1 2009-01-09 Jakub Jelinek <jakub@redhat.com>
4 * dojump.c (do_jump_by_parts_zero_rtx): Use mode instead of
5 GET_MODE (op0) in operand_subword_force calls.
8 * fold-const.c (fold_unary): For COMPOUND_EXPR and COND_EXPR,
9 fold_convert arg0 operands to TREE_TYPE (op0) first.
11 2009-01-08 Vladimir Makarov <vmakarov@redhat.com>
13 * params.def (ira-max-conflict-table-size): Decrease default value
16 2009-01-08 Jakub Jelinek <jakub@redhat.com>
18 PR tree-optimization/37031
19 * lambda-code.c (lambda_collect_parameters): Call pointer_set_destroy
21 (build_access_matrix): Reserve correct size for AM_MATRIX vector,
22 allocate it using gc instead of heap, use VEC_quick_push instead of
24 * graphite.c (build_access_matrix): Allocate AM_MATRIX vector using gc
25 instead of heap, use VEC_quick_push instead of VEC_safe_push.
26 * tree-data-ref.h (struct access_matrix): Change matrix to gc
27 allocated vector from heap allocated.
28 * lambda.h: Add DEF_VEC_ALLOC_P for gc allocated lambda_vector.
29 * tree-loop-linear.c (linear_transform_loops): Allocate nest
30 vector only after perfect_loop_nest_depth call.
32 2009-01-08 Sebastian Pop <sebastian.pop@amd.com>
33 Jan Sjodin <jan.sjodin@amd.com>
35 PR tree-optimization/38559
36 * graphite.c (debug_value, copy_constraint,
37 swap_constraint_variables, scale_constraint_variable, ): New.
38 (get_lower_bound, get_upper_bound): Removed.
39 (graphite_trans_bb_strip_mine): Clean up this code that works
40 only for constant number of iterations. Fully copy upper and
41 lower bound constraints, not only the constant part of them.
42 * graphite.h (debug_value): Declared.
44 2009-01-08 Ira Rosen <irar@il.ibm.com>
46 PR tree-optimization/37194
47 * tree-vect-transform.c (vect_estimate_min_profitable_iters):
48 Don't add the cost of cost model guard in prologue to scalar
49 outside cost in case of known number of iterations.
51 2009-01-07 Nathan Froyd <froydnj@codesourcery.com>
52 Alan Modra <amodra@bigpond.net.au>
54 * config/rs6000/rs6000.c (rs6000_legitimize_address): Check for
55 non-word-aligned REG+CONST addressing.
57 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
60 * config/alpha/alpha.c (alpha_end_function): For TARGET_ABI_OSF, call
61 free_after_compilation when outputting a thunk.
62 (alpha_output_mi_thunk_osf): Assert that we are processing a thunk.
63 Do not call free_after_compilation here.
65 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
67 * config/i386/i386.c (ix86_target_string): Use ARRAY_SIZE.
68 (ix86_valid_target_attribute_inner_p): Ditto.
70 2009-01-07 Jan Sjodin <jan.sjodin@amd.com>
72 PR tree-optimization/38492
73 PR tree-optimization/38498
74 * tree-check.c (operator_is_linear, scev_is_linear_expression): New.
75 * tree-chrec.h (scev_is_linear_expression): Declared.
76 * graphite.c (graphite_cannot_represent_loop_niter): New.
77 (scopdet_basic_block_info): Call graphite_cannot_represent_loop_niter.
78 (graphite_loop_normal_form): Use gcc_assert.
79 (scan_tree_for_params): Use CASE_CONVERT.
80 (phi_node_is_iv, bb_contains_non_iv_scalar_phi_nodes): New.
81 (build_scop_conditions_1): Call bb_contains_non_iv_scalar_phi_nodes.
82 Use gcc_assert. Discard scops that contain unhandled cases.
83 (build_scop_conditions): Return a boolean status for unhandled cases.
84 (strip_mine_profitable_p): Print the loop number, not its depth.
85 (is_interchange_valid): Pass the depth of the loop nest, don't
87 (graphite_trans_bb_block): Same.
