1 2005-02-06 Joseph S. Myers <joseph@codesourcery.com>
4 * doc/invoke.texi: Mention for each option included in -Wall that
5 it is included in -Wall.
6 * doc/gcc.texi: Update copyright and last modification date.
8 2005-02-06 Zdenek Dvorak <dvorakz@suse.cz>
10 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Do not add
11 unnecessary cast to original induction variable increments.
13 2005-02-06 Zdenek Dvorak <dvorakz@suse.cz>
15 PR tree-optimization/18219
16 * tree-ssa-loop-ivopts.c (get_computation_at): Produce computations
19 2005-02-06 Richard Sandiford <rsandifo@redhat.com>
21 * expmed.c (store_bit_field): Make the SUBREG code adjust bitnum.
22 Set bitpos and offset later in the function. Do nothing if the
23 target is a register and if the bitfield lies completely outside
25 (extract_bit_field): Make the same SUBREG, bitpos and offset changes
26 here. Return an uninitialised register if the source value is stored
27 in a register and the bitfield lies completely outside that register.
29 2005-02-06 Steven Bosscher <stevenb@suse.de>
31 * df.c (df_insn_refs_record): Use XEXP to get the operand of a USE,
34 2005-02-06 Marcin Dalecki <martin@dalecki.de>
36 * diagnostic.c (default_diagnostic_finalizer): Fix use of attribute.
38 2005-02-06 Geoffrey Keating <geoffk@apple.com>
40 * config.gcc (*-*-darwin*): Separate darwin7 and darwin8 configs.
41 * config/darwin8.h: New.
42 * config/darwin7.h: Update comment.
44 2005-02-03 Leehod Baruch <leehod@il.ibm.com>
45 Dorit Naishlos <dorit@il.ibm.com>
47 * tree-vectorizer.h (LOC): New type.
48 (UNKNOWN_LOC, EXPR_LOC, LOC_FILE, LOC_LINE): New macros.
49 (loop_line_number): New field in struct _loop_vec_info.
50 (LOOP_VINFO_LOC, LOOP_LOC): New macros.
52 * tree-vectorizer.c (input.h): Included.
53 (find_loop_location): New function.
54 (vect_debug_stats): Argument changed from loop to LOC. Computation of
55 loop line number removed.
56 (vect_debug_details): Likewise.
57 (new_loop_vec_info): Initialize new field LOOP_VINFO_LOC.
59 (slpeel_make_loop_iterate_ntimes): Call find_loop_location. Argument in
60 call to vect_debug_details/stats changed from loop to loop_loc.
61 (slpeel_tree_duplicate_loop_to_edge_cfg): Likewise.
62 (slpeel_tree_peel_loop_to_edge): Likewise.
64 (vect_analyze_offset_expr): Argument in call to vect_debug_details/stats
65 changed from NULL to UNKNOWN_LOC.
66 (vect_get_base_and_offset):
67 (vect_create_addr_base_for_vector_ref):
68 (get_vectype_for_scalar_type):
69 (vect_create_data_ref_ptr):
71 (vect_get_vec_def_for_operand):
72 (vect_finish_stmt_generation):
73 (vectorizable_assignment):
74 (vectorizable_operation):
77 (vect_transform_stmt):
78 (vect_update_ivs_after_vectorizer):
79 (vect_do_peeling_for_loop_bound):
80 (vect_do_peeling_for_alignment):
81 (vect_transform_loop):
83 (vect_analyze_operations):
84 (vect_is_simple_iv_evolution):
85 (vect_analyze_scalar_cycles):
86 (vect_analyze_data_ref_dependences):
87 (vect_compute_data_ref_alignment):
88 (vect_enhance_data_refs_alignment):
89 (vect_analyze_data_ref_access):
90 (vect_analyze_data_ref_accesses):
91 (vect_analyze_pointer_ref_access):
92 (vect_get_memtag_and_dr):
93 (vect_analyze_data_refs):
95 (vect_stmt_relevant_p):
96 (vect_mark_stmts_to_be_vectorized):
97 (vect_can_advance_ivs_p):
98 (vect_get_loop_niters):
100 (vectorize_loops): Likewise.
102 (vectorizable_load): Argument in call to vect_debug_details/stats
103 changed from loop to LOOP_LOC (loop_vinfo).
104 (vect_transform_loop):
105 (vect_analyze_operations):
106 (vect_analyze_scalar_cycles):
107 (vect_analyze_data_ref_dependence):
108 (vect_enhance_data_refs_alignment):
109 (vect_analyze_data_ref_accesses):
110 (vect_analyze_pointer_ref_access):
111 (vect_analyze_data_refs):
112 (vect_analyze_loop): Likewise.
114 (vect_analyze_loop_form): Argument in call to vect_debug_details/stats
115 changed from loop to loop_loc.
117 (vect_enhance_data_refs_alignment): Removed unused variable loop.
119 2005-02-03 Leehod Baruch <leehod@il.ibm.com>
120 Dorit Naishlos <dorit@il.ibm.com>
122 * tree-vectorizer.c (vect_analyze_data_ref_dependence): Function
124 (vect_analyze_data_ref_dependences): Likewise.
126 (vect_is_simple_use): Argument changed from loop structure to
127 loop_vect_info structure.
128 (vect_can_advance_ivs_p): Likewise.
129 (vect_create_index_for_vector_ref): Likewise.
130 (vect_update_ivs_after_vectorizer): Likewise.
131 (new_stmt_vec_info): Likewise.
133 (new_loop_vec_info): Second argument in call to new_stmt_vec_info was
134 changed from loop to loop_vinfo.
135 (vect_create_data_ref_ptr): First argument in call to
136 vect_create_index_for_vector_ref was changed from loop to loop_vinfo.
137 (vectorizable_assignment): Second argument in call to vect_is_simple_use
138 was changed from loop to loop_vinfo.
139 (vectorizable_operation): Likewise.
140 (vectorizable_store): Likewise.
141 (vect_mark_stmts_to_be_vectorized): Likewise.
142 (vect_do_peeling_for_loop_bound): First argument in call to
143 vect_update_ivs_after_vectorizer was changed from loop to loop_vinfo.
144 (vect_analyze_operations): Argument in call to vect_can_advance_ivs_p
145 was changed from loop to loop_vinfo.
146 (vect_analyze_data_ref_dependences): Third argument in call to
147 vect_analyze_data_ref_dependence was changed from loop to loop_vinfo.
149 (vect_create_index_for_vector_ref): Get the loop from loop_vinfo.
150 (vect_create_data_ref_ptr): Likewise.
151 (vect_init_vector): Likewise.
152 (vect_get_vec_def_for_operand): Likewise.
153 (vectorizable_assignment): Likewise.
154 (vectorizable_operation): Likewise.
155 (vectorizable_store): Likewise.
156 (vectorizable_load): Likewise.
157 (vect_update_ivs_after_vectorizer): Likewise.
158 (vect_is_simple_use): Likewise.
159 (vect_analyze_data_ref_dependence): Likewise.
160 (vect_analyze_pointer_ref_access): Likewise.
161 (vect_can_advance_ivs_p): Likewise.
163 * tree-vectorizer.h: (_loop_vec_info): Definition and macros moved
164 before _stmt_vec_info.
165 (_stmt_vec_info): Loop field replaced by loop_vec_info.
166 (STMT_VINFO_LOOP): Replaced with STMT_VINFO_LOOP_VINFO.
167 (new_stmt_vec_info): Argument changed from loop structure to
168 loop_vect_info structure
170 (vect_analyze_data_ref_dependences): Unnecessary line was removed.
171 (vect_analyze_offset_expr): Avoid 80 columns overflow.
172 (vect_create_addr_base_for_vector_ref): Likewise.
173 (vect_analyze_pointer_ref_access): Likewise.
175 2005-02-03 Andrew Pinski <pinskia@physics.uc.edu>
178 * tree-ssa-dse.c (fix_phi_uses): Update the occurs in abnormal
179 phi flag if the phi is abnormal.
181 2005-02-03 Andrew Pinski <pinskia@physics.uc.edu>
184 * tree-ssa.c (replace_immediate_uses): Update the immediate_uses
185 information for the new statement.
187 2005-02-03 Joseph S. Myers <joseph@codesourcery.com>
190 * gcc.c (main): Include the target in -v output.
192 2005-02-03 Alexandre Oliva <aoliva@redhat.com>
194 * combine.c (simplify_comparison, case SIGN_EXTEND, ZERO_EXTEND):
195 Do not drop the extend if we'd have to add a paradoxical subreg
196 later. Include optabs.h and insn-codes.h.
197 * Makefile.in (combine.o): Depend on $(OPTABS_H).
199 2005-02-02 Roger Sayle <roger@eyesopen.com>
202 * real.h (REAL_MODE_FORMAT_COMPOSITE_P): New macro.
203 * fold-const.c (const_binop): Avoid constant folding floating
204 point operations in modes that use composite representations.
205 * simplify-rtx.c (simplify_binary_operation): Likewise.
207 2005-02-02 Geoffrey Keating <geoffk@apple.com>
209 * config/rs6000/altivec.md (altivec_dst): Make the first operand
211 (altivec_dstt): Likewise.
212 (altivec_dstst): Likewise.
213 (altivec_dststt): Likewise.
214 * config/rs6000/rs6000.c (altivec_expand_dst_builtin): Adjust creation
217 2005-02-03 Dorit Naishlos <dorit@il.ibm.com>
219 * tree-vectorizer.c (slpeel_make_loop_iterate_ntimes): Copy
220 EXPR_LOCUS from orig_cond to the new cond_stmt.
221 (vect_finish_stmt_generation): Copy EXPR_LOCUS from stmt to
224 2005-02-03 Dorit Naishlos <dorit@il.ibm.com>
226 * tree-vectorizer.c (slpeel_make_loop_iterate_ntimes): Call
227 standard_iv_increment_position. Remove call to bsi_next
228 (no need to bump the iterator anymore).
229 (vect_create_index_for_vector_ref): Call
230 standard_iv_increment_position. Remove second function argument.
231 (vect_finish_stmt_generation): Remove call to bsi_next
232 (no need to bump the iterator anymore).
233 (vect_create_data_ref_ptr): Remove second argument (bsi) in call
234 to vect_create_index_for_vector_ref.
236 2005-02-03 Dorit Naishlos <dorit@il.ibm.com>
238 * tree-vectorizer.c (vect_mark_stmts_to_be_vectorized): Check if
239 phis are used out of the loop.
241 2005-02-03 Joseph S. Myers <joseph@codesourcery.com>
244 * c-decl.c (undef_nested_function): New variable.
245 (pop_scope): Diagnose undefined nested functions.
246 (finish_function): Don't attempt cgraph processing or genericizing
247 if current top-level function contained an undefined nested
248 function. Reset undef_nested_function at the end of a top-level
251 2005-02-02 Zdenek Dvorak <dvorakz@suse.cz>
253 * tree.c (build_int_cst_type): Take sign of the value into account
254 when deciding whether sign extend the value.
256 2005-02-02 Joseph S. Myers <joseph@codesourcery.com>
259 * gcc.c (cpp_unique_options): Remove %{trigraphs}.
260 (cpp_options, cc1_options): Change %{std*&ansi} to
261 %{std*&ansi&trigraphs}.
263 2005-02-02 Kazu Hirata <kazu@cs.umass.edu>
265 * cse.c: Update comments.
267 2005-02-02 Aldy Hernandez <aldyh@redhat.com>
269 * config/rs6000/rs6000.c (rs6000_spe_function_arg): Multi-reg
270 arguments go on the stack. Do not put complex doubles in odd/even
273 2005-02-02 Steven Bosscher <stevenb@suse.de>
275 * doc/invoke.texi: Update the options enabled at -O1, -O2,
276 and -O3. Update the documentation for -fgcse-sm, -fgcse-las,
277 -ftree-pre, and -fthread-jumps.
279 2005-02-02 Richard Sandiford <rsandifo@redhat.com>
281 PR tree-optimization/19578
282 * tree-flow.h (modified_noreturn_calls): Declare.
283 (noreturn_call_p): Declare.
284 * tree-flow-inline.h (noreturn_call_p): New function.
285 (modify_stmt): Add modified noreturn calls to modified_noreturn_calls.
286 * tree-cfg.c (modified_noreturn_calls): New variable.
287 (cleanup_control_flow): Use noreturn_call_p. Split basic blocks
288 that contain a mid-block noreturn call.
289 * tree-ssa.c (delete_tree_ssa): Clear modified_noreturn_calls.
291 2005-02-02 Kazu Hirata <kazu@cs.umass.edu>
293 * df.c (df_def_record_1, df_uses_record): Don't use
295 * df.h (DF_FOR_REGALLOC): Remove.
297 2005-02-02 Joseph S. Myers <joseph@codesourcery.com>
300 * c-typeck.c (really_start_incremental_init): Reset
301 constructor_max_index for arrays of incomplete type.
303 2005-02-02 Jeff Law <law@redhat.com>
305 * gcse.c (struct reg_set): Store the block index where the register
306 was set rather than the setting insn.
307 (record_one_set, compute_transp): Corresponding changes.
308 (pre_insert_copy_insn): Do not call replace_one_set.
309 (replace_one_set): Remove.
311 2005-02-02 Joseph S. Myers <joseph@codesourcery.com>
314 * doc/gcov.texi: Fix typo.
316 2005-02-01 Richard Henderson <rth@redhat.com
319 * config/i386/i386.h (MODES_TIEABLE_P): Use ix86_modes_tieable_p.
320 * config/i386/i386.c (ix86_hard_regno_mode_ok): Change return
322 (ix86_tieable_integer_mode_p, ix86_modes_tieable_p): New.
323 * config/i386/i386-protos.h: Update.
325 2005-02-01 Steven Bosscher <stevenb@suse.de>
327 PR tree-optimization/19217
328 * tree-cfg.c (verify_expr): Use the data field to see if TP was
329 seen inside a PHI node. Do not do the ADDR_EXPR check if it was.
330 (verify_stmts): Pass (void*)1 as data to verify_expr to signal
331 that it is walking a PHI node.
333 2005-02-01 Joseph S. Myers <joseph@codesourcery.com>
335 * doc/extend.texi (Nested Functions): Update.
337 2005-02-01 Richard Henderson <rth@redhat.com>
340 * optabs.c (expand_copysign_absneg): Export.
341 * optabs.h (expand_copysign_absneg): Declare.
342 * config/rs6000/rs6000.md (copysigntf3): New.
344 2005-02-01 Diego Novillo <dnovillo@redhat.com>
346 PR tree-optimization/19633
347 * tree-flow.h (struct ptr_info_def): Add field 'pt_null'.
348 * tree-ssa-alias.c (init_alias_info): Initialize.
349 (merge_pointed_to_info): Set.
350 (add_pointed_to_expr): Set pt_null if EXPR is a NULL pointer.
351 (dump_points_to_info_for): Show value of pt_null.
352 (struct count_ptr_d): Declare.
353 (find_ptr_dereference): Remove.
354 (ptr_is_dereferenced_by): Remove.
355 (count_ptr_derefs): New local function.
356 (count_uses_and_derefs): New local function.
357 (compute_points_to_and_addr_escape): Call it. If the number
358 of dereferences is greater than zero, mark the pointer as
359 dereferenced. If there are fewer dereferences than uses of
360 the pointer, the pointer's value escapes.
362 2005-02-01 Diego Novillo <dnovillo@redhat.com>
364 PR tree-optimization/19670
365 * tree-ssa.c (verify_ssa_name): Don't set TREE_VISITED
367 (verify_use): ... set it here, instead.
369 2005-02-01 Janis Johnson <janis187@us.ibm.com>
371 * doc/sourcebuild.texi (Test directives): Fix formatting.
373 * doc/sourcebuild.texi (Test directives): Describe selector
376 2005-02-01 Segher Boessenkool <segher@kernel.crashing.org>
378 * config.gcc (powerpc64-*-linux*): Default to -m64 also for
379 specific 64-bit CPUs.
381 2005-02-01 Richard Earnshaw <rearnsha@arm.com>
384 * arm.c (arm_eliminable_register): New function.
385 (adjacent_mem_locations): Don't allow eliminable registers. Use
386 HOST_WIDE_INT for address offsets.
387 * arm-protos.h (arm_eliminable_register): Add prototype.
389 2005-02-01 Steven Bosscher <stevenb@suse.de>
391 PR optimization/15242
392 * params.def (PARAM_MAX_GOTO_DUPLICATION_INSNS): New param.
393 * basic-block.h (duplicate_computed_gotos): Add prototype.
394 * bb-reorder.c (duplicate_computed_gotos): New function to
395 duplicate sufficiently small blocks ending in a computed jump.
396 * passes.c (rest_of_compilation): Call duplicate_computed_gotos
397 if not optimizing for size.
398 * cfgcleanup.c (try_crossjump_bb): If not optimizing for size,
399 never do tail merging for blocks ending in a computed jump.
400 * doc/invoke.texi: Document the max-goto-duplication-insns param.
