1 2005-02-07 Eric Botcazou <ebotcazou@libertysurf.fr>
4 * dbxout.c (dbxout_symbol) <FUNCTION_DECL>: Skip inline instance
7 2005-02-07 Jonathan Wakely <redi@gcc.gnu.org>
9 * doc/extend.texi: Document deprecated extensions allowing
10 static floating-point members to have initializers and allowing
11 floating-point literals in integral constant expressions.
13 2005-02-07 Kazu Hirata <kazu@cs.umass.edu>
15 * cfgcleanup.c, df.h, diagnostic.c, rtl.c, tree-vectorizer.h,
16 config/darwin7.h, config/mips/mips.c,
17 config/rs6000/altivec.md, config/rs6000/darwin-tramp.asm:
20 * cse.c: Update comments.
22 2005-02-07 Richard Guenther <rguenth@gcc.gnu.org>
25 * builtins.c (fold_builtin_sqrt): Transform
26 sqrt(pow(x,y)) to pow(fabs(x),y*0.5), not
29 2005-02-07 Leehod Baruch <leehod@il.ibm.com>
30 Dorit Naishlos <dorit@il.ibm.com>
32 * doc/invoke.texi (ftree-vectorizer-verbose): New.
33 * opts.c (OPT_ftree_vectorizer_verbose_): New case for switch.
34 * common.opt (ftree-vectorizer-verbose): New Flag for the vectorizer
36 * tree.h (vect_set_verbosity_level): New extern function declaration
38 * tree-vectorizer.h (verbosity_levels): New enum type.
39 * tree-vectorizer.c (vect_debug_stats): Function removed.
40 (vect_debug_details): Likewise.
41 (vect_verbosity_level): Global variable was defined and initialized.
42 (vect_dump): Global variable definition.
43 (vect_print_dump_info): New function.
44 (vect_set_dump_settings): New function.
45 (vect_set_verbosity_level): New function.
46 (vectorize_loops): Add call to vect_set_dump_settings.
48 (slpeel_make_loop_iterate_ntimes): Dump condition was changed.
49 (slpeel_tree_duplicate_loop_to_edge_cfg): Likewise.
50 (slpeel_tree_peel_loop_to_edge): Likewise.
52 (vect_analyze_offset_expr): Call to vect_print_dump_info with
53 appropriate verbosity level instead of call to vect_debug_details
55 (vect_get_base_and_offset):
56 (vect_create_addr_base_for_vector_ref):
57 (get_vectype_for_scalar_type):
58 (vect_create_data_ref_ptr):
60 (vect_get_vec_def_for_operand):
61 (vect_finish_stmt_generation):
62 (vectorizable_assignment):
63 (vectorizable_operation):
66 (vect_transform_stmt):
67 (vect_update_ivs_after_vectorizer):
68 (vect_do_peeling_for_loop_bound):
69 (vect_update_inits_of_drs):
70 (vect_do_peeling_for_alignment):
71 (vect_transform_loop):
73 (vect_analyze_operations):
74 (vect_is_simple_iv_evolution):
75 (vect_analyze_scalar_cycles):
76 (vect_analyze_data_ref_dependence):
77 (vect_analyze_data_ref_dependences):
78 (vect_compute_data_ref_alignment):
79 (vect_enhance_data_refs_alignment):
80 (vect_analyze_data_refs_alignment):
81 (vect_analyze_data_ref_access):
82 (vect_analyze_data_ref_accesses):
83 (vect_analyze_pointer_ref_access):
84 (vect_get_memtag_and_dr):
85 (vect_analyze_data_refs):
87 (vect_stmt_relevant_p):
88 (vect_mark_stmts_to_be_vectorized):
89 (vect_can_advance_ivs_p):
90 (vect_get_loop_niters):
91 (vect_analyze_loop_form):
93 (vectorize_loops): Likewise.
95 (vect_do_peeling_for_loop_bound): Dump format slightly changed.
96 (vect_update_inits_of_drs):
97 (vect_do_peeling_for_alignment):
98 (vect_transform_loop):
99 (vect_analyze_operations):
100 (vect_analyze_scalar_cycles):
101 (vect_analyze_data_ref_dependences):
102 (vect_analyze_data_refs_alignment):
103 (vect_analyze_data_ref_accesses):
104 (vect_analyze_data_refs):
105 (vect_mark_stmts_to_be_vectorized):
106 (vect_get_loop_niters):
107 (vect_analyze_loop_form):
108 (vect_analyze_loop): Likewise.
109 (vect_mark_stmts_to_be_vectorized): Add call to print_generic_expr.
111 2005-02-07 Richard Sandiford <rsandifo@redhat.com>
115 2005-01-26 Richard Sandiford <rsandifo@redhat.com>
116 * config/mips/iris6.h (ENDFILE_SPEC): Don't link in irix-csr.o if
117 either -ffast-math or -funsafe-math-optimizations is in use.
119 2005-01-24 Richard Sandiford <rsandifo@redhat.com>
120 * config/mips/irix-csr.c: New file.
121 * config/mips/t-iris6 (irix-csr.o): New rule to build it.
122 (EXTRA_MULTILIB_PARTS): Add irix-csr.o.
123 * config/mips/iris6.h (ENDFILE_SPEC): Include it in n32 and n64
126 2005-02-07 Alexandre Oliva <aoliva@redhat.com>
128 * config/mn10300/linux.h (LIB_SPEC, STARTFILE_SPEC): Remove
129 redundant, incorrect definitions.
131 * config/frv/frv.md ("tls_indirect_call"): Turn into libcall-like
133 ("*tls_indirect_call"): New, reference hard regs directly.
134 * config/frv/frv.c (gen_inlined_tls_plt): Adjust.
136 2005-02-07 Ralf Corsepius <ralf.corsepius@rtems.org>
138 * config/arm/rtems-elf.h (SUBTARGET_EXTRA_ASM_SPEC):
139 Let !-mhard-float !-msoft-float imply -mfpu=softfpa.
141 2005-02-06 Roger Sayle <roger@eyesopen.com>
143 PR rtl-optimization/19800
144 * simplify_rtx.c (simplify_relational_operation_1): Explicitly
145 call gen_lowpart_common and gen_lowpart_SUBREG instead of calling
148 2005-02-07 Ralf Corsepius <ralf.corsepius@rtems.org>
150 * config/m68k/t-rtems (MULTILIB_MATCHES): Let m528x match m5200.
152 2005-02-07 Joseph S. Myers <joseph@codesourcery.com>
155 * doc/invoke.texi: Correct documentation of -freorder-functions.
157 2005-02-06 Joseph S. Myers <joseph@codesourcery.com>
160 * doc/invoke.texi: Don't mention gccbug.
162 2005-02-06 Roger Sayle <roger@eyesopen.com>
163 Andrew Pinski <pinskia@physics.uc.edu>
164 Paolo Bonzini <paolo.bonzini@lu.unisi.ch>
166 * simplify-rtx.c (simplify_relational_operation_1): Simplify
167 (ne:SI (zero_extract:SI FOO (const_int 1) BAR) (const_int 0))
168 into just (zero_extract:SI FOO (const_int 1) BAR).
170 2005-02-06 Joseph S. Myers <joseph@codesourcery.com>
172 * doc/cpp.texi: Include gcc-common.texi. Don't define gcctabopt
173 macro locally. Don't give last revision date. Use GCC version
174 number from gcc-common.texi.
175 * doc/cppinternals.texi: Include gcc-common.texi. Don't give last
176 revision date. Use GCC version number from gcc-common.texi.
177 Describe being a library as current state rather than as 3.x
179 * doc/gcc.texi: Remove last update date.
180 * doc/gccint.texi: Likewise. Update copyright dates.
181 * doc/install.texi: Update copyright dates.
183 2005-02-06 Joseph S. Myers <joseph@codesourcery.com>
186 * doc/invoke.texi: Mention for each option included in -Wall that
187 it is included in -Wall.
188 * doc/gcc.texi: Update copyright and last modification date.
190 2005-02-06 Zdenek Dvorak <dvorakz@suse.cz>
192 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Do not add
193 unnecessary cast to original induction variable increments.
195 2005-02-06 Zdenek Dvorak <dvorakz@suse.cz>
197 PR tree-optimization/18219
198 * tree-ssa-loop-ivopts.c (get_computation_at): Produce computations
201 2005-02-06 Richard Sandiford <rsandifo@redhat.com>
203 * expmed.c (store_bit_field): Make the SUBREG code adjust bitnum.
204 Set bitpos and offset later in the function. Do nothing if the
205 target is a register and if the bitfield lies completely outside
207 (extract_bit_field): Make the same SUBREG, bitpos and offset changes
208 here. Return an uninitialised register if the source value is stored
209 in a register and the bitfield lies completely outside that register.
211 2005-02-06 Steven Bosscher <stevenb@suse.de>
213 * df.c (df_insn_refs_record): Use XEXP to get the operand of a USE,
216 2005-02-06 Marcin Dalecki <martin@dalecki.de>
218 * diagnostic.c (default_diagnostic_finalizer): Fix use of attribute.
220 2005-02-06 Geoffrey Keating <geoffk@apple.com>
222 * config.gcc (*-*-darwin*): Separate darwin7 and darwin8 configs.
223 * config/darwin8.h: New.
224 * config/darwin7.h: Update comment.
226 2005-02-03 Leehod Baruch <leehod@il.ibm.com>
227 Dorit Naishlos <dorit@il.ibm.com>
229 * tree-vectorizer.h (LOC): New type.
230 (UNKNOWN_LOC, EXPR_LOC, LOC_FILE, LOC_LINE): New macros.
231 (loop_line_number): New field in struct _loop_vec_info.
232 (LOOP_VINFO_LOC, LOOP_LOC): New macros.
234 * tree-vectorizer.c (input.h): Included.
235 (find_loop_location): New function.
236 (vect_debug_stats): Argument changed from loop to LOC. Computation of
237 loop line number removed.
238 (vect_debug_details): Likewise.
239 (new_loop_vec_info): Initialize new field LOOP_VINFO_LOC.
241 (slpeel_make_loop_iterate_ntimes): Call find_loop_location. Argument in
242 call to vect_debug_details/stats changed from loop to loop_loc.
243 (slpeel_tree_duplicate_loop_to_edge_cfg): Likewise.
244 (slpeel_tree_peel_loop_to_edge): Likewise.
246 (vect_analyze_offset_expr): Argument in call to vect_debug_details/stats
247 changed from NULL to UNKNOWN_LOC.
248 (vect_get_base_and_offset):
249 (vect_create_addr_base_for_vector_ref):
250 (get_vectype_for_scalar_type):
251 (vect_create_data_ref_ptr):
253 (vect_get_vec_def_for_operand):
254 (vect_finish_stmt_generation):
255 (vectorizable_assignment):
256 (vectorizable_operation):
257 (vectorizable_store):
259 (vect_transform_stmt):
260 (vect_update_ivs_after_vectorizer):
261 (vect_do_peeling_for_loop_bound):
262 (vect_do_peeling_for_alignment):
263 (vect_transform_loop):
264 (vect_is_simple_use):
265 (vect_analyze_operations):
266 (vect_is_simple_iv_evolution):
267 (vect_analyze_scalar_cycles):
268 (vect_analyze_data_ref_dependences):
269 (vect_compute_data_ref_alignment):
270 (vect_enhance_data_refs_alignment):
271 (vect_analyze_data_ref_access):
272 (vect_analyze_data_ref_accesses):
273 (vect_analyze_pointer_ref_access):
274 (vect_get_memtag_and_dr):
275 (vect_analyze_data_refs):
276 (vect_mark_relevant):
277 (vect_stmt_relevant_p):
278 (vect_mark_stmts_to_be_vectorized):
279 (vect_can_advance_ivs_p):
280 (vect_get_loop_niters):
282 (vectorize_loops): Likewise.
284 (vectorizable_load): Argument in call to vect_debug_details/stats
285 changed from loop to LOOP_LOC (loop_vinfo).
286 (vect_transform_loop):
287 (vect_analyze_operations):
288 (vect_analyze_scalar_cycles):
289 (vect_analyze_data_ref_dependence):
290 (vect_enhance_data_refs_alignment):
291 (vect_analyze_data_ref_accesses):
292 (vect_analyze_pointer_ref_access):
293 (vect_analyze_data_refs):
294 (vect_analyze_loop): Likewise.
296 (vect_analyze_loop_form): Argument in call to vect_debug_details/stats
297 changed from loop to loop_loc.
299 (vect_enhance_data_refs_alignment): Removed unused variable loop.
301 2005-02-03 Leehod Baruch <leehod@il.ibm.com>
302 Dorit Naishlos <dorit@il.ibm.com>
304 * tree-vectorizer.c (vect_analyze_data_ref_dependence): Function
306 (vect_analyze_data_ref_dependences): Likewise.
308 (vect_is_simple_use): Argument changed from loop structure to
309 loop_vect_info structure.
310 (vect_can_advance_ivs_p): Likewise.
311 (vect_create_index_for_vector_ref): Likewise.
312 (vect_update_ivs_after_vectorizer): Likewise.
313 (new_stmt_vec_info): Likewise.
315 (new_loop_vec_info): Second argument in call to new_stmt_vec_info was
316 changed from loop to loop_vinfo.
317 (vect_create_data_ref_ptr): First argument in call to
318 vect_create_index_for_vector_ref was changed from loop to loop_vinfo.
319 (vectorizable_assignment): Second argument in call to vect_is_simple_use
320 was changed from loop to loop_vinfo.
321 (vectorizable_operation): Likewise.
322 (vectorizable_store): Likewise.
323 (vect_mark_stmts_to_be_vectorized): Likewise.
324 (vect_do_peeling_for_loop_bound): First argument in call to
325 vect_update_ivs_after_vectorizer was changed from loop to loop_vinfo.
326 (vect_analyze_operations): Argument in call to vect_can_advance_ivs_p
327 was changed from loop to loop_vinfo.
328 (vect_analyze_data_ref_dependences): Third argument in call to
329 vect_analyze_data_ref_dependence was changed from loop to loop_vinfo.
331 (vect_create_index_for_vector_ref): Get the loop from loop_vinfo.
332 (vect_create_data_ref_ptr): Likewise.
333 (vect_init_vector): Likewise.
334 (vect_get_vec_def_for_operand): Likewise.
335 (vectorizable_assignment): Likewise.
336 (vectorizable_operation): Likewise.
337 (vectorizable_store): Likewise.
338 (vectorizable_load): Likewise.
339 (vect_update_ivs_after_vectorizer): Likewise.
340 (vect_is_simple_use): Likewise.
341 (vect_analyze_data_ref_dependence): Likewise.
342 (vect_analyze_pointer_ref_access): Likewise.
343 (vect_can_advance_ivs_p): Likewise.
345 * tree-vectorizer.h: (_loop_vec_info): Definition and macros moved
346 before _stmt_vec_info.
347 (_stmt_vec_info): Loop field replaced by loop_vec_info.
348 (STMT_VINFO_LOOP): Replaced with STMT_VINFO_LOOP_VINFO.
349 (new_stmt_vec_info): Argument changed from loop structure to
350 loop_vect_info structure
352 (vect_analyze_data_ref_dependences): Unnecessary line was removed.
353 (vect_analyze_offset_expr): Avoid 80 columns overflow.
354 (vect_create_addr_base_for_vector_ref): Likewise.
355 (vect_analyze_pointer_ref_access): Likewise.
357 2005-02-03 Andrew Pinski <pinskia@physics.uc.edu>
360 * tree-ssa-dse.c (fix_phi_uses): Update the occurs in abnormal
361 phi flag if the phi is abnormal.
363 2005-02-03 Andrew Pinski <pinskia@physics.uc.edu>
366 * tree-ssa.c (replace_immediate_uses): Update the immediate_uses
367 information for the new statement.
369 2005-02-03 Joseph S. Myers <joseph@codesourcery.com>
372 * gcc.c (main): Include the target in -v output.
374 2005-02-03 Alexandre Oliva <aoliva@redhat.com>
376 * combine.c (simplify_comparison, case SIGN_EXTEND, ZERO_EXTEND):
377 Do not drop the extend if we'd have to add a paradoxical subreg
378 later. Include optabs.h and insn-codes.h.
379 * Makefile.in (combine.o): Depend on $(OPTABS_H).
381 2005-02-02 Roger Sayle <roger@eyesopen.com>
384 * real.h (REAL_MODE_FORMAT_COMPOSITE_P): New macro.
385 * fold-const.c (const_binop): Avoid constant folding floating
386 point operations in modes that use composite representations.
387 * simplify-rtx.c (simplify_binary_operation): Likewise.
389 2005-02-02 Geoffrey Keating <geoffk@apple.com>
391 * config/rs6000/altivec.md (altivec_dst): Make the first operand
393 (altivec_dstt): Likewise.
394 (altivec_dstst): Likewise.
395 (altivec_dststt): Likewise.
396 * config/rs6000/rs6000.c (altivec_expand_dst_builtin): Adjust creation
399 2005-02-03 Dorit Naishlos <dorit@il.ibm.com>
401 * tree-vectorizer.c (slpeel_make_loop_iterate_ntimes): Copy
402 EXPR_LOCUS from orig_cond to the new cond_stmt.
403 (vect_finish_stmt_generation): Copy EXPR_LOCUS from stmt to
406 2005-02-03 Dorit Naishlos <dorit@il.ibm.com>
408 * tree-vectorizer.c (slpeel_make_loop_iterate_ntimes): Call
409 standard_iv_increment_position. Remove call to bsi_next
410 (no need to bump the iterator anymore).
411 (vect_create_index_for_vector_ref): Call
412 standard_iv_increment_position. Remove second function argument.
413 (vect_finish_stmt_generation): Remove call to bsi_next
414 (no need to bump the iterator anymore).
415 (vect_create_data_ref_ptr): Remove second argument (bsi) in call
416 to vect_create_index_for_vector_ref.
418 2005-02-03 Dorit Naishlos <dorit@il.ibm.com>
420 * tree-vectorizer.c (vect_mark_stmts_to_be_vectorized): Check if
421 phis are used out of the loop.
