1 2009-01-08 Ira Rosen <irar@il.ibm.com>
3 PR tree-optimization/37194
4 * tree-vect-transform.c (vect_estimate_min_profitable_iters):
5 Don't add the cost of cost model guard in prologue to scalar
6 outside cost in case of known number of iterations.
8 2009-01-07 Nathan Froyd <froydnj@codesourcery.com>
9 Alan Modra <amodra@bigpond.net.au>
11 * config/rs6000/rs6000.c (rs6000_legitimize_address): Check for
12 non-word-aligned REG+CONST addressing.
14 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
17 * config/alpha/alpha.c (alpha_end_function): For TARGET_ABI_OSF, call
18 free_after_compilation when outputting a thunk.
19 (alpha_output_mi_thunk_osf): Assert that we are processing a thunk.
20 Do not call free_after_compilation here.
22 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
24 * config/i386/i386.c (ix86_target_string): Use ARRAY_SIZE.
25 (ix86_valid_target_attribute_inner_p): Ditto.
27 2009-01-07 Jan Sjodin <jan.sjodin@amd.com>
29 PR tree-optimization/38492
30 PR tree-optimization/38498
31 * tree-check.c (operator_is_linear, scev_is_linear_expression): New.
32 * tree-chrec.h (scev_is_linear_expression): Declared.
33 * graphite.c (graphite_cannot_represent_loop_niter): New.
34 (scopdet_basic_block_info): Call graphite_cannot_represent_loop_niter.
35 (graphite_loop_normal_form): Use gcc_assert.
36 (scan_tree_for_params): Use CASE_CONVERT.
37 (phi_node_is_iv, bb_contains_non_iv_scalar_phi_nodes): New.
38 (build_scop_conditions_1): Call bb_contains_non_iv_scalar_phi_nodes.
39 Use gcc_assert. Discard scops that contain unhandled cases.
40 (build_scop_conditions): Return a boolean status for unhandled cases.
41 (strip_mine_profitable_p): Print the loop number, not its depth.
42 (is_interchange_valid): Pass the depth of the loop nest, don't
44 (graphite_trans_bb_block): Same.
45 (graphite_trans_bb_block): Print tentative of loop blocking.
46 (graphite_trans_scop_block): Do not print that the loop has been
48 (graphite_transform_loops): Do not handle scops that contain condition
51 2009-01-07 H.J. Lu <hongjiu.lu@intel.com>
53 AVX Programming Reference (December, 2008)
54 * config/i386/avxintrin.h (_mm256_stream_si256): New.
55 (_mm256_stream_pd): Likewise.
56 (_mm256_stream_ps): Likewise.
58 * config/i386/i386.c (ix86_builtins): Add IX86_BUILTIN_MOVNTDQ256,
59 IX86_BUILTIN_MOVNTPD256 and IX86_BUILTIN_MOVNTPS256.
60 (ix86_special_builtin_type): Add VOID_FTYPE_PV4DI_V4DI.
61 (bdesc_special_args): Add __builtin_ia32_movntdq256,
62 __builtin_ia32_movntpd256 and __builtin_ia32_movntps256.
63 (ix86_init_mmx_sse_builtins): Handle VOID_FTYPE_PV4DI_V4DI.
64 (ix86_expand_special_args_builtin): Likewise.
66 * config/i386/sse.md (AVXMODEDI): New.
67 (avx_movnt<mode>): Likewise.
68 (avx_movnt<mode>): Likewise.
69 (<sse>_movnt<mode>): Remove AVX support.
70 (sse2_movntv2di): Likewise.
72 2009-01-07 Richard Guenther <rguenther@suse.de>
75 * fold-const.c (extract_muldiv): Remove obsolete comment.
76 (fold_plusminus_mult_expr): Undo MINUS_EXPR
77 to PLUS_EXPR canonicalization for the canonicalization.
79 2009-01-07 Gerald Pfeifer <gerald@pfeifer.com>
81 * doc/install.texi (alpha*-dec-osf*): Remove note on 32-bit
82 hosted cross-compilers generating less efficient code.
84 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
86 * function.h (rtl_data): Add a dbr_scheduled_p field.
87 * reorg.c (dbr_schedule): Set it.
88 (gate_handle_delay_slots): Check it.
89 * config/mips/mips.c (mips_base_delayed_branch): Delete.
90 (mips_reorg): Check flag_delayed_branch instead of
91 mips_base_delayed_branch.
