1 2005-02-01 Kazu Hirata <kazu@cs.umass.edu>
3 * cse.c (cse_reg_info): Remove hash_next, next, regno. Add
5 (cse_reg_info_list, cse_reg_info_list_free, REGHASH_SHIFT,
6 REGHASH_SIZE, REGHASH_MASK, reg_hash, REGHASH_FN,
7 cached_cse_reg_info, GET_CSE_REG_INFO): Remove.
8 (cached_regno): Initialize to INVALID_REGNUM.
9 (cse_reg_info_table_size,
10 cse_reg_info_table_first_uninitialized,
11 cse_reg_info_timestamp): New.
12 (REG_TICK, REG_IN_TABLE, SUBREG_TICKED, REG_QTY): Use
14 (init_cse_reg_info, get_cse_reg_info_1): New.
15 (get_cse_reg_info): Cache the last look-up.
16 (new_basic_block): Update the code to clear mappings from
17 registers to cse_reg_info entries.
18 (cse_main): Call init_cse_reg_info.
20 * cse.c (get_cse_reg_info): Update a comment.
22 2005-01-31 Steven Bosscher <stevenb@suse.de>
25 * c-decl.c (start_decl): Do not warn about arrays of elements with
26 an incomplete type here.
27 (grokdeclarator): Do it here by making a pedwarn an error.
28 * c-typeck.c (push_init_level): If there were previous errors with
29 the constructor type, do not warn about braces for initializers.
30 (process_init_element): Likewise for excess initializer elements.
32 2005-01-31 Kazu Hirata <kazu@cs.umass.edu>
34 * cse.c (delete_trivially_dead_insn): Don't iterate.
36 2005-01-31 Andrew Pinski <pinskia@physics.uc.edu>
38 * config/rs6000/rs6000.md (copysignsf3): New expand.
39 (copysigndf3): Likewise.
41 2005-01-31 Steven Bosscher <stevenb@suse.de>
43 * recog.c (constrain_operands): Only look for earlyclobber operand
44 conflicts if an '&' constraint was seen.
46 2005-01-31 Marc Espie <espie@openbsd.org>
48 * config.gcc: Don't include embedded systems fragment, switches default
49 debugging format to ELF.
50 * config/i386/openbsdelf.h: Add DBX_REGISTER_NUMBER since we no
51 longer pick it up there.
53 2005-01-31 Dale Johannesen <dalej@apple.com>
55 * doc/extend.texi (nested functions): Fix linkage description.
56 Clarify that static is not allowed.
58 2005-01-31 Dale Johannesen <dalej@apple.com>
60 * config/rs6000/darwin.md (movsf_low_di): Make work.
61 (movdf_low_di): Make work.
63 2005-01-31 Dale Johannesen <dalej@apple.com>
65 * config/rs6000/darwin-tramp.asm (__trampoline_setup):
68 2005-01-31 Roger Sayle <roger@eyesopen.com>
69 Dale Johannesen <dalej@apple.com>
72 * fold-const.c (fold_binary_op_with_conditional_arg):
73 Make types match original operands, before STRIP_NOPS.
75 2005-01-31 Alan Modra <amodra@bigpond.net.au>
77 * config/rs6000/linux-unwind.h (struct gcc_vregs): New.
78 (struct gcc_regs): Rename from gcc_pt_regs. Add more regs.
79 (struct gcc_sigcontext): Delete. Merge contents to..
80 (struct gcc_ucontext): ..here.
81 (get_sigcontext): Delete.
82 (get_regs): New function, like get_sigcontext but return regs ptr.
83 64-bit version finds regs from r1 to support vdso.
84 (ppc_linux_aux_vector): New function.
85 (ppc_fallback_frame_state): Modify for get_regs. Restore fprs
88 2005-01-31 Ira Rosen <irar@il.ibm.com>
90 * tree-vectorizer.c (vect_analyze_offset_expr): Use ssizetype instead
92 (vect_get_base_and_offset): Use ssizetype instead sizetype. Remove
93 redundant fold. Fix misalignment for MINUS_EXPR.
94 (vect_compute_data_ref_alignment): Use ssizetype instead sizetype.
95 (vect_analyze_pointer_ref_access): Likewise.
96 (vect_get_memtag_and_dr): Likewise.
98 2005-01-31 Richard Henderson <rth@redhat.com>
100 * config/i386/i386.c (absnegsf2_mixed): Remove all of the # hackery.
101 (absnegsf2_sse, absnegsf2_i387, absnegdf2_mixed, absnegdf2_sse,
102 absnegdf2_i387, absnegxf2_i387): Likewise.
104 2005-01-31 Richard Henderson <rth@redhat.com>
107 * optabs.c (expand_copysign_absneg): New.
108 (expand_copysign_bit): Split out from ...
109 (expand_copysign): ... here. Use expand_copysign_absneg.
111 2005-01-30 Roger Sayle <roger@eyesopen.com>
114 * config/pa/pa.md (anddi3, iordi3): On HPPA64, disallow an integer
115 constant as the second operand and a register as the third.
117 2005-01-31 Danny Smith <dannysmith@users.sourceforge.net>
120 * config/i386/i386.c (ix86_function_ok_for_sibcall): Also check
121 that dllimport'd functions do not use all call-clobbered registers
124 2005-01-30 Richard Henderson <rth@redhat.com>
127 * optabs.c (expand_absneg_bit): Split out from expand_unop and
128 expand_abs_nojump and generalize. Use operand_subword and
129 emit_no_conflict_block. Support large modes.
130 (expand_abs_nojump): Use it. Use HONOR_SIGNED_ZEROS to fall
132 (expand_unop): Likewise. Use HONOR_SIGNED_ZEROS to fall back
133 negation to subtraction.
134 (expand_copysign): Check that the format has signed zeros.
135 Use operand_subword and emit_no_conflict_block. Support large modes.
137 2005-01-30 Daniel Berlin <dberlin@dberlin.org>
139 Fix PR tree-optimization/19624
141 * Makefile.in (tree-ssa-pre.o): Add CFGLOOP_H.
142 * tree-ssa-pre.c: Add cfgloop.h.
144 (pre_stats): New member, constified.
145 (inserted_exprs): New static variable.
146 (NECESSARY): New macro.
147 (create_expression_by_pieces): Fold the expression, and
148 mark it as defaulting to not necessary. Also put in
150 (fully_constant_expression): New function.
151 (insert_into_preds_of_block): Modify to not insert phis when we
152 are playing with induction variables.
153 Push phis onto the inserted_exprs vector, and mark them as not
154 necessary by default.
155 (insert_aux): Call fully_constant_expression on eprime.
156 If all edges produce the same value, mark it constant.
157 (mark_operand_necessary): New function.
158 (remove_dead_inserted_code): New function.
159 (init_pre): Init loop optimizer to get loop info.
160 (fini_pre): Free loop_optimizer, and inserted_exprs vec.
161 (execute_pre): Commit edge inserts, then remove dead code.
163 2005-01-30 Richard Henderson <rth@redhat.com>
165 * rtl.c (rtx_equal_p): No early exit for CONST_VECTOR.
166 * varasm.c (const_rtx_hash_1): Handle CONST_VECTOR.
168 2005-01-30 Richard Henderson <rth@redhat.com>
171 * config/i386/i386.c (ix86_expand_copysign): New.
172 (ix86_split_copysign_const): New.
173 (ix86_split_copysign_var): Rename from ix86_split_copysign,
174 rearrange op1/nmask operands.
175 * config/i386/i386-protos.h: Update.
176 * config/i386/i386.md (copysignsf3): Use ix86_expand_copysign.
177 (copysigndf3): Likewise.
178 (copysignsf3_const, copysigndf3_const): New.
179 (copysignsf3_var): Rename from copysignsf3, split out splitter
180 and fix split predicate for X constraint.
181 (copysigndf3_var): Similarly.
183 2005-01-30 Kazu Hirata <kazu@cs.umass.edu>
185 * optabs.c, doc/c-tree.texi, doc/install.texi, doc/md.texi,
186 doc/passes.texi, doc/rtl.texi, doc/sourcebuild.texi,
187 doc/tm.texi, doc/tree-ssa.texi: Update copyright.
189 2005-01-29 Richard Henderson <rth@redhat.com>
192 * config/i386/i386.md (movdf_nointeger, movdf_integer): Fix ordering
193 of # and * in constraints.
195 2005-01-29 Richard Henderson <rth@redhat.com>
198 * expr.c (store_field): Don't strip sub-mode cast when the input
199 data is even smaller.
201 2005-01-29 Richard Henderson <rth@redhat.com>
204 * expr.c (categorize_ctor_elements_1): Check for CONSTRUCTOR of a
207 2005-01-29 Richard Henderson <rth@redhat.com>
209 * combine.c (make_field_assignment): Fix argument order
212 2005-01-29 Richard Guenther <richard.guenther@uni-tuebingen.de>
214 PR tree-optimization/15791
215 * fold-const.c (extract_array_ref): New function.
216 (fold): Fold comparisons between &a[i] and &a[j] or
217 semantically equivalent trees.
219 2005-01-29 Jeff Law <law@redhat.com>
221 * gcse.c (insert_expr_in_table): Revamp handling of available
222 and anticipatable occurrence lists to avoid unnecessary list
224 (insert_set_in_table): Similarly.
226 2005-01-29 Joseph S. Myers <joseph@codesourcery.com>
228 * c-common.c (fix_string_type): Just use c_build_qualified_type to
230 (c_build_qualified_type): Build qualified array types with
231 TYPE_MAIN_VARIANT pointing to corresponding unqualified type.
232 * c-decl.c (c_make_fname_decl): Build unqualified array type
233 before qualified type.
234 (grokdeclarator): Use TYPE_MAIN_VARIANT of typedef type if element
235 type is qualified, not just if type itself is. Don't apply
236 qualifiers to array type when declarator is processed. Apply
237 qualifiers to field type whether or not it is an array type.
238 Don't handle array types specially for applying qualifiers to
240 * c-typeck.c (composite_type): Build unqualified element type and
241 array type when forming composite of array types.
242 (common_pointer_type, comptypes, comp_target_types,
243 type_lists_compatible_p, build_indirect_ref, build_array_ref,
244 convert_for_assignment): Don't apply TYPE_MAIN_VARIANT to array
246 (type_lists_compatible_p): Cache TREE_VALUE (args1) and TREE_VALUE
247 (args2) in variables a1 and a2.
249 2005-01-29 Kazu Hirata <kazu@cs.umass.edu>
251 * cse.c (n_elements_made, max_elements_made): Remove.
252 (insert): Don't touch n_elements_made.
253 (cse_main): Don't touch n_elements_made or max_elements_made.
255 2005-01-28 Stephane Carrez <stcarrez@nerim.fr>
258 * config/m68hc11/t-m68hc11-gas (dp-bit.c): Fix typo causing a
259 configuration part of dp-bit.c to be lost.
261 2005-01-28 Roger Sayle <roger@eyesopen.com>
263 * expmed.c (expand_mult_highpart_optab): When attempting to use
264 a non-widening multiplication in a wider mode, the operands need
265 to be converted (zero or sign extended) to that mode.
267 2005-01-28 Ian Lance Taylor <ian@airs.com>
271 * gimple-low.c (block_may_fallthru): TRY_FINALLY_EXPR only falls
272 through if both operands fall through.
274 2005-01-28 Kazu Hirata <kazu@cs.umass.edu>
276 * cse.c (fold_rtx) <PC>: Don't optimize.
278 2005-01-28 Jeff Law <law@redhat.com>
280 * fold-const.c (fold, case CEIL_MOD_EXPR): Optimize 0 % X.
281 (case FLOOR_MOD_EXPR, ROUND_MOD_EXPR, TRUNC_MOD_EXPR): Similarly.
283 2005-01-28 Kazu Hirata <kazu@cs.umass.edu>
285 * cse.c (cse_reg_info_free_list, cse_reg_info_used_list,
286 cse_reg_info_used_list_end): Remove.
287 (cse_reg_info_list, cse_reg_info_list_free): New.
288 (get_cse_reg_info): When allocating an instance of
289 cse_reg_info, add it to the beginning of the cse_reg_info_list
290 list. Remove code to maintain cse_reg_info_used_list.
291 (new_basic_block): Reset the free list to the beginning of
294 2005-01-28 Richard Sandiford <rsandifo@redhat.com>
296 * config/frv/frv.c (bdesc_2arg): Remove __MDPACKH.
297 (frv_init_builtins): Change its prototype to take 4 uhalf arguments.
298 (frv_expand_mdpackh_builtin): New function.
299 (frv_expand_builtin): Use it to expand __MDPACKH.
300 * doc/extend.texi (__MDPACKH): Update documentation.
302 2005-01-28 Uros Bizjak <uros@kss-loka.si>
304 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use
305 SSE_FLOAT_MODE_P for use_sse computation.
307 2005-01-28 Ralf Corsepius <ralf.corsepius@rtems.org>
310 * config/sparc/rtemself.h (LINK_GCC_C_SEQUENCE_SPEC): undefine.
312 2005-01-27 Richard Henderson <rth@redhat.com>
314 * builtins.c (expand_builtin_copysign): New.
315 (expand_builtin): Call it.
316 * genopinit.c (optabs): Add copysign_optab.
317 * optabs.c (init_optabs): Initialize it.
318 (expand_copysign): New.
319 * optabs.h (OTI_copysign, copysign_optab): New.
320 (expand_copysign): Declare.
322 * config/alpha/alpha.md (UNSPEC_COPYSIGN): New.
323 (copysignsf3, ncopysignsf3, copysigndf3, ncopysigndf3): New.
325 * config/i386/i386.c (ix86_build_signbit_mask): Split from ...
326 (ix86_expand_fp_absneg_operator): ... here.
327 (ix86_split_copysign): New.
328 * config/i386/i386-protos.h: Update.
329 * config/i386/i386.md (UNSPEC_COPYSIGN): New.
330 (copysignsf3, copysigndf3): New.
332 * config/ia64/ia64.md (UNSPEC_COPYSIGN): New.
333 (copysignsf3, ncopysignsf3): New.
334 (copysigndf3, ncopysigndf3): New.
335 (copysignxf3, ncopysignxf3): New.
336 * config/ia64/ia64.c (rtx_needs_barrier): Handle UNSPEC_COPYSIGN.
338 2005-01-27 Arend Bayer <arend.bayer@web.de>
339 Kazu Hirata <kazu@cs.umass.edu>
341 * cse.c: (find_best_addr): Don't call copy_rtx before calling
342 fold_rtx. Save cost recomputation if fold_rtx did nothing.
