1 2005-02-07 Alexandre Oliva <aoliva@redhat.com>
3 * config/frv/frv.md ("tls_indirect_call"): Turn into libcall-like
5 ("*tls_indirect_call"): New, reference hard regs directly.
6 * config/frv/frv.c (gen_inlined_tls_plt): Adjust.
8 2005-02-07 Ralf Corsepius <ralf.corsepius@rtems.org>
10 * config/arm/rtems-elf.h (SUBTARGET_EXTRA_ASM_SPEC):
11 Let !-mhard-float !-msoft-float imply -mfpu=softfpa.
13 2005-02-06 Roger Sayle <roger@eyesopen.com>
15 PR rtl-optimization/19800
16 * simplify_rtx.c (simplify_relational_operaration_1): Explicitly
17 call gen_lowpart_common and gen_lowpart_SUBREG instead of calling
20 2005-02-07 Ralf Corsepius <ralf.corsepius@rtems.org>
22 * config/m68k/t-rtems (MULTILIB_MATCHES): Let m528x match m5200.
24 2005-02-07 Joseph S. Myers <joseph@codesourcery.com>
27 * doc/invoke.texi: Correct documentation of -freorder-functions.
29 2005-02-06 Joseph S. Myers <joseph@codesourcery.com>
32 * doc/invoke.texi: Don't mention gccbug.
34 2005-02-06 Roger Sayle <roger@eyesopen.com>
35 Andrew Pinski <pinskia@physics.uc.edu>
36 Paolo Bonzini <paolo.bonzini@lu.unisi.ch>
38 * simplify-rtx.c (simplify_relational_operation_1): Simplify
39 (ne:SI (zero_extract:SI FOO (const_int 1) BAR) (const_int 0))
40 into just (zero_extract:SI FOO (const_int 1) BAR).
42 2005-02-06 Joseph S. Myers <joseph@codesourcery.com>
44 * doc/cpp.texi: Include gcc-common.texi. Don't define gcctabopt
45 macro locally. Don't give last revision date. Use GCC version
46 number from gcc-common.texi.
47 * doc/cppinternals.texi: Include gcc-common.texi. Don't give last
48 revision date. Use GCC version number from gcc-common.texi.
49 Describe being a library as current state rather than as 3.x
51 * doc/gcc.texi: Remove last update date.
52 * doc/gccint.texi: Likewise. Update copyright dates.
53 * doc/install.texi: Update copyright dates.
55 2005-02-06 Joseph S. Myers <joseph@codesourcery.com>
58 * doc/invoke.texi: Mention for each option included in -Wall that
59 it is included in -Wall.
60 * doc/gcc.texi: Update copyright and last modification date.
62 2005-02-06 Zdenek Dvorak <dvorakz@suse.cz>
64 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Do not add
65 unnecessary cast to original induction variable increments.
67 2005-02-06 Zdenek Dvorak <dvorakz@suse.cz>
69 PR tree-optimization/18219
70 * tree-ssa-loop-ivopts.c (get_computation_at): Produce computations
73 2005-02-06 Richard Sandiford <rsandifo@redhat.com>
75 * expmed.c (store_bit_field): Make the SUBREG code adjust bitnum.
76 Set bitpos and offset later in the function. Do nothing if the
77 target is a register and if the bitfield lies completely outside
79 (extract_bit_field): Make the same SUBREG, bitpos and offset changes
80 here. Return an uninitialised register if the source value is stored
81 in a register and the bitfield lies completely outside that register.
83 2005-02-06 Steven Bosscher <stevenb@suse.de>
85 * df.c (df_insn_refs_record): Use XEXP to get the operand of a USE,
88 2005-02-06 Marcin Dalecki <martin@dalecki.de>
90 * diagnostic.c (default_diagnostic_finalizer): Fix use of attribute.
92 2005-02-06 Geoffrey Keating <geoffk@apple.com>
94 * config.gcc (*-*-darwin*): Separate darwin7 and darwin8 configs.
95 * config/darwin8.h: New.
96 * config/darwin7.h: Update comment.
98 2005-02-03 Leehod Baruch <leehod@il.ibm.com>
99 Dorit Naishlos <dorit@il.ibm.com>
101 * tree-vectorizer.h (LOC): New type.
102 (UNKNOWN_LOC, EXPR_LOC, LOC_FILE, LOC_LINE): New macros.
103 (loop_line_number): New field in struct _loop_vec_info.
104 (LOOP_VINFO_LOC, LOOP_LOC): New macros.
106 * tree-vectorizer.c (input.h): Included.
107 (find_loop_location): New function.
108 (vect_debug_stats): Argument changed from loop to LOC. Computation of
109 loop line number removed.
110 (vect_debug_details): Likewise.
111 (new_loop_vec_info): Initialize new field LOOP_VINFO_LOC.
113 (slpeel_make_loop_iterate_ntimes): Call find_loop_location. Argument in
114 call to vect_debug_details/stats changed from loop to loop_loc.
115 (slpeel_tree_duplicate_loop_to_edge_cfg): Likewise.
116 (slpeel_tree_peel_loop_to_edge): Likewise.
118 (vect_analyze_offset_expr): Argument in call to vect_debug_details/stats
119 changed from NULL to UNKNOWN_LOC.
120 (vect_get_base_and_offset):
121 (vect_create_addr_base_for_vector_ref):
122 (get_vectype_for_scalar_type):
123 (vect_create_data_ref_ptr):
125 (vect_get_vec_def_for_operand):
126 (vect_finish_stmt_generation):
127 (vectorizable_assignment):
128 (vectorizable_operation):
129 (vectorizable_store):
131 (vect_transform_stmt):
132 (vect_update_ivs_after_vectorizer):
133 (vect_do_peeling_for_loop_bound):
134 (vect_do_peeling_for_alignment):
135 (vect_transform_loop):
136 (vect_is_simple_use):
137 (vect_analyze_operations):
138 (vect_is_simple_iv_evolution):
139 (vect_analyze_scalar_cycles):
140 (vect_analyze_data_ref_dependences):
141 (vect_compute_data_ref_alignment):
142 (vect_enhance_data_refs_alignment):
143 (vect_analyze_data_ref_access):
144 (vect_analyze_data_ref_accesses):
145 (vect_analyze_pointer_ref_access):
146 (vect_get_memtag_and_dr):
147 (vect_analyze_data_refs):
148 (vect_mark_relevant):
149 (vect_stmt_relevant_p):
150 (vect_mark_stmts_to_be_vectorized):
151 (vect_can_advance_ivs_p):
152 (vect_get_loop_niters):
154 (vectorize_loops): Likewise.
156 (vectorizable_load): Argument in call to vect_debug_details/stats
157 changed from loop to LOOP_LOC (loop_vinfo).
158 (vect_transform_loop):
159 (vect_analyze_operations):
160 (vect_analyze_scalar_cycles):
161 (vect_analyze_data_ref_dependence):
162 (vect_enhance_data_refs_alignment):
163 (vect_analyze_data_ref_accesses):
164 (vect_analyze_pointer_ref_access):
165 (vect_analyze_data_refs):
166 (vect_analyze_loop): Likewise.
168 (vect_analyze_loop_form): Argument in call to vect_debug_details/stats
169 changed from loop to loop_loc.
171 (vect_enhance_data_refs_alignment): Removed unused variable loop.
173 2005-02-03 Leehod Baruch <leehod@il.ibm.com>
174 Dorit Naishlos <dorit@il.ibm.com>
176 * tree-vectorizer.c (vect_analyze_data_ref_dependence): Function
178 (vect_analyze_data_ref_dependences): Likewise.
180 (vect_is_simple_use): Argument changed from loop structure to
181 loop_vect_info structure.
182 (vect_can_advance_ivs_p): Likewise.
183 (vect_create_index_for_vector_ref): Likewise.
184 (vect_update_ivs_after_vectorizer): Likewise.
185 (new_stmt_vec_info): Likewise.
187 (new_loop_vec_info): Second argument in call to new_stmt_vec_info was
188 changed from loop to loop_vinfo.
189 (vect_create_data_ref_ptr): First argument in call to
190 vect_create_index_for_vector_ref was changed from loop to loop_vinfo.
191 (vectorizable_assignment): Second argument in call to vect_is_simple_use
192 was changed from loop to loop_vinfo.
193 (vectorizable_operation): Likewise.
194 (vectorizable_store): Likewise.
195 (vect_mark_stmts_to_be_vectorized): Likewise.
196 (vect_do_peeling_for_loop_bound): First argument in call to
197 vect_update_ivs_after_vectorizer was changed from loop to loop_vinfo.
198 (vect_analyze_operations): Argument in call to vect_can_advance_ivs_p
199 was changed from loop to loop_vinfo.
200 (vect_analyze_data_ref_dependences): Third argument in call to
201 vect_analyze_data_ref_dependence was changed from loop to loop_vinfo.
203 (vect_create_index_for_vector_ref): Get the loop from loop_vinfo.
204 (vect_create_data_ref_ptr): Likewise.
205 (vect_init_vector): Likewise.
206 (vect_get_vec_def_for_operand): Likewise.
207 (vectorizable_assignment): Likewise.
208 (vectorizable_operation): Likewise.
209 (vectorizable_store): Likewise.
210 (vectorizable_load): Likewise.
211 (vect_update_ivs_after_vectorizer): Likewise.
212 (vect_is_simple_use): Likewise.
213 (vect_analyze_data_ref_dependence): Likewise.
214 (vect_analyze_pointer_ref_access): Likewise.
215 (vect_can_advance_ivs_p): Likewise.
217 * tree-vectorizer.h: (_loop_vec_info): Definition and macros moved
218 before _stmt_vec_info.
219 (_stmt_vec_info): Loop field replaced by loop_vec_info.
220 (STMT_VINFO_LOOP): Replaced with STMT_VINFO_LOOP_VINFO.
221 (new_stmt_vec_info): Argument changed from loop structure to
222 loop_vect_info structure
224 (vect_analyze_data_ref_dependences): Unnecessary line was removed.
225 (vect_analyze_offset_expr): Avoid 80 columns overflow.
226 (vect_create_addr_base_for_vector_ref): Likewise.
227 (vect_analyze_pointer_ref_access): Likewise.
229 2005-02-03 Andrew Pinski <pinskia@physics.uc.edu>
232 * tree-ssa-dse.c (fix_phi_uses): Update the occurs in abnormal
233 phi flag if the phi is abnormal.
235 2005-02-03 Andrew Pinski <pinskia@physics.uc.edu>
238 * tree-ssa.c (replace_immediate_uses): Update the immediate_uses
239 information for the new statement.
241 2005-02-03 Joseph S. Myers <joseph@codesourcery.com>
244 * gcc.c (main): Include the target in -v output.
246 2005-02-03 Alexandre Oliva <aoliva@redhat.com>
248 * combine.c (simplify_comparison, case SIGN_EXTEND, ZERO_EXTEND):
249 Do not drop the extend if we'd have to add a paradoxical subreg
250 later. Include optabs.h and insn-codes.h.
251 * Makefile.in (combine.o): Depend on $(OPTABS_H).
253 2005-02-02 Roger Sayle <roger@eyesopen.com>
256 * real.h (REAL_MODE_FORMAT_COMPOSITE_P): New macro.
257 * fold-const.c (const_binop): Avoid constant folding floating
258 point operations in modes that use composite representations.
259 * simplify-rtx.c (simplify_binary_operation): Likewise.
261 2005-02-02 Geoffrey Keating <geoffk@apple.com>
263 * config/rs6000/altivec.md (altivec_dst): Make the first operand
265 (altivec_dstt): Likewise.
266 (altivec_dstst): Likewise.
267 (altivec_dststt): Likewise.
268 * config/rs6000/rs6000.c (altivec_expand_dst_builtin): Adjust creation
271 2005-02-03 Dorit Naishlos <dorit@il.ibm.com>
273 * tree-vectorizer.c (slpeel_make_loop_iterate_ntimes): Copy
274 EXPR_LOCUS from orig_cond to the new cond_stmt.
275 (vect_finish_stmt_generation): Copy EXPR_LOCUS from stmt to
278 2005-02-03 Dorit Naishlos <dorit@il.ibm.com>
280 * tree-vectorizer.c (slpeel_make_loop_iterate_ntimes): Call
281 standard_iv_increment_position. Remove call to bsi_next
282 (no need to bump the iterator anymore).
283 (vect_create_index_for_vector_ref): Call
284 standard_iv_increment_position. Remove second function argument.
285 (vect_finish_stmt_generation): Remove call to bsi_next
286 (no need to bump the iterator anymore).
287 (vect_create_data_ref_ptr): Remove second argument (bsi) in call
288 to vect_create_index_for_vector_ref.
290 2005-02-03 Dorit Naishlos <dorit@il.ibm.com>
292 * tree-vectorizer.c (vect_mark_stmts_to_be_vectorized): Check if
293 phis are used out of the loop.
295 2005-02-03 Joseph S. Myers <joseph@codesourcery.com>
298 * c-decl.c (undef_nested_function): New variable.
299 (pop_scope): Diagnose undefined nested functions.
300 (finish_function): Don't attempt cgraph processing or genericizing
301 if current top-level function contained an undefined nested
302 function. Reset undef_nested_function at the end of a top-level
305 2005-02-02 Zdenek Dvorak <dvorakz@suse.cz>
307 * tree.c (build_int_cst_type): Take sign of the value into account
308 when deciding whether sign extend the value.
310 2005-02-02 Joseph S. Myers <joseph@codesourcery.com>
313 * gcc.c (cpp_unique_options): Remove %{trigraphs}.
314 (cpp_options, cc1_options): Change %{std*&ansi} to
315 %{std*&ansi&trigraphs}.
317 2005-02-02 Kazu Hirata <kazu@cs.umass.edu>
319 * cse.c: Update comments.
321 2005-02-02 Aldy Hernandez <aldyh@redhat.com>
323 * config/rs6000/rs6000.c (rs6000_spe_function_arg): Multi-reg
324 arguments go on the stack. Do not put complex doubles in odd/even
327 2005-02-02 Steven Bosscher <stevenb@suse.de>
329 * doc/invoke.texi: Update the options enabled at -O1, -O2,
330 and -O3. Update the documentation for -fgcse-sm, -fgcse-las,
331 -ftree-pre, and -fthread-jumps.
333 2005-02-02 Richard Sandiford <rsandifo@redhat.com>
335 PR tree-optimization/19578
336 * tree-flow.h (modified_noreturn_calls): Declare.
337 (noreturn_call_p): Declare.
338 * tree-flow-inline.h (noreturn_call_p): New function.
339 (modify_stmt): Add modified noreturn calls to modified_noreturn_calls.
340 * tree-cfg.c (modified_noreturn_calls): New variable.
341 (cleanup_control_flow): Use noreturn_call_p. Split basic blocks
342 that contain a mid-block noreturn call.
343 * tree-ssa.c (delete_tree_ssa): Clear modified_noreturn_calls.
345 2005-02-02 Kazu Hirata <kazu@cs.umass.edu>
347 * df.c (df_def_record_1, df_uses_record): Don't use
349 * df.h (DF_FOR_REGALLOC): Remove.
351 2005-02-02 Joseph S. Myers <joseph@codesourcery.com>
354 * c-typeck.c (really_start_incremental_init): Reset
355 constructor_max_index for arrays of incomplete type.
357 2005-02-02 Jeff Law <law@redhat.com>
359 * gcse.c (struct reg_set): Store the block index where the register
360 was set rather than the setting insn.
361 (record_one_set, compute_transp): Corresponding changes.
362 (pre_insert_copy_insn): Do not call replace_one_set.
363 (replace_one_set): Remove.
365 2005-02-02 Joseph S. Myers <joseph@codesourcery.com>
368 * doc/gcov.texi: Fix typo.
370 2005-02-01 Richard Henderson <rth@redhat.com
373 * config/i386/i386.h (MODES_TIEABLE_P): Use ix86_modes_tieable_p.
374 * config/i386/i386.c (ix86_hard_regno_mode_ok): Change return
376 (ix86_tieable_integer_mode_p, ix86_modes_tieable_p): New.
377 * config/i386/i386-protos.h: Update.
379 2005-02-01 Steven Bosscher <stevenb@suse.de>
381 PR tree-optimization/19217
382 * tree-cfg.c (verify_expr): Use the data field to see if TP was
383 seen inside a PHI node. Do not do the ADDR_EXPR check if it was.
384 (verify_stmts): Pass (void*)1 as data to verify_expr to signal
385 that it is walking a PHI node.
387 2005-02-01 Joseph S. Myers <joseph@codesourcery.com>
389 * doc/extend.texi (Nested Functions): Update.
391 2005-02-01 Richard Henderson <rth@redhat.com>
394 * optabs.c (expand_copysign_absneg): Export.
395 * optabs.h (expand_copysign_absneg): Declare.
396 * config/rs6000/rs6000.md (copysigntf3): New.
398 2005-02-01 Diego Novillo <dnovillo@redhat.com>
400 PR tree-optimization/19633
401 * tree-flow.h (struct ptr_info_def): Add field 'pt_null'.
402 * tree-ssa-alias.c (init_alias_info): Initialize.
403 (merge_pointed_to_info): Set.
404 (add_pointed_to_expr): Set pt_null if EXPR is a NULL pointer.
405 (dump_points_to_info_for): Show value of pt_null.
406 (struct count_ptr_d): Declare.
407 (find_ptr_dereference): Remove.
408 (ptr_is_dereferenced_by): Remove.
