1 2009-01-14 Vladimir Makarov <vmakarov@redhat.com>
3 * testsuite/g++.dg/torture/pr38811.C: New file.
5 2009-01-14 Basile Starynkevitch <basile@starynkevitch.net>
6 * doc/gty.texi (Invoking the garbage collector): Added new node
7 and section documenting ggc_collect.
9 2009-01-14 Richard Guenther <rguenther@suse.de>
11 PR tree-optimization/38826
13 * tree-ssa-structalias.c (emit_alias_warning): Emit the pointer
14 initialization notes only if we actually emitted a warning.
15 (intra_create_variable_infos): Add constraints for a result decl
16 that is passed by hidden reference.
17 (build_pred_graph): Mark all related variables non-direct on
20 2009-01-14 Nick Clifton <nickc@redhat.com>
22 * ira-conflicts.c: Include addresses.h for the definition of
24 (ira_build_conflicts): Use base_reg_class instead of
26 * Makefile.in: Add a dependency of ira-conflicts.o on
29 2009-01-13 Vladimir Makarov <vmakarov@redhat.com>
32 * Makefile.in (ira-lives.o): Add except.h.
34 * ira-lives.c: Include except.h.
35 (process_bb_node_lives): Process can_throw_internal.
37 2009-01-13 Jakub Jelinek <jakub@redhat.com>
39 PR rtl-optimization/38774
40 * combine.c (simplify_set): When undoing cc_use change, don't do
41 PUT_CODE on the newly created comparison, but instead put back the
44 2009-01-13 Joseph Myers <joseph@codesourcery.com>
46 * doc/invoke.texi (ARM Options): Update lists of -mcpu and -march
47 values. Remove duplicate arm8 entry.
49 2009-01-13 Sebastian Pop <sebastian.pop@amd.com>
51 PR tree-optimization/38786
52 * graphite.c (expand_scalar_variables_ssa_name): New, outlined from
53 the SSA_NAME case of expand_scalar_variables_expr.
54 Set the type of an expression to the type of its assign statement.
55 (expand_scalar_variables_expr): Also gather the scalar computation
56 used to index the memory access. Do not pass loop_p.
57 Fix comment. Stop recursion on tcc_constant or tcc_declaration.
58 (expand_scalar_variables_stmt): Pass to expand_scalar_variables_expr
59 the gimple_stmt_iterator where it inserts new code. Do not pass loop_p.
60 (copy_bb_and_scalar_dependences): Do not pass loop_p.
61 (translate_clast): Update call to copy_bb_and_scalar_dependences.
63 2009-01-13 Sebastian Pop <sebastian.pop@amd.com>
65 * graphite.h (debug_value): Removed.
66 * graphite.c (debug_value): Removed.
68 2009-01-13 Richard Earnshaw <rearnsha@arm.com>
70 * arm.c (output_move_double): Don't synthesize thumb-2 ldrd/strd with
71 two 32-bit instructions.
73 2009-01-13 Richard Earnshaw <rearnsha@arm.com>
75 * arm.c (struct processors): Pass for speed down into cost helper
77 (const_ok_for_op): Handle COMPARE and inequality nodes.
78 (arm_rtx_costs_1): Rewrite.
79 (arm_size_rtx_costs): Update prototype.
80 (arm_rtx_costs): Pass speed down to helper functions.
81 (arm_slowmul_rtx_costs): Rework cost calculations.
82 (arm_fastmul_rtx_costs, arm_xscale_rtx_costs): Likewise.
83 (arm_9e_rtx_costs): Likewise.
85 2009-01-13 Uros Bizjak <ubizjak@gmail.com>
87 * config/alpha/alpha.c (alpha_legitimate_address_p): Explicit
88 relocations of local symbols wider than UNITS_PER_WORD are not valid.
89 (alpha_legitimize_address): Do not split local symbols wider than
90 UNITS_PER_WORD into HIGH/LO_SUM parts.
92 2009-01-13 Danny Smith <dannysmith@users.sourceforge.net>
95 * gcc.c (process_command): Replace call to execvp with calls
98 2009-01-03 Anatoly Sokolov <aesok@post.ru>
101 * config/avr/t-avr (LIB1ASMFUNCS): Add _tablejump_elpm.