402 2005-02-01 Eric Botcazou <ebotcazou@libertysurf.fr>
404 Patch from Richard Sandiford <rsandifo@redhat.com>
405 * reload1.c (choose_reload_regs): Prevent the find_equiv_reg() code
406 from inheriting a subreg equivalence with a non-spill register.
408 * tree.h (DECL_FUNCTION_CODE): Document that it is overloaded.
410 2005-02-01 Alexandre Oliva <aoliva@redhat.com>
412 * config/frv/frv.c (movcc_fp_destination_operand): New.
413 (gpr_or_memory_operand): Fix typo in comment.
414 (gpr_or_memory_operand_with_scratch): New.
415 * config/frv/frv.h (PREDICATE_CODES): Add the two new predicates.
416 * config/frv/frv.md (movcc_fp, movcc_fp_internal): Use
417 movcc_fp_destination_operand.
418 (reload_incc_fp): Use gpr_or_memory_operand_with_scratch.
419 Legitimize memory addresses using a scratch register.
421 2005-01-31 Jeff Law <law@redhat.com>
423 * tree-into-ssa.c (mark_def_sites_global_data): Make KILLS
424 bitmap a sparse bitmap instead of a simple bitmap.
425 (mark_def_sites_initialize_block): Corresponding changes.
426 (ssa_mark_def_sites_initialize_block): Likewise.
427 (ssa_mark_phi_uses): Likewise.
428 (mark_def_site, ssa_mark_def_sites): Likewise.
429 (mark_def_site_blocks): Likewise.
430 (rewrite_ssa_into_ssa): Likewise.
432 * tree-ssa-dom.c (record_cond): Pass correct variable type
433 for last argument to htab_find_slot_with_hash.
435 * fold-const.c (fold, case CEIL_MOD_EXPR): Do not lose side
436 effects when optimizing 0 % X. Do not try to optimize X % 0.
438 2005-01-31 James E. Wilson <wilson@specifixinc.com>
440 * config/ia64/itanium1.md (1_scall bypass): Change 2_mmalua to
443 2005-02-01 Eric Christopher <echristo@redhat.com>
445 * config/mips/mips.c (override_options): Warn if -mint64
447 * doc/invoke.texi (MIPS Options): Document that -mint64 is
450 2005-02-01 Kazu Hirata <kazu@cs.umass.edu>
452 * cse.c (cse_reg_info): Remove hash_next, next, regno. Add
454 (cse_reg_info_list, cse_reg_info_list_free, REGHASH_SHIFT,
455 REGHASH_SIZE, REGHASH_MASK, reg_hash, REGHASH_FN,
456 cached_cse_reg_info, GET_CSE_REG_INFO): Remove.
457 (cached_regno): Initialize to INVALID_REGNUM.
458 (cse_reg_info_table_size,
459 cse_reg_info_table_first_uninitialized,
460 cse_reg_info_timestamp): New.
461 (REG_TICK, REG_IN_TABLE, SUBREG_TICKED, REG_QTY): Use
463 (init_cse_reg_info, get_cse_reg_info_1): New.
464 (get_cse_reg_info): Cache the last look-up.
465 (new_basic_block): Update the code to clear mappings from
466 registers to cse_reg_info entries.
467 (cse_main): Call init_cse_reg_info.
469 * cse.c (get_cse_reg_info): Update a comment.
471 2005-01-31 Steven Bosscher <stevenb@suse.de>
474 * c-decl.c (start_decl): Do not warn about arrays of elements with
475 an incomplete type here.
476 (grokdeclarator): Do it here by making a pedwarn an error.
477 * c-typeck.c (push_init_level): If there were previous errors with
478 the constructor type, do not warn about braces for initializers.
479 (process_init_element): Likewise for excess initializer elements.
481 2005-01-31 Kazu Hirata <kazu@cs.umass.edu>
483 * cse.c (delete_trivially_dead_insn): Don't iterate.
485 2005-01-31 Andrew Pinski <pinskia@physics.uc.edu>
487 * config/rs6000/rs6000.md (copysignsf3): New expand.
488 (copysigndf3): Likewise.
490 2005-01-31 Steven Bosscher <stevenb@suse.de>
492 * recog.c (constrain_operands): Only look for earlyclobber operand
493 conflicts if an '&' constraint was seen.
495 2005-01-31 Marc Espie <espie@openbsd.org>
497 * config.gcc: Don't include embedded systems fragment, switches default
498 debugging format to ELF.
499 * config/i386/openbsdelf.h: Add DBX_REGISTER_NUMBER since we no
500 longer pick it up there.
502 2005-01-31 Dale Johannesen <dalej@apple.com>
504 * doc/extend.texi (nested functions): Fix linkage description.
505 Clarify that static is not allowed.
507 2005-01-31 Dale Johannesen <dalej@apple.com>
509 * config/rs6000/darwin.md (movsf_low_di): Make work.
510 (movdf_low_di): Make work.
512 2005-01-31 Dale Johannesen <dalej@apple.com>
514 * config/rs6000/darwin-tramp.asm (__trampoline_setup):
515 Make work for 64 bit.
517 2005-01-31 Roger Sayle <roger@eyesopen.com>
518 Dale Johannesen <dalej@apple.com>
521 * fold-const.c (fold_binary_op_with_conditional_arg):
522 Make types match original operands, before STRIP_NOPS.
524 2005-01-31 Alan Modra <amodra@bigpond.net.au>
526 * config/rs6000/linux-unwind.h (struct gcc_vregs): New.
527 (struct gcc_regs): Rename from gcc_pt_regs. Add more regs.
528 (struct gcc_sigcontext): Delete. Merge contents to..
529 (struct gcc_ucontext): ..here.
530 (get_sigcontext): Delete.
531 (get_regs): New function, like get_sigcontext but return regs ptr.
532 64-bit version finds regs from r1 to support vdso.
533 (ppc_linux_aux_vector): New function.
534 (ppc_fallback_frame_state): Modify for get_regs. Restore fprs
537 2005-01-31 Ira Rosen <irar@il.ibm.com>
539 * tree-vectorizer.c (vect_analyze_offset_expr): Use ssizetype instead
541 (vect_get_base_and_offset): Use ssizetype instead sizetype. Remove
542 redundant fold. Fix misalignment for MINUS_EXPR.
543 (vect_compute_data_ref_alignment): Use ssizetype instead sizetype.
544 (vect_analyze_pointer_ref_access): Likewise.
545 (vect_get_memtag_and_dr): Likewise.
547 2005-01-31 Richard Henderson <rth@redhat.com>
549 * config/i386/i386.c (absnegsf2_mixed): Remove all of the # hackery.
550 (absnegsf2_sse, absnegsf2_i387, absnegdf2_mixed, absnegdf2_sse,
551 absnegdf2_i387, absnegxf2_i387): Likewise.
553 2005-01-31 Richard Henderson <rth@redhat.com>
556 * optabs.c (expand_copysign_absneg): New.
557 (expand_copysign_bit): Split out from ...
558 (expand_copysign): ... here. Use expand_copysign_absneg.
560 2005-01-30 Roger Sayle <roger@eyesopen.com>
563 * config/pa/pa.md (anddi3, iordi3): On HPPA64, disallow an integer
564 constant as the second operand and a register as the third.
566 2005-01-31 Danny Smith <dannysmith@users.sourceforge.net>
569 * config/i386/i386.c (ix86_function_ok_for_sibcall): Also check
570 that dllimport'd functions do not use all call-clobbered registers
573 2005-01-30 Richard Henderson <rth@redhat.com>
576 * optabs.c (expand_absneg_bit): Split out from expand_unop and
577 expand_abs_nojump and generalize. Use operand_subword and
578 emit_no_conflict_block. Support large modes.
579 (expand_abs_nojump): Use it. Use HONOR_SIGNED_ZEROS to fall
581 (expand_unop): Likewise. Use HONOR_SIGNED_ZEROS to fall back
582 negation to subtraction.
583 (expand_copysign): Check that the format has signed zeros.
584 Use operand_subword and emit_no_conflict_block. Support large modes.
586 2005-01-30 Daniel Berlin <dberlin@dberlin.org>
588 Fix PR tree-optimization/19624
590 * Makefile.in (tree-ssa-pre.o): Add CFGLOOP_H.
591 * tree-ssa-pre.c: Add cfgloop.h.
593 (pre_stats): New member, constified.
594 (inserted_exprs): New static variable.
595 (NECESSARY): New macro.
596 (create_expression_by_pieces): Fold the expression, and
597 mark it as defaulting to not necessary. Also put in
599 (fully_constant_expression): New function.
600 (insert_into_preds_of_block): Modify to not insert phis when we
601 are playing with induction variables.
602 Push phis onto the inserted_exprs vector, and mark them as not
603 necessary by default.
604 (insert_aux): Call fully_constant_expression on eprime.
605 If all edges produce the same value, mark it constant.
606 (mark_operand_necessary): New function.
607 (remove_dead_inserted_code): New function.
608 (init_pre): Init loop optimizer to get loop info.
609 (fini_pre): Free loop_optimizer, and inserted_exprs vec.
610 (execute_pre): Commit edge inserts, then remove dead code.
612 2005-01-30 Richard Henderson <rth@redhat.com>
614 * rtl.c (rtx_equal_p): No early exit for CONST_VECTOR.
615 * varasm.c (const_rtx_hash_1): Handle CONST_VECTOR.
617 2005-01-30 Richard Henderson <rth@redhat.com>
620 * config/i386/i386.c (ix86_expand_copysign): New.
621 (ix86_split_copysign_const): New.
622 (ix86_split_copysign_var): Rename from ix86_split_copysign,
623 rearrange op1/nmask operands.
624 * config/i386/i386-protos.h: Update.
625 * config/i386/i386.md (copysignsf3): Use ix86_expand_copysign.
626 (copysigndf3): Likewise.
627 (copysignsf3_const, copysigndf3_const): New.
628 (copysignsf3_var): Rename from copysignsf3, split out splitter
629 and fix split predicate for X constraint.
630 (copysigndf3_var): Similarly.
632 2005-01-30 Kazu Hirata <kazu@cs.umass.edu>
634 * optabs.c, doc/c-tree.texi, doc/install.texi, doc/md.texi,
635 doc/passes.texi, doc/rtl.texi, doc/sourcebuild.texi,
636 doc/tm.texi, doc/tree-ssa.texi: Update copyright.
638 2005-01-29 Richard Henderson <rth@redhat.com>
641 * config/i386/i386.md (movdf_nointeger, movdf_integer): Fix ordering
642 of # and * in constraints.
644 2005-01-29 Richard Henderson <rth@redhat.com>
647 * expr.c (store_field): Don't strip sub-mode cast when the input
648 data is even smaller.
650 2005-01-29 Richard Henderson <rth@redhat.com>
653 * expr.c (categorize_ctor_elements_1): Check for CONSTRUCTOR of a
656 2005-01-29 Richard Henderson <rth@redhat.com>
658 * combine.c (make_field_assignment): Fix argument order
661 2005-01-29 Richard Guenther <richard.guenther@uni-tuebingen.de>
663 PR tree-optimization/15791
664 * fold-const.c (extract_array_ref): New function.
665 (fold): Fold comparisons between &a[i] and &a[j] or
666 semantically equivalent trees.
668 2005-01-29 Jeff Law <law@redhat.com>
670 * gcse.c (insert_expr_in_table): Revamp handling of available
671 and anticipatable occurrence lists to avoid unnecessary list
673 (insert_set_in_table): Similarly.
675 2005-01-29 Joseph S. Myers <joseph@codesourcery.com>
677 * c-common.c (fix_string_type): Just use c_build_qualified_type to
679 (c_build_qualified_type): Build qualified array types with
680 TYPE_MAIN_VARIANT pointing to corresponding unqualified type.
681 * c-decl.c (c_make_fname_decl): Build unqualified array type
682 before qualified type.
683 (grokdeclarator): Use TYPE_MAIN_VARIANT of typedef type if element
684 type is qualified, not just if type itself is. Don't apply
685 qualifiers to array type when declarator is processed. Apply
686 qualifiers to field type whether or not it is an array type.
687 Don't handle array types specially for applying qualifiers to
689 * c-typeck.c (composite_type): Build unqualified element type and
690 array type when forming composite of array types.
691 (common_pointer_type, comptypes, comp_target_types,
692 type_lists_compatible_p, build_indirect_ref, build_array_ref,
693 convert_for_assignment): Don't apply TYPE_MAIN_VARIANT to array
695 (type_lists_compatible_p): Cache TREE_VALUE (args1) and TREE_VALUE
696 (args2) in variables a1 and a2.
698 2005-01-29 Kazu Hirata <kazu@cs.umass.edu>
700 * cse.c (n_elements_made, max_elements_made): Remove.
701 (insert): Don't touch n_elements_made.
702 (cse_main): Don't touch n_elements_made or max_elements_made.
704 2005-01-28 Stephane Carrez <stcarrez@nerim.fr>
707 * config/m68hc11/t-m68hc11-gas (dp-bit.c): Fix typo causing a
708 configuration part of dp-bit.c to be lost.
710 2005-01-28 Roger Sayle <roger@eyesopen.com>
712 * expmed.c (expand_mult_highpart_optab): When attempting to use
713 a non-widening multiplication in a wider mode, the operands need
714 to be converted (zero or sign extended) to that mode.
716 2005-01-28 Ian Lance Taylor <ian@airs.com>
720 * gimple-low.c (block_may_fallthru): TRY_FINALLY_EXPR only falls
721 through if both operands fall through.
723 2005-01-28 Kazu Hirata <kazu@cs.umass.edu>
725 * cse.c (fold_rtx) <PC>: Don't optimize.
727 2005-01-28 Jeff Law <law@redhat.com>
729 * fold-const.c (fold, case CEIL_MOD_EXPR): Optimize 0 % X.
730 (case FLOOR_MOD_EXPR, ROUND_MOD_EXPR, TRUNC_MOD_EXPR): Similarly.
732 2005-01-28 Kazu Hirata <kazu@cs.umass.edu>
734 * cse.c (cse_reg_info_free_list, cse_reg_info_used_list,
735 cse_reg_info_used_list_end): Remove.
736 (cse_reg_info_list, cse_reg_info_list_free): New.
737 (get_cse_reg_info): When allocating an instance of
738 cse_reg_info, add it to the beginning of the cse_reg_info_list
739 list. Remove code to maintain cse_reg_info_used_list.
740 (new_basic_block): Reset the free list to the beginning of
743 2005-01-28 Richard Sandiford <rsandifo@redhat.com>
745 * config/frv/frv.c (bdesc_2arg): Remove __MDPACKH.
746 (frv_init_builtins): Change its prototype to take 4 uhalf arguments.
747 (frv_expand_mdpackh_builtin): New function.
748 (frv_expand_builtin): Use it to expand __MDPACKH.
749 * doc/extend.texi (__MDPACKH): Update documentation.
751 2005-01-28 Uros Bizjak <uros@kss-loka.si>
753 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use
754 SSE_FLOAT_MODE_P for use_sse computation.
756 2005-01-28 Ralf Corsepius <ralf.corsepius@rtems.org>
759 * config/sparc/rtemself.h (LINK_GCC_C_SEQUENCE_SPEC): undefine.
761 2005-01-27 Richard Henderson <rth@redhat.com>
763 * builtins.c (expand_builtin_copysign): New.
764 (expand_builtin): Call it.
765 * genopinit.c (optabs): Add copysign_optab.
766 * optabs.c (init_optabs): Initialize it.
767 (expand_copysign): New.
768 * optabs.h (OTI_copysign, copysign_optab): New.
769 (expand_copysign): Declare.
771 * config/alpha/alpha.md (UNSPEC_COPYSIGN): New.
772 (copysignsf3, ncopysignsf3, copysigndf3, ncopysigndf3): New.
774 * config/i386/i386.c (ix86_build_signbit_mask): Split from ...
775 (ix86_expand_fp_absneg_operator): ... here.
776 (ix86_split_copysign): New.
777 * config/i386/i386-protos.h: Update.
778 * config/i386/i386.md (UNSPEC_COPYSIGN): New.
779 (copysignsf3, copysigndf3): New.
781 * config/ia64/ia64.md (UNSPEC_COPYSIGN): New.
782 (copysignsf3, ncopysignsf3): New.
783 (copysigndf3, ncopysigndf3): New.
784 (copysignxf3, ncopysignxf3): New.
785 * config/ia64/ia64.c (rtx_needs_barrier): Handle UNSPEC_COPYSIGN.
787 2005-01-27 Arend Bayer <arend.bayer@web.de>
788 Kazu Hirata <kazu@cs.umass.edu>
790 * cse.c: (find_best_addr): Don't call copy_rtx before calling
791 fold_rtx. Save cost recomputation if fold_rtx did nothing.
792 (fold_rtx) <ASM_OPERANDS>: Don't do anything if INSN is
795 2005-01-27 Jeff Law <law@redhat.com>
797 * tree-into-ssa.c (ssa_rewrite_initialize_block): Do not register
798 definitions of SSA_NAMEs which are not being rewritten.
799 (rewrite_ssa_into_ssa): Only initialize the current definition
800 of an SSA_NAME if that SSA_NAME has been marked for rewriting.