343 (fold_rtx) <ASM_OPERANDS>: Don't do anything if INSN is
346 2005-01-27 Jeff Law <law@redhat.com>
348 * tree-into-ssa.c (ssa_rewrite_initialize_block): Do not register
349 definitions of SSA_NAMEs which are not being rewritten.
350 (rewrite_ssa_into_ssa): Only initialize the current definition
351 of an SSA_NAME if that SSA_NAME has been marked for rewriting.
352 If checking is enabled, assert that SSA_NAME_AUX is clear for all
353 SSA_NAMEs before returning.
355 2005-01-27 Kazu Hirata <kazu@cs.umass.edu>
357 * c-common.def, c-dump.c, c-gimplify.c, c-objc-common.c,
358 cfgexpand.c, dbxout.c, function.h, opts.c, tree-flow-inline.h,
359 tree-ssa-operands.h, tree-ssa-threadupdate.c, tree-tailcall.c,
360 config/avr/avr.md, config/cris/aout.h, config/cris/cris.h,
361 config/mips/iris6.h, config/sh/sh.c: Update copyright.
363 2005-01-27 Steven Bosscher <stevenb@suse.de>
366 * opts.c (decode_options): Move flag_thread_jumps from -O1 and
367 higher to -O2 and higher. Likewise for tree PRE.
368 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Only iterate at -O2
371 2005-01-27 Ian Lance Taylor <ian@airs.com>
374 * gimple-low.c (try_catch_may_fallthru): New static function.
375 (block_may_fallthru): Handle TRY_CATCH_EXPR.
376 * tree-inline.c (expand_call_inline): Don't warn about reaching
377 the end of a non-void function being inlined if the function uses
380 2005-01-27 Jakub Jelinek <jakub@redhat.com>
382 * config/i386/i386.h (CALL_USED_REGISTERS): Fix comment pastos.
385 * c-decl.c (warn_if_shadowing): Handle old_decl error_mark_node.
386 (pushdecl): Only use DECL_FILE_SCOPE_P if DECL_P.
387 (implicitly_declare): Handle error_mark_node.
389 2005-01-27 Richard Henderson <rth@redhat.com>
392 * tree.h (struct tree_decl): Add debug_expr_is_from.
393 (DECL_DEBUG_EXPR_IS_FROM): New.
394 (DECL_DEBUG_EXPR): Rename from DECL_DEBUG_ALIAS_OF.
395 * dwarf2out.c (dwarf2out_var_location): Update to match.
396 * tree-outof-ssa.c (create_temp): Likewise.
397 * var-tracking.c (track_expr_p): Likewise.
398 * tree-sra.c (instantiate_element): Set DECL_DEBUG_EXPR.
399 * c-objc-common.c (c_tree_printer) <'D'>: Handle DECL_DEBUG_EXPR.
400 * toplev.c (default_tree_printer): Likewise.
402 2005-01-27 Alexandre Oliva <aoliva@redhat.com>
404 * config/frv/frv.c (frv_class_likely_spilled_p): Add GR8_REGS,
405 GR9_REGS, GR89_REGS, FDPIC_FPTR_REGS, FDPIC_REGS.
407 2005-01-27 Steven Bosscher <stevenb@suse.de>
409 * tree.h (SWITCH_COND, SWITCH_BODY, SWITCH_LABELS, CASE_LOW,
410 CASE_HIGH, CASE_LABEL, ASM_STRING, ASM_OUTPUTS, ASM_INPUTS,
411 ASM_CLOBBERS): Add tree checks.
413 * c-common.h (SWITCH_TYPE): Rename to SWITCH_STMT_TYPE.
414 (SWITCH_STMT_COND, SWITCH_STMT_BODY): New.
415 * c-common.def (SWITCH_STMT): Update to match.
416 * c-common.c (c_do_switch_warnings): Use SWITCH_STMT accessor
417 macros instead of SWITCH_EXPR ones.
418 * c-dump.c (c_dump_tree): Likewise.
419 * c-gimplify.c (gimplify_switch_stmt): Likewise.
420 * c-typeck.c (c_start_case, do_case, c_finish_case): Likewise.
422 * doc/c-tree.texi (SWITCH_STMT): Update accessor macro names.
424 2005-01-27 Alan Modra <amodra@bigpond.net.au>
426 * unwind-dw2.c (execute_stack_op): Add missing cases for
427 DW_OP_shl, DW_OP_shr, DW_OP_shra, DW_OP_xor.
429 2005-01-27 Eric Botcazou <ebotcazou@libertysurf.fr>
431 * config.gcc (Obsolete configurations): Remove 'dummy', add
432 sparclite-*-coff*, sparclite-*-elf*, sparc86x-*-elf* and
435 2005-01-26 Diego Novillo <dnovillo@redhat.com>
437 PR tree-optimization/19633
438 * tree-ssa-alias.c (ptr_is_dereferenced_by): Also handle
440 (maybe_create_global_var): Do not create .GLOBAL_VAR if there
441 are no call-clobbered variables.
442 * tree-outof-ssa.c (check_replaceable): Return false for calls
445 2005-01-26 Ulrich Weigand <uweigand@de.ibm.com>
447 * dbxout.c (dbxout_symbol_location): Resolve constant pool references
448 even for variables with NULL DECL_INITIAL.
450 2005-01-26 Stuart Hastings <stuart@apple.com>
452 * gimplify.c (shortcut_cond_expr): Re-compute side-effects.
454 2005-01-26 Richard Henderson <rth@redhat.com>
457 * c-decl.c (finish_struct): Set DECL_MODE after resetting a
459 * expr.c (store_field): Strip conversions to odd-bit-sized types
460 if the destination field width matches.
462 2005-01-26 Richard Henderson <rth@redhat.com>
464 * c-decl.c, expmed.c, expr.c: Revert last change.
466 2005-01-26 Richard Henderson <rth@redhat.com>
469 * c-decl.c (finish_struct): Set DECL_MODE after resetting a
471 * expmed.c (store_fixed_bit_field): Create a paradoxical subreg
472 if we don't need the bits above those present in the current mode.
473 * expr.c (store_field): Strip conversions to odd-bit-sized types
474 if the destination field width matches.
476 2005-01-26 Richard Sandiford <rsandifo@redhat.com>
478 * config/mips/iris6.h (ENDFILE_SPEC): Don't link in irix-csr.o if
479 either -ffast-math or -funsafe-math-optimizations is in use.
481 2005-01-26 Marek Michalkiewicz <marekm@amelek.gda.pl>
485 * config/avr/avr.c (notice_update_cc): Only set condition code for
486 ashrqi3 if shift count > 0.
487 (out_shift_with_cnt): Handle shift count <= 0 as a no-op.
488 (ashlqi3_out, ashlhi3_out, ashlsi3_out, ashrqi3_out, ashrhi3_out,
489 ashrsi3_out, lshrqi3_out, lshrhi3_out, lshrsi3_out): Handle shift
490 count <= 0 as a no-op, and shift count >= width by copying zero
491 or sign bit to all bits of the result.
492 * config/avr/avr.md (all shifts): Add alternatives for zero shift
493 count, with attribute "length" set to 0 and "cc" set to "none".
495 2005-01-26 Aldy Hernandez <aldyh@redhat.com>
497 * doc/invoke.texi: Document -mTLS.
499 * testsuite/gcc.target/frv/all-tls-global-dynamic.c: New.
500 * testsuite/gcc.target/frv/all-tls-initial-exec.c: New.
501 * testsuite/gcc.target/frv/all-tls-initial-exec-pic.c: New.
502 * testsuite/gcc.target/frv/all-tls-local-dynamic.c: New.
503 * testsuite/gcc.target/frv/all-tls-local-dynamic-plt.c: New.
504 * testsuite/gcc.target/frv/all-tls-local-dynamic-plt-pic.c: New.
505 * testsuite/gcc.target/frv/all-tls-local-exec.c: New.
506 * testsuite/gcc.target/frv/all-tls-local-exec-TLS.c: New.
508 * configure.ac: Check for a TLS capable gas.
510 * configure: Regenerate.
512 2005-01-26 Richard Henderson <rth@redhat.com>
515 * combine.c (make_field_assignment): Simplify store to zero_extract
516 from a source with an overlapping mask.
518 2005-01-26 Aldy Hernandez <aldyh@redhat.com>
520 2004-11-11 Eric Christopher <echristo@redhat.com>
522 * config/frv/frv.md: Add fr400_integer automaton. Don't
523 allow TYPE_MUL, TYPE_MACC, or TYPE_CUT to issue in parallel
526 2005-01-26 Steven Bosscher <stevenb@suse.de>
529 * cfgbuild.c (make_edges): Do not clear or set
530 current_function_has_computed_jump.
531 * function.h (struct function): Remove the has_computed_jump field.
532 (current_function_has_computed_jump): Do not define.
533 * sched-rgn.c (is_cfg_nonregular): Return true if a basic block ends
534 in a computed jump. Ignore current_function_has_computed_jump.
536 2005-01-26 Richard Henderson <rth@redhat.com>
539 * expr.c (categorize_ctor_elements): New argument p_must_clear.
540 (categorize_ctor_elements_1): Likewise. Detect a union that isn't
542 (mostly_zeros_p): Update for new categorize_ctor_elements argument.
543 * gimplify.c (gimplify_init_constructor): Likewise. Only shove
544 objects into static storage if they have more than one non-zero value.
545 * tree.h (categorize_ctor_elements): Update decl.
547 2005-01-26 Steven Bosscher <stevenb@suse.de>
550 * tree.h (CALL_EXPR_TAILCALL): Add comment.
551 * calls.c (check_sibcall_argument_overlap_1): Revert the change
552 to this function from 2004-07-10.
553 * tree-tailcall.c (suitable_for_tail_opt_p): Do not consider the
554 the current function for tail call optimizations if the address
555 of one of it its arguments is taken.
557 2005-01-26 Kazu Hirata <kazu@cs.umass.edu>
559 * cse.c (fold_rtx): Call equiv_constant only when necessary.
561 2005-01-26 Bernd Schmidt <bernd.schmidt@analog.com>
563 * function.c (expand_function_end): If current_function_calls_alloca,
564 emit stack restore in a place that is reached when the function falls
567 2005-01-26 Alexandre Oliva <aoliva@redhat.com>
569 * Makefile.in (../$(build_subdir)/fixincludes/fixincl): Add dummy
570 rule to avoid matching inappropriate suffix rule.
572 2005-01-26 Jan Hubicka <jh@suse.cz>
574 PR tree-optimization/19241
575 * tree-inline.c (copy_body_r): Do not walk subtrees after substituting.
577 2005-01-26 Hans-Peter Nilsson <hp@axis.com>
579 * config/cris/cris.h (LIB_SPEC) <sim*>: Use -lsyslinux, not
580 -lsyssim. Wrap -lc -lsyssim in --start-group --end-group instead
582 (STARTFILE_SPEC): Add crti.o, before crtbegin.o.
583 <sim*>: Always use crt1.o, regardless of N in simN.
584 <!sim>: Drop support for -pg and -p variants; always use crt0.o.
585 (ENDFILE_SPEC): Add crtn.o, after crtend.o.
586 (CRT_CALL_STATIC_FUNCTION): Remove.
587 * config/cris/aout.h (STARTFILE_SPEC): <!melinux && sim*>: Always
588 use crt1.o, regardless of N in simN.
589 <!sim>: Drop support for -pg and -p variants; always use crt0.o.
590 (LIB_SPEC): Do not define; default to the one in config/cris/cris.h.
591 (CRIS_CPP_SUBTARGET_SPEC) <melinux && !nostdinc>: Remove special
592 case isystem setting.
593 (MAX_OFILE_ALIGNMENT): Correct to 32, not 16.
595 2005-01-25 Ian Lance Taylor <ian@airs.com>
597 * gimple-low.c (block_may_fallthru): Correct handling of
598 SWITCH_EXPR--may fall through if SWITCH_LABELS is NULL.
600 2005-01-26 Hans-Peter Nilsson <hp@axis.com>
602 * config/cris/cris.h (CPP_SPEC): Adjust the various mtune=
603 handlers to match indentation; nested under the mtune=* handler.
604 Similar for mcpu= and march=.
606 2005-01-25 Daniel Jacobowitz <dan@codesourcery.com>
608 * config/mips/linux.h (LIB_SPEC): Remove -rpath-link.
609 * config/mips/linux64.h (LIB_SPEC): Likewise.
610 * config/mn10300/linux.h (LIB_SPEC): Likewise.
611 * config/sh/linux.h (LIB_SPEC): Likewise.
613 2005-01-25 Aldy Hernandez <aldyh@redhat.com>
615 * config/frv/frv.c (frv_legitimize_tls_address): New.
616 (TARGET_HAVE_TLS): Define.
617 (FRV_SYMBOL_REF_TLS_P): Define.
618 (frv_override_options): Handle new register classes.
619 (frv_legitimate_address_p): Reject tls addresses.
620 (frv_legitimize_address): Handle TLS addresses.
621 (gen_inlined_tls_plt): New.
623 (frv_legitimize_tls_address): New.
624 (unspec_got_name): Add TLS entries.
625 (got12_operand): Add R_FRV_TLSMOFF12 case.
626 (frv_emit_move): Fixup TLS addresses.
627 (frv_emit_movsi): Legitimize TLS addresses.
629 * config/frv/frv.h (MASK_BIG_TLS): New.
630 (TARGET_BIG_TLS): New.
631 (HAVE_AS_TLS): Define.
632 (TARGET_SWITCHES): Add -mTLS and -mtls options.
633 (enum reg_class): Add GR8_REGS, GR9_REGS, GR89_REGS.
634 (REG_CLASS_NAMES): Same.
635 (REG_CLASS_CONTENTS): Same.
636 (CONSTRAINT_LEN): New.
637 (REG_CLASS_FROM_CONSTRAINT): New.
638 (PREDICATE_CODES): Add symbolic_operand.
640 * config/frv/frv.md (define_constants): Add UNSPEC_GETTLSOFF,
641 UNSPEC_TLS_LOAD_GOTTLSOFF12, UNSPEC_TLS_INDIRECT_CALL,
642 UNSPEC_TLS_TLSDESC_LDD, UNSPEC_TLS_TLSDESC_LDD_AUX,
643 UNSPEC_TLS_TLSOFF_LD, UNSPEC_TLS_LDDI, UNSPEC_TLSOFF_HILO,
644 R_FRV_GOTTLSOFF_HI, R_FRV_GOTTLSOFF_LO, R_FRV_TLSMOFFHI,
645 R_FRV_TLSMOFFLO, R_FRV_TLSMOFF12, R_FRV_TLSDESCHI,
646 R_FRV_TLSDESCLO, R_FRV_GOTTLSDESCHI, R_FRV_GOTTLSDESCLO, GR8_REG,
647 GR9_REG, GR14_REG, LRREG.
648 (type): Add load_or_call attribute.