92 (mips_override_options): Don't set mips_base_delayed_branch
93 or flag_delayed_branch.
95 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
97 PR rtl-optimization/38426.
98 * ira.c (ira): Set current_function_is_leaf earlier.
100 2009-01-06 Jakub Jelinek <jakub@redhat.com>
102 PR rtl-optimization/38722
103 * combine.c (try_combine): Don't modify PATTERN (i3) and notes
104 too early, only set a flag and modify after last possible
107 2009-01-06 Janis Johnson <janis187@us.ibm.com>
110 * ginclude/float.h: Rename DECnn_DEN to DECnn_SUBNORMAL_MIN.
111 * real.c (decimal_single_format): Correct values of emin and emax.
112 (decimal_double_format): Ditto.
113 (decimal_quad_format): Ditto.
114 * c-cppbuiltin.c (builtin_define_decimal_float_constants): Adjust
115 computation of DECnn_MIN and DECnn_MAX for corrected values of
116 emin and emax. Define __DECnn_SUBNORMAL_MIN__ instead of
117 __DECnn_MIN__, and adjust its computation for the corrected value
120 2009-01-06 Jan Hubicka <jh@suse.cz>
123 * i386.c (ix86_expand_call): Use ARRAY_SIZE.
125 2009-01-06 Gerald Pfeifer <gerald@pfeifer.com>
127 * doc/contrib.texi (Contributors): Slightly adjust the end note.
128 Add Robert Clark to the list of testers.
130 2009-01-06 Jan Hubicka <jh@suse.cz>
131 Kai Tietz <kai.tietz@onevision.com>
133 * i386.md (*msabi_syvabi): Add SSE regs clobbers.
134 * i386.c (ix86_expand_call): Add clobbers.
136 2009-01-06 Jan Hubicka <jh@suse.cz>
137 Kai Tietz <kai.tietz@onevision.com>
139 * i386.h (CONDITIONAL_CALL_USAGE): SSE regs are not used for w64 ABI.
140 * i386.c (struct ix86_frame): Add padding0 and nsseregs.
141 (ix86_nsaved_regs): Count only general purpose regs.
142 (ix86_nsaved_sseregs): New.
143 (ix86_compute_frame_layout): Update nsseregs; set preferred alignment
144 to 16 for w64; compute padding and size of sse reg save area.
145 (ix86_emit_save_regs, ix86_emit_save_regs_using_mov): Save only
146 general purpose regs.
147 (ix86_emit_save_sse_regs_using_mov): New.
148 (ix86_expand_prologue): Save SSE regs if needed.
149 (ix86_emit_restore_regs_using_mov): Use only general purpose regs.
150 (ix86_emit_restore_sse_regs_using_mov): New.
151 (ix86_expand_epilogue): Save SSE regs if needed.
153 2009-01-06 Jan Hubicka <jh@suse.cz>
154 Kai Tietz <kai.tietz@onevision.com>
156 * i386.h (ACCUMULATE_OUTGOING_ARGS): Enable for MSABI
157 * i386.c (init_cumulative_args): Disallow calls of MSABI functions
158 when accumulate outgoing args is off.
160 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
163 * ira-color.c (ira_reuse_stack_slot): Check ENABLE_IRA_CHECKING
164 before using pseudos_have_intersected_live_ranges_p.
166 * ira-int.h (ira_assert): Always define.
168 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
170 AVX Programming Reference (December, 2008)
171 * config/i386/avxintrin.h (_mm_permute2_pd): Removed.
172 (_mm256_permute2_pd): Likewise.
173 (_mm_permute2_ps): Likewise.
174 (_mm256_permute2_ps): Likewise.
175 * config/i386/i386.md (UNSPEC_VPERMIL2): Likewise.
176 * config/i386/sse.md (avx_vpermil2<mode>3): Likewise.
178 * config/i386/i386.c (ix86_builtins): Remove
179 IX86_BUILTIN_VPERMIL2PD, IX86_BUILTIN_VPERMIL2PS,
180 IX86_BUILTIN_VPERMIL2PD256 and IX86_BUILTIN_VPERMIL2PS256.
181 (ix86_builtin_type): Remove V8SF_FTYPE_V8SF_V8SF_V8SI_INT,
182 V4DF_FTYPE_V4DF_V4DF_V4DI_INT, V4SF_FTYPE_V4SF_V4SF_V4SI_INT
183 and V2DF_FTYPE_V2DF_V2DF_V2DI_INT.
