1 2009-01-09 Nick Clifton <nickc@redhat.com>
3 * config/sh/symbian.c: Replace uses of DECL_INLINE with
4 DECL_DECLARED_INLINE_P.
6 2009-01-09 Jakub Jelinek <jakub@redhat.com>
9 * dojump.c (do_jump_by_parts_zero_rtx): Use mode instead of
10 GET_MODE (op0) in operand_subword_force calls.
13 * fold-const.c (fold_unary): For COMPOUND_EXPR and COND_EXPR,
14 fold_convert arg0 operands to TREE_TYPE (op0) first.
16 2009-01-08 Vladimir Makarov <vmakarov@redhat.com>
18 * params.def (ira-max-conflict-table-size): Decrease default value
21 2009-01-08 Jakub Jelinek <jakub@redhat.com>
23 PR tree-optimization/37031
24 * lambda-code.c (lambda_collect_parameters): Call pointer_set_destroy
26 (build_access_matrix): Reserve correct size for AM_MATRIX vector,
27 allocate it using gc instead of heap, use VEC_quick_push instead of
29 * graphite.c (build_access_matrix): Allocate AM_MATRIX vector using gc
30 instead of heap, use VEC_quick_push instead of VEC_safe_push.
31 * tree-data-ref.h (struct access_matrix): Change matrix to gc
32 allocated vector from heap allocated.
33 * lambda.h: Add DEF_VEC_ALLOC_P for gc allocated lambda_vector.
34 * tree-loop-linear.c (linear_transform_loops): Allocate nest
35 vector only after perfect_loop_nest_depth call.
37 2009-01-08 Sebastian Pop <sebastian.pop@amd.com>
38 Jan Sjodin <jan.sjodin@amd.com>
40 PR tree-optimization/38559
41 * graphite.c (debug_value, copy_constraint,
42 swap_constraint_variables, scale_constraint_variable, ): New.
43 (get_lower_bound, get_upper_bound): Removed.
44 (graphite_trans_bb_strip_mine): Clean up this code that works
45 only for constant number of iterations. Fully copy upper and
46 lower bound constraints, not only the constant part of them.
47 * graphite.h (debug_value): Declared.
49 2009-01-08 Ira Rosen <irar@il.ibm.com>
51 PR tree-optimization/37194
52 * tree-vect-transform.c (vect_estimate_min_profitable_iters):
53 Don't add the cost of cost model guard in prologue to scalar
54 outside cost in case of known number of iterations.
56 2009-01-07 Nathan Froyd <froydnj@codesourcery.com>
57 Alan Modra <amodra@bigpond.net.au>
59 * config/rs6000/rs6000.c (rs6000_legitimize_address): Check for
60 non-word-aligned REG+CONST addressing.
62 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
65 * config/alpha/alpha.c (alpha_end_function): For TARGET_ABI_OSF, call
66 free_after_compilation when outputting a thunk.
67 (alpha_output_mi_thunk_osf): Assert that we are processing a thunk.
68 Do not call free_after_compilation here.
70 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
72 * config/i386/i386.c (ix86_target_string): Use ARRAY_SIZE.
73 (ix86_valid_target_attribute_inner_p): Ditto.
75 2009-01-07 Jan Sjodin <jan.sjodin@amd.com>
77 PR tree-optimization/38492
78 PR tree-optimization/38498
79 * tree-check.c (operator_is_linear, scev_is_linear_expression): New.
80 * tree-chrec.h (scev_is_linear_expression): Declared.
81 * graphite.c (graphite_cannot_represent_loop_niter): New.
82 (scopdet_basic_block_info): Call graphite_cannot_represent_loop_niter.
83 (graphite_loop_normal_form): Use gcc_assert.
84 (scan_tree_for_params): Use CASE_CONVERT.
85 (phi_node_is_iv, bb_contains_non_iv_scalar_phi_nodes): New.
86 (build_scop_conditions_1): Call bb_contains_non_iv_scalar_phi_nodes.
87 Use gcc_assert. Discard scops that contain unhandled cases.
88 (build_scop_conditions): Return a boolean status for unhandled cases.