649 ("load_or_call"): New reservation.
650 ("call_gettlsoff"): New.
651 ("tls_indirect_call"): New.
652 ("tls_load_gottlsoff12"): New.
653 ("tlsoff_hilo"): New.
654 ("tls_tlsdesc_ldd"): New.
655 ("tls_tlsoff_ld"): New.
658 * config/frv/frv-protos.h (symbolic_operand): Protoize.
660 2005-01-25 Alexandre Oliva <aoliva@redhat.com>
662 * config/frv/frv.h (ASM_OUTPUT_DWARF_DTPREL): Define.
663 * config/frv/frv-protos.h (frv_output_dwarf_dtprel): Declare.
664 * config/frv/frv.c (TLS_BIAS): Define.
665 (frv_output_dwarf_dtprel): New.
667 2005-01-26 Jakub Jelinek <jakub@redhat.com>
669 PR rtl-optimization/19579
670 * ifcvt.c (noce_try_cmove_arith): If emitting instructions to set up
671 both A and B, see if they don't clobber registers the other expr uses.
673 2005-01-25 J"orn Rennecke <joern.rennecke@st.com>
675 * real.c (do_add): Initialize signalling and canonical members.
677 * real.c (real_from_integer): Zero out destination.
679 2005-01-18 Jan Hubicka <jh@suse.cz>
681 PR tree-optimize/19337
682 * tree-inline.c (initialize_inlined_parameters): Copy
683 saved_static_chain_decl only.
685 2005-01-25 Richard Henderson <rth@redhat.com>
689 * config/i386/i386.c (x86_inter_unit_moves): Zero.
690 (ix86_preferred_reload_class): Rewrite fp-constant section, with
691 80387 enabled, to return a proper subclass. Return the subset that
692 overlaps with GENERAL_REGS for PLUS.
693 * config/i386/i386.md (movsi_1, movdi_2): Set type to "mmx" for pxor.
695 2005-01-25 Richard Henderson <rth@redhat.com>
698 * tree-complex.c (expand_complex_div_wide): Use the correct formulae.
700 2005-01-25 Roger Sayle <roger@eyesopen.com>
702 * expmed.c (choose_multiplier): Change interface to return the
703 multiplier lowpart as an rtx instead of an unsigned HOST_WIDE_INT.
704 (expand_divmod): Update calls to choose_multiplier.
706 2005-01-24 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
709 * expmed.c (extract_bit_field): Use adjust_address instead of
710 gen_lowpart when op0 is a MEM.
712 2005-01-24 Steven Bosscher <stevenb@suse.de>
714 * cfgexpand.c (expand_gimple_tailcall): Fix typo.
716 2005-01-24 Richard Sandiford <rsandifo@redhat.com>
718 * config/mips/irix-csr.c: New file.
719 * config/mips/t-iris6 (irix-csr.o): New rule to build it.
720 (EXTRA_MULTILIB_PARTS): Add irix-csr.o.
721 * config/mips/iris6.h (ENDFILE_SPEC): Include it in n32 and n64
724 2005-01-24 Eric Botcazou <ebotcazou@libertysurf.fr>
727 * config.gcc (sparc-*-elf*): Remove sol2.h, sparc/sol2.h and
728 sparc/elf.h, add sparc/sp-elf.h.
729 (sparc-*-rtems*): Likewise.
730 (sparclite-*-elf*): Remove sol2.h, sparc/sol2.h, sparc/elf.h and
731 tm-dwarf2.h, add sparc/sp-elf.h.
732 (sparc86x-*-elf): Likewise.
733 (sparc64-*-elf*): Remove sol2.h, sparc/sol2.h and tm-dwarf2.h.
734 * config/sparc/liteelf.h (TARGET_SUB_OS_CPP_BUILTINS): Rename into
735 TARGET_OS_CPP_BUILTINS.
736 * config/sparc/sp86x-elf (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
737 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
739 * config/sparc/openbsd64.h (NO_IMPLICIT_EXTERN_C): Undefine.
740 * config/sparc/sp64-elf.h (NO_IMPLICIT_EXTERN_C): New macro.
741 (SWITCH_TAKES_ARG): Likewise.
742 (LOCAL_LABEL_PREFIX): Likewise.
743 (ASM_GENERATE_INTERNAL_LABEL): Likewise.
744 (TARGET_N_FORMAT_TYPES): Delete.
745 (TARGET_FORMAT_TYPES): Likewise.
746 (ASM_DECLARE_FUNCTION_SIZE): Likewise.
747 * config/sparc/elf.h: Delete.
748 * config/sparc/sp-elf.h: New file.
750 2005-01-24 Kazu Hirata <kazu@cs.umass.edu>
752 * tree-flow-inline.h (phi_arg_from_edge): Remove.
753 * tree-flow.h: Remove the corresponding prototype.
754 * tree-ssa-operands.h, tree-ssa-threadupdate.c, tree-ssa.c,
755 tree-vectorizer.c: Use dest_idx instead of phi_arg_from_edge.
757 2005-01-24 Jakub Jelinek <jakub@redhat.com>
760 * config/rs6000/rs6000.c (rs6000_file_start): Create toc section
761 for AIX ABI or ELF -fPIC.
762 (rs6000_emit_load_toc_table): Don't create toc_section here.
763 (rs6000_xcoff_file_start): Nor here.
765 2005-01-24 J"orn Rennecke <joern.rennecke@st.com>
767 * sh.c (ra.h): Don't #include.
768 (hard_regs_intersect_p): New function, resurrected from ra.c.
770 * sh.c: Fix 1996 Copyright.
772 2005-01-24 Richard Henderson <rth@redhat.com>
773 Aldy Hernandez <aldyh@redhat.com>
775 * regrename.c (note_sets): Handle subregs.
777 2005-01-24 Joseph S. Myers <joseph@codesourcery.com>
780 * genconditions.c (write_header, write_conditions): Elide file if
783 2005-01-24 Kazu Hirata <kazu@cs.umass.edu>
785 * c-opts.c, expmed.c, gengtype.c, tree-complex.c,
786 config/arm/iwmmxt.md, config/avr/avr.c, config/s390/s390.h:
789 2005-01-24 Jakub Jelinek <jakub@redhat.com>
791 * builtins.c (get_pointer_alignment, gimplify_va_arg_expr,
792 expand_builtin_printf, expand_builtin_fprintf,
793 expand_builtin_sprintf): Use POINTER_TYPE_P instead TREE_CODE
794 checking against POINTER_TYPE.
795 (validate_arglist): Handle POINTER_TYPE code by checking
798 2005-01-24 Paolo Bonzini <bonzini@gnu.org>
800 * rtlanal.c (non_rtx_starting_operands, for_each_rtx_1,
802 (for_each_rtx): Call for_each_rtx_1.
803 * rtl.h (init_rtlanal): Declare.
804 * toplev.c (backend_init): Call init_rtlanal.
806 2005-01-24 Jakub Jelinek <jakub@redhat.com>
808 * flow.c (propagate_one_insn): Formatting.
811 * flow.c (libcall_dead_p): Be more conservative if unsure.
812 If there are any instructions between insn and call, see if they are
813 all dead before saying the libcall is dead.
815 2005-01-24 Uros Bizjak <uros@kss-loka.si>
817 * config/i386/i386.md (*extendsfdf2_sse): Use "nonimmediate_operand"
818 operand constraint for operand 0.
820 2005-01-23 Richard Henderson <rth@redhat.com>
823 * tree-complex.c (expand_complex_div_wide): Set EDGE_FALLTHRU on
824 edges leading to join block.
826 2005-01-23 Roger Sayle <roger@eyesopen.com>
828 * expmed.c (expand_mult_highpart): Make static. Change type of
829 constant multiplier argument from unsigned HOST_WIDE_INT to rtx.
830 (expand_divmod): Updates calls to expand_mult_highpart by using
831 gen_int_mode to pass a CONST_INT rtx.
832 * rtl.h (expand_mult_highpart): Remove prototype.
834 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
836 * flow.c (regset_bytes, regset_size): Remove.
838 * cfganal.c: Fix a reference to Harvey's paper.
840 2005-01-23 Steven Bosscher <stevenb@suse.de>
842 PR rtl-optimization/19464
843 * tree-optimize.c (init_tree_optimization_passes): Add one more
844 copyrename pass just before out-of-ssa.
846 2005-01-23 Joseph S. Myers <joseph@codesourcery.com>
849 * recog.c (recog_memoized): Don't define if GENERATOR_FILE.
850 * ggc-none.c (ggc_free): Define.
852 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
854 * cse.c (max_reg, max_insn_uid): Remove.
855 (cse_main): Don't access max_reg or max_insn_uid.
857 * c-common.c (explicit_flag_signed_bitfields,
858 lang_statement_code_p, lang_expand_function_end): Remove.
859 * c-common.h: Remove the corresponding declarations.
860 * c-opts.c (c_common_handle_option): Remove all write access
861 to explicit_flag_signed_bitfields.
863 * cgraph.c (cgraph_varpool_n_nodes): Remove.
864 (cgraph_varpool_node): Don't access cgraph_varpool_n_nodes.
865 * cgraph.h: Remove the corresponding declaration.
867 * gcse.c (null_pointer_info): Remove.
869 2005-01-23 Roger Sayle <roger@eyesopen.com>
870 Eric Botcazou <ebotcazou@libertysurf.fr>
872 * combine.c (simplify_logical): Only simplify logical expressions
873 of the form ior(and(x,y),z) by the inverse distributive law if the
874 result is cheaper than the original.
876 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
878 * alias.c, c-common.h, c-incpath.c, c-incpath.h, expr.c,
879 fold-const.c, gimplify.c, params.h, tree-data-ref.c,
880 tree-if-conv.c, tree-nested.c, tree-outof-ssa.c,
881 tree-ssa-dom.c, tree-vectorizer.c, tree.def, config/darwin.c,
882 config/freebsd-spec.h, config/arm/arm.h,
883 config/h8300/h8300.md, config/i386/i386.md,
884 config/i386/predicates.md, config/i386/sse.md,
885 config/ia64/ia64.c, config/ip2k/ip2k.c, config/s390/s390.c,
886 config/vax/vax.md: Fix comment typos. Follow spelling
889 2005-01-23 Steven Bosscher <stevenb@suse.de>
891 * gengtype.c (walk_type): Produce `not equal to' compares for loop
892 tests, instead of `less than'.
894 2005-01-23 Richard Sandiford <rsandifo@redhat.com>
896 * function.c (assign_parm_setup_block): When creating a new stack slot
897 for a parameter, get its alignment from the parameter's DECL_ALIGN
898 rather than the type's TYPE_ALIGN. Make sure that the parameter
899 is at least word aligned.
901 2005-01-22 Roger Sayle <roger@eyesopen.com>
904 * config/avr/avr.c (avr_hard_regno_mode_ok): Rewrite.
906 2005-01-22 Richard Henderson <rth@redhat.com>
909 * config/i386/i386.md (movsfcc_1_sse_max): Use nonimmediate_operand
910 in both compare operands.
911 (movdfcc_1_sse_max): Likewise.
912 (movsfcc_1_sse): Likewise. Add earlyclobber for scratch.
913 (movdfcc_1_sse): Likewise.
914 * config/i386/i386.c (ix86_split_sse_movcc): Emit copies into the
915 scratch register as needed.
917 2005-01-22 Richard Henderson <rth@redhat.com>
919 * config/i386/i386.md (smaxsf3): Fix mnemonic typo.
921 * config/i386/i386.c (ix86_prepare_fp_compare_args): Fix is_sse test.
923 2005-01-22 Richard Henderson <rth@redhat.com>
925 * genopinit.c (optabs): Use smin/smax for floating point too.
926 * doc/md.texi: Update to match. Clarify that floating point
927 results are undefined for +0/-0 and NaN.
928 * doc/rtl.texi: Likewise.
929 * rtl.def (SMIN, SMAX): Likewise
930 * tree.def (MIN_EXPR, MAX_EXPR): Likewise.
932 * config/alpha/alpha.md (smaxdf3, smindf3, smaxsf3, sminsf3): Add
933 leading 's' to the name.
934 * config/ia64/ia64.md (smaxsf3, sminsf3, smaxdf3, smindf3,
935 smaxxf3, sminxf3): Likewise.
936 * config/rs6000/rs6000.md (smaxdf3, smindf3, smaxsf3, sminsf3):
939 2005-01-22 Paul Brook <paul@codesourcery.com>
941 * config/arm/arm.md: Use "Uy" constraint for wcgr load/stores.
943 2005-01-21 Mark Dettinger <dettinge@de.ibm.com>
945 * config/s390/s390.c (struct processor_costs): New fields
946 dlgr, dlr, dr, dsgfr, dsgr.
947 (z900_cost, z990_cost): Values for new fields.
948 (s390_rtx_costs): New cases MEM und COMPARE in switch
949 statement. Modified handling of SIGN_EXTEND, ZERO_EXTEND,
950 DIV, MOD, UDIV, UMOD.
952 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
954 * config/s390/s390.md ("doloop_si64"): Reload input value directly
955 into the register being decremented.
956 ("doloop_si31", "doloop_di"): Likewise.
957 ("*doloop_si_long"): Adapt pattern.
959 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
961 * config/s390/s390.h (HARD_REGNO_NREGS): Fix computation for
963 (CLASS_MAX_NREGS): Likewise.
965 2005-01-21 Daniel Berlin <dberlin@dberlin.org>
967 * doc/tree-ssa.texi (Statement Operands): Add example for new
968 must-def macro. Note deprecation of old operands interface.
970 2005-01-22 Richard Sandiford <rsandifo@redhat.com>
972 PR tree-optimization/19484
973 * tree-cfg.c (remove_fallthru_edge): New function.
974 (cleanup_control_flow): Remove fallthru edges from calls that are
975 now known not to return.
977 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
979 Fix PR tree-optimization/19038
980 * tree-ssa-dom.c (cprop_operand): Don't replace loop invaeriant
981 copies with loop variant ones.
983 2005-01-22 Kazu Hirata <kazu@cs.umass.edu>
985 * cfganal.c, real.h, reorg.c, timevar.def, tree-ssa-ccp.c,
986 config/alpha/alpha-protos.h, config/alpha/alpha.h,
987 config/alpha/alpha.md, config/alpha/predicates.md,
988 config/sparc/freebsd.h, config/sparc/netbsd-elf.h,
989 config/sparc/sol2.h: Update copyright.
991 * tree-cfg.c (remove_forwarder_block_with_phi): Look at the
992 first label to see if it is a nonlocal label.