184 (bdesc_args): Remove __builtin_ia32_vpermil2pd,
185 __builtin_ia32_vpermil2ps, __builtin_ia32_vpermil2pd256 and
186 __builtin_ia32_vpermil2ps256.
187 (ix86_init_mmx_sse_builtins): Updated.
188 (ix86_expand_args_builtin): Likewise.
190 2009-01-05 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
192 * pa.c (output_call): Relocate non-jump insns in the delay slot of
193 long absolute calls when generating PA 2.0 code.
195 2009-01-05 Vladimir Makarov <vmakarov@redhat.com>
197 PR rtl-optimization/38583
198 * params.h (IRA_MAX_CONFLICT_TABLE_SIZE): New macro.
200 * params.def (ira-max-conflict-table-size): New.
202 * doc/invoke.texi (ira-max-conflict-table-size): Decribe.
204 * ira.h (ira_conflicts_p): New external definition.
206 * ira-conflicts.c (build_conflict_bit_table): Do not build too big
207 table. Report this. Return result of building.
208 (ira_build_conflicts): Use ira_conflicts_p. Check result of
209 building conflict table.
211 * ira-color.c (fast_allocation): Use num instead of ira_allocnos_num.
212 (ira_color): Use ira_conflicts_p.
214 * global.c: Include ira.h.
215 (pseudo_for_reload_consideration_p, build_insn_chain): Use
218 * Makefile.in (global.o): Add ira.h.
220 * ira-build.c (mark_all_loops_for_removal,
221 propagate_some_info_from_allocno): New.
222 (remove_unnecessary_allocnos): Call
223 propagate_some_info_from_allocno.
224 (remove_low_level_allocnos): New.
225 (remove_unnecessary_regions): Add parameter. Call
226 mark_all_loops_for_removal and remove_low_level_allocnos. Pass
227 parameter to remove_unnecessary_regions.
228 (ira_build): Remove all regions but root if the conflict table was
229 not built. Update conflict hard regs for allocnos crossing calls.
231 * ira.c (ira_conflicts_p): New global.
232 (ira): Define and use ira_conflicts_p.
234 * reload1.c (compute_use_by_pseudos, reload, count_pseudo,
235 count_spilled_pseudo, find_reg, alter_reg, finish_spills,
236 emit_input_reload_insns, delete_output_reload): Use ira_conflicts_p.
238 2009-01-06 Ben Elliston <bje@au.ibm.com>
240 * gengtype-lex.l (YY_NO_INPUT): Define.
242 2009-01-05 Andrew Pinski <andrew_pinski@playstation.sony.com>
245 * c-common.c (handle_vector_size_attribute): Also reject
248 2009-01-05 Sebastian Pop <sebastian.pop@amd.com>
250 PR tree-optimization/38492
251 * graphite.c (rename_map_elt, debug_rename_elt,
252 debug_rename_map_1, debug_rename_map, new_rename_map_elt,
253 rename_map_elt_info, eq_rename_map_elts,
254 get_new_name_from_old_name, bb_in_sese_p): Moved around.
255 (sese_find_uses_to_rename_use): Renamed sese_build_livein_liveouts_use.
256 (sese_find_uses_to_rename_bb): Renamed sese_build_livein_liveouts_bb.
257 (sese_build_livein_liveouts): New.
258 (new_sese, free_sese): New.
259 (new_scop): Call new_sese.
260 (free_scop): Call free_sese.
261 (rename_variables_from_edge, rename_phis_end_scop): Removed.
262 (register_old_new_names): Renamed register_old_and_new_names.
263 (register_scop_liveout_renames, add_loop_exit_phis,
264 insert_loop_close_phis, struct igp,
265 default_liveout_before_guard, add_guard_exit_phis,
266 insert_guard_phis, copy_renames): New.
267 (translate_clast): Call insert_loop_close_phis and insert_guard_phis.
268 (sese_add_exit_phis_edge): Renamed scop_add_exit_phis_edge.
269 (rewrite_into_sese_closed_ssa): Renamed scop_insert_phis_for_liveouts.
270 (scop_adjust_phis_for_liveouts): New.
271 (gloog): Call scop_adjust_phis_for_liveouts.
273 * graphite.h (struct sese): Documented. Added fields liveout,
275 (SESE_LIVEOUT, SESE_LIVEIN, SESE_LIVEIN_VER, SESE_NUM_VER): New.
276 (new_sese, free_sese, sese_build_livein_liveouts): Declared.
277 (struct scop): Added field liveout_renames.
