1 2009-01-17 Dave Korn <dave.korn.cygwin@gmail.com>
3 * Makefile.in (BACKENDLIBS): Reorder to match dependencies.
5 2009-01-17 Sebastian Pop <sebastian.pop@amd.com>
6 Tobias Grosser <tobi.grosser@amd.com>
8 * graphite.c (graphite_trans_scop_block): Do not block single
11 2009-01-16 Alexandre Oliva <aoliva@redhat.com>
13 * ebitmap.h (ebitmap_iter_init): Initialize all fields.
14 * ipa-struct-reorg.c (gen_struct_type): Replace known-true
17 2009-01-16 Richard Guenther <rguenther@suse.de>
19 PR tree-optimization/38835
21 * fold-const.c (fold_binary): Remove PTR + INT -> (INT)(PTR p+ INT)
22 and INT + PTR -> (INT)(PTR p+ INT) folding.
23 * tree-ssa-address.c (create_mem_ref): Properly use POINTER_PLUS_EXPR.
25 2009-01-16 Adam Nemet <anemet@caviumnetworks.com>
28 * expmed.c (expand_shift): With SHIFT_COUNT_TRUNCATED, don't lift
29 the subreg from a lowpart subreg if it is also casting the value.
31 2009-01-16 Sebastian Pop <sebastian.pop@amd.com>
32 Tobias Grosser <tobi.grosser@amd.com>
34 * graphite.c (compare_prefix_loops): New.
35 (build_scop_canonical_schedules): Rewritten.
36 (graphite_transform_loops): Move build_scop_canonical_schedules
37 after build_scop_iteration_domain.
39 2009-01-16 Sebastian Pop <sebastian.pop@amd.com>
40 Tobias Grosser <tobi.grosser@amd.com>
42 * graphite.c (add_conditions_to_domain): Add the loops to
43 the dimension of the iteration domain. Do copy the domain
45 (build_scop_conditions_1): Do not call add_conditions_to_domain.
46 (add_conditions_to_constraints): New.
47 (can_generate_code_stmt, can_generate_code): Removed.
48 (gloog): Do not call can_generate_code.
49 (graphite_transform_loops): Call add_conditions_to_constraints
50 after building the iteration domain.
52 2009-01-16 Jakub Jelinek <jakub@redhat.com>
54 PR tree-optimization/38789
55 * tree-ssa-threadedge.c
56 (record_temporary_equivalences_from_stmts_at_dest): Ignore calls to
59 2009-01-16 Kenneth Zadeck <zadeck@naturalbridge.com>
61 * dce.c (delete_unmarked_insns): Reversed the order that insns are
62 examined before deleting them.
64 2009-01-16 Richard Earnshaw <rearnsha@arm.com>
66 * function.c (aggregate_value_p): Correctly extract the function
67 type from CALL_EXPR_FN lookup.
69 2009-01-16 Hariharan Sandanagobalane <hariharan@picochip.com>
71 * config/picochip/picochip.c (picochip_override_options): Revert
72 CFI asm flag disable commited previously.
74 2009-01-15 Sebastian Pop <sebastian.pop@amd.com>
75 Tobias Grosser <tobi.grosser@amd.com>
76 Jan Sjodin <jan.sjodin@amd.com>
78 * graphite.c (scan_tree_for_params): On substractions negate
79 all the coefficients of the term.
80 (clast_to_gcc_expression_red): New. Handle reduction expressions
81 of more than two operands.
82 (clast_to_gcc_expression): Call clast_to_gcc_expression_red.
83 (get_vdef_before_scop): Handle also the case of default definitions.
85 2009-01-15 Richard Sandiford <rdsandiford@googlemail.com>
87 * caller-save.c (add_used_regs_1, add_used_regs): New functions.
88 (insert_one_insn): Use them instead of REG_DEAD and REG_INC notes.
89 Also use them when walking CALL_INSN_FUNCTION_USAGE.
91 2009-01-15 H.J. Lu <hongjiu.lu@intel.com>
92 Joey Ye <joey.ye@intel.com>
95 * cfgexpand.c (expand_stack_alignment): Don't update stack
96 boundary nor check incoming stack boundary here.
97 (gimple_expand_cfg): Update stack boundary and check incoming
100 2009-01-15 Kenneth Zadeck <zadeck@naturalbridge.com>
102 * dce.c (find_call_stack_args, delete_unmarked_insns): Fixed
105 2009-01-14 Jakub Jelinek <jakub@redhat.com>
107 PR rtl-optimization/38245
108 * calls.c (expand_call): Add stack arguments to
109 CALL_INSN_FUNCTION_USAGE even for pure calls (when
110 ACCUMULATE_OUTGOING_ARGS) and even for args partially passed
111 in regs and partially in memory or BLKmode arguments.
112 (emit_library_call_value_1): Add stack arguments to
113 CALL_INSN_FUNCTION_USAGE even for pure calls (when
114 ACCUMULATE_OUTGOING_ARGS).
115 * dce.c: Include tm_p.h.