89 (strip_mine_profitable_p): Print the loop number, not its depth.
90 (is_interchange_valid): Pass the depth of the loop nest, don't
92 (graphite_trans_bb_block): Same.
93 (graphite_trans_bb_block): Print tentative of loop blocking.
94 (graphite_trans_scop_block): Do not print that the loop has been
96 (graphite_transform_loops): Do not handle scops that contain condition
99 2009-01-07 H.J. Lu <hongjiu.lu@intel.com>
101 AVX Programming Reference (December, 2008)
102 * config/i386/avxintrin.h (_mm256_stream_si256): New.
103 (_mm256_stream_pd): Likewise.
104 (_mm256_stream_ps): Likewise.
106 * config/i386/i386.c (ix86_builtins): Add IX86_BUILTIN_MOVNTDQ256,
107 IX86_BUILTIN_MOVNTPD256 and IX86_BUILTIN_MOVNTPS256.
108 (ix86_special_builtin_type): Add VOID_FTYPE_PV4DI_V4DI.
109 (bdesc_special_args): Add __builtin_ia32_movntdq256,
110 __builtin_ia32_movntpd256 and __builtin_ia32_movntps256.
111 (ix86_init_mmx_sse_builtins): Handle VOID_FTYPE_PV4DI_V4DI.
112 (ix86_expand_special_args_builtin): Likewise.
114 * config/i386/sse.md (AVXMODEDI): New.
115 (avx_movnt<mode>): Likewise.
116 (avx_movnt<mode>): Likewise.
117 (<sse>_movnt<mode>): Remove AVX support.
118 (sse2_movntv2di): Likewise.
120 2009-01-07 Richard Guenther <rguenther@suse.de>
123 * fold-const.c (extract_muldiv): Remove obsolete comment.
124 (fold_plusminus_mult_expr): Undo MINUS_EXPR
125 to PLUS_EXPR canonicalization for the canonicalization.
127 2009-01-07 Gerald Pfeifer <gerald@pfeifer.com>
129 * doc/install.texi (alpha*-dec-osf*): Remove note on 32-bit
130 hosted cross-compilers generating less efficient code.
132 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
134 * function.h (rtl_data): Add a dbr_scheduled_p field.
135 * reorg.c (dbr_schedule): Set it.
136 (gate_handle_delay_slots): Check it.
137 * config/mips/mips.c (mips_base_delayed_branch): Delete.
138 (mips_reorg): Check flag_delayed_branch instead of
139 mips_base_delayed_branch.
140 (mips_override_options): Don't set mips_base_delayed_branch
141 or flag_delayed_branch.
143 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
145 PR rtl-optimization/38426.
146 * ira.c (ira): Set current_function_is_leaf earlier.
148 2009-01-06 Jakub Jelinek <jakub@redhat.com>
150 PR rtl-optimization/38722
151 * combine.c (try_combine): Don't modify PATTERN (i3) and notes
152 too early, only set a flag and modify after last possible
155 2009-01-06 Janis Johnson <janis187@us.ibm.com>
158 * ginclude/float.h: Rename DECnn_DEN to DECnn_SUBNORMAL_MIN.
159 * real.c (decimal_single_format): Correct values of emin and emax.
160 (decimal_double_format): Ditto.
161 (decimal_quad_format): Ditto.
162 * c-cppbuiltin.c (builtin_define_decimal_float_constants): Adjust
163 computation of DECnn_MIN and DECnn_MAX for corrected values of
164 emin and emax. Define __DECnn_SUBNORMAL_MIN__ instead of
165 __DECnn_MIN__, and adjust its computation for the corrected value
168 2009-01-06 Jan Hubicka <jh@suse.cz>
171 * i386.c (ix86_expand_call): Use ARRAY_SIZE.
173 2009-01-06 Gerald Pfeifer <gerald@pfeifer.com>
175 * doc/contrib.texi (Contributors): Slightly adjust the end note.