102 * config/avr/libgcc.S (__do_global_ctors, __do_global_dtors): Add
103 variant for devices with 3-byte PC.
104 (__tablejump_elpm__) : New.
106 2009-01-12 Jakub Jelinek <jakub@redhat.com>
109 * c-parser.c (c_parser_postfix_expression): Allow `->' in
110 offsetof member-designator, handle it as `[0].'.
112 2009-01-12 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
114 * pa.c (pa_asm_output_mi_thunk): Use pc-relative branch to thunk
115 function when not using named sections on targets with named sections
116 if branch distance is less than 262132.
118 2009-01-12 Richard Earnshaw <rearnsha@arm.com>
120 * combine.c (combine_instructions): Recompute
121 optimize_this_for_speed_p for each BB in the main combine loop.
123 2009-01-12 Tomas Bily <tbily@suse.cz>
126 * tree-loop-distribution.c (prop_phis): New function.
127 (generate_builtin): Call prop_phis.
128 * testsuite/gcc.dg/tree-ssa/pr38385.c: New file.
130 2009-01-12 Jakub Jelinek <jakub@redhat.com>
132 PR tree-optimization/38807
133 * tree-ssa-reassoc.c (remove_visited_stmt_chain): Don't look at
134 gimple_visited_p unless stmt is GIMPLE_ASSIGN.
136 2009-01-11 Adam Nemet <anemet@caviumnetworks.com>
138 * expmed.c (store_bit_field_1): Properly truncate the paradoxical
139 subreg of op0 to the original op0.
141 2009-01-11 Laurent GUERBY <laurent@guerby.net>
143 * doc/sourcebuild.texi (Source Tree): Move up intl and fixinc.
145 2009-01-11 Markus Schoepflin <markus.schoepflin@comsoft.de>
148 * gcc/mips-tfile.c (parse_def): Fix parsing of def strings
149 starting with digits.
151 2009-01-10 Jakub Jelinek <jakub@redhat.com>
154 * config/arm/arm.c (arm_is_long_call_p): Don't call
155 arm_function_in_section_p if decl isn't a FUNCTION_DECL.
157 2009-01-09 Steven Bosscher <steven@gcc.gnu.org>
159 * regrename.c (regrename_optimize): Fix dumping.
160 (find_oldest_value_reg): Preserve REG_POINTER.
161 (copy_hardreg_forward_1): Likewise.
163 2009-01-09 Diego Novillo <dnovillo@google.com>
165 * gimple.h (struct gimple_statement_base)<uid>: Document
166 the restrictions on its use.
168 (gimple_set_uid): Tidy.
170 2009-01-09 Jakub Jelinek <jakub@redhat.com>
172 * config/i386/i386.c (ix86_expand_movmem, ix86_expand_setmem): Add
173 zero guard even if align_bytes != 0 and count is smaller than
176 2008-01-09 Vladimir Makarov <vmakarov@redhat.com>
178 PR rtl-optimization/38495
179 * ira-emit.c (print_move_list, ira_debug_move_list): New functions.
180 (add_range_and_copies_from_move_list): Print all added ranges.
181 Add ranges to memory optimized destination.
183 2009-01-09 Jakub Jelinek <jakub@redhat.com>
187 * config/i386/i386.c (override_options): Reject
188 -mstringop-strategy=rep_8byte with -m32.
189 (ix86_expand_movmem): For size_needed == 1 set epilogue_size_needed
190 to 1. Do count comparison against epilogue_size_needed at compile
191 time even when count_exp was constant forced into register. For
192 size_needed don't jump to epilogue, instead just avoid aligning
193 and invoke the body algorithm. If need_zero_guard, add zero guard
194 even if count is non-zero, but smaller than size_needed + number of
195 bytes that could be stored for alignment.
196 (ix86_expand_setmem): For size_needed == 1 set epilogue_size_needed
197 to 1. If need_zero_guard, add zero guard even if count is non-zero,
198 but smaller than size_needed + number of bytes that could be stored
199 for alignment. Compare size_needed with epilogue_size_needed instead
200 of desired_align - align, don't adjust size_needed, pass
201 epilogue_size_needed to the epilogue expanders.
