1 2005-01-26 Jan Hubicka <jh@suse.cz>
3 PR tree-optimization/19421
4 * tree-inline.c (copy_body_r): Do not walk subtrees after substituting.
6 2005-01-26 Hans-Peter Nilsson <hp@axis.com>
8 * config/cris/cris.h (LIB_SPEC) <sim*>: Use -lsyslinux, not
9 -lsyssim. Wrap -lc -lsyssim in --start-group --end-group instead
11 (STARTFILE_SPEC): Add crti.o, before crtbegin.o.
12 <sim*>: Always use crt1.o, regardless of N in simN.
13 <!sim>: Drop support for -pg and -p variants; always use crt0.o.
14 (ENDFILE_SPEC): Add crtn.o, after crtend.o.
15 (CRT_CALL_STATIC_FUNCTION): Remove.
16 * config/cris/aout.h (STARTFILE_SPEC): <!melinux && sim*>: Always
17 use crt1.o, regardless of N in simN.
18 <!sim>: Drop support for -pg and -p variants; always use crt0.o.
19 (LIB_SPEC): Do not define; default to the one in config/cris/cris.h.
20 (CRIS_CPP_SUBTARGET_SPEC) <melinux && !nostdinc>: Remove special
22 (MAX_OFILE_ALIGNMENT): Correct to 32, not 16.
24 2005-01-25 Ian Lance Taylor <ian@airs.com>
26 * gimple-low.c (block_may_fallthru): Correct handling of
27 SWITCH_EXPR--may fall through if SWITCH_LABELS is NULL.
29 2005-01-26 Hans-Peter Nilsson <hp@axis.com>
31 * config/cris/cris.h (CPP_SPEC): Adjust the various mtune=
32 handlers to match indentation; nested under the mtune=* handler.
33 Similar for mcpu= and march=.
35 2005-01-25 Daniel Jacobowitz <dan@codesourcery.com>
37 * config/mips/linux.h (LIB_SPEC): Remove -rpath-link.
38 * config/mips/linux64.h (LIB_SPEC): Likewise.
39 * config/mn10300/linux.h (LIB_SPEC): Likewise.
40 * config/sh/linux.h (LIB_SPEC): Likewise.
42 2005-01-25 Aldy Hernandez <aldyh@redhat.com>
44 * config/frv/frv.c (frv_legitimize_tls_address): New.
45 (TARGET_HAVE_TLS): Define.
46 (FRV_SYMBOL_REF_TLS_P): Define.
47 (frv_override_options): Handle new register classes.
48 (frv_legitimate_address_p): Reject tls addresses.
49 (frv_legitimize_address): Handle TLS addresses.
50 (gen_inlined_tls_plt): New.
52 (frv_legitimize_tls_address): New.
53 (unspec_got_name): Add TLS entries.
54 (got12_operand): Add R_FRV_TLSMOFF12 case.
55 (frv_emit_move): Fixup TLS addresses.
56 (frv_emit_movsi): Legitimize TLS addresses.
58 * config/frv/frv.h (MASK_BIG_TLS): New.
59 (TARGET_BIG_TLS): New.
60 (HAVE_AS_TLS): Define.
61 (TARGET_SWITCHES): Add -mTLS and -mtls options.
62 (enum reg_class): Add GR8_REGS, GR9_REGS, GR89_REGS.
63 (REG_CLASS_NAMES): Same.
64 (REG_CLASS_CONTENTS): Same.
65 (CONSTRAINT_LEN): New.
66 (REG_CLASS_FROM_CONSTRAINT): New.
67 (PREDICATE_CODES): Add symbolic_operand.
69 * config/frv/frv.md (define_constants): Add UNSPEC_GETTLSOFF,
70 UNSPEC_TLS_LOAD_GOTTLSOFF12, UNSPEC_TLS_INDIRECT_CALL,
71 UNSPEC_TLS_TLSDESC_LDD, UNSPEC_TLS_TLSDESC_LDD_AUX,
72 UNSPEC_TLS_TLSOFF_LD, UNSPEC_TLS_LDDI, UNSPEC_TLSOFF_HILO,
73 R_FRV_GOTTLSOFF_HI, R_FRV_GOTTLSOFF_LO, R_FRV_TLSMOFFHI,
74 R_FRV_TLSMOFFLO, R_FRV_TLSMOFF12, R_FRV_TLSDESCHI,
75 R_FRV_TLSDESCLO, R_FRV_GOTTLSDESCHI, R_FRV_GOTTLSDESCLO, GR8_REG,
76 GR9_REG, GR14_REG, LRREG.
77 (type): Add load_or_call attribute.
78 ("load_or_call"): New reservation.
79 ("call_gettlsoff"): New.
80 ("tls_indirect_call"): New.
81 ("tls_load_gottlsoff12"): New.
83 ("tls_tlsdesc_ldd"): New.
84 ("tls_tlsoff_ld"): New.
87 * config/frv/frv-protos.h (symbolic_operand): Protoize.
89 2005-01-25 Alexandre Oliva <aoliva@redhat.com>
91 * config/frv/frv.h (ASM_OUTPUT_DWARF_DTPREL): Define.
92 * config/frv/frv-protos.h (frv_output_dwarf_dtprel): Declare.
93 * config/frv/frv.c (TLS_BIAS): Define.
94 (frv_output_dwarf_dtprel): New.
96 2005-01-26 Jakub Jelinek <jakub@redhat.com>
98 PR rtl-optimization/19579
99 * ifcvt.c (noce_try_cmove_arith): If emitting instructions to set up
100 both A and B, see if they don't clobber registers the other expr uses.
102 2005-01-25 J"orn Rennecke <joern.rennecke@st.com>
104 * real.c (do_add): Initialize signalling and canonical members.
106 * real.c (real_from_integer): Zero out destination.
108 2005-01-18 Jan Hubicka <jh@suse.cz>
110 PR tree-optimize/19337
111 * tree-inline.c (initialize_inlined_parameters): Copy
112 saved_static_chain_decl only.
114 2005-01-25 Richard Henderson <rth@redhat.com>
118 * config/i386/i386.c (x86_inter_unit_moves): Zero.
119 (ix86_preferred_reload_class): Rewrite fp-constant section, with
120 80387 enabled, to return a proper subclass. Return the subset that
121 overlaps with GENERAL_REGS for PLUS.
122 * config/i386/i386.md (movsi_1, movdi_2): Set type to "mmx" for pxor.
124 2005-01-25 Richard Henderson <rth@redhat.com>
127 * tree-complex.c (expand_complex_div_wide): Use the correct formulae.
129 2005-01-25 Roger Sayle <roger@eyesopen.com>
131 * expmed.c (choose_multiplier): Change interface to return the
132 multiplier lowpart as an rtx instead of an unsigned HOST_WIDE_INT.
133 (expand_divmod): Update calls to choose_multiplier.
135 2005-01-24 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
138 * expmed.c (extract_bit_field): Use adjust_address instead of
139 gen_lowpart when op0 is a MEM.
141 2005-01-24 Steven Bosscher <stevenb@suse.de>
143 * cfgexpand.c (expand_gimple_tailcall): Fix typo.
145 2005-01-24 Richard Sandiford <rsandifo@redhat.com>
147 * config/mips/irix-csr.c: New file.
148 * config/mips/t-iris6 (irix-csr.o): New rule to build it.
149 (EXTRA_MULTILIB_PARTS): Add irix-csr.o.
150 * config/mips/iris6.h (ENDFILE_SPEC): Include it in n32 and n64
153 2005-01-24 Eric Botcazou <ebotcazou@libertysurf.fr>
156 * config.gcc (sparc-*-elf*): Remove sol2.h, sparc/sol2.h and
157 sparc/elf.h, add sparc/sp-elf.h.
158 (sparc-*-rtems*): Likewise.
159 (sparclite-*-elf*): Remove sol2.h, sparc/sol2.h, sparc/elf.h and
160 tm-dwarf2.h, add sparc/sp-elf.h.
161 (sparc86x-*-elf): Likewise.
162 (sparc64-*-elf*): Remove sol2.h, sparc/sol2.h and tm-dwarf2.h.
163 * config/sparc/liteelf.h (TARGET_SUB_OS_CPP_BUILTINS): Rename into
164 TARGET_OS_CPP_BUILTINS.
165 * config/sparc/sp86x-elf (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
166 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
168 * config/sparc/openbsd64.h (NO_IMPLICIT_EXTERN_C): Undefine.
169 * config/sparc/sp64-elf.h (NO_IMPLICIT_EXTERN_C): New macro.
170 (SWITCH_TAKES_ARG): Likewise.
171 (LOCAL_LABEL_PREFIX): Likewise.
172 (ASM_GENERATE_INTERNAL_LABEL): Likewise.
173 (TARGET_N_FORMAT_TYPES): Delete.
174 (TARGET_FORMAT_TYPES): Likewise.
175 (ASM_DECLARE_FUNCTION_SIZE): Likewise.
176 * config/sparc/elf.h: Delete.
177 * config/sparc/sp-elf.h: New file.
179 2005-01-24 Kazu Hirata <kazu@cs.umass.edu>
181 * tree-flow-inline.h (phi_arg_from_edge): Remove.
182 * tree-flow.h: Remove the corresponding prototype.
183 * tree-ssa-operands.h, tree-ssa-threadupdate.c, tree-ssa.c,
184 tree-vectorizer.c: Use dest_idx instead of phi_arg_from_edge.
186 2005-01-24 Jakub Jelinek <jakub@redhat.com>
189 * config/rs6000/rs6000.c (rs6000_file_start): Create toc section
190 for AIX ABI or ELF -fPIC.
191 (rs6000_emit_load_toc_table): Don't create toc_section here.
192 (rs6000_xcoff_file_start): Nor here.
194 2005-01-24 J"orn Rennecke <joern.rennecke@st.com>
196 * sh.c (ra.h): Don't #include.
197 (hard_regs_intersect_p): New function, resurrected from ra.c.
199 * sh.c: Fix 1996 Copyright.
201 2005-01-24 Richard Henderson <rth@redhat.com>
202 Aldy Hernandez <aldyh@redhat.com>
204 * regrename.c (note_sets): Handle subregs.
206 2005-01-24 Joseph S. Myers <joseph@codesourcery.com>
209 * genconditions.c (write_header, write_conditions): Elide file if
212 2005-01-24 Kazu Hirata <kazu@cs.umass.edu>
214 * c-opts.c, expmed.c, gengtype.c, tree-complex.c,
215 config/arm/iwmmxt.md, config/avr/avr.c, config/s390/s390.h:
218 2005-01-24 Jakub Jelinek <jakub@redhat.com>
220 * builtins.c (get_pointer_alignment, gimplify_va_arg_expr,
221 expand_builtin_printf, expand_builtin_fprintf,
222 expand_builtin_sprintf): Use POINTER_TYPE_P instead TREE_CODE
223 checking against POINTER_TYPE.
224 (validate_arglist): Handle POINTER_TYPE code by checking
227 2005-01-24 Paolo Bonzini <bonzini@gnu.org>
229 * rtlanal.c (non_rtx_starting_operands, for_each_rtx_1,
231 (for_each_rtx): Call for_each_rtx_1.
232 * rtl.h (init_rtlanal): Declare.
233 * toplev.c (backend_init): Call init_rtlanal.
235 2005-01-24 Jakub Jelinek <jakub@redhat.com>
237 * flow.c (propagate_one_insn): Formatting.
240 * flow.c (libcall_dead_p): Be more conservative if unsure.
