1 2009-01-13 Sebastian Pop <sebastian.pop@amd.com>
3 PR tree-optimization/38786
4 * graphite.c (expand_scalar_variables_ssa_name): New, outlined from
5 the SSA_NAME case of expand_scalar_variables_expr.
6 Set the type of an expression to the type of its assign statement.
7 (expand_scalar_variables_expr): Also gather the scalar computation
8 used to index the memory access. Do not pass loop_p.
9 Fix comment. Stop recursion on tcc_constant or tcc_declaration.
10 (expand_scalar_variables_stmt): Pass to expand_scalar_variables_expr
11 the gimple_stmt_iterator where it inserts new code. Do not pass loop_p.
12 (copy_bb_and_scalar_dependences): Do not pass loop_p.
13 (translate_clast): Update call to copy_bb_and_scalar_dependences.
15 2009-01-13 Sebastian Pop <sebastian.pop@amd.com>
17 * graphite.h (debug_value): Removed.
18 * graphite.c (debug_value): Removed.
20 2009-01-13 Richard Earnshaw <rearnsha@arm.com>
22 * arm.c (output_move_double): Don't synthesize thumb-2 ldrd/strd with
23 two 32-bit instructions.
25 2009-01-13 Richard Earnshaw <rearnsha@arm.com>
27 * arm.c (struct processors): Pass for speed down into cost helper
29 (const_ok_for_op): Handle COMPARE and inequality nodes.
30 (arm_rtx_costs_1): Rewrite.
31 (arm_size_rtx_costs): Update prototype.
32 (arm_rtx_costs): Pass speed down to helper functions.
33 (arm_slowmul_rtx_costs): Rework cost calculations.
34 (arm_fastmul_rtx_costs, arm_xscale_rtx_costs): Likewise.
35 (arm_9e_rtx_costs): Likewise.
37 2009-01-13 Uros Bizjak <ubizjak@gmail.com>
39 * config/alpha/alpha.c (alpha_legitimate_address_p): Explicit
40 relocations of local symbols wider than UNITS_PER_WORD are not valid.
41 (alpha_legitimize_address): Do not split local symbols wider than
42 UNITS_PER_WORD into HIGH/LO_SUM parts.
44 2009-01-13 Danny Smith <dannysmith@users.sourceforge.net>
47 * gcc.c (process_command): Replace call to execvp with calls
50 2009-01-03 Anatoly Sokolov <aesok@post.ru>
53 * config/avr/t-avr (LIB1ASMFUNCS): Add _tablejump_elpm.
54 * config/avr/libgcc.S (__do_global_ctors, __do_global_dtors): Add
55 variant for devices with 3-byte PC.
56 (__tablejump_elpm__) : New.
58 2009-01-12 Jakub Jelinek <jakub@redhat.com>
61 * c-parser.c (c_parser_postfix_expression): Allow `->' in
62 offsetof member-designator, handle it as `[0].'.
64 2009-01-12 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
66 * pa.c (pa_asm_output_mi_thunk): Use pc-relative branch to thunk
67 function when not using named sections on targets with named sections
68 if branch distance is less than 262132.
70 2009-01-12 Richard Earnshaw <rearnsha@arm.com>
72 * combine.c (combine_instructions): Recompute
73 optimize_this_for_speed_p for each BB in the main combine loop.
75 2009-01-12 Tomas Bily <tbily@suse.cz>
78 * tree-loop-distribution.c (prop_phis): New function.
79 (generate_builtin): Call prop_phis.
80 * testsuite/gcc.dg/tree-ssa/pr38385.c: New file.
82 2009-01-12 Jakub Jelinek <jakub@redhat.com>
84 PR tree-optimization/38807
85 * tree-ssa-reassoc.c (remove_visited_stmt_chain): Don't look at
86 gimple_visited_p unless stmt is GIMPLE_ASSIGN.
88 2009-01-11 Adam Nemet <anemet@caviumnetworks.com>
90 * expmed.c (store_bit_field_1): Properly truncate the paradoxical
91 subreg of op0 to the original op0.
