1 2005-01-26 Richard Henderson <rth@redhat.com>
4 * combine.c (make_field_assignment): Simplify store to zero_extract
5 from a source with an overlapping mask.
7 2005-01-26 Aldy Hernandez <aldyh@redhat.com>
9 2004-11-11 Eric Christopher <echristo@redhat.com>
11 * config/frv/frv.md: Add fr400_integer automaton. Don't
12 allow TYPE_MUL, TYPE_MACC, or TYPE_CUT to issue in parallel
15 2005-01-26 Steven Bosscher <stevenb@suse.de>
18 * cfgbuild.c (make_edges): Do not clear or set
19 current_function_has_computed_jump.
20 * function.h (struct function): Remove the has_computed_jump field.
21 (current_function_has_computed_jump): Do not define.
22 * sched-rgn.c (is_cfg_nonregular): Return true if a basic block ends
23 in a computed jump. Ignore current_function_has_computed_jump.
25 2005-01-26 Richard Henderson <rth@redhat.com>
28 * expr.c (categorize_ctor_elements): New argument p_must_clear.
29 (categorize_ctor_elements_1): Likewise. Detect a union that isn't
31 (mostly_zeros_p): Update for new categorize_ctor_elements argument.
32 * gimplify.c (gimplify_init_constructor): Likewise. Only shove
33 objects into static storage if they have more than one non-zero value.
34 * tree.h (categorize_ctor_elements): Update decl.
36 2005-01-26 Steven Bosscher <stevenb@suse.de>
39 * tree.h (CALL_EXPR_TAILCALL): Add comment.
40 * calls.c (check_sibcall_argument_overlap_1): Revert the change
41 to this function from 2004-07-10.
42 * tree-tailcall.c (suitable_for_tail_opt_p): Do not consider the
43 the current function for tail call optimizations if the address
44 of one of it its arguments is taken.
46 2005-01-26 Kazu Hirata <kazu@cs.umass.edu>
48 * cse.c (fold_rtx): Call equiv_constant only when necessary.
50 2005-01-26 Bernd Schmidt <bernd.schmidt@analog.com>
52 * function.c (expand_function_end): If current_function_calls_alloca,
53 emit stack restore in a place that is reached when the function falls
56 2005-01-26 Alexandre Oliva <aoliva@redhat.com>
58 * Makefile.in (../$(build_subdir)/fixincludes/fixincl): Add dummy
59 rule to avoid matching inappropriate suffix rule.
61 2005-01-26 Jan Hubicka <jh@suse.cz>
63 PR tree-optimization/19241
64 * tree-inline.c (copy_body_r): Do not walk subtrees after substituting.
66 2005-01-26 Hans-Peter Nilsson <hp@axis.com>
68 * config/cris/cris.h (LIB_SPEC) <sim*>: Use -lsyslinux, not
69 -lsyssim. Wrap -lc -lsyssim in --start-group --end-group instead
71 (STARTFILE_SPEC): Add crti.o, before crtbegin.o.
72 <sim*>: Always use crt1.o, regardless of N in simN.
73 <!sim>: Drop support for -pg and -p variants; always use crt0.o.
74 (ENDFILE_SPEC): Add crtn.o, after crtend.o.
75 (CRT_CALL_STATIC_FUNCTION): Remove.
76 * config/cris/aout.h (STARTFILE_SPEC): <!melinux && sim*>: Always
77 use crt1.o, regardless of N in simN.
78 <!sim>: Drop support for -pg and -p variants; always use crt0.o.
79 (LIB_SPEC): Do not define; default to the one in config/cris/cris.h.
80 (CRIS_CPP_SUBTARGET_SPEC) <melinux && !nostdinc>: Remove special
82 (MAX_OFILE_ALIGNMENT): Correct to 32, not 16.
84 2005-01-25 Ian Lance Taylor <ian@airs.com>
86 * gimple-low.c (block_may_fallthru): Correct handling of
87 SWITCH_EXPR--may fall through if SWITCH_LABELS is NULL.
89 2005-01-26 Hans-Peter Nilsson <hp@axis.com>
91 * config/cris/cris.h (CPP_SPEC): Adjust the various mtune=
92 handlers to match indentation; nested under the mtune=* handler.
93 Similar for mcpu= and march=.
95 2005-01-25 Daniel Jacobowitz <dan@codesourcery.com>
97 * config/mips/linux.h (LIB_SPEC): Remove -rpath-link.
98 * config/mips/linux64.h (LIB_SPEC): Likewise.
99 * config/mn10300/linux.h (LIB_SPEC): Likewise.
100 * config/sh/linux.h (LIB_SPEC): Likewise.
102 2005-01-25 Aldy Hernandez <aldyh@redhat.com>
104 * config/frv/frv.c (frv_legitimize_tls_address): New.
105 (TARGET_HAVE_TLS): Define.
106 (FRV_SYMBOL_REF_TLS_P): Define.
107 (frv_override_options): Handle new register classes.
108 (frv_legitimate_address_p): Reject tls addresses.
109 (frv_legitimize_address): Handle TLS addresses.
110 (gen_inlined_tls_plt): New.
112 (frv_legitimize_tls_address): New.
113 (unspec_got_name): Add TLS entries.
114 (got12_operand): Add R_FRV_TLSMOFF12 case.
115 (frv_emit_move): Fixup TLS addresses.
116 (frv_emit_movsi): Legitimize TLS addresses.
118 * config/frv/frv.h (MASK_BIG_TLS): New.
119 (TARGET_BIG_TLS): New.
120 (HAVE_AS_TLS): Define.
121 (TARGET_SWITCHES): Add -mTLS and -mtls options.
122 (enum reg_class): Add GR8_REGS, GR9_REGS, GR89_REGS.
123 (REG_CLASS_NAMES): Same.
124 (REG_CLASS_CONTENTS): Same.
125 (CONSTRAINT_LEN): New.
126 (REG_CLASS_FROM_CONSTRAINT): New.
127 (PREDICATE_CODES): Add symbolic_operand.
129 * config/frv/frv.md (define_constants): Add UNSPEC_GETTLSOFF,
130 UNSPEC_TLS_LOAD_GOTTLSOFF12, UNSPEC_TLS_INDIRECT_CALL,
131 UNSPEC_TLS_TLSDESC_LDD, UNSPEC_TLS_TLSDESC_LDD_AUX,
132 UNSPEC_TLS_TLSOFF_LD, UNSPEC_TLS_LDDI, UNSPEC_TLSOFF_HILO,
133 R_FRV_GOTTLSOFF_HI, R_FRV_GOTTLSOFF_LO, R_FRV_TLSMOFFHI,
134 R_FRV_TLSMOFFLO, R_FRV_TLSMOFF12, R_FRV_TLSDESCHI,
135 R_FRV_TLSDESCLO, R_FRV_GOTTLSDESCHI, R_FRV_GOTTLSDESCLO, GR8_REG,
136 GR9_REG, GR14_REG, LRREG.
137 (type): Add load_or_call attribute.
138 ("load_or_call"): New reservation.
139 ("call_gettlsoff"): New.
140 ("tls_indirect_call"): New.
141 ("tls_load_gottlsoff12"): New.
142 ("tlsoff_hilo"): New.
143 ("tls_tlsdesc_ldd"): New.
144 ("tls_tlsoff_ld"): New.
147 * config/frv/frv-protos.h (symbolic_operand): Protoize.
149 2005-01-25 Alexandre Oliva <aoliva@redhat.com>
151 * config/frv/frv.h (ASM_OUTPUT_DWARF_DTPREL): Define.
152 * config/frv/frv-protos.h (frv_output_dwarf_dtprel): Declare.
153 * config/frv/frv.c (TLS_BIAS): Define.
154 (frv_output_dwarf_dtprel): New.
156 2005-01-26 Jakub Jelinek <jakub@redhat.com>
158 PR rtl-optimization/19579
159 * ifcvt.c (noce_try_cmove_arith): If emitting instructions to set up
160 both A and B, see if they don't clobber registers the other expr uses.
162 2005-01-25 J"orn Rennecke <joern.rennecke@st.com>
164 * real.c (do_add): Initialize signalling and canonical members.
166 * real.c (real_from_integer): Zero out destination.
168 2005-01-18 Jan Hubicka <jh@suse.cz>
170 PR tree-optimize/19337
171 * tree-inline.c (initialize_inlined_parameters): Copy
172 saved_static_chain_decl only.
174 2005-01-25 Richard Henderson <rth@redhat.com>
178 * config/i386/i386.c (x86_inter_unit_moves): Zero.
179 (ix86_preferred_reload_class): Rewrite fp-constant section, with
180 80387 enabled, to return a proper subclass. Return the subset that
181 overlaps with GENERAL_REGS for PLUS.
182 * config/i386/i386.md (movsi_1, movdi_2): Set type to "mmx" for pxor.
184 2005-01-25 Richard Henderson <rth@redhat.com>
187 * tree-complex.c (expand_complex_div_wide): Use the correct formulae.
189 2005-01-25 Roger Sayle <roger@eyesopen.com>
191 * expmed.c (choose_multiplier): Change interface to return the
192 multiplier lowpart as an rtx instead of an unsigned HOST_WIDE_INT.
193 (expand_divmod): Update calls to choose_multiplier.
195 2005-01-24 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
198 * expmed.c (extract_bit_field): Use adjust_address instead of
199 gen_lowpart when op0 is a MEM.
201 2005-01-24 Steven Bosscher <stevenb@suse.de>
203 * cfgexpand.c (expand_gimple_tailcall): Fix typo.
205 2005-01-24 Richard Sandiford <rsandifo@redhat.com>
207 * config/mips/irix-csr.c: New file.
208 * config/mips/t-iris6 (irix-csr.o): New rule to build it.
209 (EXTRA_MULTILIB_PARTS): Add irix-csr.o.
210 * config/mips/iris6.h (ENDFILE_SPEC): Include it in n32 and n64
213 2005-01-24 Eric Botcazou <ebotcazou@libertysurf.fr>
216 * config.gcc (sparc-*-elf*): Remove sol2.h, sparc/sol2.h and
217 sparc/elf.h, add sparc/sp-elf.h.
218 (sparc-*-rtems*): Likewise.
219 (sparclite-*-elf*): Remove sol2.h, sparc/sol2.h, sparc/elf.h and
220 tm-dwarf2.h, add sparc/sp-elf.h.
221 (sparc86x-*-elf): Likewise.
222 (sparc64-*-elf*): Remove sol2.h, sparc/sol2.h and tm-dwarf2.h.
223 * config/sparc/liteelf.h (TARGET_SUB_OS_CPP_BUILTINS): Rename into
224 TARGET_OS_CPP_BUILTINS.
225 * config/sparc/sp86x-elf (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
226 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
228 * config/sparc/openbsd64.h (NO_IMPLICIT_EXTERN_C): Undefine.
229 * config/sparc/sp64-elf.h (NO_IMPLICIT_EXTERN_C): New macro.
230 (SWITCH_TAKES_ARG): Likewise.
231 (LOCAL_LABEL_PREFIX): Likewise.
232 (ASM_GENERATE_INTERNAL_LABEL): Likewise.
233 (TARGET_N_FORMAT_TYPES): Delete.
234 (TARGET_FORMAT_TYPES): Likewise.
235 (ASM_DECLARE_FUNCTION_SIZE): Likewise.
236 * config/sparc/elf.h: Delete.
237 * config/sparc/sp-elf.h: New file.
239 2005-01-24 Kazu Hirata <kazu@cs.umass.edu>
241 * tree-flow-inline.h (phi_arg_from_edge): Remove.
242 * tree-flow.h: Remove the corresponding prototype.
243 * tree-ssa-operands.h, tree-ssa-threadupdate.c, tree-ssa.c,
244 tree-vectorizer.c: Use dest_idx instead of phi_arg_from_edge.