423 2005-02-03 Joseph S. Myers <joseph@codesourcery.com>
426 * c-decl.c (undef_nested_function): New variable.
427 (pop_scope): Diagnose undefined nested functions.
428 (finish_function): Don't attempt cgraph processing or genericizing
429 if current top-level function contained an undefined nested
430 function. Reset undef_nested_function at the end of a top-level
433 2005-02-02 Zdenek Dvorak <dvorakz@suse.cz>
435 * tree.c (build_int_cst_type): Take sign of the value into account
436 when deciding whether sign extend the value.
438 2005-02-02 Joseph S. Myers <joseph@codesourcery.com>
441 * gcc.c (cpp_unique_options): Remove %{trigraphs}.
442 (cpp_options, cc1_options): Change %{std*&ansi} to
443 %{std*&ansi&trigraphs}.
445 2005-02-02 Kazu Hirata <kazu@cs.umass.edu>
447 * cse.c: Update comments.
449 2005-02-02 Aldy Hernandez <aldyh@redhat.com>
451 * config/rs6000/rs6000.c (rs6000_spe_function_arg): Multi-reg
452 arguments go on the stack. Do not put complex doubles in odd/even
455 2005-02-02 Steven Bosscher <stevenb@suse.de>
457 * doc/invoke.texi: Update the options enabled at -O1, -O2,
458 and -O3. Update the documentation for -fgcse-sm, -fgcse-las,
459 -ftree-pre, and -fthread-jumps.
461 2005-02-02 Richard Sandiford <rsandifo@redhat.com>
463 PR tree-optimization/19578
464 * tree-flow.h (modified_noreturn_calls): Declare.
465 (noreturn_call_p): Declare.
466 * tree-flow-inline.h (noreturn_call_p): New function.
467 (modify_stmt): Add modified noreturn calls to modified_noreturn_calls.
468 * tree-cfg.c (modified_noreturn_calls): New variable.
469 (cleanup_control_flow): Use noreturn_call_p. Split basic blocks
470 that contain a mid-block noreturn call.
471 * tree-ssa.c (delete_tree_ssa): Clear modified_noreturn_calls.
473 2005-02-02 Kazu Hirata <kazu@cs.umass.edu>
475 * df.c (df_def_record_1, df_uses_record): Don't use
477 * df.h (DF_FOR_REGALLOC): Remove.
479 2005-02-02 Joseph S. Myers <joseph@codesourcery.com>
482 * c-typeck.c (really_start_incremental_init): Reset
483 constructor_max_index for arrays of incomplete type.
485 2005-02-02 Jeff Law <law@redhat.com>
487 * gcse.c (struct reg_set): Store the block index where the register
488 was set rather than the setting insn.
489 (record_one_set, compute_transp): Corresponding changes.
490 (pre_insert_copy_insn): Do not call replace_one_set.
491 (replace_one_set): Remove.
493 2005-02-02 Joseph S. Myers <joseph@codesourcery.com>
496 * doc/gcov.texi: Fix typo.
498 2005-02-01 Richard Henderson <rth@redhat.com
501 * config/i386/i386.h (MODES_TIEABLE_P): Use ix86_modes_tieable_p.
502 * config/i386/i386.c (ix86_hard_regno_mode_ok): Change return
504 (ix86_tieable_integer_mode_p, ix86_modes_tieable_p): New.
505 * config/i386/i386-protos.h: Update.
507 2005-02-01 Steven Bosscher <stevenb@suse.de>
509 PR tree-optimization/19217
510 * tree-cfg.c (verify_expr): Use the data field to see if TP was
511 seen inside a PHI node. Do not do the ADDR_EXPR check if it was.
512 (verify_stmts): Pass (void*)1 as data to verify_expr to signal
513 that it is walking a PHI node.
515 2005-02-01 Joseph S. Myers <joseph@codesourcery.com>
517 * doc/extend.texi (Nested Functions): Update.
519 2005-02-01 Richard Henderson <rth@redhat.com>
522 * optabs.c (expand_copysign_absneg): Export.
523 * optabs.h (expand_copysign_absneg): Declare.
524 * config/rs6000/rs6000.md (copysigntf3): New.
526 2005-02-01 Diego Novillo <dnovillo@redhat.com>
528 PR tree-optimization/19633
529 * tree-flow.h (struct ptr_info_def): Add field 'pt_null'.
530 * tree-ssa-alias.c (init_alias_info): Initialize.
531 (merge_pointed_to_info): Set.
532 (add_pointed_to_expr): Set pt_null if EXPR is a NULL pointer.
533 (dump_points_to_info_for): Show value of pt_null.
534 (struct count_ptr_d): Declare.
535 (find_ptr_dereference): Remove.
536 (ptr_is_dereferenced_by): Remove.
537 (count_ptr_derefs): New local function.
538 (count_uses_and_derefs): New local function.
539 (compute_points_to_and_addr_escape): Call it. If the number
540 of dereferences is greater than zero, mark the pointer as
541 dereferenced. If there are fewer dereferences than uses of
542 the pointer, the pointer's value escapes.
544 2005-02-01 Diego Novillo <dnovillo@redhat.com>
546 PR tree-optimization/19670
547 * tree-ssa.c (verify_ssa_name): Don't set TREE_VISITED
549 (verify_use): ... set it here, instead.
551 2005-02-01 Janis Johnson <janis187@us.ibm.com>
553 * doc/sourcebuild.texi (Test directives): Fix formatting.
555 * doc/sourcebuild.texi (Test directives): Describe selector
558 2005-02-01 Segher Boessenkool <segher@kernel.crashing.org>
560 * config.gcc (powerpc64-*-linux*): Default to -m64 also for
561 specific 64-bit CPUs.
563 2005-02-01 Richard Earnshaw <rearnsha@arm.com>
566 * arm.c (arm_eliminable_register): New function.
567 (adjacent_mem_locations): Don't allow eliminable registers. Use
568 HOST_WIDE_INT for address offsets.
569 * arm-protos.h (arm_eliminable_register): Add prototype.
571 2005-02-01 Steven Bosscher <stevenb@suse.de>
573 PR optimization/15242
574 * params.def (PARAM_MAX_GOTO_DUPLICATION_INSNS): New param.
575 * basic-block.h (duplicate_computed_gotos): Add prototype.
576 * bb-reorder.c (duplicate_computed_gotos): New function to
577 duplicate sufficiently small blocks ending in a computed jump.
578 * passes.c (rest_of_compilation): Call duplicate_computed_gotos
579 if not optimizing for size.
580 * cfgcleanup.c (try_crossjump_bb): If not optimizing for size,
581 never do tail merging for blocks ending in a computed jump.
582 * doc/invoke.texi: Document the max-goto-duplication-insns param.
584 2005-02-01 Eric Botcazou <ebotcazou@libertysurf.fr>
586 Patch from Richard Sandiford <rsandifo@redhat.com>
587 * reload1.c (choose_reload_regs): Prevent the find_equiv_reg() code
588 from inheriting a subreg equivalence with a non-spill register.
590 * tree.h (DECL_FUNCTION_CODE): Document that it is overloaded.
592 2005-02-01 Alexandre Oliva <aoliva@redhat.com>
594 * config/frv/frv.c (movcc_fp_destination_operand): New.
595 (gpr_or_memory_operand): Fix typo in comment.
596 (gpr_or_memory_operand_with_scratch): New.
597 * config/frv/frv.h (PREDICATE_CODES): Add the two new predicates.
598 * config/frv/frv.md (movcc_fp, movcc_fp_internal): Use
599 movcc_fp_destination_operand.
600 (reload_incc_fp): Use gpr_or_memory_operand_with_scratch.
601 Legitimize memory addresses using a scratch register.
603 2005-01-31 Jeff Law <law@redhat.com>
605 * tree-into-ssa.c (mark_def_sites_global_data): Make KILLS
606 bitmap a sparse bitmap instead of a simple bitmap.
607 (mark_def_sites_initialize_block): Corresponding changes.
608 (ssa_mark_def_sites_initialize_block): Likewise.
609 (ssa_mark_phi_uses): Likewise.
610 (mark_def_site, ssa_mark_def_sites): Likewise.
611 (mark_def_site_blocks): Likewise.
612 (rewrite_ssa_into_ssa): Likewise.
614 * tree-ssa-dom.c (record_cond): Pass correct variable type
615 for last argument to htab_find_slot_with_hash.
617 * fold-const.c (fold, case CEIL_MOD_EXPR): Do not lose side
618 effects when optimizing 0 % X. Do not try to optimize X % 0.
620 2005-01-31 James E. Wilson <wilson@specifixinc.com>
622 * config/ia64/itanium1.md (1_scall bypass): Change 2_mmalua to
625 2005-02-01 Eric Christopher <echristo@redhat.com>
627 * config/mips/mips.c (override_options): Warn if -mint64
629 * doc/invoke.texi (MIPS Options): Document that -mint64 is
632 2005-02-01 Kazu Hirata <kazu@cs.umass.edu>
634 * cse.c (cse_reg_info): Remove hash_next, next, regno. Add
636 (cse_reg_info_list, cse_reg_info_list_free, REGHASH_SHIFT,
637 REGHASH_SIZE, REGHASH_MASK, reg_hash, REGHASH_FN,
638 cached_cse_reg_info, GET_CSE_REG_INFO): Remove.
639 (cached_regno): Initialize to INVALID_REGNUM.
640 (cse_reg_info_table_size,
641 cse_reg_info_table_first_uninitialized,
642 cse_reg_info_timestamp): New.
643 (REG_TICK, REG_IN_TABLE, SUBREG_TICKED, REG_QTY): Use
645 (init_cse_reg_info, get_cse_reg_info_1): New.
646 (get_cse_reg_info): Cache the last look-up.
647 (new_basic_block): Update the code to clear mappings from
648 registers to cse_reg_info entries.
649 (cse_main): Call init_cse_reg_info.
651 * cse.c (get_cse_reg_info): Update a comment.
653 2005-01-31 Steven Bosscher <stevenb@suse.de>
656 * c-decl.c (start_decl): Do not warn about arrays of elements with
657 an incomplete type here.
658 (grokdeclarator): Do it here by making a pedwarn an error.
659 * c-typeck.c (push_init_level): If there were previous errors with
660 the constructor type, do not warn about braces for initializers.
661 (process_init_element): Likewise for excess initializer elements.
663 2005-01-31 Kazu Hirata <kazu@cs.umass.edu>
665 * cse.c (delete_trivially_dead_insn): Don't iterate.
667 2005-01-31 Andrew Pinski <pinskia@physics.uc.edu>
669 * config/rs6000/rs6000.md (copysignsf3): New expand.
670 (copysigndf3): Likewise.
672 2005-01-31 Steven Bosscher <stevenb@suse.de>
674 * recog.c (constrain_operands): Only look for earlyclobber operand
675 conflicts if an '&' constraint was seen.
677 2005-01-31 Marc Espie <espie@openbsd.org>
679 * config.gcc: Don't include embedded systems fragment, switches default
680 debugging format to ELF.
681 * config/i386/openbsdelf.h: Add DBX_REGISTER_NUMBER since we no
682 longer pick it up there.
684 2005-01-31 Dale Johannesen <dalej@apple.com>
686 * doc/extend.texi (nested functions): Fix linkage description.
687 Clarify that static is not allowed.
689 2005-01-31 Dale Johannesen <dalej@apple.com>
691 * config/rs6000/darwin.md (movsf_low_di): Make work.
692 (movdf_low_di): Make work.
694 2005-01-31 Dale Johannesen <dalej@apple.com>
696 * config/rs6000/darwin-tramp.asm (__trampoline_setup):
697 Make work for 64 bit.
699 2005-01-31 Roger Sayle <roger@eyesopen.com>
700 Dale Johannesen <dalej@apple.com>
703 * fold-const.c (fold_binary_op_with_conditional_arg):
704 Make types match original operands, before STRIP_NOPS.
706 2005-01-31 Alan Modra <amodra@bigpond.net.au>
708 * config/rs6000/linux-unwind.h (struct gcc_vregs): New.
709 (struct gcc_regs): Rename from gcc_pt_regs. Add more regs.
710 (struct gcc_sigcontext): Delete. Merge contents to..
711 (struct gcc_ucontext): ..here.
712 (get_sigcontext): Delete.
713 (get_regs): New function, like get_sigcontext but return regs ptr.
714 64-bit version finds regs from r1 to support vdso.
715 (ppc_linux_aux_vector): New function.
716 (ppc_fallback_frame_state): Modify for get_regs. Restore fprs
719 2005-01-31 Ira Rosen <irar@il.ibm.com>
721 * tree-vectorizer.c (vect_analyze_offset_expr): Use ssizetype instead
723 (vect_get_base_and_offset): Use ssizetype instead sizetype. Remove
724 redundant fold. Fix misalignment for MINUS_EXPR.
725 (vect_compute_data_ref_alignment): Use ssizetype instead sizetype.
726 (vect_analyze_pointer_ref_access): Likewise.
727 (vect_get_memtag_and_dr): Likewise.
729 2005-01-31 Richard Henderson <rth@redhat.com>
731 * config/i386/i386.c (absnegsf2_mixed): Remove all of the # hackery.
732 (absnegsf2_sse, absnegsf2_i387, absnegdf2_mixed, absnegdf2_sse,
733 absnegdf2_i387, absnegxf2_i387): Likewise.
735 2005-01-31 Richard Henderson <rth@redhat.com>
738 * optabs.c (expand_copysign_absneg): New.
739 (expand_copysign_bit): Split out from ...
740 (expand_copysign): ... here. Use expand_copysign_absneg.
742 2005-01-30 Roger Sayle <roger@eyesopen.com>
745 * config/pa/pa.md (anddi3, iordi3): On HPPA64, disallow an integer
746 constant as the second operand and a register as the third.
748 2005-01-31 Danny Smith <dannysmith@users.sourceforge.net>
751 * config/i386/i386.c (ix86_function_ok_for_sibcall): Also check
752 that dllimport'd functions do not use all call-clobbered registers
755 2005-01-30 Richard Henderson <rth@redhat.com>
758 * optabs.c (expand_absneg_bit): Split out from expand_unop and
759 expand_abs_nojump and generalize. Use operand_subword and
760 emit_no_conflict_block. Support large modes.
761 (expand_abs_nojump): Use it. Use HONOR_SIGNED_ZEROS to fall
763 (expand_unop): Likewise. Use HONOR_SIGNED_ZEROS to fall back
764 negation to subtraction.
765 (expand_copysign): Check that the format has signed zeros.
766 Use operand_subword and emit_no_conflict_block. Support large modes.
768 2005-01-30 Daniel Berlin <dberlin@dberlin.org>
770 Fix PR tree-optimization/19624
772 * Makefile.in (tree-ssa-pre.o): Add CFGLOOP_H.
773 * tree-ssa-pre.c: Add cfgloop.h.
775 (pre_stats): New member, constified.
776 (inserted_exprs): New static variable.
777 (NECESSARY): New macro.
778 (create_expression_by_pieces): Fold the expression, and
779 mark it as defaulting to not necessary. Also put in
781 (fully_constant_expression): New function.
782 (insert_into_preds_of_block): Modify to not insert phis when we
783 are playing with induction variables.
784 Push phis onto the inserted_exprs vector, and mark them as not
785 necessary by default.
786 (insert_aux): Call fully_constant_expression on eprime.
787 If all edges produce the same value, mark it constant.
788 (mark_operand_necessary): New function.
789 (remove_dead_inserted_code): New function.
790 (init_pre): Init loop optimizer to get loop info.
791 (fini_pre): Free loop_optimizer, and inserted_exprs vec.
792 (execute_pre): Commit edge inserts, then remove dead code.
794 2005-01-30 Richard Henderson <rth@redhat.com>
796 * rtl.c (rtx_equal_p): No early exit for CONST_VECTOR.
797 * varasm.c (const_rtx_hash_1): Handle CONST_VECTOR.
799 2005-01-30 Richard Henderson <rth@redhat.com>
802 * config/i386/i386.c (ix86_expand_copysign): New.
803 (ix86_split_copysign_const): New.
804 (ix86_split_copysign_var): Rename from ix86_split_copysign,
805 rearrange op1/nmask operands.
806 * config/i386/i386-protos.h: Update.
807 * config/i386/i386.md (copysignsf3): Use ix86_expand_copysign.
808 (copysigndf3): Likewise.
809 (copysignsf3_const, copysigndf3_const): New.
810 (copysignsf3_var): Rename from copysignsf3, split out splitter
811 and fix split predicate for X constraint.
812 (copysigndf3_var): Similarly.
814 2005-01-30 Kazu Hirata <kazu@cs.umass.edu>
816 * optabs.c, doc/c-tree.texi, doc/install.texi, doc/md.texi,
817 doc/passes.texi, doc/rtl.texi, doc/sourcebuild.texi,
818 doc/tm.texi, doc/tree-ssa.texi: Update copyright.
820 2005-01-29 Richard Henderson <rth@redhat.com>
823 * config/i386/i386.md (movdf_nointeger, movdf_integer): Fix ordering
824 of # and * in constraints.
826 2005-01-29 Richard Henderson <rth@redhat.com>
829 * expr.c (store_field): Don't strip sub-mode cast when the input
830 data is even smaller.
832 2005-01-29 Richard Henderson <rth@redhat.com>
835 * expr.c (categorize_ctor_elements_1): Check for CONSTRUCTOR of a
838 2005-01-29 Richard Henderson <rth@redhat.com>
840 * combine.c (make_field_assignment): Fix argument order
843 2005-01-29 Richard Guenther <richard.guenther@uni-tuebingen.de>
845 PR tree-optimization/15791
846 * fold-const.c (extract_array_ref): New function.
847 (fold): Fold comparisons between &a[i] and &a[j] or
848 semantically equivalent trees.
850 2005-01-29 Jeff Law <law@redhat.com>
852 * gcse.c (insert_expr_in_table): Revamp handling of available
853 and anticipatable occurrence lists to avoid unnecessary list
855 (insert_set_in_table): Similarly.