204 * c-pretty-print.c (pp_c_expression): Handle GOTO_EXPR like BIND_EXPR.
206 2009-01-09 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
208 * pa.c (last_address): Change to unsigned.
209 (update_total_code_bytes): Change argument to unsigned. Don't
210 check if insn addresses are set.
211 (pa_output_function_epilogue): Set last_address to UINT_MAX if insn
212 addresses are not set.
213 (pa_asm_output_mi_thunk): Handle wrap when updating last_address.
215 2009-01-09 Nick Clifton <nickc@redhat.com>
217 * config/sh/symbian.c: Replace uses of DECL_INLINE with
218 DECL_DECLARED_INLINE_P.
220 2009-01-09 Jakub Jelinek <jakub@redhat.com>
223 * dojump.c (do_jump_by_parts_zero_rtx): Use mode instead of
224 GET_MODE (op0) in operand_subword_force calls.
227 * fold-const.c (fold_unary): For COMPOUND_EXPR and COND_EXPR,
228 fold_convert arg0 operands to TREE_TYPE (op0) first.
230 2009-01-08 Vladimir Makarov <vmakarov@redhat.com>
232 * params.def (ira-max-conflict-table-size): Decrease default value
235 2009-01-08 Jakub Jelinek <jakub@redhat.com>
237 PR tree-optimization/37031
238 * lambda-code.c (lambda_collect_parameters): Call pointer_set_destroy
240 (build_access_matrix): Reserve correct size for AM_MATRIX vector,
241 allocate it using gc instead of heap, use VEC_quick_push instead of
243 * graphite.c (build_access_matrix): Allocate AM_MATRIX vector using gc
244 instead of heap, use VEC_quick_push instead of VEC_safe_push.
245 * tree-data-ref.h (struct access_matrix): Change matrix to gc
246 allocated vector from heap allocated.
247 * lambda.h: Add DEF_VEC_ALLOC_P for gc allocated lambda_vector.
248 * tree-loop-linear.c (linear_transform_loops): Allocate nest
249 vector only after perfect_loop_nest_depth call.
251 2009-01-08 Sebastian Pop <sebastian.pop@amd.com>
252 Jan Sjodin <jan.sjodin@amd.com>
254 PR tree-optimization/38559
255 * graphite.c (debug_value, copy_constraint,
256 swap_constraint_variables, scale_constraint_variable, ): New.
257 (get_lower_bound, get_upper_bound): Removed.
258 (graphite_trans_bb_strip_mine): Clean up this code that works
259 only for constant number of iterations. Fully copy upper and
260 lower bound constraints, not only the constant part of them.
261 * graphite.h (debug_value): Declared.
263 2009-01-08 Ira Rosen <irar@il.ibm.com>
265 PR tree-optimization/37194
266 * tree-vect-transform.c (vect_estimate_min_profitable_iters):
267 Don't add the cost of cost model guard in prologue to scalar
268 outside cost in case of known number of iterations.
270 2009-01-07 Nathan Froyd <froydnj@codesourcery.com>
271 Alan Modra <amodra@bigpond.net.au>
273 * config/rs6000/rs6000.c (rs6000_legitimize_address): Check for
274 non-word-aligned REG+CONST addressing.
276 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
279 * config/alpha/alpha.c (alpha_end_function): For TARGET_ABI_OSF, call
280 free_after_compilation when outputting a thunk.
281 (alpha_output_mi_thunk_osf): Assert that we are processing a thunk.
282 Do not call free_after_compilation here.
284 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
286 * config/i386/i386.c (ix86_target_string): Use ARRAY_SIZE.
287 (ix86_valid_target_attribute_inner_p): Ditto.
289 2009-01-07 Jan Sjodin <jan.sjodin@amd.com>
291 PR tree-optimization/38492
292 PR tree-optimization/38498
293 * tree-check.c (operator_is_linear, scev_is_linear_expression): New.
294 * tree-chrec.h (scev_is_linear_expression): Declared.
295 * graphite.c (graphite_cannot_represent_loop_niter): New.
296 (scopdet_basic_block_info): Call graphite_cannot_represent_loop_niter.