246 2005-01-24 Jakub Jelinek <jakub@redhat.com>
249 * config/rs6000/rs6000.c (rs6000_file_start): Create toc section
250 for AIX ABI or ELF -fPIC.
251 (rs6000_emit_load_toc_table): Don't create toc_section here.
252 (rs6000_xcoff_file_start): Nor here.
254 2005-01-24 J"orn Rennecke <joern.rennecke@st.com>
256 * sh.c (ra.h): Don't #include.
257 (hard_regs_intersect_p): New function, resurrected from ra.c.
259 * sh.c: Fix 1996 Copyright.
261 2005-01-24 Richard Henderson <rth@redhat.com>
262 Aldy Hernandez <aldyh@redhat.com>
264 * regrename.c (note_sets): Handle subregs.
266 2005-01-24 Joseph S. Myers <joseph@codesourcery.com>
269 * genconditions.c (write_header, write_conditions): Elide file if
272 2005-01-24 Kazu Hirata <kazu@cs.umass.edu>
274 * c-opts.c, expmed.c, gengtype.c, tree-complex.c,
275 config/arm/iwmmxt.md, config/avr/avr.c, config/s390/s390.h:
278 2005-01-24 Jakub Jelinek <jakub@redhat.com>
280 * builtins.c (get_pointer_alignment, gimplify_va_arg_expr,
281 expand_builtin_printf, expand_builtin_fprintf,
282 expand_builtin_sprintf): Use POINTER_TYPE_P instead TREE_CODE
283 checking against POINTER_TYPE.
284 (validate_arglist): Handle POINTER_TYPE code by checking
287 2005-01-24 Paolo Bonzini <bonzini@gnu.org>
289 * rtlanal.c (non_rtx_starting_operands, for_each_rtx_1,
291 (for_each_rtx): Call for_each_rtx_1.
292 * rtl.h (init_rtlanal): Declare.
293 * toplev.c (backend_init): Call init_rtlanal.
295 2005-01-24 Jakub Jelinek <jakub@redhat.com>
297 * flow.c (propagate_one_insn): Formatting.
300 * flow.c (libcall_dead_p): Be more conservative if unsure.
301 If there are any instructions between insn and call, see if they are
302 all dead before saying the libcall is dead.
304 2005-01-24 Uros Bizjak <uros@kss-loka.si>
306 * config/i386/i386.md (*extendsfdf2_sse): Use "nonimmediate_operand"
307 operand constraint for operand 0.
309 2005-01-23 Richard Henderson <rth@redhat.com>
312 * tree-complex.c (expand_complex_div_wide): Set EDGE_FALLTHRU on
313 edges leading to join block.
315 2005-01-23 Roger Sayle <roger@eyesopen.com>
317 * expmed.c (expand_mult_highpart): Make static. Change type of
318 constant multiplier argument from unsigned HOST_WIDE_INT to rtx.
319 (expand_divmod): Updates calls to expand_mult_highpart by using
320 gen_int_mode to pass a CONST_INT rtx.
321 * rtl.h (expand_mult_highpart): Remove prototype.
323 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
325 * flow.c (regset_bytes, regset_size): Remove.
327 * cfganal.c: Fix a reference to Harvey's paper.
329 2005-01-23 Steven Bosscher <stevenb@suse.de>
331 PR rtl-optimization/19464
332 * tree-optimize.c (init_tree_optimization_passes): Add one more
333 copyrename pass just before out-of-ssa.
335 2005-01-23 Joseph S. Myers <joseph@codesourcery.com>
338 * recog.c (recog_memoized): Don't define if GENERATOR_FILE.
339 * ggc-none.c (ggc_free): Define.
341 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
343 * cse.c (max_reg, max_insn_uid): Remove.
344 (cse_main): Don't access max_reg or max_insn_uid.
346 * c-common.c (explicit_flag_signed_bitfields,
347 lang_statement_code_p, lang_expand_function_end): Remove.
348 * c-common.h: Remove the corresponding declarations.
349 * c-opts.c (c_common_handle_option): Remove all write access
350 to explicit_flag_signed_bitfields.
352 * cgraph.c (cgraph_varpool_n_nodes): Remove.
353 (cgraph_varpool_node): Don't access cgraph_varpool_n_nodes.
354 * cgraph.h: Remove the corresponding declaration.
356 * gcse.c (null_pointer_info): Remove.
358 2005-01-23 Roger Sayle <roger@eyesopen.com>
359 Eric Botcazou <ebotcazou@libertysurf.fr>
361 * combine.c (simplify_logical): Only simplify logical expressions
362 of the form ior(and(x,y),z) by the inverse distributive law if the
363 result is cheaper than the original.
365 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
367 * alias.c, c-common.h, c-incpath.c, c-incpath.h, expr.c,
368 fold-const.c, gimplify.c, params.h, tree-data-ref.c,
369 tree-if-conv.c, tree-nested.c, tree-outof-ssa.c,
370 tree-ssa-dom.c, tree-vectorizer.c, tree.def, config/darwin.c,
371 config/freebsd-spec.h, config/arm/arm.h,
372 config/h8300/h8300.md, config/i386/i386.md,
373 config/i386/predicates.md, config/i386/sse.md,
374 config/ia64/ia64.c, config/ip2k/ip2k.c, config/s390/s390.c,
375 config/vax/vax.md: Fix comment typos. Follow spelling
378 2005-01-23 Steven Bosscher <stevenb@suse.de>
380 * gengtype.c (walk_type): Produce `not equal to' compares for loop
381 tests, instead of `less than'.
383 2005-01-23 Richard Sandiford <rsandifo@redhat.com>
385 * function.c (assign_parm_setup_block): When creating a new stack slot
386 for a parameter, get its alignment from the parameter's DECL_ALIGN
387 rather than the type's TYPE_ALIGN. Make sure that the parameter
388 is at least word aligned.
390 2005-01-22 Roger Sayle <roger@eyesopen.com>
393 * config/avr/avr.c (avr_hard_regno_mode_ok): Rewrite.
395 2005-01-22 Richard Henderson <rth@redhat.com>
398 * config/i386/i386.md (movsfcc_1_sse_max): Use nonimmediate_operand
399 in both compare operands.
400 (movdfcc_1_sse_max): Likewise.
401 (movsfcc_1_sse): Likewise. Add earlyclobber for scratch.
402 (movdfcc_1_sse): Likewise.
403 * config/i386/i386.c (ix86_split_sse_movcc): Emit copies into the
404 scratch register as needed.
406 2005-01-22 Richard Henderson <rth@redhat.com>
408 * config/i386/i386.md (smaxsf3): Fix mnemonic typo.
410 * config/i386/i386.c (ix86_prepare_fp_compare_args): Fix is_sse test.
412 2005-01-22 Richard Henderson <rth@redhat.com>
414 * genopinit.c (optabs): Use smin/smax for floating point too.
415 * doc/md.texi: Update to match. Clarify that floating point
416 results are undefined for +0/-0 and NaN.
417 * doc/rtl.texi: Likewise.
418 * rtl.def (SMIN, SMAX): Likewise
419 * tree.def (MIN_EXPR, MAX_EXPR): Likewise.
421 * config/alpha/alpha.md (smaxdf3, smindf3, smaxsf3, sminsf3): Add
422 leading 's' to the name.
423 * config/ia64/ia64.md (smaxsf3, sminsf3, smaxdf3, smindf3,
424 smaxxf3, sminxf3): Likewise.
425 * config/rs6000/rs6000.md (smaxdf3, smindf3, smaxsf3, sminsf3):
428 2005-01-22 Paul Brook <paul@codesourcery.com>
430 * config/arm/arm.md: Use "Uy" constraint for wcgr load/stores.
432 2005-01-21 Mark Dettinger <dettinge@de.ibm.com>
434 * config/s390/s390.c (struct processor_costs): New fields
435 dlgr, dlr, dr, dsgfr, dsgr.
436 (z900_cost, z990_cost): Values for new fields.
437 (s390_rtx_costs): New cases MEM und COMPARE in switch
438 statement. Modified handling of SIGN_EXTEND, ZERO_EXTEND,
439 DIV, MOD, UDIV, UMOD.
441 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
443 * config/s390/s390.md ("doloop_si64"): Reload input value directly
444 into the register being decremented.
445 ("doloop_si31", "doloop_di"): Likewise.
446 ("*doloop_si_long"): Adapt pattern.
448 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
450 * config/s390/s390.h (HARD_REGNO_NREGS): Fix computation for
452 (CLASS_MAX_NREGS): Likewise.
454 2005-01-21 Daniel Berlin <dberlin@dberlin.org>
456 * doc/tree-ssa.texi (Statement Operands): Add example for new
457 must-def macro. Note deprecation of old operands interface.
459 2005-01-22 Richard Sandiford <rsandifo@redhat.com>
461 PR tree-optimization/19484
462 * tree-cfg.c (remove_fallthru_edge): New function.
463 (cleanup_control_flow): Remove fallthru edges from calls that are
464 now known not to return.
466 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
468 Fix PR tree-optimization/19038
469 * tree-ssa-dom.c (cprop_operand): Don't replace loop invaeriant
470 copies with loop variant ones.
472 2005-01-22 Kazu Hirata <kazu@cs.umass.edu>
474 * cfganal.c, real.h, reorg.c, timevar.def, tree-ssa-ccp.c,
475 config/alpha/alpha-protos.h, config/alpha/alpha.h,
476 config/alpha/alpha.md, config/alpha/predicates.md,
477 config/sparc/freebsd.h, config/sparc/netbsd-elf.h,
478 config/sparc/sol2.h: Update copyright.
480 * tree-cfg.c (remove_forwarder_block_with_phi): Look at the
481 first label to see if it is a nonlocal label.
483 2005-01-22 David Edelsohn <edelsohn@gnu.org>
484 Andrew Pinski <pinskia@physics.uc.edu>
487 * config/rs6000/rs6000.c (rs6000_va_start): Saturate n_gpr at
488 maximum number of GPRs. Saturate n_fpr at maximum number of FPRs.
490 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
492 * config/rs6000/rtems.h: Update copyright.
494 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
497 * config/rs6000/rtems.h: Resurrect cpp_os_rtems_spec from gcc < 3.4.
498 (CPP_OS_RTEMS_SPEC): New (From gcc-3.3's config/rs6000/sys4.h).
499 (SUBSUBTARGET_EXTRA_SPECS): Use CPP_OS_RTEMS_SPEC.
501 2005-01-22 Volker Reichelt <reichelt@igpm.rwth-aachen.de>
504 * c-typeck.c (convert_arguments): Check for error_mark_node.
506 2005-01-21 Ian Lance Taylor <ian@airs.com>
508 PR tree-optimization/13000
509 * tree-inline.c: Include "tree-flow.h".
510 (expand_call_inline): If warn_return_type, warn if non-void inline
511 function falls through.
512 * tree-cfg.c (execute_warn_function_return): Don't warn about
513 control reaching end if TREE_NO_WARNING is set. Set
515 * gimple-low.c (block_may_fallthru): Don't assume that SWITCH_EXPR
517 * gimplify.c (shortcut_cond_expr): Don't emit a jump over the else
518 branch if we don't need one.
519 * c-typeck.c: Include "tree-flow.h"
520 (c_finish_bc_stmt): Don't add a goto if the current statement
521 list doesn't fall through to the current point.
523 2005-01-21 Roger Sayle <roger@eyesopen.com>
525 PR rtl-optimization/576
526 * real.c (real_arithmetic): Change return type from void to bool
527 to return an indication that the result may be inexact.
528 * real.h (real_arithmeric): Update prototype.
529 * fold-const.c (const_binop): Don't constant fold floating
530 point expressions when the user specifies -frounding-math and
531 the result may depend upon the run-time rounding mode.