93 2009-01-11 Laurent GUERBY <laurent@guerby.net>
95 * doc/sourcebuild.texi (Source Tree): Move up intl and fixinc.
97 2009-01-11 Markus Schoepflin <markus.schoepflin@comsoft.de>
100 * gcc/mips-tfile.c (parse_def): Fix parsing of def strings
101 starting with digits.
103 2009-01-10 Jakub Jelinek <jakub@redhat.com>
106 * config/arm/arm.c (arm_is_long_call_p): Don't call
107 arm_function_in_section_p if decl isn't a FUNCTION_DECL.
109 2009-01-09 Steven Bosscher <steven@gcc.gnu.org>
111 * regrename.c (regrename_optimize): Fix dumping.
112 (find_oldest_value_reg): Preserve REG_POINTER.
113 (copy_hardreg_forward_1): Likewise.
115 2009-01-09 Diego Novillo <dnovillo@google.com>
117 * gimple.h (struct gimple_statement_base)<uid>: Document
118 the restrictions on its use.
120 (gimple_set_uid): Tidy.
122 2009-01-09 Jakub Jelinek <jakub@redhat.com>
124 * config/i386/i386.c (ix86_expand_movmem, ix86_expand_setmem): Add
125 zero guard even if align_bytes != 0 and count is smaller than
128 2008-01-09 Vladimir Makarov <vmakarov@redhat.com>
130 PR rtl-optimization/38495
131 * ira-emit.c (print_move_list, ira_debug_move_list): New functions.
132 (add_range_and_copies_from_move_list): Print all added ranges.
133 Add ranges to memory optimized destination.
135 2009-01-09 Jakub Jelinek <jakub@redhat.com>
139 * config/i386/i386.c (override_options): Reject
140 -mstringop-strategy=rep_8byte with -m32.
141 (ix86_expand_movmem): For size_needed == 1 set epilogue_size_needed
142 to 1. Do count comparison against epilogue_size_needed at compile
143 time even when count_exp was constant forced into register. For
144 size_needed don't jump to epilogue, instead just avoid aligning
145 and invoke the body algorithm. If need_zero_guard, add zero guard
146 even if count is non-zero, but smaller than size_needed + number of
147 bytes that could be stored for alignment.
148 (ix86_expand_setmem): For size_needed == 1 set epilogue_size_needed
149 to 1. If need_zero_guard, add zero guard even if count is non-zero,
150 but smaller than size_needed + number of bytes that could be stored
151 for alignment. Compare size_needed with epilogue_size_needed instead
152 of desired_align - align, don't adjust size_needed, pass
153 epilogue_size_needed to the epilogue expanders.
156 * c-pretty-print.c (pp_c_expression): Handle GOTO_EXPR like BIND_EXPR.
158 2009-01-09 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
160 * pa.c (last_address): Change to unsigned.
161 (update_total_code_bytes): Change argument to unsigned. Don't
162 check if insn addresses are set.
163 (pa_output_function_epilogue): Set last_address to UINT_MAX if insn
164 addresses are not set.
165 (pa_asm_output_mi_thunk): Handle wrap when updating last_address.
167 2009-01-09 Nick Clifton <nickc@redhat.com>
169 * config/sh/symbian.c: Replace uses of DECL_INLINE with
170 DECL_DECLARED_INLINE_P.
172 2009-01-09 Jakub Jelinek <jakub@redhat.com>
175 * dojump.c (do_jump_by_parts_zero_rtx): Use mode instead of
176 GET_MODE (op0) in operand_subword_force calls.
179 * fold-const.c (fold_unary): For COMPOUND_EXPR and COND_EXPR,
180 fold_convert arg0 operands to TREE_TYPE (op0) first.
