1 2005-02-09 Zdenek Dvorak <dvorakz@suse.cz>
3 PR tree-optimization/18687
4 * tree-flow.h (find_loop_niter): Declare.
5 * tree-ssa-loop-ivcanon.c (canonicalize_loop_induction_variables):
6 Try using scev even for loops with more than one exit.
7 * tree-ssa-loop-ivopts.c (struct loop_data): Removed niter field.
8 (struct ivopts_data): Added niters field.
9 (struct nfe_cache_elt): New.
10 (nfe_hash, nfe_eq, niter_for_exit, niter_for_single_dom_exit): New
12 (tree_ssa_iv_optimize_init): Initialize niters cache.
13 (determine_number_of_iterations): Removed.
14 (find_induction_variables): Do not call determine_number_of_iterations.
15 Access niters for single exit through niter_for_single_dom_exit.
16 (add_iv_outer_candidates): Access niters for single exit through
17 niter_for_single_dom_exit.
18 (may_eliminate_iv): Take data argument. Use niter_for_exit. Do not use
19 number_of_iterations_cond.
20 (iv_period): New function.
21 (determine_use_iv_cost_condition): Pass data to may_eliminate_iv.
22 (may_replace_final_value): Take data argument. Use
23 niter_for_single_dom_exit.
24 (determine_use_iv_cost_outer): Pass data to may_replace_final_value.
25 (rewrite_use_compare): Pass data to may_eliminate_iv.
26 (rewrite_use_outer): Pass data to may_replace_final_value.
27 (free_loop_data): Clean up the niters cache.
28 (tree_ssa_iv_optimize_finalize): Free the niters cache.
29 (tree_ssa_iv_optimize_loop): Do not call loop_commit_inserts.
30 * tree-ssa-loop-niter.c (find_loop_niter): New function.
31 (find_loop_niter_by_eval): Use tree_int_cst_lt.
32 (num_ending_zeros): Moved to tree.c.
33 * tree.h (num_ending_zeros): Declare.
34 * tree.c (num_ending_zeros): Moved from tree.c.
36 2005-02-09 Richard Henderson <rth@redhat.com>
38 * builtins.c (DEF_BUILTIN): Add COND argument.
39 * tree.h (DEF_BUILTIN): Likewise.
40 * builtins.def (DEF_GCC_BUILTIN, DEF_LIB_BUILTIN, DEF_EXT_LIB_BUILTIN,
41 DEF_C94_BUILTIN, DEF_C99_BUILTIN, DEF_C99_C90RES_BUILTIN): Update to
43 (DEF_BUILTIN_STUB): New.
44 (BUILT_IN_STACK_SAVE, BUILT_IN_STACK_RESTORE, BUILT_IN_INIT_TRAMPOLINE,
45 BUILT_IN_ADJUST_TRAMPOLINE, BUILT_IN_NONLOCAL_GOTO,
46 BUILT_IN_PROFILE_FUNC_ENTER, BUILT_IN_PROFILE_FUNC_EXIT): Use it.
47 * c-common.c (DEF_BUILTIN): Add COND argument.
48 * tree.c (local_define_builtin): New.
49 (build_common_builtin_nodes): New.
51 2005-02-09 Roger Sayle <roger@eyesopen.com>
53 * fold-const.c (fold_strip_sign_ops): New function to simplify a
54 floating point expression ignoring the sign of the result.
55 (fold) <ABS_EXPR>: Use it to simplify fabs(x).
56 (fold) <MULT_EXPR>: Use it to simplify x*x.
57 * tree.h (fold_strip_sign_ops): Prototype here.
58 * builtins.c (fold_builtin_copysign): Take an additional FNDECL
59 argument. Use fold_strip_sign_ops to simplify the first argument.
60 (fold_builtin_pow): Use fold_strip_sign_ops to simplify the
61 first argument when the second argument is an even integer
62 constant, but only with -funsafe_math_optimizations.
63 (fold_builtin_1): Update call to fold_builtin_copysign.
65 2005-02-09 Ian Lance Taylor <ian@airs.com>
68 * gimple-low.c (try_catch_may_fallthru): In EH_FILTER_EXPR case,
69 just check whether EH_FILTER_FAILURE falls through.
71 2005-02-09 Andreas Krebbel <krebbel1@de.ibm.com>
73 * gcc/haifa-sched.c (schedule_block): Make queued sched group
74 insns return to ready list in the next turn.
76 2005-02-09 Richard Guenther <rguenth@gcc.gnu.org>
79 * builtins.def: New __builtin_powi[lf].
80 * builtins.c (mathfn_built_in): Handle BUILT_IN_POWI.
81 (expand_builtin_powi): New function.
82 (expand_builtin): Dispatch to expand_builtin_powi.
83 * libgcc2.h: Add prototypes for __builtin_powi[lf].
84 * libgcc2.c: Add __builtin_powi[lf] implementation.
85 * mklibgcc.in: Add __builtin_powi[lf] to lib2funcs.
86 * optabs.h: Add powi_optab.
87 * optabs.c (init_optabs): Initialize powi_optab.
88 * doc/extend.texi: Document __builtin_powi[lf].
90 2005-02-09 Dorit Naishlos <dorit@il.ibm.com>
92 * tree-vectorizer.c (vect_set_dump_settings): Check that dump_file
95 2005-02-09 Richard Guenther <rguenth@gcc.gnu.org>
98 * fold-const.c (try_move_mult_to_index): Remove redundant
99 type argument. Create ADDR_EXPR with correct type.
100 (fold): Update callers of try_move_mult_to_index. Convert
101 result to the appropriate type.
103 2005-02-09 Roger Sayle <roger@eyesopen.com>
106 * config/avr/avr.c (default_rtx_costs): Delete.
107 (avr_operand_rtx_cost): New function.
108 (avr_rtx_costs): Completely rewrite.
110 2005-02-08 Hans-Peter Nilsson <hp@axis.com>
113 * config/cris/cris.c (in_code): New variable.
114 (cris_output_addr_const): Now a static function, a wrapper for
116 (cris_asm_output_symbol_ref): New function, broken out SYMBOL_REF
117 case from old cris_output_addr_const.
118 (cris_asm_output_label_ref): Similar for LABEL_REF.
119 (cris_output_addr_const_extra): Similar for UNSPEC.
120 * config/cris/cris.h (OUTPUT_ADDR_CONST_EXTRA)
121 (ASM_OUTPUT_SYMBOL_REF, ASM_OUTPUT_LABEL_REF): Define.
122 * config/cris/cris-protos.h (cris_output_addr_const): Remove
124 (cris_asm_output_symbol_ref, cris_output_addr_const_extra)
125 (cris_asm_output_label_ref): Declare.
127 2005-02-08 Paolo Bonzini <bonzini@gnu.org>
129 PR preprocessor/19801
130 * doc/cppinternals.texi (Conventions, Lexer, Files): Adjust
131 filenames that changed when libcpp was moved to the toplevel.
133 2005-02-07 Roger Sayle <roger@eyesopen.com>
135 * simplify-rtx.c (simplify_relational_operation_1): Avoid creating
136 zero extensions of BImode operands. Call lowpart_subreg instead
137 of gen_lowpart_common and gen_lowpart_SUBREG.
139 2005-02-07 Nathanael Nerode <neroden@gcc.gnu.org>
142 * libada-mk.in: Use cc_for_cross_gnattools, not cc_set_by_configure.
143 * configure.ac: Set cc_for_cross_gnattools.
144 * configure: Regenerate.
146 2005-02-08 Alan Modra <amodra@bigpond.net.au>
149 * predict.c (PROB_VERY_UNLIKELY): Use 1% instead of 10%.
151 2005-02-07 Eric Botcazou <ebotcazou@libertysurf.fr>
154 * dbxout.c (dbxout_symbol) <FUNCTION_DECL>: Skip inline instance
157 2005-02-07 Jonathan Wakely <redi@gcc.gnu.org>
159 * doc/extend.texi: Document deprecated extensions allowing
160 static floating-point members to have initializers and allowing
161 floating-point literals in integral constant expressions.
163 2005-02-07 Kazu Hirata <kazu@cs.umass.edu>
165 * cfgcleanup.c, df.h, diagnostic.c, rtl.c, tree-vectorizer.h,
166 config/darwin7.h, config/mips/mips.c,
167 config/rs6000/altivec.md, config/rs6000/darwin-tramp.asm:
170 * cse.c: Update comments.
172 2005-02-07 Richard Guenther <rguenth@gcc.gnu.org>
175 * builtins.c (fold_builtin_sqrt): Transform
176 sqrt(pow(x,y)) to pow(fabs(x),y*0.5), not
179 2005-02-07 Leehod Baruch <leehod@il.ibm.com>
180 Dorit Naishlos <dorit@il.ibm.com>
182 * doc/invoke.texi (ftree-vectorizer-verbose): New.
183 * opts.c (OPT_ftree_vectorizer_verbose_): New case for switch.
184 * common.opt (ftree-vectorizer-verbose): New Flag for the vectorizer
186 * tree.h (vect_set_verbosity_level): New extern function declaration
188 * tree-vectorizer.h (verbosity_levels): New enum type.
189 * tree-vectorizer.c (vect_debug_stats): Function removed.
190 (vect_debug_details): Likewise.
191 (vect_verbosity_level): Global variable was defined and initialized.
192 (vect_dump): Global variable definition.
193 (vect_print_dump_info): New function.
194 (vect_set_dump_settings): New function.
195 (vect_set_verbosity_level): New function.
196 (vectorize_loops): Add call to vect_set_dump_settings.
198 (slpeel_make_loop_iterate_ntimes): Dump condition was changed.
199 (slpeel_tree_duplicate_loop_to_edge_cfg): Likewise.
200 (slpeel_tree_peel_loop_to_edge): Likewise.
202 (vect_analyze_offset_expr): Call to vect_print_dump_info with
203 appropriate verbosity level instead of call to vect_debug_details
205 (vect_get_base_and_offset):
206 (vect_create_addr_base_for_vector_ref):
207 (get_vectype_for_scalar_type):
208 (vect_create_data_ref_ptr):
210 (vect_get_vec_def_for_operand):
211 (vect_finish_stmt_generation):
212 (vectorizable_assignment):
213 (vectorizable_operation):
214 (vectorizable_store):
216 (vect_transform_stmt):
217 (vect_update_ivs_after_vectorizer):
218 (vect_do_peeling_for_loop_bound):
219 (vect_update_inits_of_drs):
220 (vect_do_peeling_for_alignment):
221 (vect_transform_loop):
222 (vect_is_simple_use):
223 (vect_analyze_operations):
224 (vect_is_simple_iv_evolution):
225 (vect_analyze_scalar_cycles):
226 (vect_analyze_data_ref_dependence):
227 (vect_analyze_data_ref_dependences):
228 (vect_compute_data_ref_alignment):
229 (vect_enhance_data_refs_alignment):
230 (vect_analyze_data_refs_alignment):
231 (vect_analyze_data_ref_access):
232 (vect_analyze_data_ref_accesses):
233 (vect_analyze_pointer_ref_access):
234 (vect_get_memtag_and_dr):
235 (vect_analyze_data_refs):
236 (vect_mark_relevant):
237 (vect_stmt_relevant_p):
238 (vect_mark_stmts_to_be_vectorized):
239 (vect_can_advance_ivs_p):
240 (vect_get_loop_niters):
241 (vect_analyze_loop_form):
243 (vectorize_loops): Likewise.
245 (vect_do_peeling_for_loop_bound): Dump format slightly changed.
246 (vect_update_inits_of_drs):
247 (vect_do_peeling_for_alignment):
248 (vect_transform_loop):
249 (vect_analyze_operations):
250 (vect_analyze_scalar_cycles):
251 (vect_analyze_data_ref_dependences):
252 (vect_analyze_data_refs_alignment):
253 (vect_analyze_data_ref_accesses):
254 (vect_analyze_data_refs):
255 (vect_mark_stmts_to_be_vectorized):
256 (vect_get_loop_niters):
257 (vect_analyze_loop_form):
258 (vect_analyze_loop): Likewise.
259 (vect_mark_stmts_to_be_vectorized): Add call to print_generic_expr.
261 2005-02-07 Richard Sandiford <rsandifo@redhat.com>
265 2005-01-26 Richard Sandiford <rsandifo@redhat.com>
266 * config/mips/iris6.h (ENDFILE_SPEC): Don't link in irix-csr.o if
267 either -ffast-math or -funsafe-math-optimizations is in use.
269 2005-01-24 Richard Sandiford <rsandifo@redhat.com>
270 * config/mips/irix-csr.c: New file.
271 * config/mips/t-iris6 (irix-csr.o): New rule to build it.
272 (EXTRA_MULTILIB_PARTS): Add irix-csr.o.
273 * config/mips/iris6.h (ENDFILE_SPEC): Include it in n32 and n64
276 2005-02-07 Alexandre Oliva <aoliva@redhat.com>
278 * config/mn10300/linux.h (LIB_SPEC, STARTFILE_SPEC): Remove
279 redundant, incorrect definitions.
281 * config/frv/frv.md ("tls_indirect_call"): Turn into libcall-like
283 ("*tls_indirect_call"): New, reference hard regs directly.
284 * config/frv/frv.c (gen_inlined_tls_plt): Adjust.
286 2005-02-07 Ralf Corsepius <ralf.corsepius@rtems.org>
288 * config/arm/rtems-elf.h (SUBTARGET_EXTRA_ASM_SPEC):
289 Let !-mhard-float !-msoft-float imply -mfpu=softfpa.
291 2005-02-06 Roger Sayle <roger@eyesopen.com>
293 PR rtl-optimization/19800
294 * simplify_rtx.c (simplify_relational_operation_1): Explicitly
295 call gen_lowpart_common and gen_lowpart_SUBREG instead of calling
298 2005-02-07 Ralf Corsepius <ralf.corsepius@rtems.org>
300 * config/m68k/t-rtems (MULTILIB_MATCHES): Let m528x match m5200.
302 2005-02-07 Joseph S. Myers <joseph@codesourcery.com>
305 * doc/invoke.texi: Correct documentation of -freorder-functions.
307 2005-02-06 Joseph S. Myers <joseph@codesourcery.com>
310 * doc/invoke.texi: Don't mention gccbug.
312 2005-02-06 Roger Sayle <roger@eyesopen.com>
313 Andrew Pinski <pinskia@physics.uc.edu>
314 Paolo Bonzini <paolo.bonzini@lu.unisi.ch>
316 * simplify-rtx.c (simplify_relational_operation_1): Simplify
317 (ne:SI (zero_extract:SI FOO (const_int 1) BAR) (const_int 0))
318 into just (zero_extract:SI FOO (const_int 1) BAR).
320 2005-02-06 Joseph S. Myers <joseph@codesourcery.com>
322 * doc/cpp.texi: Include gcc-common.texi. Don't define gcctabopt
323 macro locally. Don't give last revision date. Use GCC version
324 number from gcc-common.texi.
325 * doc/cppinternals.texi: Include gcc-common.texi. Don't give last
326 revision date. Use GCC version number from gcc-common.texi.
327 Describe being a library as current state rather than as 3.x
329 * doc/gcc.texi: Remove last update date.
330 * doc/gccint.texi: Likewise. Update copyright dates.
331 * doc/install.texi: Update copyright dates.
333 2005-02-06 Joseph S. Myers <joseph@codesourcery.com>
336 * doc/invoke.texi: Mention for each option included in -Wall that
337 it is included in -Wall.
338 * doc/gcc.texi: Update copyright and last modification date.
340 2005-02-06 Zdenek Dvorak <dvorakz@suse.cz>
342 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Do not add
343 unnecessary cast to original induction variable increments.
345 2005-02-06 Zdenek Dvorak <dvorakz@suse.cz>
347 PR tree-optimization/18219
348 * tree-ssa-loop-ivopts.c (get_computation_at): Produce computations
351 2005-02-06 Richard Sandiford <rsandifo@redhat.com>
353 * expmed.c (store_bit_field): Make the SUBREG code adjust bitnum.
354 Set bitpos and offset later in the function. Do nothing if the
355 target is a register and if the bitfield lies completely outside
357 (extract_bit_field): Make the same SUBREG, bitpos and offset changes
358 here. Return an uninitialised register if the source value is stored
359 in a register and the bitfield lies completely outside that register.
361 2005-02-06 Steven Bosscher <stevenb@suse.de>
363 * df.c (df_insn_refs_record): Use XEXP to get the operand of a USE,
366 2005-02-06 Marcin Dalecki <martin@dalecki.de>
368 * diagnostic.c (default_diagnostic_finalizer): Fix use of attribute.
370 2005-02-06 Geoffrey Keating <geoffk@apple.com>
372 * config.gcc (*-*-darwin*): Separate darwin7 and darwin8 configs.
373 * config/darwin8.h: New.
374 * config/darwin7.h: Update comment.
376 2005-02-03 Leehod Baruch <leehod@il.ibm.com>
377 Dorit Naishlos <dorit@il.ibm.com>
379 * tree-vectorizer.h (LOC): New type.
380 (UNKNOWN_LOC, EXPR_LOC, LOC_FILE, LOC_LINE): New macros.
381 (loop_line_number): New field in struct _loop_vec_info.
382 (LOOP_VINFO_LOC, LOOP_LOC): New macros.
384 * tree-vectorizer.c (input.h): Included.
385 (find_loop_location): New function.
386 (vect_debug_stats): Argument changed from loop to LOC. Computation of
387 loop line number removed.
388 (vect_debug_details): Likewise.
389 (new_loop_vec_info): Initialize new field LOOP_VINFO_LOC.
391 (slpeel_make_loop_iterate_ntimes): Call find_loop_location. Argument in
392 call to vect_debug_details/stats changed from loop to loop_loc.
393 (slpeel_tree_duplicate_loop_to_edge_cfg): Likewise.
394 (slpeel_tree_peel_loop_to_edge): Likewise.
396 (vect_analyze_offset_expr): Argument in call to vect_debug_details/stats
397 changed from NULL to UNKNOWN_LOC.
398 (vect_get_base_and_offset):
399 (vect_create_addr_base_for_vector_ref):
400 (get_vectype_for_scalar_type):
401 (vect_create_data_ref_ptr):
403 (vect_get_vec_def_for_operand):
404 (vect_finish_stmt_generation):
405 (vectorizable_assignment):
406 (vectorizable_operation):
407 (vectorizable_store):
409 (vect_transform_stmt):
410 (vect_update_ivs_after_vectorizer):
411 (vect_do_peeling_for_loop_bound):
412 (vect_do_peeling_for_alignment):
413 (vect_transform_loop):
414 (vect_is_simple_use):
415 (vect_analyze_operations):
416 (vect_is_simple_iv_evolution):
417 (vect_analyze_scalar_cycles):
418 (vect_analyze_data_ref_dependences):
419 (vect_compute_data_ref_alignment):
420 (vect_enhance_data_refs_alignment):
421 (vect_analyze_data_ref_access):
422 (vect_analyze_data_ref_accesses):
423 (vect_analyze_pointer_ref_access):
424 (vect_get_memtag_and_dr):
425 (vect_analyze_data_refs):
426 (vect_mark_relevant):
427 (vect_stmt_relevant_p):
428 (vect_mark_stmts_to_be_vectorized):
429 (vect_can_advance_ivs_p):
430 (vect_get_loop_niters):
432 (vectorize_loops): Likewise.
