1 2009-01-18 Richard Guenther <rguenther@suse.de>
3 PR tree-optimization/38819
4 * tree-flow.h (operation_could_trap_helper_p): Declare.
5 * tree-eh.c (operation_could_trap_helper_p): Export.
6 * tree-ssa-sccvn.h (vn_nary_may_trap): Declare.
7 * tree-ssa-sccvn.c (vn_nary_may_trap): New function.
8 * tree-ssa-pre.c (insert_into_preds_of_block): Check if we
9 are about to insert a possibly trapping instruction and fail
12 2009-01-18 Andreas Schwab <schwab@suse.de>
14 * doc/install.texi (Configuration): Remove obsolete paragraph
15 about use of --with-gnu-ld with --with-gnu-as.
17 2009-01-18 Kazu Hirata <kazu@codesourcery.com>
19 * doc/extend.texi, doc/gimple.texi, doc/invoke.texi,
20 doc/md.texi, doc/sourcebuild.texi, doc/tm.texi: Fix typos.
21 Follow spelling conventions.
23 2009-01-18 Ben Elliston <bje@au.ibm.com>
25 * bitmap.c (bitmap_obstack_alloc_stat): Adjust cast to eliminate
26 C++ warning about implicit conversion from void * to struct
28 (bitmap_obstack_free): Likewise for bitmap_element *.
29 * Makefile.in (bitmap.o-warn): Remove goal.
31 2009-01-17 Dave Korn <dave.korn.cygwin@gmail.com>
33 * Makefile.in (BACKENDLIBS): Reorder to match dependencies.
35 2009-01-17 Sebastian Pop <sebastian.pop@amd.com>
36 Tobias Grosser <tobi.grosser@amd.com>
38 * graphite.c (graphite_trans_scop_block): Do not block single
41 2009-01-16 Alexandre Oliva <aoliva@redhat.com>
43 * ebitmap.h (ebitmap_iter_init): Initialize all fields.
44 * ipa-struct-reorg.c (gen_struct_type): Replace known-true
47 2009-01-16 Richard Guenther <rguenther@suse.de>
49 PR tree-optimization/38835
51 * fold-const.c (fold_binary): Remove PTR + INT -> (INT)(PTR p+ INT)
52 and INT + PTR -> (INT)(PTR p+ INT) folding.
53 * tree-ssa-address.c (create_mem_ref): Properly use POINTER_PLUS_EXPR.
55 2009-01-16 Adam Nemet <anemet@caviumnetworks.com>
58 * expmed.c (expand_shift): With SHIFT_COUNT_TRUNCATED, don't lift
59 the subreg from a lowpart subreg if it is also casting the value.
61 2009-01-16 Sebastian Pop <sebastian.pop@amd.com>
62 Tobias Grosser <tobi.grosser@amd.com>
64 * graphite.c (compare_prefix_loops): New.
65 (build_scop_canonical_schedules): Rewritten.
66 (graphite_transform_loops): Move build_scop_canonical_schedules
67 after build_scop_iteration_domain.
69 2009-01-16 Sebastian Pop <sebastian.pop@amd.com>
70 Tobias Grosser <tobi.grosser@amd.com>
72 * graphite.c (add_conditions_to_domain): Add the loops to
73 the dimension of the iteration domain. Do copy the domain
75 (build_scop_conditions_1): Do not call add_conditions_to_domain.
76 (add_conditions_to_constraints): New.
77 (can_generate_code_stmt, can_generate_code): Removed.
78 (gloog): Do not call can_generate_code.
79 (graphite_transform_loops): Call add_conditions_to_constraints
80 after building the iteration domain.
82 2009-01-16 Jakub Jelinek <jakub@redhat.com>
84 PR tree-optimization/38789
85 * tree-ssa-threadedge.c
86 (record_temporary_equivalences_from_stmts_at_dest): Ignore calls to
89 2009-01-16 Kenneth Zadeck <zadeck@naturalbridge.com>
91 * dce.c (delete_unmarked_insns): Reversed the order that insns are
92 examined before deleting them.
94 2009-01-16 Richard Earnshaw <rearnsha@arm.com>
96 * function.c (aggregate_value_p): Correctly extract the function
97 type from CALL_EXPR_FN lookup.
99 2009-01-16 Hariharan Sandanagobalane <hariharan@picochip.com>
101 * config/picochip/picochip.c (picochip_override_options): Revert
102 CFI asm flag disable commited previously.
104 2009-01-15 Sebastian Pop <sebastian.pop@amd.com>
105 Tobias Grosser <tobi.grosser@amd.com>
106 Jan Sjodin <jan.sjodin@amd.com>
108 * graphite.c (scan_tree_for_params): On substractions negate
109 all the coefficients of the term.
110 (clast_to_gcc_expression_red): New. Handle reduction expressions
111 of more than two operands.
112 (clast_to_gcc_expression): Call clast_to_gcc_expression_red.
113 (get_vdef_before_scop): Handle also the case of default definitions.
115 2009-01-15 Richard Sandiford <rdsandiford@googlemail.com>
117 * caller-save.c (add_used_regs_1, add_used_regs): New functions.
118 (insert_one_insn): Use them instead of REG_DEAD and REG_INC notes.