176 Add Robert Clark to the list of testers.
178 2009-01-06 Jan Hubicka <jh@suse.cz>
179 Kai Tietz <kai.tietz@onevision.com>
181 * i386.md (*msabi_syvabi): Add SSE regs clobbers.
182 * i386.c (ix86_expand_call): Add clobbers.
184 2009-01-06 Jan Hubicka <jh@suse.cz>
185 Kai Tietz <kai.tietz@onevision.com>
187 * i386.h (CONDITIONAL_CALL_USAGE): SSE regs are not used for w64 ABI.
188 * i386.c (struct ix86_frame): Add padding0 and nsseregs.
189 (ix86_nsaved_regs): Count only general purpose regs.
190 (ix86_nsaved_sseregs): New.
191 (ix86_compute_frame_layout): Update nsseregs; set preferred alignment
192 to 16 for w64; compute padding and size of sse reg save area.
193 (ix86_emit_save_regs, ix86_emit_save_regs_using_mov): Save only
194 general purpose regs.
195 (ix86_emit_save_sse_regs_using_mov): New.
196 (ix86_expand_prologue): Save SSE regs if needed.
197 (ix86_emit_restore_regs_using_mov): Use only general purpose regs.
198 (ix86_emit_restore_sse_regs_using_mov): New.
199 (ix86_expand_epilogue): Save SSE regs if needed.
201 2009-01-06 Jan Hubicka <jh@suse.cz>
202 Kai Tietz <kai.tietz@onevision.com>
204 * i386.h (ACCUMULATE_OUTGOING_ARGS): Enable for MSABI
205 * i386.c (init_cumulative_args): Disallow calls of MSABI functions
206 when accumulate outgoing args is off.
208 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
211 * ira-color.c (ira_reuse_stack_slot): Check ENABLE_IRA_CHECKING
212 before using pseudos_have_intersected_live_ranges_p.
214 * ira-int.h (ira_assert): Always define.
216 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
218 AVX Programming Reference (December, 2008)
219 * config/i386/avxintrin.h (_mm_permute2_pd): Removed.
220 (_mm256_permute2_pd): Likewise.
221 (_mm_permute2_ps): Likewise.
222 (_mm256_permute2_ps): Likewise.
223 * config/i386/i386.md (UNSPEC_VPERMIL2): Likewise.
224 * config/i386/sse.md (avx_vpermil2<mode>3): Likewise.
226 * config/i386/i386.c (ix86_builtins): Remove
227 IX86_BUILTIN_VPERMIL2PD, IX86_BUILTIN_VPERMIL2PS,
228 IX86_BUILTIN_VPERMIL2PD256 and IX86_BUILTIN_VPERMIL2PS256.
229 (ix86_builtin_type): Remove V8SF_FTYPE_V8SF_V8SF_V8SI_INT,
230 V4DF_FTYPE_V4DF_V4DF_V4DI_INT, V4SF_FTYPE_V4SF_V4SF_V4SI_INT
231 and V2DF_FTYPE_V2DF_V2DF_V2DI_INT.
232 (bdesc_args): Remove __builtin_ia32_vpermil2pd,
233 __builtin_ia32_vpermil2ps, __builtin_ia32_vpermil2pd256 and
234 __builtin_ia32_vpermil2ps256.
235 (ix86_init_mmx_sse_builtins): Updated.
236 (ix86_expand_args_builtin): Likewise.
238 2009-01-05 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
240 * pa.c (output_call): Relocate non-jump insns in the delay slot of
241 long absolute calls when generating PA 2.0 code.
243 2009-01-05 Vladimir Makarov <vmakarov@redhat.com>
245 PR rtl-optimization/38583
246 * params.h (IRA_MAX_CONFLICT_TABLE_SIZE): New macro.
248 * params.def (ira-max-conflict-table-size): New.
250 * doc/invoke.texi (ira-max-conflict-table-size): Decribe.
252 * ira.h (ira_conflicts_p): New external definition.
254 * ira-conflicts.c (build_conflict_bit_table): Do not build too big
255 table. Report this. Return result of building.