434 (vectorizable_load): Argument in call to vect_debug_details/stats
435 changed from loop to LOOP_LOC (loop_vinfo).
436 (vect_transform_loop):
437 (vect_analyze_operations):
438 (vect_analyze_scalar_cycles):
439 (vect_analyze_data_ref_dependence):
440 (vect_enhance_data_refs_alignment):
441 (vect_analyze_data_ref_accesses):
442 (vect_analyze_pointer_ref_access):
443 (vect_analyze_data_refs):
444 (vect_analyze_loop): Likewise.
446 (vect_analyze_loop_form): Argument in call to vect_debug_details/stats
447 changed from loop to loop_loc.
449 (vect_enhance_data_refs_alignment): Removed unused variable loop.
451 2005-02-03 Leehod Baruch <leehod@il.ibm.com>
452 Dorit Naishlos <dorit@il.ibm.com>
454 * tree-vectorizer.c (vect_analyze_data_ref_dependence): Function
456 (vect_analyze_data_ref_dependences): Likewise.
458 (vect_is_simple_use): Argument changed from loop structure to
459 loop_vect_info structure.
460 (vect_can_advance_ivs_p): Likewise.
461 (vect_create_index_for_vector_ref): Likewise.
462 (vect_update_ivs_after_vectorizer): Likewise.
463 (new_stmt_vec_info): Likewise.
465 (new_loop_vec_info): Second argument in call to new_stmt_vec_info was
466 changed from loop to loop_vinfo.
467 (vect_create_data_ref_ptr): First argument in call to
468 vect_create_index_for_vector_ref was changed from loop to loop_vinfo.
469 (vectorizable_assignment): Second argument in call to vect_is_simple_use
470 was changed from loop to loop_vinfo.
471 (vectorizable_operation): Likewise.
472 (vectorizable_store): Likewise.
473 (vect_mark_stmts_to_be_vectorized): Likewise.
474 (vect_do_peeling_for_loop_bound): First argument in call to
475 vect_update_ivs_after_vectorizer was changed from loop to loop_vinfo.
476 (vect_analyze_operations): Argument in call to vect_can_advance_ivs_p
477 was changed from loop to loop_vinfo.
478 (vect_analyze_data_ref_dependences): Third argument in call to
479 vect_analyze_data_ref_dependence was changed from loop to loop_vinfo.
481 (vect_create_index_for_vector_ref): Get the loop from loop_vinfo.
482 (vect_create_data_ref_ptr): Likewise.
483 (vect_init_vector): Likewise.
484 (vect_get_vec_def_for_operand): Likewise.
485 (vectorizable_assignment): Likewise.
486 (vectorizable_operation): Likewise.
487 (vectorizable_store): Likewise.
488 (vectorizable_load): Likewise.
489 (vect_update_ivs_after_vectorizer): Likewise.
490 (vect_is_simple_use): Likewise.
491 (vect_analyze_data_ref_dependence): Likewise.
492 (vect_analyze_pointer_ref_access): Likewise.
493 (vect_can_advance_ivs_p): Likewise.
495 * tree-vectorizer.h: (_loop_vec_info): Definition and macros moved
496 before _stmt_vec_info.
497 (_stmt_vec_info): Loop field replaced by loop_vec_info.
498 (STMT_VINFO_LOOP): Replaced with STMT_VINFO_LOOP_VINFO.
499 (new_stmt_vec_info): Argument changed from loop structure to
500 loop_vect_info structure
502 (vect_analyze_data_ref_dependences): Unnecessary line was removed.
503 (vect_analyze_offset_expr): Avoid 80 columns overflow.
504 (vect_create_addr_base_for_vector_ref): Likewise.
505 (vect_analyze_pointer_ref_access): Likewise.
507 2005-02-03 Andrew Pinski <pinskia@physics.uc.edu>
510 * tree-ssa-dse.c (fix_phi_uses): Update the occurs in abnormal
511 phi flag if the phi is abnormal.
513 2005-02-03 Andrew Pinski <pinskia@physics.uc.edu>
516 * tree-ssa.c (replace_immediate_uses): Update the immediate_uses
517 information for the new statement.
519 2005-02-03 Joseph S. Myers <joseph@codesourcery.com>
522 * gcc.c (main): Include the target in -v output.
524 2005-02-03 Alexandre Oliva <aoliva@redhat.com>
526 * combine.c (simplify_comparison, case SIGN_EXTEND, ZERO_EXTEND):
527 Do not drop the extend if we'd have to add a paradoxical subreg
528 later. Include optabs.h and insn-codes.h.
529 * Makefile.in (combine.o): Depend on $(OPTABS_H).
531 2005-02-02 Roger Sayle <roger@eyesopen.com>
534 * real.h (REAL_MODE_FORMAT_COMPOSITE_P): New macro.
535 * fold-const.c (const_binop): Avoid constant folding floating
536 point operations in modes that use composite representations.
537 * simplify-rtx.c (simplify_binary_operation): Likewise.
539 2005-02-02 Geoffrey Keating <geoffk@apple.com>
541 * config/rs6000/altivec.md (altivec_dst): Make the first operand
543 (altivec_dstt): Likewise.
544 (altivec_dstst): Likewise.
545 (altivec_dststt): Likewise.
546 * config/rs6000/rs6000.c (altivec_expand_dst_builtin): Adjust creation
549 2005-02-03 Dorit Naishlos <dorit@il.ibm.com>
551 * tree-vectorizer.c (slpeel_make_loop_iterate_ntimes): Copy
552 EXPR_LOCUS from orig_cond to the new cond_stmt.
553 (vect_finish_stmt_generation): Copy EXPR_LOCUS from stmt to
556 2005-02-03 Dorit Naishlos <dorit@il.ibm.com>
558 * tree-vectorizer.c (slpeel_make_loop_iterate_ntimes): Call
559 standard_iv_increment_position. Remove call to bsi_next
560 (no need to bump the iterator anymore).
561 (vect_create_index_for_vector_ref): Call
562 standard_iv_increment_position. Remove second function argument.
563 (vect_finish_stmt_generation): Remove call to bsi_next
564 (no need to bump the iterator anymore).
565 (vect_create_data_ref_ptr): Remove second argument (bsi) in call
566 to vect_create_index_for_vector_ref.
568 2005-02-03 Dorit Naishlos <dorit@il.ibm.com>
570 * tree-vectorizer.c (vect_mark_stmts_to_be_vectorized): Check if
571 phis are used out of the loop.
573 2005-02-03 Joseph S. Myers <joseph@codesourcery.com>
576 * c-decl.c (undef_nested_function): New variable.
577 (pop_scope): Diagnose undefined nested functions.
578 (finish_function): Don't attempt cgraph processing or genericizing
579 if current top-level function contained an undefined nested
580 function. Reset undef_nested_function at the end of a top-level
583 2005-02-02 Zdenek Dvorak <dvorakz@suse.cz>
585 * tree.c (build_int_cst_type): Take sign of the value into account
586 when deciding whether sign extend the value.
588 2005-02-02 Joseph S. Myers <joseph@codesourcery.com>
591 * gcc.c (cpp_unique_options): Remove %{trigraphs}.
592 (cpp_options, cc1_options): Change %{std*&ansi} to
593 %{std*&ansi&trigraphs}.
595 2005-02-02 Kazu Hirata <kazu@cs.umass.edu>
597 * cse.c: Update comments.
599 2005-02-02 Aldy Hernandez <aldyh@redhat.com>
601 * config/rs6000/rs6000.c (rs6000_spe_function_arg): Multi-reg
602 arguments go on the stack. Do not put complex doubles in odd/even
605 2005-02-02 Steven Bosscher <stevenb@suse.de>
607 * doc/invoke.texi: Update the options enabled at -O1, -O2,
608 and -O3. Update the documentation for -fgcse-sm, -fgcse-las,
609 -ftree-pre, and -fthread-jumps.
611 2005-02-02 Richard Sandiford <rsandifo@redhat.com>
613 PR tree-optimization/19578
614 * tree-flow.h (modified_noreturn_calls): Declare.
615 (noreturn_call_p): Declare.
616 * tree-flow-inline.h (noreturn_call_p): New function.
617 (modify_stmt): Add modified noreturn calls to modified_noreturn_calls.
618 * tree-cfg.c (modified_noreturn_calls): New variable.
619 (cleanup_control_flow): Use noreturn_call_p. Split basic blocks
620 that contain a mid-block noreturn call.
621 * tree-ssa.c (delete_tree_ssa): Clear modified_noreturn_calls.
623 2005-02-02 Kazu Hirata <kazu@cs.umass.edu>
625 * df.c (df_def_record_1, df_uses_record): Don't use
627 * df.h (DF_FOR_REGALLOC): Remove.
629 2005-02-02 Joseph S. Myers <joseph@codesourcery.com>
632 * c-typeck.c (really_start_incremental_init): Reset
633 constructor_max_index for arrays of incomplete type.
635 2005-02-02 Jeff Law <law@redhat.com>
637 * gcse.c (struct reg_set): Store the block index where the register
638 was set rather than the setting insn.
639 (record_one_set, compute_transp): Corresponding changes.
640 (pre_insert_copy_insn): Do not call replace_one_set.
641 (replace_one_set): Remove.
643 2005-02-02 Joseph S. Myers <joseph@codesourcery.com>
646 * doc/gcov.texi: Fix typo.
648 2005-02-01 Richard Henderson <rth@redhat.com
651 * config/i386/i386.h (MODES_TIEABLE_P): Use ix86_modes_tieable_p.
652 * config/i386/i386.c (ix86_hard_regno_mode_ok): Change return
654 (ix86_tieable_integer_mode_p, ix86_modes_tieable_p): New.
655 * config/i386/i386-protos.h: Update.
657 2005-02-01 Steven Bosscher <stevenb@suse.de>
659 PR tree-optimization/19217
660 * tree-cfg.c (verify_expr): Use the data field to see if TP was
661 seen inside a PHI node. Do not do the ADDR_EXPR check if it was.
662 (verify_stmts): Pass (void*)1 as data to verify_expr to signal
663 that it is walking a PHI node.
665 2005-02-01 Joseph S. Myers <joseph@codesourcery.com>
667 * doc/extend.texi (Nested Functions): Update.
669 2005-02-01 Richard Henderson <rth@redhat.com>
672 * optabs.c (expand_copysign_absneg): Export.
673 * optabs.h (expand_copysign_absneg): Declare.
674 * config/rs6000/rs6000.md (copysigntf3): New.
676 2005-02-01 Diego Novillo <dnovillo@redhat.com>
678 PR tree-optimization/19633
679 * tree-flow.h (struct ptr_info_def): Add field 'pt_null'.
680 * tree-ssa-alias.c (init_alias_info): Initialize.
681 (merge_pointed_to_info): Set.
682 (add_pointed_to_expr): Set pt_null if EXPR is a NULL pointer.
683 (dump_points_to_info_for): Show value of pt_null.
684 (struct count_ptr_d): Declare.
685 (find_ptr_dereference): Remove.
686 (ptr_is_dereferenced_by): Remove.
687 (count_ptr_derefs): New local function.
688 (count_uses_and_derefs): New local function.
689 (compute_points_to_and_addr_escape): Call it. If the number
690 of dereferences is greater than zero, mark the pointer as
691 dereferenced. If there are fewer dereferences than uses of
692 the pointer, the pointer's value escapes.
694 2005-02-01 Diego Novillo <dnovillo@redhat.com>
696 PR tree-optimization/19670
697 * tree-ssa.c (verify_ssa_name): Don't set TREE_VISITED
699 (verify_use): ... set it here, instead.
701 2005-02-01 Janis Johnson <janis187@us.ibm.com>
703 * doc/sourcebuild.texi (Test directives): Fix formatting.
705 * doc/sourcebuild.texi (Test directives): Describe selector
708 2005-02-01 Segher Boessenkool <segher@kernel.crashing.org>
710 * config.gcc (powerpc64-*-linux*): Default to -m64 also for
711 specific 64-bit CPUs.
713 2005-02-01 Richard Earnshaw <rearnsha@arm.com>
716 * arm.c (arm_eliminable_register): New function.
717 (adjacent_mem_locations): Don't allow eliminable registers. Use
718 HOST_WIDE_INT for address offsets.
719 * arm-protos.h (arm_eliminable_register): Add prototype.
721 2005-02-01 Steven Bosscher <stevenb@suse.de>
723 PR optimization/15242
724 * params.def (PARAM_MAX_GOTO_DUPLICATION_INSNS): New param.
725 * basic-block.h (duplicate_computed_gotos): Add prototype.
726 * bb-reorder.c (duplicate_computed_gotos): New function to
727 duplicate sufficiently small blocks ending in a computed jump.
728 * passes.c (rest_of_compilation): Call duplicate_computed_gotos
729 if not optimizing for size.
730 * cfgcleanup.c (try_crossjump_bb): If not optimizing for size,
731 never do tail merging for blocks ending in a computed jump.
732 * doc/invoke.texi: Document the max-goto-duplication-insns param.
734 2005-02-01 Eric Botcazou <ebotcazou@libertysurf.fr>
736 Patch from Richard Sandiford <rsandifo@redhat.com>
737 * reload1.c (choose_reload_regs): Prevent the find_equiv_reg() code
738 from inheriting a subreg equivalence with a non-spill register.
740 * tree.h (DECL_FUNCTION_CODE): Document that it is overloaded.
742 2005-02-01 Alexandre Oliva <aoliva@redhat.com>
744 * config/frv/frv.c (movcc_fp_destination_operand): New.
745 (gpr_or_memory_operand): Fix typo in comment.
746 (gpr_or_memory_operand_with_scratch): New.
747 * config/frv/frv.h (PREDICATE_CODES): Add the two new predicates.
748 * config/frv/frv.md (movcc_fp, movcc_fp_internal): Use
749 movcc_fp_destination_operand.
750 (reload_incc_fp): Use gpr_or_memory_operand_with_scratch.
751 Legitimize memory addresses using a scratch register.
753 2005-01-31 Jeff Law <law@redhat.com>
755 * tree-into-ssa.c (mark_def_sites_global_data): Make KILLS
756 bitmap a sparse bitmap instead of a simple bitmap.
757 (mark_def_sites_initialize_block): Corresponding changes.
758 (ssa_mark_def_sites_initialize_block): Likewise.
759 (ssa_mark_phi_uses): Likewise.
760 (mark_def_site, ssa_mark_def_sites): Likewise.
761 (mark_def_site_blocks): Likewise.
762 (rewrite_ssa_into_ssa): Likewise.
764 * tree-ssa-dom.c (record_cond): Pass correct variable type
765 for last argument to htab_find_slot_with_hash.
767 * fold-const.c (fold, case CEIL_MOD_EXPR): Do not lose side
768 effects when optimizing 0 % X. Do not try to optimize X % 0.
770 2005-01-31 James E. Wilson <wilson@specifixinc.com>
772 * config/ia64/itanium1.md (1_scall bypass): Change 2_mmalua to
775 2005-02-01 Eric Christopher <echristo@redhat.com>
777 * config/mips/mips.c (override_options): Warn if -mint64
779 * doc/invoke.texi (MIPS Options): Document that -mint64 is
782 2005-02-01 Kazu Hirata <kazu@cs.umass.edu>
784 * cse.c (cse_reg_info): Remove hash_next, next, regno. Add
786 (cse_reg_info_list, cse_reg_info_list_free, REGHASH_SHIFT,
787 REGHASH_SIZE, REGHASH_MASK, reg_hash, REGHASH_FN,
788 cached_cse_reg_info, GET_CSE_REG_INFO): Remove.
789 (cached_regno): Initialize to INVALID_REGNUM.
790 (cse_reg_info_table_size,
791 cse_reg_info_table_first_uninitialized,
792 cse_reg_info_timestamp): New.
793 (REG_TICK, REG_IN_TABLE, SUBREG_TICKED, REG_QTY): Use
795 (init_cse_reg_info, get_cse_reg_info_1): New.
796 (get_cse_reg_info): Cache the last look-up.
797 (new_basic_block): Update the code to clear mappings from
798 registers to cse_reg_info entries.
799 (cse_main): Call init_cse_reg_info.
801 * cse.c (get_cse_reg_info): Update a comment.
803 2005-01-31 Steven Bosscher <stevenb@suse.de>
806 * c-decl.c (start_decl): Do not warn about arrays of elements with
807 an incomplete type here.
808 (grokdeclarator): Do it here by making a pedwarn an error.
809 * c-typeck.c (push_init_level): If there were previous errors with
810 the constructor type, do not warn about braces for initializers.
811 (process_init_element): Likewise for excess initializer elements.
813 2005-01-31 Kazu Hirata <kazu@cs.umass.edu>
815 * cse.c (delete_trivially_dead_insn): Don't iterate.
817 2005-01-31 Andrew Pinski <pinskia@physics.uc.edu>
819 * config/rs6000/rs6000.md (copysignsf3): New expand.
820 (copysigndf3): Likewise.
822 2005-01-31 Steven Bosscher <stevenb@suse.de>
824 * recog.c (constrain_operands): Only look for earlyclobber operand
825 conflicts if an '&' constraint was seen.
827 2005-01-31 Marc Espie <espie@openbsd.org>
829 * config.gcc: Don't include embedded systems fragment, switches default
830 debugging format to ELF.
831 * config/i386/openbsdelf.h: Add DBX_REGISTER_NUMBER since we no
832 longer pick it up there.
834 2005-01-31 Dale Johannesen <dalej@apple.com>
836 * doc/extend.texi (nested functions): Fix linkage description.
837 Clarify that static is not allowed.
839 2005-01-31 Dale Johannesen <dalej@apple.com>
841 * config/rs6000/darwin.md (movsf_low_di): Make work.
842 (movdf_low_di): Make work.
844 2005-01-31 Dale Johannesen <dalej@apple.com>
846 * config/rs6000/darwin-tramp.asm (__trampoline_setup):
847 Make work for 64 bit.
849 2005-01-31 Roger Sayle <roger@eyesopen.com>
850 Dale Johannesen <dalej@apple.com>
853 * fold-const.c (fold_binary_op_with_conditional_arg):
854 Make types match original operands, before STRIP_NOPS.
856 2005-01-31 Alan Modra <amodra@bigpond.net.au>
858 * config/rs6000/linux-unwind.h (struct gcc_vregs): New.
859 (struct gcc_regs): Rename from gcc_pt_regs. Add more regs.
860 (struct gcc_sigcontext): Delete. Merge contents to..
861 (struct gcc_ucontext): ..here.
862 (get_sigcontext): Delete.
863 (get_regs): New function, like get_sigcontext but return regs ptr.
864 64-bit version finds regs from r1 to support vdso.
865 (ppc_linux_aux_vector): New function.