297 (graphite_loop_normal_form): Use gcc_assert.
298 (scan_tree_for_params): Use CASE_CONVERT.
299 (phi_node_is_iv, bb_contains_non_iv_scalar_phi_nodes): New.
300 (build_scop_conditions_1): Call bb_contains_non_iv_scalar_phi_nodes.
301 Use gcc_assert. Discard scops that contain unhandled cases.
302 (build_scop_conditions): Return a boolean status for unhandled cases.
303 (strip_mine_profitable_p): Print the loop number, not its depth.
304 (is_interchange_valid): Pass the depth of the loop nest, don't
305 recompute it wrongly.
306 (graphite_trans_bb_block): Same.
307 (graphite_trans_bb_block): Print tentative of loop blocking.
308 (graphite_trans_scop_block): Do not print that the loop has been
310 (graphite_transform_loops): Do not handle scops that contain condition
313 2009-01-07 H.J. Lu <hongjiu.lu@intel.com>
315 AVX Programming Reference (December, 2008)
316 * config/i386/avxintrin.h (_mm256_stream_si256): New.
317 (_mm256_stream_pd): Likewise.
318 (_mm256_stream_ps): Likewise.
320 * config/i386/i386.c (ix86_builtins): Add IX86_BUILTIN_MOVNTDQ256,
321 IX86_BUILTIN_MOVNTPD256 and IX86_BUILTIN_MOVNTPS256.
322 (ix86_special_builtin_type): Add VOID_FTYPE_PV4DI_V4DI.
323 (bdesc_special_args): Add __builtin_ia32_movntdq256,
324 __builtin_ia32_movntpd256 and __builtin_ia32_movntps256.
325 (ix86_init_mmx_sse_builtins): Handle VOID_FTYPE_PV4DI_V4DI.
326 (ix86_expand_special_args_builtin): Likewise.
328 * config/i386/sse.md (AVXMODEDI): New.
329 (avx_movnt<mode>): Likewise.
330 (avx_movnt<mode>): Likewise.
331 (<sse>_movnt<mode>): Remove AVX support.
332 (sse2_movntv2di): Likewise.
334 2009-01-07 Richard Guenther <rguenther@suse.de>
337 * fold-const.c (extract_muldiv): Remove obsolete comment.
338 (fold_plusminus_mult_expr): Undo MINUS_EXPR
339 to PLUS_EXPR canonicalization for the canonicalization.
341 2009-01-07 Gerald Pfeifer <gerald@pfeifer.com>
343 * doc/install.texi (alpha*-dec-osf*): Remove note on 32-bit
344 hosted cross-compilers generating less efficient code.
346 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
348 * function.h (rtl_data): Add a dbr_scheduled_p field.
349 * reorg.c (dbr_schedule): Set it.
350 (gate_handle_delay_slots): Check it.
351 * config/mips/mips.c (mips_base_delayed_branch): Delete.
352 (mips_reorg): Check flag_delayed_branch instead of
353 mips_base_delayed_branch.
354 (mips_override_options): Don't set mips_base_delayed_branch
355 or flag_delayed_branch.
357 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
359 PR rtl-optimization/38426.
360 * ira.c (ira): Set current_function_is_leaf earlier.
362 2009-01-06 Jakub Jelinek <jakub@redhat.com>
364 PR rtl-optimization/38722
365 * combine.c (try_combine): Don't modify PATTERN (i3) and notes
366 too early, only set a flag and modify after last possible
369 2009-01-06 Janis Johnson <janis187@us.ibm.com>
372 * ginclude/float.h: Rename DECnn_DEN to DECnn_SUBNORMAL_MIN.
373 * real.c (decimal_single_format): Correct values of emin and emax.
374 (decimal_double_format): Ditto.
375 (decimal_quad_format): Ditto.
376 * c-cppbuiltin.c (builtin_define_decimal_float_constants): Adjust
377 computation of DECnn_MIN and DECnn_MAX for corrected values of
378 emin and emax. Define __DECnn_SUBNORMAL_MIN__ instead of
379 __DECnn_MIN__, and adjust its computation for the corrected value
382 2009-01-06 Jan Hubicka <jh@suse.cz>
385 * i386.c (ix86_expand_call): Use ARRAY_SIZE.