119 Also use them when walking CALL_INSN_FUNCTION_USAGE.
121 2009-01-15 H.J. Lu <hongjiu.lu@intel.com>
122 Joey Ye <joey.ye@intel.com>
125 * cfgexpand.c (expand_stack_alignment): Don't update stack
126 boundary nor check incoming stack boundary here.
127 (gimple_expand_cfg): Update stack boundary and check incoming
130 2009-01-15 Kenneth Zadeck <zadeck@naturalbridge.com>
132 * dce.c (find_call_stack_args, delete_unmarked_insns): Fixed
135 2009-01-14 Jakub Jelinek <jakub@redhat.com>
137 PR rtl-optimization/38245
138 * calls.c (expand_call): Add stack arguments to
139 CALL_INSN_FUNCTION_USAGE even for pure calls (when
140 ACCUMULATE_OUTGOING_ARGS) and even for args partially passed
141 in regs and partially in memory or BLKmode arguments.
142 (emit_library_call_value_1): Add stack arguments to
143 CALL_INSN_FUNCTION_USAGE even for pure calls (when
144 ACCUMULATE_OUTGOING_ARGS).
145 * dce.c: Include tm_p.h.
146 (find_call_stack_args): New function.
147 (deletable_insn_p): Call it for CALL_P insns. Add ARG_STORES
149 (mark_insn): Call find_call_stack_args for CALL_Ps.
150 (prescan_insns_for_dce): Walk insns backwards in bb rather than
151 forwards. Allocate and free arg_stores bitmap if needed, pass it
152 down to deletable_insn_p, don't mark stores set in arg_stores
153 bitmap, clear the bitmap at the beginning of each bb.
154 * Makefile.in (dce.o): Depend on $(TM_P_H).
156 2009-01-14 Michael Meissner <gnu@the-meissners.org>
159 * i386.c (print_operand): Add tests for 'D', 'C', 'F', 'f' to make
160 sure the insn is a conditional test (bug 22599). Reformat a few long
163 2009-01-14 Sebastian Pop <sebastian.pop@amd.com>
166 * graphite.c (get_vdef_before_scop, scop_adjust_vphi): New.
167 (scop_adjust_phis_for_liveouts): Call scop_adjust_vphi.
168 (gloog): Do not call cleanup_tree_cfg.
169 (graphite_transform_loops): Call cleanup_tree_cfg after all
170 scops have been code generated.
172 2009-01-14 Vladimir Makarov <vmakarov@redhat.com>
174 * testsuite/g++.dg/torture/pr38811.C: New file.
176 2009-01-14 Basile Starynkevitch <basile@starynkevitch.net>
177 * doc/gty.texi (Invoking the garbage collector): Added new node
178 and section documenting ggc_collect.
180 2009-01-14 Richard Guenther <rguenther@suse.de>
182 PR tree-optimization/38826
184 * tree-ssa-structalias.c (emit_alias_warning): Emit the pointer
185 initialization notes only if we actually emitted a warning.
186 (intra_create_variable_infos): Add constraints for a result decl
187 that is passed by hidden reference.
188 (build_pred_graph): Mark all related variables non-direct on
191 2009-01-14 Nick Clifton <nickc@redhat.com>
193 * ira-conflicts.c: Include addresses.h for the definition of
195 (ira_build_conflicts): Use base_reg_class instead of
197 * Makefile.in: Add a dependency of ira-conflicts.o on
200 2009-01-13 Vladimir Makarov <vmakarov@redhat.com>
203 * Makefile.in (ira-lives.o): Add except.h.
205 * ira-lives.c: Include except.h.
206 (process_bb_node_lives): Process can_throw_internal.
208 2009-01-13 Jakub Jelinek <jakub@redhat.com>
210 PR rtl-optimization/38774
211 * combine.c (simplify_set): When undoing cc_use change, don't do
212 PUT_CODE on the newly created comparison, but instead put back the
215 2009-01-13 Joseph Myers <joseph@codesourcery.com>
217 * doc/invoke.texi (ARM Options): Update lists of -mcpu and -march
218 values. Remove duplicate arm8 entry.
220 2009-01-13 Sebastian Pop <sebastian.pop@amd.com>
222 PR tree-optimization/38786
223 * graphite.c (expand_scalar_variables_ssa_name): New, outlined from
224 the SSA_NAME case of expand_scalar_variables_expr.
225 Set the type of an expression to the type of its assign statement.
226 (expand_scalar_variables_expr): Also gather the scalar computation
227 used to index the memory access. Do not pass loop_p.
228 Fix comment. Stop recursion on tcc_constant or tcc_declaration.
229 (expand_scalar_variables_stmt): Pass to expand_scalar_variables_expr
230 the gimple_stmt_iterator where it inserts new code. Do not pass loop_p.
231 (copy_bb_and_scalar_dependences): Do not pass loop_p.
232 (translate_clast): Update call to copy_bb_and_scalar_dependences.
234 2009-01-13 Sebastian Pop <sebastian.pop@amd.com>
236 * graphite.h (debug_value): Removed.
237 * graphite.c (debug_value): Removed.