866 (ppc_fallback_frame_state): Modify for get_regs. Restore fprs
869 2005-01-31 Ira Rosen <irar@il.ibm.com>
871 * tree-vectorizer.c (vect_analyze_offset_expr): Use ssizetype instead
873 (vect_get_base_and_offset): Use ssizetype instead sizetype. Remove
874 redundant fold. Fix misalignment for MINUS_EXPR.
875 (vect_compute_data_ref_alignment): Use ssizetype instead sizetype.
876 (vect_analyze_pointer_ref_access): Likewise.
877 (vect_get_memtag_and_dr): Likewise.
879 2005-01-31 Richard Henderson <rth@redhat.com>
881 * config/i386/i386.c (absnegsf2_mixed): Remove all of the # hackery.
882 (absnegsf2_sse, absnegsf2_i387, absnegdf2_mixed, absnegdf2_sse,
883 absnegdf2_i387, absnegxf2_i387): Likewise.
885 2005-01-31 Richard Henderson <rth@redhat.com>
888 * optabs.c (expand_copysign_absneg): New.
889 (expand_copysign_bit): Split out from ...
890 (expand_copysign): ... here. Use expand_copysign_absneg.
892 2005-01-30 Roger Sayle <roger@eyesopen.com>
895 * config/pa/pa.md (anddi3, iordi3): On HPPA64, disallow an integer
896 constant as the second operand and a register as the third.
898 2005-01-31 Danny Smith <dannysmith@users.sourceforge.net>
901 * config/i386/i386.c (ix86_function_ok_for_sibcall): Also check
902 that dllimport'd functions do not use all call-clobbered registers
905 2005-01-30 Richard Henderson <rth@redhat.com>
908 * optabs.c (expand_absneg_bit): Split out from expand_unop and
909 expand_abs_nojump and generalize. Use operand_subword and
910 emit_no_conflict_block. Support large modes.
911 (expand_abs_nojump): Use it. Use HONOR_SIGNED_ZEROS to fall
913 (expand_unop): Likewise. Use HONOR_SIGNED_ZEROS to fall back
914 negation to subtraction.
915 (expand_copysign): Check that the format has signed zeros.
916 Use operand_subword and emit_no_conflict_block. Support large modes.
918 2005-01-30 Daniel Berlin <dberlin@dberlin.org>
920 Fix PR tree-optimization/19624
922 * Makefile.in (tree-ssa-pre.o): Add CFGLOOP_H.
923 * tree-ssa-pre.c: Add cfgloop.h.
925 (pre_stats): New member, constified.
926 (inserted_exprs): New static variable.
927 (NECESSARY): New macro.
928 (create_expression_by_pieces): Fold the expression, and
929 mark it as defaulting to not necessary. Also put in
931 (fully_constant_expression): New function.
932 (insert_into_preds_of_block): Modify to not insert phis when we
933 are playing with induction variables.
934 Push phis onto the inserted_exprs vector, and mark them as not
935 necessary by default.
936 (insert_aux): Call fully_constant_expression on eprime.
937 If all edges produce the same value, mark it constant.
938 (mark_operand_necessary): New function.
939 (remove_dead_inserted_code): New function.
940 (init_pre): Init loop optimizer to get loop info.
941 (fini_pre): Free loop_optimizer, and inserted_exprs vec.
942 (execute_pre): Commit edge inserts, then remove dead code.
944 2005-01-30 Richard Henderson <rth@redhat.com>
946 * rtl.c (rtx_equal_p): No early exit for CONST_VECTOR.
947 * varasm.c (const_rtx_hash_1): Handle CONST_VECTOR.
949 2005-01-30 Richard Henderson <rth@redhat.com>
952 * config/i386/i386.c (ix86_expand_copysign): New.
953 (ix86_split_copysign_const): New.
954 (ix86_split_copysign_var): Rename from ix86_split_copysign,
955 rearrange op1/nmask operands.
956 * config/i386/i386-protos.h: Update.
957 * config/i386/i386.md (copysignsf3): Use ix86_expand_copysign.
958 (copysigndf3): Likewise.
959 (copysignsf3_const, copysigndf3_const): New.
960 (copysignsf3_var): Rename from copysignsf3, split out splitter
961 and fix split predicate for X constraint.
962 (copysigndf3_var): Similarly.
964 2005-01-30 Kazu Hirata <kazu@cs.umass.edu>
966 * optabs.c, doc/c-tree.texi, doc/install.texi, doc/md.texi,
967 doc/passes.texi, doc/rtl.texi, doc/sourcebuild.texi,
968 doc/tm.texi, doc/tree-ssa.texi: Update copyright.
970 2005-01-29 Richard Henderson <rth@redhat.com>
973 * config/i386/i386.md (movdf_nointeger, movdf_integer): Fix ordering
974 of # and * in constraints.
976 2005-01-29 Richard Henderson <rth@redhat.com>
979 * expr.c (store_field): Don't strip sub-mode cast when the input
980 data is even smaller.
982 2005-01-29 Richard Henderson <rth@redhat.com>
985 * expr.c (categorize_ctor_elements_1): Check for CONSTRUCTOR of a
988 2005-01-29 Richard Henderson <rth@redhat.com>
990 * combine.c (make_field_assignment): Fix argument order
993 2005-01-29 Richard Guenther <richard.guenther@uni-tuebingen.de>
995 PR tree-optimization/15791
996 * fold-const.c (extract_array_ref): New function.
997 (fold): Fold comparisons between &a[i] and &a[j] or
998 semantically equivalent trees.
1000 2005-01-29 Jeff Law <law@redhat.com>
1002 * gcse.c (insert_expr_in_table): Revamp handling of available
1003 and anticipatable occurrence lists to avoid unnecessary list
1005 (insert_set_in_table): Similarly.
1007 2005-01-29 Joseph S. Myers <joseph@codesourcery.com>
1009 * c-common.c (fix_string_type): Just use c_build_qualified_type to
1011 (c_build_qualified_type): Build qualified array types with
1012 TYPE_MAIN_VARIANT pointing to corresponding unqualified type.
1013 * c-decl.c (c_make_fname_decl): Build unqualified array type
1014 before qualified type.
1015 (grokdeclarator): Use TYPE_MAIN_VARIANT of typedef type if element
1016 type is qualified, not just if type itself is. Don't apply
1017 qualifiers to array type when declarator is processed. Apply
1018 qualifiers to field type whether or not it is an array type.
1019 Don't handle array types specially for applying qualifiers to
1021 * c-typeck.c (composite_type): Build unqualified element type and
1022 array type when forming composite of array types.
1023 (common_pointer_type, comptypes, comp_target_types,
1024 type_lists_compatible_p, build_indirect_ref, build_array_ref,
1025 convert_for_assignment): Don't apply TYPE_MAIN_VARIANT to array
1027 (type_lists_compatible_p): Cache TREE_VALUE (args1) and TREE_VALUE
1028 (args2) in variables a1 and a2.
1030 2005-01-29 Kazu Hirata <kazu@cs.umass.edu>
1032 * cse.c (n_elements_made, max_elements_made): Remove.
1033 (insert): Don't touch n_elements_made.
1034 (cse_main): Don't touch n_elements_made or max_elements_made.
1036 2005-01-28 Stephane Carrez <stcarrez@nerim.fr>
1039 * config/m68hc11/t-m68hc11-gas (dp-bit.c): Fix typo causing a
1040 configuration part of dp-bit.c to be lost.
1042 2005-01-28 Roger Sayle <roger@eyesopen.com>
1044 * expmed.c (expand_mult_highpart_optab): When attempting to use
1045 a non-widening multiplication in a wider mode, the operands need
1046 to be converted (zero or sign extended) to that mode.
1048 2005-01-28 Ian Lance Taylor <ian@airs.com>
1052 * gimple-low.c (block_may_fallthru): TRY_FINALLY_EXPR only falls
1053 through if both operands fall through.
1055 2005-01-28 Kazu Hirata <kazu@cs.umass.edu>
1057 * cse.c (fold_rtx) <PC>: Don't optimize.
1059 2005-01-28 Jeff Law <law@redhat.com>
1061 * fold-const.c (fold, case CEIL_MOD_EXPR): Optimize 0 % X.
1062 (case FLOOR_MOD_EXPR, ROUND_MOD_EXPR, TRUNC_MOD_EXPR): Similarly.
1064 2005-01-28 Kazu Hirata <kazu@cs.umass.edu>
1066 * cse.c (cse_reg_info_free_list, cse_reg_info_used_list,
1067 cse_reg_info_used_list_end): Remove.
1068 (cse_reg_info_list, cse_reg_info_list_free): New.
1069 (get_cse_reg_info): When allocating an instance of
1070 cse_reg_info, add it to the beginning of the cse_reg_info_list
1071 list. Remove code to maintain cse_reg_info_used_list.
1072 (new_basic_block): Reset the free list to the beginning of
1075 2005-01-28 Richard Sandiford <rsandifo@redhat.com>
1077 * config/frv/frv.c (bdesc_2arg): Remove __MDPACKH.
1078 (frv_init_builtins): Change its prototype to take 4 uhalf arguments.
1079 (frv_expand_mdpackh_builtin): New function.
1080 (frv_expand_builtin): Use it to expand __MDPACKH.
1081 * doc/extend.texi (__MDPACKH): Update documentation.
1083 2005-01-28 Uros Bizjak <uros@kss-loka.si>
1085 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use
1086 SSE_FLOAT_MODE_P for use_sse computation.
1088 2005-01-28 Ralf Corsepius <ralf.corsepius@rtems.org>
1091 * config/sparc/rtemself.h (LINK_GCC_C_SEQUENCE_SPEC): undefine.
1093 2005-01-27 Richard Henderson <rth@redhat.com>
1095 * builtins.c (expand_builtin_copysign): New.
1096 (expand_builtin): Call it.
1097 * genopinit.c (optabs): Add copysign_optab.
1098 * optabs.c (init_optabs): Initialize it.
1099 (expand_copysign): New.
1100 * optabs.h (OTI_copysign, copysign_optab): New.
1101 (expand_copysign): Declare.
1103 * config/alpha/alpha.md (UNSPEC_COPYSIGN): New.
1104 (copysignsf3, ncopysignsf3, copysigndf3, ncopysigndf3): New.
1106 * config/i386/i386.c (ix86_build_signbit_mask): Split from ...
1107 (ix86_expand_fp_absneg_operator): ... here.
1108 (ix86_split_copysign): New.
1109 * config/i386/i386-protos.h: Update.
1110 * config/i386/i386.md (UNSPEC_COPYSIGN): New.
1111 (copysignsf3, copysigndf3): New.
1113 * config/ia64/ia64.md (UNSPEC_COPYSIGN): New.
1114 (copysignsf3, ncopysignsf3): New.
1115 (copysigndf3, ncopysigndf3): New.
1116 (copysignxf3, ncopysignxf3): New.
1117 * config/ia64/ia64.c (rtx_needs_barrier): Handle UNSPEC_COPYSIGN.
1119 2005-01-27 Arend Bayer <arend.bayer@web.de>
1120 Kazu Hirata <kazu@cs.umass.edu>
1122 * cse.c: (find_best_addr): Don't call copy_rtx before calling
1123 fold_rtx. Save cost recomputation if fold_rtx did nothing.
1124 (fold_rtx) <ASM_OPERANDS>: Don't do anything if INSN is
1127 2005-01-27 Jeff Law <law@redhat.com>
1129 * tree-into-ssa.c (ssa_rewrite_initialize_block): Do not register
1130 definitions of SSA_NAMEs which are not being rewritten.
1131 (rewrite_ssa_into_ssa): Only initialize the current definition
1132 of an SSA_NAME if that SSA_NAME has been marked for rewriting.
1133 If checking is enabled, assert that SSA_NAME_AUX is clear for all
1134 SSA_NAMEs before returning.
1136 2005-01-27 Kazu Hirata <kazu@cs.umass.edu>
1138 * c-common.def, c-dump.c, c-gimplify.c, c-objc-common.c,
1139 cfgexpand.c, dbxout.c, function.h, opts.c, tree-flow-inline.h,
1140 tree-ssa-operands.h, tree-ssa-threadupdate.c, tree-tailcall.c,
1141 config/avr/avr.md, config/cris/aout.h, config/cris/cris.h,
1142 config/mips/iris6.h, config/sh/sh.c: Update copyright.
1144 2005-01-27 Steven Bosscher <stevenb@suse.de>
1147 * opts.c (decode_options): Move flag_thread_jumps from -O1 and
1148 higher to -O2 and higher. Likewise for tree PRE.
1149 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Only iterate at -O2
1152 2005-01-27 Ian Lance Taylor <ian@airs.com>
1155 * gimple-low.c (try_catch_may_fallthru): New static function.
1156 (block_may_fallthru): Handle TRY_CATCH_EXPR.
1157 * tree-inline.c (expand_call_inline): Don't warn about reaching
1158 the end of a non-void function being inlined if the function uses
1161 2005-01-27 Jakub Jelinek <jakub@redhat.com>
1163 * config/i386/i386.h (CALL_USED_REGISTERS): Fix comment pastos.
1166 * c-decl.c (warn_if_shadowing): Handle old_decl error_mark_node.
1167 (pushdecl): Only use DECL_FILE_SCOPE_P if DECL_P.
1168 (implicitly_declare): Handle error_mark_node.
1170 2005-01-27 Richard Henderson <rth@redhat.com>
1173 * tree.h (struct tree_decl): Add debug_expr_is_from.
1174 (DECL_DEBUG_EXPR_IS_FROM): New.
1175 (DECL_DEBUG_EXPR): Rename from DECL_DEBUG_ALIAS_OF.
1176 * dwarf2out.c (dwarf2out_var_location): Update to match.
1177 * tree-outof-ssa.c (create_temp): Likewise.
1178 * var-tracking.c (track_expr_p): Likewise.
1179 * tree-sra.c (instantiate_element): Set DECL_DEBUG_EXPR.
1180 * c-objc-common.c (c_tree_printer) <'D'>: Handle DECL_DEBUG_EXPR.
1181 * toplev.c (default_tree_printer): Likewise.
1183 2005-01-27 Alexandre Oliva <aoliva@redhat.com>
1185 * config/frv/frv.c (frv_class_likely_spilled_p): Add GR8_REGS,
1186 GR9_REGS, GR89_REGS, FDPIC_FPTR_REGS, FDPIC_REGS.
1188 2005-01-27 Steven Bosscher <stevenb@suse.de>
1190 * tree.h (SWITCH_COND, SWITCH_BODY, SWITCH_LABELS, CASE_LOW,
1191 CASE_HIGH, CASE_LABEL, ASM_STRING, ASM_OUTPUTS, ASM_INPUTS,
1192 ASM_CLOBBERS): Add tree checks.
1194 * c-common.h (SWITCH_TYPE): Rename to SWITCH_STMT_TYPE.
1195 (SWITCH_STMT_COND, SWITCH_STMT_BODY): New.
1196 * c-common.def (SWITCH_STMT): Update to match.
1197 * c-common.c (c_do_switch_warnings): Use SWITCH_STMT accessor
1198 macros instead of SWITCH_EXPR ones.
1199 * c-dump.c (c_dump_tree): Likewise.
1200 * c-gimplify.c (gimplify_switch_stmt): Likewise.
1201 * c-typeck.c (c_start_case, do_case, c_finish_case): Likewise.
1203 * doc/c-tree.texi (SWITCH_STMT): Update accessor macro names.
1205 2005-01-27 Alan Modra <amodra@bigpond.net.au>
1207 * unwind-dw2.c (execute_stack_op): Add missing cases for
1208 DW_OP_shl, DW_OP_shr, DW_OP_shra, DW_OP_xor.
1210 2005-01-27 Eric Botcazou <ebotcazou@libertysurf.fr>
1212 * config.gcc (Obsolete configurations): Remove 'dummy', add
1213 sparclite-*-coff*, sparclite-*-elf*, sparc86x-*-elf* and
1216 2005-01-26 Diego Novillo <dnovillo@redhat.com>
1218 PR tree-optimization/19633
1219 * tree-ssa-alias.c (ptr_is_dereferenced_by): Also handle
1221 (maybe_create_global_var): Do not create .GLOBAL_VAR if there
1222 are no call-clobbered variables.
1223 * tree-outof-ssa.c (check_replaceable): Return false for calls
1226 2005-01-26 Ulrich Weigand <uweigand@de.ibm.com>
1228 * dbxout.c (dbxout_symbol_location): Resolve constant pool references
1229 even for variables with NULL DECL_INITIAL.
1231 2005-01-26 Stuart Hastings <stuart@apple.com>
1233 * gimplify.c (shortcut_cond_expr): Re-compute side-effects.
1235 2005-01-26 Richard Henderson <rth@redhat.com>
1238 * c-decl.c (finish_struct): Set DECL_MODE after resetting a
1240 * expr.c (store_field): Strip conversions to odd-bit-sized types
1241 if the destination field width matches.
1243 2005-01-26 Richard Henderson <rth@redhat.com>
1245 * c-decl.c, expmed.c, expr.c: Revert last change.
1247 2005-01-26 Richard Henderson <rth@redhat.com>
1250 * c-decl.c (finish_struct): Set DECL_MODE after resetting a
1252 * expmed.c (store_fixed_bit_field): Create a paradoxical subreg
1253 if we don't need the bits above those present in the current mode.
1254 * expr.c (store_field): Strip conversions to odd-bit-sized types
1255 if the destination field width matches.
1257 2005-01-26 Richard Sandiford <rsandifo@redhat.com>
1259 * config/mips/iris6.h (ENDFILE_SPEC): Don't link in irix-csr.o if
1260 either -ffast-math or -funsafe-math-optimizations is in use.
1262 2005-01-26 Marek Michalkiewicz <marekm@amelek.gda.pl>
1266 * config/avr/avr.c (notice_update_cc): Only set condition code for
1267 ashrqi3 if shift count > 0.
1268 (out_shift_with_cnt): Handle shift count <= 0 as a no-op.
1269 (ashlqi3_out, ashlhi3_out, ashlsi3_out, ashrqi3_out, ashrhi3_out,
1270 ashrsi3_out, lshrqi3_out, lshrhi3_out, lshrsi3_out): Handle shift
1271 count <= 0 as a no-op, and shift count >= width by copying zero
1272 or sign bit to all bits of the result.
1273 * config/avr/avr.md (all shifts): Add alternatives for zero shift
1274 count, with attribute "length" set to 0 and "cc" set to "none".
1276 2005-01-26 Aldy Hernandez <aldyh@redhat.com>
1278 * doc/invoke.texi: Document -mTLS.
1280 * testsuite/gcc.target/frv/all-tls-global-dynamic.c: New.
1281 * testsuite/gcc.target/frv/all-tls-initial-exec.c: New.
1282 * testsuite/gcc.target/frv/all-tls-initial-exec-pic.c: New.
1283 * testsuite/gcc.target/frv/all-tls-local-dynamic.c: New.
1284 * testsuite/gcc.target/frv/all-tls-local-dynamic-plt.c: New.
1285 * testsuite/gcc.target/frv/all-tls-local-dynamic-plt-pic.c: New.
1286 * testsuite/gcc.target/frv/all-tls-local-exec.c: New.