116 (find_call_stack_args): New function.
117 (deletable_insn_p): Call it for CALL_P insns. Add ARG_STORES
119 (mark_insn): Call find_call_stack_args for CALL_Ps.
120 (prescan_insns_for_dce): Walk insns backwards in bb rather than
121 forwards. Allocate and free arg_stores bitmap if needed, pass it
122 down to deletable_insn_p, don't mark stores set in arg_stores
123 bitmap, clear the bitmap at the beginning of each bb.
124 * Makefile.in (dce.o): Depend on $(TM_P_H).
126 2009-01-14 Michael Meissner <gnu@the-meissners.org>
129 * i386.c (print_operand): Add tests for 'D', 'C', 'F', 'f' to make
130 sure the insn is a conditional test (bug 22599). Reformat a few long
133 2009-01-14 Sebastian Pop <sebastian.pop@amd.com>
136 * graphite.c (get_vdef_before_scop, scop_adjust_vphi): New.
137 (scop_adjust_phis_for_liveouts): Call scop_adjust_vphi.
138 (gloog): Do not call cleanup_tree_cfg.
139 (graphite_transform_loops): Call cleanup_tree_cfg after all
140 scops have been code generated.
142 2009-01-14 Vladimir Makarov <vmakarov@redhat.com>
144 * testsuite/g++.dg/torture/pr38811.C: New file.
146 2009-01-14 Basile Starynkevitch <basile@starynkevitch.net>
147 * doc/gty.texi (Invoking the garbage collector): Added new node
148 and section documenting ggc_collect.
150 2009-01-14 Richard Guenther <rguenther@suse.de>
152 PR tree-optimization/38826
154 * tree-ssa-structalias.c (emit_alias_warning): Emit the pointer
155 initialization notes only if we actually emitted a warning.
156 (intra_create_variable_infos): Add constraints for a result decl
157 that is passed by hidden reference.
158 (build_pred_graph): Mark all related variables non-direct on
161 2009-01-14 Nick Clifton <nickc@redhat.com>
163 * ira-conflicts.c: Include addresses.h for the definition of
165 (ira_build_conflicts): Use base_reg_class instead of
167 * Makefile.in: Add a dependency of ira-conflicts.o on
170 2009-01-13 Vladimir Makarov <vmakarov@redhat.com>
173 * Makefile.in (ira-lives.o): Add except.h.
175 * ira-lives.c: Include except.h.
176 (process_bb_node_lives): Process can_throw_internal.
178 2009-01-13 Jakub Jelinek <jakub@redhat.com>
180 PR rtl-optimization/38774
181 * combine.c (simplify_set): When undoing cc_use change, don't do
182 PUT_CODE on the newly created comparison, but instead put back the
185 2009-01-13 Joseph Myers <joseph@codesourcery.com>
187 * doc/invoke.texi (ARM Options): Update lists of -mcpu and -march
188 values. Remove duplicate arm8 entry.
190 2009-01-13 Sebastian Pop <sebastian.pop@amd.com>
192 PR tree-optimization/38786
193 * graphite.c (expand_scalar_variables_ssa_name): New, outlined from
194 the SSA_NAME case of expand_scalar_variables_expr.
195 Set the type of an expression to the type of its assign statement.
196 (expand_scalar_variables_expr): Also gather the scalar computation
197 used to index the memory access. Do not pass loop_p.
198 Fix comment. Stop recursion on tcc_constant or tcc_declaration.
199 (expand_scalar_variables_stmt): Pass to expand_scalar_variables_expr
200 the gimple_stmt_iterator where it inserts new code. Do not pass loop_p.
201 (copy_bb_and_scalar_dependences): Do not pass loop_p.
202 (translate_clast): Update call to copy_bb_and_scalar_dependences.
204 2009-01-13 Sebastian Pop <sebastian.pop@amd.com>
206 * graphite.h (debug_value): Removed.
207 * graphite.c (debug_value): Removed.
209 2009-01-13 Richard Earnshaw <rearnsha@arm.com>
211 * arm.c (output_move_double): Don't synthesize thumb-2 ldrd/strd with
212 two 32-bit instructions.
214 2009-01-13 Richard Earnshaw <rearnsha@arm.com>
216 * arm.c (struct processors): Pass for speed down into cost helper
218 (const_ok_for_op): Handle COMPARE and inequality nodes.
219 (arm_rtx_costs_1): Rewrite.
220 (arm_size_rtx_costs): Update prototype.
221 (arm_rtx_costs): Pass speed down to helper functions.
222 (arm_slowmul_rtx_costs): Rework cost calculations.
223 (arm_fastmul_rtx_costs, arm_xscale_rtx_costs): Likewise.
224 (arm_9e_rtx_costs): Likewise.
226 2009-01-13 Uros Bizjak <ubizjak@gmail.com>
228 * config/alpha/alpha.c (alpha_legitimate_address_p): Explicit
229 relocations of local symbols wider than UNITS_PER_WORD are not valid.