801 If checking is enabled, assert that SSA_NAME_AUX is clear for all
802 SSA_NAMEs before returning.
804 2005-01-27 Kazu Hirata <kazu@cs.umass.edu>
806 * c-common.def, c-dump.c, c-gimplify.c, c-objc-common.c,
807 cfgexpand.c, dbxout.c, function.h, opts.c, tree-flow-inline.h,
808 tree-ssa-operands.h, tree-ssa-threadupdate.c, tree-tailcall.c,
809 config/avr/avr.md, config/cris/aout.h, config/cris/cris.h,
810 config/mips/iris6.h, config/sh/sh.c: Update copyright.
812 2005-01-27 Steven Bosscher <stevenb@suse.de>
815 * opts.c (decode_options): Move flag_thread_jumps from -O1 and
816 higher to -O2 and higher. Likewise for tree PRE.
817 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Only iterate at -O2
820 2005-01-27 Ian Lance Taylor <ian@airs.com>
823 * gimple-low.c (try_catch_may_fallthru): New static function.
824 (block_may_fallthru): Handle TRY_CATCH_EXPR.
825 * tree-inline.c (expand_call_inline): Don't warn about reaching
826 the end of a non-void function being inlined if the function uses
829 2005-01-27 Jakub Jelinek <jakub@redhat.com>
831 * config/i386/i386.h (CALL_USED_REGISTERS): Fix comment pastos.
834 * c-decl.c (warn_if_shadowing): Handle old_decl error_mark_node.
835 (pushdecl): Only use DECL_FILE_SCOPE_P if DECL_P.
836 (implicitly_declare): Handle error_mark_node.
838 2005-01-27 Richard Henderson <rth@redhat.com>
841 * tree.h (struct tree_decl): Add debug_expr_is_from.
842 (DECL_DEBUG_EXPR_IS_FROM): New.
843 (DECL_DEBUG_EXPR): Rename from DECL_DEBUG_ALIAS_OF.
844 * dwarf2out.c (dwarf2out_var_location): Update to match.
845 * tree-outof-ssa.c (create_temp): Likewise.
846 * var-tracking.c (track_expr_p): Likewise.
847 * tree-sra.c (instantiate_element): Set DECL_DEBUG_EXPR.
848 * c-objc-common.c (c_tree_printer) <'D'>: Handle DECL_DEBUG_EXPR.
849 * toplev.c (default_tree_printer): Likewise.
851 2005-01-27 Alexandre Oliva <aoliva@redhat.com>
853 * config/frv/frv.c (frv_class_likely_spilled_p): Add GR8_REGS,
854 GR9_REGS, GR89_REGS, FDPIC_FPTR_REGS, FDPIC_REGS.
856 2005-01-27 Steven Bosscher <stevenb@suse.de>
858 * tree.h (SWITCH_COND, SWITCH_BODY, SWITCH_LABELS, CASE_LOW,
859 CASE_HIGH, CASE_LABEL, ASM_STRING, ASM_OUTPUTS, ASM_INPUTS,
860 ASM_CLOBBERS): Add tree checks.
862 * c-common.h (SWITCH_TYPE): Rename to SWITCH_STMT_TYPE.
863 (SWITCH_STMT_COND, SWITCH_STMT_BODY): New.
864 * c-common.def (SWITCH_STMT): Update to match.
865 * c-common.c (c_do_switch_warnings): Use SWITCH_STMT accessor
866 macros instead of SWITCH_EXPR ones.
867 * c-dump.c (c_dump_tree): Likewise.
868 * c-gimplify.c (gimplify_switch_stmt): Likewise.
869 * c-typeck.c (c_start_case, do_case, c_finish_case): Likewise.
871 * doc/c-tree.texi (SWITCH_STMT): Update accessor macro names.
873 2005-01-27 Alan Modra <amodra@bigpond.net.au>
875 * unwind-dw2.c (execute_stack_op): Add missing cases for
876 DW_OP_shl, DW_OP_shr, DW_OP_shra, DW_OP_xor.
878 2005-01-27 Eric Botcazou <ebotcazou@libertysurf.fr>
880 * config.gcc (Obsolete configurations): Remove 'dummy', add
881 sparclite-*-coff*, sparclite-*-elf*, sparc86x-*-elf* and
884 2005-01-26 Diego Novillo <dnovillo@redhat.com>
886 PR tree-optimization/19633
887 * tree-ssa-alias.c (ptr_is_dereferenced_by): Also handle
889 (maybe_create_global_var): Do not create .GLOBAL_VAR if there
890 are no call-clobbered variables.
891 * tree-outof-ssa.c (check_replaceable): Return false for calls
894 2005-01-26 Ulrich Weigand <uweigand@de.ibm.com>
896 * dbxout.c (dbxout_symbol_location): Resolve constant pool references
897 even for variables with NULL DECL_INITIAL.
899 2005-01-26 Stuart Hastings <stuart@apple.com>
901 * gimplify.c (shortcut_cond_expr): Re-compute side-effects.
903 2005-01-26 Richard Henderson <rth@redhat.com>
906 * c-decl.c (finish_struct): Set DECL_MODE after resetting a
908 * expr.c (store_field): Strip conversions to odd-bit-sized types
909 if the destination field width matches.
911 2005-01-26 Richard Henderson <rth@redhat.com>
913 * c-decl.c, expmed.c, expr.c: Revert last change.
915 2005-01-26 Richard Henderson <rth@redhat.com>
918 * c-decl.c (finish_struct): Set DECL_MODE after resetting a
920 * expmed.c (store_fixed_bit_field): Create a paradoxical subreg
921 if we don't need the bits above those present in the current mode.
922 * expr.c (store_field): Strip conversions to odd-bit-sized types
923 if the destination field width matches.
925 2005-01-26 Richard Sandiford <rsandifo@redhat.com>
927 * config/mips/iris6.h (ENDFILE_SPEC): Don't link in irix-csr.o if
928 either -ffast-math or -funsafe-math-optimizations is in use.
930 2005-01-26 Marek Michalkiewicz <marekm@amelek.gda.pl>
934 * config/avr/avr.c (notice_update_cc): Only set condition code for
935 ashrqi3 if shift count > 0.
936 (out_shift_with_cnt): Handle shift count <= 0 as a no-op.
937 (ashlqi3_out, ashlhi3_out, ashlsi3_out, ashrqi3_out, ashrhi3_out,
938 ashrsi3_out, lshrqi3_out, lshrhi3_out, lshrsi3_out): Handle shift
939 count <= 0 as a no-op, and shift count >= width by copying zero
940 or sign bit to all bits of the result.
941 * config/avr/avr.md (all shifts): Add alternatives for zero shift
942 count, with attribute "length" set to 0 and "cc" set to "none".
944 2005-01-26 Aldy Hernandez <aldyh@redhat.com>
946 * doc/invoke.texi: Document -mTLS.
948 * testsuite/gcc.target/frv/all-tls-global-dynamic.c: New.
949 * testsuite/gcc.target/frv/all-tls-initial-exec.c: New.
950 * testsuite/gcc.target/frv/all-tls-initial-exec-pic.c: New.
951 * testsuite/gcc.target/frv/all-tls-local-dynamic.c: New.
952 * testsuite/gcc.target/frv/all-tls-local-dynamic-plt.c: New.
953 * testsuite/gcc.target/frv/all-tls-local-dynamic-plt-pic.c: New.
954 * testsuite/gcc.target/frv/all-tls-local-exec.c: New.
955 * testsuite/gcc.target/frv/all-tls-local-exec-TLS.c: New.
957 * configure.ac: Check for a TLS capable gas.
959 * configure: Regenerate.
961 2005-01-26 Richard Henderson <rth@redhat.com>
964 * combine.c (make_field_assignment): Simplify store to zero_extract
965 from a source with an overlapping mask.
967 2005-01-26 Aldy Hernandez <aldyh@redhat.com>
969 2004-11-11 Eric Christopher <echristo@redhat.com>
971 * config/frv/frv.md: Add fr400_integer automaton. Don't
972 allow TYPE_MUL, TYPE_MACC, or TYPE_CUT to issue in parallel
975 2005-01-26 Steven Bosscher <stevenb@suse.de>
978 * cfgbuild.c (make_edges): Do not clear or set
979 current_function_has_computed_jump.
980 * function.h (struct function): Remove the has_computed_jump field.
981 (current_function_has_computed_jump): Do not define.
982 * sched-rgn.c (is_cfg_nonregular): Return true if a basic block ends
983 in a computed jump. Ignore current_function_has_computed_jump.
985 2005-01-26 Richard Henderson <rth@redhat.com>
988 * expr.c (categorize_ctor_elements): New argument p_must_clear.
989 (categorize_ctor_elements_1): Likewise. Detect a union that isn't
991 (mostly_zeros_p): Update for new categorize_ctor_elements argument.
992 * gimplify.c (gimplify_init_constructor): Likewise. Only shove
993 objects into static storage if they have more than one non-zero value.
994 * tree.h (categorize_ctor_elements): Update decl.
996 2005-01-26 Steven Bosscher <stevenb@suse.de>
999 * tree.h (CALL_EXPR_TAILCALL): Add comment.
1000 * calls.c (check_sibcall_argument_overlap_1): Revert the change
1001 to this function from 2004-07-10.
1002 * tree-tailcall.c (suitable_for_tail_opt_p): Do not consider the
1003 the current function for tail call optimizations if the address
1004 of one of it its arguments is taken.
1006 2005-01-26 Kazu Hirata <kazu@cs.umass.edu>
1008 * cse.c (fold_rtx): Call equiv_constant only when necessary.
1010 2005-01-26 Bernd Schmidt <bernd.schmidt@analog.com>
1012 * function.c (expand_function_end): If current_function_calls_alloca,
1013 emit stack restore in a place that is reached when the function falls
1016 2005-01-26 Alexandre Oliva <aoliva@redhat.com>
1018 * Makefile.in (../$(build_subdir)/fixincludes/fixincl): Add dummy
1019 rule to avoid matching inappropriate suffix rule.
1021 2005-01-26 Jan Hubicka <jh@suse.cz>
1023 PR tree-optimization/19241
1024 * tree-inline.c (copy_body_r): Do not walk subtrees after substituting.
1026 2005-01-26 Hans-Peter Nilsson <hp@axis.com>
1028 * config/cris/cris.h (LIB_SPEC) <sim*>: Use -lsyslinux, not
1029 -lsyssim. Wrap -lc -lsyssim in --start-group --end-group instead
1031 (STARTFILE_SPEC): Add crti.o, before crtbegin.o.
1032 <sim*>: Always use crt1.o, regardless of N in simN.
1033 <!sim>: Drop support for -pg and -p variants; always use crt0.o.
1034 (ENDFILE_SPEC): Add crtn.o, after crtend.o.
1035 (CRT_CALL_STATIC_FUNCTION): Remove.
1036 * config/cris/aout.h (STARTFILE_SPEC): <!melinux && sim*>: Always
1037 use crt1.o, regardless of N in simN.
1038 <!sim>: Drop support for -pg and -p variants; always use crt0.o.
1039 (LIB_SPEC): Do not define; default to the one in config/cris/cris.h.
1040 (CRIS_CPP_SUBTARGET_SPEC) <melinux && !nostdinc>: Remove special
1041 case isystem setting.
1042 (MAX_OFILE_ALIGNMENT): Correct to 32, not 16.
1044 2005-01-25 Ian Lance Taylor <ian@airs.com>
1046 * gimple-low.c (block_may_fallthru): Correct handling of
1047 SWITCH_EXPR--may fall through if SWITCH_LABELS is NULL.
1049 2005-01-26 Hans-Peter Nilsson <hp@axis.com>
1051 * config/cris/cris.h (CPP_SPEC): Adjust the various mtune=
1052 handlers to match indentation; nested under the mtune=* handler.
1053 Similar for mcpu= and march=.
1055 2005-01-25 Daniel Jacobowitz <dan@codesourcery.com>
1057 * config/mips/linux.h (LIB_SPEC): Remove -rpath-link.
1058 * config/mips/linux64.h (LIB_SPEC): Likewise.
1059 * config/mn10300/linux.h (LIB_SPEC): Likewise.
1060 * config/sh/linux.h (LIB_SPEC): Likewise.
1062 2005-01-25 Aldy Hernandez <aldyh@redhat.com>
1064 * config/frv/frv.c (frv_legitimize_tls_address): New.
1065 (TARGET_HAVE_TLS): Define.
1066 (FRV_SYMBOL_REF_TLS_P): Define.
1067 (frv_override_options): Handle new register classes.
1068 (frv_legitimate_address_p): Reject tls addresses.
1069 (frv_legitimize_address): Handle TLS addresses.
1070 (gen_inlined_tls_plt): New.
1072 (frv_legitimize_tls_address): New.
1073 (unspec_got_name): Add TLS entries.
1074 (got12_operand): Add R_FRV_TLSMOFF12 case.
1075 (frv_emit_move): Fixup TLS addresses.
1076 (frv_emit_movsi): Legitimize TLS addresses.
1078 * config/frv/frv.h (MASK_BIG_TLS): New.
1079 (TARGET_BIG_TLS): New.
1080 (HAVE_AS_TLS): Define.
1081 (TARGET_SWITCHES): Add -mTLS and -mtls options.
1082 (enum reg_class): Add GR8_REGS, GR9_REGS, GR89_REGS.
1083 (REG_CLASS_NAMES): Same.
1084 (REG_CLASS_CONTENTS): Same.
1085 (CONSTRAINT_LEN): New.
1086 (REG_CLASS_FROM_CONSTRAINT): New.
1087 (PREDICATE_CODES): Add symbolic_operand.
1089 * config/frv/frv.md (define_constants): Add UNSPEC_GETTLSOFF,
1090 UNSPEC_TLS_LOAD_GOTTLSOFF12, UNSPEC_TLS_INDIRECT_CALL,
1091 UNSPEC_TLS_TLSDESC_LDD, UNSPEC_TLS_TLSDESC_LDD_AUX,
1092 UNSPEC_TLS_TLSOFF_LD, UNSPEC_TLS_LDDI, UNSPEC_TLSOFF_HILO,
1093 R_FRV_GOTTLSOFF_HI, R_FRV_GOTTLSOFF_LO, R_FRV_TLSMOFFHI,
1094 R_FRV_TLSMOFFLO, R_FRV_TLSMOFF12, R_FRV_TLSDESCHI,
1095 R_FRV_TLSDESCLO, R_FRV_GOTTLSDESCHI, R_FRV_GOTTLSDESCLO, GR8_REG,
1096 GR9_REG, GR14_REG, LRREG.
1097 (type): Add load_or_call attribute.
1098 ("load_or_call"): New reservation.
1099 ("call_gettlsoff"): New.
1100 ("tls_indirect_call"): New.
1101 ("tls_load_gottlsoff12"): New.
1102 ("tlsoff_hilo"): New.
1103 ("tls_tlsdesc_ldd"): New.
1104 ("tls_tlsoff_ld"): New.
1107 * config/frv/frv-protos.h (symbolic_operand): Protoize.
1109 2005-01-25 Alexandre Oliva <aoliva@redhat.com>
1111 * config/frv/frv.h (ASM_OUTPUT_DWARF_DTPREL): Define.
1112 * config/frv/frv-protos.h (frv_output_dwarf_dtprel): Declare.
1113 * config/frv/frv.c (TLS_BIAS): Define.
1114 (frv_output_dwarf_dtprel): New.
1116 2005-01-26 Jakub Jelinek <jakub@redhat.com>
1118 PR rtl-optimization/19579
1119 * ifcvt.c (noce_try_cmove_arith): If emitting instructions to set up
1120 both A and B, see if they don't clobber registers the other expr uses.
1122 2005-01-25 J"orn Rennecke <joern.rennecke@st.com>
1124 * real.c (do_add): Initialize signalling and canonical members.
1126 * real.c (real_from_integer): Zero out destination.
1128 2005-01-18 Jan Hubicka <jh@suse.cz>
1130 PR tree-optimize/19337
1131 * tree-inline.c (initialize_inlined_parameters): Copy
1132 saved_static_chain_decl only.
1134 2005-01-25 Richard Henderson <rth@redhat.com>
1138 * config/i386/i386.c (x86_inter_unit_moves): Zero.
1139 (ix86_preferred_reload_class): Rewrite fp-constant section, with
1140 80387 enabled, to return a proper subclass. Return the subset that
1141 overlaps with GENERAL_REGS for PLUS.
1142 * config/i386/i386.md (movsi_1, movdi_2): Set type to "mmx" for pxor.
1144 2005-01-25 Richard Henderson <rth@redhat.com>
1147 * tree-complex.c (expand_complex_div_wide): Use the correct formulae.
1149 2005-01-25 Roger Sayle <roger@eyesopen.com>
1151 * expmed.c (choose_multiplier): Change interface to return the
1152 multiplier lowpart as an rtx instead of an unsigned HOST_WIDE_INT.
1153 (expand_divmod): Update calls to choose_multiplier.
1155 2005-01-24 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1158 * expmed.c (extract_bit_field): Use adjust_address instead of
1159 gen_lowpart when op0 is a MEM.
1161 2005-01-24 Steven Bosscher <stevenb@suse.de>
1163 * cfgexpand.c (expand_gimple_tailcall): Fix typo.
