1 2009-01-08 Jakub Jelinek <jakub@redhat.com>
3 PR tree-optimization/37031
4 * lambda-code.c (lambda_collect_parameters): Call pointer_set_destroy
6 (build_access_matrix): Reserve correct size for AM_MATRIX vector,
7 allocate it using gc instead of heap, use VEC_quick_push instead of
9 * graphite.c (build_access_matrix): Allocate AM_MATRIX vector using gc
10 instead of heap, use VEC_quick_push instead of VEC_safe_push.
11 * tree-data-ref.h (struct access_matrix): Change matrix to gc
12 allocated vector from heap allocated.
13 * lambda.h: Add DEF_VEC_ALLOC_P for gc allocated lambda_vector.
14 * tree-loop-linear.c (linear_transform_loops): Allocate nest
15 vector only after perfect_loop_nest_depth call.
17 2009-01-08 Sebastian Pop <sebastian.pop@amd.com>
18 Jan Sjodin <jan.sjodin@amd.com>
20 PR tree-optimization/38559
21 * graphite.c (debug_value, copy_constraint,
22 swap_constraint_variables, scale_constraint_variable, ): New.
23 (get_lower_bound, get_upper_bound): Removed.
24 (graphite_trans_bb_strip_mine): Clean up this code that works
25 only for constant number of iterations. Fully copy upper and
26 lower bound constraints, not only the constant part of them.
27 * graphite.h (debug_value): Declared.
29 2009-01-08 Ira Rosen <irar@il.ibm.com>
31 PR tree-optimization/37194
32 * tree-vect-transform.c (vect_estimate_min_profitable_iters):
33 Don't add the cost of cost model guard in prologue to scalar
34 outside cost in case of known number of iterations.
36 2009-01-07 Nathan Froyd <froydnj@codesourcery.com>
37 Alan Modra <amodra@bigpond.net.au>
39 * config/rs6000/rs6000.c (rs6000_legitimize_address): Check for
40 non-word-aligned REG+CONST addressing.
42 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
45 * config/alpha/alpha.c (alpha_end_function): For TARGET_ABI_OSF, call
46 free_after_compilation when outputting a thunk.
47 (alpha_output_mi_thunk_osf): Assert that we are processing a thunk.
48 Do not call free_after_compilation here.
50 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
52 * config/i386/i386.c (ix86_target_string): Use ARRAY_SIZE.
53 (ix86_valid_target_attribute_inner_p): Ditto.
55 2009-01-07 Jan Sjodin <jan.sjodin@amd.com>
57 PR tree-optimization/38492
58 PR tree-optimization/38498
59 * tree-check.c (operator_is_linear, scev_is_linear_expression): New.
60 * tree-chrec.h (scev_is_linear_expression): Declared.
61 * graphite.c (graphite_cannot_represent_loop_niter): New.
62 (scopdet_basic_block_info): Call graphite_cannot_represent_loop_niter.
63 (graphite_loop_normal_form): Use gcc_assert.
64 (scan_tree_for_params): Use CASE_CONVERT.
65 (phi_node_is_iv, bb_contains_non_iv_scalar_phi_nodes): New.
66 (build_scop_conditions_1): Call bb_contains_non_iv_scalar_phi_nodes.
67 Use gcc_assert. Discard scops that contain unhandled cases.
68 (build_scop_conditions): Return a boolean status for unhandled cases.
69 (strip_mine_profitable_p): Print the loop number, not its depth.
70 (is_interchange_valid): Pass the depth of the loop nest, don't
72 (graphite_trans_bb_block): Same.
73 (graphite_trans_bb_block): Print tentative of loop blocking.
74 (graphite_trans_scop_block): Do not print that the loop has been
76 (graphite_transform_loops): Do not handle scops that contain condition
79 2009-01-07 H.J. Lu <hongjiu.lu@intel.com>
81 AVX Programming Reference (December, 2008)
82 * config/i386/avxintrin.h (_mm256_stream_si256): New.