532 (fold_convert_const_real_from_real): Clean-up.
533 (fold_initializer): Ignore flag_rounding_math for initializers.
534 * simplify-rtx.c (simplify_binary_operation): Likewise, don't
535 constant fold FP operations with flag_rounding_math if the
536 result may depend upon the run-time rounding mode.
538 2005-01-21 Tom Tromey <tromey@redhat.com>
540 * c-cppbuiltin.c (define__GNUC__): Correct assertion.
542 2005-01-21 Kazu Hirata <kazu@cs.umass.edu>
544 * tree-cfg.c: Fix comment typos.
546 2005-01-21 J"orn Rennecke <joern.rennecke@st.com>
548 * passes.c (rest_of_handle_flow2): Close / open the flow2 dump file
549 around the call to rest_of_handle_branch_target_load_optimize.
551 2005-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
553 * config.gcc (sparc-*-netbsdelf*): Include dbxelf.h
554 (sparc64-*-openbsd*): Likewise.
555 (sparclite-*-elf*): Include tm-dwarf2.h.
556 (sparc86x-*-elf*): Likewise.
557 (sparc64-*-elf*): Likewise.
558 * config/sparc/linux64.h (DWARF2_DEBUGGING_INFO): Delete.
559 (DBX_DEBUGGING_INFO): Likewise.
560 * config/sparc/liteelf.h (DWARF2_DEBUGGING_INFO): Likewise.
561 (PREFERRED_DEBUGGING_TYPE): Likewise.
562 * config/sparc/netbsd-elf.h (PREFERRED_DEBUGGING_TYPE): Likewise.
563 * config/sparc/sp64-elf.h (DBX_DEBUGGING_INFO): Likewise.
564 (PREFERRED_DEBUGGING_TYPE): Likewise.
565 * config/sparc/sp86x-elf.h (DWARF2_DEBUGGING_INFO): Likewise.
566 (PREFERRED_DEBUGGING_TYPE): Likewise.
567 * config/sparc/sparc.h (DBX_DEBUGGING_INFO): Likewise.
569 2005-01-21 Hans-Peter Nilsson <hp@bitrange.com>
572 * combine.c (combine_simplify_rtx): Revert change of 2004-12-31.
574 2005-01-20 Paul Brook <paul@codesourcery.com>
576 * doc/rtl.texi: Document value extension requirements for CONST_INT.
578 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
580 * tree-cfg.c (tree_verify_flow_info): Check that a nonlocal
581 label is first in a sequence of labels.
583 * tree-cfg.c (tree_verify_flow_info): Fix a typo.
585 2005-01-20 Janis Johnson <janis187@us.ibm.com>
586 Giovanni Bajo <giovannibajo@gcc.gnu.org>
588 * doc/sourcebuild.texi (Test Directives): New.
590 2005-01-20 Roger Sayle <roger@eyesopen.com>
592 * real.c (real_floor): Don't leave the result uninitialized when
593 mode is VOIDMode, but return the unrounded intermediate value.
594 (real_ceil): Likewise.
596 2005-01-20 Paul Brook <paul@codesourcery.com>
598 * config/arm/arm.md (insv): Use gen_int_mode.
600 2005-01-20 Eric Botcazou <ebotcazou@libertysurf.fr>
602 * config/sparc/freebsd.h (ASM_OUTPUT_INTERNAL_LABELREF): Delete.
603 * config/sparc/linux.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
604 * config/sparc/linux64.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
605 * config/sparc/netbsd-elf.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
606 * config/sparc/sol2.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
608 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
611 * dwarf2out.c (concat_loc_descriptor): We don't know we can use
612 fbreg, so we have to assume we can't.
613 (loc_descriptor_from_tree_1): Ditto.
614 (containing_function_has_frame_base): New function.
615 (add_location_or_const_value_attribute): Use it.
616 Also try to generate a frame_base from a single element location
619 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
621 PR tree-optimization/15349
622 * timevar.def (TV_TREE_MERGE_PHI): New.
623 * tree-cfg.c (tree_forwarder_block_p): Add a new argument
625 (remove_forwarder_block, cleanup_forwarder_blocks): Adjust the
626 calls to tree_forwarder_block_p.
627 (remove_forwarder_block_with_phi, merge_phi_nodes,
628 gate_merge_phi, pass_merge_phi): New.
629 * tree-optimize.c (init_tree_optimization_passes): Add
631 * tree-pass.h: Add an extern for pass_merge_phi;
633 2005-01-20 Richard Henderson <rth@redhat.com>
636 * config/i386/emmintrin.h (_mm_castpd_ps, _mm_castpd_si128): New.
637 (_mm_castps_pd, _mm_castps_si128): New.
638 (_mm_castsi128_ps, _mm_castsi128_pd): New.
640 2005-01-20 Richard Henderson <rth@redhat.com>
643 * config/i386/mmintrin.h (_mm_cvtsi32_si64): Use
644 __builtin_ia32_vec_init_v2si.
645 (_mm_cvtsi64_si32): Use __builtin_ia32_vec_ext_v2si.
646 * config/i386/i386.c (IX86_BUILTIN_VEC_EXT_V2SI): New.
647 (ix86_init_mmx_sse_builtins): Create it.
648 (ix86_expand_builtin): Expand it.
649 (ix86_expand_vector_set): Handle V2SFmode and V2SImode.
650 * config/i386/mmx.md (vec_extractv2sf_0, vec_extractv2sf_1): New.
651 (vec_extractv2si_0, vec_extractv2si_1): New.
653 2005-01-20 Richard Henderson <rth@redhat.com>
656 * config/i386/i386.c (ix86_expand_push): New.
657 * config/i386/mmx.md (push<MMXMODE>1): New.
658 * config/i386/sse.md (push<SSEMODE>1): New.
659 * config/i386/i386-protos.h: Update.
661 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
663 * params.def, params.h, predict.c, tree-ssa-loop-im.c: Update
666 2005-01-20 Richard Henderson <rth@redhat.com>
668 * rtl.def (CONST_VECTOR): Use RTX_CONST_OBJ.
669 * rtl.h (CONSTANT_P): Don't special case CONST_VECTOR.
671 2005-01-19 Richard Henderson <rth@redhat.com>
674 * config/i386/i386.c (ix86_expand_vector_move_misalign): Convert
675 to V4SFmode in SSE1 fallback load path.
677 2005-01-19 Richard Henderson <rth@redhat.com>
679 * config/i386/i386.c (ix86_expand_vector_init_one_var): Fix typo
680 in QImode expansion to ix86_expand_vector_set.
681 (ix86_expand_vector_init_general): Fix typo in V8HImode recursive call.
683 2005-01-19 Richard Henderson <rth@redhat.com>
686 * config/i386/i386.c (ix86_preferred_reload_class): Return a proper
687 subclass of the input class.
688 (ix86_secondary_memory_needed): Always true for cross-MMX classes.
689 Always true for cross-SSE1 classes. Rationalize conditionals.
690 * config/i386/i386.h (SSE_CLASS_P, MMX_CLASS_P): Use straight equality.
691 * config/i386/i386.md (movsi_1): Add MMX/SSE zeros. Fix alternatives
692 for SSE1. Don't check TARGET_INTER_UNIT_MOVES.
693 (movdi_2): Add MMX/SSE zeros.
694 (movdi_1_rex64): Likewise. Don't check TARGET_INTER_UNIT_MOVES.
695 (movsf_1): Don't check TARGET_INTER_UNIT_MOVES.
696 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
697 (movsi_1_nointernunit, movdi_1_rex64_nointerunit): Remove.
698 (movsf_1_nointerunit, zero_extendsidi2_32_1): Remove.
699 (zero_extendsidi2_rex64_1): Remove.
700 (MOV0 peephole): Check GENERAL_REG_P.
702 2005-01-19 Richard Henderson <rth@redhat.com>
705 * config/i386/i386.c (ix86_expand_vector_set): Fix third and fourth
707 (ix86_expand_vector_extract): Likewise.
709 2005-01-19 Richard Henderson <rth@redhat.com>
712 * expr.c (emit_move_change_mode): New argument force; use
713 simplify_gen_subreg if true.
714 (emit_move_via_alt_mode): Merge into ...
715 (emit_move_via_integer): ... here.
716 (emit_move_ccmode): Use emit_move_change_mode directly.
718 2005-01-19 Richard Henderson <rth@redhat.com>
721 * config/alpha/alpha.c (alpha_rtx_costs): Handle HIGH.
722 (alpha_preferred_reload_class): Handle CONST_VECTOR.
723 (alpha_emit_set_const_1): Add no_output parameter; don't emit
725 (alpha_emit_set_const): Likewise. Make static.
726 (alpha_emit_set_long_const): Make static.
727 (alpha_extract_integer): Split out from alpha_expand_mov.
728 (alpha_split_const_mov): Likewise.
729 (alpha_expand_mov): Use them. Handle CONST_VECTOR.
730 (alpha_legitimate_constant_p): New.
731 * config/alpha/alpha-protos.h: Update.
732 * config/alpha/alpha.h (REGISTER_MOVE_COST): Correct fp<->gp cost.
733 (LEGITIMATE_CONSTANT_P): Re-implement with a function.
734 * config/alpha/alpha.md (movsi): Add n alternative.
735 (movsi_nt_vms, movdi_er_nofix, movdi_er_fix, movdi_fix): Likewise.
736 (mov<VEC>_fix, mov<VEC>_nofix): Add i alternative.
737 (splitters for all of the above): Use alpha_split_const_mov.
738 * config/alpha/predicates.md (non_add_const_operand): New.
739 (non_zero_const_operand): New.
740 (input_operand): Use alpha_legitimate_constant_p after reload.
742 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
744 PR tree-optimization/19038
745 * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): New function.
746 (add_candidate): Add ivs with increment in latch only if
747 allow_ip_end_pos_p is true.
748 (determine_iv_cost): Use empty_block_p.
750 2005-01-19 Daniel Berlin <dberlin@dberlin.org>
752 * cfganal.c (compute_dominance_frontiers_1): Replace with new algorithm
753 (compute_dominance_frontiers): Ditto.
755 2005-01-19 Ralf Corsepius <ralf.corsepius@rtems.org>
758 * config/sh/t-rtems: New.
759 * config.gcc (sh-*-rtems*): Reflect having added config/sh/t-rtems.
761 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
763 * tree-ssa-ccp.c (maybe_fold_offset_to_component_ref): Always subtract
764 the offset of the selected field.
766 2005-01-19 Kazu Hirata <kazu@cs.umass.edu>
768 * tree-cfg.c (remove_forwarder_block): Fix the check to
769 prevent a nonlocal label from appearing in the middle of a
772 2005-01-19 Hans-Peter Nilsson <hp@axis.com>
774 PR rtl-optimization/19462
775 * reorg.c (find_end_label): Create return insn only if
776 current_function_epilogue_delay_list is empty.
778 2005-01-19 Paolo Bonzini <bonzini@gnu.org>
780 * doc/passes.texi: Remove paragraph mentioning new-ra.
782 2005-01-19 Jakub Jelinek <jakub@redhat.com>
784 PR rtl-optimization/15139
785 * combine.c: Include params.h.
786 (count_rtxs): New function.
787 (record_value_for_reg): If replace_rtx would replace at least
788 2 occurrences of REG in VALUE and TEM is really large, replace REG with
789 (clobber (const_int 0)) instead of TEM.
790 * params.def (PARAM_MAX_LAST_VALUE_RTL): New.
791 * params.h (MAX_LAST_VALUE_RTL): New.
792 * Makefile.in (combine.o): Depend on $(PARAMS_H).
793 * doc/invoke.texi (--param max-last-value-rtl=N): Document.
796 * c-typeck.c (digest_init): Only call build_vector if all constructor
797 elements are *_CST nodes.