1165 2005-01-24 Richard Sandiford <rsandifo@redhat.com>
1167 * config/mips/irix-csr.c: New file.
1168 * config/mips/t-iris6 (irix-csr.o): New rule to build it.
1169 (EXTRA_MULTILIB_PARTS): Add irix-csr.o.
1170 * config/mips/iris6.h (ENDFILE_SPEC): Include it in n32 and n64
1173 2005-01-24 Eric Botcazou <ebotcazou@libertysurf.fr>
1176 * config.gcc (sparc-*-elf*): Remove sol2.h, sparc/sol2.h and
1177 sparc/elf.h, add sparc/sp-elf.h.
1178 (sparc-*-rtems*): Likewise.
1179 (sparclite-*-elf*): Remove sol2.h, sparc/sol2.h, sparc/elf.h and
1180 tm-dwarf2.h, add sparc/sp-elf.h.
1181 (sparc86x-*-elf): Likewise.
1182 (sparc64-*-elf*): Remove sol2.h, sparc/sol2.h and tm-dwarf2.h.
1183 * config/sparc/liteelf.h (TARGET_SUB_OS_CPP_BUILTINS): Rename into
1184 TARGET_OS_CPP_BUILTINS.
1185 * config/sparc/sp86x-elf (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1186 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1188 * config/sparc/openbsd64.h (NO_IMPLICIT_EXTERN_C): Undefine.
1189 * config/sparc/sp64-elf.h (NO_IMPLICIT_EXTERN_C): New macro.
1190 (SWITCH_TAKES_ARG): Likewise.
1191 (LOCAL_LABEL_PREFIX): Likewise.
1192 (ASM_GENERATE_INTERNAL_LABEL): Likewise.
1193 (TARGET_N_FORMAT_TYPES): Delete.
1194 (TARGET_FORMAT_TYPES): Likewise.
1195 (ASM_DECLARE_FUNCTION_SIZE): Likewise.
1196 * config/sparc/elf.h: Delete.
1197 * config/sparc/sp-elf.h: New file.
1199 2005-01-24 Kazu Hirata <kazu@cs.umass.edu>
1201 * tree-flow-inline.h (phi_arg_from_edge): Remove.
1202 * tree-flow.h: Remove the corresponding prototype.
1203 * tree-ssa-operands.h, tree-ssa-threadupdate.c, tree-ssa.c,
1204 tree-vectorizer.c: Use dest_idx instead of phi_arg_from_edge.
1206 2005-01-24 Jakub Jelinek <jakub@redhat.com>
1209 * config/rs6000/rs6000.c (rs6000_file_start): Create toc section
1210 for AIX ABI or ELF -fPIC.
1211 (rs6000_emit_load_toc_table): Don't create toc_section here.
1212 (rs6000_xcoff_file_start): Nor here.
1214 2005-01-24 J"orn Rennecke <joern.rennecke@st.com>
1216 * sh.c (ra.h): Don't #include.
1217 (hard_regs_intersect_p): New function, resurrected from ra.c.
1219 * sh.c: Fix 1996 Copyright.
1221 2005-01-24 Richard Henderson <rth@redhat.com>
1222 Aldy Hernandez <aldyh@redhat.com>
1224 * regrename.c (note_sets): Handle subregs.
1226 2005-01-24 Joseph S. Myers <joseph@codesourcery.com>
1229 * genconditions.c (write_header, write_conditions): Elide file if
1232 2005-01-24 Kazu Hirata <kazu@cs.umass.edu>
1234 * c-opts.c, expmed.c, gengtype.c, tree-complex.c,
1235 config/arm/iwmmxt.md, config/avr/avr.c, config/s390/s390.h:
1238 2005-01-24 Jakub Jelinek <jakub@redhat.com>
1240 * builtins.c (get_pointer_alignment, gimplify_va_arg_expr,
1241 expand_builtin_printf, expand_builtin_fprintf,
1242 expand_builtin_sprintf): Use POINTER_TYPE_P instead TREE_CODE
1243 checking against POINTER_TYPE.
1244 (validate_arglist): Handle POINTER_TYPE code by checking
1247 2005-01-24 Paolo Bonzini <bonzini@gnu.org>
1249 * rtlanal.c (non_rtx_starting_operands, for_each_rtx_1,
1251 (for_each_rtx): Call for_each_rtx_1.
1252 * rtl.h (init_rtlanal): Declare.
1253 * toplev.c (backend_init): Call init_rtlanal.
1255 2005-01-24 Jakub Jelinek <jakub@redhat.com>
1257 * flow.c (propagate_one_insn): Formatting.
1260 * flow.c (libcall_dead_p): Be more conservative if unsure.
1261 If there are any instructions between insn and call, see if they are
1262 all dead before saying the libcall is dead.
1264 2005-01-24 Uros Bizjak <uros@kss-loka.si>
1266 * config/i386/i386.md (*extendsfdf2_sse): Use "nonimmediate_operand"
1267 operand constraint for operand 0.
1269 2005-01-23 Richard Henderson <rth@redhat.com>
1272 * tree-complex.c (expand_complex_div_wide): Set EDGE_FALLTHRU on
1273 edges leading to join block.
1275 2005-01-23 Roger Sayle <roger@eyesopen.com>
1277 * expmed.c (expand_mult_highpart): Make static. Change type of
1278 constant multiplier argument from unsigned HOST_WIDE_INT to rtx.
1279 (expand_divmod): Updates calls to expand_mult_highpart by using
1280 gen_int_mode to pass a CONST_INT rtx.
1281 * rtl.h (expand_mult_highpart): Remove prototype.
1283 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
1285 * flow.c (regset_bytes, regset_size): Remove.
1287 * cfganal.c: Fix a reference to Harvey's paper.
1289 2005-01-23 Steven Bosscher <stevenb@suse.de>
1291 PR rtl-optimization/19464
1292 * tree-optimize.c (init_tree_optimization_passes): Add one more
1293 copyrename pass just before out-of-ssa.
1295 2005-01-23 Joseph S. Myers <joseph@codesourcery.com>
1298 * recog.c (recog_memoized): Don't define if GENERATOR_FILE.
1299 * ggc-none.c (ggc_free): Define.
1301 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
1303 * cse.c (max_reg, max_insn_uid): Remove.
1304 (cse_main): Don't access max_reg or max_insn_uid.
1306 * c-common.c (explicit_flag_signed_bitfields,
1307 lang_statement_code_p, lang_expand_function_end): Remove.
1308 * c-common.h: Remove the corresponding declarations.
1309 * c-opts.c (c_common_handle_option): Remove all write access
1310 to explicit_flag_signed_bitfields.
1312 * cgraph.c (cgraph_varpool_n_nodes): Remove.
1313 (cgraph_varpool_node): Don't access cgraph_varpool_n_nodes.
1314 * cgraph.h: Remove the corresponding declaration.
1316 * gcse.c (null_pointer_info): Remove.
1318 2005-01-23 Roger Sayle <roger@eyesopen.com>
1319 Eric Botcazou <ebotcazou@libertysurf.fr>
1321 * combine.c (simplify_logical): Only simplify logical expressions
1322 of the form ior(and(x,y),z) by the inverse distributive law if the
1323 result is cheaper than the original.
1325 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
1327 * alias.c, c-common.h, c-incpath.c, c-incpath.h, expr.c,
1328 fold-const.c, gimplify.c, params.h, tree-data-ref.c,
1329 tree-if-conv.c, tree-nested.c, tree-outof-ssa.c,
1330 tree-ssa-dom.c, tree-vectorizer.c, tree.def, config/darwin.c,
1331 config/freebsd-spec.h, config/arm/arm.h,
1332 config/h8300/h8300.md, config/i386/i386.md,
1333 config/i386/predicates.md, config/i386/sse.md,
1334 config/ia64/ia64.c, config/ip2k/ip2k.c, config/s390/s390.c,
1335 config/vax/vax.md: Fix comment typos. Follow spelling
1338 2005-01-23 Steven Bosscher <stevenb@suse.de>
1340 * gengtype.c (walk_type): Produce `not equal to' compares for loop
1341 tests, instead of `less than'.
1343 2005-01-23 Richard Sandiford <rsandifo@redhat.com>
1345 * function.c (assign_parm_setup_block): When creating a new stack slot
1346 for a parameter, get its alignment from the parameter's DECL_ALIGN
1347 rather than the type's TYPE_ALIGN. Make sure that the parameter
1348 is at least word aligned.
1350 2005-01-22 Roger Sayle <roger@eyesopen.com>
1353 * config/avr/avr.c (avr_hard_regno_mode_ok): Rewrite.
1355 2005-01-22 Richard Henderson <rth@redhat.com>
1358 * config/i386/i386.md (movsfcc_1_sse_max): Use nonimmediate_operand
1359 in both compare operands.
1360 (movdfcc_1_sse_max): Likewise.
1361 (movsfcc_1_sse): Likewise. Add earlyclobber for scratch.
1362 (movdfcc_1_sse): Likewise.
1363 * config/i386/i386.c (ix86_split_sse_movcc): Emit copies into the
1364 scratch register as needed.
1366 2005-01-22 Richard Henderson <rth@redhat.com>
1368 * config/i386/i386.md (smaxsf3): Fix mnemonic typo.
1370 * config/i386/i386.c (ix86_prepare_fp_compare_args): Fix is_sse test.
1372 2005-01-22 Richard Henderson <rth@redhat.com>
1374 * genopinit.c (optabs): Use smin/smax for floating point too.
1375 * doc/md.texi: Update to match. Clarify that floating point
1376 results are undefined for +0/-0 and NaN.
1377 * doc/rtl.texi: Likewise.
1378 * rtl.def (SMIN, SMAX): Likewise
1379 * tree.def (MIN_EXPR, MAX_EXPR): Likewise.
1381 * config/alpha/alpha.md (smaxdf3, smindf3, smaxsf3, sminsf3): Add
1382 leading 's' to the name.
1383 * config/ia64/ia64.md (smaxsf3, sminsf3, smaxdf3, smindf3,
1384 smaxxf3, sminxf3): Likewise.
1385 * config/rs6000/rs6000.md (smaxdf3, smindf3, smaxsf3, sminsf3):
1388 2005-01-22 Paul Brook <paul@codesourcery.com>
1390 * config/arm/arm.md: Use "Uy" constraint for wcgr load/stores.
1392 2005-01-21 Mark Dettinger <dettinge@de.ibm.com>
1394 * config/s390/s390.c (struct processor_costs): New fields
1395 dlgr, dlr, dr, dsgfr, dsgr.
1396 (z900_cost, z990_cost): Values for new fields.
1397 (s390_rtx_costs): New cases MEM und COMPARE in switch
1398 statement. Modified handling of SIGN_EXTEND, ZERO_EXTEND,
1399 DIV, MOD, UDIV, UMOD.
1401 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
1403 * config/s390/s390.md ("doloop_si64"): Reload input value directly
1404 into the register being decremented.
1405 ("doloop_si31", "doloop_di"): Likewise.
1406 ("*doloop_si_long"): Adapt pattern.
1408 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
1410 * config/s390/s390.h (HARD_REGNO_NREGS): Fix computation for
1412 (CLASS_MAX_NREGS): Likewise.
1414 2005-01-21 Daniel Berlin <dberlin@dberlin.org>
1416 * doc/tree-ssa.texi (Statement Operands): Add example for new
1417 must-def macro. Note deprecation of old operands interface.
1419 2005-01-22 Richard Sandiford <rsandifo@redhat.com>
1421 PR tree-optimization/19484
1422 * tree-cfg.c (remove_fallthru_edge): New function.
1423 (cleanup_control_flow): Remove fallthru edges from calls that are
1424 now known not to return.
1426 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
1428 Fix PR tree-optimization/19038
1429 * tree-ssa-dom.c (cprop_operand): Don't replace loop invaeriant
1430 copies with loop variant ones.
1432 2005-01-22 Kazu Hirata <kazu@cs.umass.edu>
1434 * cfganal.c, real.h, reorg.c, timevar.def, tree-ssa-ccp.c,
1435 config/alpha/alpha-protos.h, config/alpha/alpha.h,
1436 config/alpha/alpha.md, config/alpha/predicates.md,
1437 config/sparc/freebsd.h, config/sparc/netbsd-elf.h,
1438 config/sparc/sol2.h: Update copyright.
1440 * tree-cfg.c (remove_forwarder_block_with_phi): Look at the
1441 first label to see if it is a nonlocal label.
1443 2005-01-22 David Edelsohn <edelsohn@gnu.org>
1444 Andrew Pinski <pinskia@physics.uc.edu>
1447 * config/rs6000/rs6000.c (rs6000_va_start): Saturate n_gpr at
1448 maximum number of GPRs. Saturate n_fpr at maximum number of FPRs.
1450 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
1452 * config/rs6000/rtems.h: Update copyright.
1454 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
1457 * config/rs6000/rtems.h: Resurrect cpp_os_rtems_spec from gcc < 3.4.
1458 (CPP_OS_RTEMS_SPEC): New (From gcc-3.3's config/rs6000/sys4.h).
1459 (SUBSUBTARGET_EXTRA_SPECS): Use CPP_OS_RTEMS_SPEC.
1461 2005-01-22 Volker Reichelt <reichelt@igpm.rwth-aachen.de>
1464 * c-typeck.c (convert_arguments): Check for error_mark_node.
1466 2005-01-21 Ian Lance Taylor <ian@airs.com>
1468 PR tree-optimization/13000
1469 * tree-inline.c: Include "tree-flow.h".
1470 (expand_call_inline): If warn_return_type, warn if non-void inline
1471 function falls through.
1472 * tree-cfg.c (execute_warn_function_return): Don't warn about
1473 control reaching end if TREE_NO_WARNING is set. Set
1475 * gimple-low.c (block_may_fallthru): Don't assume that SWITCH_EXPR
1477 * gimplify.c (shortcut_cond_expr): Don't emit a jump over the else
1478 branch if we don't need one.
1479 * c-typeck.c: Include "tree-flow.h"
1480 (c_finish_bc_stmt): Don't add a goto if the current statement
1481 list doesn't fall through to the current point.
1483 2005-01-21 Roger Sayle <roger@eyesopen.com>
1485 PR rtl-optimization/576
1486 * real.c (real_arithmetic): Change return type from void to bool
1487 to return an indication that the result may be inexact.
1488 * real.h (real_arithmeric): Update prototype.
1489 * fold-const.c (const_binop): Don't constant fold floating
1490 point expressions when the user specifies -frounding-math and
1491 the result may depend upon the run-time rounding mode.
1492 (fold_convert_const_real_from_real): Clean-up.
1493 (fold_initializer): Ignore flag_rounding_math for initializers.
1494 * simplify-rtx.c (simplify_binary_operation): Likewise, don't
1495 constant fold FP operations with flag_rounding_math if the
1496 result may depend upon the run-time rounding mode.
1498 2005-01-21 Tom Tromey <tromey@redhat.com>
1500 * c-cppbuiltin.c (define__GNUC__): Correct assertion.
1502 2005-01-21 Kazu Hirata <kazu@cs.umass.edu>
1504 * tree-cfg.c: Fix comment typos.
1506 2005-01-21 J"orn Rennecke <joern.rennecke@st.com>
1508 * passes.c (rest_of_handle_flow2): Close / open the flow2 dump file
1509 around the call to rest_of_handle_branch_target_load_optimize.
1511 2005-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
1513 * config.gcc (sparc-*-netbsdelf*): Include dbxelf.h
1514 (sparc64-*-openbsd*): Likewise.
1515 (sparclite-*-elf*): Include tm-dwarf2.h.
1516 (sparc86x-*-elf*): Likewise.
1517 (sparc64-*-elf*): Likewise.
1518 * config/sparc/linux64.h (DWARF2_DEBUGGING_INFO): Delete.
1519 (DBX_DEBUGGING_INFO): Likewise.
1520 * config/sparc/liteelf.h (DWARF2_DEBUGGING_INFO): Likewise.
1521 (PREFERRED_DEBUGGING_TYPE): Likewise.
1522 * config/sparc/netbsd-elf.h (PREFERRED_DEBUGGING_TYPE): Likewise.
1523 * config/sparc/sp64-elf.h (DBX_DEBUGGING_INFO): Likewise.
1524 (PREFERRED_DEBUGGING_TYPE): Likewise.
1525 * config/sparc/sp86x-elf.h (DWARF2_DEBUGGING_INFO): Likewise.
1526 (PREFERRED_DEBUGGING_TYPE): Likewise.
1527 * config/sparc/sparc.h (DBX_DEBUGGING_INFO): Likewise.
1529 2005-01-21 Hans-Peter Nilsson <hp@bitrange.com>
1532 * combine.c (combine_simplify_rtx): Revert change of 2004-12-31.
1534 2005-01-20 Paul Brook <paul@codesourcery.com>
1536 * doc/rtl.texi: Document value extension requirements for CONST_INT.
1538 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
1540 * tree-cfg.c (tree_verify_flow_info): Check that a nonlocal
1541 label is first in a sequence of labels.
1543 * tree-cfg.c (tree_verify_flow_info): Fix a typo.