409 (count_ptr_derefs): New local function.
410 (count_uses_and_derefs): New local function.
411 (compute_points_to_and_addr_escape): Call it. If the number
412 of dereferences is greater than zero, mark the pointer as
413 dereferenced. If there are fewer dereferences than uses of
414 the pointer, the pointer's value escapes.
416 2005-02-01 Diego Novillo <dnovillo@redhat.com>
418 PR tree-optimization/19670
419 * tree-ssa.c (verify_ssa_name): Don't set TREE_VISITED
421 (verify_use): ... set it here, instead.
423 2005-02-01 Janis Johnson <janis187@us.ibm.com>
425 * doc/sourcebuild.texi (Test directives): Fix formatting.
427 * doc/sourcebuild.texi (Test directives): Describe selector
430 2005-02-01 Segher Boessenkool <segher@kernel.crashing.org>
432 * config.gcc (powerpc64-*-linux*): Default to -m64 also for
433 specific 64-bit CPUs.
435 2005-02-01 Richard Earnshaw <rearnsha@arm.com>
438 * arm.c (arm_eliminable_register): New function.
439 (adjacent_mem_locations): Don't allow eliminable registers. Use
440 HOST_WIDE_INT for address offsets.
441 * arm-protos.h (arm_eliminable_register): Add prototype.
443 2005-02-01 Steven Bosscher <stevenb@suse.de>
445 PR optimization/15242
446 * params.def (PARAM_MAX_GOTO_DUPLICATION_INSNS): New param.
447 * basic-block.h (duplicate_computed_gotos): Add prototype.
448 * bb-reorder.c (duplicate_computed_gotos): New function to
449 duplicate sufficiently small blocks ending in a computed jump.
450 * passes.c (rest_of_compilation): Call duplicate_computed_gotos
451 if not optimizing for size.
452 * cfgcleanup.c (try_crossjump_bb): If not optimizing for size,
453 never do tail merging for blocks ending in a computed jump.
454 * doc/invoke.texi: Document the max-goto-duplication-insns param.
456 2005-02-01 Eric Botcazou <ebotcazou@libertysurf.fr>
458 Patch from Richard Sandiford <rsandifo@redhat.com>
459 * reload1.c (choose_reload_regs): Prevent the find_equiv_reg() code
460 from inheriting a subreg equivalence with a non-spill register.
462 * tree.h (DECL_FUNCTION_CODE): Document that it is overloaded.
464 2005-02-01 Alexandre Oliva <aoliva@redhat.com>
466 * config/frv/frv.c (movcc_fp_destination_operand): New.
467 (gpr_or_memory_operand): Fix typo in comment.
468 (gpr_or_memory_operand_with_scratch): New.
469 * config/frv/frv.h (PREDICATE_CODES): Add the two new predicates.
470 * config/frv/frv.md (movcc_fp, movcc_fp_internal): Use
471 movcc_fp_destination_operand.
472 (reload_incc_fp): Use gpr_or_memory_operand_with_scratch.
473 Legitimize memory addresses using a scratch register.
475 2005-01-31 Jeff Law <law@redhat.com>
477 * tree-into-ssa.c (mark_def_sites_global_data): Make KILLS
478 bitmap a sparse bitmap instead of a simple bitmap.
479 (mark_def_sites_initialize_block): Corresponding changes.
480 (ssa_mark_def_sites_initialize_block): Likewise.
481 (ssa_mark_phi_uses): Likewise.
482 (mark_def_site, ssa_mark_def_sites): Likewise.
483 (mark_def_site_blocks): Likewise.
484 (rewrite_ssa_into_ssa): Likewise.
486 * tree-ssa-dom.c (record_cond): Pass correct variable type
487 for last argument to htab_find_slot_with_hash.
489 * fold-const.c (fold, case CEIL_MOD_EXPR): Do not lose side
490 effects when optimizing 0 % X. Do not try to optimize X % 0.
492 2005-01-31 James E. Wilson <wilson@specifixinc.com>
494 * config/ia64/itanium1.md (1_scall bypass): Change 2_mmalua to
497 2005-02-01 Eric Christopher <echristo@redhat.com>
499 * config/mips/mips.c (override_options): Warn if -mint64
501 * doc/invoke.texi (MIPS Options): Document that -mint64 is
504 2005-02-01 Kazu Hirata <kazu@cs.umass.edu>
506 * cse.c (cse_reg_info): Remove hash_next, next, regno. Add
508 (cse_reg_info_list, cse_reg_info_list_free, REGHASH_SHIFT,
509 REGHASH_SIZE, REGHASH_MASK, reg_hash, REGHASH_FN,
510 cached_cse_reg_info, GET_CSE_REG_INFO): Remove.
511 (cached_regno): Initialize to INVALID_REGNUM.
512 (cse_reg_info_table_size,
513 cse_reg_info_table_first_uninitialized,
514 cse_reg_info_timestamp): New.
515 (REG_TICK, REG_IN_TABLE, SUBREG_TICKED, REG_QTY): Use
517 (init_cse_reg_info, get_cse_reg_info_1): New.
518 (get_cse_reg_info): Cache the last look-up.
519 (new_basic_block): Update the code to clear mappings from
520 registers to cse_reg_info entries.
521 (cse_main): Call init_cse_reg_info.
523 * cse.c (get_cse_reg_info): Update a comment.
525 2005-01-31 Steven Bosscher <stevenb@suse.de>
528 * c-decl.c (start_decl): Do not warn about arrays of elements with
529 an incomplete type here.
530 (grokdeclarator): Do it here by making a pedwarn an error.
531 * c-typeck.c (push_init_level): If there were previous errors with
532 the constructor type, do not warn about braces for initializers.
533 (process_init_element): Likewise for excess initializer elements.
535 2005-01-31 Kazu Hirata <kazu@cs.umass.edu>
537 * cse.c (delete_trivially_dead_insn): Don't iterate.
539 2005-01-31 Andrew Pinski <pinskia@physics.uc.edu>
541 * config/rs6000/rs6000.md (copysignsf3): New expand.
542 (copysigndf3): Likewise.
544 2005-01-31 Steven Bosscher <stevenb@suse.de>
546 * recog.c (constrain_operands): Only look for earlyclobber operand
547 conflicts if an '&' constraint was seen.
549 2005-01-31 Marc Espie <espie@openbsd.org>
551 * config.gcc: Don't include embedded systems fragment, switches default
552 debugging format to ELF.
553 * config/i386/openbsdelf.h: Add DBX_REGISTER_NUMBER since we no
554 longer pick it up there.
556 2005-01-31 Dale Johannesen <dalej@apple.com>
558 * doc/extend.texi (nested functions): Fix linkage description.
559 Clarify that static is not allowed.
561 2005-01-31 Dale Johannesen <dalej@apple.com>
563 * config/rs6000/darwin.md (movsf_low_di): Make work.
564 (movdf_low_di): Make work.
566 2005-01-31 Dale Johannesen <dalej@apple.com>
568 * config/rs6000/darwin-tramp.asm (__trampoline_setup):
569 Make work for 64 bit.
571 2005-01-31 Roger Sayle <roger@eyesopen.com>
572 Dale Johannesen <dalej@apple.com>
575 * fold-const.c (fold_binary_op_with_conditional_arg):
576 Make types match original operands, before STRIP_NOPS.
578 2005-01-31 Alan Modra <amodra@bigpond.net.au>
580 * config/rs6000/linux-unwind.h (struct gcc_vregs): New.
581 (struct gcc_regs): Rename from gcc_pt_regs. Add more regs.
582 (struct gcc_sigcontext): Delete. Merge contents to..
583 (struct gcc_ucontext): ..here.
584 (get_sigcontext): Delete.
585 (get_regs): New function, like get_sigcontext but return regs ptr.
586 64-bit version finds regs from r1 to support vdso.
587 (ppc_linux_aux_vector): New function.
588 (ppc_fallback_frame_state): Modify for get_regs. Restore fprs
591 2005-01-31 Ira Rosen <irar@il.ibm.com>
593 * tree-vectorizer.c (vect_analyze_offset_expr): Use ssizetype instead
595 (vect_get_base_and_offset): Use ssizetype instead sizetype. Remove
596 redundant fold. Fix misalignment for MINUS_EXPR.
597 (vect_compute_data_ref_alignment): Use ssizetype instead sizetype.
598 (vect_analyze_pointer_ref_access): Likewise.
599 (vect_get_memtag_and_dr): Likewise.
601 2005-01-31 Richard Henderson <rth@redhat.com>
603 * config/i386/i386.c (absnegsf2_mixed): Remove all of the # hackery.
604 (absnegsf2_sse, absnegsf2_i387, absnegdf2_mixed, absnegdf2_sse,
605 absnegdf2_i387, absnegxf2_i387): Likewise.
607 2005-01-31 Richard Henderson <rth@redhat.com>
610 * optabs.c (expand_copysign_absneg): New.
611 (expand_copysign_bit): Split out from ...
612 (expand_copysign): ... here. Use expand_copysign_absneg.
614 2005-01-30 Roger Sayle <roger@eyesopen.com>
617 * config/pa/pa.md (anddi3, iordi3): On HPPA64, disallow an integer
618 constant as the second operand and a register as the third.
620 2005-01-31 Danny Smith <dannysmith@users.sourceforge.net>
623 * config/i386/i386.c (ix86_function_ok_for_sibcall): Also check
624 that dllimport'd functions do not use all call-clobbered registers
627 2005-01-30 Richard Henderson <rth@redhat.com>
630 * optabs.c (expand_absneg_bit): Split out from expand_unop and
631 expand_abs_nojump and generalize. Use operand_subword and
632 emit_no_conflict_block. Support large modes.
633 (expand_abs_nojump): Use it. Use HONOR_SIGNED_ZEROS to fall
635 (expand_unop): Likewise. Use HONOR_SIGNED_ZEROS to fall back
636 negation to subtraction.
637 (expand_copysign): Check that the format has signed zeros.
638 Use operand_subword and emit_no_conflict_block. Support large modes.
640 2005-01-30 Daniel Berlin <dberlin@dberlin.org>
642 Fix PR tree-optimization/19624
644 * Makefile.in (tree-ssa-pre.o): Add CFGLOOP_H.
645 * tree-ssa-pre.c: Add cfgloop.h.
647 (pre_stats): New member, constified.
648 (inserted_exprs): New static variable.
649 (NECESSARY): New macro.
650 (create_expression_by_pieces): Fold the expression, and
651 mark it as defaulting to not necessary. Also put in
653 (fully_constant_expression): New function.
654 (insert_into_preds_of_block): Modify to not insert phis when we
655 are playing with induction variables.
656 Push phis onto the inserted_exprs vector, and mark them as not
657 necessary by default.
658 (insert_aux): Call fully_constant_expression on eprime.
659 If all edges produce the same value, mark it constant.
660 (mark_operand_necessary): New function.
661 (remove_dead_inserted_code): New function.
662 (init_pre): Init loop optimizer to get loop info.
663 (fini_pre): Free loop_optimizer, and inserted_exprs vec.
664 (execute_pre): Commit edge inserts, then remove dead code.
666 2005-01-30 Richard Henderson <rth@redhat.com>
668 * rtl.c (rtx_equal_p): No early exit for CONST_VECTOR.
669 * varasm.c (const_rtx_hash_1): Handle CONST_VECTOR.
671 2005-01-30 Richard Henderson <rth@redhat.com>
674 * config/i386/i386.c (ix86_expand_copysign): New.
675 (ix86_split_copysign_const): New.
676 (ix86_split_copysign_var): Rename from ix86_split_copysign,
677 rearrange op1/nmask operands.
678 * config/i386/i386-protos.h: Update.
679 * config/i386/i386.md (copysignsf3): Use ix86_expand_copysign.
680 (copysigndf3): Likewise.
681 (copysignsf3_const, copysigndf3_const): New.
682 (copysignsf3_var): Rename from copysignsf3, split out splitter
683 and fix split predicate for X constraint.
684 (copysigndf3_var): Similarly.
686 2005-01-30 Kazu Hirata <kazu@cs.umass.edu>
688 * optabs.c, doc/c-tree.texi, doc/install.texi, doc/md.texi,
689 doc/passes.texi, doc/rtl.texi, doc/sourcebuild.texi,
690 doc/tm.texi, doc/tree-ssa.texi: Update copyright.
692 2005-01-29 Richard Henderson <rth@redhat.com>
695 * config/i386/i386.md (movdf_nointeger, movdf_integer): Fix ordering
696 of # and * in constraints.
698 2005-01-29 Richard Henderson <rth@redhat.com>
701 * expr.c (store_field): Don't strip sub-mode cast when the input
702 data is even smaller.
704 2005-01-29 Richard Henderson <rth@redhat.com>
707 * expr.c (categorize_ctor_elements_1): Check for CONSTRUCTOR of a
710 2005-01-29 Richard Henderson <rth@redhat.com>
712 * combine.c (make_field_assignment): Fix argument order
715 2005-01-29 Richard Guenther <richard.guenther@uni-tuebingen.de>
717 PR tree-optimization/15791
718 * fold-const.c (extract_array_ref): New function.
719 (fold): Fold comparisons between &a[i] and &a[j] or
720 semantically equivalent trees.
722 2005-01-29 Jeff Law <law@redhat.com>
724 * gcse.c (insert_expr_in_table): Revamp handling of available
725 and anticipatable occurrence lists to avoid unnecessary list
727 (insert_set_in_table): Similarly.
729 2005-01-29 Joseph S. Myers <joseph@codesourcery.com>
731 * c-common.c (fix_string_type): Just use c_build_qualified_type to
733 (c_build_qualified_type): Build qualified array types with
734 TYPE_MAIN_VARIANT pointing to corresponding unqualified type.
735 * c-decl.c (c_make_fname_decl): Build unqualified array type
736 before qualified type.
737 (grokdeclarator): Use TYPE_MAIN_VARIANT of typedef type if element
738 type is qualified, not just if type itself is. Don't apply
739 qualifiers to array type when declarator is processed. Apply
740 qualifiers to field type whether or not it is an array type.
741 Don't handle array types specially for applying qualifiers to
743 * c-typeck.c (composite_type): Build unqualified element type and
744 array type when forming composite of array types.
745 (common_pointer_type, comptypes, comp_target_types,
746 type_lists_compatible_p, build_indirect_ref, build_array_ref,
747 convert_for_assignment): Don't apply TYPE_MAIN_VARIANT to array
749 (type_lists_compatible_p): Cache TREE_VALUE (args1) and TREE_VALUE
750 (args2) in variables a1 and a2.
752 2005-01-29 Kazu Hirata <kazu@cs.umass.edu>
754 * cse.c (n_elements_made, max_elements_made): Remove.
755 (insert): Don't touch n_elements_made.
756 (cse_main): Don't touch n_elements_made or max_elements_made.
758 2005-01-28 Stephane Carrez <stcarrez@nerim.fr>
761 * config/m68hc11/t-m68hc11-gas (dp-bit.c): Fix typo causing a
762 configuration part of dp-bit.c to be lost.
764 2005-01-28 Roger Sayle <roger@eyesopen.com>
766 * expmed.c (expand_mult_highpart_optab): When attempting to use
767 a non-widening multiplication in a wider mode, the operands need
768 to be converted (zero or sign extended) to that mode.
770 2005-01-28 Ian Lance Taylor <ian@airs.com>
774 * gimple-low.c (block_may_fallthru): TRY_FINALLY_EXPR only falls
775 through if both operands fall through.
777 2005-01-28 Kazu Hirata <kazu@cs.umass.edu>
779 * cse.c (fold_rtx) <PC>: Don't optimize.
781 2005-01-28 Jeff Law <law@redhat.com>
783 * fold-const.c (fold, case CEIL_MOD_EXPR): Optimize 0 % X.
784 (case FLOOR_MOD_EXPR, ROUND_MOD_EXPR, TRUNC_MOD_EXPR): Similarly.
786 2005-01-28 Kazu Hirata <kazu@cs.umass.edu>
788 * cse.c (cse_reg_info_free_list, cse_reg_info_used_list,
789 cse_reg_info_used_list_end): Remove.
790 (cse_reg_info_list, cse_reg_info_list_free): New.
791 (get_cse_reg_info): When allocating an instance of
792 cse_reg_info, add it to the beginning of the cse_reg_info_list
793 list. Remove code to maintain cse_reg_info_used_list.
794 (new_basic_block): Reset the free list to the beginning of
797 2005-01-28 Richard Sandiford <rsandifo@redhat.com>
799 * config/frv/frv.c (bdesc_2arg): Remove __MDPACKH.
800 (frv_init_builtins): Change its prototype to take 4 uhalf arguments.
801 (frv_expand_mdpackh_builtin): New function.
802 (frv_expand_builtin): Use it to expand __MDPACKH.
803 * doc/extend.texi (__MDPACKH): Update documentation.
805 2005-01-28 Uros Bizjak <uros@kss-loka.si>
807 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use
808 SSE_FLOAT_MODE_P for use_sse computation.
810 2005-01-28 Ralf Corsepius <ralf.corsepius@rtems.org>
813 * config/sparc/rtemself.h (LINK_GCC_C_SEQUENCE_SPEC): undefine.
815 2005-01-27 Richard Henderson <rth@redhat.com>
817 * builtins.c (expand_builtin_copysign): New.
818 (expand_builtin): Call it.
819 * genopinit.c (optabs): Add copysign_optab.
820 * optabs.c (init_optabs): Initialize it.