230 (alpha_legitimize_address): Do not split local symbols wider than
231 UNITS_PER_WORD into HIGH/LO_SUM parts.
233 2009-01-13 Danny Smith <dannysmith@users.sourceforge.net>
236 * gcc.c (process_command): Replace call to execvp with calls
239 2009-01-03 Anatoly Sokolov <aesok@post.ru>
242 * config/avr/t-avr (LIB1ASMFUNCS): Add _tablejump_elpm.
243 * config/avr/libgcc.S (__do_global_ctors, __do_global_dtors): Add
244 variant for devices with 3-byte PC.
245 (__tablejump_elpm__) : New.
247 2009-01-12 Jakub Jelinek <jakub@redhat.com>
250 * c-parser.c (c_parser_postfix_expression): Allow `->' in
251 offsetof member-designator, handle it as `[0].'.
253 2009-01-12 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
255 * pa.c (pa_asm_output_mi_thunk): Use pc-relative branch to thunk
256 function when not using named sections on targets with named sections
257 if branch distance is less than 262132.
259 2009-01-12 Richard Earnshaw <rearnsha@arm.com>
261 * combine.c (combine_instructions): Recompute
262 optimize_this_for_speed_p for each BB in the main combine loop.
264 2009-01-12 Tomas Bily <tbily@suse.cz>
267 * tree-loop-distribution.c (prop_phis): New function.
268 (generate_builtin): Call prop_phis.
269 * testsuite/gcc.dg/tree-ssa/pr38385.c: New file.
271 2009-01-12 Jakub Jelinek <jakub@redhat.com>
273 PR tree-optimization/38807
274 * tree-ssa-reassoc.c (remove_visited_stmt_chain): Don't look at
275 gimple_visited_p unless stmt is GIMPLE_ASSIGN.
277 2009-01-11 Adam Nemet <anemet@caviumnetworks.com>
279 * expmed.c (store_bit_field_1): Properly truncate the paradoxical
280 subreg of op0 to the original op0.
282 2009-01-11 Laurent GUERBY <laurent@guerby.net>
284 * doc/sourcebuild.texi (Source Tree): Move up intl and fixinc.
286 2009-01-11 Markus Schoepflin <markus.schoepflin@comsoft.de>
289 * gcc/mips-tfile.c (parse_def): Fix parsing of def strings
290 starting with digits.
292 2009-01-10 Jakub Jelinek <jakub@redhat.com>
295 * config/arm/arm.c (arm_is_long_call_p): Don't call
296 arm_function_in_section_p if decl isn't a FUNCTION_DECL.
298 2009-01-09 Steven Bosscher <steven@gcc.gnu.org>
300 * regrename.c (regrename_optimize): Fix dumping.
301 (find_oldest_value_reg): Preserve REG_POINTER.
302 (copy_hardreg_forward_1): Likewise.
304 2009-01-09 Diego Novillo <dnovillo@google.com>
306 * gimple.h (struct gimple_statement_base)<uid>: Document
307 the restrictions on its use.
309 (gimple_set_uid): Tidy.
311 2009-01-09 Jakub Jelinek <jakub@redhat.com>
313 * config/i386/i386.c (ix86_expand_movmem, ix86_expand_setmem): Add
314 zero guard even if align_bytes != 0 and count is smaller than
317 2008-01-09 Vladimir Makarov <vmakarov@redhat.com>
319 PR rtl-optimization/38495
320 * ira-emit.c (print_move_list, ira_debug_move_list): New functions.
321 (add_range_and_copies_from_move_list): Print all added ranges.
322 Add ranges to memory optimized destination.
324 2009-01-09 Jakub Jelinek <jakub@redhat.com>
328 * config/i386/i386.c (override_options): Reject
329 -mstringop-strategy=rep_8byte with -m32.
330 (ix86_expand_movmem): For size_needed == 1 set epilogue_size_needed
331 to 1. Do count comparison against epilogue_size_needed at compile
332 time even when count_exp was constant forced into register. For
333 size_needed don't jump to epilogue, instead just avoid aligning
334 and invoke the body algorithm. If need_zero_guard, add zero guard
335 even if count is non-zero, but smaller than size_needed + number of
336 bytes that could be stored for alignment.
337 (ix86_expand_setmem): For size_needed == 1 set epilogue_size_needed
338 to 1. If need_zero_guard, add zero guard even if count is non-zero,
339 but smaller than size_needed + number of bytes that could be stored
340 for alignment. Compare size_needed with epilogue_size_needed instead
341 of desired_align - align, don't adjust size_needed, pass
342 epilogue_size_needed to the epilogue expanders.
345 * c-pretty-print.c (pp_c_expression): Handle GOTO_EXPR like BIND_EXPR.
347 2009-01-09 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
349 * pa.c (last_address): Change to unsigned.
350 (update_total_code_bytes): Change argument to unsigned. Don't
351 check if insn addresses are set.
