1 2005-02-09 Richard Guenther <rguenth@gcc.gnu.org>
4 * fold-const.c (try_move_mult_to_index): Remove redundant
5 type argument. Create ADDR_EXPR with correct type.
6 (fold): Update callers of try_move_mult_to_index. Convert
7 result to the appropriate type.
9 2005-02-09 Roger Sayle <roger@eyesopen.com>
12 * config/avr/avr.c (default_rtx_costs): Delete.
13 (avr_operand_rtx_cost): New function.
14 (avr_rtx_costs): Completely rewrite.
16 2005-02-08 Hans-Peter Nilsson <hp@axis.com>
19 * config/cris/cris.c (in_code): New variable.
20 (cris_output_addr_const): Now a static function, a wrapper for
22 (cris_asm_output_symbol_ref): New function, broken out SYMBOL_REF
23 case from old cris_output_addr_const.
24 (cris_asm_output_label_ref): Similar for LABEL_REF.
25 (cris_output_addr_const_extra): Similar for UNSPEC.
26 * config/cris/cris.h (OUTPUT_ADDR_CONST_EXTRA)
27 (ASM_OUTPUT_SYMBOL_REF, ASM_OUTPUT_LABEL_REF): Define.
28 * config/cris/cris-protos.h (cris_output_addr_const): Remove
30 (cris_asm_output_symbol_ref, cris_output_addr_const_extra)
31 (cris_asm_output_label_ref): Declare.
33 2005-02-08 Paolo Bonzini <bonzini@gnu.org>
36 * doc/cppinternals.texi (Conventions, Lexer, Files): Adjust
37 filenames that changed when libcpp was moved to the toplevel.
39 2005-02-07 Roger Sayle <roger@eyesopen.com>
41 * simplify-rtx.c (simplify_relational_operation_1): Avoid creating
42 zero extensions of BImode operands. Call lowpart_subreg instead
43 of gen_lowpart_common and gen_lowpart_SUBREG.
45 2005-02-07 Nathanael Nerode <neroden@gcc.gnu.org>
48 * libada-mk.in: Use cc_for_cross_gnattools, not cc_set_by_configure.
49 * configure.ac: Set cc_for_cross_gnattools.
50 * configure: Regenerate.
52 2005-02-08 Alan Modra <amodra@bigpond.net.au>
55 * predict.c (PROB_VERY_UNLIKELY): Use 1% instead of 10%.
57 2005-02-07 Eric Botcazou <ebotcazou@libertysurf.fr>
60 * dbxout.c (dbxout_symbol) <FUNCTION_DECL>: Skip inline instance
63 2005-02-07 Jonathan Wakely <redi@gcc.gnu.org>
65 * doc/extend.texi: Document deprecated extensions allowing
66 static floating-point members to have initializers and allowing
67 floating-point literals in integral constant expressions.
69 2005-02-07 Kazu Hirata <kazu@cs.umass.edu>
71 * cfgcleanup.c, df.h, diagnostic.c, rtl.c, tree-vectorizer.h,
72 config/darwin7.h, config/mips/mips.c,
73 config/rs6000/altivec.md, config/rs6000/darwin-tramp.asm:
76 * cse.c: Update comments.
78 2005-02-07 Richard Guenther <rguenth@gcc.gnu.org>
81 * builtins.c (fold_builtin_sqrt): Transform
82 sqrt(pow(x,y)) to pow(fabs(x),y*0.5), not
85 2005-02-07 Leehod Baruch <leehod@il.ibm.com>
86 Dorit Naishlos <dorit@il.ibm.com>
88 * doc/invoke.texi (ftree-vectorizer-verbose): New.
89 * opts.c (OPT_ftree_vectorizer_verbose_): New case for switch.
90 * common.opt (ftree-vectorizer-verbose): New Flag for the vectorizer
92 * tree.h (vect_set_verbosity_level): New extern function declaration
94 * tree-vectorizer.h (verbosity_levels): New enum type.
95 * tree-vectorizer.c (vect_debug_stats): Function removed.
96 (vect_debug_details): Likewise.
97 (vect_verbosity_level): Global variable was defined and initialized.
98 (vect_dump): Global variable definition.
99 (vect_print_dump_info): New function.
100 (vect_set_dump_settings): New function.
101 (vect_set_verbosity_level): New function.
102 (vectorize_loops): Add call to vect_set_dump_settings.
104 (slpeel_make_loop_iterate_ntimes): Dump condition was changed.
105 (slpeel_tree_duplicate_loop_to_edge_cfg): Likewise.
106 (slpeel_tree_peel_loop_to_edge): Likewise.
108 (vect_analyze_offset_expr): Call to vect_print_dump_info with
109 appropriate verbosity level instead of call to vect_debug_details
111 (vect_get_base_and_offset):
112 (vect_create_addr_base_for_vector_ref):
113 (get_vectype_for_scalar_type):
114 (vect_create_data_ref_ptr):
116 (vect_get_vec_def_for_operand):
117 (vect_finish_stmt_generation):
118 (vectorizable_assignment):
119 (vectorizable_operation):
120 (vectorizable_store):
122 (vect_transform_stmt):
123 (vect_update_ivs_after_vectorizer):
124 (vect_do_peeling_for_loop_bound):
125 (vect_update_inits_of_drs):
126 (vect_do_peeling_for_alignment):
127 (vect_transform_loop):
128 (vect_is_simple_use):
129 (vect_analyze_operations):
130 (vect_is_simple_iv_evolution):
131 (vect_analyze_scalar_cycles):
132 (vect_analyze_data_ref_dependence):
133 (vect_analyze_data_ref_dependences):
134 (vect_compute_data_ref_alignment):
135 (vect_enhance_data_refs_alignment):
136 (vect_analyze_data_refs_alignment):
137 (vect_analyze_data_ref_access):
138 (vect_analyze_data_ref_accesses):
139 (vect_analyze_pointer_ref_access):
140 (vect_get_memtag_and_dr):
141 (vect_analyze_data_refs):
142 (vect_mark_relevant):
143 (vect_stmt_relevant_p):
144 (vect_mark_stmts_to_be_vectorized):
145 (vect_can_advance_ivs_p):
146 (vect_get_loop_niters):
147 (vect_analyze_loop_form):
149 (vectorize_loops): Likewise.
151 (vect_do_peeling_for_loop_bound): Dump format slightly changed.
152 (vect_update_inits_of_drs):
153 (vect_do_peeling_for_alignment):
154 (vect_transform_loop):
155 (vect_analyze_operations):
156 (vect_analyze_scalar_cycles):
157 (vect_analyze_data_ref_dependences):
158 (vect_analyze_data_refs_alignment):
159 (vect_analyze_data_ref_accesses):
160 (vect_analyze_data_refs):
161 (vect_mark_stmts_to_be_vectorized):
162 (vect_get_loop_niters):
163 (vect_analyze_loop_form):
164 (vect_analyze_loop): Likewise.
165 (vect_mark_stmts_to_be_vectorized): Add call to print_generic_expr.
167 2005-02-07 Richard Sandiford <rsandifo@redhat.com>
171 2005-01-26 Richard Sandiford <rsandifo@redhat.com>
172 * config/mips/iris6.h (ENDFILE_SPEC): Don't link in irix-csr.o if
173 either -ffast-math or -funsafe-math-optimizations is in use.
175 2005-01-24 Richard Sandiford <rsandifo@redhat.com>
176 * config/mips/irix-csr.c: New file.
177 * config/mips/t-iris6 (irix-csr.o): New rule to build it.
178 (EXTRA_MULTILIB_PARTS): Add irix-csr.o.
179 * config/mips/iris6.h (ENDFILE_SPEC): Include it in n32 and n64
182 2005-02-07 Alexandre Oliva <aoliva@redhat.com>
184 * config/mn10300/linux.h (LIB_SPEC, STARTFILE_SPEC): Remove
185 redundant, incorrect definitions.
187 * config/frv/frv.md ("tls_indirect_call"): Turn into libcall-like
189 ("*tls_indirect_call"): New, reference hard regs directly.
190 * config/frv/frv.c (gen_inlined_tls_plt): Adjust.
192 2005-02-07 Ralf Corsepius <ralf.corsepius@rtems.org>
194 * config/arm/rtems-elf.h (SUBTARGET_EXTRA_ASM_SPEC):
195 Let !-mhard-float !-msoft-float imply -mfpu=softfpa.
197 2005-02-06 Roger Sayle <roger@eyesopen.com>
199 PR rtl-optimization/19800
200 * simplify_rtx.c (simplify_relational_operation_1): Explicitly
201 call gen_lowpart_common and gen_lowpart_SUBREG instead of calling
204 2005-02-07 Ralf Corsepius <ralf.corsepius@rtems.org>
206 * config/m68k/t-rtems (MULTILIB_MATCHES): Let m528x match m5200.
208 2005-02-07 Joseph S. Myers <joseph@codesourcery.com>
211 * doc/invoke.texi: Correct documentation of -freorder-functions.
213 2005-02-06 Joseph S. Myers <joseph@codesourcery.com>
216 * doc/invoke.texi: Don't mention gccbug.
218 2005-02-06 Roger Sayle <roger@eyesopen.com>
219 Andrew Pinski <pinskia@physics.uc.edu>
220 Paolo Bonzini <paolo.bonzini@lu.unisi.ch>
222 * simplify-rtx.c (simplify_relational_operation_1): Simplify
223 (ne:SI (zero_extract:SI FOO (const_int 1) BAR) (const_int 0))
224 into just (zero_extract:SI FOO (const_int 1) BAR).
226 2005-02-06 Joseph S. Myers <joseph@codesourcery.com>
228 * doc/cpp.texi: Include gcc-common.texi. Don't define gcctabopt
229 macro locally. Don't give last revision date. Use GCC version
230 number from gcc-common.texi.
231 * doc/cppinternals.texi: Include gcc-common.texi. Don't give last
232 revision date. Use GCC version number from gcc-common.texi.
233 Describe being a library as current state rather than as 3.x
235 * doc/gcc.texi: Remove last update date.
236 * doc/gccint.texi: Likewise. Update copyright dates.
237 * doc/install.texi: Update copyright dates.
239 2005-02-06 Joseph S. Myers <joseph@codesourcery.com>
242 * doc/invoke.texi: Mention for each option included in -Wall that
243 it is included in -Wall.
244 * doc/gcc.texi: Update copyright and last modification date.
246 2005-02-06 Zdenek Dvorak <dvorakz@suse.cz>
248 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Do not add
249 unnecessary cast to original induction variable increments.
251 2005-02-06 Zdenek Dvorak <dvorakz@suse.cz>
253 PR tree-optimization/18219
254 * tree-ssa-loop-ivopts.c (get_computation_at): Produce computations
257 2005-02-06 Richard Sandiford <rsandifo@redhat.com>
259 * expmed.c (store_bit_field): Make the SUBREG code adjust bitnum.
260 Set bitpos and offset later in the function. Do nothing if the
261 target is a register and if the bitfield lies completely outside
263 (extract_bit_field): Make the same SUBREG, bitpos and offset changes
264 here. Return an uninitialised register if the source value is stored
265 in a register and the bitfield lies completely outside that register.
267 2005-02-06 Steven Bosscher <stevenb@suse.de>
269 * df.c (df_insn_refs_record): Use XEXP to get the operand of a USE,
272 2005-02-06 Marcin Dalecki <martin@dalecki.de>
274 * diagnostic.c (default_diagnostic_finalizer): Fix use of attribute.
276 2005-02-06 Geoffrey Keating <geoffk@apple.com>
278 * config.gcc (*-*-darwin*): Separate darwin7 and darwin8 configs.
279 * config/darwin8.h: New.
280 * config/darwin7.h: Update comment.
282 2005-02-03 Leehod Baruch <leehod@il.ibm.com>
283 Dorit Naishlos <dorit@il.ibm.com>
285 * tree-vectorizer.h (LOC): New type.
286 (UNKNOWN_LOC, EXPR_LOC, LOC_FILE, LOC_LINE): New macros.
287 (loop_line_number): New field in struct _loop_vec_info.
288 (LOOP_VINFO_LOC, LOOP_LOC): New macros.
290 * tree-vectorizer.c (input.h): Included.
291 (find_loop_location): New function.
292 (vect_debug_stats): Argument changed from loop to LOC. Computation of
293 loop line number removed.
294 (vect_debug_details): Likewise.
295 (new_loop_vec_info): Initialize new field LOOP_VINFO_LOC.
297 (slpeel_make_loop_iterate_ntimes): Call find_loop_location. Argument in
298 call to vect_debug_details/stats changed from loop to loop_loc.
299 (slpeel_tree_duplicate_loop_to_edge_cfg): Likewise.
300 (slpeel_tree_peel_loop_to_edge): Likewise.
302 (vect_analyze_offset_expr): Argument in call to vect_debug_details/stats
303 changed from NULL to UNKNOWN_LOC.
304 (vect_get_base_and_offset):
305 (vect_create_addr_base_for_vector_ref):
306 (get_vectype_for_scalar_type):
307 (vect_create_data_ref_ptr):
309 (vect_get_vec_def_for_operand):
310 (vect_finish_stmt_generation):
311 (vectorizable_assignment):
312 (vectorizable_operation):
313 (vectorizable_store):
315 (vect_transform_stmt):
316 (vect_update_ivs_after_vectorizer):
317 (vect_do_peeling_for_loop_bound):
318 (vect_do_peeling_for_alignment):
319 (vect_transform_loop):
320 (vect_is_simple_use):
321 (vect_analyze_operations):
322 (vect_is_simple_iv_evolution):
323 (vect_analyze_scalar_cycles):
324 (vect_analyze_data_ref_dependences):
325 (vect_compute_data_ref_alignment):
326 (vect_enhance_data_refs_alignment):
327 (vect_analyze_data_ref_access):
328 (vect_analyze_data_ref_accesses):
329 (vect_analyze_pointer_ref_access):
330 (vect_get_memtag_and_dr):
331 (vect_analyze_data_refs):
332 (vect_mark_relevant):
333 (vect_stmt_relevant_p):
334 (vect_mark_stmts_to_be_vectorized):
335 (vect_can_advance_ivs_p):
336 (vect_get_loop_niters):
338 (vectorize_loops): Likewise.
340 (vectorizable_load): Argument in call to vect_debug_details/stats
341 changed from loop to LOOP_LOC (loop_vinfo).
342 (vect_transform_loop):
343 (vect_analyze_operations):
344 (vect_analyze_scalar_cycles):
345 (vect_analyze_data_ref_dependence):
346 (vect_enhance_data_refs_alignment):
347 (vect_analyze_data_ref_accesses):
348 (vect_analyze_pointer_ref_access):
349 (vect_analyze_data_refs):
350 (vect_analyze_loop): Likewise.
352 (vect_analyze_loop_form): Argument in call to vect_debug_details/stats
353 changed from loop to loop_loc.
355 (vect_enhance_data_refs_alignment): Removed unused variable loop.
357 2005-02-03 Leehod Baruch <leehod@il.ibm.com>
358 Dorit Naishlos <dorit@il.ibm.com>
360 * tree-vectorizer.c (vect_analyze_data_ref_dependence): Function
362 (vect_analyze_data_ref_dependences): Likewise.
364 (vect_is_simple_use): Argument changed from loop structure to
365 loop_vect_info structure.
366 (vect_can_advance_ivs_p): Likewise.
367 (vect_create_index_for_vector_ref): Likewise.
368 (vect_update_ivs_after_vectorizer): Likewise.
369 (new_stmt_vec_info): Likewise.
371 (new_loop_vec_info): Second argument in call to new_stmt_vec_info was
372 changed from loop to loop_vinfo.
373 (vect_create_data_ref_ptr): First argument in call to
374 vect_create_index_for_vector_ref was changed from loop to loop_vinfo.
375 (vectorizable_assignment): Second argument in call to vect_is_simple_use
376 was changed from loop to loop_vinfo.
377 (vectorizable_operation): Likewise.
378 (vectorizable_store): Likewise.
379 (vect_mark_stmts_to_be_vectorized): Likewise.
380 (vect_do_peeling_for_loop_bound): First argument in call to
381 vect_update_ivs_after_vectorizer was changed from loop to loop_vinfo.
382 (vect_analyze_operations): Argument in call to vect_can_advance_ivs_p
383 was changed from loop to loop_vinfo.
384 (vect_analyze_data_ref_dependences): Third argument in call to
385 vect_analyze_data_ref_dependence was changed from loop to loop_vinfo.
387 (vect_create_index_for_vector_ref): Get the loop from loop_vinfo.
388 (vect_create_data_ref_ptr): Likewise.
389 (vect_init_vector): Likewise.
390 (vect_get_vec_def_for_operand): Likewise.
391 (vectorizable_assignment): Likewise.
392 (vectorizable_operation): Likewise.
393 (vectorizable_store): Likewise.
394 (vectorizable_load): Likewise.
395 (vect_update_ivs_after_vectorizer): Likewise.
396 (vect_is_simple_use): Likewise.
397 (vect_analyze_data_ref_dependence): Likewise.
398 (vect_analyze_pointer_ref_access): Likewise.
399 (vect_can_advance_ivs_p): Likewise.
401 * tree-vectorizer.h: (_loop_vec_info): Definition and macros moved
402 before _stmt_vec_info.
403 (_stmt_vec_info): Loop field replaced by loop_vec_info.
404 (STMT_VINFO_LOOP): Replaced with STMT_VINFO_LOOP_VINFO.
405 (new_stmt_vec_info): Argument changed from loop structure to
406 loop_vect_info structure
408 (vect_analyze_data_ref_dependences): Unnecessary line was removed.
409 (vect_analyze_offset_expr): Avoid 80 columns overflow.
410 (vect_create_addr_base_for_vector_ref): Likewise.
411 (vect_analyze_pointer_ref_access): Likewise.
413 2005-02-03 Andrew Pinski <pinskia@physics.uc.edu>
416 * tree-ssa-dse.c (fix_phi_uses): Update the occurs in abnormal
417 phi flag if the phi is abnormal.
419 2005-02-03 Andrew Pinski <pinskia@physics.uc.edu>
422 * tree-ssa.c (replace_immediate_uses): Update the immediate_uses
423 information for the new statement.
425 2005-02-03 Joseph S. Myers <joseph@codesourcery.com>
428 * gcc.c (main): Include the target in -v output.