994 2005-01-22 David Edelsohn <edelsohn@gnu.org>
995 Andrew Pinski <pinskia@physics.uc.edu>
998 * config/rs6000/rs6000.c (rs6000_va_start): Saturate n_gpr at
999 maximum number of GPRs. Saturate n_fpr at maximum number of FPRs.
1001 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
1003 * config/rs6000/rtems.h: Update copyright.
1005 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
1008 * config/rs6000/rtems.h: Resurrect cpp_os_rtems_spec from gcc < 3.4.
1009 (CPP_OS_RTEMS_SPEC): New (From gcc-3.3's config/rs6000/sys4.h).
1010 (SUBSUBTARGET_EXTRA_SPECS): Use CPP_OS_RTEMS_SPEC.
1012 2005-01-22 Volker Reichelt <reichelt@igpm.rwth-aachen.de>
1015 * c-typeck.c (convert_arguments): Check for error_mark_node.
1017 2005-01-21 Ian Lance Taylor <ian@airs.com>
1019 PR tree-optimization/13000
1020 * tree-inline.c: Include "tree-flow.h".
1021 (expand_call_inline): If warn_return_type, warn if non-void inline
1022 function falls through.
1023 * tree-cfg.c (execute_warn_function_return): Don't warn about
1024 control reaching end if TREE_NO_WARNING is set. Set
1026 * gimple-low.c (block_may_fallthru): Don't assume that SWITCH_EXPR
1028 * gimplify.c (shortcut_cond_expr): Don't emit a jump over the else
1029 branch if we don't need one.
1030 * c-typeck.c: Include "tree-flow.h"
1031 (c_finish_bc_stmt): Don't add a goto if the current statement
1032 list doesn't fall through to the current point.
1034 2005-01-21 Roger Sayle <roger@eyesopen.com>
1036 PR rtl-optimization/576
1037 * real.c (real_arithmetic): Change return type from void to bool
1038 to return an indication that the result may be inexact.
1039 * real.h (real_arithmeric): Update prototype.
1040 * fold-const.c (const_binop): Don't constant fold floating
1041 point expressions when the user specifies -frounding-math and
1042 the result may depend upon the run-time rounding mode.
1043 (fold_convert_const_real_from_real): Clean-up.
1044 (fold_initializer): Ignore flag_rounding_math for initializers.
1045 * simplify-rtx.c (simplify_binary_operation): Likewise, don't
1046 constant fold FP operations with flag_rounding_math if the
1047 result may depend upon the run-time rounding mode.
1049 2005-01-21 Tom Tromey <tromey@redhat.com>
1051 * c-cppbuiltin.c (define__GNUC__): Correct assertion.
1053 2005-01-21 Kazu Hirata <kazu@cs.umass.edu>
1055 * tree-cfg.c: Fix comment typos.
1057 2005-01-21 J"orn Rennecke <joern.rennecke@st.com>
1059 * passes.c (rest_of_handle_flow2): Close / open the flow2 dump file
1060 around the call to rest_of_handle_branch_target_load_optimize.
1062 2005-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
1064 * config.gcc (sparc-*-netbsdelf*): Include dbxelf.h
1065 (sparc64-*-openbsd*): Likewise.
1066 (sparclite-*-elf*): Include tm-dwarf2.h.
1067 (sparc86x-*-elf*): Likewise.
1068 (sparc64-*-elf*): Likewise.
1069 * config/sparc/linux64.h (DWARF2_DEBUGGING_INFO): Delete.
1070 (DBX_DEBUGGING_INFO): Likewise.
1071 * config/sparc/liteelf.h (DWARF2_DEBUGGING_INFO): Likewise.
1072 (PREFERRED_DEBUGGING_TYPE): Likewise.
1073 * config/sparc/netbsd-elf.h (PREFERRED_DEBUGGING_TYPE): Likewise.
1074 * config/sparc/sp64-elf.h (DBX_DEBUGGING_INFO): Likewise.
1075 (PREFERRED_DEBUGGING_TYPE): Likewise.
1076 * config/sparc/sp86x-elf.h (DWARF2_DEBUGGING_INFO): Likewise.
1077 (PREFERRED_DEBUGGING_TYPE): Likewise.
1078 * config/sparc/sparc.h (DBX_DEBUGGING_INFO): Likewise.
1080 2005-01-21 Hans-Peter Nilsson <hp@bitrange.com>
1083 * combine.c (combine_simplify_rtx): Revert change of 2004-12-31.
1085 2005-01-20 Paul Brook <paul@codesourcery.com>
1087 * doc/rtl.texi: Document value extension requirements for CONST_INT.
1089 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
1091 * tree-cfg.c (tree_verify_flow_info): Check that a nonlocal
1092 label is first in a sequence of labels.
1094 * tree-cfg.c (tree_verify_flow_info): Fix a typo.
1096 2005-01-20 Janis Johnson <janis187@us.ibm.com>
1097 Giovanni Bajo <giovannibajo@gcc.gnu.org>
1099 * doc/sourcebuild.texi (Test Directives): New.
1101 2005-01-20 Roger Sayle <roger@eyesopen.com>
1103 * real.c (real_floor): Don't leave the result uninitialized when
1104 mode is VOIDMode, but return the unrounded intermediate value.
1105 (real_ceil): Likewise.
1107 2005-01-20 Paul Brook <paul@codesourcery.com>
1109 * config/arm/arm.md (insv): Use gen_int_mode.
1111 2005-01-20 Eric Botcazou <ebotcazou@libertysurf.fr>
1113 * config/sparc/freebsd.h (ASM_OUTPUT_INTERNAL_LABELREF): Delete.
1114 * config/sparc/linux.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1115 * config/sparc/linux64.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1116 * config/sparc/netbsd-elf.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1117 * config/sparc/sol2.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
1119 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
1122 * dwarf2out.c (concat_loc_descriptor): We don't know we can use
1123 fbreg, so we have to assume we can't.
1124 (loc_descriptor_from_tree_1): Ditto.
1125 (containing_function_has_frame_base): New function.
1126 (add_location_or_const_value_attribute): Use it.
1127 Also try to generate a frame_base from a single element location
1130 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
1132 PR tree-optimization/15349
1133 * timevar.def (TV_TREE_MERGE_PHI): New.
1134 * tree-cfg.c (tree_forwarder_block_p): Add a new argument
1136 (remove_forwarder_block, cleanup_forwarder_blocks): Adjust the
1137 calls to tree_forwarder_block_p.
1138 (remove_forwarder_block_with_phi, merge_phi_nodes,
1139 gate_merge_phi, pass_merge_phi): New.
1140 * tree-optimize.c (init_tree_optimization_passes): Add
1142 * tree-pass.h: Add an extern for pass_merge_phi;
1144 2005-01-20 Richard Henderson <rth@redhat.com>
1147 * config/i386/emmintrin.h (_mm_castpd_ps, _mm_castpd_si128): New.
1148 (_mm_castps_pd, _mm_castps_si128): New.
1149 (_mm_castsi128_ps, _mm_castsi128_pd): New.
1151 2005-01-20 Richard Henderson <rth@redhat.com>
1154 * config/i386/mmintrin.h (_mm_cvtsi32_si64): Use
1155 __builtin_ia32_vec_init_v2si.
1156 (_mm_cvtsi64_si32): Use __builtin_ia32_vec_ext_v2si.
1157 * config/i386/i386.c (IX86_BUILTIN_VEC_EXT_V2SI): New.
1158 (ix86_init_mmx_sse_builtins): Create it.
1159 (ix86_expand_builtin): Expand it.
1160 (ix86_expand_vector_set): Handle V2SFmode and V2SImode.
1161 * config/i386/mmx.md (vec_extractv2sf_0, vec_extractv2sf_1): New.
1162 (vec_extractv2si_0, vec_extractv2si_1): New.
1164 2005-01-20 Richard Henderson <rth@redhat.com>
1167 * config/i386/i386.c (ix86_expand_push): New.
1168 * config/i386/mmx.md (push<MMXMODE>1): New.
1169 * config/i386/sse.md (push<SSEMODE>1): New.
1170 * config/i386/i386-protos.h: Update.
1172 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
1174 * params.def, params.h, predict.c, tree-ssa-loop-im.c: Update
1177 2005-01-20 Richard Henderson <rth@redhat.com>
1179 * rtl.def (CONST_VECTOR): Use RTX_CONST_OBJ.
1180 * rtl.h (CONSTANT_P): Don't special case CONST_VECTOR.
1182 2005-01-19 Richard Henderson <rth@redhat.com>
1185 * config/i386/i386.c (ix86_expand_vector_move_misalign): Convert
1186 to V4SFmode in SSE1 fallback load path.
1188 2005-01-19 Richard Henderson <rth@redhat.com>
1190 * config/i386/i386.c (ix86_expand_vector_init_one_var): Fix typo
1191 in QImode expansion to ix86_expand_vector_set.
1192 (ix86_expand_vector_init_general): Fix typo in V8HImode recursive call.
1194 2005-01-19 Richard Henderson <rth@redhat.com>
1197 * config/i386/i386.c (ix86_preferred_reload_class): Return a proper
1198 subclass of the input class.
1199 (ix86_secondary_memory_needed): Always true for cross-MMX classes.
1200 Always true for cross-SSE1 classes. Rationalize conditionals.
1201 * config/i386/i386.h (SSE_CLASS_P, MMX_CLASS_P): Use straight equality.
1202 * config/i386/i386.md (movsi_1): Add MMX/SSE zeros. Fix alternatives
1203 for SSE1. Don't check TARGET_INTER_UNIT_MOVES.
1204 (movdi_2): Add MMX/SSE zeros.
1205 (movdi_1_rex64): Likewise. Don't check TARGET_INTER_UNIT_MOVES.
1206 (movsf_1): Don't check TARGET_INTER_UNIT_MOVES.
1207 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
1208 (movsi_1_nointernunit, movdi_1_rex64_nointerunit): Remove.
1209 (movsf_1_nointerunit, zero_extendsidi2_32_1): Remove.
1210 (zero_extendsidi2_rex64_1): Remove.
1211 (MOV0 peephole): Check GENERAL_REG_P.
1213 2005-01-19 Richard Henderson <rth@redhat.com>
1216 * config/i386/i386.c (ix86_expand_vector_set): Fix third and fourth
1218 (ix86_expand_vector_extract): Likewise.
1220 2005-01-19 Richard Henderson <rth@redhat.com>
1223 * expr.c (emit_move_change_mode): New argument force; use
1224 simplify_gen_subreg if true.
1225 (emit_move_via_alt_mode): Merge into ...
1226 (emit_move_via_integer): ... here.
1227 (emit_move_ccmode): Use emit_move_change_mode directly.
1229 2005-01-19 Richard Henderson <rth@redhat.com>
1232 * config/alpha/alpha.c (alpha_rtx_costs): Handle HIGH.
1233 (alpha_preferred_reload_class): Handle CONST_VECTOR.
1234 (alpha_emit_set_const_1): Add no_output parameter; don't emit
1236 (alpha_emit_set_const): Likewise. Make static.
1237 (alpha_emit_set_long_const): Make static.
1238 (alpha_extract_integer): Split out from alpha_expand_mov.
1239 (alpha_split_const_mov): Likewise.
1240 (alpha_expand_mov): Use them. Handle CONST_VECTOR.
1241 (alpha_legitimate_constant_p): New.
1242 * config/alpha/alpha-protos.h: Update.
1243 * config/alpha/alpha.h (REGISTER_MOVE_COST): Correct fp<->gp cost.
1244 (LEGITIMATE_CONSTANT_P): Re-implement with a function.
1245 * config/alpha/alpha.md (movsi): Add n alternative.
1246 (movsi_nt_vms, movdi_er_nofix, movdi_er_fix, movdi_fix): Likewise.
1247 (mov<VEC>_fix, mov<VEC>_nofix): Add i alternative.
1248 (splitters for all of the above): Use alpha_split_const_mov.
1249 * config/alpha/predicates.md (non_add_const_operand): New.
1250 (non_zero_const_operand): New.
1251 (input_operand): Use alpha_legitimate_constant_p after reload.
1253 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
1255 PR tree-optimization/19038
1256 * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): New function.
1257 (add_candidate): Add ivs with increment in latch only if
1258 allow_ip_end_pos_p is true.
1259 (determine_iv_cost): Use empty_block_p.
1261 2005-01-19 Daniel Berlin <dberlin@dberlin.org>
1263 * cfganal.c (compute_dominance_frontiers_1): Replace with new algorithm
1264 (compute_dominance_frontiers): Ditto.
1266 2005-01-19 Ralf Corsepius <ralf.corsepius@rtems.org>
1269 * config/sh/t-rtems: New.
1270 * config.gcc (sh-*-rtems*): Reflect having added config/sh/t-rtems.
1272 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
1274 * tree-ssa-ccp.c (maybe_fold_offset_to_component_ref): Always subtract
1275 the offset of the selected field.
1277 2005-01-19 Kazu Hirata <kazu@cs.umass.edu>
1279 * tree-cfg.c (remove_forwarder_block): Fix the check to
1280 prevent a nonlocal label from appearing in the middle of a
1283 2005-01-19 Hans-Peter Nilsson <hp@axis.com>
1285 PR rtl-optimization/19462
1286 * reorg.c (find_end_label): Create return insn only if
1287 current_function_epilogue_delay_list is empty.
1289 2005-01-19 Paolo Bonzini <bonzini@gnu.org>
1291 * doc/passes.texi: Remove paragraph mentioning new-ra.
1293 2005-01-19 Jakub Jelinek <jakub@redhat.com>
1295 PR rtl-optimization/15139
1296 * combine.c: Include params.h.
1297 (count_rtxs): New function.
1298 (record_value_for_reg): If replace_rtx would replace at least
1299 2 occurrences of REG in VALUE and TEM is really large, replace REG with
1300 (clobber (const_int 0)) instead of TEM.
1301 * params.def (PARAM_MAX_LAST_VALUE_RTL): New.
1302 * params.h (MAX_LAST_VALUE_RTL): New.
1303 * Makefile.in (combine.o): Depend on $(PARAMS_H).
1304 * doc/invoke.texi (--param max-last-value-rtl=N): Document.
1307 * c-typeck.c (digest_init): Only call build_vector if all constructor
1308 elements are *_CST nodes.
1309 * gimplify.c (gimplify_init_constructor): Likewise.
1312 * c-typeck.c (digest_init): Only call build_vector if inside_init
1315 2005-01-18 Tobias Schl"uter <tobias.schlueter@physik.uni-muenchen.de>
1317 * toplev.c (init_asm_output): Remove dead #ifdef.
