1 2005-01-26 Richard Henderson <rth@redhat.com>
4 * c-decl.c (finish_struct): Set DECL_MODE after resetting a
6 * expmed.c (store_fixed_bit_field): Create a paradoxical subreg
7 if we don't need the bits above those present in the current mode.
8 * expr.c (store_field): Strip conversions to odd-bit-sized types
9 if the destination field width matches.
11 2005-01-26 Richard Sandiford <rsandifo@redhat.com>
13 * config/mips/iris6.h (ENDFILE_SPEC): Don't link in irix-csr.o if
14 either -ffast-math or -funsafe-math-optimizations is in use.
16 2005-01-26 Marek Michalkiewicz <marekm@amelek.gda.pl>
20 * config/avr/avr.c (notice_update_cc): Only set condition code for
21 ashrqi3 if shift count > 0.
22 (out_shift_with_cnt): Handle shift count <= 0 as a no-op.
23 (ashlqi3_out, ashlhi3_out, ashlsi3_out, ashrqi3_out, ashrhi3_out,
24 ashrsi3_out, lshrqi3_out, lshrhi3_out, lshrsi3_out): Handle shift
25 count <= 0 as a no-op, and shift count >= width by copying zero
26 or sign bit to all bits of the result.
27 * config/avr/avr.md (all shifts): Add alternatives for zero shift
28 count, with attribute "length" set to 0 and "cc" set to "none".
30 2005-01-26 Aldy Hernandez <aldyh@redhat.com>
32 * doc/invoke.texi: Document -mTLS.
34 * testsuite/gcc.target/frv/all-tls-global-dynamic.c: New.
35 * testsuite/gcc.target/frv/all-tls-initial-exec.c: New.
36 * testsuite/gcc.target/frv/all-tls-initial-exec-pic.c: New.
37 * testsuite/gcc.target/frv/all-tls-local-dynamic.c: New.
38 * testsuite/gcc.target/frv/all-tls-local-dynamic-plt.c: New.
39 * testsuite/gcc.target/frv/all-tls-local-dynamic-plt-pic.c: New.
40 * testsuite/gcc.target/frv/all-tls-local-exec.c: New.
41 * testsuite/gcc.target/frv/all-tls-local-exec-TLS.c: New.
43 * configure.ac: Check for a TLS capable gas.
45 * configure: Regenerate.
47 2005-01-26 Richard Henderson <rth@redhat.com>
50 * combine.c (make_field_assignment): Simplify store to zero_extract
51 from a source with an overlapping mask.
53 2005-01-26 Aldy Hernandez <aldyh@redhat.com>
55 2004-11-11 Eric Christopher <echristo@redhat.com>
57 * config/frv/frv.md: Add fr400_integer automaton. Don't
58 allow TYPE_MUL, TYPE_MACC, or TYPE_CUT to issue in parallel
61 2005-01-26 Steven Bosscher <stevenb@suse.de>
64 * cfgbuild.c (make_edges): Do not clear or set
65 current_function_has_computed_jump.
66 * function.h (struct function): Remove the has_computed_jump field.
67 (current_function_has_computed_jump): Do not define.
68 * sched-rgn.c (is_cfg_nonregular): Return true if a basic block ends
69 in a computed jump. Ignore current_function_has_computed_jump.
71 2005-01-26 Richard Henderson <rth@redhat.com>
74 * expr.c (categorize_ctor_elements): New argument p_must_clear.
75 (categorize_ctor_elements_1): Likewise. Detect a union that isn't
77 (mostly_zeros_p): Update for new categorize_ctor_elements argument.
78 * gimplify.c (gimplify_init_constructor): Likewise. Only shove
79 objects into static storage if they have more than one non-zero value.
80 * tree.h (categorize_ctor_elements): Update decl.
82 2005-01-26 Steven Bosscher <stevenb@suse.de>
85 * tree.h (CALL_EXPR_TAILCALL): Add comment.
86 * calls.c (check_sibcall_argument_overlap_1): Revert the change
87 to this function from 2004-07-10.
88 * tree-tailcall.c (suitable_for_tail_opt_p): Do not consider the
89 the current function for tail call optimizations if the address
90 of one of it its arguments is taken.
92 2005-01-26 Kazu Hirata <kazu@cs.umass.edu>
94 * cse.c (fold_rtx): Call equiv_constant only when necessary.
96 2005-01-26 Bernd Schmidt <bernd.schmidt@analog.com>
98 * function.c (expand_function_end): If current_function_calls_alloca,
99 emit stack restore in a place that is reached when the function falls
102 2005-01-26 Alexandre Oliva <aoliva@redhat.com>
104 * Makefile.in (../$(build_subdir)/fixincludes/fixincl): Add dummy
105 rule to avoid matching inappropriate suffix rule.
107 2005-01-26 Jan Hubicka <jh@suse.cz>
109 PR tree-optimization/19241
110 * tree-inline.c (copy_body_r): Do not walk subtrees after substituting.
112 2005-01-26 Hans-Peter Nilsson <hp@axis.com>
114 * config/cris/cris.h (LIB_SPEC) <sim*>: Use -lsyslinux, not
115 -lsyssim. Wrap -lc -lsyssim in --start-group --end-group instead
117 (STARTFILE_SPEC): Add crti.o, before crtbegin.o.
118 <sim*>: Always use crt1.o, regardless of N in simN.
119 <!sim>: Drop support for -pg and -p variants; always use crt0.o.
120 (ENDFILE_SPEC): Add crtn.o, after crtend.o.
121 (CRT_CALL_STATIC_FUNCTION): Remove.
122 * config/cris/aout.h (STARTFILE_SPEC): <!melinux && sim*>: Always
123 use crt1.o, regardless of N in simN.
124 <!sim>: Drop support for -pg and -p variants; always use crt0.o.
125 (LIB_SPEC): Do not define; default to the one in config/cris/cris.h.
126 (CRIS_CPP_SUBTARGET_SPEC) <melinux && !nostdinc>: Remove special
127 case isystem setting.
128 (MAX_OFILE_ALIGNMENT): Correct to 32, not 16.
130 2005-01-25 Ian Lance Taylor <ian@airs.com>
132 * gimple-low.c (block_may_fallthru): Correct handling of
133 SWITCH_EXPR--may fall through if SWITCH_LABELS is NULL.
135 2005-01-26 Hans-Peter Nilsson <hp@axis.com>
137 * config/cris/cris.h (CPP_SPEC): Adjust the various mtune=
138 handlers to match indentation; nested under the mtune=* handler.
139 Similar for mcpu= and march=.
141 2005-01-25 Daniel Jacobowitz <dan@codesourcery.com>
143 * config/mips/linux.h (LIB_SPEC): Remove -rpath-link.
144 * config/mips/linux64.h (LIB_SPEC): Likewise.
145 * config/mn10300/linux.h (LIB_SPEC): Likewise.
146 * config/sh/linux.h (LIB_SPEC): Likewise.
148 2005-01-25 Aldy Hernandez <aldyh@redhat.com>
150 * config/frv/frv.c (frv_legitimize_tls_address): New.
151 (TARGET_HAVE_TLS): Define.
152 (FRV_SYMBOL_REF_TLS_P): Define.
153 (frv_override_options): Handle new register classes.
154 (frv_legitimate_address_p): Reject tls addresses.
155 (frv_legitimize_address): Handle TLS addresses.
156 (gen_inlined_tls_plt): New.
158 (frv_legitimize_tls_address): New.
159 (unspec_got_name): Add TLS entries.
160 (got12_operand): Add R_FRV_TLSMOFF12 case.
161 (frv_emit_move): Fixup TLS addresses.
162 (frv_emit_movsi): Legitimize TLS addresses.
164 * config/frv/frv.h (MASK_BIG_TLS): New.
165 (TARGET_BIG_TLS): New.
166 (HAVE_AS_TLS): Define.
167 (TARGET_SWITCHES): Add -mTLS and -mtls options.
168 (enum reg_class): Add GR8_REGS, GR9_REGS, GR89_REGS.
169 (REG_CLASS_NAMES): Same.
170 (REG_CLASS_CONTENTS): Same.
171 (CONSTRAINT_LEN): New.
172 (REG_CLASS_FROM_CONSTRAINT): New.
173 (PREDICATE_CODES): Add symbolic_operand.
175 * config/frv/frv.md (define_constants): Add UNSPEC_GETTLSOFF,
176 UNSPEC_TLS_LOAD_GOTTLSOFF12, UNSPEC_TLS_INDIRECT_CALL,
177 UNSPEC_TLS_TLSDESC_LDD, UNSPEC_TLS_TLSDESC_LDD_AUX,
178 UNSPEC_TLS_TLSOFF_LD, UNSPEC_TLS_LDDI, UNSPEC_TLSOFF_HILO,
179 R_FRV_GOTTLSOFF_HI, R_FRV_GOTTLSOFF_LO, R_FRV_TLSMOFFHI,
180 R_FRV_TLSMOFFLO, R_FRV_TLSMOFF12, R_FRV_TLSDESCHI,
181 R_FRV_TLSDESCLO, R_FRV_GOTTLSDESCHI, R_FRV_GOTTLSDESCLO, GR8_REG,
182 GR9_REG, GR14_REG, LRREG.
183 (type): Add load_or_call attribute.
184 ("load_or_call"): New reservation.
185 ("call_gettlsoff"): New.
186 ("tls_indirect_call"): New.
187 ("tls_load_gottlsoff12"): New.
188 ("tlsoff_hilo"): New.
189 ("tls_tlsdesc_ldd"): New.
190 ("tls_tlsoff_ld"): New.
193 * config/frv/frv-protos.h (symbolic_operand): Protoize.
195 2005-01-25 Alexandre Oliva <aoliva@redhat.com>
197 * config/frv/frv.h (ASM_OUTPUT_DWARF_DTPREL): Define.
198 * config/frv/frv-protos.h (frv_output_dwarf_dtprel): Declare.
199 * config/frv/frv.c (TLS_BIAS): Define.
200 (frv_output_dwarf_dtprel): New.
202 2005-01-26 Jakub Jelinek <jakub@redhat.com>
204 PR rtl-optimization/19579
205 * ifcvt.c (noce_try_cmove_arith): If emitting instructions to set up
206 both A and B, see if they don't clobber registers the other expr uses.
208 2005-01-25 J"orn Rennecke <joern.rennecke@st.com>
210 * real.c (do_add): Initialize signalling and canonical members.
212 * real.c (real_from_integer): Zero out destination.
214 2005-01-18 Jan Hubicka <jh@suse.cz>
216 PR tree-optimize/19337
217 * tree-inline.c (initialize_inlined_parameters): Copy
218 saved_static_chain_decl only.
220 2005-01-25 Richard Henderson <rth@redhat.com>
224 * config/i386/i386.c (x86_inter_unit_moves): Zero.
225 (ix86_preferred_reload_class): Rewrite fp-constant section, with
226 80387 enabled, to return a proper subclass. Return the subset that
227 overlaps with GENERAL_REGS for PLUS.
228 * config/i386/i386.md (movsi_1, movdi_2): Set type to "mmx" for pxor.
230 2005-01-25 Richard Henderson <rth@redhat.com>
233 * tree-complex.c (expand_complex_div_wide): Use the correct formulae.
235 2005-01-25 Roger Sayle <roger@eyesopen.com>
237 * expmed.c (choose_multiplier): Change interface to return the
238 multiplier lowpart as an rtx instead of an unsigned HOST_WIDE_INT.
239 (expand_divmod): Update calls to choose_multiplier.
241 2005-01-24 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
244 * expmed.c (extract_bit_field): Use adjust_address instead of
245 gen_lowpart when op0 is a MEM.
247 2005-01-24 Steven Bosscher <stevenb@suse.de>
249 * cfgexpand.c (expand_gimple_tailcall): Fix typo.
251 2005-01-24 Richard Sandiford <rsandifo@redhat.com>
253 * config/mips/irix-csr.c: New file.
254 * config/mips/t-iris6 (irix-csr.o): New rule to build it.
255 (EXTRA_MULTILIB_PARTS): Add irix-csr.o.