798 * gimplify.c (gimplify_init_constructor): Likewise.
801 * c-typeck.c (digest_init): Only call build_vector if inside_init
804 2005-01-18 Tobias Schl"uter <tobias.schlueter@physik.uni-muenchen.de>
806 * toplev.c (init_asm_output): Remove dead #ifdef.
808 2005-01-18 Kaz Kojima <kkojima@gcc.gnu.org>
810 * config/sh/linux.h (TARGET_C99_FUNCTIONS): Define.
812 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
814 * calls.c (expand_call): Check DECL_BUILT_IN_CLASS before
815 accessing DECL_FUNCTION_CODE.
816 * dojump.c (do_jump): Likewise.
817 * gimplify.c (gimplify_call_expr): Likewise.
818 * predict.c (expr_expected_value): Likewise.
819 (strip_builtin_expect): Likewise.
820 * tree-inline.c (estimate_num_insns_1): Likewise.
821 * tree-ssa-loop-im.c (stmt_cost): Likewise
822 * fold-const.c (fold): Test for BUILT_IN_NORMAL.
823 (tree_expr_nonnegative_p): Likewise.
825 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
827 * config/sparc/sparc.c (load_pic_register): Emit the appropriate
828 variant of the load_pcrel_sym pattern.
829 * config/sparc/sparc.md (P macro): Move to the top.
830 (load_pcrel_sym): Macroize using P.
831 (save_register_window): Likewise.
833 2005-01-18 Aldy Hernandez <aldyh@redhat.com>
835 * config/rs6000/rs6000.md ("sunordered"): Disable for e500.
838 2005-01-18 Andrew Pinski <pinskia@physics.uc.edu>
841 * c-typeck.c (build_asm_expr): Strip nops off
842 input memory operands.
844 2005-01-18 David Edelsohn <edelsohn@gnu.org>
846 * config/rs6000/aix43.h (CPLUSPLUS_CPP_SPEC): Delete
847 _XOPEN_SOURCE, _XOPEN_SOURCE_EXTENDED, _LARGE_FILE_API.
848 * config/rs6000/aix51.h (CPLUSPLUS_CPP_SPEC): Same.
849 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Same.
851 2005-01-18 Joel Sherrill <joel.sherrill@oarcorp.com>
854 * config/i386/i386.c (override_options): If the 80387 is disabled,
855 then do not return FP values using FP registers.
857 2005-01-18 Kazu Hirata <kazu@cs.umass.edu>
859 * basic-block.h, c-common.c, c-cppbuiltin.c, c-lang.c,
860 c-tree.h, cfgbuild.c, cgraph.c, cgraph.h, collect2.c,
861 combine.c, config.gcc, coverage.h, cse.c, cselib.c,
862 defaults.h, df.c, dwarf2asm.c, dwarf2out.c, explow.c, expr.c,
863 flow.c, fold-const.c, gcse.c, ggc-page.c, gimple-low.c,
864 gimplify.c, ifcvt.c, langhooks-def.h, lcm.c, optabs.h,
865 output.h, postreload-gcse.c, postreload.c, recog.c,
866 resource.c, rtl.def, rtlanal.c, sched-deps.c, sched-rgn.c,
867 targhooks.h, toplev.c, tree-data-ref.c, tree-eh.c,
868 tree-flow.h, tree-loop-linear.c, tree-mudflap.h, tree-nrv.c,
869 tree-optimize.c, tree-outof-ssa.c, tree-pass.h,
870 tree-scalar-evolution.c, tree-ssa-copy.c, tree-ssa-dce.c,
871 tree-ssa-dse.c, tree-ssa-loop-ivopts.c, tree-ssa-operands.c,
872 tree-ssa-pre.c, tree-ssa.c, tree-vectorizer.c, tree.def,
873 unwind-dw2-fde-darwin.c, var-tracking.c: Update copyright.
875 * config/darwin.c, config/darwin.h, config/freebsd-spec.h,
876 config/alpha/alpha.c, config/arm/vfp.md,
877 config/i386/emmintrin.h, config/i386/i386-modes.def,
878 config/i386/i386-protos.h, config/i386/i386.h,
879 config/i386/pmmintrin.h, config/i386/ppro.md,
880 config/i386/predicates.md, config/i386/xmmintrin.h,
881 config/i860/i860.c, config/ia64/ia64-modes.def,
882 config/ia64/ia64-protos.h, config/ia64/ia64.h,
883 config/ia64/ia64.md, config/ia64/itanium1.md,
884 config/ia64/itanium2.md, config/ia64/predicates.md,
885 config/iq2000/iq2000.h, config/mips/linux64.h,
886 config/rs6000/aix41.h, config/rs6000/aix43.h,
887 config/rs6000/aix52.h, config/rs6000/darwin-fallback.c,
888 config/rs6000/rs6000.c, config/rs6000/rs6000.h,
889 config/rs6000/spe.md, config/sh/sh.md, config/sparc/linux.h,
890 config/sparc/linux64.h, config/sparc/litecoff.h,
891 config/sparc/sp64-elf.h, config/sparc/sparc.c,
892 config/sparc/sparc.h, config/sparc/sparc.md,
893 config/sparc/sysv4-only.h: Update copyright.
895 2005-01-18 Richard Henderson <rth@redhat.com>
898 * config/i386/i386.c (ix86_expand_fp_movcc): Fail for LTGT and UNEQ.
900 2005-01-18 Marc Espie <espie@openbsd.org>
901 * config/i386/openbsdelf.h: Typo.
903 2005-01-18 Richard Henderson <rth@redhat.com>
905 * config/i386/i386.c (ix86_expand_fp_movcc): Remove TARGET_IEEE_FP
906 special case for sse.
907 * config/i386/predicates.md (sse_comparison_operator): Likewise.
909 2005-01-18 Dorit Naishlos <dorit@il.ibm.com>
911 * tree-inline.c (estimate_num_insns_1): Added cases for
912 ALIGN_INDIRECT_REF, MISALIGNED_INDIRECT_REF, and REALIGN_LOAD_EXPR.
914 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
916 PR rtl-optimization/19296
917 * combine.c (simplify_comparison): Rewrite the condition under
918 which a non-paradoxical SUBREG of a PLUS can be lifted when
919 compared against a constant.
921 2005-01-18 Andi Kleen <ak@muc.de>
923 * c-typeck.c: (convert_for_assignment): Check warn_pointer_sign.
924 * c.opt (-Wpointer-sign): Add.
925 * doc/invoke.texi: (-Wpointer-sign): Add.
927 2005-01-18 Uros Bizjak <uros@kss-loka.si>
929 * config/i386/i386.c (override_options): Revert 2004-11-24 change.
930 * config/i386/i386.md (fmodsf3, dremsf3, *sinsf2, *cossf2, sincossf3,
931 *tansf3_1, tansf2, atan2sf3_1, atan2sf3, atansf2, asinsf2, acossf2,
932 logsf2, log10sf2, log2sf2, log1psf2, logbsf2, ilogbsf2, expsf2,
933 exp10sf2, exp2sf2, expm1sf2, rintsf2, floorsf2, ceilsf2, btruncsf2,
934 nearbyintsf2): Disable for TARGET_SSE_MATH. Leave patterns enabled
935 for TARGET_MIX_SSE_I387.
936 (fmoddf3, dremdf3, *sindf2, *sinextendsfdf2, *cosdf2, *cosextendsfdf2,
937 sincosdf3, *sincosextendsfdf3, *tandf3_1, tandf2, atan2df3_1,
938 atan2df3, atandf2, asindf2, acosdf2, logdf2, log10df2, log2df2,
939 log1pdf2, logbdf2, expdf2, exp10df2, exp2df2, expm1df2, rintdf2,
940 floordf2, ceildf2, btruncdf2, nearbyintdf2): Disable for
941 (TARGET_SSE2 && TARGET_SSE_MATH). Leave patterns enabled for
943 (atan2sf3, atan2df3, atan2xf3): Remove register constraints
946 2005-01-18 Uros Bizjak <uros@kss-loka.si>
949 * config/i386/mmx.md (*movv2sf_internal_rex64, *movv2sf_internal):
950 Add movaps alternative for xmm reg->reg move.
952 2005-01-17 Jeff Law <law@redhat.com>
954 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Clear SSA_NAME_VALUE
955 at the end of the main DOM loop rather than just before DOM exits.
957 2005-01-17 Diego Novillo <dnovillo@redhat.com>
959 PR tree-optimization/19121
960 * tree-ssa-alias.c (compute_flow_sensitive_aliasing): When
961 adding aliases to a name tag, also add them to the pointer's
963 * tree-ssa-copy.c (merge_alias_info): Do not merge flow
964 sensitive alias info at all. Only check that the two pointers
965 have compatible pointed-to sets.
966 * tree-ssa.c (verify_name_tags): Verify that the alias set of
967 a pointer's type tag is a superset of the alias set of the
970 2005-01-17 James E Wilson <wilson@specifixinc.com>
973 * config/ia64/ia64.md (movxf): Handle general register source. Adjust
974 comment to document why.
976 2005-01-17 Richard Henderson <rth@redhat.com>
978 * config/i386/sse.md (smaxv4sf3_finite, sse_vmsmaxv4sf3_finite,
979 sminv4sf3_finite, sse_vmsminv4sf3_finite, smaxv2df3_finite,
980 sse2_vmsmaxv2df3_finite, sminv2df3_finite,
981 sse2_vmsminv2df3_finite): New.
982 (smaxv4sf3, sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, smaxv2df3,
983 sse2_vmsmaxv2df3, sminv2df3, sse2_vmsminv2df3): Remove commutative.
984 Force op1 into register.
986 2005-01-17 Kelley Cook <kcook@gcc.gnu.org>
988 * Makefile.in (STAGEMOVESTUFF): Stage all the stamp files to prevent
989 bubblestrap from always rebuilding build directory.
991 2005-01-17 Ulrich Weigand <uweigand@de.ibm.com>
993 * config/s390/s390.md ("*clc"): Remove incorrect '='.
994 ("*cmpmem_short"): Likewise.
996 2005-01-17 Steven Bosscher <stevenb@suse.de>
998 * basic-block.h: Document BB_* flags.
999 * regrename.c (copyprop_hardreg_forward): Don't use BB_VISITED,
1000 use an sbitmap instead.
1001 * sched-rgn.c (compute_trg_info): Likewise.
1003 2005-01-17 Richard Sandiford <rsandifo@redhat.com>
1005 * config.gcc (mips64*-*-linux*): Set the default abi to n32. Remove
1006 redundant target_cpu_default and extra_parts lines. Use the default
1008 * config/mips/linux64.h (DRIVER_DEFAULT_ABI_SELF_SPEC): Delete.
1009 (SUBTARGET_EXTRA_SPECS): Delete.
1010 (DRIVER_SELF_SPECS): Remove architecture lines. Set the ABI directly.
1011 (SUBTARGET_ASM_SPEC): Remove -mabi=n32 and -mabi=64 mappings.
1012 * config/mips/t-linux64 (MULTILIB_DIRNAMES): Use "32" for -mabi=32
1013 and "n32" for -mabi=n32.
1015 2005-01-17 Ian Lance Taylor <ian@airs.com>
1018 * c-typeck.c (build_c_cast): Revert patch of 2003-02-16: don't
1019 fold constant variables into initial values.
1021 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
1023 * tree-cfg.c (tree_can_merge_blocks_p): Reorder two checks.
1025 * tree-cfg.c (tree_forwarder_block_p): Speed up by walking
1026 through the statements backward.
1028 2005-01-17 Ian Lance Taylor <ian@airs.com>
1030 PR middle-end/13127:
1031 * tree-inline.c (expand_call_inline): Set TREE_NO_WARNING on
1032 a variable set to the return value of the inlined function.