241 If there are any instructions between insn and call, see if they are
242 all dead before saying the libcall is dead.
244 2005-01-24 Uros Bizjak <uros@kss-loka.si>
246 * config/i386/i386.md (*extendsfdf2_sse): Use "nonimmediate_operand"
247 operand constraint for operand 0.
249 2005-01-23 Richard Henderson <rth@redhat.com>
252 * tree-complex.c (expand_complex_div_wide): Set EDGE_FALLTHRU on
253 edges leading to join block.
255 2005-01-23 Roger Sayle <roger@eyesopen.com>
257 * expmed.c (expand_mult_highpart): Make static. Change type of
258 constant multiplier argument from unsigned HOST_WIDE_INT to rtx.
259 (expand_divmod): Updates calls to expand_mult_highpart by using
260 gen_int_mode to pass a CONST_INT rtx.
261 * rtl.h (expand_mult_highpart): Remove prototype.
263 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
265 * flow.c (regset_bytes, regset_size): Remove.
267 * cfganal.c: Fix a reference to Harvey's paper.
269 2005-01-23 Steven Bosscher <stevenb@suse.de>
271 PR rtl-optimization/19464
272 * tree-optimize.c (init_tree_optimization_passes): Add one more
273 copyrename pass just before out-of-ssa.
275 2005-01-23 Joseph S. Myers <joseph@codesourcery.com>
278 * recog.c (recog_memoized): Don't define if GENERATOR_FILE.
279 * ggc-none.c (ggc_free): Define.
281 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
283 * cse.c (max_reg, max_insn_uid): Remove.
284 (cse_main): Don't access max_reg or max_insn_uid.
286 * c-common.c (explicit_flag_signed_bitfields,
287 lang_statement_code_p, lang_expand_function_end): Remove.
288 * c-common.h: Remove the corresponding declarations.
289 * c-opts.c (c_common_handle_option): Remove all write access
290 to explicit_flag_signed_bitfields.
292 * cgraph.c (cgraph_varpool_n_nodes): Remove.
293 (cgraph_varpool_node): Don't access cgraph_varpool_n_nodes.
294 * cgraph.h: Remove the corresponding declaration.
296 * gcse.c (null_pointer_info): Remove.
298 2005-01-23 Roger Sayle <roger@eyesopen.com>
299 Eric Botcazou <ebotcazou@libertysurf.fr>
301 * combine.c (simplify_logical): Only simplify logical expressions
302 of the form ior(and(x,y),z) by the inverse distributive law if the
303 result is cheaper than the original.
305 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
307 * alias.c, c-common.h, c-incpath.c, c-incpath.h, expr.c,
308 fold-const.c, gimplify.c, params.h, tree-data-ref.c,
309 tree-if-conv.c, tree-nested.c, tree-outof-ssa.c,
310 tree-ssa-dom.c, tree-vectorizer.c, tree.def, config/darwin.c,
311 config/freebsd-spec.h, config/arm/arm.h,
312 config/h8300/h8300.md, config/i386/i386.md,
313 config/i386/predicates.md, config/i386/sse.md,
314 config/ia64/ia64.c, config/ip2k/ip2k.c, config/s390/s390.c,
315 config/vax/vax.md: Fix comment typos. Follow spelling
318 2005-01-23 Steven Bosscher <stevenb@suse.de>
320 * gengtype.c (walk_type): Produce `not equal to' compares for loop
321 tests, instead of `less than'.
323 2005-01-23 Richard Sandiford <rsandifo@redhat.com>
325 * function.c (assign_parm_setup_block): When creating a new stack slot
326 for a parameter, get its alignment from the parameter's DECL_ALIGN
327 rather than the type's TYPE_ALIGN. Make sure that the parameter
328 is at least word aligned.
330 2005-01-22 Roger Sayle <roger@eyesopen.com>
333 * config/avr/avr.c (avr_hard_regno_mode_ok): Rewrite.
335 2005-01-22 Richard Henderson <rth@redhat.com>
338 * config/i386/i386.md (movsfcc_1_sse_max): Use nonimmediate_operand
339 in both compare operands.
340 (movdfcc_1_sse_max): Likewise.
341 (movsfcc_1_sse): Likewise. Add earlyclobber for scratch.
342 (movdfcc_1_sse): Likewise.
343 * config/i386/i386.c (ix86_split_sse_movcc): Emit copies into the
344 scratch register as needed.
346 2005-01-22 Richard Henderson <rth@redhat.com>
348 * config/i386/i386.md (smaxsf3): Fix mnemonic typo.
350 * config/i386/i386.c (ix86_prepare_fp_compare_args): Fix is_sse test.
352 2005-01-22 Richard Henderson <rth@redhat.com>
354 * genopinit.c (optabs): Use smin/smax for floating point too.
355 * doc/md.texi: Update to match. Clarify that floating point
356 results are undefined for +0/-0 and NaN.
357 * doc/rtl.texi: Likewise.
358 * rtl.def (SMIN, SMAX): Likewise
359 * tree.def (MIN_EXPR, MAX_EXPR): Likewise.
361 * config/alpha/alpha.md (smaxdf3, smindf3, smaxsf3, sminsf3): Add
362 leading 's' to the name.
363 * config/ia64/ia64.md (smaxsf3, sminsf3, smaxdf3, smindf3,
364 smaxxf3, sminxf3): Likewise.
365 * config/rs6000/rs6000.md (smaxdf3, smindf3, smaxsf3, sminsf3):
368 2005-01-22 Paul Brook <paul@codesourcery.com>
370 * config/arm/arm.md: Use "Uy" constraint for wcgr load/stores.
372 2005-01-21 Mark Dettinger <dettinge@de.ibm.com>
374 * config/s390/s390.c (struct processor_costs): New fields
375 dlgr, dlr, dr, dsgfr, dsgr.
376 (z900_cost, z990_cost): Values for new fields.
377 (s390_rtx_costs): New cases MEM und COMPARE in switch
378 statement. Modified handling of SIGN_EXTEND, ZERO_EXTEND,
379 DIV, MOD, UDIV, UMOD.
381 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
383 * config/s390/s390.md ("doloop_si64"): Reload input value directly
384 into the register being decremented.
385 ("doloop_si31", "doloop_di"): Likewise.
386 ("*doloop_si_long"): Adapt pattern.
388 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
390 * config/s390/s390.h (HARD_REGNO_NREGS): Fix computation for
392 (CLASS_MAX_NREGS): Likewise.
394 2005-01-21 Daniel Berlin <dberlin@dberlin.org>
396 * doc/tree-ssa.texi (Statement Operands): Add example for new
397 must-def macro. Note deprecation of old operands interface.
399 2005-01-22 Richard Sandiford <rsandifo@redhat.com>
401 PR tree-optimization/19484
402 * tree-cfg.c (remove_fallthru_edge): New function.
403 (cleanup_control_flow): Remove fallthru edges from calls that are
404 now known not to return.
406 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
408 Fix PR tree-optimization/19038
409 * tree-ssa-dom.c (cprop_operand): Don't replace loop invaeriant
410 copies with loop variant ones.
412 2005-01-22 Kazu Hirata <kazu@cs.umass.edu>
414 * cfganal.c, real.h, reorg.c, timevar.def, tree-ssa-ccp.c,
415 config/alpha/alpha-protos.h, config/alpha/alpha.h,
416 config/alpha/alpha.md, config/alpha/predicates.md,
417 config/sparc/freebsd.h, config/sparc/netbsd-elf.h,
418 config/sparc/sol2.h: Update copyright.
420 * tree-cfg.c (remove_forwarder_block_with_phi): Look at the
421 first label to see if it is a nonlocal label.
423 2005-01-22 David Edelsohn <edelsohn@gnu.org>
424 Andrew Pinski <pinskia@physics.uc.edu>
427 * config/rs6000/rs6000.c (rs6000_va_start): Saturate n_gpr at
428 maximum number of GPRs. Saturate n_fpr at maximum number of FPRs.
430 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
432 * config/rs6000/rtems.h: Update copyright.
434 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
437 * config/rs6000/rtems.h: Resurrect cpp_os_rtems_spec from gcc < 3.4.
438 (CPP_OS_RTEMS_SPEC): New (From gcc-3.3's config/rs6000/sys4.h).
439 (SUBSUBTARGET_EXTRA_SPECS): Use CPP_OS_RTEMS_SPEC.
441 2005-01-22 Volker Reichelt <reichelt@igpm.rwth-aachen.de>
444 * c-typeck.c (convert_arguments): Check for error_mark_node.
446 2005-01-21 Ian Lance Taylor <ian@airs.com>
448 PR tree-optimization/13000
449 * tree-inline.c: Include "tree-flow.h".
450 (expand_call_inline): If warn_return_type, warn if non-void inline
451 function falls through.
452 * tree-cfg.c (execute_warn_function_return): Don't warn about
453 control reaching end if TREE_NO_WARNING is set. Set
455 * gimple-low.c (block_may_fallthru): Don't assume that SWITCH_EXPR
457 * gimplify.c (shortcut_cond_expr): Don't emit a jump over the else
458 branch if we don't need one.
459 * c-typeck.c: Include "tree-flow.h"
460 (c_finish_bc_stmt): Don't add a goto if the current statement
461 list doesn't fall through to the current point.
463 2005-01-21 Roger Sayle <roger@eyesopen.com>
465 PR rtl-optimization/576
466 * real.c (real_arithmetic): Change return type from void to bool
467 to return an indication that the result may be inexact.
468 * real.h (real_arithmeric): Update prototype.
469 * fold-const.c (const_binop): Don't constant fold floating
470 point expressions when the user specifies -frounding-math and
471 the result may depend upon the run-time rounding mode.
472 (fold_convert_const_real_from_real): Clean-up.
473 (fold_initializer): Ignore flag_rounding_math for initializers.
474 * simplify-rtx.c (simplify_binary_operation): Likewise, don't
475 constant fold FP operations with flag_rounding_math if the
476 result may depend upon the run-time rounding mode.
478 2005-01-21 Tom Tromey <tromey@redhat.com>
480 * c-cppbuiltin.c (define__GNUC__): Correct assertion.
482 2005-01-21 Kazu Hirata <kazu@cs.umass.edu>
484 * tree-cfg.c: Fix comment typos.
486 2005-01-21 J"orn Rennecke <joern.rennecke@st.com>
488 * passes.c (rest_of_handle_flow2): Close / open the flow2 dump file
489 around the call to rest_of_handle_branch_target_load_optimize.
491 2005-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
493 * config.gcc (sparc-*-netbsdelf*): Include dbxelf.h
494 (sparc64-*-openbsd*): Likewise.
495 (sparclite-*-elf*): Include tm-dwarf2.h.
496 (sparc86x-*-elf*): Likewise.
497 (sparc64-*-elf*): Likewise.
498 * config/sparc/linux64.h (DWARF2_DEBUGGING_INFO): Delete.
499 (DBX_DEBUGGING_INFO): Likewise.
500 * config/sparc/liteelf.h (DWARF2_DEBUGGING_INFO): Likewise.
501 (PREFERRED_DEBUGGING_TYPE): Likewise.
502 * config/sparc/netbsd-elf.h (PREFERRED_DEBUGGING_TYPE): Likewise.
503 * config/sparc/sp64-elf.h (DBX_DEBUGGING_INFO): Likewise.
504 (PREFERRED_DEBUGGING_TYPE): Likewise.
505 * config/sparc/sp86x-elf.h (DWARF2_DEBUGGING_INFO): Likewise.
