1 2009-01-14 Nick Clifton <nickc@redhat.com>
3 * ira-conflicts.c: Include addresses.h for the definition of
5 (ira_build_conflicts): Use base_reg_class instead of
7 * Makefile.in: Add a dependency of ira-conflicts.o on
10 2009-01-13 Vladimir Makarov <vmakarov@redhat.com>
13 * Makefile.in (ira-lives.o): Add except.h.
15 * ira-lives.c: Include except.h.
16 (process_bb_node_lives): Process can_throw_internal.
18 2009-01-13 Jakub Jelinek <jakub@redhat.com>
20 PR rtl-optimization/38774
21 * combine.c (simplify_set): When undoing cc_use change, don't do
22 PUT_CODE on the newly created comparison, but instead put back the
25 2009-01-13 Joseph Myers <joseph@codesourcery.com>
27 * doc/invoke.texi (ARM Options): Update lists of -mcpu and -march
28 values. Remove duplicate arm8 entry.
30 2009-01-13 Sebastian Pop <sebastian.pop@amd.com>
32 PR tree-optimization/38786
33 * graphite.c (expand_scalar_variables_ssa_name): New, outlined from
34 the SSA_NAME case of expand_scalar_variables_expr.
35 Set the type of an expression to the type of its assign statement.
36 (expand_scalar_variables_expr): Also gather the scalar computation
37 used to index the memory access. Do not pass loop_p.
38 Fix comment. Stop recursion on tcc_constant or tcc_declaration.
39 (expand_scalar_variables_stmt): Pass to expand_scalar_variables_expr
40 the gimple_stmt_iterator where it inserts new code. Do not pass loop_p.
41 (copy_bb_and_scalar_dependences): Do not pass loop_p.
42 (translate_clast): Update call to copy_bb_and_scalar_dependences.
44 2009-01-13 Sebastian Pop <sebastian.pop@amd.com>
46 * graphite.h (debug_value): Removed.
47 * graphite.c (debug_value): Removed.
49 2009-01-13 Richard Earnshaw <rearnsha@arm.com>
51 * arm.c (output_move_double): Don't synthesize thumb-2 ldrd/strd with
52 two 32-bit instructions.
54 2009-01-13 Richard Earnshaw <rearnsha@arm.com>
56 * arm.c (struct processors): Pass for speed down into cost helper
58 (const_ok_for_op): Handle COMPARE and inequality nodes.
59 (arm_rtx_costs_1): Rewrite.
60 (arm_size_rtx_costs): Update prototype.
61 (arm_rtx_costs): Pass speed down to helper functions.
62 (arm_slowmul_rtx_costs): Rework cost calculations.
63 (arm_fastmul_rtx_costs, arm_xscale_rtx_costs): Likewise.
64 (arm_9e_rtx_costs): Likewise.
66 2009-01-13 Uros Bizjak <ubizjak@gmail.com>
68 * config/alpha/alpha.c (alpha_legitimate_address_p): Explicit
69 relocations of local symbols wider than UNITS_PER_WORD are not valid.
70 (alpha_legitimize_address): Do not split local symbols wider than
71 UNITS_PER_WORD into HIGH/LO_SUM parts.
73 2009-01-13 Danny Smith <dannysmith@users.sourceforge.net>
76 * gcc.c (process_command): Replace call to execvp with calls
79 2009-01-03 Anatoly Sokolov <aesok@post.ru>
82 * config/avr/t-avr (LIB1ASMFUNCS): Add _tablejump_elpm.
83 * config/avr/libgcc.S (__do_global_ctors, __do_global_dtors): Add
84 variant for devices with 3-byte PC.
85 (__tablejump_elpm__) : New.
87 2009-01-12 Jakub Jelinek <jakub@redhat.com>
90 * c-parser.c (c_parser_postfix_expression): Allow `->' in
91 offsetof member-designator, handle it as `[0].'.
93 2009-01-12 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
95 * pa.c (pa_asm_output_mi_thunk): Use pc-relative branch to thunk
96 function when not using named sections on targets with named sections
97 if branch distance is less than 262132.