256 * config/mips/iris6.h (ENDFILE_SPEC): Include it in n32 and n64
259 2005-01-24 Eric Botcazou <ebotcazou@libertysurf.fr>
262 * config.gcc (sparc-*-elf*): Remove sol2.h, sparc/sol2.h and
263 sparc/elf.h, add sparc/sp-elf.h.
264 (sparc-*-rtems*): Likewise.
265 (sparclite-*-elf*): Remove sol2.h, sparc/sol2.h, sparc/elf.h and
266 tm-dwarf2.h, add sparc/sp-elf.h.
267 (sparc86x-*-elf): Likewise.
268 (sparc64-*-elf*): Remove sol2.h, sparc/sol2.h and tm-dwarf2.h.
269 * config/sparc/liteelf.h (TARGET_SUB_OS_CPP_BUILTINS): Rename into
270 TARGET_OS_CPP_BUILTINS.
271 * config/sparc/sp86x-elf (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
272 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
274 * config/sparc/openbsd64.h (NO_IMPLICIT_EXTERN_C): Undefine.
275 * config/sparc/sp64-elf.h (NO_IMPLICIT_EXTERN_C): New macro.
276 (SWITCH_TAKES_ARG): Likewise.
277 (LOCAL_LABEL_PREFIX): Likewise.
278 (ASM_GENERATE_INTERNAL_LABEL): Likewise.
279 (TARGET_N_FORMAT_TYPES): Delete.
280 (TARGET_FORMAT_TYPES): Likewise.
281 (ASM_DECLARE_FUNCTION_SIZE): Likewise.
282 * config/sparc/elf.h: Delete.
283 * config/sparc/sp-elf.h: New file.
285 2005-01-24 Kazu Hirata <kazu@cs.umass.edu>
287 * tree-flow-inline.h (phi_arg_from_edge): Remove.
288 * tree-flow.h: Remove the corresponding prototype.
289 * tree-ssa-operands.h, tree-ssa-threadupdate.c, tree-ssa.c,
290 tree-vectorizer.c: Use dest_idx instead of phi_arg_from_edge.
292 2005-01-24 Jakub Jelinek <jakub@redhat.com>
295 * config/rs6000/rs6000.c (rs6000_file_start): Create toc section
296 for AIX ABI or ELF -fPIC.
297 (rs6000_emit_load_toc_table): Don't create toc_section here.
298 (rs6000_xcoff_file_start): Nor here.
300 2005-01-24 J"orn Rennecke <joern.rennecke@st.com>
302 * sh.c (ra.h): Don't #include.
303 (hard_regs_intersect_p): New function, resurrected from ra.c.
305 * sh.c: Fix 1996 Copyright.
307 2005-01-24 Richard Henderson <rth@redhat.com>
308 Aldy Hernandez <aldyh@redhat.com>
310 * regrename.c (note_sets): Handle subregs.
312 2005-01-24 Joseph S. Myers <joseph@codesourcery.com>
315 * genconditions.c (write_header, write_conditions): Elide file if
318 2005-01-24 Kazu Hirata <kazu@cs.umass.edu>
320 * c-opts.c, expmed.c, gengtype.c, tree-complex.c,
321 config/arm/iwmmxt.md, config/avr/avr.c, config/s390/s390.h:
324 2005-01-24 Jakub Jelinek <jakub@redhat.com>
326 * builtins.c (get_pointer_alignment, gimplify_va_arg_expr,
327 expand_builtin_printf, expand_builtin_fprintf,
328 expand_builtin_sprintf): Use POINTER_TYPE_P instead TREE_CODE
329 checking against POINTER_TYPE.
330 (validate_arglist): Handle POINTER_TYPE code by checking
333 2005-01-24 Paolo Bonzini <bonzini@gnu.org>
335 * rtlanal.c (non_rtx_starting_operands, for_each_rtx_1,
337 (for_each_rtx): Call for_each_rtx_1.
338 * rtl.h (init_rtlanal): Declare.
339 * toplev.c (backend_init): Call init_rtlanal.
341 2005-01-24 Jakub Jelinek <jakub@redhat.com>
343 * flow.c (propagate_one_insn): Formatting.
346 * flow.c (libcall_dead_p): Be more conservative if unsure.
347 If there are any instructions between insn and call, see if they are
348 all dead before saying the libcall is dead.
350 2005-01-24 Uros Bizjak <uros@kss-loka.si>
352 * config/i386/i386.md (*extendsfdf2_sse): Use "nonimmediate_operand"
353 operand constraint for operand 0.
355 2005-01-23 Richard Henderson <rth@redhat.com>
358 * tree-complex.c (expand_complex_div_wide): Set EDGE_FALLTHRU on
359 edges leading to join block.
361 2005-01-23 Roger Sayle <roger@eyesopen.com>
363 * expmed.c (expand_mult_highpart): Make static. Change type of
364 constant multiplier argument from unsigned HOST_WIDE_INT to rtx.
365 (expand_divmod): Updates calls to expand_mult_highpart by using
366 gen_int_mode to pass a CONST_INT rtx.
367 * rtl.h (expand_mult_highpart): Remove prototype.
369 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
371 * flow.c (regset_bytes, regset_size): Remove.
373 * cfganal.c: Fix a reference to Harvey's paper.
375 2005-01-23 Steven Bosscher <stevenb@suse.de>
377 PR rtl-optimization/19464
378 * tree-optimize.c (init_tree_optimization_passes): Add one more
379 copyrename pass just before out-of-ssa.
381 2005-01-23 Joseph S. Myers <joseph@codesourcery.com>
384 * recog.c (recog_memoized): Don't define if GENERATOR_FILE.
385 * ggc-none.c (ggc_free): Define.
387 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
389 * cse.c (max_reg, max_insn_uid): Remove.
390 (cse_main): Don't access max_reg or max_insn_uid.
392 * c-common.c (explicit_flag_signed_bitfields,
393 lang_statement_code_p, lang_expand_function_end): Remove.
394 * c-common.h: Remove the corresponding declarations.
395 * c-opts.c (c_common_handle_option): Remove all write access
396 to explicit_flag_signed_bitfields.
398 * cgraph.c (cgraph_varpool_n_nodes): Remove.
399 (cgraph_varpool_node): Don't access cgraph_varpool_n_nodes.
400 * cgraph.h: Remove the corresponding declaration.
402 * gcse.c (null_pointer_info): Remove.
404 2005-01-23 Roger Sayle <roger@eyesopen.com>
405 Eric Botcazou <ebotcazou@libertysurf.fr>
407 * combine.c (simplify_logical): Only simplify logical expressions
408 of the form ior(and(x,y),z) by the inverse distributive law if the
409 result is cheaper than the original.
411 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
413 * alias.c, c-common.h, c-incpath.c, c-incpath.h, expr.c,
414 fold-const.c, gimplify.c, params.h, tree-data-ref.c,
415 tree-if-conv.c, tree-nested.c, tree-outof-ssa.c,
416 tree-ssa-dom.c, tree-vectorizer.c, tree.def, config/darwin.c,
417 config/freebsd-spec.h, config/arm/arm.h,
418 config/h8300/h8300.md, config/i386/i386.md,
419 config/i386/predicates.md, config/i386/sse.md,
420 config/ia64/ia64.c, config/ip2k/ip2k.c, config/s390/s390.c,
421 config/vax/vax.md: Fix comment typos. Follow spelling
424 2005-01-23 Steven Bosscher <stevenb@suse.de>
426 * gengtype.c (walk_type): Produce `not equal to' compares for loop
427 tests, instead of `less than'.
429 2005-01-23 Richard Sandiford <rsandifo@redhat.com>
431 * function.c (assign_parm_setup_block): When creating a new stack slot
432 for a parameter, get its alignment from the parameter's DECL_ALIGN
433 rather than the type's TYPE_ALIGN. Make sure that the parameter
434 is at least word aligned.
436 2005-01-22 Roger Sayle <roger@eyesopen.com>
439 * config/avr/avr.c (avr_hard_regno_mode_ok): Rewrite.
441 2005-01-22 Richard Henderson <rth@redhat.com>
444 * config/i386/i386.md (movsfcc_1_sse_max): Use nonimmediate_operand
445 in both compare operands.
446 (movdfcc_1_sse_max): Likewise.
447 (movsfcc_1_sse): Likewise. Add earlyclobber for scratch.
448 (movdfcc_1_sse): Likewise.
449 * config/i386/i386.c (ix86_split_sse_movcc): Emit copies into the
450 scratch register as needed.
452 2005-01-22 Richard Henderson <rth@redhat.com>
454 * config/i386/i386.md (smaxsf3): Fix mnemonic typo.
456 * config/i386/i386.c (ix86_prepare_fp_compare_args): Fix is_sse test.
458 2005-01-22 Richard Henderson <rth@redhat.com>
460 * genopinit.c (optabs): Use smin/smax for floating point too.
461 * doc/md.texi: Update to match. Clarify that floating point
462 results are undefined for +0/-0 and NaN.
463 * doc/rtl.texi: Likewise.
464 * rtl.def (SMIN, SMAX): Likewise
465 * tree.def (MIN_EXPR, MAX_EXPR): Likewise.
467 * config/alpha/alpha.md (smaxdf3, smindf3, smaxsf3, sminsf3): Add
468 leading 's' to the name.
469 * config/ia64/ia64.md (smaxsf3, sminsf3, smaxdf3, smindf3,
470 smaxxf3, sminxf3): Likewise.
471 * config/rs6000/rs6000.md (smaxdf3, smindf3, smaxsf3, sminsf3):
474 2005-01-22 Paul Brook <paul@codesourcery.com>
476 * config/arm/arm.md: Use "Uy" constraint for wcgr load/stores.
478 2005-01-21 Mark Dettinger <dettinge@de.ibm.com>
480 * config/s390/s390.c (struct processor_costs): New fields
481 dlgr, dlr, dr, dsgfr, dsgr.
482 (z900_cost, z990_cost): Values for new fields.
483 (s390_rtx_costs): New cases MEM und COMPARE in switch
484 statement. Modified handling of SIGN_EXTEND, ZERO_EXTEND,
485 DIV, MOD, UDIV, UMOD.
487 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
489 * config/s390/s390.md ("doloop_si64"): Reload input value directly
490 into the register being decremented.
491 ("doloop_si31", "doloop_di"): Likewise.
492 ("*doloop_si_long"): Adapt pattern.
494 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
496 * config/s390/s390.h (HARD_REGNO_NREGS): Fix computation for
498 (CLASS_MAX_NREGS): Likewise.
500 2005-01-21 Daniel Berlin <dberlin@dberlin.org>
502 * doc/tree-ssa.texi (Statement Operands): Add example for new
503 must-def macro. Note deprecation of old operands interface.
505 2005-01-22 Richard Sandiford <rsandifo@redhat.com>
507 PR tree-optimization/19484
508 * tree-cfg.c (remove_fallthru_edge): New function.
509 (cleanup_control_flow): Remove fallthru edges from calls that are
510 now known not to return.
512 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
514 Fix PR tree-optimization/19038
515 * tree-ssa-dom.c (cprop_operand): Don't replace loop invaeriant
516 copies with loop variant ones.
518 2005-01-22 Kazu Hirata <kazu@cs.umass.edu>
520 * cfganal.c, real.h, reorg.c, timevar.def, tree-ssa-ccp.c,
521 config/alpha/alpha-protos.h, config/alpha/alpha.h,
522 config/alpha/alpha.md, config/alpha/predicates.md,
523 config/sparc/freebsd.h, config/sparc/netbsd-elf.h,
524 config/sparc/sol2.h: Update copyright.
526 * tree-cfg.c (remove_forwarder_block_with_phi): Look at the
527 first label to see if it is a nonlocal label.
529 2005-01-22 David Edelsohn <edelsohn@gnu.org>
530 Andrew Pinski <pinskia@physics.uc.edu>
533 * config/rs6000/rs6000.c (rs6000_va_start): Saturate n_gpr at
534 maximum number of GPRs. Saturate n_fpr at maximum number of FPRs.
536 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
538 * config/rs6000/rtems.h: Update copyright.