88 (graphite_trans_bb_block): Print tentative of loop blocking.
89 (graphite_trans_scop_block): Do not print that the loop has been
91 (graphite_transform_loops): Do not handle scops that contain condition
94 2009-01-07 H.J. Lu <hongjiu.lu@intel.com>
96 AVX Programming Reference (December, 2008)
97 * config/i386/avxintrin.h (_mm256_stream_si256): New.
98 (_mm256_stream_pd): Likewise.
99 (_mm256_stream_ps): Likewise.
101 * config/i386/i386.c (ix86_builtins): Add IX86_BUILTIN_MOVNTDQ256,
102 IX86_BUILTIN_MOVNTPD256 and IX86_BUILTIN_MOVNTPS256.
103 (ix86_special_builtin_type): Add VOID_FTYPE_PV4DI_V4DI.
104 (bdesc_special_args): Add __builtin_ia32_movntdq256,
105 __builtin_ia32_movntpd256 and __builtin_ia32_movntps256.
106 (ix86_init_mmx_sse_builtins): Handle VOID_FTYPE_PV4DI_V4DI.
107 (ix86_expand_special_args_builtin): Likewise.
109 * config/i386/sse.md (AVXMODEDI): New.
110 (avx_movnt<mode>): Likewise.
111 (avx_movnt<mode>): Likewise.
112 (<sse>_movnt<mode>): Remove AVX support.
113 (sse2_movntv2di): Likewise.
115 2009-01-07 Richard Guenther <rguenther@suse.de>
118 * fold-const.c (extract_muldiv): Remove obsolete comment.
119 (fold_plusminus_mult_expr): Undo MINUS_EXPR
120 to PLUS_EXPR canonicalization for the canonicalization.
122 2009-01-07 Gerald Pfeifer <gerald@pfeifer.com>
124 * doc/install.texi (alpha*-dec-osf*): Remove note on 32-bit
125 hosted cross-compilers generating less efficient code.
127 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
129 * function.h (rtl_data): Add a dbr_scheduled_p field.
130 * reorg.c (dbr_schedule): Set it.
131 (gate_handle_delay_slots): Check it.
132 * config/mips/mips.c (mips_base_delayed_branch): Delete.
133 (mips_reorg): Check flag_delayed_branch instead of
134 mips_base_delayed_branch.
135 (mips_override_options): Don't set mips_base_delayed_branch
136 or flag_delayed_branch.
138 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
140 PR rtl-optimization/38426.
141 * ira.c (ira): Set current_function_is_leaf earlier.
143 2009-01-06 Jakub Jelinek <jakub@redhat.com>
145 PR rtl-optimization/38722
146 * combine.c (try_combine): Don't modify PATTERN (i3) and notes
147 too early, only set a flag and modify after last possible
150 2009-01-06 Janis Johnson <janis187@us.ibm.com>
153 * ginclude/float.h: Rename DECnn_DEN to DECnn_SUBNORMAL_MIN.
154 * real.c (decimal_single_format): Correct values of emin and emax.
155 (decimal_double_format): Ditto.
156 (decimal_quad_format): Ditto.
157 * c-cppbuiltin.c (builtin_define_decimal_float_constants): Adjust
158 computation of DECnn_MIN and DECnn_MAX for corrected values of
159 emin and emax. Define __DECnn_SUBNORMAL_MIN__ instead of
160 __DECnn_MIN__, and adjust its computation for the corrected value
163 2009-01-06 Jan Hubicka <jh@suse.cz>
166 * i386.c (ix86_expand_call): Use ARRAY_SIZE.
168 2009-01-06 Gerald Pfeifer <gerald@pfeifer.com>
170 * doc/contrib.texi (Contributors): Slightly adjust the end note.
171 Add Robert Clark to the list of testers.
173 2009-01-06 Jan Hubicka <jh@suse.cz>
174 Kai Tietz <kai.tietz@onevision.com>
176 * i386.md (*msabi_syvabi): Add SSE regs clobbers.
177 * i386.c (ix86_expand_call): Add clobbers.