430 2005-02-03 Alexandre Oliva <aoliva@redhat.com>
432 * combine.c (simplify_comparison, case SIGN_EXTEND, ZERO_EXTEND):
433 Do not drop the extend if we'd have to add a paradoxical subreg
434 later. Include optabs.h and insn-codes.h.
435 * Makefile.in (combine.o): Depend on $(OPTABS_H).
437 2005-02-02 Roger Sayle <roger@eyesopen.com>
440 * real.h (REAL_MODE_FORMAT_COMPOSITE_P): New macro.
441 * fold-const.c (const_binop): Avoid constant folding floating
442 point operations in modes that use composite representations.
443 * simplify-rtx.c (simplify_binary_operation): Likewise.
445 2005-02-02 Geoffrey Keating <geoffk@apple.com>
447 * config/rs6000/altivec.md (altivec_dst): Make the first operand
449 (altivec_dstt): Likewise.
450 (altivec_dstst): Likewise.
451 (altivec_dststt): Likewise.
452 * config/rs6000/rs6000.c (altivec_expand_dst_builtin): Adjust creation
455 2005-02-03 Dorit Naishlos <dorit@il.ibm.com>
457 * tree-vectorizer.c (slpeel_make_loop_iterate_ntimes): Copy
458 EXPR_LOCUS from orig_cond to the new cond_stmt.
459 (vect_finish_stmt_generation): Copy EXPR_LOCUS from stmt to
462 2005-02-03 Dorit Naishlos <dorit@il.ibm.com>
464 * tree-vectorizer.c (slpeel_make_loop_iterate_ntimes): Call
465 standard_iv_increment_position. Remove call to bsi_next
466 (no need to bump the iterator anymore).
467 (vect_create_index_for_vector_ref): Call
468 standard_iv_increment_position. Remove second function argument.
469 (vect_finish_stmt_generation): Remove call to bsi_next
470 (no need to bump the iterator anymore).
471 (vect_create_data_ref_ptr): Remove second argument (bsi) in call
472 to vect_create_index_for_vector_ref.
474 2005-02-03 Dorit Naishlos <dorit@il.ibm.com>
476 * tree-vectorizer.c (vect_mark_stmts_to_be_vectorized): Check if
477 phis are used out of the loop.
479 2005-02-03 Joseph S. Myers <joseph@codesourcery.com>
482 * c-decl.c (undef_nested_function): New variable.
483 (pop_scope): Diagnose undefined nested functions.
484 (finish_function): Don't attempt cgraph processing or genericizing
485 if current top-level function contained an undefined nested
486 function. Reset undef_nested_function at the end of a top-level
489 2005-02-02 Zdenek Dvorak <dvorakz@suse.cz>
491 * tree.c (build_int_cst_type): Take sign of the value into account
492 when deciding whether sign extend the value.
494 2005-02-02 Joseph S. Myers <joseph@codesourcery.com>
497 * gcc.c (cpp_unique_options): Remove %{trigraphs}.
498 (cpp_options, cc1_options): Change %{std*&ansi} to
499 %{std*&ansi&trigraphs}.
501 2005-02-02 Kazu Hirata <kazu@cs.umass.edu>
503 * cse.c: Update comments.
505 2005-02-02 Aldy Hernandez <aldyh@redhat.com>
507 * config/rs6000/rs6000.c (rs6000_spe_function_arg): Multi-reg
508 arguments go on the stack. Do not put complex doubles in odd/even
511 2005-02-02 Steven Bosscher <stevenb@suse.de>
513 * doc/invoke.texi: Update the options enabled at -O1, -O2,
514 and -O3. Update the documentation for -fgcse-sm, -fgcse-las,
515 -ftree-pre, and -fthread-jumps.
517 2005-02-02 Richard Sandiford <rsandifo@redhat.com>
519 PR tree-optimization/19578
520 * tree-flow.h (modified_noreturn_calls): Declare.
521 (noreturn_call_p): Declare.
522 * tree-flow-inline.h (noreturn_call_p): New function.
523 (modify_stmt): Add modified noreturn calls to modified_noreturn_calls.
524 * tree-cfg.c (modified_noreturn_calls): New variable.
525 (cleanup_control_flow): Use noreturn_call_p. Split basic blocks
526 that contain a mid-block noreturn call.
527 * tree-ssa.c (delete_tree_ssa): Clear modified_noreturn_calls.
529 2005-02-02 Kazu Hirata <kazu@cs.umass.edu>
531 * df.c (df_def_record_1, df_uses_record): Don't use
533 * df.h (DF_FOR_REGALLOC): Remove.
535 2005-02-02 Joseph S. Myers <joseph@codesourcery.com>
538 * c-typeck.c (really_start_incremental_init): Reset
539 constructor_max_index for arrays of incomplete type.
541 2005-02-02 Jeff Law <law@redhat.com>
543 * gcse.c (struct reg_set): Store the block index where the register
544 was set rather than the setting insn.
545 (record_one_set, compute_transp): Corresponding changes.
546 (pre_insert_copy_insn): Do not call replace_one_set.
547 (replace_one_set): Remove.
549 2005-02-02 Joseph S. Myers <joseph@codesourcery.com>
552 * doc/gcov.texi: Fix typo.
554 2005-02-01 Richard Henderson <rth@redhat.com
557 * config/i386/i386.h (MODES_TIEABLE_P): Use ix86_modes_tieable_p.
558 * config/i386/i386.c (ix86_hard_regno_mode_ok): Change return
560 (ix86_tieable_integer_mode_p, ix86_modes_tieable_p): New.
561 * config/i386/i386-protos.h: Update.
563 2005-02-01 Steven Bosscher <stevenb@suse.de>
565 PR tree-optimization/19217
566 * tree-cfg.c (verify_expr): Use the data field to see if TP was
567 seen inside a PHI node. Do not do the ADDR_EXPR check if it was.
568 (verify_stmts): Pass (void*)1 as data to verify_expr to signal
569 that it is walking a PHI node.
571 2005-02-01 Joseph S. Myers <joseph@codesourcery.com>
573 * doc/extend.texi (Nested Functions): Update.
575 2005-02-01 Richard Henderson <rth@redhat.com>
578 * optabs.c (expand_copysign_absneg): Export.
579 * optabs.h (expand_copysign_absneg): Declare.
580 * config/rs6000/rs6000.md (copysigntf3): New.
582 2005-02-01 Diego Novillo <dnovillo@redhat.com>
584 PR tree-optimization/19633
585 * tree-flow.h (struct ptr_info_def): Add field 'pt_null'.
586 * tree-ssa-alias.c (init_alias_info): Initialize.
587 (merge_pointed_to_info): Set.
588 (add_pointed_to_expr): Set pt_null if EXPR is a NULL pointer.
589 (dump_points_to_info_for): Show value of pt_null.
590 (struct count_ptr_d): Declare.
591 (find_ptr_dereference): Remove.
592 (ptr_is_dereferenced_by): Remove.
593 (count_ptr_derefs): New local function.
594 (count_uses_and_derefs): New local function.
595 (compute_points_to_and_addr_escape): Call it. If the number
596 of dereferences is greater than zero, mark the pointer as
597 dereferenced. If there are fewer dereferences than uses of
598 the pointer, the pointer's value escapes.
600 2005-02-01 Diego Novillo <dnovillo@redhat.com>
602 PR tree-optimization/19670
603 * tree-ssa.c (verify_ssa_name): Don't set TREE_VISITED
605 (verify_use): ... set it here, instead.
607 2005-02-01 Janis Johnson <janis187@us.ibm.com>
609 * doc/sourcebuild.texi (Test directives): Fix formatting.
611 * doc/sourcebuild.texi (Test directives): Describe selector
614 2005-02-01 Segher Boessenkool <segher@kernel.crashing.org>
616 * config.gcc (powerpc64-*-linux*): Default to -m64 also for
617 specific 64-bit CPUs.
619 2005-02-01 Richard Earnshaw <rearnsha@arm.com>
622 * arm.c (arm_eliminable_register): New function.
623 (adjacent_mem_locations): Don't allow eliminable registers. Use
624 HOST_WIDE_INT for address offsets.
625 * arm-protos.h (arm_eliminable_register): Add prototype.
627 2005-02-01 Steven Bosscher <stevenb@suse.de>
629 PR optimization/15242
630 * params.def (PARAM_MAX_GOTO_DUPLICATION_INSNS): New param.
631 * basic-block.h (duplicate_computed_gotos): Add prototype.
632 * bb-reorder.c (duplicate_computed_gotos): New function to
633 duplicate sufficiently small blocks ending in a computed jump.
634 * passes.c (rest_of_compilation): Call duplicate_computed_gotos
635 if not optimizing for size.
636 * cfgcleanup.c (try_crossjump_bb): If not optimizing for size,
637 never do tail merging for blocks ending in a computed jump.
638 * doc/invoke.texi: Document the max-goto-duplication-insns param.
640 2005-02-01 Eric Botcazou <ebotcazou@libertysurf.fr>
642 Patch from Richard Sandiford <rsandifo@redhat.com>
643 * reload1.c (choose_reload_regs): Prevent the find_equiv_reg() code
644 from inheriting a subreg equivalence with a non-spill register.
646 * tree.h (DECL_FUNCTION_CODE): Document that it is overloaded.
648 2005-02-01 Alexandre Oliva <aoliva@redhat.com>
650 * config/frv/frv.c (movcc_fp_destination_operand): New.
651 (gpr_or_memory_operand): Fix typo in comment.
652 (gpr_or_memory_operand_with_scratch): New.
653 * config/frv/frv.h (PREDICATE_CODES): Add the two new predicates.
654 * config/frv/frv.md (movcc_fp, movcc_fp_internal): Use
655 movcc_fp_destination_operand.
656 (reload_incc_fp): Use gpr_or_memory_operand_with_scratch.
657 Legitimize memory addresses using a scratch register.
659 2005-01-31 Jeff Law <law@redhat.com>
661 * tree-into-ssa.c (mark_def_sites_global_data): Make KILLS
662 bitmap a sparse bitmap instead of a simple bitmap.
663 (mark_def_sites_initialize_block): Corresponding changes.
664 (ssa_mark_def_sites_initialize_block): Likewise.
665 (ssa_mark_phi_uses): Likewise.
666 (mark_def_site, ssa_mark_def_sites): Likewise.
667 (mark_def_site_blocks): Likewise.
668 (rewrite_ssa_into_ssa): Likewise.
670 * tree-ssa-dom.c (record_cond): Pass correct variable type
671 for last argument to htab_find_slot_with_hash.
673 * fold-const.c (fold, case CEIL_MOD_EXPR): Do not lose side
674 effects when optimizing 0 % X. Do not try to optimize X % 0.
676 2005-01-31 James E. Wilson <wilson@specifixinc.com>
678 * config/ia64/itanium1.md (1_scall bypass): Change 2_mmalua to
681 2005-02-01 Eric Christopher <echristo@redhat.com>
683 * config/mips/mips.c (override_options): Warn if -mint64
685 * doc/invoke.texi (MIPS Options): Document that -mint64 is
688 2005-02-01 Kazu Hirata <kazu@cs.umass.edu>
690 * cse.c (cse_reg_info): Remove hash_next, next, regno. Add
692 (cse_reg_info_list, cse_reg_info_list_free, REGHASH_SHIFT,
693 REGHASH_SIZE, REGHASH_MASK, reg_hash, REGHASH_FN,
694 cached_cse_reg_info, GET_CSE_REG_INFO): Remove.
695 (cached_regno): Initialize to INVALID_REGNUM.
696 (cse_reg_info_table_size,
697 cse_reg_info_table_first_uninitialized,
698 cse_reg_info_timestamp): New.
699 (REG_TICK, REG_IN_TABLE, SUBREG_TICKED, REG_QTY): Use
701 (init_cse_reg_info, get_cse_reg_info_1): New.
702 (get_cse_reg_info): Cache the last look-up.
703 (new_basic_block): Update the code to clear mappings from
704 registers to cse_reg_info entries.
705 (cse_main): Call init_cse_reg_info.
707 * cse.c (get_cse_reg_info): Update a comment.
709 2005-01-31 Steven Bosscher <stevenb@suse.de>
712 * c-decl.c (start_decl): Do not warn about arrays of elements with
713 an incomplete type here.
714 (grokdeclarator): Do it here by making a pedwarn an error.
715 * c-typeck.c (push_init_level): If there were previous errors with
716 the constructor type, do not warn about braces for initializers.
717 (process_init_element): Likewise for excess initializer elements.
719 2005-01-31 Kazu Hirata <kazu@cs.umass.edu>
721 * cse.c (delete_trivially_dead_insn): Don't iterate.
723 2005-01-31 Andrew Pinski <pinskia@physics.uc.edu>
725 * config/rs6000/rs6000.md (copysignsf3): New expand.
726 (copysigndf3): Likewise.
728 2005-01-31 Steven Bosscher <stevenb@suse.de>
730 * recog.c (constrain_operands): Only look for earlyclobber operand
731 conflicts if an '&' constraint was seen.
733 2005-01-31 Marc Espie <espie@openbsd.org>
735 * config.gcc: Don't include embedded systems fragment, switches default
736 debugging format to ELF.
737 * config/i386/openbsdelf.h: Add DBX_REGISTER_NUMBER since we no
738 longer pick it up there.
740 2005-01-31 Dale Johannesen <dalej@apple.com>
742 * doc/extend.texi (nested functions): Fix linkage description.
743 Clarify that static is not allowed.
745 2005-01-31 Dale Johannesen <dalej@apple.com>
747 * config/rs6000/darwin.md (movsf_low_di): Make work.
748 (movdf_low_di): Make work.
750 2005-01-31 Dale Johannesen <dalej@apple.com>
752 * config/rs6000/darwin-tramp.asm (__trampoline_setup):
753 Make work for 64 bit.
755 2005-01-31 Roger Sayle <roger@eyesopen.com>
756 Dale Johannesen <dalej@apple.com>
759 * fold-const.c (fold_binary_op_with_conditional_arg):
760 Make types match original operands, before STRIP_NOPS.
762 2005-01-31 Alan Modra <amodra@bigpond.net.au>
764 * config/rs6000/linux-unwind.h (struct gcc_vregs): New.
765 (struct gcc_regs): Rename from gcc_pt_regs. Add more regs.
766 (struct gcc_sigcontext): Delete. Merge contents to..
767 (struct gcc_ucontext): ..here.
768 (get_sigcontext): Delete.
769 (get_regs): New function, like get_sigcontext but return regs ptr.
770 64-bit version finds regs from r1 to support vdso.
771 (ppc_linux_aux_vector): New function.
772 (ppc_fallback_frame_state): Modify for get_regs. Restore fprs
775 2005-01-31 Ira Rosen <irar@il.ibm.com>
777 * tree-vectorizer.c (vect_analyze_offset_expr): Use ssizetype instead
779 (vect_get_base_and_offset): Use ssizetype instead sizetype. Remove
780 redundant fold. Fix misalignment for MINUS_EXPR.
781 (vect_compute_data_ref_alignment): Use ssizetype instead sizetype.
782 (vect_analyze_pointer_ref_access): Likewise.
783 (vect_get_memtag_and_dr): Likewise.
785 2005-01-31 Richard Henderson <rth@redhat.com>
787 * config/i386/i386.c (absnegsf2_mixed): Remove all of the # hackery.
788 (absnegsf2_sse, absnegsf2_i387, absnegdf2_mixed, absnegdf2_sse,
789 absnegdf2_i387, absnegxf2_i387): Likewise.
791 2005-01-31 Richard Henderson <rth@redhat.com>
794 * optabs.c (expand_copysign_absneg): New.
795 (expand_copysign_bit): Split out from ...
796 (expand_copysign): ... here. Use expand_copysign_absneg.
798 2005-01-30 Roger Sayle <roger@eyesopen.com>
801 * config/pa/pa.md (anddi3, iordi3): On HPPA64, disallow an integer
802 constant as the second operand and a register as the third.
804 2005-01-31 Danny Smith <dannysmith@users.sourceforge.net>
807 * config/i386/i386.c (ix86_function_ok_for_sibcall): Also check
808 that dllimport'd functions do not use all call-clobbered registers
811 2005-01-30 Richard Henderson <rth@redhat.com>
814 * optabs.c (expand_absneg_bit): Split out from expand_unop and
815 expand_abs_nojump and generalize. Use operand_subword and
816 emit_no_conflict_block. Support large modes.
817 (expand_abs_nojump): Use it. Use HONOR_SIGNED_ZEROS to fall
819 (expand_unop): Likewise. Use HONOR_SIGNED_ZEROS to fall back
820 negation to subtraction.
821 (expand_copysign): Check that the format has signed zeros.
822 Use operand_subword and emit_no_conflict_block. Support large modes.
824 2005-01-30 Daniel Berlin <dberlin@dberlin.org>
826 Fix PR tree-optimization/19624
828 * Makefile.in (tree-ssa-pre.o): Add CFGLOOP_H.
829 * tree-ssa-pre.c: Add cfgloop.h.
831 (pre_stats): New member, constified.
832 (inserted_exprs): New static variable.
833 (NECESSARY): New macro.
834 (create_expression_by_pieces): Fold the expression, and
835 mark it as defaulting to not necessary. Also put in
837 (fully_constant_expression): New function.
838 (insert_into_preds_of_block): Modify to not insert phis when we
839 are playing with induction variables.
840 Push phis onto the inserted_exprs vector, and mark them as not
841 necessary by default.
842 (insert_aux): Call fully_constant_expression on eprime.
843 If all edges produce the same value, mark it constant.
844 (mark_operand_necessary): New function.
845 (remove_dead_inserted_code): New function.
846 (init_pre): Init loop optimizer to get loop info.
847 (fini_pre): Free loop_optimizer, and inserted_exprs vec.
848 (execute_pre): Commit edge inserts, then remove dead code.
850 2005-01-30 Richard Henderson <rth@redhat.com>
852 * rtl.c (rtx_equal_p): No early exit for CONST_VECTOR.
853 * varasm.c (const_rtx_hash_1): Handle CONST_VECTOR.
855 2005-01-30 Richard Henderson <rth@redhat.com>
858 * config/i386/i386.c (ix86_expand_copysign): New.
859 (ix86_split_copysign_const): New.
860 (ix86_split_copysign_var): Rename from ix86_split_copysign,
861 rearrange op1/nmask operands.
862 * config/i386/i386-protos.h: Update.
863 * config/i386/i386.md (copysignsf3): Use ix86_expand_copysign.
864 (copysigndf3): Likewise.