83 (_mm256_stream_pd): Likewise.
84 (_mm256_stream_ps): Likewise.
86 * config/i386/i386.c (ix86_builtins): Add IX86_BUILTIN_MOVNTDQ256,
87 IX86_BUILTIN_MOVNTPD256 and IX86_BUILTIN_MOVNTPS256.
88 (ix86_special_builtin_type): Add VOID_FTYPE_PV4DI_V4DI.
89 (bdesc_special_args): Add __builtin_ia32_movntdq256,
90 __builtin_ia32_movntpd256 and __builtin_ia32_movntps256.
91 (ix86_init_mmx_sse_builtins): Handle VOID_FTYPE_PV4DI_V4DI.
92 (ix86_expand_special_args_builtin): Likewise.
94 * config/i386/sse.md (AVXMODEDI): New.
95 (avx_movnt<mode>): Likewise.
96 (avx_movnt<mode>): Likewise.
97 (<sse>_movnt<mode>): Remove AVX support.
98 (sse2_movntv2di): Likewise.
100 2009-01-07 Richard Guenther <rguenther@suse.de>
103 * fold-const.c (extract_muldiv): Remove obsolete comment.
104 (fold_plusminus_mult_expr): Undo MINUS_EXPR
105 to PLUS_EXPR canonicalization for the canonicalization.
107 2009-01-07 Gerald Pfeifer <gerald@pfeifer.com>
109 * doc/install.texi (alpha*-dec-osf*): Remove note on 32-bit
110 hosted cross-compilers generating less efficient code.
112 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
114 * function.h (rtl_data): Add a dbr_scheduled_p field.
115 * reorg.c (dbr_schedule): Set it.
116 (gate_handle_delay_slots): Check it.
117 * config/mips/mips.c (mips_base_delayed_branch): Delete.
118 (mips_reorg): Check flag_delayed_branch instead of
119 mips_base_delayed_branch.
120 (mips_override_options): Don't set mips_base_delayed_branch
121 or flag_delayed_branch.
123 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
125 PR rtl-optimization/38426.
126 * ira.c (ira): Set current_function_is_leaf earlier.
128 2009-01-06 Jakub Jelinek <jakub@redhat.com>
130 PR rtl-optimization/38722
131 * combine.c (try_combine): Don't modify PATTERN (i3) and notes
132 too early, only set a flag and modify after last possible
135 2009-01-06 Janis Johnson <janis187@us.ibm.com>
138 * ginclude/float.h: Rename DECnn_DEN to DECnn_SUBNORMAL_MIN.
139 * real.c (decimal_single_format): Correct values of emin and emax.
140 (decimal_double_format): Ditto.
141 (decimal_quad_format): Ditto.
142 * c-cppbuiltin.c (builtin_define_decimal_float_constants): Adjust
143 computation of DECnn_MIN and DECnn_MAX for corrected values of
144 emin and emax. Define __DECnn_SUBNORMAL_MIN__ instead of
145 __DECnn_MIN__, and adjust its computation for the corrected value
148 2009-01-06 Jan Hubicka <jh@suse.cz>
151 * i386.c (ix86_expand_call): Use ARRAY_SIZE.
153 2009-01-06 Gerald Pfeifer <gerald@pfeifer.com>
155 * doc/contrib.texi (Contributors): Slightly adjust the end note.
156 Add Robert Clark to the list of testers.
158 2009-01-06 Jan Hubicka <jh@suse.cz>
159 Kai Tietz <kai.tietz@onevision.com>
161 * i386.md (*msabi_syvabi): Add SSE regs clobbers.
162 * i386.c (ix86_expand_call): Add clobbers.
164 2009-01-06 Jan Hubicka <jh@suse.cz>
165 Kai Tietz <kai.tietz@onevision.com>
167 * i386.h (CONDITIONAL_CALL_USAGE): SSE regs are not used for w64 ABI.