1034 2005-01-17 Mark Dettinger <dettinge@de.ibm.com>
1036 * rtlanal.c (rtx_cost): Assign cost of 0 to a SUBREG
1037 when modes are tieable.
1039 2005-01-17 Ranjit Mathew <rmathew@hotmail.com>
1041 * gthr-posix.h (__gthread_active_p): Use pthread_cancel instead
1042 of pthread_create to find out if threads are enabled.
1043 * gthr-posix95.h (__gthread_active_p): Likewise.
1045 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
1047 * common.opt (-fnew-ra): Remove.
1049 * toplev.h (flag_new_regalloc): Remove.
1050 * Makefile.in (ra*.*): Don't mention.
1051 * passes.c (rest_of_handle_new_regalloc): Remove.
1052 (rest_of_handle_combine, rest_of_compilation): Always consider
1053 flag_new_regalloc as false.
1054 * doc/invoke.texi: Don't document -fnew-ra.
1056 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
1058 * bb-reorder.c (fix_edges_for_rarely_executed_code): Remove
1059 last parameter to reg_scan.
1060 * loop.c (loop_optimize): Likewise.
1061 * passes.c (rest_of_handle_tracer, rest_of_handle_if_conversion,
1062 rest_of_handle_web, rest_of_handle_cfg, rest_of_handle_jump_bypass,
1063 rest_of_handle_life, rest_of_handle_cse, rest_of_handle_cse2,
1064 rest_of_handle_gcse, rest_of_handle_loop_optimize,
1065 rest_of_handle_loop2, rest_of_handle_jump2): Likewise.
1066 * regclass.c (reg_scan): Likewise, for the declaration.
1067 * rtl.h (reg_scan): Likewise, for the prototype.
1069 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
1071 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Don't call
1072 mark_dfs_back_edges.
1074 2005-01-16 Daniel Berlin <dberlin@dberlin.org>
1076 * tree-ssa-pre.c (add_to_sets): s1 may be NULL.
1077 (compute_avail): Uses don't go in tmp_gen.
1079 2005-01-16 Steven Bosscher <stevenb@suse.de>
1081 * ggc-page.c (ggc_alloc_stat): Use __builtin_ctzl instead of a
1082 loop to look for a free slot in a page entry.
1084 2005-01-16 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1087 * defaults.h (TARGET_DEFERRED_OUTPUT_DEFS): Provide default.
1088 * toplev.c (compile_file): Call process_pending_assemble_output_defs
1089 just before targetm.asm_out.file_end.
1090 * tree.h (process_pending_assemble_output_defs): Declare.
1091 * varasm.c (assemble_output_def, process_pending_assemble_output_defs):
1093 (assemble_alias): Defer generation of assembly code for defines when
1094 TARGET_DEFERRED_OUTPUT_DEFS is true.
1095 * config/rs6000/aix41.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
1096 * config/rs6000/aix43.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
1097 * doc/tm.texi (TARGET_DEFERRED_OUTPUT_DEFS): document.
1099 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1102 * pa.c (pa_scalar_mode_supported_p): New function.
1103 (TARGET_SCALAR_MODE_SUPPORTED_P): Define.
1104 * pa.h (MIN_UNITS_PER_WORD): Add comment.
1106 * pa-protos.h (prefetch_operand): Delete.
1107 (prefetch_cc_operand, prefetch_nocc_operand): New declations.
1108 * pa.c (prefetch_operand): Delete.
1109 (prefetch_cc_operand, prefetch_nocc_operand): New functions.
1110 * pa.h (EXTRA_CONSTRAINT): Add `W' constraint.
1111 (PREDICATE_CODES): Delete prefetch_operand. Add prefetch_cc_operand
1112 and prefetch_nocc_operand.
1113 * pa.md (prefetch): Rework to avoid reload problems handling short
1114 displacements when a cache control completer needs to be provided.
1115 (prefetch_32, prefetch_64): Delete.
1116 (prefetch_cc, prefetch_nocc): New patterns.
1118 2005-01-15 David Edelsohn <edelsohn@gnu.org>
1120 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Revert previous change.
1122 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1125 * pa.c (pa_asm_output_aligned_common): Warn if specified alignment
1126 exceeds maximum alignment for global common data.
1127 * pa64-hpux.h (MAX_OFILE_ALIGNMENT): Define to 32768.
1128 * som.h (MAX_OFILE_ALIGNMENT): Likewise.
1130 2004-01-15 Roger Sayle <roger@eyesopen.com>
1132 * tree-ssa-dom.c (extract_range_from_cond): Correct condition.
1134 2004-01-15 Roger Sayle <roger@eyesopen.com>
1136 * harg-reg-set.h (reg_class_names): Prototype global array.
1137 * regclass.c (reg_class_names): Declare here and initialize to
1139 (dump_regclass): Remove local declaration of reg_class_names.
1140 (regclass): Likewise.
1141 * cfg.c (dump_flow_info): Likewise.
1142 * ra-debug.c (reg_class_names): Likewise.
1143 * regrename.c (reg_class_names): Likewise.
1144 * reload.c (reg_class_names): Likewise.
1145 * reload1.c (spill_failure): Likewise.
1146 * config/m68hc11/m68hc11.c (reg_class_names): Likewise.
1148 2005-01-15 Ulrich Weigand <uweigand@de.ibm.com>
1150 * config/s390/s390.md ("reload_outti"): Remove predicate for
1151 output operand. Abort if operand is not a MEM.
1152 ("reload_outdi", "reload_outdf"): Likewise.
1154 2005-01-15 Marc Espie <espie@openbsd.org>
1156 * config.gcc (*-*-openbsd*): Set HAS_LIBC_R for OpenBSD <= 3.2.
1157 (i[34567]86-*-openbsd*): Switch to ELF for OpenBSD >= 3.4.
1158 * config/openbsd.h: Fix C++ includes for native configurations.
1159 Add proper OS_CPP_BUILTINS. Fix libspec for recent OpenBSD.
1160 Add trampoline support.
1161 * config/i386/openbsdelf.h: New.
1163 2005-01-15 Marc Espie <espie@openbsd.org>
1165 * collect2.c (main): Explicitly parse -dynamic-linker option.
1167 2005-01-15 Jakub Jelinek <jakub@redhat.com>
1169 PR tree-optimization/19060
1170 * tree-ssa-dom.c (extract_range_from_cond) <case LT_EXPR, GT_EXPR>:
1171 Return 0 if op1 <= TYPE_MIN_VALUE () resp. op1 >= TYPE_MAX_VALUE ().
1172 (simplify_cond_and_lookup_avail_expr): Add assert for dummy == 0
1173 and handle extract_range_from_cond returning false.
1174 * fold-const.c (fold): Optimize comparisons with min/max even for
1175 width > HOST_BITS_PER_WIDE_INT.
1177 2005-01-15 Ralf Corsepius <ralf.corsepius@rtems.org>
1179 * config/mips/rtems.h (MIPS_DEFAULT_GVALUE): Set to 0.
1180 * config/mips/t-rtems (MULTILIBS_DIRNAMES,MULTILIB_OPTIONS):
1181 Remove little endian multilib variants.
1182 Add mips32 multilib variant.
1184 2005-01-15 Kazu Hirata <kazu@cs.umass.edu>
1186 * explow.c (copy_all_reg, stabilize): Remove.
1187 * expr.h: Remove the corresponding prototypes.
1189 * rtlanal.c (regs_set_between_p): Remove.
1190 * rtl.h: Remove the corresponding prototype.
1192 2005-01-14 David Edelsohn <edelsohn@gnu.org>
1194 * config/rs6000/rs6000.c (rs6k_nonimmediate_operand): Rename
1195 to rs6000_nonimmediate_operand.
1196 * config/rs6000/rs6000.h (PREDICATE_CODES): Rename
1197 rs6k_nonimmediate_operand to rs6000_nonimmediate_operand.
1198 * config/rs6000/rs6000.md (movsi_internal1): Rename predicate to
1199 rs6000_nonimmediate_operand.
1200 * config/rs6000/spe.md (movdf_e500_double): Rename predicate to
1201 rs6000_nonimmediate_operand.
1203 2005-01-14 Andrew Pinski <pinskia@physics.uc.edu>
1205 * tree-ssa-alias.c (pass_may_alias): Add TODO_verify_stmts.
1207 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1209 * config/rs6000/rs6000.h (CLASS_MAX_NREGS): DF goes in 1 register
1211 (CANNOT_CHANGE_MODE_CLASS): Restrict DI mode changes on e500v2.
1212 (PREDICATE_CODES): Add rs6k_nonimmediate_operand.
1214 * config/rs6000/rs6000.c (invalid_e500_subreg): New.
1215 (rs6k_nonimmediate_operand): New.
1216 (rs6000_legitimate_offset_address_p): Handle DI modes on e500v2
1218 (legitimate_lo_sum_address_p): Same.
1219 (rs6000_legitimize_address): Same.
1220 (rs6000_legitimize_reload_address): Same.
1221 (rs6000_legitimate_address): Same.
1222 (spe_build_register_parallel): Pass DF and DC modes in a DI
1225 * config/rs6000/rs6000.md ("*movsi_internal1"): Change predicate
1226 to rs6k_nonimmediate_operand.
1228 * config/rs6000/spe.md ("*frob_df_di"): New.
1229 ("*frob_di_df"): New.
1230 ("*frob_di_df_2"): New.
1231 ("*mov_sidf_e500_subreg0"): New.
1232 ("*mov_sidf_e500_subreg4"): New.
1233 ("*movdf_e500_double"): Change predicate to
1234 rs6k_nonimmediate_operand.
1236 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1238 * postreload.c (move2add_note_store): Only call
1239 trunc_int_for_mode on scalar integers.
1241 2005-01-14 Steven Bosscher <stevenb@suse.de>
1243 * tree-ssa-dce.c (visited_control_parents): New sbitmap to
1244 replace BB_VISITED uses.
1245 (find_obviously_necessary_stmts): Don't clear BB_VISITED.
1246 (propagate_necessity): Check the bitmap instead of BB_VISITED.
1247 (tree_dce_done): Free visited_control_parents.
1248 (perform_tree_ssa_dce): Allocate and clear it.
1249 * tree-ssa-pre.c (compute_antic_aux): Make non-recursive.
1250 (compute_antic): Iterate from here using a DFS. Use an sbitmap
1251 instead of BB_VISITED.
1253 2005-01-14 Kazu Hirata <kazu@cs.umass.edu>
1255 * c-tree.h, coverage.h, langhooks-def.h, optabs.h, output.h,
1256 rtl.h, targhooks.h, tree-flow.h, tree-mudflap.h, tree.h:
1257 Remove unused prototypes.
1259 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1261 * config/rs6000/rs6000.c (rs6000_stack_info): Fix caching of SPE
1262 64-bit register usage.
1263 (spe_func_has_64bit_regs_p): Add FIXME note.
1265 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1267 * config/rs6000/rs6000.c (rs6000_generate_compare): Replace
1268 flag_finite_math_only with flag_unsafe_math_optimizations.
1270 2005-01-14 J. D. Johnston <jjohnst@us.ibm.com>
1272 * config/s390/tpf-unwind.h (s390_fallback_frame_state): Correct
1274 (__tpf_eh_return): Copy TPF private stack area from the module
1275 boundary stack frame. Add check for when module addresses are
1276 equal, but no stub address is found.
1278 2005-01-14 Richard Earnshaw <rearnsha@arm.com>
1281 * arm.h (struct machine_function): Add call_via field.
1282 (thumb_call_via_label): Declare.
1283 * arm.c (thumb_call_via_label): New variable.
1284 (thumb_call_reg_needed): New variable.
1285 (arm_output_function_epilogue): For Thumb code, output any per-function
1286 call-indirect trampolines.
1287 (thumb_call_via_reg): New function.