506 (PREFERRED_DEBUGGING_TYPE): Likewise.
507 * config/sparc/sparc.h (DBX_DEBUGGING_INFO): Likewise.
509 2005-01-21 Hans-Peter Nilsson <hp@bitrange.com>
512 * combine.c (combine_simplify_rtx): Revert change of 2004-12-31.
514 2005-01-20 Paul Brook <paul@codesourcery.com>
516 * doc/rtl.texi: Document value extension requirements for CONST_INT.
518 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
520 * tree-cfg.c (tree_verify_flow_info): Check that a nonlocal
521 label is first in a sequence of labels.
523 * tree-cfg.c (tree_verify_flow_info): Fix a typo.
525 2005-01-20 Janis Johnson <janis187@us.ibm.com>
526 Giovanni Bajo <giovannibajo@gcc.gnu.org>
528 * doc/sourcebuild.texi (Test Directives): New.
530 2005-01-20 Roger Sayle <roger@eyesopen.com>
532 * real.c (real_floor): Don't leave the result uninitialized when
533 mode is VOIDMode, but return the unrounded intermediate value.
534 (real_ceil): Likewise.
536 2005-01-20 Paul Brook <paul@codesourcery.com>
538 * config/arm/arm.md (insv): Use gen_int_mode.
540 2005-01-20 Eric Botcazou <ebotcazou@libertysurf.fr>
542 * config/sparc/freebsd.h (ASM_OUTPUT_INTERNAL_LABELREF): Delete.
543 * config/sparc/linux.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
544 * config/sparc/linux64.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
545 * config/sparc/netbsd-elf.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
546 * config/sparc/sol2.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
548 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
551 * dwarf2out.c (concat_loc_descriptor): We don't know we can use
552 fbreg, so we have to assume we can't.
553 (loc_descriptor_from_tree_1): Ditto.
554 (containing_function_has_frame_base): New function.
555 (add_location_or_const_value_attribute): Use it.
556 Also try to generate a frame_base from a single element location
559 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
561 PR tree-optimization/15349
562 * timevar.def (TV_TREE_MERGE_PHI): New.
563 * tree-cfg.c (tree_forwarder_block_p): Add a new argument
565 (remove_forwarder_block, cleanup_forwarder_blocks): Adjust the
566 calls to tree_forwarder_block_p.
567 (remove_forwarder_block_with_phi, merge_phi_nodes,
568 gate_merge_phi, pass_merge_phi): New.
569 * tree-optimize.c (init_tree_optimization_passes): Add
571 * tree-pass.h: Add an extern for pass_merge_phi;
573 2005-01-20 Richard Henderson <rth@redhat.com>
576 * config/i386/emmintrin.h (_mm_castpd_ps, _mm_castpd_si128): New.
577 (_mm_castps_pd, _mm_castps_si128): New.
578 (_mm_castsi128_ps, _mm_castsi128_pd): New.
580 2005-01-20 Richard Henderson <rth@redhat.com>
583 * config/i386/mmintrin.h (_mm_cvtsi32_si64): Use
584 __builtin_ia32_vec_init_v2si.
585 (_mm_cvtsi64_si32): Use __builtin_ia32_vec_ext_v2si.
586 * config/i386/i386.c (IX86_BUILTIN_VEC_EXT_V2SI): New.
587 (ix86_init_mmx_sse_builtins): Create it.
588 (ix86_expand_builtin): Expand it.
589 (ix86_expand_vector_set): Handle V2SFmode and V2SImode.
590 * config/i386/mmx.md (vec_extractv2sf_0, vec_extractv2sf_1): New.
591 (vec_extractv2si_0, vec_extractv2si_1): New.
593 2005-01-20 Richard Henderson <rth@redhat.com>
596 * config/i386/i386.c (ix86_expand_push): New.
597 * config/i386/mmx.md (push<MMXMODE>1): New.
598 * config/i386/sse.md (push<SSEMODE>1): New.
599 * config/i386/i386-protos.h: Update.
601 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
603 * params.def, params.h, predict.c, tree-ssa-loop-im.c: Update
606 2005-01-20 Richard Henderson <rth@redhat.com>
608 * rtl.def (CONST_VECTOR): Use RTX_CONST_OBJ.
609 * rtl.h (CONSTANT_P): Don't special case CONST_VECTOR.
611 2005-01-19 Richard Henderson <rth@redhat.com>
614 * config/i386/i386.c (ix86_expand_vector_move_misalign): Convert
615 to V4SFmode in SSE1 fallback load path.
617 2005-01-19 Richard Henderson <rth@redhat.com>
619 * config/i386/i386.c (ix86_expand_vector_init_one_var): Fix typo
620 in QImode expansion to ix86_expand_vector_set.
621 (ix86_expand_vector_init_general): Fix typo in V8HImode recursive call.
623 2005-01-19 Richard Henderson <rth@redhat.com>
626 * config/i386/i386.c (ix86_preferred_reload_class): Return a proper
627 subclass of the input class.
628 (ix86_secondary_memory_needed): Always true for cross-MMX classes.
629 Always true for cross-SSE1 classes. Rationalize conditionals.
630 * config/i386/i386.h (SSE_CLASS_P, MMX_CLASS_P): Use straight equality.
631 * config/i386/i386.md (movsi_1): Add MMX/SSE zeros. Fix alternatives
632 for SSE1. Don't check TARGET_INTER_UNIT_MOVES.
633 (movdi_2): Add MMX/SSE zeros.
634 (movdi_1_rex64): Likewise. Don't check TARGET_INTER_UNIT_MOVES.
635 (movsf_1): Don't check TARGET_INTER_UNIT_MOVES.
636 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
637 (movsi_1_nointernunit, movdi_1_rex64_nointerunit): Remove.
638 (movsf_1_nointerunit, zero_extendsidi2_32_1): Remove.
639 (zero_extendsidi2_rex64_1): Remove.
640 (MOV0 peephole): Check GENERAL_REG_P.
642 2005-01-19 Richard Henderson <rth@redhat.com>
645 * config/i386/i386.c (ix86_expand_vector_set): Fix third and fourth
647 (ix86_expand_vector_extract): Likewise.
649 2005-01-19 Richard Henderson <rth@redhat.com>
652 * expr.c (emit_move_change_mode): New argument force; use
653 simplify_gen_subreg if true.
654 (emit_move_via_alt_mode): Merge into ...
655 (emit_move_via_integer): ... here.
656 (emit_move_ccmode): Use emit_move_change_mode directly.
658 2005-01-19 Richard Henderson <rth@redhat.com>
661 * config/alpha/alpha.c (alpha_rtx_costs): Handle HIGH.
662 (alpha_preferred_reload_class): Handle CONST_VECTOR.
663 (alpha_emit_set_const_1): Add no_output parameter; don't emit
665 (alpha_emit_set_const): Likewise. Make static.
666 (alpha_emit_set_long_const): Make static.
667 (alpha_extract_integer): Split out from alpha_expand_mov.
668 (alpha_split_const_mov): Likewise.
669 (alpha_expand_mov): Use them. Handle CONST_VECTOR.
670 (alpha_legitimate_constant_p): New.
671 * config/alpha/alpha-protos.h: Update.
672 * config/alpha/alpha.h (REGISTER_MOVE_COST): Correct fp<->gp cost.
673 (LEGITIMATE_CONSTANT_P): Re-implement with a function.
674 * config/alpha/alpha.md (movsi): Add n alternative.
675 (movsi_nt_vms, movdi_er_nofix, movdi_er_fix, movdi_fix): Likewise.
676 (mov<VEC>_fix, mov<VEC>_nofix): Add i alternative.
677 (splitters for all of the above): Use alpha_split_const_mov.
678 * config/alpha/predicates.md (non_add_const_operand): New.
679 (non_zero_const_operand): New.
680 (input_operand): Use alpha_legitimate_constant_p after reload.
682 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
684 PR tree-optimization/19038
685 * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): New function.
686 (add_candidate): Add ivs with increment in latch only if
687 allow_ip_end_pos_p is true.
688 (determine_iv_cost): Use empty_block_p.
690 2005-01-19 Daniel Berlin <dberlin@dberlin.org>
692 * cfganal.c (compute_dominance_frontiers_1): Replace with new algorithm
693 (compute_dominance_frontiers): Ditto.
695 2005-01-19 Ralf Corsepius <ralf.corsepius@rtems.org>
698 * config/sh/t-rtems: New.
699 * config.gcc (sh-*-rtems*): Reflect having added config/sh/t-rtems.
701 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
703 * tree-ssa-ccp.c (maybe_fold_offset_to_component_ref): Always subtract
704 the offset of the selected field.
706 2005-01-19 Kazu Hirata <kazu@cs.umass.edu>
708 * tree-cfg.c (remove_forwarder_block): Fix the check to
709 prevent a nonlocal label from appearing in the middle of a
712 2005-01-19 Hans-Peter Nilsson <hp@axis.com>
714 PR rtl-optimization/19462
715 * reorg.c (find_end_label): Create return insn only if
716 current_function_epilogue_delay_list is empty.
718 2005-01-19 Paolo Bonzini <bonzini@gnu.org>
720 * doc/passes.texi: Remove paragraph mentioning new-ra.
722 2005-01-19 Jakub Jelinek <jakub@redhat.com>
724 PR rtl-optimization/15139
725 * combine.c: Include params.h.
726 (count_rtxs): New function.
727 (record_value_for_reg): If replace_rtx would replace at least
728 2 occurrences of REG in VALUE and TEM is really large, replace REG with
729 (clobber (const_int 0)) instead of TEM.
730 * params.def (PARAM_MAX_LAST_VALUE_RTL): New.
731 * params.h (MAX_LAST_VALUE_RTL): New.
732 * Makefile.in (combine.o): Depend on $(PARAMS_H).
733 * doc/invoke.texi (--param max-last-value-rtl=N): Document.
736 * c-typeck.c (digest_init): Only call build_vector if all constructor
737 elements are *_CST nodes.
738 * gimplify.c (gimplify_init_constructor): Likewise.
741 * c-typeck.c (digest_init): Only call build_vector if inside_init
744 2005-01-18 Tobias Schl"uter <tobias.schlueter@physik.uni-muenchen.de>
746 * toplev.c (init_asm_output): Remove dead #ifdef.
748 2005-01-18 Kaz Kojima <kkojima@gcc.gnu.org>
750 * config/sh/linux.h (TARGET_C99_FUNCTIONS): Define.
752 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
754 * calls.c (expand_call): Check DECL_BUILT_IN_CLASS before
755 accessing DECL_FUNCTION_CODE.
756 * dojump.c (do_jump): Likewise.
757 * gimplify.c (gimplify_call_expr): Likewise.
758 * predict.c (expr_expected_value): Likewise.
759 (strip_builtin_expect): Likewise.
760 * tree-inline.c (estimate_num_insns_1): Likewise.
761 * tree-ssa-loop-im.c (stmt_cost): Likewise
762 * fold-const.c (fold): Test for BUILT_IN_NORMAL.
763 (tree_expr_nonnegative_p): Likewise.
765 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
767 * config/sparc/sparc.c (load_pic_register): Emit the appropriate
768 variant of the load_pcrel_sym pattern.
769 * config/sparc/sparc.md (P macro): Move to the top.
770 (load_pcrel_sym): Macroize using P.
771 (save_register_window): Likewise.
773 2005-01-18 Aldy Hernandez <aldyh@redhat.com>
775 * config/rs6000/rs6000.md ("sunordered"): Disable for e500.