540 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
543 * config/rs6000/rtems.h: Resurrect cpp_os_rtems_spec from gcc < 3.4.
544 (CPP_OS_RTEMS_SPEC): New (From gcc-3.3's config/rs6000/sys4.h).
545 (SUBSUBTARGET_EXTRA_SPECS): Use CPP_OS_RTEMS_SPEC.
547 2005-01-22 Volker Reichelt <reichelt@igpm.rwth-aachen.de>
550 * c-typeck.c (convert_arguments): Check for error_mark_node.
552 2005-01-21 Ian Lance Taylor <ian@airs.com>
554 PR tree-optimization/13000
555 * tree-inline.c: Include "tree-flow.h".
556 (expand_call_inline): If warn_return_type, warn if non-void inline
557 function falls through.
558 * tree-cfg.c (execute_warn_function_return): Don't warn about
559 control reaching end if TREE_NO_WARNING is set. Set
561 * gimple-low.c (block_may_fallthru): Don't assume that SWITCH_EXPR
563 * gimplify.c (shortcut_cond_expr): Don't emit a jump over the else
564 branch if we don't need one.
565 * c-typeck.c: Include "tree-flow.h"
566 (c_finish_bc_stmt): Don't add a goto if the current statement
567 list doesn't fall through to the current point.
569 2005-01-21 Roger Sayle <roger@eyesopen.com>
571 PR rtl-optimization/576
572 * real.c (real_arithmetic): Change return type from void to bool
573 to return an indication that the result may be inexact.
574 * real.h (real_arithmeric): Update prototype.
575 * fold-const.c (const_binop): Don't constant fold floating
576 point expressions when the user specifies -frounding-math and
577 the result may depend upon the run-time rounding mode.
578 (fold_convert_const_real_from_real): Clean-up.
579 (fold_initializer): Ignore flag_rounding_math for initializers.
580 * simplify-rtx.c (simplify_binary_operation): Likewise, don't
581 constant fold FP operations with flag_rounding_math if the
582 result may depend upon the run-time rounding mode.
584 2005-01-21 Tom Tromey <tromey@redhat.com>
586 * c-cppbuiltin.c (define__GNUC__): Correct assertion.
588 2005-01-21 Kazu Hirata <kazu@cs.umass.edu>
590 * tree-cfg.c: Fix comment typos.
592 2005-01-21 J"orn Rennecke <joern.rennecke@st.com>
594 * passes.c (rest_of_handle_flow2): Close / open the flow2 dump file
595 around the call to rest_of_handle_branch_target_load_optimize.
597 2005-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
599 * config.gcc (sparc-*-netbsdelf*): Include dbxelf.h
600 (sparc64-*-openbsd*): Likewise.
601 (sparclite-*-elf*): Include tm-dwarf2.h.
602 (sparc86x-*-elf*): Likewise.
603 (sparc64-*-elf*): Likewise.
604 * config/sparc/linux64.h (DWARF2_DEBUGGING_INFO): Delete.
605 (DBX_DEBUGGING_INFO): Likewise.
606 * config/sparc/liteelf.h (DWARF2_DEBUGGING_INFO): Likewise.
607 (PREFERRED_DEBUGGING_TYPE): Likewise.
608 * config/sparc/netbsd-elf.h (PREFERRED_DEBUGGING_TYPE): Likewise.
609 * config/sparc/sp64-elf.h (DBX_DEBUGGING_INFO): Likewise.
610 (PREFERRED_DEBUGGING_TYPE): Likewise.
611 * config/sparc/sp86x-elf.h (DWARF2_DEBUGGING_INFO): Likewise.
612 (PREFERRED_DEBUGGING_TYPE): Likewise.
613 * config/sparc/sparc.h (DBX_DEBUGGING_INFO): Likewise.
615 2005-01-21 Hans-Peter Nilsson <hp@bitrange.com>
618 * combine.c (combine_simplify_rtx): Revert change of 2004-12-31.
620 2005-01-20 Paul Brook <paul@codesourcery.com>
622 * doc/rtl.texi: Document value extension requirements for CONST_INT.
624 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
626 * tree-cfg.c (tree_verify_flow_info): Check that a nonlocal
627 label is first in a sequence of labels.
629 * tree-cfg.c (tree_verify_flow_info): Fix a typo.
631 2005-01-20 Janis Johnson <janis187@us.ibm.com>
632 Giovanni Bajo <giovannibajo@gcc.gnu.org>
634 * doc/sourcebuild.texi (Test Directives): New.
636 2005-01-20 Roger Sayle <roger@eyesopen.com>
638 * real.c (real_floor): Don't leave the result uninitialized when
639 mode is VOIDMode, but return the unrounded intermediate value.
640 (real_ceil): Likewise.
642 2005-01-20 Paul Brook <paul@codesourcery.com>
644 * config/arm/arm.md (insv): Use gen_int_mode.
646 2005-01-20 Eric Botcazou <ebotcazou@libertysurf.fr>
648 * config/sparc/freebsd.h (ASM_OUTPUT_INTERNAL_LABELREF): Delete.
649 * config/sparc/linux.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
650 * config/sparc/linux64.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
651 * config/sparc/netbsd-elf.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
652 * config/sparc/sol2.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
654 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
657 * dwarf2out.c (concat_loc_descriptor): We don't know we can use
658 fbreg, so we have to assume we can't.
659 (loc_descriptor_from_tree_1): Ditto.
660 (containing_function_has_frame_base): New function.
661 (add_location_or_const_value_attribute): Use it.
662 Also try to generate a frame_base from a single element location
665 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
667 PR tree-optimization/15349
668 * timevar.def (TV_TREE_MERGE_PHI): New.
669 * tree-cfg.c (tree_forwarder_block_p): Add a new argument
671 (remove_forwarder_block, cleanup_forwarder_blocks): Adjust the
672 calls to tree_forwarder_block_p.
673 (remove_forwarder_block_with_phi, merge_phi_nodes,
674 gate_merge_phi, pass_merge_phi): New.
675 * tree-optimize.c (init_tree_optimization_passes): Add
677 * tree-pass.h: Add an extern for pass_merge_phi;
679 2005-01-20 Richard Henderson <rth@redhat.com>
682 * config/i386/emmintrin.h (_mm_castpd_ps, _mm_castpd_si128): New.
683 (_mm_castps_pd, _mm_castps_si128): New.
684 (_mm_castsi128_ps, _mm_castsi128_pd): New.
686 2005-01-20 Richard Henderson <rth@redhat.com>
689 * config/i386/mmintrin.h (_mm_cvtsi32_si64): Use
690 __builtin_ia32_vec_init_v2si.
691 (_mm_cvtsi64_si32): Use __builtin_ia32_vec_ext_v2si.
692 * config/i386/i386.c (IX86_BUILTIN_VEC_EXT_V2SI): New.
693 (ix86_init_mmx_sse_builtins): Create it.
694 (ix86_expand_builtin): Expand it.
695 (ix86_expand_vector_set): Handle V2SFmode and V2SImode.
696 * config/i386/mmx.md (vec_extractv2sf_0, vec_extractv2sf_1): New.
697 (vec_extractv2si_0, vec_extractv2si_1): New.
699 2005-01-20 Richard Henderson <rth@redhat.com>
702 * config/i386/i386.c (ix86_expand_push): New.
703 * config/i386/mmx.md (push<MMXMODE>1): New.
704 * config/i386/sse.md (push<SSEMODE>1): New.
705 * config/i386/i386-protos.h: Update.
707 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
709 * params.def, params.h, predict.c, tree-ssa-loop-im.c: Update
712 2005-01-20 Richard Henderson <rth@redhat.com>
714 * rtl.def (CONST_VECTOR): Use RTX_CONST_OBJ.
715 * rtl.h (CONSTANT_P): Don't special case CONST_VECTOR.
717 2005-01-19 Richard Henderson <rth@redhat.com>
720 * config/i386/i386.c (ix86_expand_vector_move_misalign): Convert
721 to V4SFmode in SSE1 fallback load path.
723 2005-01-19 Richard Henderson <rth@redhat.com>
725 * config/i386/i386.c (ix86_expand_vector_init_one_var): Fix typo
726 in QImode expansion to ix86_expand_vector_set.
727 (ix86_expand_vector_init_general): Fix typo in V8HImode recursive call.
729 2005-01-19 Richard Henderson <rth@redhat.com>
732 * config/i386/i386.c (ix86_preferred_reload_class): Return a proper
733 subclass of the input class.
734 (ix86_secondary_memory_needed): Always true for cross-MMX classes.
735 Always true for cross-SSE1 classes. Rationalize conditionals.
736 * config/i386/i386.h (SSE_CLASS_P, MMX_CLASS_P): Use straight equality.
737 * config/i386/i386.md (movsi_1): Add MMX/SSE zeros. Fix alternatives
738 for SSE1. Don't check TARGET_INTER_UNIT_MOVES.
739 (movdi_2): Add MMX/SSE zeros.
740 (movdi_1_rex64): Likewise. Don't check TARGET_INTER_UNIT_MOVES.
741 (movsf_1): Don't check TARGET_INTER_UNIT_MOVES.
742 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
743 (movsi_1_nointernunit, movdi_1_rex64_nointerunit): Remove.
744 (movsf_1_nointerunit, zero_extendsidi2_32_1): Remove.
745 (zero_extendsidi2_rex64_1): Remove.
746 (MOV0 peephole): Check GENERAL_REG_P.
748 2005-01-19 Richard Henderson <rth@redhat.com>
751 * config/i386/i386.c (ix86_expand_vector_set): Fix third and fourth
753 (ix86_expand_vector_extract): Likewise.
755 2005-01-19 Richard Henderson <rth@redhat.com>
758 * expr.c (emit_move_change_mode): New argument force; use
759 simplify_gen_subreg if true.
760 (emit_move_via_alt_mode): Merge into ...
761 (emit_move_via_integer): ... here.
762 (emit_move_ccmode): Use emit_move_change_mode directly.
764 2005-01-19 Richard Henderson <rth@redhat.com>
767 * config/alpha/alpha.c (alpha_rtx_costs): Handle HIGH.
768 (alpha_preferred_reload_class): Handle CONST_VECTOR.
769 (alpha_emit_set_const_1): Add no_output parameter; don't emit
771 (alpha_emit_set_const): Likewise. Make static.
772 (alpha_emit_set_long_const): Make static.
773 (alpha_extract_integer): Split out from alpha_expand_mov.
774 (alpha_split_const_mov): Likewise.
775 (alpha_expand_mov): Use them. Handle CONST_VECTOR.
776 (alpha_legitimate_constant_p): New.
777 * config/alpha/alpha-protos.h: Update.
778 * config/alpha/alpha.h (REGISTER_MOVE_COST): Correct fp<->gp cost.
779 (LEGITIMATE_CONSTANT_P): Re-implement with a function.
780 * config/alpha/alpha.md (movsi): Add n alternative.
781 (movsi_nt_vms, movdi_er_nofix, movdi_er_fix, movdi_fix): Likewise.
782 (mov<VEC>_fix, mov<VEC>_nofix): Add i alternative.
783 (splitters for all of the above): Use alpha_split_const_mov.
784 * config/alpha/predicates.md (non_add_const_operand): New.
785 (non_zero_const_operand): New.
786 (input_operand): Use alpha_legitimate_constant_p after reload.
788 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
790 PR tree-optimization/19038
791 * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): New function.
792 (add_candidate): Add ivs with increment in latch only if
793 allow_ip_end_pos_p is true.
794 (determine_iv_cost): Use empty_block_p.
796 2005-01-19 Daniel Berlin <dberlin@dberlin.org>
798 * cfganal.c (compute_dominance_frontiers_1): Replace with new algorithm
799 (compute_dominance_frontiers): Ditto.
801 2005-01-19 Ralf Corsepius <ralf.corsepius@rtems.org>
804 * config/sh/t-rtems: New.
805 * config.gcc (sh-*-rtems*): Reflect having added config/sh/t-rtems.
807 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
809 * tree-ssa-ccp.c (maybe_fold_offset_to_component_ref): Always subtract
810 the offset of the selected field.