1319 2005-01-18 Kaz Kojima <kkojima@gcc.gnu.org>
1321 * config/sh/linux.h (TARGET_C99_FUNCTIONS): Define.
1323 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
1325 * calls.c (expand_call): Check DECL_BUILT_IN_CLASS before
1326 accessing DECL_FUNCTION_CODE.
1327 * dojump.c (do_jump): Likewise.
1328 * gimplify.c (gimplify_call_expr): Likewise.
1329 * predict.c (expr_expected_value): Likewise.
1330 (strip_builtin_expect): Likewise.
1331 * tree-inline.c (estimate_num_insns_1): Likewise.
1332 * tree-ssa-loop-im.c (stmt_cost): Likewise
1333 * fold-const.c (fold): Test for BUILT_IN_NORMAL.
1334 (tree_expr_nonnegative_p): Likewise.
1336 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
1338 * config/sparc/sparc.c (load_pic_register): Emit the appropriate
1339 variant of the load_pcrel_sym pattern.
1340 * config/sparc/sparc.md (P macro): Move to the top.
1341 (load_pcrel_sym): Macroize using P.
1342 (save_register_window): Likewise.
1344 2005-01-18 Aldy Hernandez <aldyh@redhat.com>
1346 * config/rs6000/rs6000.md ("sunordered"): Disable for e500.
1349 2005-01-18 Andrew Pinski <pinskia@physics.uc.edu>
1352 * c-typeck.c (build_asm_expr): Strip nops off
1353 input memory operands.
1355 2005-01-18 David Edelsohn <edelsohn@gnu.org>
1357 * config/rs6000/aix43.h (CPLUSPLUS_CPP_SPEC): Delete
1358 _XOPEN_SOURCE, _XOPEN_SOURCE_EXTENDED, _LARGE_FILE_API.
1359 * config/rs6000/aix51.h (CPLUSPLUS_CPP_SPEC): Same.
1360 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Same.
1362 2005-01-18 Joel Sherrill <joel.sherrill@oarcorp.com>
1365 * config/i386/i386.c (override_options): If the 80387 is disabled,
1366 then do not return FP values using FP registers.
1368 2005-01-18 Kazu Hirata <kazu@cs.umass.edu>
1370 * basic-block.h, c-common.c, c-cppbuiltin.c, c-lang.c,
1371 c-tree.h, cfgbuild.c, cgraph.c, cgraph.h, collect2.c,
1372 combine.c, config.gcc, coverage.h, cse.c, cselib.c,
1373 defaults.h, df.c, dwarf2asm.c, dwarf2out.c, explow.c, expr.c,
1374 flow.c, fold-const.c, gcse.c, ggc-page.c, gimple-low.c,
1375 gimplify.c, ifcvt.c, langhooks-def.h, lcm.c, optabs.h,
1376 output.h, postreload-gcse.c, postreload.c, recog.c,
1377 resource.c, rtl.def, rtlanal.c, sched-deps.c, sched-rgn.c,
1378 targhooks.h, toplev.c, tree-data-ref.c, tree-eh.c,
1379 tree-flow.h, tree-loop-linear.c, tree-mudflap.h, tree-nrv.c,
1380 tree-optimize.c, tree-outof-ssa.c, tree-pass.h,
1381 tree-scalar-evolution.c, tree-ssa-copy.c, tree-ssa-dce.c,
1382 tree-ssa-dse.c, tree-ssa-loop-ivopts.c, tree-ssa-operands.c,
1383 tree-ssa-pre.c, tree-ssa.c, tree-vectorizer.c, tree.def,
1384 unwind-dw2-fde-darwin.c, var-tracking.c: Update copyright.
1386 * config/darwin.c, config/darwin.h, config/freebsd-spec.h,
1387 config/alpha/alpha.c, config/arm/vfp.md,
1388 config/i386/emmintrin.h, config/i386/i386-modes.def,
1389 config/i386/i386-protos.h, config/i386/i386.h,
1390 config/i386/pmmintrin.h, config/i386/ppro.md,
1391 config/i386/predicates.md, config/i386/xmmintrin.h,
1392 config/i860/i860.c, config/ia64/ia64-modes.def,
1393 config/ia64/ia64-protos.h, config/ia64/ia64.h,
1394 config/ia64/ia64.md, config/ia64/itanium1.md,
1395 config/ia64/itanium2.md, config/ia64/predicates.md,
1396 config/iq2000/iq2000.h, config/mips/linux64.h,
1397 config/rs6000/aix41.h, config/rs6000/aix43.h,
1398 config/rs6000/aix52.h, config/rs6000/darwin-fallback.c,
1399 config/rs6000/rs6000.c, config/rs6000/rs6000.h,
1400 config/rs6000/spe.md, config/sh/sh.md, config/sparc/linux.h,
1401 config/sparc/linux64.h, config/sparc/litecoff.h,
1402 config/sparc/sp64-elf.h, config/sparc/sparc.c,
1403 config/sparc/sparc.h, config/sparc/sparc.md,
1404 config/sparc/sysv4-only.h: Update copyright.
1406 2005-01-18 Richard Henderson <rth@redhat.com>
1409 * config/i386/i386.c (ix86_expand_fp_movcc): Fail for LTGT and UNEQ.
1411 2005-01-18 Marc Espie <espie@openbsd.org>
1412 * config/i386/openbsdelf.h: Typo.
1414 2005-01-18 Richard Henderson <rth@redhat.com>
1416 * config/i386/i386.c (ix86_expand_fp_movcc): Remove TARGET_IEEE_FP
1417 special case for sse.
1418 * config/i386/predicates.md (sse_comparison_operator): Likewise.
1420 2005-01-18 Dorit Naishlos <dorit@il.ibm.com>
1422 * tree-inline.c (estimate_num_insns_1): Added cases for
1423 ALIGN_INDIRECT_REF, MISALIGNED_INDIRECT_REF, and REALIGN_LOAD_EXPR.
1425 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
1427 PR rtl-optimization/19296
1428 * combine.c (simplify_comparison): Rewrite the condition under
1429 which a non-paradoxical SUBREG of a PLUS can be lifted when
1430 compared against a constant.
1432 2005-01-18 Andi Kleen <ak@muc.de>
1434 * c-typeck.c: (convert_for_assignment): Check warn_pointer_sign.
1435 * c.opt (-Wpointer-sign): Add.
1436 * doc/invoke.texi: (-Wpointer-sign): Add.
1438 2005-01-18 Uros Bizjak <uros@kss-loka.si>
1440 * config/i386/i386.c (override_options): Revert 2004-11-24 change.
1441 * config/i386/i386.md (fmodsf3, dremsf3, *sinsf2, *cossf2, sincossf3,
1442 *tansf3_1, tansf2, atan2sf3_1, atan2sf3, atansf2, asinsf2, acossf2,
1443 logsf2, log10sf2, log2sf2, log1psf2, logbsf2, ilogbsf2, expsf2,
1444 exp10sf2, exp2sf2, expm1sf2, rintsf2, floorsf2, ceilsf2, btruncsf2,
1445 nearbyintsf2): Disable for TARGET_SSE_MATH. Leave patterns enabled
1446 for TARGET_MIX_SSE_I387.
1447 (fmoddf3, dremdf3, *sindf2, *sinextendsfdf2, *cosdf2, *cosextendsfdf2,
1448 sincosdf3, *sincosextendsfdf3, *tandf3_1, tandf2, atan2df3_1,
1449 atan2df3, atandf2, asindf2, acosdf2, logdf2, log10df2, log2df2,
1450 log1pdf2, logbdf2, expdf2, exp10df2, exp2df2, expm1df2, rintdf2,
1451 floordf2, ceildf2, btruncdf2, nearbyintdf2): Disable for
1452 (TARGET_SSE2 && TARGET_SSE_MATH). Leave patterns enabled for
1453 TARGET_MIX_SSE_I387.
1454 (atan2sf3, atan2df3, atan2xf3): Remove register constraints
1457 2005-01-18 Uros Bizjak <uros@kss-loka.si>
1460 * config/i386/mmx.md (*movv2sf_internal_rex64, *movv2sf_internal):
1461 Add movaps alternative for xmm reg->reg move.
1463 2005-01-17 Jeff Law <law@redhat.com>
1465 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Clear SSA_NAME_VALUE
1466 at the end of the main DOM loop rather than just before DOM exits.
1468 2005-01-17 Diego Novillo <dnovillo@redhat.com>
1470 PR tree-optimization/19121
1471 * tree-ssa-alias.c (compute_flow_sensitive_aliasing): When
1472 adding aliases to a name tag, also add them to the pointer's
1474 * tree-ssa-copy.c (merge_alias_info): Do not merge flow
1475 sensitive alias info at all. Only check that the two pointers
1476 have compatible pointed-to sets.
1477 * tree-ssa.c (verify_name_tags): Verify that the alias set of
1478 a pointer's type tag is a superset of the alias set of the
1481 2005-01-17 James E Wilson <wilson@specifixinc.com>
1484 * config/ia64/ia64.md (movxf): Handle general register source. Adjust
1485 comment to document why.
1487 2005-01-17 Richard Henderson <rth@redhat.com>
1489 * config/i386/sse.md (smaxv4sf3_finite, sse_vmsmaxv4sf3_finite,
1490 sminv4sf3_finite, sse_vmsminv4sf3_finite, smaxv2df3_finite,
1491 sse2_vmsmaxv2df3_finite, sminv2df3_finite,
1492 sse2_vmsminv2df3_finite): New.
1493 (smaxv4sf3, sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, smaxv2df3,
1494 sse2_vmsmaxv2df3, sminv2df3, sse2_vmsminv2df3): Remove commutative.
1495 Force op1 into register.
1497 2005-01-17 Kelley Cook <kcook@gcc.gnu.org>
1499 * Makefile.in (STAGEMOVESTUFF): Stage all the stamp files to prevent
1500 bubblestrap from always rebuilding build directory.
1502 2005-01-17 Ulrich Weigand <uweigand@de.ibm.com>
1504 * config/s390/s390.md ("*clc"): Remove incorrect '='.
1505 ("*cmpmem_short"): Likewise.
1507 2005-01-17 Steven Bosscher <stevenb@suse.de>
1509 * basic-block.h: Document BB_* flags.
1510 * regrename.c (copyprop_hardreg_forward): Don't use BB_VISITED,
1511 use an sbitmap instead.
1512 * sched-rgn.c (compute_trg_info): Likewise.
1514 2005-01-17 Richard Sandiford <rsandifo@redhat.com>
1516 * config.gcc (mips64*-*-linux*): Set the default abi to n32. Remove
1517 redundant target_cpu_default and extra_parts lines. Use the default
1519 * config/mips/linux64.h (DRIVER_DEFAULT_ABI_SELF_SPEC): Delete.
1520 (SUBTARGET_EXTRA_SPECS): Delete.
1521 (DRIVER_SELF_SPECS): Remove architecture lines. Set the ABI directly.
1522 (SUBTARGET_ASM_SPEC): Remove -mabi=n32 and -mabi=64 mappings.
1523 * config/mips/t-linux64 (MULTILIB_DIRNAMES): Use "32" for -mabi=32
1524 and "n32" for -mabi=n32.
1526 2005-01-17 Ian Lance Taylor <ian@airs.com>
1529 * c-typeck.c (build_c_cast): Revert patch of 2003-02-16: don't
1530 fold constant variables into initial values.
1532 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
1534 * tree-cfg.c (tree_can_merge_blocks_p): Reorder two checks.
1536 * tree-cfg.c (tree_forwarder_block_p): Speed up by walking
1537 through the statements backward.
1539 2005-01-17 Ian Lance Taylor <ian@airs.com>
1541 PR middle-end/13127:
1542 * tree-inline.c (expand_call_inline): Set TREE_NO_WARNING on
1543 a variable set to the return value of the inlined function.
1545 2005-01-17 Mark Dettinger <dettinge@de.ibm.com>
1547 * rtlanal.c (rtx_cost): Assign cost of 0 to a SUBREG
1548 when modes are tieable.
1550 2005-01-17 Ranjit Mathew <rmathew@hotmail.com>
1552 * gthr-posix.h (__gthread_active_p): Use pthread_cancel instead
1553 of pthread_create to find out if threads are enabled.
1554 * gthr-posix95.h (__gthread_active_p): Likewise.
1556 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
1558 * common.opt (-fnew-ra): Remove.
1560 * toplev.h (flag_new_regalloc): Remove.
1561 * Makefile.in (ra*.*): Don't mention.
1562 * passes.c (rest_of_handle_new_regalloc): Remove.
1563 (rest_of_handle_combine, rest_of_compilation): Always consider
1564 flag_new_regalloc as false.
1565 * doc/invoke.texi: Don't document -fnew-ra.
1567 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
1569 * bb-reorder.c (fix_edges_for_rarely_executed_code): Remove
1570 last parameter to reg_scan.
1571 * loop.c (loop_optimize): Likewise.
1572 * passes.c (rest_of_handle_tracer, rest_of_handle_if_conversion,
1573 rest_of_handle_web, rest_of_handle_cfg, rest_of_handle_jump_bypass,
1574 rest_of_handle_life, rest_of_handle_cse, rest_of_handle_cse2,
1575 rest_of_handle_gcse, rest_of_handle_loop_optimize,
1576 rest_of_handle_loop2, rest_of_handle_jump2): Likewise.
1577 * regclass.c (reg_scan): Likewise, for the declaration.
1578 * rtl.h (reg_scan): Likewise, for the prototype.
1580 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
1582 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Don't call
1583 mark_dfs_back_edges.
1585 2005-01-16 Daniel Berlin <dberlin@dberlin.org>
1587 * tree-ssa-pre.c (add_to_sets): s1 may be NULL.
1588 (compute_avail): Uses don't go in tmp_gen.
1590 2005-01-16 Steven Bosscher <stevenb@suse.de>
1592 * ggc-page.c (ggc_alloc_stat): Use __builtin_ctzl instead of a
1593 loop to look for a free slot in a page entry.
1595 2005-01-16 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1598 * defaults.h (TARGET_DEFERRED_OUTPUT_DEFS): Provide default.
1599 * toplev.c (compile_file): Call process_pending_assemble_output_defs
1600 just before targetm.asm_out.file_end.
1601 * tree.h (process_pending_assemble_output_defs): Declare.
1602 * varasm.c (assemble_output_def, process_pending_assemble_output_defs):
1604 (assemble_alias): Defer generation of assembly code for defines when
1605 TARGET_DEFERRED_OUTPUT_DEFS is true.
1606 * config/rs6000/aix41.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
1607 * config/rs6000/aix43.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
1608 * doc/tm.texi (TARGET_DEFERRED_OUTPUT_DEFS): document.
1610 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1613 * pa.c (pa_scalar_mode_supported_p): New function.