857 2005-01-29 Joseph S. Myers <joseph@codesourcery.com>
859 * c-common.c (fix_string_type): Just use c_build_qualified_type to
861 (c_build_qualified_type): Build qualified array types with
862 TYPE_MAIN_VARIANT pointing to corresponding unqualified type.
863 * c-decl.c (c_make_fname_decl): Build unqualified array type
864 before qualified type.
865 (grokdeclarator): Use TYPE_MAIN_VARIANT of typedef type if element
866 type is qualified, not just if type itself is. Don't apply
867 qualifiers to array type when declarator is processed. Apply
868 qualifiers to field type whether or not it is an array type.
869 Don't handle array types specially for applying qualifiers to
871 * c-typeck.c (composite_type): Build unqualified element type and
872 array type when forming composite of array types.
873 (common_pointer_type, comptypes, comp_target_types,
874 type_lists_compatible_p, build_indirect_ref, build_array_ref,
875 convert_for_assignment): Don't apply TYPE_MAIN_VARIANT to array
877 (type_lists_compatible_p): Cache TREE_VALUE (args1) and TREE_VALUE
878 (args2) in variables a1 and a2.
880 2005-01-29 Kazu Hirata <kazu@cs.umass.edu>
882 * cse.c (n_elements_made, max_elements_made): Remove.
883 (insert): Don't touch n_elements_made.
884 (cse_main): Don't touch n_elements_made or max_elements_made.
886 2005-01-28 Stephane Carrez <stcarrez@nerim.fr>
889 * config/m68hc11/t-m68hc11-gas (dp-bit.c): Fix typo causing a
890 configuration part of dp-bit.c to be lost.
892 2005-01-28 Roger Sayle <roger@eyesopen.com>
894 * expmed.c (expand_mult_highpart_optab): When attempting to use
895 a non-widening multiplication in a wider mode, the operands need
896 to be converted (zero or sign extended) to that mode.
898 2005-01-28 Ian Lance Taylor <ian@airs.com>
902 * gimple-low.c (block_may_fallthru): TRY_FINALLY_EXPR only falls
903 through if both operands fall through.
905 2005-01-28 Kazu Hirata <kazu@cs.umass.edu>
907 * cse.c (fold_rtx) <PC>: Don't optimize.
909 2005-01-28 Jeff Law <law@redhat.com>
911 * fold-const.c (fold, case CEIL_MOD_EXPR): Optimize 0 % X.
912 (case FLOOR_MOD_EXPR, ROUND_MOD_EXPR, TRUNC_MOD_EXPR): Similarly.
914 2005-01-28 Kazu Hirata <kazu@cs.umass.edu>
916 * cse.c (cse_reg_info_free_list, cse_reg_info_used_list,
917 cse_reg_info_used_list_end): Remove.
918 (cse_reg_info_list, cse_reg_info_list_free): New.
919 (get_cse_reg_info): When allocating an instance of
920 cse_reg_info, add it to the beginning of the cse_reg_info_list
921 list. Remove code to maintain cse_reg_info_used_list.
922 (new_basic_block): Reset the free list to the beginning of
925 2005-01-28 Richard Sandiford <rsandifo@redhat.com>
927 * config/frv/frv.c (bdesc_2arg): Remove __MDPACKH.
928 (frv_init_builtins): Change its prototype to take 4 uhalf arguments.
929 (frv_expand_mdpackh_builtin): New function.
930 (frv_expand_builtin): Use it to expand __MDPACKH.
931 * doc/extend.texi (__MDPACKH): Update documentation.
933 2005-01-28 Uros Bizjak <uros@kss-loka.si>
935 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use
936 SSE_FLOAT_MODE_P for use_sse computation.
938 2005-01-28 Ralf Corsepius <ralf.corsepius@rtems.org>
941 * config/sparc/rtemself.h (LINK_GCC_C_SEQUENCE_SPEC): undefine.
943 2005-01-27 Richard Henderson <rth@redhat.com>
945 * builtins.c (expand_builtin_copysign): New.
946 (expand_builtin): Call it.
947 * genopinit.c (optabs): Add copysign_optab.
948 * optabs.c (init_optabs): Initialize it.
949 (expand_copysign): New.
950 * optabs.h (OTI_copysign, copysign_optab): New.
951 (expand_copysign): Declare.
953 * config/alpha/alpha.md (UNSPEC_COPYSIGN): New.
954 (copysignsf3, ncopysignsf3, copysigndf3, ncopysigndf3): New.
956 * config/i386/i386.c (ix86_build_signbit_mask): Split from ...
957 (ix86_expand_fp_absneg_operator): ... here.
958 (ix86_split_copysign): New.
959 * config/i386/i386-protos.h: Update.
960 * config/i386/i386.md (UNSPEC_COPYSIGN): New.
961 (copysignsf3, copysigndf3): New.
963 * config/ia64/ia64.md (UNSPEC_COPYSIGN): New.
964 (copysignsf3, ncopysignsf3): New.
965 (copysigndf3, ncopysigndf3): New.
966 (copysignxf3, ncopysignxf3): New.
967 * config/ia64/ia64.c (rtx_needs_barrier): Handle UNSPEC_COPYSIGN.
969 2005-01-27 Arend Bayer <arend.bayer@web.de>
970 Kazu Hirata <kazu@cs.umass.edu>
972 * cse.c: (find_best_addr): Don't call copy_rtx before calling
973 fold_rtx. Save cost recomputation if fold_rtx did nothing.
974 (fold_rtx) <ASM_OPERANDS>: Don't do anything if INSN is
977 2005-01-27 Jeff Law <law@redhat.com>
979 * tree-into-ssa.c (ssa_rewrite_initialize_block): Do not register
980 definitions of SSA_NAMEs which are not being rewritten.
981 (rewrite_ssa_into_ssa): Only initialize the current definition
982 of an SSA_NAME if that SSA_NAME has been marked for rewriting.
983 If checking is enabled, assert that SSA_NAME_AUX is clear for all
984 SSA_NAMEs before returning.
986 2005-01-27 Kazu Hirata <kazu@cs.umass.edu>
988 * c-common.def, c-dump.c, c-gimplify.c, c-objc-common.c,
989 cfgexpand.c, dbxout.c, function.h, opts.c, tree-flow-inline.h,
990 tree-ssa-operands.h, tree-ssa-threadupdate.c, tree-tailcall.c,
991 config/avr/avr.md, config/cris/aout.h, config/cris/cris.h,
992 config/mips/iris6.h, config/sh/sh.c: Update copyright.
994 2005-01-27 Steven Bosscher <stevenb@suse.de>
997 * opts.c (decode_options): Move flag_thread_jumps from -O1 and
998 higher to -O2 and higher. Likewise for tree PRE.
999 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Only iterate at -O2
1002 2005-01-27 Ian Lance Taylor <ian@airs.com>
1005 * gimple-low.c (try_catch_may_fallthru): New static function.
1006 (block_may_fallthru): Handle TRY_CATCH_EXPR.
1007 * tree-inline.c (expand_call_inline): Don't warn about reaching
1008 the end of a non-void function being inlined if the function uses
1011 2005-01-27 Jakub Jelinek <jakub@redhat.com>
1013 * config/i386/i386.h (CALL_USED_REGISTERS): Fix comment pastos.
1016 * c-decl.c (warn_if_shadowing): Handle old_decl error_mark_node.
1017 (pushdecl): Only use DECL_FILE_SCOPE_P if DECL_P.
1018 (implicitly_declare): Handle error_mark_node.
1020 2005-01-27 Richard Henderson <rth@redhat.com>
1023 * tree.h (struct tree_decl): Add debug_expr_is_from.
1024 (DECL_DEBUG_EXPR_IS_FROM): New.
1025 (DECL_DEBUG_EXPR): Rename from DECL_DEBUG_ALIAS_OF.
1026 * dwarf2out.c (dwarf2out_var_location): Update to match.
1027 * tree-outof-ssa.c (create_temp): Likewise.
1028 * var-tracking.c (track_expr_p): Likewise.
1029 * tree-sra.c (instantiate_element): Set DECL_DEBUG_EXPR.
1030 * c-objc-common.c (c_tree_printer) <'D'>: Handle DECL_DEBUG_EXPR.
1031 * toplev.c (default_tree_printer): Likewise.
1033 2005-01-27 Alexandre Oliva <aoliva@redhat.com>
1035 * config/frv/frv.c (frv_class_likely_spilled_p): Add GR8_REGS,
1036 GR9_REGS, GR89_REGS, FDPIC_FPTR_REGS, FDPIC_REGS.
1038 2005-01-27 Steven Bosscher <stevenb@suse.de>
1040 * tree.h (SWITCH_COND, SWITCH_BODY, SWITCH_LABELS, CASE_LOW,
1041 CASE_HIGH, CASE_LABEL, ASM_STRING, ASM_OUTPUTS, ASM_INPUTS,
1042 ASM_CLOBBERS): Add tree checks.
1044 * c-common.h (SWITCH_TYPE): Rename to SWITCH_STMT_TYPE.
1045 (SWITCH_STMT_COND, SWITCH_STMT_BODY): New.
1046 * c-common.def (SWITCH_STMT): Update to match.
1047 * c-common.c (c_do_switch_warnings): Use SWITCH_STMT accessor
1048 macros instead of SWITCH_EXPR ones.
1049 * c-dump.c (c_dump_tree): Likewise.
1050 * c-gimplify.c (gimplify_switch_stmt): Likewise.
1051 * c-typeck.c (c_start_case, do_case, c_finish_case): Likewise.
1053 * doc/c-tree.texi (SWITCH_STMT): Update accessor macro names.
1055 2005-01-27 Alan Modra <amodra@bigpond.net.au>
1057 * unwind-dw2.c (execute_stack_op): Add missing cases for
1058 DW_OP_shl, DW_OP_shr, DW_OP_shra, DW_OP_xor.
1060 2005-01-27 Eric Botcazou <ebotcazou@libertysurf.fr>
1062 * config.gcc (Obsolete configurations): Remove 'dummy', add
1063 sparclite-*-coff*, sparclite-*-elf*, sparc86x-*-elf* and
1066 2005-01-26 Diego Novillo <dnovillo@redhat.com>
1068 PR tree-optimization/19633
1069 * tree-ssa-alias.c (ptr_is_dereferenced_by): Also handle
1071 (maybe_create_global_var): Do not create .GLOBAL_VAR if there
1072 are no call-clobbered variables.
1073 * tree-outof-ssa.c (check_replaceable): Return false for calls
1076 2005-01-26 Ulrich Weigand <uweigand@de.ibm.com>
1078 * dbxout.c (dbxout_symbol_location): Resolve constant pool references
1079 even for variables with NULL DECL_INITIAL.
1081 2005-01-26 Stuart Hastings <stuart@apple.com>
1083 * gimplify.c (shortcut_cond_expr): Re-compute side-effects.
1085 2005-01-26 Richard Henderson <rth@redhat.com>
1088 * c-decl.c (finish_struct): Set DECL_MODE after resetting a
1090 * expr.c (store_field): Strip conversions to odd-bit-sized types
1091 if the destination field width matches.
1093 2005-01-26 Richard Henderson <rth@redhat.com>
1095 * c-decl.c, expmed.c, expr.c: Revert last change.
1097 2005-01-26 Richard Henderson <rth@redhat.com>
1100 * c-decl.c (finish_struct): Set DECL_MODE after resetting a
1102 * expmed.c (store_fixed_bit_field): Create a paradoxical subreg
1103 if we don't need the bits above those present in the current mode.
1104 * expr.c (store_field): Strip conversions to odd-bit-sized types
1105 if the destination field width matches.
1107 2005-01-26 Richard Sandiford <rsandifo@redhat.com>
1109 * config/mips/iris6.h (ENDFILE_SPEC): Don't link in irix-csr.o if
1110 either -ffast-math or -funsafe-math-optimizations is in use.
1112 2005-01-26 Marek Michalkiewicz <marekm@amelek.gda.pl>
1116 * config/avr/avr.c (notice_update_cc): Only set condition code for
1117 ashrqi3 if shift count > 0.
1118 (out_shift_with_cnt): Handle shift count <= 0 as a no-op.
1119 (ashlqi3_out, ashlhi3_out, ashlsi3_out, ashrqi3_out, ashrhi3_out,
1120 ashrsi3_out, lshrqi3_out, lshrhi3_out, lshrsi3_out): Handle shift
1121 count <= 0 as a no-op, and shift count >= width by copying zero
1122 or sign bit to all bits of the result.
1123 * config/avr/avr.md (all shifts): Add alternatives for zero shift
1124 count, with attribute "length" set to 0 and "cc" set to "none".
1126 2005-01-26 Aldy Hernandez <aldyh@redhat.com>
1128 * doc/invoke.texi: Document -mTLS.
1130 * testsuite/gcc.target/frv/all-tls-global-dynamic.c: New.
1131 * testsuite/gcc.target/frv/all-tls-initial-exec.c: New.
1132 * testsuite/gcc.target/frv/all-tls-initial-exec-pic.c: New.
1133 * testsuite/gcc.target/frv/all-tls-local-dynamic.c: New.
1134 * testsuite/gcc.target/frv/all-tls-local-dynamic-plt.c: New.
1135 * testsuite/gcc.target/frv/all-tls-local-dynamic-plt-pic.c: New.
1136 * testsuite/gcc.target/frv/all-tls-local-exec.c: New.
1137 * testsuite/gcc.target/frv/all-tls-local-exec-TLS.c: New.
1139 * configure.ac: Check for a TLS capable gas.
1141 * configure: Regenerate.
1143 2005-01-26 Richard Henderson <rth@redhat.com>
1146 * combine.c (make_field_assignment): Simplify store to zero_extract
1147 from a source with an overlapping mask.
1149 2005-01-26 Aldy Hernandez <aldyh@redhat.com>
1151 2004-11-11 Eric Christopher <echristo@redhat.com>
1153 * config/frv/frv.md: Add fr400_integer automaton. Don't
1154 allow TYPE_MUL, TYPE_MACC, or TYPE_CUT to issue in parallel
1157 2005-01-26 Steven Bosscher <stevenb@suse.de>
1160 * cfgbuild.c (make_edges): Do not clear or set
1161 current_function_has_computed_jump.
1162 * function.h (struct function): Remove the has_computed_jump field.
1163 (current_function_has_computed_jump): Do not define.
1164 * sched-rgn.c (is_cfg_nonregular): Return true if a basic block ends
1165 in a computed jump. Ignore current_function_has_computed_jump.
1167 2005-01-26 Richard Henderson <rth@redhat.com>
1170 * expr.c (categorize_ctor_elements): New argument p_must_clear.
1171 (categorize_ctor_elements_1): Likewise. Detect a union that isn't
1173 (mostly_zeros_p): Update for new categorize_ctor_elements argument.
1174 * gimplify.c (gimplify_init_constructor): Likewise. Only shove
1175 objects into static storage if they have more than one non-zero value.
1176 * tree.h (categorize_ctor_elements): Update decl.
1178 2005-01-26 Steven Bosscher <stevenb@suse.de>
1181 * tree.h (CALL_EXPR_TAILCALL): Add comment.
1182 * calls.c (check_sibcall_argument_overlap_1): Revert the change
1183 to this function from 2004-07-10.
1184 * tree-tailcall.c (suitable_for_tail_opt_p): Do not consider the
1185 the current function for tail call optimizations if the address
1186 of one of it its arguments is taken.
1188 2005-01-26 Kazu Hirata <kazu@cs.umass.edu>
1190 * cse.c (fold_rtx): Call equiv_constant only when necessary.
1192 2005-01-26 Bernd Schmidt <bernd.schmidt@analog.com>
1194 * function.c (expand_function_end): If current_function_calls_alloca,
1195 emit stack restore in a place that is reached when the function falls
1198 2005-01-26 Alexandre Oliva <aoliva@redhat.com>
1200 * Makefile.in (../$(build_subdir)/fixincludes/fixincl): Add dummy
1201 rule to avoid matching inappropriate suffix rule.
1203 2005-01-26 Jan Hubicka <jh@suse.cz>
1205 PR tree-optimization/19241
1206 * tree-inline.c (copy_body_r): Do not walk subtrees after substituting.
1208 2005-01-26 Hans-Peter Nilsson <hp@axis.com>
1210 * config/cris/cris.h (LIB_SPEC) <sim*>: Use -lsyslinux, not
1211 -lsyssim. Wrap -lc -lsyssim in --start-group --end-group instead
1213 (STARTFILE_SPEC): Add crti.o, before crtbegin.o.
1214 <sim*>: Always use crt1.o, regardless of N in simN.
1215 <!sim>: Drop support for -pg and -p variants; always use crt0.o.
1216 (ENDFILE_SPEC): Add crtn.o, after crtend.o.
1217 (CRT_CALL_STATIC_FUNCTION): Remove.
1218 * config/cris/aout.h (STARTFILE_SPEC): <!melinux && sim*>: Always
1219 use crt1.o, regardless of N in simN.
1220 <!sim>: Drop support for -pg and -p variants; always use crt0.o.
1221 (LIB_SPEC): Do not define; default to the one in config/cris/cris.h.
1222 (CRIS_CPP_SUBTARGET_SPEC) <melinux && !nostdinc>: Remove special
1223 case isystem setting.
1224 (MAX_OFILE_ALIGNMENT): Correct to 32, not 16.
1226 2005-01-25 Ian Lance Taylor <ian@airs.com>
1228 * gimple-low.c (block_may_fallthru): Correct handling of
1229 SWITCH_EXPR--may fall through if SWITCH_LABELS is NULL.
1231 2005-01-26 Hans-Peter Nilsson <hp@axis.com>
1233 * config/cris/cris.h (CPP_SPEC): Adjust the various mtune=
1234 handlers to match indentation; nested under the mtune=* handler.
1235 Similar for mcpu= and march=.
1237 2005-01-25 Daniel Jacobowitz <dan@codesourcery.com>
1239 * config/mips/linux.h (LIB_SPEC): Remove -rpath-link.
1240 * config/mips/linux64.h (LIB_SPEC): Likewise.
1241 * config/mn10300/linux.h (LIB_SPEC): Likewise.
1242 * config/sh/linux.h (LIB_SPEC): Likewise.