812 2005-01-19 Kazu Hirata <kazu@cs.umass.edu>
814 * tree-cfg.c (remove_forwarder_block): Fix the check to
815 prevent a nonlocal label from appearing in the middle of a
818 2005-01-19 Hans-Peter Nilsson <hp@axis.com>
820 PR rtl-optimization/19462
821 * reorg.c (find_end_label): Create return insn only if
822 current_function_epilogue_delay_list is empty.
824 2005-01-19 Paolo Bonzini <bonzini@gnu.org>
826 * doc/passes.texi: Remove paragraph mentioning new-ra.
828 2005-01-19 Jakub Jelinek <jakub@redhat.com>
830 PR rtl-optimization/15139
831 * combine.c: Include params.h.
832 (count_rtxs): New function.
833 (record_value_for_reg): If replace_rtx would replace at least
834 2 occurrences of REG in VALUE and TEM is really large, replace REG with
835 (clobber (const_int 0)) instead of TEM.
836 * params.def (PARAM_MAX_LAST_VALUE_RTL): New.
837 * params.h (MAX_LAST_VALUE_RTL): New.
838 * Makefile.in (combine.o): Depend on $(PARAMS_H).
839 * doc/invoke.texi (--param max-last-value-rtl=N): Document.
842 * c-typeck.c (digest_init): Only call build_vector if all constructor
843 elements are *_CST nodes.
844 * gimplify.c (gimplify_init_constructor): Likewise.
847 * c-typeck.c (digest_init): Only call build_vector if inside_init
850 2005-01-18 Tobias Schl"uter <tobias.schlueter@physik.uni-muenchen.de>
852 * toplev.c (init_asm_output): Remove dead #ifdef.
854 2005-01-18 Kaz Kojima <kkojima@gcc.gnu.org>
856 * config/sh/linux.h (TARGET_C99_FUNCTIONS): Define.
858 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
860 * calls.c (expand_call): Check DECL_BUILT_IN_CLASS before
861 accessing DECL_FUNCTION_CODE.
862 * dojump.c (do_jump): Likewise.
863 * gimplify.c (gimplify_call_expr): Likewise.
864 * predict.c (expr_expected_value): Likewise.
865 (strip_builtin_expect): Likewise.
866 * tree-inline.c (estimate_num_insns_1): Likewise.
867 * tree-ssa-loop-im.c (stmt_cost): Likewise
868 * fold-const.c (fold): Test for BUILT_IN_NORMAL.
869 (tree_expr_nonnegative_p): Likewise.
871 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
873 * config/sparc/sparc.c (load_pic_register): Emit the appropriate
874 variant of the load_pcrel_sym pattern.
875 * config/sparc/sparc.md (P macro): Move to the top.
876 (load_pcrel_sym): Macroize using P.
877 (save_register_window): Likewise.
879 2005-01-18 Aldy Hernandez <aldyh@redhat.com>
881 * config/rs6000/rs6000.md ("sunordered"): Disable for e500.
884 2005-01-18 Andrew Pinski <pinskia@physics.uc.edu>
887 * c-typeck.c (build_asm_expr): Strip nops off
888 input memory operands.
890 2005-01-18 David Edelsohn <edelsohn@gnu.org>
892 * config/rs6000/aix43.h (CPLUSPLUS_CPP_SPEC): Delete
893 _XOPEN_SOURCE, _XOPEN_SOURCE_EXTENDED, _LARGE_FILE_API.
894 * config/rs6000/aix51.h (CPLUSPLUS_CPP_SPEC): Same.
895 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Same.
897 2005-01-18 Joel Sherrill <joel.sherrill@oarcorp.com>
900 * config/i386/i386.c (override_options): If the 80387 is disabled,
901 then do not return FP values using FP registers.
903 2005-01-18 Kazu Hirata <kazu@cs.umass.edu>
905 * basic-block.h, c-common.c, c-cppbuiltin.c, c-lang.c,
906 c-tree.h, cfgbuild.c, cgraph.c, cgraph.h, collect2.c,
907 combine.c, config.gcc, coverage.h, cse.c, cselib.c,
908 defaults.h, df.c, dwarf2asm.c, dwarf2out.c, explow.c, expr.c,
909 flow.c, fold-const.c, gcse.c, ggc-page.c, gimple-low.c,
910 gimplify.c, ifcvt.c, langhooks-def.h, lcm.c, optabs.h,
911 output.h, postreload-gcse.c, postreload.c, recog.c,
912 resource.c, rtl.def, rtlanal.c, sched-deps.c, sched-rgn.c,
913 targhooks.h, toplev.c, tree-data-ref.c, tree-eh.c,
914 tree-flow.h, tree-loop-linear.c, tree-mudflap.h, tree-nrv.c,
915 tree-optimize.c, tree-outof-ssa.c, tree-pass.h,
916 tree-scalar-evolution.c, tree-ssa-copy.c, tree-ssa-dce.c,
917 tree-ssa-dse.c, tree-ssa-loop-ivopts.c, tree-ssa-operands.c,
918 tree-ssa-pre.c, tree-ssa.c, tree-vectorizer.c, tree.def,
919 unwind-dw2-fde-darwin.c, var-tracking.c: Update copyright.
921 * config/darwin.c, config/darwin.h, config/freebsd-spec.h,
922 config/alpha/alpha.c, config/arm/vfp.md,
923 config/i386/emmintrin.h, config/i386/i386-modes.def,
924 config/i386/i386-protos.h, config/i386/i386.h,
925 config/i386/pmmintrin.h, config/i386/ppro.md,
926 config/i386/predicates.md, config/i386/xmmintrin.h,
927 config/i860/i860.c, config/ia64/ia64-modes.def,
928 config/ia64/ia64-protos.h, config/ia64/ia64.h,
929 config/ia64/ia64.md, config/ia64/itanium1.md,
930 config/ia64/itanium2.md, config/ia64/predicates.md,
931 config/iq2000/iq2000.h, config/mips/linux64.h,
932 config/rs6000/aix41.h, config/rs6000/aix43.h,
933 config/rs6000/aix52.h, config/rs6000/darwin-fallback.c,
934 config/rs6000/rs6000.c, config/rs6000/rs6000.h,
935 config/rs6000/spe.md, config/sh/sh.md, config/sparc/linux.h,
936 config/sparc/linux64.h, config/sparc/litecoff.h,
937 config/sparc/sp64-elf.h, config/sparc/sparc.c,
938 config/sparc/sparc.h, config/sparc/sparc.md,
939 config/sparc/sysv4-only.h: Update copyright.
941 2005-01-18 Richard Henderson <rth@redhat.com>
944 * config/i386/i386.c (ix86_expand_fp_movcc): Fail for LTGT and UNEQ.
946 2005-01-18 Marc Espie <espie@openbsd.org>
947 * config/i386/openbsdelf.h: Typo.
949 2005-01-18 Richard Henderson <rth@redhat.com>
951 * config/i386/i386.c (ix86_expand_fp_movcc): Remove TARGET_IEEE_FP
952 special case for sse.
953 * config/i386/predicates.md (sse_comparison_operator): Likewise.
955 2005-01-18 Dorit Naishlos <dorit@il.ibm.com>
957 * tree-inline.c (estimate_num_insns_1): Added cases for
958 ALIGN_INDIRECT_REF, MISALIGNED_INDIRECT_REF, and REALIGN_LOAD_EXPR.
960 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
962 PR rtl-optimization/19296
963 * combine.c (simplify_comparison): Rewrite the condition under
964 which a non-paradoxical SUBREG of a PLUS can be lifted when
965 compared against a constant.
967 2005-01-18 Andi Kleen <ak@muc.de>
969 * c-typeck.c: (convert_for_assignment): Check warn_pointer_sign.
970 * c.opt (-Wpointer-sign): Add.
971 * doc/invoke.texi: (-Wpointer-sign): Add.
973 2005-01-18 Uros Bizjak <uros@kss-loka.si>
975 * config/i386/i386.c (override_options): Revert 2004-11-24 change.
976 * config/i386/i386.md (fmodsf3, dremsf3, *sinsf2, *cossf2, sincossf3,
977 *tansf3_1, tansf2, atan2sf3_1, atan2sf3, atansf2, asinsf2, acossf2,
978 logsf2, log10sf2, log2sf2, log1psf2, logbsf2, ilogbsf2, expsf2,
979 exp10sf2, exp2sf2, expm1sf2, rintsf2, floorsf2, ceilsf2, btruncsf2,
980 nearbyintsf2): Disable for TARGET_SSE_MATH. Leave patterns enabled
981 for TARGET_MIX_SSE_I387.
982 (fmoddf3, dremdf3, *sindf2, *sinextendsfdf2, *cosdf2, *cosextendsfdf2,
983 sincosdf3, *sincosextendsfdf3, *tandf3_1, tandf2, atan2df3_1,
984 atan2df3, atandf2, asindf2, acosdf2, logdf2, log10df2, log2df2,
985 log1pdf2, logbdf2, expdf2, exp10df2, exp2df2, expm1df2, rintdf2,
986 floordf2, ceildf2, btruncdf2, nearbyintdf2): Disable for
987 (TARGET_SSE2 && TARGET_SSE_MATH). Leave patterns enabled for
989 (atan2sf3, atan2df3, atan2xf3): Remove register constraints
992 2005-01-18 Uros Bizjak <uros@kss-loka.si>
995 * config/i386/mmx.md (*movv2sf_internal_rex64, *movv2sf_internal):
996 Add movaps alternative for xmm reg->reg move.
998 2005-01-17 Jeff Law <law@redhat.com>
1000 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Clear SSA_NAME_VALUE
1001 at the end of the main DOM loop rather than just before DOM exits.
1003 2005-01-17 Diego Novillo <dnovillo@redhat.com>
1005 PR tree-optimization/19121
1006 * tree-ssa-alias.c (compute_flow_sensitive_aliasing): When
1007 adding aliases to a name tag, also add them to the pointer's
1009 * tree-ssa-copy.c (merge_alias_info): Do not merge flow
1010 sensitive alias info at all. Only check that the two pointers
1011 have compatible pointed-to sets.
1012 * tree-ssa.c (verify_name_tags): Verify that the alias set of
1013 a pointer's type tag is a superset of the alias set of the
1016 2005-01-17 James E Wilson <wilson@specifixinc.com>
1019 * config/ia64/ia64.md (movxf): Handle general register source. Adjust
1020 comment to document why.
1022 2005-01-17 Richard Henderson <rth@redhat.com>
1024 * config/i386/sse.md (smaxv4sf3_finite, sse_vmsmaxv4sf3_finite,
1025 sminv4sf3_finite, sse_vmsminv4sf3_finite, smaxv2df3_finite,
1026 sse2_vmsmaxv2df3_finite, sminv2df3_finite,
1027 sse2_vmsminv2df3_finite): New.
1028 (smaxv4sf3, sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, smaxv2df3,
1029 sse2_vmsmaxv2df3, sminv2df3, sse2_vmsminv2df3): Remove commutative.
1030 Force op1 into register.
1032 2005-01-17 Kelley Cook <kcook@gcc.gnu.org>
1034 * Makefile.in (STAGEMOVESTUFF): Stage all the stamp files to prevent
1035 bubblestrap from always rebuilding build directory.
1037 2005-01-17 Ulrich Weigand <uweigand@de.ibm.com>
1039 * config/s390/s390.md ("*clc"): Remove incorrect '='.
1040 ("*cmpmem_short"): Likewise.
1042 2005-01-17 Steven Bosscher <stevenb@suse.de>
1044 * basic-block.h: Document BB_* flags.
1045 * regrename.c (copyprop_hardreg_forward): Don't use BB_VISITED,
1046 use an sbitmap instead.
1047 * sched-rgn.c (compute_trg_info): Likewise.
1049 2005-01-17 Richard Sandiford <rsandifo@redhat.com>
1051 * config.gcc (mips64*-*-linux*): Set the default abi to n32. Remove
1052 redundant target_cpu_default and extra_parts lines. Use the default
1054 * config/mips/linux64.h (DRIVER_DEFAULT_ABI_SELF_SPEC): Delete.