1545 2005-01-20 Janis Johnson <janis187@us.ibm.com>
1546 Giovanni Bajo <giovannibajo@gcc.gnu.org>
1548 * doc/sourcebuild.texi (Test Directives): New.
1550 2005-01-20 Roger Sayle <roger@eyesopen.com>
1552 * real.c (real_floor): Don't leave the result uninitialized when
1553 mode is VOIDMode, but return the unrounded intermediate value.
1554 (real_ceil): Likewise.
1556 2005-01-20 Paul Brook <paul@codesourcery.com>
1558 * config/arm/arm.md (insv): Use gen_int_mode.
1560 2005-01-20 Eric Botcazou <ebotcazou@libertysurf.fr>
1562 * config/sparc/freebsd.h (ASM_OUTPUT_INTERNAL_LABELREF): Delete.
1563 * config/sparc/linux.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1564 * config/sparc/linux64.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1565 * config/sparc/netbsd-elf.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1566 * config/sparc/sol2.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1568 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
1571 * dwarf2out.c (concat_loc_descriptor): We don't know we can use
1572 fbreg, so we have to assume we can't.
1573 (loc_descriptor_from_tree_1): Ditto.
1574 (containing_function_has_frame_base): New function.
1575 (add_location_or_const_value_attribute): Use it.
1576 Also try to generate a frame_base from a single element location
1579 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
1581 PR tree-optimization/15349
1582 * timevar.def (TV_TREE_MERGE_PHI): New.
1583 * tree-cfg.c (tree_forwarder_block_p): Add a new argument
1585 (remove_forwarder_block, cleanup_forwarder_blocks): Adjust the
1586 calls to tree_forwarder_block_p.
1587 (remove_forwarder_block_with_phi, merge_phi_nodes,
1588 gate_merge_phi, pass_merge_phi): New.
1589 * tree-optimize.c (init_tree_optimization_passes): Add
1591 * tree-pass.h: Add an extern for pass_merge_phi;
1593 2005-01-20 Richard Henderson <rth@redhat.com>
1596 * config/i386/emmintrin.h (_mm_castpd_ps, _mm_castpd_si128): New.
1597 (_mm_castps_pd, _mm_castps_si128): New.
1598 (_mm_castsi128_ps, _mm_castsi128_pd): New.
1600 2005-01-20 Richard Henderson <rth@redhat.com>
1603 * config/i386/mmintrin.h (_mm_cvtsi32_si64): Use
1604 __builtin_ia32_vec_init_v2si.
1605 (_mm_cvtsi64_si32): Use __builtin_ia32_vec_ext_v2si.
1606 * config/i386/i386.c (IX86_BUILTIN_VEC_EXT_V2SI): New.
1607 (ix86_init_mmx_sse_builtins): Create it.
1608 (ix86_expand_builtin): Expand it.
1609 (ix86_expand_vector_set): Handle V2SFmode and V2SImode.
1610 * config/i386/mmx.md (vec_extractv2sf_0, vec_extractv2sf_1): New.
1611 (vec_extractv2si_0, vec_extractv2si_1): New.
1613 2005-01-20 Richard Henderson <rth@redhat.com>
1616 * config/i386/i386.c (ix86_expand_push): New.
1617 * config/i386/mmx.md (push<MMXMODE>1): New.
1618 * config/i386/sse.md (push<SSEMODE>1): New.
1619 * config/i386/i386-protos.h: Update.
1621 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
1623 * params.def, params.h, predict.c, tree-ssa-loop-im.c: Update
1626 2005-01-20 Richard Henderson <rth@redhat.com>
1628 * rtl.def (CONST_VECTOR): Use RTX_CONST_OBJ.
1629 * rtl.h (CONSTANT_P): Don't special case CONST_VECTOR.
1631 2005-01-19 Richard Henderson <rth@redhat.com>
1634 * config/i386/i386.c (ix86_expand_vector_move_misalign): Convert
1635 to V4SFmode in SSE1 fallback load path.
1637 2005-01-19 Richard Henderson <rth@redhat.com>
1639 * config/i386/i386.c (ix86_expand_vector_init_one_var): Fix typo
1640 in QImode expansion to ix86_expand_vector_set.
1641 (ix86_expand_vector_init_general): Fix typo in V8HImode recursive call.
1643 2005-01-19 Richard Henderson <rth@redhat.com>
1646 * config/i386/i386.c (ix86_preferred_reload_class): Return a proper
1647 subclass of the input class.
1648 (ix86_secondary_memory_needed): Always true for cross-MMX classes.
1649 Always true for cross-SSE1 classes. Rationalize conditionals.
1650 * config/i386/i386.h (SSE_CLASS_P, MMX_CLASS_P): Use straight equality.
1651 * config/i386/i386.md (movsi_1): Add MMX/SSE zeros. Fix alternatives
1652 for SSE1. Don't check TARGET_INTER_UNIT_MOVES.
1653 (movdi_2): Add MMX/SSE zeros.
1654 (movdi_1_rex64): Likewise. Don't check TARGET_INTER_UNIT_MOVES.
1655 (movsf_1): Don't check TARGET_INTER_UNIT_MOVES.
1656 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
1657 (movsi_1_nointernunit, movdi_1_rex64_nointerunit): Remove.
1658 (movsf_1_nointerunit, zero_extendsidi2_32_1): Remove.
1659 (zero_extendsidi2_rex64_1): Remove.
1660 (MOV0 peephole): Check GENERAL_REG_P.
1662 2005-01-19 Richard Henderson <rth@redhat.com>
1665 * config/i386/i386.c (ix86_expand_vector_set): Fix third and fourth
1667 (ix86_expand_vector_extract): Likewise.
1669 2005-01-19 Richard Henderson <rth@redhat.com>
1672 * expr.c (emit_move_change_mode): New argument force; use
1673 simplify_gen_subreg if true.
1674 (emit_move_via_alt_mode): Merge into ...
1675 (emit_move_via_integer): ... here.
1676 (emit_move_ccmode): Use emit_move_change_mode directly.
1678 2005-01-19 Richard Henderson <rth@redhat.com>
1681 * config/alpha/alpha.c (alpha_rtx_costs): Handle HIGH.
1682 (alpha_preferred_reload_class): Handle CONST_VECTOR.
1683 (alpha_emit_set_const_1): Add no_output parameter; don't emit
1685 (alpha_emit_set_const): Likewise. Make static.
1686 (alpha_emit_set_long_const): Make static.
1687 (alpha_extract_integer): Split out from alpha_expand_mov.
1688 (alpha_split_const_mov): Likewise.
1689 (alpha_expand_mov): Use them. Handle CONST_VECTOR.
1690 (alpha_legitimate_constant_p): New.
1691 * config/alpha/alpha-protos.h: Update.
1692 * config/alpha/alpha.h (REGISTER_MOVE_COST): Correct fp<->gp cost.
1693 (LEGITIMATE_CONSTANT_P): Re-implement with a function.
1694 * config/alpha/alpha.md (movsi): Add n alternative.
1695 (movsi_nt_vms, movdi_er_nofix, movdi_er_fix, movdi_fix): Likewise.
1696 (mov<VEC>_fix, mov<VEC>_nofix): Add i alternative.
1697 (splitters for all of the above): Use alpha_split_const_mov.
1698 * config/alpha/predicates.md (non_add_const_operand): New.
1699 (non_zero_const_operand): New.
1700 (input_operand): Use alpha_legitimate_constant_p after reload.
1702 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
1704 PR tree-optimization/19038
1705 * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): New function.
1706 (add_candidate): Add ivs with increment in latch only if
1707 allow_ip_end_pos_p is true.
1708 (determine_iv_cost): Use empty_block_p.
1710 2005-01-19 Daniel Berlin <dberlin@dberlin.org>
1712 * cfganal.c (compute_dominance_frontiers_1): Replace with new algorithm
1713 (compute_dominance_frontiers): Ditto.
1715 2005-01-19 Ralf Corsepius <ralf.corsepius@rtems.org>
1718 * config/sh/t-rtems: New.
1719 * config.gcc (sh-*-rtems*): Reflect having added config/sh/t-rtems.
1721 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
1723 * tree-ssa-ccp.c (maybe_fold_offset_to_component_ref): Always subtract
1724 the offset of the selected field.
1726 2005-01-19 Kazu Hirata <kazu@cs.umass.edu>
1728 * tree-cfg.c (remove_forwarder_block): Fix the check to
1729 prevent a nonlocal label from appearing in the middle of a
1732 2005-01-19 Hans-Peter Nilsson <hp@axis.com>
1734 PR rtl-optimization/19462
1735 * reorg.c (find_end_label): Create return insn only if
1736 current_function_epilogue_delay_list is empty.
1738 2005-01-19 Paolo Bonzini <bonzini@gnu.org>
1740 * doc/passes.texi: Remove paragraph mentioning new-ra.
1742 2005-01-19 Jakub Jelinek <jakub@redhat.com>
1744 PR rtl-optimization/15139
1745 * combine.c: Include params.h.
1746 (count_rtxs): New function.
1747 (record_value_for_reg): If replace_rtx would replace at least
1748 2 occurrences of REG in VALUE and TEM is really large, replace REG with
1749 (clobber (const_int 0)) instead of TEM.
1750 * params.def (PARAM_MAX_LAST_VALUE_RTL): New.
1751 * params.h (MAX_LAST_VALUE_RTL): New.
1752 * Makefile.in (combine.o): Depend on $(PARAMS_H).
1753 * doc/invoke.texi (--param max-last-value-rtl=N): Document.
1756 * c-typeck.c (digest_init): Only call build_vector if all constructor
1757 elements are *_CST nodes.
1758 * gimplify.c (gimplify_init_constructor): Likewise.
1761 * c-typeck.c (digest_init): Only call build_vector if inside_init
1764 2005-01-18 Tobias Schl"uter <tobias.schlueter@physik.uni-muenchen.de>
1766 * toplev.c (init_asm_output): Remove dead #ifdef.
1768 2005-01-18 Kaz Kojima <kkojima@gcc.gnu.org>
1770 * config/sh/linux.h (TARGET_C99_FUNCTIONS): Define.
1772 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
1774 * calls.c (expand_call): Check DECL_BUILT_IN_CLASS before
1775 accessing DECL_FUNCTION_CODE.
1776 * dojump.c (do_jump): Likewise.
1777 * gimplify.c (gimplify_call_expr): Likewise.
1778 * predict.c (expr_expected_value): Likewise.
1779 (strip_builtin_expect): Likewise.
1780 * tree-inline.c (estimate_num_insns_1): Likewise.
1781 * tree-ssa-loop-im.c (stmt_cost): Likewise
1782 * fold-const.c (fold): Test for BUILT_IN_NORMAL.
1783 (tree_expr_nonnegative_p): Likewise.
1785 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
1787 * config/sparc/sparc.c (load_pic_register): Emit the appropriate
1788 variant of the load_pcrel_sym pattern.
1789 * config/sparc/sparc.md (P macro): Move to the top.
1790 (load_pcrel_sym): Macroize using P.
1791 (save_register_window): Likewise.
1793 2005-01-18 Aldy Hernandez <aldyh@redhat.com>
1795 * config/rs6000/rs6000.md ("sunordered"): Disable for e500.
1798 2005-01-18 Andrew Pinski <pinskia@physics.uc.edu>
1801 * c-typeck.c (build_asm_expr): Strip nops off
1802 input memory operands.
1804 2005-01-18 David Edelsohn <edelsohn@gnu.org>
1806 * config/rs6000/aix43.h (CPLUSPLUS_CPP_SPEC): Delete
1807 _XOPEN_SOURCE, _XOPEN_SOURCE_EXTENDED, _LARGE_FILE_API.
1808 * config/rs6000/aix51.h (CPLUSPLUS_CPP_SPEC): Same.
1809 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Same.
1811 2005-01-18 Joel Sherrill <joel.sherrill@oarcorp.com>
1814 * config/i386/i386.c (override_options): If the 80387 is disabled,
1815 then do not return FP values using FP registers.
1817 2005-01-18 Kazu Hirata <kazu@cs.umass.edu>
1819 * basic-block.h, c-common.c, c-cppbuiltin.c, c-lang.c,
1820 c-tree.h, cfgbuild.c, cgraph.c, cgraph.h, collect2.c,
1821 combine.c, config.gcc, coverage.h, cse.c, cselib.c,
1822 defaults.h, df.c, dwarf2asm.c, dwarf2out.c, explow.c, expr.c,
1823 flow.c, fold-const.c, gcse.c, ggc-page.c, gimple-low.c,
1824 gimplify.c, ifcvt.c, langhooks-def.h, lcm.c, optabs.h,
1825 output.h, postreload-gcse.c, postreload.c, recog.c,
1826 resource.c, rtl.def, rtlanal.c, sched-deps.c, sched-rgn.c,
1827 targhooks.h, toplev.c, tree-data-ref.c, tree-eh.c,
1828 tree-flow.h, tree-loop-linear.c, tree-mudflap.h, tree-nrv.c,
1829 tree-optimize.c, tree-outof-ssa.c, tree-pass.h,
1830 tree-scalar-evolution.c, tree-ssa-copy.c, tree-ssa-dce.c,
1831 tree-ssa-dse.c, tree-ssa-loop-ivopts.c, tree-ssa-operands.c,
1832 tree-ssa-pre.c, tree-ssa.c, tree-vectorizer.c, tree.def,
1833 unwind-dw2-fde-darwin.c, var-tracking.c: Update copyright.
1835 * config/darwin.c, config/darwin.h, config/freebsd-spec.h,
1836 config/alpha/alpha.c, config/arm/vfp.md,
1837 config/i386/emmintrin.h, config/i386/i386-modes.def,
1838 config/i386/i386-protos.h, config/i386/i386.h,
1839 config/i386/pmmintrin.h, config/i386/ppro.md,
1840 config/i386/predicates.md, config/i386/xmmintrin.h,
1841 config/i860/i860.c, config/ia64/ia64-modes.def,
1842 config/ia64/ia64-protos.h, config/ia64/ia64.h,
1843 config/ia64/ia64.md, config/ia64/itanium1.md,
1844 config/ia64/itanium2.md, config/ia64/predicates.md,
1845 config/iq2000/iq2000.h, config/mips/linux64.h,
1846 config/rs6000/aix41.h, config/rs6000/aix43.h,
1847 config/rs6000/aix52.h, config/rs6000/darwin-fallback.c,
1848 config/rs6000/rs6000.c, config/rs6000/rs6000.h,
1849 config/rs6000/spe.md, config/sh/sh.md, config/sparc/linux.h,
1850 config/sparc/linux64.h, config/sparc/litecoff.h,
1851 config/sparc/sp64-elf.h, config/sparc/sparc.c,
1852 config/sparc/sparc.h, config/sparc/sparc.md,
1853 config/sparc/sysv4-only.h: Update copyright.
1855 2005-01-18 Richard Henderson <rth@redhat.com>
1858 * config/i386/i386.c (ix86_expand_fp_movcc): Fail for LTGT and UNEQ.
1860 2005-01-18 Marc Espie <espie@openbsd.org>
1861 * config/i386/openbsdelf.h: Typo.
1863 2005-01-18 Richard Henderson <rth@redhat.com>
1865 * config/i386/i386.c (ix86_expand_fp_movcc): Remove TARGET_IEEE_FP
1866 special case for sse.
1867 * config/i386/predicates.md (sse_comparison_operator): Likewise.
1869 2005-01-18 Dorit Naishlos <dorit@il.ibm.com>
1871 * tree-inline.c (estimate_num_insns_1): Added cases for
1872 ALIGN_INDIRECT_REF, MISALIGNED_INDIRECT_REF, and REALIGN_LOAD_EXPR.
1874 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
1876 PR rtl-optimization/19296
1877 * combine.c (simplify_comparison): Rewrite the condition under
1878 which a non-paradoxical SUBREG of a PLUS can be lifted when
1879 compared against a constant.
1881 2005-01-18 Andi Kleen <ak@muc.de>
1883 * c-typeck.c: (convert_for_assignment): Check warn_pointer_sign.
1884 * c.opt (-Wpointer-sign): Add.
1885 * doc/invoke.texi: (-Wpointer-sign): Add.
1887 2005-01-18 Uros Bizjak <uros@kss-loka.si>
1889 * config/i386/i386.c (override_options): Revert 2004-11-24 change.
1890 * config/i386/i386.md (fmodsf3, dremsf3, *sinsf2, *cossf2, sincossf3,
1891 *tansf3_1, tansf2, atan2sf3_1, atan2sf3, atansf2, asinsf2, acossf2,
1892 logsf2, log10sf2, log2sf2, log1psf2, logbsf2, ilogbsf2, expsf2,
1893 exp10sf2, exp2sf2, expm1sf2, rintsf2, floorsf2, ceilsf2, btruncsf2,
1894 nearbyintsf2): Disable for TARGET_SSE_MATH. Leave patterns enabled
1895 for TARGET_MIX_SSE_I387.