387 2009-01-06 Gerald Pfeifer <gerald@pfeifer.com>
389 * doc/contrib.texi (Contributors): Slightly adjust the end note.
390 Add Robert Clark to the list of testers.
392 2009-01-06 Jan Hubicka <jh@suse.cz>
393 Kai Tietz <kai.tietz@onevision.com>
395 * i386.md (*msabi_syvabi): Add SSE regs clobbers.
396 * i386.c (ix86_expand_call): Add clobbers.
398 2009-01-06 Jan Hubicka <jh@suse.cz>
399 Kai Tietz <kai.tietz@onevision.com>
401 * i386.h (CONDITIONAL_CALL_USAGE): SSE regs are not used for w64 ABI.
402 * i386.c (struct ix86_frame): Add padding0 and nsseregs.
403 (ix86_nsaved_regs): Count only general purpose regs.
404 (ix86_nsaved_sseregs): New.
405 (ix86_compute_frame_layout): Update nsseregs; set preferred alignment
406 to 16 for w64; compute padding and size of sse reg save area.
407 (ix86_emit_save_regs, ix86_emit_save_regs_using_mov): Save only
408 general purpose regs.
409 (ix86_emit_save_sse_regs_using_mov): New.
410 (ix86_expand_prologue): Save SSE regs if needed.
411 (ix86_emit_restore_regs_using_mov): Use only general purpose regs.
412 (ix86_emit_restore_sse_regs_using_mov): New.
413 (ix86_expand_epilogue): Save SSE regs if needed.
415 2009-01-06 Jan Hubicka <jh@suse.cz>
416 Kai Tietz <kai.tietz@onevision.com>
418 * i386.h (ACCUMULATE_OUTGOING_ARGS): Enable for MSABI
419 * i386.c (init_cumulative_args): Disallow calls of MSABI functions
420 when accumulate outgoing args is off.
422 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
425 * ira-color.c (ira_reuse_stack_slot): Check ENABLE_IRA_CHECKING
426 before using pseudos_have_intersected_live_ranges_p.
428 * ira-int.h (ira_assert): Always define.
430 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
432 AVX Programming Reference (December, 2008)
433 * config/i386/avxintrin.h (_mm_permute2_pd): Removed.
434 (_mm256_permute2_pd): Likewise.
435 (_mm_permute2_ps): Likewise.
436 (_mm256_permute2_ps): Likewise.
437 * config/i386/i386.md (UNSPEC_VPERMIL2): Likewise.
438 * config/i386/sse.md (avx_vpermil2<mode>3): Likewise.
440 * config/i386/i386.c (ix86_builtins): Remove
441 IX86_BUILTIN_VPERMIL2PD, IX86_BUILTIN_VPERMIL2PS,
442 IX86_BUILTIN_VPERMIL2PD256 and IX86_BUILTIN_VPERMIL2PS256.
443 (ix86_builtin_type): Remove V8SF_FTYPE_V8SF_V8SF_V8SI_INT,
444 V4DF_FTYPE_V4DF_V4DF_V4DI_INT, V4SF_FTYPE_V4SF_V4SF_V4SI_INT
445 and V2DF_FTYPE_V2DF_V2DF_V2DI_INT.
446 (bdesc_args): Remove __builtin_ia32_vpermil2pd,
447 __builtin_ia32_vpermil2ps, __builtin_ia32_vpermil2pd256 and
448 __builtin_ia32_vpermil2ps256.
449 (ix86_init_mmx_sse_builtins): Updated.
450 (ix86_expand_args_builtin): Likewise.
452 2009-01-05 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
454 * pa.c (output_call): Relocate non-jump insns in the delay slot of
455 long absolute calls when generating PA 2.0 code.
457 2009-01-05 Vladimir Makarov <vmakarov@redhat.com>
459 PR rtl-optimization/38583
460 * params.h (IRA_MAX_CONFLICT_TABLE_SIZE): New macro.
462 * params.def (ira-max-conflict-table-size): New.
464 * doc/invoke.texi (ira-max-conflict-table-size): Decribe.
466 * ira.h (ira_conflicts_p): New external definition.