1244 2005-01-25 Aldy Hernandez <aldyh@redhat.com>
1246 * config/frv/frv.c (frv_legitimize_tls_address): New.
1247 (TARGET_HAVE_TLS): Define.
1248 (FRV_SYMBOL_REF_TLS_P): Define.
1249 (frv_override_options): Handle new register classes.
1250 (frv_legitimate_address_p): Reject tls addresses.
1251 (frv_legitimize_address): Handle TLS addresses.
1252 (gen_inlined_tls_plt): New.
1254 (frv_legitimize_tls_address): New.
1255 (unspec_got_name): Add TLS entries.
1256 (got12_operand): Add R_FRV_TLSMOFF12 case.
1257 (frv_emit_move): Fixup TLS addresses.
1258 (frv_emit_movsi): Legitimize TLS addresses.
1260 * config/frv/frv.h (MASK_BIG_TLS): New.
1261 (TARGET_BIG_TLS): New.
1262 (HAVE_AS_TLS): Define.
1263 (TARGET_SWITCHES): Add -mTLS and -mtls options.
1264 (enum reg_class): Add GR8_REGS, GR9_REGS, GR89_REGS.
1265 (REG_CLASS_NAMES): Same.
1266 (REG_CLASS_CONTENTS): Same.
1267 (CONSTRAINT_LEN): New.
1268 (REG_CLASS_FROM_CONSTRAINT): New.
1269 (PREDICATE_CODES): Add symbolic_operand.
1271 * config/frv/frv.md (define_constants): Add UNSPEC_GETTLSOFF,
1272 UNSPEC_TLS_LOAD_GOTTLSOFF12, UNSPEC_TLS_INDIRECT_CALL,
1273 UNSPEC_TLS_TLSDESC_LDD, UNSPEC_TLS_TLSDESC_LDD_AUX,
1274 UNSPEC_TLS_TLSOFF_LD, UNSPEC_TLS_LDDI, UNSPEC_TLSOFF_HILO,
1275 R_FRV_GOTTLSOFF_HI, R_FRV_GOTTLSOFF_LO, R_FRV_TLSMOFFHI,
1276 R_FRV_TLSMOFFLO, R_FRV_TLSMOFF12, R_FRV_TLSDESCHI,
1277 R_FRV_TLSDESCLO, R_FRV_GOTTLSDESCHI, R_FRV_GOTTLSDESCLO, GR8_REG,
1278 GR9_REG, GR14_REG, LRREG.
1279 (type): Add load_or_call attribute.
1280 ("load_or_call"): New reservation.
1281 ("call_gettlsoff"): New.
1282 ("tls_indirect_call"): New.
1283 ("tls_load_gottlsoff12"): New.
1284 ("tlsoff_hilo"): New.
1285 ("tls_tlsdesc_ldd"): New.
1286 ("tls_tlsoff_ld"): New.
1289 * config/frv/frv-protos.h (symbolic_operand): Protoize.
1291 2005-01-25 Alexandre Oliva <aoliva@redhat.com>
1293 * config/frv/frv.h (ASM_OUTPUT_DWARF_DTPREL): Define.
1294 * config/frv/frv-protos.h (frv_output_dwarf_dtprel): Declare.
1295 * config/frv/frv.c (TLS_BIAS): Define.
1296 (frv_output_dwarf_dtprel): New.
1298 2005-01-26 Jakub Jelinek <jakub@redhat.com>
1300 PR rtl-optimization/19579
1301 * ifcvt.c (noce_try_cmove_arith): If emitting instructions to set up
1302 both A and B, see if they don't clobber registers the other expr uses.
1304 2005-01-25 J"orn Rennecke <joern.rennecke@st.com>
1306 * real.c (do_add): Initialize signalling and canonical members.
1308 * real.c (real_from_integer): Zero out destination.
1310 2005-01-18 Jan Hubicka <jh@suse.cz>
1312 PR tree-optimize/19337
1313 * tree-inline.c (initialize_inlined_parameters): Copy
1314 saved_static_chain_decl only.
1316 2005-01-25 Richard Henderson <rth@redhat.com>
1320 * config/i386/i386.c (x86_inter_unit_moves): Zero.
1321 (ix86_preferred_reload_class): Rewrite fp-constant section, with
1322 80387 enabled, to return a proper subclass. Return the subset that
1323 overlaps with GENERAL_REGS for PLUS.
1324 * config/i386/i386.md (movsi_1, movdi_2): Set type to "mmx" for pxor.
1326 2005-01-25 Richard Henderson <rth@redhat.com>
1329 * tree-complex.c (expand_complex_div_wide): Use the correct formulae.
1331 2005-01-25 Roger Sayle <roger@eyesopen.com>
1333 * expmed.c (choose_multiplier): Change interface to return the
1334 multiplier lowpart as an rtx instead of an unsigned HOST_WIDE_INT.
1335 (expand_divmod): Update calls to choose_multiplier.
1337 2005-01-24 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1340 * expmed.c (extract_bit_field): Use adjust_address instead of
1341 gen_lowpart when op0 is a MEM.
1343 2005-01-24 Steven Bosscher <stevenb@suse.de>
1345 * cfgexpand.c (expand_gimple_tailcall): Fix typo.
1347 2005-01-24 Richard Sandiford <rsandifo@redhat.com>
1349 * config/mips/irix-csr.c: New file.
1350 * config/mips/t-iris6 (irix-csr.o): New rule to build it.
1351 (EXTRA_MULTILIB_PARTS): Add irix-csr.o.
1352 * config/mips/iris6.h (ENDFILE_SPEC): Include it in n32 and n64
1355 2005-01-24 Eric Botcazou <ebotcazou@libertysurf.fr>
1358 * config.gcc (sparc-*-elf*): Remove sol2.h, sparc/sol2.h and
1359 sparc/elf.h, add sparc/sp-elf.h.
1360 (sparc-*-rtems*): Likewise.
1361 (sparclite-*-elf*): Remove sol2.h, sparc/sol2.h, sparc/elf.h and
1362 tm-dwarf2.h, add sparc/sp-elf.h.
1363 (sparc86x-*-elf): Likewise.
1364 (sparc64-*-elf*): Remove sol2.h, sparc/sol2.h and tm-dwarf2.h.
1365 * config/sparc/liteelf.h (TARGET_SUB_OS_CPP_BUILTINS): Rename into
1366 TARGET_OS_CPP_BUILTINS.
1367 * config/sparc/sp86x-elf (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1368 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1370 * config/sparc/openbsd64.h (NO_IMPLICIT_EXTERN_C): Undefine.
1371 * config/sparc/sp64-elf.h (NO_IMPLICIT_EXTERN_C): New macro.
1372 (SWITCH_TAKES_ARG): Likewise.
1373 (LOCAL_LABEL_PREFIX): Likewise.
1374 (ASM_GENERATE_INTERNAL_LABEL): Likewise.
1375 (TARGET_N_FORMAT_TYPES): Delete.
1376 (TARGET_FORMAT_TYPES): Likewise.
1377 (ASM_DECLARE_FUNCTION_SIZE): Likewise.
1378 * config/sparc/elf.h: Delete.
1379 * config/sparc/sp-elf.h: New file.
1381 2005-01-24 Kazu Hirata <kazu@cs.umass.edu>
1383 * tree-flow-inline.h (phi_arg_from_edge): Remove.
1384 * tree-flow.h: Remove the corresponding prototype.
1385 * tree-ssa-operands.h, tree-ssa-threadupdate.c, tree-ssa.c,
1386 tree-vectorizer.c: Use dest_idx instead of phi_arg_from_edge.
1388 2005-01-24 Jakub Jelinek <jakub@redhat.com>
1391 * config/rs6000/rs6000.c (rs6000_file_start): Create toc section
1392 for AIX ABI or ELF -fPIC.
1393 (rs6000_emit_load_toc_table): Don't create toc_section here.
1394 (rs6000_xcoff_file_start): Nor here.
1396 2005-01-24 J"orn Rennecke <joern.rennecke@st.com>
1398 * sh.c (ra.h): Don't #include.
1399 (hard_regs_intersect_p): New function, resurrected from ra.c.
1401 * sh.c: Fix 1996 Copyright.
1403 2005-01-24 Richard Henderson <rth@redhat.com>
1404 Aldy Hernandez <aldyh@redhat.com>
1406 * regrename.c (note_sets): Handle subregs.
1408 2005-01-24 Joseph S. Myers <joseph@codesourcery.com>
1411 * genconditions.c (write_header, write_conditions): Elide file if
1414 2005-01-24 Kazu Hirata <kazu@cs.umass.edu>
1416 * c-opts.c, expmed.c, gengtype.c, tree-complex.c,
1417 config/arm/iwmmxt.md, config/avr/avr.c, config/s390/s390.h:
1420 2005-01-24 Jakub Jelinek <jakub@redhat.com>
1422 * builtins.c (get_pointer_alignment, gimplify_va_arg_expr,
1423 expand_builtin_printf, expand_builtin_fprintf,
1424 expand_builtin_sprintf): Use POINTER_TYPE_P instead TREE_CODE
1425 checking against POINTER_TYPE.
1426 (validate_arglist): Handle POINTER_TYPE code by checking
1429 2005-01-24 Paolo Bonzini <bonzini@gnu.org>
1431 * rtlanal.c (non_rtx_starting_operands, for_each_rtx_1,
1433 (for_each_rtx): Call for_each_rtx_1.
1434 * rtl.h (init_rtlanal): Declare.
1435 * toplev.c (backend_init): Call init_rtlanal.
1437 2005-01-24 Jakub Jelinek <jakub@redhat.com>
1439 * flow.c (propagate_one_insn): Formatting.
1442 * flow.c (libcall_dead_p): Be more conservative if unsure.
1443 If there are any instructions between insn and call, see if they are
1444 all dead before saying the libcall is dead.
1446 2005-01-24 Uros Bizjak <uros@kss-loka.si>
1448 * config/i386/i386.md (*extendsfdf2_sse): Use "nonimmediate_operand"
1449 operand constraint for operand 0.
1451 2005-01-23 Richard Henderson <rth@redhat.com>
1454 * tree-complex.c (expand_complex_div_wide): Set EDGE_FALLTHRU on
1455 edges leading to join block.
1457 2005-01-23 Roger Sayle <roger@eyesopen.com>
1459 * expmed.c (expand_mult_highpart): Make static. Change type of
1460 constant multiplier argument from unsigned HOST_WIDE_INT to rtx.
1461 (expand_divmod): Updates calls to expand_mult_highpart by using
1462 gen_int_mode to pass a CONST_INT rtx.
1463 * rtl.h (expand_mult_highpart): Remove prototype.
1465 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
1467 * flow.c (regset_bytes, regset_size): Remove.
1469 * cfganal.c: Fix a reference to Harvey's paper.
1471 2005-01-23 Steven Bosscher <stevenb@suse.de>
1473 PR rtl-optimization/19464
1474 * tree-optimize.c (init_tree_optimization_passes): Add one more
1475 copyrename pass just before out-of-ssa.
1477 2005-01-23 Joseph S. Myers <joseph@codesourcery.com>
1480 * recog.c (recog_memoized): Don't define if GENERATOR_FILE.
1481 * ggc-none.c (ggc_free): Define.
1483 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
1485 * cse.c (max_reg, max_insn_uid): Remove.
1486 (cse_main): Don't access max_reg or max_insn_uid.
1488 * c-common.c (explicit_flag_signed_bitfields,
1489 lang_statement_code_p, lang_expand_function_end): Remove.
1490 * c-common.h: Remove the corresponding declarations.
1491 * c-opts.c (c_common_handle_option): Remove all write access
1492 to explicit_flag_signed_bitfields.
1494 * cgraph.c (cgraph_varpool_n_nodes): Remove.
1495 (cgraph_varpool_node): Don't access cgraph_varpool_n_nodes.
1496 * cgraph.h: Remove the corresponding declaration.
1498 * gcse.c (null_pointer_info): Remove.
1500 2005-01-23 Roger Sayle <roger@eyesopen.com>
1501 Eric Botcazou <ebotcazou@libertysurf.fr>
1503 * combine.c (simplify_logical): Only simplify logical expressions
1504 of the form ior(and(x,y),z) by the inverse distributive law if the
1505 result is cheaper than the original.
1507 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
1509 * alias.c, c-common.h, c-incpath.c, c-incpath.h, expr.c,
1510 fold-const.c, gimplify.c, params.h, tree-data-ref.c,
1511 tree-if-conv.c, tree-nested.c, tree-outof-ssa.c,
1512 tree-ssa-dom.c, tree-vectorizer.c, tree.def, config/darwin.c,
1513 config/freebsd-spec.h, config/arm/arm.h,
1514 config/h8300/h8300.md, config/i386/i386.md,
1515 config/i386/predicates.md, config/i386/sse.md,
1516 config/ia64/ia64.c, config/ip2k/ip2k.c, config/s390/s390.c,
1517 config/vax/vax.md: Fix comment typos. Follow spelling
1520 2005-01-23 Steven Bosscher <stevenb@suse.de>
1522 * gengtype.c (walk_type): Produce `not equal to' compares for loop
1523 tests, instead of `less than'.
1525 2005-01-23 Richard Sandiford <rsandifo@redhat.com>
1527 * function.c (assign_parm_setup_block): When creating a new stack slot
1528 for a parameter, get its alignment from the parameter's DECL_ALIGN
1529 rather than the type's TYPE_ALIGN. Make sure that the parameter
1530 is at least word aligned.
1532 2005-01-22 Roger Sayle <roger@eyesopen.com>
1535 * config/avr/avr.c (avr_hard_regno_mode_ok): Rewrite.
1537 2005-01-22 Richard Henderson <rth@redhat.com>
1540 * config/i386/i386.md (movsfcc_1_sse_max): Use nonimmediate_operand
1541 in both compare operands.
1542 (movdfcc_1_sse_max): Likewise.
1543 (movsfcc_1_sse): Likewise. Add earlyclobber for scratch.
1544 (movdfcc_1_sse): Likewise.
1545 * config/i386/i386.c (ix86_split_sse_movcc): Emit copies into the
1546 scratch register as needed.
1548 2005-01-22 Richard Henderson <rth@redhat.com>
1550 * config/i386/i386.md (smaxsf3): Fix mnemonic typo.
1552 * config/i386/i386.c (ix86_prepare_fp_compare_args): Fix is_sse test.
1554 2005-01-22 Richard Henderson <rth@redhat.com>
1556 * genopinit.c (optabs): Use smin/smax for floating point too.
1557 * doc/md.texi: Update to match. Clarify that floating point
1558 results are undefined for +0/-0 and NaN.
1559 * doc/rtl.texi: Likewise.
1560 * rtl.def (SMIN, SMAX): Likewise
1561 * tree.def (MIN_EXPR, MAX_EXPR): Likewise.
1563 * config/alpha/alpha.md (smaxdf3, smindf3, smaxsf3, sminsf3): Add
1564 leading 's' to the name.
1565 * config/ia64/ia64.md (smaxsf3, sminsf3, smaxdf3, smindf3,
1566 smaxxf3, sminxf3): Likewise.
1567 * config/rs6000/rs6000.md (smaxdf3, smindf3, smaxsf3, sminsf3):
1570 2005-01-22 Paul Brook <paul@codesourcery.com>
1572 * config/arm/arm.md: Use "Uy" constraint for wcgr load/stores.
1574 2005-01-21 Mark Dettinger <dettinge@de.ibm.com>
1576 * config/s390/s390.c (struct processor_costs): New fields
1577 dlgr, dlr, dr, dsgfr, dsgr.
1578 (z900_cost, z990_cost): Values for new fields.
1579 (s390_rtx_costs): New cases MEM und COMPARE in switch
1580 statement. Modified handling of SIGN_EXTEND, ZERO_EXTEND,
1581 DIV, MOD, UDIV, UMOD.
1583 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
1585 * config/s390/s390.md ("doloop_si64"): Reload input value directly
1586 into the register being decremented.
1587 ("doloop_si31", "doloop_di"): Likewise.
1588 ("*doloop_si_long"): Adapt pattern.
1590 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
1592 * config/s390/s390.h (HARD_REGNO_NREGS): Fix computation for
1594 (CLASS_MAX_NREGS): Likewise.
1596 2005-01-21 Daniel Berlin <dberlin@dberlin.org>
1598 * doc/tree-ssa.texi (Statement Operands): Add example for new
1599 must-def macro. Note deprecation of old operands interface.
1601 2005-01-22 Richard Sandiford <rsandifo@redhat.com>
1603 PR tree-optimization/19484
1604 * tree-cfg.c (remove_fallthru_edge): New function.
1605 (cleanup_control_flow): Remove fallthru edges from calls that are
1606 now known not to return.
1608 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
1610 Fix PR tree-optimization/19038
1611 * tree-ssa-dom.c (cprop_operand): Don't replace loop invaeriant
1612 copies with loop variant ones.
1614 2005-01-22 Kazu Hirata <kazu@cs.umass.edu>
1616 * cfganal.c, real.h, reorg.c, timevar.def, tree-ssa-ccp.c,
1617 config/alpha/alpha-protos.h, config/alpha/alpha.h,
1618 config/alpha/alpha.md, config/alpha/predicates.md,
1619 config/sparc/freebsd.h, config/sparc/netbsd-elf.h,
1620 config/sparc/sol2.h: Update copyright.
1622 * tree-cfg.c (remove_forwarder_block_with_phi): Look at the
1623 first label to see if it is a nonlocal label.
1625 2005-01-22 David Edelsohn <edelsohn@gnu.org>
1626 Andrew Pinski <pinskia@physics.uc.edu>
1629 * config/rs6000/rs6000.c (rs6000_va_start): Saturate n_gpr at
1630 maximum number of GPRs. Saturate n_fpr at maximum number of FPRs.
1632 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
1634 * config/rs6000/rtems.h: Update copyright.
1636 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
1639 * config/rs6000/rtems.h: Resurrect cpp_os_rtems_spec from gcc < 3.4.
1640 (CPP_OS_RTEMS_SPEC): New (From gcc-3.3's config/rs6000/sys4.h).
1641 (SUBSUBTARGET_EXTRA_SPECS): Use CPP_OS_RTEMS_SPEC.