821 (expand_copysign): New.
822 * optabs.h (OTI_copysign, copysign_optab): New.
823 (expand_copysign): Declare.
825 * config/alpha/alpha.md (UNSPEC_COPYSIGN): New.
826 (copysignsf3, ncopysignsf3, copysigndf3, ncopysigndf3): New.
828 * config/i386/i386.c (ix86_build_signbit_mask): Split from ...
829 (ix86_expand_fp_absneg_operator): ... here.
830 (ix86_split_copysign): New.
831 * config/i386/i386-protos.h: Update.
832 * config/i386/i386.md (UNSPEC_COPYSIGN): New.
833 (copysignsf3, copysigndf3): New.
835 * config/ia64/ia64.md (UNSPEC_COPYSIGN): New.
836 (copysignsf3, ncopysignsf3): New.
837 (copysigndf3, ncopysigndf3): New.
838 (copysignxf3, ncopysignxf3): New.
839 * config/ia64/ia64.c (rtx_needs_barrier): Handle UNSPEC_COPYSIGN.
841 2005-01-27 Arend Bayer <arend.bayer@web.de>
842 Kazu Hirata <kazu@cs.umass.edu>
844 * cse.c: (find_best_addr): Don't call copy_rtx before calling
845 fold_rtx. Save cost recomputation if fold_rtx did nothing.
846 (fold_rtx) <ASM_OPERANDS>: Don't do anything if INSN is
849 2005-01-27 Jeff Law <law@redhat.com>
851 * tree-into-ssa.c (ssa_rewrite_initialize_block): Do not register
852 definitions of SSA_NAMEs which are not being rewritten.
853 (rewrite_ssa_into_ssa): Only initialize the current definition
854 of an SSA_NAME if that SSA_NAME has been marked for rewriting.
855 If checking is enabled, assert that SSA_NAME_AUX is clear for all
856 SSA_NAMEs before returning.
858 2005-01-27 Kazu Hirata <kazu@cs.umass.edu>
860 * c-common.def, c-dump.c, c-gimplify.c, c-objc-common.c,
861 cfgexpand.c, dbxout.c, function.h, opts.c, tree-flow-inline.h,
862 tree-ssa-operands.h, tree-ssa-threadupdate.c, tree-tailcall.c,
863 config/avr/avr.md, config/cris/aout.h, config/cris/cris.h,
864 config/mips/iris6.h, config/sh/sh.c: Update copyright.
866 2005-01-27 Steven Bosscher <stevenb@suse.de>
869 * opts.c (decode_options): Move flag_thread_jumps from -O1 and
870 higher to -O2 and higher. Likewise for tree PRE.
871 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Only iterate at -O2
874 2005-01-27 Ian Lance Taylor <ian@airs.com>
877 * gimple-low.c (try_catch_may_fallthru): New static function.
878 (block_may_fallthru): Handle TRY_CATCH_EXPR.
879 * tree-inline.c (expand_call_inline): Don't warn about reaching
880 the end of a non-void function being inlined if the function uses
883 2005-01-27 Jakub Jelinek <jakub@redhat.com>
885 * config/i386/i386.h (CALL_USED_REGISTERS): Fix comment pastos.
888 * c-decl.c (warn_if_shadowing): Handle old_decl error_mark_node.
889 (pushdecl): Only use DECL_FILE_SCOPE_P if DECL_P.
890 (implicitly_declare): Handle error_mark_node.
892 2005-01-27 Richard Henderson <rth@redhat.com>
895 * tree.h (struct tree_decl): Add debug_expr_is_from.
896 (DECL_DEBUG_EXPR_IS_FROM): New.
897 (DECL_DEBUG_EXPR): Rename from DECL_DEBUG_ALIAS_OF.
898 * dwarf2out.c (dwarf2out_var_location): Update to match.
899 * tree-outof-ssa.c (create_temp): Likewise.
900 * var-tracking.c (track_expr_p): Likewise.
901 * tree-sra.c (instantiate_element): Set DECL_DEBUG_EXPR.
902 * c-objc-common.c (c_tree_printer) <'D'>: Handle DECL_DEBUG_EXPR.
903 * toplev.c (default_tree_printer): Likewise.
905 2005-01-27 Alexandre Oliva <aoliva@redhat.com>
907 * config/frv/frv.c (frv_class_likely_spilled_p): Add GR8_REGS,
908 GR9_REGS, GR89_REGS, FDPIC_FPTR_REGS, FDPIC_REGS.
910 2005-01-27 Steven Bosscher <stevenb@suse.de>
912 * tree.h (SWITCH_COND, SWITCH_BODY, SWITCH_LABELS, CASE_LOW,
913 CASE_HIGH, CASE_LABEL, ASM_STRING, ASM_OUTPUTS, ASM_INPUTS,
914 ASM_CLOBBERS): Add tree checks.
916 * c-common.h (SWITCH_TYPE): Rename to SWITCH_STMT_TYPE.
917 (SWITCH_STMT_COND, SWITCH_STMT_BODY): New.
918 * c-common.def (SWITCH_STMT): Update to match.
919 * c-common.c (c_do_switch_warnings): Use SWITCH_STMT accessor
920 macros instead of SWITCH_EXPR ones.
921 * c-dump.c (c_dump_tree): Likewise.
922 * c-gimplify.c (gimplify_switch_stmt): Likewise.
923 * c-typeck.c (c_start_case, do_case, c_finish_case): Likewise.
925 * doc/c-tree.texi (SWITCH_STMT): Update accessor macro names.
927 2005-01-27 Alan Modra <amodra@bigpond.net.au>
929 * unwind-dw2.c (execute_stack_op): Add missing cases for
930 DW_OP_shl, DW_OP_shr, DW_OP_shra, DW_OP_xor.
932 2005-01-27 Eric Botcazou <ebotcazou@libertysurf.fr>
934 * config.gcc (Obsolete configurations): Remove 'dummy', add
935 sparclite-*-coff*, sparclite-*-elf*, sparc86x-*-elf* and
938 2005-01-26 Diego Novillo <dnovillo@redhat.com>
940 PR tree-optimization/19633
941 * tree-ssa-alias.c (ptr_is_dereferenced_by): Also handle
943 (maybe_create_global_var): Do not create .GLOBAL_VAR if there
944 are no call-clobbered variables.
945 * tree-outof-ssa.c (check_replaceable): Return false for calls
948 2005-01-26 Ulrich Weigand <uweigand@de.ibm.com>
950 * dbxout.c (dbxout_symbol_location): Resolve constant pool references
951 even for variables with NULL DECL_INITIAL.
953 2005-01-26 Stuart Hastings <stuart@apple.com>
955 * gimplify.c (shortcut_cond_expr): Re-compute side-effects.
957 2005-01-26 Richard Henderson <rth@redhat.com>
960 * c-decl.c (finish_struct): Set DECL_MODE after resetting a
962 * expr.c (store_field): Strip conversions to odd-bit-sized types
963 if the destination field width matches.
965 2005-01-26 Richard Henderson <rth@redhat.com>
967 * c-decl.c, expmed.c, expr.c: Revert last change.
969 2005-01-26 Richard Henderson <rth@redhat.com>
972 * c-decl.c (finish_struct): Set DECL_MODE after resetting a
974 * expmed.c (store_fixed_bit_field): Create a paradoxical subreg
975 if we don't need the bits above those present in the current mode.
976 * expr.c (store_field): Strip conversions to odd-bit-sized types
977 if the destination field width matches.
979 2005-01-26 Richard Sandiford <rsandifo@redhat.com>
981 * config/mips/iris6.h (ENDFILE_SPEC): Don't link in irix-csr.o if
982 either -ffast-math or -funsafe-math-optimizations is in use.
984 2005-01-26 Marek Michalkiewicz <marekm@amelek.gda.pl>
988 * config/avr/avr.c (notice_update_cc): Only set condition code for
989 ashrqi3 if shift count > 0.
990 (out_shift_with_cnt): Handle shift count <= 0 as a no-op.
991 (ashlqi3_out, ashlhi3_out, ashlsi3_out, ashrqi3_out, ashrhi3_out,
992 ashrsi3_out, lshrqi3_out, lshrhi3_out, lshrsi3_out): Handle shift
993 count <= 0 as a no-op, and shift count >= width by copying zero
994 or sign bit to all bits of the result.
995 * config/avr/avr.md (all shifts): Add alternatives for zero shift
996 count, with attribute "length" set to 0 and "cc" set to "none".
998 2005-01-26 Aldy Hernandez <aldyh@redhat.com>
1000 * doc/invoke.texi: Document -mTLS.
1002 * testsuite/gcc.target/frv/all-tls-global-dynamic.c: New.
1003 * testsuite/gcc.target/frv/all-tls-initial-exec.c: New.
1004 * testsuite/gcc.target/frv/all-tls-initial-exec-pic.c: New.
1005 * testsuite/gcc.target/frv/all-tls-local-dynamic.c: New.
1006 * testsuite/gcc.target/frv/all-tls-local-dynamic-plt.c: New.
1007 * testsuite/gcc.target/frv/all-tls-local-dynamic-plt-pic.c: New.
1008 * testsuite/gcc.target/frv/all-tls-local-exec.c: New.
1009 * testsuite/gcc.target/frv/all-tls-local-exec-TLS.c: New.
1011 * configure.ac: Check for a TLS capable gas.
1013 * configure: Regenerate.
1015 2005-01-26 Richard Henderson <rth@redhat.com>
1018 * combine.c (make_field_assignment): Simplify store to zero_extract
1019 from a source with an overlapping mask.
1021 2005-01-26 Aldy Hernandez <aldyh@redhat.com>
1023 2004-11-11 Eric Christopher <echristo@redhat.com>
1025 * config/frv/frv.md: Add fr400_integer automaton. Don't
1026 allow TYPE_MUL, TYPE_MACC, or TYPE_CUT to issue in parallel
1029 2005-01-26 Steven Bosscher <stevenb@suse.de>
1032 * cfgbuild.c (make_edges): Do not clear or set
1033 current_function_has_computed_jump.
1034 * function.h (struct function): Remove the has_computed_jump field.
1035 (current_function_has_computed_jump): Do not define.
1036 * sched-rgn.c (is_cfg_nonregular): Return true if a basic block ends
1037 in a computed jump. Ignore current_function_has_computed_jump.
1039 2005-01-26 Richard Henderson <rth@redhat.com>
1042 * expr.c (categorize_ctor_elements): New argument p_must_clear.
1043 (categorize_ctor_elements_1): Likewise. Detect a union that isn't
1045 (mostly_zeros_p): Update for new categorize_ctor_elements argument.
1046 * gimplify.c (gimplify_init_constructor): Likewise. Only shove
1047 objects into static storage if they have more than one non-zero value.
1048 * tree.h (categorize_ctor_elements): Update decl.
1050 2005-01-26 Steven Bosscher <stevenb@suse.de>
1053 * tree.h (CALL_EXPR_TAILCALL): Add comment.
1054 * calls.c (check_sibcall_argument_overlap_1): Revert the change
1055 to this function from 2004-07-10.
1056 * tree-tailcall.c (suitable_for_tail_opt_p): Do not consider the
1057 the current function for tail call optimizations if the address
1058 of one of it its arguments is taken.
1060 2005-01-26 Kazu Hirata <kazu@cs.umass.edu>
1062 * cse.c (fold_rtx): Call equiv_constant only when necessary.
1064 2005-01-26 Bernd Schmidt <bernd.schmidt@analog.com>
1066 * function.c (expand_function_end): If current_function_calls_alloca,
1067 emit stack restore in a place that is reached when the function falls
1070 2005-01-26 Alexandre Oliva <aoliva@redhat.com>
1072 * Makefile.in (../$(build_subdir)/fixincludes/fixincl): Add dummy
1073 rule to avoid matching inappropriate suffix rule.
1075 2005-01-26 Jan Hubicka <jh@suse.cz>
1077 PR tree-optimization/19241
1078 * tree-inline.c (copy_body_r): Do not walk subtrees after substituting.
1080 2005-01-26 Hans-Peter Nilsson <hp@axis.com>
1082 * config/cris/cris.h (LIB_SPEC) <sim*>: Use -lsyslinux, not
1083 -lsyssim. Wrap -lc -lsyssim in --start-group --end-group instead
1085 (STARTFILE_SPEC): Add crti.o, before crtbegin.o.
1086 <sim*>: Always use crt1.o, regardless of N in simN.
1087 <!sim>: Drop support for -pg and -p variants; always use crt0.o.
1088 (ENDFILE_SPEC): Add crtn.o, after crtend.o.
1089 (CRT_CALL_STATIC_FUNCTION): Remove.
1090 * config/cris/aout.h (STARTFILE_SPEC): <!melinux && sim*>: Always
1091 use crt1.o, regardless of N in simN.
1092 <!sim>: Drop support for -pg and -p variants; always use crt0.o.
1093 (LIB_SPEC): Do not define; default to the one in config/cris/cris.h.
1094 (CRIS_CPP_SUBTARGET_SPEC) <melinux && !nostdinc>: Remove special
1095 case isystem setting.
1096 (MAX_OFILE_ALIGNMENT): Correct to 32, not 16.
1098 2005-01-25 Ian Lance Taylor <ian@airs.com>
1100 * gimple-low.c (block_may_fallthru): Correct handling of
1101 SWITCH_EXPR--may fall through if SWITCH_LABELS is NULL.
1103 2005-01-26 Hans-Peter Nilsson <hp@axis.com>
1105 * config/cris/cris.h (CPP_SPEC): Adjust the various mtune=
1106 handlers to match indentation; nested under the mtune=* handler.
1107 Similar for mcpu= and march=.
1109 2005-01-25 Daniel Jacobowitz <dan@codesourcery.com>
1111 * config/mips/linux.h (LIB_SPEC): Remove -rpath-link.
1112 * config/mips/linux64.h (LIB_SPEC): Likewise.
1113 * config/mn10300/linux.h (LIB_SPEC): Likewise.
1114 * config/sh/linux.h (LIB_SPEC): Likewise.
1116 2005-01-25 Aldy Hernandez <aldyh@redhat.com>
1118 * config/frv/frv.c (frv_legitimize_tls_address): New.
1119 (TARGET_HAVE_TLS): Define.
1120 (FRV_SYMBOL_REF_TLS_P): Define.
1121 (frv_override_options): Handle new register classes.
1122 (frv_legitimate_address_p): Reject tls addresses.
1123 (frv_legitimize_address): Handle TLS addresses.
1124 (gen_inlined_tls_plt): New.
1126 (frv_legitimize_tls_address): New.
1127 (unspec_got_name): Add TLS entries.
1128 (got12_operand): Add R_FRV_TLSMOFF12 case.
1129 (frv_emit_move): Fixup TLS addresses.
1130 (frv_emit_movsi): Legitimize TLS addresses.
1132 * config/frv/frv.h (MASK_BIG_TLS): New.
1133 (TARGET_BIG_TLS): New.
1134 (HAVE_AS_TLS): Define.
1135 (TARGET_SWITCHES): Add -mTLS and -mtls options.
1136 (enum reg_class): Add GR8_REGS, GR9_REGS, GR89_REGS.
1137 (REG_CLASS_NAMES): Same.
1138 (REG_CLASS_CONTENTS): Same.
1139 (CONSTRAINT_LEN): New.
1140 (REG_CLASS_FROM_CONSTRAINT): New.
1141 (PREDICATE_CODES): Add symbolic_operand.
1143 * config/frv/frv.md (define_constants): Add UNSPEC_GETTLSOFF,
1144 UNSPEC_TLS_LOAD_GOTTLSOFF12, UNSPEC_TLS_INDIRECT_CALL,
1145 UNSPEC_TLS_TLSDESC_LDD, UNSPEC_TLS_TLSDESC_LDD_AUX,
1146 UNSPEC_TLS_TLSOFF_LD, UNSPEC_TLS_LDDI, UNSPEC_TLSOFF_HILO,
1147 R_FRV_GOTTLSOFF_HI, R_FRV_GOTTLSOFF_LO, R_FRV_TLSMOFFHI,
1148 R_FRV_TLSMOFFLO, R_FRV_TLSMOFF12, R_FRV_TLSDESCHI,
1149 R_FRV_TLSDESCLO, R_FRV_GOTTLSDESCHI, R_FRV_GOTTLSDESCLO, GR8_REG,
1150 GR9_REG, GR14_REG, LRREG.
1151 (type): Add load_or_call attribute.
1152 ("load_or_call"): New reservation.
1153 ("call_gettlsoff"): New.
1154 ("tls_indirect_call"): New.
1155 ("tls_load_gottlsoff12"): New.
1156 ("tlsoff_hilo"): New.
1157 ("tls_tlsdesc_ldd"): New.
1158 ("tls_tlsoff_ld"): New.
1161 * config/frv/frv-protos.h (symbolic_operand): Protoize.
1163 2005-01-25 Alexandre Oliva <aoliva@redhat.com>
1165 * config/frv/frv.h (ASM_OUTPUT_DWARF_DTPREL): Define.
1166 * config/frv/frv-protos.h (frv_output_dwarf_dtprel): Declare.
1167 * config/frv/frv.c (TLS_BIAS): Define.
1168 (frv_output_dwarf_dtprel): New.
1170 2005-01-26 Jakub Jelinek <jakub@redhat.com>
1172 PR rtl-optimization/19579
1173 * ifcvt.c (noce_try_cmove_arith): If emitting instructions to set up
1174 both A and B, see if they don't clobber registers the other expr uses.
