations. Fully copy upper and
252 lower bound constraints, not only the constant part of them.
253 * graphite.h (debug_value): Declared.
255 2009-01-08 Ira Rosen <irar@il.ibm.com>
257 PR tree-optimization/37194
258 * tree-vect-transform.c (vect_estimate_min_profitable_iters):
259 Don't add the cost of cost model guard in prologue to scalar
260 outside cost in case of known number of iterations.
262 2009-01-07 Nathan Froyd <froydnj@codesourcery.com>
263 Alan Modra <amodra@bigpond.net.au>
265 * config/rs6000/rs6000.c (rs6000_legitimize_address): Check for
266 non-word-aligned REG+CONST addressing.
268 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
271 * config/alpha/alpha.c (alpha_end_function): For TARGET_ABI_OSF, call
272 free_after_compilation when outputting a thunk.
273 (alpha_output_mi_thunk_osf): Assert that we are processing a thunk.
274 Do not call free_after_compilation here.
276 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
278 * config/i386/i386.c (ix86_target_string): Use ARRAY_SIZE.
279 (ix86_valid_target_attribute_inner_p): Ditto.
281 2009-01-07 Jan Sjodin <jan.sjodin@amd.com>
283 PR tree-optimization/38492
284 PR tree-optimization/38498
285 * tree-check.c (operator_is_linear, scev_is_linear_expression): New.
286 * tree-chrec.h (scev_is_linear_expression): Declared.
287 * graphite.c (graphite_cannot_represent_loop_niter): New.
288 (scopdet_basic_block_info): Call graphite_cannot_represent_loop_niter.
289 (graphite_loop_normal_form): Use gcc_assert.
290 (scan_tree_for_params): Use CASE_CONVERT.
291 (phi_node_is_iv, bb_contains_non_iv_scalar_phi_nodes): New.
292 (build_scop_conditions_1): Call bb_contains_non_iv_scalar_phi_nodes.
293 Use gcc_assert. Discard scops that contain unhandled cases.
294 (build_scop_conditions): Return a boolean status for unhandled cases.
295 (strip_mine_profitable_p): Print the loop number, not its depth.
296 (is_interchange_valid): Pass the depth of the loop nest, don't
297 recompute it wrongly.
298 (graphite_trans_bb_block): Same.
299 (graphite_trans_bb_block): Print tentative of loop blocking.
300 (graphite_trans_scop_block): Do not print that the loop has been
302 (graphite_transform_loops): Do not handle scops that contain condition
305 2009-01-07 H.J. Lu <hongjiu.lu@intel.com>
307 AVX Programming Reference (December, 2008)
308 * config/i386/avxintrin.h (_mm256_stream_si256): New.
309 (_mm256_stream_pd): Likewise.
310 (_mm256_stream_ps): Likewise.
312 * config/i386/i386.c (ix86_builtins): Add IX86_BUILTIN_MOVNTDQ256,
313 IX86_BUILTIN_MOVNTPD256 and IX86_BUILTIN_MOVNTPS256.
314 (ix86_special_builtin_type): Add VOID_FTYPE_PV4DI_V4DI.
315 (bdesc_special_args): Add __builtin_ia32_movntdq256,
316 __builtin_ia32_movntpd256 and __builtin_ia32_movntps256.
317 (ix86_init_mmx_sse_builtins): Handle VOID_FTYPE_PV4DI_V4DI.
318 (ix86_expand_special_args_builtin): Likewise.
320 * config/i386/sse.md (AVXMODEDI): New.
321 (avx_movnt<mode>): Likewise.
322 (avx_movnt<mode>): Likewise.
323 (<sse>_movnt<mode>): Remove AVX support.
324 (sse2_movntv2di): Likewise.
326 2009-01-07 Richard Guenther <rguenther@suse.de>
329 * fold-const.c (extract_muldiv): Remove obsolete comment.
330 (fold_plusminus_mult_expr): Undo MINUS_EXPR
331 to PLUS_EXPR canonicalization for the canonicalization.
333 2009-01-07 Gerald Pfeifer <gerald@pfeifer.com>
335 * doc/install.texi (alpha*-dec-osf*): Remove note on 32-bit
336 hosted cross-compilers generating less efficient code.
338 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
340 * function.h (rtl_data): Add a dbr_scheduled_p field.
341 * reorg.c (dbr_schedule): Set it.
342 (gate_handle_delay_slots): Check it.
343 * config/mips/mips.c (mips_base_delayed_branch): Delete.