179 2009-01-06 Jan Hubicka <jh@suse.cz>
180 Kai Tietz <kai.tietz@onevision.com>
182 * i386.h (CONDITIONAL_CALL_USAGE): SSE regs are not used for w64 ABI.
183 * i386.c (struct ix86_frame): Add padding0 and nsseregs.
184 (ix86_nsaved_regs): Count only general purpose regs.
185 (ix86_nsaved_sseregs): New.
186 (ix86_compute_frame_layout): Update nsseregs; set preferred alignment
187 to 16 for w64; compute padding and size of sse reg save area.
188 (ix86_emit_save_regs, ix86_emit_save_regs_using_mov): Save only
189 general purpose regs.
190 (ix86_emit_save_sse_regs_using_mov): New.
191 (ix86_expand_prologue): Save SSE regs if needed.
192 (ix86_emit_restore_regs_using_mov): Use only general purpose regs.
193 (ix86_emit_restore_sse_regs_using_mov): New.
194 (ix86_expand_epilogue): Save SSE regs if needed.
196 2009-01-06 Jan Hubicka <jh@suse.cz>
197 Kai Tietz <kai.tietz@onevision.com>
199 * i386.h (ACCUMULATE_OUTGOING_ARGS): Enable for MSABI
200 * i386.c (init_cumulative_args): Disallow calls of MSABI functions
201 when accumulate outgoing args is off.
203 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
206 * ira-color.c (ira_reuse_stack_slot): Check ENABLE_IRA_CHECKING
207 before using pseudos_have_intersected_live_ranges_p.
209 * ira-int.h (ira_assert): Always define.
211 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
213 AVX Programming Reference (December, 2008)
214 * config/i386/avxintrin.h (_mm_permute2_pd): Removed.
215 (_mm256_permute2_pd): Likewise.
216 (_mm_permute2_ps): Likewise.
217 (_mm256_permute2_ps): Likewise.
218 * config/i386/i386.md (UNSPEC_VPERMIL2): Likewise.
219 * config/i386/sse.md (avx_vpermil2<mode>3): Likewise.
221 * config/i386/i386.c (ix86_builtins): Remove
222 IX86_BUILTIN_VPERMIL2PD, IX86_BUILTIN_VPERMIL2PS,
223 IX86_BUILTIN_VPERMIL2PD256 and IX86_BUILTIN_VPERMIL2PS256.
224 (ix86_builtin_type): Remove V8SF_FTYPE_V8SF_V8SF_V8SI_INT,
225 V4DF_FTYPE_V4DF_V4DF_V4DI_INT, V4SF_FTYPE_V4SF_V4SF_V4SI_INT
226 and V2DF_FTYPE_V2DF_V2DF_V2DI_INT.
227 (bdesc_args): Remove __builtin_ia32_vpermil2pd,
228 __builtin_ia32_vpermil2ps, __builtin_ia32_vpermil2pd256 and
229 __builtin_ia32_vpermil2ps256.
230 (ix86_init_mmx_sse_builtins): Updated.
231 (ix86_expand_args_builtin): Likewise.
233 2009-01-05 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
235 * pa.c (output_call): Relocate non-jump insns in the delay slot of
236 long absolute calls when generating PA 2.0 code.
238 2009-01-05 Vladimir Makarov <vmakarov@redhat.com>
240 PR rtl-optimization/38583
241 * params.h (IRA_MAX_CONFLICT_TABLE_SIZE): New macro.
243 * params.def (ira-max-conflict-table-size): New.
245 * doc/invoke.texi (ira-max-conflict-table-size): Decribe.
247 * ira.h (ira_conflicts_p): New external definition.
249 * ira-conflicts.c (build_conflict_bit_table): Do not build too big
250 table. Report this. Return result of building.
251 (ira_build_conflicts): Use ira_conflicts_p. Check result of
252 building conflict table.
254 * ira-color.c (fast_allocation): Use num instead of ira_allocnos_num.
255 (ira_color): Use ira_conflicts_p.
257 * global.c: Include ira.h.