1176 2005-01-25 J"orn Rennecke <joern.rennecke@st.com>
1178 * real.c (do_add): Initialize signalling and canonical members.
1180 * real.c (real_from_integer): Zero out destination.
1182 2005-01-18 Jan Hubicka <jh@suse.cz>
1184 PR tree-optimize/19337
1185 * tree-inline.c (initialize_inlined_parameters): Copy
1186 saved_static_chain_decl only.
1188 2005-01-25 Richard Henderson <rth@redhat.com>
1192 * config/i386/i386.c (x86_inter_unit_moves): Zero.
1193 (ix86_preferred_reload_class): Rewrite fp-constant section, with
1194 80387 enabled, to return a proper subclass. Return the subset that
1195 overlaps with GENERAL_REGS for PLUS.
1196 * config/i386/i386.md (movsi_1, movdi_2): Set type to "mmx" for pxor.
1198 2005-01-25 Richard Henderson <rth@redhat.com>
1201 * tree-complex.c (expand_complex_div_wide): Use the correct formulae.
1203 2005-01-25 Roger Sayle <roger@eyesopen.com>
1205 * expmed.c (choose_multiplier): Change interface to return the
1206 multiplier lowpart as an rtx instead of an unsigned HOST_WIDE_INT.
1207 (expand_divmod): Update calls to choose_multiplier.
1209 2005-01-24 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1212 * expmed.c (extract_bit_field): Use adjust_address instead of
1213 gen_lowpart when op0 is a MEM.
1215 2005-01-24 Steven Bosscher <stevenb@suse.de>
1217 * cfgexpand.c (expand_gimple_tailcall): Fix typo.
1219 2005-01-24 Richard Sandiford <rsandifo@redhat.com>
1221 * config/mips/irix-csr.c: New file.
1222 * config/mips/t-iris6 (irix-csr.o): New rule to build it.
1223 (EXTRA_MULTILIB_PARTS): Add irix-csr.o.
1224 * config/mips/iris6.h (ENDFILE_SPEC): Include it in n32 and n64
1227 2005-01-24 Eric Botcazou <ebotcazou@libertysurf.fr>
1230 * config.gcc (sparc-*-elf*): Remove sol2.h, sparc/sol2.h and
1231 sparc/elf.h, add sparc/sp-elf.h.
1232 (sparc-*-rtems*): Likewise.
1233 (sparclite-*-elf*): Remove sol2.h, sparc/sol2.h, sparc/elf.h and
1234 tm-dwarf2.h, add sparc/sp-elf.h.
1235 (sparc86x-*-elf): Likewise.
1236 (sparc64-*-elf*): Remove sol2.h, sparc/sol2.h and tm-dwarf2.h.
1237 * config/sparc/liteelf.h (TARGET_SUB_OS_CPP_BUILTINS): Rename into
1238 TARGET_OS_CPP_BUILTINS.
1239 * config/sparc/sp86x-elf (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1240 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1242 * config/sparc/openbsd64.h (NO_IMPLICIT_EXTERN_C): Undefine.
1243 * config/sparc/sp64-elf.h (NO_IMPLICIT_EXTERN_C): New macro.
1244 (SWITCH_TAKES_ARG): Likewise.
1245 (LOCAL_LABEL_PREFIX): Likewise.
1246 (ASM_GENERATE_INTERNAL_LABEL): Likewise.
1247 (TARGET_N_FORMAT_TYPES): Delete.
1248 (TARGET_FORMAT_TYPES): Likewise.
1249 (ASM_DECLARE_FUNCTION_SIZE): Likewise.
1250 * config/sparc/elf.h: Delete.
1251 * config/sparc/sp-elf.h: New file.
1253 2005-01-24 Kazu Hirata <kazu@cs.umass.edu>
1255 * tree-flow-inline.h (phi_arg_from_edge): Remove.
1256 * tree-flow.h: Remove the corresponding prototype.
1257 * tree-ssa-operands.h, tree-ssa-threadupdate.c, tree-ssa.c,
1258 tree-vectorizer.c: Use dest_idx instead of phi_arg_from_edge.
1260 2005-01-24 Jakub Jelinek <jakub@redhat.com>
1263 * config/rs6000/rs6000.c (rs6000_file_start): Create toc section
1264 for AIX ABI or ELF -fPIC.
1265 (rs6000_emit_load_toc_table): Don't create toc_section here.
1266 (rs6000_xcoff_file_start): Nor here.
1268 2005-01-24 J"orn Rennecke <joern.rennecke@st.com>
1270 * sh.c (ra.h): Don't #include.
1271 (hard_regs_intersect_p): New function, resurrected from ra.c.
1273 * sh.c: Fix 1996 Copyright.
1275 2005-01-24 Richard Henderson <rth@redhat.com>
1276 Aldy Hernandez <aldyh@redhat.com>
1278 * regrename.c (note_sets): Handle subregs.
1280 2005-01-24 Joseph S. Myers <joseph@codesourcery.com>
1283 * genconditions.c (write_header, write_conditions): Elide file if
1286 2005-01-24 Kazu Hirata <kazu@cs.umass.edu>
1288 * c-opts.c, expmed.c, gengtype.c, tree-complex.c,
1289 config/arm/iwmmxt.md, config/avr/avr.c, config/s390/s390.h:
1292 2005-01-24 Jakub Jelinek <jakub@redhat.com>
1294 * builtins.c (get_pointer_alignment, gimplify_va_arg_expr,
1295 expand_builtin_printf, expand_builtin_fprintf,
1296 expand_builtin_sprintf): Use POINTER_TYPE_P instead TREE_CODE
1297 checking against POINTER_TYPE.
1298 (validate_arglist): Handle POINTER_TYPE code by checking
1301 2005-01-24 Paolo Bonzini <bonzini@gnu.org>
1303 * rtlanal.c (non_rtx_starting_operands, for_each_rtx_1,
1305 (for_each_rtx): Call for_each_rtx_1.
1306 * rtl.h (init_rtlanal): Declare.
1307 * toplev.c (backend_init): Call init_rtlanal.
1309 2005-01-24 Jakub Jelinek <jakub@redhat.com>
1311 * flow.c (propagate_one_insn): Formatting.
1314 * flow.c (libcall_dead_p): Be more conservative if unsure.
1315 If there are any instructions between insn and call, see if they are
1316 all dead before saying the libcall is dead.
1318 2005-01-24 Uros Bizjak <uros@kss-loka.si>
1320 * config/i386/i386.md (*extendsfdf2_sse): Use "nonimmediate_operand"
1321 operand constraint for operand 0.
1323 2005-01-23 Richard Henderson <rth@redhat.com>
1326 * tree-complex.c (expand_complex_div_wide): Set EDGE_FALLTHRU on
1327 edges leading to join block.
1329 2005-01-23 Roger Sayle <roger@eyesopen.com>
1331 * expmed.c (expand_mult_highpart): Make static. Change type of
1332 constant multiplier argument from unsigned HOST_WIDE_INT to rtx.
1333 (expand_divmod): Updates calls to expand_mult_highpart by using
1334 gen_int_mode to pass a CONST_INT rtx.
1335 * rtl.h (expand_mult_highpart): Remove prototype.
1337 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
1339 * flow.c (regset_bytes, regset_size): Remove.
1341 * cfganal.c: Fix a reference to Harvey's paper.
1343 2005-01-23 Steven Bosscher <stevenb@suse.de>
1345 PR rtl-optimization/19464
1346 * tree-optimize.c (init_tree_optimization_passes): Add one more
1347 copyrename pass just before out-of-ssa.
1349 2005-01-23 Joseph S. Myers <joseph@codesourcery.com>
1352 * recog.c (recog_memoized): Don't define if GENERATOR_FILE.
1353 * ggc-none.c (ggc_free): Define.
1355 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
1357 * cse.c (max_reg, max_insn_uid): Remove.
1358 (cse_main): Don't access max_reg or max_insn_uid.
1360 * c-common.c (explicit_flag_signed_bitfields,
1361 lang_statement_code_p, lang_expand_function_end): Remove.
1362 * c-common.h: Remove the corresponding declarations.
1363 * c-opts.c (c_common_handle_option): Remove all write access
1364 to explicit_flag_signed_bitfields.
1366 * cgraph.c (cgraph_varpool_n_nodes): Remove.
1367 (cgraph_varpool_node): Don't access cgraph_varpool_n_nodes.
1368 * cgraph.h: Remove the corresponding declaration.
1370 * gcse.c (null_pointer_info): Remove.
1372 2005-01-23 Roger Sayle <roger@eyesopen.com>
1373 Eric Botcazou <ebotcazou@libertysurf.fr>
1375 * combine.c (simplify_logical): Only simplify logical expressions
1376 of the form ior(and(x,y),z) by the inverse distributive law if the
1377 result is cheaper than the original.
1379 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
1381 * alias.c, c-common.h, c-incpath.c, c-incpath.h, expr.c,
1382 fold-const.c, gimplify.c, params.h, tree-data-ref.c,
1383 tree-if-conv.c, tree-nested.c, tree-outof-ssa.c,
1384 tree-ssa-dom.c, tree-vectorizer.c, tree.def, config/darwin.c,
1385 config/freebsd-spec.h, config/arm/arm.h,
1386 config/h8300/h8300.md, config/i386/i386.md,
1387 config/i386/predicates.md, config/i386/sse.md,
1388 config/ia64/ia64.c, config/ip2k/ip2k.c, config/s390/s390.c,
1389 config/vax/vax.md: Fix comment typos. Follow spelling
1392 2005-01-23 Steven Bosscher <stevenb@suse.de>
1394 * gengtype.c (walk_type): Produce `not equal to' compares for loop
1395 tests, instead of `less than'.
1397 2005-01-23 Richard Sandiford <rsandifo@redhat.com>
1399 * function.c (assign_parm_setup_block): When creating a new stack slot
1400 for a parameter, get its alignment from the parameter's DECL_ALIGN
1401 rather than the type's TYPE_ALIGN. Make sure that the parameter
1402 is at least word aligned.
1404 2005-01-22 Roger Sayle <roger@eyesopen.com>
1407 * config/avr/avr.c (avr_hard_regno_mode_ok): Rewrite.
1409 2005-01-22 Richard Henderson <rth@redhat.com>
1412 * config/i386/i386.md (movsfcc_1_sse_max): Use nonimmediate_operand
1413 in both compare operands.
1414 (movdfcc_1_sse_max): Likewise.
1415 (movsfcc_1_sse): Likewise. Add earlyclobber for scratch.
1416 (movdfcc_1_sse): Likewise.
1417 * config/i386/i386.c (ix86_split_sse_movcc): Emit copies into the
1418 scratch register as needed.
1420 2005-01-22 Richard Henderson <rth@redhat.com>
1422 * config/i386/i386.md (smaxsf3): Fix mnemonic typo.
1424 * config/i386/i386.c (ix86_prepare_fp_compare_args): Fix is_sse test.
1426 2005-01-22 Richard Henderson <rth@redhat.com>
1428 * genopinit.c (optabs): Use smin/smax for floating point too.
1429 * doc/md.texi: Update to match. Clarify that floating point
1430 results are undefined for +0/-0 and NaN.
1431 * doc/rtl.texi: Likewise.
1432 * rtl.def (SMIN, SMAX): Likewise
1433 * tree.def (MIN_EXPR, MAX_EXPR): Likewise.
1435 * config/alpha/alpha.md (smaxdf3, smindf3, smaxsf3, sminsf3): Add
1436 leading 's' to the name.
1437 * config/ia64/ia64.md (smaxsf3, sminsf3, smaxdf3, smindf3,
1438 smaxxf3, sminxf3): Likewise.
1439 * config/rs6000/rs6000.md (smaxdf3, smindf3, smaxsf3, sminsf3):
1442 2005-01-22 Paul Brook <paul@codesourcery.com>
1444 * config/arm/arm.md: Use "Uy" constraint for wcgr load/stores.
1446 2005-01-21 Mark Dettinger <dettinge@de.ibm.com>
1448 * config/s390/s390.c (struct processor_costs): New fields
1449 dlgr, dlr, dr, dsgfr, dsgr.
1450 (z900_cost, z990_cost): Values for new fields.
1451 (s390_rtx_costs): New cases MEM und COMPARE in switch
1452 statement. Modified handling of SIGN_EXTEND, ZERO_EXTEND,
1453 DIV, MOD, UDIV, UMOD.
1455 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
1457 * config/s390/s390.md ("doloop_si64"): Reload input value directly
1458 into the register being decremented.
1459 ("doloop_si31", "doloop_di"): Likewise.
1460 ("*doloop_si_long"): Adapt pattern.
1462 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
1464 * config/s390/s390.h (HARD_REGNO_NREGS): Fix computation for
1466 (CLASS_MAX_NREGS): Likewise.
1468 2005-01-21 Daniel Berlin <dberlin@dberlin.org>
1470 * doc/tree-ssa.texi (Statement Operands): Add example for new
1471 must-def macro. Note deprecation of old operands interface.
1473 2005-01-22 Richard Sandiford <rsandifo@redhat.com>
1475 PR tree-optimization/19484
1476 * tree-cfg.c (remove_fallthru_edge): New function.
1477 (cleanup_control_flow): Remove fallthru edges from calls that are
1478 now known not to return.
1480 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
1482 Fix PR tree-optimization/19038
1483 * tree-ssa-dom.c (cprop_operand): Don't replace loop invaeriant
1484 copies with loop variant ones.
1486 2005-01-22 Kazu Hirata <kazu@cs.umass.edu>
1488 * cfganal.c, real.h, reorg.c, timevar.def, tree-ssa-ccp.c,
1489 config/alpha/alpha-protos.h, config/alpha/alpha.h,
1490 config/alpha/alpha.md, config/alpha/predicates.md,
1491 config/sparc/freebsd.h, config/sparc/netbsd-elf.h,
1492 config/sparc/sol2.h: Update copyright.
1494 * tree-cfg.c (remove_forwarder_block_with_phi): Look at the
1495 first label to see if it is a nonlocal label.
1497 2005-01-22 David Edelsohn <edelsohn@gnu.org>
1498 Andrew Pinski <pinskia@physics.uc.edu>
1501 * config/rs6000/rs6000.c (rs6000_va_start): Saturate n_gpr at
1502 maximum number of GPRs. Saturate n_fpr at maximum number of FPRs.
1504 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
1506 * config/rs6000/rtems.h: Update copyright.
1508 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
1511 * config/rs6000/rtems.h: Resurrect cpp_os_rtems_spec from gcc < 3.4.
1512 (CPP_OS_RTEMS_SPEC): New (From gcc-3.3's config/rs6000/sys4.h).
1513 (SUBSUBTARGET_EXTRA_SPECS): Use CPP_OS_RTEMS_SPEC.
1515 2005-01-22 Volker Reichelt <reichelt@igpm.rwth-aachen.de>
1518 * c-typeck.c (convert_arguments): Check for error_mark_node.
1520 2005-01-21 Ian Lance Taylor <ian@airs.com>
1522 PR tree-optimization/13000
1523 * tree-inline.c: Include "tree-flow.h".
1524 (expand_call_inline): If warn_return_type, warn if non-void inline
1525 function falls through.
1526 * tree-cfg.c (execute_warn_function_return): Don't warn about
1527 control reaching end if TREE_NO_WARNING is set. Set
1529 * gimple-low.c (block_may_fallthru): Don't assume that SWITCH_EXPR
1531 * gimplify.c (shortcut_cond_expr): Don't emit a jump over the else
1532 branch if we don't need one.
1533 * c-typeck.c: Include "tree-flow.h"
1534 (c_finish_bc_stmt): Don't add a goto if the current statement
1535 list doesn't fall through to the current point.
1537 2005-01-21 Roger Sayle <roger@eyesopen.com>
1539 PR rtl-optimization/576
1540 * real.c (real_arithmetic): Change return type from void to bool
1541 to return an indication that the result may be inexact.
1542 * real.h (real_arithmeric): Update prototype.
1543 * fold-const.c (const_binop): Don't constant fold floating
1544 point expressions when the user specifies -frounding-math and
1545 the result may depend upon the run-time rounding mode.
1546 (fold_convert_const_real_from_real): Clean-up.
1547 (fold_initializer): Ignore flag_rounding_math for initializers.
1548 * simplify-rtx.c (simplify_binary_operation): Likewise, don't
1549 constant fold FP operations with flag_rounding_math if the
1550 result may depend upon the run-time rounding mode.
1552 2005-01-21 Tom Tromey <tromey@redhat.com>
1554 * c-cppbuiltin.c (define__GNUC__): Correct assertion.
1556 2005-01-21 Kazu Hirata <kazu@cs.umass.edu>
1558 * tree-cfg.c: Fix comment typos.
1560 2005-01-21 J"orn Rennecke <joern.rennecke@st.com>
1562 * passes.c (rest_of_handle_flow2): Close / open the flow2 dump file
1563 around the call to rest_of_handle_branch_target_load_optimize.
1565 2005-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
1567 * config.gcc (sparc-*-netbsdelf*): Include dbxelf.h
1568 (sparc64-*-openbsd*): Likewise.
1569 (sparclite-*-elf*): Include tm-dwarf2.h.
1570 (sparc86x-*-elf*): Likewise.
1571 (sparc64-*-elf*): Likewise.
1572 * config/sparc/linux64.h (DWARF2_DEBUGGING_INFO): Delete.
1573 (DBX_DEBUGGING_INFO): Likewise.