239 2009-01-13 Richard Earnshaw <rearnsha@arm.com>
241 * arm.c (output_move_double): Don't synthesize thumb-2 ldrd/strd with
242 two 32-bit instructions.
244 2009-01-13 Richard Earnshaw <rearnsha@arm.com>
246 * arm.c (struct processors): Pass for speed down into cost helper
248 (const_ok_for_op): Handle COMPARE and inequality nodes.
249 (arm_rtx_costs_1): Rewrite.
250 (arm_size_rtx_costs): Update prototype.
251 (arm_rtx_costs): Pass speed down to helper functions.
252 (arm_slowmul_rtx_costs): Rework cost calculations.
253 (arm_fastmul_rtx_costs, arm_xscale_rtx_costs): Likewise.
254 (arm_9e_rtx_costs): Likewise.
256 2009-01-13 Uros Bizjak <ubizjak@gmail.com>
258 * config/alpha/alpha.c (alpha_legitimate_address_p): Explicit
259 relocations of local symbols wider than UNITS_PER_WORD are not valid.
260 (alpha_legitimize_address): Do not split local symbols wider than
261 UNITS_PER_WORD into HIGH/LO_SUM parts.
263 2009-01-13 Danny Smith <dannysmith@users.sourceforge.net>
266 * gcc.c (process_command): Replace call to execvp with calls
269 2009-01-03 Anatoly Sokolov <aesok@post.ru>
272 * config/avr/t-avr (LIB1ASMFUNCS): Add _tablejump_elpm.
273 * config/avr/libgcc.S (__do_global_ctors, __do_global_dtors): Add
274 variant for devices with 3-byte PC.
275 (__tablejump_elpm__) : New.
277 2009-01-12 Jakub Jelinek <jakub@redhat.com>
280 * c-parser.c (c_parser_postfix_expression): Allow `->' in
281 offsetof member-designator, handle it as `[0].'.
283 2009-01-12 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
285 * pa.c (pa_asm_output_mi_thunk): Use pc-relative branch to thunk
286 function when not using named sections on targets with named sections
287 if branch distance is less than 262132.
289 2009-01-12 Richard Earnshaw <rearnsha@arm.com>
291 * combine.c (combine_instructions): Recompute
292 optimize_this_for_speed_p for each BB in the main combine loop.
294 2009-01-12 Tomas Bily <tbily@suse.cz>
297 * tree-loop-distribution.c (prop_phis): New function.
298 (generate_builtin): Call prop_phis.
299 * testsuite/gcc.dg/tree-ssa/pr38385.c: New file.
301 2009-01-12 Jakub Jelinek <jakub@redhat.com>
303 PR tree-optimization/38807
304 * tree-ssa-reassoc.c (remove_visited_stmt_chain): Don't look at
305 gimple_visited_p unless stmt is GIMPLE_ASSIGN.
307 2009-01-11 Adam Nemet <anemet@caviumnetworks.com>
309 * expmed.c (store_bit_field_1): Properly truncate the paradoxical
310 subreg of op0 to the original op0.
312 2009-01-11 Laurent GUERBY <laurent@guerby.net>
314 * doc/sourcebuild.texi (Source Tree): Move up intl and fixinc.
316 2009-01-11 Markus Schoepflin <markus.schoepflin@comsoft.de>
319 * gcc/mips-tfile.c (parse_def): Fix parsing of def strings
320 starting with digits.
322 2009-01-10 Jakub Jelinek <jakub@redhat.com>
325 * config/arm/arm.c (arm_is_long_call_p): Don't call
326 arm_function_in_section_p if decl isn't a FUNCTION_DECL.
328 2009-01-09 Steven Bosscher <steven@gcc.gnu.org>
330 * regrename.c (regrename_optimize): Fix dumping.
331 (find_oldest_value_reg): Preserve REG_POINTER.
332 (copy_hardreg_forward_1): Likewise.
334 2009-01-09 Diego Novillo <dnovillo@google.com>
336 * gimple.h (struct gimple_statement_base)<uid>: Document
337 the restrictions on its use.
339 (gimple_set_uid): Tidy.
341 2009-01-09 Jakub Jelinek <jakub@redhat.com>
343 * config/i386/i386.c (ix86_expand_movmem, ix86_expand_setmem): Add
344 zero guard even if align_bytes != 0 and count is smaller than
347 2008-01-09 Vladimir Makarov <vmakarov@redhat.com>
349 PR rtl-optimization/38495
350 * ira-emit.c (print_move_list, ira_debug_move_list): New functions.
351 (add_range_and_copies_from_move_list): Print all added ranges.
352 Add ranges to memory optimized destination.
354 2009-01-09 Jakub Jelinek <jakub@redhat.com>
358 * config/i386/i386.c (override_options): Reject
359 -mstringop-strategy=rep_8byte with -m32.
360 (ix86_expand_movmem): For size_needed == 1 set epilogue_size_needed
361 to 1. Do count comparison against epilogue_size_needed at compile
362 time even when count_exp was constant forced into register. For
363 size_needed don't jump to epilogue, instead just avoid aligning
364 and invoke the body algorithm. If need_zero_guard, add zero guard
365 even if count is non-zero, but smaller than size_needed + number of
366 bytes that could be stored for alignment.
