1 2005-01-31 Andrew Pinski <pinskia@physics.uc.edu>
3 * config/rs6000/rs6000.md (copysignsf3): New expand.
4 (copysigndf3): Likewise.
6 2005-01-31 Steven Bosscher <stevenb@suse.de>
8 * recog.c (constrain_operands): Only look for earlyclobber operand
9 conflicts if an '&' constraint was seen.
11 2005-01-31 Marc Espie <espie@openbsd.org>
13 * config.gcc: Don't include embedded systems fragment, switches default
14 debugging format to ELF.
15 * config/i386/openbsdelf.h: Add DBX_REGISTER_NUMBER since we no
16 longer pick it up there.
18 2005-01-31 Dale Johannesen <dalej@apple.com>
20 * config/rs6000/darwin.md (movsf_low_di): Make work.
21 (movdf_low_di): Make work.
23 2005-01-31 Dale Johannesen <dalej@apple.com>
25 * config/rs6000/darwin-tramp.asm (__trampoline_setup):
28 2005-01-31 Roger Sayle <roger@eyesopen.com>
29 Dale Johannesen <dalej@apple.com>
32 * fold-const.c (fold_binary_op_with_conditional_arg):
33 Make types match original operands, before STRIP_NOPS.
35 2005-01-31 Alan Modra <amodra@bigpond.net.au>
37 * config/rs6000/linux-unwind.h (struct gcc_vregs): New.
38 (struct gcc_regs): Rename from gcc_pt_regs. Add more regs.
39 (struct gcc_sigcontext): Delete. Merge contents to..
40 (struct gcc_ucontext): ..here.
41 (get_sigcontext): Delete.
42 (get_regs): New function, like get_sigcontext but return regs ptr.
43 64-bit version finds regs from r1 to support vdso.
44 (ppc_linux_aux_vector): New function.
45 (ppc_fallback_frame_state): Modify for get_regs. Restore fprs
48 2005-01-31 Ira Rosen <irar@il.ibm.com>
50 * tree-vectorizer.c (vect_analyze_offset_expr): Use ssizetype instead
52 (vect_get_base_and_offset): Use ssizetype instead sizetype. Remove
53 redundant fold. Fix misalignment for MINUS_EXPR.
54 (vect_compute_data_ref_alignment): Use ssizetype instead sizetype.
55 (vect_analyze_pointer_ref_access): Likewise.
56 (vect_get_memtag_and_dr): Likewise.
58 2005-01-31 Richard Henderson <rth@redhat.com>
60 * config/i386/i386.c (absnegsf2_mixed): Remove all of the # hackery.
61 (absnegsf2_sse, absnegsf2_i387, absnegdf2_mixed, absnegdf2_sse,
62 absnegdf2_i387, absnegxf2_i387): Likewise.
64 2005-01-31 Richard Henderson <rth@redhat.com>
67 * optabs.c (expand_copysign_absneg): New.
68 (expand_copysign_bit): Split out from ...
69 (expand_copysign): ... here. Use expand_copysign_absneg.
71 2005-01-30 Roger Sayle <roger@eyesopen.com>
74 * config/pa/pa.md (anddi3, iordi3): On HPPA64, disallow an integer
75 constant as the second operand and a register as the third.
77 2005-01-31 Danny Smith <dannysmith@users.sourceforge.net>
80 * config/i386/i386.c (ix86_function_ok_for_sibcall): Also check
81 that dllimport'd functions do not use all call-clobbered registers
84 2005-01-30 Richard Henderson <rth@redhat.com>
87 * optabs.c (expand_absneg_bit): Split out from expand_unop and
88 expand_abs_nojump and generalize. Use operand_subword and
89 emit_no_conflict_block. Support large modes.
90 (expand_abs_nojump): Use it. Use HONOR_SIGNED_ZEROS to fall
92 (expand_unop): Likewise. Use HONOR_SIGNED_ZEROS to fall back
93 negation to subtraction.
94 (expand_copysign): Check that the format has signed zeros.
95 Use operand_subword and emit_no_conflict_block. Support large modes.
97 2005-01-30 Daniel Berlin <dberlin@dberlin.org>
99 Fix PR tree-optimization/19624
101 * Makefile.in (tree-ssa-pre.o): Add CFGLOOP_H.
102 * tree-ssa-pre.c: Add cfgloop.h.
104 (pre_stats): New member, constified.
105 (inserted_exprs): New static variable.
106 (NECESSARY): New macro.
107 (create_expression_by_pieces): Fold the expression, and
108 mark it as defaulting to not necessary. Also put in
110 (fully_constant_expression): New function.
111 (insert_into_preds_of_block): Modify to not insert phis when we
112 are playing with induction variables.
113 Push phis onto the inserted_exprs vector, and mark them as not
114 necessary by default.
115 (insert_aux): Call fully_constant_expression on eprime.
116 If all edges produce the same value, mark it constant.
117 (mark_operand_necessary): New function.
118 (remove_dead_inserted_code): New function.
119 (init_pre): Init loop optimizer to get loop info.
120 (fini_pre): Free loop_optimizer, and inserted_exprs vec.
121 (execute_pre): Commit edge inserts, then remove dead code.
123 2005-01-30 Richard Henderson <rth@redhat.com>
125 * rtl.c (rtx_equal_p): No early exit for CONST_VECTOR.
126 * varasm.c (const_rtx_hash_1): Handle CONST_VECTOR.
128 2005-01-30 Richard Henderson <rth@redhat.com>
131 * config/i386/i386.c (ix86_expand_copysign): New.
132 (ix86_split_copysign_const): New.
133 (ix86_split_copysign_var): Rename from ix86_split_copysign,
134 rearrange op1/nmask operands.
135 * config/i386/i386-protos.h: Update.
136 * config/i386/i386.md (copysignsf3): Use ix86_expand_copysign.
137 (copysigndf3): Likewise.
138 (copysignsf3_const, copysigndf3_const): New.
139 (copysignsf3_var): Rename from copysignsf3, split out splitter
140 and fix split predicate for X constraint.
141 (copysigndf3_var): Similarly.
143 2005-01-30 Kazu Hirata <kazu@cs.umass.edu>
145 * optabs.c, doc/c-tree.texi, doc/install.texi, doc/md.texi,
146 doc/passes.texi, doc/rtl.texi, doc/sourcebuild.texi,
147 doc/tm.texi, doc/tree-ssa.texi: Update copyright.
149 2005-01-29 Richard Henderson <rth@redhat.com>
152 * config/i386/i386.md (movdf_nointeger, movdf_integer): Fix ordering
153 of # and * in constraints.
155 2005-01-29 Richard Henderson <rth@redhat.com>
158 * expr.c (store_field): Don't strip sub-mode cast when the input
159 data is even smaller.
161 2005-01-29 Richard Henderson <rth@redhat.com>
164 * expr.c (categorize_ctor_elements_1): Check for CONSTRUCTOR of a
167 2005-01-29 Richard Henderson <rth@redhat.com>
169 * combine.c (make_field_assignment): Fix argument order
172 2005-01-29 Richard Guenther <richard.guenther@uni-tuebingen.de>
174 PR tree-optimization/15791
175 * fold-const.c (extract_array_ref): New function.
176 (fold): Fold comparisons between &a[i] and &a[j] or
177 semantically equivalent trees.
179 2005-01-29 Jeff Law <law@redhat.com>
181 * gcse.c (insert_expr_in_table): Revamp handling of available
182 and anticipatable occurrence lists to avoid unnecessary list
184 (insert_set_in_table): Similarly.
186 2005-01-29 Joseph S. Myers <joseph@codesourcery.com>
188 * c-common.c (fix_string_type): Just use c_build_qualified_type to
190 (c_build_qualified_type): Build qualified array types with
191 TYPE_MAIN_VARIANT pointing to corresponding unqualified type.
192 * c-decl.c (c_make_fname_decl): Build unqualified array type
193 before qualified type.
194 (grokdeclarator): Use TYPE_MAIN_VARIANT of typedef type if element
195 type is qualified, not just if type itself is. Don't apply
196 qualifiers to array type when declarator is processed. Apply
197 qualifiers to field type whether or not it is an array type.
198 Don't handle array types specially for applying qualifiers to
200 * c-typeck.c (composite_type): Build unqualified element type and
201 array type when forming composite of array types.
202 (common_pointer_type, comptypes, comp_target_types,
203 type_lists_compatible_p, build_indirect_ref, build_array_ref,
204 convert_for_assignment): Don't apply TYPE_MAIN_VARIANT to array
206 (type_lists_compatible_p): Cache TREE_VALUE (args1) and TREE_VALUE
207 (args2) in variables a1 and a2.
209 2005-01-29 Kazu Hirata <kazu@cs.umass.edu>
211 * cse.c (n_elements_made, max_elements_made): Remove.
212 (insert): Don't touch n_elements_made.
213 (cse_main): Don't touch n_elements_made or max_elements_made.
215 2005-01-28 Stephane Carrez <stcarrez@nerim.fr>
218 * config/m68hc11/t-m68hc11-gas (dp-bit.c): Fix typo causing a
219 configuration part of dp-bit.c to be lost.
221 2005-01-28 Roger Sayle <roger@eyesopen.com>
223 * expmed.c (expand_mult_highpart_optab): When attempting to use
224 a non-widening multiplication in a wider mode, the operands need
225 to be converted (zero or sign extended) to that mode.
227 2005-01-28 Ian Lance Taylor <ian@airs.com>
231 * gimple-low.c (block_may_fallthru): TRY_FINALLY_EXPR only falls
232 through if both operands fall through.
234 2005-01-28 Kazu Hirata <kazu@cs.umass.edu>
236 * cse.c (fold_rtx) <PC>: Don't optimize.
238 2005-01-28 Jeff Law <law@redhat.com>
240 * fold-const.c (fold, case CEIL_MOD_EXPR): Optimize 0 % X.
241 (case FLOOR_MOD_EXPR, ROUND_MOD_EXPR, TRUNC_MOD_EXPR): Similarly.
243 2005-01-28 Kazu Hirata <kazu@cs.umass.edu>
245 * cse.c (cse_reg_info_free_list, cse_reg_info_used_list,
246 cse_reg_info_used_list_end): Remove.
247 (cse_reg_info_list, cse_reg_info_list_free): New.
248 (get_cse_reg_info): When allocating an instance of
249 cse_reg_info, add it to the beginning of the cse_reg_info_list
250 list. Remove code to maintain cse_reg_info_used_list.
251 (new_basic_block): Reset the free list to the beginning of
254 2005-01-28 Richard Sandiford <rsandifo@redhat.com>
256 * config/frv/frv.c (bdesc_2arg): Remove __MDPACKH.
257 (frv_init_builtins): Change its prototype to take 4 uhalf arguments.
258 (frv_expand_mdpackh_builtin): New function.
259 (frv_expand_builtin): Use it to expand __MDPACKH.
260 * doc/extend.texi (__MDPACKH): Update documentation.
262 2005-01-28 Uros Bizjak <uros@kss-loka.si>
264 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use
265 SSE_FLOAT_MODE_P for use_sse computation.
267 2005-01-28 Ralf Corsepius <ralf.corsepius@rtems.org>
270 * config/sparc/rtemself.h (LINK_GCC_C_SEQUENCE_SPEC): undefine.
272 2005-01-27 Richard Henderson <rth@redhat.com>
274 * builtins.c (expand_builtin_copysign): New.
275 (expand_builtin): Call it.
276 * genopinit.c (optabs): Add copysign_optab.
277 * optabs.c (init_optabs): Initialize it.
278 (expand_copysign): New.
279 * optabs.h (OTI_copysign, copysign_optab): New.
280 (expand_copysign): Declare.
282 * config/alpha/alpha.md (UNSPEC_COPYSIGN): New.
283 (copysignsf3, ncopysignsf3, copysigndf3, ncopysigndf3): New.
285 * config/i386/i386.c (ix86_build_signbit_mask): Split from ...
286 (ix86_expand_fp_absneg_operator): ... here.
287 (ix86_split_copysign): New.
288 * config/i386/i386-protos.h: Update.
289 * config/i386/i386.md (UNSPEC_COPYSIGN): New.
290 (copysignsf3, copysigndf3): New.
292 * config/ia64/ia64.md (UNSPEC_COPYSIGN): New.
293 (copysignsf3, ncopysignsf3): New.
294 (copysigndf3, ncopysigndf3): New.
295 (copysignxf3, ncopysignxf3): New.
296 * config/ia64/ia64.c (rtx_needs_barrier): Handle UNSPEC_COPYSIGN.
298 2005-01-27 Arend Bayer <arend.bayer@web.de>
299 Kazu Hirata <kazu@cs.umass.edu>
301 * cse.c: (find_best_addr): Don't call copy_rtx before calling
302 fold_rtx. Save cost recomputation if fold_rtx did nothing.
303 (fold_rtx) <ASM_OPERANDS>: Don't do anything if INSN is
306 2005-01-27 Jeff Law <law@redhat.com>
308 * tree-into-ssa.c (ssa_rewrite_initialize_block): Do not register
309 definitions of SSA_NAMEs which are not being rewritten.
310 (rewrite_ssa_into_ssa): Only initialize the current definition
311 of an SSA_NAME if that SSA_NAME has been marked for rewriting.
312 If checking is enabled, assert that SSA_NAME_AUX is clear for all
313 SSA_NAMEs before returning.
315 2005-01-27 Kazu Hirata <kazu@cs.umass.edu>
317 * c-common.def, c-dump.c, c-gimplify.c, c-objc-common.c,
318 cfgexpand.c, dbxout.c, function.h, opts.c, tree-flow-inline.h,
319 tree-ssa-operands.h, tree-ssa-threadupdate.c, tree-tailcall.c,
320 config/avr/avr.md, config/cris/aout.h, config/cris/cris.h,
321 config/mips/iris6.h, config/sh/sh.c: Update copyright.
323 2005-01-27 Steven Bosscher <stevenb@suse.de>
326 * opts.c (decode_options): Move flag_thread_jumps from -O1 and
327 higher to -O2 and higher. Likewise for tree PRE.
328 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Only iterate at -O2
331 2005-01-27 Ian Lance Taylor <ian@airs.com>
334 * gimple-low.c (try_catch_may_fallthru): New static function.
335 (block_may_fallthru): Handle TRY_CATCH_EXPR.