1288 (arm_file_end): New function.
1289 (TARGET_ASM_FILE_END): Call arm_file_end.
1290 (aof_file_end): Likewise.
1291 * arm-protos.h (thumb_call_via_reg): Declare.
1292 * arm.md (call_reg_thumb, call_value_reg_thumb): Call
1293 thumb_call_via_reg in normal case.
1295 2005-01-14 Jakub Jelinek <jakub@redhat.com>
1298 PR rtl-optimization/19348
1299 * recog.c (peephole2_optimize): Do global life update if some peephole
1300 decides it doesn't need at least one of its inputs and that change
1301 influences liveness at the start of the basic block.
1303 * basic-block.h (EXECUTE_IF_AND_COMPL_IN_REG_SET): Needs 2 REGSET
1304 arguments instead of 1.
1306 2005-01-14 Eric Botcazou <ebotcazou@libertysurf.fr>
1309 * varasm.c (initializer_constant_valid_p) <ADDR_EXPR>: Return
1310 zero for nested functions needing a static chain or functions
1311 with a non-constant address.
1313 2005-01-13 Roger Sayle <roger@eyesopen.com>
1315 * simplify-rtx.c (simplify_binary_operation) <AND>: Optimize
1316 (and (sign_extend X) C) into (zero_extend (and X C)).
1318 2005-01-13 David O'Brien <obrien@FreeBSD.org>
1320 * config/freebsd-spec.h: Make KSE pthread lib logic the default.
1322 2005-01-13 Richard Henderson <rth@redhat.com>
1327 * config/i386/i386.md (cmpdf, cmpsf, bunordered, bordered, buneq,
1328 bunge, bungt, bunle, bunlt, bltgt): Enable for TARGET_SSE_MATH,
1329 not just TARGET_SSE.
1330 (cmpfp_i_387): Rename from cmpfp_i. Move after sse patterns.
1331 (cmpfp_i_mixed): Rename from cmpfp_i_sse; use for TARGET_MIX_SSE_I387.
1332 (cmpfp_i_sse): Rename from cmpfp_i_sse_only; use for TARGET_SSE_MATH.
1333 (cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387): Similarly.
1334 (fp_jcc_1_mixed, fp_jcc_1_sse, fp_jcc_1_387): Similarly.
1335 (fp_jcc_2_mixed, fp_jcc_2_sse, fp_jcc_2_387): Similarly.
1336 (fp_jcc_3_387, fp_jcc_4_387, fp_jcc_5_387, fp_jcc_6_387,
1337 fp_jcc_7_387, fp_jcc_8_387): Rename from fp_jcc_N.
1338 (movdicc_c_rex64): Rename with '*'.
1339 (movsfcc, movdfcc): Add checks for 387 and sse math to condition.
1340 (movsfcc_1_sse_min, movsfcc_1_sse_max, movsfcc_1_sse): New.
1341 (movsfcc_1_387): Rename from movsfcc_1.
1342 (movdfcc_1_sse_min, movdfcc_1_sse_max, movdfcc_1_sse): New.
1343 (movdfcc_1, movdfcc_1_rex64): Add check for 387.
1344 (sminsf3, smaxsf3, smindf3, smaxdf3): New.
1345 (minsf3, minsf, minsf_nonieee, minsf_sse, mindf3, mindf,
1346 mindf_nonieee, mindf_sse, maxsf3, maxsf, maxsf_nonieee, maxsf_sse,
1347 maxdf3, maxdf, maxdf_nonieee, maxdf_sse, sse_movsfcc, sse_movsfcc_eq,
1348 sse_movdfcc, sse_movdfcc_eq, sse_movsfcc_const0_1,
1349 sse_movsfcc_const0_2, sse_movsfcc_const0_3, sse_movsfcc_const0_4,
1350 sse_movdfcc_const0_1, sse_movdfcc_const0_2, sse_movdfcc_const0_3,
1351 sse_movdfcc_const0_4): Remove.
1352 * config/i386/i386.c (ix86_expand_fp_movcc): For TARGET_SSE_MATH,
1353 recognize min/max early. Update for changed sse cmove patterns.
1354 (ix86_split_sse_movcc): New.
1355 * config/i386/i386-protos.h: Update.
1357 2005-01-13 Steven Bosscher <stevenb@suse.de>
1359 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
1360 (fix_stmt_v_may_defs): Likewise.
1362 2005-01-13 Richard Henderson <rth@redhat.com>
1364 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
1365 for converting the mask.
1367 2005-01-13 David Edelsohn <edelsohn@gnu.org>
1369 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
1372 2005-01-13 Richard Henderson <rth@redhat.com>
1374 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
1375 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
1376 IX86_BUILTIN_MOVDQ2Q): Remove.
1377 (IX86_BUILTIN_VEC_EXT_V4SI): New.
1378 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
1379 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
1381 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
1382 (sse2_pextrw): Fix immediate constraint.
1383 (sse2_loadq, sse2_loadq_rex64): Remove.
1384 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
1385 Use __builtin_ia32_vec_ext_<size>.
1386 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
1388 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
1390 * function.c (assign_parm_setup_block): Look inside original
1391 entry_parm when inspecting PARALLEL.
1393 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
1394 Joel Sherrill <joel@oarcorp.com>
1397 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
1398 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
1399 rtems_gxx_recursive_mutex_init.
1400 (__gthread_recursive_mutex_lock): New function.
1401 (__gthread_recursive_mutex_trylock): Likewise.
1402 (__gthread_recursive_mutex_unlock): Likewise.
1404 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
1406 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
1408 2005-01-13 Jan Beulich <jbeulich@novell.com>
1410 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
1411 (ashrti3_internal): Indicate output is early clobber. Generate result
1412 into output rather than first input. Use move for low word of output
1413 if shift count is exactly 64.
1414 (lshrti3_internal): Likewise.
1416 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
1420 * reload1.c (delete_output_reload): Don't delete an output reload
1421 if the pseudo lives longer than a single basic block. Adjust and
1422 improve wording of comment.
1424 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
1427 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
1428 of TYPE_SIZE where we mean the number of bytes.
1430 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
1432 * function.c (assign_parm_setup_block): Relax condition on
1433 multi-register optimization.
1435 2005-01-12 Nick Clifton <nickc@redhat.com>
1437 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
1438 an "in_delay_slot" attribute of "no" to prevent them being used in
1439 delay slots. This is forbidden because they might generate
1442 2005-01-12 Alan Modra <amodra@bigpond.net.au>
1445 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
1448 2005-01-12 Nick Clifton <nickc@redhat.com>
1450 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
1451 passed on to GAS) but do not define, as GAS no longer supports or
1452 needs the -m2000 option.
1454 2005-01-11 Roger Sayle <roger@eyesopen.com>
1456 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
1457 type as args of unprototyped builtins aren't automatically promoted.
1458 (fold_builtin_abs): Likewise.
1460 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1462 * Makefile.in: Set a `build-warn' variable.
1464 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
1466 PR tree-optimization/17949
1467 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
1468 (find_interesting_uses_address): Use it.
1470 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
1472 * regrename.c (kill_value): Handle subreg's that won't simplify.
1474 2005-01-11 Richard Henderson <rth@redhat.com>
1477 * config/i386/i386.h (enum ix86_builtins): Move ...
1478 * config/i386/i386.c: ... here.
1479 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
1480 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
1481 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
1482 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
1483 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
1484 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
1485 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
1486 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
1487 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
1488 IX86_BUILTIN_LOADDDUP): Remove.
1489 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
1490 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
1491 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
1492 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
1493 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
1494 (ix86_init_builtins): Make static.
1495 (ix86_init_mmx_sse_builtins): Update for changed builtins.
1496 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
1497 if all the modes match. Otherwise, fake it.
1498 (get_element_number, ix86_expand_vec_init_builtin,
1499 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
1500 (ix86_expand_builtin): Make static. Update for changed builtins.
1501 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
1502 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
1503 (ix86_expand_vector_init_duplicate): New.
1504 (ix86_expand_vector_init_low_nonzero): New.
1505 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
1506 Split out from ix86_expand_vector_init; handle integer modes.
1507 (ix86_expand_vector_init): Use them.
1508 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
1509 * config/i386/i386-protos.h: Update.
1510 * config/i386/predicates.md (reg_or_0_operand): New.
1511 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
1512 (movv2sf_internal): Likewise. And a splitter to match them all.
1513 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
1514 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
1515 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
1516 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
1517 (mmx_pinsrw): Fix operand ordering.
1518 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
1519 rather than sse_loadss expander.
1520 (movv2df splitter): Similarly.
1521 (sse_loadss, sse_loadlss): Remove.
1522 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
1523 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
1524 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
1525 (sse3_movddup): Rename with '*'.
1526 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
1527 (sse2_loadsd): Remove.
1528 (vec_dupv2df_sse3): Rename from sse3_loadddup.
1529 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
1530 (sse2_pinsrw): Fix argument ordering.
1531 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
1532 (sse2_stored): Remove 'r' destination.
1533 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
1534 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
1535 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
1536 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
1537 vec_extractv16qi, vec_initv16qi): New.
1539 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
1540 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
1541 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
1543 (_mm_load_pd, _mm_store_pd): Use plain dereference.
1544 (_mm_load_si128, _mm_store_si128): Likewise.
1545 (_mm_load1_pd): Use _mm_set1_pd.
1546 (_mm_load_sd): Use _mm_set_sd.
1547 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
1548 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
1549 (_mm_set_epi64): Use _mm_set_epi64x.
1550 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
1551 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
1552 _mm_setr_epi8): Use _mm_set_foo form.
1553 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
1554 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
1555 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
1556 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
1557 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
1558 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
1559 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
1560 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
1561 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
1562 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
1563 (_mm_movedup_pd): Use _mm_shuffle_pd.
1564 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
1565 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
1566 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
1567 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
1569 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
1570 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
1571 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
1572 (_mm_store_ps): Use raw dereference.
1573 (_mm_store1_ps): Use _mm_storeu_ps.
1574 (_mm_storer_ps): Use _mm_store_ps.
1575 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
1576 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
1578 2005-01-11 Stan Shebs <shebs@apple.com>
1580 * config/rs6000/rs6000.c (machopic_output_stub): Issue
1581 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
1583 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
1586 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
1587 Skip all DECLs except for FIELD_DECLs.
1589 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
1591 * config/s390/s390.c (override_options): Return error if
1592 -mbackchain, -mpacked-stack and -mhard-float are used together.
1593 (s390_va_start): Remove the backchain && packed-stack special case.
1594 (s390_gimplify_va_arg): Likewise.
1595 * doc/invoke.texi: Remove the ABI incompatibility note.
1597 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
1599 * config/s390/s390.c (struct s390_frame_layout): Remove
1601 (s390_frame_info, s390_emit_prologue): Replace occurrences of
1602 save_backchain_p with TARGET_BACKCHAIN.
1604 2005-01-11 Alan Modra <amodra@bigpond.net.au>
1607 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
1608 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
1609 * function.c (locate_and_pad_parm): Set new field.
1610 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
1611 Tweak where_pad test to include "none". Always set mem align for
1613 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
1614 not sufficient for type.
1615 (assign_parm_setup_block): If stack_parm is zero on entry, always
1616 make a new stack local. Block move old stack parm if necessary
1617 to new aligned stack local.
1618 (assign_parm_setup_stack): Use a block move to handle
1619 potentially misaligned entry_parm.
1620 (assign_parms_unsplit_complex): Specify required alignment when
1621 creating stack local.
1622 * calls.c (compute_argument_addresses): Override alignment of stack
1623 arg calculated from its type with the alignment given by
1624 FUNCTION_ARG_BOUNDARY.
1625 (store_one_arg): Likewise.
1627 2005-01-11 Jan Beulich <jbeulich@novell.com>
1629 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
1630 Change respective itanium_class attribute to ialu.
1631 (shladdp4_internal): New.