367 (ix86_expand_setmem): For size_needed == 1 set epilogue_size_needed
368 to 1. If need_zero_guard, add zero guard even if count is non-zero,
369 but smaller than size_needed + number of bytes that could be stored
370 for alignment. Compare size_needed with epilogue_size_needed instead
371 of desired_align - align, don't adjust size_needed, pass
372 epilogue_size_needed to the epilogue expanders.
375 * c-pretty-print.c (pp_c_expression): Handle GOTO_EXPR like BIND_EXPR.
377 2009-01-09 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
379 * pa.c (last_address): Change to unsigned.
380 (update_total_code_bytes): Change argument to unsigned. Don't
381 check if insn addresses are set.
382 (pa_output_function_epilogue): Set last_address to UINT_MAX if insn
383 addresses are not set.
384 (pa_asm_output_mi_thunk): Handle wrap when updating last_address.
386 2009-01-09 Nick Clifton <nickc@redhat.com>
388 * config/sh/symbian.c: Replace uses of DECL_INLINE with
389 DECL_DECLARED_INLINE_P.
391 2009-01-09 Jakub Jelinek <jakub@redhat.com>
394 * dojump.c (do_jump_by_parts_zero_rtx): Use mode instead of
395 GET_MODE (op0) in operand_subword_force calls.
398 * fold-const.c (fold_unary): For COMPOUND_EXPR and COND_EXPR,
399 fold_convert arg0 operands to TREE_TYPE (op0) first.
401 2009-01-08 Vladimir Makarov <vmakarov@redhat.com>
403 * params.def (ira-max-conflict-table-size): Decrease default value
406 2009-01-08 Jakub Jelinek <jakub@redhat.com>
408 PR tree-optimization/37031
409 * lambda-code.c (lambda_collect_parameters): Call pointer_set_destroy
411 (build_access_matrix): Reserve correct size for AM_MATRIX vector,
412 allocate it using gc instead of heap, use VEC_quick_push instead of
414 * graphite.c (build_access_matrix): Allocate AM_MATRIX vector using gc
415 instead of heap, use VEC_quick_push instead of VEC_safe_push.
416 * tree-data-ref.h (struct access_matrix): Change matrix to gc
417 allocated vector from heap allocated.
418 * lambda.h: Add DEF_VEC_ALLOC_P for gc allocated lambda_vector.
419 * tree-loop-linear.c (linear_transform_loops): Allocate nest
420 vector only after perfect_loop_nest_depth call.
422 2009-01-08 Sebastian Pop <sebastian.pop@amd.com>
423 Jan Sjodin <jan.sjodin@amd.com>
425 PR tree-optimization/38559
426 * graphite.c (debug_value, copy_constraint,
427 swap_constraint_variables, scale_constraint_variable, ): New.
428 (get_lower_bound, get_upper_bound): Removed.
429 (graphite_trans_bb_strip_mine): Clean up this code that works
430 only for constant number of iterations. Fully copy upper and
431 lower bound constraints, not only the constant part of them.
432 * graphite.h (debug_value): Declared.
434 2009-01-08 Ira Rosen <irar@il.ibm.com>
436 PR tree-optimization/37194
437 * tree-vect-transform.c (vect_estimate_min_profitable_iters):
438 Don't add the cost of cost model guard in prologue to scalar
439 outside cost in case of known number of iterations.
441 2009-01-07 Nathan Froyd <froydnj@codesourcery.com>
442 Alan Modra <amodra@bigpond.net.au>
444 * config/rs6000/rs6000.c (rs6000_legitimize_address): Check for
445 non-word-aligned REG+CONST addressing.
447 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
450 * config/alpha/alpha.c (alpha_end_function): For TARGET_ABI_OSF, call
451 free_after_compilation when outputting a thunk.
452 (alpha_output_mi_thunk_osf): Assert that we are processing a thunk.
453 Do not call free_after_compilation here.
455 2009-01-07 Uros Bizjak <ubizjak@gmail.com>
457 * config/i386/i386.c (ix86_target_string): Use ARRAY_SIZE.
458 (ix86_valid_target_attribute_inner_p): Ditto.
460 2009-01-07 Jan Sjodin <jan.sjodin@amd.com>
462 PR tree-optimization/38492
463 PR tree-optimization/38498
464 * tree-check.c (operator_is_linear, scev_is_linear_expression): New.
465 * tree-chrec.h (scev_is_linear_expression): Declared.
466 * graphite.c (graphite_cannot_represent_loop_niter): New.
467 (scopdet_basic_block_info): Call graphite_cannot_represent_loop_niter.
468 (graphite_loop_normal_form): Use gcc_assert.
469 (scan_tree_for_params): Use CASE_CONVERT.
470 (phi_node_is_iv, bb_contains_non_iv_scalar_phi_nodes): New.
471 (build_scop_conditions_1): Call bb_contains_non_iv_scalar_phi_nodes.
472 Use gcc_assert. Discard scops that contain unhandled cases.
473 (build_scop_conditions): Return a boolean status for unhandled cases.
474 (strip_mine_profitable_p): Print the loop number, not its depth.