258 (pseudo_for_reload_consideration_p, build_insn_chain): Use
261 * Makefile.in (global.o): Add ira.h.
263 * ira-build.c (mark_all_loops_for_removal,
264 propagate_some_info_from_allocno): New.
265 (remove_unnecessary_allocnos): Call
266 propagate_some_info_from_allocno.
267 (remove_low_level_allocnos): New.
268 (remove_unnecessary_regions): Add parameter. Call
269 mark_all_loops_for_removal and remove_low_level_allocnos. Pass
270 parameter to remove_unnecessary_regions.
271 (ira_build): Remove all regions but root if the conflict table was
272 not built. Update conflict hard regs for allocnos crossing calls.
274 * ira.c (ira_conflicts_p): New global.
275 (ira): Define and use ira_conflicts_p.
277 * reload1.c (compute_use_by_pseudos, reload, count_pseudo,
278 count_spilled_pseudo, find_reg, alter_reg, finish_spills,
279 emit_input_reload_insns, delete_output_reload): Use ira_conflicts_p.
281 2009-01-06 Ben Elliston <bje@au.ibm.com>
283 * gengtype-lex.l (YY_NO_INPUT): Define.
285 2009-01-05 Andrew Pinski <andrew_pinski@playstation.sony.com>
288 * c-common.c (handle_vector_size_attribute): Also reject
291 2009-01-05 Sebastian Pop <sebastian.pop@amd.com>
293 PR tree-optimization/38492
294 * graphite.c (rename_map_elt, debug_rename_elt,
295 debug_rename_map_1, debug_rename_map, new_rename_map_elt,
296 rename_map_elt_info, eq_rename_map_elts,
297 get_new_name_from_old_name, bb_in_sese_p): Moved around.
298 (sese_find_uses_to_rename_use): Renamed sese_build_livein_liveouts_use.
299 (sese_find_uses_to_rename_bb): Renamed sese_build_livein_liveouts_bb.
300 (sese_build_livein_liveouts): New.
301 (new_sese, free_sese): New.
302 (new_scop): Call new_sese.
303 (free_scop): Call free_sese.
304 (rename_variables_from_edge, rename_phis_end_scop): Removed.
305 (register_old_new_names): Renamed register_old_and_new_names.
306 (register_scop_liveout_renames, add_loop_exit_phis,
307 insert_loop_close_phis, struct igp,
308 default_liveout_before_guard, add_guard_exit_phis,
309 insert_guard_phis, copy_renames): New.
310 (translate_clast): Call insert_loop_close_phis and insert_guard_phis.
311 (sese_add_exit_phis_edge): Renamed scop_add_exit_phis_edge.
312 (rewrite_into_sese_closed_ssa): Renamed scop_insert_phis_for_liveouts.
313 (scop_adjust_phis_for_liveouts): New.
314 (gloog): Call scop_adjust_phis_for_liveouts.
316 * graphite.h (struct sese): Documented. Added fields liveout,
318 (SESE_LIVEOUT, SESE_LIVEIN, SESE_LIVEIN_VER, SESE_NUM_VER): New.
319 (new_sese, free_sese, sese_build_livein_liveouts): Declared.
320 (struct scop): Added field liveout_renames.
321 (SCOP_LIVEOUT_RENAMES): New.
323 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
325 PR tree-optimization/38510
326 * graphite.c (recompute_all_dominators): Call mark_irreducible_loops.
327 (translate_clast): Call recompute_all_dominators before
329 (gloog): Call recompute_all_dominators before graphite_verify.
331 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
332 Jan Sjodin <jan.sjodin@amd.com>
334 PR tree-optimization/38500
335 * graphite.c (create_sese_edges): Call fix_loop_structure after
338 2009-01-05 Joel Sherrill <joel.sherrill@oarcorp.com>
340 * config.gcc: Add m32r*-*-rtems*.
341 * config/m32r/rtems.h: New file.
343 2009-01-05 Ben Elliston <bje@au.ibm.com>
345 * Makefile.in (.po.gmo): Use mkinstalldirs, not test -d || mkdir.