344 (mips_reorg): Check flag_delayed_branch instead of
345 mips_base_delayed_branch.
346 (mips_override_options): Don't set mips_base_delayed_branch
347 or flag_delayed_branch.
349 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
351 PR rtl-optimization/38426.
352 * ira.c (ira): Set current_function_is_leaf earlier.
354 2009-01-06 Jakub Jelinek <jakub@redhat.com>
356 PR rtl-optimization/38722
357 * combine.c (try_combine): Don't modify PATTERN (i3) and notes
358 too early, only set a flag and modify after last possible
361 2009-01-06 Janis Johnson <janis187@us.ibm.com>
364 * ginclude/float.h: Rename DECnn_DEN to DECnn_SUBNORMAL_MIN.
365 * real.c (decimal_single_format): Correct values of emin and emax.
366 (decimal_double_format): Ditto.
367 (decimal_quad_format): Ditto.
368 * c-cppbuiltin.c (builtin_define_decimal_float_constants): Adjust
369 computation of DECnn_MIN and DECnn_MAX for corrected values of
370 emin and emax. Define __DECnn_SUBNORMAL_MIN__ instead of
371 __DECnn_MIN__, and adjust its computation for the corrected value
374 2009-01-06 Jan Hubicka <jh@suse.cz>
377 * i386.c (ix86_expand_call): Use ARRAY_SIZE.
379 2009-01-06 Gerald Pfeifer <gerald@pfeifer.com>
381 * doc/contrib.texi (Contributors): Slightly adjust the end note.
382 Add Robert Clark to the list of testers.
384 2009-01-06 Jan Hubicka <jh@suse.cz>
385 Kai Tietz <kai.tietz@onevision.com>
387 * i386.md (*msabi_syvabi): Add SSE regs clobbers.
388 * i386.c (ix86_expand_call): Add clobbers.
390 2009-01-06 Jan Hubicka <jh@suse.cz>
391 Kai Tietz <kai.tietz@onevision.com>
393 * i386.h (CONDITIONAL_CALL_USAGE): SSE regs are not used for w64 ABI.
394 * i386.c (struct ix86_frame): Add padding0 and nsseregs.
395 (ix86_nsaved_regs): Count only general purpose regs.
396 (ix86_nsaved_sseregs): New.
397 (ix86_compute_frame_layout): Update nsseregs; set preferred alignment
398 to 16 for w64; compute padding and size of sse reg save area.
399 (ix86_emit_save_regs, ix86_emit_save_regs_using_mov): Save only
400 general purpose regs.
401 (ix86_emit_save_sse_regs_using_mov): New.
402 (ix86_expand_prologue): Save SSE regs if needed.
403 (ix86_emit_restore_regs_using_mov): Use only general purpose regs.
404 (ix86_emit_restore_sse_regs_using_mov): New.
405 (ix86_expand_epilogue): Save SSE regs if needed.
407 2009-01-06 Jan Hubicka <jh@suse.cz>
408 Kai Tietz <kai.tietz@onevision.com>
410 * i386.h (ACCUMULATE_OUTGOING_ARGS): Enable for MSABI
411 * i386.c (init_cumulative_args): Disallow calls of MSABI functions
412 when accumulate outgoing args is off.
414 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
417 * ira-color.c (ira_reuse_stack_slot): Check ENABLE_IRA_CHECKING
418 before using pseudos_have_intersected_live_ranges_p.
420 * ira-int.h (ira_assert): Always define.
422 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
424 AVX Programming Reference (December, 2008)
425 * config/i386/avxintrin.h (_mm_permute2_pd): Removed.
426 (_mm256_permute2_pd): Likewise.
427 (_mm_permute2_ps): Likewise.
428 (_mm256_permute2_ps): Likewise.
429 * config/i386/i386.md (UNSPEC_VPERMIL2): Likewise.
430 * config/i386/sse.md (avx_vpermil2<mode>3): Likewise.
432 * config/i386/i386.c (ix86_builtins): Remove
433 IX86_BUILTIN_VPERMIL2PD, IX86_BUILTIN_VPERMIL2PS,
434 IX86_BUILTIN_VPERMIL2PD256 and IX86_BUILTIN_VPERMIL2PS256.
435 (ix86_builtin_type): Remove V8SF_FTYPE_V8SF_V8SF_V8SI_INT,
436 V4DF_FTYPE_V4DF_V4DF_V4DI_INT, V4SF_FTYPE_V4SF_V4SF_V4SI_INT
437 and V2DF_FTYPE_V2DF_V2DF_V2DI_INT.