182 2009-01-08 Vladimir Makarov <vmakarov@redhat.com>
184 * params.def (ira-max-conflict-table-size): Decrease default value
187 2009-01-08 Jakub Jelinek <jakub@redhat.com>
189 PR tree-optimization/37031
190 * lambda-code.c (lambda_collect_parameters): Call pointer_set_destroy
192 (build_access_matrix): Reserve correct size for AM_MATRIX vector,
193 allocate it using gc instead of heap, use VEC_quick_push instead of
195 * graphite.c (build_access_matrix): Allocate AM_MATRIX vector using gc
196 instead of heap, use VEC_quick_push instead of VEC_safe_push.
197 * tree-data-ref.h (struct access_matrix): Change matrix to gc
198 allocated vector from heap allocated.
199 * lambda.h: Add DEF_VEC_ALLOC_P for gc allocated lambda_vector.
200 * tree-loop-linear.c (linear_transform_loops): Allocate nest
201 vector only after perfect_loop_nest_depth call.
203 2009-01-08 Sebastian Pop <sebastian.pop@amd.com>
204 Jan Sjodin <jan.sjodin@amd.com>
206 PR tree-optimization/38559
207 * graphite.c (debug_value, copy_constraint,
208 swap_constraint_variables, scale_constraint_variable, ): New.
209 (get_lower_bound, get_upper_bound): Removed.
210 (graphite_trans_bb_strip_mine): Clean up this code that works
211 only for constant number of iterations. Fully copy upper and
212 lower bound constraints, not only the constant part of them.
213 * graphite.h (debug_value): Declared.
215 2009-01-08 Ira Rosen <irar@il.ibm.com>
217 PR tree-optimization/37194
218 * tree-vect-transform.c (vect_estimate_min_profitable_iters):
219 Don't add the cost of cost model guard in prologue to scalar
220 outside cost in case of known number of iterations.
222 2009-01-07 Nathan Froyd <froydnj@codesourcery.com>
223 Alan Modra <amodra@bigpond.net.au>
225 * config/rs6000/rs6000.c (rs6000_legitimize_address): Check for
226 non-word-aligned REG+CONST addressing.
228 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
231 * config/alpha/alpha.c (alpha_end_function): For TARGET_ABI_OSF, call
232 free_after_compilation when outputting a thunk.
233 (alpha_output_mi_thunk_osf): Assert that we are processing a thunk.
234 Do not call free_after_compilation here.
236 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
238 * config/i386/i386.c (ix86_target_string): Use ARRAY_SIZE.
239 (ix86_valid_target_attribute_inner_p): Ditto.
241 2009-01-07 Jan Sjodin <jan.sjodin@amd.com>
243 PR tree-optimization/38492
244 PR tree-optimization/38498
245 * tree-check.c (operator_is_linear, scev_is_linear_expression): New.
246 * tree-chrec.h (scev_is_linear_expression): Declared.
247 * graphite.c (graphite_cannot_represent_loop_niter): New.
248 (scopdet_basic_block_info): Call graphite_cannot_represent_loop_niter.
249 (graphite_loop_normal_form): Use gcc_assert.
250 (scan_tree_for_params): Use CASE_CONVERT.
251 (phi_node_is_iv, bb_contains_non_iv_scalar_phi_nodes): New.
252 (build_scop_conditions_1): Call bb_contains_non_iv_scalar_phi_nodes.
253 Use gcc_assert. Discard scops that contain unhandled cases.
254 (build_scop_conditions): Return a boolean status for unhandled cases.
255 (strip_mine_profitable_p): Print the loop number, not its depth.
256 (is_interchange_valid): Pass the depth of the loop nest, don't
257 recompute it wrongly.
258 (graphite_trans_bb_block): Same.
259 (graphite_trans_bb_block): Print tentative of loop blocking.
260 (graphite_trans_scop_block): Do not print that the loop has been
262 (graphite_transform_loops): Do not handle scops that contain condition
265 2009-01-07 H.J. Lu <hongjiu.lu@intel.com>
267 AVX Programming Reference (December, 2008)
268 * config/i386/avxintrin.h (_mm256_stream_si256): New.