99 2009-01-12 Richard Earnshaw <rearnsha@arm.com>
101 * combine.c (combine_instructions): Recompute
102 optimize_this_for_speed_p for each BB in the main combine loop.
104 2009-01-12 Tomas Bily <tbily@suse.cz>
107 * tree-loop-distribution.c (prop_phis): New function.
108 (generate_builtin): Call prop_phis.
109 * testsuite/gcc.dg/tree-ssa/pr38385.c: New file.
111 2009-01-12 Jakub Jelinek <jakub@redhat.com>
113 PR tree-optimization/38807
114 * tree-ssa-reassoc.c (remove_visited_stmt_chain): Don't look at
115 gimple_visited_p unless stmt is GIMPLE_ASSIGN.
117 2009-01-11 Adam Nemet <anemet@caviumnetworks.com>
119 * expmed.c (store_bit_field_1): Properly truncate the paradoxical
120 subreg of op0 to the original op0.
122 2009-01-11 Laurent GUERBY <laurent@guerby.net>
124 * doc/sourcebuild.texi (Source Tree): Move up intl and fixinc.
126 2009-01-11 Markus Schoepflin <markus.schoepflin@comsoft.de>
129 * gcc/mips-tfile.c (parse_def): Fix parsing of def strings
130 starting with digits.
132 2009-01-10 Jakub Jelinek <jakub@redhat.com>
135 * config/arm/arm.c (arm_is_long_call_p): Don't call
136 arm_function_in_section_p if decl isn't a FUNCTION_DECL.
138 2009-01-09 Steven Bosscher <steven@gcc.gnu.org>
140 * regrename.c (regrename_optimize): Fix dumping.
141 (find_oldest_value_reg): Preserve REG_POINTER.
142 (copy_hardreg_forward_1): Likewise.
144 2009-01-09 Diego Novillo <dnovillo@google.com>
146 * gimple.h (struct gimple_statement_base)<uid>: Document
147 the restrictions on its use.
149 (gimple_set_uid): Tidy.
151 2009-01-09 Jakub Jelinek <jakub@redhat.com>
153 * config/i386/i386.c (ix86_expand_movmem, ix86_expand_setmem): Add
154 zero guard even if align_bytes != 0 and count is smaller than
157 2008-01-09 Vladimir Makarov <vmakarov@redhat.com>
159 PR rtl-optimization/38495
160 * ira-emit.c (print_move_list, ira_debug_move_list): New functions.
161 (add_range_and_copies_from_move_list): Print all added ranges.
162 Add ranges to memory optimized destination.
164 2009-01-09 Jakub Jelinek <jakub@redhat.com>
168 * config/i386/i386.c (override_options): Reject
169 -mstringop-strategy=rep_8byte with -m32.
170 (ix86_expand_movmem): For size_needed == 1 set epilogue_size_needed
171 to 1. Do count comparison against epilogue_size_needed at compile
172 time even when count_exp was constant forced into register. For
173 size_needed don't jump to epilogue, instead just avoid aligning
174 and invoke the body algorithm. If need_zero_guard, add zero guard
175 even if count is non-zero, but smaller than size_needed + number of
176 bytes that could be stored for alignment.
177 (ix86_expand_setmem): For size_needed == 1 set epilogue_size_needed
178 to 1. If need_zero_guard, add zero guard even if count is non-zero,
179 but smaller than size_needed + number of bytes that could be stored
180 for alignment. Compare size_needed with epilogue_size_needed instead
181 of desired_align - align, don't adjust size_needed, pass
182 epilogue_size_needed to the epilogue expanders.
185 * c-pretty-print.c (pp_c_expression): Handle GOTO_EXPR like BIND_EXPR.
187 2009-01-09 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
189 * pa.c (last_address): Change to unsigned.
190 (update_total_code_bytes): Change argument to unsigned. Don't
191 check if insn addresses are set.
192 (pa_output_function_epilogue): Set last_address to UINT_MAX if insn
193 addresses are not set.
194 (pa_asm_output_mi_thunk): Handle wrap when updating last_address.