475 (is_interchange_valid): Pass the depth of the loop nest, don't
476 recompute it wrongly.
477 (graphite_trans_bb_block): Same.
478 (graphite_trans_bb_block): Print tentative of loop blocking.
479 (graphite_trans_scop_block): Do not print that the loop has been
481 (graphite_transform_loops): Do not handle scops that contain condition
484 2009-01-07 H.J. Lu <hongjiu.lu@intel.com>
486 AVX Programming Reference (December, 2008)
487 * config/i386/avxintrin.h (_mm256_stream_si256): New.
488 (_mm256_stream_pd): Likewise.
489 (_mm256_stream_ps): Likewise.
491 * config/i386/i386.c (ix86_builtins): Add IX86_BUILTIN_MOVNTDQ256,
492 IX86_BUILTIN_MOVNTPD256 and IX86_BUILTIN_MOVNTPS256.
493 (ix86_special_builtin_type): Add VOID_FTYPE_PV4DI_V4DI.
494 (bdesc_special_args): Add __builtin_ia32_movntdq256,
495 __builtin_ia32_movntpd256 and __builtin_ia32_movntps256.
496 (ix86_init_mmx_sse_builtins): Handle VOID_FTYPE_PV4DI_V4DI.
497 (ix86_expand_special_args_builtin): Likewise.
499 * config/i386/sse.md (AVXMODEDI): New.
500 (avx_movnt<mode>): Likewise.
501 (avx_movnt<mode>): Likewise.
502 (<sse>_movnt<mode>): Remove AVX support.
503 (sse2_movntv2di): Likewise.
505 2009-01-07 Richard Guenther <rguenther@suse.de>
508 * fold-const.c (extract_muldiv): Remove obsolete comment.
509 (fold_plusminus_mult_expr): Undo MINUS_EXPR
510 to PLUS_EXPR canonicalization for the canonicalization.
512 2009-01-07 Gerald Pfeifer <gerald@pfeifer.com>
514 * doc/install.texi (alpha*-dec-osf*): Remove note on 32-bit
515 hosted cross-compilers generating less efficient code.
517 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
519 * function.h (rtl_data): Add a dbr_scheduled_p field.
520 * reorg.c (dbr_schedule): Set it.
521 (gate_handle_delay_slots): Check it.
522 * config/mips/mips.c (mips_base_delayed_branch): Delete.
523 (mips_reorg): Check flag_delayed_branch instead of
524 mips_base_delayed_branch.
525 (mips_override_options): Don't set mips_base_delayed_branch
526 or flag_delayed_branch.
528 2009-01-06 Richard Sandiford <rdsandiford@googlemail.com>
530 PR rtl-optimization/38426.
531 * ira.c (ira): Set current_function_is_leaf earlier.
533 2009-01-06 Jakub Jelinek <jakub@redhat.com>
535 PR rtl-optimization/38722
536 * combine.c (try_combine): Don't modify PATTERN (i3) and notes
537 too early, only set a flag and modify after last possible
540 2009-01-06 Janis Johnson <janis187@us.ibm.com>
543 * ginclude/float.h: Rename DECnn_DEN to DECnn_SUBNORMAL_MIN.
544 * real.c (decimal_single_format): Correct values of emin and emax.
545 (decimal_double_format): Ditto.
546 (decimal_quad_format): Ditto.
547 * c-cppbuiltin.c (builtin_define_decimal_float_constants): Adjust
548 computation of DECnn_MIN and DECnn_MAX for corrected values of
549 emin and emax. Define __DECnn_SUBNORMAL_MIN__ instead of
550 __DECnn_MIN__, and adjust its computation for the corrected value
553 2009-01-06 Jan Hubicka <jh@suse.cz>
556 * i386.c (ix86_expand_call): Use ARRAY_SIZE.
558 2009-01-06 Gerald Pfeifer <gerald@pfeifer.com>
560 * doc/contrib.texi (Contributors): Slightly adjust the end note.
561 Add Robert Clark to the list of testers.
563 2009-01-06 Jan Hubicka <jh@suse.cz>
564 Kai Tietz <kai.tietz@onevision.com>
566 * i386.md (*msabi_syvabi): Add SSE regs clobbers.
567 * i386.c (ix86_expand_call): Add clobbers.
569 2009-01-06 Jan Hubicka <jh@suse.cz>
570 Kai Tietz <kai.tietz@onevision.com>
572 * i386.h (CONDITIONAL_CALL_USAGE): SSE regs are not used for w64 ABI.
573 * i386.c (struct ix86_frame): Add padding0 and nsseregs.
574 (ix86_nsaved_regs): Count only general purpose regs.
575 (ix86_nsaved_sseregs): New.
576 (ix86_compute_frame_layout): Update nsseregs; set preferred alignment
577 to 16 for w64; compute padding and size of sse reg save area.
578 (ix86_emit_save_regs, ix86_emit_save_regs_using_mov): Save only
579 general purpose regs.
580 (ix86_emit_save_sse_regs_using_mov): New.
581 (ix86_expand_prologue): Save SSE regs if needed.
582 (ix86_emit_restore_regs_using_mov): Use only general purpose regs.
583 (ix86_emit_restore_sse_regs_using_mov): New.