865 (copysignsf3_const, copysigndf3_const): New.
866 (copysignsf3_var): Rename from copysignsf3, split out splitter
867 and fix split predicate for X constraint.
868 (copysigndf3_var): Similarly.
870 2005-01-30 Kazu Hirata <kazu@cs.umass.edu>
872 * optabs.c, doc/c-tree.texi, doc/install.texi, doc/md.texi,
873 doc/passes.texi, doc/rtl.texi, doc/sourcebuild.texi,
874 doc/tm.texi, doc/tree-ssa.texi: Update copyright.
876 2005-01-29 Richard Henderson <rth@redhat.com>
879 * config/i386/i386.md (movdf_nointeger, movdf_integer): Fix ordering
880 of # and * in constraints.
882 2005-01-29 Richard Henderson <rth@redhat.com>
885 * expr.c (store_field): Don't strip sub-mode cast when the input
886 data is even smaller.
888 2005-01-29 Richard Henderson <rth@redhat.com>
891 * expr.c (categorize_ctor_elements_1): Check for CONSTRUCTOR of a
894 2005-01-29 Richard Henderson <rth@redhat.com>
896 * combine.c (make_field_assignment): Fix argument order
899 2005-01-29 Richard Guenther <richard.guenther@uni-tuebingen.de>
901 PR tree-optimization/15791
902 * fold-const.c (extract_array_ref): New function.
903 (fold): Fold comparisons between &a[i] and &a[j] or
904 semantically equivalent trees.
906 2005-01-29 Jeff Law <law@redhat.com>
908 * gcse.c (insert_expr_in_table): Revamp handling of available
909 and anticipatable occurrence lists to avoid unnecessary list
911 (insert_set_in_table): Similarly.
913 2005-01-29 Joseph S. Myers <joseph@codesourcery.com>
915 * c-common.c (fix_string_type): Just use c_build_qualified_type to
917 (c_build_qualified_type): Build qualified array types with
918 TYPE_MAIN_VARIANT pointing to corresponding unqualified type.
919 * c-decl.c (c_make_fname_decl): Build unqualified array type
920 before qualified type.
921 (grokdeclarator): Use TYPE_MAIN_VARIANT of typedef type if element
922 type is qualified, not just if type itself is. Don't apply
923 qualifiers to array type when declarator is processed. Apply
924 qualifiers to field type whether or not it is an array type.
925 Don't handle array types specially for applying qualifiers to
927 * c-typeck.c (composite_type): Build unqualified element type and
928 array type when forming composite of array types.
929 (common_pointer_type, comptypes, comp_target_types,
930 type_lists_compatible_p, build_indirect_ref, build_array_ref,
931 convert_for_assignment): Don't apply TYPE_MAIN_VARIANT to array
933 (type_lists_compatible_p): Cache TREE_VALUE (args1) and TREE_VALUE
934 (args2) in variables a1 and a2.
936 2005-01-29 Kazu Hirata <kazu@cs.umass.edu>
938 * cse.c (n_elements_made, max_elements_made): Remove.
939 (insert): Don't touch n_elements_made.
940 (cse_main): Don't touch n_elements_made or max_elements_made.
942 2005-01-28 Stephane Carrez <stcarrez@nerim.fr>
945 * config/m68hc11/t-m68hc11-gas (dp-bit.c): Fix typo causing a
946 configuration part of dp-bit.c to be lost.
948 2005-01-28 Roger Sayle <roger@eyesopen.com>
950 * expmed.c (expand_mult_highpart_optab): When attempting to use
951 a non-widening multiplication in a wider mode, the operands need
952 to be converted (zero or sign extended) to that mode.
954 2005-01-28 Ian Lance Taylor <ian@airs.com>
958 * gimple-low.c (block_may_fallthru): TRY_FINALLY_EXPR only falls
959 through if both operands fall through.
961 2005-01-28 Kazu Hirata <kazu@cs.umass.edu>
963 * cse.c (fold_rtx) <PC>: Don't optimize.
965 2005-01-28 Jeff Law <law@redhat.com>
967 * fold-const.c (fold, case CEIL_MOD_EXPR): Optimize 0 % X.
968 (case FLOOR_MOD_EXPR, ROUND_MOD_EXPR, TRUNC_MOD_EXPR): Similarly.
970 2005-01-28 Kazu Hirata <kazu@cs.umass.edu>
972 * cse.c (cse_reg_info_free_list, cse_reg_info_used_list,
973 cse_reg_info_used_list_end): Remove.
974 (cse_reg_info_list, cse_reg_info_list_free): New.
975 (get_cse_reg_info): When allocating an instance of
976 cse_reg_info, add it to the beginning of the cse_reg_info_list
977 list. Remove code to maintain cse_reg_info_used_list.
978 (new_basic_block): Reset the free list to the beginning of
981 2005-01-28 Richard Sandiford <rsandifo@redhat.com>
983 * config/frv/frv.c (bdesc_2arg): Remove __MDPACKH.
984 (frv_init_builtins): Change its prototype to take 4 uhalf arguments.
985 (frv_expand_mdpackh_builtin): New function.
986 (frv_expand_builtin): Use it to expand __MDPACKH.
987 * doc/extend.texi (__MDPACKH): Update documentation.
989 2005-01-28 Uros Bizjak <uros@kss-loka.si>
991 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use
992 SSE_FLOAT_MODE_P for use_sse computation.
994 2005-01-28 Ralf Corsepius <ralf.corsepius@rtems.org>
997 * config/sparc/rtemself.h (LINK_GCC_C_SEQUENCE_SPEC): undefine.
999 2005-01-27 Richard Henderson <rth@redhat.com>
1001 * builtins.c (expand_builtin_copysign): New.
1002 (expand_builtin): Call it.
1003 * genopinit.c (optabs): Add copysign_optab.
1004 * optabs.c (init_optabs): Initialize it.
1005 (expand_copysign): New.
1006 * optabs.h (OTI_copysign, copysign_optab): New.
1007 (expand_copysign): Declare.
1009 * config/alpha/alpha.md (UNSPEC_COPYSIGN): New.
1010 (copysignsf3, ncopysignsf3, copysigndf3, ncopysigndf3): New.
1012 * config/i386/i386.c (ix86_build_signbit_mask): Split from ...
1013 (ix86_expand_fp_absneg_operator): ... here.
1014 (ix86_split_copysign): New.
1015 * config/i386/i386-protos.h: Update.
1016 * config/i386/i386.md (UNSPEC_COPYSIGN): New.
1017 (copysignsf3, copysigndf3): New.
1019 * config/ia64/ia64.md (UNSPEC_COPYSIGN): New.
1020 (copysignsf3, ncopysignsf3): New.
1021 (copysigndf3, ncopysigndf3): New.
1022 (copysignxf3, ncopysignxf3): New.
1023 * config/ia64/ia64.c (rtx_needs_barrier): Handle UNSPEC_COPYSIGN.
1025 2005-01-27 Arend Bayer <arend.bayer@web.de>
1026 Kazu Hirata <kazu@cs.umass.edu>
1028 * cse.c: (find_best_addr): Don't call copy_rtx before calling
1029 fold_rtx. Save cost recomputation if fold_rtx did nothing.
1030 (fold_rtx) <ASM_OPERANDS>: Don't do anything if INSN is
1033 2005-01-27 Jeff Law <law@redhat.com>
1035 * tree-into-ssa.c (ssa_rewrite_initialize_block): Do not register
1036 definitions of SSA_NAMEs which are not being rewritten.
1037 (rewrite_ssa_into_ssa): Only initialize the current definition
1038 of an SSA_NAME if that SSA_NAME has been marked for rewriting.
1039 If checking is enabled, assert that SSA_NAME_AUX is clear for all
1040 SSA_NAMEs before returning.
1042 2005-01-27 Kazu Hirata <kazu@cs.umass.edu>
1044 * c-common.def, c-dump.c, c-gimplify.c, c-objc-common.c,
1045 cfgexpand.c, dbxout.c, function.h, opts.c, tree-flow-inline.h,
1046 tree-ssa-operands.h, tree-ssa-threadupdate.c, tree-tailcall.c,
1047 config/avr/avr.md, config/cris/aout.h, config/cris/cris.h,
1048 config/mips/iris6.h, config/sh/sh.c: Update copyright.
1050 2005-01-27 Steven Bosscher <stevenb@suse.de>
1053 * opts.c (decode_options): Move flag_thread_jumps from -O1 and
1054 higher to -O2 and higher. Likewise for tree PRE.
1055 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Only iterate at -O2
1058 2005-01-27 Ian Lance Taylor <ian@airs.com>
1061 * gimple-low.c (try_catch_may_fallthru): New static function.
1062 (block_may_fallthru): Handle TRY_CATCH_EXPR.
1063 * tree-inline.c (expand_call_inline): Don't warn about reaching
1064 the end of a non-void function being inlined if the function uses
1067 2005-01-27 Jakub Jelinek <jakub@redhat.com>
1069 * config/i386/i386.h (CALL_USED_REGISTERS): Fix comment pastos.
1072 * c-decl.c (warn_if_shadowing): Handle old_decl error_mark_node.
1073 (pushdecl): Only use DECL_FILE_SCOPE_P if DECL_P.
1074 (implicitly_declare): Handle error_mark_node.
1076 2005-01-27 Richard Henderson <rth@redhat.com>
1079 * tree.h (struct tree_decl): Add debug_expr_is_from.
1080 (DECL_DEBUG_EXPR_IS_FROM): New.
1081 (DECL_DEBUG_EXPR): Rename from DECL_DEBUG_ALIAS_OF.
1082 * dwarf2out.c (dwarf2out_var_location): Update to match.
1083 * tree-outof-ssa.c (create_temp): Likewise.
1084 * var-tracking.c (track_expr_p): Likewise.
1085 * tree-sra.c (instantiate_element): Set DECL_DEBUG_EXPR.
1086 * c-objc-common.c (c_tree_printer) <'D'>: Handle DECL_DEBUG_EXPR.
1087 * toplev.c (default_tree_printer): Likewise.
1089 2005-01-27 Alexandre Oliva <aoliva@redhat.com>
1091 * config/frv/frv.c (frv_class_likely_spilled_p): Add GR8_REGS,
1092 GR9_REGS, GR89_REGS, FDPIC_FPTR_REGS, FDPIC_REGS.
1094 2005-01-27 Steven Bosscher <stevenb@suse.de>
1096 * tree.h (SWITCH_COND, SWITCH_BODY, SWITCH_LABELS, CASE_LOW,
1097 CASE_HIGH, CASE_LABEL, ASM_STRING, ASM_OUTPUTS, ASM_INPUTS,
1098 ASM_CLOBBERS): Add tree checks.
1100 * c-common.h (SWITCH_TYPE): Rename to SWITCH_STMT_TYPE.
1101 (SWITCH_STMT_COND, SWITCH_STMT_BODY): New.
1102 * c-common.def (SWITCH_STMT): Update to match.
1103 * c-common.c (c_do_switch_warnings): Use SWITCH_STMT accessor
1104 macros instead of SWITCH_EXPR ones.
1105 * c-dump.c (c_dump_tree): Likewise.
1106 * c-gimplify.c (gimplify_switch_stmt): Likewise.
1107 * c-typeck.c (c_start_case, do_case, c_finish_case): Likewise.
1109 * doc/c-tree.texi (SWITCH_STMT): Update accessor macro names.
1111 2005-01-27 Alan Modra <amodra@bigpond.net.au>
1113 * unwind-dw2.c (execute_stack_op): Add missing cases for
1114 DW_OP_shl, DW_OP_shr, DW_OP_shra, DW_OP_xor.
1116 2005-01-27 Eric Botcazou <ebotcazou@libertysurf.fr>
1118 * config.gcc (Obsolete configurations): Remove 'dummy', add
1119 sparclite-*-coff*, sparclite-*-elf*, sparc86x-*-elf* and
1122 2005-01-26 Diego Novillo <dnovillo@redhat.com>
1124 PR tree-optimization/19633
1125 * tree-ssa-alias.c (ptr_is_dereferenced_by): Also handle
1127 (maybe_create_global_var): Do not create .GLOBAL_VAR if there
1128 are no call-clobbered variables.
1129 * tree-outof-ssa.c (check_replaceable): Return false for calls
1132 2005-01-26 Ulrich Weigand <uweigand@de.ibm.com>
1134 * dbxout.c (dbxout_symbol_location): Resolve constant pool references
1135 even for variables with NULL DECL_INITIAL.
1137 2005-01-26 Stuart Hastings <stuart@apple.com>
1139 * gimplify.c (shortcut_cond_expr): Re-compute side-effects.
1141 2005-01-26 Richard Henderson <rth@redhat.com>
1144 * c-decl.c (finish_struct): Set DECL_MODE after resetting a
1146 * expr.c (store_field): Strip conversions to odd-bit-sized types
1147 if the destination field width matches.
1149 2005-01-26 Richard Henderson <rth@redhat.com>
1151 * c-decl.c, expmed.c, expr.c: Revert last change.
1153 2005-01-26 Richard Henderson <rth@redhat.com>
1156 * c-decl.c (finish_struct): Set DECL_MODE after resetting a
1158 * expmed.c (store_fixed_bit_field): Create a paradoxical subreg
1159 if we don't need the bits above those present in the current mode.
1160 * expr.c (store_field): Strip conversions to odd-bit-sized types
1161 if the destination field width matches.
1163 2005-01-26 Richard Sandiford <rsandifo@redhat.com>
1165 * config/mips/iris6.h (ENDFILE_SPEC): Don't link in irix-csr.o if
1166 either -ffast-math or -funsafe-math-optimizations is in use.
1168 2005-01-26 Marek Michalkiewicz <marekm@amelek.gda.pl>
1172 * config/avr/avr.c (notice_update_cc): Only set condition code for
1173 ashrqi3 if shift count > 0.
1174 (out_shift_with_cnt): Handle shift count <= 0 as a no-op.
1175 (ashlqi3_out, ashlhi3_out, ashlsi3_out, ashrqi3_out, ashrhi3_out,
1176 ashrsi3_out, lshrqi3_out, lshrhi3_out, lshrsi3_out): Handle shift
1177 count <= 0 as a no-op, and shift count >= width by copying zero
1178 or sign bit to all bits of the result.
1179 * config/avr/avr.md (all shifts): Add alternatives for zero shift
1180 count, with attribute "length" set to 0 and "cc" set to "none".
1182 2005-01-26 Aldy Hernandez <aldyh@redhat.com>
1184 * doc/invoke.texi: Document -mTLS.
1186 * testsuite/gcc.target/frv/all-tls-global-dynamic.c: New.
1187 * testsuite/gcc.target/frv/all-tls-initial-exec.c: New.
1188 * testsuite/gcc.target/frv/all-tls-initial-exec-pic.c: New.
1189 * testsuite/gcc.target/frv/all-tls-local-dynamic.c: New.
1190 * testsuite/gcc.target/frv/all-tls-local-dynamic-plt.c: New.
1191 * testsuite/gcc.target/frv/all-tls-local-dynamic-plt-pic.c: New.
1192 * testsuite/gcc.target/frv/all-tls-local-exec.c: New.
1193 * testsuite/gcc.target/frv/all-tls-local-exec-TLS.c: New.
1195 * configure.ac: Check for a TLS capable gas.
1197 * configure: Regenerate.
1199 2005-01-26 Richard Henderson <rth@redhat.com>
1202 * combine.c (make_field_assignment): Simplify store to zero_extract
1203 from a source with an overlapping mask.
1205 2005-01-26 Aldy Hernandez <aldyh@redhat.com>
1207 2004-11-11 Eric Christopher <echristo@redhat.com>
1209 * config/frv/frv.md: Add fr400_integer automaton. Don't
1210 allow TYPE_MUL, TYPE_MACC, or TYPE_CUT to issue in parallel
1213 2005-01-26 Steven Bosscher <stevenb@suse.de>
1216 * cfgbuild.c (make_edges): Do not clear or set
1217 current_function_has_computed_jump.
1218 * function.h (struct function): Remove the has_computed_jump field.
1219 (current_function_has_computed_jump): Do not define.
1220 * sched-rgn.c (is_cfg_nonregular): Return true if a basic block ends
1221 in a computed jump. Ignore current_function_has_computed_jump.
1223 2005-01-26 Richard Henderson <rth@redhat.com>
1226 * expr.c (categorize_ctor_elements): New argument p_must_clear.
1227 (categorize_ctor_elements_1): Likewise. Detect a union that isn't
1229 (mostly_zeros_p): Update for new categorize_ctor_elements argument.
1230 * gimplify.c (gimplify_init_constructor): Likewise. Only shove
1231 objects into static storage if they have more than one non-zero value.
1232 * tree.h (categorize_ctor_elements): Update decl.
1234 2005-01-26 Steven Bosscher <stevenb@suse.de>
1237 * tree.h (CALL_EXPR_TAILCALL): Add comment.
1238 * calls.c (check_sibcall_argument_overlap_1): Revert the change
1239 to this function from 2004-07-10.
1240 * tree-tailcall.c (suitable_for_tail_opt_p): Do not consider the
1241 the current function for tail call optimizations if the address
1242 of one of it its arguments is taken.
1244 2005-01-26 Kazu Hirata <kazu@cs.umass.edu>
1246 * cse.c (fold_rtx): Call equiv_constant only when necessary.
1248 2005-01-26 Bernd Schmidt <bernd.schmidt@analog.com>
1250 * function.c (expand_function_end): If current_function_calls_alloca,
1251 emit stack restore in a place that is reached when the function falls
1254 2005-01-26 Alexandre Oliva <aoliva@redhat.com>
1256 * Makefile.in (../$(build_subdir)/fixincludes/fixincl): Add dummy
1257 rule to avoid matching inappropriate suffix rule.
1259 2005-01-26 Jan Hubicka <jh@suse.cz>
1261 PR tree-optimization/19241
1262 * tree-inline.c (copy_body_r): Do not walk subtrees after substituting.
1264 2005-01-26 Hans-Peter Nilsson <hp@axis.com>
1266 * config/cris/cris.h (LIB_SPEC) <sim*>: Use -lsyslinux, not
1267 -lsyssim. Wrap -lc -lsyssim in --start-group --end-group instead
1269 (STARTFILE_SPEC): Add crti.o, before crtbegin.o.
1270 <sim*>: Always use crt1.o, regardless of N in simN.
1271 <!sim>: Drop support for -pg and -p variants; always use crt0.o.