336 * tree-inline.c (expand_call_inline): Don't warn about reaching
337 the end of a non-void function being inlined if the function uses
340 2005-01-27 Jakub Jelinek <jakub@redhat.com>
342 * config/i386/i386.h (CALL_USED_REGISTERS): Fix comment pastos.
345 * c-decl.c (warn_if_shadowing): Handle old_decl error_mark_node.
346 (pushdecl): Only use DECL_FILE_SCOPE_P if DECL_P.
347 (implicitly_declare): Handle error_mark_node.
349 2005-01-27 Richard Henderson <rth@redhat.com>
352 * tree.h (struct tree_decl): Add debug_expr_is_from.
353 (DECL_DEBUG_EXPR_IS_FROM): New.
354 (DECL_DEBUG_EXPR): Rename from DECL_DEBUG_ALIAS_OF.
355 * dwarf2out.c (dwarf2out_var_location): Update to match.
356 * tree-outof-ssa.c (create_temp): Likewise.
357 * var-tracking.c (track_expr_p): Likewise.
358 * tree-sra.c (instantiate_element): Set DECL_DEBUG_EXPR.
359 * c-objc-common.c (c_tree_printer) <'D'>: Handle DECL_DEBUG_EXPR.
360 * toplev.c (default_tree_printer): Likewise.
362 2005-01-27 Alexandre Oliva <aoliva@redhat.com>
364 * config/frv/frv.c (frv_class_likely_spilled_p): Add GR8_REGS,
365 GR9_REGS, GR89_REGS, FDPIC_FPTR_REGS, FDPIC_REGS.
367 2005-01-27 Steven Bosscher <stevenb@suse.de>
369 * tree.h (SWITCH_COND, SWITCH_BODY, SWITCH_LABELS, CASE_LOW,
370 CASE_HIGH, CASE_LABEL, ASM_STRING, ASM_OUTPUTS, ASM_INPUTS,
371 ASM_CLOBBERS): Add tree checks.
373 * c-common.h (SWITCH_TYPE): Rename to SWITCH_STMT_TYPE.
374 (SWITCH_STMT_COND, SWITCH_STMT_BODY): New.
375 * c-common.def (SWITCH_STMT): Update to match.
376 * c-common.c (c_do_switch_warnings): Use SWITCH_STMT accessor
377 macros instead of SWITCH_EXPR ones.
378 * c-dump.c (c_dump_tree): Likewise.
379 * c-gimplify.c (gimplify_switch_stmt): Likewise.
380 * c-typeck.c (c_start_case, do_case, c_finish_case): Likewise.
382 * doc/c-tree.texi (SWITCH_STMT): Update accessor macro names.
384 2005-01-27 Alan Modra <amodra@bigpond.net.au>
386 * unwind-dw2.c (execute_stack_op): Add missing cases for
387 DW_OP_shl, DW_OP_shr, DW_OP_shra, DW_OP_xor.
389 2005-01-27 Eric Botcazou <ebotcazou@libertysurf.fr>
391 * config.gcc (Obsolete configurations): Remove 'dummy', add
392 sparclite-*-coff*, sparclite-*-elf*, sparc86x-*-elf* and
395 2005-01-26 Diego Novillo <dnovillo@redhat.com>
397 PR tree-optimization/19633
398 * tree-ssa-alias.c (ptr_is_dereferenced_by): Also handle
400 (maybe_create_global_var): Do not create .GLOBAL_VAR if there
401 are no call-clobbered variables.
402 * tree-outof-ssa.c (check_replaceable): Return false for calls
405 2005-01-26 Ulrich Weigand <uweigand@de.ibm.com>
407 * dbxout.c (dbxout_symbol_location): Resolve constant pool references
408 even for variables with NULL DECL_INITIAL.
410 2005-01-26 Stuart Hastings <stuart@apple.com>
412 * gimplify.c (shortcut_cond_expr): Re-compute side-effects.
414 2005-01-26 Richard Henderson <rth@redhat.com>
417 * c-decl.c (finish_struct): Set DECL_MODE after resetting a
419 * expr.c (store_field): Strip conversions to odd-bit-sized types
420 if the destination field width matches.
422 2005-01-26 Richard Henderson <rth@redhat.com>
424 * c-decl.c, expmed.c, expr.c: Revert last change.
426 2005-01-26 Richard Henderson <rth@redhat.com>
429 * c-decl.c (finish_struct): Set DECL_MODE after resetting a
431 * expmed.c (store_fixed_bit_field): Create a paradoxical subreg
432 if we don't need the bits above those present in the current mode.
433 * expr.c (store_field): Strip conversions to odd-bit-sized types
434 if the destination field width matches.
436 2005-01-26 Richard Sandiford <rsandifo@redhat.com>
438 * config/mips/iris6.h (ENDFILE_SPEC): Don't link in irix-csr.o if
439 either -ffast-math or -funsafe-math-optimizations is in use.
441 2005-01-26 Marek Michalkiewicz <marekm@amelek.gda.pl>
445 * config/avr/avr.c (notice_update_cc): Only set condition code for
446 ashrqi3 if shift count > 0.
447 (out_shift_with_cnt): Handle shift count <= 0 as a no-op.
448 (ashlqi3_out, ashlhi3_out, ashlsi3_out, ashrqi3_out, ashrhi3_out,
449 ashrsi3_out, lshrqi3_out, lshrhi3_out, lshrsi3_out): Handle shift
450 count <= 0 as a no-op, and shift count >= width by copying zero
451 or sign bit to all bits of the result.
452 * config/avr/avr.md (all shifts): Add alternatives for zero shift
453 count, with attribute "length" set to 0 and "cc" set to "none".
455 2005-01-26 Aldy Hernandez <aldyh@redhat.com>
457 * doc/invoke.texi: Document -mTLS.
459 * testsuite/gcc.target/frv/all-tls-global-dynamic.c: New.
460 * testsuite/gcc.target/frv/all-tls-initial-exec.c: New.
461 * testsuite/gcc.target/frv/all-tls-initial-exec-pic.c: New.
462 * testsuite/gcc.target/frv/all-tls-local-dynamic.c: New.
463 * testsuite/gcc.target/frv/all-tls-local-dynamic-plt.c: New.
464 * testsuite/gcc.target/frv/all-tls-local-dynamic-plt-pic.c: New.
465 * testsuite/gcc.target/frv/all-tls-local-exec.c: New.
466 * testsuite/gcc.target/frv/all-tls-local-exec-TLS.c: New.
468 * configure.ac: Check for a TLS capable gas.
470 * configure: Regenerate.
472 2005-01-26 Richard Henderson <rth@redhat.com>
475 * combine.c (make_field_assignment): Simplify store to zero_extract
476 from a source with an overlapping mask.
478 2005-01-26 Aldy Hernandez <aldyh@redhat.com>
480 2004-11-11 Eric Christopher <echristo@redhat.com>
482 * config/frv/frv.md: Add fr400_integer automaton. Don't
483 allow TYPE_MUL, TYPE_MACC, or TYPE_CUT to issue in parallel
486 2005-01-26 Steven Bosscher <stevenb@suse.de>
489 * cfgbuild.c (make_edges): Do not clear or set
490 current_function_has_computed_jump.
491 * function.h (struct function): Remove the has_computed_jump field.
492 (current_function_has_computed_jump): Do not define.
493 * sched-rgn.c (is_cfg_nonregular): Return true if a basic block ends
494 in a computed jump. Ignore current_function_has_computed_jump.
496 2005-01-26 Richard Henderson <rth@redhat.com>
499 * expr.c (categorize_ctor_elements): New argument p_must_clear.
500 (categorize_ctor_elements_1): Likewise. Detect a union that isn't
502 (mostly_zeros_p): Update for new categorize_ctor_elements argument.
503 * gimplify.c (gimplify_init_constructor): Likewise. Only shove
504 objects into static storage if they have more than one non-zero value.
505 * tree.h (categorize_ctor_elements): Update decl.
507 2005-01-26 Steven Bosscher <stevenb@suse.de>
510 * tree.h (CALL_EXPR_TAILCALL): Add comment.
511 * calls.c (check_sibcall_argument_overlap_1): Revert the change
512 to this function from 2004-07-10.
513 * tree-tailcall.c (suitable_for_tail_opt_p): Do not consider the
514 the current function for tail call optimizations if the address
515 of one of it its arguments is taken.
517 2005-01-26 Kazu Hirata <kazu@cs.umass.edu>
519 * cse.c (fold_rtx): Call equiv_constant only when necessary.
521 2005-01-26 Bernd Schmidt <bernd.schmidt@analog.com>
523 * function.c (expand_function_end): If current_function_calls_alloca,
524 emit stack restore in a place that is reached when the function falls
527 2005-01-26 Alexandre Oliva <aoliva@redhat.com>
529 * Makefile.in (../$(build_subdir)/fixincludes/fixincl): Add dummy
530 rule to avoid matching inappropriate suffix rule.
532 2005-01-26 Jan Hubicka <jh@suse.cz>
534 PR tree-optimization/19241
535 * tree-inline.c (copy_body_r): Do not walk subtrees after substituting.
537 2005-01-26 Hans-Peter Nilsson <hp@axis.com>
539 * config/cris/cris.h (LIB_SPEC) <sim*>: Use -lsyslinux, not
540 -lsyssim. Wrap -lc -lsyssim in --start-group --end-group instead
542 (STARTFILE_SPEC): Add crti.o, before crtbegin.o.
543 <sim*>: Always use crt1.o, regardless of N in simN.
544 <!sim>: Drop support for -pg and -p variants; always use crt0.o.
545 (ENDFILE_SPEC): Add crtn.o, after crtend.o.
546 (CRT_CALL_STATIC_FUNCTION): Remove.
547 * config/cris/aout.h (STARTFILE_SPEC): <!melinux && sim*>: Always
548 use crt1.o, regardless of N in simN.
549 <!sim>: Drop support for -pg and -p variants; always use crt0.o.
550 (LIB_SPEC): Do not define; default to the one in config/cris/cris.h.
551 (CRIS_CPP_SUBTARGET_SPEC) <melinux && !nostdinc>: Remove special
552 case isystem setting.
553 (MAX_OFILE_ALIGNMENT): Correct to 32, not 16.
555 2005-01-25 Ian Lance Taylor <ian@airs.com>
557 * gimple-low.c (block_may_fallthru): Correct handling of
558 SWITCH_EXPR--may fall through if SWITCH_LABELS is NULL.
560 2005-01-26 Hans-Peter Nilsson <hp@axis.com>
562 * config/cris/cris.h (CPP_SPEC): Adjust the various mtune=
563 handlers to match indentation; nested under the mtune=* handler.
564 Similar for mcpu= and march=.
566 2005-01-25 Daniel Jacobowitz <dan@codesourcery.com>
568 * config/mips/linux.h (LIB_SPEC): Remove -rpath-link.
569 * config/mips/linux64.h (LIB_SPEC): Likewise.
570 * config/mn10300/linux.h (LIB_SPEC): Likewise.
571 * config/sh/linux.h (LIB_SPEC): Likewise.
573 2005-01-25 Aldy Hernandez <aldyh@redhat.com>
575 * config/frv/frv.c (frv_legitimize_tls_address): New.
576 (TARGET_HAVE_TLS): Define.
577 (FRV_SYMBOL_REF_TLS_P): Define.
578 (frv_override_options): Handle new register classes.
579 (frv_legitimate_address_p): Reject tls addresses.
580 (frv_legitimize_address): Handle TLS addresses.
581 (gen_inlined_tls_plt): New.
583 (frv_legitimize_tls_address): New.
584 (unspec_got_name): Add TLS entries.
585 (got12_operand): Add R_FRV_TLSMOFF12 case.
586 (frv_emit_move): Fixup TLS addresses.
587 (frv_emit_movsi): Legitimize TLS addresses.
589 * config/frv/frv.h (MASK_BIG_TLS): New.
590 (TARGET_BIG_TLS): New.
591 (HAVE_AS_TLS): Define.
592 (TARGET_SWITCHES): Add -mTLS and -mtls options.
593 (enum reg_class): Add GR8_REGS, GR9_REGS, GR89_REGS.
594 (REG_CLASS_NAMES): Same.
595 (REG_CLASS_CONTENTS): Same.
596 (CONSTRAINT_LEN): New.
597 (REG_CLASS_FROM_CONSTRAINT): New.
598 (PREDICATE_CODES): Add symbolic_operand.
600 * config/frv/frv.md (define_constants): Add UNSPEC_GETTLSOFF,
601 UNSPEC_TLS_LOAD_GOTTLSOFF12, UNSPEC_TLS_INDIRECT_CALL,
602 UNSPEC_TLS_TLSDESC_LDD, UNSPEC_TLS_TLSDESC_LDD_AUX,
603 UNSPEC_TLS_TLSOFF_LD, UNSPEC_TLS_LDDI, UNSPEC_TLSOFF_HILO,
604 R_FRV_GOTTLSOFF_HI, R_FRV_GOTTLSOFF_LO, R_FRV_TLSMOFFHI,
605 R_FRV_TLSMOFFLO, R_FRV_TLSMOFF12, R_FRV_TLSDESCHI,
606 R_FRV_TLSDESCLO, R_FRV_GOTTLSDESCHI, R_FRV_GOTTLSDESCLO, GR8_REG,
607 GR9_REG, GR14_REG, LRREG.
608 (type): Add load_or_call attribute.
609 ("load_or_call"): New reservation.
610 ("call_gettlsoff"): New.
611 ("tls_indirect_call"): New.
612 ("tls_load_gottlsoff12"): New.
613 ("tlsoff_hilo"): New.
614 ("tls_tlsdesc_ldd"): New.
615 ("tls_tlsoff_ld"): New.
618 * config/frv/frv-protos.h (symbolic_operand): Protoize.
620 2005-01-25 Alexandre Oliva <aoliva@redhat.com>
622 * config/frv/frv.h (ASM_OUTPUT_DWARF_DTPREL): Define.
623 * config/frv/frv-protos.h (frv_output_dwarf_dtprel): Declare.
624 * config/frv/frv.c (TLS_BIAS): Define.
625 (frv_output_dwarf_dtprel): New.
627 2005-01-26 Jakub Jelinek <jakub@redhat.com>
629 PR rtl-optimization/19579
630 * ifcvt.c (noce_try_cmove_arith): If emitting instructions to set up
631 both A and B, see if they don't clobber registers the other expr uses.
633 2005-01-25 J"orn Rennecke <joern.rennecke@st.com>
635 * real.c (do_add): Initialize signalling and canonical members.
637 * real.c (real_from_integer): Zero out destination.