1896 (fmoddf3, dremdf3, *sindf2, *sinextendsfdf2, *cosdf2, *cosextendsfdf2,
1897 sincosdf3, *sincosextendsfdf3, *tandf3_1, tandf2, atan2df3_1,
1898 atan2df3, atandf2, asindf2, acosdf2, logdf2, log10df2, log2df2,
1899 log1pdf2, logbdf2, expdf2, exp10df2, exp2df2, expm1df2, rintdf2,
1900 floordf2, ceildf2, btruncdf2, nearbyintdf2): Disable for
1901 (TARGET_SSE2 && TARGET_SSE_MATH). Leave patterns enabled for
1902 TARGET_MIX_SSE_I387.
1903 (atan2sf3, atan2df3, atan2xf3): Remove register constraints
1906 2005-01-18 Uros Bizjak <uros@kss-loka.si>
1909 * config/i386/mmx.md (*movv2sf_internal_rex64, *movv2sf_internal):
1910 Add movaps alternative for xmm reg->reg move.
1912 2005-01-17 Jeff Law <law@redhat.com>
1914 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Clear SSA_NAME_VALUE
1915 at the end of the main DOM loop rather than just before DOM exits.
1917 2005-01-17 Diego Novillo <dnovillo@redhat.com>
1919 PR tree-optimization/19121
1920 * tree-ssa-alias.c (compute_flow_sensitive_aliasing): When
1921 adding aliases to a name tag, also add them to the pointer's
1923 * tree-ssa-copy.c (merge_alias_info): Do not merge flow
1924 sensitive alias info at all. Only check that the two pointers
1925 have compatible pointed-to sets.
1926 * tree-ssa.c (verify_name_tags): Verify that the alias set of
1927 a pointer's type tag is a superset of the alias set of the
1930 2005-01-17 James E Wilson <wilson@specifixinc.com>
1933 * config/ia64/ia64.md (movxf): Handle general register source. Adjust
1934 comment to document why.
1936 2005-01-17 Richard Henderson <rth@redhat.com>
1938 * config/i386/sse.md (smaxv4sf3_finite, sse_vmsmaxv4sf3_finite,
1939 sminv4sf3_finite, sse_vmsminv4sf3_finite, smaxv2df3_finite,
1940 sse2_vmsmaxv2df3_finite, sminv2df3_finite,
1941 sse2_vmsminv2df3_finite): New.
1942 (smaxv4sf3, sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, smaxv2df3,
1943 sse2_vmsmaxv2df3, sminv2df3, sse2_vmsminv2df3): Remove commutative.
1944 Force op1 into register.
1946 2005-01-17 Kelley Cook <kcook@gcc.gnu.org>
1948 * Makefile.in (STAGEMOVESTUFF): Stage all the stamp files to prevent
1949 bubblestrap from always rebuilding build directory.
1951 2005-01-17 Ulrich Weigand <uweigand@de.ibm.com>
1953 * config/s390/s390.md ("*clc"): Remove incorrect '='.
1954 ("*cmpmem_short"): Likewise.
1956 2005-01-17 Steven Bosscher <stevenb@suse.de>
1958 * basic-block.h: Document BB_* flags.
1959 * regrename.c (copyprop_hardreg_forward): Don't use BB_VISITED,
1960 use an sbitmap instead.
1961 * sched-rgn.c (compute_trg_info): Likewise.
1963 2005-01-17 Richard Sandiford <rsandifo@redhat.com>
1965 * config.gcc (mips64*-*-linux*): Set the default abi to n32. Remove
1966 redundant target_cpu_default and extra_parts lines. Use the default
1968 * config/mips/linux64.h (DRIVER_DEFAULT_ABI_SELF_SPEC): Delete.
1969 (SUBTARGET_EXTRA_SPECS): Delete.
1970 (DRIVER_SELF_SPECS): Remove architecture lines. Set the ABI directly.
1971 (SUBTARGET_ASM_SPEC): Remove -mabi=n32 and -mabi=64 mappings.
1972 * config/mips/t-linux64 (MULTILIB_DIRNAMES): Use "32" for -mabi=32
1973 and "n32" for -mabi=n32.
1975 2005-01-17 Ian Lance Taylor <ian@airs.com>
1978 * c-typeck.c (build_c_cast): Revert patch of 2003-02-16: don't
1979 fold constant variables into initial values.
1981 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
1983 * tree-cfg.c (tree_can_merge_blocks_p): Reorder two checks.
1985 * tree-cfg.c (tree_forwarder_block_p): Speed up by walking
1986 through the statements backward.
1988 2005-01-17 Ian Lance Taylor <ian@airs.com>
1990 PR middle-end/13127:
1991 * tree-inline.c (expand_call_inline): Set TREE_NO_WARNING on
1992 a variable set to the return value of the inlined function.
1994 2005-01-17 Mark Dettinger <dettinge@de.ibm.com>
1996 * rtlanal.c (rtx_cost): Assign cost of 0 to a SUBREG
1997 when modes are tieable.
1999 2005-01-17 Ranjit Mathew <rmathew@hotmail.com>
2001 * gthr-posix.h (__gthread_active_p): Use pthread_cancel instead
2002 of pthread_create to find out if threads are enabled.
2003 * gthr-posix95.h (__gthread_active_p): Likewise.
2005 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
2007 * common.opt (-fnew-ra): Remove.
2009 * toplev.h (flag_new_regalloc): Remove.
2010 * Makefile.in (ra*.*): Don't mention.
2011 * passes.c (rest_of_handle_new_regalloc): Remove.
2012 (rest_of_handle_combine, rest_of_compilation): Always consider
2013 flag_new_regalloc as false.
2014 * doc/invoke.texi: Don't document -fnew-ra.
2016 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
2018 * bb-reorder.c (fix_edges_for_rarely_executed_code): Remove
2019 last parameter to reg_scan.
2020 * loop.c (loop_optimize): Likewise.
2021 * passes.c (rest_of_handle_tracer, rest_of_handle_if_conversion,
2022 rest_of_handle_web, rest_of_handle_cfg, rest_of_handle_jump_bypass,
2023 rest_of_handle_life, rest_of_handle_cse, rest_of_handle_cse2,
2024 rest_of_handle_gcse, rest_of_handle_loop_optimize,
2025 rest_of_handle_loop2, rest_of_handle_jump2): Likewise.
2026 * regclass.c (reg_scan): Likewise, for the declaration.
2027 * rtl.h (reg_scan): Likewise, for the prototype.
2029 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
2031 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Don't call
2032 mark_dfs_back_edges.
2034 2005-01-16 Daniel Berlin <dberlin@dberlin.org>
2036 * tree-ssa-pre.c (add_to_sets): s1 may be NULL.
2037 (compute_avail): Uses don't go in tmp_gen.
2039 2005-01-16 Steven Bosscher <stevenb@suse.de>
2041 * ggc-page.c (ggc_alloc_stat): Use __builtin_ctzl instead of a
2042 loop to look for a free slot in a page entry.
2044 2005-01-16 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2047 * defaults.h (TARGET_DEFERRED_OUTPUT_DEFS): Provide default.
2048 * toplev.c (compile_file): Call process_pending_assemble_output_defs
2049 just before targetm.asm_out.file_end.
2050 * tree.h (process_pending_assemble_output_defs): Declare.
2051 * varasm.c (assemble_output_def, process_pending_assemble_output_defs):
2053 (assemble_alias): Defer generation of assembly code for defines when
2054 TARGET_DEFERRED_OUTPUT_DEFS is true.
2055 * config/rs6000/aix41.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
2056 * config/rs6000/aix43.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
2057 * doc/tm.texi (TARGET_DEFERRED_OUTPUT_DEFS): document.
2059 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2062 * pa.c (pa_scalar_mode_supported_p): New function.
2063 (TARGET_SCALAR_MODE_SUPPORTED_P): Define.
2064 * pa.h (MIN_UNITS_PER_WORD): Add comment.
2066 * pa-protos.h (prefetch_operand): Delete.
2067 (prefetch_cc_operand, prefetch_nocc_operand): New declations.
2068 * pa.c (prefetch_operand): Delete.
2069 (prefetch_cc_operand, prefetch_nocc_operand): New functions.
2070 * pa.h (EXTRA_CONSTRAINT): Add `W' constraint.
2071 (PREDICATE_CODES): Delete prefetch_operand. Add prefetch_cc_operand
2072 and prefetch_nocc_operand.
2073 * pa.md (prefetch): Rework to avoid reload problems handling short
2074 displacements when a cache control completer needs to be provided.
2075 (prefetch_32, prefetch_64): Delete.
2076 (prefetch_cc, prefetch_nocc): New patterns.
2078 2005-01-15 David Edelsohn <edelsohn@gnu.org>
2080 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Revert previous change.
2082 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2085 * pa.c (pa_asm_output_aligned_common): Warn if specified alignment
2086 exceeds maximum alignment for global common data.
2087 * pa64-hpux.h (MAX_OFILE_ALIGNMENT): Define to 32768.
2088 * som.h (MAX_OFILE_ALIGNMENT): Likewise.
2090 2004-01-15 Roger Sayle <roger@eyesopen.com>
2092 * tree-ssa-dom.c (extract_range_from_cond): Correct condition.
2094 2004-01-15 Roger Sayle <roger@eyesopen.com>
2096 * harg-reg-set.h (reg_class_names): Prototype global array.
2097 * regclass.c (reg_class_names): Declare here and initialize to
2099 (dump_regclass): Remove local declaration of reg_class_names.
2100 (regclass): Likewise.
2101 * cfg.c (dump_flow_info): Likewise.
2102 * ra-debug.c (reg_class_names): Likewise.
2103 * regrename.c (reg_class_names): Likewise.
2104 * reload.c (reg_class_names): Likewise.
2105 * reload1.c (spill_failure): Likewise.
2106 * config/m68hc11/m68hc11.c (reg_class_names): Likewise.
2108 2005-01-15 Ulrich Weigand <uweigand@de.ibm.com>
2110 * config/s390/s390.md ("reload_outti"): Remove predicate for
2111 output operand. Abort if operand is not a MEM.
2112 ("reload_outdi", "reload_outdf"): Likewise.
2114 2005-01-15 Marc Espie <espie@openbsd.org>
2116 * config.gcc (*-*-openbsd*): Set HAS_LIBC_R for OpenBSD <= 3.2.
2117 (i[34567]86-*-openbsd*): Switch to ELF for OpenBSD >= 3.4.
2118 * config/openbsd.h: Fix C++ includes for native configurations.
2119 Add proper OS_CPP_BUILTINS. Fix libspec for recent OpenBSD.
2120 Add trampoline support.
2121 * config/i386/openbsdelf.h: New.
2123 2005-01-15 Marc Espie <espie@openbsd.org>
2125 * collect2.c (main): Explicitly parse -dynamic-linker option.
2127 2005-01-15 Jakub Jelinek <jakub@redhat.com>
2129 PR tree-optimization/19060
2130 * tree-ssa-dom.c (extract_range_from_cond) <case LT_EXPR, GT_EXPR>:
2131 Return 0 if op1 <= TYPE_MIN_VALUE () resp. op1 >= TYPE_MAX_VALUE ().
2132 (simplify_cond_and_lookup_avail_expr): Add assert for dummy == 0
2133 and handle extract_range_from_cond returning false.
2134 * fold-const.c (fold): Optimize comparisons with min/max even for
2135 width > HOST_BITS_PER_WIDE_INT.
2137 2005-01-15 Ralf Corsepius <ralf.corsepius@rtems.org>
2139 * config/mips/rtems.h (MIPS_DEFAULT_GVALUE): Set to 0.
2140 * config/mips/t-rtems (MULTILIBS_DIRNAMES,MULTILIB_OPTIONS):
2141 Remove little endian multilib variants.
2142 Add mips32 multilib variant.
2144 2005-01-15 Kazu Hirata <kazu@cs.umass.edu>
2146 * explow.c (copy_all_reg, stabilize): Remove.
2147 * expr.h: Remove the corresponding prototypes.
2149 * rtlanal.c (regs_set_between_p): Remove.
2150 * rtl.h: Remove the corresponding prototype.
2152 2005-01-14 David Edelsohn <edelsohn@gnu.org>
2154 * config/rs6000/rs6000.c (rs6k_nonimmediate_operand): Rename
2155 to rs6000_nonimmediate_operand.
2156 * config/rs6000/rs6000.h (PREDICATE_CODES): Rename
2157 rs6k_nonimmediate_operand to rs6000_nonimmediate_operand.
2158 * config/rs6000/rs6000.md (movsi_internal1): Rename predicate to
2159 rs6000_nonimmediate_operand.
2160 * config/rs6000/spe.md (movdf_e500_double): Rename predicate to
2161 rs6000_nonimmediate_operand.
2163 2005-01-14 Andrew Pinski <pinskia@physics.uc.edu>
2165 * tree-ssa-alias.c (pass_may_alias): Add TODO_verify_stmts.
2167 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
2169 * config/rs6000/rs6000.h (CLASS_MAX_NREGS): DF goes in 1 register
2171 (CANNOT_CHANGE_MODE_CLASS): Restrict DI mode changes on e500v2.
2172 (PREDICATE_CODES): Add rs6k_nonimmediate_operand.
2174 * config/rs6000/rs6000.c (invalid_e500_subreg): New.
2175 (rs6k_nonimmediate_operand): New.
2176 (rs6000_legitimate_offset_address_p): Handle DI modes on e500v2
2178 (legitimate_lo_sum_address_p): Same.
2179 (rs6000_legitimize_address): Same.
2180 (rs6000_legitimize_reload_address): Same.
2181 (rs6000_legitimate_address): Same.
2182 (spe_build_register_parallel): Pass DF and DC modes in a DI
2185 * config/rs6000/rs6000.md ("*movsi_internal1"): Change predicate
2186 to rs6k_nonimmediate_operand.
2188 * config/rs6000/spe.md ("*frob_df_di"): New.
2189 ("*frob_di_df"): New.
2190 ("*frob_di_df_2"): New.
2191 ("*mov_sidf_e500_subreg0"): New.
2192 ("*mov_sidf_e500_subreg4"): New.
2193 ("*movdf_e500_double"): Change predicate to
2194 rs6k_nonimmediate_operand.
2196 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
2198 * postreload.c (move2add_note_store): Only call
2199 trunc_int_for_mode on scalar integers.
2201 2005-01-14 Steven Bosscher <stevenb@suse.de>
2203 * tree-ssa-dce.c (visited_control_parents): New sbitmap to
2204 replace BB_VISITED uses.
2205 (find_obviously_necessary_stmts): Don't clear BB_VISITED.
2206 (propagate_necessity): Check the bitmap instead of BB_VISITED.
2207 (tree_dce_done): Free visited_control_parents.
2208 (perform_tree_ssa_dce): Allocate and clear it.
2209 * tree-ssa-pre.c (compute_antic_aux): Make non-recursive.
2210 (compute_antic): Iterate from here using a DFS. Use an sbitmap
2211 instead of BB_VISITED.
2213 2005-01-14 Kazu Hirata <kazu@cs.umass.edu>
2215 * c-tree.h, coverage.h, langhooks-def.h, optabs.h, output.h,
2216 rtl.h, targhooks.h, tree-flow.h, tree-mudflap.h, tree.h:
2217 Remove unused prototypes.
2219 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
2221 * config/rs6000/rs6000.c (rs6000_stack_info): Fix caching of SPE
2222 64-bit register usage.
2223 (spe_func_has_64bit_regs_p): Add FIXME note.
2225 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
2227 * config/rs6000/rs6000.c (rs6000_generate_compare): Replace
2228 flag_finite_math_only with flag_unsafe_math_optimizations.
2230 2005-01-14 J. D. Johnston <jjohnst@us.ibm.com>
2232 * config/s390/tpf-unwind.h (s390_fallback_frame_state): Correct
2234 (__tpf_eh_return): Copy TPF private stack area from the module
2235 boundary stack frame. Add check for when module addresses are
2236 equal, but no stub address is found.
2238 2005-01-14 Richard Earnshaw <rearnsha@arm.com>
2241 * arm.h (struct machine_function): Add call_via field.
2242 (thumb_call_via_label): Declare.
2243 * arm.c (thumb_call_via_label): New variable.
2244 (thumb_call_reg_needed): New variable.
2245 (arm_output_function_epilogue): For Thumb code, output any per-function
2246 call-indirect trampolines.
2247 (thumb_call_via_reg): New function.
2248 (arm_file_end): New function.
2249 (TARGET_ASM_FILE_END): Call arm_file_end.
2250 (aof_file_end): Likewise.
2251 * arm-protos.h (thumb_call_via_reg): Declare.
2252 * arm.md (call_reg_thumb, call_value_reg_thumb): Call
2253 thumb_call_via_reg in normal case.
2255 2005-01-14 Jakub Jelinek <jakub@redhat.com>
2258 PR rtl-optimization/19348
2259 * recog.c (peephole2_optimize): Do global life update if some peephole
2260 decides it doesn't need at least one of its inputs and that change
2261 influences liveness at the start of the basic block.
2263 * basic-block.h (EXECUTE_IF_AND_COMPL_IN_REG_SET): Needs 2 REGSET
2264 arguments instead of 1.