438 (bdesc_args): Remove __builtin_ia32_vpermil2pd,
439 __builtin_ia32_vpermil2ps, __builtin_ia32_vpermil2pd256 and
440 __builtin_ia32_vpermil2ps256.
441 (ix86_init_mmx_sse_builtins): Updated.
442 (ix86_expand_args_builtin): Likewise.
444 2009-01-05 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
446 * pa.c (output_call): Relocate non-jump insns in the delay slot of
447 long absolute calls when generating PA 2.0 code.
449 2009-01-05 Vladimir Makarov <vmakarov@redhat.com>
451 PR rtl-optimization/38583
452 * params.h (IRA_MAX_CONFLICT_TABLE_SIZE): New macro.
454 * params.def (ira-max-conflict-table-size): New.
456 * doc/invoke.texi (ira-max-conflict-table-size): Decribe.
458 * ira.h (ira_conflicts_p): New external definition.
460 * ira-conflicts.c (build_conflict_bit_table): Do not build too big
461 table. Report this. Return result of building.
462 (ira_build_conflicts): Use ira_conflicts_p. Check result of
463 building conflict table.
465 * ira-color.c (fast_allocation): Use num instead of ira_allocnos_num.
466 (ira_color): Use ira_conflicts_p.
468 * global.c: Include ira.h.
469 (pseudo_for_reload_consideration_p, build_insn_chain): Use
472 * Makefile.in (global.o): Add ira.h.
474 * ira-build.c (mark_all_loops_for_removal,
475 propagate_some_info_from_allocno): New.
476 (remove_unnecessary_allocnos): Call
477 propagate_some_info_from_allocno.
478 (remove_low_level_allocnos): New.
479 (remove_unnecessary_regions): Add parameter. Call
480 mark_all_loops_for_removal and remove_low_level_allocnos. Pass
481 parameter to remove_unnecessary_regions.
482 (ira_build): Remove all regions but root if the conflict table was
483 not built. Update conflict hard regs for allocnos crossing calls.
485 * ira.c (ira_conflicts_p): New global.
486 (ira): Define and use ira_conflicts_p.
488 * reload1.c (compute_use_by_pseudos, reload, count_pseudo,
489 count_spilled_pseudo, find_reg, alter_reg, finish_spills,
490 emit_input_reload_insns, delete_output_reload): Use ira_conflicts_p.
492 2009-01-06 Ben Elliston <bje@au.ibm.com>
494 * gengtype-lex.l (YY_NO_INPUT): Define.
496 2009-01-05 Andrew Pinski <andrew_pinski@playstation.sony.com>
499 * c-common.c (handle_vector_size_attribute): Also reject
502 2009-01-05 Sebastian Pop <sebastian.pop@amd.com>
504 PR tree-optimization/38492
505 * graphite.c (rename_map_elt, debug_rename_elt,
506 debug_rename_map_1, debug_rename_map, new_rename_map_elt,
507 rename_map_elt_info, eq_rename_map_elts,
508 get_new_name_from_old_name, bb_in_sese_p): Moved around.
509 (sese_find_uses_to_rename_use): Renamed sese_build_livein_liveouts_use.
510 (sese_find_uses_to_rename_bb): Renamed sese_build_livein_liveouts_bb.
511 (sese_build_livein_liveouts): New.
512 (new_sese, free_sese): New.
513 (new_scop): Call new_sese.
514 (free_scop): Call free_sese.
515 (rename_variables_from_edge, rename_phis_end_scop): Removed.
516 (register_old_new_names): Renamed register_old_and_new_names.
517 (register_scop_liveout_renames, add_loop_exit_phis,
518 insert_loop_close_phis, struct igp,
519 default_liveout_before_guard, add_guard_exit_phis,
520 insert_guard_phis, copy_renames): New.
521 (translate_clast): Call insert_loop_close_phis and insert_guard_phis.
522 (sese_add_exit_phis_edge): Renamed scop_add_exit_phis_edge.
523 (rewrite_into_sese_closed_ssa): Renamed scop_insert_phis_for_liveouts.
524 (scop_adjust_phis_for_liveouts): New.
525 (gloog): Call scop_adjust_phis_for_liveouts.
527 * graphite.h (struct sese): Documented. Added fields liveout,
529 (SESE_LIVEOUT, SESE_LIVEIN, SESE_LIVEIN_VER, SESE_NUM_VER): New.
530 (new_sese, free_sese, sese_build_livein_liveouts): Declared.