1574 * config/sparc/liteelf.h (DWARF2_DEBUGGING_INFO): Likewise.
1575 (PREFERRED_DEBUGGING_TYPE): Likewise.
1576 * config/sparc/netbsd-elf.h (PREFERRED_DEBUGGING_TYPE): Likewise.
1577 * config/sparc/sp64-elf.h (DBX_DEBUGGING_INFO): Likewise.
1578 (PREFERRED_DEBUGGING_TYPE): Likewise.
1579 * config/sparc/sp86x-elf.h (DWARF2_DEBUGGING_INFO): Likewise.
1580 (PREFERRED_DEBUGGING_TYPE): Likewise.
1581 * config/sparc/sparc.h (DBX_DEBUGGING_INFO): Likewise.
1583 2005-01-21 Hans-Peter Nilsson <hp@bitrange.com>
1586 * combine.c (combine_simplify_rtx): Revert change of 2004-12-31.
1588 2005-01-20 Paul Brook <paul@codesourcery.com>
1590 * doc/rtl.texi: Document value extension requirements for CONST_INT.
1592 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
1594 * tree-cfg.c (tree_verify_flow_info): Check that a nonlocal
1595 label is first in a sequence of labels.
1597 * tree-cfg.c (tree_verify_flow_info): Fix a typo.
1599 2005-01-20 Janis Johnson <janis187@us.ibm.com>
1600 Giovanni Bajo <giovannibajo@gcc.gnu.org>
1602 * doc/sourcebuild.texi (Test Directives): New.
1604 2005-01-20 Roger Sayle <roger@eyesopen.com>
1606 * real.c (real_floor): Don't leave the result uninitialized when
1607 mode is VOIDMode, but return the unrounded intermediate value.
1608 (real_ceil): Likewise.
1610 2005-01-20 Paul Brook <paul@codesourcery.com>
1612 * config/arm/arm.md (insv): Use gen_int_mode.
1614 2005-01-20 Eric Botcazou <ebotcazou@libertysurf.fr>
1616 * config/sparc/freebsd.h (ASM_OUTPUT_INTERNAL_LABELREF): Delete.
1617 * config/sparc/linux.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1618 * config/sparc/linux64.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1619 * config/sparc/netbsd-elf.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1620 * config/sparc/sol2.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1622 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
1625 * dwarf2out.c (concat_loc_descriptor): We don't know we can use
1626 fbreg, so we have to assume we can't.
1627 (loc_descriptor_from_tree_1): Ditto.
1628 (containing_function_has_frame_base): New function.
1629 (add_location_or_const_value_attribute): Use it.
1630 Also try to generate a frame_base from a single element location
1633 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
1635 PR tree-optimization/15349
1636 * timevar.def (TV_TREE_MERGE_PHI): New.
1637 * tree-cfg.c (tree_forwarder_block_p): Add a new argument
1639 (remove_forwarder_block, cleanup_forwarder_blocks): Adjust the
1640 calls to tree_forwarder_block_p.
1641 (remove_forwarder_block_with_phi, merge_phi_nodes,
1642 gate_merge_phi, pass_merge_phi): New.
1643 * tree-optimize.c (init_tree_optimization_passes): Add
1645 * tree-pass.h: Add an extern for pass_merge_phi;
1647 2005-01-20 Richard Henderson <rth@redhat.com>
1650 * config/i386/emmintrin.h (_mm_castpd_ps, _mm_castpd_si128): New.
1651 (_mm_castps_pd, _mm_castps_si128): New.
1652 (_mm_castsi128_ps, _mm_castsi128_pd): New.
1654 2005-01-20 Richard Henderson <rth@redhat.com>
1657 * config/i386/mmintrin.h (_mm_cvtsi32_si64): Use
1658 __builtin_ia32_vec_init_v2si.
1659 (_mm_cvtsi64_si32): Use __builtin_ia32_vec_ext_v2si.
1660 * config/i386/i386.c (IX86_BUILTIN_VEC_EXT_V2SI): New.
1661 (ix86_init_mmx_sse_builtins): Create it.
1662 (ix86_expand_builtin): Expand it.
1663 (ix86_expand_vector_set): Handle V2SFmode and V2SImode.
1664 * config/i386/mmx.md (vec_extractv2sf_0, vec_extractv2sf_1): New.
1665 (vec_extractv2si_0, vec_extractv2si_1): New.
1667 2005-01-20 Richard Henderson <rth@redhat.com>
1670 * config/i386/i386.c (ix86_expand_push): New.
1671 * config/i386/mmx.md (push<MMXMODE>1): New.
1672 * config/i386/sse.md (push<SSEMODE>1): New.
1673 * config/i386/i386-protos.h: Update.
1675 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
1677 * params.def, params.h, predict.c, tree-ssa-loop-im.c: Update
1680 2005-01-20 Richard Henderson <rth@redhat.com>
1682 * rtl.def (CONST_VECTOR): Use RTX_CONST_OBJ.
1683 * rtl.h (CONSTANT_P): Don't special case CONST_VECTOR.
1685 2005-01-19 Richard Henderson <rth@redhat.com>
1688 * config/i386/i386.c (ix86_expand_vector_move_misalign): Convert
1689 to V4SFmode in SSE1 fallback load path.
1691 2005-01-19 Richard Henderson <rth@redhat.com>
1693 * config/i386/i386.c (ix86_expand_vector_init_one_var): Fix typo
1694 in QImode expansion to ix86_expand_vector_set.
1695 (ix86_expand_vector_init_general): Fix typo in V8HImode recursive call.
1697 2005-01-19 Richard Henderson <rth@redhat.com>
1700 * config/i386/i386.c (ix86_preferred_reload_class): Return a proper
1701 subclass of the input class.
1702 (ix86_secondary_memory_needed): Always true for cross-MMX classes.
1703 Always true for cross-SSE1 classes. Rationalize conditionals.
1704 * config/i386/i386.h (SSE_CLASS_P, MMX_CLASS_P): Use straight equality.
1705 * config/i386/i386.md (movsi_1): Add MMX/SSE zeros. Fix alternatives
1706 for SSE1. Don't check TARGET_INTER_UNIT_MOVES.
1707 (movdi_2): Add MMX/SSE zeros.
1708 (movdi_1_rex64): Likewise. Don't check TARGET_INTER_UNIT_MOVES.
1709 (movsf_1): Don't check TARGET_INTER_UNIT_MOVES.
1710 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
1711 (movsi_1_nointernunit, movdi_1_rex64_nointerunit): Remove.
1712 (movsf_1_nointerunit, zero_extendsidi2_32_1): Remove.
1713 (zero_extendsidi2_rex64_1): Remove.
1714 (MOV0 peephole): Check GENERAL_REG_P.
1716 2005-01-19 Richard Henderson <rth@redhat.com>
1719 * config/i386/i386.c (ix86_expand_vector_set): Fix third and fourth
1721 (ix86_expand_vector_extract): Likewise.
1723 2005-01-19 Richard Henderson <rth@redhat.com>
1726 * expr.c (emit_move_change_mode): New argument force; use
1727 simplify_gen_subreg if true.
1728 (emit_move_via_alt_mode): Merge into ...
1729 (emit_move_via_integer): ... here.
1730 (emit_move_ccmode): Use emit_move_change_mode directly.
1732 2005-01-19 Richard Henderson <rth@redhat.com>
1735 * config/alpha/alpha.c (alpha_rtx_costs): Handle HIGH.
1736 (alpha_preferred_reload_class): Handle CONST_VECTOR.
1737 (alpha_emit_set_const_1): Add no_output parameter; don't emit
1739 (alpha_emit_set_const): Likewise. Make static.
1740 (alpha_emit_set_long_const): Make static.
1741 (alpha_extract_integer): Split out from alpha_expand_mov.
1742 (alpha_split_const_mov): Likewise.
1743 (alpha_expand_mov): Use them. Handle CONST_VECTOR.
1744 (alpha_legitimate_constant_p): New.
1745 * config/alpha/alpha-protos.h: Update.
1746 * config/alpha/alpha.h (REGISTER_MOVE_COST): Correct fp<->gp cost.
1747 (LEGITIMATE_CONSTANT_P): Re-implement with a function.
1748 * config/alpha/alpha.md (movsi): Add n alternative.
1749 (movsi_nt_vms, movdi_er_nofix, movdi_er_fix, movdi_fix): Likewise.
1750 (mov<VEC>_fix, mov<VEC>_nofix): Add i alternative.
1751 (splitters for all of the above): Use alpha_split_const_mov.
1752 * config/alpha/predicates.md (non_add_const_operand): New.
1753 (non_zero_const_operand): New.
1754 (input_operand): Use alpha_legitimate_constant_p after reload.
1756 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
1758 PR tree-optimization/19038
1759 * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): New function.
1760 (add_candidate): Add ivs with increment in latch only if
1761 allow_ip_end_pos_p is true.
1762 (determine_iv_cost): Use empty_block_p.
1764 2005-01-19 Daniel Berlin <dberlin@dberlin.org>
1766 * cfganal.c (compute_dominance_frontiers_1): Replace with new algorithm
1767 (compute_dominance_frontiers): Ditto.
1769 2005-01-19 Ralf Corsepius <ralf.corsepius@rtems.org>
1772 * config/sh/t-rtems: New.
1773 * config.gcc (sh-*-rtems*): Reflect having added config/sh/t-rtems.
1775 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
1777 * tree-ssa-ccp.c (maybe_fold_offset_to_component_ref): Always subtract
1778 the offset of the selected field.
1780 2005-01-19 Kazu Hirata <kazu@cs.umass.edu>
1782 * tree-cfg.c (remove_forwarder_block): Fix the check to
1783 prevent a nonlocal label from appearing in the middle of a
1786 2005-01-19 Hans-Peter Nilsson <hp@axis.com>
1788 PR rtl-optimization/19462
1789 * reorg.c (find_end_label): Create return insn only if
1790 current_function_epilogue_delay_list is empty.
1792 2005-01-19 Paolo Bonzini <bonzini@gnu.org>
1794 * doc/passes.texi: Remove paragraph mentioning new-ra.
1796 2005-01-19 Jakub Jelinek <jakub@redhat.com>
1798 PR rtl-optimization/15139
1799 * combine.c: Include params.h.
1800 (count_rtxs): New function.
1801 (record_value_for_reg): If replace_rtx would replace at least
1802 2 occurrences of REG in VALUE and TEM is really large, replace REG with
1803 (clobber (const_int 0)) instead of TEM.
1804 * params.def (PARAM_MAX_LAST_VALUE_RTL): New.
1805 * params.h (MAX_LAST_VALUE_RTL): New.
1806 * Makefile.in (combine.o): Depend on $(PARAMS_H).
1807 * doc/invoke.texi (--param max-last-value-rtl=N): Document.
1810 * c-typeck.c (digest_init): Only call build_vector if all constructor
1811 elements are *_CST nodes.
1812 * gimplify.c (gimplify_init_constructor): Likewise.
1815 * c-typeck.c (digest_init): Only call build_vector if inside_init
1818 2005-01-18 Tobias Schl"uter <tobias.schlueter@physik.uni-muenchen.de>
1820 * toplev.c (init_asm_output): Remove dead #ifdef.
1822 2005-01-18 Kaz Kojima <kkojima@gcc.gnu.org>
1824 * config/sh/linux.h (TARGET_C99_FUNCTIONS): Define.
1826 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
1828 * calls.c (expand_call): Check DECL_BUILT_IN_CLASS before
1829 accessing DECL_FUNCTION_CODE.
1830 * dojump.c (do_jump): Likewise.
1831 * gimplify.c (gimplify_call_expr): Likewise.
1832 * predict.c (expr_expected_value): Likewise.
1833 (strip_builtin_expect): Likewise.
1834 * tree-inline.c (estimate_num_insns_1): Likewise.
1835 * tree-ssa-loop-im.c (stmt_cost): Likewise
1836 * fold-const.c (fold): Test for BUILT_IN_NORMAL.
1837 (tree_expr_nonnegative_p): Likewise.
1839 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
1841 * config/sparc/sparc.c (load_pic_register): Emit the appropriate
1842 variant of the load_pcrel_sym pattern.
1843 * config/sparc/sparc.md (P macro): Move to the top.
1844 (load_pcrel_sym): Macroize using P.
1845 (save_register_window): Likewise.
1847 2005-01-18 Aldy Hernandez <aldyh@redhat.com>
1849 * config/rs6000/rs6000.md ("sunordered"): Disable for e500.
1852 2005-01-18 Andrew Pinski <pinskia@physics.uc.edu>
1855 * c-typeck.c (build_asm_expr): Strip nops off
1856 input memory operands.
1858 2005-01-18 David Edelsohn <edelsohn@gnu.org>
1860 * config/rs6000/aix43.h (CPLUSPLUS_CPP_SPEC): Delete
1861 _XOPEN_SOURCE, _XOPEN_SOURCE_EXTENDED, _LARGE_FILE_API.
1862 * config/rs6000/aix51.h (CPLUSPLUS_CPP_SPEC): Same.
1863 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Same.
1865 2005-01-18 Joel Sherrill <joel.sherrill@oarcorp.com>
1868 * config/i386/i386.c (override_options): If the 80387 is disabled,
1869 then do not return FP values using FP registers.
1871 2005-01-18 Kazu Hirata <kazu@cs.umass.edu>
1873 * basic-block.h, c-common.c, c-cppbuiltin.c, c-lang.c,
1874 c-tree.h, cfgbuild.c, cgraph.c, cgraph.h, collect2.c,
1875 combine.c, config.gcc, coverage.h, cse.c, cselib.c,
1876 defaults.h, df.c, dwarf2asm.c, dwarf2out.c, explow.c, expr.c,
1877 flow.c, fold-const.c, gcse.c, ggc-page.c, gimple-low.c,
1878 gimplify.c, ifcvt.c, langhooks-def.h, lcm.c, optabs.h,
1879 output.h, postreload-gcse.c, postreload.c, recog.c,
1880 resource.c, rtl.def, rtlanal.c, sched-deps.c, sched-rgn.c,
1881 targhooks.h, toplev.c, tree-data-ref.c, tree-eh.c,
1882 tree-flow.h, tree-loop-linear.c, tree-mudflap.h, tree-nrv.c,
1883 tree-optimize.c, tree-outof-ssa.c, tree-pass.h,
1884 tree-scalar-evolution.c, tree-ssa-copy.c, tree-ssa-dce.c,
1885 tree-ssa-dse.c, tree-ssa-loop-ivopts.c, tree-ssa-operands.c,
1886 tree-ssa-pre.c, tree-ssa.c, tree-vectorizer.c, tree.def,
1887 unwind-dw2-fde-darwin.c, var-tracking.c: Update copyright.
1889 * config/darwin.c, config/darwin.h, config/freebsd-spec.h,
1890 config/alpha/alpha.c, config/arm/vfp.md,
1891 config/i386/emmintrin.h, config/i386/i386-modes.def,
1892 config/i386/i386-protos.h, config/i386/i386.h,
1893 config/i386/pmmintrin.h, config/i386/ppro.md,
1894 config/i386/predicates.md, config/i386/xmmintrin.h,
1895 config/i860/i860.c, config/ia64/ia64-modes.def,
1896 config/ia64/ia64-protos.h, config/ia64/ia64.h,
1897 config/ia64/ia64.md, config/ia64/itanium1.md,
1898 config/ia64/itanium2.md, config/ia64/predicates.md,
1899 config/iq2000/iq2000.h, config/mips/linux64.h,
1900 config/rs6000/aix41.h, config/rs6000/aix43.h,
1901 config/rs6000/aix52.h, config/rs6000/darwin-fallback.c,
1902 config/rs6000/rs6000.c, config/rs6000/rs6000.h,
1903 config/rs6000/spe.md, config/sh/sh.md, config/sparc/linux.h,
1904 config/sparc/linux64.h, config/sparc/litecoff.h,
1905 config/sparc/sp64-elf.h, config/sparc/sparc.c,
1906 config/sparc/sparc.h, config/sparc/sparc.md,
1907 config/sparc/sysv4-only.h: Update copyright.
1909 2005-01-18 Richard Henderson <rth@redhat.com>
1912 * config/i386/i386.c (ix86_expand_fp_movcc): Fail for LTGT and UNEQ.
1914 2005-01-18 Marc Espie <espie@openbsd.org>
1915 * config/i386/openbsdelf.h: Typo.
1917 2005-01-18 Richard Henderson <rth@redhat.com>
1919 * config/i386/i386.c (ix86_expand_fp_movcc): Remove TARGET_IEEE_FP
1920 special case for sse.
1921 * config/i386/predicates.md (sse_comparison_operator): Likewise.
1923 2005-01-18 Dorit Naishlos <dorit@il.ibm.com>
1925 * tree-inline.c (estimate_num_insns_1): Added cases for
1926 ALIGN_INDIRECT_REF, MISALIGNED_INDIRECT_REF, and REALIGN_LOAD_EXPR.
1928 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
1930 PR rtl-optimization/19296
1931 * combine.c (simplify_comparison): Rewrite the condition under
1932 which a non-paradoxical SUBREG of a PLUS can be lifted when
1933 compared against a constant.
1935 2005-01-18 Andi Kleen <ak@muc.de>
1937 * c-typeck.c: (convert_for_assignment): Check warn_pointer_sign.
1938 * c.opt (-Wpointer-sign): Add.
1939 * doc/invoke.texi: (-Wpointer-sign): Add.