468 * ira-conflicts.c (build_conflict_bit_table): Do not build too big
469 table. Report this. Return result of building.
470 (ira_build_conflicts): Use ira_conflicts_p. Check result of
471 building conflict table.
473 * ira-color.c (fast_allocation): Use num instead of ira_allocnos_num.
474 (ira_color): Use ira_conflicts_p.
476 * global.c: Include ira.h.
477 (pseudo_for_reload_consideration_p, build_insn_chain): Use
480 * Makefile.in (global.o): Add ira.h.
482 * ira-build.c (mark_all_loops_for_removal,
483 propagate_some_info_from_allocno): New.
484 (remove_unnecessary_allocnos): Call
485 propagate_some_info_from_allocno.
486 (remove_low_level_allocnos): New.
487 (remove_unnecessary_regions): Add parameter. Call
488 mark_all_loops_for_removal and remove_low_level_allocnos. Pass
489 parameter to remove_unnecessary_regions.
490 (ira_build): Remove all regions but root if the conflict table was
491 not built. Update conflict hard regs for allocnos crossing calls.
493 * ira.c (ira_conflicts_p): New global.
494 (ira): Define and use ira_conflicts_p.
496 * reload1.c (compute_use_by_pseudos, reload, count_pseudo,
497 count_spilled_pseudo, find_reg, alter_reg, finish_spills,
498 emit_input_reload_insns, delete_output_reload): Use ira_conflicts_p.
500 2009-01-06 Ben Elliston <bje@au.ibm.com>
502 * gengtype-lex.l (YY_NO_INPUT): Define.
504 2009-01-05 Andrew Pinski <andrew_pinski@playstation.sony.com>
507 * c-common.c (handle_vector_size_attribute): Also reject
510 2009-01-05 Sebastian Pop <sebastian.pop@amd.com>
512 PR tree-optimization/38492
513 * graphite.c (rename_map_elt, debug_rename_elt,
514 debug_rename_map_1, debug_rename_map, new_rename_map_elt,
515 rename_map_elt_info, eq_rename_map_elts,
516 get_new_name_from_old_name, bb_in_sese_p): Moved around.
517 (sese_find_uses_to_rename_use): Renamed sese_build_livein_liveouts_use.
518 (sese_find_uses_to_rename_bb): Renamed sese_build_livein_liveouts_bb.
519 (sese_build_livein_liveouts): New.
520 (new_sese, free_sese): New.
521 (new_scop): Call new_sese.
522 (free_scop): Call free_sese.
523 (rename_variables_from_edge, rename_phis_end_scop): Removed.
524 (register_old_new_names): Renamed register_old_and_new_names.
525 (register_scop_liveout_renames, add_loop_exit_phis,
526 insert_loop_close_phis, struct igp,
527 default_liveout_before_guard, add_guard_exit_phis,
528 insert_guard_phis, copy_renames): New.
529 (translate_clast): Call insert_loop_close_phis and insert_guard_phis.
530 (sese_add_exit_phis_edge): Renamed scop_add_exit_phis_edge.
531 (rewrite_into_sese_closed_ssa): Renamed scop_insert_phis_for_liveouts.
532 (scop_adjust_phis_for_liveouts): New.
533 (gloog): Call scop_adjust_phis_for_liveouts.
535 * graphite.h (struct sese): Documented. Added fields liveout,
537 (SESE_LIVEOUT, SESE_LIVEIN, SESE_LIVEIN_VER, SESE_NUM_VER): New.
538 (new_sese, free_sese, sese_build_livein_liveouts): Declared.
539 (struct scop): Added field liveout_renames.
540 (SCOP_LIVEOUT_RENAMES): New.
542 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
544 PR tree-optimization/38510
545 * graphite.c (recompute_all_dominators): Call mark_irreducible_loops.
546 (translate_clast): Call recompute_all_dominators before
548 (gloog): Call recompute_all_dominators before graphite_verify.
550 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
551 Jan Sjodin <jan.sjodin@amd.com>
553 PR tree-optimization/38500
554 * graphite.c (create_sese_edges): Call fix_loop_structure after
557 2009-01-05 Joel Sherrill <joel.sherrill@oarcorp.com>
559 * config.gcc: Add m32r*-*-rtems*.