256 (ira_build_conflicts): Use ira_conflicts_p. Check result of
257 building conflict table.
259 * ira-color.c (fast_allocation): Use num instead of ira_allocnos_num.
260 (ira_color): Use ira_conflicts_p.
262 * global.c: Include ira.h.
263 (pseudo_for_reload_consideration_p, build_insn_chain): Use
266 * Makefile.in (global.o): Add ira.h.
268 * ira-build.c (mark_all_loops_for_removal,
269 propagate_some_info_from_allocno): New.
270 (remove_unnecessary_allocnos): Call
271 propagate_some_info_from_allocno.
272 (remove_low_level_allocnos): New.
273 (remove_unnecessary_regions): Add parameter. Call
274 mark_all_loops_for_removal and remove_low_level_allocnos. Pass
275 parameter to remove_unnecessary_regions.
276 (ira_build): Remove all regions but root if the conflict table was
277 not built. Update conflict hard regs for allocnos crossing calls.
279 * ira.c (ira_conflicts_p): New global.
280 (ira): Define and use ira_conflicts_p.
282 * reload1.c (compute_use_by_pseudos, reload, count_pseudo,
283 count_spilled_pseudo, find_reg, alter_reg, finish_spills,
284 emit_input_reload_insns, delete_output_reload): Use ira_conflicts_p.
286 2009-01-06 Ben Elliston <bje@au.ibm.com>
288 * gengtype-lex.l (YY_NO_INPUT): Define.
290 2009-01-05 Andrew Pinski <andrew_pinski@playstation.sony.com>
293 * c-common.c (handle_vector_size_attribute): Also reject
296 2009-01-05 Sebastian Pop <sebastian.pop@amd.com>
298 PR tree-optimization/38492
299 * graphite.c (rename_map_elt, debug_rename_elt,
300 debug_rename_map_1, debug_rename_map, new_rename_map_elt,
301 rename_map_elt_info, eq_rename_map_elts,
302 get_new_name_from_old_name, bb_in_sese_p): Moved around.
303 (sese_find_uses_to_rename_use): Renamed sese_build_livein_liveouts_use.
304 (sese_find_uses_to_rename_bb): Renamed sese_build_livein_liveouts_bb.
305 (sese_build_livein_liveouts): New.
306 (new_sese, free_sese): New.
307 (new_scop): Call new_sese.
308 (free_scop): Call free_sese.
309 (rename_variables_from_edge, rename_phis_end_scop): Removed.
310 (register_old_new_names): Renamed register_old_and_new_names.
311 (register_scop_liveout_renames, add_loop_exit_phis,
312 insert_loop_close_phis, struct igp,
313 default_liveout_before_guard, add_guard_exit_phis,
314 insert_guard_phis, copy_renames): New.
315 (translate_clast): Call insert_loop_close_phis and insert_guard_phis.
316 (sese_add_exit_phis_edge): Renamed scop_add_exit_phis_edge.
317 (rewrite_into_sese_closed_ssa): Renamed scop_insert_phis_for_liveouts.
318 (scop_adjust_phis_for_liveouts): New.
319 (gloog): Call scop_adjust_phis_for_liveouts.
321 * graphite.h (struct sese): Documented. Added fields liveout,
323 (SESE_LIVEOUT, SESE_LIVEIN, SESE_LIVEIN_VER, SESE_NUM_VER): New.
324 (new_sese, free_sese, sese_build_livein_liveouts): Declared.
325 (struct scop): Added field liveout_renames.
326 (SCOP_LIVEOUT_RENAMES): New.
328 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
330 PR tree-optimization/38510
331 * graphite.c (recompute_all_dominators): Call mark_irreducible_loops.
332 (translate_clast): Call recompute_all_dominators before
334 (gloog): Call recompute_all_dominators before graphite_verify.
336 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
337 Jan Sjodin <jan.sjodin@amd.com>
339 PR tree-optimization/38500
340 * graphite.c (create_sese_edges): Call fix_loop_structure after
343 2009-01-05 Joel Sherrill <joel.sherrill@oarcorp.com>
345 * config.gcc: Add m32r*-*-rtems*.