778 2005-01-18 Andrew Pinski <pinskia@physics.uc.edu>
781 * c-typeck.c (build_asm_expr): Strip nops off
782 input memory operands.
784 2005-01-18 David Edelsohn <edelsohn@gnu.org>
786 * config/rs6000/aix43.h (CPLUSPLUS_CPP_SPEC): Delete
787 _XOPEN_SOURCE, _XOPEN_SOURCE_EXTENDED, _LARGE_FILE_API.
788 * config/rs6000/aix51.h (CPLUSPLUS_CPP_SPEC): Same.
789 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Same.
791 2005-01-18 Joel Sherrill <joel.sherrill@oarcorp.com>
794 * config/i386/i386.c (override_options): If the 80387 is disabled,
795 then do not return FP values using FP registers.
797 2005-01-18 Kazu Hirata <kazu@cs.umass.edu>
799 * basic-block.h, c-common.c, c-cppbuiltin.c, c-lang.c,
800 c-tree.h, cfgbuild.c, cgraph.c, cgraph.h, collect2.c,
801 combine.c, config.gcc, coverage.h, cse.c, cselib.c,
802 defaults.h, df.c, dwarf2asm.c, dwarf2out.c, explow.c, expr.c,
803 flow.c, fold-const.c, gcse.c, ggc-page.c, gimple-low.c,
804 gimplify.c, ifcvt.c, langhooks-def.h, lcm.c, optabs.h,
805 output.h, postreload-gcse.c, postreload.c, recog.c,
806 resource.c, rtl.def, rtlanal.c, sched-deps.c, sched-rgn.c,
807 targhooks.h, toplev.c, tree-data-ref.c, tree-eh.c,
808 tree-flow.h, tree-loop-linear.c, tree-mudflap.h, tree-nrv.c,
809 tree-optimize.c, tree-outof-ssa.c, tree-pass.h,
810 tree-scalar-evolution.c, tree-ssa-copy.c, tree-ssa-dce.c,
811 tree-ssa-dse.c, tree-ssa-loop-ivopts.c, tree-ssa-operands.c,
812 tree-ssa-pre.c, tree-ssa.c, tree-vectorizer.c, tree.def,
813 unwind-dw2-fde-darwin.c, var-tracking.c: Update copyright.
815 * config/darwin.c, config/darwin.h, config/freebsd-spec.h,
816 config/alpha/alpha.c, config/arm/vfp.md,
817 config/i386/emmintrin.h, config/i386/i386-modes.def,
818 config/i386/i386-protos.h, config/i386/i386.h,
819 config/i386/pmmintrin.h, config/i386/ppro.md,
820 config/i386/predicates.md, config/i386/xmmintrin.h,
821 config/i860/i860.c, config/ia64/ia64-modes.def,
822 config/ia64/ia64-protos.h, config/ia64/ia64.h,
823 config/ia64/ia64.md, config/ia64/itanium1.md,
824 config/ia64/itanium2.md, config/ia64/predicates.md,
825 config/iq2000/iq2000.h, config/mips/linux64.h,
826 config/rs6000/aix41.h, config/rs6000/aix43.h,
827 config/rs6000/aix52.h, config/rs6000/darwin-fallback.c,
828 config/rs6000/rs6000.c, config/rs6000/rs6000.h,
829 config/rs6000/spe.md, config/sh/sh.md, config/sparc/linux.h,
830 config/sparc/linux64.h, config/sparc/litecoff.h,
831 config/sparc/sp64-elf.h, config/sparc/sparc.c,
832 config/sparc/sparc.h, config/sparc/sparc.md,
833 config/sparc/sysv4-only.h: Update copyright.
835 2005-01-18 Richard Henderson <rth@redhat.com>
838 * config/i386/i386.c (ix86_expand_fp_movcc): Fail for LTGT and UNEQ.
840 2005-01-18 Marc Espie <espie@openbsd.org>
841 * config/i386/openbsdelf.h: Typo.
843 2005-01-18 Richard Henderson <rth@redhat.com>
845 * config/i386/i386.c (ix86_expand_fp_movcc): Remove TARGET_IEEE_FP
846 special case for sse.
847 * config/i386/predicates.md (sse_comparison_operator): Likewise.
849 2005-01-18 Dorit Naishlos <dorit@il.ibm.com>
851 * tree-inline.c (estimate_num_insns_1): Added cases for
852 ALIGN_INDIRECT_REF, MISALIGNED_INDIRECT_REF, and REALIGN_LOAD_EXPR.
854 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
856 PR rtl-optimization/19296
857 * combine.c (simplify_comparison): Rewrite the condition under
858 which a non-paradoxical SUBREG of a PLUS can be lifted when
859 compared against a constant.
861 2005-01-18 Andi Kleen <ak@muc.de>
863 * c-typeck.c: (convert_for_assignment): Check warn_pointer_sign.
864 * c.opt (-Wpointer-sign): Add.
865 * doc/invoke.texi: (-Wpointer-sign): Add.
867 2005-01-18 Uros Bizjak <uros@kss-loka.si>
869 * config/i386/i386.c (override_options): Revert 2004-11-24 change.
870 * config/i386/i386.md (fmodsf3, dremsf3, *sinsf2, *cossf2, sincossf3,
871 *tansf3_1, tansf2, atan2sf3_1, atan2sf3, atansf2, asinsf2, acossf2,
872 logsf2, log10sf2, log2sf2, log1psf2, logbsf2, ilogbsf2, expsf2,
873 exp10sf2, exp2sf2, expm1sf2, rintsf2, floorsf2, ceilsf2, btruncsf2,
874 nearbyintsf2): Disable for TARGET_SSE_MATH. Leave patterns enabled
875 for TARGET_MIX_SSE_I387.
876 (fmoddf3, dremdf3, *sindf2, *sinextendsfdf2, *cosdf2, *cosextendsfdf2,
877 sincosdf3, *sincosextendsfdf3, *tandf3_1, tandf2, atan2df3_1,
878 atan2df3, atandf2, asindf2, acosdf2, logdf2, log10df2, log2df2,
879 log1pdf2, logbdf2, expdf2, exp10df2, exp2df2, expm1df2, rintdf2,
880 floordf2, ceildf2, btruncdf2, nearbyintdf2): Disable for
881 (TARGET_SSE2 && TARGET_SSE_MATH). Leave patterns enabled for
883 (atan2sf3, atan2df3, atan2xf3): Remove register constraints
886 2005-01-18 Uros Bizjak <uros@kss-loka.si>
889 * config/i386/mmx.md (*movv2sf_internal_rex64, *movv2sf_internal):
890 Add movaps alternative for xmm reg->reg move.
892 2005-01-17 Jeff Law <law@redhat.com>
894 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Clear SSA_NAME_VALUE
895 at the end of the main DOM loop rather than just before DOM exits.
897 2005-01-17 Diego Novillo <dnovillo@redhat.com>
899 PR tree-optimization/19121
900 * tree-ssa-alias.c (compute_flow_sensitive_aliasing): When
901 adding aliases to a name tag, also add them to the pointer's
903 * tree-ssa-copy.c (merge_alias_info): Do not merge flow
904 sensitive alias info at all. Only check that the two pointers
905 have compatible pointed-to sets.
906 * tree-ssa.c (verify_name_tags): Verify that the alias set of
907 a pointer's type tag is a superset of the alias set of the
910 2005-01-17 James E Wilson <wilson@specifixinc.com>
913 * config/ia64/ia64.md (movxf): Handle general register source. Adjust
914 comment to document why.
916 2005-01-17 Richard Henderson <rth@redhat.com>
918 * config/i386/sse.md (smaxv4sf3_finite, sse_vmsmaxv4sf3_finite,
919 sminv4sf3_finite, sse_vmsminv4sf3_finite, smaxv2df3_finite,
920 sse2_vmsmaxv2df3_finite, sminv2df3_finite,
921 sse2_vmsminv2df3_finite): New.
922 (smaxv4sf3, sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, smaxv2df3,
923 sse2_vmsmaxv2df3, sminv2df3, sse2_vmsminv2df3): Remove commutative.
924 Force op1 into register.
926 2005-01-17 Kelley Cook <kcook@gcc.gnu.org>
928 * Makefile.in (STAGEMOVESTUFF): Stage all the stamp files to prevent
929 bubblestrap from always rebuilding build directory.
931 2005-01-17 Ulrich Weigand <uweigand@de.ibm.com>
933 * config/s390/s390.md ("*clc"): Remove incorrect '='.
934 ("*cmpmem_short"): Likewise.
936 2005-01-17 Steven Bosscher <stevenb@suse.de>
938 * basic-block.h: Document BB_* flags.
939 * regrename.c (copyprop_hardreg_forward): Don't use BB_VISITED,
940 use an sbitmap instead.
941 * sched-rgn.c (compute_trg_info): Likewise.
943 2005-01-17 Richard Sandiford <rsandifo@redhat.com>
945 * config.gcc (mips64*-*-linux*): Set the default abi to n32. Remove
946 redundant target_cpu_default and extra_parts lines. Use the default
948 * config/mips/linux64.h (DRIVER_DEFAULT_ABI_SELF_SPEC): Delete.
949 (SUBTARGET_EXTRA_SPECS): Delete.
950 (DRIVER_SELF_SPECS): Remove architecture lines. Set the ABI directly.
951 (SUBTARGET_ASM_SPEC): Remove -mabi=n32 and -mabi=64 mappings.
952 * config/mips/t-linux64 (MULTILIB_DIRNAMES): Use "32" for -mabi=32
953 and "n32" for -mabi=n32.
955 2005-01-17 Ian Lance Taylor <ian@airs.com>
958 * c-typeck.c (build_c_cast): Revert patch of 2003-02-16: don't
959 fold constant variables into initial values.
961 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
963 * tree-cfg.c (tree_can_merge_blocks_p): Reorder two checks.
965 * tree-cfg.c (tree_forwarder_block_p): Speed up by walking
966 through the statements backward.
968 2005-01-17 Ian Lance Taylor <ian@airs.com>
971 * tree-inline.c (expand_call_inline): Set TREE_NO_WARNING on
972 a variable set to the return value of the inlined function.
974 2005-01-17 Mark Dettinger <dettinge@de.ibm.com>
976 * rtlanal.c (rtx_cost): Assign cost of 0 to a SUBREG
977 when modes are tieable.
979 2005-01-17 Ranjit Mathew <rmathew@hotmail.com>
981 * gthr-posix.h (__gthread_active_p): Use pthread_cancel instead
982 of pthread_create to find out if threads are enabled.
983 * gthr-posix95.h (__gthread_active_p): Likewise.
985 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
987 * common.opt (-fnew-ra): Remove.
989 * toplev.h (flag_new_regalloc): Remove.
990 * Makefile.in (ra*.*): Don't mention.
991 * passes.c (rest_of_handle_new_regalloc): Remove.
992 (rest_of_handle_combine, rest_of_compilation): Always consider
993 flag_new_regalloc as false.
994 * doc/invoke.texi: Don't document -fnew-ra.
996 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
998 * bb-reorder.c (fix_edges_for_rarely_executed_code): Remove
999 last parameter to reg_scan.
1000 * loop.c (loop_optimize): Likewise.