560 * config/m32r/rtems.h: New file.
562 2009-01-05 Ben Elliston <bje@au.ibm.com>
564 * Makefile.in (.po.gmo): Use mkinstalldirs, not test -d || mkdir.
566 (po/gcc.pot): Likewise.
568 2009-01-04 David S. Miller <davem@davemloft.net>
570 * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
571 (STARTING_FRAME_OFFSET): Always set to zero.
573 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
575 * tree.def (LSHIFT_EXPR, RSHIFT_EXPR): Add commentary.
576 * tree-cfg.c (verify_gimple_assign_binary): Allow shifts of
577 fixed-point types, and vectors of the same.
579 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
581 * config/mips/sync.md (*mb_barrier): Rename to...
582 (*memory_barrier): ...this.
584 2009-01-04 Jonathan Wakely <jwakely.gcc@gmail.com>
586 * doc/extend.texi (Function Attributes): Move @cindex after @item
587 for 'artificial' and 'flatten'. Fix grammar for 'externally_visible'
588 and put in alphabetical order. Fix 'target' name and put in order.
589 * doc/invoke.texi (-Wstrict-null-sentinel, -fipa-matrix-reorg): Fix
592 2009-01-04 Uros Bizjak <ubizjak@gmail.com>
594 * config/s390/s390.md (UNSPEC_MB): Rename from UNSPECV_MB.
595 (memory_barrier): Expand as unspec instead of unspec_volatile.
596 Remove mem:BLK from insn operands. Use Pmode scratch register.
597 (*memory_barrier): Define as unspec instead of unspec_volatile.
598 Use (match_dup 0) as input operand.
600 * config/sparc/sparc.md (UNSPEC_MEMBAR): Rename from UNSPECV_MEMBAR.
601 * config/sparc/sync.md (memory_barrier): Expand as unspec instead of
602 unspec_volatile. Remove mem:BLK from insn operands. Use Pmode
603 scratch register. Remove operand 1.
604 (*stbar): Define as unspec instead of unspec_volatile.
605 Use (match_dup 0) as input operand, remove (const_int 8).
606 (*membar): Define as unspec instead of unspec_volatile.
607 Use (match_dup 0) as input operand, remove input operand 2.
609 * config/xtensa/xtensa.md (UNSPEC_MEMW): Rename from UNSPECV_MEMW.
610 (memory_barrier): Expand as unspec instead of unspec_volatile.
611 Remove mem:BLK from insn operands. Use Pmode scratch register.
612 (*memory_barrier): Define as unspec instead of unspec_volatile.
613 Use (match_dup 0) as input operand.
615 * config/ia64/sync.md (memory_barrier): Redefine as expander pattern.
616 Remove mem:BLK from insn operands. Use Pmode scratch register.
617 Set volatile flag on operand 0.
618 (*memory_barrier): New insn pattern.
620 * config/rs6000/sync.md (memory_barrier): Remove mem:BLK from
622 (*memory_barrier): Use (match_dup 0) as input operand.
624 * config/mips/sync.md (memory_barrier): Redefine as expander pattern.
625 Remove mem:BLK from insn operands. Use Pmode scratch register.
626 Set volatile flag on operand 0.
627 (*mb_internal): New insn pattern.
629 * config/alpha/sync.md (*memory_barrier): Rename from *mb_internal.
631 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
634 * function.c (struct temp_slot): Move to the section of the file
635 that deals with temp slots. Remove field 'address'.
636 (temp_slot_address_table): New hash table of address -> temp slot.
637 (struct temp_slot_address_entry): New struct, items for the table.
638 (temp_slot_address_compute_hash, temp_slot_address_hash,
639 temp_slot_address_eq, insert_temp_slot_address): Support functions
641 (find_temp_slot_from_address): Rewrite to use the new hash table.
642 (remove_unused_temp_slot_addresses): Remove addresses of temp
643 slots that have been made available.
644 (remove_unused_temp_slot_addresses_1): Call-back for htab_traverse,
645 worker function for remove_unused_temp_slot_addresses.
646 (assign_stack_temp_for_type): Don't clear the temp slot address list.