196 2009-01-09 Nick Clifton <nickc@redhat.com>
198 * config/sh/symbian.c: Replace uses of DECL_INLINE with
199 DECL_DECLARED_INLINE_P.
201 2009-01-09 Jakub Jelinek <jakub@redhat.com>
204 * dojump.c (do_jump_by_parts_zero_rtx): Use mode instead of
205 GET_MODE (op0) in operand_subword_force calls.
208 * fold-const.c (fold_unary): For COMPOUND_EXPR and COND_EXPR,
209 fold_convert arg0 operands to TREE_TYPE (op0) first.
211 2009-01-08 Vladimir Makarov <vmakarov@redhat.com>
213 * params.def (ira-max-conflict-table-size): Decrease default value
216 2009-01-08 Jakub Jelinek <jakub@redhat.com>
218 PR tree-optimization/37031
219 * lambda-code.c (lambda_collect_parameters): Call pointer_set_destroy
221 (build_access_matrix): Reserve correct size for AM_MATRIX vector,
222 allocate it using gc instead of heap, use VEC_quick_push instead of
224 * graphite.c (build_access_matrix): Allocate AM_MATRIX vector using gc
225 instead of heap, use VEC_quick_push instead of VEC_safe_push.
226 * tree-data-ref.h (struct access_matrix): Change matrix to gc
227 allocated vector from heap allocated.
228 * lambda.h: Add DEF_VEC_ALLOC_P for gc allocated lambda_vector.
229 * tree-loop-linear.c (linear_transform_loops): Allocate nest
230 vector only after perfect_loop_nest_depth call.
232 2009-01-08 Sebastian Pop <sebastian.pop@amd.com>
233 Jan Sjodin <jan.sjodin@amd.com>
235 PR tree-optimization/38559
236 * graphite.c (debug_value, copy_constraint,
237 swap_constraint_variables, scale_constraint_variable, ): New.
238 (get_lower_bound, get_upper_bound): Removed.
239 (graphite_trans_bb_strip_mine): Clean up this code that works
240 only for constant number of iterations. Fully copy upper and
241 lower bound constraints, not only the constant part of them.
242 * graphite.h (debug_value): Declared.
244 2009-01-08 Ira Rosen <irar@il.ibm.com>
246 PR tree-optimization/37194
247 * tree-vect-transform.c (vect_estimate_min_profitable_iters):
248 Don't add the cost of cost model guard in prologue to scalar
249 outside cost in case of known number of iterations.
251 2009-01-07 Nathan Froyd <froydnj@codesourcery.com>
252 Alan Modra <amodra@bigpond.net.au>
254 * config/rs6000/rs6000.c (rs6000_legitimize_address): Check for
255 non-word-aligned REG+CONST addressing.
257 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
260 * config/alpha/alpha.c (alpha_end_function): For TARGET_ABI_OSF, call
261 free_after_compilation when outputting a thunk.
262 (alpha_output_mi_thunk_osf): Assert that we are processing a thunk.
263 Do not call free_after_compilation here.
265 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
267 * config/i386/i386.c (ix86_target_string): Use ARRAY_SIZE.
268 (ix86_valid_target_attribute_inner_p): Ditto.
270 2009-01-07 Jan Sjodin <jan.sjodin@amd.com>
272 PR tree-optimization/38492
273 PR tree-optimization/38498
274 * tree-check.c (operator_is_linear, scev_is_linear_expression): New.
275 * tree-chrec.h (scev_is_linear_expression): Declared.
276 * graphite.c (graphite_cannot_represent_loop_niter): New.
277 (scopdet_basic_block_info): Call graphite_cannot_represent_loop_niter.
278 (graphite_loop_normal_form): Use gcc_assert.
279 (scan_tree_for_params): Use CASE_CONVERT.
280 (phi_node_is_iv, bb_contains_non_iv_scalar_phi_nodes): New.
281 (build_scop_conditions_1): Call bb_contains_non_iv_scalar_phi_nodes.
282 Use gcc_assert. Discard scops that contain unhandled cases.
283 (build_scop_conditions): Return a boolean status for unhandled cases.