1001 * passes.c (rest_of_handle_tracer, rest_of_handle_if_conversion,
1002 rest_of_handle_web, rest_of_handle_cfg, rest_of_handle_jump_bypass,
1003 rest_of_handle_life, rest_of_handle_cse, rest_of_handle_cse2,
1004 rest_of_handle_gcse, rest_of_handle_loop_optimize,
1005 rest_of_handle_loop2, rest_of_handle_jump2): Likewise.
1006 * regclass.c (reg_scan): Likewise, for the declaration.
1007 * rtl.h (reg_scan): Likewise, for the prototype.
1009 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
1011 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Don't call
1012 mark_dfs_back_edges.
1014 2005-01-16 Daniel Berlin <dberlin@dberlin.org>
1016 * tree-ssa-pre.c (add_to_sets): s1 may be NULL.
1017 (compute_avail): Uses don't go in tmp_gen.
1019 2005-01-16 Steven Bosscher <stevenb@suse.de>
1021 * ggc-page.c (ggc_alloc_stat): Use __builtin_ctzl instead of a
1022 loop to look for a free slot in a page entry.
1024 2005-01-16 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1027 * defaults.h (TARGET_DEFERRED_OUTPUT_DEFS): Provide default.
1028 * toplev.c (compile_file): Call process_pending_assemble_output_defs
1029 just before targetm.asm_out.file_end.
1030 * tree.h (process_pending_assemble_output_defs): Declare.
1031 * varasm.c (assemble_output_def, process_pending_assemble_output_defs):
1033 (assemble_alias): Defer generation of assembly code for defines when
1034 TARGET_DEFERRED_OUTPUT_DEFS is true.
1035 * config/rs6000/aix41.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
1036 * config/rs6000/aix43.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
1037 * doc/tm.texi (TARGET_DEFERRED_OUTPUT_DEFS): document.
1039 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1042 * pa.c (pa_scalar_mode_supported_p): New function.
1043 (TARGET_SCALAR_MODE_SUPPORTED_P): Define.
1044 * pa.h (MIN_UNITS_PER_WORD): Add comment.
1046 * pa-protos.h (prefetch_operand): Delete.
1047 (prefetch_cc_operand, prefetch_nocc_operand): New declations.
1048 * pa.c (prefetch_operand): Delete.
1049 (prefetch_cc_operand, prefetch_nocc_operand): New functions.
1050 * pa.h (EXTRA_CONSTRAINT): Add `W' constraint.
1051 (PREDICATE_CODES): Delete prefetch_operand. Add prefetch_cc_operand
1052 and prefetch_nocc_operand.
1053 * pa.md (prefetch): Rework to avoid reload problems handling short
1054 displacements when a cache control completer needs to be provided.
1055 (prefetch_32, prefetch_64): Delete.
1056 (prefetch_cc, prefetch_nocc): New patterns.
1058 2005-01-15 David Edelsohn <edelsohn@gnu.org>
1060 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Revert previous change.
1062 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1065 * pa.c (pa_asm_output_aligned_common): Warn if specified alignment
1066 exceeds maximum alignment for global common data.
1067 * pa64-hpux.h (MAX_OFILE_ALIGNMENT): Define to 32768.
1068 * som.h (MAX_OFILE_ALIGNMENT): Likewise.
1070 2004-01-15 Roger Sayle <roger@eyesopen.com>
1072 * tree-ssa-dom.c (extract_range_from_cond): Correct condition.
1074 2004-01-15 Roger Sayle <roger@eyesopen.com>
1076 * harg-reg-set.h (reg_class_names): Prototype global array.
1077 * regclass.c (reg_class_names): Declare here and initialize to
1079 (dump_regclass): Remove local declaration of reg_class_names.
1080 (regclass): Likewise.
1081 * cfg.c (dump_flow_info): Likewise.
1082 * ra-debug.c (reg_class_names): Likewise.
1083 * regrename.c (reg_class_names): Likewise.
1084 * reload.c (reg_class_names): Likewise.
1085 * reload1.c (spill_failure): Likewise.
1086 * config/m68hc11/m68hc11.c (reg_class_names): Likewise.
1088 2005-01-15 Ulrich Weigand <uweigand@de.ibm.com>
1090 * config/s390/s390.md ("reload_outti"): Remove predicate for
1091 output operand. Abort if operand is not a MEM.
1092 ("reload_outdi", "reload_outdf"): Likewise.
1094 2005-01-15 Marc Espie <espie@openbsd.org>
1096 * config.gcc (*-*-openbsd*): Set HAS_LIBC_R for OpenBSD <= 3.2.
1097 (i[34567]86-*-openbsd*): Switch to ELF for OpenBSD >= 3.4.
1098 * config/openbsd.h: Fix C++ includes for native configurations.
1099 Add proper OS_CPP_BUILTINS. Fix libspec for recent OpenBSD.
1100 Add trampoline support.
1101 * config/i386/openbsdelf.h: New.
1103 2005-01-15 Marc Espie <espie@openbsd.org>
1105 * collect2.c (main): Explicitly parse -dynamic-linker option.
1107 2005-01-15 Jakub Jelinek <jakub@redhat.com>
1109 PR tree-optimization/19060
1110 * tree-ssa-dom.c (extract_range_from_cond) <case LT_EXPR, GT_EXPR>:
1111 Return 0 if op1 <= TYPE_MIN_VALUE () resp. op1 >= TYPE_MAX_VALUE ().
1112 (simplify_cond_and_lookup_avail_expr): Add assert for dummy == 0
1113 and handle extract_range_from_cond returning false.
1114 * fold-const.c (fold): Optimize comparisons with min/max even for
1115 width > HOST_BITS_PER_WIDE_INT.
1117 2005-01-15 Ralf Corsepius <ralf.corsepius@rtems.org>
1119 * config/mips/rtems.h (MIPS_DEFAULT_GVALUE): Set to 0.
1120 * config/mips/t-rtems (MULTILIBS_DIRNAMES,MULTILIB_OPTIONS):
1121 Remove little endian multilib variants.
1122 Add mips32 multilib variant.
1124 2005-01-15 Kazu Hirata <kazu@cs.umass.edu>
1126 * explow.c (copy_all_reg, stabilize): Remove.
1127 * expr.h: Remove the corresponding prototypes.
1129 * rtlanal.c (regs_set_between_p): Remove.
1130 * rtl.h: Remove the corresponding prototype.
1132 2005-01-14 David Edelsohn <edelsohn@gnu.org>
1134 * config/rs6000/rs6000.c (rs6k_nonimmediate_operand): Rename
1135 to rs6000_nonimmediate_operand.
1136 * config/rs6000/rs6000.h (PREDICATE_CODES): Rename
1137 rs6k_nonimmediate_operand to rs6000_nonimmediate_operand.
1138 * config/rs6000/rs6000.md (movsi_internal1): Rename predicate to
1139 rs6000_nonimmediate_operand.
1140 * config/rs6000/spe.md (movdf_e500_double): Rename predicate to
1141 rs6000_nonimmediate_operand.
1143 2005-01-14 Andrew Pinski <pinskia@physics.uc.edu>
1145 * tree-ssa-alias.c (pass_may_alias): Add TODO_verify_stmts.
1147 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1149 * config/rs6000/rs6000.h (CLASS_MAX_NREGS): DF goes in 1 register
1151 (CANNOT_CHANGE_MODE_CLASS): Restrict DI mode changes on e500v2.
1152 (PREDICATE_CODES): Add rs6k_nonimmediate_operand.
1154 * config/rs6000/rs6000.c (invalid_e500_subreg): New.
1155 (rs6k_nonimmediate_operand): New.
1156 (rs6000_legitimate_offset_address_p): Handle DI modes on e500v2
1158 (legitimate_lo_sum_address_p): Same.
1159 (rs6000_legitimize_address): Same.
1160 (rs6000_legitimize_reload_address): Same.
1161 (rs6000_legitimate_address): Same.
1162 (spe_build_register_parallel): Pass DF and DC modes in a DI
1165 * config/rs6000/rs6000.md ("*movsi_internal1"): Change predicate
1166 to rs6k_nonimmediate_operand.
1168 * config/rs6000/spe.md ("*frob_df_di"): New.
1169 ("*frob_di_df"): New.
1170 ("*frob_di_df_2"): New.
1171 ("*mov_sidf_e500_subreg0"): New.
1172 ("*mov_sidf_e500_subreg4"): New.
1173 ("*movdf_e500_double"): Change predicate to
1174 rs6k_nonimmediate_operand.
1176 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1178 * postreload.c (move2add_note_store): Only call
1179 trunc_int_for_mode on scalar integers.
1181 2005-01-14 Steven Bosscher <stevenb@suse.de>
1183 * tree-ssa-dce.c (visited_control_parents): New sbitmap to
1184 replace BB_VISITED uses.
1185 (find_obviously_necessary_stmts): Don't clear BB_VISITED.
1186 (propagate_necessity): Check the bitmap instead of BB_VISITED.
1187 (tree_dce_done): Free visited_control_parents.
1188 (perform_tree_ssa_dce): Allocate and clear it.
1189 * tree-ssa-pre.c (compute_antic_aux): Make non-recursive.
1190 (compute_antic): Iterate from here using a DFS. Use an sbitmap
1191 instead of BB_VISITED.
1193 2005-01-14 Kazu Hirata <kazu@cs.umass.edu>
1195 * c-tree.h, coverage.h, langhooks-def.h, optabs.h, output.h,
1196 rtl.h, targhooks.h, tree-flow.h, tree-mudflap.h, tree.h:
1197 Remove unused prototypes.
1199 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1201 * config/rs6000/rs6000.c (rs6000_stack_info): Fix caching of SPE
1202 64-bit register usage.
1203 (spe_func_has_64bit_regs_p): Add FIXME note.
1205 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1207 * config/rs6000/rs6000.c (rs6000_generate_compare): Replace
1208 flag_finite_math_only with flag_unsafe_math_optimizations.
1210 2005-01-14 J. D. Johnston <jjohnst@us.ibm.com>
1212 * config/s390/tpf-unwind.h (s390_fallback_frame_state): Correct
1214 (__tpf_eh_return): Copy TPF private stack area from the module
1215 boundary stack frame. Add check for when module addresses are
1216 equal, but no stub address is found.
1218 2005-01-14 Richard Earnshaw <rearnsha@arm.com>
1221 * arm.h (struct machine_function): Add call_via field.
1222 (thumb_call_via_label): Declare.
1223 * arm.c (thumb_call_via_label): New variable.
1224 (thumb_call_reg_needed): New variable.
1225 (arm_output_function_epilogue): For Thumb code, output any per-function
1226 call-indirect trampolines.
1227 (thumb_call_via_reg): New function.
1228 (arm_file_end): New function.
1229 (TARGET_ASM_FILE_END): Call arm_file_end.
1230 (aof_file_end): Likewise.
1231 * arm-protos.h (thumb_call_via_reg): Declare.
1232 * arm.md (call_reg_thumb, call_value_reg_thumb): Call
1233 thumb_call_via_reg in normal case.
1235 2005-01-14 Jakub Jelinek <jakub@redhat.com>
1238 PR rtl-optimization/19348
1239 * recog.c (peephole2_optimize): Do global life update if some peephole
1240 decides it doesn't need at least one of its inputs and that change
1241 influences liveness at the start of the basic block.
1243 * basic-block.h (EXECUTE_IF_AND_COMPL_IN_REG_SET): Needs 2 REGSET
1244 arguments instead of 1.
1246 2005-01-14 Eric Botcazou <ebotcazou@libertysurf.fr>
1249 * varasm.c (initializer_constant_valid_p) <ADDR_EXPR>: Return
1250 zero for nested functions needing a static chain or functions
1251 with a non-constant address.