168 * i386.c (struct ix86_frame): Add padding0 and nsseregs.
169 (ix86_nsaved_regs): Count only general purpose regs.
170 (ix86_nsaved_sseregs): New.
171 (ix86_compute_frame_layout): Update nsseregs; set preferred alignment
172 to 16 for w64; compute padding and size of sse reg save area.
173 (ix86_emit_save_regs, ix86_emit_save_regs_using_mov): Save only
174 general purpose regs.
175 (ix86_emit_save_sse_regs_using_mov): New.
176 (ix86_expand_prologue): Save SSE regs if needed.
177 (ix86_emit_restore_regs_using_mov): Use only general purpose regs.
178 (ix86_emit_restore_sse_regs_using_mov): New.
179 (ix86_expand_epilogue): Save SSE regs if needed.
181 2009-01-06 Jan Hubicka <jh@suse.cz>
182 Kai Tietz <kai.tietz@onevision.com>
184 * i386.h (ACCUMULATE_OUTGOING_ARGS): Enable for MSABI
185 * i386.c (init_cumulative_args): Disallow calls of MSABI functions
186 when accumulate outgoing args is off.
188 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
191 * ira-color.c (ira_reuse_stack_slot): Check ENABLE_IRA_CHECKING
192 before using pseudos_have_intersected_live_ranges_p.
194 * ira-int.h (ira_assert): Always define.
196 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
198 AVX Programming Reference (December, 2008)
199 * config/i386/avxintrin.h (_mm_permute2_pd): Removed.
200 (_mm256_permute2_pd): Likewise.
201 (_mm_permute2_ps): Likewise.
202 (_mm256_permute2_ps): Likewise.
203 * config/i386/i386.md (UNSPEC_VPERMIL2): Likewise.
204 * config/i386/sse.md (avx_vpermil2<mode>3): Likewise.
206 * config/i386/i386.c (ix86_builtins): Remove
207 IX86_BUILTIN_VPERMIL2PD, IX86_BUILTIN_VPERMIL2PS,
208 IX86_BUILTIN_VPERMIL2PD256 and IX86_BUILTIN_VPERMIL2PS256.
209 (ix86_builtin_type): Remove V8SF_FTYPE_V8SF_V8SF_V8SI_INT,
210 V4DF_FTYPE_V4DF_V4DF_V4DI_INT, V4SF_FTYPE_V4SF_V4SF_V4SI_INT
211 and V2DF_FTYPE_V2DF_V2DF_V2DI_INT.
212 (bdesc_args): Remove __builtin_ia32_vpermil2pd,
213 __builtin_ia32_vpermil2ps, __builtin_ia32_vpermil2pd256 and
214 __builtin_ia32_vpermil2ps256.
215 (ix86_init_mmx_sse_builtins): Updated.
216 (ix86_expand_args_builtin): Likewise.
218 2009-01-05 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
220 * pa.c (output_call): Relocate non-jump insns in the delay slot of
221 long absolute calls when generating PA 2.0 code.
223 2009-01-05 Vladimir Makarov <vmakarov@redhat.com>
225 PR rtl-optimization/38583
226 * params.h (IRA_MAX_CONFLICT_TABLE_SIZE): New macro.
228 * params.def (ira-max-conflict-table-size): New.
230 * doc/invoke.texi (ira-max-conflict-table-size): Decribe.
232 * ira.h (ira_conflicts_p): New external definition.
234 * ira-conflicts.c (build_conflict_bit_table): Do not build too big
235 table. Report this. Return result of building.
236 (ira_build_conflicts): Use ira_conflicts_p. Check result of
237 building conflict table.
239 * ira-color.c (fast_allocation): Use num instead of ira_allocnos_num.
240 (ira_color): Use ira_conflicts_p.
242 * global.c: Include ira.h.
243 (pseudo_for_reload_consideration_p, build_insn_chain): Use
246 * Makefile.in (global.o): Add ira.h.