531 (struct scop): Added field liveout_renames.
532 (SCOP_LIVEOUT_RENAMES): New.
534 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
536 PR tree-optimization/38510
537 * graphite.c (recompute_all_dominators): Call mark_irreducible_loops.
538 (translate_clast): Call recompute_all_dominators before
540 (gloog): Call recompute_all_dominators before graphite_verify.
542 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
543 Jan Sjodin <jan.sjodin@amd.com>
545 PR tree-optimization/38500
546 * graphite.c (create_sese_edges): Call fix_loop_structure after
549 2009-01-05 Joel Sherrill <joel.sherrill@oarcorp.com>
551 * config.gcc: Add m32r*-*-rtems*.
552 * config/m32r/rtems.h: New file.
554 2009-01-05 Ben Elliston <bje@au.ibm.com>
556 * Makefile.in (.po.gmo): Use mkinstalldirs, not test -d || mkdir.
558 (po/gcc.pot): Likewise.
560 2009-01-04 David S. Miller <davem@davemloft.net>
562 * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
563 (STARTING_FRAME_OFFSET): Always set to zero.
565 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
567 * tree.def (LSHIFT_EXPR, RSHIFT_EXPR): Add commentary.
568 * tree-cfg.c (verify_gimple_assign_binary): Allow shifts of
569 fixed-point types, and vectors of the same.
571 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
573 * config/mips/sync.md (*mb_barrier): Rename to...
574 (*memory_barrier): ...this.
576 2009-01-04 Jonathan Wakely <jwakely.gcc@gmail.com>
578 * doc/extend.texi (Function Attributes): Move @cindex after @item
579 for 'artificial' and 'flatten'. Fix grammar for 'externally_visible'
580 and put in alphabetical order. Fix 'target' name and put in order.
581 * doc/invoke.texi (-Wstrict-null-sentinel, -fipa-matrix-reorg): Fix
584 2009-01-04 Uros Bizjak <ubizjak@gmail.com>
586 * config/s390/s390.md (UNSPEC_MB): Rename from UNSPECV_MB.
587 (memory_barrier): Expand as unspec instead of unspec_volatile.
588 Remove mem:BLK from insn operands. Use Pmode scratch register.
589 (*memory_barrier): Define as unspec instead of unspec_volatile.
590 Use (match_dup 0) as input operand.
592 * config/sparc/sparc.md (UNSPEC_MEMBAR): Rename from UNSPECV_MEMBAR.
593 * config/sparc/sync.md (memory_barrier): Expand as unspec instead of
594 unspec_volatile. Remove mem:BLK from insn operands. Use Pmode
595 scratch register. Remove operand 1.
596 (*stbar): Define as unspec instead of unspec_volatile.
597 Use (match_dup 0) as input operand, remove (const_int 8).
598 (*membar): Define as unspec instead of unspec_volatile.
599 Use (match_dup 0) as input operand, remove input operand 2.
601 * config/xtensa/xtensa.md (UNSPEC_MEMW): Rename from UNSPECV_MEMW.
602 (memory_barrier): Expand as unspec instead of unspec_volatile.
603 Remove mem:BLK from insn operands. Use Pmode scratch register.
604 (*memory_barrier): Define as unspec instead of unspec_volatile.
605 Use (match_dup 0) as input operand.
607 * config/ia64/sync.md (memory_barrier): Redefine as expander pattern.
608 Remove mem:BLK from insn operands. Use Pmode scratch register.
609 Set volatile flag on operand 0.
610 (*memory_barrier): New insn pattern.
612 * config/rs6000/sync.md (memory_barrier): Remove mem:BLK from
614 (*memory_barrier): Use (match_dup 0) as input operand.
616 * config/mips/sync.md (memory_barrier): Redefine as expander pattern.
617 Remove mem:BLK from insn operands. Use Pmode scratch register.
618 Set volatile flag on operand 0.
619 (*mb_internal): New insn pattern.
621 * config/alpha/sync.md (*memory_barrier): Rename from *mb_internal.
623 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
626 * function.c (struct temp_slot): Move to the section of the file
627 that deals with temp slots. Remove field 'address'.
628 (temp_slot_address_table): New hash table of address -> temp slot.
629 (struct temp_slot_address_entry): New struct, items for the table.
630 (temp_slot_address_compute_hash, temp_slot_address_hash,
631 temp_slot_address_eq, insert_temp_slot_address): Support functions
633 (find_temp_slot_from_address): Rewrite to use the new hash table.