639 2005-01-18 Jan Hubicka <jh@suse.cz>
641 PR tree-optimize/19337
642 * tree-inline.c (initialize_inlined_parameters): Copy
643 saved_static_chain_decl only.
645 2005-01-25 Richard Henderson <rth@redhat.com>
649 * config/i386/i386.c (x86_inter_unit_moves): Zero.
650 (ix86_preferred_reload_class): Rewrite fp-constant section, with
651 80387 enabled, to return a proper subclass. Return the subset that
652 overlaps with GENERAL_REGS for PLUS.
653 * config/i386/i386.md (movsi_1, movdi_2): Set type to "mmx" for pxor.
655 2005-01-25 Richard Henderson <rth@redhat.com>
658 * tree-complex.c (expand_complex_div_wide): Use the correct formulae.
660 2005-01-25 Roger Sayle <roger@eyesopen.com>
662 * expmed.c (choose_multiplier): Change interface to return the
663 multiplier lowpart as an rtx instead of an unsigned HOST_WIDE_INT.
664 (expand_divmod): Update calls to choose_multiplier.
666 2005-01-24 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
669 * expmed.c (extract_bit_field): Use adjust_address instead of
670 gen_lowpart when op0 is a MEM.
672 2005-01-24 Steven Bosscher <stevenb@suse.de>
674 * cfgexpand.c (expand_gimple_tailcall): Fix typo.
676 2005-01-24 Richard Sandiford <rsandifo@redhat.com>
678 * config/mips/irix-csr.c: New file.
679 * config/mips/t-iris6 (irix-csr.o): New rule to build it.
680 (EXTRA_MULTILIB_PARTS): Add irix-csr.o.
681 * config/mips/iris6.h (ENDFILE_SPEC): Include it in n32 and n64
684 2005-01-24 Eric Botcazou <ebotcazou@libertysurf.fr>
687 * config.gcc (sparc-*-elf*): Remove sol2.h, sparc/sol2.h and
688 sparc/elf.h, add sparc/sp-elf.h.
689 (sparc-*-rtems*): Likewise.
690 (sparclite-*-elf*): Remove sol2.h, sparc/sol2.h, sparc/elf.h and
691 tm-dwarf2.h, add sparc/sp-elf.h.
692 (sparc86x-*-elf): Likewise.
693 (sparc64-*-elf*): Remove sol2.h, sparc/sol2.h and tm-dwarf2.h.
694 * config/sparc/liteelf.h (TARGET_SUB_OS_CPP_BUILTINS): Rename into
695 TARGET_OS_CPP_BUILTINS.
696 * config/sparc/sp86x-elf (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
697 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
699 * config/sparc/openbsd64.h (NO_IMPLICIT_EXTERN_C): Undefine.
700 * config/sparc/sp64-elf.h (NO_IMPLICIT_EXTERN_C): New macro.
701 (SWITCH_TAKES_ARG): Likewise.
702 (LOCAL_LABEL_PREFIX): Likewise.
703 (ASM_GENERATE_INTERNAL_LABEL): Likewise.
704 (TARGET_N_FORMAT_TYPES): Delete.
705 (TARGET_FORMAT_TYPES): Likewise.
706 (ASM_DECLARE_FUNCTION_SIZE): Likewise.
707 * config/sparc/elf.h: Delete.
708 * config/sparc/sp-elf.h: New file.
710 2005-01-24 Kazu Hirata <kazu@cs.umass.edu>
712 * tree-flow-inline.h (phi_arg_from_edge): Remove.
713 * tree-flow.h: Remove the corresponding prototype.
714 * tree-ssa-operands.h, tree-ssa-threadupdate.c, tree-ssa.c,
715 tree-vectorizer.c: Use dest_idx instead of phi_arg_from_edge.
717 2005-01-24 Jakub Jelinek <jakub@redhat.com>
720 * config/rs6000/rs6000.c (rs6000_file_start): Create toc section
721 for AIX ABI or ELF -fPIC.
722 (rs6000_emit_load_toc_table): Don't create toc_section here.
723 (rs6000_xcoff_file_start): Nor here.
725 2005-01-24 J"orn Rennecke <joern.rennecke@st.com>
727 * sh.c (ra.h): Don't #include.
728 (hard_regs_intersect_p): New function, resurrected from ra.c.
730 * sh.c: Fix 1996 Copyright.
732 2005-01-24 Richard Henderson <rth@redhat.com>
733 Aldy Hernandez <aldyh@redhat.com>
735 * regrename.c (note_sets): Handle subregs.
737 2005-01-24 Joseph S. Myers <joseph@codesourcery.com>
740 * genconditions.c (write_header, write_conditions): Elide file if
743 2005-01-24 Kazu Hirata <kazu@cs.umass.edu>
745 * c-opts.c, expmed.c, gengtype.c, tree-complex.c,
746 config/arm/iwmmxt.md, config/avr/avr.c, config/s390/s390.h:
749 2005-01-24 Jakub Jelinek <jakub@redhat.com>
751 * builtins.c (get_pointer_alignment, gimplify_va_arg_expr,
752 expand_builtin_printf, expand_builtin_fprintf,
753 expand_builtin_sprintf): Use POINTER_TYPE_P instead TREE_CODE
754 checking against POINTER_TYPE.
755 (validate_arglist): Handle POINTER_TYPE code by checking
758 2005-01-24 Paolo Bonzini <bonzini@gnu.org>
760 * rtlanal.c (non_rtx_starting_operands, for_each_rtx_1,
762 (for_each_rtx): Call for_each_rtx_1.
763 * rtl.h (init_rtlanal): Declare.
764 * toplev.c (backend_init): Call init_rtlanal.
766 2005-01-24 Jakub Jelinek <jakub@redhat.com>
768 * flow.c (propagate_one_insn): Formatting.
771 * flow.c (libcall_dead_p): Be more conservative if unsure.
772 If there are any instructions between insn and call, see if they are
773 all dead before saying the libcall is dead.
775 2005-01-24 Uros Bizjak <uros@kss-loka.si>
777 * config/i386/i386.md (*extendsfdf2_sse): Use "nonimmediate_operand"
778 operand constraint for operand 0.
780 2005-01-23 Richard Henderson <rth@redhat.com>
783 * tree-complex.c (expand_complex_div_wide): Set EDGE_FALLTHRU on
784 edges leading to join block.
786 2005-01-23 Roger Sayle <roger@eyesopen.com>
788 * expmed.c (expand_mult_highpart): Make static. Change type of
789 constant multiplier argument from unsigned HOST_WIDE_INT to rtx.
790 (expand_divmod): Updates calls to expand_mult_highpart by using
791 gen_int_mode to pass a CONST_INT rtx.
792 * rtl.h (expand_mult_highpart): Remove prototype.
794 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
796 * flow.c (regset_bytes, regset_size): Remove.
798 * cfganal.c: Fix a reference to Harvey's paper.
800 2005-01-23 Steven Bosscher <stevenb@suse.de>
802 PR rtl-optimization/19464
803 * tree-optimize.c (init_tree_optimization_passes): Add one more
804 copyrename pass just before out-of-ssa.
806 2005-01-23 Joseph S. Myers <joseph@codesourcery.com>
809 * recog.c (recog_memoized): Don't define if GENERATOR_FILE.
810 * ggc-none.c (ggc_free): Define.
812 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
814 * cse.c (max_reg, max_insn_uid): Remove.
815 (cse_main): Don't access max_reg or max_insn_uid.
817 * c-common.c (explicit_flag_signed_bitfields,
818 lang_statement_code_p, lang_expand_function_end): Remove.
819 * c-common.h: Remove the corresponding declarations.
820 * c-opts.c (c_common_handle_option): Remove all write access
821 to explicit_flag_signed_bitfields.
823 * cgraph.c (cgraph_varpool_n_nodes): Remove.
824 (cgraph_varpool_node): Don't access cgraph_varpool_n_nodes.
825 * cgraph.h: Remove the corresponding declaration.
827 * gcse.c (null_pointer_info): Remove.
829 2005-01-23 Roger Sayle <roger@eyesopen.com>
830 Eric Botcazou <ebotcazou@libertysurf.fr>
832 * combine.c (simplify_logical): Only simplify logical expressions
833 of the form ior(and(x,y),z) by the inverse distributive law if the
834 result is cheaper than the original.
836 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
838 * alias.c, c-common.h, c-incpath.c, c-incpath.h, expr.c,
839 fold-const.c, gimplify.c, params.h, tree-data-ref.c,
840 tree-if-conv.c, tree-nested.c, tree-outof-ssa.c,
841 tree-ssa-dom.c, tree-vectorizer.c, tree.def, config/darwin.c,
842 config/freebsd-spec.h, config/arm/arm.h,
843 config/h8300/h8300.md, config/i386/i386.md,
844 config/i386/predicates.md, config/i386/sse.md,
845 config/ia64/ia64.c, config/ip2k/ip2k.c, config/s390/s390.c,
846 config/vax/vax.md: Fix comment typos. Follow spelling
849 2005-01-23 Steven Bosscher <stevenb@suse.de>
851 * gengtype.c (walk_type): Produce `not equal to' compares for loop
852 tests, instead of `less than'.
854 2005-01-23 Richard Sandiford <rsandifo@redhat.com>
856 * function.c (assign_parm_setup_block): When creating a new stack slot
857 for a parameter, get its alignment from the parameter's DECL_ALIGN
858 rather than the type's TYPE_ALIGN. Make sure that the parameter
859 is at least word aligned.
861 2005-01-22 Roger Sayle <roger@eyesopen.com>
864 * config/avr/avr.c (avr_hard_regno_mode_ok): Rewrite.
866 2005-01-22 Richard Henderson <rth@redhat.com>
869 * config/i386/i386.md (movsfcc_1_sse_max): Use nonimmediate_operand
870 in both compare operands.
871 (movdfcc_1_sse_max): Likewise.
872 (movsfcc_1_sse): Likewise. Add earlyclobber for scratch.
873 (movdfcc_1_sse): Likewise.
874 * config/i386/i386.c (ix86_split_sse_movcc): Emit copies into the
875 scratch register as needed.
877 2005-01-22 Richard Henderson <rth@redhat.com>
879 * config/i386/i386.md (smaxsf3): Fix mnemonic typo.
881 * config/i386/i386.c (ix86_prepare_fp_compare_args): Fix is_sse test.
883 2005-01-22 Richard Henderson <rth@redhat.com>
885 * genopinit.c (optabs): Use smin/smax for floating point too.
886 * doc/md.texi: Update to match. Clarify that floating point
887 results are undefined for +0/-0 and NaN.
888 * doc/rtl.texi: Likewise.
889 * rtl.def (SMIN, SMAX): Likewise
890 * tree.def (MIN_EXPR, MAX_EXPR): Likewise.
892 * config/alpha/alpha.md (smaxdf3, smindf3, smaxsf3, sminsf3): Add
893 leading 's' to the name.
894 * config/ia64/ia64.md (smaxsf3, sminsf3, smaxdf3, smindf3,
895 smaxxf3, sminxf3): Likewise.
896 * config/rs6000/rs6000.md (smaxdf3, smindf3, smaxsf3, sminsf3):
899 2005-01-22 Paul Brook <paul@codesourcery.com>
901 * config/arm/arm.md: Use "Uy" constraint for wcgr load/stores.
903 2005-01-21 Mark Dettinger <dettinge@de.ibm.com>
905 * config/s390/s390.c (struct processor_costs): New fields
906 dlgr, dlr, dr, dsgfr, dsgr.
907 (z900_cost, z990_cost): Values for new fields.
908 (s390_rtx_costs): New cases MEM und COMPARE in switch
909 statement. Modified handling of SIGN_EXTEND, ZERO_EXTEND,
910 DIV, MOD, UDIV, UMOD.
912 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
914 * config/s390/s390.md ("doloop_si64"): Reload input value directly
915 into the register being decremented.
916 ("doloop_si31", "doloop_di"): Likewise.
917 ("*doloop_si_long"): Adapt pattern.
919 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
921 * config/s390/s390.h (HARD_REGNO_NREGS): Fix computation for
923 (CLASS_MAX_NREGS): Likewise.
925 2005-01-21 Daniel Berlin <dberlin@dberlin.org>
927 * doc/tree-ssa.texi (Statement Operands): Add example for new
928 must-def macro. Note deprecation of old operands interface.
930 2005-01-22 Richard Sandiford <rsandifo@redhat.com>
932 PR tree-optimization/19484
933 * tree-cfg.c (remove_fallthru_edge): New function.
934 (cleanup_control_flow): Remove fallthru edges from calls that are
935 now known not to return.
937 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
939 Fix PR tree-optimization/19038
940 * tree-ssa-dom.c (cprop_operand): Don't replace loop invaeriant
941 copies with loop variant ones.
943 2005-01-22 Kazu Hirata <kazu@cs.umass.edu>
945 * cfganal.c, real.h, reorg.c, timevar.def, tree-ssa-ccp.c,
946 config/alpha/alpha-protos.h, config/alpha/alpha.h,
947 config/alpha/alpha.md, config/alpha/predicates.md,
948 config/sparc/freebsd.h, config/sparc/netbsd-elf.h,
949 config/sparc/sol2.h: Update copyright.
951 * tree-cfg.c (remove_forwarder_block_with_phi): Look at the
952 first label to see if it is a nonlocal label.
954 2005-01-22 David Edelsohn <edelsohn@gnu.org>
955 Andrew Pinski <pinskia@physics.uc.edu>
958 * config/rs6000/rs6000.c (rs6000_va_start): Saturate n_gpr at
959 maximum number of GPRs. Saturate n_fpr at maximum number of FPRs.
961 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
963 * config/rs6000/rtems.h: Update copyright.
965 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
968 * config/rs6000/rtems.h: Resurrect cpp_os_rtems_spec from gcc < 3.4.
969 (CPP_OS_RTEMS_SPEC): New (From gcc-3.3's config/rs6000/sys4.h).
970 (SUBSUBTARGET_EXTRA_SPECS): Use CPP_OS_RTEMS_SPEC.
972 2005-01-22 Volker Reichelt <reichelt@igpm.rwth-aachen.de>
975 * c-typeck.c (convert_arguments): Check for error_mark_node.
977 2005-01-21 Ian Lance Taylor <ian@airs.com>
979 PR tree-optimization/13000
980 * tree-inline.c: Include "tree-flow.h".
981 (expand_call_inline): If warn_return_type, warn if non-void inline
982 function falls through.