1643 2005-01-22 Volker Reichelt <reichelt@igpm.rwth-aachen.de>
1646 * c-typeck.c (convert_arguments): Check for error_mark_node.
1648 2005-01-21 Ian Lance Taylor <ian@airs.com>
1650 PR tree-optimization/13000
1651 * tree-inline.c: Include "tree-flow.h".
1652 (expand_call_inline): If warn_return_type, warn if non-void inline
1653 function falls through.
1654 * tree-cfg.c (execute_warn_function_return): Don't warn about
1655 control reaching end if TREE_NO_WARNING is set. Set
1657 * gimple-low.c (block_may_fallthru): Don't assume that SWITCH_EXPR
1659 * gimplify.c (shortcut_cond_expr): Don't emit a jump over the else
1660 branch if we don't need one.
1661 * c-typeck.c: Include "tree-flow.h"
1662 (c_finish_bc_stmt): Don't add a goto if the current statement
1663 list doesn't fall through to the current point.
1665 2005-01-21 Roger Sayle <roger@eyesopen.com>
1667 PR rtl-optimization/576
1668 * real.c (real_arithmetic): Change return type from void to bool
1669 to return an indication that the result may be inexact.
1670 * real.h (real_arithmeric): Update prototype.
1671 * fold-const.c (const_binop): Don't constant fold floating
1672 point expressions when the user specifies -frounding-math and
1673 the result may depend upon the run-time rounding mode.
1674 (fold_convert_const_real_from_real): Clean-up.
1675 (fold_initializer): Ignore flag_rounding_math for initializers.
1676 * simplify-rtx.c (simplify_binary_operation): Likewise, don't
1677 constant fold FP operations with flag_rounding_math if the
1678 result may depend upon the run-time rounding mode.
1680 2005-01-21 Tom Tromey <tromey@redhat.com>
1682 * c-cppbuiltin.c (define__GNUC__): Correct assertion.
1684 2005-01-21 Kazu Hirata <kazu@cs.umass.edu>
1686 * tree-cfg.c: Fix comment typos.
1688 2005-01-21 J"orn Rennecke <joern.rennecke@st.com>
1690 * passes.c (rest_of_handle_flow2): Close / open the flow2 dump file
1691 around the call to rest_of_handle_branch_target_load_optimize.
1693 2005-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
1695 * config.gcc (sparc-*-netbsdelf*): Include dbxelf.h
1696 (sparc64-*-openbsd*): Likewise.
1697 (sparclite-*-elf*): Include tm-dwarf2.h.
1698 (sparc86x-*-elf*): Likewise.
1699 (sparc64-*-elf*): Likewise.
1700 * config/sparc/linux64.h (DWARF2_DEBUGGING_INFO): Delete.
1701 (DBX_DEBUGGING_INFO): Likewise.
1702 * config/sparc/liteelf.h (DWARF2_DEBUGGING_INFO): Likewise.
1703 (PREFERRED_DEBUGGING_TYPE): Likewise.
1704 * config/sparc/netbsd-elf.h (PREFERRED_DEBUGGING_TYPE): Likewise.
1705 * config/sparc/sp64-elf.h (DBX_DEBUGGING_INFO): Likewise.
1706 (PREFERRED_DEBUGGING_TYPE): Likewise.
1707 * config/sparc/sp86x-elf.h (DWARF2_DEBUGGING_INFO): Likewise.
1708 (PREFERRED_DEBUGGING_TYPE): Likewise.
1709 * config/sparc/sparc.h (DBX_DEBUGGING_INFO): Likewise.
1711 2005-01-21 Hans-Peter Nilsson <hp@bitrange.com>
1714 * combine.c (combine_simplify_rtx): Revert change of 2004-12-31.
1716 2005-01-20 Paul Brook <paul@codesourcery.com>
1718 * doc/rtl.texi: Document value extension requirements for CONST_INT.
1720 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
1722 * tree-cfg.c (tree_verify_flow_info): Check that a nonlocal
1723 label is first in a sequence of labels.
1725 * tree-cfg.c (tree_verify_flow_info): Fix a typo.
1727 2005-01-20 Janis Johnson <janis187@us.ibm.com>
1728 Giovanni Bajo <giovannibajo@gcc.gnu.org>
1730 * doc/sourcebuild.texi (Test Directives): New.
1732 2005-01-20 Roger Sayle <roger@eyesopen.com>
1734 * real.c (real_floor): Don't leave the result uninitialized when
1735 mode is VOIDMode, but return the unrounded intermediate value.
1736 (real_ceil): Likewise.
1738 2005-01-20 Paul Brook <paul@codesourcery.com>
1740 * config/arm/arm.md (insv): Use gen_int_mode.
1742 2005-01-20 Eric Botcazou <ebotcazou@libertysurf.fr>
1744 * config/sparc/freebsd.h (ASM_OUTPUT_INTERNAL_LABELREF): Delete.
1745 * config/sparc/linux.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1746 * config/sparc/linux64.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1747 * config/sparc/netbsd-elf.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1748 * config/sparc/sol2.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1750 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
1753 * dwarf2out.c (concat_loc_descriptor): We don't know we can use
1754 fbreg, so we have to assume we can't.
1755 (loc_descriptor_from_tree_1): Ditto.
1756 (containing_function_has_frame_base): New function.
1757 (add_location_or_const_value_attribute): Use it.
1758 Also try to generate a frame_base from a single element location
1761 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
1763 PR tree-optimization/15349
1764 * timevar.def (TV_TREE_MERGE_PHI): New.
1765 * tree-cfg.c (tree_forwarder_block_p): Add a new argument
1767 (remove_forwarder_block, cleanup_forwarder_blocks): Adjust the
1768 calls to tree_forwarder_block_p.
1769 (remove_forwarder_block_with_phi, merge_phi_nodes,
1770 gate_merge_phi, pass_merge_phi): New.
1771 * tree-optimize.c (init_tree_optimization_passes): Add
1773 * tree-pass.h: Add an extern for pass_merge_phi;
1775 2005-01-20 Richard Henderson <rth@redhat.com>
1778 * config/i386/emmintrin.h (_mm_castpd_ps, _mm_castpd_si128): New.
1779 (_mm_castps_pd, _mm_castps_si128): New.
1780 (_mm_castsi128_ps, _mm_castsi128_pd): New.
1782 2005-01-20 Richard Henderson <rth@redhat.com>
1785 * config/i386/mmintrin.h (_mm_cvtsi32_si64): Use
1786 __builtin_ia32_vec_init_v2si.
1787 (_mm_cvtsi64_si32): Use __builtin_ia32_vec_ext_v2si.
1788 * config/i386/i386.c (IX86_BUILTIN_VEC_EXT_V2SI): New.
1789 (ix86_init_mmx_sse_builtins): Create it.
1790 (ix86_expand_builtin): Expand it.
1791 (ix86_expand_vector_set): Handle V2SFmode and V2SImode.
1792 * config/i386/mmx.md (vec_extractv2sf_0, vec_extractv2sf_1): New.
1793 (vec_extractv2si_0, vec_extractv2si_1): New.
1795 2005-01-20 Richard Henderson <rth@redhat.com>
1798 * config/i386/i386.c (ix86_expand_push): New.
1799 * config/i386/mmx.md (push<MMXMODE>1): New.
1800 * config/i386/sse.md (push<SSEMODE>1): New.
1801 * config/i386/i386-protos.h: Update.
1803 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
1805 * params.def, params.h, predict.c, tree-ssa-loop-im.c: Update
1808 2005-01-20 Richard Henderson <rth@redhat.com>
1810 * rtl.def (CONST_VECTOR): Use RTX_CONST_OBJ.
1811 * rtl.h (CONSTANT_P): Don't special case CONST_VECTOR.
1813 2005-01-19 Richard Henderson <rth@redhat.com>
1816 * config/i386/i386.c (ix86_expand_vector_move_misalign): Convert
1817 to V4SFmode in SSE1 fallback load path.
1819 2005-01-19 Richard Henderson <rth@redhat.com>
1821 * config/i386/i386.c (ix86_expand_vector_init_one_var): Fix typo
1822 in QImode expansion to ix86_expand_vector_set.
1823 (ix86_expand_vector_init_general): Fix typo in V8HImode recursive call.
1825 2005-01-19 Richard Henderson <rth@redhat.com>
1828 * config/i386/i386.c (ix86_preferred_reload_class): Return a proper
1829 subclass of the input class.
1830 (ix86_secondary_memory_needed): Always true for cross-MMX classes.
1831 Always true for cross-SSE1 classes. Rationalize conditionals.
1832 * config/i386/i386.h (SSE_CLASS_P, MMX_CLASS_P): Use straight equality.
1833 * config/i386/i386.md (movsi_1): Add MMX/SSE zeros. Fix alternatives
1834 for SSE1. Don't check TARGET_INTER_UNIT_MOVES.
1835 (movdi_2): Add MMX/SSE zeros.
1836 (movdi_1_rex64): Likewise. Don't check TARGET_INTER_UNIT_MOVES.
1837 (movsf_1): Don't check TARGET_INTER_UNIT_MOVES.
1838 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
1839 (movsi_1_nointernunit, movdi_1_rex64_nointerunit): Remove.
1840 (movsf_1_nointerunit, zero_extendsidi2_32_1): Remove.
1841 (zero_extendsidi2_rex64_1): Remove.
1842 (MOV0 peephole): Check GENERAL_REG_P.
1844 2005-01-19 Richard Henderson <rth@redhat.com>
1847 * config/i386/i386.c (ix86_expand_vector_set): Fix third and fourth
1849 (ix86_expand_vector_extract): Likewise.
1851 2005-01-19 Richard Henderson <rth@redhat.com>
1854 * expr.c (emit_move_change_mode): New argument force; use
1855 simplify_gen_subreg if true.
1856 (emit_move_via_alt_mode): Merge into ...
1857 (emit_move_via_integer): ... here.
1858 (emit_move_ccmode): Use emit_move_change_mode directly.
1860 2005-01-19 Richard Henderson <rth@redhat.com>
1863 * config/alpha/alpha.c (alpha_rtx_costs): Handle HIGH.
1864 (alpha_preferred_reload_class): Handle CONST_VECTOR.
1865 (alpha_emit_set_const_1): Add no_output parameter; don't emit
1867 (alpha_emit_set_const): Likewise. Make static.
1868 (alpha_emit_set_long_const): Make static.
1869 (alpha_extract_integer): Split out from alpha_expand_mov.
1870 (alpha_split_const_mov): Likewise.
1871 (alpha_expand_mov): Use them. Handle CONST_VECTOR.
1872 (alpha_legitimate_constant_p): New.
1873 * config/alpha/alpha-protos.h: Update.
1874 * config/alpha/alpha.h (REGISTER_MOVE_COST): Correct fp<->gp cost.
1875 (LEGITIMATE_CONSTANT_P): Re-implement with a function.
1876 * config/alpha/alpha.md (movsi): Add n alternative.
1877 (movsi_nt_vms, movdi_er_nofix, movdi_er_fix, movdi_fix): Likewise.
1878 (mov<VEC>_fix, mov<VEC>_nofix): Add i alternative.
1879 (splitters for all of the above): Use alpha_split_const_mov.
1880 * config/alpha/predicates.md (non_add_const_operand): New.
1881 (non_zero_const_operand): New.
1882 (input_operand): Use alpha_legitimate_constant_p after reload.
1884 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
1886 PR tree-optimization/19038
1887 * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): New function.
1888 (add_candidate): Add ivs with increment in latch only if
1889 allow_ip_end_pos_p is true.
1890 (determine_iv_cost): Use empty_block_p.
1892 2005-01-19 Daniel Berlin <dberlin@dberlin.org>
1894 * cfganal.c (compute_dominance_frontiers_1): Replace with new algorithm
1895 (compute_dominance_frontiers): Ditto.
1897 2005-01-19 Ralf Corsepius <ralf.corsepius@rtems.org>
1900 * config/sh/t-rtems: New.
1901 * config.gcc (sh-*-rtems*): Reflect having added config/sh/t-rtems.
1903 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
1905 * tree-ssa-ccp.c (maybe_fold_offset_to_component_ref): Always subtract
1906 the offset of the selected field.
1908 2005-01-19 Kazu Hirata <kazu@cs.umass.edu>
1910 * tree-cfg.c (remove_forwarder_block): Fix the check to
1911 prevent a nonlocal label from appearing in the middle of a
1914 2005-01-19 Hans-Peter Nilsson <hp@axis.com>
1916 PR rtl-optimization/19462
1917 * reorg.c (find_end_label): Create return insn only if
1918 current_function_epilogue_delay_list is empty.
1920 2005-01-19 Paolo Bonzini <bonzini@gnu.org>
1922 * doc/passes.texi: Remove paragraph mentioning new-ra.
1924 2005-01-19 Jakub Jelinek <jakub@redhat.com>
1926 PR rtl-optimization/15139
1927 * combine.c: Include params.h.
1928 (count_rtxs): New function.
1929 (record_value_for_reg): If replace_rtx would replace at least
1930 2 occurrences of REG in VALUE and TEM is really large, replace REG with
1931 (clobber (const_int 0)) instead of TEM.
1932 * params.def (PARAM_MAX_LAST_VALUE_RTL): New.
1933 * params.h (MAX_LAST_VALUE_RTL): New.
1934 * Makefile.in (combine.o): Depend on $(PARAMS_H).
1935 * doc/invoke.texi (--param max-last-value-rtl=N): Document.
1938 * c-typeck.c (digest_init): Only call build_vector if all constructor
1939 elements are *_CST nodes.
1940 * gimplify.c (gimplify_init_constructor): Likewise.
1943 * c-typeck.c (digest_init): Only call build_vector if inside_init
1946 2005-01-18 Tobias Schl"uter <tobias.schlueter@physik.uni-muenchen.de>
1948 * toplev.c (init_asm_output): Remove dead #ifdef.
1950 2005-01-18 Kaz Kojima <kkojima@gcc.gnu.org>
1952 * config/sh/linux.h (TARGET_C99_FUNCTIONS): Define.
1954 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
1956 * calls.c (expand_call): Check DECL_BUILT_IN_CLASS before
1957 accessing DECL_FUNCTION_CODE.
1958 * dojump.c (do_jump): Likewise.
1959 * gimplify.c (gimplify_call_expr): Likewise.
1960 * predict.c (expr_expected_value): Likewise.
1961 (strip_builtin_expect): Likewise.
1962 * tree-inline.c (estimate_num_insns_1): Likewise.
1963 * tree-ssa-loop-im.c (stmt_cost): Likewise
1964 * fold-const.c (fold): Test for BUILT_IN_NORMAL.
1965 (tree_expr_nonnegative_p): Likewise.
1967 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
1969 * config/sparc/sparc.c (load_pic_register): Emit the appropriate
1970 variant of the load_pcrel_sym pattern.
1971 * config/sparc/sparc.md (P macro): Move to the top.
1972 (load_pcrel_sym): Macroize using P.
1973 (save_register_window): Likewise.
1975 2005-01-18 Aldy Hernandez <aldyh@redhat.com>
1977 * config/rs6000/rs6000.md ("sunordered"): Disable for e500.
1980 2005-01-18 Andrew Pinski <pinskia@physics.uc.edu>
1983 * c-typeck.c (build_asm_expr): Strip nops off
1984 input memory operands.
1986 2005-01-18 David Edelsohn <edelsohn@gnu.org>
1988 * config/rs6000/aix43.h (CPLUSPLUS_CPP_SPEC): Delete
1989 _XOPEN_SOURCE, _XOPEN_SOURCE_EXTENDED, _LARGE_FILE_API.
1990 * config/rs6000/aix51.h (CPLUSPLUS_CPP_SPEC): Same.
1991 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Same.
1993 2005-01-18 Joel Sherrill <joel.sherrill@oarcorp.com>
1996 * config/i386/i386.c (override_options): If the 80387 is disabled,
1997 then do not return FP values using FP registers.
1999 2005-01-18 Kazu Hirata <kazu@cs.umass.edu>
2001 * basic-block.h, c-common.c, c-cppbuiltin.c, c-lang.c,
2002 c-tree.h, cfgbuild.c, cgraph.c, cgraph.h, collect2.c,
2003 combine.c, config.gcc, coverage.h, cse.c, cselib.c,
2004 defaults.h, df.c, dwarf2asm.c, dwarf2out.c, explow.c, expr.c,
2005 flow.c, fold-const.c, gcse.c, ggc-page.c, gimple-low.c,
2006 gimplify.c, ifcvt.c, langhooks-def.h, lcm.c, optabs.h,
2007 output.h, postreload-gcse.c, postreload.c, recog.c,
2008 resource.c, rtl.def, rtlanal.c, sched-deps.c, sched-rgn.c,
2009 targhooks.h, toplev.c, tree-data-ref.c, tree-eh.c,
2010 tree-flow.h, tree-loop-linear.c, tree-mudflap.h, tree-nrv.c,
2011 tree-optimize.c, tree-outof-ssa.c, tree-pass.h,
2012 tree-scalar-evolution.c, tree-ssa-copy.c, tree-ssa-dce.c,
2013 tree-ssa-dse.c, tree-ssa-loop-ivopts.c, tree-ssa-operands.c,
2014 tree-ssa-pre.c, tree-ssa.c, tree-vectorizer.c, tree.def,
2015 unwind-dw2-fde-darwin.c, var-tracking.c: Update copyright.