284 (strip_mine_profitable_p): Print the loop number, not its depth.
285 (is_interchange_valid): Pass the depth of the loop nest, don't
286 recompute it wrongly.
287 (graphite_trans_bb_block): Same.
288 (graphite_trans_bb_block): Print tentative of loop blocking.
289 (graphite_trans_scop_block): Do not print that the loop has been
291 (graphite_transform_loops): Do not handle scops that contain condition
294 2009-01-07 H.J. Lu <hongjiu.lu@intel.com>
296 AVX Programming Reference (December, 2008)
297 * config/i386/avxintrin.h (_mm256_stream_si256): New.
298 (_mm256_stream_pd): Likewise.
299 (_mm256_stream_ps): Likewise.
301 * config/i386/i386.c (ix86_builtins): Add IX86_BUILTIN_MOVNTDQ256,
302 IX86_BUILTIN_MOVNTPD256 and IX86_BUILTIN_MOVNTPS256.
303 (ix86_special_builtin_type): Add VOID_FTYPE_PV4DI_V4DI.
304 (bdesc_special_args): Add __builtin_ia32_movntdq256,
305 __builtin_ia32_movntpd256 and __builtin_ia32_movntps256.
306 (ix86_init_mmx_sse_builtins): Handle VOID_FTYPE_PV4DI_V4DI.
307 (ix86_expand_special_args_builtin): Likewise.
309 * config/i386/sse.md (AVXMODEDI): New.
310 (avx_movnt<mode>): Likewise.
311 (avx_movnt<mode>): Likewise.
312 (<sse>_movnt<mode>): Remove AVX support.
313 (sse2_movntv2di): Likewise.
315 2009-01-07 Richard Guenther <rguenther@suse.de>
318 * fold-const.c (extract_muldiv): Remove obsolete comment.
319 (fold_plusminus_mult_expr): Undo MINUS_EXPR
320 to PLUS_EXPR canonicalization for the canonicalization.
322 2009-01-07 Gerald Pfeifer <gerald@pfeifer.com>
324 * doc/install.texi (alpha*-dec-osf*): Remove note on 32-bit
325 hosted cross-compilers generating less efficient code.
327 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
329 * function.h (rtl_data): Add a dbr_scheduled_p field.
330 * reorg.c (dbr_schedule): Set it.
331 (gate_handle_delay_slots): Check it.
332 * config/mips/mips.c (mips_base_delayed_branch): Delete.
333 (mips_reorg): Check flag_delayed_branch instead of
334 mips_base_delayed_branch.
335 (mips_override_options): Don't set mips_base_delayed_branch
336 or flag_delayed_branch.
338 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
340 PR rtl-optimization/38426.
341 * ira.c (ira): Set current_function_is_leaf earlier.
343 2009-01-06 Jakub Jelinek <jakub@redhat.com>
345 PR rtl-optimization/38722
346 * combine.c (try_combine): Don't modify PATTERN (i3) and notes
347 too early, only set a flag and modify after last possible
350 2009-01-06 Janis Johnson <janis187@us.ibm.com>
353 * ginclude/float.h: Rename DECnn_DEN to DECnn_SUBNORMAL_MIN.
354 * real.c (decimal_single_format): Correct values of emin and emax.
355 (decimal_double_format): Ditto.
356 (decimal_quad_format): Ditto.
357 * c-cppbuiltin.c (builtin_define_decimal_float_constants): Adjust
358 computation of DECnn_MIN and DECnn_MAX for corrected values of
359 emin and emax. Define __DECnn_SUBNORMAL_MIN__ instead of
360 __DECnn_MIN__, and adjust its computation for the corrected value
363 2009-01-06 Jan Hubicka <jh@suse.cz>
366 * i386.c (ix86_expand_call): Use ARRAY_SIZE.
368 2009-01-06 Gerald Pfeifer <gerald@pfeifer.com>
370 * doc/contrib.texi (Contributors): Slightly adjust the end note.
371 Add Robert Clark to the list of testers.
373 2009-01-06 Jan Hubicka <jh@suse.cz>
374 Kai Tietz <kai.tietz@onevision.com>
376 * i386.md (*msabi_syvabi): Add SSE regs clobbers.