1614 (TARGET_SCALAR_MODE_SUPPORTED_P): Define.
1615 * pa.h (MIN_UNITS_PER_WORD): Add comment.
1617 * pa-protos.h (prefetch_operand): Delete.
1618 (prefetch_cc_operand, prefetch_nocc_operand): New declations.
1619 * pa.c (prefetch_operand): Delete.
1620 (prefetch_cc_operand, prefetch_nocc_operand): New functions.
1621 * pa.h (EXTRA_CONSTRAINT): Add `W' constraint.
1622 (PREDICATE_CODES): Delete prefetch_operand. Add prefetch_cc_operand
1623 and prefetch_nocc_operand.
1624 * pa.md (prefetch): Rework to avoid reload problems handling short
1625 displacements when a cache control completer needs to be provided.
1626 (prefetch_32, prefetch_64): Delete.
1627 (prefetch_cc, prefetch_nocc): New patterns.
1629 2005-01-15 David Edelsohn <edelsohn@gnu.org>
1631 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Revert previous change.
1633 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1636 * pa.c (pa_asm_output_aligned_common): Warn if specified alignment
1637 exceeds maximum alignment for global common data.
1638 * pa64-hpux.h (MAX_OFILE_ALIGNMENT): Define to 32768.
1639 * som.h (MAX_OFILE_ALIGNMENT): Likewise.
1641 2004-01-15 Roger Sayle <roger@eyesopen.com>
1643 * tree-ssa-dom.c (extract_range_from_cond): Correct condition.
1645 2004-01-15 Roger Sayle <roger@eyesopen.com>
1647 * harg-reg-set.h (reg_class_names): Prototype global array.
1648 * regclass.c (reg_class_names): Declare here and initialize to
1650 (dump_regclass): Remove local declaration of reg_class_names.
1651 (regclass): Likewise.
1652 * cfg.c (dump_flow_info): Likewise.
1653 * ra-debug.c (reg_class_names): Likewise.
1654 * regrename.c (reg_class_names): Likewise.
1655 * reload.c (reg_class_names): Likewise.
1656 * reload1.c (spill_failure): Likewise.
1657 * config/m68hc11/m68hc11.c (reg_class_names): Likewise.
1659 2005-01-15 Ulrich Weigand <uweigand@de.ibm.com>
1661 * config/s390/s390.md ("reload_outti"): Remove predicate for
1662 output operand. Abort if operand is not a MEM.
1663 ("reload_outdi", "reload_outdf"): Likewise.
1665 2005-01-15 Marc Espie <espie@openbsd.org>
1667 * config.gcc (*-*-openbsd*): Set HAS_LIBC_R for OpenBSD <= 3.2.
1668 (i[34567]86-*-openbsd*): Switch to ELF for OpenBSD >= 3.4.
1669 * config/openbsd.h: Fix C++ includes for native configurations.
1670 Add proper OS_CPP_BUILTINS. Fix libspec for recent OpenBSD.
1671 Add trampoline support.
1672 * config/i386/openbsdelf.h: New.
1674 2005-01-15 Marc Espie <espie@openbsd.org>
1676 * collect2.c (main): Explicitly parse -dynamic-linker option.
1678 2005-01-15 Jakub Jelinek <jakub@redhat.com>
1680 PR tree-optimization/19060
1681 * tree-ssa-dom.c (extract_range_from_cond) <case LT_EXPR, GT_EXPR>:
1682 Return 0 if op1 <= TYPE_MIN_VALUE () resp. op1 >= TYPE_MAX_VALUE ().
1683 (simplify_cond_and_lookup_avail_expr): Add assert for dummy == 0
1684 and handle extract_range_from_cond returning false.
1685 * fold-const.c (fold): Optimize comparisons with min/max even for
1686 width > HOST_BITS_PER_WIDE_INT.
1688 2005-01-15 Ralf Corsepius <ralf.corsepius@rtems.org>
1690 * config/mips/rtems.h (MIPS_DEFAULT_GVALUE): Set to 0.
1691 * config/mips/t-rtems (MULTILIBS_DIRNAMES,MULTILIB_OPTIONS):
1692 Remove little endian multilib variants.
1693 Add mips32 multilib variant.
1695 2005-01-15 Kazu Hirata <kazu@cs.umass.edu>
1697 * explow.c (copy_all_reg, stabilize): Remove.
1698 * expr.h: Remove the corresponding prototypes.
1700 * rtlanal.c (regs_set_between_p): Remove.
1701 * rtl.h: Remove the corresponding prototype.
1703 2005-01-14 David Edelsohn <edelsohn@gnu.org>
1705 * config/rs6000/rs6000.c (rs6k_nonimmediate_operand): Rename
1706 to rs6000_nonimmediate_operand.
1707 * config/rs6000/rs6000.h (PREDICATE_CODES): Rename
1708 rs6k_nonimmediate_operand to rs6000_nonimmediate_operand.
1709 * config/rs6000/rs6000.md (movsi_internal1): Rename predicate to
1710 rs6000_nonimmediate_operand.
1711 * config/rs6000/spe.md (movdf_e500_double): Rename predicate to
1712 rs6000_nonimmediate_operand.
1714 2005-01-14 Andrew Pinski <pinskia@physics.uc.edu>
1716 * tree-ssa-alias.c (pass_may_alias): Add TODO_verify_stmts.
1718 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1720 * config/rs6000/rs6000.h (CLASS_MAX_NREGS): DF goes in 1 register
1722 (CANNOT_CHANGE_MODE_CLASS): Restrict DI mode changes on e500v2.
1723 (PREDICATE_CODES): Add rs6k_nonimmediate_operand.
1725 * config/rs6000/rs6000.c (invalid_e500_subreg): New.
1726 (rs6k_nonimmediate_operand): New.
1727 (rs6000_legitimate_offset_address_p): Handle DI modes on e500v2
1729 (legitimate_lo_sum_address_p): Same.
1730 (rs6000_legitimize_address): Same.
1731 (rs6000_legitimize_reload_address): Same.
1732 (rs6000_legitimate_address): Same.
1733 (spe_build_register_parallel): Pass DF and DC modes in a DI
1736 * config/rs6000/rs6000.md ("*movsi_internal1"): Change predicate
1737 to rs6k_nonimmediate_operand.
1739 * config/rs6000/spe.md ("*frob_df_di"): New.
1740 ("*frob_di_df"): New.
1741 ("*frob_di_df_2"): New.
1742 ("*mov_sidf_e500_subreg0"): New.
1743 ("*mov_sidf_e500_subreg4"): New.
1744 ("*movdf_e500_double"): Change predicate to
1745 rs6k_nonimmediate_operand.
1747 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1749 * postreload.c (move2add_note_store): Only call
1750 trunc_int_for_mode on scalar integers.
1752 2005-01-14 Steven Bosscher <stevenb@suse.de>
1754 * tree-ssa-dce.c (visited_control_parents): New sbitmap to
1755 replace BB_VISITED uses.
1756 (find_obviously_necessary_stmts): Don't clear BB_VISITED.
1757 (propagate_necessity): Check the bitmap instead of BB_VISITED.
1758 (tree_dce_done): Free visited_control_parents.
1759 (perform_tree_ssa_dce): Allocate and clear it.
1760 * tree-ssa-pre.c (compute_antic_aux): Make non-recursive.
1761 (compute_antic): Iterate from here using a DFS. Use an sbitmap
1762 instead of BB_VISITED.
1764 2005-01-14 Kazu Hirata <kazu@cs.umass.edu>
1766 * c-tree.h, coverage.h, langhooks-def.h, optabs.h, output.h,
1767 rtl.h, targhooks.h, tree-flow.h, tree-mudflap.h, tree.h:
1768 Remove unused prototypes.
1770 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1772 * config/rs6000/rs6000.c (rs6000_stack_info): Fix caching of SPE
1773 64-bit register usage.
1774 (spe_func_has_64bit_regs_p): Add FIXME note.
1776 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1778 * config/rs6000/rs6000.c (rs6000_generate_compare): Replace
1779 flag_finite_math_only with flag_unsafe_math_optimizations.
1781 2005-01-14 J. D. Johnston <jjohnst@us.ibm.com>
1783 * config/s390/tpf-unwind.h (s390_fallback_frame_state): Correct
1785 (__tpf_eh_return): Copy TPF private stack area from the module
1786 boundary stack frame. Add check for when module addresses are
1787 equal, but no stub address is found.
1789 2005-01-14 Richard Earnshaw <rearnsha@arm.com>
1792 * arm.h (struct machine_function): Add call_via field.
1793 (thumb_call_via_label): Declare.
1794 * arm.c (thumb_call_via_label): New variable.
1795 (thumb_call_reg_needed): New variable.
1796 (arm_output_function_epilogue): For Thumb code, output any per-function
1797 call-indirect trampolines.
1798 (thumb_call_via_reg): New function.
1799 (arm_file_end): New function.
1800 (TARGET_ASM_FILE_END): Call arm_file_end.
1801 (aof_file_end): Likewise.
1802 * arm-protos.h (thumb_call_via_reg): Declare.
1803 * arm.md (call_reg_thumb, call_value_reg_thumb): Call
1804 thumb_call_via_reg in normal case.
1806 2005-01-14 Jakub Jelinek <jakub@redhat.com>
1809 PR rtl-optimization/19348
1810 * recog.c (peephole2_optimize): Do global life update if some peephole
1811 decides it doesn't need at least one of its inputs and that change
1812 influences liveness at the start of the basic block.
1814 * basic-block.h (EXECUTE_IF_AND_COMPL_IN_REG_SET): Needs 2 REGSET
1815 arguments instead of 1.
1817 2005-01-14 Eric Botcazou <ebotcazou@libertysurf.fr>
1820 * varasm.c (initializer_constant_valid_p) <ADDR_EXPR>: Return
1821 zero for nested functions needing a static chain or functions
1822 with a non-constant address.
1824 2005-01-13 Roger Sayle <roger@eyesopen.com>
1826 * simplify-rtx.c (simplify_binary_operation) <AND>: Optimize
1827 (and (sign_extend X) C) into (zero_extend (and X C)).
1829 2005-01-13 David O'Brien <obrien@FreeBSD.org>
1831 * config/freebsd-spec.h: Make KSE pthread lib logic the default.
1833 2005-01-13 Richard Henderson <rth@redhat.com>
1838 * config/i386/i386.md (cmpdf, cmpsf, bunordered, bordered, buneq,
1839 bunge, bungt, bunle, bunlt, bltgt): Enable for TARGET_SSE_MATH,
1840 not just TARGET_SSE.
1841 (cmpfp_i_387): Rename from cmpfp_i. Move after sse patterns.
1842 (cmpfp_i_mixed): Rename from cmpfp_i_sse; use for TARGET_MIX_SSE_I387.
1843 (cmpfp_i_sse): Rename from cmpfp_i_sse_only; use for TARGET_SSE_MATH.
1844 (cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387): Similarly.
1845 (fp_jcc_1_mixed, fp_jcc_1_sse, fp_jcc_1_387): Similarly.
1846 (fp_jcc_2_mixed, fp_jcc_2_sse, fp_jcc_2_387): Similarly.
1847 (fp_jcc_3_387, fp_jcc_4_387, fp_jcc_5_387, fp_jcc_6_387,
1848 fp_jcc_7_387, fp_jcc_8_387): Rename from fp_jcc_N.
1849 (movdicc_c_rex64): Rename with '*'.
1850 (movsfcc, movdfcc): Add checks for 387 and sse math to condition.
1851 (movsfcc_1_sse_min, movsfcc_1_sse_max, movsfcc_1_sse): New.
1852 (movsfcc_1_387): Rename from movsfcc_1.
1853 (movdfcc_1_sse_min, movdfcc_1_sse_max, movdfcc_1_sse): New.
1854 (movdfcc_1, movdfcc_1_rex64): Add check for 387.
1855 (sminsf3, smaxsf3, smindf3, smaxdf3): New.
1856 (minsf3, minsf, minsf_nonieee, minsf_sse, mindf3, mindf,
1857 mindf_nonieee, mindf_sse, maxsf3, maxsf, maxsf_nonieee, maxsf_sse,
1858 maxdf3, maxdf, maxdf_nonieee, maxdf_sse, sse_movsfcc, sse_movsfcc_eq,
1859 sse_movdfcc, sse_movdfcc_eq, sse_movsfcc_const0_1,
1860 sse_movsfcc_const0_2, sse_movsfcc_const0_3, sse_movsfcc_const0_4,
1861 sse_movdfcc_const0_1, sse_movdfcc_const0_2, sse_movdfcc_const0_3,
1862 sse_movdfcc_const0_4): Remove.
1863 * config/i386/i386.c (ix86_expand_fp_movcc): For TARGET_SSE_MATH,
1864 recognize min/max early. Update for changed sse cmove patterns.
1865 (ix86_split_sse_movcc): New.
1866 * config/i386/i386-protos.h: Update.
1868 2005-01-13 Steven Bosscher <stevenb@suse.de>
1870 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
1871 (fix_stmt_v_may_defs): Likewise.
1873 2005-01-13 Richard Henderson <rth@redhat.com>
1875 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
1876 for converting the mask.
1878 2005-01-13 David Edelsohn <edelsohn@gnu.org>
1880 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
1883 2005-01-13 Richard Henderson <rth@redhat.com>
1885 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
1886 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
1887 IX86_BUILTIN_MOVDQ2Q): Remove.
1888 (IX86_BUILTIN_VEC_EXT_V4SI): New.
1889 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
1890 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
1892 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
1893 (sse2_pextrw): Fix immediate constraint.
1894 (sse2_loadq, sse2_loadq_rex64): Remove.
1895 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
1896 Use __builtin_ia32_vec_ext_<size>.
1897 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
1899 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
1901 * function.c (assign_parm_setup_block): Look inside original
1902 entry_parm when inspecting PARALLEL.
1904 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
1905 Joel Sherrill <joel@oarcorp.com>
1908 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
1909 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
1910 rtems_gxx_recursive_mutex_init.
1911 (__gthread_recursive_mutex_lock): New function.
1912 (__gthread_recursive_mutex_trylock): Likewise.
1913 (__gthread_recursive_mutex_unlock): Likewise.
1915 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
1917 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
1919 2005-01-13 Jan Beulich <jbeulich@novell.com>
1921 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
1922 (ashrti3_internal): Indicate output is early clobber. Generate result
1923 into output rather than first input. Use move for low word of output
1924 if shift count is exactly 64.
1925 (lshrti3_internal): Likewise.
1927 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
1931 * reload1.c (delete_output_reload): Don't delete an output reload
1932 if the pseudo lives longer than a single basic block. Adjust and
1933 improve wording of comment.