1055 (SUBTARGET_EXTRA_SPECS): Delete.
1056 (DRIVER_SELF_SPECS): Remove architecture lines. Set the ABI directly.
1057 (SUBTARGET_ASM_SPEC): Remove -mabi=n32 and -mabi=64 mappings.
1058 * config/mips/t-linux64 (MULTILIB_DIRNAMES): Use "32" for -mabi=32
1059 and "n32" for -mabi=n32.
1061 2005-01-17 Ian Lance Taylor <ian@airs.com>
1064 * c-typeck.c (build_c_cast): Revert patch of 2003-02-16: don't
1065 fold constant variables into initial values.
1067 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
1069 * tree-cfg.c (tree_can_merge_blocks_p): Reorder two checks.
1071 * tree-cfg.c (tree_forwarder_block_p): Speed up by walking
1072 through the statements backward.
1074 2005-01-17 Ian Lance Taylor <ian@airs.com>
1076 PR middle-end/13127:
1077 * tree-inline.c (expand_call_inline): Set TREE_NO_WARNING on
1078 a variable set to the return value of the inlined function.
1080 2005-01-17 Mark Dettinger <dettinge@de.ibm.com>
1082 * rtlanal.c (rtx_cost): Assign cost of 0 to a SUBREG
1083 when modes are tieable.
1085 2005-01-17 Ranjit Mathew <rmathew@hotmail.com>
1087 * gthr-posix.h (__gthread_active_p): Use pthread_cancel instead
1088 of pthread_create to find out if threads are enabled.
1089 * gthr-posix95.h (__gthread_active_p): Likewise.
1091 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
1093 * common.opt (-fnew-ra): Remove.
1095 * toplev.h (flag_new_regalloc): Remove.
1096 * Makefile.in (ra*.*): Don't mention.
1097 * passes.c (rest_of_handle_new_regalloc): Remove.
1098 (rest_of_handle_combine, rest_of_compilation): Always consider
1099 flag_new_regalloc as false.
1100 * doc/invoke.texi: Don't document -fnew-ra.
1102 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
1104 * bb-reorder.c (fix_edges_for_rarely_executed_code): Remove
1105 last parameter to reg_scan.
1106 * loop.c (loop_optimize): Likewise.
1107 * passes.c (rest_of_handle_tracer, rest_of_handle_if_conversion,
1108 rest_of_handle_web, rest_of_handle_cfg, rest_of_handle_jump_bypass,
1109 rest_of_handle_life, rest_of_handle_cse, rest_of_handle_cse2,
1110 rest_of_handle_gcse, rest_of_handle_loop_optimize,
1111 rest_of_handle_loop2, rest_of_handle_jump2): Likewise.
1112 * regclass.c (reg_scan): Likewise, for the declaration.
1113 * rtl.h (reg_scan): Likewise, for the prototype.
1115 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
1117 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Don't call
1118 mark_dfs_back_edges.
1120 2005-01-16 Daniel Berlin <dberlin@dberlin.org>
1122 * tree-ssa-pre.c (add_to_sets): s1 may be NULL.
1123 (compute_avail): Uses don't go in tmp_gen.
1125 2005-01-16 Steven Bosscher <stevenb@suse.de>
1127 * ggc-page.c (ggc_alloc_stat): Use __builtin_ctzl instead of a
1128 loop to look for a free slot in a page entry.
1130 2005-01-16 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1133 * defaults.h (TARGET_DEFERRED_OUTPUT_DEFS): Provide default.
1134 * toplev.c (compile_file): Call process_pending_assemble_output_defs
1135 just before targetm.asm_out.file_end.
1136 * tree.h (process_pending_assemble_output_defs): Declare.
1137 * varasm.c (assemble_output_def, process_pending_assemble_output_defs):
1139 (assemble_alias): Defer generation of assembly code for defines when
1140 TARGET_DEFERRED_OUTPUT_DEFS is true.
1141 * config/rs6000/aix41.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
1142 * config/rs6000/aix43.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
1143 * doc/tm.texi (TARGET_DEFERRED_OUTPUT_DEFS): document.
1145 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1148 * pa.c (pa_scalar_mode_supported_p): New function.
1149 (TARGET_SCALAR_MODE_SUPPORTED_P): Define.
1150 * pa.h (MIN_UNITS_PER_WORD): Add comment.
1152 * pa-protos.h (prefetch_operand): Delete.
1153 (prefetch_cc_operand, prefetch_nocc_operand): New declations.
1154 * pa.c (prefetch_operand): Delete.
1155 (prefetch_cc_operand, prefetch_nocc_operand): New functions.
1156 * pa.h (EXTRA_CONSTRAINT): Add `W' constraint.
1157 (PREDICATE_CODES): Delete prefetch_operand. Add prefetch_cc_operand
1158 and prefetch_nocc_operand.
1159 * pa.md (prefetch): Rework to avoid reload problems handling short
1160 displacements when a cache control completer needs to be provided.
1161 (prefetch_32, prefetch_64): Delete.
1162 (prefetch_cc, prefetch_nocc): New patterns.
1164 2005-01-15 David Edelsohn <edelsohn@gnu.org>
1166 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Revert previous change.
1168 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1171 * pa.c (pa_asm_output_aligned_common): Warn if specified alignment
1172 exceeds maximum alignment for global common data.
1173 * pa64-hpux.h (MAX_OFILE_ALIGNMENT): Define to 32768.
1174 * som.h (MAX_OFILE_ALIGNMENT): Likewise.
1176 2004-01-15 Roger Sayle <roger@eyesopen.com>
1178 * tree-ssa-dom.c (extract_range_from_cond): Correct condition.
1180 2004-01-15 Roger Sayle <roger@eyesopen.com>
1182 * harg-reg-set.h (reg_class_names): Prototype global array.
1183 * regclass.c (reg_class_names): Declare here and initialize to
1185 (dump_regclass): Remove local declaration of reg_class_names.
1186 (regclass): Likewise.
1187 * cfg.c (dump_flow_info): Likewise.
1188 * ra-debug.c (reg_class_names): Likewise.
1189 * regrename.c (reg_class_names): Likewise.
1190 * reload.c (reg_class_names): Likewise.
1191 * reload1.c (spill_failure): Likewise.
1192 * config/m68hc11/m68hc11.c (reg_class_names): Likewise.
1194 2005-01-15 Ulrich Weigand <uweigand@de.ibm.com>
1196 * config/s390/s390.md ("reload_outti"): Remove predicate for
1197 output operand. Abort if operand is not a MEM.
1198 ("reload_outdi", "reload_outdf"): Likewise.
1200 2005-01-15 Marc Espie <espie@openbsd.org>
1202 * config.gcc (*-*-openbsd*): Set HAS_LIBC_R for OpenBSD <= 3.2.
1203 (i[34567]86-*-openbsd*): Switch to ELF for OpenBSD >= 3.4.
1204 * config/openbsd.h: Fix C++ includes for native configurations.
1205 Add proper OS_CPP_BUILTINS. Fix libspec for recent OpenBSD.
1206 Add trampoline support.
1207 * config/i386/openbsdelf.h: New.
1209 2005-01-15 Marc Espie <espie@openbsd.org>
1211 * collect2.c (main): Explicitly parse -dynamic-linker option.
1213 2005-01-15 Jakub Jelinek <jakub@redhat.com>
1215 PR tree-optimization/19060
1216 * tree-ssa-dom.c (extract_range_from_cond) <case LT_EXPR, GT_EXPR>:
1217 Return 0 if op1 <= TYPE_MIN_VALUE () resp. op1 >= TYPE_MAX_VALUE ().
1218 (simplify_cond_and_lookup_avail_expr): Add assert for dummy == 0
1219 and handle extract_range_from_cond returning false.
1220 * fold-const.c (fold): Optimize comparisons with min/max even for
1221 width > HOST_BITS_PER_WIDE_INT.
1223 2005-01-15 Ralf Corsepius <ralf.corsepius@rtems.org>
1225 * config/mips/rtems.h (MIPS_DEFAULT_GVALUE): Set to 0.
1226 * config/mips/t-rtems (MULTILIBS_DIRNAMES,MULTILIB_OPTIONS):
1227 Remove little endian multilib variants.
1228 Add mips32 multilib variant.
1230 2005-01-15 Kazu Hirata <kazu@cs.umass.edu>
1232 * explow.c (copy_all_reg, stabilize): Remove.
1233 * expr.h: Remove the corresponding prototypes.
1235 * rtlanal.c (regs_set_between_p): Remove.
1236 * rtl.h: Remove the corresponding prototype.
1238 2005-01-14 David Edelsohn <edelsohn@gnu.org>
1240 * config/rs6000/rs6000.c (rs6k_nonimmediate_operand): Rename
1241 to rs6000_nonimmediate_operand.
1242 * config/rs6000/rs6000.h (PREDICATE_CODES): Rename
1243 rs6k_nonimmediate_operand to rs6000_nonimmediate_operand.
1244 * config/rs6000/rs6000.md (movsi_internal1): Rename predicate to
1245 rs6000_nonimmediate_operand.
1246 * config/rs6000/spe.md (movdf_e500_double): Rename predicate to
1247 rs6000_nonimmediate_operand.
1249 2005-01-14 Andrew Pinski <pinskia@physics.uc.edu>
1251 * tree-ssa-alias.c (pass_may_alias): Add TODO_verify_stmts.
1253 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1255 * config/rs6000/rs6000.h (CLASS_MAX_NREGS): DF goes in 1 register
1257 (CANNOT_CHANGE_MODE_CLASS): Restrict DI mode changes on e500v2.
1258 (PREDICATE_CODES): Add rs6k_nonimmediate_operand.
1260 * config/rs6000/rs6000.c (invalid_e500_subreg): New.
1261 (rs6k_nonimmediate_operand): New.
1262 (rs6000_legitimate_offset_address_p): Handle DI modes on e500v2
1264 (legitimate_lo_sum_address_p): Same.
1265 (rs6000_legitimize_address): Same.
1266 (rs6000_legitimize_reload_address): Same.
1267 (rs6000_legitimate_address): Same.
1268 (spe_build_register_parallel): Pass DF and DC modes in a DI
1271 * config/rs6000/rs6000.md ("*movsi_internal1"): Change predicate
1272 to rs6k_nonimmediate_operand.
1274 * config/rs6000/spe.md ("*frob_df_di"): New.
1275 ("*frob_di_df"): New.
1276 ("*frob_di_df_2"): New.
1277 ("*mov_sidf_e500_subreg0"): New.
1278 ("*mov_sidf_e500_subreg4"): New.
1279 ("*movdf_e500_double"): Change predicate to
1280 rs6k_nonimmediate_operand.
1282 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1284 * postreload.c (move2add_note_store): Only call
1285 trunc_int_for_mode on scalar integers.
1287 2005-01-14 Steven Bosscher <stevenb@suse.de>
1289 * tree-ssa-dce.c (visited_control_parents): New sbitmap to
1290 replace BB_VISITED uses.
1291 (find_obviously_necessary_stmts): Don't clear BB_VISITED.
1292 (propagate_necessity): Check the bitmap instead of BB_VISITED.
1293 (tree_dce_done): Free visited_control_parents.
1294 (perform_tree_ssa_dce): Allocate and clear it.
1295 * tree-ssa-pre.c (compute_antic_aux): Make non-recursive.
1296 (compute_antic): Iterate from here using a DFS. Use an sbitmap
1297 instead of BB_VISITED.
1299 2005-01-14 Kazu Hirata <kazu@cs.umass.edu>
1301 * c-tree.h, coverage.h, langhooks-def.h, optabs.h, output.h,
1302 rtl.h, targhooks.h, tree-flow.h, tree-mudflap.h, tree.h:
1303 Remove unused prototypes.
1305 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1307 * config/rs6000/rs6000.c (rs6000_stack_info): Fix caching of SPE
1308 64-bit register usage.
1309 (spe_func_has_64bit_regs_p): Add FIXME note.