346 * config/m32r/rtems.h: New file.
348 2009-01-05 Ben Elliston <bje@au.ibm.com>
350 * Makefile.in (.po.gmo): Use mkinstalldirs, not test -d || mkdir.
352 (po/gcc.pot): Likewise.
354 2009-01-04 David S. Miller <davem@davemloft.net>
356 * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
357 (STARTING_FRAME_OFFSET): Always set to zero.
359 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
361 * tree.def (LSHIFT_EXPR, RSHIFT_EXPR): Add commentary.
362 * tree-cfg.c (verify_gimple_assign_binary): Allow shifts of
363 fixed-point types, and vectors of the same.
365 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
367 * config/mips/sync.md (*mb_barrier): Rename to...
368 (*memory_barrier): ...this.
370 2009-01-04 Jonathan Wakely <jwakely.gcc@gmail.com>
372 * doc/extend.texi (Function Attributes): Move @cindex after @item
373 for 'artificial' and 'flatten'. Fix grammar for 'externally_visible'
374 and put in alphabetical order. Fix 'target' name and put in order.
375 * doc/invoke.texi (-Wstrict-null-sentinel, -fipa-matrix-reorg): Fix
378 2009-01-04 Uros Bizjak <ubizjak@gmail.com>
380 * config/s390/s390.md (UNSPEC_MB): Rename from UNSPECV_MB.
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/sparc/sparc.md (UNSPEC_MEMBAR): Rename from UNSPECV_MEMBAR.
387 * config/sparc/sync.md (memory_barrier): Expand as unspec instead of
388 unspec_volatile. Remove mem:BLK from insn operands. Use Pmode
389 scratch register. Remove operand 1.
390 (*stbar): Define as unspec instead of unspec_volatile.
391 Use (match_dup 0) as input operand, remove (const_int 8).
392 (*membar): Define as unspec instead of unspec_volatile.
393 Use (match_dup 0) as input operand, remove input operand 2.
395 * config/xtensa/xtensa.md (UNSPEC_MEMW): Rename from UNSPECV_MEMW.
396 (memory_barrier): Expand as unspec instead of unspec_volatile.
397 Remove mem:BLK from insn operands. Use Pmode scratch register.
398 (*memory_barrier): Define as unspec instead of unspec_volatile.
399 Use (match_dup 0) as input operand.
401 * config/ia64/sync.md (memory_barrier): Redefine as expander pattern.
402 Remove mem:BLK from insn operands. Use Pmode scratch register.
403 Set volatile flag on operand 0.
404 (*memory_barrier): New insn pattern.
406 * config/rs6000/sync.md (memory_barrier): Remove mem:BLK from
408 (*memory_barrier): Use (match_dup 0) as input operand.
410 * config/mips/sync.md (memory_barrier): Redefine as expander pattern.
411 Remove mem:BLK from insn operands. Use Pmode scratch register.
412 Set volatile flag on operand 0.
413 (*mb_internal): New insn pattern.
415 * config/alpha/sync.md (*memory_barrier): Rename from *mb_internal.
417 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
420 * function.c (struct temp_slot): Move to the section of the file
421 that deals with temp slots. Remove field 'address'.
422 (temp_slot_address_table): New hash table of address -> temp slot.
423 (struct temp_slot_address_entry): New struct, items for the table.
424 (temp_slot_address_compute_hash, temp_slot_address_hash,
425 temp_slot_address_eq, insert_temp_slot_address): Support functions
427 (find_temp_slot_from_address): Rewrite to use the new hash table.
428 (remove_unused_temp_slot_addresses): Remove addresses of temp
429 slots that have been made available.
430 (remove_unused_temp_slot_addresses_1): Call-back for htab_traverse,
431 worker function for remove_unused_temp_slot_addresses.
432 (assign_stack_temp_for_type): Don't clear the temp slot address list.
433 Add the temp slot address to the address -> temp slot map.
434 (update_temp_slot_address): Update via insert_temp_slot_address.