1935 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
1938 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
1939 of TYPE_SIZE where we mean the number of bytes.
1941 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
1943 * function.c (assign_parm_setup_block): Relax condition on
1944 multi-register optimization.
1946 2005-01-12 Nick Clifton <nickc@redhat.com>
1948 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
1949 an "in_delay_slot" attribute of "no" to prevent them being used in
1950 delay slots. This is forbidden because they might generate
1953 2005-01-12 Alan Modra <amodra@bigpond.net.au>
1956 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
1959 2005-01-12 Nick Clifton <nickc@redhat.com>
1961 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
1962 passed on to GAS) but do not define, as GAS no longer supports or
1963 needs the -m2000 option.
1965 2005-01-11 Roger Sayle <roger@eyesopen.com>
1967 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
1968 type as args of unprototyped builtins aren't automatically promoted.
1969 (fold_builtin_abs): Likewise.
1971 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1973 * Makefile.in: Set a `build-warn' variable.
1975 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
1977 PR tree-optimization/17949
1978 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
1979 (find_interesting_uses_address): Use it.
1981 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
1983 * regrename.c (kill_value): Handle subreg's that won't simplify.
1985 2005-01-11 Richard Henderson <rth@redhat.com>
1988 * config/i386/i386.h (enum ix86_builtins): Move ...
1989 * config/i386/i386.c: ... here.
1990 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
1991 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
1992 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
1993 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
1994 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
1995 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
1996 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
1997 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
1998 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
1999 IX86_BUILTIN_LOADDDUP): Remove.
2000 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
2001 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
2002 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
2003 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
2004 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
2005 (ix86_init_builtins): Make static.
2006 (ix86_init_mmx_sse_builtins): Update for changed builtins.
2007 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
2008 if all the modes match. Otherwise, fake it.
2009 (get_element_number, ix86_expand_vec_init_builtin,
2010 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
2011 (ix86_expand_builtin): Make static. Update for changed builtins.
2012 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
2013 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
2014 (ix86_expand_vector_init_duplicate): New.
2015 (ix86_expand_vector_init_low_nonzero): New.
2016 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
2017 Split out from ix86_expand_vector_init; handle integer modes.
2018 (ix86_expand_vector_init): Use them.
2019 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
2020 * config/i386/i386-protos.h: Update.
2021 * config/i386/predicates.md (reg_or_0_operand): New.
2022 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
2023 (movv2sf_internal): Likewise. And a splitter to match them all.
2024 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
2025 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
2026 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
2027 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
2028 (mmx_pinsrw): Fix operand ordering.
2029 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
2030 rather than sse_loadss expander.
2031 (movv2df splitter): Similarly.
2032 (sse_loadss, sse_loadlss): Remove.
2033 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
2034 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
2035 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
2036 (sse3_movddup): Rename with '*'.
2037 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
2038 (sse2_loadsd): Remove.
2039 (vec_dupv2df_sse3): Rename from sse3_loadddup.
2040 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
2041 (sse2_pinsrw): Fix argument ordering.
2042 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
2043 (sse2_stored): Remove 'r' destination.
2044 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
2045 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
2046 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
2047 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
2048 vec_extractv16qi, vec_initv16qi): New.
2050 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
2051 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
2052 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
2054 (_mm_load_pd, _mm_store_pd): Use plain dereference.
2055 (_mm_load_si128, _mm_store_si128): Likewise.
2056 (_mm_load1_pd): Use _mm_set1_pd.
2057 (_mm_load_sd): Use _mm_set_sd.
2058 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
2059 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
2060 (_mm_set_epi64): Use _mm_set_epi64x.
2061 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
2062 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
2063 _mm_setr_epi8): Use _mm_set_foo form.
2064 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
2065 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
2066 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
2067 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
2068 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
2069 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
2070 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
2071 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
2072 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
2073 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
2074 (_mm_movedup_pd): Use _mm_shuffle_pd.
2075 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
2076 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
2077 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
2078 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
2080 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
2081 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
2082 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
2083 (_mm_store_ps): Use raw dereference.
2084 (_mm_store1_ps): Use _mm_storeu_ps.
2085 (_mm_storer_ps): Use _mm_store_ps.
2086 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
2087 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
2089 2005-01-11 Stan Shebs <shebs@apple.com>
2091 * config/rs6000/rs6000.c (machopic_output_stub): Issue
2092 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
2094 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
2097 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
2098 Skip all DECLs except for FIELD_DECLs.
2100 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
2102 * config/s390/s390.c (override_options): Return error if
2103 -mbackchain, -mpacked-stack and -mhard-float are used together.
2104 (s390_va_start): Remove the backchain && packed-stack special case.
2105 (s390_gimplify_va_arg): Likewise.
2106 * doc/invoke.texi: Remove the ABI incompatibility note.
2108 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
2110 * config/s390/s390.c (struct s390_frame_layout): Remove
2112 (s390_frame_info, s390_emit_prologue): Replace occurrences of
2113 save_backchain_p with TARGET_BACKCHAIN.
2115 2005-01-11 Alan Modra <amodra@bigpond.net.au>
2118 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
2119 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
2120 * function.c (locate_and_pad_parm): Set new field.
2121 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
2122 Tweak where_pad test to include "none". Always set mem align for
2124 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
2125 not sufficient for type.
2126 (assign_parm_setup_block): If stack_parm is zero on entry, always
2127 make a new stack local. Block move old stack parm if necessary
2128 to new aligned stack local.
2129 (assign_parm_setup_stack): Use a block move to handle
2130 potentially misaligned entry_parm.
2131 (assign_parms_unsplit_complex): Specify required alignment when
2132 creating stack local.
2133 * calls.c (compute_argument_addresses): Override alignment of stack
2134 arg calculated from its type with the alignment given by
2135 FUNCTION_ARG_BOUNDARY.
2136 (store_one_arg): Likewise.
2138 2005-01-11 Jan Beulich <jbeulich@novell.com>
2140 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
2141 Change respective itanium_class attribute to ialu.
2142 (shladdp4_internal): New.
2143 * config/ia64/predicates.md (shladd_log2_operand): New.
2145 2005-01-11 Richard Henderson <rth@redhat.com>
2147 * expr.c (store_constructor): Use rtvec_alloc instead of
2148 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
2150 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
2152 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
2154 2005-01-11 Alan Modra <amodra@bigpond.net.au>
2156 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
2157 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
2158 (TARGET_RWRELOC): Define.
2159 (ia64_rwreloc_section_type_flags): Delete.
2160 (ia64_section_type_flags): New function.
2161 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
2162 (TARGET_RWRELOC): Define.
2164 2005-01-10 David Mosberger <davidm@hpl.hp.com>
2167 * config/ia64/ia64.c (process_set): For alloc insn, only call
2168 process_epilogue is !frame_pointer_needed.
2170 2005-01-10 Roger Sayle <roger@eyesopen.com>
2173 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
2174 unary operator and can't be treated as a binary/comparison operator.
2176 2005-01-10 Richard Henderson <rth@redhat.com>
2178 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
2179 (ix86_return_in_memory): Likewise.
2180 (function_arg_advance): Likewise. Mirror structure in function_arg
2181 for choosing register to advance.
2183 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
2185 * tree-vectorizer.c, tree.def: Fix comment typos.
2187 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
2189 * config/s390/s390.c (struct processor_costs): 4 new fields:
2190 ddbr, ddr, debr, der.
2191 (s390_rtx_costs): More precise handling of divide instructions.
2193 2005-01-10 David Edelsohn <edelsohn@gnu.org>
2195 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
2197 2005-01-10 Jan Beulich <jbeulich@novell.com>
2199 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
2200 names resulting from -ffunction-sections/-fdata-sections and linkonce
2202 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
2203 based on the section name. Rearrange the section name comparison logic
2204 slightly so that each section name is compared against at most once.
2206 2005-01-10 Ben Elliston <bje@au.ibm.com>
2208 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
2209 targets for which -fPIC reduces limitations on the GOT size.
2211 2005-01-09 Falk Hueffner <falk@debian.org>
2213 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
2216 2005-01-09 David Edelsohn <edelsohn@gnu.org>
2219 * collect2.c (main): Set aixrtl_flag for -brtl option.
2220 (resolve_lib_name): Search for .so file extension before .a
2223 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
2225 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
2226 (vect_analyze_data_refs_alignment): Add dump prints.
2228 2005-01-09 Ira Rosen <irar@il.ibm.com>
2230 * tree-vectorizer.c (vect_analyze_offset_expr): Use
2231 expr_invariant_in_loop_p.
2232 Initialize outputs first thing in the function.
2233 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
2234 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
2235 (vect_analyze_pointer_ref_access): Check that the initial condition of
2236 the access function is loop invariant.
2238 2005-01-09 Richard Henderson <rth@redhat.com>
2240 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
2241 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
2242 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
2243 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
2244 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
2245 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
2246 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
2247 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
2248 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
2249 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
2250 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
2251 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
2252 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
2253 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
2254 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
2255 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
2256 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
2257 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
2258 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
2259 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
2260 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
2261 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
2262 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
2264 (mmx_clrdi, pavgusb): Remove.
2265 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
2266 with leading sse_ prefix.
2267 * config/i386/sse.md: Receive them.
2268 * config/i386/mmx.md: New file.
2269 (MMXMODE12, MMXMODE24, mmxvecsize): New.
2270 (subrv2sf3): Turn into expander for normal subtraction.
2271 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
2272 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
2273 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
2274 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
2275 commutative; use ix86_binary_operator_ok.
2276 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
2277 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
2278 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
2279 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
2280 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
2281 Macroize from existing patterns; use ix86_binary_operator_ok.
2282 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
2283 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
2284 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
2285 with vec_select+vec_concat.
2286 (mmx_pshufw, mmx_pshufw_1): Likewise.
2287 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
2288 (mmx_uavgv4hi3): Model correctly.
2289 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
2290 _mm_xor_si64): Remove casts.
2292 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
2294 PR tree-optimization/19224
2295 * tree-scalar-evolution.c (get_instantiated_value,
2296 set_instantiated_value): New functions.
2297 (instantiate_parameters_1): Cache the results.
2298 (instantiate_parameters, resolve_mixers): Initialize and free
2301 2005-01-08 David Edelsohn <edelsohn@gnu.org>
2303 * config/i386/i386.md (addhi_4): Correct reference in comment.
2306 2005-01-08 Richard Henderson <rth@redhat.com>
2308 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
2309 (_mm_cvtsi128_si64x): Likewise.
2310 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
2311 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
2312 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
2313 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
2314 * config/i386/i386-protos.h: Update.
2315 * config/i386/i386.c (print_operand): Add 'H'.
2316 (ix86_fixup_binary_operands): Split out from ...
2317 (ix86_expand_binary_operator): ... here.
2318 (ix86_fixup_binary_operands_no_copy): New.
2319 (ix86_expand_fp_absneg_operator): Handle vector mode results.
2320 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
2321 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
2322 (safe_vector_operand): Use CONST0_RTX.
2323 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
2324 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
2325 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
2326 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
2328 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
2329 (IX86_BUILTIN_CMPNGESS): New.
2330 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
2331 (attr type): Add sselog1.
2332 (attr unit, attr memory): Handle it.
2333 (movti, movti_internal, movti_rex64): Move near other integer moves.
2334 (movtf, movtf_internal): Move near other fp moves.
2335 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
2336 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
2337 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
2338 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
2339 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
2340 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
2341 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
2342 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
2343 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
2344 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
2345 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
2346 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
2347 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
2348 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
2349 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
2350 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
2351 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
2352 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
2353 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
2354 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
2355 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
2356 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
2357 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
2358 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
2359 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
2360 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
2361 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
2362 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
2363 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
2364 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
2365 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
2366 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
2367 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
2368 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
2369 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
2370 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
2371 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
2372 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
2373 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
2374 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
2375 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
2376 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
2377 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
2378 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
2380 (SSEPUSH, push<SSEPUSH>): Remove.
2381 (MMXPUSH, push<MMXPUSH>): Remove.
2382 (sse_movaps, sse_movaps_1, sse_movups): Remove.
2383 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
2384 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
2385 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
2386 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
2387 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
2388 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
2389 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
2390 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
2391 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
2392 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
2393 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
2394 (ppro_sse_log_V4SF): Handle sselog1.
2395 * config/i386/predicates.md (const_0_to_1_operand): New.
2396 (const_0_to_255_mul_8_operand): New.
2397 (const_1_to_31_operand): Rename from const_int_1_31_operand.
2398 (const_2_to_3_operand, const_4_to_7_operand): New.
2399 * config/i386/sse.md: New file.
2400 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
2401 (sse_movups): Rename from sse_movups_1.
2402 (sse_loadlss): Rename from sse_loadss_1.
2403 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
2405 (negv4sf2): Use ix86_expand_fp_absneg_operator.
2406 (absv4sf2, negv2df, absv2df): New.
2407 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
2408 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
2409 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
2410 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
2411 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
2412 (sse3_addsubv4sf3): Model correctly.
2413 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
2414 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
2415 (sse_movhlps): Model with vec_select+vec_concat.
2416 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
2417 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
2418 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
2419 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
2420 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
2421 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
2422 sse2_pshufhw_1): Likewise.
2423 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
2424 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
2425 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
2426 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
2427 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
2428 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
2429 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
2430 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
2431 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
2432 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
2433 sminv8hi3): Mark commutative
2434 operands. Use ix86_binary_operator_ok.
2435 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
2436 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
2437 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
2438 sse2_punpcklqdq): Allow operand2 in memory.
2439 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
2440 sse2_movsd): Add memory alternatives.
2441 (sse_storelps): Turn expander into an insn; split after reload.
2442 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
2443 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
2445 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
2447 * configure.ac (DWARF-2 debug_line): Use objdump.
2448 * configure: Regenerate.
2450 2005-01-08 Jeff Law <law@redhat.com>
2451 Diego Novillo <dnovillo@redhat.com>
2453 PR tree-optimization/18241
2454 * tree-nrv.c (tree_nrv): Ignore volatile return values.
2455 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
2456 statements with volatile operands.
2457 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
2458 operands after marking a statement with has_volatile_ops.
2460 2005-01-08 Roger Sayle <roger@eyesopen.com>
2462 * tree.c (int_fits_type_p): Always honor integer constant
2463 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
2465 2005-01-08 Roger Sayle <roger@eyesopen.com>
2467 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
2468 the latent bug in rtl_delete_block has been resolved.
2470 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
2472 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
2473 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
2475 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
2477 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
2478 library as @multilib_dir@/$(SHLIB_SONAME).
2480 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
2482 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
2483 * configure: Regenerate.