269 (_mm256_stream_pd): Likewise.
270 (_mm256_stream_ps): Likewise.
272 * config/i386/i386.c (ix86_builtins): Add IX86_BUILTIN_MOVNTDQ256,
273 IX86_BUILTIN_MOVNTPD256 and IX86_BUILTIN_MOVNTPS256.
274 (ix86_special_builtin_type): Add VOID_FTYPE_PV4DI_V4DI.
275 (bdesc_special_args): Add __builtin_ia32_movntdq256,
276 __builtin_ia32_movntpd256 and __builtin_ia32_movntps256.
277 (ix86_init_mmx_sse_builtins): Handle VOID_FTYPE_PV4DI_V4DI.
278 (ix86_expand_special_args_builtin): Likewise.
280 * config/i386/sse.md (AVXMODEDI): New.
281 (avx_movnt<mode>): Likewise.
282 (avx_movnt<mode>): Likewise.
283 (<sse>_movnt<mode>): Remove AVX support.
284 (sse2_movntv2di): Likewise.
286 2009-01-07 Richard Guenther <rguenther@suse.de>
289 * fold-const.c (extract_muldiv): Remove obsolete comment.
290 (fold_plusminus_mult_expr): Undo MINUS_EXPR
291 to PLUS_EXPR canonicalization for the canonicalization.
293 2009-01-07 Gerald Pfeifer <gerald@pfeifer.com>
295 * doc/install.texi (alpha*-dec-osf*): Remove note on 32-bit
296 hosted cross-compilers generating less efficient code.
298 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
300 * function.h (rtl_data): Add a dbr_scheduled_p field.
301 * reorg.c (dbr_schedule): Set it.
302 (gate_handle_delay_slots): Check it.
303 * config/mips/mips.c (mips_base_delayed_branch): Delete.
304 (mips_reorg): Check flag_delayed_branch instead of
305 mips_base_delayed_branch.
306 (mips_override_options): Don't set mips_base_delayed_branch
307 or flag_delayed_branch.
309 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
311 PR rtl-optimization/38426.
312 * ira.c (ira): Set current_function_is_leaf earlier.
314 2009-01-06 Jakub Jelinek <jakub@redhat.com>
316 PR rtl-optimization/38722
317 * combine.c (try_combine): Don't modify PATTERN (i3) and notes
318 too early, only set a flag and modify after last possible
321 2009-01-06 Janis Johnson <janis187@us.ibm.com>
324 * ginclude/float.h: Rename DECnn_DEN to DECnn_SUBNORMAL_MIN.
325 * real.c (decimal_single_format): Correct values of emin and emax.
326 (decimal_double_format): Ditto.
327 (decimal_quad_format): Ditto.
328 * c-cppbuiltin.c (builtin_define_decimal_float_constants): Adjust
329 computation of DECnn_MIN and DECnn_MAX for corrected values of
330 emin and emax. Define __DECnn_SUBNORMAL_MIN__ instead of
331 __DECnn_MIN__, and adjust its computation for the corrected value
334 2009-01-06 Jan Hubicka <jh@suse.cz>
337 * i386.c (ix86_expand_call): Use ARRAY_SIZE.
339 2009-01-06 Gerald Pfeifer <gerald@pfeifer.com>
341 * doc/contrib.texi (Contributors): Slightly adjust the end note.
342 Add Robert Clark to the list of testers.
344 2009-01-06 Jan Hubicka <jh@suse.cz>
345 Kai Tietz <kai.tietz@onevision.com>
347 * i386.md (*msabi_syvabi): Add SSE regs clobbers.
348 * i386.c (ix86_expand_call): Add clobbers.
350 2009-01-06 Jan Hubicka <jh@suse.cz>
351 Kai Tietz <kai.tietz@onevision.com>
353 * i386.h (CONDITIONAL_CALL_USAGE): SSE regs are not used for w64 ABI.
354 * i386.c (struct ix86_frame): Add padding0 and nsseregs.