2017 * config/darwin.c, config/darwin.h, config/freebsd-spec.h,
2018 config/alpha/alpha.c, config/arm/vfp.md,
2019 config/i386/emmintrin.h, config/i386/i386-modes.def,
2020 config/i386/i386-protos.h, config/i386/i386.h,
2021 config/i386/pmmintrin.h, config/i386/ppro.md,
2022 config/i386/predicates.md, config/i386/xmmintrin.h,
2023 config/i860/i860.c, config/ia64/ia64-modes.def,
2024 config/ia64/ia64-protos.h, config/ia64/ia64.h,
2025 config/ia64/ia64.md, config/ia64/itanium1.md,
2026 config/ia64/itanium2.md, config/ia64/predicates.md,
2027 config/iq2000/iq2000.h, config/mips/linux64.h,
2028 config/rs6000/aix41.h, config/rs6000/aix43.h,
2029 config/rs6000/aix52.h, config/rs6000/darwin-fallback.c,
2030 config/rs6000/rs6000.c, config/rs6000/rs6000.h,
2031 config/rs6000/spe.md, config/sh/sh.md, config/sparc/linux.h,
2032 config/sparc/linux64.h, config/sparc/litecoff.h,
2033 config/sparc/sp64-elf.h, config/sparc/sparc.c,
2034 config/sparc/sparc.h, config/sparc/sparc.md,
2035 config/sparc/sysv4-only.h: Update copyright.
2037 2005-01-18 Richard Henderson <rth@redhat.com>
2040 * config/i386/i386.c (ix86_expand_fp_movcc): Fail for LTGT and UNEQ.
2042 2005-01-18 Marc Espie <espie@openbsd.org>
2043 * config/i386/openbsdelf.h: Typo.
2045 2005-01-18 Richard Henderson <rth@redhat.com>
2047 * config/i386/i386.c (ix86_expand_fp_movcc): Remove TARGET_IEEE_FP
2048 special case for sse.
2049 * config/i386/predicates.md (sse_comparison_operator): Likewise.
2051 2005-01-18 Dorit Naishlos <dorit@il.ibm.com>
2053 * tree-inline.c (estimate_num_insns_1): Added cases for
2054 ALIGN_INDIRECT_REF, MISALIGNED_INDIRECT_REF, and REALIGN_LOAD_EXPR.
2056 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
2058 PR rtl-optimization/19296
2059 * combine.c (simplify_comparison): Rewrite the condition under
2060 which a non-paradoxical SUBREG of a PLUS can be lifted when
2061 compared against a constant.
2063 2005-01-18 Andi Kleen <ak@muc.de>
2065 * c-typeck.c: (convert_for_assignment): Check warn_pointer_sign.
2066 * c.opt (-Wpointer-sign): Add.
2067 * doc/invoke.texi: (-Wpointer-sign): Add.
2069 2005-01-18 Uros Bizjak <uros@kss-loka.si>
2071 * config/i386/i386.c (override_options): Revert 2004-11-24 change.
2072 * config/i386/i386.md (fmodsf3, dremsf3, *sinsf2, *cossf2, sincossf3,
2073 *tansf3_1, tansf2, atan2sf3_1, atan2sf3, atansf2, asinsf2, acossf2,
2074 logsf2, log10sf2, log2sf2, log1psf2, logbsf2, ilogbsf2, expsf2,
2075 exp10sf2, exp2sf2, expm1sf2, rintsf2, floorsf2, ceilsf2, btruncsf2,
2076 nearbyintsf2): Disable for TARGET_SSE_MATH. Leave patterns enabled
2077 for TARGET_MIX_SSE_I387.
2078 (fmoddf3, dremdf3, *sindf2, *sinextendsfdf2, *cosdf2, *cosextendsfdf2,
2079 sincosdf3, *sincosextendsfdf3, *tandf3_1, tandf2, atan2df3_1,
2080 atan2df3, atandf2, asindf2, acosdf2, logdf2, log10df2, log2df2,
2081 log1pdf2, logbdf2, expdf2, exp10df2, exp2df2, expm1df2, rintdf2,
2082 floordf2, ceildf2, btruncdf2, nearbyintdf2): Disable for
2083 (TARGET_SSE2 && TARGET_SSE_MATH). Leave patterns enabled for
2084 TARGET_MIX_SSE_I387.
2085 (atan2sf3, atan2df3, atan2xf3): Remove register constraints
2088 2005-01-18 Uros Bizjak <uros@kss-loka.si>
2091 * config/i386/mmx.md (*movv2sf_internal_rex64, *movv2sf_internal):
2092 Add movaps alternative for xmm reg->reg move.
2094 2005-01-17 Jeff Law <law@redhat.com>
2096 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Clear SSA_NAME_VALUE
2097 at the end of the main DOM loop rather than just before DOM exits.
2099 2005-01-17 Diego Novillo <dnovillo@redhat.com>
2101 PR tree-optimization/19121
2102 * tree-ssa-alias.c (compute_flow_sensitive_aliasing): When
2103 adding aliases to a name tag, also add them to the pointer's
2105 * tree-ssa-copy.c (merge_alias_info): Do not merge flow
2106 sensitive alias info at all. Only check that the two pointers
2107 have compatible pointed-to sets.
2108 * tree-ssa.c (verify_name_tags): Verify that the alias set of
2109 a pointer's type tag is a superset of the alias set of the
2112 2005-01-17 James E Wilson <wilson@specifixinc.com>
2115 * config/ia64/ia64.md (movxf): Handle general register source. Adjust
2116 comment to document why.
2118 2005-01-17 Richard Henderson <rth@redhat.com>
2120 * config/i386/sse.md (smaxv4sf3_finite, sse_vmsmaxv4sf3_finite,
2121 sminv4sf3_finite, sse_vmsminv4sf3_finite, smaxv2df3_finite,
2122 sse2_vmsmaxv2df3_finite, sminv2df3_finite,
2123 sse2_vmsminv2df3_finite): New.
2124 (smaxv4sf3, sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, smaxv2df3,
2125 sse2_vmsmaxv2df3, sminv2df3, sse2_vmsminv2df3): Remove commutative.
2126 Force op1 into register.
2128 2005-01-17 Kelley Cook <kcook@gcc.gnu.org>
2130 * Makefile.in (STAGEMOVESTUFF): Stage all the stamp files to prevent
2131 bubblestrap from always rebuilding build directory.
2133 2005-01-17 Ulrich Weigand <uweigand@de.ibm.com>
2135 * config/s390/s390.md ("*clc"): Remove incorrect '='.
2136 ("*cmpmem_short"): Likewise.
2138 2005-01-17 Steven Bosscher <stevenb@suse.de>
2140 * basic-block.h: Document BB_* flags.
2141 * regrename.c (copyprop_hardreg_forward): Don't use BB_VISITED,
2142 use an sbitmap instead.
2143 * sched-rgn.c (compute_trg_info): Likewise.
2145 2005-01-17 Richard Sandiford <rsandifo@redhat.com>
2147 * config.gcc (mips64*-*-linux*): Set the default abi to n32. Remove
2148 redundant target_cpu_default and extra_parts lines. Use the default
2150 * config/mips/linux64.h (DRIVER_DEFAULT_ABI_SELF_SPEC): Delete.
2151 (SUBTARGET_EXTRA_SPECS): Delete.
2152 (DRIVER_SELF_SPECS): Remove architecture lines. Set the ABI directly.
2153 (SUBTARGET_ASM_SPEC): Remove -mabi=n32 and -mabi=64 mappings.
2154 * config/mips/t-linux64 (MULTILIB_DIRNAMES): Use "32" for -mabi=32
2155 and "n32" for -mabi=n32.
2157 2005-01-17 Ian Lance Taylor <ian@airs.com>
2160 * c-typeck.c (build_c_cast): Revert patch of 2003-02-16: don't
2161 fold constant variables into initial values.
2163 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
2165 * tree-cfg.c (tree_can_merge_blocks_p): Reorder two checks.
2167 * tree-cfg.c (tree_forwarder_block_p): Speed up by walking
2168 through the statements backward.
2170 2005-01-17 Ian Lance Taylor <ian@airs.com>
2172 PR middle-end/13127:
2173 * tree-inline.c (expand_call_inline): Set TREE_NO_WARNING on
2174 a variable set to the return value of the inlined function.
2176 2005-01-17 Mark Dettinger <dettinge@de.ibm.com>
2178 * rtlanal.c (rtx_cost): Assign cost of 0 to a SUBREG
2179 when modes are tieable.
2181 2005-01-17 Ranjit Mathew <rmathew@hotmail.com>
2183 * gthr-posix.h (__gthread_active_p): Use pthread_cancel instead
2184 of pthread_create to find out if threads are enabled.
2185 * gthr-posix95.h (__gthread_active_p): Likewise.
2187 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
2189 * common.opt (-fnew-ra): Remove.
2191 * toplev.h (flag_new_regalloc): Remove.
2192 * Makefile.in (ra*.*): Don't mention.
2193 * passes.c (rest_of_handle_new_regalloc): Remove.
2194 (rest_of_handle_combine, rest_of_compilation): Always consider
2195 flag_new_regalloc as false.
2196 * doc/invoke.texi: Don't document -fnew-ra.
2198 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
2200 * bb-reorder.c (fix_edges_for_rarely_executed_code): Remove
2201 last parameter to reg_scan.
2202 * loop.c (loop_optimize): Likewise.
2203 * passes.c (rest_of_handle_tracer, rest_of_handle_if_conversion,
2204 rest_of_handle_web, rest_of_handle_cfg, rest_of_handle_jump_bypass,
2205 rest_of_handle_life, rest_of_handle_cse, rest_of_handle_cse2,
2206 rest_of_handle_gcse, rest_of_handle_loop_optimize,
2207 rest_of_handle_loop2, rest_of_handle_jump2): Likewise.
2208 * regclass.c (reg_scan): Likewise, for the declaration.
2209 * rtl.h (reg_scan): Likewise, for the prototype.
2211 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
2213 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Don't call
2214 mark_dfs_back_edges.
2216 2005-01-16 Daniel Berlin <dberlin@dberlin.org>
2218 * tree-ssa-pre.c (add_to_sets): s1 may be NULL.
2219 (compute_avail): Uses don't go in tmp_gen.
2221 2005-01-16 Steven Bosscher <stevenb@suse.de>
2223 * ggc-page.c (ggc_alloc_stat): Use __builtin_ctzl instead of a
2224 loop to look for a free slot in a page entry.
2226 2005-01-16 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2229 * defaults.h (TARGET_DEFERRED_OUTPUT_DEFS): Provide default.
2230 * toplev.c (compile_file): Call process_pending_assemble_output_defs
2231 just before targetm.asm_out.file_end.
2232 * tree.h (process_pending_assemble_output_defs): Declare.
2233 * varasm.c (assemble_output_def, process_pending_assemble_output_defs):
2235 (assemble_alias): Defer generation of assembly code for defines when
2236 TARGET_DEFERRED_OUTPUT_DEFS is true.
2237 * config/rs6000/aix41.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
2238 * config/rs6000/aix43.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
2239 * doc/tm.texi (TARGET_DEFERRED_OUTPUT_DEFS): document.
2241 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2244 * pa.c (pa_scalar_mode_supported_p): New function.
2245 (TARGET_SCALAR_MODE_SUPPORTED_P): Define.
2246 * pa.h (MIN_UNITS_PER_WORD): Add comment.
2248 * pa-protos.h (prefetch_operand): Delete.
2249 (prefetch_cc_operand, prefetch_nocc_operand): New declations.
2250 * pa.c (prefetch_operand): Delete.
2251 (prefetch_cc_operand, prefetch_nocc_operand): New functions.
2252 * pa.h (EXTRA_CONSTRAINT): Add `W' constraint.
2253 (PREDICATE_CODES): Delete prefetch_operand. Add prefetch_cc_operand
2254 and prefetch_nocc_operand.
2255 * pa.md (prefetch): Rework to avoid reload problems handling short
2256 displacements when a cache control completer needs to be provided.
2257 (prefetch_32, prefetch_64): Delete.
2258 (prefetch_cc, prefetch_nocc): New patterns.
2260 2005-01-15 David Edelsohn <edelsohn@gnu.org>
2262 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Revert previous change.
2264 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2267 * pa.c (pa_asm_output_aligned_common): Warn if specified alignment
2268 exceeds maximum alignment for global common data.
2269 * pa64-hpux.h (MAX_OFILE_ALIGNMENT): Define to 32768.
2270 * som.h (MAX_OFILE_ALIGNMENT): Likewise.
2272 2004-01-15 Roger Sayle <roger@eyesopen.com>
2274 * tree-ssa-dom.c (extract_range_from_cond): Correct condition.
2276 2004-01-15 Roger Sayle <roger@eyesopen.com>
2278 * harg-reg-set.h (reg_class_names): Prototype global array.
2279 * regclass.c (reg_class_names): Declare here and initialize to
2281 (dump_regclass): Remove local declaration of reg_class_names.
2282 (regclass): Likewise.
2283 * cfg.c (dump_flow_info): Likewise.
2284 * ra-debug.c (reg_class_names): Likewise.
2285 * regrename.c (reg_class_names): Likewise.
2286 * reload.c (reg_class_names): Likewise.
2287 * reload1.c (spill_failure): Likewise.
2288 * config/m68hc11/m68hc11.c (reg_class_names): Likewise.
2290 2005-01-15 Ulrich Weigand <uweigand@de.ibm.com>
2292 * config/s390/s390.md ("reload_outti"): Remove predicate for
2293 output operand. Abort if operand is not a MEM.
2294 ("reload_outdi", "reload_outdf"): Likewise.
2296 2005-01-15 Marc Espie <espie@openbsd.org>
2298 * config.gcc (*-*-openbsd*): Set HAS_LIBC_R for OpenBSD <= 3.2.
2299 (i[34567]86-*-openbsd*): Switch to ELF for OpenBSD >= 3.4.
2300 * config/openbsd.h: Fix C++ includes for native configurations.
2301 Add proper OS_CPP_BUILTINS. Fix libspec for recent OpenBSD.
2302 Add trampoline support.
2303 * config/i386/openbsdelf.h: New.
2305 2005-01-15 Marc Espie <espie@openbsd.org>
2307 * collect2.c (main): Explicitly parse -dynamic-linker option.
2309 2005-01-15 Jakub Jelinek <jakub@redhat.com>
2311 PR tree-optimization/19060
2312 * tree-ssa-dom.c (extract_range_from_cond) <case LT_EXPR, GT_EXPR>:
2313 Return 0 if op1 <= TYPE_MIN_VALUE () resp. op1 >= TYPE_MAX_VALUE ().
2314 (simplify_cond_and_lookup_avail_expr): Add assert for dummy == 0
2315 and handle extract_range_from_cond returning false.
2316 * fold-const.c (fold): Optimize comparisons with min/max even for
2317 width > HOST_BITS_PER_WIDE_INT.
2319 2005-01-15 Ralf Corsepius <ralf.corsepius@rtems.org>
2321 * config/mips/rtems.h (MIPS_DEFAULT_GVALUE): Set to 0.
2322 * config/mips/t-rtems (MULTILIBS_DIRNAMES,MULTILIB_OPTIONS):
2323 Remove little endian multilib variants.
2324 Add mips32 multilib variant.
2326 2005-01-15 Kazu Hirata <kazu@cs.umass.edu>
2328 * explow.c (copy_all_reg, stabilize): Remove.
2329 * expr.h: Remove the corresponding prototypes.
2331 * rtlanal.c (regs_set_between_p): Remove.
2332 * rtl.h: Remove the corresponding prototype.
2334 2005-01-14 David Edelsohn <edelsohn@gnu.org>
2336 * config/rs6000/rs6000.c (rs6k_nonimmediate_operand): Rename
2337 to rs6000_nonimmediate_operand.
2338 * config/rs6000/rs6000.h (PREDICATE_CODES): Rename
2339 rs6k_nonimmediate_operand to rs6000_nonimmediate_operand.
2340 * config/rs6000/rs6000.md (movsi_internal1): Rename predicate to
2341 rs6000_nonimmediate_operand.
2342 * config/rs6000/spe.md (movdf_e500_double): Rename predicate to
2343 rs6000_nonimmediate_operand.
2345 2005-01-14 Andrew Pinski <pinskia@physics.uc.edu>
2347 * tree-ssa-alias.c (pass_may_alias): Add TODO_verify_stmts.
2349 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
2351 * config/rs6000/rs6000.h (CLASS_MAX_NREGS): DF goes in 1 register
2353 (CANNOT_CHANGE_MODE_CLASS): Restrict DI mode changes on e500v2.
2354 (PREDICATE_CODES): Add rs6k_nonimmediate_operand.
2356 * config/rs6000/rs6000.c (invalid_e500_subreg): New.
2357 (rs6k_nonimmediate_operand): New.
2358 (rs6000_legitimate_offset_address_p): Handle DI modes on e500v2
2360 (legitimate_lo_sum_address_p): Same.
2361 (rs6000_legitimize_address): Same.
2362 (rs6000_legitimize_reload_address): Same.
2363 (rs6000_legitimate_address): Same.
2364 (spe_build_register_parallel): Pass DF and DC modes in a DI
2367 * config/rs6000/rs6000.md ("*movsi_internal1"): Change predicate
2368 to rs6k_nonimmediate_operand.
2370 * config/rs6000/spe.md ("*frob_df_di"): New.
2371 ("*frob_di_df"): New.
2372 ("*frob_di_df_2"): New.
2373 ("*mov_sidf_e500_subreg0"): New.
2374 ("*mov_sidf_e500_subreg4"): New.
2375 ("*movdf_e500_double"): Change predicate to
2376 rs6k_nonimmediate_operand.
2378 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
2380 * postreload.c (move2add_note_store): Only call
2381 trunc_int_for_mode on scalar integers.
2383 2005-01-14 Steven Bosscher <stevenb@suse.de>
2385 * tree-ssa-dce.c (visited_control_parents): New sbitmap to
2386 replace BB_VISITED uses.
2387 (find_obviously_necessary_stmts): Don't clear BB_VISITED.
2388 (propagate_necessity): Check the bitmap instead of BB_VISITED.
2389 (tree_dce_done): Free visited_control_parents.
2390 (perform_tree_ssa_dce): Allocate and clear it.
2391 * tree-ssa-pre.c (compute_antic_aux): Make non-recursive.
2392 (compute_antic): Iterate from here using a DFS. Use an sbitmap
2393 instead of BB_VISITED.
2395 2005-01-14 Kazu Hirata <kazu@cs.umass.edu>
2397 * c-tree.h, coverage.h, langhooks-def.h, optabs.h, output.h,
2398 rtl.h, targhooks.h, tree-flow.h, tree-mudflap.h, tree.h:
2399 Remove unused prototypes.
2401 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
2403 * config/rs6000/rs6000.c (rs6000_stack_info): Fix caching of SPE
2404 64-bit register usage.