634 (remove_unused_temp_slot_addresses): Remove addresses of temp
635 slots that have been made available.
636 (remove_unused_temp_slot_addresses_1): Call-back for htab_traverse,
637 worker function for remove_unused_temp_slot_addresses.
638 (assign_stack_temp_for_type): Don't clear the temp slot address list.
639 Add the temp slot address to the address -> temp slot map.
640 (update_temp_slot_address): Update via insert_temp_slot_address.
641 (free_temp_slots): Call remove_unused_temp_slot_addresses.
642 (pop_temp_slots): Likewise.
643 (init_temp_slots): Allocate the address -> temp slot map, or empty
644 the map if it is already allocated.
645 (prepare_function_start): Initialize temp slot processing.
647 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
650 * cfgexpand.c (estimate_stack_frame_size): Simplify the estimate:
651 Calculate the size of all stack vars assuming no packing of stack
652 vars will happen, replacing a quadratic algorithm with a linear one.
654 2009-01-03 Jakub Jelinek <jakub@redhat.com>
657 * expmed.c (store_bit_field_1): Don't modify op0 if movstrict insn
660 2009-01-03 Diego Novillo <dnovillo@google.com>
662 * doc/contrib.texi: Update contributions.
664 2009-01-03 Jakub Jelinek <jakub@redhat.com>
667 * builtins.c (fold_builtin_memory_op): Give up if either operand
668 is volatile. Set srctype or desttype to non-qualified version
672 * builtins.c (fold_builtin_expect): Only check DECL_WEAK for VAR_DECLs
675 2009-01-02 Kenneth Zadeck <zadeck@naturalbridge.com>
677 PR rtl-optimization/35805
678 * df-problems.c (df_lr_finalize): Add recursive call to resolve lr
679 problem if fast dce is able to remove any instructions.
680 * dce.c (dce_process_block): Fix dump message.
682 2009-01-02 Mark Mitchell <mark@codesourcery.com>
685 * tree-ssa-pre.c (compute_antic): Correct loop bounds.
687 2009-01-02 Jakub Jelinek <jakub@redhat.com>
690 * tree-flow.h (op_code_prio, op_prio): New prototypes.
691 * tree-pretty-print.c (op_code_prio): New function.
692 (op_prio): No longer static. Use op_code_prio.
693 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs):
694 Use op_prio and op_code_prio to determine if () should be
695 printed around operand(s) or not.
697 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs,
698 dump_gimple_call, dump_gimple_switch, dump_gimple_cond,
699 dump_gimple_label, dump_gimple_try, dump_symbols, dump_gimple_phi,
700 dump_gimple_mem_ops, dump_bb_header, dump_bb_end, pp_cfg_jump): Use
701 pp_character instead of pp_string for single letter printing.
703 2009-01-02 Richard Sandiford <rdsandiford@googlemail.com>
705 * doc/extend.texi: Fix '#pragma GCC option' typo.
707 2009-01-02 Richard Guenther <rguenther@suse.de>
709 * doc/install.texi (--enable-checking): Mention different
711 (--enable-stage1-checking): Document.
713 2009-01-01 Andrew Pinski <pinskia@gmail.com>
716 * tree-cfg.c (verify_expr): Add INDIRECT_REF case. Change MODIFY_EXPR
719 2009-01-02 Ben Elliston <bje@au.ibm.com>
721 * config/fp-bit.h (pack_d): Constify argument.
722 * config/fp-bit.c (makenan): Constify return type. Remove casts.
723 (isnan): Constify argument.
727 (_fpadd_parts): Constify return type.
728 (_fpmul_parts): Likewise.
729 (_fpdiv_parts): Likewise.
731 2009-01-01 Jakub Jelinek <jakub@redhat.com>
734 * c-typeck.c (add_pending_init): Add IMPLICIT argument. Only
735 warn about overwriting initializer with side-effects or
736 -Woverride-init if !IMPLICIT.
737 (output_init_element): Likewise. Pass IMPLICIT down to
739 (process_init_element): Add IMPLICIT argument. Pass it down
740 to output_init_element.
741 (push_init_element, pop_init_level, set_designator): Adjust
742 process_init_element callers.
743 (set_nonincremental_init, set_nonincremental_init_from_string):
744 Adjust add_pending_init callers.
745 (output_pending_init_elements): Adjust output_init_element callers.
746 * c-tree.h (process_init_element): Adjust prototype.
747 * c-parser.c (c_parser_initelt, c_parser_initval): Adjust
748 process_init_element callers.