1 2009-01-08 Vladimir Makarov <vmakarov@redhat.com>
3 * params.def (ira-max-conflict-table-size): Decrease default value
6 2009-01-08 Jakub Jelinek <jakub@redhat.com>
8 PR tree-optimization/37031
9 * lambda-code.c (lambda_collect_parameters): Call pointer_set_destroy
11 (build_access_matrix): Reserve correct size for AM_MATRIX vector,
12 allocate it using gc instead of heap, use VEC_quick_push instead of
14 * graphite.c (build_access_matrix): Allocate AM_MATRIX vector using gc
15 instead of heap, use VEC_quick_push instead of VEC_safe_push.
16 * tree-data-ref.h (struct access_matrix): Change matrix to gc
17 allocated vector from heap allocated.
18 * lambda.h: Add DEF_VEC_ALLOC_P for gc allocated lambda_vector.
19 * tree-loop-linear.c (linear_transform_loops): Allocate nest
20 vector only after perfect_loop_nest_depth call.
22 2009-01-08 Sebastian Pop <sebastian.pop@amd.com>
23 Jan Sjodin <jan.sjodin@amd.com>
25 PR tree-optimization/38559
26 * graphite.c (debug_value, copy_constraint,
27 swap_constraint_variables, scale_constraint_variable, ): New.
28 (get_lower_bound, get_upper_bound): Removed.
29 (graphite_trans_bb_strip_mine): Clean up this code that works
30 only for constant number of iterations. Fully copy upper and
31 lower bound constraints, not only the constant part of them.
32 * graphite.h (debug_value): Declared.
34 2009-01-08 Ira Rosen <irar@il.ibm.com>
36 PR tree-optimization/37194
37 * tree-vect-transform.c (vect_estimate_min_profitable_iters):
38 Don't add the cost of cost model guard in prologue to scalar
39 outside cost in case of known number of iterations.
41 2009-01-07 Nathan Froyd <froydnj@codesourcery.com>
42 Alan Modra <amodra@bigpond.net.au>
44 * config/rs6000/rs6000.c (rs6000_legitimize_address): Check for
45 non-word-aligned REG+CONST addressing.
47 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
50 * config/alpha/alpha.c (alpha_end_function): For TARGET_ABI_OSF, call
51 free_after_compilation when outputting a thunk.
52 (alpha_output_mi_thunk_osf): Assert that we are processing a thunk.
53 Do not call free_after_compilation here.
55 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
57 * config/i386/i386.c (ix86_target_string): Use ARRAY_SIZE.
58 (ix86_valid_target_attribute_inner_p): Ditto.
60 2009-01-07 Jan Sjodin <jan.sjodin@amd.com>
62 PR tree-optimization/38492
63 PR tree-optimization/38498
64 * tree-check.c (operator_is_linear, scev_is_linear_expression): New.
65 * tree-chrec.h (scev_is_linear_expression): Declared.
66 * graphite.c (graphite_cannot_represent_loop_niter): New.
67 (scopdet_basic_block_info): Call graphite_cannot_represent_loop_niter.
68 (graphite_loop_normal_form): Use gcc_assert.
69 (scan_tree_for_params): Use CASE_CONVERT.
70 (phi_node_is_iv, bb_contains_non_iv_scalar_phi_nodes): New.
71 (build_scop_conditions_1): Call bb_contains_non_iv_scalar_phi_nodes.
72 Use gcc_assert. Discard scops that contain unhandled cases.
73 (build_scop_conditions): Return a boolean status for unhandled cases.
74 (strip_mine_profitable_p): Print the loop number, not its depth.
75 (is_interchange_valid): Pass the depth of the loop nest, don't
77 (graphite_trans_bb_block): Same.
78 (graphite_trans_bb_block): Print tentative of loop blocking.
79 (graphite_trans_scop_block): Do not print that the loop has been
81 (graphite_transform_loops): Do not handle scops that contain condition
84 2009-01-07 H.J. Lu <hongjiu.lu@intel.com>
86 AVX Programming Reference (December, 2008)
87 * config/i386/avxintrin.h (_mm256_stream_si256): New.