355 (ix86_nsaved_regs): Count only general purpose regs.
356 (ix86_nsaved_sseregs): New.
357 (ix86_compute_frame_layout): Update nsseregs; set preferred alignment
358 to 16 for w64; compute padding and size of sse reg save area.
359 (ix86_emit_save_regs, ix86_emit_save_regs_using_mov): Save only
360 general purpose regs.
361 (ix86_emit_save_sse_regs_using_mov): New.
362 (ix86_expand_prologue): Save SSE regs if needed.
363 (ix86_emit_restore_regs_using_mov): Use only general purpose regs.
364 (ix86_emit_restore_sse_regs_using_mov): New.
365 (ix86_expand_epilogue): Save SSE regs if needed.
367 2009-01-06 Jan Hubicka <jh@suse.cz>
368 Kai Tietz <kai.tietz@onevision.com>
370 * i386.h (ACCUMULATE_OUTGOING_ARGS): Enable for MSABI
371 * i386.c (init_cumulative_args): Disallow calls of MSABI functions
372 when accumulate outgoing args is off.
374 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
377 * ira-color.c (ira_reuse_stack_slot): Check ENABLE_IRA_CHECKING
378 before using pseudos_have_intersected_live_ranges_p.
380 * ira-int.h (ira_assert): Always define.
382 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
384 AVX Programming Reference (December, 2008)
385 * config/i386/avxintrin.h (_mm_permute2_pd): Removed.
386 (_mm256_permute2_pd): Likewise.
387 (_mm_permute2_ps): Likewise.
388 (_mm256_permute2_ps): Likewise.
389 * config/i386/i386.md (UNSPEC_VPERMIL2): Likewise.
390 * config/i386/sse.md (avx_vpermil2<mode>3): Likewise.
392 * config/i386/i386.c (ix86_builtins): Remove
393 IX86_BUILTIN_VPERMIL2PD, IX86_BUILTIN_VPERMIL2PS,
394 IX86_BUILTIN_VPERMIL2PD256 and IX86_BUILTIN_VPERMIL2PS256.
395 (ix86_builtin_type): Remove V8SF_FTYPE_V8SF_V8SF_V8SI_INT,
396 V4DF_FTYPE_V4DF_V4DF_V4DI_INT, V4SF_FTYPE_V4SF_V4SF_V4SI_INT
397 and V2DF_FTYPE_V2DF_V2DF_V2DI_INT.
398 (bdesc_args): Remove __builtin_ia32_vpermil2pd,
399 __builtin_ia32_vpermil2ps, __builtin_ia32_vpermil2pd256 and
400 __builtin_ia32_vpermil2ps256.
401 (ix86_init_mmx_sse_builtins): Updated.
402 (ix86_expand_args_builtin): Likewise.
404 2009-01-05 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
406 * pa.c (output_call): Relocate non-jump insns in the delay slot of
407 long absolute calls when generating PA 2.0 code.
409 2009-01-05 Vladimir Makarov <vmakarov@redhat.com>
411 PR rtl-optimization/38583
412 * params.h (IRA_MAX_CONFLICT_TABLE_SIZE): New macro.
414 * params.def (ira-max-conflict-table-size): New.
416 * doc/invoke.texi (ira-max-conflict-table-size): Decribe.
418 * ira.h (ira_conflicts_p): New external definition.
420 * ira-conflicts.c (build_conflict_bit_table): Do not build too big
421 table. Report this. Return result of building.
422 (ira_build_conflicts): Use ira_conflicts_p. Check result of
423 building conflict table.
425 * ira-color.c (fast_allocation): Use num instead of ira_allocnos_num.
426 (ira_color): Use ira_conflicts_p.
428 * global.c: Include ira.h.
429 (pseudo_for_reload_consideration_p, build_insn_chain): Use
432 * Makefile.in (global.o): Add ira.h.
434 * ira-build.c (mark_all_loops_for_removal,
435 propagate_some_info_from_allocno): New.