347 (po/gcc.pot): Likewise.
349 2009-01-04 David S. Miller <davem@davemloft.net>
351 * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
352 (STARTING_FRAME_OFFSET): Always set to zero.
354 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
356 * tree.def (LSHIFT_EXPR, RSHIFT_EXPR): Add commentary.
357 * tree-cfg.c (verify_gimple_assign_binary): Allow shifts of
358 fixed-point types, and vectors of the same.
360 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
362 * config/mips/sync.md (*mb_barrier): Rename to...
363 (*memory_barrier): ...this.
365 2009-01-04 Jonathan Wakely <jwakely.gcc@gmail.com>
367 * doc/extend.texi (Function Attributes): Move @cindex after @item
368 for 'artificial' and 'flatten'. Fix grammar for 'externally_visible'
369 and put in alphabetical order. Fix 'target' name and put in order.
370 * doc/invoke.texi (-Wstrict-null-sentinel, -fipa-matrix-reorg): Fix
373 2009-01-04 Uros Bizjak <ubizjak@gmail.com>
375 * config/s390/s390.md (UNSPEC_MB): Rename from UNSPECV_MB.
376 (memory_barrier): Expand as unspec instead of unspec_volatile.
377 Remove mem:BLK from insn operands. Use Pmode scratch register.
378 (*memory_barrier): Define as unspec instead of unspec_volatile.
379 Use (match_dup 0) as input operand.
381 * config/sparc/sparc.md (UNSPEC_MEMBAR): Rename from UNSPECV_MEMBAR.
382 * config/sparc/sync.md (memory_barrier): Expand as unspec instead of
383 unspec_volatile. Remove mem:BLK from insn operands. Use Pmode
384 scratch register. Remove operand 1.
385 (*stbar): Define as unspec instead of unspec_volatile.
386 Use (match_dup 0) as input operand, remove (const_int 8).
387 (*membar): Define as unspec instead of unspec_volatile.
388 Use (match_dup 0) as input operand, remove input operand 2.
390 * config/xtensa/xtensa.md (UNSPEC_MEMW): Rename from UNSPECV_MEMW.
391 (memory_barrier): Expand as unspec instead of unspec_volatile.
392 Remove mem:BLK from insn operands. Use Pmode scratch register.
393 (*memory_barrier): Define as unspec instead of unspec_volatile.
394 Use (match_dup 0) as input operand.
396 * config/ia64/sync.md (memory_barrier): Redefine as expander pattern.
397 Remove mem:BLK from insn operands. Use Pmode scratch register.
398 Set volatile flag on operand 0.
399 (*memory_barrier): New insn pattern.
401 * config/rs6000/sync.md (memory_barrier): Remove mem:BLK from
403 (*memory_barrier): Use (match_dup 0) as input operand.
405 * config/mips/sync.md (memory_barrier): Redefine as expander pattern.
406 Remove mem:BLK from insn operands. Use Pmode scratch register.
407 Set volatile flag on operand 0.
408 (*mb_internal): New insn pattern.
410 * config/alpha/sync.md (*memory_barrier): Rename from *mb_internal.
412 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
415 * function.c (struct temp_slot): Move to the section of the file
416 that deals with temp slots. Remove field 'address'.
417 (temp_slot_address_table): New hash table of address -> temp slot.
418 (struct temp_slot_address_entry): New struct, items for the table.
419 (temp_slot_address_compute_hash, temp_slot_address_hash,
420 temp_slot_address_eq, insert_temp_slot_address): Support functions
422 (find_temp_slot_from_address): Rewrite to use the new hash table.
423 (remove_unused_temp_slot_addresses): Remove addresses of temp
424 slots that have been made available.
425 (remove_unused_temp_slot_addresses_1): Call-back for htab_traverse,
426 worker function for remove_unused_temp_slot_addresses.
427 (assign_stack_temp_for_type): Don't clear the temp slot address list.
428 Add the temp slot address to the address -> temp slot map.
429 (update_temp_slot_address): Update via insert_temp_slot_address.