352 (pa_output_function_epilogue): Set last_address to UINT_MAX if insn
353 addresses are not set.
354 (pa_asm_output_mi_thunk): Handle wrap when updating last_address.
356 2009-01-09 Nick Clifton <nickc@redhat.com>
358 * config/sh/symbian.c: Replace uses of DECL_INLINE with
359 DECL_DECLARED_INLINE_P.
361 2009-01-09 Jakub Jelinek <jakub@redhat.com>
364 * dojump.c (do_jump_by_parts_zero_rtx): Use mode instead of
365 GET_MODE (op0) in operand_subword_force calls.
368 * fold-const.c (fold_unary): For COMPOUND_EXPR and COND_EXPR,
369 fold_convert arg0 operands to TREE_TYPE (op0) first.
371 2009-01-08 Vladimir Makarov <vmakarov@redhat.com>
373 * params.def (ira-max-conflict-table-size): Decrease default value
376 2009-01-08 Jakub Jelinek <jakub@redhat.com>
378 PR tree-optimization/37031
379 * lambda-code.c (lambda_collect_parameters): Call pointer_set_destroy
381 (build_access_matrix): Reserve correct size for AM_MATRIX vector,
382 allocate it using gc instead of heap, use VEC_quick_push instead of
384 * graphite.c (build_access_matrix): Allocate AM_MATRIX vector using gc
385 instead of heap, use VEC_quick_push instead of VEC_safe_push.
386 * tree-data-ref.h (struct access_matrix): Change matrix to gc
387 allocated vector from heap allocated.
388 * lambda.h: Add DEF_VEC_ALLOC_P for gc allocated lambda_vector.
389 * tree-loop-linear.c (linear_transform_loops): Allocate nest
390 vector only after perfect_loop_nest_depth call.
392 2009-01-08 Sebastian Pop <sebastian.pop@amd.com>
393 Jan Sjodin <jan.sjodin@amd.com>
395 PR tree-optimization/38559
396 * graphite.c (debug_value, copy_constraint,
397 swap_constraint_variables, scale_constraint_variable, ): New.
398 (get_lower_bound, get_upper_bound): Removed.
399 (graphite_trans_bb_strip_mine): Clean up this code that works
400 only for constant number of iterations. Fully copy upper and
401 lower bound constraints, not only the constant part of them.
402 * graphite.h (debug_value): Declared.
404 2009-01-08 Ira Rosen <irar@il.ibm.com>
406 PR tree-optimization/37194
407 * tree-vect-transform.c (vect_estimate_min_profitable_iters):
408 Don't add the cost of cost model guard in prologue to scalar
409 outside cost in case of known number of iterations.
411 2009-01-07 Nathan Froyd <froydnj@codesourcery.com>
412 Alan Modra <amodra@bigpond.net.au>
414 * config/rs6000/rs6000.c (rs6000_legitimize_address): Check for
415 non-word-aligned REG+CONST addressing.
417 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
420 * config/alpha/alpha.c (alpha_end_function): For TARGET_ABI_OSF, call
421 free_after_compilation when outputting a thunk.
422 (alpha_output_mi_thunk_osf): Assert that we are processing a thunk.
423 Do not call free_after_compilation here.
425 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
427 * config/i386/i386.c (ix86_target_string): Use ARRAY_SIZE.
428 (ix86_valid_target_attribute_inner_p): Ditto.
430 2009-01-07 Jan Sjodin <jan.sjodin@amd.com>
432 PR tree-optimization/38492
433 PR tree-optimization/38498
434 * tree-check.c (operator_is_linear, scev_is_linear_expression): New.
435 * tree-chrec.h (scev_is_linear_expression): Declared.
436 * graphite.c (graphite_cannot_represent_loop_niter): New.
437 (scopdet_basic_block_info): Call graphite_cannot_represent_loop_niter.
438 (graphite_loop_normal_form): Use gcc_assert.
439 (scan_tree_for_params): Use CASE_CONVERT.
440 (phi_node_is_iv, bb_contains_non_iv_scalar_phi_nodes): New.
441 (build_scop_conditions_1): Call bb_contains_non_iv_scalar_phi_nodes.
442 Use gcc_assert. Discard scops that contain unhandled cases.
443 (build_scop_conditions): Return a boolean status for unhandled cases.
444 (strip_mine_profitable_p): Print the loop number, not its depth.
445 (is_interchange_valid): Pass the depth of the loop nest, don't
446 recompute it wrongly.
447 (graphite_trans_bb_block): Same.
448 (graphite_trans_bb_block): Print tentative of loop blocking.
449 (graphite_trans_scop_block): Do not print that the loop has been
451 (graphite_transform_loops): Do not handle scops that contain condition
454 2009-01-07 H.J. Lu <hongjiu.lu@intel.com>
456 AVX Programming Reference (December, 2008)
457 * config/i386/avxintrin.h (_mm256_stream_si256): New.