1272 (ENDFILE_SPEC): Add crtn.o, after crtend.o.
1273 (CRT_CALL_STATIC_FUNCTION): Remove.
1274 * config/cris/aout.h (STARTFILE_SPEC): <!melinux && sim*>: Always
1275 use crt1.o, regardless of N in simN.
1276 <!sim>: Drop support for -pg and -p variants; always use crt0.o.
1277 (LIB_SPEC): Do not define; default to the one in config/cris/cris.h.
1278 (CRIS_CPP_SUBTARGET_SPEC) <melinux && !nostdinc>: Remove special
1279 case isystem setting.
1280 (MAX_OFILE_ALIGNMENT): Correct to 32, not 16.
1282 2005-01-25 Ian Lance Taylor <ian@airs.com>
1284 * gimple-low.c (block_may_fallthru): Correct handling of
1285 SWITCH_EXPR--may fall through if SWITCH_LABELS is NULL.
1287 2005-01-26 Hans-Peter Nilsson <hp@axis.com>
1289 * config/cris/cris.h (CPP_SPEC): Adjust the various mtune=
1290 handlers to match indentation; nested under the mtune=* handler.
1291 Similar for mcpu= and march=.
1293 2005-01-25 Daniel Jacobowitz <dan@codesourcery.com>
1295 * config/mips/linux.h (LIB_SPEC): Remove -rpath-link.
1296 * config/mips/linux64.h (LIB_SPEC): Likewise.
1297 * config/mn10300/linux.h (LIB_SPEC): Likewise.
1298 * config/sh/linux.h (LIB_SPEC): Likewise.
1300 2005-01-25 Aldy Hernandez <aldyh@redhat.com>
1302 * config/frv/frv.c (frv_legitimize_tls_address): New.
1303 (TARGET_HAVE_TLS): Define.
1304 (FRV_SYMBOL_REF_TLS_P): Define.
1305 (frv_override_options): Handle new register classes.
1306 (frv_legitimate_address_p): Reject tls addresses.
1307 (frv_legitimize_address): Handle TLS addresses.
1308 (gen_inlined_tls_plt): New.
1310 (frv_legitimize_tls_address): New.
1311 (unspec_got_name): Add TLS entries.
1312 (got12_operand): Add R_FRV_TLSMOFF12 case.
1313 (frv_emit_move): Fixup TLS addresses.
1314 (frv_emit_movsi): Legitimize TLS addresses.
1316 * config/frv/frv.h (MASK_BIG_TLS): New.
1317 (TARGET_BIG_TLS): New.
1318 (HAVE_AS_TLS): Define.
1319 (TARGET_SWITCHES): Add -mTLS and -mtls options.
1320 (enum reg_class): Add GR8_REGS, GR9_REGS, GR89_REGS.
1321 (REG_CLASS_NAMES): Same.
1322 (REG_CLASS_CONTENTS): Same.
1323 (CONSTRAINT_LEN): New.
1324 (REG_CLASS_FROM_CONSTRAINT): New.
1325 (PREDICATE_CODES): Add symbolic_operand.
1327 * config/frv/frv.md (define_constants): Add UNSPEC_GETTLSOFF,
1328 UNSPEC_TLS_LOAD_GOTTLSOFF12, UNSPEC_TLS_INDIRECT_CALL,
1329 UNSPEC_TLS_TLSDESC_LDD, UNSPEC_TLS_TLSDESC_LDD_AUX,
1330 UNSPEC_TLS_TLSOFF_LD, UNSPEC_TLS_LDDI, UNSPEC_TLSOFF_HILO,
1331 R_FRV_GOTTLSOFF_HI, R_FRV_GOTTLSOFF_LO, R_FRV_TLSMOFFHI,
1332 R_FRV_TLSMOFFLO, R_FRV_TLSMOFF12, R_FRV_TLSDESCHI,
1333 R_FRV_TLSDESCLO, R_FRV_GOTTLSDESCHI, R_FRV_GOTTLSDESCLO, GR8_REG,
1334 GR9_REG, GR14_REG, LRREG.
1335 (type): Add load_or_call attribute.
1336 ("load_or_call"): New reservation.
1337 ("call_gettlsoff"): New.
1338 ("tls_indirect_call"): New.
1339 ("tls_load_gottlsoff12"): New.
1340 ("tlsoff_hilo"): New.
1341 ("tls_tlsdesc_ldd"): New.
1342 ("tls_tlsoff_ld"): New.
1345 * config/frv/frv-protos.h (symbolic_operand): Protoize.
1347 2005-01-25 Alexandre Oliva <aoliva@redhat.com>
1349 * config/frv/frv.h (ASM_OUTPUT_DWARF_DTPREL): Define.
1350 * config/frv/frv-protos.h (frv_output_dwarf_dtprel): Declare.
1351 * config/frv/frv.c (TLS_BIAS): Define.
1352 (frv_output_dwarf_dtprel): New.
1354 2005-01-26 Jakub Jelinek <jakub@redhat.com>
1356 PR rtl-optimization/19579
1357 * ifcvt.c (noce_try_cmove_arith): If emitting instructions to set up
1358 both A and B, see if they don't clobber registers the other expr uses.
1360 2005-01-25 J"orn Rennecke <joern.rennecke@st.com>
1362 * real.c (do_add): Initialize signalling and canonical members.
1364 * real.c (real_from_integer): Zero out destination.
1366 2005-01-18 Jan Hubicka <jh@suse.cz>
1368 PR tree-optimize/19337
1369 * tree-inline.c (initialize_inlined_parameters): Copy
1370 saved_static_chain_decl only.
1372 2005-01-25 Richard Henderson <rth@redhat.com>
1376 * config/i386/i386.c (x86_inter_unit_moves): Zero.
1377 (ix86_preferred_reload_class): Rewrite fp-constant section, with
1378 80387 enabled, to return a proper subclass. Return the subset that
1379 overlaps with GENERAL_REGS for PLUS.
1380 * config/i386/i386.md (movsi_1, movdi_2): Set type to "mmx" for pxor.
1382 2005-01-25 Richard Henderson <rth@redhat.com>
1385 * tree-complex.c (expand_complex_div_wide): Use the correct formulae.
1387 2005-01-25 Roger Sayle <roger@eyesopen.com>
1389 * expmed.c (choose_multiplier): Change interface to return the
1390 multiplier lowpart as an rtx instead of an unsigned HOST_WIDE_INT.
1391 (expand_divmod): Update calls to choose_multiplier.
1393 2005-01-24 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1396 * expmed.c (extract_bit_field): Use adjust_address instead of
1397 gen_lowpart when op0 is a MEM.
1399 2005-01-24 Steven Bosscher <stevenb@suse.de>
1401 * cfgexpand.c (expand_gimple_tailcall): Fix typo.
1403 2005-01-24 Richard Sandiford <rsandifo@redhat.com>
1405 * config/mips/irix-csr.c: New file.
1406 * config/mips/t-iris6 (irix-csr.o): New rule to build it.
1407 (EXTRA_MULTILIB_PARTS): Add irix-csr.o.
1408 * config/mips/iris6.h (ENDFILE_SPEC): Include it in n32 and n64
1411 2005-01-24 Eric Botcazou <ebotcazou@libertysurf.fr>
1414 * config.gcc (sparc-*-elf*): Remove sol2.h, sparc/sol2.h and
1415 sparc/elf.h, add sparc/sp-elf.h.
1416 (sparc-*-rtems*): Likewise.
1417 (sparclite-*-elf*): Remove sol2.h, sparc/sol2.h, sparc/elf.h and
1418 tm-dwarf2.h, add sparc/sp-elf.h.
1419 (sparc86x-*-elf): Likewise.
1420 (sparc64-*-elf*): Remove sol2.h, sparc/sol2.h and tm-dwarf2.h.
1421 * config/sparc/liteelf.h (TARGET_SUB_OS_CPP_BUILTINS): Rename into
1422 TARGET_OS_CPP_BUILTINS.
1423 * config/sparc/sp86x-elf (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1424 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1426 * config/sparc/openbsd64.h (NO_IMPLICIT_EXTERN_C): Undefine.
1427 * config/sparc/sp64-elf.h (NO_IMPLICIT_EXTERN_C): New macro.
1428 (SWITCH_TAKES_ARG): Likewise.
1429 (LOCAL_LABEL_PREFIX): Likewise.
1430 (ASM_GENERATE_INTERNAL_LABEL): Likewise.
1431 (TARGET_N_FORMAT_TYPES): Delete.
1432 (TARGET_FORMAT_TYPES): Likewise.
1433 (ASM_DECLARE_FUNCTION_SIZE): Likewise.
1434 * config/sparc/elf.h: Delete.
1435 * config/sparc/sp-elf.h: New file.
1437 2005-01-24 Kazu Hirata <kazu@cs.umass.edu>
1439 * tree-flow-inline.h (phi_arg_from_edge): Remove.
1440 * tree-flow.h: Remove the corresponding prototype.
1441 * tree-ssa-operands.h, tree-ssa-threadupdate.c, tree-ssa.c,
1442 tree-vectorizer.c: Use dest_idx instead of phi_arg_from_edge.
1444 2005-01-24 Jakub Jelinek <jakub@redhat.com>
1447 * config/rs6000/rs6000.c (rs6000_file_start): Create toc section
1448 for AIX ABI or ELF -fPIC.
1449 (rs6000_emit_load_toc_table): Don't create toc_section here.
1450 (rs6000_xcoff_file_start): Nor here.
1452 2005-01-24 J"orn Rennecke <joern.rennecke@st.com>
1454 * sh.c (ra.h): Don't #include.
1455 (hard_regs_intersect_p): New function, resurrected from ra.c.
1457 * sh.c: Fix 1996 Copyright.
1459 2005-01-24 Richard Henderson <rth@redhat.com>
1460 Aldy Hernandez <aldyh@redhat.com>
1462 * regrename.c (note_sets): Handle subregs.
1464 2005-01-24 Joseph S. Myers <joseph@codesourcery.com>
1467 * genconditions.c (write_header, write_conditions): Elide file if
1470 2005-01-24 Kazu Hirata <kazu@cs.umass.edu>
1472 * c-opts.c, expmed.c, gengtype.c, tree-complex.c,
1473 config/arm/iwmmxt.md, config/avr/avr.c, config/s390/s390.h:
1476 2005-01-24 Jakub Jelinek <jakub@redhat.com>
1478 * builtins.c (get_pointer_alignment, gimplify_va_arg_expr,
1479 expand_builtin_printf, expand_builtin_fprintf,
1480 expand_builtin_sprintf): Use POINTER_TYPE_P instead TREE_CODE
1481 checking against POINTER_TYPE.
1482 (validate_arglist): Handle POINTER_TYPE code by checking
1485 2005-01-24 Paolo Bonzini <bonzini@gnu.org>
1487 * rtlanal.c (non_rtx_starting_operands, for_each_rtx_1,
1489 (for_each_rtx): Call for_each_rtx_1.
1490 * rtl.h (init_rtlanal): Declare.
1491 * toplev.c (backend_init): Call init_rtlanal.
1493 2005-01-24 Jakub Jelinek <jakub@redhat.com>
1495 * flow.c (propagate_one_insn): Formatting.
1498 * flow.c (libcall_dead_p): Be more conservative if unsure.
1499 If there are any instructions between insn and call, see if they are
1500 all dead before saying the libcall is dead.
1502 2005-01-24 Uros Bizjak <uros@kss-loka.si>
1504 * config/i386/i386.md (*extendsfdf2_sse): Use "nonimmediate_operand"
1505 operand constraint for operand 0.
1507 2005-01-23 Richard Henderson <rth@redhat.com>
1510 * tree-complex.c (expand_complex_div_wide): Set EDGE_FALLTHRU on
1511 edges leading to join block.
1513 2005-01-23 Roger Sayle <roger@eyesopen.com>
1515 * expmed.c (expand_mult_highpart): Make static. Change type of
1516 constant multiplier argument from unsigned HOST_WIDE_INT to rtx.
1517 (expand_divmod): Updates calls to expand_mult_highpart by using
1518 gen_int_mode to pass a CONST_INT rtx.
1519 * rtl.h (expand_mult_highpart): Remove prototype.
1521 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
1523 * flow.c (regset_bytes, regset_size): Remove.
1525 * cfganal.c: Fix a reference to Harvey's paper.
1527 2005-01-23 Steven Bosscher <stevenb@suse.de>
1529 PR rtl-optimization/19464
1530 * tree-optimize.c (init_tree_optimization_passes): Add one more
1531 copyrename pass just before out-of-ssa.
1533 2005-01-23 Joseph S. Myers <joseph@codesourcery.com>
1536 * recog.c (recog_memoized): Don't define if GENERATOR_FILE.
1537 * ggc-none.c (ggc_free): Define.
1539 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
1541 * cse.c (max_reg, max_insn_uid): Remove.
1542 (cse_main): Don't access max_reg or max_insn_uid.
1544 * c-common.c (explicit_flag_signed_bitfields,
1545 lang_statement_code_p, lang_expand_function_end): Remove.
1546 * c-common.h: Remove the corresponding declarations.
1547 * c-opts.c (c_common_handle_option): Remove all write access
1548 to explicit_flag_signed_bitfields.
1550 * cgraph.c (cgraph_varpool_n_nodes): Remove.
1551 (cgraph_varpool_node): Don't access cgraph_varpool_n_nodes.
1552 * cgraph.h: Remove the corresponding declaration.
1554 * gcse.c (null_pointer_info): Remove.
1556 2005-01-23 Roger Sayle <roger@eyesopen.com>
1557 Eric Botcazou <ebotcazou@libertysurf.fr>
1559 * combine.c (simplify_logical): Only simplify logical expressions
1560 of the form ior(and(x,y),z) by the inverse distributive law if the
1561 result is cheaper than the original.
1563 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
1565 * alias.c, c-common.h, c-incpath.c, c-incpath.h, expr.c,
1566 fold-const.c, gimplify.c, params.h, tree-data-ref.c,
1567 tree-if-conv.c, tree-nested.c, tree-outof-ssa.c,
1568 tree-ssa-dom.c, tree-vectorizer.c, tree.def, config/darwin.c,
1569 config/freebsd-spec.h, config/arm/arm.h,
1570 config/h8300/h8300.md, config/i386/i386.md,
1571 config/i386/predicates.md, config/i386/sse.md,
1572 config/ia64/ia64.c, config/ip2k/ip2k.c, config/s390/s390.c,
1573 config/vax/vax.md: Fix comment typos. Follow spelling
1576 2005-01-23 Steven Bosscher <stevenb@suse.de>
1578 * gengtype.c (walk_type): Produce `not equal to' compares for loop
1579 tests, instead of `less than'.
1581 2005-01-23 Richard Sandiford <rsandifo@redhat.com>
1583 * function.c (assign_parm_setup_block): When creating a new stack slot
1584 for a parameter, get its alignment from the parameter's DECL_ALIGN
1585 rather than the type's TYPE_ALIGN. Make sure that the parameter
1586 is at least word aligned.
1588 2005-01-22 Roger Sayle <roger@eyesopen.com>
1591 * config/avr/avr.c (avr_hard_regno_mode_ok): Rewrite.
1593 2005-01-22 Richard Henderson <rth@redhat.com>
1596 * config/i386/i386.md (movsfcc_1_sse_max): Use nonimmediate_operand
1597 in both compare operands.
1598 (movdfcc_1_sse_max): Likewise.
1599 (movsfcc_1_sse): Likewise. Add earlyclobber for scratch.
1600 (movdfcc_1_sse): Likewise.
1601 * config/i386/i386.c (ix86_split_sse_movcc): Emit copies into the
1602 scratch register as needed.
1604 2005-01-22 Richard Henderson <rth@redhat.com>
1606 * config/i386/i386.md (smaxsf3): Fix mnemonic typo.
1608 * config/i386/i386.c (ix86_prepare_fp_compare_args): Fix is_sse test.
1610 2005-01-22 Richard Henderson <rth@redhat.com>
1612 * genopinit.c (optabs): Use smin/smax for floating point too.
1613 * doc/md.texi: Update to match. Clarify that floating point
1614 results are undefined for +0/-0 and NaN.
1615 * doc/rtl.texi: Likewise.
1616 * rtl.def (SMIN, SMAX): Likewise
1617 * tree.def (MIN_EXPR, MAX_EXPR): Likewise.
1619 * config/alpha/alpha.md (smaxdf3, smindf3, smaxsf3, sminsf3): Add
1620 leading 's' to the name.
1621 * config/ia64/ia64.md (smaxsf3, sminsf3, smaxdf3, smindf3,
1622 smaxxf3, sminxf3): Likewise.
1623 * config/rs6000/rs6000.md (smaxdf3, smindf3, smaxsf3, sminsf3):
1626 2005-01-22 Paul Brook <paul@codesourcery.com>
1628 * config/arm/arm.md: Use "Uy" constraint for wcgr load/stores.
1630 2005-01-21 Mark Dettinger <dettinge@de.ibm.com>
1632 * config/s390/s390.c (struct processor_costs): New fields
1633 dlgr, dlr, dr, dsgfr, dsgr.
1634 (z900_cost, z990_cost): Values for new fields.
1635 (s390_rtx_costs): New cases MEM und COMPARE in switch
1636 statement. Modified handling of SIGN_EXTEND, ZERO_EXTEND,
1637 DIV, MOD, UDIV, UMOD.
1639 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
1641 * config/s390/s390.md ("doloop_si64"): Reload input value directly
1642 into the register being decremented.
1643 ("doloop_si31", "doloop_di"): Likewise.
1644 ("*doloop_si_long"): Adapt pattern.
1646 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
1648 * config/s390/s390.h (HARD_REGNO_NREGS): Fix computation for
1650 (CLASS_MAX_NREGS): Likewise.
1652 2005-01-21 Daniel Berlin <dberlin@dberlin.org>
1654 * doc/tree-ssa.texi (Statement Operands): Add example for new
1655 must-def macro. Note deprecation of old operands interface.
1657 2005-01-22 Richard Sandiford <rsandifo@redhat.com>
1659 PR tree-optimization/19484
1660 * tree-cfg.c (remove_fallthru_edge): New function.
1661 (cleanup_control_flow): Remove fallthru edges from calls that are
1662 now known not to return.
1664 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
1666 Fix PR tree-optimization/19038
1667 * tree-ssa-dom.c (cprop_operand): Don't replace loop invaeriant
1668 copies with loop variant ones.