1287 * testsuite/gcc.target/frv/all-tls-local-exec-TLS.c: New.
1289 * configure.ac: Check for a TLS capable gas.
1291 * configure: Regenerate.
1293 2005-01-26 Richard Henderson <rth@redhat.com>
1296 * combine.c (make_field_assignment): Simplify store to zero_extract
1297 from a source with an overlapping mask.
1299 2005-01-26 Aldy Hernandez <aldyh@redhat.com>
1301 2004-11-11 Eric Christopher <echristo@redhat.com>
1303 * config/frv/frv.md: Add fr400_integer automaton. Don't
1304 allow TYPE_MUL, TYPE_MACC, or TYPE_CUT to issue in parallel
1307 2005-01-26 Steven Bosscher <stevenb@suse.de>
1310 * cfgbuild.c (make_edges): Do not clear or set
1311 current_function_has_computed_jump.
1312 * function.h (struct function): Remove the has_computed_jump field.
1313 (current_function_has_computed_jump): Do not define.
1314 * sched-rgn.c (is_cfg_nonregular): Return true if a basic block ends
1315 in a computed jump. Ignore current_function_has_computed_jump.
1317 2005-01-26 Richard Henderson <rth@redhat.com>
1320 * expr.c (categorize_ctor_elements): New argument p_must_clear.
1321 (categorize_ctor_elements_1): Likewise. Detect a union that isn't
1323 (mostly_zeros_p): Update for new categorize_ctor_elements argument.
1324 * gimplify.c (gimplify_init_constructor): Likewise. Only shove
1325 objects into static storage if they have more than one non-zero value.
1326 * tree.h (categorize_ctor_elements): Update decl.
1328 2005-01-26 Steven Bosscher <stevenb@suse.de>
1331 * tree.h (CALL_EXPR_TAILCALL): Add comment.
1332 * calls.c (check_sibcall_argument_overlap_1): Revert the change
1333 to this function from 2004-07-10.
1334 * tree-tailcall.c (suitable_for_tail_opt_p): Do not consider the
1335 the current function for tail call optimizations if the address
1336 of one of it its arguments is taken.
1338 2005-01-26 Kazu Hirata <kazu@cs.umass.edu>
1340 * cse.c (fold_rtx): Call equiv_constant only when necessary.
1342 2005-01-26 Bernd Schmidt <bernd.schmidt@analog.com>
1344 * function.c (expand_function_end): If current_function_calls_alloca,
1345 emit stack restore in a place that is reached when the function falls
1348 2005-01-26 Alexandre Oliva <aoliva@redhat.com>
1350 * Makefile.in (../$(build_subdir)/fixincludes/fixincl): Add dummy
1351 rule to avoid matching inappropriate suffix rule.
1353 2005-01-26 Jan Hubicka <jh@suse.cz>
1355 PR tree-optimization/19241
1356 * tree-inline.c (copy_body_r): Do not walk subtrees after substituting.
1358 2005-01-26 Hans-Peter Nilsson <hp@axis.com>
1360 * config/cris/cris.h (LIB_SPEC) <sim*>: Use -lsyslinux, not
1361 -lsyssim. Wrap -lc -lsyssim in --start-group --end-group instead
1363 (STARTFILE_SPEC): Add crti.o, before crtbegin.o.
1364 <sim*>: Always use crt1.o, regardless of N in simN.
1365 <!sim>: Drop support for -pg and -p variants; always use crt0.o.
1366 (ENDFILE_SPEC): Add crtn.o, after crtend.o.
1367 (CRT_CALL_STATIC_FUNCTION): Remove.
1368 * config/cris/aout.h (STARTFILE_SPEC): <!melinux && sim*>: Always
1369 use crt1.o, regardless of N in simN.
1370 <!sim>: Drop support for -pg and -p variants; always use crt0.o.
1371 (LIB_SPEC): Do not define; default to the one in config/cris/cris.h.
1372 (CRIS_CPP_SUBTARGET_SPEC) <melinux && !nostdinc>: Remove special
1373 case isystem setting.
1374 (MAX_OFILE_ALIGNMENT): Correct to 32, not 16.
1376 2005-01-25 Ian Lance Taylor <ian@airs.com>
1378 * gimple-low.c (block_may_fallthru): Correct handling of
1379 SWITCH_EXPR--may fall through if SWITCH_LABELS is NULL.
1381 2005-01-26 Hans-Peter Nilsson <hp@axis.com>
1383 * config/cris/cris.h (CPP_SPEC): Adjust the various mtune=
1384 handlers to match indentation; nested under the mtune=* handler.
1385 Similar for mcpu= and march=.
1387 2005-01-25 Daniel Jacobowitz <dan@codesourcery.com>
1389 * config/mips/linux.h (LIB_SPEC): Remove -rpath-link.
1390 * config/mips/linux64.h (LIB_SPEC): Likewise.
1391 * config/mn10300/linux.h (LIB_SPEC): Likewise.
1392 * config/sh/linux.h (LIB_SPEC): Likewise.
1394 2005-01-25 Aldy Hernandez <aldyh@redhat.com>
1396 * config/frv/frv.c (frv_legitimize_tls_address): New.
1397 (TARGET_HAVE_TLS): Define.
1398 (FRV_SYMBOL_REF_TLS_P): Define.
1399 (frv_override_options): Handle new register classes.
1400 (frv_legitimate_address_p): Reject tls addresses.
1401 (frv_legitimize_address): Handle TLS addresses.
1402 (gen_inlined_tls_plt): New.
1404 (frv_legitimize_tls_address): New.
1405 (unspec_got_name): Add TLS entries.
1406 (got12_operand): Add R_FRV_TLSMOFF12 case.
1407 (frv_emit_move): Fixup TLS addresses.
1408 (frv_emit_movsi): Legitimize TLS addresses.
1410 * config/frv/frv.h (MASK_BIG_TLS): New.
1411 (TARGET_BIG_TLS): New.
1412 (HAVE_AS_TLS): Define.
1413 (TARGET_SWITCHES): Add -mTLS and -mtls options.
1414 (enum reg_class): Add GR8_REGS, GR9_REGS, GR89_REGS.
1415 (REG_CLASS_NAMES): Same.
1416 (REG_CLASS_CONTENTS): Same.
1417 (CONSTRAINT_LEN): New.
1418 (REG_CLASS_FROM_CONSTRAINT): New.
1419 (PREDICATE_CODES): Add symbolic_operand.
1421 * config/frv/frv.md (define_constants): Add UNSPEC_GETTLSOFF,
1422 UNSPEC_TLS_LOAD_GOTTLSOFF12, UNSPEC_TLS_INDIRECT_CALL,
1423 UNSPEC_TLS_TLSDESC_LDD, UNSPEC_TLS_TLSDESC_LDD_AUX,
1424 UNSPEC_TLS_TLSOFF_LD, UNSPEC_TLS_LDDI, UNSPEC_TLSOFF_HILO,
1425 R_FRV_GOTTLSOFF_HI, R_FRV_GOTTLSOFF_LO, R_FRV_TLSMOFFHI,
1426 R_FRV_TLSMOFFLO, R_FRV_TLSMOFF12, R_FRV_TLSDESCHI,
1427 R_FRV_TLSDESCLO, R_FRV_GOTTLSDESCHI, R_FRV_GOTTLSDESCLO, GR8_REG,
1428 GR9_REG, GR14_REG, LRREG.
1429 (type): Add load_or_call attribute.
1430 ("load_or_call"): New reservation.
1431 ("call_gettlsoff"): New.
1432 ("tls_indirect_call"): New.
1433 ("tls_load_gottlsoff12"): New.
1434 ("tlsoff_hilo"): New.
1435 ("tls_tlsdesc_ldd"): New.
1436 ("tls_tlsoff_ld"): New.
1439 * config/frv/frv-protos.h (symbolic_operand): Protoize.
1441 2005-01-25 Alexandre Oliva <aoliva@redhat.com>
1443 * config/frv/frv.h (ASM_OUTPUT_DWARF_DTPREL): Define.
1444 * config/frv/frv-protos.h (frv_output_dwarf_dtprel): Declare.
1445 * config/frv/frv.c (TLS_BIAS): Define.
1446 (frv_output_dwarf_dtprel): New.
1448 2005-01-26 Jakub Jelinek <jakub@redhat.com>
1450 PR rtl-optimization/19579
1451 * ifcvt.c (noce_try_cmove_arith): If emitting instructions to set up
1452 both A and B, see if they don't clobber registers the other expr uses.
1454 2005-01-25 J"orn Rennecke <joern.rennecke@st.com>
1456 * real.c (do_add): Initialize signalling and canonical members.
1458 * real.c (real_from_integer): Zero out destination.
1460 2005-01-18 Jan Hubicka <jh@suse.cz>
1462 PR tree-optimize/19337
1463 * tree-inline.c (initialize_inlined_parameters): Copy
1464 saved_static_chain_decl only.
1466 2005-01-25 Richard Henderson <rth@redhat.com>
1470 * config/i386/i386.c (x86_inter_unit_moves): Zero.
1471 (ix86_preferred_reload_class): Rewrite fp-constant section, with
1472 80387 enabled, to return a proper subclass. Return the subset that
1473 overlaps with GENERAL_REGS for PLUS.
1474 * config/i386/i386.md (movsi_1, movdi_2): Set type to "mmx" for pxor.
1476 2005-01-25 Richard Henderson <rth@redhat.com>
1479 * tree-complex.c (expand_complex_div_wide): Use the correct formulae.
1481 2005-01-25 Roger Sayle <roger@eyesopen.com>
1483 * expmed.c (choose_multiplier): Change interface to return the
1484 multiplier lowpart as an rtx instead of an unsigned HOST_WIDE_INT.
1485 (expand_divmod): Update calls to choose_multiplier.
1487 2005-01-24 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1490 * expmed.c (extract_bit_field): Use adjust_address instead of
1491 gen_lowpart when op0 is a MEM.
1493 2005-01-24 Steven Bosscher <stevenb@suse.de>
1495 * cfgexpand.c (expand_gimple_tailcall): Fix typo.
1497 2005-01-24 Richard Sandiford <rsandifo@redhat.com>
1499 * config/mips/irix-csr.c: New file.
1500 * config/mips/t-iris6 (irix-csr.o): New rule to build it.
1501 (EXTRA_MULTILIB_PARTS): Add irix-csr.o.
1502 * config/mips/iris6.h (ENDFILE_SPEC): Include it in n32 and n64
1505 2005-01-24 Eric Botcazou <ebotcazou@libertysurf.fr>
1508 * config.gcc (sparc-*-elf*): Remove sol2.h, sparc/sol2.h and
1509 sparc/elf.h, add sparc/sp-elf.h.
1510 (sparc-*-rtems*): Likewise.
1511 (sparclite-*-elf*): Remove sol2.h, sparc/sol2.h, sparc/elf.h and
1512 tm-dwarf2.h, add sparc/sp-elf.h.
1513 (sparc86x-*-elf): Likewise.
1514 (sparc64-*-elf*): Remove sol2.h, sparc/sol2.h and tm-dwarf2.h.
1515 * config/sparc/liteelf.h (TARGET_SUB_OS_CPP_BUILTINS): Rename into
1516 TARGET_OS_CPP_BUILTINS.
1517 * config/sparc/sp86x-elf (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1518 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1520 * config/sparc/openbsd64.h (NO_IMPLICIT_EXTERN_C): Undefine.
1521 * config/sparc/sp64-elf.h (NO_IMPLICIT_EXTERN_C): New macro.
1522 (SWITCH_TAKES_ARG): Likewise.
1523 (LOCAL_LABEL_PREFIX): Likewise.
1524 (ASM_GENERATE_INTERNAL_LABEL): Likewise.
1525 (TARGET_N_FORMAT_TYPES): Delete.
1526 (TARGET_FORMAT_TYPES): Likewise.
1527 (ASM_DECLARE_FUNCTION_SIZE): Likewise.
1528 * config/sparc/elf.h: Delete.
1529 * config/sparc/sp-elf.h: New file.
1531 2005-01-24 Kazu Hirata <kazu@cs.umass.edu>
1533 * tree-flow-inline.h (phi_arg_from_edge): Remove.
1534 * tree-flow.h: Remove the corresponding prototype.
1535 * tree-ssa-operands.h, tree-ssa-threadupdate.c, tree-ssa.c,
1536 tree-vectorizer.c: Use dest_idx instead of phi_arg_from_edge.
1538 2005-01-24 Jakub Jelinek <jakub@redhat.com>
1541 * config/rs6000/rs6000.c (rs6000_file_start): Create toc section
1542 for AIX ABI or ELF -fPIC.
1543 (rs6000_emit_load_toc_table): Don't create toc_section here.
1544 (rs6000_xcoff_file_start): Nor here.
1546 2005-01-24 J"orn Rennecke <joern.rennecke@st.com>
1548 * sh.c (ra.h): Don't #include.
1549 (hard_regs_intersect_p): New function, resurrected from ra.c.
1551 * sh.c: Fix 1996 Copyright.
1553 2005-01-24 Richard Henderson <rth@redhat.com>
1554 Aldy Hernandez <aldyh@redhat.com>
1556 * regrename.c (note_sets): Handle subregs.
1558 2005-01-24 Joseph S. Myers <joseph@codesourcery.com>
1561 * genconditions.c (write_header, write_conditions): Elide file if
1564 2005-01-24 Kazu Hirata <kazu@cs.umass.edu>
1566 * c-opts.c, expmed.c, gengtype.c, tree-complex.c,
1567 config/arm/iwmmxt.md, config/avr/avr.c, config/s390/s390.h:
1570 2005-01-24 Jakub Jelinek <jakub@redhat.com>
1572 * builtins.c (get_pointer_alignment, gimplify_va_arg_expr,
1573 expand_builtin_printf, expand_builtin_fprintf,
1574 expand_builtin_sprintf): Use POINTER_TYPE_P instead TREE_CODE
1575 checking against POINTER_TYPE.
1576 (validate_arglist): Handle POINTER_TYPE code by checking
1579 2005-01-24 Paolo Bonzini <bonzini@gnu.org>
1581 * rtlanal.c (non_rtx_starting_operands, for_each_rtx_1,
1583 (for_each_rtx): Call for_each_rtx_1.
1584 * rtl.h (init_rtlanal): Declare.
1585 * toplev.c (backend_init): Call init_rtlanal.
1587 2005-01-24 Jakub Jelinek <jakub@redhat.com>
1589 * flow.c (propagate_one_insn): Formatting.
1592 * flow.c (libcall_dead_p): Be more conservative if unsure.
1593 If there are any instructions between insn and call, see if they are
1594 all dead before saying the libcall is dead.
1596 2005-01-24 Uros Bizjak <uros@kss-loka.si>
1598 * config/i386/i386.md (*extendsfdf2_sse): Use "nonimmediate_operand"
1599 operand constraint for operand 0.
1601 2005-01-23 Richard Henderson <rth@redhat.com>
1604 * tree-complex.c (expand_complex_div_wide): Set EDGE_FALLTHRU on
1605 edges leading to join block.
1607 2005-01-23 Roger Sayle <roger@eyesopen.com>
1609 * expmed.c (expand_mult_highpart): Make static. Change type of
1610 constant multiplier argument from unsigned HOST_WIDE_INT to rtx.
1611 (expand_divmod): Updates calls to expand_mult_highpart by using
1612 gen_int_mode to pass a CONST_INT rtx.
1613 * rtl.h (expand_mult_highpart): Remove prototype.
1615 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
1617 * flow.c (regset_bytes, regset_size): Remove.
1619 * cfganal.c: Fix a reference to Harvey's paper.
1621 2005-01-23 Steven Bosscher <stevenb@suse.de>
1623 PR rtl-optimization/19464
1624 * tree-optimize.c (init_tree_optimization_passes): Add one more
1625 copyrename pass just before out-of-ssa.
1627 2005-01-23 Joseph S. Myers <joseph@codesourcery.com>
1630 * recog.c (recog_memoized): Don't define if GENERATOR_FILE.
1631 * ggc-none.c (ggc_free): Define.
1633 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
1635 * cse.c (max_reg, max_insn_uid): Remove.
1636 (cse_main): Don't access max_reg or max_insn_uid.
1638 * c-common.c (explicit_flag_signed_bitfields,
1639 lang_statement_code_p, lang_expand_function_end): Remove.
1640 * c-common.h: Remove the corresponding declarations.
1641 * c-opts.c (c_common_handle_option): Remove all write access
1642 to explicit_flag_signed_bitfields.
1644 * cgraph.c (cgraph_varpool_n_nodes): Remove.
1645 (cgraph_varpool_node): Don't access cgraph_varpool_n_nodes.
1646 * cgraph.h: Remove the corresponding declaration.
1648 * gcse.c (null_pointer_info): Remove.
1650 2005-01-23 Roger Sayle <roger@eyesopen.com>
1651 Eric Botcazou <ebotcazou@libertysurf.fr>
1653 * combine.c (simplify_logical): Only simplify logical expressions
1654 of the form ior(and(x,y),z) by the inverse distributive law if the
1655 result is cheaper than the original.
1657 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
1659 * alias.c, c-common.h, c-incpath.c, c-incpath.h, expr.c,
1660 fold-const.c, gimplify.c, params.h, tree-data-ref.c,
1661 tree-if-conv.c, tree-nested.c, tree-outof-ssa.c,
1662 tree-ssa-dom.c, tree-vectorizer.c, tree.def, config/darwin.c,
1663 config/freebsd-spec.h, config/arm/arm.h,
1664 config/h8300/h8300.md, config/i386/i386.md,
1665 config/i386/predicates.md, config/i386/sse.md,
1666 config/ia64/ia64.c, config/ip2k/ip2k.c, config/s390/s390.c,
1667 config/vax/vax.md: Fix comment typos. Follow spelling
1670 2005-01-23 Steven Bosscher <stevenb@suse.de>
1672 * gengtype.c (walk_type): Produce `not equal to' compares for loop
1673 tests, instead of `less than'.
1675 2005-01-23 Richard Sandiford <rsandifo@redhat.com>
1677 * function.c (assign_parm_setup_block): When creating a new stack slot
1678 for a parameter, get its alignment from the parameter's DECL_ALIGN
1679 rather than the type's TYPE_ALIGN. Make sure that the parameter
1680 is at least word aligned.
1682 2005-01-22 Roger Sayle <roger@eyesopen.com>
1685 * config/avr/avr.c (avr_hard_regno_mode_ok): Rewrite.
1687 2005-01-22 Richard Henderson <rth@redhat.com>
1690 * config/i386/i386.md (movsfcc_1_sse_max): Use nonimmediate_operand
1691 in both compare operands.
1692 (movdfcc_1_sse_max): Likewise.
1693 (movsfcc_1_sse): Likewise. Add earlyclobber for scratch.
1694 (movdfcc_1_sse): Likewise.
1695 * config/i386/i386.c (ix86_split_sse_movcc): Emit copies into the
1696 scratch register as needed.
1698 2005-01-22 Richard Henderson <rth@redhat.com>
1700 * config/i386/i386.md (smaxsf3): Fix mnemonic typo.