88 (_mm256_stream_pd): Likewise.
89 (_mm256_stream_ps): Likewise.
91 * config/i386/i386.c (ix86_builtins): Add IX86_BUILTIN_MOVNTDQ256,
92 IX86_BUILTIN_MOVNTPD256 and IX86_BUILTIN_MOVNTPS256.
93 (ix86_special_builtin_type): Add VOID_FTYPE_PV4DI_V4DI.
94 (bdesc_special_args): Add __builtin_ia32_movntdq256,
95 __builtin_ia32_movntpd256 and __builtin_ia32_movntps256.
96 (ix86_init_mmx_sse_builtins): Handle VOID_FTYPE_PV4DI_V4DI.
97 (ix86_expand_special_args_builtin): Likewise.
99 * config/i386/sse.md (AVXMODEDI): New.
100 (avx_movnt<mode>): Likewise.
101 (avx_movnt<mode>): Likewise.
102 (<sse>_movnt<mode>): Remove AVX support.
103 (sse2_movntv2di): Likewise.
105 2009-01-07 Richard Guenther <rguenther@suse.de>
108 * fold-const.c (extract_muldiv): Remove obsolete comment.
109 (fold_plusminus_mult_expr): Undo MINUS_EXPR
110 to PLUS_EXPR canonicalization for the canonicalization.
112 2009-01-07 Gerald Pfeifer <gerald@pfeifer.com>
114 * doc/install.texi (alpha*-dec-osf*): Remove note on 32-bit
115 hosted cross-compilers generating less efficient code.
117 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
119 * function.h (rtl_data): Add a dbr_scheduled_p field.
120 * reorg.c (dbr_schedule): Set it.
121 (gate_handle_delay_slots): Check it.
122 * config/mips/mips.c (mips_base_delayed_branch): Delete.
123 (mips_reorg): Check flag_delayed_branch instead of
124 mips_base_delayed_branch.
125 (mips_override_options): Don't set mips_base_delayed_branch
126 or flag_delayed_branch.
128 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
130 PR rtl-optimization/38426.
131 * ira.c (ira): Set current_function_is_leaf earlier.
133 2009-01-06 Jakub Jelinek <jakub@redhat.com>
135 PR rtl-optimization/38722
136 * combine.c (try_combine): Don't modify PATTERN (i3) and notes
137 too early, only set a flag and modify after last possible
140 2009-01-06 Janis Johnson <janis187@us.ibm.com>
143 * ginclude/float.h: Rename DECnn_DEN to DECnn_SUBNORMAL_MIN.
144 * real.c (decimal_single_format): Correct values of emin and emax.
145 (decimal_double_format): Ditto.
146 (decimal_quad_format): Ditto.
147 * c-cppbuiltin.c (builtin_define_decimal_float_constants): Adjust
148 computation of DECnn_MIN and DECnn_MAX for corrected values of
149 emin and emax. Define __DECnn_SUBNORMAL_MIN__ instead of
150 __DECnn_MIN__, and adjust its computation for the corrected value
153 2009-01-06 Jan Hubicka <jh@suse.cz>
156 * i386.c (ix86_expand_call): Use ARRAY_SIZE.
158 2009-01-06 Gerald Pfeifer <gerald@pfeifer.com>
160 * doc/contrib.texi (Contributors): Slightly adjust the end note.
161 Add Robert Clark to the list of testers.
163 2009-01-06 Jan Hubicka <jh@suse.cz>
164 Kai Tietz <kai.tietz@onevision.com>
166 * i386.md (*msabi_syvabi): Add SSE regs clobbers.
167 * i386.c (ix86_expand_call): Add clobbers.
169 2009-01-06 Jan Hubicka <jh@suse.cz>
170 Kai Tietz <kai.tietz@onevision.com>
172 * i386.h (CONDITIONAL_CALL_USAGE): SSE regs are not used for w64 ABI.