377 * i386.c (ix86_expand_call): Add clobbers.
379 2009-01-06 Jan Hubicka <jh@suse.cz>
380 Kai Tietz <kai.tietz@onevision.com>
382 * i386.h (CONDITIONAL_CALL_USAGE): SSE regs are not used for w64 ABI.
383 * i386.c (struct ix86_frame): Add padding0 and nsseregs.
384 (ix86_nsaved_regs): Count only general purpose regs.
385 (ix86_nsaved_sseregs): New.
386 (ix86_compute_frame_layout): Update nsseregs; set preferred alignment
387 to 16 for w64; compute padding and size of sse reg save area.
388 (ix86_emit_save_regs, ix86_emit_save_regs_using_mov): Save only
389 general purpose regs.
390 (ix86_emit_save_sse_regs_using_mov): New.
391 (ix86_expand_prologue): Save SSE regs if needed.
392 (ix86_emit_restore_regs_using_mov): Use only general purpose regs.
393 (ix86_emit_restore_sse_regs_using_mov): New.
394 (ix86_expand_epilogue): Save SSE regs if needed.
396 2009-01-06 Jan Hubicka <jh@suse.cz>
397 Kai Tietz <kai.tietz@onevision.com>
399 * i386.h (ACCUMULATE_OUTGOING_ARGS): Enable for MSABI
400 * i386.c (init_cumulative_args): Disallow calls of MSABI functions
401 when accumulate outgoing args is off.
403 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
406 * ira-color.c (ira_reuse_stack_slot): Check ENABLE_IRA_CHECKING
407 before using pseudos_have_intersected_live_ranges_p.
409 * ira-int.h (ira_assert): Always define.
411 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
413 AVX Programming Reference (December, 2008)
414 * config/i386/avxintrin.h (_mm_permute2_pd): Removed.
415 (_mm256_permute2_pd): Likewise.
416 (_mm_permute2_ps): Likewise.
417 (_mm256_permute2_ps): Likewise.
418 * config/i386/i386.md (UNSPEC_VPERMIL2): Likewise.
419 * config/i386/sse.md (avx_vpermil2<mode>3): Likewise.
421 * config/i386/i386.c (ix86_builtins): Remove
422 IX86_BUILTIN_VPERMIL2PD, IX86_BUILTIN_VPERMIL2PS,
423 IX86_BUILTIN_VPERMIL2PD256 and IX86_BUILTIN_VPERMIL2PS256.
424 (ix86_builtin_type): Remove V8SF_FTYPE_V8SF_V8SF_V8SI_INT,
425 V4DF_FTYPE_V4DF_V4DF_V4DI_INT, V4SF_FTYPE_V4SF_V4SF_V4SI_INT
426 and V2DF_FTYPE_V2DF_V2DF_V2DI_INT.
427 (bdesc_args): Remove __builtin_ia32_vpermil2pd,
428 __builtin_ia32_vpermil2ps, __builtin_ia32_vpermil2pd256 and
429 __builtin_ia32_vpermil2ps256.
430 (ix86_init_mmx_sse_builtins): Updated.
431 (ix86_expand_args_builtin): Likewise.
433 2009-01-05 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
435 * pa.c (output_call): Relocate non-jump insns in the delay slot of
436 long absolute calls when generating PA 2.0 code.
438 2009-01-05 Vladimir Makarov <vmakarov@redhat.com>
440 PR rtl-optimization/38583
441 * params.h (IRA_MAX_CONFLICT_TABLE_SIZE): New macro.
443 * params.def (ira-max-conflict-table-size): New.
445 * doc/invoke.texi (ira-max-conflict-table-size): Decribe.
447 * ira.h (ira_conflicts_p): New external definition.
449 * ira-conflicts.c (build_conflict_bit_table): Do not build too big
450 table. Report this. Return result of building.
451 (ira_build_conflicts): Use ira_conflicts_p. Check result of
452 building conflict table.
454 * ira-color.c (fast_allocation): Use num instead of ira_allocnos_num.