248 * ira-build.c (mark_all_loops_for_removal,
249 propagate_some_info_from_allocno): New.
250 (remove_unnecessary_allocnos): Call
251 propagate_some_info_from_allocno.
252 (remove_low_level_allocnos): New.
253 (remove_unnecessary_regions): Add parameter. Call
254 mark_all_loops_for_removal and remove_low_level_allocnos. Pass
255 parameter to remove_unnecessary_regions.
256 (ira_build): Remove all regions but root if the conflict table was
257 not built. Update conflict hard regs for allocnos crossing calls.
259 * ira.c (ira_conflicts_p): New global.
260 (ira): Define and use ira_conflicts_p.
262 * reload1.c (compute_use_by_pseudos, reload, count_pseudo,
263 count_spilled_pseudo, find_reg, alter_reg, finish_spills,
264 emit_input_reload_insns, delete_output_reload): Use ira_conflicts_p.
266 2009-01-06 Ben Elliston <bje@au.ibm.com>
268 * gengtype-lex.l (YY_NO_INPUT): Define.
270 2009-01-05 Andrew Pinski <andrew_pinski@playstation.sony.com>
273 * c-common.c (handle_vector_size_attribute): Also reject
276 2009-01-05 Sebastian Pop <sebastian.pop@amd.com>
278 PR tree-optimization/38492
279 * graphite.c (rename_map_elt, debug_rename_elt,
280 debug_rename_map_1, debug_rename_map, new_rename_map_elt,
281 rename_map_elt_info, eq_rename_map_elts,
282 get_new_name_from_old_name, bb_in_sese_p): Moved around.
283 (sese_find_uses_to_rename_use): Renamed sese_build_livein_liveouts_use.
284 (sese_find_uses_to_rename_bb): Renamed sese_build_livein_liveouts_bb.
285 (sese_build_livein_liveouts): New.
286 (new_sese, free_sese): New.
287 (new_scop): Call new_sese.
288 (free_scop): Call free_sese.
289 (rename_variables_from_edge, rename_phis_end_scop): Removed.
290 (register_old_new_names): Renamed register_old_and_new_names.
291 (register_scop_liveout_renames, add_loop_exit_phis,
292 insert_loop_close_phis, struct igp,
293 default_liveout_before_guard, add_guard_exit_phis,
294 insert_guard_phis, copy_renames): New.
295 (translate_clast): Call insert_loop_close_phis and insert_guard_phis.
296 (sese_add_exit_phis_edge): Renamed scop_add_exit_phis_edge.
297 (rewrite_into_sese_closed_ssa): Renamed scop_insert_phis_for_liveouts.
298 (scop_adjust_phis_for_liveouts): New.
299 (gloog): Call scop_adjust_phis_for_liveouts.
301 * graphite.h (struct sese): Documented. Added fields liveout,
303 (SESE_LIVEOUT, SESE_LIVEIN, SESE_LIVEIN_VER, SESE_NUM_VER): New.
304 (new_sese, free_sese, sese_build_livein_liveouts): Declared.
305 (struct scop): Added field liveout_renames.
306 (SCOP_LIVEOUT_RENAMES): New.
308 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
310 PR tree-optimization/38510
311 * graphite.c (recompute_all_dominators): Call mark_irreducible_loops.
312 (translate_clast): Call recompute_all_dominators before
314 (gloog): Call recompute_all_dominators before graphite_verify.
316 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
317 Jan Sjodin <jan.sjodin@amd.com>
319 PR tree-optimization/38500
320 * graphite.c (create_sese_edges): Call fix_loop_structure after
323 2009-01-05 Joel Sherrill <joel.sherrill@oarcorp.com>
325 * config.gcc: Add m32r*-*-rtems*.
326 * config/m32r/rtems.h: New file.
328 2009-01-05 Ben Elliston <bje@au.ibm.com>
330 * Makefile.in (.po.gmo): Use mkinstalldirs, not test -d || mkdir.