983 * tree-cfg.c (execute_warn_function_return): Don't warn about
984 control reaching end if TREE_NO_WARNING is set. Set
986 * gimple-low.c (block_may_fallthru): Don't assume that SWITCH_EXPR
988 * gimplify.c (shortcut_cond_expr): Don't emit a jump over the else
989 branch if we don't need one.
990 * c-typeck.c: Include "tree-flow.h"
991 (c_finish_bc_stmt): Don't add a goto if the current statement
992 list doesn't fall through to the current point.
994 2005-01-21 Roger Sayle <roger@eyesopen.com>
996 PR rtl-optimization/576
997 * real.c (real_arithmetic): Change return type from void to bool
998 to return an indication that the result may be inexact.
999 * real.h (real_arithmeric): Update prototype.
1000 * fold-const.c (const_binop): Don't constant fold floating
1001 point expressions when the user specifies -frounding-math and
1002 the result may depend upon the run-time rounding mode.
1003 (fold_convert_const_real_from_real): Clean-up.
1004 (fold_initializer): Ignore flag_rounding_math for initializers.
1005 * simplify-rtx.c (simplify_binary_operation): Likewise, don't
1006 constant fold FP operations with flag_rounding_math if the
1007 result may depend upon the run-time rounding mode.
1009 2005-01-21 Tom Tromey <tromey@redhat.com>
1011 * c-cppbuiltin.c (define__GNUC__): Correct assertion.
1013 2005-01-21 Kazu Hirata <kazu@cs.umass.edu>
1015 * tree-cfg.c: Fix comment typos.
1017 2005-01-21 J"orn Rennecke <joern.rennecke@st.com>
1019 * passes.c (rest_of_handle_flow2): Close / open the flow2 dump file
1020 around the call to rest_of_handle_branch_target_load_optimize.
1022 2005-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
1024 * config.gcc (sparc-*-netbsdelf*): Include dbxelf.h
1025 (sparc64-*-openbsd*): Likewise.
1026 (sparclite-*-elf*): Include tm-dwarf2.h.
1027 (sparc86x-*-elf*): Likewise.
1028 (sparc64-*-elf*): Likewise.
1029 * config/sparc/linux64.h (DWARF2_DEBUGGING_INFO): Delete.
1030 (DBX_DEBUGGING_INFO): Likewise.
1031 * config/sparc/liteelf.h (DWARF2_DEBUGGING_INFO): Likewise.
1032 (PREFERRED_DEBUGGING_TYPE): Likewise.
1033 * config/sparc/netbsd-elf.h (PREFERRED_DEBUGGING_TYPE): Likewise.
1034 * config/sparc/sp64-elf.h (DBX_DEBUGGING_INFO): Likewise.
1035 (PREFERRED_DEBUGGING_TYPE): Likewise.
1036 * config/sparc/sp86x-elf.h (DWARF2_DEBUGGING_INFO): Likewise.
1037 (PREFERRED_DEBUGGING_TYPE): Likewise.
1038 * config/sparc/sparc.h (DBX_DEBUGGING_INFO): Likewise.
1040 2005-01-21 Hans-Peter Nilsson <hp@bitrange.com>
1043 * combine.c (combine_simplify_rtx): Revert change of 2004-12-31.
1045 2005-01-20 Paul Brook <paul@codesourcery.com>
1047 * doc/rtl.texi: Document value extension requirements for CONST_INT.
1049 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
1051 * tree-cfg.c (tree_verify_flow_info): Check that a nonlocal
1052 label is first in a sequence of labels.
1054 * tree-cfg.c (tree_verify_flow_info): Fix a typo.
1056 2005-01-20 Janis Johnson <janis187@us.ibm.com>
1057 Giovanni Bajo <giovannibajo@gcc.gnu.org>
1059 * doc/sourcebuild.texi (Test Directives): New.
1061 2005-01-20 Roger Sayle <roger@eyesopen.com>
1063 * real.c (real_floor): Don't leave the result uninitialized when
1064 mode is VOIDMode, but return the unrounded intermediate value.
1065 (real_ceil): Likewise.
1067 2005-01-20 Paul Brook <paul@codesourcery.com>
1069 * config/arm/arm.md (insv): Use gen_int_mode.
1071 2005-01-20 Eric Botcazou <ebotcazou@libertysurf.fr>
1073 * config/sparc/freebsd.h (ASM_OUTPUT_INTERNAL_LABELREF): Delete.
1074 * config/sparc/linux.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1075 * config/sparc/linux64.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1076 * config/sparc/netbsd-elf.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1077 * config/sparc/sol2.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1079 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
1082 * dwarf2out.c (concat_loc_descriptor): We don't know we can use
1083 fbreg, so we have to assume we can't.
1084 (loc_descriptor_from_tree_1): Ditto.
1085 (containing_function_has_frame_base): New function.
1086 (add_location_or_const_value_attribute): Use it.
1087 Also try to generate a frame_base from a single element location
1090 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
1092 PR tree-optimization/15349
1093 * timevar.def (TV_TREE_MERGE_PHI): New.
1094 * tree-cfg.c (tree_forwarder_block_p): Add a new argument
1096 (remove_forwarder_block, cleanup_forwarder_blocks): Adjust the
1097 calls to tree_forwarder_block_p.
1098 (remove_forwarder_block_with_phi, merge_phi_nodes,
1099 gate_merge_phi, pass_merge_phi): New.
1100 * tree-optimize.c (init_tree_optimization_passes): Add
1102 * tree-pass.h: Add an extern for pass_merge_phi;
1104 2005-01-20 Richard Henderson <rth@redhat.com>
1107 * config/i386/emmintrin.h (_mm_castpd_ps, _mm_castpd_si128): New.
1108 (_mm_castps_pd, _mm_castps_si128): New.
1109 (_mm_castsi128_ps, _mm_castsi128_pd): New.
1111 2005-01-20 Richard Henderson <rth@redhat.com>
1114 * config/i386/mmintrin.h (_mm_cvtsi32_si64): Use
1115 __builtin_ia32_vec_init_v2si.
1116 (_mm_cvtsi64_si32): Use __builtin_ia32_vec_ext_v2si.
1117 * config/i386/i386.c (IX86_BUILTIN_VEC_EXT_V2SI): New.
1118 (ix86_init_mmx_sse_builtins): Create it.
1119 (ix86_expand_builtin): Expand it.
1120 (ix86_expand_vector_set): Handle V2SFmode and V2SImode.
1121 * config/i386/mmx.md (vec_extractv2sf_0, vec_extractv2sf_1): New.
1122 (vec_extractv2si_0, vec_extractv2si_1): New.
1124 2005-01-20 Richard Henderson <rth@redhat.com>
1127 * config/i386/i386.c (ix86_expand_push): New.
1128 * config/i386/mmx.md (push<MMXMODE>1): New.
1129 * config/i386/sse.md (push<SSEMODE>1): New.
1130 * config/i386/i386-protos.h: Update.
1132 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
1134 * params.def, params.h, predict.c, tree-ssa-loop-im.c: Update
1137 2005-01-20 Richard Henderson <rth@redhat.com>
1139 * rtl.def (CONST_VECTOR): Use RTX_CONST_OBJ.
1140 * rtl.h (CONSTANT_P): Don't special case CONST_VECTOR.
1142 2005-01-19 Richard Henderson <rth@redhat.com>
1145 * config/i386/i386.c (ix86_expand_vector_move_misalign): Convert
1146 to V4SFmode in SSE1 fallback load path.
1148 2005-01-19 Richard Henderson <rth@redhat.com>
1150 * config/i386/i386.c (ix86_expand_vector_init_one_var): Fix typo
1151 in QImode expansion to ix86_expand_vector_set.
1152 (ix86_expand_vector_init_general): Fix typo in V8HImode recursive call.
1154 2005-01-19 Richard Henderson <rth@redhat.com>
1157 * config/i386/i386.c (ix86_preferred_reload_class): Return a proper
1158 subclass of the input class.
1159 (ix86_secondary_memory_needed): Always true for cross-MMX classes.
1160 Always true for cross-SSE1 classes. Rationalize conditionals.
1161 * config/i386/i386.h (SSE_CLASS_P, MMX_CLASS_P): Use straight equality.
1162 * config/i386/i386.md (movsi_1): Add MMX/SSE zeros. Fix alternatives
1163 for SSE1. Don't check TARGET_INTER_UNIT_MOVES.
1164 (movdi_2): Add MMX/SSE zeros.
1165 (movdi_1_rex64): Likewise. Don't check TARGET_INTER_UNIT_MOVES.
1166 (movsf_1): Don't check TARGET_INTER_UNIT_MOVES.
1167 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
1168 (movsi_1_nointernunit, movdi_1_rex64_nointerunit): Remove.
1169 (movsf_1_nointerunit, zero_extendsidi2_32_1): Remove.
1170 (zero_extendsidi2_rex64_1): Remove.
1171 (MOV0 peephole): Check GENERAL_REG_P.
1173 2005-01-19 Richard Henderson <rth@redhat.com>
1176 * config/i386/i386.c (ix86_expand_vector_set): Fix third and fourth
1178 (ix86_expand_vector_extract): Likewise.
1180 2005-01-19 Richard Henderson <rth@redhat.com>
1183 * expr.c (emit_move_change_mode): New argument force; use
1184 simplify_gen_subreg if true.
1185 (emit_move_via_alt_mode): Merge into ...
1186 (emit_move_via_integer): ... here.
1187 (emit_move_ccmode): Use emit_move_change_mode directly.
1189 2005-01-19 Richard Henderson <rth@redhat.com>
1192 * config/alpha/alpha.c (alpha_rtx_costs): Handle HIGH.
1193 (alpha_preferred_reload_class): Handle CONST_VECTOR.
1194 (alpha_emit_set_const_1): Add no_output parameter; don't emit
1196 (alpha_emit_set_const): Likewise. Make static.
1197 (alpha_emit_set_long_const): Make static.
1198 (alpha_extract_integer): Split out from alpha_expand_mov.
1199 (alpha_split_const_mov): Likewise.
1200 (alpha_expand_mov): Use them. Handle CONST_VECTOR.
1201 (alpha_legitimate_constant_p): New.
1202 * config/alpha/alpha-protos.h: Update.
1203 * config/alpha/alpha.h (REGISTER_MOVE_COST): Correct fp<->gp cost.
1204 (LEGITIMATE_CONSTANT_P): Re-implement with a function.
1205 * config/alpha/alpha.md (movsi): Add n alternative.
1206 (movsi_nt_vms, movdi_er_nofix, movdi_er_fix, movdi_fix): Likewise.
1207 (mov<VEC>_fix, mov<VEC>_nofix): Add i alternative.
1208 (splitters for all of the above): Use alpha_split_const_mov.
1209 * config/alpha/predicates.md (non_add_const_operand): New.
1210 (non_zero_const_operand): New.
1211 (input_operand): Use alpha_legitimate_constant_p after reload.
1213 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
1215 PR tree-optimization/19038
1216 * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): New function.
1217 (add_candidate): Add ivs with increment in latch only if
1218 allow_ip_end_pos_p is true.
1219 (determine_iv_cost): Use empty_block_p.
1221 2005-01-19 Daniel Berlin <dberlin@dberlin.org>
1223 * cfganal.c (compute_dominance_frontiers_1): Replace with new algorithm
1224 (compute_dominance_frontiers): Ditto.
1226 2005-01-19 Ralf Corsepius <ralf.corsepius@rtems.org>
1229 * config/sh/t-rtems: New.
1230 * config.gcc (sh-*-rtems*): Reflect having added config/sh/t-rtems.
1232 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
1234 * tree-ssa-ccp.c (maybe_fold_offset_to_component_ref): Always subtract
1235 the offset of the selected field.
1237 2005-01-19 Kazu Hirata <kazu@cs.umass.edu>
1239 * tree-cfg.c (remove_forwarder_block): Fix the check to
1240 prevent a nonlocal label from appearing in the middle of a
1243 2005-01-19 Hans-Peter Nilsson <hp@axis.com>
1245 PR rtl-optimization/19462
1246 * reorg.c (find_end_label): Create return insn only if
1247 current_function_epilogue_delay_list is empty.
1249 2005-01-19 Paolo Bonzini <bonzini@gnu.org>
1251 * doc/passes.texi: Remove paragraph mentioning new-ra.
1253 2005-01-19 Jakub Jelinek <jakub@redhat.com>
1255 PR rtl-optimization/15139
1256 * combine.c: Include params.h.
1257 (count_rtxs): New function.
1258 (record_value_for_reg): If replace_rtx would replace at least
1259 2 occurrences of REG in VALUE and TEM is really large, replace REG with
1260 (clobber (const_int 0)) instead of TEM.
1261 * params.def (PARAM_MAX_LAST_VALUE_RTL): New.
1262 * params.h (MAX_LAST_VALUE_RTL): New.
1263 * Makefile.in (combine.o): Depend on $(PARAMS_H).
1264 * doc/invoke.texi (--param max-last-value-rtl=N): Document.
1267 * c-typeck.c (digest_init): Only call build_vector if all constructor
1268 elements are *_CST nodes.
1269 * gimplify.c (gimplify_init_constructor): Likewise.
1272 * c-typeck.c (digest_init): Only call build_vector if inside_init
1275 2005-01-18 Tobias Schl"uter <tobias.schlueter@physik.uni-muenchen.de>
1277 * toplev.c (init_asm_output): Remove dead #ifdef.
1279 2005-01-18 Kaz Kojima <kkojima@gcc.gnu.org>
1281 * config/sh/linux.h (TARGET_C99_FUNCTIONS): Define.
1283 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
1285 * calls.c (expand_call): Check DECL_BUILT_IN_CLASS before
1286 accessing DECL_FUNCTION_CODE.
1287 * dojump.c (do_jump): Likewise.
1288 * gimplify.c (gimplify_call_expr): Likewise.
1289 * predict.c (expr_expected_value): Likewise.
1290 (strip_builtin_expect): Likewise.
1291 * tree-inline.c (estimate_num_insns_1): Likewise.
1292 * tree-ssa-loop-im.c (stmt_cost): Likewise
1293 * fold-const.c (fold): Test for BUILT_IN_NORMAL.
1294 (tree_expr_nonnegative_p): Likewise.
1296 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
1298 * config/sparc/sparc.c (load_pic_register): Emit the appropriate
1299 variant of the load_pcrel_sym pattern.