1632 * config/ia64/predicates.md (shladd_log2_operand): New.
1634 2005-01-11 Richard Henderson <rth@redhat.com>
1636 * expr.c (store_constructor): Use rtvec_alloc instead of
1637 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
1639 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
1641 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
1643 2005-01-11 Alan Modra <amodra@bigpond.net.au>
1645 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
1646 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
1647 (TARGET_RWRELOC): Define.
1648 (ia64_rwreloc_section_type_flags): Delete.
1649 (ia64_section_type_flags): New function.
1650 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
1651 (TARGET_RWRELOC): Define.
1653 2005-01-10 David Mosberger <davidm@hpl.hp.com>
1656 * config/ia64/ia64.c (process_set): For alloc insn, only call
1657 process_epilogue is !frame_pointer_needed.
1659 2005-01-10 Roger Sayle <roger@eyesopen.com>
1662 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
1663 unary operator and can't be treated as a binary/comparison operator.
1665 2005-01-10 Richard Henderson <rth@redhat.com>
1667 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
1668 (ix86_return_in_memory): Likewise.
1669 (function_arg_advance): Likewise. Mirror structure in function_arg
1670 for choosing register to advance.
1672 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
1674 * tree-vectorizer.c, tree.def: Fix comment typos.
1676 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
1678 * config/s390/s390.c (struct processor_costs): 4 new fields:
1679 ddbr, ddr, debr, der.
1680 (s390_rtx_costs): More precise handling of divide instructions.
1682 2005-01-10 David Edelsohn <edelsohn@gnu.org>
1684 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
1686 2005-01-10 Jan Beulich <jbeulich@novell.com>
1688 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
1689 names resulting from -ffunction-sections/-fdata-sections and linkonce
1691 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
1692 based on the section name. Rearrange the section name comparison logic
1693 slightly so that each section name is compared against at most once.
1695 2005-01-10 Ben Elliston <bje@au.ibm.com>
1697 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
1698 targets for which -fPIC reduces limitations on the GOT size.
1700 2005-01-09 Falk Hueffner <falk@debian.org>
1702 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
1705 2005-01-09 David Edelsohn <edelsohn@gnu.org>
1708 * collect2.c (main): Set aixrtl_flag for -brtl option.
1709 (resolve_lib_name): Search for .so file extension before .a
1712 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
1714 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
1715 (vect_analyze_data_refs_alignment): Add dump prints.
1717 2005-01-09 Ira Rosen <irar@il.ibm.com>
1719 * tree-vectorizer.c (vect_analyze_offset_expr): Use
1720 expr_invariant_in_loop_p.
1721 Initialize outputs first thing in the function.
1722 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
1723 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
1724 (vect_analyze_pointer_ref_access): Check that the initial condition of
1725 the access function is loop invariant.
1727 2005-01-09 Richard Henderson <rth@redhat.com>
1729 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
1730 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
1731 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
1732 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
1733 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
1734 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
1735 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
1736 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
1737 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
1738 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
1739 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
1740 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
1741 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
1742 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
1743 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
1744 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
1745 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
1746 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
1747 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
1748 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
1749 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
1750 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
1751 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
1753 (mmx_clrdi, pavgusb): Remove.
1754 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
1755 with leading sse_ prefix.
1756 * config/i386/sse.md: Receive them.
1757 * config/i386/mmx.md: New file.
1758 (MMXMODE12, MMXMODE24, mmxvecsize): New.
1759 (subrv2sf3): Turn into expander for normal subtraction.
1760 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
1761 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
1762 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
1763 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
1764 commutative; use ix86_binary_operator_ok.
1765 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
1766 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
1767 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
1768 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
1769 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
1770 Macroize from existing patterns; use ix86_binary_operator_ok.
1771 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
1772 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
1773 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
1774 with vec_select+vec_concat.
1775 (mmx_pshufw, mmx_pshufw_1): Likewise.
1776 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
1777 (mmx_uavgv4hi3): Model correctly.
1778 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
1779 _mm_xor_si64): Remove casts.
1781 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
1783 PR tree-optimization/19224
1784 * tree-scalar-evolution.c (get_instantiated_value,
1785 set_instantiated_value): New functions.
1786 (instantiate_parameters_1): Cache the results.
1787 (instantiate_parameters, resolve_mixers): Initialize and free
1790 2005-01-08 David Edelsohn <edelsohn@gnu.org>
1792 * config/i386/i386.md (addhi_4): Correct reference in comment.
1795 2005-01-08 Richard Henderson <rth@redhat.com>
1797 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
1798 (_mm_cvtsi128_si64x): Likewise.
1799 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
1800 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
1801 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
1802 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
1803 * config/i386/i386-protos.h: Update.
1804 * config/i386/i386.c (print_operand): Add 'H'.
1805 (ix86_fixup_binary_operands): Split out from ...
1806 (ix86_expand_binary_operator): ... here.
1807 (ix86_fixup_binary_operands_no_copy): New.
1808 (ix86_expand_fp_absneg_operator): Handle vector mode results.
1809 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
1810 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
1811 (safe_vector_operand): Use CONST0_RTX.
1812 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
1813 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
1814 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
1815 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
1817 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
1818 (IX86_BUILTIN_CMPNGESS): New.
1819 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
1820 (attr type): Add sselog1.
1821 (attr unit, attr memory): Handle it.
1822 (movti, movti_internal, movti_rex64): Move near other integer moves.
1823 (movtf, movtf_internal): Move near other fp moves.
1824 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
1825 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
1826 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
1827 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
1828 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
1829 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
1830 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
1831 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
1832 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
1833 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
1834 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
1835 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
1836 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
1837 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
1838 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
1839 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
1840 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
1841 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
1842 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
1843 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
1844 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
1845 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
1846 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
1847 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
1848 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
1849 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
1850 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
1851 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
1852 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
1853 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
1854 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
1855 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
1856 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
1857 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
1858 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
1859 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
1860 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
1861 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
1862 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
1863 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
1864 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
1865 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
1866 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
1867 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
1869 (SSEPUSH, push<SSEPUSH>): Remove.
1870 (MMXPUSH, push<MMXPUSH>): Remove.
1871 (sse_movaps, sse_movaps_1, sse_movups): Remove.
1872 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
1873 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
1874 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
1875 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
1876 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
1877 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
1878 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
1879 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
1880 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
1881 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
1882 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
1883 (ppro_sse_log_V4SF): Handle sselog1.
1884 * config/i386/predicates.md (const_0_to_1_operand): New.
1885 (const_0_to_255_mul_8_operand): New.
1886 (const_1_to_31_operand): Rename from const_int_1_31_operand.
1887 (const_2_to_3_operand, const_4_to_7_operand): New.
1888 * config/i386/sse.md: New file.
1889 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
1890 (sse_movups): Rename from sse_movups_1.
1891 (sse_loadlss): Rename from sse_loadss_1.
1892 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
1894 (negv4sf2): Use ix86_expand_fp_absneg_operator.
1895 (absv4sf2, negv2df, absv2df): New.
1896 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
1897 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
1898 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
1899 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
1900 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
1901 (sse3_addsubv4sf3): Model correctly.
1902 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
1903 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
1904 (sse_movhlps): Model with vec_select+vec_concat.
1905 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
1906 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
1907 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
1908 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
1909 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
1910 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
1911 sse2_pshufhw_1): Likewise.
1912 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
1913 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
1914 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
1915 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
1916 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
1917 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
1918 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
1919 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
1920 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
1921 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
1922 sminv8hi3): Mark commutative
1923 operands. Use ix86_binary_operator_ok.
1924 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
1925 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
1926 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
1927 sse2_punpcklqdq): Allow operand2 in memory.
1928 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
1929 sse2_movsd): Add memory alternatives.
1930 (sse_storelps): Turn expander into an insn; split after reload.
1931 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
1932 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
1934 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
1936 * configure.ac (DWARF-2 debug_line): Use objdump.
1937 * configure: Regenerate.
1939 2005-01-08 Jeff Law <law@redhat.com>
1940 Diego Novillo <dnovillo@redhat.com>
1942 PR tree-optimization/18241
1943 * tree-nrv.c (tree_nrv): Ignore volatile return values.
1944 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
1945 statements with volatile operands.
1946 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
1947 operands after marking a statement with has_volatile_ops.
1949 2005-01-08 Roger Sayle <roger@eyesopen.com>
1951 * tree.c (int_fits_type_p): Always honor integer constant
1952 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
1954 2005-01-08 Roger Sayle <roger@eyesopen.com>
1956 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
1957 the latent bug in rtl_delete_block has been resolved.
1959 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
1961 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
1962 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
1964 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
1966 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
1967 library as @multilib_dir@/$(SHLIB_SONAME).
1969 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
1971 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
1972 * configure: Regenerate.
1974 2005-01-07 Jakub Jelinek <jakub@redhat.com>
1976 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
1979 2005-01-07 David Edelsohn <edelsohn@gnu.org>
1982 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
1983 Convert non-word aligned offset address using ld/std into
1986 2005-01-07 Richard Henderson <rth@redhat.com>
1988 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
1990 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
1992 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
1994 2005-01-07 Jakub Jelinek <jakub@redhat.com>
1996 PR tree-optimization/19283
1997 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
1998 is not shorter than the original type.
2000 PR rtl-optimization/19012
2001 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
2003 PR rtl-optimization/18861
2004 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
2005 (FULL_STATE): Define.
2006 (mark_tablejump_edge): New function.
2007 (purge_dead_tablejump_edges): New function.
2008 (find_bb_boundaries): Use it.
2010 PR tree-optimization/18828
2011 * builtins.c (expand_builtin_next_arg): Remove argument and all
2012 the argument checking.
2013 (expand_builtin): Adjust caller.
2014 (expand_builtin_va_start): Likewise. Remove error for too many
2016 (fold_builtin_next_arg): Issue error for too many arguments.
2017 After checking arguments, replace them with magic arguments that
2018 prevent further checking of the args.
2020 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2022 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
2024 2005-01-06 Roger Sayle <roger@eyesopen.com>
2027 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
2028 and %{.S:...} (and their negative variants) to test whether the
2029 input file is assembler or pre-processed-assembler independent of
2030 the actual filename extension.
2032 2005-01-06 Roger Sayle <roger@eyesopen.com>
2034 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
2035 of sign or zero extended values.
2037 2005-01-06 Geoffrey Keating <geoffk@apple.com>
2039 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
2040 for IBM long double format correctly.
2042 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
2044 Fix PR tree-optimization/18792
2046 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
2047 to first_loop_depth, and use loop depth instead of loop number.
2048 (build_classic_dir_vector): Ditto.
2049 (compute_data_dependences_for_loop): Use depth, not loop number.
2050 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
2051 number. Pass in loops, instead of loop numbers.
2052 (gather_interchange_stats): Ditto.
2053 (linear_transform_loops): Ditto.
2055 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
2058 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
2059 biv_fits_mode_p, extension_within_bounds_p): New functions.
2060 (check_ext_dependent_givs): Use them.
2062 2005-01-06 Roger Sayle <roger@eyesopen.com>
2064 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
2065 more than one barrier, in which case we should delete them all.
2067 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
2069 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
2071 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
2073 Fix PR middle-end/19286
2075 * dwarf2out.c (gen_subprogram_die): If we've already tried to
2076 output this subprogram, simply ignore this attempt to do it again.
2077 (add_abstract_origin_attribute): Don't abort trying to add the abstract
2078 origin attribute if it's not possible.
2079 (gen_block_die): Don't ignore subblocks of "unused" blocks.
2080 (decls_for_scope): Ditto.
2081 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
2082 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
2084 2005-01-05 Richard Henderson <rth@redhat.com>
2087 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
2088 (bdesc_2arg): Use it.