332 (po/gcc.pot): Likewise.
334 2009-01-04 David S. Miller <davem@davemloft.net>
336 * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
337 (STARTING_FRAME_OFFSET): Always set to zero.
339 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
341 * tree.def (LSHIFT_EXPR, RSHIFT_EXPR): Add commentary.
342 * tree-cfg.c (verify_gimple_assign_binary): Allow shifts of
343 fixed-point types, and vectors of the same.
345 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
347 * config/mips/sync.md (*mb_barrier): Rename to...
348 (*memory_barrier): ...this.
350 2009-01-04 Jonathan Wakely <jwakely.gcc@gmail.com>
352 * doc/extend.texi (Function Attributes): Move @cindex after @item
353 for 'artificial' and 'flatten'. Fix grammar for 'externally_visible'
354 and put in alphabetical order. Fix 'target' name and put in order.
355 * doc/invoke.texi (-Wstrict-null-sentinel, -fipa-matrix-reorg): Fix
358 2009-01-04 Uros Bizjak <ubizjak@gmail.com>
360 * config/s390/s390.md (UNSPEC_MB): Rename from UNSPECV_MB.
361 (memory_barrier): Expand as unspec instead of unspec_volatile.
362 Remove mem:BLK from insn operands. Use Pmode scratch register.
363 (*memory_barrier): Define as unspec instead of unspec_volatile.
364 Use (match_dup 0) as input operand.
366 * config/sparc/sparc.md (UNSPEC_MEMBAR): Rename from UNSPECV_MEMBAR.
367 * config/sparc/sync.md (memory_barrier): Expand as unspec instead of
368 unspec_volatile. Remove mem:BLK from insn operands. Use Pmode
369 scratch register. Remove operand 1.
370 (*stbar): Define as unspec instead of unspec_volatile.
371 Use (match_dup 0) as input operand, remove (const_int 8).
372 (*membar): Define as unspec instead of unspec_volatile.
373 Use (match_dup 0) as input operand, remove input operand 2.
375 * config/xtensa/xtensa.md (UNSPEC_MEMW): Rename from UNSPECV_MEMW.
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/ia64/sync.md (memory_barrier): Redefine as expander pattern.
382 Remove mem:BLK from insn operands. Use Pmode scratch register.
383 Set volatile flag on operand 0.
384 (*memory_barrier): New insn pattern.
386 * config/rs6000/sync.md (memory_barrier): Remove mem:BLK from
388 (*memory_barrier): Use (match_dup 0) as input operand.
390 * config/mips/sync.md (memory_barrier): Redefine as expander pattern.
391 Remove mem:BLK from insn operands. Use Pmode scratch register.
392 Set volatile flag on operand 0.
393 (*mb_internal): New insn pattern.
395 * config/alpha/sync.md (*memory_barrier): Rename from *mb_internal.
397 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
400 * function.c (struct temp_slot): Move to the section of the file
401 that deals with temp slots. Remove field 'address'.
402 (temp_slot_address_table): New hash table of address -> temp slot.
403 (struct temp_slot_address_entry): New struct, items for the table.
404 (temp_slot_address_compute_hash, temp_slot_address_hash,
405 temp_slot_address_eq, insert_temp_slot_address): Support functions
407 (find_temp_slot_from_address): Rewrite to use the new hash table.
408 (remove_unused_temp_slot_addresses): Remove addresses of temp
409 slots that have been made available.
410 (remove_unused_temp_slot_addresses_1): Call-back for htab_traverse,
411 worker function for remove_unused_temp_slot_addresses.
412 (assign_stack_temp_for_type): Don't clear the temp slot address list.
413 Add the temp slot address to the address -> temp slot map.
414 (update_temp_slot_address): Update via insert_temp_slot_address.
415 (free_temp_slots): Call remove_unused_temp_slot_addresses.
416 (pop_temp_slots): Likewise.