1253 2005-01-13 Roger Sayle <roger@eyesopen.com>
1255 * simplify-rtx.c (simplify_binary_operation) <AND>: Optimize
1256 (and (sign_extend X) C) into (zero_extend (and X C)).
1258 2005-01-13 David O'Brien <obrien@FreeBSD.org>
1260 * config/freebsd-spec.h: Make KSE pthread lib logic the default.
1262 2005-01-13 Richard Henderson <rth@redhat.com>
1267 * config/i386/i386.md (cmpdf, cmpsf, bunordered, bordered, buneq,
1268 bunge, bungt, bunle, bunlt, bltgt): Enable for TARGET_SSE_MATH,
1269 not just TARGET_SSE.
1270 (cmpfp_i_387): Rename from cmpfp_i. Move after sse patterns.
1271 (cmpfp_i_mixed): Rename from cmpfp_i_sse; use for TARGET_MIX_SSE_I387.
1272 (cmpfp_i_sse): Rename from cmpfp_i_sse_only; use for TARGET_SSE_MATH.
1273 (cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387): Similarly.
1274 (fp_jcc_1_mixed, fp_jcc_1_sse, fp_jcc_1_387): Similarly.
1275 (fp_jcc_2_mixed, fp_jcc_2_sse, fp_jcc_2_387): Similarly.
1276 (fp_jcc_3_387, fp_jcc_4_387, fp_jcc_5_387, fp_jcc_6_387,
1277 fp_jcc_7_387, fp_jcc_8_387): Rename from fp_jcc_N.
1278 (movdicc_c_rex64): Rename with '*'.
1279 (movsfcc, movdfcc): Add checks for 387 and sse math to condition.
1280 (movsfcc_1_sse_min, movsfcc_1_sse_max, movsfcc_1_sse): New.
1281 (movsfcc_1_387): Rename from movsfcc_1.
1282 (movdfcc_1_sse_min, movdfcc_1_sse_max, movdfcc_1_sse): New.
1283 (movdfcc_1, movdfcc_1_rex64): Add check for 387.
1284 (sminsf3, smaxsf3, smindf3, smaxdf3): New.
1285 (minsf3, minsf, minsf_nonieee, minsf_sse, mindf3, mindf,
1286 mindf_nonieee, mindf_sse, maxsf3, maxsf, maxsf_nonieee, maxsf_sse,
1287 maxdf3, maxdf, maxdf_nonieee, maxdf_sse, sse_movsfcc, sse_movsfcc_eq,
1288 sse_movdfcc, sse_movdfcc_eq, sse_movsfcc_const0_1,
1289 sse_movsfcc_const0_2, sse_movsfcc_const0_3, sse_movsfcc_const0_4,
1290 sse_movdfcc_const0_1, sse_movdfcc_const0_2, sse_movdfcc_const0_3,
1291 sse_movdfcc_const0_4): Remove.
1292 * config/i386/i386.c (ix86_expand_fp_movcc): For TARGET_SSE_MATH,
1293 recognize min/max early. Update for changed sse cmove patterns.
1294 (ix86_split_sse_movcc): New.
1295 * config/i386/i386-protos.h: Update.
1297 2005-01-13 Steven Bosscher <stevenb@suse.de>
1299 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
1300 (fix_stmt_v_may_defs): Likewise.
1302 2005-01-13 Richard Henderson <rth@redhat.com>
1304 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
1305 for converting the mask.
1307 2005-01-13 David Edelsohn <edelsohn@gnu.org>
1309 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
1312 2005-01-13 Richard Henderson <rth@redhat.com>
1314 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
1315 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
1316 IX86_BUILTIN_MOVDQ2Q): Remove.
1317 (IX86_BUILTIN_VEC_EXT_V4SI): New.
1318 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
1319 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
1321 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
1322 (sse2_pextrw): Fix immediate constraint.
1323 (sse2_loadq, sse2_loadq_rex64): Remove.
1324 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
1325 Use __builtin_ia32_vec_ext_<size>.
1326 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
1328 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
1330 * function.c (assign_parm_setup_block): Look inside original
1331 entry_parm when inspecting PARALLEL.
1333 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
1334 Joel Sherrill <joel@oarcorp.com>
1337 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
1338 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
1339 rtems_gxx_recursive_mutex_init.
1340 (__gthread_recursive_mutex_lock): New function.
1341 (__gthread_recursive_mutex_trylock): Likewise.
1342 (__gthread_recursive_mutex_unlock): Likewise.
1344 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
1346 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
1348 2005-01-13 Jan Beulich <jbeulich@novell.com>
1350 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
1351 (ashrti3_internal): Indicate output is early clobber. Generate result
1352 into output rather than first input. Use move for low word of output
1353 if shift count is exactly 64.
1354 (lshrti3_internal): Likewise.
1356 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
1360 * reload1.c (delete_output_reload): Don't delete an output reload
1361 if the pseudo lives longer than a single basic block. Adjust and
1362 improve wording of comment.
1364 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
1367 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
1368 of TYPE_SIZE where we mean the number of bytes.
1370 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
1372 * function.c (assign_parm_setup_block): Relax condition on
1373 multi-register optimization.
1375 2005-01-12 Nick Clifton <nickc@redhat.com>
1377 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
1378 an "in_delay_slot" attribute of "no" to prevent them being used in
1379 delay slots. This is forbidden because they might generate
1382 2005-01-12 Alan Modra <amodra@bigpond.net.au>
1385 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
1388 2005-01-12 Nick Clifton <nickc@redhat.com>
1390 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
1391 passed on to GAS) but do not define, as GAS no longer supports or
1392 needs the -m2000 option.
1394 2005-01-11 Roger Sayle <roger@eyesopen.com>
1396 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
1397 type as args of unprototyped builtins aren't automatically promoted.
1398 (fold_builtin_abs): Likewise.
1400 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1402 * Makefile.in: Set a `build-warn' variable.
1404 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
1406 PR tree-optimization/17949
1407 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
1408 (find_interesting_uses_address): Use it.
1410 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
1412 * regrename.c (kill_value): Handle subreg's that won't simplify.
1414 2005-01-11 Richard Henderson <rth@redhat.com>
1417 * config/i386/i386.h (enum ix86_builtins): Move ...
1418 * config/i386/i386.c: ... here.
1419 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
1420 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
1421 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
1422 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
1423 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
1424 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
1425 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
1426 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
1427 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
1428 IX86_BUILTIN_LOADDDUP): Remove.
1429 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
1430 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
1431 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
1432 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
1433 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
1434 (ix86_init_builtins): Make static.
1435 (ix86_init_mmx_sse_builtins): Update for changed builtins.
1436 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
1437 if all the modes match. Otherwise, fake it.
1438 (get_element_number, ix86_expand_vec_init_builtin,
1439 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
1440 (ix86_expand_builtin): Make static. Update for changed builtins.
1441 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
1442 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
1443 (ix86_expand_vector_init_duplicate): New.
1444 (ix86_expand_vector_init_low_nonzero): New.
1445 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
1446 Split out from ix86_expand_vector_init; handle integer modes.
1447 (ix86_expand_vector_init): Use them.
1448 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
1449 * config/i386/i386-protos.h: Update.
1450 * config/i386/predicates.md (reg_or_0_operand): New.
1451 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
1452 (movv2sf_internal): Likewise. And a splitter to match them all.
1453 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
1454 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
1455 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
1456 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
1457 (mmx_pinsrw): Fix operand ordering.
1458 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
1459 rather than sse_loadss expander.
1460 (movv2df splitter): Similarly.
1461 (sse_loadss, sse_loadlss): Remove.
1462 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
1463 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
1464 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
1465 (sse3_movddup): Rename with '*'.
1466 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
1467 (sse2_loadsd): Remove.
1468 (vec_dupv2df_sse3): Rename from sse3_loadddup.
1469 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
1470 (sse2_pinsrw): Fix argument ordering.
1471 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
1472 (sse2_stored): Remove 'r' destination.
1473 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
1474 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
1475 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
1476 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
1477 vec_extractv16qi, vec_initv16qi): New.
1479 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
1480 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
1481 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
1483 (_mm_load_pd, _mm_store_pd): Use plain dereference.
1484 (_mm_load_si128, _mm_store_si128): Likewise.
1485 (_mm_load1_pd): Use _mm_set1_pd.
1486 (_mm_load_sd): Use _mm_set_sd.
1487 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
1488 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
1489 (_mm_set_epi64): Use _mm_set_epi64x.
1490 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
1491 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
1492 _mm_setr_epi8): Use _mm_set_foo form.
1493 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
1494 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
1495 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
1496 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
1497 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
1498 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
1499 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
1500 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
1501 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
1502 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
1503 (_mm_movedup_pd): Use _mm_shuffle_pd.
1504 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
1505 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
1506 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
1507 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
1509 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
1510 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
1511 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
1512 (_mm_store_ps): Use raw dereference.
1513 (_mm_store1_ps): Use _mm_storeu_ps.
1514 (_mm_storer_ps): Use _mm_store_ps.
1515 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
1516 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
1518 2005-01-11 Stan Shebs <shebs@apple.com>
1520 * config/rs6000/rs6000.c (machopic_output_stub): Issue
1521 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
1523 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
1526 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
1527 Skip all DECLs except for FIELD_DECLs.
1529 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
1531 * config/s390/s390.c (override_options): Return error if
1532 -mbackchain, -mpacked-stack and -mhard-float are used together.
1533 (s390_va_start): Remove the backchain && packed-stack special case.
1534 (s390_gimplify_va_arg): Likewise.
1535 * doc/invoke.texi: Remove the ABI incompatibility note.
1537 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
1539 * config/s390/s390.c (struct s390_frame_layout): Remove
1541 (s390_frame_info, s390_emit_prologue): Replace occurrences of
1542 save_backchain_p with TARGET_BACKCHAIN.
1544 2005-01-11 Alan Modra <amodra@bigpond.net.au>
1547 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
1548 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
1549 * function.c (locate_and_pad_parm): Set new field.
1550 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
1551 Tweak where_pad test to include "none". Always set mem align for
1553 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
1554 not sufficient for type.
1555 (assign_parm_setup_block): If stack_parm is zero on entry, always
1556 make a new stack local. Block move old stack parm if necessary
1557 to new aligned stack local.
1558 (assign_parm_setup_stack): Use a block move to handle
1559 potentially misaligned entry_parm.
1560 (assign_parms_unsplit_complex): Specify required alignment when
1561 creating stack local.
1562 * calls.c (compute_argument_addresses): Override alignment of stack
1563 arg calculated from its type with the alignment given by
1564 FUNCTION_ARG_BOUNDARY.
1565 (store_one_arg): Likewise.
1567 2005-01-11 Jan Beulich <jbeulich@novell.com>
1569 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
1570 Change respective itanium_class attribute to ialu.
1571 (shladdp4_internal): New.
1572 * config/ia64/predicates.md (shladd_log2_operand): New.
1574 2005-01-11 Richard Henderson <rth@redhat.com>
1576 * expr.c (store_constructor): Use rtvec_alloc instead of
1577 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
1579 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
1581 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
1583 2005-01-11 Alan Modra <amodra@bigpond.net.au>
1585 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
1586 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
1587 (TARGET_RWRELOC): Define.
1588 (ia64_rwreloc_section_type_flags): Delete.
1589 (ia64_section_type_flags): New function.
1590 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
1591 (TARGET_RWRELOC): Define.