2405 (spe_func_has_64bit_regs_p): Add FIXME note.
2407 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
2409 * config/rs6000/rs6000.c (rs6000_generate_compare): Replace
2410 flag_finite_math_only with flag_unsafe_math_optimizations.
2412 2005-01-14 J. D. Johnston <jjohnst@us.ibm.com>
2414 * config/s390/tpf-unwind.h (s390_fallback_frame_state): Correct
2416 (__tpf_eh_return): Copy TPF private stack area from the module
2417 boundary stack frame. Add check for when module addresses are
2418 equal, but no stub address is found.
2420 2005-01-14 Richard Earnshaw <rearnsha@arm.com>
2423 * arm.h (struct machine_function): Add call_via field.
2424 (thumb_call_via_label): Declare.
2425 * arm.c (thumb_call_via_label): New variable.
2426 (thumb_call_reg_needed): New variable.
2427 (arm_output_function_epilogue): For Thumb code, output any per-function
2428 call-indirect trampolines.
2429 (thumb_call_via_reg): New function.
2430 (arm_file_end): New function.
2431 (TARGET_ASM_FILE_END): Call arm_file_end.
2432 (aof_file_end): Likewise.
2433 * arm-protos.h (thumb_call_via_reg): Declare.
2434 * arm.md (call_reg_thumb, call_value_reg_thumb): Call
2435 thumb_call_via_reg in normal case.
2437 2005-01-14 Jakub Jelinek <jakub@redhat.com>
2440 PR rtl-optimization/19348
2441 * recog.c (peephole2_optimize): Do global life update if some peephole
2442 decides it doesn't need at least one of its inputs and that change
2443 influences liveness at the start of the basic block.
2445 * basic-block.h (EXECUTE_IF_AND_COMPL_IN_REG_SET): Needs 2 REGSET
2446 arguments instead of 1.
2448 2005-01-14 Eric Botcazou <ebotcazou@libertysurf.fr>
2451 * varasm.c (initializer_constant_valid_p) <ADDR_EXPR>: Return
2452 zero for nested functions needing a static chain or functions
2453 with a non-constant address.
2455 2005-01-13 Roger Sayle <roger@eyesopen.com>
2457 * simplify-rtx.c (simplify_binary_operation) <AND>: Optimize
2458 (and (sign_extend X) C) into (zero_extend (and X C)).
2460 2005-01-13 David O'Brien <obrien@FreeBSD.org>
2462 * config/freebsd-spec.h: Make KSE pthread lib logic the default.
2464 2005-01-13 Richard Henderson <rth@redhat.com>
2469 * config/i386/i386.md (cmpdf, cmpsf, bunordered, bordered, buneq,
2470 bunge, bungt, bunle, bunlt, bltgt): Enable for TARGET_SSE_MATH,
2471 not just TARGET_SSE.
2472 (cmpfp_i_387): Rename from cmpfp_i. Move after sse patterns.
2473 (cmpfp_i_mixed): Rename from cmpfp_i_sse; use for TARGET_MIX_SSE_I387.
2474 (cmpfp_i_sse): Rename from cmpfp_i_sse_only; use for TARGET_SSE_MATH.
2475 (cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387): Similarly.
2476 (fp_jcc_1_mixed, fp_jcc_1_sse, fp_jcc_1_387): Similarly.
2477 (fp_jcc_2_mixed, fp_jcc_2_sse, fp_jcc_2_387): Similarly.
2478 (fp_jcc_3_387, fp_jcc_4_387, fp_jcc_5_387, fp_jcc_6_387,
2479 fp_jcc_7_387, fp_jcc_8_387): Rename from fp_jcc_N.
2480 (movdicc_c_rex64): Rename with '*'.
2481 (movsfcc, movdfcc): Add checks for 387 and sse math to condition.
2482 (movsfcc_1_sse_min, movsfcc_1_sse_max, movsfcc_1_sse): New.
2483 (movsfcc_1_387): Rename from movsfcc_1.
2484 (movdfcc_1_sse_min, movdfcc_1_sse_max, movdfcc_1_sse): New.
2485 (movdfcc_1, movdfcc_1_rex64): Add check for 387.
2486 (sminsf3, smaxsf3, smindf3, smaxdf3): New.
2487 (minsf3, minsf, minsf_nonieee, minsf_sse, mindf3, mindf,
2488 mindf_nonieee, mindf_sse, maxsf3, maxsf, maxsf_nonieee, maxsf_sse,
2489 maxdf3, maxdf, maxdf_nonieee, maxdf_sse, sse_movsfcc, sse_movsfcc_eq,
2490 sse_movdfcc, sse_movdfcc_eq, sse_movsfcc_const0_1,
2491 sse_movsfcc_const0_2, sse_movsfcc_const0_3, sse_movsfcc_const0_4,
2492 sse_movdfcc_const0_1, sse_movdfcc_const0_2, sse_movdfcc_const0_3,
2493 sse_movdfcc_const0_4): Remove.
2494 * config/i386/i386.c (ix86_expand_fp_movcc): For TARGET_SSE_MATH,
2495 recognize min/max early. Update for changed sse cmove patterns.
2496 (ix86_split_sse_movcc): New.
2497 * config/i386/i386-protos.h: Update.
2499 2005-01-13 Steven Bosscher <stevenb@suse.de>
2501 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
2502 (fix_stmt_v_may_defs): Likewise.
2504 2005-01-13 Richard Henderson <rth@redhat.com>
2506 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
2507 for converting the mask.
2509 2005-01-13 David Edelsohn <edelsohn@gnu.org>
2511 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
2514 2005-01-13 Richard Henderson <rth@redhat.com>
2516 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
2517 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
2518 IX86_BUILTIN_MOVDQ2Q): Remove.
2519 (IX86_BUILTIN_VEC_EXT_V4SI): New.
2520 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
2521 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
2523 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
2524 (sse2_pextrw): Fix immediate constraint.
2525 (sse2_loadq, sse2_loadq_rex64): Remove.
2526 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
2527 Use __builtin_ia32_vec_ext_<size>.
2528 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
2530 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
2532 * function.c (assign_parm_setup_block): Look inside original
2533 entry_parm when inspecting PARALLEL.
2535 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
2536 Joel Sherrill <joel@oarcorp.com>
2539 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
2540 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
2541 rtems_gxx_recursive_mutex_init.
2542 (__gthread_recursive_mutex_lock): New function.
2543 (__gthread_recursive_mutex_trylock): Likewise.
2544 (__gthread_recursive_mutex_unlock): Likewise.
2546 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
2548 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
2550 2005-01-13 Jan Beulich <jbeulich@novell.com>
2552 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
2553 (ashrti3_internal): Indicate output is early clobber. Generate result
2554 into output rather than first input. Use move for low word of output
2555 if shift count is exactly 64.
2556 (lshrti3_internal): Likewise.
2558 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
2562 * reload1.c (delete_output_reload): Don't delete an output reload
2563 if the pseudo lives longer than a single basic block. Adjust and
2564 improve wording of comment.
2566 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
2569 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
2570 of TYPE_SIZE where we mean the number of bytes.
2572 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
2574 * function.c (assign_parm_setup_block): Relax condition on
2575 multi-register optimization.
2577 2005-01-12 Nick Clifton <nickc@redhat.com>
2579 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
2580 an "in_delay_slot" attribute of "no" to prevent them being used in
2581 delay slots. This is forbidden because they might generate
2584 2005-01-12 Alan Modra <amodra@bigpond.net.au>
2587 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
2590 2005-01-12 Nick Clifton <nickc@redhat.com>
2592 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
2593 passed on to GAS) but do not define, as GAS no longer supports or
2594 needs the -m2000 option.
2596 2005-01-11 Roger Sayle <roger@eyesopen.com>
2598 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
2599 type as args of unprototyped builtins aren't automatically promoted.
2600 (fold_builtin_abs): Likewise.
2602 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2604 * Makefile.in: Set a `build-warn' variable.
2606 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
2608 PR tree-optimization/17949
2609 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
2610 (find_interesting_uses_address): Use it.
2612 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
2614 * regrename.c (kill_value): Handle subreg's that won't simplify.
2616 2005-01-11 Richard Henderson <rth@redhat.com>
2619 * config/i386/i386.h (enum ix86_builtins): Move ...
2620 * config/i386/i386.c: ... here.
2621 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
2622 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
2623 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
2624 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
2625 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
2626 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
2627 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
2628 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
2629 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
2630 IX86_BUILTIN_LOADDDUP): Remove.
2631 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
2632 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
2633 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
2634 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
2635 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
2636 (ix86_init_builtins): Make static.
2637 (ix86_init_mmx_sse_builtins): Update for changed builtins.
2638 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
2639 if all the modes match. Otherwise, fake it.
2640 (get_element_number, ix86_expand_vec_init_builtin,
2641 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
2642 (ix86_expand_builtin): Make static. Update for changed builtins.
2643 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
2644 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
2645 (ix86_expand_vector_init_duplicate): New.
2646 (ix86_expand_vector_init_low_nonzero): New.
2647 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
2648 Split out from ix86_expand_vector_init; handle integer modes.
2649 (ix86_expand_vector_init): Use them.
2650 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
2651 * config/i386/i386-protos.h: Update.
2652 * config/i386/predicates.md (reg_or_0_operand): New.
2653 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
2654 (movv2sf_internal): Likewise. And a splitter to match them all.
2655 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
2656 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
2657 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
2658 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
2659 (mmx_pinsrw): Fix operand ordering.
2660 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
2661 rather than sse_loadss expander.
2662 (movv2df splitter): Similarly.
2663 (sse_loadss, sse_loadlss): Remove.
2664 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
2665 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
2666 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
2667 (sse3_movddup): Rename with '*'.
2668 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
2669 (sse2_loadsd): Remove.
2670 (vec_dupv2df_sse3): Rename from sse3_loadddup.
2671 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
2672 (sse2_pinsrw): Fix argument ordering.
2673 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
2674 (sse2_stored): Remove 'r' destination.
2675 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
2676 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
2677 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
2678 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
2679 vec_extractv16qi, vec_initv16qi): New.
2681 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
2682 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
2683 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
2685 (_mm_load_pd, _mm_store_pd): Use plain dereference.
2686 (_mm_load_si128, _mm_store_si128): Likewise.
2687 (_mm_load1_pd): Use _mm_set1_pd.
2688 (_mm_load_sd): Use _mm_set_sd.
2689 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
2690 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
2691 (_mm_set_epi64): Use _mm_set_epi64x.
2692 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
2693 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
2694 _mm_setr_epi8): Use _mm_set_foo form.
2695 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
2696 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
2697 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
2698 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
2699 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
2700 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
2701 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
2702 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
2703 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
2704 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
2705 (_mm_movedup_pd): Use _mm_shuffle_pd.
2706 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
2707 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
2708 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
2709 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
2711 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
2712 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
2713 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
2714 (_mm_store_ps): Use raw dereference.
2715 (_mm_store1_ps): Use _mm_storeu_ps.
2716 (_mm_storer_ps): Use _mm_store_ps.
2717 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
2718 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
2720 2005-01-11 Stan Shebs <shebs@apple.com>
2722 * config/rs6000/rs6000.c (machopic_output_stub): Issue
2723 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
2725 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
2728 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
2729 Skip all DECLs except for FIELD_DECLs.
2731 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
2733 * config/s390/s390.c (override_options): Return error if
2734 -mbackchain, -mpacked-stack and -mhard-float are used together.
2735 (s390_va_start): Remove the backchain && packed-stack special case.
2736 (s390_gimplify_va_arg): Likewise.
2737 * doc/invoke.texi: Remove the ABI incompatibility note.
2739 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
2741 * config/s390/s390.c (struct s390_frame_layout): Remove
2743 (s390_frame_info, s390_emit_prologue): Replace occurrences of
2744 save_backchain_p with TARGET_BACKCHAIN.
2746 2005-01-11 Alan Modra <amodra@bigpond.net.au>
2749 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
2750 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
2751 * function.c (locate_and_pad_parm): Set new field.
2752 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
2753 Tweak where_pad test to include "none". Always set mem align for
2755 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
2756 not sufficient for type.
2757 (assign_parm_setup_block): If stack_parm is zero on entry, always
2758 make a new stack local. Block move old stack parm if necessary
2759 to new aligned stack local.
2760 (assign_parm_setup_stack): Use a block move to handle
2761 potentially misaligned entry_parm.
2762 (assign_parms_unsplit_complex): Specify required alignment when
2763 creating stack local.
2764 * calls.c (compute_argument_addresses): Override alignment of stack
2765 arg calculated from its type with the alignment given by
2766 FUNCTION_ARG_BOUNDARY.
2767 (store_one_arg): Likewise.
2769 2005-01-11 Jan Beulich <jbeulich@novell.com>
2771 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
2772 Change respective itanium_class attribute to ialu.
2773 (shladdp4_internal): New.
2774 * config/ia64/predicates.md (shladd_log2_operand): New.
2776 2005-01-11 Richard Henderson <rth@redhat.com>
2778 * expr.c (store_constructor): Use rtvec_alloc instead of
2779 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
2781 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
2783 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
2785 2005-01-11 Alan Modra <amodra@bigpond.net.au>
2787 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
2788 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
2789 (TARGET_RWRELOC): Define.
2790 (ia64_rwreloc_section_type_flags): Delete.
2791 (ia64_section_type_flags): New function.
2792 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
2793 (TARGET_RWRELOC): Define.
2795 2005-01-10 David Mosberger <davidm@hpl.hp.com>
2798 * config/ia64/ia64.c (process_set): For alloc insn, only call
2799 process_epilogue is !frame_pointer_needed.
2801 2005-01-10 Roger Sayle <roger@eyesopen.com>
2804 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
2805 unary operator and can't be treated as a binary/comparison operator.
2807 2005-01-10 Richard Henderson <rth@redhat.com>
2809 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
2810 (ix86_return_in_memory): Likewise.
2811 (function_arg_advance): Likewise. Mirror structure in function_arg
2812 for choosing register to advance.
2814 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
2816 * tree-vectorizer.c, tree.def: Fix comment typos.
2818 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
2820 * config/s390/s390.c (struct processor_costs): 4 new fields:
2821 ddbr, ddr, debr, der.
2822 (s390_rtx_costs): More precise handling of divide instructions.
2824 2005-01-10 David Edelsohn <edelsohn@gnu.org>
2826 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
2828 2005-01-10 Jan Beulich <jbeulich@novell.com>
2830 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
2831 names resulting from -ffunction-sections/-fdata-sections and linkonce
2833 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
2834 based on the section name. Rearrange the section name comparison logic
2835 slightly so that each section name is compared against at most once.
2837 2005-01-10 Ben Elliston <bje@au.ibm.com>
2839 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
2840 targets for which -fPIC reduces limitations on the GOT size.
2842 2005-01-09 Falk Hueffner <falk@debian.org>
2844 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
2847 2005-01-09 David Edelsohn <edelsohn@gnu.org>
2850 * collect2.c (main): Set aixrtl_flag for -brtl option.
2851 (resolve_lib_name): Search for .so file extension before .a
2854 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
2856 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
2857 (vect_analyze_data_refs_alignment): Add dump prints.
2859 2005-01-09 Ira Rosen <irar@il.ibm.com>
2861 * tree-vectorizer.c (vect_analyze_offset_expr): Use
2862 expr_invariant_in_loop_p.
2863 Initialize outputs first thing in the function.
2864 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
2865 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
2866 (vect_analyze_pointer_ref_access): Check that the initial condition of
2867 the access function is loop invariant.
2869 2005-01-09 Richard Henderson <rth@redhat.com>
2871 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
2872 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
2873 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
2874 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
2875 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
2876 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
2877 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
2878 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
2879 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
2880 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
2881 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
2882 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
2883 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
2884 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
2885 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
2886 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
2887 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
2888 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
2889 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
2890 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
2891 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
2892 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
2893 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
2895 (mmx_clrdi, pavgusb): Remove.
2896 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
2897 with leading sse_ prefix.
2898 * config/i386/sse.md: Receive them.
2899 * config/i386/mmx.md: New file.
2900 (MMXMODE12, MMXMODE24, mmxvecsize): New.
2901 (subrv2sf3): Turn into expander for normal subtraction.
2902 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
2903 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
2904 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
2905 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
2906 commutative; use ix86_binary_operator_ok.
2907 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
2908 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
2909 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
2910 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
2911 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
2912 Macroize from existing patterns; use ix86_binary_operator_ok.
2913 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
2914 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
2915 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
2916 with vec_select+vec_concat.
2917 (mmx_pshufw, mmx_pshufw_1): Likewise.
2918 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
2919 (mmx_uavgv4hi3): Model correctly.
2920 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
2921 _mm_xor_si64): Remove casts.
2923 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
2925 PR tree-optimization/19224
2926 * tree-scalar-evolution.c (get_instantiated_value,
2927 set_instantiated_value): New functions.
2928 (instantiate_parameters_1): Cache the results.
2929 (instantiate_parameters, resolve_mixers): Initialize and free
2932 2005-01-08 David Edelsohn <edelsohn@gnu.org>
2934 * config/i386/i386.md (addhi_4): Correct reference in comment.
2937 2005-01-08 Richard Henderson <rth@redhat.com>
2939 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
2940 (_mm_cvtsi128_si64x): Likewise.
2941 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
2942 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
2943 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
2944 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
2945 * config/i386/i386-protos.h: Update.
2946 * config/i386/i386.c (print_operand): Add 'H'.
2947 (ix86_fixup_binary_operands): Split out from ...
2948 (ix86_expand_binary_operator): ... here.
2949 (ix86_fixup_binary_operands_no_copy): New.
2950 (ix86_expand_fp_absneg_operator): Handle vector mode results.
2951 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
2952 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
2953 (safe_vector_operand): Use CONST0_RTX.
2954 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
2955 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
2956 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
2957 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
2959 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
2960 (IX86_BUILTIN_CMPNGESS): New.
2961 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
2962 (attr type): Add sselog1.
2963 (attr unit, attr memory): Handle it.