2266 2005-01-14 Eric Botcazou <ebotcazou@libertysurf.fr>
2269 * varasm.c (initializer_constant_valid_p) <ADDR_EXPR>: Return
2270 zero for nested functions needing a static chain or functions
2271 with a non-constant address.
2273 2005-01-13 Roger Sayle <roger@eyesopen.com>
2275 * simplify-rtx.c (simplify_binary_operation) <AND>: Optimize
2276 (and (sign_extend X) C) into (zero_extend (and X C)).
2278 2005-01-13 David O'Brien <obrien@FreeBSD.org>
2280 * config/freebsd-spec.h: Make KSE pthread lib logic the default.
2282 2005-01-13 Richard Henderson <rth@redhat.com>
2287 * config/i386/i386.md (cmpdf, cmpsf, bunordered, bordered, buneq,
2288 bunge, bungt, bunle, bunlt, bltgt): Enable for TARGET_SSE_MATH,
2289 not just TARGET_SSE.
2290 (cmpfp_i_387): Rename from cmpfp_i. Move after sse patterns.
2291 (cmpfp_i_mixed): Rename from cmpfp_i_sse; use for TARGET_MIX_SSE_I387.
2292 (cmpfp_i_sse): Rename from cmpfp_i_sse_only; use for TARGET_SSE_MATH.
2293 (cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387): Similarly.
2294 (fp_jcc_1_mixed, fp_jcc_1_sse, fp_jcc_1_387): Similarly.
2295 (fp_jcc_2_mixed, fp_jcc_2_sse, fp_jcc_2_387): Similarly.
2296 (fp_jcc_3_387, fp_jcc_4_387, fp_jcc_5_387, fp_jcc_6_387,
2297 fp_jcc_7_387, fp_jcc_8_387): Rename from fp_jcc_N.
2298 (movdicc_c_rex64): Rename with '*'.
2299 (movsfcc, movdfcc): Add checks for 387 and sse math to condition.
2300 (movsfcc_1_sse_min, movsfcc_1_sse_max, movsfcc_1_sse): New.
2301 (movsfcc_1_387): Rename from movsfcc_1.
2302 (movdfcc_1_sse_min, movdfcc_1_sse_max, movdfcc_1_sse): New.
2303 (movdfcc_1, movdfcc_1_rex64): Add check for 387.
2304 (sminsf3, smaxsf3, smindf3, smaxdf3): New.
2305 (minsf3, minsf, minsf_nonieee, minsf_sse, mindf3, mindf,
2306 mindf_nonieee, mindf_sse, maxsf3, maxsf, maxsf_nonieee, maxsf_sse,
2307 maxdf3, maxdf, maxdf_nonieee, maxdf_sse, sse_movsfcc, sse_movsfcc_eq,
2308 sse_movdfcc, sse_movdfcc_eq, sse_movsfcc_const0_1,
2309 sse_movsfcc_const0_2, sse_movsfcc_const0_3, sse_movsfcc_const0_4,
2310 sse_movdfcc_const0_1, sse_movdfcc_const0_2, sse_movdfcc_const0_3,
2311 sse_movdfcc_const0_4): Remove.
2312 * config/i386/i386.c (ix86_expand_fp_movcc): For TARGET_SSE_MATH,
2313 recognize min/max early. Update for changed sse cmove patterns.
2314 (ix86_split_sse_movcc): New.
2315 * config/i386/i386-protos.h: Update.
2317 2005-01-13 Steven Bosscher <stevenb@suse.de>
2319 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
2320 (fix_stmt_v_may_defs): Likewise.
2322 2005-01-13 Richard Henderson <rth@redhat.com>
2324 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
2325 for converting the mask.
2327 2005-01-13 David Edelsohn <edelsohn@gnu.org>
2329 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
2332 2005-01-13 Richard Henderson <rth@redhat.com>
2334 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
2335 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
2336 IX86_BUILTIN_MOVDQ2Q): Remove.
2337 (IX86_BUILTIN_VEC_EXT_V4SI): New.
2338 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
2339 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
2341 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
2342 (sse2_pextrw): Fix immediate constraint.
2343 (sse2_loadq, sse2_loadq_rex64): Remove.
2344 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
2345 Use __builtin_ia32_vec_ext_<size>.
2346 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
2348 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
2350 * function.c (assign_parm_setup_block): Look inside original
2351 entry_parm when inspecting PARALLEL.
2353 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
2354 Joel Sherrill <joel@oarcorp.com>
2357 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
2358 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
2359 rtems_gxx_recursive_mutex_init.
2360 (__gthread_recursive_mutex_lock): New function.
2361 (__gthread_recursive_mutex_trylock): Likewise.
2362 (__gthread_recursive_mutex_unlock): Likewise.
2364 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
2366 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
2368 2005-01-13 Jan Beulich <jbeulich@novell.com>
2370 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
2371 (ashrti3_internal): Indicate output is early clobber. Generate result
2372 into output rather than first input. Use move for low word of output
2373 if shift count is exactly 64.
2374 (lshrti3_internal): Likewise.
2376 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
2380 * reload1.c (delete_output_reload): Don't delete an output reload
2381 if the pseudo lives longer than a single basic block. Adjust and
2382 improve wording of comment.
2384 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
2387 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
2388 of TYPE_SIZE where we mean the number of bytes.
2390 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
2392 * function.c (assign_parm_setup_block): Relax condition on
2393 multi-register optimization.
2395 2005-01-12 Nick Clifton <nickc@redhat.com>
2397 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
2398 an "in_delay_slot" attribute of "no" to prevent them being used in
2399 delay slots. This is forbidden because they might generate
2402 2005-01-12 Alan Modra <amodra@bigpond.net.au>
2405 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
2408 2005-01-12 Nick Clifton <nickc@redhat.com>
2410 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
2411 passed on to GAS) but do not define, as GAS no longer supports or
2412 needs the -m2000 option.
2414 2005-01-11 Roger Sayle <roger@eyesopen.com>
2416 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
2417 type as args of unprototyped builtins aren't automatically promoted.
2418 (fold_builtin_abs): Likewise.
2420 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2422 * Makefile.in: Set a `build-warn' variable.
2424 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
2426 PR tree-optimization/17949
2427 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
2428 (find_interesting_uses_address): Use it.
2430 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
2432 * regrename.c (kill_value): Handle subreg's that won't simplify.
2434 2005-01-11 Richard Henderson <rth@redhat.com>
2437 * config/i386/i386.h (enum ix86_builtins): Move ...
2438 * config/i386/i386.c: ... here.
2439 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
2440 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
2441 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
2442 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
2443 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
2444 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
2445 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
2446 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
2447 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
2448 IX86_BUILTIN_LOADDDUP): Remove.
2449 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
2450 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
2451 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
2452 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
2453 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
2454 (ix86_init_builtins): Make static.
2455 (ix86_init_mmx_sse_builtins): Update for changed builtins.
2456 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
2457 if all the modes match. Otherwise, fake it.
2458 (get_element_number, ix86_expand_vec_init_builtin,
2459 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
2460 (ix86_expand_builtin): Make static. Update for changed builtins.
2461 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
2462 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
2463 (ix86_expand_vector_init_duplicate): New.
2464 (ix86_expand_vector_init_low_nonzero): New.
2465 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
2466 Split out from ix86_expand_vector_init; handle integer modes.
2467 (ix86_expand_vector_init): Use them.
2468 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
2469 * config/i386/i386-protos.h: Update.
2470 * config/i386/predicates.md (reg_or_0_operand): New.
2471 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
2472 (movv2sf_internal): Likewise. And a splitter to match them all.
2473 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
2474 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
2475 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
2476 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
2477 (mmx_pinsrw): Fix operand ordering.
2478 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
2479 rather than sse_loadss expander.
2480 (movv2df splitter): Similarly.
2481 (sse_loadss, sse_loadlss): Remove.
2482 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
2483 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
2484 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
2485 (sse3_movddup): Rename with '*'.
2486 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
2487 (sse2_loadsd): Remove.
2488 (vec_dupv2df_sse3): Rename from sse3_loadddup.
2489 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
2490 (sse2_pinsrw): Fix argument ordering.
2491 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
2492 (sse2_stored): Remove 'r' destination.
2493 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
2494 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
2495 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
2496 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
2497 vec_extractv16qi, vec_initv16qi): New.
2499 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
2500 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
2501 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
2503 (_mm_load_pd, _mm_store_pd): Use plain dereference.
2504 (_mm_load_si128, _mm_store_si128): Likewise.
2505 (_mm_load1_pd): Use _mm_set1_pd.
2506 (_mm_load_sd): Use _mm_set_sd.
2507 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
2508 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
2509 (_mm_set_epi64): Use _mm_set_epi64x.
2510 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
2511 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
2512 _mm_setr_epi8): Use _mm_set_foo form.
2513 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
2514 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
2515 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
2516 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
2517 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
2518 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
2519 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
2520 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
2521 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
2522 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
2523 (_mm_movedup_pd): Use _mm_shuffle_pd.
2524 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
2525 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
2526 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
2527 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
2529 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
2530 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
2531 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
2532 (_mm_store_ps): Use raw dereference.
2533 (_mm_store1_ps): Use _mm_storeu_ps.
2534 (_mm_storer_ps): Use _mm_store_ps.
2535 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
2536 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
2538 2005-01-11 Stan Shebs <shebs@apple.com>
2540 * config/rs6000/rs6000.c (machopic_output_stub): Issue
2541 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
2543 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
2546 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
2547 Skip all DECLs except for FIELD_DECLs.
2549 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
2551 * config/s390/s390.c (override_options): Return error if
2552 -mbackchain, -mpacked-stack and -mhard-float are used together.
2553 (s390_va_start): Remove the backchain && packed-stack special case.
2554 (s390_gimplify_va_arg): Likewise.
2555 * doc/invoke.texi: Remove the ABI incompatibility note.
2557 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
2559 * config/s390/s390.c (struct s390_frame_layout): Remove
2561 (s390_frame_info, s390_emit_prologue): Replace occurrences of
2562 save_backchain_p with TARGET_BACKCHAIN.
2564 2005-01-11 Alan Modra <amodra@bigpond.net.au>
2567 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
2568 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
2569 * function.c (locate_and_pad_parm): Set new field.
2570 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
2571 Tweak where_pad test to include "none". Always set mem align for
2573 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
2574 not sufficient for type.
2575 (assign_parm_setup_block): If stack_parm is zero on entry, always
2576 make a new stack local. Block move old stack parm if necessary
2577 to new aligned stack local.
2578 (assign_parm_setup_stack): Use a block move to handle
2579 potentially misaligned entry_parm.
2580 (assign_parms_unsplit_complex): Specify required alignment when
2581 creating stack local.
2582 * calls.c (compute_argument_addresses): Override alignment of stack
2583 arg calculated from its type with the alignment given by
2584 FUNCTION_ARG_BOUNDARY.
2585 (store_one_arg): Likewise.
2587 2005-01-11 Jan Beulich <jbeulich@novell.com>
2589 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
2590 Change respective itanium_class attribute to ialu.
2591 (shladdp4_internal): New.
2592 * config/ia64/predicates.md (shladd_log2_operand): New.
2594 2005-01-11 Richard Henderson <rth@redhat.com>
2596 * expr.c (store_constructor): Use rtvec_alloc instead of
2597 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
2599 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
2601 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
2603 2005-01-11 Alan Modra <amodra@bigpond.net.au>
2605 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
2606 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
2607 (TARGET_RWRELOC): Define.
2608 (ia64_rwreloc_section_type_flags): Delete.
2609 (ia64_section_type_flags): New function.
2610 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
2611 (TARGET_RWRELOC): Define.
2613 2005-01-10 David Mosberger <davidm@hpl.hp.com>
2616 * config/ia64/ia64.c (process_set): For alloc insn, only call
2617 process_epilogue is !frame_pointer_needed.
2619 2005-01-10 Roger Sayle <roger@eyesopen.com>
2622 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
2623 unary operator and can't be treated as a binary/comparison operator.
2625 2005-01-10 Richard Henderson <rth@redhat.com>
2627 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
2628 (ix86_return_in_memory): Likewise.
2629 (function_arg_advance): Likewise. Mirror structure in function_arg
2630 for choosing register to advance.
2632 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
2634 * tree-vectorizer.c, tree.def: Fix comment typos.
2636 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
2638 * config/s390/s390.c (struct processor_costs): 4 new fields:
2639 ddbr, ddr, debr, der.
2640 (s390_rtx_costs): More precise handling of divide instructions.
2642 2005-01-10 David Edelsohn <edelsohn@gnu.org>
2644 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
2646 2005-01-10 Jan Beulich <jbeulich@novell.com>
2648 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
2649 names resulting from -ffunction-sections/-fdata-sections and linkonce
2651 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
2652 based on the section name. Rearrange the section name comparison logic
2653 slightly so that each section name is compared against at most once.
2655 2005-01-10 Ben Elliston <bje@au.ibm.com>
2657 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
2658 targets for which -fPIC reduces limitations on the GOT size.
2660 2005-01-09 Falk Hueffner <falk@debian.org>
2662 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
2665 2005-01-09 David Edelsohn <edelsohn@gnu.org>
2668 * collect2.c (main): Set aixrtl_flag for -brtl option.
2669 (resolve_lib_name): Search for .so file extension before .a
2672 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
2674 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
2675 (vect_analyze_data_refs_alignment): Add dump prints.
2677 2005-01-09 Ira Rosen <irar@il.ibm.com>
2679 * tree-vectorizer.c (vect_analyze_offset_expr): Use
2680 expr_invariant_in_loop_p.
2681 Initialize outputs first thing in the function.
2682 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
2683 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
2684 (vect_analyze_pointer_ref_access): Check that the initial condition of
2685 the access function is loop invariant.
2687 2005-01-09 Richard Henderson <rth@redhat.com>
2689 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
2690 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
2691 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
2692 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
2693 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
2694 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
2695 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
2696 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
2697 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
2698 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
2699 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
2700 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
2701 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
2702 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
2703 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
2704 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
2705 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
2706 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
2707 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
2708 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
2709 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
2710 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
2711 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
2713 (mmx_clrdi, pavgusb): Remove.
2714 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
2715 with leading sse_ prefix.
2716 * config/i386/sse.md: Receive them.
2717 * config/i386/mmx.md: New file.
2718 (MMXMODE12, MMXMODE24, mmxvecsize): New.
2719 (subrv2sf3): Turn into expander for normal subtraction.
2720 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
2721 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
2722 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
2723 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
2724 commutative; use ix86_binary_operator_ok.
2725 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
2726 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
2727 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
2728 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
2729 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
2730 Macroize from existing patterns; use ix86_binary_operator_ok.
2731 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
2732 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
2733 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
2734 with vec_select+vec_concat.
2735 (mmx_pshufw, mmx_pshufw_1): Likewise.
2736 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
2737 (mmx_uavgv4hi3): Model correctly.
2738 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
2739 _mm_xor_si64): Remove casts.
2741 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
2743 PR tree-optimization/19224
2744 * tree-scalar-evolution.c (get_instantiated_value,
2745 set_instantiated_value): New functions.
2746 (instantiate_parameters_1): Cache the results.
2747 (instantiate_parameters, resolve_mixers): Initialize and free
2750 2005-01-08 David Edelsohn <edelsohn@gnu.org>
2752 * config/i386/i386.md (addhi_4): Correct reference in comment.
2755 2005-01-08 Richard Henderson <rth@redhat.com>
2757 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
2758 (_mm_cvtsi128_si64x): Likewise.
2759 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
2760 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
2761 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
2762 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
2763 * config/i386/i386-protos.h: Update.
2764 * config/i386/i386.c (print_operand): Add 'H'.
2765 (ix86_fixup_binary_operands): Split out from ...
2766 (ix86_expand_binary_operator): ... here.
2767 (ix86_fixup_binary_operands_no_copy): New.
2768 (ix86_expand_fp_absneg_operator): Handle vector mode results.
2769 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
2770 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
2771 (safe_vector_operand): Use CONST0_RTX.
2772 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
2773 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
2774 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
2775 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
2777 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
2778 (IX86_BUILTIN_CMPNGESS): New.
2779 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
2780 (attr type): Add sselog1.
2781 (attr unit, attr memory): Handle it.
2782 (movti, movti_internal, movti_rex64): Move near other integer moves.
2783 (movtf, movtf_internal): Move near other fp moves.
2784 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
2785 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
2786 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
2787 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
2788 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
2789 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
2790 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
2791 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
2792 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
2793 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
2794 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
2795 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
2796 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
2797 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
2798 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
2799 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
2800 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
2801 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
2802 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
2803 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
2804 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
2805 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
2806 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
2807 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
2808 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
2809 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
2810 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
2811 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
2812 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
2813 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
2814 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
2815 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
2816 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
2817 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
2818 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
2819 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
2820 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
2821 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
2822 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
2823 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
2824 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
2825 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
2826 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
2827 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
2829 (SSEPUSH, push<SSEPUSH>): Remove.
2830 (MMXPUSH, push<MMXPUSH>): Remove.
2831 (sse_movaps, sse_movaps_1, sse_movups): Remove.
2832 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
2833 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
2834 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
2835 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
2836 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
2837 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
2838 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
2839 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
2840 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
2841 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
2842 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
2843 (ppro_sse_log_V4SF): Handle sselog1.