2089 (ix86_expand_binop_builtin): Force operands into registers
2091 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
2092 ix86_expand_sse_compare, ix86_expand_sse_comi,
2093 ix86_expand_builtin): Likewise.
2095 2005-01-05 Richard Henderson <rth@redhat.com>
2097 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
2099 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
2100 (mulv4hi3): Set itanium_class mmmul.
2101 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
2102 (fpack): Rename from fpack_sfsf.
2104 2005-01-05 Richard Henderson <rth@redhat.com>
2107 * reload1.c (do_input_reload): Restrict the optimization deleteing
2108 a previous output reload to RELOAD_FOR_INPUT.
2110 2005-01-05 Steven Bosscher <stevenb@suse.de>
2112 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
2113 comment that we fall through after case.
2114 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
2115 (distribute_links): Likewise.
2116 * cse.c (cse_insn): Likewise.
2117 * cselib.c (cselib_invalidate_mem): Likewise.
2118 * df.c: Update comments at the top of the file.
2119 (read_modify_subreg_p): Update comments here too.
2120 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
2121 * flow.c (mark_set_1): Likewise.
2122 (mark_used_regs): Likewise.
2123 * gcse.c (mems_conflict_for_gcse_p): Likewise.
2124 (canon_list_insert): Likewise.
2125 (mark_set): Likewise.
2126 (try_replace_reg): Likewise.
2127 (store_killed_in_insn): Likewise.
2128 * loop.c (count_one_set): Likewise.
2129 (basic_induction_var): Likewise.
2130 * postreload-gcse.c (find_mem_conflicts): Likewise.
2131 * postreload.c (reload_combine_note_store): Likewise.
2132 (move2add_note_store): Likewise.
2133 * reload.c (find_equiv_reg): Likewise.
2134 (mark_referenced_resources): Likewise.
2135 * rtlanal.c (set_noop_p): Likewise.
2136 (note_stores): Likewise.
2137 (note_uses): Likewise.
2138 * sched-deps.c (sched_analyze_1): Likewise.
2139 * sched-rgn.c (check_live_1): Likewise.
2140 (update_live_1): Likewise.
2141 * config/i860/i860.c: Likewise.
2143 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
2145 (ZERO_EXTRACT): Mention that this one can be an lvalue.
2147 * doc/rtl.texi: Update documentation for bit-fields and SET.
2149 2005-01-05 Roger Sayle <roger@eyesopen.com>
2151 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
2153 2005-01-05 Jan Hubicka <jh@suse.cz>
2154 Richard Henderson <rth@redhat.com>
2157 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
2160 2005-01-05 Richard Henderson <rth@redhat.com>
2163 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
2164 (sse_shufps): Change operand 3 to const_int_operand.
2165 (sse2_storelps): Fix typo in template.
2166 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
2167 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
2168 (ix86_expand_builtin): Likewise.
2170 2005-01-05 Julian Brown <julian@codesourcery.com>
2172 * config/arm/arm.c (arm_return_in_memory): Treat complex types
2173 as aggregates for AAPCS ABIs.
2175 2005-01-05 Stan Shebs <shebs@apple.com>
2177 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
2178 Mach-O getters if ppc64.
2179 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
2181 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
2182 signed four-byte field for global code case.
2183 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
2185 2005-01-05 Roger Sayle <roger@eyesopen.com>
2188 * c-common.c: Include real.h.
2189 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
2190 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
2191 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
2192 * Makefile.in (c-common.o): Update dependencies.
2194 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
2196 * c-parse.in (asm_string): Add trailing semicolon.
2198 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
2200 * c-parse.in (asm_string): New. Don't allow wide strings in
2202 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
2203 asm_string instead of STRING.
2205 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
2207 * c-typeck.c (constructor_no_implicit): Remove.
2208 (set_designator, process_init_element): Don't check
2209 constructor_no_implicit.
2211 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
2212 Kaz Kojima <kkojima@gcc.gnu.org>
2215 * lcm.c (create_pre_exit): New.
2216 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
2217 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
2219 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
2221 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
2222 -msoft-float in target_float_switch.
2223 * arm.c (arm_override_options): Fix processing of target_float_switch.
2225 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
2227 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
2230 2004-01-05 Uros Bizjak <uros@kss-loka.si>
2232 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
2233 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
2235 2005-01-04 Roger Sayle <roger@eyesopen.com>
2237 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
2238 if deleting the then-block allows the test-block to fallthru to the
2241 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
2244 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
2245 after the full declaration if the are in two different TUs.
2247 2005-01-04 Richard Henderson <rth@redhat.com>
2250 * tree-sra.c (generate_one_element_init): Just
2251 call gimplify_and_add.
2252 (generate_element_init): Record the
2253 new referenced variables and mark them for renaming
2254 and split out to ...
2255 (generate_element_init_1): This.
2256 (scalarize_init): Don't call push_gimplify_context/
2257 pop_gimplify_context.
2259 2005-01-04 Geoffrey Keating <geoffk@apple.com>
2261 * toplev.c (get_src_pwd): Handle failure of getpwd().
2263 2005-01-04 Roger Sayle <roger@eyesopen.com>
2265 * fold-const.c (fold_single_bit_test): Delete unreachable handling
2267 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
2268 all the cases handled by it are inverted by invert_truthvalue.
2270 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
2272 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
2273 (save_register_windowsi): Likewise.
2275 2005-01-04 Richard Henderson <rth@redhat.com>
2277 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
2279 2005-01-04 Uros Bizjak <uros@kss-loka.si>
2282 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
2283 any simplifications of vector mode comparison operators.
2284 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
2286 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
2287 Devang Patel <dpatel@apple.com>
2289 PR tree-optimization/18308
2290 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
2291 the operands before creating a new expression.
2292 * dojump.c (do_jump): Make drop_through_label available
2293 for all cases. Add expansion of COND_EXPR.
2295 2005-01-04 Ira Rosen <irar@il.ibm.com>
2297 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
2298 INTEGER_CST instead of TREE_CONSTANT.
2299 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
2300 instead of TREE_CONSTANT.
2301 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
2302 instead of TREE_CONSTANT.
2304 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
2306 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
2308 2005-01-04 Uros Bizjak <uros@kss-loka.si>
2311 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
2312 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
2314 2005-01-03 Richard Henderson <rth@redhat.com>
2316 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
2317 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
2318 fold_convert_const_real_from_real): Split out from ...
2319 (fold_convert_const): ... here.
2321 2005-01-03 Richard Henderson <rth@redhat.com>
2324 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
2325 (mov<MMXMODEI>_internal): Likewise.
2326 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
2327 Add V2SF case; use it for SSE1; don't use TI.
2328 (movdf_integer): Likewise.
2329 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
2331 2005-01-03 Ira Rosen <irar@il.ibm.com>
2333 * tree-vectorizer.c (vect_strip_conversions): New function.
2334 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
2335 check for binary class.
2337 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
2341 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
2343 * gimple-low.c (mark_blocks_with_used_vars): New function.
2344 (mark_blocks_with_used_subblocks): Ditto.
2345 (mark_used_blocks): Ditto.
2346 (pass_mark_used_blocks): New pass.
2347 * tree-inline.c: Include debug.h.
2348 (expand_call_inline): Call outlining_inline_function here.
2349 * tree-optimize.c (init_tree_optimization_passes): Add
2350 pass_mark_used_blocks.
2351 * tree-pass.h (pass_mark_used_blocks): New.
2352 * Makefile.in (tree-inline.o): Add debug.h dependency.
2354 2005-01-03 Geoffrey Keating <geoffk@apple.com>
2356 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
2357 VAR_DECLs to have weak_import attribute.
2359 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
2362 2004-01-03 Steven Bosscher <stevenb@suse.de>
2364 * passes.c (rest_of_compilation): Don't run regmove if only
2365 flag_expensive_optimizations, require flag_regmove instead.
2367 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
2369 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
2371 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
2372 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
2373 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
2374 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
2376 2005-01-03 Richard Henderson <rth@redhat.com>
2378 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
2379 (ia64_const_ok_for_letter_p): New.
2380 (ia64_const_double_ok_for_letter_p): New.
2381 (ia64_extra_constraint): New.
2382 (ia64_expand_vecint_compare): New.
2383 (ia64_expand_vcondu_v2si): New.
2384 (ia64_expand_vecint_cmov): New.
2385 (ia64_expand_vecint_minmax): New.
2386 (ia64_print_operand): Add 'v'.
2387 (ia64_preferred_reload_class): New.
2388 (ia64_vector_mode_supported_p): New.
2389 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
2390 (PREFERRED_RELOAD_CLASS): Move to function.
2391 (CONST_OK_FOR_LETTER_P): Move to function.
2392 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
2393 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
2394 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
2395 (EXTRA_CONSTRAINT): Move to function.
2396 * config/ia64/ia64.md: Include vect.md.
2397 (itanium_class): Add mmalua.
2399 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
2401 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
2402 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
2404 (const_int_2bit_operand): New.
2405 (fr_reg_or_0_operand): New.
2406 * config/ia64/ia64-modes.def: Add vector modes.
2407 * config/ia64/ia64-protos.h: Update.
2408 * config/ia64/vect.md: New file.
2410 2005-01-03 Richard Henderson <rth@redhat.com>
2412 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
2414 2005-01-03 Uros Bizjak <uros@kss-loka.si>
2417 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
2419 (log1pdf2): Change mode of operands[0,1] to DFmode.
2421 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
2423 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
2424 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
2425 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
2426 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
2427 SPARC_RELAXED_ORDERING.
2429 2005-01-03 Richard Henderson <rth@redhat.com>
2430 Uros Bizjak <uros@kss-loka.si>
2433 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
2434 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
2435 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
2437 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
2438 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
2439 const_pow2_1_to_128_operand): New predicates.
2441 2005-01-02 Greg McGary <greg@mcgary.org>
2443 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
2445 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
2446 object. [ptr_type]: Remove unused variable.
2447 Remember nearest addressable array-element or record-component
2448 when checking bitfield components. Tolerate empty BIND_EXPR.
2450 2005-01-01 Richard Henderson <rth@redhat.com>
2452 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
2453 that don't narrow the value. Fail for other conversions.
2455 2005-01-01 Richard Henderson <rth@redhat.com>
2458 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
2459 * c-lang.c (finish_file): Don't do it here.
2460 * objc/objc-act.c (objc_finish_file): Likewise.
2462 * cgraph.c (decl_assembler_name_equal): New.
2463 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
2464 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
2465 * cgraph.h (struct cgraph_varpool_node): Add next.
2466 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
2467 * varasm.c (assemble_alias): Mark the target as needed.
2469 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
2472 * function.c (get_arg_pointer_save_area): Use entry_of_function
2473 instead of get_insns.
2475 2005-01-01 Roger Sayle <roger@eyesopen.com>
2476 Andrew Pinski <pinskia@physics.uc.edu>
2477 James E. Wilson <wilson@specifixinc.com>
2479 PR rtl-optimization/12092
2480 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
2483 2005-01-01 Roger Sayle <roger@eyesopen.com>
2484 Olivier Hainque <hainque@act-europe.fr>
2486 * tree.c (int_fits_type_p): A narrower type always fits in a
2487 wider one, except for negative values into unsigned types.
2489 2005-01-01 Roger Sayle <roger@eyesopen.com>
2491 * tree.c (int_fits_type_p): Compare the result of force_fit_type
2492 with the original constant rather than require TREE_OVERFLOW.
2494 2005-01-01 Steven Bosscher <stevenb@suse.de>
2497 * c-decl.c (finish_function): If compiling C99, annotate the
2498 compiler generated return with the current file name and line 0.
2499 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
2500 the source line is greater than 0.
2501 (remove_bb): Likewise.
2503 See ChangeLog.12 for earlier changes.