1670 2005-01-22 Kazu Hirata <kazu@cs.umass.edu>
1672 * cfganal.c, real.h, reorg.c, timevar.def, tree-ssa-ccp.c,
1673 config/alpha/alpha-protos.h, config/alpha/alpha.h,
1674 config/alpha/alpha.md, config/alpha/predicates.md,
1675 config/sparc/freebsd.h, config/sparc/netbsd-elf.h,
1676 config/sparc/sol2.h: Update copyright.
1678 * tree-cfg.c (remove_forwarder_block_with_phi): Look at the
1679 first label to see if it is a nonlocal label.
1681 2005-01-22 David Edelsohn <edelsohn@gnu.org>
1682 Andrew Pinski <pinskia@physics.uc.edu>
1685 * config/rs6000/rs6000.c (rs6000_va_start): Saturate n_gpr at
1686 maximum number of GPRs. Saturate n_fpr at maximum number of FPRs.
1688 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
1690 * config/rs6000/rtems.h: Update copyright.
1692 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
1695 * config/rs6000/rtems.h: Resurrect cpp_os_rtems_spec from gcc < 3.4.
1696 (CPP_OS_RTEMS_SPEC): New (From gcc-3.3's config/rs6000/sys4.h).
1697 (SUBSUBTARGET_EXTRA_SPECS): Use CPP_OS_RTEMS_SPEC.
1699 2005-01-22 Volker Reichelt <reichelt@igpm.rwth-aachen.de>
1702 * c-typeck.c (convert_arguments): Check for error_mark_node.
1704 2005-01-21 Ian Lance Taylor <ian@airs.com>
1706 PR tree-optimization/13000
1707 * tree-inline.c: Include "tree-flow.h".
1708 (expand_call_inline): If warn_return_type, warn if non-void inline
1709 function falls through.
1710 * tree-cfg.c (execute_warn_function_return): Don't warn about
1711 control reaching end if TREE_NO_WARNING is set. Set
1713 * gimple-low.c (block_may_fallthru): Don't assume that SWITCH_EXPR
1715 * gimplify.c (shortcut_cond_expr): Don't emit a jump over the else
1716 branch if we don't need one.
1717 * c-typeck.c: Include "tree-flow.h"
1718 (c_finish_bc_stmt): Don't add a goto if the current statement
1719 list doesn't fall through to the current point.
1721 2005-01-21 Roger Sayle <roger@eyesopen.com>
1723 PR rtl-optimization/576
1724 * real.c (real_arithmetic): Change return type from void to bool
1725 to return an indication that the result may be inexact.
1726 * real.h (real_arithmeric): Update prototype.
1727 * fold-const.c (const_binop): Don't constant fold floating
1728 point expressions when the user specifies -frounding-math and
1729 the result may depend upon the run-time rounding mode.
1730 (fold_convert_const_real_from_real): Clean-up.
1731 (fold_initializer): Ignore flag_rounding_math for initializers.
1732 * simplify-rtx.c (simplify_binary_operation): Likewise, don't
1733 constant fold FP operations with flag_rounding_math if the
1734 result may depend upon the run-time rounding mode.
1736 2005-01-21 Tom Tromey <tromey@redhat.com>
1738 * c-cppbuiltin.c (define__GNUC__): Correct assertion.
1740 2005-01-21 Kazu Hirata <kazu@cs.umass.edu>
1742 * tree-cfg.c: Fix comment typos.
1744 2005-01-21 J"orn Rennecke <joern.rennecke@st.com>
1746 * passes.c (rest_of_handle_flow2): Close / open the flow2 dump file
1747 around the call to rest_of_handle_branch_target_load_optimize.
1749 2005-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
1751 * config.gcc (sparc-*-netbsdelf*): Include dbxelf.h
1752 (sparc64-*-openbsd*): Likewise.
1753 (sparclite-*-elf*): Include tm-dwarf2.h.
1754 (sparc86x-*-elf*): Likewise.
1755 (sparc64-*-elf*): Likewise.
1756 * config/sparc/linux64.h (DWARF2_DEBUGGING_INFO): Delete.
1757 (DBX_DEBUGGING_INFO): Likewise.
1758 * config/sparc/liteelf.h (DWARF2_DEBUGGING_INFO): Likewise.
1759 (PREFERRED_DEBUGGING_TYPE): Likewise.
1760 * config/sparc/netbsd-elf.h (PREFERRED_DEBUGGING_TYPE): Likewise.
1761 * config/sparc/sp64-elf.h (DBX_DEBUGGING_INFO): Likewise.
1762 (PREFERRED_DEBUGGING_TYPE): Likewise.
1763 * config/sparc/sp86x-elf.h (DWARF2_DEBUGGING_INFO): Likewise.
1764 (PREFERRED_DEBUGGING_TYPE): Likewise.
1765 * config/sparc/sparc.h (DBX_DEBUGGING_INFO): Likewise.
1767 2005-01-21 Hans-Peter Nilsson <hp@bitrange.com>
1770 * combine.c (combine_simplify_rtx): Revert change of 2004-12-31.
1772 2005-01-20 Paul Brook <paul@codesourcery.com>
1774 * doc/rtl.texi: Document value extension requirements for CONST_INT.
1776 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
1778 * tree-cfg.c (tree_verify_flow_info): Check that a nonlocal
1779 label is first in a sequence of labels.
1781 * tree-cfg.c (tree_verify_flow_info): Fix a typo.
1783 2005-01-20 Janis Johnson <janis187@us.ibm.com>
1784 Giovanni Bajo <giovannibajo@gcc.gnu.org>
1786 * doc/sourcebuild.texi (Test Directives): New.
1788 2005-01-20 Roger Sayle <roger@eyesopen.com>
1790 * real.c (real_floor): Don't leave the result uninitialized when
1791 mode is VOIDMode, but return the unrounded intermediate value.
1792 (real_ceil): Likewise.
1794 2005-01-20 Paul Brook <paul@codesourcery.com>
1796 * config/arm/arm.md (insv): Use gen_int_mode.
1798 2005-01-20 Eric Botcazou <ebotcazou@libertysurf.fr>
1800 * config/sparc/freebsd.h (ASM_OUTPUT_INTERNAL_LABELREF): Delete.
1801 * config/sparc/linux.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1802 * config/sparc/linux64.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1803 * config/sparc/netbsd-elf.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1804 * config/sparc/sol2.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1806 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
1809 * dwarf2out.c (concat_loc_descriptor): We don't know we can use
1810 fbreg, so we have to assume we can't.
1811 (loc_descriptor_from_tree_1): Ditto.
1812 (containing_function_has_frame_base): New function.
1813 (add_location_or_const_value_attribute): Use it.
1814 Also try to generate a frame_base from a single element location
1817 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
1819 PR tree-optimization/15349
1820 * timevar.def (TV_TREE_MERGE_PHI): New.
1821 * tree-cfg.c (tree_forwarder_block_p): Add a new argument
1823 (remove_forwarder_block, cleanup_forwarder_blocks): Adjust the
1824 calls to tree_forwarder_block_p.
1825 (remove_forwarder_block_with_phi, merge_phi_nodes,
1826 gate_merge_phi, pass_merge_phi): New.
1827 * tree-optimize.c (init_tree_optimization_passes): Add
1829 * tree-pass.h: Add an extern for pass_merge_phi;
1831 2005-01-20 Richard Henderson <rth@redhat.com>
1834 * config/i386/emmintrin.h (_mm_castpd_ps, _mm_castpd_si128): New.
1835 (_mm_castps_pd, _mm_castps_si128): New.
1836 (_mm_castsi128_ps, _mm_castsi128_pd): New.
1838 2005-01-20 Richard Henderson <rth@redhat.com>
1841 * config/i386/mmintrin.h (_mm_cvtsi32_si64): Use
1842 __builtin_ia32_vec_init_v2si.
1843 (_mm_cvtsi64_si32): Use __builtin_ia32_vec_ext_v2si.
1844 * config/i386/i386.c (IX86_BUILTIN_VEC_EXT_V2SI): New.
1845 (ix86_init_mmx_sse_builtins): Create it.
1846 (ix86_expand_builtin): Expand it.
1847 (ix86_expand_vector_set): Handle V2SFmode and V2SImode.
1848 * config/i386/mmx.md (vec_extractv2sf_0, vec_extractv2sf_1): New.
1849 (vec_extractv2si_0, vec_extractv2si_1): New.
1851 2005-01-20 Richard Henderson <rth@redhat.com>
1854 * config/i386/i386.c (ix86_expand_push): New.
1855 * config/i386/mmx.md (push<MMXMODE>1): New.
1856 * config/i386/sse.md (push<SSEMODE>1): New.
1857 * config/i386/i386-protos.h: Update.
1859 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
1861 * params.def, params.h, predict.c, tree-ssa-loop-im.c: Update
1864 2005-01-20 Richard Henderson <rth@redhat.com>
1866 * rtl.def (CONST_VECTOR): Use RTX_CONST_OBJ.
1867 * rtl.h (CONSTANT_P): Don't special case CONST_VECTOR.
1869 2005-01-19 Richard Henderson <rth@redhat.com>
1872 * config/i386/i386.c (ix86_expand_vector_move_misalign): Convert
1873 to V4SFmode in SSE1 fallback load path.
1875 2005-01-19 Richard Henderson <rth@redhat.com>
1877 * config/i386/i386.c (ix86_expand_vector_init_one_var): Fix typo
1878 in QImode expansion to ix86_expand_vector_set.
1879 (ix86_expand_vector_init_general): Fix typo in V8HImode recursive call.
1881 2005-01-19 Richard Henderson <rth@redhat.com>
1884 * config/i386/i386.c (ix86_preferred_reload_class): Return a proper
1885 subclass of the input class.
1886 (ix86_secondary_memory_needed): Always true for cross-MMX classes.
1887 Always true for cross-SSE1 classes. Rationalize conditionals.
1888 * config/i386/i386.h (SSE_CLASS_P, MMX_CLASS_P): Use straight equality.
1889 * config/i386/i386.md (movsi_1): Add MMX/SSE zeros. Fix alternatives
1890 for SSE1. Don't check TARGET_INTER_UNIT_MOVES.
1891 (movdi_2): Add MMX/SSE zeros.
1892 (movdi_1_rex64): Likewise. Don't check TARGET_INTER_UNIT_MOVES.
1893 (movsf_1): Don't check TARGET_INTER_UNIT_MOVES.
1894 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
1895 (movsi_1_nointernunit, movdi_1_rex64_nointerunit): Remove.
1896 (movsf_1_nointerunit, zero_extendsidi2_32_1): Remove.
1897 (zero_extendsidi2_rex64_1): Remove.
1898 (MOV0 peephole): Check GENERAL_REG_P.
1900 2005-01-19 Richard Henderson <rth@redhat.com>
1903 * config/i386/i386.c (ix86_expand_vector_set): Fix third and fourth
1905 (ix86_expand_vector_extract): Likewise.
1907 2005-01-19 Richard Henderson <rth@redhat.com>
1910 * expr.c (emit_move_change_mode): New argument force; use
1911 simplify_gen_subreg if true.
1912 (emit_move_via_alt_mode): Merge into ...
1913 (emit_move_via_integer): ... here.
1914 (emit_move_ccmode): Use emit_move_change_mode directly.
1916 2005-01-19 Richard Henderson <rth@redhat.com>
1919 * config/alpha/alpha.c (alpha_rtx_costs): Handle HIGH.
1920 (alpha_preferred_reload_class): Handle CONST_VECTOR.
1921 (alpha_emit_set_const_1): Add no_output parameter; don't emit
1923 (alpha_emit_set_const): Likewise. Make static.
1924 (alpha_emit_set_long_const): Make static.
1925 (alpha_extract_integer): Split out from alpha_expand_mov.
1926 (alpha_split_const_mov): Likewise.
1927 (alpha_expand_mov): Use them. Handle CONST_VECTOR.
1928 (alpha_legitimate_constant_p): New.
1929 * config/alpha/alpha-protos.h: Update.
1930 * config/alpha/alpha.h (REGISTER_MOVE_COST): Correct fp<->gp cost.
1931 (LEGITIMATE_CONSTANT_P): Re-implement with a function.
1932 * config/alpha/alpha.md (movsi): Add n alternative.
1933 (movsi_nt_vms, movdi_er_nofix, movdi_er_fix, movdi_fix): Likewise.
1934 (mov<VEC>_fix, mov<VEC>_nofix): Add i alternative.
1935 (splitters for all of the above): Use alpha_split_const_mov.
1936 * config/alpha/predicates.md (non_add_const_operand): New.
1937 (non_zero_const_operand): New.
1938 (input_operand): Use alpha_legitimate_constant_p after reload.
1940 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
1942 PR tree-optimization/19038
1943 * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): New function.
1944 (add_candidate): Add ivs with increment in latch only if
1945 allow_ip_end_pos_p is true.
1946 (determine_iv_cost): Use empty_block_p.
1948 2005-01-19 Daniel Berlin <dberlin@dberlin.org>
1950 * cfganal.c (compute_dominance_frontiers_1): Replace with new algorithm
1951 (compute_dominance_frontiers): Ditto.
1953 2005-01-19 Ralf Corsepius <ralf.corsepius@rtems.org>
1956 * config/sh/t-rtems: New.
1957 * config.gcc (sh-*-rtems*): Reflect having added config/sh/t-rtems.
1959 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
1961 * tree-ssa-ccp.c (maybe_fold_offset_to_component_ref): Always subtract
1962 the offset of the selected field.
1964 2005-01-19 Kazu Hirata <kazu@cs.umass.edu>
1966 * tree-cfg.c (remove_forwarder_block): Fix the check to
1967 prevent a nonlocal label from appearing in the middle of a
1970 2005-01-19 Hans-Peter Nilsson <hp@axis.com>
1972 PR rtl-optimization/19462
1973 * reorg.c (find_end_label): Create return insn only if
1974 current_function_epilogue_delay_list is empty.
1976 2005-01-19 Paolo Bonzini <bonzini@gnu.org>
1978 * doc/passes.texi: Remove paragraph mentioning new-ra.
1980 2005-01-19 Jakub Jelinek <jakub@redhat.com>
1982 PR rtl-optimization/15139
1983 * combine.c: Include params.h.
1984 (count_rtxs): New function.
1985 (record_value_for_reg): If replace_rtx would replace at least
1986 2 occurrences of REG in VALUE and TEM is really large, replace REG with
1987 (clobber (const_int 0)) instead of TEM.
1988 * params.def (PARAM_MAX_LAST_VALUE_RTL): New.
1989 * params.h (MAX_LAST_VALUE_RTL): New.
1990 * Makefile.in (combine.o): Depend on $(PARAMS_H).
1991 * doc/invoke.texi (--param max-last-value-rtl=N): Document.
1994 * c-typeck.c (digest_init): Only call build_vector if all constructor
1995 elements are *_CST nodes.
1996 * gimplify.c (gimplify_init_constructor): Likewise.
1999 * c-typeck.c (digest_init): Only call build_vector if inside_init
2002 2005-01-18 Tobias Schl"uter <tobias.schlueter@physik.uni-muenchen.de>
2004 * toplev.c (init_asm_output): Remove dead #ifdef.
2006 2005-01-18 Kaz Kojima <kkojima@gcc.gnu.org>
2008 * config/sh/linux.h (TARGET_C99_FUNCTIONS): Define.
2010 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
2012 * calls.c (expand_call): Check DECL_BUILT_IN_CLASS before
2013 accessing DECL_FUNCTION_CODE.
2014 * dojump.c (do_jump): Likewise.
2015 * gimplify.c (gimplify_call_expr): Likewise.
2016 * predict.c (expr_expected_value): Likewise.
2017 (strip_builtin_expect): Likewise.
2018 * tree-inline.c (estimate_num_insns_1): Likewise.
2019 * tree-ssa-loop-im.c (stmt_cost): Likewise
2020 * fold-const.c (fold): Test for BUILT_IN_NORMAL.
2021 (tree_expr_nonnegative_p): Likewise.
2023 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
2025 * config/sparc/sparc.c (load_pic_register): Emit the appropriate
2026 variant of the load_pcrel_sym pattern.
2027 * config/sparc/sparc.md (P macro): Move to the top.
2028 (load_pcrel_sym): Macroize using P.
2029 (save_register_window): Likewise.
2031 2005-01-18 Aldy Hernandez <aldyh@redhat.com>
2033 * config/rs6000/rs6000.md ("sunordered"): Disable for e500.
2036 2005-01-18 Andrew Pinski <pinskia@physics.uc.edu>
2039 * c-typeck.c (build_asm_expr): Strip nops off
2040 input memory operands.
2042 2005-01-18 David Edelsohn <edelsohn@gnu.org>
2044 * config/rs6000/aix43.h (CPLUSPLUS_CPP_SPEC): Delete
2045 _XOPEN_SOURCE, _XOPEN_SOURCE_EXTENDED, _LARGE_FILE_API.
2046 * config/rs6000/aix51.h (CPLUSPLUS_CPP_SPEC): Same.
2047 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Same.
2049 2005-01-18 Joel Sherrill <joel.sherrill@oarcorp.com>
2052 * config/i386/i386.c (override_options): If the 80387 is disabled,
2053 then do not return FP values using FP registers.
2055 2005-01-18 Kazu Hirata <kazu@cs.umass.edu>
2057 * basic-block.h, c-common.c, c-cppbuiltin.c, c-lang.c,
2058 c-tree.h, cfgbuild.c, cgraph.c, cgraph.h, collect2.c,
2059 combine.c, config.gcc, coverage.h, cse.c, cselib.c,
2060 defaults.h, df.c, dwarf2asm.c, dwarf2out.c, explow.c, expr.c,
2061 flow.c, fold-const.c, gcse.c, ggc-page.c, gimple-low.c,
2062 gimplify.c, ifcvt.c, langhooks-def.h, lcm.c, optabs.h,
2063 output.h, postreload-gcse.c, postreload.c, recog.c,
2064 resource.c, rtl.def, rtlanal.c, sched-deps.c, sched-rgn.c,
2065 targhooks.h, toplev.c, tree-data-ref.c, tree-eh.c,
2066 tree-flow.h, tree-loop-linear.c, tree-mudflap.h, tree-nrv.c,
2067 tree-optimize.c, tree-outof-ssa.c, tree-pass.h,
2068 tree-scalar-evolution.c, tree-ssa-copy.c, tree-ssa-dce.c,
2069 tree-ssa-dse.c, tree-ssa-loop-ivopts.c, tree-ssa-operands.c,
2070 tree-ssa-pre.c, tree-ssa.c, tree-vectorizer.c, tree.def,
2071 unwind-dw2-fde-darwin.c, var-tracking.c: Update copyright.