278 (SCOP_LIVEOUT_RENAMES): New.
280 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
282 PR tree-optimization/38510
283 * graphite.c (recompute_all_dominators): Call mark_irreducible_loops.
284 (translate_clast): Call recompute_all_dominators before
286 (gloog): Call recompute_all_dominators before graphite_verify.
288 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
289 Jan Sjodin <jan.sjodin@amd.com>
291 PR tree-optimization/38500
292 * graphite.c (create_sese_edges): Call fix_loop_structure after
295 2009-01-05 Joel Sherrill <joel.sherrill@oarcorp.com>
297 * config.gcc: Add m32r*-*-rtems*.
298 * config/m32r/rtems.h: New file.
300 2009-01-05 Ben Elliston <bje@au.ibm.com>
302 * Makefile.in (.po.gmo): Use mkinstalldirs, not test -d || mkdir.
304 (po/gcc.pot): Likewise.
306 2009-01-04 David S. Miller <davem@davemloft.net>
308 * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
309 (STARTING_FRAME_OFFSET): Always set to zero.
311 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
313 * tree.def (LSHIFT_EXPR, RSHIFT_EXPR): Add commentary.
314 * tree-cfg.c (verify_gimple_assign_binary): Allow shifts of
315 fixed-point types, and vectors of the same.
317 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
319 * config/mips/sync.md (*mb_barrier): Rename to...
320 (*memory_barrier): ...this.
322 2009-01-04 Jonathan Wakely <jwakely.gcc@gmail.com>
324 * doc/extend.texi (Function Attributes): Move @cindex after @item
325 for 'artificial' and 'flatten'. Fix grammar for 'externally_visible'
326 and put in alphabetical order. Fix 'target' name and put in order.
327 * doc/invoke.texi (-Wstrict-null-sentinel, -fipa-matrix-reorg): Fix
330 2009-01-04 Uros Bizjak <ubizjak@gmail.com>
332 * config/s390/s390.md (UNSPEC_MB): Rename from UNSPECV_MB.
333 (memory_barrier): Expand as unspec instead of unspec_volatile.
334 Remove mem:BLK from insn operands. Use Pmode scratch register.
335 (*memory_barrier): Define as unspec instead of unspec_volatile.
336 Use (match_dup 0) as input operand.
338 * config/sparc/sparc.md (UNSPEC_MEMBAR): Rename from UNSPECV_MEMBAR.
339 * config/sparc/sync.md (memory_barrier): Expand as unspec instead of
340 unspec_volatile. Remove mem:BLK from insn operands. Use Pmode
341 scratch register. Remove operand 1.
342 (*stbar): Define as unspec instead of unspec_volatile.
343 Use (match_dup 0) as input operand, remove (const_int 8).
344 (*membar): Define as unspec instead of unspec_volatile.
345 Use (match_dup 0) as input operand, remove input operand 2.
347 * config/xtensa/xtensa.md (UNSPEC_MEMW): Rename from UNSPECV_MEMW.
348 (memory_barrier): Expand as unspec instead of unspec_volatile.
349 Remove mem:BLK from insn operands. Use Pmode scratch register.
350 (*memory_barrier): Define as unspec instead of unspec_volatile.
351 Use (match_dup 0) as input operand.
353 * config/ia64/sync.md (memory_barrier): Redefine as expander pattern.
354 Remove mem:BLK from insn operands. Use Pmode scratch register.
355 Set volatile flag on operand 0.
356 (*memory_barrier): New insn pattern.
358 * config/rs6000/sync.md (memory_barrier): Remove mem:BLK from
360 (*memory_barrier): Use (match_dup 0) as input operand.
362 * config/mips/sync.md (memory_barrier): Redefine as expander pattern.
363 Remove mem:BLK from insn operands. Use Pmode scratch register.
364 Set volatile flag on operand 0.
365 (*mb_internal): New insn pattern.
367 * config/alpha/sync.md (*memory_barrier): Rename from *mb_internal.
369 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
372 * function.c (struct temp_slot): Move to the section of the file
373 that deals with temp slots. Remove field 'address'.
374 (temp_slot_address_table): New hash table of address -> temp slot.
375 (struct temp_slot_address_entry): New struct, items for the table.
376 (temp_slot_address_compute_hash, temp_slot_address_hash,
377 temp_slot_address_eq, insert_temp_slot_address): Support functions
379 (find_temp_slot_from_address): Rewrite to use the new hash table.