584 (ix86_expand_epilogue): Save SSE regs if needed.
586 2009-01-06 Jan Hubicka <jh@suse.cz>
587 Kai Tietz <kai.tietz@onevision.com>
589 * i386.h (ACCUMULATE_OUTGOING_ARGS): Enable for MSABI
590 * i386.c (init_cumulative_args): Disallow calls of MSABI functions
591 when accumulate outgoing args is off.
593 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
596 * ira-color.c (ira_reuse_stack_slot): Check ENABLE_IRA_CHECKING
597 before using pseudos_have_intersected_live_ranges_p.
599 * ira-int.h (ira_assert): Always define.
601 2009-01-06 H.J. Lu <hongjiu.lu@intel.com>
603 AVX Programming Reference (December, 2008)
604 * config/i386/avxintrin.h (_mm_permute2_pd): Removed.
605 (_mm256_permute2_pd): Likewise.
606 (_mm_permute2_ps): Likewise.
607 (_mm256_permute2_ps): Likewise.
608 * config/i386/i386.md (UNSPEC_VPERMIL2): Likewise.
609 * config/i386/sse.md (avx_vpermil2<mode>3): Likewise.
611 * config/i386/i386.c (ix86_builtins): Remove
612 IX86_BUILTIN_VPERMIL2PD, IX86_BUILTIN_VPERMIL2PS,
613 IX86_BUILTIN_VPERMIL2PD256 and IX86_BUILTIN_VPERMIL2PS256.
614 (ix86_builtin_type): Remove V8SF_FTYPE_V8SF_V8SF_V8SI_INT,
615 V4DF_FTYPE_V4DF_V4DF_V4DI_INT, V4SF_FTYPE_V4SF_V4SF_V4SI_INT
616 and V2DF_FTYPE_V2DF_V2DF_V2DI_INT.
617 (bdesc_args): Remove __builtin_ia32_vpermil2pd,
618 __builtin_ia32_vpermil2ps, __builtin_ia32_vpermil2pd256 and
619 __builtin_ia32_vpermil2ps256.
620 (ix86_init_mmx_sse_builtins): Updated.
621 (ix86_expand_args_builtin): Likewise.
623 2009-01-05 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
625 * pa.c (output_call): Relocate non-jump insns in the delay slot of
626 long absolute calls when generating PA 2.0 code.
628 2009-01-05 Vladimir Makarov <vmakarov@redhat.com>
630 PR rtl-optimization/38583
631 * params.h (IRA_MAX_CONFLICT_TABLE_SIZE): New macro.
633 * params.def (ira-max-conflict-table-size): New.
635 * doc/invoke.texi (ira-max-conflict-table-size): Decribe.
637 * ira.h (ira_conflicts_p): New external definition.
639 * ira-conflicts.c (build_conflict_bit_table): Do not build too big
640 table. Report this. Return result of building.
641 (ira_build_conflicts): Use ira_conflicts_p. Check result of
642 building conflict table.
644 * ira-color.c (fast_allocation): Use num instead of ira_allocnos_num.
645 (ira_color): Use ira_conflicts_p.
647 * global.c: Include ira.h.
648 (pseudo_for_reload_consideration_p, build_insn_chain): Use
651 * Makefile.in (global.o): Add ira.h.
653 * ira-build.c (mark_all_loops_for_removal,
654 propagate_some_info_from_allocno): New.
655 (remove_unnecessary_allocnos): Call
656 propagate_some_info_from_allocno.
657 (remove_low_level_allocnos): New.
658 (remove_unnecessary_regions): Add parameter. Call
659 mark_all_loops_for_removal and remove_low_level_allocnos. Pass
660 parameter to remove_unnecessary_regions.
661 (ira_build): Remove all regions but root if the conflict table was
662 not built. Update conflict hard regs for allocnos crossing calls.
664 * ira.c (ira_conflicts_p): New global.
665 (ira): Define and use ira_conflicts_p.
667 * reload1.c (compute_use_by_pseudos, reload, count_pseudo,
668 count_spilled_pseudo, find_reg, alter_reg, finish_spills,
669 emit_input_reload_insns, delete_output_reload): Use ira_conflicts_p.
671 2009-01-06 Ben Elliston <bje@au.ibm.com>
673 * gengtype-lex.l (YY_NO_INPUT): Define.
675 2009-01-05 Andrew Pinski <andrew_pinski@playstation.sony.com>
678 * c-common.c (handle_vector_size_attribute): Also reject
681 2009-01-05 Sebastian Pop <sebastian.pop@amd.com>
683 PR tree-optimization/38492
684 * graphite.c (rename_map_elt, debug_rename_elt,
685 debug_rename_map_1, debug_rename_map, new_rename_map_elt,
686 rename_map_elt_info, eq_rename_map_elts,
687 get_new_name_from_old_name, bb_in_sese_p): Moved around.
688 (sese_find_uses_to_rename_use): Renamed sese_build_livein_liveouts_use.
689 (sese_find_uses_to_rename_bb): Renamed sese_build_livein_liveouts_bb.
690 (sese_build_livein_liveouts): New.
691 (new_sese, free_sese): New.
692 (new_scop): Call new_sese.