458 (_mm256_stream_pd): Likewise.
459 (_mm256_stream_ps): Likewise.
461 * config/i386/i386.c (ix86_builtins): Add IX86_BUILTIN_MOVNTDQ256,
462 IX86_BUILTIN_MOVNTPD256 and IX86_BUILTIN_MOVNTPS256.
463 (ix86_special_builtin_type): Add VOID_FTYPE_PV4DI_V4DI.
464 (bdesc_special_args): Add __builtin_ia32_movntdq256,
465 __builtin_ia32_movntpd256 and __builtin_ia32_movntps256.
466 (ix86_init_mmx_sse_builtins): Handle VOID_FTYPE_PV4DI_V4DI.
467 (ix86_expand_special_args_builtin): Likewise.
469 * config/i386/sse.md (AVXMODEDI): New.
470 (avx_movnt<mode>): Likewise.
471 (avx_movnt<mode>): Likewise.
472 (<sse>_movnt<mode>): Remove AVX support.
473 (sse2_movntv2di): Likewise.
475 2009-01-07 Richard Guenther <rguenther@suse.de>
478 * fold-const.c (extract_muldiv): Remove obsolete comment.
479 (fold_plusminus_mult_expr): Undo MINUS_EXPR
480 to PLUS_EXPR canonicalization for the canonicalization.
482 2009-01-07 Gerald Pfeifer <gerald@pfeifer.com>
484 * doc/install.texi (alpha*-dec-osf*): Remove note on 32-bit
485 hosted cross-compilers generating less efficient code.
487 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
489 * function.h (rtl_data): Add a dbr_scheduled_p field.
490 * reorg.c (dbr_schedule): Set it.
491 (gate_handle_delay_slots): Check it.
492 * config/mips/mips.c (mips_base_delayed_branch): Delete.
493 (mips_reorg): Check flag_delayed_branch instead of
494 mips_base_delayed_branch.
495 (mips_override_options): Don't set mips_base_delayed_branch
496 or flag_delayed_branch.
498 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
500 PR rtl-optimization/38426.
501 * ira.c (ira): Set current_function_is_leaf earlier.
503 2009-01-06 Jakub Jelinek <jakub@redhat.com>
505 PR rtl-optimization/38722
506 * combine.c (try_combine): Don't modify PATTERN (i3) and notes
507 too early, only set a flag and modify after last possible
510 2009-01-06 Janis Johnson <janis187@us.ibm.com>
513 * ginclude/float.h: Rename DECnn_DEN to DECnn_SUBNORMAL_MIN.
514 * real.c (decimal_single_format): Correct values of emin and emax.
515 (decimal_double_format): Ditto.
516 (decimal_quad_format): Ditto.
517 * c-cppbuiltin.c (builtin_define_decimal_float_constants): Adjust
518 computation of DECnn_MIN and DECnn_MAX for corrected values of
519 emin and emax. Define __DECnn_SUBNORMAL_MIN__ instead of
520 __DECnn_MIN__, and adjust its computation for the corrected value
523 2009-01-06 Jan Hubicka <jh@suse.cz>
526 * i386.c (ix86_expand_call): Use ARRAY_SIZE.
528 2009-01-06 Gerald Pfeifer <gerald@pfeifer.com>
530 * doc/contrib.texi (Contributors): Slightly adjust the end note.
531 Add Robert Clark to the list of testers.
533 2009-01-06 Jan Hubicka <jh@suse.cz>
534 Kai Tietz <kai.tietz@onevision.com>
536 * i386.md (*msabi_syvabi): Add SSE regs clobbers.
537 * i386.c (ix86_expand_call): Add clobbers.
539 2009-01-06 Jan Hubicka <jh@suse.cz>
540 Kai Tietz <kai.tietz@onevision.com>
542 * i386.h (CONDITIONAL_CALL_USAGE): SSE regs are not used for w64 ABI.
543 * i386.c (struct ix86_frame): Add padding0 and nsseregs.
544 (ix86_nsaved_regs): Count only general purpose regs.
545 (ix86_nsaved_sseregs): New.
546 (ix86_compute_frame_layout): Update nsseregs; set preferred alignment
547 to 16 for w64; compute padding and size of sse reg save area.
548 (ix86_emit_save_regs, ix86_emit_save_regs_using_mov): Save only
549 general purpose regs.
550 (ix86_emit_save_sse_regs_using_mov): New.
551 (ix86_expand_prologue): Save SSE regs if needed.
552 (ix86_emit_restore_regs_using_mov): Use only general purpose regs.
553 (ix86_emit_restore_sse_regs_using_mov): New.
554 (ix86_expand_epilogue): Save SSE regs if needed.
556 2009-01-06 Jan Hubicka <jh@suse.cz>
557 Kai Tietz <kai.tietz@onevision.com>
559 * i386.h (ACCUMULATE_OUTGOING_ARGS): Enable for MSABI
560 * i386.c (init_cumulative_args): Disallow calls of MSABI functions
561 when accumulate outgoing args is off.