2073 * config/darwin.c, config/darwin.h, config/freebsd-spec.h,
2074 config/alpha/alpha.c, config/arm/vfp.md,
2075 config/i386/emmintrin.h, config/i386/i386-modes.def,
2076 config/i386/i386-protos.h, config/i386/i386.h,
2077 config/i386/pmmintrin.h, config/i386/ppro.md,
2078 config/i386/predicates.md, config/i386/xmmintrin.h,
2079 config/i860/i860.c, config/ia64/ia64-modes.def,
2080 config/ia64/ia64-protos.h, config/ia64/ia64.h,
2081 config/ia64/ia64.md, config/ia64/itanium1.md,
2082 config/ia64/itanium2.md, config/ia64/predicates.md,
2083 config/iq2000/iq2000.h, config/mips/linux64.h,
2084 config/rs6000/aix41.h, config/rs6000/aix43.h,
2085 config/rs6000/aix52.h, config/rs6000/darwin-fallback.c,
2086 config/rs6000/rs6000.c, config/rs6000/rs6000.h,
2087 config/rs6000/spe.md, config/sh/sh.md, config/sparc/linux.h,
2088 config/sparc/linux64.h, config/sparc/litecoff.h,
2089 config/sparc/sp64-elf.h, config/sparc/sparc.c,
2090 config/sparc/sparc.h, config/sparc/sparc.md,
2091 config/sparc/sysv4-only.h: Update copyright.
2093 2005-01-18 Richard Henderson <rth@redhat.com>
2096 * config/i386/i386.c (ix86_expand_fp_movcc): Fail for LTGT and UNEQ.
2098 2005-01-18 Marc Espie <espie@openbsd.org>
2099 * config/i386/openbsdelf.h: Typo.
2101 2005-01-18 Richard Henderson <rth@redhat.com>
2103 * config/i386/i386.c (ix86_expand_fp_movcc): Remove TARGET_IEEE_FP
2104 special case for sse.
2105 * config/i386/predicates.md (sse_comparison_operator): Likewise.
2107 2005-01-18 Dorit Naishlos <dorit@il.ibm.com>
2109 * tree-inline.c (estimate_num_insns_1): Added cases for
2110 ALIGN_INDIRECT_REF, MISALIGNED_INDIRECT_REF, and REALIGN_LOAD_EXPR.
2112 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
2114 PR rtl-optimization/19296
2115 * combine.c (simplify_comparison): Rewrite the condition under
2116 which a non-paradoxical SUBREG of a PLUS can be lifted when
2117 compared against a constant.
2119 2005-01-18 Andi Kleen <ak@muc.de>
2121 * c-typeck.c: (convert_for_assignment): Check warn_pointer_sign.
2122 * c.opt (-Wpointer-sign): Add.
2123 * doc/invoke.texi: (-Wpointer-sign): Add.
2125 2005-01-18 Uros Bizjak <uros@kss-loka.si>
2127 * config/i386/i386.c (override_options): Revert 2004-11-24 change.
2128 * config/i386/i386.md (fmodsf3, dremsf3, *sinsf2, *cossf2, sincossf3,
2129 *tansf3_1, tansf2, atan2sf3_1, atan2sf3, atansf2, asinsf2, acossf2,
2130 logsf2, log10sf2, log2sf2, log1psf2, logbsf2, ilogbsf2, expsf2,
2131 exp10sf2, exp2sf2, expm1sf2, rintsf2, floorsf2, ceilsf2, btruncsf2,
2132 nearbyintsf2): Disable for TARGET_SSE_MATH. Leave patterns enabled
2133 for TARGET_MIX_SSE_I387.
2134 (fmoddf3, dremdf3, *sindf2, *sinextendsfdf2, *cosdf2, *cosextendsfdf2,
2135 sincosdf3, *sincosextendsfdf3, *tandf3_1, tandf2, atan2df3_1,
2136 atan2df3, atandf2, asindf2, acosdf2, logdf2, log10df2, log2df2,
2137 log1pdf2, logbdf2, expdf2, exp10df2, exp2df2, expm1df2, rintdf2,
2138 floordf2, ceildf2, btruncdf2, nearbyintdf2): Disable for
2139 (TARGET_SSE2 && TARGET_SSE_MATH). Leave patterns enabled for
2140 TARGET_MIX_SSE_I387.
2141 (atan2sf3, atan2df3, atan2xf3): Remove register constraints
2144 2005-01-18 Uros Bizjak <uros@kss-loka.si>
2147 * config/i386/mmx.md (*movv2sf_internal_rex64, *movv2sf_internal):
2148 Add movaps alternative for xmm reg->reg move.
2150 2005-01-17 Jeff Law <law@redhat.com>
2152 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Clear SSA_NAME_VALUE
2153 at the end of the main DOM loop rather than just before DOM exits.
2155 2005-01-17 Diego Novillo <dnovillo@redhat.com>
2157 PR tree-optimization/19121
2158 * tree-ssa-alias.c (compute_flow_sensitive_aliasing): When
2159 adding aliases to a name tag, also add them to the pointer's
2161 * tree-ssa-copy.c (merge_alias_info): Do not merge flow
2162 sensitive alias info at all. Only check that the two pointers
2163 have compatible pointed-to sets.
2164 * tree-ssa.c (verify_name_tags): Verify that the alias set of
2165 a pointer's type tag is a superset of the alias set of the
2168 2005-01-17 James E Wilson <wilson@specifixinc.com>
2171 * config/ia64/ia64.md (movxf): Handle general register source. Adjust
2172 comment to document why.
2174 2005-01-17 Richard Henderson <rth@redhat.com>
2176 * config/i386/sse.md (smaxv4sf3_finite, sse_vmsmaxv4sf3_finite,
2177 sminv4sf3_finite, sse_vmsminv4sf3_finite, smaxv2df3_finite,
2178 sse2_vmsmaxv2df3_finite, sminv2df3_finite,
2179 sse2_vmsminv2df3_finite): New.
2180 (smaxv4sf3, sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, smaxv2df3,
2181 sse2_vmsmaxv2df3, sminv2df3, sse2_vmsminv2df3): Remove commutative.
2182 Force op1 into register.
2184 2005-01-17 Kelley Cook <kcook@gcc.gnu.org>
2186 * Makefile.in (STAGEMOVESTUFF): Stage all the stamp files to prevent
2187 bubblestrap from always rebuilding build directory.
2189 2005-01-17 Ulrich Weigand <uweigand@de.ibm.com>
2191 * config/s390/s390.md ("*clc"): Remove incorrect '='.
2192 ("*cmpmem_short"): Likewise.
2194 2005-01-17 Steven Bosscher <stevenb@suse.de>
2196 * basic-block.h: Document BB_* flags.
2197 * regrename.c (copyprop_hardreg_forward): Don't use BB_VISITED,
2198 use an sbitmap instead.
2199 * sched-rgn.c (compute_trg_info): Likewise.
2201 2005-01-17 Richard Sandiford <rsandifo@redhat.com>
2203 * config.gcc (mips64*-*-linux*): Set the default abi to n32. Remove
2204 redundant target_cpu_default and extra_parts lines. Use the default
2206 * config/mips/linux64.h (DRIVER_DEFAULT_ABI_SELF_SPEC): Delete.
2207 (SUBTARGET_EXTRA_SPECS): Delete.
2208 (DRIVER_SELF_SPECS): Remove architecture lines. Set the ABI directly.
2209 (SUBTARGET_ASM_SPEC): Remove -mabi=n32 and -mabi=64 mappings.
2210 * config/mips/t-linux64 (MULTILIB_DIRNAMES): Use "32" for -mabi=32
2211 and "n32" for -mabi=n32.
2213 2005-01-17 Ian Lance Taylor <ian@airs.com>
2216 * c-typeck.c (build_c_cast): Revert patch of 2003-02-16: don't
2217 fold constant variables into initial values.
2219 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
2221 * tree-cfg.c (tree_can_merge_blocks_p): Reorder two checks.
2223 * tree-cfg.c (tree_forwarder_block_p): Speed up by walking
2224 through the statements backward.
2226 2005-01-17 Ian Lance Taylor <ian@airs.com>
2228 PR middle-end/13127:
2229 * tree-inline.c (expand_call_inline): Set TREE_NO_WARNING on
2230 a variable set to the return value of the inlined function.
2232 2005-01-17 Mark Dettinger <dettinge@de.ibm.com>
2234 * rtlanal.c (rtx_cost): Assign cost of 0 to a SUBREG
2235 when modes are tieable.
2237 2005-01-17 Ranjit Mathew <rmathew@hotmail.com>
2239 * gthr-posix.h (__gthread_active_p): Use pthread_cancel instead
2240 of pthread_create to find out if threads are enabled.
2241 * gthr-posix95.h (__gthread_active_p): Likewise.
2243 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
2245 * common.opt (-fnew-ra): Remove.
2247 * toplev.h (flag_new_regalloc): Remove.
2248 * Makefile.in (ra*.*): Don't mention.
2249 * passes.c (rest_of_handle_new_regalloc): Remove.
2250 (rest_of_handle_combine, rest_of_compilation): Always consider
2251 flag_new_regalloc as false.
2252 * doc/invoke.texi: Don't document -fnew-ra.
2254 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
2256 * bb-reorder.c (fix_edges_for_rarely_executed_code): Remove
2257 last parameter to reg_scan.
2258 * loop.c (loop_optimize): Likewise.
2259 * passes.c (rest_of_handle_tracer, rest_of_handle_if_conversion,
2260 rest_of_handle_web, rest_of_handle_cfg, rest_of_handle_jump_bypass,
2261 rest_of_handle_life, rest_of_handle_cse, rest_of_handle_cse2,
2262 rest_of_handle_gcse, rest_of_handle_loop_optimize,
2263 rest_of_handle_loop2, rest_of_handle_jump2): Likewise.
2264 * regclass.c (reg_scan): Likewise, for the declaration.
2265 * rtl.h (reg_scan): Likewise, for the prototype.
2267 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
2269 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Don't call
2270 mark_dfs_back_edges.
2272 2005-01-16 Daniel Berlin <dberlin@dberlin.org>
2274 * tree-ssa-pre.c (add_to_sets): s1 may be NULL.
2275 (compute_avail): Uses don't go in tmp_gen.
2277 2005-01-16 Steven Bosscher <stevenb@suse.de>
2279 * ggc-page.c (ggc_alloc_stat): Use __builtin_ctzl instead of a
2280 loop to look for a free slot in a page entry.
2282 2005-01-16 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2285 * defaults.h (TARGET_DEFERRED_OUTPUT_DEFS): Provide default.
2286 * toplev.c (compile_file): Call process_pending_assemble_output_defs
2287 just before targetm.asm_out.file_end.
2288 * tree.h (process_pending_assemble_output_defs): Declare.
2289 * varasm.c (assemble_output_def, process_pending_assemble_output_defs):
2291 (assemble_alias): Defer generation of assembly code for defines when
2292 TARGET_DEFERRED_OUTPUT_DEFS is true.
2293 * config/rs6000/aix41.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
2294 * config/rs6000/aix43.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
2295 * doc/tm.texi (TARGET_DEFERRED_OUTPUT_DEFS): document.
2297 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2300 * pa.c (pa_scalar_mode_supported_p): New function.
2301 (TARGET_SCALAR_MODE_SUPPORTED_P): Define.
2302 * pa.h (MIN_UNITS_PER_WORD): Add comment.
2304 * pa-protos.h (prefetch_operand): Delete.
2305 (prefetch_cc_operand, prefetch_nocc_operand): New declations.
2306 * pa.c (prefetch_operand): Delete.
2307 (prefetch_cc_operand, prefetch_nocc_operand): New functions.
2308 * pa.h (EXTRA_CONSTRAINT): Add `W' constraint.
2309 (PREDICATE_CODES): Delete prefetch_operand. Add prefetch_cc_operand
2310 and prefetch_nocc_operand.
2311 * pa.md (prefetch): Rework to avoid reload problems handling short
2312 displacements when a cache control completer needs to be provided.
2313 (prefetch_32, prefetch_64): Delete.
2314 (prefetch_cc, prefetch_nocc): New patterns.
2316 2005-01-15 David Edelsohn <edelsohn@gnu.org>
2318 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Revert previous change.
2320 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2323 * pa.c (pa_asm_output_aligned_common): Warn if specified alignment
2324 exceeds maximum alignment for global common data.
2325 * pa64-hpux.h (MAX_OFILE_ALIGNMENT): Define to 32768.
2326 * som.h (MAX_OFILE_ALIGNMENT): Likewise.
2328 2004-01-15 Roger Sayle <roger@eyesopen.com>
2330 * tree-ssa-dom.c (extract_range_from_cond): Correct condition.
2332 2004-01-15 Roger Sayle <roger@eyesopen.com>
2334 * harg-reg-set.h (reg_class_names): Prototype global array.
2335 * regclass.c (reg_class_names): Declare here and initialize to
2337 (dump_regclass): Remove local declaration of reg_class_names.
2338 (regclass): Likewise.
2339 * cfg.c (dump_flow_info): Likewise.
2340 * ra-debug.c (reg_class_names): Likewise.
2341 * regrename.c (reg_class_names): Likewise.
2342 * reload.c (reg_class_names): Likewise.
2343 * reload1.c (spill_failure): Likewise.
2344 * config/m68hc11/m68hc11.c (reg_class_names): Likewise.
2346 2005-01-15 Ulrich Weigand <uweigand@de.ibm.com>
2348 * config/s390/s390.md ("reload_outti"): Remove predicate for
2349 output operand. Abort if operand is not a MEM.
2350 ("reload_outdi", "reload_outdf"): Likewise.
2352 2005-01-15 Marc Espie <espie@openbsd.org>
2354 * config.gcc (*-*-openbsd*): Set HAS_LIBC_R for OpenBSD <= 3.2.
2355 (i[34567]86-*-openbsd*): Switch to ELF for OpenBSD >= 3.4.
2356 * config/openbsd.h: Fix C++ includes for native configurations.
2357 Add proper OS_CPP_BUILTINS. Fix libspec for recent OpenBSD.
2358 Add trampoline support.
2359 * config/i386/openbsdelf.h: New.
2361 2005-01-15 Marc Espie <espie@openbsd.org>
2363 * collect2.c (main): Explicitly parse -dynamic-linker option.
2365 2005-01-15 Jakub Jelinek <jakub@redhat.com>
2367 PR tree-optimization/19060
2368 * tree-ssa-dom.c (extract_range_from_cond) <case LT_EXPR, GT_EXPR>:
2369 Return 0 if op1 <= TYPE_MIN_VALUE () resp. op1 >= TYPE_MAX_VALUE ().
2370 (simplify_cond_and_lookup_avail_expr): Add assert for dummy == 0
2371 and handle extract_range_from_cond returning false.
2372 * fold-const.c (fold): Optimize comparisons with min/max even for
2373 width > HOST_BITS_PER_WIDE_INT.
2375 2005-01-15 Ralf Corsepius <ralf.corsepius@rtems.org>
2377 * config/mips/rtems.h (MIPS_DEFAULT_GVALUE): Set to 0.
2378 * config/mips/t-rtems (MULTILIBS_DIRNAMES,MULTILIB_OPTIONS):
2379 Remove little endian multilib variants.
2380 Add mips32 multilib variant.
2382 2005-01-15 Kazu Hirata <kazu@cs.umass.edu>
2384 * explow.c (copy_all_reg, stabilize): Remove.
2385 * expr.h: Remove the corresponding prototypes.
2387 * rtlanal.c (regs_set_between_p): Remove.
2388 * rtl.h: Remove the corresponding prototype.
2390 2005-01-14 David Edelsohn <edelsohn@gnu.org>
2392 * config/rs6000/rs6000.c (rs6k_nonimmediate_operand): Rename
2393 to rs6000_nonimmediate_operand.
2394 * config/rs6000/rs6000.h (PREDICATE_CODES): Rename
2395 rs6k_nonimmediate_operand to rs6000_nonimmediate_operand.
2396 * config/rs6000/rs6000.md (movsi_internal1): Rename predicate to
2397 rs6000_nonimmediate_operand.
2398 * config/rs6000/spe.md (movdf_e500_double): Rename predicate to
2399 rs6000_nonimmediate_operand.
2401 2005-01-14 Andrew Pinski <pinskia@physics.uc.edu>
2403 * tree-ssa-alias.c (pass_may_alias): Add TODO_verify_stmts.
2405 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
2407 * config/rs6000/rs6000.h (CLASS_MAX_NREGS): DF goes in 1 register
2409 (CANNOT_CHANGE_MODE_CLASS): Restrict DI mode changes on e500v2.
2410 (PREDICATE_CODES): Add rs6k_nonimmediate_operand.
2412 * config/rs6000/rs6000.c (invalid_e500_subreg): New.
2413 (rs6k_nonimmediate_operand): New.
2414 (rs6000_legitimate_offset_address_p): Handle DI modes on e500v2
2416 (legitimate_lo_sum_address_p): Same.
2417 (rs6000_legitimize_address): Same.
2418 (rs6000_legitimize_reload_address): Same.
2419 (rs6000_legitimate_address): Same.
2420 (spe_build_register_parallel): Pass DF and DC modes in a DI
2423 * config/rs6000/rs6000.md ("*movsi_internal1"): Change predicate
2424 to rs6k_nonimmediate_operand.
2426 * config/rs6000/spe.md ("*frob_df_di"): New.
2427 ("*frob_di_df"): New.
2428 ("*frob_di_df_2"): New.
2429 ("*mov_sidf_e500_subreg0"): New.
2430 ("*mov_sidf_e500_subreg4"): New.
2431 ("*movdf_e500_double"): Change predicate to
2432 rs6k_nonimmediate_operand.
2434 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
2436 * postreload.c (move2add_note_store): Only call
2437 trunc_int_for_mode on scalar integers.
2439 2005-01-14 Steven Bosscher <stevenb@suse.de>
2441 * tree-ssa-dce.c (visited_control_parents): New sbitmap to
2442 replace BB_VISITED uses.
2443 (find_obviously_necessary_stmts): Don't clear BB_VISITED.
2444 (propagate_necessity): Check the bitmap instead of BB_VISITED.