1702 * config/i386/i386.c (ix86_prepare_fp_compare_args): Fix is_sse test.
1704 2005-01-22 Richard Henderson <rth@redhat.com>
1706 * genopinit.c (optabs): Use smin/smax for floating point too.
1707 * doc/md.texi: Update to match. Clarify that floating point
1708 results are undefined for +0/-0 and NaN.
1709 * doc/rtl.texi: Likewise.
1710 * rtl.def (SMIN, SMAX): Likewise
1711 * tree.def (MIN_EXPR, MAX_EXPR): Likewise.
1713 * config/alpha/alpha.md (smaxdf3, smindf3, smaxsf3, sminsf3): Add
1714 leading 's' to the name.
1715 * config/ia64/ia64.md (smaxsf3, sminsf3, smaxdf3, smindf3,
1716 smaxxf3, sminxf3): Likewise.
1717 * config/rs6000/rs6000.md (smaxdf3, smindf3, smaxsf3, sminsf3):
1720 2005-01-22 Paul Brook <paul@codesourcery.com>
1722 * config/arm/arm.md: Use "Uy" constraint for wcgr load/stores.
1724 2005-01-21 Mark Dettinger <dettinge@de.ibm.com>
1726 * config/s390/s390.c (struct processor_costs): New fields
1727 dlgr, dlr, dr, dsgfr, dsgr.
1728 (z900_cost, z990_cost): Values for new fields.
1729 (s390_rtx_costs): New cases MEM und COMPARE in switch
1730 statement. Modified handling of SIGN_EXTEND, ZERO_EXTEND,
1731 DIV, MOD, UDIV, UMOD.
1733 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
1735 * config/s390/s390.md ("doloop_si64"): Reload input value directly
1736 into the register being decremented.
1737 ("doloop_si31", "doloop_di"): Likewise.
1738 ("*doloop_si_long"): Adapt pattern.
1740 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
1742 * config/s390/s390.h (HARD_REGNO_NREGS): Fix computation for
1744 (CLASS_MAX_NREGS): Likewise.
1746 2005-01-21 Daniel Berlin <dberlin@dberlin.org>
1748 * doc/tree-ssa.texi (Statement Operands): Add example for new
1749 must-def macro. Note deprecation of old operands interface.
1751 2005-01-22 Richard Sandiford <rsandifo@redhat.com>
1753 PR tree-optimization/19484
1754 * tree-cfg.c (remove_fallthru_edge): New function.
1755 (cleanup_control_flow): Remove fallthru edges from calls that are
1756 now known not to return.
1758 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
1760 Fix PR tree-optimization/19038
1761 * tree-ssa-dom.c (cprop_operand): Don't replace loop invaeriant
1762 copies with loop variant ones.
1764 2005-01-22 Kazu Hirata <kazu@cs.umass.edu>
1766 * cfganal.c, real.h, reorg.c, timevar.def, tree-ssa-ccp.c,
1767 config/alpha/alpha-protos.h, config/alpha/alpha.h,
1768 config/alpha/alpha.md, config/alpha/predicates.md,
1769 config/sparc/freebsd.h, config/sparc/netbsd-elf.h,
1770 config/sparc/sol2.h: Update copyright.
1772 * tree-cfg.c (remove_forwarder_block_with_phi): Look at the
1773 first label to see if it is a nonlocal label.
1775 2005-01-22 David Edelsohn <edelsohn@gnu.org>
1776 Andrew Pinski <pinskia@physics.uc.edu>
1779 * config/rs6000/rs6000.c (rs6000_va_start): Saturate n_gpr at
1780 maximum number of GPRs. Saturate n_fpr at maximum number of FPRs.
1782 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
1784 * config/rs6000/rtems.h: Update copyright.
1786 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
1789 * config/rs6000/rtems.h: Resurrect cpp_os_rtems_spec from gcc < 3.4.
1790 (CPP_OS_RTEMS_SPEC): New (From gcc-3.3's config/rs6000/sys4.h).
1791 (SUBSUBTARGET_EXTRA_SPECS): Use CPP_OS_RTEMS_SPEC.
1793 2005-01-22 Volker Reichelt <reichelt@igpm.rwth-aachen.de>
1796 * c-typeck.c (convert_arguments): Check for error_mark_node.
1798 2005-01-21 Ian Lance Taylor <ian@airs.com>
1800 PR tree-optimization/13000
1801 * tree-inline.c: Include "tree-flow.h".
1802 (expand_call_inline): If warn_return_type, warn if non-void inline
1803 function falls through.
1804 * tree-cfg.c (execute_warn_function_return): Don't warn about
1805 control reaching end if TREE_NO_WARNING is set. Set
1807 * gimple-low.c (block_may_fallthru): Don't assume that SWITCH_EXPR
1809 * gimplify.c (shortcut_cond_expr): Don't emit a jump over the else
1810 branch if we don't need one.
1811 * c-typeck.c: Include "tree-flow.h"
1812 (c_finish_bc_stmt): Don't add a goto if the current statement
1813 list doesn't fall through to the current point.
1815 2005-01-21 Roger Sayle <roger@eyesopen.com>
1817 PR rtl-optimization/576
1818 * real.c (real_arithmetic): Change return type from void to bool
1819 to return an indication that the result may be inexact.
1820 * real.h (real_arithmeric): Update prototype.
1821 * fold-const.c (const_binop): Don't constant fold floating
1822 point expressions when the user specifies -frounding-math and
1823 the result may depend upon the run-time rounding mode.
1824 (fold_convert_const_real_from_real): Clean-up.
1825 (fold_initializer): Ignore flag_rounding_math for initializers.
1826 * simplify-rtx.c (simplify_binary_operation): Likewise, don't
1827 constant fold FP operations with flag_rounding_math if the
1828 result may depend upon the run-time rounding mode.
1830 2005-01-21 Tom Tromey <tromey@redhat.com>
1832 * c-cppbuiltin.c (define__GNUC__): Correct assertion.
1834 2005-01-21 Kazu Hirata <kazu@cs.umass.edu>
1836 * tree-cfg.c: Fix comment typos.
1838 2005-01-21 J"orn Rennecke <joern.rennecke@st.com>
1840 * passes.c (rest_of_handle_flow2): Close / open the flow2 dump file
1841 around the call to rest_of_handle_branch_target_load_optimize.
1843 2005-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
1845 * config.gcc (sparc-*-netbsdelf*): Include dbxelf.h
1846 (sparc64-*-openbsd*): Likewise.
1847 (sparclite-*-elf*): Include tm-dwarf2.h.
1848 (sparc86x-*-elf*): Likewise.
1849 (sparc64-*-elf*): Likewise.
1850 * config/sparc/linux64.h (DWARF2_DEBUGGING_INFO): Delete.
1851 (DBX_DEBUGGING_INFO): Likewise.
1852 * config/sparc/liteelf.h (DWARF2_DEBUGGING_INFO): Likewise.
1853 (PREFERRED_DEBUGGING_TYPE): Likewise.
1854 * config/sparc/netbsd-elf.h (PREFERRED_DEBUGGING_TYPE): Likewise.
1855 * config/sparc/sp64-elf.h (DBX_DEBUGGING_INFO): Likewise.
1856 (PREFERRED_DEBUGGING_TYPE): Likewise.
1857 * config/sparc/sp86x-elf.h (DWARF2_DEBUGGING_INFO): Likewise.
1858 (PREFERRED_DEBUGGING_TYPE): Likewise.
1859 * config/sparc/sparc.h (DBX_DEBUGGING_INFO): Likewise.
1861 2005-01-21 Hans-Peter Nilsson <hp@bitrange.com>
1864 * combine.c (combine_simplify_rtx): Revert change of 2004-12-31.
1866 2005-01-20 Paul Brook <paul@codesourcery.com>
1868 * doc/rtl.texi: Document value extension requirements for CONST_INT.
1870 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
1872 * tree-cfg.c (tree_verify_flow_info): Check that a nonlocal
1873 label is first in a sequence of labels.
1875 * tree-cfg.c (tree_verify_flow_info): Fix a typo.
1877 2005-01-20 Janis Johnson <janis187@us.ibm.com>
1878 Giovanni Bajo <giovannibajo@gcc.gnu.org>
1880 * doc/sourcebuild.texi (Test Directives): New.
1882 2005-01-20 Roger Sayle <roger@eyesopen.com>
1884 * real.c (real_floor): Don't leave the result uninitialized when
1885 mode is VOIDMode, but return the unrounded intermediate value.
1886 (real_ceil): Likewise.
1888 2005-01-20 Paul Brook <paul@codesourcery.com>
1890 * config/arm/arm.md (insv): Use gen_int_mode.
1892 2005-01-20 Eric Botcazou <ebotcazou@libertysurf.fr>
1894 * config/sparc/freebsd.h (ASM_OUTPUT_INTERNAL_LABELREF): Delete.
1895 * config/sparc/linux.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1896 * config/sparc/linux64.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1897 * config/sparc/netbsd-elf.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1898 * config/sparc/sol2.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1900 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
1903 * dwarf2out.c (concat_loc_descriptor): We don't know we can use
1904 fbreg, so we have to assume we can't.
1905 (loc_descriptor_from_tree_1): Ditto.
1906 (containing_function_has_frame_base): New function.
1907 (add_location_or_const_value_attribute): Use it.
1908 Also try to generate a frame_base from a single element location
1911 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
1913 PR tree-optimization/15349
1914 * timevar.def (TV_TREE_MERGE_PHI): New.
1915 * tree-cfg.c (tree_forwarder_block_p): Add a new argument
1917 (remove_forwarder_block, cleanup_forwarder_blocks): Adjust the
1918 calls to tree_forwarder_block_p.
1919 (remove_forwarder_block_with_phi, merge_phi_nodes,
1920 gate_merge_phi, pass_merge_phi): New.
1921 * tree-optimize.c (init_tree_optimization_passes): Add
1923 * tree-pass.h: Add an extern for pass_merge_phi;
1925 2005-01-20 Richard Henderson <rth@redhat.com>
1928 * config/i386/emmintrin.h (_mm_castpd_ps, _mm_castpd_si128): New.
1929 (_mm_castps_pd, _mm_castps_si128): New.
1930 (_mm_castsi128_ps, _mm_castsi128_pd): New.
1932 2005-01-20 Richard Henderson <rth@redhat.com>
1935 * config/i386/mmintrin.h (_mm_cvtsi32_si64): Use
1936 __builtin_ia32_vec_init_v2si.
1937 (_mm_cvtsi64_si32): Use __builtin_ia32_vec_ext_v2si.
1938 * config/i386/i386.c (IX86_BUILTIN_VEC_EXT_V2SI): New.
1939 (ix86_init_mmx_sse_builtins): Create it.
1940 (ix86_expand_builtin): Expand it.
1941 (ix86_expand_vector_set): Handle V2SFmode and V2SImode.
1942 * config/i386/mmx.md (vec_extractv2sf_0, vec_extractv2sf_1): New.
1943 (vec_extractv2si_0, vec_extractv2si_1): New.
1945 2005-01-20 Richard Henderson <rth@redhat.com>
1948 * config/i386/i386.c (ix86_expand_push): New.
1949 * config/i386/mmx.md (push<MMXMODE>1): New.
1950 * config/i386/sse.md (push<SSEMODE>1): New.
1951 * config/i386/i386-protos.h: Update.
1953 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
1955 * params.def, params.h, predict.c, tree-ssa-loop-im.c: Update
1958 2005-01-20 Richard Henderson <rth@redhat.com>
1960 * rtl.def (CONST_VECTOR): Use RTX_CONST_OBJ.
1961 * rtl.h (CONSTANT_P): Don't special case CONST_VECTOR.
1963 2005-01-19 Richard Henderson <rth@redhat.com>
1966 * config/i386/i386.c (ix86_expand_vector_move_misalign): Convert
1967 to V4SFmode in SSE1 fallback load path.
1969 2005-01-19 Richard Henderson <rth@redhat.com>
1971 * config/i386/i386.c (ix86_expand_vector_init_one_var): Fix typo
1972 in QImode expansion to ix86_expand_vector_set.
1973 (ix86_expand_vector_init_general): Fix typo in V8HImode recursive call.
1975 2005-01-19 Richard Henderson <rth@redhat.com>
1978 * config/i386/i386.c (ix86_preferred_reload_class): Return a proper
1979 subclass of the input class.
1980 (ix86_secondary_memory_needed): Always true for cross-MMX classes.
1981 Always true for cross-SSE1 classes. Rationalize conditionals.
1982 * config/i386/i386.h (SSE_CLASS_P, MMX_CLASS_P): Use straight equality.
1983 * config/i386/i386.md (movsi_1): Add MMX/SSE zeros. Fix alternatives
1984 for SSE1. Don't check TARGET_INTER_UNIT_MOVES.
1985 (movdi_2): Add MMX/SSE zeros.
1986 (movdi_1_rex64): Likewise. Don't check TARGET_INTER_UNIT_MOVES.
1987 (movsf_1): Don't check TARGET_INTER_UNIT_MOVES.
1988 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
1989 (movsi_1_nointernunit, movdi_1_rex64_nointerunit): Remove.
1990 (movsf_1_nointerunit, zero_extendsidi2_32_1): Remove.
1991 (zero_extendsidi2_rex64_1): Remove.
1992 (MOV0 peephole): Check GENERAL_REG_P.
1994 2005-01-19 Richard Henderson <rth@redhat.com>
1997 * config/i386/i386.c (ix86_expand_vector_set): Fix third and fourth
1999 (ix86_expand_vector_extract): Likewise.
2001 2005-01-19 Richard Henderson <rth@redhat.com>
2004 * expr.c (emit_move_change_mode): New argument force; use
2005 simplify_gen_subreg if true.
2006 (emit_move_via_alt_mode): Merge into ...
2007 (emit_move_via_integer): ... here.
2008 (emit_move_ccmode): Use emit_move_change_mode directly.
2010 2005-01-19 Richard Henderson <rth@redhat.com>
2013 * config/alpha/alpha.c (alpha_rtx_costs): Handle HIGH.
2014 (alpha_preferred_reload_class): Handle CONST_VECTOR.
2015 (alpha_emit_set_const_1): Add no_output parameter; don't emit
2017 (alpha_emit_set_const): Likewise. Make static.
2018 (alpha_emit_set_long_const): Make static.
2019 (alpha_extract_integer): Split out from alpha_expand_mov.
2020 (alpha_split_const_mov): Likewise.
2021 (alpha_expand_mov): Use them. Handle CONST_VECTOR.
2022 (alpha_legitimate_constant_p): New.
2023 * config/alpha/alpha-protos.h: Update.
2024 * config/alpha/alpha.h (REGISTER_MOVE_COST): Correct fp<->gp cost.
2025 (LEGITIMATE_CONSTANT_P): Re-implement with a function.
2026 * config/alpha/alpha.md (movsi): Add n alternative.
2027 (movsi_nt_vms, movdi_er_nofix, movdi_er_fix, movdi_fix): Likewise.
2028 (mov<VEC>_fix, mov<VEC>_nofix): Add i alternative.
2029 (splitters for all of the above): Use alpha_split_const_mov.
2030 * config/alpha/predicates.md (non_add_const_operand): New.
2031 (non_zero_const_operand): New.
2032 (input_operand): Use alpha_legitimate_constant_p after reload.
2034 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
2036 PR tree-optimization/19038
2037 * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): New function.
2038 (add_candidate): Add ivs with increment in latch only if
2039 allow_ip_end_pos_p is true.
2040 (determine_iv_cost): Use empty_block_p.
2042 2005-01-19 Daniel Berlin <dberlin@dberlin.org>
2044 * cfganal.c (compute_dominance_frontiers_1): Replace with new algorithm
2045 (compute_dominance_frontiers): Ditto.
2047 2005-01-19 Ralf Corsepius <ralf.corsepius@rtems.org>
2050 * config/sh/t-rtems: New.
2051 * config.gcc (sh-*-rtems*): Reflect having added config/sh/t-rtems.
2053 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
2055 * tree-ssa-ccp.c (maybe_fold_offset_to_component_ref): Always subtract
2056 the offset of the selected field.
2058 2005-01-19 Kazu Hirata <kazu@cs.umass.edu>
2060 * tree-cfg.c (remove_forwarder_block): Fix the check to
2061 prevent a nonlocal label from appearing in the middle of a
2064 2005-01-19 Hans-Peter Nilsson <hp@axis.com>
2066 PR rtl-optimization/19462
2067 * reorg.c (find_end_label): Create return insn only if
2068 current_function_epilogue_delay_list is empty.
2070 2005-01-19 Paolo Bonzini <bonzini@gnu.org>
2072 * doc/passes.texi: Remove paragraph mentioning new-ra.
2074 2005-01-19 Jakub Jelinek <jakub@redhat.com>
2076 PR rtl-optimization/15139
2077 * combine.c: Include params.h.
2078 (count_rtxs): New function.
2079 (record_value_for_reg): If replace_rtx would replace at least
2080 2 occurrences of REG in VALUE and TEM is really large, replace REG with
2081 (clobber (const_int 0)) instead of TEM.
2082 * params.def (PARAM_MAX_LAST_VALUE_RTL): New.
2083 * params.h (MAX_LAST_VALUE_RTL): New.
2084 * Makefile.in (combine.o): Depend on $(PARAMS_H).
2085 * doc/invoke.texi (--param max-last-value-rtl=N): Document.
2088 * c-typeck.c (digest_init): Only call build_vector if all constructor
2089 elements are *_CST nodes.
2090 * gimplify.c (gimplify_init_constructor): Likewise.
2093 * c-typeck.c (digest_init): Only call build_vector if inside_init
2096 2005-01-18 Tobias Schl"uter <tobias.schlueter@physik.uni-muenchen.de>
2098 * toplev.c (init_asm_output): Remove dead #ifdef.
2100 2005-01-18 Kaz Kojima <kkojima@gcc.gnu.org>
2102 * config/sh/linux.h (TARGET_C99_FUNCTIONS): Define.
2104 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
2106 * calls.c (expand_call): Check DECL_BUILT_IN_CLASS before
2107 accessing DECL_FUNCTION_CODE.
2108 * dojump.c (do_jump): Likewise.
2109 * gimplify.c (gimplify_call_expr): Likewise.
2110 * predict.c (expr_expected_value): Likewise.
2111 (strip_builtin_expect): Likewise.
2112 * tree-inline.c (estimate_num_insns_1): Likewise.
2113 * tree-ssa-loop-im.c (stmt_cost): Likewise
2114 * fold-const.c (fold): Test for BUILT_IN_NORMAL.
2115 (tree_expr_nonnegative_p): Likewise.
2117 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
2119 * config/sparc/sparc.c (load_pic_register): Emit the appropriate
2120 variant of the load_pcrel_sym pattern.
2121 * config/sparc/sparc.md (P macro): Move to the top.
2122 (load_pcrel_sym): Macroize using P.
2123 (save_register_window): Likewise.
2125 2005-01-18 Aldy Hernandez <aldyh@redhat.com>
2127 * config/rs6000/rs6000.md ("sunordered"): Disable for e500.