455 (ira_color): Use ira_conflicts_p.
457 * global.c: Include ira.h.
458 (pseudo_for_reload_consideration_p, build_insn_chain): Use
461 * Makefile.in (global.o): Add ira.h.
463 * ira-build.c (mark_all_loops_for_removal,
464 propagate_some_info_from_allocno): New.
465 (remove_unnecessary_allocnos): Call
466 propagate_some_info_from_allocno.
467 (remove_low_level_allocnos): New.
468 (remove_unnecessary_regions): Add parameter. Call
469 mark_all_loops_for_removal and remove_low_level_allocnos. Pass
470 parameter to remove_unnecessary_regions.
471 (ira_build): Remove all regions but root if the conflict table was
472 not built. Update conflict hard regs for allocnos crossing calls.
474 * ira.c (ira_conflicts_p): New global.
475 (ira): Define and use ira_conflicts_p.
477 * reload1.c (compute_use_by_pseudos, reload, count_pseudo,
478 count_spilled_pseudo, find_reg, alter_reg, finish_spills,
479 emit_input_reload_insns, delete_output_reload): Use ira_conflicts_p.
481 2009-01-06 Ben Elliston <bje@au.ibm.com>
483 * gengtype-lex.l (YY_NO_INPUT): Define.
485 2009-01-05 Andrew Pinski <andrew_pinski@playstation.sony.com>
488 * c-common.c (handle_vector_size_attribute): Also reject
491 2009-01-05 Sebastian Pop <sebastian.pop@amd.com>
493 PR tree-optimization/38492
494 * graphite.c (rename_map_elt, debug_rename_elt,
495 debug_rename_map_1, debug_rename_map, new_rename_map_elt,
496 rename_map_elt_info, eq_rename_map_elts,
497 get_new_name_from_old_name, bb_in_sese_p): Moved around.
498 (sese_find_uses_to_rename_use): Renamed sese_build_livein_liveouts_use.
499 (sese_find_uses_to_rename_bb): Renamed sese_build_livein_liveouts_bb.
500 (sese_build_livein_liveouts): New.
501 (new_sese, free_sese): New.
502 (new_scop): Call new_sese.
503 (free_scop): Call free_sese.
504 (rename_variables_from_edge, rename_phis_end_scop): Removed.
505 (register_old_new_names): Renamed register_old_and_new_names.
506 (register_scop_liveout_renames, add_loop_exit_phis,
507 insert_loop_close_phis, struct igp,
508 default_liveout_before_guard, add_guard_exit_phis,
509 insert_guard_phis, copy_renames): New.
510 (translate_clast): Call insert_loop_close_phis and insert_guard_phis.
511 (sese_add_exit_phis_edge): Renamed scop_add_exit_phis_edge.
512 (rewrite_into_sese_closed_ssa): Renamed scop_insert_phis_for_liveouts.
513 (scop_adjust_phis_for_liveouts): New.
514 (gloog): Call scop_adjust_phis_for_liveouts.
516 * graphite.h (struct sese): Documented. Added fields liveout,
518 (SESE_LIVEOUT, SESE_LIVEIN, SESE_LIVEIN_VER, SESE_NUM_VER): New.
519 (new_sese, free_sese, sese_build_livein_liveouts): Declared.
520 (struct scop): Added field liveout_renames.
521 (SCOP_LIVEOUT_RENAMES): New.
523 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
525 PR tree-optimization/38510
526 * graphite.c (recompute_all_dominators): Call mark_irreducible_loops.
527 (translate_clast): Call recompute_all_dominators before
529 (gloog): Call recompute_all_dominators before graphite_verify.
531 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
532 Jan Sjodin <jan.sjodin@amd.com>
534 PR tree-optimization/38500
535 * graphite.c (create_sese_edges): Call fix_loop_structure after
538 2009-01-05 Joel Sherrill <joel.sherrill@oarcorp.com>
540 * config.gcc: Add m32r*-*-rtems*.
541 * config/m32r/rtems.h: New file.