436 (remove_unnecessary_allocnos): Call
437 propagate_some_info_from_allocno.
438 (remove_low_level_allocnos): New.
439 (remove_unnecessary_regions): Add parameter. Call
440 mark_all_loops_for_removal and remove_low_level_allocnos. Pass
441 parameter to remove_unnecessary_regions.
442 (ira_build): Remove all regions but root if the conflict table was
443 not built. Update conflict hard regs for allocnos crossing calls.
445 * ira.c (ira_conflicts_p): New global.
446 (ira): Define and use ira_conflicts_p.
448 * reload1.c (compute_use_by_pseudos, reload, count_pseudo,
449 count_spilled_pseudo, find_reg, alter_reg, finish_spills,
450 emit_input_reload_insns, delete_output_reload): Use ira_conflicts_p.
452 2009-01-06 Ben Elliston <bje@au.ibm.com>
454 * gengtype-lex.l (YY_NO_INPUT): Define.
456 2009-01-05 Andrew Pinski <andrew_pinski@playstation.sony.com>
459 * c-common.c (handle_vector_size_attribute): Also reject
462 2009-01-05 Sebastian Pop <sebastian.pop@amd.com>
464 PR tree-optimization/38492
465 * graphite.c (rename_map_elt, debug_rename_elt,
466 debug_rename_map_1, debug_rename_map, new_rename_map_elt,
467 rename_map_elt_info, eq_rename_map_elts,
468 get_new_name_from_old_name, bb_in_sese_p): Moved around.
469 (sese_find_uses_to_rename_use): Renamed sese_build_livein_liveouts_use.
470 (sese_find_uses_to_rename_bb): Renamed sese_build_livein_liveouts_bb.
471 (sese_build_livein_liveouts): New.
472 (new_sese, free_sese): New.
473 (new_scop): Call new_sese.
474 (free_scop): Call free_sese.
475 (rename_variables_from_edge, rename_phis_end_scop): Removed.
476 (register_old_new_names): Renamed register_old_and_new_names.
477 (register_scop_liveout_renames, add_loop_exit_phis,
478 insert_loop_close_phis, struct igp,
479 default_liveout_before_guard, add_guard_exit_phis,
480 insert_guard_phis, copy_renames): New.
481 (translate_clast): Call insert_loop_close_phis and insert_guard_phis.
482 (sese_add_exit_phis_edge): Renamed scop_add_exit_phis_edge.
483 (rewrite_into_sese_closed_ssa): Renamed scop_insert_phis_for_liveouts.
484 (scop_adjust_phis_for_liveouts): New.
485 (gloog): Call scop_adjust_phis_for_liveouts.
487 * graphite.h (struct sese): Documented. Added fields liveout,
489 (SESE_LIVEOUT, SESE_LIVEIN, SESE_LIVEIN_VER, SESE_NUM_VER): New.
490 (new_sese, free_sese, sese_build_livein_liveouts): Declared.
491 (struct scop): Added field liveout_renames.
492 (SCOP_LIVEOUT_RENAMES): New.
494 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
496 PR tree-optimization/38510
497 * graphite.c (recompute_all_dominators): Call mark_irreducible_loops.
498 (translate_clast): Call recompute_all_dominators before
500 (gloog): Call recompute_all_dominators before graphite_verify.
502 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
503 Jan Sjodin <jan.sjodin@amd.com>
505 PR tree-optimization/38500
506 * graphite.c (create_sese_edges): Call fix_loop_structure after
509 2009-01-05 Joel Sherrill <joel.sherrill@oarcorp.com>
511 * config.gcc: Add m32r*-*-rtems*.
512 * config/m32r/rtems.h: New file.
514 2009-01-05 Ben Elliston <bje@au.ibm.com>
516 * Makefile.in (.po.gmo): Use mkinstalldirs, not test -d || mkdir.
518 (po/gcc.pot): Likewise.
520 2009-01-04 David S. Miller <davem@davemloft.net>
522 * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
523 (STARTING_FRAME_OFFSET): Always set to zero.