2445 (tree_dce_done): Free visited_control_parents.
2446 (perform_tree_ssa_dce): Allocate and clear it.
2447 * tree-ssa-pre.c (compute_antic_aux): Make non-recursive.
2448 (compute_antic): Iterate from here using a DFS. Use an sbitmap
2449 instead of BB_VISITED.
2451 2005-01-14 Kazu Hirata <kazu@cs.umass.edu>
2453 * c-tree.h, coverage.h, langhooks-def.h, optabs.h, output.h,
2454 rtl.h, targhooks.h, tree-flow.h, tree-mudflap.h, tree.h:
2455 Remove unused prototypes.
2457 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
2459 * config/rs6000/rs6000.c (rs6000_stack_info): Fix caching of SPE
2460 64-bit register usage.
2461 (spe_func_has_64bit_regs_p): Add FIXME note.
2463 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
2465 * config/rs6000/rs6000.c (rs6000_generate_compare): Replace
2466 flag_finite_math_only with flag_unsafe_math_optimizations.
2468 2005-01-14 J. D. Johnston <jjohnst@us.ibm.com>
2470 * config/s390/tpf-unwind.h (s390_fallback_frame_state): Correct
2472 (__tpf_eh_return): Copy TPF private stack area from the module
2473 boundary stack frame. Add check for when module addresses are
2474 equal, but no stub address is found.
2476 2005-01-14 Richard Earnshaw <rearnsha@arm.com>
2479 * arm.h (struct machine_function): Add call_via field.
2480 (thumb_call_via_label): Declare.
2481 * arm.c (thumb_call_via_label): New variable.
2482 (thumb_call_reg_needed): New variable.
2483 (arm_output_function_epilogue): For Thumb code, output any per-function
2484 call-indirect trampolines.
2485 (thumb_call_via_reg): New function.
2486 (arm_file_end): New function.
2487 (TARGET_ASM_FILE_END): Call arm_file_end.
2488 (aof_file_end): Likewise.
2489 * arm-protos.h (thumb_call_via_reg): Declare.
2490 * arm.md (call_reg_thumb, call_value_reg_thumb): Call
2491 thumb_call_via_reg in normal case.
2493 2005-01-14 Jakub Jelinek <jakub@redhat.com>
2496 PR rtl-optimization/19348
2497 * recog.c (peephole2_optimize): Do global life update if some peephole
2498 decides it doesn't need at least one of its inputs and that change
2499 influences liveness at the start of the basic block.
2501 * basic-block.h (EXECUTE_IF_AND_COMPL_IN_REG_SET): Needs 2 REGSET
2502 arguments instead of 1.
2504 2005-01-14 Eric Botcazou <ebotcazou@libertysurf.fr>
2507 * varasm.c (initializer_constant_valid_p) <ADDR_EXPR>: Return
2508 zero for nested functions needing a static chain or functions
2509 with a non-constant address.
2511 2005-01-13 Roger Sayle <roger@eyesopen.com>
2513 * simplify-rtx.c (simplify_binary_operation) <AND>: Optimize
2514 (and (sign_extend X) C) into (zero_extend (and X C)).
2516 2005-01-13 David O'Brien <obrien@FreeBSD.org>
2518 * config/freebsd-spec.h: Make KSE pthread lib logic the default.
2520 2005-01-13 Richard Henderson <rth@redhat.com>
2525 * config/i386/i386.md (cmpdf, cmpsf, bunordered, bordered, buneq,
2526 bunge, bungt, bunle, bunlt, bltgt): Enable for TARGET_SSE_MATH,
2527 not just TARGET_SSE.
2528 (cmpfp_i_387): Rename from cmpfp_i. Move after sse patterns.
2529 (cmpfp_i_mixed): Rename from cmpfp_i_sse; use for TARGET_MIX_SSE_I387.
2530 (cmpfp_i_sse): Rename from cmpfp_i_sse_only; use for TARGET_SSE_MATH.
2531 (cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387): Similarly.
2532 (fp_jcc_1_mixed, fp_jcc_1_sse, fp_jcc_1_387): Similarly.
2533 (fp_jcc_2_mixed, fp_jcc_2_sse, fp_jcc_2_387): Similarly.
2534 (fp_jcc_3_387, fp_jcc_4_387, fp_jcc_5_387, fp_jcc_6_387,
2535 fp_jcc_7_387, fp_jcc_8_387): Rename from fp_jcc_N.
2536 (movdicc_c_rex64): Rename with '*'.
2537 (movsfcc, movdfcc): Add checks for 387 and sse math to condition.
2538 (movsfcc_1_sse_min, movsfcc_1_sse_max, movsfcc_1_sse): New.
2539 (movsfcc_1_387): Rename from movsfcc_1.
2540 (movdfcc_1_sse_min, movdfcc_1_sse_max, movdfcc_1_sse): New.
2541 (movdfcc_1, movdfcc_1_rex64): Add check for 387.
2542 (sminsf3, smaxsf3, smindf3, smaxdf3): New.
2543 (minsf3, minsf, minsf_nonieee, minsf_sse, mindf3, mindf,
2544 mindf_nonieee, mindf_sse, maxsf3, maxsf, maxsf_nonieee, maxsf_sse,
2545 maxdf3, maxdf, maxdf_nonieee, maxdf_sse, sse_movsfcc, sse_movsfcc_eq,
2546 sse_movdfcc, sse_movdfcc_eq, sse_movsfcc_const0_1,
2547 sse_movsfcc_const0_2, sse_movsfcc_const0_3, sse_movsfcc_const0_4,
2548 sse_movdfcc_const0_1, sse_movdfcc_const0_2, sse_movdfcc_const0_3,
2549 sse_movdfcc_const0_4): Remove.
2550 * config/i386/i386.c (ix86_expand_fp_movcc): For TARGET_SSE_MATH,
2551 recognize min/max early. Update for changed sse cmove patterns.
2552 (ix86_split_sse_movcc): New.
2553 * config/i386/i386-protos.h: Update.
2555 2005-01-13 Steven Bosscher <stevenb@suse.de>
2557 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
2558 (fix_stmt_v_may_defs): Likewise.
2560 2005-01-13 Richard Henderson <rth@redhat.com>
2562 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
2563 for converting the mask.
2565 2005-01-13 David Edelsohn <edelsohn@gnu.org>
2567 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
2570 2005-01-13 Richard Henderson <rth@redhat.com>
2572 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
2573 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
2574 IX86_BUILTIN_MOVDQ2Q): Remove.
2575 (IX86_BUILTIN_VEC_EXT_V4SI): New.
2576 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
2577 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
2579 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
2580 (sse2_pextrw): Fix immediate constraint.
2581 (sse2_loadq, sse2_loadq_rex64): Remove.
2582 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
2583 Use __builtin_ia32_vec_ext_<size>.
2584 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
2586 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
2588 * function.c (assign_parm_setup_block): Look inside original
2589 entry_parm when inspecting PARALLEL.
2591 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
2592 Joel Sherrill <joel@oarcorp.com>
2595 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
2596 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
2597 rtems_gxx_recursive_mutex_init.
2598 (__gthread_recursive_mutex_lock): New function.
2599 (__gthread_recursive_mutex_trylock): Likewise.
2600 (__gthread_recursive_mutex_unlock): Likewise.
2602 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
2604 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
2606 2005-01-13 Jan Beulich <jbeulich@novell.com>
2608 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
2609 (ashrti3_internal): Indicate output is early clobber. Generate result
2610 into output rather than first input. Use move for low word of output
2611 if shift count is exactly 64.
2612 (lshrti3_internal): Likewise.
2614 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
2618 * reload1.c (delete_output_reload): Don't delete an output reload
2619 if the pseudo lives longer than a single basic block. Adjust and
2620 improve wording of comment.
2622 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
2625 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
2626 of TYPE_SIZE where we mean the number of bytes.
2628 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
2630 * function.c (assign_parm_setup_block): Relax condition on
2631 multi-register optimization.
2633 2005-01-12 Nick Clifton <nickc@redhat.com>
2635 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
2636 an "in_delay_slot" attribute of "no" to prevent them being used in
2637 delay slots. This is forbidden because they might generate
2640 2005-01-12 Alan Modra <amodra@bigpond.net.au>
2643 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
2646 2005-01-12 Nick Clifton <nickc@redhat.com>
2648 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
2649 passed on to GAS) but do not define, as GAS no longer supports or
2650 needs the -m2000 option.
2652 2005-01-11 Roger Sayle <roger@eyesopen.com>
2654 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
2655 type as args of unprototyped builtins aren't automatically promoted.
2656 (fold_builtin_abs): Likewise.
2658 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
2660 * Makefile.in: Set a `build-warn' variable.
2662 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
2664 PR tree-optimization/17949
2665 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
2666 (find_interesting_uses_address): Use it.
2668 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
2670 * regrename.c (kill_value): Handle subreg's that won't simplify.
2672 2005-01-11 Richard Henderson <rth@redhat.com>
2675 * config/i386/i386.h (enum ix86_builtins): Move ...
2676 * config/i386/i386.c: ... here.
2677 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
2678 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
2679 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
2680 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
2681 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
2682 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
2683 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
2684 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
2685 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
2686 IX86_BUILTIN_LOADDDUP): Remove.
2687 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
2688 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
2689 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
2690 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
2691 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
2692 (ix86_init_builtins): Make static.
2693 (ix86_init_mmx_sse_builtins): Update for changed builtins.
2694 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
2695 if all the modes match. Otherwise, fake it.
2696 (get_element_number, ix86_expand_vec_init_builtin,
2697 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
2698 (ix86_expand_builtin): Make static. Update for changed builtins.
2699 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
2700 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
2701 (ix86_expand_vector_init_duplicate): New.
2702 (ix86_expand_vector_init_low_nonzero): New.
2703 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
2704 Split out from ix86_expand_vector_init; handle integer modes.
2705 (ix86_expand_vector_init): Use them.
2706 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
2707 * config/i386/i386-protos.h: Update.
2708 * config/i386/predicates.md (reg_or_0_operand): New.
2709 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
2710 (movv2sf_internal): Likewise. And a splitter to match them all.
2711 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
2712 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
2713 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
2714 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
2715 (mmx_pinsrw): Fix operand ordering.
2716 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
2717 rather than sse_loadss expander.
2718 (movv2df splitter): Similarly.
2719 (sse_loadss, sse_loadlss): Remove.
2720 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
2721 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
2722 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
2723 (sse3_movddup): Rename with '*'.
2724 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
2725 (sse2_loadsd): Remove.
2726 (vec_dupv2df_sse3): Rename from sse3_loadddup.
2727 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
2728 (sse2_pinsrw): Fix argument ordering.
2729 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
2730 (sse2_stored): Remove 'r' destination.
2731 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
2732 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
2733 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
2734 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
2735 vec_extractv16qi, vec_initv16qi): New.
2737 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
2738 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
2739 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
2741 (_mm_load_pd, _mm_store_pd): Use plain dereference.
2742 (_mm_load_si128, _mm_store_si128): Likewise.
2743 (_mm_load1_pd): Use _mm_set1_pd.
2744 (_mm_load_sd): Use _mm_set_sd.
2745 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
2746 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
2747 (_mm_set_epi64): Use _mm_set_epi64x.
2748 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
2749 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
2750 _mm_setr_epi8): Use _mm_set_foo form.
2751 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
2752 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
2753 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
2754 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
2755 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
2756 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
2757 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
2758 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
2759 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
2760 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
2761 (_mm_movedup_pd): Use _mm_shuffle_pd.
2762 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
2763 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
2764 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
2765 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
2767 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
2768 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
2769 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
2770 (_mm_store_ps): Use raw dereference.
2771 (_mm_store1_ps): Use _mm_storeu_ps.
2772 (_mm_storer_ps): Use _mm_store_ps.
2773 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
2774 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
2776 2005-01-11 Stan Shebs <shebs@apple.com>
2778 * config/rs6000/rs6000.c (machopic_output_stub): Issue
2779 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
2781 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
2784 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
2785 Skip all DECLs except for FIELD_DECLs.
2787 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
2789 * config/s390/s390.c (override_options): Return error if
2790 -mbackchain, -mpacked-stack and -mhard-float are used together.
2791 (s390_va_start): Remove the backchain && packed-stack special case.
2792 (s390_gimplify_va_arg): Likewise.
2793 * doc/invoke.texi: Remove the ABI incompatibility note.
2795 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
2797 * config/s390/s390.c (struct s390_frame_layout): Remove
2799 (s390_frame_info, s390_emit_prologue): Replace occurrences of
2800 save_backchain_p with TARGET_BACKCHAIN.
2802 2005-01-11 Alan Modra <amodra@bigpond.net.au>
2805 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
2806 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
2807 * function.c (locate_and_pad_parm): Set new field.
2808 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
2809 Tweak where_pad test to include "none". Always set mem align for
2811 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
2812 not sufficient for type.
2813 (assign_parm_setup_block): If stack_parm is zero on entry, always
2814 make a new stack local. Block move old stack parm if necessary
2815 to new aligned stack local.
2816 (assign_parm_setup_stack): Use a block move to handle
2817 potentially misaligned entry_parm.
2818 (assign_parms_unsplit_complex): Specify required alignment when
2819 creating stack local.
2820 * calls.c (compute_argument_addresses): Override alignment of stack
2821 arg calculated from its type with the alignment given by
2822 FUNCTION_ARG_BOUNDARY.
2823 (store_one_arg): Likewise.
2825 2005-01-11 Jan Beulich <jbeulich@novell.com>
2827 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
2828 Change respective itanium_class attribute to ialu.
2829 (shladdp4_internal): New.
2830 * config/ia64/predicates.md (shladd_log2_operand): New.
2832 2005-01-11 Richard Henderson <rth@redhat.com>
2834 * expr.c (store_constructor): Use rtvec_alloc instead of
2835 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
2837 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
2839 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
2841 2005-01-11 Alan Modra <amodra@bigpond.net.au>
2843 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
2844 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
2845 (TARGET_RWRELOC): Define.
2846 (ia64_rwreloc_section_type_flags): Delete.
2847 (ia64_section_type_flags): New function.
2848 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
2849 (TARGET_RWRELOC): Define.
2851 2005-01-10 David Mosberger <davidm@hpl.hp.com>
2854 * config/ia64/ia64.c (process_set): For alloc insn, only call
2855 process_epilogue is !frame_pointer_needed.
2857 2005-01-10 Roger Sayle <roger@eyesopen.com>
2860 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
2861 unary operator and can't be treated as a binary/comparison operator.
2863 2005-01-10 Richard Henderson <rth@redhat.com>
2865 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
2866 (ix86_return_in_memory): Likewise.
2867 (function_arg_advance): Likewise. Mirror structure in function_arg
2868 for choosing register to advance.
2870 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
2872 * tree-vectorizer.c, tree.def: Fix comment typos.
2874 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
2876 * config/s390/s390.c (struct processor_costs): 4 new fields:
2877 ddbr, ddr, debr, der.
2878 (s390_rtx_costs): More precise handling of divide instructions.
2880 2005-01-10 David Edelsohn <edelsohn@gnu.org>
2882 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
2884 2005-01-10 Jan Beulich <jbeulich@novell.com>
2886 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
2887 names resulting from -ffunction-sections/-fdata-sections and linkonce
2889 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
2890 based on the section name. Rearrange the section name comparison logic
2891 slightly so that each section name is compared against at most once.
2893 2005-01-10 Ben Elliston <bje@au.ibm.com>
2895 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
2896 targets for which -fPIC reduces limitations on the GOT size.
2898 2005-01-09 Falk Hueffner <falk@debian.org>
2900 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
2903 2005-01-09 David Edelsohn <edelsohn@gnu.org>
2906 * collect2.c (main): Set aixrtl_flag for -brtl option.
2907 (resolve_lib_name): Search for .so file extension before .a
2910 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
2912 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
2913 (vect_analyze_data_refs_alignment): Add dump prints.
2915 2005-01-09 Ira Rosen <irar@il.ibm.com>
2917 * tree-vectorizer.c (vect_analyze_offset_expr): Use
2918 expr_invariant_in_loop_p.
2919 Initialize outputs first thing in the function.
2920 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
2921 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
2922 (vect_analyze_pointer_ref_access): Check that the initial condition of
2923 the access function is loop invariant.
2925 2005-01-09 Richard Henderson <rth@redhat.com>
2927 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
2928 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
2929 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
2930 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
2931 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
2932 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
2933 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
2934 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
2935 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
2936 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
2937 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
2938 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
2939 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
2940 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
2941 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
2942 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
2943 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
2944 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
2945 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
2946 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
2947 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
2948 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
2949 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
2951 (mmx_clrdi, pavgusb): Remove.
2952 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
2953 with leading sse_ prefix.
2954 * config/i386/sse.md: Receive them.
2955 * config/i386/mmx.md: New file.
2956 (MMXMODE12, MMXMODE24, mmxvecsize): New.
2957 (subrv2sf3): Turn into expander for normal subtraction.
2958 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
2959 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
2960 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
2961 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
2962 commutative; use ix86_binary_operator_ok.
2963 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
2964 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
2965 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
2966 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
2967 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
2968 Macroize from existing patterns; use ix86_binary_operator_ok.
2969 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
2970 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
2971 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
2972 with vec_select+vec_concat.
2973 (mmx_pshufw, mmx_pshufw_1): Likewise.
2974 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
2975 (mmx_uavgv4hi3): Model correctly.
2976 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
2977 _mm_xor_si64): Remove casts.
2979 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
2981 PR tree-optimization/19224
2982 * tree-scalar-evolution.c (get_instantiated_value,
2983 set_instantiated_value): New functions.
2984 (instantiate_parameters_1): Cache the results.
2985 (instantiate_parameters, resolve_mixers): Initialize and free
2988 2005-01-08 David Edelsohn <edelsohn@gnu.org>
2990 * config/i386/i386.md (addhi_4): Correct reference in comment.
2993 2005-01-08 Richard Henderson <rth@redhat.com>
2995 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
2996 (_mm_cvtsi128_si64x): Likewise.
2997 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
2998 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
2999 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
3000 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
3001 * config/i386/i386-protos.h: Update.
3002 * config/i386/i386.c (print_operand): Add 'H'.
3003 (ix86_fixup_binary_operands): Split out from ...
3004 (ix86_expand_binary_operator): ... here.
3005 (ix86_fixup_binary_operands_no_copy): New.
3006 (ix86_expand_fp_absneg_operator): Handle vector mode results.