1300 * config/sparc/sparc.md (P macro): Move to the top.
1301 (load_pcrel_sym): Macroize using P.
1302 (save_register_window): Likewise.
1304 2005-01-18 Aldy Hernandez <aldyh@redhat.com>
1306 * config/rs6000/rs6000.md ("sunordered"): Disable for e500.
1309 2005-01-18 Andrew Pinski <pinskia@physics.uc.edu>
1312 * c-typeck.c (build_asm_expr): Strip nops off
1313 input memory operands.
1315 2005-01-18 David Edelsohn <edelsohn@gnu.org>
1317 * config/rs6000/aix43.h (CPLUSPLUS_CPP_SPEC): Delete
1318 _XOPEN_SOURCE, _XOPEN_SOURCE_EXTENDED, _LARGE_FILE_API.
1319 * config/rs6000/aix51.h (CPLUSPLUS_CPP_SPEC): Same.
1320 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Same.
1322 2005-01-18 Joel Sherrill <joel.sherrill@oarcorp.com>
1325 * config/i386/i386.c (override_options): If the 80387 is disabled,
1326 then do not return FP values using FP registers.
1328 2005-01-18 Kazu Hirata <kazu@cs.umass.edu>
1330 * basic-block.h, c-common.c, c-cppbuiltin.c, c-lang.c,
1331 c-tree.h, cfgbuild.c, cgraph.c, cgraph.h, collect2.c,
1332 combine.c, config.gcc, coverage.h, cse.c, cselib.c,
1333 defaults.h, df.c, dwarf2asm.c, dwarf2out.c, explow.c, expr.c,
1334 flow.c, fold-const.c, gcse.c, ggc-page.c, gimple-low.c,
1335 gimplify.c, ifcvt.c, langhooks-def.h, lcm.c, optabs.h,
1336 output.h, postreload-gcse.c, postreload.c, recog.c,
1337 resource.c, rtl.def, rtlanal.c, sched-deps.c, sched-rgn.c,
1338 targhooks.h, toplev.c, tree-data-ref.c, tree-eh.c,
1339 tree-flow.h, tree-loop-linear.c, tree-mudflap.h, tree-nrv.c,
1340 tree-optimize.c, tree-outof-ssa.c, tree-pass.h,
1341 tree-scalar-evolution.c, tree-ssa-copy.c, tree-ssa-dce.c,
1342 tree-ssa-dse.c, tree-ssa-loop-ivopts.c, tree-ssa-operands.c,
1343 tree-ssa-pre.c, tree-ssa.c, tree-vectorizer.c, tree.def,
1344 unwind-dw2-fde-darwin.c, var-tracking.c: Update copyright.
1346 * config/darwin.c, config/darwin.h, config/freebsd-spec.h,
1347 config/alpha/alpha.c, config/arm/vfp.md,
1348 config/i386/emmintrin.h, config/i386/i386-modes.def,
1349 config/i386/i386-protos.h, config/i386/i386.h,
1350 config/i386/pmmintrin.h, config/i386/ppro.md,
1351 config/i386/predicates.md, config/i386/xmmintrin.h,
1352 config/i860/i860.c, config/ia64/ia64-modes.def,
1353 config/ia64/ia64-protos.h, config/ia64/ia64.h,
1354 config/ia64/ia64.md, config/ia64/itanium1.md,
1355 config/ia64/itanium2.md, config/ia64/predicates.md,
1356 config/iq2000/iq2000.h, config/mips/linux64.h,
1357 config/rs6000/aix41.h, config/rs6000/aix43.h,
1358 config/rs6000/aix52.h, config/rs6000/darwin-fallback.c,
1359 config/rs6000/rs6000.c, config/rs6000/rs6000.h,
1360 config/rs6000/spe.md, config/sh/sh.md, config/sparc/linux.h,
1361 config/sparc/linux64.h, config/sparc/litecoff.h,
1362 config/sparc/sp64-elf.h, config/sparc/sparc.c,
1363 config/sparc/sparc.h, config/sparc/sparc.md,
1364 config/sparc/sysv4-only.h: Update copyright.
1366 2005-01-18 Richard Henderson <rth@redhat.com>
1369 * config/i386/i386.c (ix86_expand_fp_movcc): Fail for LTGT and UNEQ.
1371 2005-01-18 Marc Espie <espie@openbsd.org>
1372 * config/i386/openbsdelf.h: Typo.
1374 2005-01-18 Richard Henderson <rth@redhat.com>
1376 * config/i386/i386.c (ix86_expand_fp_movcc): Remove TARGET_IEEE_FP
1377 special case for sse.
1378 * config/i386/predicates.md (sse_comparison_operator): Likewise.
1380 2005-01-18 Dorit Naishlos <dorit@il.ibm.com>
1382 * tree-inline.c (estimate_num_insns_1): Added cases for
1383 ALIGN_INDIRECT_REF, MISALIGNED_INDIRECT_REF, and REALIGN_LOAD_EXPR.
1385 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
1387 PR rtl-optimization/19296
1388 * combine.c (simplify_comparison): Rewrite the condition under
1389 which a non-paradoxical SUBREG of a PLUS can be lifted when
1390 compared against a constant.
1392 2005-01-18 Andi Kleen <ak@muc.de>
1394 * c-typeck.c: (convert_for_assignment): Check warn_pointer_sign.
1395 * c.opt (-Wpointer-sign): Add.
1396 * doc/invoke.texi: (-Wpointer-sign): Add.
1398 2005-01-18 Uros Bizjak <uros@kss-loka.si>
1400 * config/i386/i386.c (override_options): Revert 2004-11-24 change.
1401 * config/i386/i386.md (fmodsf3, dremsf3, *sinsf2, *cossf2, sincossf3,
1402 *tansf3_1, tansf2, atan2sf3_1, atan2sf3, atansf2, asinsf2, acossf2,
1403 logsf2, log10sf2, log2sf2, log1psf2, logbsf2, ilogbsf2, expsf2,
1404 exp10sf2, exp2sf2, expm1sf2, rintsf2, floorsf2, ceilsf2, btruncsf2,
1405 nearbyintsf2): Disable for TARGET_SSE_MATH. Leave patterns enabled
1406 for TARGET_MIX_SSE_I387.
1407 (fmoddf3, dremdf3, *sindf2, *sinextendsfdf2, *cosdf2, *cosextendsfdf2,
1408 sincosdf3, *sincosextendsfdf3, *tandf3_1, tandf2, atan2df3_1,
1409 atan2df3, atandf2, asindf2, acosdf2, logdf2, log10df2, log2df2,
1410 log1pdf2, logbdf2, expdf2, exp10df2, exp2df2, expm1df2, rintdf2,
1411 floordf2, ceildf2, btruncdf2, nearbyintdf2): Disable for
1412 (TARGET_SSE2 && TARGET_SSE_MATH). Leave patterns enabled for
1413 TARGET_MIX_SSE_I387.
1414 (atan2sf3, atan2df3, atan2xf3): Remove register constraints
1417 2005-01-18 Uros Bizjak <uros@kss-loka.si>
1420 * config/i386/mmx.md (*movv2sf_internal_rex64, *movv2sf_internal):
1421 Add movaps alternative for xmm reg->reg move.
1423 2005-01-17 Jeff Law <law@redhat.com>
1425 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Clear SSA_NAME_VALUE
1426 at the end of the main DOM loop rather than just before DOM exits.
1428 2005-01-17 Diego Novillo <dnovillo@redhat.com>
1430 PR tree-optimization/19121
1431 * tree-ssa-alias.c (compute_flow_sensitive_aliasing): When
1432 adding aliases to a name tag, also add them to the pointer's
1434 * tree-ssa-copy.c (merge_alias_info): Do not merge flow
1435 sensitive alias info at all. Only check that the two pointers
1436 have compatible pointed-to sets.
1437 * tree-ssa.c (verify_name_tags): Verify that the alias set of
1438 a pointer's type tag is a superset of the alias set of the
1441 2005-01-17 James E Wilson <wilson@specifixinc.com>
1444 * config/ia64/ia64.md (movxf): Handle general register source. Adjust
1445 comment to document why.
1447 2005-01-17 Richard Henderson <rth@redhat.com>
1449 * config/i386/sse.md (smaxv4sf3_finite, sse_vmsmaxv4sf3_finite,
1450 sminv4sf3_finite, sse_vmsminv4sf3_finite, smaxv2df3_finite,
1451 sse2_vmsmaxv2df3_finite, sminv2df3_finite,
1452 sse2_vmsminv2df3_finite): New.
1453 (smaxv4sf3, sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, smaxv2df3,
1454 sse2_vmsmaxv2df3, sminv2df3, sse2_vmsminv2df3): Remove commutative.
1455 Force op1 into register.
1457 2005-01-17 Kelley Cook <kcook@gcc.gnu.org>
1459 * Makefile.in (STAGEMOVESTUFF): Stage all the stamp files to prevent
1460 bubblestrap from always rebuilding build directory.
1462 2005-01-17 Ulrich Weigand <uweigand@de.ibm.com>
1464 * config/s390/s390.md ("*clc"): Remove incorrect '='.
1465 ("*cmpmem_short"): Likewise.
1467 2005-01-17 Steven Bosscher <stevenb@suse.de>
1469 * basic-block.h: Document BB_* flags.
1470 * regrename.c (copyprop_hardreg_forward): Don't use BB_VISITED,
1471 use an sbitmap instead.
1472 * sched-rgn.c (compute_trg_info): Likewise.
1474 2005-01-17 Richard Sandiford <rsandifo@redhat.com>
1476 * config.gcc (mips64*-*-linux*): Set the default abi to n32. Remove
1477 redundant target_cpu_default and extra_parts lines. Use the default
1479 * config/mips/linux64.h (DRIVER_DEFAULT_ABI_SELF_SPEC): Delete.
1480 (SUBTARGET_EXTRA_SPECS): Delete.
1481 (DRIVER_SELF_SPECS): Remove architecture lines. Set the ABI directly.
1482 (SUBTARGET_ASM_SPEC): Remove -mabi=n32 and -mabi=64 mappings.
1483 * config/mips/t-linux64 (MULTILIB_DIRNAMES): Use "32" for -mabi=32
1484 and "n32" for -mabi=n32.
1486 2005-01-17 Ian Lance Taylor <ian@airs.com>
1489 * c-typeck.c (build_c_cast): Revert patch of 2003-02-16: don't
1490 fold constant variables into initial values.
1492 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
1494 * tree-cfg.c (tree_can_merge_blocks_p): Reorder two checks.
1496 * tree-cfg.c (tree_forwarder_block_p): Speed up by walking
1497 through the statements backward.
1499 2005-01-17 Ian Lance Taylor <ian@airs.com>
1501 PR middle-end/13127:
1502 * tree-inline.c (expand_call_inline): Set TREE_NO_WARNING on
1503 a variable set to the return value of the inlined function.
1505 2005-01-17 Mark Dettinger <dettinge@de.ibm.com>
1507 * rtlanal.c (rtx_cost): Assign cost of 0 to a SUBREG
1508 when modes are tieable.
1510 2005-01-17 Ranjit Mathew <rmathew@hotmail.com>
1512 * gthr-posix.h (__gthread_active_p): Use pthread_cancel instead
1513 of pthread_create to find out if threads are enabled.
1514 * gthr-posix95.h (__gthread_active_p): Likewise.
1516 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
1518 * common.opt (-fnew-ra): Remove.
1520 * toplev.h (flag_new_regalloc): Remove.
1521 * Makefile.in (ra*.*): Don't mention.
1522 * passes.c (rest_of_handle_new_regalloc): Remove.
1523 (rest_of_handle_combine, rest_of_compilation): Always consider
1524 flag_new_regalloc as false.
1525 * doc/invoke.texi: Don't document -fnew-ra.
1527 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
1529 * bb-reorder.c (fix_edges_for_rarely_executed_code): Remove
1530 last parameter to reg_scan.
1531 * loop.c (loop_optimize): Likewise.
1532 * passes.c (rest_of_handle_tracer, rest_of_handle_if_conversion,
1533 rest_of_handle_web, rest_of_handle_cfg, rest_of_handle_jump_bypass,
1534 rest_of_handle_life, rest_of_handle_cse, rest_of_handle_cse2,
1535 rest_of_handle_gcse, rest_of_handle_loop_optimize,
1536 rest_of_handle_loop2, rest_of_handle_jump2): Likewise.
1537 * regclass.c (reg_scan): Likewise, for the declaration.
1538 * rtl.h (reg_scan): Likewise, for the prototype.
1540 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
1542 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Don't call
1543 mark_dfs_back_edges.
1545 2005-01-16 Daniel Berlin <dberlin@dberlin.org>
1547 * tree-ssa-pre.c (add_to_sets): s1 may be NULL.
1548 (compute_avail): Uses don't go in tmp_gen.
1550 2005-01-16 Steven Bosscher <stevenb@suse.de>
1552 * ggc-page.c (ggc_alloc_stat): Use __builtin_ctzl instead of a
1553 loop to look for a free slot in a page entry.
1555 2005-01-16 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1558 * defaults.h (TARGET_DEFERRED_OUTPUT_DEFS): Provide default.
1559 * toplev.c (compile_file): Call process_pending_assemble_output_defs
1560 just before targetm.asm_out.file_end.
1561 * tree.h (process_pending_assemble_output_defs): Declare.
1562 * varasm.c (assemble_output_def, process_pending_assemble_output_defs):
1564 (assemble_alias): Defer generation of assembly code for defines when
1565 TARGET_DEFERRED_OUTPUT_DEFS is true.
1566 * config/rs6000/aix41.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
1567 * config/rs6000/aix43.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
1568 * doc/tm.texi (TARGET_DEFERRED_OUTPUT_DEFS): document.
1570 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1573 * pa.c (pa_scalar_mode_supported_p): New function.
1574 (TARGET_SCALAR_MODE_SUPPORTED_P): Define.
1575 * pa.h (MIN_UNITS_PER_WORD): Add comment.
1577 * pa-protos.h (prefetch_operand): Delete.
1578 (prefetch_cc_operand, prefetch_nocc_operand): New declations.
1579 * pa.c (prefetch_operand): Delete.
1580 (prefetch_cc_operand, prefetch_nocc_operand): New functions.
1581 * pa.h (EXTRA_CONSTRAINT): Add `W' constraint.
1582 (PREDICATE_CODES): Delete prefetch_operand. Add prefetch_cc_operand
1583 and prefetch_nocc_operand.