173 * i386.c (struct ix86_frame): Add padding0 and nsseregs.
174 (ix86_nsaved_regs): Count only general purpose regs.
175 (ix86_nsaved_sseregs): New.
176 (ix86_compute_frame_layout): Update nsseregs; set preferred alignment
177 to 16 for w64; compute padding and size of sse reg save area.
178 (ix86_emit_save_regs, ix86_emit_save_regs_using_mov): Save only
179 general purpose regs.
180 (ix86_emit_save_sse_regs_using_mov): New.
181 (ix86_expand_prologue): Save SSE regs if needed.
182 (ix86_emit_restore_regs_using_mov): Use only general purpose regs.
183 (ix86_emit_restore_sse_regs_using_mov): New.
184 (ix86_expand_epilogue): Save SSE regs if needed.
186 2009-01-06 Jan Hubicka <jh@suse.cz>
187 Kai Tietz <kai.tietz@onevision.com>
189 * i386.h (ACCUMULATE_OUTGOING_ARGS): Enable for MSABI
190 * i386.c (init_cumulative_args): Disallow calls of MSABI functions
191 when accumulate outgoing args is off.
193 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
196 * ira-color.c (ira_reuse_stack_slot): Check ENABLE_IRA_CHECKING
197 before using pseudos_have_intersected_live_ranges_p.
199 * ira-int.h (ira_assert): Always define.
201 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
203 AVX Programming Reference (December, 2008)
204 * config/i386/avxintrin.h (_mm_permute2_pd): Removed.
205 (_mm256_permute2_pd): Likewise.
206 (_mm_permute2_ps): Likewise.
207 (_mm256_permute2_ps): Likewise.
208 * config/i386/i386.md (UNSPEC_VPERMIL2): Likewise.
209 * config/i386/sse.md (avx_vpermil2<mode>3): Likewise.
211 * config/i386/i386.c (ix86_builtins): Remove
212 IX86_BUILTIN_VPERMIL2PD, IX86_BUILTIN_VPERMIL2PS,
213 IX86_BUILTIN_VPERMIL2PD256 and IX86_BUILTIN_VPERMIL2PS256.
214 (ix86_builtin_type): Remove V8SF_FTYPE_V8SF_V8SF_V8SI_INT,
215 V4DF_FTYPE_V4DF_V4DF_V4DI_INT, V4SF_FTYPE_V4SF_V4SF_V4SI_INT
216 and V2DF_FTYPE_V2DF_V2DF_V2DI_INT.
217 (bdesc_args): Remove __builtin_ia32_vpermil2pd,
218 __builtin_ia32_vpermil2ps, __builtin_ia32_vpermil2pd256 and
219 __builtin_ia32_vpermil2ps256.
220 (ix86_init_mmx_sse_builtins): Updated.
221 (ix86_expand_args_builtin): Likewise.
223 2009-01-05 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
225 * pa.c (output_call): Relocate non-jump insns in the delay slot of
226 long absolute calls when generating PA 2.0 code.
228 2009-01-05 Vladimir Makarov <vmakarov@redhat.com>
230 PR rtl-optimization/38583
231 * params.h (IRA_MAX_CONFLICT_TABLE_SIZE): New macro.
233 * params.def (ira-max-conflict-table-size): New.
235 * doc/invoke.texi (ira-max-conflict-table-size): Decribe.
237 * ira.h (ira_conflicts_p): New external definition.
239 * ira-conflicts.c (build_conflict_bit_table): Do not build too big
240 table. Report this. Return result of building.
241 (ira_build_conflicts): Use ira_conflicts_p. Check result of
242 building conflict table.
244 * ira-color.c (fast_allocation): Use num instead of ira_allocnos_num.
245 (ira_color): Use ira_conflicts_p.
247 * global.c: Include ira.h.
248 (pseudo_for_reload_consideration_p, build_insn_chain): Use
251 * Makefile.in (global.o): Add ira.h.
253 * ira-build.c (mark_all_loops_for_removal,
254 propagate_some_info_from_allocno): New.