3007 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
3008 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
3009 (safe_vector_operand): Use CONST0_RTX.
3010 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
3011 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
3012 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
3013 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
3015 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
3016 (IX86_BUILTIN_CMPNGESS): New.
3017 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
3018 (attr type): Add sselog1.
3019 (attr unit, attr memory): Handle it.
3020 (movti, movti_internal, movti_rex64): Move near other integer moves.
3021 (movtf, movtf_internal): Move near other fp moves.
3022 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
3023 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
3024 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
3025 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
3026 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
3027 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
3028 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
3029 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
3030 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
3031 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
3032 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
3033 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
3034 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
3035 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
3036 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
3037 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
3038 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
3039 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
3040 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
3041 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
3042 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
3043 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
3044 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
3045 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
3046 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
3047 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
3048 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
3049 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
3050 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
3051 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
3052 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
3053 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
3054 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
3055 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
3056 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
3057 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
3058 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
3059 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
3060 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
3061 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
3062 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
3063 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
3064 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
3065 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
3067 (SSEPUSH, push<SSEPUSH>): Remove.
3068 (MMXPUSH, push<MMXPUSH>): Remove.
3069 (sse_movaps, sse_movaps_1, sse_movups): Remove.
3070 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
3071 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
3072 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
3073 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
3074 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
3075 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
3076 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
3077 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
3078 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
3079 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
3080 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
3081 (ppro_sse_log_V4SF): Handle sselog1.
3082 * config/i386/predicates.md (const_0_to_1_operand): New.
3083 (const_0_to_255_mul_8_operand): New.
3084 (const_1_to_31_operand): Rename from const_int_1_31_operand.
3085 (const_2_to_3_operand, const_4_to_7_operand): New.
3086 * config/i386/sse.md: New file.
3087 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
3088 (sse_movups): Rename from sse_movups_1.
3089 (sse_loadlss): Rename from sse_loadss_1.
3090 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
3092 (negv4sf2): Use ix86_expand_fp_absneg_operator.
3093 (absv4sf2, negv2df, absv2df): New.
3094 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
3095 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
3096 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
3097 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
3098 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
3099 (sse3_addsubv4sf3): Model correctly.
3100 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
3101 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
3102 (sse_movhlps): Model with vec_select+vec_concat.
3103 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
3104 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
3105 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
3106 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
3107 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
3108 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
3109 sse2_pshufhw_1): Likewise.
3110 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
3111 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
3112 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
3113 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
3114 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
3115 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
3116 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
3117 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
3118 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
3119 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
3120 sminv8hi3): Mark commutative
3121 operands. Use ix86_binary_operator_ok.
3122 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
3123 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
3124 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
3125 sse2_punpcklqdq): Allow operand2 in memory.
3126 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
3127 sse2_movsd): Add memory alternatives.
3128 (sse_storelps): Turn expander into an insn; split after reload.
3129 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
3130 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
3132 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
3134 * configure.ac (DWARF-2 debug_line): Use objdump.
3135 * configure: Regenerate.
3137 2005-01-08 Jeff Law <law@redhat.com>
3138 Diego Novillo <dnovillo@redhat.com>
3140 PR tree-optimization/18241
3141 * tree-nrv.c (tree_nrv): Ignore volatile return values.
3142 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
3143 statements with volatile operands.
3144 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
3145 operands after marking a statement with has_volatile_ops.
3147 2005-01-08 Roger Sayle <roger@eyesopen.com>
3149 * tree.c (int_fits_type_p): Always honor integer constant
3150 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
3152 2005-01-08 Roger Sayle <roger@eyesopen.com>
3154 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
3155 the latent bug in rtl_delete_block has been resolved.
3157 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
3159 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
3160 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
3162 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
3164 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
3165 library as @multilib_dir@/$(SHLIB_SONAME).
3167 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
3169 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
3170 * configure: Regenerate.
3172 2005-01-07 Jakub Jelinek <jakub@redhat.com>
3174 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
3177 2005-01-07 David Edelsohn <edelsohn@gnu.org>
3180 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
3181 Convert non-word aligned offset address using ld/std into
3184 2005-01-07 Richard Henderson <rth@redhat.com>
3186 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
3188 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
3190 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
3192 2005-01-07 Jakub Jelinek <jakub@redhat.com>
3194 PR tree-optimization/19283
3195 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
3196 is not shorter than the original type.
3198 PR rtl-optimization/19012
3199 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
3201 PR rtl-optimization/18861
3202 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
3203 (FULL_STATE): Define.
3204 (mark_tablejump_edge): New function.
3205 (purge_dead_tablejump_edges): New function.
3206 (find_bb_boundaries): Use it.
3208 PR tree-optimization/18828
3209 * builtins.c (expand_builtin_next_arg): Remove argument and all
3210 the argument checking.
3211 (expand_builtin): Adjust caller.
3212 (expand_builtin_va_start): Likewise. Remove error for too many
3214 (fold_builtin_next_arg): Issue error for too many arguments.
3215 After checking arguments, replace them with magic arguments that
3216 prevent further checking of the args.
3218 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
3220 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
3222 2005-01-06 Roger Sayle <roger@eyesopen.com>
3225 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
3226 and %{.S:...} (and their negative variants) to test whether the
3227 input file is assembler or pre-processed-assembler independent of
3228 the actual filename extension.
3230 2005-01-06 Roger Sayle <roger@eyesopen.com>
3232 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
3233 of sign or zero extended values.
3235 2005-01-06 Geoffrey Keating <geoffk@apple.com>
3237 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
3238 for IBM long double format correctly.
3240 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
3242 Fix PR tree-optimization/18792
3244 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
3245 to first_loop_depth, and use loop depth instead of loop number.
3246 (build_classic_dir_vector): Ditto.
3247 (compute_data_dependences_for_loop): Use depth, not loop number.
3248 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
3249 number. Pass in loops, instead of loop numbers.
3250 (gather_interchange_stats): Ditto.
3251 (linear_transform_loops): Ditto.
3253 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
3256 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
3257 biv_fits_mode_p, extension_within_bounds_p): New functions.
3258 (check_ext_dependent_givs): Use them.
3260 2005-01-06 Roger Sayle <roger@eyesopen.com>
3262 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
3263 more than one barrier, in which case we should delete them all.
3265 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
3267 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
3269 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
3271 Fix PR middle-end/19286
3273 * dwarf2out.c (gen_subprogram_die): If we've already tried to
3274 output this subprogram, simply ignore this attempt to do it again.
3275 (add_abstract_origin_attribute): Don't abort trying to add the abstract
3276 origin attribute if it's not possible.
3277 (gen_block_die): Don't ignore subblocks of "unused" blocks.
3278 (decls_for_scope): Ditto.
3279 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
3280 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
3282 2005-01-05 Richard Henderson <rth@redhat.com>
3285 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
3286 (bdesc_2arg): Use it.
3287 (ix86_expand_binop_builtin): Force operands into registers
3289 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
3290 ix86_expand_sse_compare, ix86_expand_sse_comi,
3291 ix86_expand_builtin): Likewise.
3293 2005-01-05 Richard Henderson <rth@redhat.com>
3295 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
3297 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
3298 (mulv4hi3): Set itanium_class mmmul.
3299 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
3300 (fpack): Rename from fpack_sfsf.
3302 2005-01-05 Richard Henderson <rth@redhat.com>
3305 * reload1.c (do_input_reload): Restrict the optimization deleteing
3306 a previous output reload to RELOAD_FOR_INPUT.
3308 2005-01-05 Steven Bosscher <stevenb@suse.de>
3310 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
3311 comment that we fall through after case.
3312 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
3313 (distribute_links): Likewise.
3314 * cse.c (cse_insn): Likewise.
3315 * cselib.c (cselib_invalidate_mem): Likewise.
3316 * df.c: Update comments at the top of the file.
3317 (read_modify_subreg_p): Update comments here too.
3318 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
3319 * flow.c (mark_set_1): Likewise.
3320 (mark_used_regs): Likewise.
3321 * gcse.c (mems_conflict_for_gcse_p): Likewise.
3322 (canon_list_insert): Likewise.
3323 (mark_set): Likewise.
3324 (try_replace_reg): Likewise.
3325 (store_killed_in_insn): Likewise.
3326 * loop.c (count_one_set): Likewise.
3327 (basic_induction_var): Likewise.
3328 * postreload-gcse.c (find_mem_conflicts): Likewise.
3329 * postreload.c (reload_combine_note_store): Likewise.
3330 (move2add_note_store): Likewise.
3331 * reload.c (find_equiv_reg): Likewise.
3332 (mark_referenced_resources): Likewise.
3333 * rtlanal.c (set_noop_p): Likewise.
3334 (note_stores): Likewise.
3335 (note_uses): Likewise.
3336 * sched-deps.c (sched_analyze_1): Likewise.
3337 * sched-rgn.c (check_live_1): Likewise.
3338 (update_live_1): Likewise.
3339 * config/i860/i860.c: Likewise.
3341 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
3343 (ZERO_EXTRACT): Mention that this one can be an lvalue.
3345 * doc/rtl.texi: Update documentation for bit-fields and SET.
3347 2005-01-05 Roger Sayle <roger@eyesopen.com>
3349 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
3351 2005-01-05 Jan Hubicka <jh@suse.cz>
3352 Richard Henderson <rth@redhat.com>
3355 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
3358 2005-01-05 Richard Henderson <rth@redhat.com>
3361 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
3362 (sse_shufps): Change operand 3 to const_int_operand.
3363 (sse2_storelps): Fix typo in template.
3364 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
3365 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
3366 (ix86_expand_builtin): Likewise.
3368 2005-01-05 Julian Brown <julian@codesourcery.com>
3370 * config/arm/arm.c (arm_return_in_memory): Treat complex types
3371 as aggregates for AAPCS ABIs.
3373 2005-01-05 Stan Shebs <shebs@apple.com>
3375 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
3376 Mach-O getters if ppc64.
3377 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
3379 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
3380 signed four-byte field for global code case.
3381 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
3383 2005-01-05 Roger Sayle <roger@eyesopen.com>
3386 * c-common.c: Include real.h.
3387 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
3388 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
3389 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
3390 * Makefile.in (c-common.o): Update dependencies.
3392 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
3394 * c-parse.in (asm_string): Add trailing semicolon.
3396 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
3398 * c-parse.in (asm_string): New. Don't allow wide strings in
3400 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
3401 asm_string instead of STRING.
3403 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
3405 * c-typeck.c (constructor_no_implicit): Remove.
3406 (set_designator, process_init_element): Don't check
3407 constructor_no_implicit.
3409 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
3410 Kaz Kojima <kkojima@gcc.gnu.org>
3413 * lcm.c (create_pre_exit): New.
3414 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
3415 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
3417 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
3419 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
3420 -msoft-float in target_float_switch.
3421 * arm.c (arm_override_options): Fix processing of target_float_switch.
3423 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
3425 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
3428 2004-01-05 Uros Bizjak <uros@kss-loka.si>
3430 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
3431 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
3433 2005-01-04 Roger Sayle <roger@eyesopen.com>
3435 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
3436 if deleting the then-block allows the test-block to fallthru to the
3439 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
3442 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
3443 after the full declaration if the are in two different TUs.
3445 2005-01-04 Richard Henderson <rth@redhat.com>
3448 * tree-sra.c (generate_one_element_init): Just
3449 call gimplify_and_add.
3450 (generate_element_init): Record the
3451 new referenced variables and mark them for renaming
3452 and split out to ...
3453 (generate_element_init_1): This.
3454 (scalarize_init): Don't call push_gimplify_context/
3455 pop_gimplify_context.
3457 2005-01-04 Geoffrey Keating <geoffk@apple.com>
3459 * toplev.c (get_src_pwd): Handle failure of getpwd().
3461 2005-01-04 Roger Sayle <roger@eyesopen.com>
3463 * fold-const.c (fold_single_bit_test): Delete unreachable handling
3465 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
3466 all the cases handled by it are inverted by invert_truthvalue.
3468 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
3470 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
3471 (save_register_windowsi): Likewise.
3473 2005-01-04 Richard Henderson <rth@redhat.com>
3475 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
3477 2005-01-04 Uros Bizjak <uros@kss-loka.si>
3480 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
3481 any simplifications of vector mode comparison operators.
3482 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
3484 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
3485 Devang Patel <dpatel@apple.com>
3487 PR tree-optimization/18308
3488 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
3489 the operands before creating a new expression.
3490 * dojump.c (do_jump): Make drop_through_label available
3491 for all cases. Add expansion of COND_EXPR.
3493 2005-01-04 Ira Rosen <irar@il.ibm.com>
3495 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
3496 INTEGER_CST instead of TREE_CONSTANT.
3497 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
3498 instead of TREE_CONSTANT.
3499 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
3500 instead of TREE_CONSTANT.
3502 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
3504 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
3506 2005-01-04 Uros Bizjak <uros@kss-loka.si>
3509 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
3510 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
3512 2005-01-03 Richard Henderson <rth@redhat.com>
3514 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
3515 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
3516 fold_convert_const_real_from_real): Split out from ...
3517 (fold_convert_const): ... here.
3519 2005-01-03 Richard Henderson <rth@redhat.com>
3522 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
3523 (mov<MMXMODEI>_internal): Likewise.
3524 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
3525 Add V2SF case; use it for SSE1; don't use TI.
3526 (movdf_integer): Likewise.
3527 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
3529 2005-01-03 Ira Rosen <irar@il.ibm.com>
3531 * tree-vectorizer.c (vect_strip_conversions): New function.
3532 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
3533 check for binary class.
3535 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
3539 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
3541 * gimple-low.c (mark_blocks_with_used_vars): New function.
3542 (mark_blocks_with_used_subblocks): Ditto.
3543 (mark_used_blocks): Ditto.
3544 (pass_mark_used_blocks): New pass.
3545 * tree-inline.c: Include debug.h.
3546 (expand_call_inline): Call outlining_inline_function here.
3547 * tree-optimize.c (init_tree_optimization_passes): Add
3548 pass_mark_used_blocks.
3549 * tree-pass.h (pass_mark_used_blocks): New.
3550 * Makefile.in (tree-inline.o): Add debug.h dependency.
3552 2005-01-03 Geoffrey Keating <geoffk@apple.com>
3554 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
3555 VAR_DECLs to have weak_import attribute.
3557 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
3560 2004-01-03 Steven Bosscher <stevenb@suse.de>
3562 * passes.c (rest_of_compilation): Don't run regmove if only
3563 flag_expensive_optimizations, require flag_regmove instead.
3565 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
3567 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
3569 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
3570 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
3571 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
3572 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
3574 2005-01-03 Richard Henderson <rth@redhat.com>
3576 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
3577 (ia64_const_ok_for_letter_p): New.
3578 (ia64_const_double_ok_for_letter_p): New.
3579 (ia64_extra_constraint): New.
3580 (ia64_expand_vecint_compare): New.
3581 (ia64_expand_vcondu_v2si): New.
3582 (ia64_expand_vecint_cmov): New.
3583 (ia64_expand_vecint_minmax): New.
3584 (ia64_print_operand): Add 'v'.
3585 (ia64_preferred_reload_class): New.
3586 (ia64_vector_mode_supported_p): New.
3587 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
3588 (PREFERRED_RELOAD_CLASS): Move to function.
3589 (CONST_OK_FOR_LETTER_P): Move to function.
3590 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
3591 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
3592 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
3593 (EXTRA_CONSTRAINT): Move to function.
3594 * config/ia64/ia64.md: Include vect.md.
3595 (itanium_class): Add mmalua.
3597 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
3599 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
3600 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
3602 (const_int_2bit_operand): New.
3603 (fr_reg_or_0_operand): New.
3604 * config/ia64/ia64-modes.def: Add vector modes.
3605 * config/ia64/ia64-protos.h: Update.
3606 * config/ia64/vect.md: New file.
3608 2005-01-03 Richard Henderson <rth@redhat.com>
3610 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
3612 2005-01-03 Uros Bizjak <uros@kss-loka.si>
3615 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
3617 (log1pdf2): Change mode of operands[0,1] to DFmode.
3619 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
3621 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
3622 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
3623 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
3624 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
3625 SPARC_RELAXED_ORDERING.
3627 2005-01-03 Richard Henderson <rth@redhat.com>
3628 Uros Bizjak <uros@kss-loka.si>
3631 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
3632 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
3633 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
3635 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
3636 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
3637 const_pow2_1_to_128_operand): New predicates.
3639 2005-01-02 Greg McGary <greg@mcgary.org>
3641 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
3643 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
3644 object. [ptr_type]: Remove unused variable.
3645 Remember nearest addressable array-element or record-component
3646 when checking bitfield components. Tolerate empty BIND_EXPR.
3648 2005-01-01 Richard Henderson <rth@redhat.com>
3650 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
3651 that don't narrow the value. Fail for other conversions.
3653 2005-01-01 Richard Henderson <rth@redhat.com>
3656 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
3657 * c-lang.c (finish_file): Don't do it here.
3658 * objc/objc-act.c (objc_finish_file): Likewise.
3660 * cgraph.c (decl_assembler_name_equal): New.
3661 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
3662 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
3663 * cgraph.h (struct cgraph_varpool_node): Add next.
3664 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
3665 * varasm.c (assemble_alias): Mark the target as needed.
3667 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
3670 * function.c (get_arg_pointer_save_area): Use entry_of_function
3671 instead of get_insns.
3673 2005-01-01 Roger Sayle <roger@eyesopen.com>
3674 Andrew Pinski <pinskia@physics.uc.edu>
3675 James E. Wilson <wilson@specifixinc.com>
3677 PR rtl-optimization/12092
3678 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
3681 2005-01-01 Roger Sayle <roger@eyesopen.com>
3682 Olivier Hainque <hainque@act-europe.fr>
3684 * tree.c (int_fits_type_p): A narrower type always fits in a
3685 wider one, except for negative values into unsigned types.
3687 2005-01-01 Roger Sayle <roger@eyesopen.com>
3689 * tree.c (int_fits_type_p): Compare the result of force_fit_type
3690 with the original constant rather than require TREE_OVERFLOW.
3692 2005-01-01 Steven Bosscher <stevenb@suse.de>
3695 * c-decl.c (finish_function): If compiling C99, annotate the
3696 compiler generated return with the current file name and line 0.
3697 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
3698 the source line is greater than 0.
3699 (remove_bb): Likewise.
3701 See ChangeLog.12 for earlier changes.