2964 (movti, movti_internal, movti_rex64): Move near other integer moves.
2965 (movtf, movtf_internal): Move near other fp moves.
2966 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
2967 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
2968 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
2969 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
2970 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
2971 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
2972 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
2973 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
2974 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
2975 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
2976 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
2977 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
2978 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
2979 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
2980 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
2981 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
2982 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
2983 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
2984 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
2985 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
2986 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
2987 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
2988 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
2989 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
2990 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
2991 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
2992 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
2993 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
2994 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
2995 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
2996 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
2997 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
2998 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
2999 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
3000 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
3001 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
3002 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
3003 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
3004 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
3005 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
3006 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
3007 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
3008 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
3009 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
3011 (SSEPUSH, push<SSEPUSH>): Remove.
3012 (MMXPUSH, push<MMXPUSH>): Remove.
3013 (sse_movaps, sse_movaps_1, sse_movups): Remove.
3014 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
3015 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
3016 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
3017 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
3018 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
3019 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
3020 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
3021 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
3022 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
3023 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
3024 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
3025 (ppro_sse_log_V4SF): Handle sselog1.
3026 * config/i386/predicates.md (const_0_to_1_operand): New.
3027 (const_0_to_255_mul_8_operand): New.
3028 (const_1_to_31_operand): Rename from const_int_1_31_operand.
3029 (const_2_to_3_operand, const_4_to_7_operand): New.
3030 * config/i386/sse.md: New file.
3031 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
3032 (sse_movups): Rename from sse_movups_1.
3033 (sse_loadlss): Rename from sse_loadss_1.
3034 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
3036 (negv4sf2): Use ix86_expand_fp_absneg_operator.
3037 (absv4sf2, negv2df, absv2df): New.
3038 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
3039 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
3040 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
3041 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
3042 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
3043 (sse3_addsubv4sf3): Model correctly.
3044 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
3045 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
3046 (sse_movhlps): Model with vec_select+vec_concat.
3047 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
3048 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
3049 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
3050 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
3051 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
3052 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
3053 sse2_pshufhw_1): Likewise.
3054 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
3055 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
3056 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
3057 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
3058 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
3059 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
3060 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
3061 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
3062 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
3063 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
3064 sminv8hi3): Mark commutative
3065 operands. Use ix86_binary_operator_ok.
3066 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
3067 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
3068 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
3069 sse2_punpcklqdq): Allow operand2 in memory.
3070 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
3071 sse2_movsd): Add memory alternatives.
3072 (sse_storelps): Turn expander into an insn; split after reload.
3073 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
3074 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
3076 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
3078 * configure.ac (DWARF-2 debug_line): Use objdump.
3079 * configure: Regenerate.
3081 2005-01-08 Jeff Law <law@redhat.com>
3082 Diego Novillo <dnovillo@redhat.com>
3084 PR tree-optimization/18241
3085 * tree-nrv.c (tree_nrv): Ignore volatile return values.
3086 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
3087 statements with volatile operands.
3088 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
3089 operands after marking a statement with has_volatile_ops.
3091 2005-01-08 Roger Sayle <roger@eyesopen.com>
3093 * tree.c (int_fits_type_p): Always honor integer constant
3094 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
3096 2005-01-08 Roger Sayle <roger@eyesopen.com>
3098 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
3099 the latent bug in rtl_delete_block has been resolved.
3101 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
3103 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
3104 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
3106 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
3108 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
3109 library as @multilib_dir@/$(SHLIB_SONAME).
3111 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
3113 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
3114 * configure: Regenerate.
3116 2005-01-07 Jakub Jelinek <jakub@redhat.com>
3118 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
3121 2005-01-07 David Edelsohn <edelsohn@gnu.org>
3124 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
3125 Convert non-word aligned offset address using ld/std into
3128 2005-01-07 Richard Henderson <rth@redhat.com>
3130 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
3132 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
3134 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
3136 2005-01-07 Jakub Jelinek <jakub@redhat.com>
3138 PR tree-optimization/19283
3139 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
3140 is not shorter than the original type.
3142 PR rtl-optimization/19012
3143 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
3145 PR rtl-optimization/18861
3146 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
3147 (FULL_STATE): Define.
3148 (mark_tablejump_edge): New function.
3149 (purge_dead_tablejump_edges): New function.
3150 (find_bb_boundaries): Use it.
3152 PR tree-optimization/18828
3153 * builtins.c (expand_builtin_next_arg): Remove argument and all
3154 the argument checking.
3155 (expand_builtin): Adjust caller.
3156 (expand_builtin_va_start): Likewise. Remove error for too many
3158 (fold_builtin_next_arg): Issue error for too many arguments.
3159 After checking arguments, replace them with magic arguments that
3160 prevent further checking of the args.
3162 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
3164 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
3166 2005-01-06 Roger Sayle <roger@eyesopen.com>
3169 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
3170 and %{.S:...} (and their negative variants) to test whether the
3171 input file is assembler or pre-processed-assembler independent of
3172 the actual filename extension.
3174 2005-01-06 Roger Sayle <roger@eyesopen.com>
3176 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
3177 of sign or zero extended values.
3179 2005-01-06 Geoffrey Keating <geoffk@apple.com>
3181 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
3182 for IBM long double format correctly.
3184 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
3186 Fix PR tree-optimization/18792
3188 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
3189 to first_loop_depth, and use loop depth instead of loop number.
3190 (build_classic_dir_vector): Ditto.
3191 (compute_data_dependences_for_loop): Use depth, not loop number.
3192 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
3193 number. Pass in loops, instead of loop numbers.
3194 (gather_interchange_stats): Ditto.
3195 (linear_transform_loops): Ditto.
3197 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
3200 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
3201 biv_fits_mode_p, extension_within_bounds_p): New functions.
3202 (check_ext_dependent_givs): Use them.
3204 2005-01-06 Roger Sayle <roger@eyesopen.com>
3206 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
3207 more than one barrier, in which case we should delete them all.
3209 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
3211 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
3213 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
3215 Fix PR middle-end/19286
3217 * dwarf2out.c (gen_subprogram_die): If we've already tried to
3218 output this subprogram, simply ignore this attempt to do it again.
3219 (add_abstract_origin_attribute): Don't abort trying to add the abstract
3220 origin attribute if it's not possible.
3221 (gen_block_die): Don't ignore subblocks of "unused" blocks.
3222 (decls_for_scope): Ditto.
3223 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
3224 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
3226 2005-01-05 Richard Henderson <rth@redhat.com>
3229 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
3230 (bdesc_2arg): Use it.
3231 (ix86_expand_binop_builtin): Force operands into registers
3233 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
3234 ix86_expand_sse_compare, ix86_expand_sse_comi,
3235 ix86_expand_builtin): Likewise.
3237 2005-01-05 Richard Henderson <rth@redhat.com>
3239 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
3241 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
3242 (mulv4hi3): Set itanium_class mmmul.
3243 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
3244 (fpack): Rename from fpack_sfsf.
3246 2005-01-05 Richard Henderson <rth@redhat.com>
3249 * reload1.c (do_input_reload): Restrict the optimization deleteing
3250 a previous output reload to RELOAD_FOR_INPUT.
3252 2005-01-05 Steven Bosscher <stevenb@suse.de>
3254 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
3255 comment that we fall through after case.
3256 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
3257 (distribute_links): Likewise.
3258 * cse.c (cse_insn): Likewise.
3259 * cselib.c (cselib_invalidate_mem): Likewise.
3260 * df.c: Update comments at the top of the file.
3261 (read_modify_subreg_p): Update comments here too.
3262 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
3263 * flow.c (mark_set_1): Likewise.
3264 (mark_used_regs): Likewise.
3265 * gcse.c (mems_conflict_for_gcse_p): Likewise.
3266 (canon_list_insert): Likewise.
3267 (mark_set): Likewise.
3268 (try_replace_reg): Likewise.
3269 (store_killed_in_insn): Likewise.
3270 * loop.c (count_one_set): Likewise.
3271 (basic_induction_var): Likewise.
3272 * postreload-gcse.c (find_mem_conflicts): Likewise.
3273 * postreload.c (reload_combine_note_store): Likewise.
3274 (move2add_note_store): Likewise.
3275 * reload.c (find_equiv_reg): Likewise.
3276 (mark_referenced_resources): Likewise.
3277 * rtlanal.c (set_noop_p): Likewise.
3278 (note_stores): Likewise.
3279 (note_uses): Likewise.
3280 * sched-deps.c (sched_analyze_1): Likewise.
3281 * sched-rgn.c (check_live_1): Likewise.
3282 (update_live_1): Likewise.
3283 * config/i860/i860.c: Likewise.
3285 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
3287 (ZERO_EXTRACT): Mention that this one can be an lvalue.
3289 * doc/rtl.texi: Update documentation for bit-fields and SET.
3291 2005-01-05 Roger Sayle <roger@eyesopen.com>
3293 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
3295 2005-01-05 Jan Hubicka <jh@suse.cz>
3296 Richard Henderson <rth@redhat.com>
3299 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
3302 2005-01-05 Richard Henderson <rth@redhat.com>
3305 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
3306 (sse_shufps): Change operand 3 to const_int_operand.
3307 (sse2_storelps): Fix typo in template.
3308 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
3309 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
3310 (ix86_expand_builtin): Likewise.
3312 2005-01-05 Julian Brown <julian@codesourcery.com>
3314 * config/arm/arm.c (arm_return_in_memory): Treat complex types
3315 as aggregates for AAPCS ABIs.
3317 2005-01-05 Stan Shebs <shebs@apple.com>
3319 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
3320 Mach-O getters if ppc64.
3321 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
3323 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
3324 signed four-byte field for global code case.
3325 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
3327 2005-01-05 Roger Sayle <roger@eyesopen.com>
3330 * c-common.c: Include real.h.
3331 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
3332 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
3333 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
3334 * Makefile.in (c-common.o): Update dependencies.
3336 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
3338 * c-parse.in (asm_string): Add trailing semicolon.
3340 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
3342 * c-parse.in (asm_string): New. Don't allow wide strings in
3344 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
3345 asm_string instead of STRING.
3347 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
3349 * c-typeck.c (constructor_no_implicit): Remove.
3350 (set_designator, process_init_element): Don't check
3351 constructor_no_implicit.
3353 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
3354 Kaz Kojima <kkojima@gcc.gnu.org>
3357 * lcm.c (create_pre_exit): New.
3358 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
3359 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
3361 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
3363 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
3364 -msoft-float in target_float_switch.
3365 * arm.c (arm_override_options): Fix processing of target_float_switch.
3367 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
3369 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
3372 2004-01-05 Uros Bizjak <uros@kss-loka.si>
3374 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
3375 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
3377 2005-01-04 Roger Sayle <roger@eyesopen.com>
3379 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
3380 if deleting the then-block allows the test-block to fallthru to the
3383 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
3386 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
3387 after the full declaration if the are in two different TUs.
3389 2005-01-04 Richard Henderson <rth@redhat.com>
3392 * tree-sra.c (generate_one_element_init): Just
3393 call gimplify_and_add.
3394 (generate_element_init): Record the
3395 new referenced variables and mark them for renaming
3396 and split out to ...
3397 (generate_element_init_1): This.
3398 (scalarize_init): Don't call push_gimplify_context/
3399 pop_gimplify_context.
3401 2005-01-04 Geoffrey Keating <geoffk@apple.com>
3403 * toplev.c (get_src_pwd): Handle failure of getpwd().
3405 2005-01-04 Roger Sayle <roger@eyesopen.com>
3407 * fold-const.c (fold_single_bit_test): Delete unreachable handling
3409 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
3410 all the cases handled by it are inverted by invert_truthvalue.
3412 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
3414 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
3415 (save_register_windowsi): Likewise.
3417 2005-01-04 Richard Henderson <rth@redhat.com>
3419 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
3421 2005-01-04 Uros Bizjak <uros@kss-loka.si>
3424 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
3425 any simplifications of vector mode comparison operators.
3426 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
3428 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
3429 Devang Patel <dpatel@apple.com>
3431 PR tree-optimization/18308
3432 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
3433 the operands before creating a new expression.
3434 * dojump.c (do_jump): Make drop_through_label available
3435 for all cases. Add expansion of COND_EXPR.
3437 2005-01-04 Ira Rosen <irar@il.ibm.com>
3439 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
3440 INTEGER_CST instead of TREE_CONSTANT.
3441 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
3442 instead of TREE_CONSTANT.
3443 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
3444 instead of TREE_CONSTANT.
3446 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
3448 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
3450 2005-01-04 Uros Bizjak <uros@kss-loka.si>
3453 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
3454 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
3456 2005-01-03 Richard Henderson <rth@redhat.com>
3458 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
3459 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
3460 fold_convert_const_real_from_real): Split out from ...
3461 (fold_convert_const): ... here.
3463 2005-01-03 Richard Henderson <rth@redhat.com>
3466 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
3467 (mov<MMXMODEI>_internal): Likewise.
3468 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
3469 Add V2SF case; use it for SSE1; don't use TI.
3470 (movdf_integer): Likewise.
3471 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
3473 2005-01-03 Ira Rosen <irar@il.ibm.com>
3475 * tree-vectorizer.c (vect_strip_conversions): New function.
3476 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
3477 check for binary class.
3479 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
3483 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
3485 * gimple-low.c (mark_blocks_with_used_vars): New function.
3486 (mark_blocks_with_used_subblocks): Ditto.
3487 (mark_used_blocks): Ditto.
3488 (pass_mark_used_blocks): New pass.
3489 * tree-inline.c: Include debug.h.
3490 (expand_call_inline): Call outlining_inline_function here.
3491 * tree-optimize.c (init_tree_optimization_passes): Add
3492 pass_mark_used_blocks.
3493 * tree-pass.h (pass_mark_used_blocks): New.
3494 * Makefile.in (tree-inline.o): Add debug.h dependency.
3496 2005-01-03 Geoffrey Keating <geoffk@apple.com>
3498 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
3499 VAR_DECLs to have weak_import attribute.
3501 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
3504 2004-01-03 Steven Bosscher <stevenb@suse.de>
3506 * passes.c (rest_of_compilation): Don't run regmove if only
3507 flag_expensive_optimizations, require flag_regmove instead.
3509 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
3511 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
3513 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
3514 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
3515 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
3516 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
3518 2005-01-03 Richard Henderson <rth@redhat.com>
3520 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
3521 (ia64_const_ok_for_letter_p): New.
3522 (ia64_const_double_ok_for_letter_p): New.
3523 (ia64_extra_constraint): New.
3524 (ia64_expand_vecint_compare): New.
3525 (ia64_expand_vcondu_v2si): New.
3526 (ia64_expand_vecint_cmov): New.
3527 (ia64_expand_vecint_minmax): New.
3528 (ia64_print_operand): Add 'v'.
3529 (ia64_preferred_reload_class): New.
3530 (ia64_vector_mode_supported_p): New.
3531 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
3532 (PREFERRED_RELOAD_CLASS): Move to function.
3533 (CONST_OK_FOR_LETTER_P): Move to function.
3534 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
3535 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
3536 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
3537 (EXTRA_CONSTRAINT): Move to function.
3538 * config/ia64/ia64.md: Include vect.md.
3539 (itanium_class): Add mmalua.
3541 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
3543 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
3544 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
3546 (const_int_2bit_operand): New.
3547 (fr_reg_or_0_operand): New.
3548 * config/ia64/ia64-modes.def: Add vector modes.
3549 * config/ia64/ia64-protos.h: Update.
3550 * config/ia64/vect.md: New file.
3552 2005-01-03 Richard Henderson <rth@redhat.com>
3554 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
3556 2005-01-03 Uros Bizjak <uros@kss-loka.si>
3559 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
3561 (log1pdf2): Change mode of operands[0,1] to DFmode.
3563 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
3565 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
3566 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
3567 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
3568 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
3569 SPARC_RELAXED_ORDERING.
3571 2005-01-03 Richard Henderson <rth@redhat.com>
3572 Uros Bizjak <uros@kss-loka.si>
3575 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
3576 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
3577 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
3579 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
3580 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
3581 const_pow2_1_to_128_operand): New predicates.
3583 2005-01-02 Greg McGary <greg@mcgary.org>
3585 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
3587 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
3588 object. [ptr_type]: Remove unused variable.
3589 Remember nearest addressable array-element or record-component
3590 when checking bitfield components. Tolerate empty BIND_EXPR.
3592 2005-01-01 Richard Henderson <rth@redhat.com>
3594 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
3595 that don't narrow the value. Fail for other conversions.
3597 2005-01-01 Richard Henderson <rth@redhat.com>
3600 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
3601 * c-lang.c (finish_file): Don't do it here.
3602 * objc/objc-act.c (objc_finish_file): Likewise.
3604 * cgraph.c (decl_assembler_name_equal): New.
3605 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
3606 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
3607 * cgraph.h (struct cgraph_varpool_node): Add next.
3608 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
3609 * varasm.c (assemble_alias): Mark the target as needed.
3611 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
3614 * function.c (get_arg_pointer_save_area): Use entry_of_function
3615 instead of get_insns.
3617 2005-01-01 Roger Sayle <roger@eyesopen.com>
3618 Andrew Pinski <pinskia@physics.uc.edu>
3619 James E. Wilson <wilson@specifixinc.com>
3621 PR rtl-optimization/12092
3622 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
3625 2005-01-01 Roger Sayle <roger@eyesopen.com>
3626 Olivier Hainque <hainque@act-europe.fr>
3628 * tree.c (int_fits_type_p): A narrower type always fits in a
3629 wider one, except for negative values into unsigned types.
3631 2005-01-01 Roger Sayle <roger@eyesopen.com>
3633 * tree.c (int_fits_type_p): Compare the result of force_fit_type
3634 with the original constant rather than require TREE_OVERFLOW.
3636 2005-01-01 Steven Bosscher <stevenb@suse.de>
3639 * c-decl.c (finish_function): If compiling C99, annotate the
3640 compiler generated return with the current file name and line 0.
3641 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
3642 the source line is greater than 0.
3643 (remove_bb): Likewise.
3645 See ChangeLog.12 for earlier changes.