255 (remove_unnecessary_allocnos): Call
256 propagate_some_info_from_allocno.
257 (remove_low_level_allocnos): New.
258 (remove_unnecessary_regions): Add parameter. Call
259 mark_all_loops_for_removal and remove_low_level_allocnos. Pass
260 parameter to remove_unnecessary_regions.
261 (ira_build): Remove all regions but root if the conflict table was
262 not built. Update conflict hard regs for allocnos crossing calls.
264 * ira.c (ira_conflicts_p): New global.
265 (ira): Define and use ira_conflicts_p.
267 * reload1.c (compute_use_by_pseudos, reload, count_pseudo,
268 count_spilled_pseudo, find_reg, alter_reg, finish_spills,
269 emit_input_reload_insns, delete_output_reload): Use ira_conflicts_p.
271 2009-01-06 Ben Elliston <bje@au.ibm.com>
273 * gengtype-lex.l (YY_NO_INPUT): Define.
275 2009-01-05 Andrew Pinski <andrew_pinski@playstation.sony.com>
278 * c-common.c (handle_vector_size_attribute): Also reject
281 2009-01-05 Sebastian Pop <sebastian.pop@amd.com>
283 PR tree-optimization/38492
284 * graphite.c (rename_map_elt, debug_rename_elt,
285 debug_rename_map_1, debug_rename_map, new_rename_map_elt,
286 rename_map_elt_info, eq_rename_map_elts,
287 get_new_name_from_old_name, bb_in_sese_p): Moved around.
288 (sese_find_uses_to_rename_use): Renamed sese_build_livein_liveouts_use.
289 (sese_find_uses_to_rename_bb): Renamed sese_build_livein_liveouts_bb.
290 (sese_build_livein_liveouts): New.
291 (new_sese, free_sese): New.
292 (new_scop): Call new_sese.
293 (free_scop): Call free_sese.
294 (rename_variables_from_edge, rename_phis_end_scop): Removed.
295 (register_old_new_names): Renamed register_old_and_new_names.
296 (register_scop_liveout_renames, add_loop_exit_phis,
297 insert_loop_close_phis, struct igp,
298 default_liveout_before_guard, add_guard_exit_phis,
299 insert_guard_phis, copy_renames): New.
300 (translate_clast): Call insert_loop_close_phis and insert_guard_phis.
301 (sese_add_exit_phis_edge): Renamed scop_add_exit_phis_edge.
302 (rewrite_into_sese_closed_ssa): Renamed scop_insert_phis_for_liveouts.
303 (scop_adjust_phis_for_liveouts): New.
304 (gloog): Call scop_adjust_phis_for_liveouts.
306 * graphite.h (struct sese): Documented. Added fields liveout,
308 (SESE_LIVEOUT, SESE_LIVEIN, SESE_LIVEIN_VER, SESE_NUM_VER): New.
309 (new_sese, free_sese, sese_build_livein_liveouts): Declared.
310 (struct scop): Added field liveout_renames.
311 (SCOP_LIVEOUT_RENAMES): New.
313 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
315 PR tree-optimization/38510
316 * graphite.c (recompute_all_dominators): Call mark_irreducible_loops.
317 (translate_clast): Call recompute_all_dominators before
319 (gloog): Call recompute_all_dominators before graphite_verify.
321 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
322 Jan Sjodin <jan.sjodin@amd.com>
324 PR tree-optimization/38500
325 * graphite.c (create_sese_edges): Call fix_loop_structure after
328 2009-01-05 Joel Sherrill <joel.sherrill@oarcorp.com>
330 * config.gcc: Add m32r*-*-rtems*.
331 * config/m32r/rtems.h: New file.
333 2009-01-05 Ben Elliston <bje@au.ibm.com>
335 * Makefile.in (.po.gmo): Use mkinstalldirs, not test -d || mkdir.
337 (po/gcc.pot): Likewise.