435 (free_temp_slots): Call remove_unused_temp_slot_addresses.
436 (pop_temp_slots): Likewise.
437 (init_temp_slots): Allocate the address -> temp slot map, or empty
438 the map if it is already allocated.
439 (prepare_function_start): Initialize temp slot processing.
441 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
444 * cfgexpand.c (estimate_stack_frame_size): Simplify the estimate:
445 Calculate the size of all stack vars assuming no packing of stack
446 vars will happen, replacing a quadratic algorithm with a linear one.
448 2009-01-03 Jakub Jelinek <jakub@redhat.com>
451 * expmed.c (store_bit_field_1): Don't modify op0 if movstrict insn
454 2009-01-03 Diego Novillo <dnovillo@google.com>
456 * doc/contrib.texi: Update contributions.
458 2009-01-03 Jakub Jelinek <jakub@redhat.com>
461 * builtins.c (fold_builtin_memory_op): Give up if either operand
462 is volatile. Set srctype or desttype to non-qualified version
466 * builtins.c (fold_builtin_expect): Only check DECL_WEAK for VAR_DECLs
469 2009-01-02 Kenneth Zadeck <zadeck@naturalbridge.com>
471 PR rtl-optimization/35805
472 * df-problems.c (df_lr_finalize): Add recursive call to resolve lr
473 problem if fast dce is able to remove any instructions.
474 * dce.c (dce_process_block): Fix dump message.
476 2009-01-02 Mark Mitchell <mark@codesourcery.com>
479 * tree-ssa-pre.c (compute_antic): Correct loop bounds.
481 2009-01-02 Jakub Jelinek <jakub@redhat.com>
484 * tree-flow.h (op_code_prio, op_prio): New prototypes.
485 * tree-pretty-print.c (op_code_prio): New function.
486 (op_prio): No longer static. Use op_code_prio.
487 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs):
488 Use op_prio and op_code_prio to determine if () should be
489 printed around operand(s) or not.
491 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs,
492 dump_gimple_call, dump_gimple_switch, dump_gimple_cond,
493 dump_gimple_label, dump_gimple_try, dump_symbols, dump_gimple_phi,
494 dump_gimple_mem_ops, dump_bb_header, dump_bb_end, pp_cfg_jump): Use
495 pp_character instead of pp_string for single letter printing.
497 2009-01-02 Richard Sandiford <rdsandiford@googlemail.com>
499 * doc/extend.texi: Fix '#pragma GCC option' typo.
501 2009-01-02 Richard Guenther <rguenther@suse.de>
503 * doc/install.texi (--enable-checking): Mention different
505 (--enable-stage1-checking): Document.
507 2009-01-01 Andrew Pinski <pinskia@gmail.com>
510 * tree-cfg.c (verify_expr): Add INDIRECT_REF case. Change MODIFY_EXPR
513 2009-01-02 Ben Elliston <bje@au.ibm.com>
515 * config/fp-bit.h (pack_d): Constify argument.
516 * config/fp-bit.c (makenan): Constify return type. Remove casts.
517 (isnan): Constify argument.
521 (_fpadd_parts): Constify return type.
522 (_fpmul_parts): Likewise.
523 (_fpdiv_parts): Likewise.
525 2009-01-01 Jakub Jelinek <jakub@redhat.com>
528 * c-typeck.c (add_pending_init): Add IMPLICIT argument. Only
529 warn about overwriting initializer with side-effects or
530 -Woverride-init if !IMPLICIT.
531 (output_init_element): Likewise. Pass IMPLICIT down to
533 (process_init_element): Add IMPLICIT argument. Pass it down
534 to output_init_element.
535 (push_init_element, pop_init_level, set_designator): Adjust
536 process_init_element callers.
537 (set_nonincremental_init, set_nonincremental_init_from_string):
538 Adjust add_pending_init callers.
539 (output_pending_init_elements): Adjust output_init_element callers.
540 * c-tree.h (process_init_element): Adjust prototype.
541 * c-parser.c (c_parser_initelt, c_parser_initval): Adjust
542 process_init_element callers.