525 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
527 * tree.def (LSHIFT_EXPR, RSHIFT_EXPR): Add commentary.
528 * tree-cfg.c (verify_gimple_assign_binary): Allow shifts of
529 fixed-point types, and vectors of the same.
531 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
533 * config/mips/sync.md (*mb_barrier): Rename to...
534 (*memory_barrier): ...this.
536 2009-01-04 Jonathan Wakely <jwakely.gcc@gmail.com>
538 * doc/extend.texi (Function Attributes): Move @cindex after @item
539 for 'artificial' and 'flatten'. Fix grammar for 'externally_visible'
540 and put in alphabetical order. Fix 'target' name and put in order.
541 * doc/invoke.texi (-Wstrict-null-sentinel, -fipa-matrix-reorg): Fix
544 2009-01-04 Uros Bizjak <ubizjak@gmail.com>
546 * config/s390/s390.md (UNSPEC_MB): Rename from UNSPECV_MB.
547 (memory_barrier): Expand as unspec instead of unspec_volatile.
548 Remove mem:BLK from insn operands. Use Pmode scratch register.
549 (*memory_barrier): Define as unspec instead of unspec_volatile.
550 Use (match_dup 0) as input operand.
552 * config/sparc/sparc.md (UNSPEC_MEMBAR): Rename from UNSPECV_MEMBAR.
553 * config/sparc/sync.md (memory_barrier): Expand as unspec instead of
554 unspec_volatile. Remove mem:BLK from insn operands. Use Pmode
555 scratch register. Remove operand 1.
556 (*stbar): Define as unspec instead of unspec_volatile.
557 Use (match_dup 0) as input operand, remove (const_int 8).
558 (*membar): Define as unspec instead of unspec_volatile.
559 Use (match_dup 0) as input operand, remove input operand 2.
561 * config/xtensa/xtensa.md (UNSPEC_MEMW): Rename from UNSPECV_MEMW.
562 (memory_barrier): Expand as unspec instead of unspec_volatile.
563 Remove mem:BLK from insn operands. Use Pmode scratch register.
564 (*memory_barrier): Define as unspec instead of unspec_volatile.
565 Use (match_dup 0) as input operand.
567 * config/ia64/sync.md (memory_barrier): Redefine as expander pattern.
568 Remove mem:BLK from insn operands. Use Pmode scratch register.
569 Set volatile flag on operand 0.
570 (*memory_barrier): New insn pattern.
572 * config/rs6000/sync.md (memory_barrier): Remove mem:BLK from
574 (*memory_barrier): Use (match_dup 0) as input operand.
576 * config/mips/sync.md (memory_barrier): Redefine as expander pattern.
577 Remove mem:BLK from insn operands. Use Pmode scratch register.
578 Set volatile flag on operand 0.
579 (*mb_internal): New insn pattern.
581 * config/alpha/sync.md (*memory_barrier): Rename from *mb_internal.
583 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
586 * function.c (struct temp_slot): Move to the section of the file
587 that deals with temp slots. Remove field 'address'.
588 (temp_slot_address_table): New hash table of address -> temp slot.
589 (struct temp_slot_address_entry): New struct, items for the table.
590 (temp_slot_address_compute_hash, temp_slot_address_hash,
591 temp_slot_address_eq, insert_temp_slot_address): Support functions
593 (find_temp_slot_from_address): Rewrite to use the new hash table.
594 (remove_unused_temp_slot_addresses): Remove addresses of temp
595 slots that have been made available.
596 (remove_unused_temp_slot_addresses_1): Call-back for htab_traverse,
597 worker function for remove_unused_temp_slot_addresses.
598 (assign_stack_temp_for_type): Don't clear the temp slot address list.
599 Add the temp slot address to the address -> temp slot map.
600 (update_temp_slot_address): Update via insert_temp_slot_address.
601 (free_temp_slots): Call remove_unused_temp_slot_addresses.
602 (pop_temp_slots): Likewise.