339 2009-01-04 David S. Miller <davem@davemloft.net>
341 * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
342 (STARTING_FRAME_OFFSET): Always set to zero.
344 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
346 * tree.def (LSHIFT_EXPR, RSHIFT_EXPR): Add commentary.
347 * tree-cfg.c (verify_gimple_assign_binary): Allow shifts of
348 fixed-point types, and vectors of the same.
350 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
352 * config/mips/sync.md (*mb_barrier): Rename to...
353 (*memory_barrier): ...this.
355 2009-01-04 Jonathan Wakely <jwakely.gcc@gmail.com>
357 * doc/extend.texi (Function Attributes): Move @cindex after @item
358 for 'artificial' and 'flatten'. Fix grammar for 'externally_visible'
359 and put in alphabetical order. Fix 'target' name and put in order.
360 * doc/invoke.texi (-Wstrict-null-sentinel, -fipa-matrix-reorg): Fix
363 2009-01-04 Uros Bizjak <ubizjak@gmail.com>
365 * config/s390/s390.md (UNSPEC_MB): Rename from UNSPECV_MB.
366 (memory_barrier): Expand as unspec instead of unspec_volatile.
367 Remove mem:BLK from insn operands. Use Pmode scratch register.
368 (*memory_barrier): Define as unspec instead of unspec_volatile.
369 Use (match_dup 0) as input operand.
371 * config/sparc/sparc.md (UNSPEC_MEMBAR): Rename from UNSPECV_MEMBAR.
372 * config/sparc/sync.md (memory_barrier): Expand as unspec instead of
373 unspec_volatile. Remove mem:BLK from insn operands. Use Pmode
374 scratch register. Remove operand 1.
375 (*stbar): Define as unspec instead of unspec_volatile.
376 Use (match_dup 0) as input operand, remove (const_int 8).
377 (*membar): Define as unspec instead of unspec_volatile.
378 Use (match_dup 0) as input operand, remove input operand 2.
380 * config/xtensa/xtensa.md (UNSPEC_MEMW): Rename from UNSPECV_MEMW.
381 (memory_barrier): Expand as unspec instead of unspec_volatile.
382 Remove mem:BLK from insn operands. Use Pmode scratch register.
383 (*memory_barrier): Define as unspec instead of unspec_volatile.
384 Use (match_dup 0) as input operand.
386 * config/ia64/sync.md (memory_barrier): Redefine as expander pattern.
387 Remove mem:BLK from insn operands. Use Pmode scratch register.
388 Set volatile flag on operand 0.
389 (*memory_barrier): New insn pattern.
391 * config/rs6000/sync.md (memory_barrier): Remove mem:BLK from
393 (*memory_barrier): Use (match_dup 0) as input operand.
395 * config/mips/sync.md (memory_barrier): Redefine as expander pattern.
396 Remove mem:BLK from insn operands. Use Pmode scratch register.
397 Set volatile flag on operand 0.
398 (*mb_internal): New insn pattern.
400 * config/alpha/sync.md (*memory_barrier): Rename from *mb_internal.
402 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
405 * function.c (struct temp_slot): Move to the section of the file
406 that deals with temp slots. Remove field 'address'.
407 (temp_slot_address_table): New hash table of address -> temp slot.
408 (struct temp_slot_address_entry): New struct, items for the table.
409 (temp_slot_address_compute_hash, temp_slot_address_hash,
410 temp_slot_address_eq, insert_temp_slot_address): Support functions
412 (find_temp_slot_from_address): Rewrite to use the new hash table.
413 (remove_unused_temp_slot_addresses): Remove addresses of temp
414 slots that have been made available.
415 (remove_unused_temp_slot_addresses_1): Call-back for htab_traverse,
416 worker function for remove_unused_temp_slot_addresses.
417 (assign_stack_temp_for_type): Don't clear the temp slot address list.
418 Add the temp slot address to the address -> temp slot map.
419 (update_temp_slot_address): Update via insert_temp_slot_address.
420 (free_temp_slots): Call remove_unused_temp_slot_addresses.