1311 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1313 * config/rs6000/rs6000.c (rs6000_generate_compare): Replace
1314 flag_finite_math_only with flag_unsafe_math_optimizations.
1316 2005-01-14 J. D. Johnston <jjohnst@us.ibm.com>
1318 * config/s390/tpf-unwind.h (s390_fallback_frame_state): Correct
1320 (__tpf_eh_return): Copy TPF private stack area from the module
1321 boundary stack frame. Add check for when module addresses are
1322 equal, but no stub address is found.
1324 2005-01-14 Richard Earnshaw <rearnsha@arm.com>
1327 * arm.h (struct machine_function): Add call_via field.
1328 (thumb_call_via_label): Declare.
1329 * arm.c (thumb_call_via_label): New variable.
1330 (thumb_call_reg_needed): New variable.
1331 (arm_output_function_epilogue): For Thumb code, output any per-function
1332 call-indirect trampolines.
1333 (thumb_call_via_reg): New function.
1334 (arm_file_end): New function.
1335 (TARGET_ASM_FILE_END): Call arm_file_end.
1336 (aof_file_end): Likewise.
1337 * arm-protos.h (thumb_call_via_reg): Declare.
1338 * arm.md (call_reg_thumb, call_value_reg_thumb): Call
1339 thumb_call_via_reg in normal case.
1341 2005-01-14 Jakub Jelinek <jakub@redhat.com>
1344 PR rtl-optimization/19348
1345 * recog.c (peephole2_optimize): Do global life update if some peephole
1346 decides it doesn't need at least one of its inputs and that change
1347 influences liveness at the start of the basic block.
1349 * basic-block.h (EXECUTE_IF_AND_COMPL_IN_REG_SET): Needs 2 REGSET
1350 arguments instead of 1.
1352 2005-01-14 Eric Botcazou <ebotcazou@libertysurf.fr>
1355 * varasm.c (initializer_constant_valid_p) <ADDR_EXPR>: Return
1356 zero for nested functions needing a static chain or functions
1357 with a non-constant address.
1359 2005-01-13 Roger Sayle <roger@eyesopen.com>
1361 * simplify-rtx.c (simplify_binary_operation) <AND>: Optimize
1362 (and (sign_extend X) C) into (zero_extend (and X C)).
1364 2005-01-13 David O'Brien <obrien@FreeBSD.org>
1366 * config/freebsd-spec.h: Make KSE pthread lib logic the default.
1368 2005-01-13 Richard Henderson <rth@redhat.com>
1373 * config/i386/i386.md (cmpdf, cmpsf, bunordered, bordered, buneq,
1374 bunge, bungt, bunle, bunlt, bltgt): Enable for TARGET_SSE_MATH,
1375 not just TARGET_SSE.
1376 (cmpfp_i_387): Rename from cmpfp_i. Move after sse patterns.
1377 (cmpfp_i_mixed): Rename from cmpfp_i_sse; use for TARGET_MIX_SSE_I387.
1378 (cmpfp_i_sse): Rename from cmpfp_i_sse_only; use for TARGET_SSE_MATH.
1379 (cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387): Similarly.
1380 (fp_jcc_1_mixed, fp_jcc_1_sse, fp_jcc_1_387): Similarly.
1381 (fp_jcc_2_mixed, fp_jcc_2_sse, fp_jcc_2_387): Similarly.
1382 (fp_jcc_3_387, fp_jcc_4_387, fp_jcc_5_387, fp_jcc_6_387,
1383 fp_jcc_7_387, fp_jcc_8_387): Rename from fp_jcc_N.
1384 (movdicc_c_rex64): Rename with '*'.
1385 (movsfcc, movdfcc): Add checks for 387 and sse math to condition.
1386 (movsfcc_1_sse_min, movsfcc_1_sse_max, movsfcc_1_sse): New.
1387 (movsfcc_1_387): Rename from movsfcc_1.
1388 (movdfcc_1_sse_min, movdfcc_1_sse_max, movdfcc_1_sse): New.
1389 (movdfcc_1, movdfcc_1_rex64): Add check for 387.
1390 (sminsf3, smaxsf3, smindf3, smaxdf3): New.
1391 (minsf3, minsf, minsf_nonieee, minsf_sse, mindf3, mindf,
1392 mindf_nonieee, mindf_sse, maxsf3, maxsf, maxsf_nonieee, maxsf_sse,
1393 maxdf3, maxdf, maxdf_nonieee, maxdf_sse, sse_movsfcc, sse_movsfcc_eq,
1394 sse_movdfcc, sse_movdfcc_eq, sse_movsfcc_const0_1,
1395 sse_movsfcc_const0_2, sse_movsfcc_const0_3, sse_movsfcc_const0_4,
1396 sse_movdfcc_const0_1, sse_movdfcc_const0_2, sse_movdfcc_const0_3,
1397 sse_movdfcc_const0_4): Remove.
1398 * config/i386/i386.c (ix86_expand_fp_movcc): For TARGET_SSE_MATH,
1399 recognize min/max early. Update for changed sse cmove patterns.
1400 (ix86_split_sse_movcc): New.
1401 * config/i386/i386-protos.h: Update.
1403 2005-01-13 Steven Bosscher <stevenb@suse.de>
1405 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
1406 (fix_stmt_v_may_defs): Likewise.
1408 2005-01-13 Richard Henderson <rth@redhat.com>
1410 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
1411 for converting the mask.
1413 2005-01-13 David Edelsohn <edelsohn@gnu.org>
1415 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
1418 2005-01-13 Richard Henderson <rth@redhat.com>
1420 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
1421 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
1422 IX86_BUILTIN_MOVDQ2Q): Remove.
1423 (IX86_BUILTIN_VEC_EXT_V4SI): New.
1424 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
1425 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
1427 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
1428 (sse2_pextrw): Fix immediate constraint.
1429 (sse2_loadq, sse2_loadq_rex64): Remove.
1430 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
1431 Use __builtin_ia32_vec_ext_<size>.
1432 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
1434 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
1436 * function.c (assign_parm_setup_block): Look inside original
1437 entry_parm when inspecting PARALLEL.
1439 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
1440 Joel Sherrill <joel@oarcorp.com>
1443 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
1444 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
1445 rtems_gxx_recursive_mutex_init.
1446 (__gthread_recursive_mutex_lock): New function.
1447 (__gthread_recursive_mutex_trylock): Likewise.
1448 (__gthread_recursive_mutex_unlock): Likewise.
1450 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
1452 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
1454 2005-01-13 Jan Beulich <jbeulich@novell.com>
1456 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
1457 (ashrti3_internal): Indicate output is early clobber. Generate result
1458 into output rather than first input. Use move for low word of output
1459 if shift count is exactly 64.
1460 (lshrti3_internal): Likewise.
1462 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
1466 * reload1.c (delete_output_reload): Don't delete an output reload
1467 if the pseudo lives longer than a single basic block. Adjust and
1468 improve wording of comment.
1470 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
1473 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
1474 of TYPE_SIZE where we mean the number of bytes.
1476 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
1478 * function.c (assign_parm_setup_block): Relax condition on
1479 multi-register optimization.
1481 2005-01-12 Nick Clifton <nickc@redhat.com>
1483 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
1484 an "in_delay_slot" attribute of "no" to prevent them being used in
1485 delay slots. This is forbidden because they might generate
1488 2005-01-12 Alan Modra <amodra@bigpond.net.au>
1491 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
1494 2005-01-12 Nick Clifton <nickc@redhat.com>
1496 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
1497 passed on to GAS) but do not define, as GAS no longer supports or
1498 needs the -m2000 option.
1500 2005-01-11 Roger Sayle <roger@eyesopen.com>
1502 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
1503 type as args of unprototyped builtins aren't automatically promoted.
1504 (fold_builtin_abs): Likewise.
1506 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1508 * Makefile.in: Set a `build-warn' variable.
1510 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
1512 PR tree-optimization/17949
1513 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
1514 (find_interesting_uses_address): Use it.
1516 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
1518 * regrename.c (kill_value): Handle subreg's that won't simplify.
1520 2005-01-11 Richard Henderson <rth@redhat.com>
1523 * config/i386/i386.h (enum ix86_builtins): Move ...
1524 * config/i386/i386.c: ... here.
1525 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
1526 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
1527 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
1528 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
1529 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
1530 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
1531 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
1532 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
1533 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
1534 IX86_BUILTIN_LOADDDUP): Remove.
1535 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
1536 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
1537 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
1538 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
1539 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
1540 (ix86_init_builtins): Make static.
1541 (ix86_init_mmx_sse_builtins): Update for changed builtins.
1542 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
1543 if all the modes match. Otherwise, fake it.
1544 (get_element_number, ix86_expand_vec_init_builtin,
1545 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
1546 (ix86_expand_builtin): Make static. Update for changed builtins.
1547 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
1548 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
1549 (ix86_expand_vector_init_duplicate): New.
1550 (ix86_expand_vector_init_low_nonzero): New.
1551 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
1552 Split out from ix86_expand_vector_init; handle integer modes.
1553 (ix86_expand_vector_init): Use them.
1554 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
1555 * config/i386/i386-protos.h: Update.
1556 * config/i386/predicates.md (reg_or_0_operand): New.
1557 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
1558 (movv2sf_internal): Likewise. And a splitter to match them all.
1559 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
1560 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
1561 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
1562 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
1563 (mmx_pinsrw): Fix operand ordering.
1564 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
1565 rather than sse_loadss expander.
1566 (movv2df splitter): Similarly.
1567 (sse_loadss, sse_loadlss): Remove.
1568 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
1569 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
1570 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
1571 (sse3_movddup): Rename with '*'.
1572 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
1573 (sse2_loadsd): Remove.
1574 (vec_dupv2df_sse3): Rename from sse3_loadddup.
1575 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
1576 (sse2_pinsrw): Fix argument ordering.
1577 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
1578 (sse2_stored): Remove 'r' destination.
1579 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
1580 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
1581 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
1582 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
1583 vec_extractv16qi, vec_initv16qi): New.
1585 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
1586 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
1587 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
1589 (_mm_load_pd, _mm_store_pd): Use plain dereference.
1590 (_mm_load_si128, _mm_store_si128): Likewise.
1591 (_mm_load1_pd): Use _mm_set1_pd.
1592 (_mm_load_sd): Use _mm_set_sd.
1593 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
1594 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
1595 (_mm_set_epi64): Use _mm_set_epi64x.
1596 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
1597 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
1598 _mm_setr_epi8): Use _mm_set_foo form.
1599 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
1600 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
1601 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
1602 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
1603 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
1604 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
1605 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
1606 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
1607 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
1608 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
1609 (_mm_movedup_pd): Use _mm_shuffle_pd.
1610 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
1611 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
1612 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
1613 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
1615 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
1616 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
1617 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
1618 (_mm_store_ps): Use raw dereference.
1619 (_mm_store1_ps): Use _mm_storeu_ps.
1620 (_mm_storer_ps): Use _mm_store_ps.
1621 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
1622 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
1624 2005-01-11 Stan Shebs <shebs@apple.com>
1626 * config/rs6000/rs6000.c (machopic_output_stub): Issue
1627 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
1629 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
1632 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
1633 Skip all DECLs except for FIELD_DECLs.
1635 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
1637 * config/s390/s390.c (override_options): Return error if
1638 -mbackchain, -mpacked-stack and -mhard-float are used together.
1639 (s390_va_start): Remove the backchain && packed-stack special case.
1640 (s390_gimplify_va_arg): Likewise.
1641 * doc/invoke.texi: Remove the ABI incompatibility note.
1643 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
1645 * config/s390/s390.c (struct s390_frame_layout): Remove
1647 (s390_frame_info, s390_emit_prologue): Replace occurrences of
1648 save_backchain_p with TARGET_BACKCHAIN.
1650 2005-01-11 Alan Modra <amodra@bigpond.net.au>
1653 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
1654 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
1655 * function.c (locate_and_pad_parm): Set new field.
1656 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
1657 Tweak where_pad test to include "none". Always set mem align for
1659 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
1660 not sufficient for type.