430 (free_temp_slots): Call remove_unused_temp_slot_addresses.
431 (pop_temp_slots): Likewise.
432 (init_temp_slots): Allocate the address -> temp slot map, or empty
433 the map if it is already allocated.
434 (prepare_function_start): Initialize temp slot processing.
436 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
439 * cfgexpand.c (estimate_stack_frame_size): Simplify the estimate:
440 Calculate the size of all stack vars assuming no packing of stack
441 vars will happen, replacing a quadratic algorithm with a linear one.
443 2009-01-03 Jakub Jelinek <jakub@redhat.com>
446 * expmed.c (store_bit_field_1): Don't modify op0 if movstrict insn
449 2009-01-03 Diego Novillo <dnovillo@google.com>
451 * doc/contrib.texi: Update contributions.
453 2009-01-03 Jakub Jelinek <jakub@redhat.com>
456 * builtins.c (fold_builtin_memory_op): Give up if either operand
457 is volatile. Set srctype or desttype to non-qualified version
461 * builtins.c (fold_builtin_expect): Only check DECL_WEAK for VAR_DECLs
464 2009-01-02 Kenneth Zadeck <zadeck@naturalbridge.com>
466 PR rtl-optimization/35805
467 * df-problems.c (df_lr_finalize): Add recursive call to resolve lr
468 problem if fast dce is able to remove any instructions.
469 * dce.c (dce_process_block): Fix dump message.
471 2009-01-02 Mark Mitchell <mark@codesourcery.com>
474 * tree-ssa-pre.c (compute_antic): Correct loop bounds.
476 2009-01-02 Jakub Jelinek <jakub@redhat.com>
479 * tree-flow.h (op_code_prio, op_prio): New prototypes.
480 * tree-pretty-print.c (op_code_prio): New function.
481 (op_prio): No longer static. Use op_code_prio.
482 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs):
483 Use op_prio and op_code_prio to determine if () should be
484 printed around operand(s) or not.
486 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs,
487 dump_gimple_call, dump_gimple_switch, dump_gimple_cond,
488 dump_gimple_label, dump_gimple_try, dump_symbols, dump_gimple_phi,
489 dump_gimple_mem_ops, dump_bb_header, dump_bb_end, pp_cfg_jump): Use
490 pp_character instead of pp_string for single letter printing.
492 2009-01-02 Richard Sandiford <rdsandiford@googlemail.com>
494 * doc/extend.texi: Fix '#pragma GCC option' typo.
496 2009-01-02 Richard Guenther <rguenther@suse.de>
498 * doc/install.texi (--enable-checking): Mention different
500 (--enable-stage1-checking): Document.
502 2009-01-01 Andrew Pinski <pinskia@gmail.com>
505 * tree-cfg.c (verify_expr): Add INDIRECT_REF case. Change MODIFY_EXPR
508 2009-01-02 Ben Elliston <bje@au.ibm.com>
510 * config/fp-bit.h (pack_d): Constify argument.
511 * config/fp-bit.c (makenan): Constify return type. Remove casts.
512 (isnan): Constify argument.
516 (_fpadd_parts): Constify return type.
517 (_fpmul_parts): Likewise.
518 (_fpdiv_parts): Likewise.
520 2009-01-01 Jakub Jelinek <jakub@redhat.com>
523 * c-typeck.c (add_pending_init): Add IMPLICIT argument. Only
524 warn about overwriting initializer with side-effects or
525 -Woverride-init if !IMPLICIT.
526 (output_init_element): Likewise. Pass IMPLICIT down to
528 (process_init_element): Add IMPLICIT argument. Pass it down
529 to output_init_element.
530 (push_init_element, pop_init_level, set_designator): Adjust
531 process_init_element callers.
532 (set_nonincremental_init, set_nonincremental_init_from_string):
533 Adjust add_pending_init callers.
534 (output_pending_init_elements): Adjust output_init_element callers.
535 * c-tree.h (process_init_element): Adjust prototype.
536 * c-parser.c (c_parser_initelt, c_parser_initval): Adjust
537 process_init_element callers.