1593 2005-01-10 David Mosberger <davidm@hpl.hp.com>
1596 * config/ia64/ia64.c (process_set): For alloc insn, only call
1597 process_epilogue is !frame_pointer_needed.
1599 2005-01-10 Roger Sayle <roger@eyesopen.com>
1602 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
1603 unary operator and can't be treated as a binary/comparison operator.
1605 2005-01-10 Richard Henderson <rth@redhat.com>
1607 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
1608 (ix86_return_in_memory): Likewise.
1609 (function_arg_advance): Likewise. Mirror structure in function_arg
1610 for choosing register to advance.
1612 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
1614 * tree-vectorizer.c, tree.def: Fix comment typos.
1616 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
1618 * config/s390/s390.c (struct processor_costs): 4 new fields:
1619 ddbr, ddr, debr, der.
1620 (s390_rtx_costs): More precise handling of divide instructions.
1622 2005-01-10 David Edelsohn <edelsohn@gnu.org>
1624 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
1626 2005-01-10 Jan Beulich <jbeulich@novell.com>
1628 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
1629 names resulting from -ffunction-sections/-fdata-sections and linkonce
1631 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
1632 based on the section name. Rearrange the section name comparison logic
1633 slightly so that each section name is compared against at most once.
1635 2005-01-10 Ben Elliston <bje@au.ibm.com>
1637 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
1638 targets for which -fPIC reduces limitations on the GOT size.
1640 2005-01-09 Falk Hueffner <falk@debian.org>
1642 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
1645 2005-01-09 David Edelsohn <edelsohn@gnu.org>
1648 * collect2.c (main): Set aixrtl_flag for -brtl option.
1649 (resolve_lib_name): Search for .so file extension before .a
1652 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
1654 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
1655 (vect_analyze_data_refs_alignment): Add dump prints.
1657 2005-01-09 Ira Rosen <irar@il.ibm.com>
1659 * tree-vectorizer.c (vect_analyze_offset_expr): Use
1660 expr_invariant_in_loop_p.
1661 Initialize outputs first thing in the function.
1662 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
1663 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
1664 (vect_analyze_pointer_ref_access): Check that the initial condition of
1665 the access function is loop invariant.
1667 2005-01-09 Richard Henderson <rth@redhat.com>
1669 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
1670 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
1671 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
1672 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
1673 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
1674 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
1675 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
1676 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
1677 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
1678 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
1679 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
1680 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
1681 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
1682 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
1683 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
1684 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
1685 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
1686 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
1687 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
1688 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
1689 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
1690 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
1691 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
1693 (mmx_clrdi, pavgusb): Remove.
1694 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
1695 with leading sse_ prefix.
1696 * config/i386/sse.md: Receive them.
1697 * config/i386/mmx.md: New file.
1698 (MMXMODE12, MMXMODE24, mmxvecsize): New.
1699 (subrv2sf3): Turn into expander for normal subtraction.
1700 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
1701 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
1702 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
1703 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
1704 commutative; use ix86_binary_operator_ok.
1705 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
1706 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
1707 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
1708 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
1709 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
1710 Macroize from existing patterns; use ix86_binary_operator_ok.
1711 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
1712 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
1713 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
1714 with vec_select+vec_concat.
1715 (mmx_pshufw, mmx_pshufw_1): Likewise.
1716 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
1717 (mmx_uavgv4hi3): Model correctly.
1718 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
1719 _mm_xor_si64): Remove casts.
1721 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
1723 PR tree-optimization/19224
1724 * tree-scalar-evolution.c (get_instantiated_value,
1725 set_instantiated_value): New functions.
1726 (instantiate_parameters_1): Cache the results.
1727 (instantiate_parameters, resolve_mixers): Initialize and free
1730 2005-01-08 David Edelsohn <edelsohn@gnu.org>
1732 * config/i386/i386.md (addhi_4): Correct reference in comment.
1735 2005-01-08 Richard Henderson <rth@redhat.com>
1737 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
1738 (_mm_cvtsi128_si64x): Likewise.
1739 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
1740 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
1741 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
1742 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
1743 * config/i386/i386-protos.h: Update.
1744 * config/i386/i386.c (print_operand): Add 'H'.
1745 (ix86_fixup_binary_operands): Split out from ...
1746 (ix86_expand_binary_operator): ... here.
1747 (ix86_fixup_binary_operands_no_copy): New.
1748 (ix86_expand_fp_absneg_operator): Handle vector mode results.
1749 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
1750 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
1751 (safe_vector_operand): Use CONST0_RTX.
1752 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
1753 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
1754 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
1755 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
1757 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
1758 (IX86_BUILTIN_CMPNGESS): New.
1759 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
1760 (attr type): Add sselog1.
1761 (attr unit, attr memory): Handle it.
1762 (movti, movti_internal, movti_rex64): Move near other integer moves.
1763 (movtf, movtf_internal): Move near other fp moves.
1764 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
1765 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
1766 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
1767 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
1768 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
1769 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
1770 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
1771 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
1772 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
1773 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
1774 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
1775 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
1776 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
1777 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
1778 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
1779 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
1780 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
1781 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
1782 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
1783 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
1784 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
1785 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
1786 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
1787 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
1788 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
1789 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
1790 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
1791 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
1792 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
1793 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
1794 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
1795 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
1796 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
1797 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
1798 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
1799 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
1800 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
1801 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
1802 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
1803 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
1804 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
1805 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
1806 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
1807 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
1809 (SSEPUSH, push<SSEPUSH>): Remove.
1810 (MMXPUSH, push<MMXPUSH>): Remove.
1811 (sse_movaps, sse_movaps_1, sse_movups): Remove.
1812 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
1813 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
1814 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
1815 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
1816 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
1817 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
1818 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
1819 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
1820 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
1821 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
1822 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
1823 (ppro_sse_log_V4SF): Handle sselog1.
1824 * config/i386/predicates.md (const_0_to_1_operand): New.
1825 (const_0_to_255_mul_8_operand): New.
1826 (const_1_to_31_operand): Rename from const_int_1_31_operand.
1827 (const_2_to_3_operand, const_4_to_7_operand): New.
1828 * config/i386/sse.md: New file.
1829 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
1830 (sse_movups): Rename from sse_movups_1.
1831 (sse_loadlss): Rename from sse_loadss_1.
1832 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
1834 (negv4sf2): Use ix86_expand_fp_absneg_operator.
1835 (absv4sf2, negv2df, absv2df): New.
1836 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
1837 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
1838 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
1839 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
1840 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
1841 (sse3_addsubv4sf3): Model correctly.
1842 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
1843 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
1844 (sse_movhlps): Model with vec_select+vec_concat.
1845 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
1846 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
1847 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
1848 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
1849 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
1850 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
1851 sse2_pshufhw_1): Likewise.
1852 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
1853 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
1854 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
1855 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
1856 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
1857 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
1858 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
1859 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
1860 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
1861 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
1862 sminv8hi3): Mark commutative
1863 operands. Use ix86_binary_operator_ok.
1864 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
1865 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
1866 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
1867 sse2_punpcklqdq): Allow operand2 in memory.
1868 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
1869 sse2_movsd): Add memory alternatives.
1870 (sse_storelps): Turn expander into an insn; split after reload.
1871 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
1872 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
1874 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
1876 * configure.ac (DWARF-2 debug_line): Use objdump.
1877 * configure: Regenerate.
1879 2005-01-08 Jeff Law <law@redhat.com>
1880 Diego Novillo <dnovillo@redhat.com>
1882 PR tree-optimization/18241
1883 * tree-nrv.c (tree_nrv): Ignore volatile return values.
1884 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
1885 statements with volatile operands.
1886 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
1887 operands after marking a statement with has_volatile_ops.
1889 2005-01-08 Roger Sayle <roger@eyesopen.com>
1891 * tree.c (int_fits_type_p): Always honor integer constant
1892 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
1894 2005-01-08 Roger Sayle <roger@eyesopen.com>
1896 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
1897 the latent bug in rtl_delete_block has been resolved.
1899 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
1901 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
1902 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
1904 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
1906 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
1907 library as @multilib_dir@/$(SHLIB_SONAME).
1909 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
1911 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
1912 * configure: Regenerate.
1914 2005-01-07 Jakub Jelinek <jakub@redhat.com>
1916 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
1919 2005-01-07 David Edelsohn <edelsohn@gnu.org>
1922 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
1923 Convert non-word aligned offset address using ld/std into
1926 2005-01-07 Richard Henderson <rth@redhat.com>
1928 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
1930 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
1932 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
1934 2005-01-07 Jakub Jelinek <jakub@redhat.com>
1936 PR tree-optimization/19283
1937 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
1938 is not shorter than the original type.
1940 PR rtl-optimization/19012
1941 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
1943 PR rtl-optimization/18861
1944 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
1945 (FULL_STATE): Define.
1946 (mark_tablejump_edge): New function.
1947 (purge_dead_tablejump_edges): New function.
1948 (find_bb_boundaries): Use it.
1950 PR tree-optimization/18828
1951 * builtins.c (expand_builtin_next_arg): Remove argument and all
1952 the argument checking.
1953 (expand_builtin): Adjust caller.
1954 (expand_builtin_va_start): Likewise. Remove error for too many
1956 (fold_builtin_next_arg): Issue error for too many arguments.
1957 After checking arguments, replace them with magic arguments that
1958 prevent further checking of the args.
1960 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1962 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
1964 2005-01-06 Roger Sayle <roger@eyesopen.com>
1967 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
1968 and %{.S:...} (and their negative variants) to test whether the
1969 input file is assembler or pre-processed-assembler independent of
1970 the actual filename extension.
1972 2005-01-06 Roger Sayle <roger@eyesopen.com>
1974 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
1975 of sign or zero extended values.
1977 2005-01-06 Geoffrey Keating <geoffk@apple.com>
1979 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
1980 for IBM long double format correctly.
1982 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
1984 Fix PR tree-optimization/18792
1986 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
1987 to first_loop_depth, and use loop depth instead of loop number.
1988 (build_classic_dir_vector): Ditto.
1989 (compute_data_dependences_for_loop): Use depth, not loop number.
1990 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
1991 number. Pass in loops, instead of loop numbers.
1992 (gather_interchange_stats): Ditto.
1993 (linear_transform_loops): Ditto.
1995 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
1998 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
1999 biv_fits_mode_p, extension_within_bounds_p): New functions.
2000 (check_ext_dependent_givs): Use them.
2002 2005-01-06 Roger Sayle <roger@eyesopen.com>
2004 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
2005 more than one barrier, in which case we should delete them all.
2007 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
2009 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
2011 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
2013 Fix PR middle-end/19286
2015 * dwarf2out.c (gen_subprogram_die): If we've already tried to
2016 output this subprogram, simply ignore this attempt to do it again.
2017 (add_abstract_origin_attribute): Don't abort trying to add the abstract
2018 origin attribute if it's not possible.
2019 (gen_block_die): Don't ignore subblocks of "unused" blocks.
2020 (decls_for_scope): Ditto.
2021 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
2022 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
2024 2005-01-05 Richard Henderson <rth@redhat.com>
2027 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
2028 (bdesc_2arg): Use it.
2029 (ix86_expand_binop_builtin): Force operands into registers
2031 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
2032 ix86_expand_sse_compare, ix86_expand_sse_comi,
2033 ix86_expand_builtin): Likewise.
2035 2005-01-05 Richard Henderson <rth@redhat.com>
2037 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
2039 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
2040 (mulv4hi3): Set itanium_class mmmul.