563 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
566 * ira-color.c (ira_reuse_stack_slot): Check ENABLE_IRA_CHECKING
567 before using pseudos_have_intersected_live_ranges_p.
569 * ira-int.h (ira_assert): Always define.
571 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
573 AVX Programming Reference (December, 2008)
574 * config/i386/avxintrin.h (_mm_permute2_pd): Removed.
575 (_mm256_permute2_pd): Likewise.
576 (_mm_permute2_ps): Likewise.
577 (_mm256_permute2_ps): Likewise.
578 * config/i386/i386.md (UNSPEC_VPERMIL2): Likewise.
579 * config/i386/sse.md (avx_vpermil2<mode>3): Likewise.
581 * config/i386/i386.c (ix86_builtins): Remove
582 IX86_BUILTIN_VPERMIL2PD, IX86_BUILTIN_VPERMIL2PS,
583 IX86_BUILTIN_VPERMIL2PD256 and IX86_BUILTIN_VPERMIL2PS256.
584 (ix86_builtin_type): Remove V8SF_FTYPE_V8SF_V8SF_V8SI_INT,
585 V4DF_FTYPE_V4DF_V4DF_V4DI_INT, V4SF_FTYPE_V4SF_V4SF_V4SI_INT
586 and V2DF_FTYPE_V2DF_V2DF_V2DI_INT.
587 (bdesc_args): Remove __builtin_ia32_vpermil2pd,
588 __builtin_ia32_vpermil2ps, __builtin_ia32_vpermil2pd256 and
589 __builtin_ia32_vpermil2ps256.
590 (ix86_init_mmx_sse_builtins): Updated.
591 (ix86_expand_args_builtin): Likewise.
593 2009-01-05 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
595 * pa.c (output_call): Relocate non-jump insns in the delay slot of
596 long absolute calls when generating PA 2.0 code.
598 2009-01-05 Vladimir Makarov <vmakarov@redhat.com>
600 PR rtl-optimization/38583
601 * params.h (IRA_MAX_CONFLICT_TABLE_SIZE): New macro.
603 * params.def (ira-max-conflict-table-size): New.
605 * doc/invoke.texi (ira-max-conflict-table-size): Decribe.
607 * ira.h (ira_conflicts_p): New external definition.
609 * ira-conflicts.c (build_conflict_bit_table): Do not build too big
610 table. Report this. Return result of building.
611 (ira_build_conflicts): Use ira_conflicts_p. Check result of
612 building conflict table.
614 * ira-color.c (fast_allocation): Use num instead of ira_allocnos_num.
615 (ira_color): Use ira_conflicts_p.
617 * global.c: Include ira.h.
618 (pseudo_for_reload_consideration_p, build_insn_chain): Use
621 * Makefile.in (global.o): Add ira.h.
623 * ira-build.c (mark_all_loops_for_removal,
624 propagate_some_info_from_allocno): New.
625 (remove_unnecessary_allocnos): Call
626 propagate_some_info_from_allocno.
627 (remove_low_level_allocnos): New.
628 (remove_unnecessary_regions): Add parameter. Call
629 mark_all_loops_for_removal and remove_low_level_allocnos. Pass
630 parameter to remove_unnecessary_regions.
631 (ira_build): Remove all regions but root if the conflict table was
632 not built. Update conflict hard regs for allocnos crossing calls.
634 * ira.c (ira_conflicts_p): New global.
635 (ira): Define and use ira_conflicts_p.
637 * reload1.c (compute_use_by_pseudos, reload, count_pseudo,
638 count_spilled_pseudo, find_reg, alter_reg, finish_spills,
639 emit_input_reload_insns, delete_output_reload): Use ira_conflicts_p.
641 2009-01-06 Ben Elliston <bje@au.ibm.com>
643 * gengtype-lex.l (YY_NO_INPUT): Define.
645 2009-01-05 Andrew Pinski <andrew_pinski@playstation.sony.com>
648 * c-common.c (handle_vector_size_attribute): Also reject
651 2009-01-05 Sebastian Pop <sebastian.pop@amd.com>
653 PR tree-optimization/38492
654 * graphite.c (rename_map_elt, debug_rename_elt,
655 debug_rename_map_1, debug_rename_map, new_rename_map_elt,
656 rename_map_elt_info, eq_rename_map_elts,
657 get_new_name_from_old_name, bb_in_sese_p): Moved around.
658 (sese_find_uses_to_rename_use): Renamed sese_build_livein_liveouts_use.
659 (sese_find_uses_to_rename_bb): Renamed sese_build_livein_liveouts_bb.
660 (sese_build_livein_liveouts): New.
661 (new_sese, free_sese): New.
662 (new_scop): Call new_sese.
663 (free_scop): Call free_sese.
664 (rename_variables_from_edge, rename_phis_end_scop): Removed.
665 (register_old_new_names): Renamed register_old_and_new_names.