1941 2005-01-18 Uros Bizjak <uros@kss-loka.si>
1943 * config/i386/i386.c (override_options): Revert 2004-11-24 change.
1944 * config/i386/i386.md (fmodsf3, dremsf3, *sinsf2, *cossf2, sincossf3,
1945 *tansf3_1, tansf2, atan2sf3_1, atan2sf3, atansf2, asinsf2, acossf2,
1946 logsf2, log10sf2, log2sf2, log1psf2, logbsf2, ilogbsf2, expsf2,
1947 exp10sf2, exp2sf2, expm1sf2, rintsf2, floorsf2, ceilsf2, btruncsf2,
1948 nearbyintsf2): Disable for TARGET_SSE_MATH. Leave patterns enabled
1949 for TARGET_MIX_SSE_I387.
1950 (fmoddf3, dremdf3, *sindf2, *sinextendsfdf2, *cosdf2, *cosextendsfdf2,
1951 sincosdf3, *sincosextendsfdf3, *tandf3_1, tandf2, atan2df3_1,
1952 atan2df3, atandf2, asindf2, acosdf2, logdf2, log10df2, log2df2,
1953 log1pdf2, logbdf2, expdf2, exp10df2, exp2df2, expm1df2, rintdf2,
1954 floordf2, ceildf2, btruncdf2, nearbyintdf2): Disable for
1955 (TARGET_SSE2 && TARGET_SSE_MATH). Leave patterns enabled for
1956 TARGET_MIX_SSE_I387.
1957 (atan2sf3, atan2df3, atan2xf3): Remove register constraints
1960 2005-01-18 Uros Bizjak <uros@kss-loka.si>
1963 * config/i386/mmx.md (*movv2sf_internal_rex64, *movv2sf_internal):
1964 Add movaps alternative for xmm reg->reg move.
1966 2005-01-17 Jeff Law <law@redhat.com>
1968 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Clear SSA_NAME_VALUE
1969 at the end of the main DOM loop rather than just before DOM exits.
1971 2005-01-17 Diego Novillo <dnovillo@redhat.com>
1973 PR tree-optimization/19121
1974 * tree-ssa-alias.c (compute_flow_sensitive_aliasing): When
1975 adding aliases to a name tag, also add them to the pointer's
1977 * tree-ssa-copy.c (merge_alias_info): Do not merge flow
1978 sensitive alias info at all. Only check that the two pointers
1979 have compatible pointed-to sets.
1980 * tree-ssa.c (verify_name_tags): Verify that the alias set of
1981 a pointer's type tag is a superset of the alias set of the
1984 2005-01-17 James E Wilson <wilson@specifixinc.com>
1987 * config/ia64/ia64.md (movxf): Handle general register source. Adjust
1988 comment to document why.
1990 2005-01-17 Richard Henderson <rth@redhat.com>
1992 * config/i386/sse.md (smaxv4sf3_finite, sse_vmsmaxv4sf3_finite,
1993 sminv4sf3_finite, sse_vmsminv4sf3_finite, smaxv2df3_finite,
1994 sse2_vmsmaxv2df3_finite, sminv2df3_finite,
1995 sse2_vmsminv2df3_finite): New.
1996 (smaxv4sf3, sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, smaxv2df3,
1997 sse2_vmsmaxv2df3, sminv2df3, sse2_vmsminv2df3): Remove commutative.
1998 Force op1 into register.
2000 2005-01-17 Kelley Cook <kcook@gcc.gnu.org>
2002 * Makefile.in (STAGEMOVESTUFF): Stage all the stamp files to prevent
2003 bubblestrap from always rebuilding build directory.
2005 2005-01-17 Ulrich Weigand <uweigand@de.ibm.com>
2007 * config/s390/s390.md ("*clc"): Remove incorrect '='.
2008 ("*cmpmem_short"): Likewise.
2010 2005-01-17 Steven Bosscher <stevenb@suse.de>
2012 * basic-block.h: Document BB_* flags.
2013 * regrename.c (copyprop_hardreg_forward): Don't use BB_VISITED,
2014 use an sbitmap instead.
2015 * sched-rgn.c (compute_trg_info): Likewise.
2017 2005-01-17 Richard Sandiford <rsandifo@redhat.com>
2019 * config.gcc (mips64*-*-linux*): Set the default abi to n32. Remove
2020 redundant target_cpu_default and extra_parts lines. Use the default
2022 * config/mips/linux64.h (DRIVER_DEFAULT_ABI_SELF_SPEC): Delete.
2023 (SUBTARGET_EXTRA_SPECS): Delete.
2024 (DRIVER_SELF_SPECS): Remove architecture lines. Set the ABI directly.
2025 (SUBTARGET_ASM_SPEC): Remove -mabi=n32 and -mabi=64 mappings.
2026 * config/mips/t-linux64 (MULTILIB_DIRNAMES): Use "32" for -mabi=32
2027 and "n32" for -mabi=n32.
2029 2005-01-17 Ian Lance Taylor <ian@airs.com>
2032 * c-typeck.c (build_c_cast): Revert patch of 2003-02-16: don't
2033 fold constant variables into initial values.
2035 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
2037 * tree-cfg.c (tree_can_merge_blocks_p): Reorder two checks.
2039 * tree-cfg.c (tree_forwarder_block_p): Speed up by walking
2040 through the statements backward.
2042 2005-01-17 Ian Lance Taylor <ian@airs.com>
2044 PR middle-end/13127:
2045 * tree-inline.c (expand_call_inline): Set TREE_NO_WARNING on
2046 a variable set to the return value of the inlined function.
2048 2005-01-17 Mark Dettinger <dettinge@de.ibm.com>
2050 * rtlanal.c (rtx_cost): Assign cost of 0 to a SUBREG
2051 when modes are tieable.
2053 2005-01-17 Ranjit Mathew <rmathew@hotmail.com>
2055 * gthr-posix.h (__gthread_active_p): Use pthread_cancel instead
2056 of pthread_create to find out if threads are enabled.
2057 * gthr-posix95.h (__gthread_active_p): Likewise.
2059 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
2061 * common.opt (-fnew-ra): Remove.
2063 * toplev.h (flag_new_regalloc): Remove.
2064 * Makefile.in (ra*.*): Don't mention.
2065 * passes.c (rest_of_handle_new_regalloc): Remove.
2066 (rest_of_handle_combine, rest_of_compilation): Always consider
2067 flag_new_regalloc as false.
2068 * doc/invoke.texi: Don't document -fnew-ra.
2070 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
2072 * bb-reorder.c (fix_edges_for_rarely_executed_code): Remove
2073 last parameter to reg_scan.
2074 * loop.c (loop_optimize): Likewise.
2075 * passes.c (rest_of_handle_tracer, rest_of_handle_if_conversion,
2076 rest_of_handle_web, rest_of_handle_cfg, rest_of_handle_jump_bypass,
2077 rest_of_handle_life, rest_of_handle_cse, rest_of_handle_cse2,
2078 rest_of_handle_gcse, rest_of_handle_loop_optimize,
2079 rest_of_handle_loop2, rest_of_handle_jump2): Likewise.
2080 * regclass.c (reg_scan): Likewise, for the declaration.
2081 * rtl.h (reg_scan): Likewise, for the prototype.
2083 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
2085 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Don't call
2086 mark_dfs_back_edges.
2088 2005-01-16 Daniel Berlin <dberlin@dberlin.org>
2090 * tree-ssa-pre.c (add_to_sets): s1 may be NULL.
2091 (compute_avail): Uses don't go in tmp_gen.
2093 2005-01-16 Steven Bosscher <stevenb@suse.de>
2095 * ggc-page.c (ggc_alloc_stat): Use __builtin_ctzl instead of a
2096 loop to look for a free slot in a page entry.
2098 2005-01-16 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2101 * defaults.h (TARGET_DEFERRED_OUTPUT_DEFS): Provide default.
2102 * toplev.c (compile_file): Call process_pending_assemble_output_defs
2103 just before targetm.asm_out.file_end.
2104 * tree.h (process_pending_assemble_output_defs): Declare.
2105 * varasm.c (assemble_output_def, process_pending_assemble_output_defs):
2107 (assemble_alias): Defer generation of assembly code for defines when
2108 TARGET_DEFERRED_OUTPUT_DEFS is true.
2109 * config/rs6000/aix41.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
2110 * config/rs6000/aix43.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
2111 * doc/tm.texi (TARGET_DEFERRED_OUTPUT_DEFS): document.
2113 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2116 * pa.c (pa_scalar_mode_supported_p): New function.
2117 (TARGET_SCALAR_MODE_SUPPORTED_P): Define.
2118 * pa.h (MIN_UNITS_PER_WORD): Add comment.
2120 * pa-protos.h (prefetch_operand): Delete.
2121 (prefetch_cc_operand, prefetch_nocc_operand): New declations.
2122 * pa.c (prefetch_operand): Delete.
2123 (prefetch_cc_operand, prefetch_nocc_operand): New functions.
2124 * pa.h (EXTRA_CONSTRAINT): Add `W' constraint.
2125 (PREDICATE_CODES): Delete prefetch_operand. Add prefetch_cc_operand
2126 and prefetch_nocc_operand.
2127 * pa.md (prefetch): Rework to avoid reload problems handling short
2128 displacements when a cache control completer needs to be provided.
2129 (prefetch_32, prefetch_64): Delete.
2130 (prefetch_cc, prefetch_nocc): New patterns.
2132 2005-01-15 David Edelsohn <edelsohn@gnu.org>
2134 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Revert previous change.
2136 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2139 * pa.c (pa_asm_output_aligned_common): Warn if specified alignment
2140 exceeds maximum alignment for global common data.
2141 * pa64-hpux.h (MAX_OFILE_ALIGNMENT): Define to 32768.
2142 * som.h (MAX_OFILE_ALIGNMENT): Likewise.
2144 2004-01-15 Roger Sayle <roger@eyesopen.com>
2146 * tree-ssa-dom.c (extract_range_from_cond): Correct condition.
2148 2004-01-15 Roger Sayle <roger@eyesopen.com>
2150 * harg-reg-set.h (reg_class_names): Prototype global array.
2151 * regclass.c (reg_class_names): Declare here and initialize to
2153 (dump_regclass): Remove local declaration of reg_class_names.
2154 (regclass): Likewise.
2155 * cfg.c (dump_flow_info): Likewise.
2156 * ra-debug.c (reg_class_names): Likewise.
2157 * regrename.c (reg_class_names): Likewise.
2158 * reload.c (reg_class_names): Likewise.
2159 * reload1.c (spill_failure): Likewise.
2160 * config/m68hc11/m68hc11.c (reg_class_names): Likewise.
2162 2005-01-15 Ulrich Weigand <uweigand@de.ibm.com>
2164 * config/s390/s390.md ("reload_outti"): Remove predicate for
2165 output operand. Abort if operand is not a MEM.
2166 ("reload_outdi", "reload_outdf"): Likewise.
2168 2005-01-15 Marc Espie <espie@openbsd.org>
2170 * config.gcc (*-*-openbsd*): Set HAS_LIBC_R for OpenBSD <= 3.2.
2171 (i[34567]86-*-openbsd*): Switch to ELF for OpenBSD >= 3.4.
2172 * config/openbsd.h: Fix C++ includes for native configurations.
2173 Add proper OS_CPP_BUILTINS. Fix libspec for recent OpenBSD.
2174 Add trampoline support.
2175 * config/i386/openbsdelf.h: New.
2177 2005-01-15 Marc Espie <espie@openbsd.org>
2179 * collect2.c (main): Explicitly parse -dynamic-linker option.
2181 2005-01-15 Jakub Jelinek <jakub@redhat.com>
2183 PR tree-optimization/19060
2184 * tree-ssa-dom.c (extract_range_from_cond) <case LT_EXPR, GT_EXPR>:
2185 Return 0 if op1 <= TYPE_MIN_VALUE () resp. op1 >= TYPE_MAX_VALUE ().
2186 (simplify_cond_and_lookup_avail_expr): Add assert for dummy == 0
2187 and handle extract_range_from_cond returning false.
2188 * fold-const.c (fold): Optimize comparisons with min/max even for
2189 width > HOST_BITS_PER_WIDE_INT.
2191 2005-01-15 Ralf Corsepius <ralf.corsepius@rtems.org>
2193 * config/mips/rtems.h (MIPS_DEFAULT_GVALUE): Set to 0.
2194 * config/mips/t-rtems (MULTILIBS_DIRNAMES,MULTILIB_OPTIONS):
2195 Remove little endian multilib variants.
2196 Add mips32 multilib variant.
2198 2005-01-15 Kazu Hirata <kazu@cs.umass.edu>
2200 * explow.c (copy_all_reg, stabilize): Remove.
2201 * expr.h: Remove the corresponding prototypes.
2203 * rtlanal.c (regs_set_between_p): Remove.
2204 * rtl.h: Remove the corresponding prototype.
2206 2005-01-14 David Edelsohn <edelsohn@gnu.org>
2208 * config/rs6000/rs6000.c (rs6k_nonimmediate_operand): Rename
2209 to rs6000_nonimmediate_operand.
2210 * config/rs6000/rs6000.h (PREDICATE_CODES): Rename
2211 rs6k_nonimmediate_operand to rs6000_nonimmediate_operand.
2212 * config/rs6000/rs6000.md (movsi_internal1): Rename predicate to
2213 rs6000_nonimmediate_operand.
2214 * config/rs6000/spe.md (movdf_e500_double): Rename predicate to
2215 rs6000_nonimmediate_operand.
2217 2005-01-14 Andrew Pinski <pinskia@physics.uc.edu>
2219 * tree-ssa-alias.c (pass_may_alias): Add TODO_verify_stmts.
2221 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
2223 * config/rs6000/rs6000.h (CLASS_MAX_NREGS): DF goes in 1 register
2225 (CANNOT_CHANGE_MODE_CLASS): Restrict DI mode changes on e500v2.
2226 (PREDICATE_CODES): Add rs6k_nonimmediate_operand.
2228 * config/rs6000/rs6000.c (invalid_e500_subreg): New.
2229 (rs6k_nonimmediate_operand): New.
2230 (rs6000_legitimate_offset_address_p): Handle DI modes on e500v2
2232 (legitimate_lo_sum_address_p): Same.
2233 (rs6000_legitimize_address): Same.
2234 (rs6000_legitimize_reload_address): Same.
2235 (rs6000_legitimate_address): Same.
2236 (spe_build_register_parallel): Pass DF and DC modes in a DI
2239 * config/rs6000/rs6000.md ("*movsi_internal1"): Change predicate
2240 to rs6k_nonimmediate_operand.
2242 * config/rs6000/spe.md ("*frob_df_di"): New.
2243 ("*frob_di_df"): New.
2244 ("*frob_di_df_2"): New.
2245 ("*mov_sidf_e500_subreg0"): New.
2246 ("*mov_sidf_e500_subreg4"): New.
2247 ("*movdf_e500_double"): Change predicate to
2248 rs6k_nonimmediate_operand.
2250 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
2252 * postreload.c (move2add_note_store): Only call
2253 trunc_int_for_mode on scalar integers.
2255 2005-01-14 Steven Bosscher <stevenb@suse.de>
2257 * tree-ssa-dce.c (visited_control_parents): New sbitmap to
2258 replace BB_VISITED uses.
2259 (find_obviously_necessary_stmts): Don't clear BB_VISITED.
2260 (propagate_necessity): Check the bitmap instead of BB_VISITED.
2261 (tree_dce_done): Free visited_control_parents.
2262 (perform_tree_ssa_dce): Allocate and clear it.
2263 * tree-ssa-pre.c (compute_antic_aux): Make non-recursive.
2264 (compute_antic): Iterate from here using a DFS. Use an sbitmap
2265 instead of BB_VISITED.
2267 2005-01-14 Kazu Hirata <kazu@cs.umass.edu>
2269 * c-tree.h, coverage.h, langhooks-def.h, optabs.h, output.h,
2270 rtl.h, targhooks.h, tree-flow.h, tree-mudflap.h, tree.h:
2271 Remove unused prototypes.
2273 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
2275 * config/rs6000/rs6000.c (rs6000_stack_info): Fix caching of SPE
2276 64-bit register usage.
2277 (spe_func_has_64bit_regs_p): Add FIXME note.
2279 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
2281 * config/rs6000/rs6000.c (rs6000_generate_compare): Replace
2282 flag_finite_math_only with flag_unsafe_math_optimizations.
2284 2005-01-14 J. D. Johnston <jjohnst@us.ibm.com>
2286 * config/s390/tpf-unwind.h (s390_fallback_frame_state): Correct
2288 (__tpf_eh_return): Copy TPF private stack area from the module
2289 boundary stack frame. Add check for when module addresses are
2290 equal, but no stub address is found.
2292 2005-01-14 Richard Earnshaw <rearnsha@arm.com>
2295 * arm.h (struct machine_function): Add call_via field.
2296 (thumb_call_via_label): Declare.
2297 * arm.c (thumb_call_via_label): New variable.
2298 (thumb_call_reg_needed): New variable.
2299 (arm_output_function_epilogue): For Thumb code, output any per-function
2300 call-indirect trampolines.
2301 (thumb_call_via_reg): New function.
2302 (arm_file_end): New function.
2303 (TARGET_ASM_FILE_END): Call arm_file_end.
2304 (aof_file_end): Likewise.
2305 * arm-protos.h (thumb_call_via_reg): Declare.