1661 (assign_parm_setup_block): If stack_parm is zero on entry, always
1662 make a new stack local. Block move old stack parm if necessary
1663 to new aligned stack local.
1664 (assign_parm_setup_stack): Use a block move to handle
1665 potentially misaligned entry_parm.
1666 (assign_parms_unsplit_complex): Specify required alignment when
1667 creating stack local.
1668 * calls.c (compute_argument_addresses): Override alignment of stack
1669 arg calculated from its type with the alignment given by
1670 FUNCTION_ARG_BOUNDARY.
1671 (store_one_arg): Likewise.
1673 2005-01-11 Jan Beulich <jbeulich@novell.com>
1675 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
1676 Change respective itanium_class attribute to ialu.
1677 (shladdp4_internal): New.
1678 * config/ia64/predicates.md (shladd_log2_operand): New.
1680 2005-01-11 Richard Henderson <rth@redhat.com>
1682 * expr.c (store_constructor): Use rtvec_alloc instead of
1683 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
1685 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
1687 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
1689 2005-01-11 Alan Modra <amodra@bigpond.net.au>
1691 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
1692 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
1693 (TARGET_RWRELOC): Define.
1694 (ia64_rwreloc_section_type_flags): Delete.
1695 (ia64_section_type_flags): New function.
1696 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
1697 (TARGET_RWRELOC): Define.
1699 2005-01-10 David Mosberger <davidm@hpl.hp.com>
1702 * config/ia64/ia64.c (process_set): For alloc insn, only call
1703 process_epilogue is !frame_pointer_needed.
1705 2005-01-10 Roger Sayle <roger@eyesopen.com>
1708 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
1709 unary operator and can't be treated as a binary/comparison operator.
1711 2005-01-10 Richard Henderson <rth@redhat.com>
1713 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
1714 (ix86_return_in_memory): Likewise.
1715 (function_arg_advance): Likewise. Mirror structure in function_arg
1716 for choosing register to advance.
1718 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
1720 * tree-vectorizer.c, tree.def: Fix comment typos.
1722 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
1724 * config/s390/s390.c (struct processor_costs): 4 new fields:
1725 ddbr, ddr, debr, der.
1726 (s390_rtx_costs): More precise handling of divide instructions.
1728 2005-01-10 David Edelsohn <edelsohn@gnu.org>
1730 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
1732 2005-01-10 Jan Beulich <jbeulich@novell.com>
1734 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
1735 names resulting from -ffunction-sections/-fdata-sections and linkonce
1737 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
1738 based on the section name. Rearrange the section name comparison logic
1739 slightly so that each section name is compared against at most once.
1741 2005-01-10 Ben Elliston <bje@au.ibm.com>
1743 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
1744 targets for which -fPIC reduces limitations on the GOT size.
1746 2005-01-09 Falk Hueffner <falk@debian.org>
1748 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
1751 2005-01-09 David Edelsohn <edelsohn@gnu.org>
1754 * collect2.c (main): Set aixrtl_flag for -brtl option.
1755 (resolve_lib_name): Search for .so file extension before .a
1758 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
1760 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
1761 (vect_analyze_data_refs_alignment): Add dump prints.
1763 2005-01-09 Ira Rosen <irar@il.ibm.com>
1765 * tree-vectorizer.c (vect_analyze_offset_expr): Use
1766 expr_invariant_in_loop_p.
1767 Initialize outputs first thing in the function.
1768 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
1769 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
1770 (vect_analyze_pointer_ref_access): Check that the initial condition of
1771 the access function is loop invariant.
1773 2005-01-09 Richard Henderson <rth@redhat.com>
1775 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
1776 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
1777 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
1778 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
1779 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
1780 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
1781 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
1782 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
1783 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
1784 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
1785 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
1786 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
1787 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
1788 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
1789 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
1790 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
1791 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
1792 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
1793 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
1794 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
1795 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
1796 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
1797 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
1799 (mmx_clrdi, pavgusb): Remove.
1800 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
1801 with leading sse_ prefix.
1802 * config/i386/sse.md: Receive them.
1803 * config/i386/mmx.md: New file.
1804 (MMXMODE12, MMXMODE24, mmxvecsize): New.
1805 (subrv2sf3): Turn into expander for normal subtraction.
1806 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
1807 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
1808 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
1809 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
1810 commutative; use ix86_binary_operator_ok.
1811 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
1812 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
1813 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
1814 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
1815 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
1816 Macroize from existing patterns; use ix86_binary_operator_ok.
1817 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
1818 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
1819 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
1820 with vec_select+vec_concat.
1821 (mmx_pshufw, mmx_pshufw_1): Likewise.
1822 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
1823 (mmx_uavgv4hi3): Model correctly.
1824 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
1825 _mm_xor_si64): Remove casts.
1827 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
1829 PR tree-optimization/19224
1830 * tree-scalar-evolution.c (get_instantiated_value,
1831 set_instantiated_value): New functions.
1832 (instantiate_parameters_1): Cache the results.
1833 (instantiate_parameters, resolve_mixers): Initialize and free
1836 2005-01-08 David Edelsohn <edelsohn@gnu.org>
1838 * config/i386/i386.md (addhi_4): Correct reference in comment.
1841 2005-01-08 Richard Henderson <rth@redhat.com>
1843 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
1844 (_mm_cvtsi128_si64x): Likewise.
1845 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
1846 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
1847 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
1848 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
1849 * config/i386/i386-protos.h: Update.
1850 * config/i386/i386.c (print_operand): Add 'H'.
1851 (ix86_fixup_binary_operands): Split out from ...
1852 (ix86_expand_binary_operator): ... here.
1853 (ix86_fixup_binary_operands_no_copy): New.
1854 (ix86_expand_fp_absneg_operator): Handle vector mode results.
1855 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
1856 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
1857 (safe_vector_operand): Use CONST0_RTX.
1858 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
1859 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
1860 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
1861 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
1863 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
1864 (IX86_BUILTIN_CMPNGESS): New.
1865 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
1866 (attr type): Add sselog1.
1867 (attr unit, attr memory): Handle it.
1868 (movti, movti_internal, movti_rex64): Move near other integer moves.
1869 (movtf, movtf_internal): Move near other fp moves.
1870 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
1871 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
1872 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
1873 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
1874 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
1875 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
1876 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
1877 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
1878 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
1879 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
1880 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
1881 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
1882 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
1883 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
1884 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
1885 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
1886 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
1887 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
1888 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
1889 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
1890 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
1891 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
1892 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
1893 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
1894 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
1895 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
1896 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
1897 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
1898 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
1899 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
1900 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
1901 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
1902 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
1903 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
1904 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
1905 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
1906 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
1907 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
1908 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
1909 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
1910 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
1911 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
1912 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
1913 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
1915 (SSEPUSH, push<SSEPUSH>): Remove.
1916 (MMXPUSH, push<MMXPUSH>): Remove.
1917 (sse_movaps, sse_movaps_1, sse_movups): Remove.
1918 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
1919 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
1920 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
1921 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
1922 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
1923 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
1924 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
1925 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
1926 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
1927 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
1928 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
1929 (ppro_sse_log_V4SF): Handle sselog1.
1930 * config/i386/predicates.md (const_0_to_1_operand): New.
1931 (const_0_to_255_mul_8_operand): New.
1932 (const_1_to_31_operand): Rename from const_int_1_31_operand.
1933 (const_2_to_3_operand, const_4_to_7_operand): New.
1934 * config/i386/sse.md: New file.
1935 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
1936 (sse_movups): Rename from sse_movups_1.
1937 (sse_loadlss): Rename from sse_loadss_1.
1938 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
1940 (negv4sf2): Use ix86_expand_fp_absneg_operator.
1941 (absv4sf2, negv2df, absv2df): New.
1942 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
1943 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
1944 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
1945 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
1946 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
1947 (sse3_addsubv4sf3): Model correctly.
1948 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
1949 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
1950 (sse_movhlps): Model with vec_select+vec_concat.
1951 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
1952 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
1953 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
1954 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
1955 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
1956 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
1957 sse2_pshufhw_1): Likewise.
1958 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
1959 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
1960 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
1961 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
1962 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
1963 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
1964 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
1965 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
1966 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
1967 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
1968 sminv8hi3): Mark commutative
1969 operands. Use ix86_binary_operator_ok.
1970 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
1971 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
1972 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
1973 sse2_punpcklqdq): Allow operand2 in memory.
1974 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
1975 sse2_movsd): Add memory alternatives.
1976 (sse_storelps): Turn expander into an insn; split after reload.
1977 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
1978 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
1980 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
1982 * configure.ac (DWARF-2 debug_line): Use objdump.
1983 * configure: Regenerate.
1985 2005-01-08 Jeff Law <law@redhat.com>
1986 Diego Novillo <dnovillo@redhat.com>
1988 PR tree-optimization/18241
1989 * tree-nrv.c (tree_nrv): Ignore volatile return values.
1990 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
1991 statements with volatile operands.
1992 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
1993 operands after marking a statement with has_volatile_ops.
1995 2005-01-08 Roger Sayle <roger@eyesopen.com>
1997 * tree.c (int_fits_type_p): Always honor integer constant
1998 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
2000 2005-01-08 Roger Sayle <roger@eyesopen.com>
2002 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
2003 the latent bug in rtl_delete_block has been resolved.
2005 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
2007 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
2008 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
2010 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
2012 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
2013 library as @multilib_dir@/$(SHLIB_SONAME).
2015 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
2017 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
2018 * configure: Regenerate.
2020 2005-01-07 Jakub Jelinek <jakub@redhat.com>
2022 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
2025 2005-01-07 David Edelsohn <edelsohn@gnu.org>
2028 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
2029 Convert non-word aligned offset address using ld/std into
2032 2005-01-07 Richard Henderson <rth@redhat.com>
2034 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
2036 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
2038 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
2040 2005-01-07 Jakub Jelinek <jakub@redhat.com>
2042 PR tree-optimization/19283
2043 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
2044 is not shorter than the original type.
2046 PR rtl-optimization/19012
2047 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
2049 PR rtl-optimization/18861
2050 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
2051 (FULL_STATE): Define.
2052 (mark_tablejump_edge): New function.
2053 (purge_dead_tablejump_edges): New function.
2054 (find_bb_boundaries): Use it.
2056 PR tree-optimization/18828
2057 * builtins.c (expand_builtin_next_arg): Remove argument and all
2058 the argument checking.
2059 (expand_builtin): Adjust caller.
2060 (expand_builtin_va_start): Likewise. Remove error for too many
2062 (fold_builtin_next_arg): Issue error for too many arguments.
2063 After checking arguments, replace them with magic arguments that
2064 prevent further checking of the args.
2066 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
2068 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
2070 2005-01-06 Roger Sayle <roger@eyesopen.com>
2073 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
2074 and %{.S:...} (and their negative variants) to test whether the
2075 input file is assembler or pre-processed-assembler independent of
2076 the actual filename extension.
2078 2005-01-06 Roger Sayle <roger@eyesopen.com>
2080 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
2081 of sign or zero extended values.
2083 2005-01-06 Geoffrey Keating <geoffk@apple.com>
2085 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
2086 for IBM long double format correctly.
2088 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
2090 Fix PR tree-optimization/18792
2092 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
2093 to first_loop_depth, and use loop depth instead of loop number.
2094 (build_classic_dir_vector): Ditto.
2095 (compute_data_dependences_for_loop): Use depth, not loop number.
2096 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
2097 number. Pass in loops, instead of loop numbers.
2098 (gather_interchange_stats): Ditto.
2099 (linear_transform_loops): Ditto.
2101 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
2104 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
2105 biv_fits_mode_p, extension_within_bounds_p): New functions.
2106 (check_ext_dependent_givs): Use them.
2108 2005-01-06 Roger Sayle <roger@eyesopen.com>
2110 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
2111 more than one barrier, in which case we should delete them all.
2113 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
2115 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
2117 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
2119 Fix PR middle-end/19286
2121 * dwarf2out.c (gen_subprogram_die): If we've already tried to
2122 output this subprogram, simply ignore this attempt to do it again.
2123 (add_abstract_origin_attribute): Don't abort trying to add the abstract
2124 origin attribute if it's not possible.