666 (register_scop_liveout_renames, add_loop_exit_phis,
667 insert_loop_close_phis, struct igp,
668 default_liveout_before_guard, add_guard_exit_phis,
669 insert_guard_phis, copy_renames): New.
670 (translate_clast): Call insert_loop_close_phis and insert_guard_phis.
671 (sese_add_exit_phis_edge): Renamed scop_add_exit_phis_edge.
672 (rewrite_into_sese_closed_ssa): Renamed scop_insert_phis_for_liveouts.
673 (scop_adjust_phis_for_liveouts): New.
674 (gloog): Call scop_adjust_phis_for_liveouts.
676 * graphite.h (struct sese): Documented. Added fields liveout,
678 (SESE_LIVEOUT, SESE_LIVEIN, SESE_LIVEIN_VER, SESE_NUM_VER): New.
679 (new_sese, free_sese, sese_build_livein_liveouts): Declared.
680 (struct scop): Added field liveout_renames.
681 (SCOP_LIVEOUT_RENAMES): New.
683 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
685 PR tree-optimization/38510
686 * graphite.c (recompute_all_dominators): Call mark_irreducible_loops.
687 (translate_clast): Call recompute_all_dominators before
689 (gloog): Call recompute_all_dominators before graphite_verify.
691 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
692 Jan Sjodin <jan.sjodin@amd.com>
694 PR tree-optimization/38500
695 * graphite.c (create_sese_edges): Call fix_loop_structure after
698 2009-01-05 Joel Sherrill <joel.sherrill@oarcorp.com>
700 * config.gcc: Add m32r*-*-rtems*.
701 * config/m32r/rtems.h: New file.
703 2009-01-05 Ben Elliston <bje@au.ibm.com>
705 * Makefile.in (.po.gmo): Use mkinstalldirs, not test -d || mkdir.
707 (po/gcc.pot): Likewise.
709 2009-01-04 David S. Miller <davem@davemloft.net>
711 * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
712 (STARTING_FRAME_OFFSET): Always set to zero.
714 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
716 * tree.def (LSHIFT_EXPR, RSHIFT_EXPR): Add commentary.
717 * tree-cfg.c (verify_gimple_assign_binary): Allow shifts of
718 fixed-point types, and vectors of the same.
720 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
722 * config/mips/sync.md (*mb_barrier): Rename to...
723 (*memory_barrier): ...this.
725 2009-01-04 Jonathan Wakely <jwakely.gcc@gmail.com>
727 * doc/extend.texi (Function Attributes): Move @cindex after @item
728 for 'artificial' and 'flatten'. Fix grammar for 'externally_visible'
729 and put in alphabetical order. Fix 'target' name and put in order.
730 * doc/invoke.texi (-Wstrict-null-sentinel, -fipa-matrix-reorg): Fix
733 2009-01-04 Uros Bizjak <ubizjak@gmail.com>
735 * config/s390/s390.md (UNSPEC_MB): Rename from UNSPECV_MB.
736 (memory_barrier): Expand as unspec instead of unspec_volatile.
737 Remove mem:BLK from insn operands. Use Pmode scratch register.
738 (*memory_barrier): Define as unspec instead of unspec_volatile.
739 Use (match_dup 0) as input operand.
741 * config/sparc/sparc.md (UNSPEC_MEMBAR): Rename from UNSPECV_MEMBAR.
742 * config/sparc/sync.md (memory_barrier): Expand as unspec instead of
743 unspec_volatile. Remove mem:BLK from insn operands. Use Pmode
744 scratch register. Remove operand 1.
745 (*stbar): Define as unspec instead of unspec_volatile.
746 Use (match_dup 0) as input operand, remove (const_int 8).
747 (*membar): Define as unspec instead of unspec_volatile.
748 Use (match_dup 0) as input operand, remove input operand 2.
750 * config/xtensa/xtensa.md (UNSPEC_MEMW): Rename from UNSPECV_MEMW.
751 (memory_barrier): Expand as unspec instead of unspec_volatile.
752 Remove mem:BLK from insn operands. Use Pmode scratch register.
753 (*memory_barrier): Define as unspec instead of unspec_volatile.
754 Use (match_dup 0) as input operand.
756 * config/ia64/sync.md (memory_barrier): Redefine as expander pattern.
757 Remove mem:BLK from insn operands. Use Pmode scratch register.
758 Set volatile flag on operand 0.
759 (*memory_barrier): New insn pattern.
761 * config/rs6000/sync.md (memory_barrier): Remove mem:BLK from
763 (*memory_barrier): Use (match_dup 0) as input operand.
765 * config/mips/sync.md (memory_barrier): Redefine as expander pattern.
766 Remove mem:BLK from insn operands. Use Pmode scratch register.
767 Set volatile flag on operand 0.
768 (*mb_internal): New insn pattern.
770 * config/alpha/sync.md (*memory_barrier): Rename from *mb_internal.