2041 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
2042 (fpack): Rename from fpack_sfsf.
2044 2005-01-05 Richard Henderson <rth@redhat.com>
2047 * reload1.c (do_input_reload): Restrict the optimization deleteing
2048 a previous output reload to RELOAD_FOR_INPUT.
2050 2005-01-05 Steven Bosscher <stevenb@suse.de>
2052 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
2053 comment that we fall through after case.
2054 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
2055 (distribute_links): Likewise.
2056 * cse.c (cse_insn): Likewise.
2057 * cselib.c (cselib_invalidate_mem): Likewise.
2058 * df.c: Update comments at the top of the file.
2059 (read_modify_subreg_p): Update comments here too.
2060 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
2061 * flow.c (mark_set_1): Likewise.
2062 (mark_used_regs): Likewise.
2063 * gcse.c (mems_conflict_for_gcse_p): Likewise.
2064 (canon_list_insert): Likewise.
2065 (mark_set): Likewise.
2066 (try_replace_reg): Likewise.
2067 (store_killed_in_insn): Likewise.
2068 * loop.c (count_one_set): Likewise.
2069 (basic_induction_var): Likewise.
2070 * postreload-gcse.c (find_mem_conflicts): Likewise.
2071 * postreload.c (reload_combine_note_store): Likewise.
2072 (move2add_note_store): Likewise.
2073 * reload.c (find_equiv_reg): Likewise.
2074 (mark_referenced_resources): Likewise.
2075 * rtlanal.c (set_noop_p): Likewise.
2076 (note_stores): Likewise.
2077 (note_uses): Likewise.
2078 * sched-deps.c (sched_analyze_1): Likewise.
2079 * sched-rgn.c (check_live_1): Likewise.
2080 (update_live_1): Likewise.
2081 * config/i860/i860.c: Likewise.
2083 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
2085 (ZERO_EXTRACT): Mention that this one can be an lvalue.
2087 * doc/rtl.texi: Update documentation for bit-fields and SET.
2089 2005-01-05 Roger Sayle <roger@eyesopen.com>
2091 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
2093 2005-01-05 Jan Hubicka <jh@suse.cz>
2094 Richard Henderson <rth@redhat.com>
2097 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
2100 2005-01-05 Richard Henderson <rth@redhat.com>
2103 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
2104 (sse_shufps): Change operand 3 to const_int_operand.
2105 (sse2_storelps): Fix typo in template.
2106 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
2107 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
2108 (ix86_expand_builtin): Likewise.
2110 2005-01-05 Julian Brown <julian@codesourcery.com>
2112 * config/arm/arm.c (arm_return_in_memory): Treat complex types
2113 as aggregates for AAPCS ABIs.
2115 2005-01-05 Stan Shebs <shebs@apple.com>
2117 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
2118 Mach-O getters if ppc64.
2119 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
2121 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
2122 signed four-byte field for global code case.
2123 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
2125 2005-01-05 Roger Sayle <roger@eyesopen.com>
2128 * c-common.c: Include real.h.
2129 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
2130 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
2131 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
2132 * Makefile.in (c-common.o): Update dependencies.
2134 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
2136 * c-parse.in (asm_string): Add trailing semicolon.
2138 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
2140 * c-parse.in (asm_string): New. Don't allow wide strings in
2142 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
2143 asm_string instead of STRING.
2145 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
2147 * c-typeck.c (constructor_no_implicit): Remove.
2148 (set_designator, process_init_element): Don't check
2149 constructor_no_implicit.
2151 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
2152 Kaz Kojima <kkojima@gcc.gnu.org>
2155 * lcm.c (create_pre_exit): New.
2156 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
2157 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
2159 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
2161 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
2162 -msoft-float in target_float_switch.
2163 * arm.c (arm_override_options): Fix processing of target_float_switch.
2165 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
2167 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
2170 2004-01-05 Uros Bizjak <uros@kss-loka.si>
2172 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
2173 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
2175 2005-01-04 Roger Sayle <roger@eyesopen.com>
2177 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
2178 if deleting the then-block allows the test-block to fallthru to the
2181 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
2184 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
2185 after the full declaration if the are in two different TUs.
2187 2005-01-04 Richard Henderson <rth@redhat.com>
2190 * tree-sra.c (generate_one_element_init): Just
2191 call gimplify_and_add.
2192 (generate_element_init): Record the
2193 new referenced variables and mark them for renaming
2194 and split out to ...
2195 (generate_element_init_1): This.
2196 (scalarize_init): Don't call push_gimplify_context/
2197 pop_gimplify_context.
2199 2005-01-04 Geoffrey Keating <geoffk@apple.com>
2201 * toplev.c (get_src_pwd): Handle failure of getpwd().
2203 2005-01-04 Roger Sayle <roger@eyesopen.com>
2205 * fold-const.c (fold_single_bit_test): Delete unreachable handling
2207 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
2208 all the cases handled by it are inverted by invert_truthvalue.
2210 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
2212 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
2213 (save_register_windowsi): Likewise.
2215 2005-01-04 Richard Henderson <rth@redhat.com>
2217 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
2219 2005-01-04 Uros Bizjak <uros@kss-loka.si>
2222 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
2223 any simplifications of vector mode comparison operators.
2224 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
2226 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
2227 Devang Patel <dpatel@apple.com>
2229 PR tree-optimization/18308
2230 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
2231 the operands before creating a new expression.
2232 * dojump.c (do_jump): Make drop_through_label available
2233 for all cases. Add expansion of COND_EXPR.
2235 2005-01-04 Ira Rosen <irar@il.ibm.com>
2237 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
2238 INTEGER_CST instead of TREE_CONSTANT.
2239 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
2240 instead of TREE_CONSTANT.
2241 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
2242 instead of TREE_CONSTANT.
2244 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
2246 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
2248 2005-01-04 Uros Bizjak <uros@kss-loka.si>
2251 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
2252 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
2254 2005-01-03 Richard Henderson <rth@redhat.com>
2256 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
2257 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
2258 fold_convert_const_real_from_real): Split out from ...
2259 (fold_convert_const): ... here.
2261 2005-01-03 Richard Henderson <rth@redhat.com>
2264 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
2265 (mov<MMXMODEI>_internal): Likewise.
2266 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
2267 Add V2SF case; use it for SSE1; don't use TI.
2268 (movdf_integer): Likewise.
2269 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
2271 2005-01-03 Ira Rosen <irar@il.ibm.com>
2273 * tree-vectorizer.c (vect_strip_conversions): New function.
2274 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
2275 check for binary class.
2277 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
2281 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
2283 * gimple-low.c (mark_blocks_with_used_vars): New function.
2284 (mark_blocks_with_used_subblocks): Ditto.
2285 (mark_used_blocks): Ditto.
2286 (pass_mark_used_blocks): New pass.
2287 * tree-inline.c: Include debug.h.
2288 (expand_call_inline): Call outlining_inline_function here.
2289 * tree-optimize.c (init_tree_optimization_passes): Add
2290 pass_mark_used_blocks.
2291 * tree-pass.h (pass_mark_used_blocks): New.
2292 * Makefile.in (tree-inline.o): Add debug.h dependency.
2294 2005-01-03 Geoffrey Keating <geoffk@apple.com>
2296 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
2297 VAR_DECLs to have weak_import attribute.
2299 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
2302 2004-01-03 Steven Bosscher <stevenb@suse.de>
2304 * passes.c (rest_of_compilation): Don't run regmove if only
2305 flag_expensive_optimizations, require flag_regmove instead.
2307 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
2309 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
2311 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
2312 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
2313 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
2314 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
2316 2005-01-03 Richard Henderson <rth@redhat.com>
2318 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
2319 (ia64_const_ok_for_letter_p): New.
2320 (ia64_const_double_ok_for_letter_p): New.
2321 (ia64_extra_constraint): New.
2322 (ia64_expand_vecint_compare): New.
2323 (ia64_expand_vcondu_v2si): New.
2324 (ia64_expand_vecint_cmov): New.
2325 (ia64_expand_vecint_minmax): New.
2326 (ia64_print_operand): Add 'v'.
2327 (ia64_preferred_reload_class): New.
2328 (ia64_vector_mode_supported_p): New.
2329 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
2330 (PREFERRED_RELOAD_CLASS): Move to function.
2331 (CONST_OK_FOR_LETTER_P): Move to function.
2332 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
2333 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
2334 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
2335 (EXTRA_CONSTRAINT): Move to function.
2336 * config/ia64/ia64.md: Include vect.md.
2337 (itanium_class): Add mmalua.
2339 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
2341 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
2342 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
2344 (const_int_2bit_operand): New.
2345 (fr_reg_or_0_operand): New.
2346 * config/ia64/ia64-modes.def: Add vector modes.
2347 * config/ia64/ia64-protos.h: Update.
2348 * config/ia64/vect.md: New file.
2350 2005-01-03 Richard Henderson <rth@redhat.com>
2352 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
2354 2005-01-03 Uros Bizjak <uros@kss-loka.si>
2357 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
2359 (log1pdf2): Change mode of operands[0,1] to DFmode.
2361 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
2363 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
2364 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
2365 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
2366 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
2367 SPARC_RELAXED_ORDERING.
2369 2005-01-03 Richard Henderson <rth@redhat.com>
2370 Uros Bizjak <uros@kss-loka.si>
2373 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
2374 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
2375 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
2377 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
2378 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
2379 const_pow2_1_to_128_operand): New predicates.
2381 2005-01-02 Greg McGary <greg@mcgary.org>
2383 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
2385 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
2386 object. [ptr_type]: Remove unused variable.
2387 Remember nearest addressable array-element or record-component
2388 when checking bitfield components. Tolerate empty BIND_EXPR.
2390 2005-01-01 Richard Henderson <rth@redhat.com>
2392 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
2393 that don't narrow the value. Fail for other conversions.
2395 2005-01-01 Richard Henderson <rth@redhat.com>
2398 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
2399 * c-lang.c (finish_file): Don't do it here.
2400 * objc/objc-act.c (objc_finish_file): Likewise.
2402 * cgraph.c (decl_assembler_name_equal): New.
2403 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
2404 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
2405 * cgraph.h (struct cgraph_varpool_node): Add next.
2406 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
2407 * varasm.c (assemble_alias): Mark the target as needed.
2409 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
2412 * function.c (get_arg_pointer_save_area): Use entry_of_function
2413 instead of get_insns.
2415 2005-01-01 Roger Sayle <roger@eyesopen.com>
2416 Andrew Pinski <pinskia@physics.uc.edu>
2417 James E. Wilson <wilson@specifixinc.com>
2419 PR rtl-optimization/12092
2420 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
2423 2005-01-01 Roger Sayle <roger@eyesopen.com>
2424 Olivier Hainque <hainque@act-europe.fr>
2426 * tree.c (int_fits_type_p): A narrower type always fits in a
2427 wider one, except for negative values into unsigned types.
2429 2005-01-01 Roger Sayle <roger@eyesopen.com>
2431 * tree.c (int_fits_type_p): Compare the result of force_fit_type
2432 with the original constant rather than require TREE_OVERFLOW.
2434 2005-01-01 Steven Bosscher <stevenb@suse.de>
2437 * c-decl.c (finish_function): If compiling C99, annotate the
2438 compiler generated return with the current file name and line 0.
2439 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
2440 the source line is greater than 0.
2441 (remove_bb): Likewise.
2443 See ChangeLog.12 for earlier changes.