380 (remove_unused_temp_slot_addresses): Remove addresses of temp
381 slots that have been made available.
382 (remove_unused_temp_slot_addresses_1): Call-back for htab_traverse,
383 worker function for remove_unused_temp_slot_addresses.
384 (assign_stack_temp_for_type): Don't clear the temp slot address list.
385 Add the temp slot address to the address -> temp slot map.
386 (update_temp_slot_address): Update via insert_temp_slot_address.
387 (free_temp_slots): Call remove_unused_temp_slot_addresses.
388 (pop_temp_slots): Likewise.
389 (init_temp_slots): Allocate the address -> temp slot map, or empty
390 the map if it is already allocated.
391 (prepare_function_start): Initialize temp slot processing.
393 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
396 * cfgexpand.c (estimate_stack_frame_size): Simplify the estimate:
397 Calculate the size of all stack vars assuming no packing of stack
398 vars will happen, replacing a quadratic algorithm with a linear one.
400 2009-01-03 Jakub Jelinek <jakub@redhat.com>
403 * expmed.c (store_bit_field_1): Don't modify op0 if movstrict insn
406 2009-01-03 Diego Novillo <dnovillo@google.com>
408 * doc/contrib.texi: Update contributions.
410 2009-01-03 Jakub Jelinek <jakub@redhat.com>
413 * builtins.c (fold_builtin_memory_op): Give up if either operand
414 is volatile. Set srctype or desttype to non-qualified version
418 * builtins.c (fold_builtin_expect): Only check DECL_WEAK for VAR_DECLs
421 2009-01-02 Kenneth Zadeck <zadeck@naturalbridge.com>
423 PR rtl-optimization/35805
424 * df-problems.c (df_lr_finalize): Add recursive call to resolve lr
425 problem if fast dce is able to remove any instructions.
426 * dce.c (dce_process_block): Fix dump message.
428 2009-01-02 Mark Mitchell <mark@codesourcery.com>
431 * tree-ssa-pre.c (compute_antic): Correct loop bounds.
433 2009-01-02 Jakub Jelinek <jakub@redhat.com>
436 * tree-flow.h (op_code_prio, op_prio): New prototypes.
437 * tree-pretty-print.c (op_code_prio): New function.
438 (op_prio): No longer static. Use op_code_prio.
439 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs):
440 Use op_prio and op_code_prio to determine if () should be
441 printed around operand(s) or not.
443 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs,
444 dump_gimple_call, dump_gimple_switch, dump_gimple_cond,
445 dump_gimple_label, dump_gimple_try, dump_symbols, dump_gimple_phi,
446 dump_gimple_mem_ops, dump_bb_header, dump_bb_end, pp_cfg_jump): Use
447 pp_character instead of pp_string for single letter printing.
449 2009-01-02 Richard Sandiford <rdsandiford@googlemail.com>
451 * doc/extend.texi: Fix '#pragma GCC option' typo.
453 2009-01-02 Richard Guenther <rguenther@suse.de>
455 * doc/install.texi (--enable-checking): Mention different
457 (--enable-stage1-checking): Document.
459 2009-01-01 Andrew Pinski <pinskia@gmail.com>
462 * tree-cfg.c (verify_expr): Add INDIRECT_REF case. Change MODIFY_EXPR
465 2009-01-02 Ben Elliston <bje@au.ibm.com>
467 * config/fp-bit.h (pack_d): Constify argument.
468 * config/fp-bit.c (makenan): Constify return type. Remove casts.
469 (isnan): Constify argument.
473 (_fpadd_parts): Constify return type.
474 (_fpmul_parts): Likewise.
475 (_fpdiv_parts): Likewise.
477 2009-01-01 Jakub Jelinek <jakub@redhat.com>
480 * c-typeck.c (add_pending_init): Add IMPLICIT argument. Only
481 warn about overwriting initializer with side-effects or
482 -Woverride-init if !IMPLICIT.
483 (output_init_element): Likewise. Pass IMPLICIT down to
485 (process_init_element): Add IMPLICIT argument. Pass it down
486 to output_init_element.
487 (push_init_element, pop_init_level, set_designator): Adjust
488 process_init_element callers.
489 (set_nonincremental_init, set_nonincremental_init_from_string):
490 Adjust add_pending_init callers.
491 (output_pending_init_elements): Adjust output_init_element callers.
492 * c-tree.h (process_init_element): Adjust prototype.
493 * c-parser.c (c_parser_initelt, c_parser_initval): Adjust
494 process_init_element callers.