647 Add the temp slot address to the address -> temp slot map.
648 (update_temp_slot_address): Update via insert_temp_slot_address.
649 (free_temp_slots): Call remove_unused_temp_slot_addresses.
650 (pop_temp_slots): Likewise.
651 (init_temp_slots): Allocate the address -> temp slot map, or empty
652 the map if it is already allocated.
653 (prepare_function_start): Initialize temp slot processing.
655 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
658 * cfgexpand.c (estimate_stack_frame_size): Simplify the estimate:
659 Calculate the size of all stack vars assuming no packing of stack
660 vars will happen, replacing a quadratic algorithm with a linear one.
662 2009-01-03 Jakub Jelinek <jakub@redhat.com>
665 * expmed.c (store_bit_field_1): Don't modify op0 if movstrict insn
668 2009-01-03 Diego Novillo <dnovillo@google.com>
670 * doc/contrib.texi: Update contributions.
672 2009-01-03 Jakub Jelinek <jakub@redhat.com>
675 * builtins.c (fold_builtin_memory_op): Give up if either operand
676 is volatile. Set srctype or desttype to non-qualified version
680 * builtins.c (fold_builtin_expect): Only check DECL_WEAK for VAR_DECLs
683 2009-01-02 Kenneth Zadeck <zadeck@naturalbridge.com>
685 PR rtl-optimization/35805
686 * df-problems.c (df_lr_finalize): Add recursive call to resolve lr
687 problem if fast dce is able to remove any instructions.
688 * dce.c (dce_process_block): Fix dump message.
690 2009-01-02 Mark Mitchell <mark@codesourcery.com>
693 * tree-ssa-pre.c (compute_antic): Correct loop bounds.
695 2009-01-02 Jakub Jelinek <jakub@redhat.com>
698 * tree-flow.h (op_code_prio, op_prio): New prototypes.
699 * tree-pretty-print.c (op_code_prio): New function.
700 (op_prio): No longer static. Use op_code_prio.
701 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs):
702 Use op_prio and op_code_prio to determine if () should be
703 printed around operand(s) or not.
705 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs,
706 dump_gimple_call, dump_gimple_switch, dump_gimple_cond,
707 dump_gimple_label, dump_gimple_try, dump_symbols, dump_gimple_phi,
708 dump_gimple_mem_ops, dump_bb_header, dump_bb_end, pp_cfg_jump): Use
709 pp_character instead of pp_string for single letter printing.
711 2009-01-02 Richard Sandiford <rdsandiford@googlemail.com>
713 * doc/extend.texi: Fix '#pragma GCC option' typo.
715 2009-01-02 Richard Guenther <rguenther@suse.de>
717 * doc/install.texi (--enable-checking): Mention different
719 (--enable-stage1-checking): Document.
721 2009-01-01 Andrew Pinski <pinskia@gmail.com>
724 * tree-cfg.c (verify_expr): Add INDIRECT_REF case. Change MODIFY_EXPR
727 2009-01-02 Ben Elliston <bje@au.ibm.com>
729 * config/fp-bit.h (pack_d): Constify argument.
730 * config/fp-bit.c (makenan): Constify return type. Remove casts.
731 (isnan): Constify argument.
735 (_fpadd_parts): Constify return type.
736 (_fpmul_parts): Likewise.
737 (_fpdiv_parts): Likewise.
739 2009-01-01 Jakub Jelinek <jakub@redhat.com>
742 * c-typeck.c (add_pending_init): Add IMPLICIT argument. Only
743 warn about overwriting initializer with side-effects or
744 -Woverride-init if !IMPLICIT.
745 (output_init_element): Likewise. Pass IMPLICIT down to
747 (process_init_element): Add IMPLICIT argument. Pass it down
748 to output_init_element.
749 (push_init_element, pop_init_level, set_designator): Adjust
750 process_init_element callers.
751 (set_nonincremental_init, set_nonincremental_init_from_string):
752 Adjust add_pending_init callers.
753 (output_pending_init_elements): Adjust output_init_element callers.
754 * c-tree.h (process_init_element): Adjust prototype.
755 * c-parser.c (c_parser_initelt, c_parser_initval): Adjust
756 process_init_element callers.