2844 * config/i386/predicates.md (const_0_to_1_operand): New.
2845 (const_0_to_255_mul_8_operand): New.
2846 (const_1_to_31_operand): Rename from const_int_1_31_operand.
2847 (const_2_to_3_operand, const_4_to_7_operand): New.
2848 * config/i386/sse.md: New file.
2849 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
2850 (sse_movups): Rename from sse_movups_1.
2851 (sse_loadlss): Rename from sse_loadss_1.
2852 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
2854 (negv4sf2): Use ix86_expand_fp_absneg_operator.
2855 (absv4sf2, negv2df, absv2df): New.
2856 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
2857 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
2858 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
2859 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
2860 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
2861 (sse3_addsubv4sf3): Model correctly.
2862 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
2863 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
2864 (sse_movhlps): Model with vec_select+vec_concat.
2865 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
2866 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
2867 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
2868 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
2869 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
2870 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
2871 sse2_pshufhw_1): Likewise.
2872 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
2873 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
2874 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
2875 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
2876 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
2877 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
2878 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
2879 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
2880 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
2881 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
2882 sminv8hi3): Mark commutative
2883 operands. Use ix86_binary_operator_ok.
2884 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
2885 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
2886 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
2887 sse2_punpcklqdq): Allow operand2 in memory.
2888 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
2889 sse2_movsd): Add memory alternatives.
2890 (sse_storelps): Turn expander into an insn; split after reload.
2891 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
2892 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
2894 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
2896 * configure.ac (DWARF-2 debug_line): Use objdump.
2897 * configure: Regenerate.
2899 2005-01-08 Jeff Law <law@redhat.com>
2900 Diego Novillo <dnovillo@redhat.com>
2902 PR tree-optimization/18241
2903 * tree-nrv.c (tree_nrv): Ignore volatile return values.
2904 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
2905 statements with volatile operands.
2906 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
2907 operands after marking a statement with has_volatile_ops.
2909 2005-01-08 Roger Sayle <roger@eyesopen.com>
2911 * tree.c (int_fits_type_p): Always honor integer constant
2912 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
2914 2005-01-08 Roger Sayle <roger@eyesopen.com>
2916 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
2917 the latent bug in rtl_delete_block has been resolved.
2919 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
2921 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
2922 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
2924 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
2926 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
2927 library as @multilib_dir@/$(SHLIB_SONAME).
2929 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
2931 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
2932 * configure: Regenerate.
2934 2005-01-07 Jakub Jelinek <jakub@redhat.com>
2936 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
2939 2005-01-07 David Edelsohn <edelsohn@gnu.org>
2942 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
2943 Convert non-word aligned offset address using ld/std into
2946 2005-01-07 Richard Henderson <rth@redhat.com>
2948 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
2950 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
2952 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
2954 2005-01-07 Jakub Jelinek <jakub@redhat.com>
2956 PR tree-optimization/19283
2957 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
2958 is not shorter than the original type.
2960 PR rtl-optimization/19012
2961 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
2963 PR rtl-optimization/18861
2964 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
2965 (FULL_STATE): Define.
2966 (mark_tablejump_edge): New function.
2967 (purge_dead_tablejump_edges): New function.
2968 (find_bb_boundaries): Use it.
2970 PR tree-optimization/18828
2971 * builtins.c (expand_builtin_next_arg): Remove argument and all
2972 the argument checking.
2973 (expand_builtin): Adjust caller.
2974 (expand_builtin_va_start): Likewise. Remove error for too many
2976 (fold_builtin_next_arg): Issue error for too many arguments.
2977 After checking arguments, replace them with magic arguments that
2978 prevent further checking of the args.
2980 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2982 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
2984 2005-01-06 Roger Sayle <roger@eyesopen.com>
2987 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
2988 and %{.S:...} (and their negative variants) to test whether the
2989 input file is assembler or pre-processed-assembler independent of
2990 the actual filename extension.
2992 2005-01-06 Roger Sayle <roger@eyesopen.com>
2994 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
2995 of sign or zero extended values.
2997 2005-01-06 Geoffrey Keating <geoffk@apple.com>
2999 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
3000 for IBM long double format correctly.
3002 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
3004 Fix PR tree-optimization/18792
3006 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
3007 to first_loop_depth, and use loop depth instead of loop number.
3008 (build_classic_dir_vector): Ditto.
3009 (compute_data_dependences_for_loop): Use depth, not loop number.
3010 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
3011 number. Pass in loops, instead of loop numbers.
3012 (gather_interchange_stats): Ditto.
3013 (linear_transform_loops): Ditto.
3015 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
3018 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
3019 biv_fits_mode_p, extension_within_bounds_p): New functions.
3020 (check_ext_dependent_givs): Use them.
3022 2005-01-06 Roger Sayle <roger@eyesopen.com>
3024 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
3025 more than one barrier, in which case we should delete them all.
3027 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
3029 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
3031 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
3033 Fix PR middle-end/19286
3035 * dwarf2out.c (gen_subprogram_die): If we've already tried to
3036 output this subprogram, simply ignore this attempt to do it again.
3037 (add_abstract_origin_attribute): Don't abort trying to add the abstract
3038 origin attribute if it's not possible.
3039 (gen_block_die): Don't ignore subblocks of "unused" blocks.
3040 (decls_for_scope): Ditto.
3041 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
3042 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
3044 2005-01-05 Richard Henderson <rth@redhat.com>
3047 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
3048 (bdesc_2arg): Use it.
3049 (ix86_expand_binop_builtin): Force operands into registers
3051 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
3052 ix86_expand_sse_compare, ix86_expand_sse_comi,
3053 ix86_expand_builtin): Likewise.
3055 2005-01-05 Richard Henderson <rth@redhat.com>
3057 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
3059 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
3060 (mulv4hi3): Set itanium_class mmmul.
3061 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
3062 (fpack): Rename from fpack_sfsf.
3064 2005-01-05 Richard Henderson <rth@redhat.com>
3067 * reload1.c (do_input_reload): Restrict the optimization deleteing
3068 a previous output reload to RELOAD_FOR_INPUT.
3070 2005-01-05 Steven Bosscher <stevenb@suse.de>
3072 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
3073 comment that we fall through after case.
3074 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
3075 (distribute_links): Likewise.
3076 * cse.c (cse_insn): Likewise.
3077 * cselib.c (cselib_invalidate_mem): Likewise.
3078 * df.c: Update comments at the top of the file.
3079 (read_modify_subreg_p): Update comments here too.
3080 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
3081 * flow.c (mark_set_1): Likewise.
3082 (mark_used_regs): Likewise.
3083 * gcse.c (mems_conflict_for_gcse_p): Likewise.
3084 (canon_list_insert): Likewise.
3085 (mark_set): Likewise.
3086 (try_replace_reg): Likewise.
3087 (store_killed_in_insn): Likewise.
3088 * loop.c (count_one_set): Likewise.
3089 (basic_induction_var): Likewise.
3090 * postreload-gcse.c (find_mem_conflicts): Likewise.
3091 * postreload.c (reload_combine_note_store): Likewise.
3092 (move2add_note_store): Likewise.
3093 * reload.c (find_equiv_reg): Likewise.
3094 (mark_referenced_resources): Likewise.
3095 * rtlanal.c (set_noop_p): Likewise.
3096 (note_stores): Likewise.
3097 (note_uses): Likewise.
3098 * sched-deps.c (sched_analyze_1): Likewise.
3099 * sched-rgn.c (check_live_1): Likewise.
3100 (update_live_1): Likewise.
3101 * config/i860/i860.c: Likewise.
3103 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
3105 (ZERO_EXTRACT): Mention that this one can be an lvalue.
3107 * doc/rtl.texi: Update documentation for bit-fields and SET.
3109 2005-01-05 Roger Sayle <roger@eyesopen.com>
3111 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
3113 2005-01-05 Jan Hubicka <jh@suse.cz>
3114 Richard Henderson <rth@redhat.com>
3117 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
3120 2005-01-05 Richard Henderson <rth@redhat.com>
3123 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
3124 (sse_shufps): Change operand 3 to const_int_operand.
3125 (sse2_storelps): Fix typo in template.
3126 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
3127 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
3128 (ix86_expand_builtin): Likewise.
3130 2005-01-05 Julian Brown <julian@codesourcery.com>
3132 * config/arm/arm.c (arm_return_in_memory): Treat complex types
3133 as aggregates for AAPCS ABIs.
3135 2005-01-05 Stan Shebs <shebs@apple.com>
3137 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
3138 Mach-O getters if ppc64.
3139 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
3141 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
3142 signed four-byte field for global code case.
3143 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
3145 2005-01-05 Roger Sayle <roger@eyesopen.com>
3148 * c-common.c: Include real.h.
3149 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
3150 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
3151 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
3152 * Makefile.in (c-common.o): Update dependencies.
3154 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
3156 * c-parse.in (asm_string): Add trailing semicolon.
3158 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
3160 * c-parse.in (asm_string): New. Don't allow wide strings in
3162 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
3163 asm_string instead of STRING.
3165 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
3167 * c-typeck.c (constructor_no_implicit): Remove.
3168 (set_designator, process_init_element): Don't check
3169 constructor_no_implicit.
3171 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
3172 Kaz Kojima <kkojima@gcc.gnu.org>
3175 * lcm.c (create_pre_exit): New.
3176 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
3177 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
3179 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
3181 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
3182 -msoft-float in target_float_switch.
3183 * arm.c (arm_override_options): Fix processing of target_float_switch.
3185 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
3187 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
3190 2004-01-05 Uros Bizjak <uros@kss-loka.si>
3192 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
3193 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
3195 2005-01-04 Roger Sayle <roger@eyesopen.com>
3197 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
3198 if deleting the then-block allows the test-block to fallthru to the
3201 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
3204 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
3205 after the full declaration if the are in two different TUs.
3207 2005-01-04 Richard Henderson <rth@redhat.com>
3210 * tree-sra.c (generate_one_element_init): Just
3211 call gimplify_and_add.
3212 (generate_element_init): Record the
3213 new referenced variables and mark them for renaming
3214 and split out to ...
3215 (generate_element_init_1): This.
3216 (scalarize_init): Don't call push_gimplify_context/
3217 pop_gimplify_context.
3219 2005-01-04 Geoffrey Keating <geoffk@apple.com>
3221 * toplev.c (get_src_pwd): Handle failure of getpwd().
3223 2005-01-04 Roger Sayle <roger@eyesopen.com>
3225 * fold-const.c (fold_single_bit_test): Delete unreachable handling
3227 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
3228 all the cases handled by it are inverted by invert_truthvalue.
3230 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
3232 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
3233 (save_register_windowsi): Likewise.
3235 2005-01-04 Richard Henderson <rth@redhat.com>
3237 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
3239 2005-01-04 Uros Bizjak <uros@kss-loka.si>
3242 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
3243 any simplifications of vector mode comparison operators.
3244 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
3246 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
3247 Devang Patel <dpatel@apple.com>
3249 PR tree-optimization/18308
3250 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
3251 the operands before creating a new expression.
3252 * dojump.c (do_jump): Make drop_through_label available
3253 for all cases. Add expansion of COND_EXPR.
3255 2005-01-04 Ira Rosen <irar@il.ibm.com>
3257 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
3258 INTEGER_CST instead of TREE_CONSTANT.
3259 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
3260 instead of TREE_CONSTANT.
3261 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
3262 instead of TREE_CONSTANT.
3264 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
3266 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
3268 2005-01-04 Uros Bizjak <uros@kss-loka.si>
3271 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
3272 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
3274 2005-01-03 Richard Henderson <rth@redhat.com>
3276 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
3277 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
3278 fold_convert_const_real_from_real): Split out from ...
3279 (fold_convert_const): ... here.
3281 2005-01-03 Richard Henderson <rth@redhat.com>
3284 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
3285 (mov<MMXMODEI>_internal): Likewise.
3286 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
3287 Add V2SF case; use it for SSE1; don't use TI.
3288 (movdf_integer): Likewise.
3289 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
3291 2005-01-03 Ira Rosen <irar@il.ibm.com>
3293 * tree-vectorizer.c (vect_strip_conversions): New function.
3294 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
3295 check for binary class.
3297 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
3301 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
3303 * gimple-low.c (mark_blocks_with_used_vars): New function.
3304 (mark_blocks_with_used_subblocks): Ditto.
3305 (mark_used_blocks): Ditto.
3306 (pass_mark_used_blocks): New pass.
3307 * tree-inline.c: Include debug.h.
3308 (expand_call_inline): Call outlining_inline_function here.
3309 * tree-optimize.c (init_tree_optimization_passes): Add
3310 pass_mark_used_blocks.
3311 * tree-pass.h (pass_mark_used_blocks): New.
3312 * Makefile.in (tree-inline.o): Add debug.h dependency.
3314 2005-01-03 Geoffrey Keating <geoffk@apple.com>
3316 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
3317 VAR_DECLs to have weak_import attribute.
3319 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
3322 2004-01-03 Steven Bosscher <stevenb@suse.de>
3324 * passes.c (rest_of_compilation): Don't run regmove if only
3325 flag_expensive_optimizations, require flag_regmove instead.
3327 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
3329 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
3331 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
3332 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
3333 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
3334 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
3336 2005-01-03 Richard Henderson <rth@redhat.com>
3338 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
3339 (ia64_const_ok_for_letter_p): New.
3340 (ia64_const_double_ok_for_letter_p): New.
3341 (ia64_extra_constraint): New.
3342 (ia64_expand_vecint_compare): New.
3343 (ia64_expand_vcondu_v2si): New.
3344 (ia64_expand_vecint_cmov): New.
3345 (ia64_expand_vecint_minmax): New.
3346 (ia64_print_operand): Add 'v'.
3347 (ia64_preferred_reload_class): New.
3348 (ia64_vector_mode_supported_p): New.
3349 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
3350 (PREFERRED_RELOAD_CLASS): Move to function.
3351 (CONST_OK_FOR_LETTER_P): Move to function.
3352 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
3353 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
3354 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
3355 (EXTRA_CONSTRAINT): Move to function.
3356 * config/ia64/ia64.md: Include vect.md.
3357 (itanium_class): Add mmalua.
3359 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
3361 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
3362 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
3364 (const_int_2bit_operand): New.
3365 (fr_reg_or_0_operand): New.
3366 * config/ia64/ia64-modes.def: Add vector modes.
3367 * config/ia64/ia64-protos.h: Update.
3368 * config/ia64/vect.md: New file.
3370 2005-01-03 Richard Henderson <rth@redhat.com>
3372 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
3374 2005-01-03 Uros Bizjak <uros@kss-loka.si>
3377 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
3379 (log1pdf2): Change mode of operands[0,1] to DFmode.
3381 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
3383 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
3384 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
3385 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
3386 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
3387 SPARC_RELAXED_ORDERING.
3389 2005-01-03 Richard Henderson <rth@redhat.com>
3390 Uros Bizjak <uros@kss-loka.si>
3393 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
3394 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
3395 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
3397 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
3398 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
3399 const_pow2_1_to_128_operand): New predicates.
3401 2005-01-02 Greg McGary <greg@mcgary.org>
3403 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
3405 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
3406 object. [ptr_type]: Remove unused variable.
3407 Remember nearest addressable array-element or record-component
3408 when checking bitfield components. Tolerate empty BIND_EXPR.
3410 2005-01-01 Richard Henderson <rth@redhat.com>
3412 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
3413 that don't narrow the value. Fail for other conversions.
3415 2005-01-01 Richard Henderson <rth@redhat.com>
3418 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
3419 * c-lang.c (finish_file): Don't do it here.
3420 * objc/objc-act.c (objc_finish_file): Likewise.
3422 * cgraph.c (decl_assembler_name_equal): New.
3423 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
3424 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
3425 * cgraph.h (struct cgraph_varpool_node): Add next.
3426 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
3427 * varasm.c (assemble_alias): Mark the target as needed.
3429 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
3432 * function.c (get_arg_pointer_save_area): Use entry_of_function
3433 instead of get_insns.
3435 2005-01-01 Roger Sayle <roger@eyesopen.com>
3436 Andrew Pinski <pinskia@physics.uc.edu>
3437 James E. Wilson <wilson@specifixinc.com>
3439 PR rtl-optimization/12092
3440 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
3443 2005-01-01 Roger Sayle <roger@eyesopen.com>
3444 Olivier Hainque <hainque@act-europe.fr>
3446 * tree.c (int_fits_type_p): A narrower type always fits in a
3447 wider one, except for negative values into unsigned types.
3449 2005-01-01 Roger Sayle <roger@eyesopen.com>
3451 * tree.c (int_fits_type_p): Compare the result of force_fit_type
3452 with the original constant rather than require TREE_OVERFLOW.
3454 2005-01-01 Steven Bosscher <stevenb@suse.de>
3457 * c-decl.c (finish_function): If compiling C99, annotate the
3458 compiler generated return with the current file name and line 0.
3459 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
3460 the source line is greater than 0.
3461 (remove_bb): Likewise.
3463 See ChangeLog.12 for earlier changes.