421 (pop_temp_slots): Likewise.
422 (init_temp_slots): Allocate the address -> temp slot map, or empty
423 the map if it is already allocated.
424 (prepare_function_start): Initialize temp slot processing.
426 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
429 * cfgexpand.c (estimate_stack_frame_size): Simplify the estimate:
430 Calculate the size of all stack vars assuming no packing of stack
431 vars will happen, replacing a quadratic algorithm with a linear one.
433 2009-01-03 Jakub Jelinek <jakub@redhat.com>
436 * expmed.c (store_bit_field_1): Don't modify op0 if movstrict insn
439 2009-01-03 Diego Novillo <dnovillo@google.com>
441 * doc/contrib.texi: Update contributions.
443 2009-01-03 Jakub Jelinek <jakub@redhat.com>
446 * builtins.c (fold_builtin_memory_op): Give up if either operand
447 is volatile. Set srctype or desttype to non-qualified version
451 * builtins.c (fold_builtin_expect): Only check DECL_WEAK for VAR_DECLs
454 2009-01-02 Kenneth Zadeck <zadeck@naturalbridge.com>
456 PR rtl-optimization/35805
457 * df-problems.c (df_lr_finalize): Add recursive call to resolve lr
458 problem if fast dce is able to remove any instructions.
459 * dce.c (dce_process_block): Fix dump message.
461 2009-01-02 Mark Mitchell <mark@codesourcery.com>
464 * tree-ssa-pre.c (compute_antic): Correct loop bounds.
466 2009-01-02 Jakub Jelinek <jakub@redhat.com>
469 * tree-flow.h (op_code_prio, op_prio): New prototypes.
470 * tree-pretty-print.c (op_code_prio): New function.
471 (op_prio): No longer static. Use op_code_prio.
472 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs):
473 Use op_prio and op_code_prio to determine if () should be
474 printed around operand(s) or not.
476 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs,
477 dump_gimple_call, dump_gimple_switch, dump_gimple_cond,
478 dump_gimple_label, dump_gimple_try, dump_symbols, dump_gimple_phi,
479 dump_gimple_mem_ops, dump_bb_header, dump_bb_end, pp_cfg_jump): Use
480 pp_character instead of pp_string for single letter printing.
482 2009-01-02 Richard Sandiford <rdsandiford@googlemail.com>
484 * doc/extend.texi: Fix '#pragma GCC option' typo.
486 2009-01-02 Richard Guenther <rguenther@suse.de>
488 * doc/install.texi (--enable-checking): Mention different
490 (--enable-stage1-checking): Document.
492 2009-01-01 Andrew Pinski <pinskia@gmail.com>
495 * tree-cfg.c (verify_expr): Add INDIRECT_REF case. Change MODIFY_EXPR
498 2009-01-02 Ben Elliston <bje@au.ibm.com>
500 * config/fp-bit.h (pack_d): Constify argument.
501 * config/fp-bit.c (makenan): Constify return type. Remove casts.
502 (isnan): Constify argument.
506 (_fpadd_parts): Constify return type.
507 (_fpmul_parts): Likewise.
508 (_fpdiv_parts): Likewise.
510 2009-01-01 Jakub Jelinek <jakub@redhat.com>
513 * c-typeck.c (add_pending_init): Add IMPLICIT argument. Only
514 warn about overwriting initializer with side-effects or
515 -Woverride-init if !IMPLICIT.
516 (output_init_element): Likewise. Pass IMPLICIT down to
518 (process_init_element): Add IMPLICIT argument. Pass it down
519 to output_init_element.
520 (push_init_element, pop_init_level, set_designator): Adjust
521 process_init_element callers.
522 (set_nonincremental_init, set_nonincremental_init_from_string):
523 Adjust add_pending_init callers.
524 (output_pending_init_elements): Adjust output_init_element callers.
525 * c-tree.h (process_init_element): Adjust prototype.
526 * c-parser.c (c_parser_initelt, c_parser_initval): Adjust
527 process_init_element callers.