2125 (gen_block_die): Don't ignore subblocks of "unused" blocks.
2126 (decls_for_scope): Ditto.
2127 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
2128 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
2130 2005-01-05 Richard Henderson <rth@redhat.com>
2133 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
2134 (bdesc_2arg): Use it.
2135 (ix86_expand_binop_builtin): Force operands into registers
2137 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
2138 ix86_expand_sse_compare, ix86_expand_sse_comi,
2139 ix86_expand_builtin): Likewise.
2141 2005-01-05 Richard Henderson <rth@redhat.com>
2143 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
2145 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
2146 (mulv4hi3): Set itanium_class mmmul.
2147 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
2148 (fpack): Rename from fpack_sfsf.
2150 2005-01-05 Richard Henderson <rth@redhat.com>
2153 * reload1.c (do_input_reload): Restrict the optimization deleteing
2154 a previous output reload to RELOAD_FOR_INPUT.
2156 2005-01-05 Steven Bosscher <stevenb@suse.de>
2158 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
2159 comment that we fall through after case.
2160 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
2161 (distribute_links): Likewise.
2162 * cse.c (cse_insn): Likewise.
2163 * cselib.c (cselib_invalidate_mem): Likewise.
2164 * df.c: Update comments at the top of the file.
2165 (read_modify_subreg_p): Update comments here too.
2166 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
2167 * flow.c (mark_set_1): Likewise.
2168 (mark_used_regs): Likewise.
2169 * gcse.c (mems_conflict_for_gcse_p): Likewise.
2170 (canon_list_insert): Likewise.
2171 (mark_set): Likewise.
2172 (try_replace_reg): Likewise.
2173 (store_killed_in_insn): Likewise.
2174 * loop.c (count_one_set): Likewise.
2175 (basic_induction_var): Likewise.
2176 * postreload-gcse.c (find_mem_conflicts): Likewise.
2177 * postreload.c (reload_combine_note_store): Likewise.
2178 (move2add_note_store): Likewise.
2179 * reload.c (find_equiv_reg): Likewise.
2180 (mark_referenced_resources): Likewise.
2181 * rtlanal.c (set_noop_p): Likewise.
2182 (note_stores): Likewise.
2183 (note_uses): Likewise.
2184 * sched-deps.c (sched_analyze_1): Likewise.
2185 * sched-rgn.c (check_live_1): Likewise.
2186 (update_live_1): Likewise.
2187 * config/i860/i860.c: Likewise.
2189 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
2191 (ZERO_EXTRACT): Mention that this one can be an lvalue.
2193 * doc/rtl.texi: Update documentation for bit-fields and SET.
2195 2005-01-05 Roger Sayle <roger@eyesopen.com>
2197 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
2199 2005-01-05 Jan Hubicka <jh@suse.cz>
2200 Richard Henderson <rth@redhat.com>
2203 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
2206 2005-01-05 Richard Henderson <rth@redhat.com>
2209 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
2210 (sse_shufps): Change operand 3 to const_int_operand.
2211 (sse2_storelps): Fix typo in template.
2212 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
2213 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
2214 (ix86_expand_builtin): Likewise.
2216 2005-01-05 Julian Brown <julian@codesourcery.com>
2218 * config/arm/arm.c (arm_return_in_memory): Treat complex types
2219 as aggregates for AAPCS ABIs.
2221 2005-01-05 Stan Shebs <shebs@apple.com>
2223 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
2224 Mach-O getters if ppc64.
2225 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
2227 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
2228 signed four-byte field for global code case.
2229 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
2231 2005-01-05 Roger Sayle <roger@eyesopen.com>
2234 * c-common.c: Include real.h.
2235 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
2236 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
2237 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
2238 * Makefile.in (c-common.o): Update dependencies.
2240 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
2242 * c-parse.in (asm_string): Add trailing semicolon.
2244 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
2246 * c-parse.in (asm_string): New. Don't allow wide strings in
2248 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
2249 asm_string instead of STRING.
2251 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
2253 * c-typeck.c (constructor_no_implicit): Remove.
2254 (set_designator, process_init_element): Don't check
2255 constructor_no_implicit.
2257 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
2258 Kaz Kojima <kkojima@gcc.gnu.org>
2261 * lcm.c (create_pre_exit): New.
2262 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
2263 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
2265 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
2267 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
2268 -msoft-float in target_float_switch.
2269 * arm.c (arm_override_options): Fix processing of target_float_switch.
2271 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
2273 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
2276 2004-01-05 Uros Bizjak <uros@kss-loka.si>
2278 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
2279 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
2281 2005-01-04 Roger Sayle <roger@eyesopen.com>
2283 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
2284 if deleting the then-block allows the test-block to fallthru to the
2287 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
2290 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
2291 after the full declaration if the are in two different TUs.
2293 2005-01-04 Richard Henderson <rth@redhat.com>
2296 * tree-sra.c (generate_one_element_init): Just
2297 call gimplify_and_add.
2298 (generate_element_init): Record the
2299 new referenced variables and mark them for renaming
2300 and split out to ...
2301 (generate_element_init_1): This.
2302 (scalarize_init): Don't call push_gimplify_context/
2303 pop_gimplify_context.
2305 2005-01-04 Geoffrey Keating <geoffk@apple.com>
2307 * toplev.c (get_src_pwd): Handle failure of getpwd().
2309 2005-01-04 Roger Sayle <roger@eyesopen.com>
2311 * fold-const.c (fold_single_bit_test): Delete unreachable handling
2313 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
2314 all the cases handled by it are inverted by invert_truthvalue.
2316 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
2318 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
2319 (save_register_windowsi): Likewise.
2321 2005-01-04 Richard Henderson <rth@redhat.com>
2323 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
2325 2005-01-04 Uros Bizjak <uros@kss-loka.si>
2328 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
2329 any simplifications of vector mode comparison operators.
2330 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
2332 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
2333 Devang Patel <dpatel@apple.com>
2335 PR tree-optimization/18308
2336 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
2337 the operands before creating a new expression.
2338 * dojump.c (do_jump): Make drop_through_label available
2339 for all cases. Add expansion of COND_EXPR.
2341 2005-01-04 Ira Rosen <irar@il.ibm.com>
2343 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
2344 INTEGER_CST instead of TREE_CONSTANT.
2345 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
2346 instead of TREE_CONSTANT.
2347 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
2348 instead of TREE_CONSTANT.
2350 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
2352 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
2354 2005-01-04 Uros Bizjak <uros@kss-loka.si>
2357 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
2358 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
2360 2005-01-03 Richard Henderson <rth@redhat.com>
2362 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
2363 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
2364 fold_convert_const_real_from_real): Split out from ...
2365 (fold_convert_const): ... here.
2367 2005-01-03 Richard Henderson <rth@redhat.com>
2370 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
2371 (mov<MMXMODEI>_internal): Likewise.
2372 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
2373 Add V2SF case; use it for SSE1; don't use TI.
2374 (movdf_integer): Likewise.
2375 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
2377 2005-01-03 Ira Rosen <irar@il.ibm.com>
2379 * tree-vectorizer.c (vect_strip_conversions): New function.
2380 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
2381 check for binary class.
2383 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
2387 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
2389 * gimple-low.c (mark_blocks_with_used_vars): New function.
2390 (mark_blocks_with_used_subblocks): Ditto.
2391 (mark_used_blocks): Ditto.
2392 (pass_mark_used_blocks): New pass.
2393 * tree-inline.c: Include debug.h.
2394 (expand_call_inline): Call outlining_inline_function here.
2395 * tree-optimize.c (init_tree_optimization_passes): Add
2396 pass_mark_used_blocks.
2397 * tree-pass.h (pass_mark_used_blocks): New.
2398 * Makefile.in (tree-inline.o): Add debug.h dependency.
2400 2005-01-03 Geoffrey Keating <geoffk@apple.com>
2402 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
2403 VAR_DECLs to have weak_import attribute.
2405 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
2408 2004-01-03 Steven Bosscher <stevenb@suse.de>
2410 * passes.c (rest_of_compilation): Don't run regmove if only
2411 flag_expensive_optimizations, require flag_regmove instead.
2413 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
2415 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
2417 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
2418 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
2419 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
2420 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
2422 2005-01-03 Richard Henderson <rth@redhat.com>
2424 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
2425 (ia64_const_ok_for_letter_p): New.
2426 (ia64_const_double_ok_for_letter_p): New.
2427 (ia64_extra_constraint): New.
2428 (ia64_expand_vecint_compare): New.
2429 (ia64_expand_vcondu_v2si): New.
2430 (ia64_expand_vecint_cmov): New.
2431 (ia64_expand_vecint_minmax): New.
2432 (ia64_print_operand): Add 'v'.
2433 (ia64_preferred_reload_class): New.
2434 (ia64_vector_mode_supported_p): New.
2435 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
2436 (PREFERRED_RELOAD_CLASS): Move to function.
2437 (CONST_OK_FOR_LETTER_P): Move to function.
2438 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
2439 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
2440 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
2441 (EXTRA_CONSTRAINT): Move to function.
2442 * config/ia64/ia64.md: Include vect.md.
2443 (itanium_class): Add mmalua.
2445 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
2447 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
2448 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
2450 (const_int_2bit_operand): New.
2451 (fr_reg_or_0_operand): New.
2452 * config/ia64/ia64-modes.def: Add vector modes.
2453 * config/ia64/ia64-protos.h: Update.
2454 * config/ia64/vect.md: New file.
2456 2005-01-03 Richard Henderson <rth@redhat.com>
2458 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
2460 2005-01-03 Uros Bizjak <uros@kss-loka.si>
2463 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
2465 (log1pdf2): Change mode of operands[0,1] to DFmode.
2467 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
2469 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
2470 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
2471 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
2472 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
2473 SPARC_RELAXED_ORDERING.
2475 2005-01-03 Richard Henderson <rth@redhat.com>
2476 Uros Bizjak <uros@kss-loka.si>
2479 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
2480 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
2481 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
2483 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
2484 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
2485 const_pow2_1_to_128_operand): New predicates.
2487 2005-01-02 Greg McGary <greg@mcgary.org>
2489 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
2491 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
2492 object. [ptr_type]: Remove unused variable.
2493 Remember nearest addressable array-element or record-component
2494 when checking bitfield components. Tolerate empty BIND_EXPR.
2496 2005-01-01 Richard Henderson <rth@redhat.com>
2498 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
2499 that don't narrow the value. Fail for other conversions.
2501 2005-01-01 Richard Henderson <rth@redhat.com>
2504 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
2505 * c-lang.c (finish_file): Don't do it here.
2506 * objc/objc-act.c (objc_finish_file): Likewise.
2508 * cgraph.c (decl_assembler_name_equal): New.
2509 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
2510 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
2511 * cgraph.h (struct cgraph_varpool_node): Add next.
2512 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
2513 * varasm.c (assemble_alias): Mark the target as needed.
2515 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
2518 * function.c (get_arg_pointer_save_area): Use entry_of_function
2519 instead of get_insns.
2521 2005-01-01 Roger Sayle <roger@eyesopen.com>
2522 Andrew Pinski <pinskia@physics.uc.edu>
2523 James E. Wilson <wilson@specifixinc.com>
2525 PR rtl-optimization/12092
2526 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
2529 2005-01-01 Roger Sayle <roger@eyesopen.com>
2530 Olivier Hainque <hainque@act-europe.fr>
2532 * tree.c (int_fits_type_p): A narrower type always fits in a
2533 wider one, except for negative values into unsigned types.
2535 2005-01-01 Roger Sayle <roger@eyesopen.com>
2537 * tree.c (int_fits_type_p): Compare the result of force_fit_type
2538 with the original constant rather than require TREE_OVERFLOW.
2540 2005-01-01 Steven Bosscher <stevenb@suse.de>
2543 * c-decl.c (finish_function): If compiling C99, annotate the
2544 compiler generated return with the current file name and line 0.
2545 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
2546 the source line is greater than 0.
2547 (remove_bb): Likewise.
2549 See ChangeLog.12 for earlier changes.