2130 2005-01-18 Andrew Pinski <pinskia@physics.uc.edu>
2133 * c-typeck.c (build_asm_expr): Strip nops off
2134 input memory operands.
2136 2005-01-18 David Edelsohn <edelsohn@gnu.org>
2138 * config/rs6000/aix43.h (CPLUSPLUS_CPP_SPEC): Delete
2139 _XOPEN_SOURCE, _XOPEN_SOURCE_EXTENDED, _LARGE_FILE_API.
2140 * config/rs6000/aix51.h (CPLUSPLUS_CPP_SPEC): Same.
2141 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Same.
2143 2005-01-18 Joel Sherrill <joel.sherrill@oarcorp.com>
2146 * config/i386/i386.c (override_options): If the 80387 is disabled,
2147 then do not return FP values using FP registers.
2149 2005-01-18 Kazu Hirata <kazu@cs.umass.edu>
2151 * basic-block.h, c-common.c, c-cppbuiltin.c, c-lang.c,
2152 c-tree.h, cfgbuild.c, cgraph.c, cgraph.h, collect2.c,
2153 combine.c, config.gcc, coverage.h, cse.c, cselib.c,
2154 defaults.h, df.c, dwarf2asm.c, dwarf2out.c, explow.c, expr.c,
2155 flow.c, fold-const.c, gcse.c, ggc-page.c, gimple-low.c,
2156 gimplify.c, ifcvt.c, langhooks-def.h, lcm.c, optabs.h,
2157 output.h, postreload-gcse.c, postreload.c, recog.c,
2158 resource.c, rtl.def, rtlanal.c, sched-deps.c, sched-rgn.c,
2159 targhooks.h, toplev.c, tree-data-ref.c, tree-eh.c,
2160 tree-flow.h, tree-loop-linear.c, tree-mudflap.h, tree-nrv.c,
2161 tree-optimize.c, tree-outof-ssa.c, tree-pass.h,
2162 tree-scalar-evolution.c, tree-ssa-copy.c, tree-ssa-dce.c,
2163 tree-ssa-dse.c, tree-ssa-loop-ivopts.c, tree-ssa-operands.c,
2164 tree-ssa-pre.c, tree-ssa.c, tree-vectorizer.c, tree.def,
2165 unwind-dw2-fde-darwin.c, var-tracking.c: Update copyright.
2167 * config/darwin.c, config/darwin.h, config/freebsd-spec.h,
2168 config/alpha/alpha.c, config/arm/vfp.md,
2169 config/i386/emmintrin.h, config/i386/i386-modes.def,
2170 config/i386/i386-protos.h, config/i386/i386.h,
2171 config/i386/pmmintrin.h, config/i386/ppro.md,
2172 config/i386/predicates.md, config/i386/xmmintrin.h,
2173 config/i860/i860.c, config/ia64/ia64-modes.def,
2174 config/ia64/ia64-protos.h, config/ia64/ia64.h,
2175 config/ia64/ia64.md, config/ia64/itanium1.md,
2176 config/ia64/itanium2.md, config/ia64/predicates.md,
2177 config/iq2000/iq2000.h, config/mips/linux64.h,
2178 config/rs6000/aix41.h, config/rs6000/aix43.h,
2179 config/rs6000/aix52.h, config/rs6000/darwin-fallback.c,
2180 config/rs6000/rs6000.c, config/rs6000/rs6000.h,
2181 config/rs6000/spe.md, config/sh/sh.md, config/sparc/linux.h,
2182 config/sparc/linux64.h, config/sparc/litecoff.h,
2183 config/sparc/sp64-elf.h, config/sparc/sparc.c,
2184 config/sparc/sparc.h, config/sparc/sparc.md,
2185 config/sparc/sysv4-only.h: Update copyright.
2187 2005-01-18 Richard Henderson <rth@redhat.com>
2190 * config/i386/i386.c (ix86_expand_fp_movcc): Fail for LTGT and UNEQ.
2192 2005-01-18 Marc Espie <espie@openbsd.org>
2193 * config/i386/openbsdelf.h: Typo.
2195 2005-01-18 Richard Henderson <rth@redhat.com>
2197 * config/i386/i386.c (ix86_expand_fp_movcc): Remove TARGET_IEEE_FP
2198 special case for sse.
2199 * config/i386/predicates.md (sse_comparison_operator): Likewise.
2201 2005-01-18 Dorit Naishlos <dorit@il.ibm.com>
2203 * tree-inline.c (estimate_num_insns_1): Added cases for
2204 ALIGN_INDIRECT_REF, MISALIGNED_INDIRECT_REF, and REALIGN_LOAD_EXPR.
2206 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
2208 PR rtl-optimization/19296
2209 * combine.c (simplify_comparison): Rewrite the condition under
2210 which a non-paradoxical SUBREG of a PLUS can be lifted when
2211 compared against a constant.
2213 2005-01-18 Andi Kleen <ak@muc.de>
2215 * c-typeck.c: (convert_for_assignment): Check warn_pointer_sign.
2216 * c.opt (-Wpointer-sign): Add.
2217 * doc/invoke.texi: (-Wpointer-sign): Add.
2219 2005-01-18 Uros Bizjak <uros@kss-loka.si>
2221 * config/i386/i386.c (override_options): Revert 2004-11-24 change.
2222 * config/i386/i386.md (fmodsf3, dremsf3, *sinsf2, *cossf2, sincossf3,
2223 *tansf3_1, tansf2, atan2sf3_1, atan2sf3, atansf2, asinsf2, acossf2,
2224 logsf2, log10sf2, log2sf2, log1psf2, logbsf2, ilogbsf2, expsf2,
2225 exp10sf2, exp2sf2, expm1sf2, rintsf2, floorsf2, ceilsf2, btruncsf2,
2226 nearbyintsf2): Disable for TARGET_SSE_MATH. Leave patterns enabled
2227 for TARGET_MIX_SSE_I387.
2228 (fmoddf3, dremdf3, *sindf2, *sinextendsfdf2, *cosdf2, *cosextendsfdf2,
2229 sincosdf3, *sincosextendsfdf3, *tandf3_1, tandf2, atan2df3_1,
2230 atan2df3, atandf2, asindf2, acosdf2, logdf2, log10df2, log2df2,
2231 log1pdf2, logbdf2, expdf2, exp10df2, exp2df2, expm1df2, rintdf2,
2232 floordf2, ceildf2, btruncdf2, nearbyintdf2): Disable for
2233 (TARGET_SSE2 && TARGET_SSE_MATH). Leave patterns enabled for
2234 TARGET_MIX_SSE_I387.
2235 (atan2sf3, atan2df3, atan2xf3): Remove register constraints
2238 2005-01-18 Uros Bizjak <uros@kss-loka.si>
2241 * config/i386/mmx.md (*movv2sf_internal_rex64, *movv2sf_internal):
2242 Add movaps alternative for xmm reg->reg move.
2244 2005-01-17 Jeff Law <law@redhat.com>
2246 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Clear SSA_NAME_VALUE
2247 at the end of the main DOM loop rather than just before DOM exits.
2249 2005-01-17 Diego Novillo <dnovillo@redhat.com>
2251 PR tree-optimization/19121
2252 * tree-ssa-alias.c (compute_flow_sensitive_aliasing): When
2253 adding aliases to a name tag, also add them to the pointer's
2255 * tree-ssa-copy.c (merge_alias_info): Do not merge flow
2256 sensitive alias info at all. Only check that the two pointers
2257 have compatible pointed-to sets.
2258 * tree-ssa.c (verify_name_tags): Verify that the alias set of
2259 a pointer's type tag is a superset of the alias set of the
2262 2005-01-17 James E Wilson <wilson@specifixinc.com>
2265 * config/ia64/ia64.md (movxf): Handle general register source. Adjust
2266 comment to document why.
2268 2005-01-17 Richard Henderson <rth@redhat.com>
2270 * config/i386/sse.md (smaxv4sf3_finite, sse_vmsmaxv4sf3_finite,
2271 sminv4sf3_finite, sse_vmsminv4sf3_finite, smaxv2df3_finite,
2272 sse2_vmsmaxv2df3_finite, sminv2df3_finite,
2273 sse2_vmsminv2df3_finite): New.
2274 (smaxv4sf3, sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, smaxv2df3,
2275 sse2_vmsmaxv2df3, sminv2df3, sse2_vmsminv2df3): Remove commutative.
2276 Force op1 into register.
2278 2005-01-17 Kelley Cook <kcook@gcc.gnu.org>
2280 * Makefile.in (STAGEMOVESTUFF): Stage all the stamp files to prevent
2281 bubblestrap from always rebuilding build directory.
2283 2005-01-17 Ulrich Weigand <uweigand@de.ibm.com>
2285 * config/s390/s390.md ("*clc"): Remove incorrect '='.
2286 ("*cmpmem_short"): Likewise.
2288 2005-01-17 Steven Bosscher <stevenb@suse.de>
2290 * basic-block.h: Document BB_* flags.
2291 * regrename.c (copyprop_hardreg_forward): Don't use BB_VISITED,
2292 use an sbitmap instead.
2293 * sched-rgn.c (compute_trg_info): Likewise.
2295 2005-01-17 Richard Sandiford <rsandifo@redhat.com>
2297 * config.gcc (mips64*-*-linux*): Set the default abi to n32. Remove
2298 redundant target_cpu_default and extra_parts lines. Use the default
2300 * config/mips/linux64.h (DRIVER_DEFAULT_ABI_SELF_SPEC): Delete.
2301 (SUBTARGET_EXTRA_SPECS): Delete.
2302 (DRIVER_SELF_SPECS): Remove architecture lines. Set the ABI directly.
2303 (SUBTARGET_ASM_SPEC): Remove -mabi=n32 and -mabi=64 mappings.
2304 * config/mips/t-linux64 (MULTILIB_DIRNAMES): Use "32" for -mabi=32
2305 and "n32" for -mabi=n32.
2307 2005-01-17 Ian Lance Taylor <ian@airs.com>
2310 * c-typeck.c (build_c_cast): Revert patch of 2003-02-16: don't
2311 fold constant variables into initial values.
2313 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
2315 * tree-cfg.c (tree_can_merge_blocks_p): Reorder two checks.
2317 * tree-cfg.c (tree_forwarder_block_p): Speed up by walking
2318 through the statements backward.
2320 2005-01-17 Ian Lance Taylor <ian@airs.com>
2322 PR middle-end/13127:
2323 * tree-inline.c (expand_call_inline): Set TREE_NO_WARNING on
2324 a variable set to the return value of the inlined function.
2326 2005-01-17 Mark Dettinger <dettinge@de.ibm.com>
2328 * rtlanal.c (rtx_cost): Assign cost of 0 to a SUBREG
2329 when modes are tieable.
2331 2005-01-17 Ranjit Mathew <rmathew@hotmail.com>
2333 * gthr-posix.h (__gthread_active_p): Use pthread_cancel instead
2334 of pthread_create to find out if threads are enabled.
2335 * gthr-posix95.h (__gthread_active_p): Likewise.
2337 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
2339 * common.opt (-fnew-ra): Remove.
2341 * toplev.h (flag_new_regalloc): Remove.
2342 * Makefile.in (ra*.*): Don't mention.
2343 * passes.c (rest_of_handle_new_regalloc): Remove.
2344 (rest_of_handle_combine, rest_of_compilation): Always consider
2345 flag_new_regalloc as false.
2346 * doc/invoke.texi: Don't document -fnew-ra.
2348 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
2350 * bb-reorder.c (fix_edges_for_rarely_executed_code): Remove
2351 last parameter to reg_scan.
2352 * loop.c (loop_optimize): Likewise.
2353 * passes.c (rest_of_handle_tracer, rest_of_handle_if_conversion,
2354 rest_of_handle_web, rest_of_handle_cfg, rest_of_handle_jump_bypass,
2355 rest_of_handle_life, rest_of_handle_cse, rest_of_handle_cse2,
2356 rest_of_handle_gcse, rest_of_handle_loop_optimize,
2357 rest_of_handle_loop2, rest_of_handle_jump2): Likewise.
2358 * regclass.c (reg_scan): Likewise, for the declaration.
2359 * rtl.h (reg_scan): Likewise, for the prototype.
2361 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
2363 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Don't call
2364 mark_dfs_back_edges.
2366 2005-01-16 Daniel Berlin <dberlin@dberlin.org>
2368 * tree-ssa-pre.c (add_to_sets): s1 may be NULL.
2369 (compute_avail): Uses don't go in tmp_gen.
2371 2005-01-16 Steven Bosscher <stevenb@suse.de>
2373 * ggc-page.c (ggc_alloc_stat): Use __builtin_ctzl instead of a
2374 loop to look for a free slot in a page entry.
2376 2005-01-16 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2379 * defaults.h (TARGET_DEFERRED_OUTPUT_DEFS): Provide default.
2380 * toplev.c (compile_file): Call process_pending_assemble_output_defs
2381 just before targetm.asm_out.file_end.
2382 * tree.h (process_pending_assemble_output_defs): Declare.
2383 * varasm.c (assemble_output_def, process_pending_assemble_output_defs):
2385 (assemble_alias): Defer generation of assembly code for defines when
2386 TARGET_DEFERRED_OUTPUT_DEFS is true.
2387 * config/rs6000/aix41.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
2388 * config/rs6000/aix43.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
2389 * doc/tm.texi (TARGET_DEFERRED_OUTPUT_DEFS): document.
2391 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2394 * pa.c (pa_scalar_mode_supported_p): New function.
2395 (TARGET_SCALAR_MODE_SUPPORTED_P): Define.
2396 * pa.h (MIN_UNITS_PER_WORD): Add comment.
2398 * pa-protos.h (prefetch_operand): Delete.
2399 (prefetch_cc_operand, prefetch_nocc_operand): New declations.
2400 * pa.c (prefetch_operand): Delete.
2401 (prefetch_cc_operand, prefetch_nocc_operand): New functions.
2402 * pa.h (EXTRA_CONSTRAINT): Add `W' constraint.
2403 (PREDICATE_CODES): Delete prefetch_operand. Add prefetch_cc_operand
2404 and prefetch_nocc_operand.
2405 * pa.md (prefetch): Rework to avoid reload problems handling short
2406 displacements when a cache control completer needs to be provided.
2407 (prefetch_32, prefetch_64): Delete.
2408 (prefetch_cc, prefetch_nocc): New patterns.
2410 2005-01-15 David Edelsohn <edelsohn@gnu.org>
2412 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Revert previous change.
2414 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2417 * pa.c (pa_asm_output_aligned_common): Warn if specified alignment
2418 exceeds maximum alignment for global common data.
2419 * pa64-hpux.h (MAX_OFILE_ALIGNMENT): Define to 32768.
2420 * som.h (MAX_OFILE_ALIGNMENT): Likewise.
2422 2004-01-15 Roger Sayle <roger@eyesopen.com>
2424 * tree-ssa-dom.c (extract_range_from_cond): Correct condition.
2426 2004-01-15 Roger Sayle <roger@eyesopen.com>
2428 * harg-reg-set.h (reg_class_names): Prototype global array.
2429 * regclass.c (reg_class_names): Declare here and initialize to
2431 (dump_regclass): Remove local declaration of reg_class_names.
2432 (regclass): Likewise.
2433 * cfg.c (dump_flow_info): Likewise.
2434 * ra-debug.c (reg_class_names): Likewise.
2435 * regrename.c (reg_class_names): Likewise.
2436 * reload.c (reg_class_names): Likewise.
2437 * reload1.c (spill_failure): Likewise.
2438 * config/m68hc11/m68hc11.c (reg_class_names): Likewise.
2440 2005-01-15 Ulrich Weigand <uweigand@de.ibm.com>
2442 * config/s390/s390.md ("reload_outti"): Remove predicate for
2443 output operand. Abort if operand is not a MEM.
2444 ("reload_outdi", "reload_outdf"): Likewise.
2446 2005-01-15 Marc Espie <espie@openbsd.org>
2448 * config.gcc (*-*-openbsd*): Set HAS_LIBC_R for OpenBSD <= 3.2.
2449 (i[34567]86-*-openbsd*): Switch to ELF for OpenBSD >= 3.4.
2450 * config/openbsd.h: Fix C++ includes for native configurations.
2451 Add proper OS_CPP_BUILTINS. Fix libspec for recent OpenBSD.
2452 Add trampoline support.
2453 * config/i386/openbsdelf.h: New.
2455 2005-01-15 Marc Espie <espie@openbsd.org>
2457 * collect2.c (main): Explicitly parse -dynamic-linker option.
2459 2005-01-15 Jakub Jelinek <jakub@redhat.com>
2461 PR tree-optimization/19060
2462 * tree-ssa-dom.c (extract_range_from_cond) <case LT_EXPR, GT_EXPR>:
2463 Return 0 if op1 <= TYPE_MIN_VALUE () resp. op1 >= TYPE_MAX_VALUE ().
2464 (simplify_cond_and_lookup_avail_expr): Add assert for dummy == 0
2465 and handle extract_range_from_cond returning false.
2466 * fold-const.c (fold): Optimize comparisons with min/max even for
2467 width > HOST_BITS_PER_WIDE_INT.
2469 2005-01-15 Ralf Corsepius <ralf.corsepius@rtems.org>
2471 * config/mips/rtems.h (MIPS_DEFAULT_GVALUE): Set to 0.
2472 * config/mips/t-rtems (MULTILIBS_DIRNAMES,MULTILIB_OPTIONS):
2473 Remove little endian multilib variants.
2474 Add mips32 multilib variant.
2476 2005-01-15 Kazu Hirata <kazu@cs.umass.edu>
2478 * explow.c (copy_all_reg, stabilize): Remove.
2479 * expr.h: Remove the corresponding prototypes.
2481 * rtlanal.c (regs_set_between_p): Remove.
2482 * rtl.h: Remove the corresponding prototype.
2484 2005-01-14 David Edelsohn <edelsohn@gnu.org>
2486 * config/rs6000/rs6000.c (rs6k_nonimmediate_operand): Rename
2487 to rs6000_nonimmediate_operand.
2488 * config/rs6000/rs6000.h (PREDICATE_CODES): Rename
2489 rs6k_nonimmediate_operand to rs6000_nonimmediate_operand.
2490 * config/rs6000/rs6000.md (movsi_internal1): Rename predicate to
2491 rs6000_nonimmediate_operand.
2492 * config/rs6000/spe.md (movdf_e500_double): Rename predicate to
2493 rs6000_nonimmediate_operand.
2495 2005-01-14 Andrew Pinski <pinskia@physics.uc.edu>
2497 * tree-ssa-alias.c (pass_may_alias): Add TODO_verify_stmts.
2499 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
2501 * config/rs6000/rs6000.h (CLASS_MAX_NREGS): DF goes in 1 register
2503 (CANNOT_CHANGE_MODE_CLASS): Restrict DI mode changes on e500v2.
2504 (PREDICATE_CODES): Add rs6k_nonimmediate_operand.
2506 * config/rs6000/rs6000.c (invalid_e500_subreg): New.
2507 (rs6k_nonimmediate_operand): New.