543 2009-01-05 Ben Elliston <bje@au.ibm.com>
545 * Makefile.in (.po.gmo): Use mkinstalldirs, not test -d || mkdir.
547 (po/gcc.pot): Likewise.
549 2009-01-04 David S. Miller <davem@davemloft.net>
551 * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
552 (STARTING_FRAME_OFFSET): Always set to zero.
554 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
556 * tree.def (LSHIFT_EXPR, RSHIFT_EXPR): Add commentary.
557 * tree-cfg.c (verify_gimple_assign_binary): Allow shifts of
558 fixed-point types, and vectors of the same.
560 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
562 * config/mips/sync.md (*mb_barrier): Rename to...
563 (*memory_barrier): ...this.
565 2009-01-04 Jonathan Wakely <jwakely.gcc@gmail.com>
567 * doc/extend.texi (Function Attributes): Move @cindex after @item
568 for 'artificial' and 'flatten'. Fix grammar for 'externally_visible'
569 and put in alphabetical order. Fix 'target' name and put in order.
570 * doc/invoke.texi (-Wstrict-null-sentinel, -fipa-matrix-reorg): Fix
573 2009-01-04 Uros Bizjak <ubizjak@gmail.com>
575 * config/s390/s390.md (UNSPEC_MB): Rename from UNSPECV_MB.
576 (memory_barrier): Expand as unspec instead of unspec_volatile.
577 Remove mem:BLK from insn operands. Use Pmode scratch register.
578 (*memory_barrier): Define as unspec instead of unspec_volatile.
579 Use (match_dup 0) as input operand.
581 * config/sparc/sparc.md (UNSPEC_MEMBAR): Rename from UNSPECV_MEMBAR.
582 * config/sparc/sync.md (memory_barrier): Expand as unspec instead of
583 unspec_volatile. Remove mem:BLK from insn operands. Use Pmode
584 scratch register. Remove operand 1.
585 (*stbar): Define as unspec instead of unspec_volatile.
586 Use (match_dup 0) as input operand, remove (const_int 8).
587 (*membar): Define as unspec instead of unspec_volatile.
588 Use (match_dup 0) as input operand, remove input operand 2.
590 * config/xtensa/xtensa.md (UNSPEC_MEMW): Rename from UNSPECV_MEMW.
591 (memory_barrier): Expand as unspec instead of unspec_volatile.
592 Remove mem:BLK from insn operands. Use Pmode scratch register.
593 (*memory_barrier): Define as unspec instead of unspec_volatile.
594 Use (match_dup 0) as input operand.
596 * config/ia64/sync.md (memory_barrier): Redefine as expander pattern.
597 Remove mem:BLK from insn operands. Use Pmode scratch register.
598 Set volatile flag on operand 0.
599 (*memory_barrier): New insn pattern.
601 * config/rs6000/sync.md (memory_barrier): Remove mem:BLK from
603 (*memory_barrier): Use (match_dup 0) as input operand.
605 * config/mips/sync.md (memory_barrier): Redefine as expander pattern.
606 Remove mem:BLK from insn operands. Use Pmode scratch register.
607 Set volatile flag on operand 0.
608 (*mb_internal): New insn pattern.
610 * config/alpha/sync.md (*memory_barrier): Rename from *mb_internal.
612 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
615 * function.c (struct temp_slot): Move to the section of the file
616 that deals with temp slots. Remove field 'address'.
617 (temp_slot_address_table): New hash table of address -> temp slot.
618 (struct temp_slot_address_entry): New struct, items for the table.
619 (temp_slot_address_compute_hash, temp_slot_address_hash,
620 temp_slot_address_eq, insert_temp_slot_address): Support functions
622 (find_temp_slot_from_address): Rewrite to use the new hash table.
623 (remove_unused_temp_slot_addresses): Remove addresses of temp
624 slots that have been made available.
625 (remove_unused_temp_slot_addresses_1): Call-back for htab_traverse,
626 worker function for remove_unused_temp_slot_addresses.
627 (assign_stack_temp_for_type): Don't clear the temp slot address list.
628 Add the temp slot address to the address -> temp slot map.
629 (update_temp_slot_address): Update via insert_temp_slot_address.