1584 * pa.md (prefetch): Rework to avoid reload problems handling short
1585 displacements when a cache control completer needs to be provided.
1586 (prefetch_32, prefetch_64): Delete.
1587 (prefetch_cc, prefetch_nocc): New patterns.
1589 2005-01-15 David Edelsohn <edelsohn@gnu.org>
1591 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Revert previous change.
1593 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1596 * pa.c (pa_asm_output_aligned_common): Warn if specified alignment
1597 exceeds maximum alignment for global common data.
1598 * pa64-hpux.h (MAX_OFILE_ALIGNMENT): Define to 32768.
1599 * som.h (MAX_OFILE_ALIGNMENT): Likewise.
1601 2004-01-15 Roger Sayle <roger@eyesopen.com>
1603 * tree-ssa-dom.c (extract_range_from_cond): Correct condition.
1605 2004-01-15 Roger Sayle <roger@eyesopen.com>
1607 * harg-reg-set.h (reg_class_names): Prototype global array.
1608 * regclass.c (reg_class_names): Declare here and initialize to
1610 (dump_regclass): Remove local declaration of reg_class_names.
1611 (regclass): Likewise.
1612 * cfg.c (dump_flow_info): Likewise.
1613 * ra-debug.c (reg_class_names): Likewise.
1614 * regrename.c (reg_class_names): Likewise.
1615 * reload.c (reg_class_names): Likewise.
1616 * reload1.c (spill_failure): Likewise.
1617 * config/m68hc11/m68hc11.c (reg_class_names): Likewise.
1619 2005-01-15 Ulrich Weigand <uweigand@de.ibm.com>
1621 * config/s390/s390.md ("reload_outti"): Remove predicate for
1622 output operand. Abort if operand is not a MEM.
1623 ("reload_outdi", "reload_outdf"): Likewise.
1625 2005-01-15 Marc Espie <espie@openbsd.org>
1627 * config.gcc (*-*-openbsd*): Set HAS_LIBC_R for OpenBSD <= 3.2.
1628 (i[34567]86-*-openbsd*): Switch to ELF for OpenBSD >= 3.4.
1629 * config/openbsd.h: Fix C++ includes for native configurations.
1630 Add proper OS_CPP_BUILTINS. Fix libspec for recent OpenBSD.
1631 Add trampoline support.
1632 * config/i386/openbsdelf.h: New.
1634 2005-01-15 Marc Espie <espie@openbsd.org>
1636 * collect2.c (main): Explicitly parse -dynamic-linker option.
1638 2005-01-15 Jakub Jelinek <jakub@redhat.com>
1640 PR tree-optimization/19060
1641 * tree-ssa-dom.c (extract_range_from_cond) <case LT_EXPR, GT_EXPR>:
1642 Return 0 if op1 <= TYPE_MIN_VALUE () resp. op1 >= TYPE_MAX_VALUE ().
1643 (simplify_cond_and_lookup_avail_expr): Add assert for dummy == 0
1644 and handle extract_range_from_cond returning false.
1645 * fold-const.c (fold): Optimize comparisons with min/max even for
1646 width > HOST_BITS_PER_WIDE_INT.
1648 2005-01-15 Ralf Corsepius <ralf.corsepius@rtems.org>
1650 * config/mips/rtems.h (MIPS_DEFAULT_GVALUE): Set to 0.
1651 * config/mips/t-rtems (MULTILIBS_DIRNAMES,MULTILIB_OPTIONS):
1652 Remove little endian multilib variants.
1653 Add mips32 multilib variant.
1655 2005-01-15 Kazu Hirata <kazu@cs.umass.edu>
1657 * explow.c (copy_all_reg, stabilize): Remove.
1658 * expr.h: Remove the corresponding prototypes.
1660 * rtlanal.c (regs_set_between_p): Remove.
1661 * rtl.h: Remove the corresponding prototype.
1663 2005-01-14 David Edelsohn <edelsohn@gnu.org>
1665 * config/rs6000/rs6000.c (rs6k_nonimmediate_operand): Rename
1666 to rs6000_nonimmediate_operand.
1667 * config/rs6000/rs6000.h (PREDICATE_CODES): Rename
1668 rs6k_nonimmediate_operand to rs6000_nonimmediate_operand.
1669 * config/rs6000/rs6000.md (movsi_internal1): Rename predicate to
1670 rs6000_nonimmediate_operand.
1671 * config/rs6000/spe.md (movdf_e500_double): Rename predicate to
1672 rs6000_nonimmediate_operand.
1674 2005-01-14 Andrew Pinski <pinskia@physics.uc.edu>
1676 * tree-ssa-alias.c (pass_may_alias): Add TODO_verify_stmts.
1678 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1680 * config/rs6000/rs6000.h (CLASS_MAX_NREGS): DF goes in 1 register
1682 (CANNOT_CHANGE_MODE_CLASS): Restrict DI mode changes on e500v2.
1683 (PREDICATE_CODES): Add rs6k_nonimmediate_operand.
1685 * config/rs6000/rs6000.c (invalid_e500_subreg): New.
1686 (rs6k_nonimmediate_operand): New.
1687 (rs6000_legitimate_offset_address_p): Handle DI modes on e500v2
1689 (legitimate_lo_sum_address_p): Same.
1690 (rs6000_legitimize_address): Same.
1691 (rs6000_legitimize_reload_address): Same.
1692 (rs6000_legitimate_address): Same.
1693 (spe_build_register_parallel): Pass DF and DC modes in a DI
1696 * config/rs6000/rs6000.md ("*movsi_internal1"): Change predicate
1697 to rs6k_nonimmediate_operand.
1699 * config/rs6000/spe.md ("*frob_df_di"): New.
1700 ("*frob_di_df"): New.
1701 ("*frob_di_df_2"): New.
1702 ("*mov_sidf_e500_subreg0"): New.
1703 ("*mov_sidf_e500_subreg4"): New.
1704 ("*movdf_e500_double"): Change predicate to
1705 rs6k_nonimmediate_operand.
1707 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1709 * postreload.c (move2add_note_store): Only call
1710 trunc_int_for_mode on scalar integers.
1712 2005-01-14 Steven Bosscher <stevenb@suse.de>
1714 * tree-ssa-dce.c (visited_control_parents): New sbitmap to
1715 replace BB_VISITED uses.
1716 (find_obviously_necessary_stmts): Don't clear BB_VISITED.
1717 (propagate_necessity): Check the bitmap instead of BB_VISITED.
1718 (tree_dce_done): Free visited_control_parents.
1719 (perform_tree_ssa_dce): Allocate and clear it.
1720 * tree-ssa-pre.c (compute_antic_aux): Make non-recursive.
1721 (compute_antic): Iterate from here using a DFS. Use an sbitmap
1722 instead of BB_VISITED.
1724 2005-01-14 Kazu Hirata <kazu@cs.umass.edu>
1726 * c-tree.h, coverage.h, langhooks-def.h, optabs.h, output.h,
1727 rtl.h, targhooks.h, tree-flow.h, tree-mudflap.h, tree.h:
1728 Remove unused prototypes.
1730 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1732 * config/rs6000/rs6000.c (rs6000_stack_info): Fix caching of SPE
1733 64-bit register usage.
1734 (spe_func_has_64bit_regs_p): Add FIXME note.
1736 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1738 * config/rs6000/rs6000.c (rs6000_generate_compare): Replace
1739 flag_finite_math_only with flag_unsafe_math_optimizations.
1741 2005-01-14 J. D. Johnston <jjohnst@us.ibm.com>
1743 * config/s390/tpf-unwind.h (s390_fallback_frame_state): Correct
1745 (__tpf_eh_return): Copy TPF private stack area from the module
1746 boundary stack frame. Add check for when module addresses are
1747 equal, but no stub address is found.
1749 2005-01-14 Richard Earnshaw <rearnsha@arm.com>
1752 * arm.h (struct machine_function): Add call_via field.
1753 (thumb_call_via_label): Declare.
1754 * arm.c (thumb_call_via_label): New variable.
1755 (thumb_call_reg_needed): New variable.
1756 (arm_output_function_epilogue): For Thumb code, output any per-function
1757 call-indirect trampolines.
1758 (thumb_call_via_reg): New function.
1759 (arm_file_end): New function.
1760 (TARGET_ASM_FILE_END): Call arm_file_end.
1761 (aof_file_end): Likewise.
1762 * arm-protos.h (thumb_call_via_reg): Declare.
1763 * arm.md (call_reg_thumb, call_value_reg_thumb): Call
1764 thumb_call_via_reg in normal case.
1766 2005-01-14 Jakub Jelinek <jakub@redhat.com>
1769 PR rtl-optimization/19348
1770 * recog.c (peephole2_optimize): Do global life update if some peephole
1771 decides it doesn't need at least one of its inputs and that change
1772 influences liveness at the start of the basic block.
1774 * basic-block.h (EXECUTE_IF_AND_COMPL_IN_REG_SET): Needs 2 REGSET
1775 arguments instead of 1.
1777 2005-01-14 Eric Botcazou <ebotcazou@libertysurf.fr>
1780 * varasm.c (initializer_constant_valid_p) <ADDR_EXPR>: Return
1781 zero for nested functions needing a static chain or functions
1782 with a non-constant address.
1784 2005-01-13 Roger Sayle <roger@eyesopen.com>
1786 * simplify-rtx.c (simplify_binary_operation) <AND>: Optimize
1787 (and (sign_extend X) C) into (zero_extend (and X C)).
1789 2005-01-13 David O'Brien <obrien@FreeBSD.org>
1791 * config/freebsd-spec.h: Make KSE pthread lib logic the default.
1793 2005-01-13 Richard Henderson <rth@redhat.com>
1798 * config/i386/i386.md (cmpdf, cmpsf, bunordered, bordered, buneq,
1799 bunge, bungt, bunle, bunlt, bltgt): Enable for TARGET_SSE_MATH,
1800 not just TARGET_SSE.
1801 (cmpfp_i_387): Rename from cmpfp_i. Move after sse patterns.
1802 (cmpfp_i_mixed): Rename from cmpfp_i_sse; use for TARGET_MIX_SSE_I387.
1803 (cmpfp_i_sse): Rename from cmpfp_i_sse_only; use for TARGET_SSE_MATH.
1804 (cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387): Similarly.
1805 (fp_jcc_1_mixed, fp_jcc_1_sse, fp_jcc_1_387): Similarly.
1806 (fp_jcc_2_mixed, fp_jcc_2_sse, fp_jcc_2_387): Similarly.
1807 (fp_jcc_3_387, fp_jcc_4_387, fp_jcc_5_387, fp_jcc_6_387,
1808 fp_jcc_7_387, fp_jcc_8_387): Rename from fp_jcc_N.
1809 (movdicc_c_rex64): Rename with '*'.
1810 (movsfcc, movdfcc): Add checks for 387 and sse math to condition.
1811 (movsfcc_1_sse_min, movsfcc_1_sse_max, movsfcc_1_sse): New.
1812 (movsfcc_1_387): Rename from movsfcc_1.
1813 (movdfcc_1_sse_min, movdfcc_1_sse_max, movdfcc_1_sse): New.
1814 (movdfcc_1, movdfcc_1_rex64): Add check for 387.
1815 (sminsf3, smaxsf3, smindf3, smaxdf3): New.
1816 (minsf3, minsf, minsf_nonieee, minsf_sse, mindf3, mindf,
1817 mindf_nonieee, mindf_sse, maxsf3, maxsf, maxsf_nonieee, maxsf_sse,
1818 maxdf3, maxdf, maxdf_nonieee, maxdf_sse, sse_movsfcc, sse_movsfcc_eq,
1819 sse_movdfcc, sse_movdfcc_eq, sse_movsfcc_const0_1,
1820 sse_movsfcc_const0_2, sse_movsfcc_const0_3, sse_movsfcc_const0_4,
1821 sse_movdfcc_const0_1, sse_movdfcc_const0_2, sse_movdfcc_const0_3,
1822 sse_movdfcc_const0_4): Remove.
1823 * config/i386/i386.c (ix86_expand_fp_movcc): For TARGET_SSE_MATH,
1824 recognize min/max early. Update for changed sse cmove patterns.
1825 (ix86_split_sse_movcc): New.
1826 * config/i386/i386-protos.h: Update.
1828 2005-01-13 Steven Bosscher <stevenb@suse.de>
1830 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
1831 (fix_stmt_v_may_defs): Likewise.
1833 2005-01-13 Richard Henderson <rth@redhat.com>
1835 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
1836 for converting the mask.
1838 2005-01-13 David Edelsohn <edelsohn@gnu.org>
1840 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
1843 2005-01-13 Richard Henderson <rth@redhat.com>
1845 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
1846 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
1847 IX86_BUILTIN_MOVDQ2Q): Remove.
1848 (IX86_BUILTIN_VEC_EXT_V4SI): New.
1849 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
1850 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
1852 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
1853 (sse2_pextrw): Fix immediate constraint.
1854 (sse2_loadq, sse2_loadq_rex64): Remove.
1855 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
1856 Use __builtin_ia32_vec_ext_<size>.
1857 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
1859 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
1861 * function.c (assign_parm_setup_block): Look inside original
1862 entry_parm when inspecting PARALLEL.
1864 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
1865 Joel Sherrill <joel@oarcorp.com>
1868 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
1869 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
1870 rtems_gxx_recursive_mutex_init.
1871 (__gthread_recursive_mutex_lock): New function.
1872 (__gthread_recursive_mutex_trylock): Likewise.
1873 (__gthread_recursive_mutex_unlock): Likewise.
1875 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
1877 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
1879 2005-01-13 Jan Beulich <jbeulich@novell.com>
1881 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
1882 (ashrti3_internal): Indicate output is early clobber. Generate result
1883 into output rather than first input. Use move for low word of output
1884 if shift count is exactly 64.
1885 (lshrti3_internal): Likewise.
1887 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
1891 * reload1.c (delete_output_reload): Don't delete an output reload
1892 if the pseudo lives longer than a single basic block. Adjust and
1893 improve wording of comment.
1895 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
1898 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
1899 of TYPE_SIZE where we mean the number of bytes.
1901 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
1903 * function.c (assign_parm_setup_block): Relax condition on
1904 multi-register optimization.
1906 2005-01-12 Nick Clifton <nickc@redhat.com>
1908 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
1909 an "in_delay_slot" attribute of "no" to prevent them being used in
1910 delay slots. This is forbidden because they might generate
1913 2005-01-12 Alan Modra <amodra@bigpond.net.au>
1916 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
1919 2005-01-12 Nick Clifton <nickc@redhat.com>
1921 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
1922 passed on to GAS) but do not define, as GAS no longer supports or
1923 needs the -m2000 option.