2485 2005-01-07 Jakub Jelinek <jakub@redhat.com>
2487 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
2490 2005-01-07 David Edelsohn <edelsohn@gnu.org>
2493 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
2494 Convert non-word aligned offset address using ld/std into
2497 2005-01-07 Richard Henderson <rth@redhat.com>
2499 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
2501 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
2503 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
2505 2005-01-07 Jakub Jelinek <jakub@redhat.com>
2507 PR tree-optimization/19283
2508 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
2509 is not shorter than the original type.
2511 PR rtl-optimization/19012
2512 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
2514 PR rtl-optimization/18861
2515 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
2516 (FULL_STATE): Define.
2517 (mark_tablejump_edge): New function.
2518 (purge_dead_tablejump_edges): New function.
2519 (find_bb_boundaries): Use it.
2521 PR tree-optimization/18828
2522 * builtins.c (expand_builtin_next_arg): Remove argument and all
2523 the argument checking.
2524 (expand_builtin): Adjust caller.
2525 (expand_builtin_va_start): Likewise. Remove error for too many
2527 (fold_builtin_next_arg): Issue error for too many arguments.
2528 After checking arguments, replace them with magic arguments that
2529 prevent further checking of the args.
2531 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2533 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
2535 2005-01-06 Roger Sayle <roger@eyesopen.com>
2538 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
2539 and %{.S:...} (and their negative variants) to test whether the
2540 input file is assembler or pre-processed-assembler independent of
2541 the actual filename extension.
2543 2005-01-06 Roger Sayle <roger@eyesopen.com>
2545 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
2546 of sign or zero extended values.
2548 2005-01-06 Geoffrey Keating <geoffk@apple.com>
2550 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
2551 for IBM long double format correctly.
2553 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
2555 Fix PR tree-optimization/18792
2557 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
2558 to first_loop_depth, and use loop depth instead of loop number.
2559 (build_classic_dir_vector): Ditto.
2560 (compute_data_dependences_for_loop): Use depth, not loop number.
2561 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
2562 number. Pass in loops, instead of loop numbers.
2563 (gather_interchange_stats): Ditto.
2564 (linear_transform_loops): Ditto.
2566 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
2569 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
2570 biv_fits_mode_p, extension_within_bounds_p): New functions.
2571 (check_ext_dependent_givs): Use them.
2573 2005-01-06 Roger Sayle <roger@eyesopen.com>
2575 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
2576 more than one barrier, in which case we should delete them all.
2578 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
2580 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
2582 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
2584 Fix PR middle-end/19286
2586 * dwarf2out.c (gen_subprogram_die): If we've already tried to
2587 output this subprogram, simply ignore this attempt to do it again.
2588 (add_abstract_origin_attribute): Don't abort trying to add the abstract
2589 origin attribute if it's not possible.
2590 (gen_block_die): Don't ignore subblocks of "unused" blocks.
2591 (decls_for_scope): Ditto.
2592 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
2593 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
2595 2005-01-05 Richard Henderson <rth@redhat.com>
2598 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
2599 (bdesc_2arg): Use it.
2600 (ix86_expand_binop_builtin): Force operands into registers
2602 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
2603 ix86_expand_sse_compare, ix86_expand_sse_comi,
2604 ix86_expand_builtin): Likewise.
2606 2005-01-05 Richard Henderson <rth@redhat.com>
2608 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
2610 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
2611 (mulv4hi3): Set itanium_class mmmul.
2612 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
2613 (fpack): Rename from fpack_sfsf.
2615 2005-01-05 Richard Henderson <rth@redhat.com>
2618 * reload1.c (do_input_reload): Restrict the optimization deleteing
2619 a previous output reload to RELOAD_FOR_INPUT.
2621 2005-01-05 Steven Bosscher <stevenb@suse.de>
2623 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
2624 comment that we fall through after case.
2625 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
2626 (distribute_links): Likewise.
2627 * cse.c (cse_insn): Likewise.
2628 * cselib.c (cselib_invalidate_mem): Likewise.
2629 * df.c: Update comments at the top of the file.
2630 (read_modify_subreg_p): Update comments here too.
2631 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
2632 * flow.c (mark_set_1): Likewise.
2633 (mark_used_regs): Likewise.
2634 * gcse.c (mems_conflict_for_gcse_p): Likewise.
2635 (canon_list_insert): Likewise.
2636 (mark_set): Likewise.
2637 (try_replace_reg): Likewise.
2638 (store_killed_in_insn): Likewise.
2639 * loop.c (count_one_set): Likewise.
2640 (basic_induction_var): Likewise.
2641 * postreload-gcse.c (find_mem_conflicts): Likewise.
2642 * postreload.c (reload_combine_note_store): Likewise.
2643 (move2add_note_store): Likewise.
2644 * reload.c (find_equiv_reg): Likewise.
2645 (mark_referenced_resources): Likewise.
2646 * rtlanal.c (set_noop_p): Likewise.
2647 (note_stores): Likewise.
2648 (note_uses): Likewise.
2649 * sched-deps.c (sched_analyze_1): Likewise.
2650 * sched-rgn.c (check_live_1): Likewise.
2651 (update_live_1): Likewise.
2652 * config/i860/i860.c: Likewise.
2654 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
2656 (ZERO_EXTRACT): Mention that this one can be an lvalue.
2658 * doc/rtl.texi: Update documentation for bit-fields and SET.
2660 2005-01-05 Roger Sayle <roger@eyesopen.com>
2662 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
2664 2005-01-05 Jan Hubicka <jh@suse.cz>
2665 Richard Henderson <rth@redhat.com>
2668 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
2671 2005-01-05 Richard Henderson <rth@redhat.com>
2674 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
2675 (sse_shufps): Change operand 3 to const_int_operand.
2676 (sse2_storelps): Fix typo in template.
2677 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
2678 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
2679 (ix86_expand_builtin): Likewise.
2681 2005-01-05 Julian Brown <julian@codesourcery.com>
2683 * config/arm/arm.c (arm_return_in_memory): Treat complex types
2684 as aggregates for AAPCS ABIs.
2686 2005-01-05 Stan Shebs <shebs@apple.com>
2688 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
2689 Mach-O getters if ppc64.
2690 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
2692 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
2693 signed four-byte field for global code case.
2694 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
2696 2005-01-05 Roger Sayle <roger@eyesopen.com>
2699 * c-common.c: Include real.h.
2700 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
2701 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
2702 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
2703 * Makefile.in (c-common.o): Update dependencies.
2705 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
2707 * c-parse.in (asm_string): Add trailing semicolon.
2709 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
2711 * c-parse.in (asm_string): New. Don't allow wide strings in
2713 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
2714 asm_string instead of STRING.
2716 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
2718 * c-typeck.c (constructor_no_implicit): Remove.
2719 (set_designator, process_init_element): Don't check
2720 constructor_no_implicit.
2722 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
2723 Kaz Kojima <kkojima@gcc.gnu.org>
2726 * lcm.c (create_pre_exit): New.
2727 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
2728 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
2730 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
2732 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
2733 -msoft-float in target_float_switch.
2734 * arm.c (arm_override_options): Fix processing of target_float_switch.
2736 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
2738 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
2741 2004-01-05 Uros Bizjak <uros@kss-loka.si>
2743 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
2744 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
2746 2005-01-04 Roger Sayle <roger@eyesopen.com>
2748 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
2749 if deleting the then-block allows the test-block to fallthru to the
2752 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
2755 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
2756 after the full declaration if the are in two different TUs.
2758 2005-01-04 Richard Henderson <rth@redhat.com>
2761 * tree-sra.c (generate_one_element_init): Just
2762 call gimplify_and_add.
2763 (generate_element_init): Record the
2764 new referenced variables and mark them for renaming
2765 and split out to ...
2766 (generate_element_init_1): This.
2767 (scalarize_init): Don't call push_gimplify_context/
2768 pop_gimplify_context.
2770 2005-01-04 Geoffrey Keating <geoffk@apple.com>
2772 * toplev.c (get_src_pwd): Handle failure of getpwd().
2774 2005-01-04 Roger Sayle <roger@eyesopen.com>
2776 * fold-const.c (fold_single_bit_test): Delete unreachable handling
2778 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
2779 all the cases handled by it are inverted by invert_truthvalue.
2781 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
2783 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
2784 (save_register_windowsi): Likewise.
2786 2005-01-04 Richard Henderson <rth@redhat.com>
2788 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
2790 2005-01-04 Uros Bizjak <uros@kss-loka.si>
2793 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
2794 any simplifications of vector mode comparison operators.
2795 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
2797 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
2798 Devang Patel <dpatel@apple.com>
2800 PR tree-optimization/18308
2801 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
2802 the operands before creating a new expression.
2803 * dojump.c (do_jump): Make drop_through_label available
2804 for all cases. Add expansion of COND_EXPR.
2806 2005-01-04 Ira Rosen <irar@il.ibm.com>
2808 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
2809 INTEGER_CST instead of TREE_CONSTANT.
2810 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
2811 instead of TREE_CONSTANT.
2812 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
2813 instead of TREE_CONSTANT.
2815 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
2817 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
2819 2005-01-04 Uros Bizjak <uros@kss-loka.si>
2822 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
2823 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
2825 2005-01-03 Richard Henderson <rth@redhat.com>
2827 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
2828 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
2829 fold_convert_const_real_from_real): Split out from ...
2830 (fold_convert_const): ... here.
2832 2005-01-03 Richard Henderson <rth@redhat.com>
2835 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
2836 (mov<MMXMODEI>_internal): Likewise.
2837 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
2838 Add V2SF case; use it for SSE1; don't use TI.
2839 (movdf_integer): Likewise.
2840 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
2842 2005-01-03 Ira Rosen <irar@il.ibm.com>
2844 * tree-vectorizer.c (vect_strip_conversions): New function.
2845 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
2846 check for binary class.
2848 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
2852 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
2854 * gimple-low.c (mark_blocks_with_used_vars): New function.
2855 (mark_blocks_with_used_subblocks): Ditto.
2856 (mark_used_blocks): Ditto.
2857 (pass_mark_used_blocks): New pass.
2858 * tree-inline.c: Include debug.h.
2859 (expand_call_inline): Call outlining_inline_function here.
2860 * tree-optimize.c (init_tree_optimization_passes): Add
2861 pass_mark_used_blocks.
2862 * tree-pass.h (pass_mark_used_blocks): New.
2863 * Makefile.in (tree-inline.o): Add debug.h dependency.
2865 2005-01-03 Geoffrey Keating <geoffk@apple.com>
2867 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
2868 VAR_DECLs to have weak_import attribute.
2870 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
2873 2004-01-03 Steven Bosscher <stevenb@suse.de>
2875 * passes.c (rest_of_compilation): Don't run regmove if only
2876 flag_expensive_optimizations, require flag_regmove instead.
2878 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
2880 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
2882 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
2883 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
2884 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
2885 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
2887 2005-01-03 Richard Henderson <rth@redhat.com>
2889 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
2890 (ia64_const_ok_for_letter_p): New.
2891 (ia64_const_double_ok_for_letter_p): New.
2892 (ia64_extra_constraint): New.
2893 (ia64_expand_vecint_compare): New.
2894 (ia64_expand_vcondu_v2si): New.
2895 (ia64_expand_vecint_cmov): New.
2896 (ia64_expand_vecint_minmax): New.
2897 (ia64_print_operand): Add 'v'.
2898 (ia64_preferred_reload_class): New.
2899 (ia64_vector_mode_supported_p): New.
2900 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
2901 (PREFERRED_RELOAD_CLASS): Move to function.
2902 (CONST_OK_FOR_LETTER_P): Move to function.
2903 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
2904 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
2905 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
2906 (EXTRA_CONSTRAINT): Move to function.
2907 * config/ia64/ia64.md: Include vect.md.
2908 (itanium_class): Add mmalua.
2910 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
2912 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
2913 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
2915 (const_int_2bit_operand): New.
2916 (fr_reg_or_0_operand): New.
2917 * config/ia64/ia64-modes.def: Add vector modes.
2918 * config/ia64/ia64-protos.h: Update.
2919 * config/ia64/vect.md: New file.
2921 2005-01-03 Richard Henderson <rth@redhat.com>
2923 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
2925 2005-01-03 Uros Bizjak <uros@kss-loka.si>
2928 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
2930 (log1pdf2): Change mode of operands[0,1] to DFmode.
2932 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
2934 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
2935 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
2936 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
2937 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
2938 SPARC_RELAXED_ORDERING.
2940 2005-01-03 Richard Henderson <rth@redhat.com>
2941 Uros Bizjak <uros@kss-loka.si>
2944 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
2945 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
2946 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
2948 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
2949 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
2950 const_pow2_1_to_128_operand): New predicates.
2952 2005-01-02 Greg McGary <greg@mcgary.org>
2954 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
2956 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
2957 object. [ptr_type]: Remove unused variable.
2958 Remember nearest addressable array-element or record-component
2959 when checking bitfield components. Tolerate empty BIND_EXPR.
2961 2005-01-01 Richard Henderson <rth@redhat.com>
2963 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
2964 that don't narrow the value. Fail for other conversions.
2966 2005-01-01 Richard Henderson <rth@redhat.com>
2969 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
2970 * c-lang.c (finish_file): Don't do it here.
2971 * objc/objc-act.c (objc_finish_file): Likewise.
2973 * cgraph.c (decl_assembler_name_equal): New.
2974 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
2975 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
2976 * cgraph.h (struct cgraph_varpool_node): Add next.
2977 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
2978 * varasm.c (assemble_alias): Mark the target as needed.
2980 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
2983 * function.c (get_arg_pointer_save_area): Use entry_of_function
2984 instead of get_insns.
2986 2005-01-01 Roger Sayle <roger@eyesopen.com>
2987 Andrew Pinski <pinskia@physics.uc.edu>
2988 James E. Wilson <wilson@specifixinc.com>
2990 PR rtl-optimization/12092
2991 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
2994 2005-01-01 Roger Sayle <roger@eyesopen.com>
2995 Olivier Hainque <hainque@act-europe.fr>
2997 * tree.c (int_fits_type_p): A narrower type always fits in a
2998 wider one, except for negative values into unsigned types.
3000 2005-01-01 Roger Sayle <roger@eyesopen.com>
3002 * tree.c (int_fits_type_p): Compare the result of force_fit_type
3003 with the original constant rather than require TREE_OVERFLOW.
3005 2005-01-01 Steven Bosscher <stevenb@suse.de>
3008 * c-decl.c (finish_function): If compiling C99, annotate the
3009 compiler generated return with the current file name and line 0.
3010 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
3011 the source line is greater than 0.
3012 (remove_bb): Likewise.
3014 See ChangeLog.12 for earlier changes.