630 (free_temp_slots): Call remove_unused_temp_slot_addresses.
631 (pop_temp_slots): Likewise.
632 (init_temp_slots): Allocate the address -> temp slot map, or empty
633 the map if it is already allocated.
634 (prepare_function_start): Initialize temp slot processing.
636 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
639 * cfgexpand.c (estimate_stack_frame_size): Simplify the estimate:
640 Calculate the size of all stack vars assuming no packing of stack
641 vars will happen, replacing a quadratic algorithm with a linear one.
643 2009-01-03 Jakub Jelinek <jakub@redhat.com>
646 * expmed.c (store_bit_field_1): Don't modify op0 if movstrict insn
649 2009-01-03 Diego Novillo <dnovillo@google.com>
651 * doc/contrib.texi: Update contributions.
653 2009-01-03 Jakub Jelinek <jakub@redhat.com>
656 * builtins.c (fold_builtin_memory_op): Give up if either operand
657 is volatile. Set srctype or desttype to non-qualified version
661 * builtins.c (fold_builtin_expect): Only check DECL_WEAK for VAR_DECLs
664 2009-01-02 Kenneth Zadeck <zadeck@naturalbridge.com>
666 PR rtl-optimization/35805
667 * df-problems.c (df_lr_finalize): Add recursive call to resolve lr
668 problem if fast dce is able to remove any instructions.
669 * dce.c (dce_process_block): Fix dump message.
671 2009-01-02 Mark Mitchell <mark@codesourcery.com>
674 * tree-ssa-pre.c (compute_antic): Correct loop bounds.
676 2009-01-02 Jakub Jelinek <jakub@redhat.com>
679 * tree-flow.h (op_code_prio, op_prio): New prototypes.
680 * tree-pretty-print.c (op_code_prio): New function.
681 (op_prio): No longer static. Use op_code_prio.
682 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs):
683 Use op_prio and op_code_prio to determine if () should be
684 printed around operand(s) or not.
686 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs,
687 dump_gimple_call, dump_gimple_switch, dump_gimple_cond,
688 dump_gimple_label, dump_gimple_try, dump_symbols, dump_gimple_phi,
689 dump_gimple_mem_ops, dump_bb_header, dump_bb_end, pp_cfg_jump): Use
690 pp_character instead of pp_string for single letter printing.
692 2009-01-02 Richard Sandiford <rdsandiford@googlemail.com>
694 * doc/extend.texi: Fix '#pragma GCC option' typo.
696 2009-01-02 Richard Guenther <rguenther@suse.de>
698 * doc/install.texi (--enable-checking): Mention different
700 (--enable-stage1-checking): Document.
702 2009-01-01 Andrew Pinski <pinskia@gmail.com>
705 * tree-cfg.c (verify_expr): Add INDIRECT_REF case. Change MODIFY_EXPR
708 2009-01-02 Ben Elliston <bje@au.ibm.com>
710 * config/fp-bit.h (pack_d): Constify argument.
711 * config/fp-bit.c (makenan): Constify return type. Remove casts.
712 (isnan): Constify argument.
716 (_fpadd_parts): Constify return type.
717 (_fpmul_parts): Likewise.
718 (_fpdiv_parts): Likewise.
720 2009-01-01 Jakub Jelinek <jakub@redhat.com>
723 * c-typeck.c (add_pending_init): Add IMPLICIT argument. Only
724 warn about overwriting initializer with side-effects or
725 -Woverride-init if !IMPLICIT.
726 (output_init_element): Likewise. Pass IMPLICIT down to
728 (process_init_element): Add IMPLICIT argument. Pass it down
729 to output_init_element.
730 (push_init_element, pop_init_level, set_designator): Adjust
731 process_init_element callers.
732 (set_nonincremental_init, set_nonincremental_init_from_string):
733 Adjust add_pending_init callers.
734 (output_pending_init_elements): Adjust output_init_element callers.
735 * c-tree.h (process_init_element): Adjust prototype.
736 * c-parser.c (c_parser_initelt, c_parser_initval): Adjust
737 process_init_element callers.