1925 2005-01-11 Roger Sayle <roger@eyesopen.com>
1927 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
1928 type as args of unprototyped builtins aren't automatically promoted.
1929 (fold_builtin_abs): Likewise.
1931 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1933 * Makefile.in: Set a `build-warn' variable.
1935 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
1937 PR tree-optimization/17949
1938 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
1939 (find_interesting_uses_address): Use it.
1941 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
1943 * regrename.c (kill_value): Handle subreg's that won't simplify.
1945 2005-01-11 Richard Henderson <rth@redhat.com>
1948 * config/i386/i386.h (enum ix86_builtins): Move ...
1949 * config/i386/i386.c: ... here.
1950 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
1951 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
1952 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
1953 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
1954 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
1955 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
1956 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
1957 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
1958 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
1959 IX86_BUILTIN_LOADDDUP): Remove.
1960 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
1961 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
1962 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
1963 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
1964 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
1965 (ix86_init_builtins): Make static.
1966 (ix86_init_mmx_sse_builtins): Update for changed builtins.
1967 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
1968 if all the modes match. Otherwise, fake it.
1969 (get_element_number, ix86_expand_vec_init_builtin,
1970 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
1971 (ix86_expand_builtin): Make static. Update for changed builtins.
1972 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
1973 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
1974 (ix86_expand_vector_init_duplicate): New.
1975 (ix86_expand_vector_init_low_nonzero): New.
1976 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
1977 Split out from ix86_expand_vector_init; handle integer modes.
1978 (ix86_expand_vector_init): Use them.
1979 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
1980 * config/i386/i386-protos.h: Update.
1981 * config/i386/predicates.md (reg_or_0_operand): New.
1982 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
1983 (movv2sf_internal): Likewise. And a splitter to match them all.
1984 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
1985 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
1986 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
1987 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
1988 (mmx_pinsrw): Fix operand ordering.
1989 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
1990 rather than sse_loadss expander.
1991 (movv2df splitter): Similarly.
1992 (sse_loadss, sse_loadlss): Remove.
1993 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
1994 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
1995 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
1996 (sse3_movddup): Rename with '*'.
1997 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
1998 (sse2_loadsd): Remove.
1999 (vec_dupv2df_sse3): Rename from sse3_loadddup.
2000 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
2001 (sse2_pinsrw): Fix argument ordering.
2002 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
2003 (sse2_stored): Remove 'r' destination.
2004 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
2005 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
2006 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
2007 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
2008 vec_extractv16qi, vec_initv16qi): New.
2010 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
2011 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
2012 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
2014 (_mm_load_pd, _mm_store_pd): Use plain dereference.
2015 (_mm_load_si128, _mm_store_si128): Likewise.
2016 (_mm_load1_pd): Use _mm_set1_pd.
2017 (_mm_load_sd): Use _mm_set_sd.
2018 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
2019 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
2020 (_mm_set_epi64): Use _mm_set_epi64x.
2021 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
2022 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
2023 _mm_setr_epi8): Use _mm_set_foo form.
2024 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
2025 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
2026 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
2027 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
2028 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
2029 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
2030 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
2031 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
2032 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
2033 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
2034 (_mm_movedup_pd): Use _mm_shuffle_pd.
2035 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
2036 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
2037 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
2038 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
2040 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
2041 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
2042 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
2043 (_mm_store_ps): Use raw dereference.
2044 (_mm_store1_ps): Use _mm_storeu_ps.
2045 (_mm_storer_ps): Use _mm_store_ps.
2046 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
2047 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
2049 2005-01-11 Stan Shebs <shebs@apple.com>
2051 * config/rs6000/rs6000.c (machopic_output_stub): Issue
2052 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
2054 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
2057 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
2058 Skip all DECLs except for FIELD_DECLs.
2060 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
2062 * config/s390/s390.c (override_options): Return error if
2063 -mbackchain, -mpacked-stack and -mhard-float are used together.
2064 (s390_va_start): Remove the backchain && packed-stack special case.
2065 (s390_gimplify_va_arg): Likewise.
2066 * doc/invoke.texi: Remove the ABI incompatibility note.
2068 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
2070 * config/s390/s390.c (struct s390_frame_layout): Remove
2072 (s390_frame_info, s390_emit_prologue): Replace occurrences of
2073 save_backchain_p with TARGET_BACKCHAIN.
2075 2005-01-11 Alan Modra <amodra@bigpond.net.au>
2078 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
2079 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
2080 * function.c (locate_and_pad_parm): Set new field.
2081 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
2082 Tweak where_pad test to include "none". Always set mem align for
2084 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
2085 not sufficient for type.
2086 (assign_parm_setup_block): If stack_parm is zero on entry, always
2087 make a new stack local. Block move old stack parm if necessary
2088 to new aligned stack local.
2089 (assign_parm_setup_stack): Use a block move to handle
2090 potentially misaligned entry_parm.
2091 (assign_parms_unsplit_complex): Specify required alignment when
2092 creating stack local.
2093 * calls.c (compute_argument_addresses): Override alignment of stack
2094 arg calculated from its type with the alignment given by
2095 FUNCTION_ARG_BOUNDARY.
2096 (store_one_arg): Likewise.
2098 2005-01-11 Jan Beulich <jbeulich@novell.com>
2100 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
2101 Change respective itanium_class attribute to ialu.
2102 (shladdp4_internal): New.
2103 * config/ia64/predicates.md (shladd_log2_operand): New.
2105 2005-01-11 Richard Henderson <rth@redhat.com>
2107 * expr.c (store_constructor): Use rtvec_alloc instead of
2108 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
2110 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
2112 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
2114 2005-01-11 Alan Modra <amodra@bigpond.net.au>
2116 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
2117 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
2118 (TARGET_RWRELOC): Define.
2119 (ia64_rwreloc_section_type_flags): Delete.
2120 (ia64_section_type_flags): New function.
2121 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
2122 (TARGET_RWRELOC): Define.
2124 2005-01-10 David Mosberger <davidm@hpl.hp.com>
2127 * config/ia64/ia64.c (process_set): For alloc insn, only call
2128 process_epilogue is !frame_pointer_needed.
2130 2005-01-10 Roger Sayle <roger@eyesopen.com>
2133 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
2134 unary operator and can't be treated as a binary/comparison operator.
2136 2005-01-10 Richard Henderson <rth@redhat.com>
2138 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
2139 (ix86_return_in_memory): Likewise.
2140 (function_arg_advance): Likewise. Mirror structure in function_arg
2141 for choosing register to advance.
2143 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
2145 * tree-vectorizer.c, tree.def: Fix comment typos.
2147 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
2149 * config/s390/s390.c (struct processor_costs): 4 new fields:
2150 ddbr, ddr, debr, der.
2151 (s390_rtx_costs): More precise handling of divide instructions.
2153 2005-01-10 David Edelsohn <edelsohn@gnu.org>
2155 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
2157 2005-01-10 Jan Beulich <jbeulich@novell.com>
2159 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
2160 names resulting from -ffunction-sections/-fdata-sections and linkonce
2162 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
2163 based on the section name. Rearrange the section name comparison logic
2164 slightly so that each section name is compared against at most once.
2166 2005-01-10 Ben Elliston <bje@au.ibm.com>
2168 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
2169 targets for which -fPIC reduces limitations on the GOT size.
2171 2005-01-09 Falk Hueffner <falk@debian.org>
2173 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
2176 2005-01-09 David Edelsohn <edelsohn@gnu.org>
2179 * collect2.c (main): Set aixrtl_flag for -brtl option.
2180 (resolve_lib_name): Search for .so file extension before .a
2183 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
2185 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
2186 (vect_analyze_data_refs_alignment): Add dump prints.
2188 2005-01-09 Ira Rosen <irar@il.ibm.com>
2190 * tree-vectorizer.c (vect_analyze_offset_expr): Use
2191 expr_invariant_in_loop_p.
2192 Initialize outputs first thing in the function.
2193 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
2194 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
2195 (vect_analyze_pointer_ref_access): Check that the initial condition of
2196 the access function is loop invariant.
2198 2005-01-09 Richard Henderson <rth@redhat.com>
2200 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
2201 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
2202 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
2203 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
2204 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
2205 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
2206 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
2207 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
2208 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
2209 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
2210 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
2211 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
2212 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
2213 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
2214 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
2215 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
2216 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
2217 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
2218 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
2219 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
2220 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
2221 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
2222 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
2224 (mmx_clrdi, pavgusb): Remove.
2225 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
2226 with leading sse_ prefix.
2227 * config/i386/sse.md: Receive them.
2228 * config/i386/mmx.md: New file.
2229 (MMXMODE12, MMXMODE24, mmxvecsize): New.
2230 (subrv2sf3): Turn into expander for normal subtraction.
2231 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
2232 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
2233 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
2234 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
2235 commutative; use ix86_binary_operator_ok.
2236 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
2237 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
2238 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
2239 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
2240 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
2241 Macroize from existing patterns; use ix86_binary_operator_ok.
2242 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
2243 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
2244 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
2245 with vec_select+vec_concat.
2246 (mmx_pshufw, mmx_pshufw_1): Likewise.
2247 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
2248 (mmx_uavgv4hi3): Model correctly.
2249 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
2250 _mm_xor_si64): Remove casts.
2252 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
2254 PR tree-optimization/19224
2255 * tree-scalar-evolution.c (get_instantiated_value,
2256 set_instantiated_value): New functions.
2257 (instantiate_parameters_1): Cache the results.
2258 (instantiate_parameters, resolve_mixers): Initialize and free
2261 2005-01-08 David Edelsohn <edelsohn@gnu.org>
2263 * config/i386/i386.md (addhi_4): Correct reference in comment.
2266 2005-01-08 Richard Henderson <rth@redhat.com>
2268 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
2269 (_mm_cvtsi128_si64x): Likewise.
2270 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
2271 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
2272 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
2273 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
2274 * config/i386/i386-protos.h: Update.
2275 * config/i386/i386.c (print_operand): Add 'H'.
2276 (ix86_fixup_binary_operands): Split out from ...
2277 (ix86_expand_binary_operator): ... here.
2278 (ix86_fixup_binary_operands_no_copy): New.
2279 (ix86_expand_fp_absneg_operator): Handle vector mode results.
2280 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
2281 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
2282 (safe_vector_operand): Use CONST0_RTX.
2283 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
2284 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
2285 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
2286 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
2288 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
2289 (IX86_BUILTIN_CMPNGESS): New.
2290 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
2291 (attr type): Add sselog1.
2292 (attr unit, attr memory): Handle it.
2293 (movti, movti_internal, movti_rex64): Move near other integer moves.
2294 (movtf, movtf_internal): Move near other fp moves.
2295 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
2296 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
2297 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
2298 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
2299 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
2300 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
2301 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
2302 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
2303 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
2304 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
2305 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
2306 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
2307 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
2308 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
2309 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
2310 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
2311 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
2312 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
2313 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
2314 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
2315 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
2316 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
2317 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
2318 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
2319 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
2320 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
2321 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
2322 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
2323 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
2324 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
2325 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
2326 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
2327 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
2328 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
2329 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
2330 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
2331 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
2332 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
2333 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
2334 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
2335 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
2336 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
2337 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
2338 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
2340 (SSEPUSH, push<SSEPUSH>): Remove.
2341 (MMXPUSH, push<MMXPUSH>): Remove.
2342 (sse_movaps, sse_movaps_1, sse_movups): Remove.
2343 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
2344 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
2345 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
2346 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
2347 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
2348 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
2349 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
2350 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
2351 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
2352 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
2353 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
2354 (ppro_sse_log_V4SF): Handle sselog1.
2355 * config/i386/predicates.md (const_0_to_1_operand): New.
2356 (const_0_to_255_mul_8_operand): New.
2357 (const_1_to_31_operand): Rename from const_int_1_31_operand.
2358 (const_2_to_3_operand, const_4_to_7_operand): New.
2359 * config/i386/sse.md: New file.
2360 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
2361 (sse_movups): Rename from sse_movups_1.
2362 (sse_loadlss): Rename from sse_loadss_1.
2363 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
2365 (negv4sf2): Use ix86_expand_fp_absneg_operator.
2366 (absv4sf2, negv2df, absv2df): New.
2367 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
2368 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
2369 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
2370 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
2371 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
2372 (sse3_addsubv4sf3): Model correctly.
2373 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
2374 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
2375 (sse_movhlps): Model with vec_select+vec_concat.
2376 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
2377 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
2378 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
2379 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
2380 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
2381 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
2382 sse2_pshufhw_1): Likewise.
2383 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
2384 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
2385 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
2386 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
2387 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
2388 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
2389 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
2390 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
2391 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
2392 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
2393 sminv8hi3): Mark commutative
2394 operands. Use ix86_binary_operator_ok.
2395 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
2396 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
2397 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
2398 sse2_punpcklqdq): Allow operand2 in memory.
2399 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
2400 sse2_movsd): Add memory alternatives.
2401 (sse_storelps): Turn expander into an insn; split after reload.
2402 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
2403 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
2405 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
2407 * configure.ac (DWARF-2 debug_line): Use objdump.
2408 * configure: Regenerate.
2410 2005-01-08 Jeff Law <law@redhat.com>
2411 Diego Novillo <dnovillo@redhat.com>
2413 PR tree-optimization/18241
2414 * tree-nrv.c (tree_nrv): Ignore volatile return values.
2415 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
2416 statements with volatile operands.
2417 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
2418 operands after marking a statement with has_volatile_ops.
2420 2005-01-08 Roger Sayle <roger@eyesopen.com>
2422 * tree.c (int_fits_type_p): Always honor integer constant
2423 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
2425 2005-01-08 Roger Sayle <roger@eyesopen.com>
2427 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
2428 the latent bug in rtl_delete_block has been resolved.
2430 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
2432 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
2433 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
2435 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
2437 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
2438 library as @multilib_dir@/$(SHLIB_SONAME).
2440 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
2442 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
2443 * configure: Regenerate.
2445 2005-01-07 Jakub Jelinek <jakub@redhat.com>
2447 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
2450 2005-01-07 David Edelsohn <edelsohn@gnu.org>
2453 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
2454 Convert non-word aligned offset address using ld/std into
2457 2005-01-07 Richard Henderson <rth@redhat.com>
2459 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
2461 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
2463 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
2465 2005-01-07 Jakub Jelinek <jakub@redhat.com>
2467 PR tree-optimization/19283
2468 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
2469 is not shorter than the original type.