2508 (rs6000_legitimate_offset_address_p): Handle DI modes on e500v2
2510 (legitimate_lo_sum_address_p): Same.
2511 (rs6000_legitimize_address): Same.
2512 (rs6000_legitimize_reload_address): Same.
2513 (rs6000_legitimate_address): Same.
2514 (spe_build_register_parallel): Pass DF and DC modes in a DI
2517 * config/rs6000/rs6000.md ("*movsi_internal1"): Change predicate
2518 to rs6k_nonimmediate_operand.
2520 * config/rs6000/spe.md ("*frob_df_di"): New.
2521 ("*frob_di_df"): New.
2522 ("*frob_di_df_2"): New.
2523 ("*mov_sidf_e500_subreg0"): New.
2524 ("*mov_sidf_e500_subreg4"): New.
2525 ("*movdf_e500_double"): Change predicate to
2526 rs6k_nonimmediate_operand.
2528 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
2530 * postreload.c (move2add_note_store): Only call
2531 trunc_int_for_mode on scalar integers.
2533 2005-01-14 Steven Bosscher <stevenb@suse.de>
2535 * tree-ssa-dce.c (visited_control_parents): New sbitmap to
2536 replace BB_VISITED uses.
2537 (find_obviously_necessary_stmts): Don't clear BB_VISITED.
2538 (propagate_necessity): Check the bitmap instead of BB_VISITED.
2539 (tree_dce_done): Free visited_control_parents.
2540 (perform_tree_ssa_dce): Allocate and clear it.
2541 * tree-ssa-pre.c (compute_antic_aux): Make non-recursive.
2542 (compute_antic): Iterate from here using a DFS. Use an sbitmap
2543 instead of BB_VISITED.
2545 2005-01-14 Kazu Hirata <kazu@cs.umass.edu>
2547 * c-tree.h, coverage.h, langhooks-def.h, optabs.h, output.h,
2548 rtl.h, targhooks.h, tree-flow.h, tree-mudflap.h, tree.h:
2549 Remove unused prototypes.
2551 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
2553 * config/rs6000/rs6000.c (rs6000_stack_info): Fix caching of SPE
2554 64-bit register usage.
2555 (spe_func_has_64bit_regs_p): Add FIXME note.
2557 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
2559 * config/rs6000/rs6000.c (rs6000_generate_compare): Replace
2560 flag_finite_math_only with flag_unsafe_math_optimizations.
2562 2005-01-14 J. D. Johnston <jjohnst@us.ibm.com>
2564 * config/s390/tpf-unwind.h (s390_fallback_frame_state): Correct
2566 (__tpf_eh_return): Copy TPF private stack area from the module
2567 boundary stack frame. Add check for when module addresses are
2568 equal, but no stub address is found.
2570 2005-01-14 Richard Earnshaw <rearnsha@arm.com>
2573 * arm.h (struct machine_function): Add call_via field.
2574 (thumb_call_via_label): Declare.
2575 * arm.c (thumb_call_via_label): New variable.
2576 (thumb_call_reg_needed): New variable.
2577 (arm_output_function_epilogue): For Thumb code, output any per-function
2578 call-indirect trampolines.
2579 (thumb_call_via_reg): New function.
2580 (arm_file_end): New function.
2581 (TARGET_ASM_FILE_END): Call arm_file_end.
2582 (aof_file_end): Likewise.
2583 * arm-protos.h (thumb_call_via_reg): Declare.
2584 * arm.md (call_reg_thumb, call_value_reg_thumb): Call
2585 thumb_call_via_reg in normal case.
2587 2005-01-14 Jakub Jelinek <jakub@redhat.com>
2590 PR rtl-optimization/19348
2591 * recog.c (peephole2_optimize): Do global life update if some peephole
2592 decides it doesn't need at least one of its inputs and that change
2593 influences liveness at the start of the basic block.
2595 * basic-block.h (EXECUTE_IF_AND_COMPL_IN_REG_SET): Needs 2 REGSET
2596 arguments instead of 1.
2598 2005-01-14 Eric Botcazou <ebotcazou@libertysurf.fr>
2601 * varasm.c (initializer_constant_valid_p) <ADDR_EXPR>: Return
2602 zero for nested functions needing a static chain or functions
2603 with a non-constant address.
2605 2005-01-13 Roger Sayle <roger@eyesopen.com>
2607 * simplify-rtx.c (simplify_binary_operation) <AND>: Optimize
2608 (and (sign_extend X) C) into (zero_extend (and X C)).
2610 2005-01-13 David O'Brien <obrien@FreeBSD.org>
2612 * config/freebsd-spec.h: Make KSE pthread lib logic the default.
2614 2005-01-13 Richard Henderson <rth@redhat.com>
2619 * config/i386/i386.md (cmpdf, cmpsf, bunordered, bordered, buneq,
2620 bunge, bungt, bunle, bunlt, bltgt): Enable for TARGET_SSE_MATH,
2621 not just TARGET_SSE.
2622 (cmpfp_i_387): Rename from cmpfp_i. Move after sse patterns.
2623 (cmpfp_i_mixed): Rename from cmpfp_i_sse; use for TARGET_MIX_SSE_I387.
2624 (cmpfp_i_sse): Rename from cmpfp_i_sse_only; use for TARGET_SSE_MATH.
2625 (cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387): Similarly.
2626 (fp_jcc_1_mixed, fp_jcc_1_sse, fp_jcc_1_387): Similarly.
2627 (fp_jcc_2_mixed, fp_jcc_2_sse, fp_jcc_2_387): Similarly.
2628 (fp_jcc_3_387, fp_jcc_4_387, fp_jcc_5_387, fp_jcc_6_387,
2629 fp_jcc_7_387, fp_jcc_8_387): Rename from fp_jcc_N.
2630 (movdicc_c_rex64): Rename with '*'.
2631 (movsfcc, movdfcc): Add checks for 387 and sse math to condition.
2632 (movsfcc_1_sse_min, movsfcc_1_sse_max, movsfcc_1_sse): New.
2633 (movsfcc_1_387): Rename from movsfcc_1.
2634 (movdfcc_1_sse_min, movdfcc_1_sse_max, movdfcc_1_sse): New.
2635 (movdfcc_1, movdfcc_1_rex64): Add check for 387.
2636 (sminsf3, smaxsf3, smindf3, smaxdf3): New.
2637 (minsf3, minsf, minsf_nonieee, minsf_sse, mindf3, mindf,
2638 mindf_nonieee, mindf_sse, maxsf3, maxsf, maxsf_nonieee, maxsf_sse,
2639 maxdf3, maxdf, maxdf_nonieee, maxdf_sse, sse_movsfcc, sse_movsfcc_eq,
2640 sse_movdfcc, sse_movdfcc_eq, sse_movsfcc_const0_1,
2641 sse_movsfcc_const0_2, sse_movsfcc_const0_3, sse_movsfcc_const0_4,
2642 sse_movdfcc_const0_1, sse_movdfcc_const0_2, sse_movdfcc_const0_3,
2643 sse_movdfcc_const0_4): Remove.
2644 * config/i386/i386.c (ix86_expand_fp_movcc): For TARGET_SSE_MATH,
2645 recognize min/max early. Update for changed sse cmove patterns.
2646 (ix86_split_sse_movcc): New.
2647 * config/i386/i386-protos.h: Update.
2649 2005-01-13 Steven Bosscher <stevenb@suse.de>
2651 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
2652 (fix_stmt_v_may_defs): Likewise.
2654 2005-01-13 Richard Henderson <rth@redhat.com>
2656 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
2657 for converting the mask.
2659 2005-01-13 David Edelsohn <edelsohn@gnu.org>
2661 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
2664 2005-01-13 Richard Henderson <rth@redhat.com>
2666 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
2667 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
2668 IX86_BUILTIN_MOVDQ2Q): Remove.
2669 (IX86_BUILTIN_VEC_EXT_V4SI): New.
2670 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
2671 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
2673 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
2674 (sse2_pextrw): Fix immediate constraint.
2675 (sse2_loadq, sse2_loadq_rex64): Remove.
2676 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
2677 Use __builtin_ia32_vec_ext_<size>.
2678 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
2680 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
2682 * function.c (assign_parm_setup_block): Look inside original
2683 entry_parm when inspecting PARALLEL.
2685 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
2686 Joel Sherrill <joel@oarcorp.com>
2689 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
2690 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
2691 rtems_gxx_recursive_mutex_init.
2692 (__gthread_recursive_mutex_lock): New function.
2693 (__gthread_recursive_mutex_trylock): Likewise.
2694 (__gthread_recursive_mutex_unlock): Likewise.
2696 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
2698 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
2700 2005-01-13 Jan Beulich <jbeulich@novell.com>
2702 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
2703 (ashrti3_internal): Indicate output is early clobber. Generate result
2704 into output rather than first input. Use move for low word of output
2705 if shift count is exactly 64.
2706 (lshrti3_internal): Likewise.
2708 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
2712 * reload1.c (delete_output_reload): Don't delete an output reload
2713 if the pseudo lives longer than a single basic block. Adjust and
2714 improve wording of comment.
2716 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
2719 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
2720 of TYPE_SIZE where we mean the number of bytes.
2722 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
2724 * function.c (assign_parm_setup_block): Relax condition on
2725 multi-register optimization.
2727 2005-01-12 Nick Clifton <nickc@redhat.com>
2729 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
2730 an "in_delay_slot" attribute of "no" to prevent them being used in
2731 delay slots. This is forbidden because they might generate
2734 2005-01-12 Alan Modra <amodra@bigpond.net.au>
2737 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
2740 2005-01-12 Nick Clifton <nickc@redhat.com>
2742 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
2743 passed on to GAS) but do not define, as GAS no longer supports or
2744 needs the -m2000 option.
2746 2005-01-11 Roger Sayle <roger@eyesopen.com>
2748 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
2749 type as args of unprototyped builtins aren't automatically promoted.
2750 (fold_builtin_abs): Likewise.
2752 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2754 * Makefile.in: Set a `build-warn' variable.
2756 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
2758 PR tree-optimization/17949
2759 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
2760 (find_interesting_uses_address): Use it.
2762 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
2764 * regrename.c (kill_value): Handle subreg's that won't simplify.
2766 2005-01-11 Richard Henderson <rth@redhat.com>
2769 * config/i386/i386.h (enum ix86_builtins): Move ...
2770 * config/i386/i386.c: ... here.
2771 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
2772 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
2773 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
2774 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
2775 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
2776 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
2777 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
2778 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
2779 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
2780 IX86_BUILTIN_LOADDDUP): Remove.
2781 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
2782 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
2783 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
2784 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
2785 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
2786 (ix86_init_builtins): Make static.
2787 (ix86_init_mmx_sse_builtins): Update for changed builtins.
2788 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
2789 if all the modes match. Otherwise, fake it.
2790 (get_element_number, ix86_expand_vec_init_builtin,
2791 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
2792 (ix86_expand_builtin): Make static. Update for changed builtins.
2793 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
2794 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
2795 (ix86_expand_vector_init_duplicate): New.
2796 (ix86_expand_vector_init_low_nonzero): New.
2797 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
2798 Split out from ix86_expand_vector_init; handle integer modes.
2799 (ix86_expand_vector_init): Use them.
2800 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
2801 * config/i386/i386-protos.h: Update.
2802 * config/i386/predicates.md (reg_or_0_operand): New.
2803 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
2804 (movv2sf_internal): Likewise. And a splitter to match them all.
2805 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
2806 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
2807 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
2808 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
2809 (mmx_pinsrw): Fix operand ordering.
2810 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
2811 rather than sse_loadss expander.
2812 (movv2df splitter): Similarly.
2813 (sse_loadss, sse_loadlss): Remove.
2814 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
2815 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
2816 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
2817 (sse3_movddup): Rename with '*'.
2818 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
2819 (sse2_loadsd): Remove.
2820 (vec_dupv2df_sse3): Rename from sse3_loadddup.
2821 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
2822 (sse2_pinsrw): Fix argument ordering.
2823 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
2824 (sse2_stored): Remove 'r' destination.
2825 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
2826 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
2827 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
2828 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
2829 vec_extractv16qi, vec_initv16qi): New.
2831 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
2832 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
2833 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
2835 (_mm_load_pd, _mm_store_pd): Use plain dereference.
2836 (_mm_load_si128, _mm_store_si128): Likewise.
2837 (_mm_load1_pd): Use _mm_set1_pd.
2838 (_mm_load_sd): Use _mm_set_sd.
2839 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
2840 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
2841 (_mm_set_epi64): Use _mm_set_epi64x.
2842 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
2843 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
2844 _mm_setr_epi8): Use _mm_set_foo form.
2845 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
2846 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
2847 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
2848 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
2849 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
2850 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
2851 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
2852 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
2853 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
2854 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
2855 (_mm_movedup_pd): Use _mm_shuffle_pd.
2856 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
2857 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
2858 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
2859 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
2861 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
2862 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
2863 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
2864 (_mm_store_ps): Use raw dereference.
2865 (_mm_store1_ps): Use _mm_storeu_ps.
2866 (_mm_storer_ps): Use _mm_store_ps.
2867 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
2868 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
2870 2005-01-11 Stan Shebs <shebs@apple.com>
2872 * config/rs6000/rs6000.c (machopic_output_stub): Issue
2873 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
2875 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
2878 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
2879 Skip all DECLs except for FIELD_DECLs.
2881 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
2883 * config/s390/s390.c (override_options): Return error if
2884 -mbackchain, -mpacked-stack and -mhard-float are used together.
2885 (s390_va_start): Remove the backchain && packed-stack special case.
2886 (s390_gimplify_va_arg): Likewise.
2887 * doc/invoke.texi: Remove the ABI incompatibility note.
2889 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
2891 * config/s390/s390.c (struct s390_frame_layout): Remove
2893 (s390_frame_info, s390_emit_prologue): Replace occurrences of
2894 save_backchain_p with TARGET_BACKCHAIN.
2896 2005-01-11 Alan Modra <amodra@bigpond.net.au>
2899 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
2900 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
2901 * function.c (locate_and_pad_parm): Set new field.
2902 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
2903 Tweak where_pad test to include "none". Always set mem align for
2905 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
2906 not sufficient for type.
2907 (assign_parm_setup_block): If stack_parm is zero on entry, always
2908 make a new stack local. Block move old stack parm if necessary
2909 to new aligned stack local.
2910 (assign_parm_setup_stack): Use a block move to handle
2911 potentially misaligned entry_parm.
2912 (assign_parms_unsplit_complex): Specify required alignment when
2913 creating stack local.
2914 * calls.c (compute_argument_addresses): Override alignment of stack
2915 arg calculated from its type with the alignment given by
2916 FUNCTION_ARG_BOUNDARY.
2917 (store_one_arg): Likewise.
2919 2005-01-11 Jan Beulich <jbeulich@novell.com>
2921 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
2922 Change respective itanium_class attribute to ialu.
2923 (shladdp4_internal): New.
2924 * config/ia64/predicates.md (shladd_log2_operand): New.
2926 2005-01-11 Richard Henderson <rth@redhat.com>
2928 * expr.c (store_constructor): Use rtvec_alloc instead of
2929 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
2931 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
2933 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
2935 2005-01-11 Alan Modra <amodra@bigpond.net.au>
2937 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
2938 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
2939 (TARGET_RWRELOC): Define.
2940 (ia64_rwreloc_section_type_flags): Delete.
2941 (ia64_section_type_flags): New function.
2942 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
2943 (TARGET_RWRELOC): Define.
2945 2005-01-10 David Mosberger <davidm@hpl.hp.com>
2948 * config/ia64/ia64.c (process_set): For alloc insn, only call
2949 process_epilogue is !frame_pointer_needed.
2951 2005-01-10 Roger Sayle <roger@eyesopen.com>
2954 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
2955 unary operator and can't be treated as a binary/comparison operator.
2957 2005-01-10 Richard Henderson <rth@redhat.com>
2959 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
2960 (ix86_return_in_memory): Likewise.
2961 (function_arg_advance): Likewise. Mirror structure in function_arg
2962 for choosing register to advance.
2964 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
2966 * tree-vectorizer.c, tree.def: Fix comment typos.
2968 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
2970 * config/s390/s390.c (struct processor_costs): 4 new fields:
2971 ddbr, ddr, debr, der.
2972 (s390_rtx_costs): More precise handling of divide instructions.
2974 2005-01-10 David Edelsohn <edelsohn@gnu.org>
2976 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
2978 2005-01-10 Jan Beulich <jbeulich@novell.com>
2980 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
2981 names resulting from -ffunction-sections/-fdata-sections and linkonce
2983 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
2984 based on the section name. Rearrange the section name comparison logic
2985 slightly so that each section name is compared against at most once.
2987 2005-01-10 Ben Elliston <bje@au.ibm.com>
2989 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
2990 targets for which -fPIC reduces limitations on the GOT size.
2992 2005-01-09 Falk Hueffner <falk@debian.org>
2994 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
2997 2005-01-09 David Edelsohn <edelsohn@gnu.org>
3000 * collect2.c (main): Set aixrtl_flag for -brtl option.
3001 (resolve_lib_name): Search for .so file extension before .a
3004 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
3006 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
3007 (vect_analyze_data_refs_alignment): Add dump prints.
3009 2005-01-09 Ira Rosen <irar@il.ibm.com>
3011 * tree-vectorizer.c (vect_analyze_offset_expr): Use
3012 expr_invariant_in_loop_p.
3013 Initialize outputs first thing in the function.
3014 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
3015 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
3016 (vect_analyze_pointer_ref_access): Check that the initial condition of
3017 the access function is loop invariant.
3019 2005-01-09 Richard Henderson <rth@redhat.com>
3021 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
3022 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
3023 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
3024 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
3025 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
3026 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
3027 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
3028 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
3029 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
3030 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
3031 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
3032 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
3033 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
3034 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
3035 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
3036 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
3037 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
3038 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
3039 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
3040 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
3041 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
3042 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
3043 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
3045 (mmx_clrdi, pavgusb): Remove.
3046 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
3047 with leading sse_ prefix.
3048 * config/i386/sse.md: Receive them.
3049 * config/i386/mmx.md: New file.
3050 (MMXMODE12, MMXMODE24, mmxvecsize): New.
3051 (subrv2sf3): Turn into expander for normal subtraction.
3052 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
3053 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
3054 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
3055 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
3056 commutative; use ix86_binary_operator_ok.
3057 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
3058 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
3059 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
3060 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
3061 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
3062 Macroize from existing patterns; use ix86_binary_operator_ok.
3063 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
3064 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
3065 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
3066 with vec_select+vec_concat.
3067 (mmx_pshufw, mmx_pshufw_1): Likewise.
3068 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
3069 (mmx_uavgv4hi3): Model correctly.
3070 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
3071 _mm_xor_si64): Remove casts.
3073 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
3075 PR tree-optimization/19224
3076 * tree-scalar-evolution.c (get_instantiated_value,
3077 set_instantiated_value): New functions.
3078 (instantiate_parameters_1): Cache the results.