417 (init_temp_slots): Allocate the address -> temp slot map, or empty
418 the map if it is already allocated.
419 (prepare_function_start): Initialize temp slot processing.
421 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
424 * cfgexpand.c (estimate_stack_frame_size): Simplify the estimate:
425 Calculate the size of all stack vars assuming no packing of stack
426 vars will happen, replacing a quadratic algorithm with a linear one.
428 2009-01-03 Jakub Jelinek <jakub@redhat.com>
431 * expmed.c (store_bit_field_1): Don't modify op0 if movstrict insn
434 2009-01-03 Diego Novillo <dnovillo@google.com>
436 * doc/contrib.texi: Update contributions.
438 2009-01-03 Jakub Jelinek <jakub@redhat.com>
441 * builtins.c (fold_builtin_memory_op): Give up if either operand
442 is volatile. Set srctype or desttype to non-qualified version
446 * builtins.c (fold_builtin_expect): Only check DECL_WEAK for VAR_DECLs
449 2009-01-02 Kenneth Zadeck <zadeck@naturalbridge.com>
451 PR rtl-optimization/35805
452 * df-problems.c (df_lr_finalize): Add recursive call to resolve lr
453 problem if fast dce is able to remove any instructions.
454 * dce.c (dce_process_block): Fix dump message.
456 2009-01-02 Mark Mitchell <mark@codesourcery.com>
459 * tree-ssa-pre.c (compute_antic): Correct loop bounds.
461 2009-01-02 Jakub Jelinek <jakub@redhat.com>
464 * tree-flow.h (op_code_prio, op_prio): New prototypes.
465 * tree-pretty-print.c (op_code_prio): New function.
466 (op_prio): No longer static. Use op_code_prio.
467 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs):
468 Use op_prio and op_code_prio to determine if () should be
469 printed around operand(s) or not.
471 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs,
472 dump_gimple_call, dump_gimple_switch, dump_gimple_cond,
473 dump_gimple_label, dump_gimple_try, dump_symbols, dump_gimple_phi,
474 dump_gimple_mem_ops, dump_bb_header, dump_bb_end, pp_cfg_jump): Use
475 pp_character instead of pp_string for single letter printing.
477 2009-01-02 Richard Sandiford <rdsandiford@googlemail.com>
479 * doc/extend.texi: Fix '#pragma GCC option' typo.
481 2009-01-02 Richard Guenther <rguenther@suse.de>
483 * doc/install.texi (--enable-checking): Mention different
485 (--enable-stage1-checking): Document.
487 2009-01-01 Andrew Pinski <pinskia@gmail.com>
490 * tree-cfg.c (verify_expr): Add INDIRECT_REF case. Change MODIFY_EXPR
493 2009-01-02 Ben Elliston <bje@au.ibm.com>
495 * config/fp-bit.h (pack_d): Constify argument.
496 * config/fp-bit.c (makenan): Constify return type. Remove casts.
497 (isnan): Constify argument.
501 (_fpadd_parts): Constify return type.
502 (_fpmul_parts): Likewise.
503 (_fpdiv_parts): Likewise.
505 2009-01-01 Jakub Jelinek <jakub@redhat.com>
508 * c-typeck.c (add_pending_init): Add IMPLICIT argument. Only
509 warn about overwriting initializer with side-effects or
510 -Woverride-init if !IMPLICIT.
511 (output_init_element): Likewise. Pass IMPLICIT down to
513 (process_init_element): Add IMPLICIT argument. Pass it down
514 to output_init_element.
515 (push_init_element, pop_init_level, set_designator): Adjust
516 process_init_element callers.
517 (set_nonincremental_init, set_nonincremental_init_from_string):
518 Adjust add_pending_init callers.
519 (output_pending_init_elements): Adjust output_init_element callers.
520 * c-tree.h (process_init_element): Adjust prototype.
521 * c-parser.c (c_parser_initelt, c_parser_initval): Adjust
522 process_init_element callers.