603 (init_temp_slots): Allocate the address -> temp slot map, or empty
604 the map if it is already allocated.
605 (prepare_function_start): Initialize temp slot processing.
607 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
610 * cfgexpand.c (estimate_stack_frame_size): Simplify the estimate:
611 Calculate the size of all stack vars assuming no packing of stack
612 vars will happen, replacing a quadratic algorithm with a linear one.
614 2009-01-03 Jakub Jelinek <jakub@redhat.com>
617 * expmed.c (store_bit_field_1): Don't modify op0 if movstrict insn
620 2009-01-03 Diego Novillo <dnovillo@google.com>
622 * doc/contrib.texi: Update contributions.
624 2009-01-03 Jakub Jelinek <jakub@redhat.com>
627 * builtins.c (fold_builtin_memory_op): Give up if either operand
628 is volatile. Set srctype or desttype to non-qualified version
632 * builtins.c (fold_builtin_expect): Only check DECL_WEAK for VAR_DECLs
635 2009-01-02 Kenneth Zadeck <zadeck@naturalbridge.com>
637 PR rtl-optimization/35805
638 * df-problems.c (df_lr_finalize): Add recursive call to resolve lr
639 problem if fast dce is able to remove any instructions.
640 * dce.c (dce_process_block): Fix dump message.
642 2009-01-02 Mark Mitchell <mark@codesourcery.com>
645 * tree-ssa-pre.c (compute_antic): Correct loop bounds.
647 2009-01-02 Jakub Jelinek <jakub@redhat.com>
650 * tree-flow.h (op_code_prio, op_prio): New prototypes.
651 * tree-pretty-print.c (op_code_prio): New function.
652 (op_prio): No longer static. Use op_code_prio.
653 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs):
654 Use op_prio and op_code_prio to determine if () should be
655 printed around operand(s) or not.
657 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs,
658 dump_gimple_call, dump_gimple_switch, dump_gimple_cond,
659 dump_gimple_label, dump_gimple_try, dump_symbols, dump_gimple_phi,
660 dump_gimple_mem_ops, dump_bb_header, dump_bb_end, pp_cfg_jump): Use
661 pp_character instead of pp_string for single letter printing.
663 2009-01-02 Richard Sandiford <rdsandiford@googlemail.com>
665 * doc/extend.texi: Fix '#pragma GCC option' typo.
667 2009-01-02 Richard Guenther <rguenther@suse.de>
669 * doc/install.texi (--enable-checking): Mention different
671 (--enable-stage1-checking): Document.
673 2009-01-01 Andrew Pinski <pinskia@gmail.com>
676 * tree-cfg.c (verify_expr): Add INDIRECT_REF case. Change MODIFY_EXPR
679 2009-01-02 Ben Elliston <bje@au.ibm.com>
681 * config/fp-bit.h (pack_d): Constify argument.
682 * config/fp-bit.c (makenan): Constify return type. Remove casts.
683 (isnan): Constify argument.
687 (_fpadd_parts): Constify return type.
688 (_fpmul_parts): Likewise.
689 (_fpdiv_parts): Likewise.
691 2009-01-01 Jakub Jelinek <jakub@redhat.com>
694 * c-typeck.c (add_pending_init): Add IMPLICIT argument. Only
695 warn about overwriting initializer with side-effects or
696 -Woverride-init if !IMPLICIT.
697 (output_init_element): Likewise. Pass IMPLICIT down to
699 (process_init_element): Add IMPLICIT argument. Pass it down
700 to output_init_element.
701 (push_init_element, pop_init_level, set_designator): Adjust
702 process_init_element callers.
703 (set_nonincremental_init, set_nonincremental_init_from_string):
704 Adjust add_pending_init callers.
705 (output_pending_init_elements): Adjust output_init_element callers.
706 * c-tree.h (process_init_element): Adjust prototype.
707 * c-parser.c (c_parser_initelt, c_parser_initval): Adjust
708 process_init_element callers.