772 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
775 * function.c (struct temp_slot): Move to the section of the file
776 that deals with temp slots. Remove field 'address'.
777 (temp_slot_address_table): New hash table of address -> temp slot.
778 (struct temp_slot_address_entry): New struct, items for the table.
779 (temp_slot_address_compute_hash, temp_slot_address_hash,
780 temp_slot_address_eq, insert_temp_slot_address): Support functions
782 (find_temp_slot_from_address): Rewrite to use the new hash table.
783 (remove_unused_temp_slot_addresses): Remove addresses of temp
784 slots that have been made available.
785 (remove_unused_temp_slot_addresses_1): Call-back for htab_traverse,
786 worker function for remove_unused_temp_slot_addresses.
787 (assign_stack_temp_for_type): Don't clear the temp slot address list.
788 Add the temp slot address to the address -> temp slot map.
789 (update_temp_slot_address): Update via insert_temp_slot_address.
790 (free_temp_slots): Call remove_unused_temp_slot_addresses.
791 (pop_temp_slots): Likewise.
792 (init_temp_slots): Allocate the address -> temp slot map, or empty
793 the map if it is already allocated.
794 (prepare_function_start): Initialize temp slot processing.
796 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
799 * cfgexpand.c (estimate_stack_frame_size): Simplify the estimate:
800 Calculate the size of all stack vars assuming no packing of stack
801 vars will happen, replacing a quadratic algorithm with a linear one.
803 2009-01-03 Jakub Jelinek <jakub@redhat.com>
806 * expmed.c (store_bit_field_1): Don't modify op0 if movstrict insn
809 2009-01-03 Diego Novillo <dnovillo@google.com>
811 * doc/contrib.texi: Update contributions.
813 2009-01-03 Jakub Jelinek <jakub@redhat.com>
816 * builtins.c (fold_builtin_memory_op): Give up if either operand
817 is volatile. Set srctype or desttype to non-qualified version
821 * builtins.c (fold_builtin_expect): Only check DECL_WEAK for VAR_DECLs
824 2009-01-02 Kenneth Zadeck <zadeck@naturalbridge.com>
826 PR rtl-optimization/35805
827 * df-problems.c (df_lr_finalize): Add recursive call to resolve lr
828 problem if fast dce is able to remove any instructions.
829 * dce.c (dce_process_block): Fix dump message.
831 2009-01-02 Mark Mitchell <mark@codesourcery.com>
834 * tree-ssa-pre.c (compute_antic): Correct loop bounds.
836 2009-01-02 Jakub Jelinek <jakub@redhat.com>
839 * tree-flow.h (op_code_prio, op_prio): New prototypes.
840 * tree-pretty-print.c (op_code_prio): New function.
841 (op_prio): No longer static. Use op_code_prio.
842 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs):
843 Use op_prio and op_code_prio to determine if () should be
844 printed around operand(s) or not.
846 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs,
847 dump_gimple_call, dump_gimple_switch, dump_gimple_cond,
848 dump_gimple_label, dump_gimple_try, dump_symbols, dump_gimple_phi,
849 dump_gimple_mem_ops, dump_bb_header, dump_bb_end, pp_cfg_jump): Use
850 pp_character instead of pp_string for single letter printing.
852 2009-01-02 Richard Sandiford <rdsandiford@googlemail.com>
854 * doc/extend.texi: Fix '#pragma GCC option' typo.
856 2009-01-02 Richard Guenther <rguenther@suse.de>
858 * doc/install.texi (--enable-checking): Mention different
860 (--enable-stage1-checking): Document.
862 2009-01-01 Andrew Pinski <pinskia@gmail.com>
865 * tree-cfg.c (verify_expr): Add INDIRECT_REF case. Change MODIFY_EXPR
868 2009-01-02 Ben Elliston <bje@au.ibm.com>
870 * config/fp-bit.h (pack_d): Constify argument.
871 * config/fp-bit.c (makenan): Constify return type. Remove casts.
872 (isnan): Constify argument.
876 (_fpadd_parts): Constify return type.
877 (_fpmul_parts): Likewise.
878 (_fpdiv_parts): Likewise.
880 2009-01-01 Jakub Jelinek <jakub@redhat.com>
883 * c-typeck.c (add_pending_init): Add IMPLICIT argument. Only
884 warn about overwriting initializer with side-effects or
885 -Woverride-init if !IMPLICIT.
886 (output_init_element): Likewise. Pass IMPLICIT down to
888 (process_init_element): Add IMPLICIT argument. Pass it down
889 to output_init_element.
890 (push_init_element, pop_init_level, set_designator): Adjust
891 process_init_element callers.
892 (set_nonincremental_init, set_nonincremental_init_from_string):
893 Adjust add_pending_init callers.
894 (output_pending_init_elements): Adjust output_init_element callers.
895 * c-tree.h (process_init_element): Adjust prototype.
896 * c-parser.c (c_parser_initelt, c_parser_initval): Adjust
897 process_init_element callers.