2471 PR rtl-optimization/19012
2472 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
2474 PR rtl-optimization/18861
2475 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
2476 (FULL_STATE): Define.
2477 (mark_tablejump_edge): New function.
2478 (purge_dead_tablejump_edges): New function.
2479 (find_bb_boundaries): Use it.
2481 PR tree-optimization/18828
2482 * builtins.c (expand_builtin_next_arg): Remove argument and all
2483 the argument checking.
2484 (expand_builtin): Adjust caller.
2485 (expand_builtin_va_start): Likewise. Remove error for too many
2487 (fold_builtin_next_arg): Issue error for too many arguments.
2488 After checking arguments, replace them with magic arguments that
2489 prevent further checking of the args.
2491 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2493 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
2495 2005-01-06 Roger Sayle <roger@eyesopen.com>
2498 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
2499 and %{.S:...} (and their negative variants) to test whether the
2500 input file is assembler or pre-processed-assembler independent of
2501 the actual filename extension.
2503 2005-01-06 Roger Sayle <roger@eyesopen.com>
2505 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
2506 of sign or zero extended values.
2508 2005-01-06 Geoffrey Keating <geoffk@apple.com>
2510 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
2511 for IBM long double format correctly.
2513 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
2515 Fix PR tree-optimization/18792
2517 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
2518 to first_loop_depth, and use loop depth instead of loop number.
2519 (build_classic_dir_vector): Ditto.
2520 (compute_data_dependences_for_loop): Use depth, not loop number.
2521 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
2522 number. Pass in loops, instead of loop numbers.
2523 (gather_interchange_stats): Ditto.
2524 (linear_transform_loops): Ditto.
2526 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
2529 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
2530 biv_fits_mode_p, extension_within_bounds_p): New functions.
2531 (check_ext_dependent_givs): Use them.
2533 2005-01-06 Roger Sayle <roger@eyesopen.com>
2535 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
2536 more than one barrier, in which case we should delete them all.
2538 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
2540 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
2542 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
2544 Fix PR middle-end/19286
2546 * dwarf2out.c (gen_subprogram_die): If we've already tried to
2547 output this subprogram, simply ignore this attempt to do it again.
2548 (add_abstract_origin_attribute): Don't abort trying to add the abstract
2549 origin attribute if it's not possible.
2550 (gen_block_die): Don't ignore subblocks of "unused" blocks.
2551 (decls_for_scope): Ditto.
2552 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
2553 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
2555 2005-01-05 Richard Henderson <rth@redhat.com>
2558 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
2559 (bdesc_2arg): Use it.
2560 (ix86_expand_binop_builtin): Force operands into registers
2562 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
2563 ix86_expand_sse_compare, ix86_expand_sse_comi,
2564 ix86_expand_builtin): Likewise.
2566 2005-01-05 Richard Henderson <rth@redhat.com>
2568 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
2570 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
2571 (mulv4hi3): Set itanium_class mmmul.
2572 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
2573 (fpack): Rename from fpack_sfsf.
2575 2005-01-05 Richard Henderson <rth@redhat.com>
2578 * reload1.c (do_input_reload): Restrict the optimization deleteing
2579 a previous output reload to RELOAD_FOR_INPUT.
2581 2005-01-05 Steven Bosscher <stevenb@suse.de>
2583 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
2584 comment that we fall through after case.
2585 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
2586 (distribute_links): Likewise.
2587 * cse.c (cse_insn): Likewise.
2588 * cselib.c (cselib_invalidate_mem): Likewise.
2589 * df.c: Update comments at the top of the file.
2590 (read_modify_subreg_p): Update comments here too.
2591 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
2592 * flow.c (mark_set_1): Likewise.
2593 (mark_used_regs): Likewise.
2594 * gcse.c (mems_conflict_for_gcse_p): Likewise.
2595 (canon_list_insert): Likewise.
2596 (mark_set): Likewise.
2597 (try_replace_reg): Likewise.
2598 (store_killed_in_insn): Likewise.
2599 * loop.c (count_one_set): Likewise.
2600 (basic_induction_var): Likewise.
2601 * postreload-gcse.c (find_mem_conflicts): Likewise.
2602 * postreload.c (reload_combine_note_store): Likewise.
2603 (move2add_note_store): Likewise.
2604 * reload.c (find_equiv_reg): Likewise.
2605 (mark_referenced_resources): Likewise.
2606 * rtlanal.c (set_noop_p): Likewise.
2607 (note_stores): Likewise.
2608 (note_uses): Likewise.
2609 * sched-deps.c (sched_analyze_1): Likewise.
2610 * sched-rgn.c (check_live_1): Likewise.
2611 (update_live_1): Likewise.
2612 * config/i860/i860.c: Likewise.
2614 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
2616 (ZERO_EXTRACT): Mention that this one can be an lvalue.
2618 * doc/rtl.texi: Update documentation for bit-fields and SET.
2620 2005-01-05 Roger Sayle <roger@eyesopen.com>
2622 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
2624 2005-01-05 Jan Hubicka <jh@suse.cz>
2625 Richard Henderson <rth@redhat.com>
2628 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
2631 2005-01-05 Richard Henderson <rth@redhat.com>
2634 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
2635 (sse_shufps): Change operand 3 to const_int_operand.
2636 (sse2_storelps): Fix typo in template.
2637 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
2638 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
2639 (ix86_expand_builtin): Likewise.
2641 2005-01-05 Julian Brown <julian@codesourcery.com>
2643 * config/arm/arm.c (arm_return_in_memory): Treat complex types
2644 as aggregates for AAPCS ABIs.
2646 2005-01-05 Stan Shebs <shebs@apple.com>
2648 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
2649 Mach-O getters if ppc64.
2650 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
2652 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
2653 signed four-byte field for global code case.
2654 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
2656 2005-01-05 Roger Sayle <roger@eyesopen.com>
2659 * c-common.c: Include real.h.
2660 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
2661 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
2662 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
2663 * Makefile.in (c-common.o): Update dependencies.
2665 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
2667 * c-parse.in (asm_string): Add trailing semicolon.
2669 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
2671 * c-parse.in (asm_string): New. Don't allow wide strings in
2673 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
2674 asm_string instead of STRING.
2676 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
2678 * c-typeck.c (constructor_no_implicit): Remove.
2679 (set_designator, process_init_element): Don't check
2680 constructor_no_implicit.
2682 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
2683 Kaz Kojima <kkojima@gcc.gnu.org>
2686 * lcm.c (create_pre_exit): New.
2687 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
2688 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
2690 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
2692 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
2693 -msoft-float in target_float_switch.
2694 * arm.c (arm_override_options): Fix processing of target_float_switch.
2696 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
2698 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
2701 2004-01-05 Uros Bizjak <uros@kss-loka.si>
2703 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
2704 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
2706 2005-01-04 Roger Sayle <roger@eyesopen.com>
2708 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
2709 if deleting the then-block allows the test-block to fallthru to the
2712 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
2715 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
2716 after the full declaration if the are in two different TUs.
2718 2005-01-04 Richard Henderson <rth@redhat.com>
2721 * tree-sra.c (generate_one_element_init): Just
2722 call gimplify_and_add.
2723 (generate_element_init): Record the
2724 new referenced variables and mark them for renaming
2725 and split out to ...
2726 (generate_element_init_1): This.
2727 (scalarize_init): Don't call push_gimplify_context/
2728 pop_gimplify_context.
2730 2005-01-04 Geoffrey Keating <geoffk@apple.com>
2732 * toplev.c (get_src_pwd): Handle failure of getpwd().
2734 2005-01-04 Roger Sayle <roger@eyesopen.com>
2736 * fold-const.c (fold_single_bit_test): Delete unreachable handling
2738 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
2739 all the cases handled by it are inverted by invert_truthvalue.
2741 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
2743 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
2744 (save_register_windowsi): Likewise.
2746 2005-01-04 Richard Henderson <rth@redhat.com>
2748 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
2750 2005-01-04 Uros Bizjak <uros@kss-loka.si>
2753 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
2754 any simplifications of vector mode comparison operators.
2755 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
2757 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
2758 Devang Patel <dpatel@apple.com>
2760 PR tree-optimization/18308
2761 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
2762 the operands before creating a new expression.
2763 * dojump.c (do_jump): Make drop_through_label available
2764 for all cases. Add expansion of COND_EXPR.
2766 2005-01-04 Ira Rosen <irar@il.ibm.com>
2768 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
2769 INTEGER_CST instead of TREE_CONSTANT.
2770 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
2771 instead of TREE_CONSTANT.
2772 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
2773 instead of TREE_CONSTANT.
2775 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
2777 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
2779 2005-01-04 Uros Bizjak <uros@kss-loka.si>
2782 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
2783 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
2785 2005-01-03 Richard Henderson <rth@redhat.com>
2787 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
2788 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
2789 fold_convert_const_real_from_real): Split out from ...
2790 (fold_convert_const): ... here.
2792 2005-01-03 Richard Henderson <rth@redhat.com>
2795 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
2796 (mov<MMXMODEI>_internal): Likewise.
2797 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
2798 Add V2SF case; use it for SSE1; don't use TI.
2799 (movdf_integer): Likewise.
2800 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
2802 2005-01-03 Ira Rosen <irar@il.ibm.com>
2804 * tree-vectorizer.c (vect_strip_conversions): New function.
2805 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
2806 check for binary class.
2808 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
2812 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
2814 * gimple-low.c (mark_blocks_with_used_vars): New function.
2815 (mark_blocks_with_used_subblocks): Ditto.
2816 (mark_used_blocks): Ditto.
2817 (pass_mark_used_blocks): New pass.
2818 * tree-inline.c: Include debug.h.
2819 (expand_call_inline): Call outlining_inline_function here.
2820 * tree-optimize.c (init_tree_optimization_passes): Add
2821 pass_mark_used_blocks.
2822 * tree-pass.h (pass_mark_used_blocks): New.
2823 * Makefile.in (tree-inline.o): Add debug.h dependency.
2825 2005-01-03 Geoffrey Keating <geoffk@apple.com>
2827 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
2828 VAR_DECLs to have weak_import attribute.
2830 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
2833 2004-01-03 Steven Bosscher <stevenb@suse.de>
2835 * passes.c (rest_of_compilation): Don't run regmove if only
2836 flag_expensive_optimizations, require flag_regmove instead.
2838 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
2840 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
2842 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
2843 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
2844 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
2845 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
2847 2005-01-03 Richard Henderson <rth@redhat.com>
2849 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
2850 (ia64_const_ok_for_letter_p): New.
2851 (ia64_const_double_ok_for_letter_p): New.
2852 (ia64_extra_constraint): New.
2853 (ia64_expand_vecint_compare): New.
2854 (ia64_expand_vcondu_v2si): New.
2855 (ia64_expand_vecint_cmov): New.
2856 (ia64_expand_vecint_minmax): New.
2857 (ia64_print_operand): Add 'v'.
2858 (ia64_preferred_reload_class): New.
2859 (ia64_vector_mode_supported_p): New.
2860 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
2861 (PREFERRED_RELOAD_CLASS): Move to function.
2862 (CONST_OK_FOR_LETTER_P): Move to function.
2863 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
2864 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
2865 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
2866 (EXTRA_CONSTRAINT): Move to function.
2867 * config/ia64/ia64.md: Include vect.md.
2868 (itanium_class): Add mmalua.
2870 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
2872 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
2873 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
2875 (const_int_2bit_operand): New.
2876 (fr_reg_or_0_operand): New.
2877 * config/ia64/ia64-modes.def: Add vector modes.
2878 * config/ia64/ia64-protos.h: Update.
2879 * config/ia64/vect.md: New file.
2881 2005-01-03 Richard Henderson <rth@redhat.com>
2883 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
2885 2005-01-03 Uros Bizjak <uros@kss-loka.si>
2888 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
2890 (log1pdf2): Change mode of operands[0,1] to DFmode.
2892 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
2894 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
2895 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
2896 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
2897 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
2898 SPARC_RELAXED_ORDERING.
2900 2005-01-03 Richard Henderson <rth@redhat.com>
2901 Uros Bizjak <uros@kss-loka.si>
2904 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
2905 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
2906 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
2908 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
2909 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
2910 const_pow2_1_to_128_operand): New predicates.
2912 2005-01-02 Greg McGary <greg@mcgary.org>
2914 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
2916 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
2917 object. [ptr_type]: Remove unused variable.
2918 Remember nearest addressable array-element or record-component
2919 when checking bitfield components. Tolerate empty BIND_EXPR.
2921 2005-01-01 Richard Henderson <rth@redhat.com>
2923 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
2924 that don't narrow the value. Fail for other conversions.
2926 2005-01-01 Richard Henderson <rth@redhat.com>
2929 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
2930 * c-lang.c (finish_file): Don't do it here.
2931 * objc/objc-act.c (objc_finish_file): Likewise.
2933 * cgraph.c (decl_assembler_name_equal): New.
2934 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
2935 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
2936 * cgraph.h (struct cgraph_varpool_node): Add next.
2937 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
2938 * varasm.c (assemble_alias): Mark the target as needed.
2940 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
2943 * function.c (get_arg_pointer_save_area): Use entry_of_function
2944 instead of get_insns.
2946 2005-01-01 Roger Sayle <roger@eyesopen.com>
2947 Andrew Pinski <pinskia@physics.uc.edu>
2948 James E. Wilson <wilson@specifixinc.com>
2950 PR rtl-optimization/12092
2951 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
2954 2005-01-01 Roger Sayle <roger@eyesopen.com>
2955 Olivier Hainque <hainque@act-europe.fr>
2957 * tree.c (int_fits_type_p): A narrower type always fits in a
2958 wider one, except for negative values into unsigned types.
2960 2005-01-01 Roger Sayle <roger@eyesopen.com>
2962 * tree.c (int_fits_type_p): Compare the result of force_fit_type
2963 with the original constant rather than require TREE_OVERFLOW.
2965 2005-01-01 Steven Bosscher <stevenb@suse.de>
2968 * c-decl.c (finish_function): If compiling C99, annotate the
2969 compiler generated return with the current file name and line 0.
2970 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
2971 the source line is greater than 0.
2972 (remove_bb): Likewise.
2974 See ChangeLog.12 for earlier changes.