3079 (instantiate_parameters, resolve_mixers): Initialize and free
3082 2005-01-08 David Edelsohn <edelsohn@gnu.org>
3084 * config/i386/i386.md (addhi_4): Correct reference in comment.
3087 2005-01-08 Richard Henderson <rth@redhat.com>
3089 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
3090 (_mm_cvtsi128_si64x): Likewise.
3091 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
3092 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
3093 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
3094 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
3095 * config/i386/i386-protos.h: Update.
3096 * config/i386/i386.c (print_operand): Add 'H'.
3097 (ix86_fixup_binary_operands): Split out from ...
3098 (ix86_expand_binary_operator): ... here.
3099 (ix86_fixup_binary_operands_no_copy): New.
3100 (ix86_expand_fp_absneg_operator): Handle vector mode results.
3101 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
3102 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
3103 (safe_vector_operand): Use CONST0_RTX.
3104 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
3105 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
3106 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
3107 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
3109 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
3110 (IX86_BUILTIN_CMPNGESS): New.
3111 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
3112 (attr type): Add sselog1.
3113 (attr unit, attr memory): Handle it.
3114 (movti, movti_internal, movti_rex64): Move near other integer moves.
3115 (movtf, movtf_internal): Move near other fp moves.
3116 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
3117 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
3118 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
3119 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
3120 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
3121 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
3122 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
3123 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
3124 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
3125 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
3126 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
3127 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
3128 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
3129 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
3130 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
3131 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
3132 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
3133 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
3134 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
3135 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
3136 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
3137 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
3138 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
3139 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
3140 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
3141 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
3142 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
3143 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
3144 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
3145 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
3146 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
3147 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
3148 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
3149 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
3150 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
3151 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
3152 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
3153 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
3154 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
3155 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
3156 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
3157 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
3158 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
3159 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
3161 (SSEPUSH, push<SSEPUSH>): Remove.
3162 (MMXPUSH, push<MMXPUSH>): Remove.
3163 (sse_movaps, sse_movaps_1, sse_movups): Remove.
3164 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
3165 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
3166 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
3167 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
3168 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
3169 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
3170 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
3171 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
3172 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
3173 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
3174 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
3175 (ppro_sse_log_V4SF): Handle sselog1.
3176 * config/i386/predicates.md (const_0_to_1_operand): New.
3177 (const_0_to_255_mul_8_operand): New.
3178 (const_1_to_31_operand): Rename from const_int_1_31_operand.
3179 (const_2_to_3_operand, const_4_to_7_operand): New.
3180 * config/i386/sse.md: New file.
3181 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
3182 (sse_movups): Rename from sse_movups_1.
3183 (sse_loadlss): Rename from sse_loadss_1.
3184 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
3186 (negv4sf2): Use ix86_expand_fp_absneg_operator.
3187 (absv4sf2, negv2df, absv2df): New.
3188 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
3189 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
3190 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
3191 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
3192 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
3193 (sse3_addsubv4sf3): Model correctly.
3194 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
3195 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
3196 (sse_movhlps): Model with vec_select+vec_concat.
3197 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
3198 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
3199 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
3200 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
3201 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
3202 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
3203 sse2_pshufhw_1): Likewise.
3204 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
3205 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
3206 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
3207 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
3208 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
3209 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
3210 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
3211 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
3212 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
3213 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
3214 sminv8hi3): Mark commutative
3215 operands. Use ix86_binary_operator_ok.
3216 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
3217 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
3218 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
3219 sse2_punpcklqdq): Allow operand2 in memory.
3220 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
3221 sse2_movsd): Add memory alternatives.
3222 (sse_storelps): Turn expander into an insn; split after reload.
3223 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
3224 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
3226 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
3228 * configure.ac (DWARF-2 debug_line): Use objdump.
3229 * configure: Regenerate.
3231 2005-01-08 Jeff Law <law@redhat.com>
3232 Diego Novillo <dnovillo@redhat.com>
3234 PR tree-optimization/18241
3235 * tree-nrv.c (tree_nrv): Ignore volatile return values.
3236 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
3237 statements with volatile operands.
3238 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
3239 operands after marking a statement with has_volatile_ops.
3241 2005-01-08 Roger Sayle <roger@eyesopen.com>
3243 * tree.c (int_fits_type_p): Always honor integer constant
3244 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
3246 2005-01-08 Roger Sayle <roger@eyesopen.com>
3248 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
3249 the latent bug in rtl_delete_block has been resolved.
3251 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
3253 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
3254 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
3256 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
3258 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
3259 library as @multilib_dir@/$(SHLIB_SONAME).
3261 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
3263 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
3264 * configure: Regenerate.
3266 2005-01-07 Jakub Jelinek <jakub@redhat.com>
3268 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
3271 2005-01-07 David Edelsohn <edelsohn@gnu.org>
3274 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
3275 Convert non-word aligned offset address using ld/std into
3278 2005-01-07 Richard Henderson <rth@redhat.com>
3280 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
3282 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
3284 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
3286 2005-01-07 Jakub Jelinek <jakub@redhat.com>
3288 PR tree-optimization/19283
3289 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
3290 is not shorter than the original type.
3292 PR rtl-optimization/19012
3293 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
3295 PR rtl-optimization/18861
3296 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
3297 (FULL_STATE): Define.
3298 (mark_tablejump_edge): New function.
3299 (purge_dead_tablejump_edges): New function.
3300 (find_bb_boundaries): Use it.
3302 PR tree-optimization/18828
3303 * builtins.c (expand_builtin_next_arg): Remove argument and all
3304 the argument checking.
3305 (expand_builtin): Adjust caller.
3306 (expand_builtin_va_start): Likewise. Remove error for too many
3308 (fold_builtin_next_arg): Issue error for too many arguments.
3309 After checking arguments, replace them with magic arguments that
3310 prevent further checking of the args.
3312 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
3314 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
3316 2005-01-06 Roger Sayle <roger@eyesopen.com>
3319 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
3320 and %{.S:...} (and their negative variants) to test whether the
3321 input file is assembler or pre-processed-assembler independent of
3322 the actual filename extension.
3324 2005-01-06 Roger Sayle <roger@eyesopen.com>
3326 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
3327 of sign or zero extended values.
3329 2005-01-06 Geoffrey Keating <geoffk@apple.com>
3331 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
3332 for IBM long double format correctly.
3334 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
3336 Fix PR tree-optimization/18792
3338 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
3339 to first_loop_depth, and use loop depth instead of loop number.
3340 (build_classic_dir_vector): Ditto.
3341 (compute_data_dependences_for_loop): Use depth, not loop number.
3342 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
3343 number. Pass in loops, instead of loop numbers.
3344 (gather_interchange_stats): Ditto.
3345 (linear_transform_loops): Ditto.
3347 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
3350 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
3351 biv_fits_mode_p, extension_within_bounds_p): New functions.
3352 (check_ext_dependent_givs): Use them.
3354 2005-01-06 Roger Sayle <roger@eyesopen.com>
3356 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
3357 more than one barrier, in which case we should delete them all.
3359 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
3361 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
3363 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
3365 Fix PR middle-end/19286
3367 * dwarf2out.c (gen_subprogram_die): If we've already tried to
3368 output this subprogram, simply ignore this attempt to do it again.
3369 (add_abstract_origin_attribute): Don't abort trying to add the abstract
3370 origin attribute if it's not possible.
3371 (gen_block_die): Don't ignore subblocks of "unused" blocks.
3372 (decls_for_scope): Ditto.
3373 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
3374 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
3376 2005-01-05 Richard Henderson <rth@redhat.com>
3379 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
3380 (bdesc_2arg): Use it.
3381 (ix86_expand_binop_builtin): Force operands into registers
3383 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
3384 ix86_expand_sse_compare, ix86_expand_sse_comi,
3385 ix86_expand_builtin): Likewise.
3387 2005-01-05 Richard Henderson <rth@redhat.com>
3389 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
3391 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
3392 (mulv4hi3): Set itanium_class mmmul.
3393 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
3394 (fpack): Rename from fpack_sfsf.
3396 2005-01-05 Richard Henderson <rth@redhat.com>
3399 * reload1.c (do_input_reload): Restrict the optimization deleteing
3400 a previous output reload to RELOAD_FOR_INPUT.
3402 2005-01-05 Steven Bosscher <stevenb@suse.de>
3404 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
3405 comment that we fall through after case.
3406 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
3407 (distribute_links): Likewise.
3408 * cse.c (cse_insn): Likewise.
3409 * cselib.c (cselib_invalidate_mem): Likewise.
3410 * df.c: Update comments at the top of the file.
3411 (read_modify_subreg_p): Update comments here too.
3412 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
3413 * flow.c (mark_set_1): Likewise.
3414 (mark_used_regs): Likewise.
3415 * gcse.c (mems_conflict_for_gcse_p): Likewise.
3416 (canon_list_insert): Likewise.
3417 (mark_set): Likewise.
3418 (try_replace_reg): Likewise.
3419 (store_killed_in_insn): Likewise.
3420 * loop.c (count_one_set): Likewise.
3421 (basic_induction_var): Likewise.
3422 * postreload-gcse.c (find_mem_conflicts): Likewise.
3423 * postreload.c (reload_combine_note_store): Likewise.
3424 (move2add_note_store): Likewise.
3425 * reload.c (find_equiv_reg): Likewise.
3426 (mark_referenced_resources): Likewise.
3427 * rtlanal.c (set_noop_p): Likewise.
3428 (note_stores): Likewise.
3429 (note_uses): Likewise.
3430 * sched-deps.c (sched_analyze_1): Likewise.
3431 * sched-rgn.c (check_live_1): Likewise.
3432 (update_live_1): Likewise.
3433 * config/i860/i860.c: Likewise.
3435 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
3437 (ZERO_EXTRACT): Mention that this one can be an lvalue.
3439 * doc/rtl.texi: Update documentation for bit-fields and SET.
3441 2005-01-05 Roger Sayle <roger@eyesopen.com>
3443 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
3445 2005-01-05 Jan Hubicka <jh@suse.cz>
3446 Richard Henderson <rth@redhat.com>
3449 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
3452 2005-01-05 Richard Henderson <rth@redhat.com>
3455 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
3456 (sse_shufps): Change operand 3 to const_int_operand.
3457 (sse2_storelps): Fix typo in template.
3458 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
3459 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
3460 (ix86_expand_builtin): Likewise.
3462 2005-01-05 Julian Brown <julian@codesourcery.com>
3464 * config/arm/arm.c (arm_return_in_memory): Treat complex types
3465 as aggregates for AAPCS ABIs.
3467 2005-01-05 Stan Shebs <shebs@apple.com>
3469 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
3470 Mach-O getters if ppc64.
3471 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
3473 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
3474 signed four-byte field for global code case.
3475 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
3477 2005-01-05 Roger Sayle <roger@eyesopen.com>
3480 * c-common.c: Include real.h.
3481 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
3482 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
3483 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
3484 * Makefile.in (c-common.o): Update dependencies.
3486 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
3488 * c-parse.in (asm_string): Add trailing semicolon.
3490 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
3492 * c-parse.in (asm_string): New. Don't allow wide strings in
3494 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
3495 asm_string instead of STRING.
3497 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
3499 * c-typeck.c (constructor_no_implicit): Remove.
3500 (set_designator, process_init_element): Don't check
3501 constructor_no_implicit.
3503 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
3504 Kaz Kojima <kkojima@gcc.gnu.org>
3507 * lcm.c (create_pre_exit): New.
3508 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
3509 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
3511 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
3513 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
3514 -msoft-float in target_float_switch.
3515 * arm.c (arm_override_options): Fix processing of target_float_switch.
3517 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
3519 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
3522 2004-01-05 Uros Bizjak <uros@kss-loka.si>
3524 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
3525 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
3527 2005-01-04 Roger Sayle <roger@eyesopen.com>
3529 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
3530 if deleting the then-block allows the test-block to fallthru to the
3533 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
3536 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
3537 after the full declaration if the are in two different TUs.
3539 2005-01-04 Richard Henderson <rth@redhat.com>
3542 * tree-sra.c (generate_one_element_init): Just
3543 call gimplify_and_add.
3544 (generate_element_init): Record the
3545 new referenced variables and mark them for renaming
3546 and split out to ...
3547 (generate_element_init_1): This.
3548 (scalarize_init): Don't call push_gimplify_context/
3549 pop_gimplify_context.
3551 2005-01-04 Geoffrey Keating <geoffk@apple.com>
3553 * toplev.c (get_src_pwd): Handle failure of getpwd().
3555 2005-01-04 Roger Sayle <roger@eyesopen.com>
3557 * fold-const.c (fold_single_bit_test): Delete unreachable handling
3559 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
3560 all the cases handled by it are inverted by invert_truthvalue.
3562 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
3564 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
3565 (save_register_windowsi): Likewise.
3567 2005-01-04 Richard Henderson <rth@redhat.com>
3569 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
3571 2005-01-04 Uros Bizjak <uros@kss-loka.si>
3574 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
3575 any simplifications of vector mode comparison operators.
3576 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
3578 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
3579 Devang Patel <dpatel@apple.com>
3581 PR tree-optimization/18308
3582 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
3583 the operands before creating a new expression.
3584 * dojump.c (do_jump): Make drop_through_label available
3585 for all cases. Add expansion of COND_EXPR.
3587 2005-01-04 Ira Rosen <irar@il.ibm.com>
3589 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
3590 INTEGER_CST instead of TREE_CONSTANT.
3591 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
3592 instead of TREE_CONSTANT.
3593 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
3594 instead of TREE_CONSTANT.
3596 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
3598 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
3600 2005-01-04 Uros Bizjak <uros@kss-loka.si>
3603 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
3604 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
3606 2005-01-03 Richard Henderson <rth@redhat.com>
3608 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
3609 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
3610 fold_convert_const_real_from_real): Split out from ...
3611 (fold_convert_const): ... here.
3613 2005-01-03 Richard Henderson <rth@redhat.com>
3616 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
3617 (mov<MMXMODEI>_internal): Likewise.
3618 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
3619 Add V2SF case; use it for SSE1; don't use TI.
3620 (movdf_integer): Likewise.
3621 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
3623 2005-01-03 Ira Rosen <irar@il.ibm.com>
3625 * tree-vectorizer.c (vect_strip_conversions): New function.
3626 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
3627 check for binary class.
3629 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
3633 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
3635 * gimple-low.c (mark_blocks_with_used_vars): New function.
3636 (mark_blocks_with_used_subblocks): Ditto.
3637 (mark_used_blocks): Ditto.
3638 (pass_mark_used_blocks): New pass.
3639 * tree-inline.c: Include debug.h.
3640 (expand_call_inline): Call outlining_inline_function here.
3641 * tree-optimize.c (init_tree_optimization_passes): Add
3642 pass_mark_used_blocks.
3643 * tree-pass.h (pass_mark_used_blocks): New.
3644 * Makefile.in (tree-inline.o): Add debug.h dependency.
3646 2005-01-03 Geoffrey Keating <geoffk@apple.com>
3648 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
3649 VAR_DECLs to have weak_import attribute.
3651 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
3654 2004-01-03 Steven Bosscher <stevenb@suse.de>
3656 * passes.c (rest_of_compilation): Don't run regmove if only
3657 flag_expensive_optimizations, require flag_regmove instead.
3659 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
3661 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
3663 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
3664 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
3665 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
3666 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
3668 2005-01-03 Richard Henderson <rth@redhat.com>
3670 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
3671 (ia64_const_ok_for_letter_p): New.
3672 (ia64_const_double_ok_for_letter_p): New.
3673 (ia64_extra_constraint): New.
3674 (ia64_expand_vecint_compare): New.
3675 (ia64_expand_vcondu_v2si): New.
3676 (ia64_expand_vecint_cmov): New.
3677 (ia64_expand_vecint_minmax): New.
3678 (ia64_print_operand): Add 'v'.
3679 (ia64_preferred_reload_class): New.
3680 (ia64_vector_mode_supported_p): New.
3681 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
3682 (PREFERRED_RELOAD_CLASS): Move to function.
3683 (CONST_OK_FOR_LETTER_P): Move to function.
3684 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
3685 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
3686 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
3687 (EXTRA_CONSTRAINT): Move to function.
3688 * config/ia64/ia64.md: Include vect.md.
3689 (itanium_class): Add mmalua.
3691 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
3693 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
3694 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
3696 (const_int_2bit_operand): New.
3697 (fr_reg_or_0_operand): New.
3698 * config/ia64/ia64-modes.def: Add vector modes.
3699 * config/ia64/ia64-protos.h: Update.
3700 * config/ia64/vect.md: New file.
3702 2005-01-03 Richard Henderson <rth@redhat.com>
3704 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
3706 2005-01-03 Uros Bizjak <uros@kss-loka.si>
3709 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
3711 (log1pdf2): Change mode of operands[0,1] to DFmode.
3713 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
3715 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
3716 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
3717 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
3718 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
3719 SPARC_RELAXED_ORDERING.
3721 2005-01-03 Richard Henderson <rth@redhat.com>
3722 Uros Bizjak <uros@kss-loka.si>
3725 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
3726 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
3727 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
3729 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
3730 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
3731 const_pow2_1_to_128_operand): New predicates.
3733 2005-01-02 Greg McGary <greg@mcgary.org>
3735 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
3737 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
3738 object. [ptr_type]: Remove unused variable.
3739 Remember nearest addressable array-element or record-component
3740 when checking bitfield components. Tolerate empty BIND_EXPR.
3742 2005-01-01 Richard Henderson <rth@redhat.com>
3744 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
3745 that don't narrow the value. Fail for other conversions.
3747 2005-01-01 Richard Henderson <rth@redhat.com>
3750 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
3751 * c-lang.c (finish_file): Don't do it here.
3752 * objc/objc-act.c (objc_finish_file): Likewise.
3754 * cgraph.c (decl_assembler_name_equal): New.
3755 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
3756 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
3757 * cgraph.h (struct cgraph_varpool_node): Add next.
3758 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
3759 * varasm.c (assemble_alias): Mark the target as needed.
3761 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
3764 * function.c (get_arg_pointer_save_area): Use entry_of_function
3765 instead of get_insns.
3767 2005-01-01 Roger Sayle <roger@eyesopen.com>
3768 Andrew Pinski <pinskia@physics.uc.edu>
3769 James E. Wilson <wilson@specifixinc.com>
3771 PR rtl-optimization/12092
3772 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
3775 2005-01-01 Roger Sayle <roger@eyesopen.com>
3776 Olivier Hainque <hainque@act-europe.fr>
3778 * tree.c (int_fits_type_p): A narrower type always fits in a
3779 wider one, except for negative values into unsigned types.
3781 2005-01-01 Roger Sayle <roger@eyesopen.com>
3783 * tree.c (int_fits_type_p): Compare the result of force_fit_type
3784 with the original constant rather than require TREE_OVERFLOW.
3786 2005-01-01 Steven Bosscher <stevenb@suse.de>
3789 * c-decl.c (finish_function): If compiling C99, annotate the
3790 compiler generated return with the current file name and line 0.
3791 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
3792 the source line is greater than 0.
3793 (remove_bb): Likewise.
3795 See ChangeLog.12 for earlier changes.