693 (free_scop): Call free_sese.
694 (rename_variables_from_edge, rename_phis_end_scop): Removed.
695 (register_old_new_names): Renamed register_old_and_new_names.
696 (register_scop_liveout_renames, add_loop_exit_phis,
697 insert_loop_close_phis, struct igp,
698 default_liveout_before_guard, add_guard_exit_phis,
699 insert_guard_phis, copy_renames): New.
700 (translate_clast): Call insert_loop_close_phis and insert_guard_phis.
701 (sese_add_exit_phis_edge): Renamed scop_add_exit_phis_edge.
702 (rewrite_into_sese_closed_ssa): Renamed scop_insert_phis_for_liveouts.
703 (scop_adjust_phis_for_liveouts): New.
704 (gloog): Call scop_adjust_phis_for_liveouts.
706 * graphite.h (struct sese): Documented. Added fields liveout,
708 (SESE_LIVEOUT, SESE_LIVEIN, SESE_LIVEIN_VER, SESE_NUM_VER): New.
709 (new_sese, free_sese, sese_build_livein_liveouts): Declared.
710 (struct scop): Added field liveout_renames.
711 (SCOP_LIVEOUT_RENAMES): New.
713 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
715 PR tree-optimization/38510
716 * graphite.c (recompute_all_dominators): Call mark_irreducible_loops.
717 (translate_clast): Call recompute_all_dominators before
719 (gloog): Call recompute_all_dominators before graphite_verify.
721 2009-01-05 Harsha Jagasia <harsha.jagasia@amd.com>
722 Jan Sjodin <jan.sjodin@amd.com>
724 PR tree-optimization/38500
725 * graphite.c (create_sese_edges): Call fix_loop_structure after
728 2009-01-05 Joel Sherrill <joel.sherrill@oarcorp.com>
730 * config.gcc: Add m32r*-*-rtems*.
731 * config/m32r/rtems.h: New file.
733 2009-01-05 Ben Elliston <bje@au.ibm.com>
735 * Makefile.in (.po.gmo): Use mkinstalldirs, not test -d || mkdir.
737 (po/gcc.pot): Likewise.
739 2009-01-04 David S. Miller <davem@davemloft.net>
741 * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
742 (STARTING_FRAME_OFFSET): Always set to zero.
744 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
746 * tree.def (LSHIFT_EXPR, RSHIFT_EXPR): Add commentary.
747 * tree-cfg.c (verify_gimple_assign_binary): Allow shifts of
748 fixed-point types, and vectors of the same.
750 2009-01-04 Richard Sandiford <rdsandiford@googlemail.com>
752 * config/mips/sync.md (*mb_barrier): Rename to...
753 (*memory_barrier): ...this.
755 2009-01-04 Jonathan Wakely <jwakely.gcc@gmail.com>
757 * doc/extend.texi (Function Attributes): Move @cindex after @item
758 for 'artificial' and 'flatten'. Fix grammar for 'externally_visible'
759 and put in alphabetical order. Fix 'target' name and put in order.
760 * doc/invoke.texi (-Wstrict-null-sentinel, -fipa-matrix-reorg): Fix
763 2009-01-04 Uros Bizjak <ubizjak@gmail.com>
765 * config/s390/s390.md (UNSPEC_MB): Rename from UNSPECV_MB.
766 (memory_barrier): Expand as unspec instead of unspec_volatile.
767 Remove mem:BLK from insn operands. Use Pmode scratch register.
768 (*memory_barrier): Define as unspec instead of unspec_volatile.
769 Use (match_dup 0) as input operand.
771 * config/sparc/sparc.md (UNSPEC_MEMBAR): Rename from UNSPECV_MEMBAR.
772 * config/sparc/sync.md (memory_barrier): Expand as unspec instead of
773 unspec_volatile. Remove mem:BLK from insn operands. Use Pmode
774 scratch register. Remove operand 1.
775 (*stbar): Define as unspec instead of unspec_volatile.
776 Use (match_dup 0) as input operand, remove (const_int 8).
777 (*membar): Define as unspec instead of unspec_volatile.
778 Use (match_dup 0) as input operand, remove input operand 2.
780 * config/xtensa/xtensa.md (UNSPEC_MEMW): Rename from UNSPECV_MEMW.
781 (memory_barrier): Expand as unspec instead of unspec_volatile.
782 Remove mem:BLK from insn operands. Use Pmode scratch register.
783 (*memory_barrier): Define as unspec instead of unspec_volatile.
784 Use (match_dup 0) as input operand.
786 * config/ia64/sync.md (memory_barrier): Redefine as expander pattern.
787 Remove mem:BLK from insn operands. Use Pmode scratch register.
788 Set volatile flag on operand 0.
789 (*memory_barrier): New insn pattern.
791 * config/rs6000/sync.md (memory_barrier): Remove mem:BLK from
793 (*memory_barrier): Use (match_dup 0) as input operand.
795 * config/mips/sync.md (memory_barrier): Redefine as expander pattern.
796 Remove mem:BLK from insn operands. Use Pmode scratch register.
797 Set volatile flag on operand 0.
798 (*mb_internal): New insn pattern.