2306 * arm.md (call_reg_thumb, call_value_reg_thumb): Call
2307 thumb_call_via_reg in normal case.
2309 2005-01-14 Jakub Jelinek <jakub@redhat.com>
2312 PR rtl-optimization/19348
2313 * recog.c (peephole2_optimize): Do global life update if some peephole
2314 decides it doesn't need at least one of its inputs and that change
2315 influences liveness at the start of the basic block.
2317 * basic-block.h (EXECUTE_IF_AND_COMPL_IN_REG_SET): Needs 2 REGSET
2318 arguments instead of 1.
2320 2005-01-14 Eric Botcazou <ebotcazou@libertysurf.fr>
2323 * varasm.c (initializer_constant_valid_p) <ADDR_EXPR>: Return
2324 zero for nested functions needing a static chain or functions
2325 with a non-constant address.
2327 2005-01-13 Roger Sayle <roger@eyesopen.com>
2329 * simplify-rtx.c (simplify_binary_operation) <AND>: Optimize
2330 (and (sign_extend X) C) into (zero_extend (and X C)).
2332 2005-01-13 David O'Brien <obrien@FreeBSD.org>
2334 * config/freebsd-spec.h: Make KSE pthread lib logic the default.
2336 2005-01-13 Richard Henderson <rth@redhat.com>
2341 * config/i386/i386.md (cmpdf, cmpsf, bunordered, bordered, buneq,
2342 bunge, bungt, bunle, bunlt, bltgt): Enable for TARGET_SSE_MATH,
2343 not just TARGET_SSE.
2344 (cmpfp_i_387): Rename from cmpfp_i. Move after sse patterns.
2345 (cmpfp_i_mixed): Rename from cmpfp_i_sse; use for TARGET_MIX_SSE_I387.
2346 (cmpfp_i_sse): Rename from cmpfp_i_sse_only; use for TARGET_SSE_MATH.
2347 (cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387): Similarly.
2348 (fp_jcc_1_mixed, fp_jcc_1_sse, fp_jcc_1_387): Similarly.
2349 (fp_jcc_2_mixed, fp_jcc_2_sse, fp_jcc_2_387): Similarly.
2350 (fp_jcc_3_387, fp_jcc_4_387, fp_jcc_5_387, fp_jcc_6_387,
2351 fp_jcc_7_387, fp_jcc_8_387): Rename from fp_jcc_N.
2352 (movdicc_c_rex64): Rename with '*'.
2353 (movsfcc, movdfcc): Add checks for 387 and sse math to condition.
2354 (movsfcc_1_sse_min, movsfcc_1_sse_max, movsfcc_1_sse): New.
2355 (movsfcc_1_387): Rename from movsfcc_1.
2356 (movdfcc_1_sse_min, movdfcc_1_sse_max, movdfcc_1_sse): New.
2357 (movdfcc_1, movdfcc_1_rex64): Add check for 387.
2358 (sminsf3, smaxsf3, smindf3, smaxdf3): New.
2359 (minsf3, minsf, minsf_nonieee, minsf_sse, mindf3, mindf,
2360 mindf_nonieee, mindf_sse, maxsf3, maxsf, maxsf_nonieee, maxsf_sse,
2361 maxdf3, maxdf, maxdf_nonieee, maxdf_sse, sse_movsfcc, sse_movsfcc_eq,
2362 sse_movdfcc, sse_movdfcc_eq, sse_movsfcc_const0_1,
2363 sse_movsfcc_const0_2, sse_movsfcc_const0_3, sse_movsfcc_const0_4,
2364 sse_movdfcc_const0_1, sse_movdfcc_const0_2, sse_movdfcc_const0_3,
2365 sse_movdfcc_const0_4): Remove.
2366 * config/i386/i386.c (ix86_expand_fp_movcc): For TARGET_SSE_MATH,
2367 recognize min/max early. Update for changed sse cmove patterns.
2368 (ix86_split_sse_movcc): New.
2369 * config/i386/i386-protos.h: Update.
2371 2005-01-13 Steven Bosscher <stevenb@suse.de>
2373 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
2374 (fix_stmt_v_may_defs): Likewise.
2376 2005-01-13 Richard Henderson <rth@redhat.com>
2378 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
2379 for converting the mask.
2381 2005-01-13 David Edelsohn <edelsohn@gnu.org>
2383 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
2386 2005-01-13 Richard Henderson <rth@redhat.com>
2388 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
2389 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
2390 IX86_BUILTIN_MOVDQ2Q): Remove.
2391 (IX86_BUILTIN_VEC_EXT_V4SI): New.
2392 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
2393 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
2395 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
2396 (sse2_pextrw): Fix immediate constraint.
2397 (sse2_loadq, sse2_loadq_rex64): Remove.
2398 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
2399 Use __builtin_ia32_vec_ext_<size>.
2400 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
2402 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
2404 * function.c (assign_parm_setup_block): Look inside original
2405 entry_parm when inspecting PARALLEL.
2407 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
2408 Joel Sherrill <joel@oarcorp.com>
2411 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
2412 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
2413 rtems_gxx_recursive_mutex_init.
2414 (__gthread_recursive_mutex_lock): New function.
2415 (__gthread_recursive_mutex_trylock): Likewise.
2416 (__gthread_recursive_mutex_unlock): Likewise.
2418 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
2420 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
2422 2005-01-13 Jan Beulich <jbeulich@novell.com>
2424 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
2425 (ashrti3_internal): Indicate output is early clobber. Generate result
2426 into output rather than first input. Use move for low word of output
2427 if shift count is exactly 64.
2428 (lshrti3_internal): Likewise.
2430 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
2434 * reload1.c (delete_output_reload): Don't delete an output reload
2435 if the pseudo lives longer than a single basic block. Adjust and
2436 improve wording of comment.
2438 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
2441 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
2442 of TYPE_SIZE where we mean the number of bytes.
2444 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
2446 * function.c (assign_parm_setup_block): Relax condition on
2447 multi-register optimization.
2449 2005-01-12 Nick Clifton <nickc@redhat.com>
2451 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
2452 an "in_delay_slot" attribute of "no" to prevent them being used in
2453 delay slots. This is forbidden because they might generate
2456 2005-01-12 Alan Modra <amodra@bigpond.net.au>
2459 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
2462 2005-01-12 Nick Clifton <nickc@redhat.com>
2464 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
2465 passed on to GAS) but do not define, as GAS no longer supports or
2466 needs the -m2000 option.
2468 2005-01-11 Roger Sayle <roger@eyesopen.com>
2470 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
2471 type as args of unprototyped builtins aren't automatically promoted.
2472 (fold_builtin_abs): Likewise.
2474 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2476 * Makefile.in: Set a `build-warn' variable.
2478 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
2480 PR tree-optimization/17949
2481 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
2482 (find_interesting_uses_address): Use it.
2484 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
2486 * regrename.c (kill_value): Handle subreg's that won't simplify.
2488 2005-01-11 Richard Henderson <rth@redhat.com>
2491 * config/i386/i386.h (enum ix86_builtins): Move ...
2492 * config/i386/i386.c: ... here.
2493 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
2494 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
2495 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
2496 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
2497 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
2498 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
2499 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
2500 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
2501 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
2502 IX86_BUILTIN_LOADDDUP): Remove.
2503 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
2504 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
2505 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
2506 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
2507 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
2508 (ix86_init_builtins): Make static.
2509 (ix86_init_mmx_sse_builtins): Update for changed builtins.
2510 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
2511 if all the modes match. Otherwise, fake it.
2512 (get_element_number, ix86_expand_vec_init_builtin,
2513 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
2514 (ix86_expand_builtin): Make static. Update for changed builtins.
2515 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
2516 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
2517 (ix86_expand_vector_init_duplicate): New.
2518 (ix86_expand_vector_init_low_nonzero): New.
2519 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
2520 Split out from ix86_expand_vector_init; handle integer modes.
2521 (ix86_expand_vector_init): Use them.
2522 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
2523 * config/i386/i386-protos.h: Update.
2524 * config/i386/predicates.md (reg_or_0_operand): New.
2525 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
2526 (movv2sf_internal): Likewise. And a splitter to match them all.
2527 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
2528 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
2529 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
2530 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
2531 (mmx_pinsrw): Fix operand ordering.
2532 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
2533 rather than sse_loadss expander.
2534 (movv2df splitter): Similarly.
2535 (sse_loadss, sse_loadlss): Remove.
2536 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
2537 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
2538 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
2539 (sse3_movddup): Rename with '*'.
2540 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
2541 (sse2_loadsd): Remove.
2542 (vec_dupv2df_sse3): Rename from sse3_loadddup.
2543 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
2544 (sse2_pinsrw): Fix argument ordering.
2545 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
2546 (sse2_stored): Remove 'r' destination.
2547 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
2548 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
2549 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
2550 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
2551 vec_extractv16qi, vec_initv16qi): New.
2553 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
2554 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
2555 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
2557 (_mm_load_pd, _mm_store_pd): Use plain dereference.
2558 (_mm_load_si128, _mm_store_si128): Likewise.
2559 (_mm_load1_pd): Use _mm_set1_pd.
2560 (_mm_load_sd): Use _mm_set_sd.
2561 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
2562 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
2563 (_mm_set_epi64): Use _mm_set_epi64x.
2564 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
2565 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
2566 _mm_setr_epi8): Use _mm_set_foo form.
2567 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
2568 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
2569 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
2570 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
2571 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
2572 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
2573 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
2574 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
2575 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
2576 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
2577 (_mm_movedup_pd): Use _mm_shuffle_pd.
2578 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
2579 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
2580 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
2581 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
2583 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
2584 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
2585 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
2586 (_mm_store_ps): Use raw dereference.
2587 (_mm_store1_ps): Use _mm_storeu_ps.
2588 (_mm_storer_ps): Use _mm_store_ps.
2589 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
2590 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
2592 2005-01-11 Stan Shebs <shebs@apple.com>
2594 * config/rs6000/rs6000.c (machopic_output_stub): Issue
2595 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
2597 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
2600 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
2601 Skip all DECLs except for FIELD_DECLs.
2603 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
2605 * config/s390/s390.c (override_options): Return error if
2606 -mbackchain, -mpacked-stack and -mhard-float are used together.
2607 (s390_va_start): Remove the backchain && packed-stack special case.
2608 (s390_gimplify_va_arg): Likewise.
2609 * doc/invoke.texi: Remove the ABI incompatibility note.
2611 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
2613 * config/s390/s390.c (struct s390_frame_layout): Remove
2615 (s390_frame_info, s390_emit_prologue): Replace occurrences of
2616 save_backchain_p with TARGET_BACKCHAIN.
2618 2005-01-11 Alan Modra <amodra@bigpond.net.au>
2621 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
2622 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
2623 * function.c (locate_and_pad_parm): Set new field.
2624 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
2625 Tweak where_pad test to include "none". Always set mem align for
2627 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
2628 not sufficient for type.
2629 (assign_parm_setup_block): If stack_parm is zero on entry, always
2630 make a new stack local. Block move old stack parm if necessary
2631 to new aligned stack local.
2632 (assign_parm_setup_stack): Use a block move to handle
2633 potentially misaligned entry_parm.
2634 (assign_parms_unsplit_complex): Specify required alignment when
2635 creating stack local.
2636 * calls.c (compute_argument_addresses): Override alignment of stack
2637 arg calculated from its type with the alignment given by
2638 FUNCTION_ARG_BOUNDARY.
2639 (store_one_arg): Likewise.
2641 2005-01-11 Jan Beulich <jbeulich@novell.com>
2643 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
2644 Change respective itanium_class attribute to ialu.
2645 (shladdp4_internal): New.
2646 * config/ia64/predicates.md (shladd_log2_operand): New.
2648 2005-01-11 Richard Henderson <rth@redhat.com>
2650 * expr.c (store_constructor): Use rtvec_alloc instead of
2651 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
2653 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
2655 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
2657 2005-01-11 Alan Modra <amodra@bigpond.net.au>
2659 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
2660 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
2661 (TARGET_RWRELOC): Define.
2662 (ia64_rwreloc_section_type_flags): Delete.
2663 (ia64_section_type_flags): New function.
2664 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
2665 (TARGET_RWRELOC): Define.
2667 2005-01-10 David Mosberger <davidm@hpl.hp.com>
2670 * config/ia64/ia64.c (process_set): For alloc insn, only call
2671 process_epilogue is !frame_pointer_needed.
2673 2005-01-10 Roger Sayle <roger@eyesopen.com>
2676 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
2677 unary operator and can't be treated as a binary/comparison operator.
2679 2005-01-10 Richard Henderson <rth@redhat.com>
2681 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
2682 (ix86_return_in_memory): Likewise.
2683 (function_arg_advance): Likewise. Mirror structure in function_arg
2684 for choosing register to advance.
2686 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
2688 * tree-vectorizer.c, tree.def: Fix comment typos.
2690 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
2692 * config/s390/s390.c (struct processor_costs): 4 new fields:
2693 ddbr, ddr, debr, der.
2694 (s390_rtx_costs): More precise handling of divide instructions.
2696 2005-01-10 David Edelsohn <edelsohn@gnu.org>
2698 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
2700 2005-01-10 Jan Beulich <jbeulich@novell.com>
2702 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
2703 names resulting from -ffunction-sections/-fdata-sections and linkonce
2705 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
2706 based on the section name. Rearrange the section name comparison logic
2707 slightly so that each section name is compared against at most once.
2709 2005-01-10 Ben Elliston <bje@au.ibm.com>
2711 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
2712 targets for which -fPIC reduces limitations on the GOT size.
2714 2005-01-09 Falk Hueffner <falk@debian.org>
2716 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
2719 2005-01-09 David Edelsohn <edelsohn@gnu.org>
2722 * collect2.c (main): Set aixrtl_flag for -brtl option.
2723 (resolve_lib_name): Search for .so file extension before .a
2726 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
2728 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
2729 (vect_analyze_data_refs_alignment): Add dump prints.
2731 2005-01-09 Ira Rosen <irar@il.ibm.com>
2733 * tree-vectorizer.c (vect_analyze_offset_expr): Use
2734 expr_invariant_in_loop_p.
2735 Initialize outputs first thing in the function.
2736 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
2737 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
2738 (vect_analyze_pointer_ref_access): Check that the initial condition of
2739 the access function is loop invariant.
2741 2005-01-09 Richard Henderson <rth@redhat.com>
2743 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
2744 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
2745 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
2746 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
2747 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
2748 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
2749 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
2750 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
2751 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
2752 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
2753 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
2754 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
2755 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
2756 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
2757 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
2758 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
2759 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
2760 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
2761 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
2762 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
2763 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
2764 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
2765 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
2767 (mmx_clrdi, pavgusb): Remove.
2768 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
2769 with leading sse_ prefix.
2770 * config/i386/sse.md: Receive them.
2771 * config/i386/mmx.md: New file.
2772 (MMXMODE12, MMXMODE24, mmxvecsize): New.
2773 (subrv2sf3): Turn into expander for normal subtraction.
2774 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
2775 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
2776 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
2777 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
2778 commutative; use ix86_binary_operator_ok.
2779 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
2780 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
2781 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
2782 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
2783 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
2784 Macroize from existing patterns; use ix86_binary_operator_ok.
2785 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
2786 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
2787 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
2788 with vec_select+vec_concat.
2789 (mmx_pshufw, mmx_pshufw_1): Likewise.
2790 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
2791 (mmx_uavgv4hi3): Model correctly.
2792 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
2793 _mm_xor_si64): Remove casts.
2795 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
2797 PR tree-optimization/19224
2798 * tree-scalar-evolution.c (get_instantiated_value,
2799 set_instantiated_value): New functions.
2800 (instantiate_parameters_1): Cache the results.
2801 (instantiate_parameters, resolve_mixers): Initialize and free
2804 2005-01-08 David Edelsohn <edelsohn@gnu.org>
2806 * config/i386/i386.md (addhi_4): Correct reference in comment.
2809 2005-01-08 Richard Henderson <rth@redhat.com>
2811 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
2812 (_mm_cvtsi128_si64x): Likewise.
2813 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
2814 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
2815 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
2816 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
2817 * config/i386/i386-protos.h: Update.
2818 * config/i386/i386.c (print_operand): Add 'H'.
2819 (ix86_fixup_binary_operands): Split out from ...
2820 (ix86_expand_binary_operator): ... here.
2821 (ix86_fixup_binary_operands_no_copy): New.
2822 (ix86_expand_fp_absneg_operator): Handle vector mode results.
2823 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
2824 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
2825 (safe_vector_operand): Use CONST0_RTX.
2826 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
2827 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
2828 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
2829 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
2831 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
2832 (IX86_BUILTIN_CMPNGESS): New.
2833 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
2834 (attr type): Add sselog1.
2835 (attr unit, attr memory): Handle it.
2836 (movti, movti_internal, movti_rex64): Move near other integer moves.
2837 (movtf, movtf_internal): Move near other fp moves.
2838 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
2839 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
2840 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
2841 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
2842 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
2843 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
2844 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
2845 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
2846 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
2847 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
2848 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
2849 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
2850 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
2851 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
2852 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
2853 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
2854 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
2855 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
2856 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
2857 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
2858 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
2859 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
2860 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
2861 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
2862 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
2863 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
2864 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
2865 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
2866 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
2867 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
2868 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
2869 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
2870 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
2871 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
2872 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
2873 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
2874 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
2875 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
2876 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
2877 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
2878 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
2879 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
2880 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
2881 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
2883 (SSEPUSH, push<SSEPUSH>): Remove.