800 * config/alpha/sync.md (*memory_barrier): Rename from *mb_internal.
802 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
805 * function.c (struct temp_slot): Move to the section of the file
806 that deals with temp slots. Remove field 'address'.
807 (temp_slot_address_table): New hash table of address -> temp slot.
808 (struct temp_slot_address_entry): New struct, items for the table.
809 (temp_slot_address_compute_hash, temp_slot_address_hash,
810 temp_slot_address_eq, insert_temp_slot_address): Support functions
812 (find_temp_slot_from_address): Rewrite to use the new hash table.
813 (remove_unused_temp_slot_addresses): Remove addresses of temp
814 slots that have been made available.
815 (remove_unused_temp_slot_addresses_1): Call-back for htab_traverse,
816 worker function for remove_unused_temp_slot_addresses.
817 (assign_stack_temp_for_type): Don't clear the temp slot address list.
818 Add the temp slot address to the address -> temp slot map.
819 (update_temp_slot_address): Update via insert_temp_slot_address.
820 (free_temp_slots): Call remove_unused_temp_slot_addresses.
821 (pop_temp_slots): Likewise.
822 (init_temp_slots): Allocate the address -> temp slot map, or empty
823 the map if it is already allocated.
824 (prepare_function_start): Initialize temp slot processing.
826 2009-01-04 Steven Bosscher <steven@gcc.gnu.org>
829 * cfgexpand.c (estimate_stack_frame_size): Simplify the estimate:
830 Calculate the size of all stack vars assuming no packing of stack
831 vars will happen, replacing a quadratic algorithm with a linear one.
833 2009-01-03 Jakub Jelinek <jakub@redhat.com>
836 * expmed.c (store_bit_field_1): Don't modify op0 if movstrict insn
839 2009-01-03 Diego Novillo <dnovillo@google.com>
841 * doc/contrib.texi: Update contributions.
843 2009-01-03 Jakub Jelinek <jakub@redhat.com>
846 * builtins.c (fold_builtin_memory_op): Give up if either operand
847 is volatile. Set srctype or desttype to non-qualified version
851 * builtins.c (fold_builtin_expect): Only check DECL_WEAK for VAR_DECLs
854 2009-01-02 Kenneth Zadeck <zadeck@naturalbridge.com>
856 PR rtl-optimization/35805
857 * df-problems.c (df_lr_finalize): Add recursive call to resolve lr
858 problem if fast dce is able to remove any instructions.
859 * dce.c (dce_process_block): Fix dump message.
861 2009-01-02 Mark Mitchell <mark@codesourcery.com>
864 * tree-ssa-pre.c (compute_antic): Correct loop bounds.
866 2009-01-02 Jakub Jelinek <jakub@redhat.com>
869 * tree-flow.h (op_code_prio, op_prio): New prototypes.
870 * tree-pretty-print.c (op_code_prio): New function.
871 (op_prio): No longer static. Use op_code_prio.
872 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs):
873 Use op_prio and op_code_prio to determine if () should be
874 printed around operand(s) or not.
876 * gimple-pretty-print.c (dump_unary_rhs, dump_binary_rhs,
877 dump_gimple_call, dump_gimple_switch, dump_gimple_cond,
878 dump_gimple_label, dump_gimple_try, dump_symbols, dump_gimple_phi,
879 dump_gimple_mem_ops, dump_bb_header, dump_bb_end, pp_cfg_jump): Use
880 pp_character instead of pp_string for single letter printing.
882 2009-01-02 Richard Sandiford <rdsandiford@googlemail.com>
884 * doc/extend.texi: Fix '#pragma GCC option' typo.
886 2009-01-02 Richard Guenther <rguenther@suse.de>
888 * doc/install.texi (--enable-checking): Mention different
890 (--enable-stage1-checking): Document.
892 2009-01-01 Andrew Pinski <pinskia@gmail.com>
895 * tree-cfg.c (verify_expr): Add INDIRECT_REF case. Change MODIFY_EXPR
898 2009-01-02 Ben Elliston <bje@au.ibm.com>
900 * config/fp-bit.h (pack_d): Constify argument.
901 * config/fp-bit.c (makenan): Constify return type. Remove casts.
902 (isnan): Constify argument.
906 (_fpadd_parts): Constify return type.
907 (_fpmul_parts): Likewise.
908 (_fpdiv_parts): Likewise.
910 2009-01-01 Jakub Jelinek <jakub@redhat.com>
913 * c-typeck.c (add_pending_init): Add IMPLICIT argument. Only
914 warn about overwriting initializer with side-effects or
915 -Woverride-init if !IMPLICIT.
916 (output_init_element): Likewise. Pass IMPLICIT down to
918 (process_init_element): Add IMPLICIT argument. Pass it down
919 to output_init_element.
920 (push_init_element, pop_init_level, set_designator): Adjust
921 process_init_element callers.
922 (set_nonincremental_init, set_nonincremental_init_from_string):
923 Adjust add_pending_init callers.
924 (output_pending_init_elements): Adjust output_init_element callers.
925 * c-tree.h (process_init_element): Adjust prototype.
926 * c-parser.c (c_parser_initelt, c_parser_initval): Adjust
927 process_init_element callers.