2884 (MMXPUSH, push<MMXPUSH>): Remove.
2885 (sse_movaps, sse_movaps_1, sse_movups): Remove.
2886 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
2887 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
2888 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
2889 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
2890 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
2891 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
2892 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
2893 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
2894 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
2895 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
2896 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
2897 (ppro_sse_log_V4SF): Handle sselog1.
2898 * config/i386/predicates.md (const_0_to_1_operand): New.
2899 (const_0_to_255_mul_8_operand): New.
2900 (const_1_to_31_operand): Rename from const_int_1_31_operand.
2901 (const_2_to_3_operand, const_4_to_7_operand): New.
2902 * config/i386/sse.md: New file.
2903 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
2904 (sse_movups): Rename from sse_movups_1.
2905 (sse_loadlss): Rename from sse_loadss_1.
2906 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
2908 (negv4sf2): Use ix86_expand_fp_absneg_operator.
2909 (absv4sf2, negv2df, absv2df): New.
2910 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
2911 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
2912 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
2913 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
2914 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
2915 (sse3_addsubv4sf3): Model correctly.
2916 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
2917 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
2918 (sse_movhlps): Model with vec_select+vec_concat.
2919 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
2920 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
2921 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
2922 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
2923 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
2924 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
2925 sse2_pshufhw_1): Likewise.
2926 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
2927 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
2928 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
2929 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
2930 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
2931 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
2932 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
2933 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
2934 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
2935 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
2936 sminv8hi3): Mark commutative
2937 operands. Use ix86_binary_operator_ok.
2938 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
2939 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
2940 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
2941 sse2_punpcklqdq): Allow operand2 in memory.
2942 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
2943 sse2_movsd): Add memory alternatives.
2944 (sse_storelps): Turn expander into an insn; split after reload.
2945 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
2946 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
2948 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
2950 * configure.ac (DWARF-2 debug_line): Use objdump.
2951 * configure: Regenerate.
2953 2005-01-08 Jeff Law <law@redhat.com>
2954 Diego Novillo <dnovillo@redhat.com>
2956 PR tree-optimization/18241
2957 * tree-nrv.c (tree_nrv): Ignore volatile return values.
2958 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
2959 statements with volatile operands.
2960 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
2961 operands after marking a statement with has_volatile_ops.
2963 2005-01-08 Roger Sayle <roger@eyesopen.com>
2965 * tree.c (int_fits_type_p): Always honor integer constant
2966 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
2968 2005-01-08 Roger Sayle <roger@eyesopen.com>
2970 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
2971 the latent bug in rtl_delete_block has been resolved.
2973 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
2975 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
2976 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
2978 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
2980 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
2981 library as @multilib_dir@/$(SHLIB_SONAME).
2983 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
2985 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
2986 * configure: Regenerate.
2988 2005-01-07 Jakub Jelinek <jakub@redhat.com>
2990 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
2993 2005-01-07 David Edelsohn <edelsohn@gnu.org>
2996 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
2997 Convert non-word aligned offset address using ld/std into
3000 2005-01-07 Richard Henderson <rth@redhat.com>
3002 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
3004 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
3006 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
3008 2005-01-07 Jakub Jelinek <jakub@redhat.com>
3010 PR tree-optimization/19283
3011 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
3012 is not shorter than the original type.
3014 PR rtl-optimization/19012
3015 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
3017 PR rtl-optimization/18861
3018 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
3019 (FULL_STATE): Define.
3020 (mark_tablejump_edge): New function.
3021 (purge_dead_tablejump_edges): New function.
3022 (find_bb_boundaries): Use it.
3024 PR tree-optimization/18828
3025 * builtins.c (expand_builtin_next_arg): Remove argument and all
3026 the argument checking.
3027 (expand_builtin): Adjust caller.
3028 (expand_builtin_va_start): Likewise. Remove error for too many
3030 (fold_builtin_next_arg): Issue error for too many arguments.
3031 After checking arguments, replace them with magic arguments that
3032 prevent further checking of the args.
3034 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
3036 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
3038 2005-01-06 Roger Sayle <roger@eyesopen.com>
3041 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
3042 and %{.S:...} (and their negative variants) to test whether the
3043 input file is assembler or pre-processed-assembler independent of
3044 the actual filename extension.
3046 2005-01-06 Roger Sayle <roger@eyesopen.com>
3048 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
3049 of sign or zero extended values.
3051 2005-01-06 Geoffrey Keating <geoffk@apple.com>
3053 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
3054 for IBM long double format correctly.
3056 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
3058 Fix PR tree-optimization/18792
3060 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
3061 to first_loop_depth, and use loop depth instead of loop number.
3062 (build_classic_dir_vector): Ditto.
3063 (compute_data_dependences_for_loop): Use depth, not loop number.
3064 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
3065 number. Pass in loops, instead of loop numbers.
3066 (gather_interchange_stats): Ditto.
3067 (linear_transform_loops): Ditto.
3069 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
3072 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
3073 biv_fits_mode_p, extension_within_bounds_p): New functions.
3074 (check_ext_dependent_givs): Use them.
3076 2005-01-06 Roger Sayle <roger@eyesopen.com>
3078 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
3079 more than one barrier, in which case we should delete them all.
3081 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
3083 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
3085 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
3087 Fix PR middle-end/19286
3089 * dwarf2out.c (gen_subprogram_die): If we've already tried to
3090 output this subprogram, simply ignore this attempt to do it again.
3091 (add_abstract_origin_attribute): Don't abort trying to add the abstract
3092 origin attribute if it's not possible.
3093 (gen_block_die): Don't ignore subblocks of "unused" blocks.
3094 (decls_for_scope): Ditto.
3095 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
3096 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
3098 2005-01-05 Richard Henderson <rth@redhat.com>
3101 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
3102 (bdesc_2arg): Use it.
3103 (ix86_expand_binop_builtin): Force operands into registers
3105 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
3106 ix86_expand_sse_compare, ix86_expand_sse_comi,
3107 ix86_expand_builtin): Likewise.
3109 2005-01-05 Richard Henderson <rth@redhat.com>
3111 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
3113 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
3114 (mulv4hi3): Set itanium_class mmmul.
3115 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
3116 (fpack): Rename from fpack_sfsf.
3118 2005-01-05 Richard Henderson <rth@redhat.com>
3121 * reload1.c (do_input_reload): Restrict the optimization deleteing
3122 a previous output reload to RELOAD_FOR_INPUT.
3124 2005-01-05 Steven Bosscher <stevenb@suse.de>
3126 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
3127 comment that we fall through after case.
3128 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
3129 (distribute_links): Likewise.
3130 * cse.c (cse_insn): Likewise.
3131 * cselib.c (cselib_invalidate_mem): Likewise.
3132 * df.c: Update comments at the top of the file.
3133 (read_modify_subreg_p): Update comments here too.
3134 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
3135 * flow.c (mark_set_1): Likewise.
3136 (mark_used_regs): Likewise.
3137 * gcse.c (mems_conflict_for_gcse_p): Likewise.
3138 (canon_list_insert): Likewise.
3139 (mark_set): Likewise.
3140 (try_replace_reg): Likewise.
3141 (store_killed_in_insn): Likewise.
3142 * loop.c (count_one_set): Likewise.
3143 (basic_induction_var): Likewise.
3144 * postreload-gcse.c (find_mem_conflicts): Likewise.
3145 * postreload.c (reload_combine_note_store): Likewise.
3146 (move2add_note_store): Likewise.
3147 * reload.c (find_equiv_reg): Likewise.
3148 (mark_referenced_resources): Likewise.
3149 * rtlanal.c (set_noop_p): Likewise.
3150 (note_stores): Likewise.
3151 (note_uses): Likewise.
3152 * sched-deps.c (sched_analyze_1): Likewise.
3153 * sched-rgn.c (check_live_1): Likewise.
3154 (update_live_1): Likewise.
3155 * config/i860/i860.c: Likewise.
3157 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
3159 (ZERO_EXTRACT): Mention that this one can be an lvalue.
3161 * doc/rtl.texi: Update documentation for bit-fields and SET.
3163 2005-01-05 Roger Sayle <roger@eyesopen.com>
3165 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
3167 2005-01-05 Jan Hubicka <jh@suse.cz>
3168 Richard Henderson <rth@redhat.com>
3171 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
3174 2005-01-05 Richard Henderson <rth@redhat.com>
3177 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
3178 (sse_shufps): Change operand 3 to const_int_operand.
3179 (sse2_storelps): Fix typo in template.
3180 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
3181 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
3182 (ix86_expand_builtin): Likewise.
3184 2005-01-05 Julian Brown <julian@codesourcery.com>
3186 * config/arm/arm.c (arm_return_in_memory): Treat complex types
3187 as aggregates for AAPCS ABIs.
3189 2005-01-05 Stan Shebs <shebs@apple.com>
3191 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
3192 Mach-O getters if ppc64.
3193 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
3195 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
3196 signed four-byte field for global code case.
3197 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
3199 2005-01-05 Roger Sayle <roger@eyesopen.com>
3202 * c-common.c: Include real.h.
3203 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
3204 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
3205 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
3206 * Makefile.in (c-common.o): Update dependencies.
3208 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
3210 * c-parse.in (asm_string): Add trailing semicolon.
3212 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
3214 * c-parse.in (asm_string): New. Don't allow wide strings in
3216 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
3217 asm_string instead of STRING.
3219 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
3221 * c-typeck.c (constructor_no_implicit): Remove.
3222 (set_designator, process_init_element): Don't check
3223 constructor_no_implicit.
3225 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
3226 Kaz Kojima <kkojima@gcc.gnu.org>
3229 * lcm.c (create_pre_exit): New.
3230 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
3231 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
3233 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
3235 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
3236 -msoft-float in target_float_switch.
3237 * arm.c (arm_override_options): Fix processing of target_float_switch.
3239 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
3241 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
3244 2004-01-05 Uros Bizjak <uros@kss-loka.si>
3246 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
3247 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
3249 2005-01-04 Roger Sayle <roger@eyesopen.com>
3251 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
3252 if deleting the then-block allows the test-block to fallthru to the
3255 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
3258 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
3259 after the full declaration if the are in two different TUs.
3261 2005-01-04 Richard Henderson <rth@redhat.com>
3264 * tree-sra.c (generate_one_element_init): Just
3265 call gimplify_and_add.
3266 (generate_element_init): Record the
3267 new referenced variables and mark them for renaming
3268 and split out to ...
3269 (generate_element_init_1): This.
3270 (scalarize_init): Don't call push_gimplify_context/
3271 pop_gimplify_context.
3273 2005-01-04 Geoffrey Keating <geoffk@apple.com>
3275 * toplev.c (get_src_pwd): Handle failure of getpwd().
3277 2005-01-04 Roger Sayle <roger@eyesopen.com>
3279 * fold-const.c (fold_single_bit_test): Delete unreachable handling
3281 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
3282 all the cases handled by it are inverted by invert_truthvalue.
3284 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
3286 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
3287 (save_register_windowsi): Likewise.
3289 2005-01-04 Richard Henderson <rth@redhat.com>
3291 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
3293 2005-01-04 Uros Bizjak <uros@kss-loka.si>
3296 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
3297 any simplifications of vector mode comparison operators.
3298 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
3300 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
3301 Devang Patel <dpatel@apple.com>
3303 PR tree-optimization/18308
3304 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
3305 the operands before creating a new expression.
3306 * dojump.c (do_jump): Make drop_through_label available
3307 for all cases. Add expansion of COND_EXPR.
3309 2005-01-04 Ira Rosen <irar@il.ibm.com>
3311 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
3312 INTEGER_CST instead of TREE_CONSTANT.
3313 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
3314 instead of TREE_CONSTANT.
3315 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
3316 instead of TREE_CONSTANT.
3318 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
3320 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
3322 2005-01-04 Uros Bizjak <uros@kss-loka.si>
3325 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
3326 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
3328 2005-01-03 Richard Henderson <rth@redhat.com>
3330 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
3331 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
3332 fold_convert_const_real_from_real): Split out from ...
3333 (fold_convert_const): ... here.
3335 2005-01-03 Richard Henderson <rth@redhat.com>
3338 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
3339 (mov<MMXMODEI>_internal): Likewise.
3340 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
3341 Add V2SF case; use it for SSE1; don't use TI.
3342 (movdf_integer): Likewise.
3343 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
3345 2005-01-03 Ira Rosen <irar@il.ibm.com>
3347 * tree-vectorizer.c (vect_strip_conversions): New function.
3348 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
3349 check for binary class.
3351 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
3355 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
3357 * gimple-low.c (mark_blocks_with_used_vars): New function.
3358 (mark_blocks_with_used_subblocks): Ditto.
3359 (mark_used_blocks): Ditto.
3360 (pass_mark_used_blocks): New pass.
3361 * tree-inline.c: Include debug.h.
3362 (expand_call_inline): Call outlining_inline_function here.
3363 * tree-optimize.c (init_tree_optimization_passes): Add
3364 pass_mark_used_blocks.
3365 * tree-pass.h (pass_mark_used_blocks): New.
3366 * Makefile.in (tree-inline.o): Add debug.h dependency.
3368 2005-01-03 Geoffrey Keating <geoffk@apple.com>
3370 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
3371 VAR_DECLs to have weak_import attribute.
3373 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
3376 2004-01-03 Steven Bosscher <stevenb@suse.de>
3378 * passes.c (rest_of_compilation): Don't run regmove if only
3379 flag_expensive_optimizations, require flag_regmove instead.
3381 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
3383 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
3385 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
3386 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
3387 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
3388 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
3390 2005-01-03 Richard Henderson <rth@redhat.com>
3392 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
3393 (ia64_const_ok_for_letter_p): New.
3394 (ia64_const_double_ok_for_letter_p): New.
3395 (ia64_extra_constraint): New.
3396 (ia64_expand_vecint_compare): New.
3397 (ia64_expand_vcondu_v2si): New.
3398 (ia64_expand_vecint_cmov): New.
3399 (ia64_expand_vecint_minmax): New.
3400 (ia64_print_operand): Add 'v'.
3401 (ia64_preferred_reload_class): New.
3402 (ia64_vector_mode_supported_p): New.
3403 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
3404 (PREFERRED_RELOAD_CLASS): Move to function.
3405 (CONST_OK_FOR_LETTER_P): Move to function.
3406 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
3407 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
3408 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
3409 (EXTRA_CONSTRAINT): Move to function.
3410 * config/ia64/ia64.md: Include vect.md.
3411 (itanium_class): Add mmalua.
3413 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
3415 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
3416 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
3418 (const_int_2bit_operand): New.
3419 (fr_reg_or_0_operand): New.
3420 * config/ia64/ia64-modes.def: Add vector modes.
3421 * config/ia64/ia64-protos.h: Update.
3422 * config/ia64/vect.md: New file.
3424 2005-01-03 Richard Henderson <rth@redhat.com>
3426 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
3428 2005-01-03 Uros Bizjak <uros@kss-loka.si>
3431 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
3433 (log1pdf2): Change mode of operands[0,1] to DFmode.
3435 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
3437 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
3438 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
3439 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
3440 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
3441 SPARC_RELAXED_ORDERING.
3443 2005-01-03 Richard Henderson <rth@redhat.com>
3444 Uros Bizjak <uros@kss-loka.si>
3447 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
3448 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
3449 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
3451 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
3452 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
3453 const_pow2_1_to_128_operand): New predicates.
3455 2005-01-02 Greg McGary <greg@mcgary.org>
3457 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
3459 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
3460 object. [ptr_type]: Remove unused variable.
3461 Remember nearest addressable array-element or record-component
3462 when checking bitfield components. Tolerate empty BIND_EXPR.
3464 2005-01-01 Richard Henderson <rth@redhat.com>
3466 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
3467 that don't narrow the value. Fail for other conversions.
3469 2005-01-01 Richard Henderson <rth@redhat.com>
3472 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
3473 * c-lang.c (finish_file): Don't do it here.
3474 * objc/objc-act.c (objc_finish_file): Likewise.
3476 * cgraph.c (decl_assembler_name_equal): New.
3477 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
3478 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
3479 * cgraph.h (struct cgraph_varpool_node): Add next.
3480 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
3481 * varasm.c (assemble_alias): Mark the target as needed.
3483 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
3486 * function.c (get_arg_pointer_save_area): Use entry_of_function
3487 instead of get_insns.
3489 2005-01-01 Roger Sayle <roger@eyesopen.com>
3490 Andrew Pinski <pinskia@physics.uc.edu>
3491 James E. Wilson <wilson@specifixinc.com>
3493 PR rtl-optimization/12092
3494 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
3497 2005-01-01 Roger Sayle <roger@eyesopen.com>
3498 Olivier Hainque <hainque@act-europe.fr>
3500 * tree.c (int_fits_type_p): A narrower type always fits in a
3501 wider one, except for negative values into unsigned types.
3503 2005-01-01 Roger Sayle <roger@eyesopen.com>
3505 * tree.c (int_fits_type_p): Compare the result of force_fit_type
3506 with the original constant rather than require TREE_OVERFLOW.
3508 2005-01-01 Steven Bosscher <stevenb@suse.de>
3511 * c-decl.c (finish_function): If compiling C99, annotate the
3512 compiler generated return with the current file name and line 0.
3513 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
3514 the source line is greater than 0.
3515 (remove_bb): Likewise.
3517 See ChangeLog.12 for earlier changes.