1 2005-01-26 Jakub Jelinek <jakub@redhat.com>
3 PR rtl-optimization/19579
4 * ifcvt.c (noce_try_cmove_arith): If emitting instructions to set up
5 both A and B, see if they don't clobber registers the other expr uses.
7 2005-01-25 J"orn Rennecke <joern.rennecke@st.com>
9 * real.c (do_add): Initialize signalling and canonical members.
11 * real.c (real_from_integer): Zero out destination.
13 2005-01-18 Jan Hubicka <jh@suse.cz>
15 PR tree-optimize/19337
16 * tree-inline.c (initialize_inlined_parameters): Copy
17 saved_static_chain_decl only.
19 2005-01-25 Richard Henderson <rth@redhat.com>
23 * config/i386/i386.c (x86_inter_unit_moves): Zero.
24 (ix86_preferred_reload_class): Rewrite fp-constant section, with
25 80387 enabled, to return a proper subclass. Return the subset that
26 overlaps with GENERAL_REGS for PLUS.
27 * config/i386/i386.md (movsi_1, movdi_2): Set type to "mmx" for pxor.
29 2005-01-25 Richard Henderson <rth@redhat.com>
32 * tree-complex.c (expand_complex_div_wide): Use the correct formulae.
34 2005-01-25 Roger Sayle <roger@eyesopen.com>
36 * expmed.c (choose_multiplier): Change interface to return the
37 multiplier lowpart as an rtx instead of an unsigned HOST_WIDE_INT.
38 (expand_divmod): Update calls to choose_multiplier.
40 2005-01-24 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
43 * expmed.c (extract_bit_field): Use adjust_address instead of
44 gen_lowpart when op0 is a MEM.
46 2005-01-24 Steven Bosscher <stevenb@suse.de>
48 * cfgexpand.c (expand_gimple_tailcall): Fix typo.
50 2005-01-24 Richard Sandiford <rsandifo@redhat.com>
52 * config/mips/irix-csr.c: New file.
53 * config/mips/t-iris6 (irix-csr.o): New rule to build it.
54 (EXTRA_MULTILIB_PARTS): Add irix-csr.o.
55 * config/mips/iris6.h (ENDFILE_SPEC): Include it in n32 and n64
58 2005-01-24 Eric Botcazou <ebotcazou@libertysurf.fr>
61 * config.gcc (sparc-*-elf*): Remove sol2.h, sparc/sol2.h and
62 sparc/elf.h, add sparc/sp-elf.h.
63 (sparc-*-rtems*): Likewise.
64 (sparclite-*-elf*): Remove sol2.h, sparc/sol2.h, sparc/elf.h and
65 tm-dwarf2.h, add sparc/sp-elf.h.
66 (sparc86x-*-elf): Likewise.
67 (sparc64-*-elf*): Remove sol2.h, sparc/sol2.h and tm-dwarf2.h.
68 * config/sparc/liteelf.h (TARGET_SUB_OS_CPP_BUILTINS): Rename into
69 TARGET_OS_CPP_BUILTINS.
70 * config/sparc/sp86x-elf (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
71 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
73 * config/sparc/openbsd64.h (NO_IMPLICIT_EXTERN_C): Undefine.
74 * config/sparc/sp64-elf.h (NO_IMPLICIT_EXTERN_C): New macro.
75 (SWITCH_TAKES_ARG): Likewise.
76 (LOCAL_LABEL_PREFIX): Likewise.
77 (ASM_GENERATE_INTERNAL_LABEL): Likewise.
78 (TARGET_N_FORMAT_TYPES): Delete.
79 (TARGET_FORMAT_TYPES): Likewise.
80 (ASM_DECLARE_FUNCTION_SIZE): Likewise.
81 * config/sparc/elf.h: Delete.
82 * config/sparc/sp-elf.h: New file.
84 2005-01-24 Kazu Hirata <kazu@cs.umass.edu>
86 * tree-flow-inline.h (phi_arg_from_edge): Remove.
87 * tree-flow.h: Remove the corresponding prototype.
88 * tree-ssa-operands.h, tree-ssa-threadupdate.c, tree-ssa.c,
89 tree-vectorizer.c: Use dest_idx instead of phi_arg_from_edge.
91 2005-01-24 Jakub Jelinek <jakub@redhat.com>
94 * config/rs6000/rs6000.c (rs6000_file_start): Create toc section
95 for AIX ABI or ELF -fPIC.
96 (rs6000_emit_load_toc_table): Don't create toc_section here.
97 (rs6000_xcoff_file_start): Nor here.
99 2005-01-24 J"orn Rennecke <joern.rennecke@st.com>
101 * sh.c (ra.h): Don't #include.
102 (hard_regs_intersect_p): New function, resurrected from ra.c.
104 * sh.c: Fix 1996 Copyright.
106 2005-01-24 Richard Henderson <rth@redhat.com>
107 Aldy Hernandez <aldyh@redhat.com>
109 * regrename.c (note_sets): Handle subregs.
111 2005-01-24 Joseph S. Myers <joseph@codesourcery.com>
114 * genconditions.c (write_header, write_conditions): Elide file if
117 2005-01-24 Kazu Hirata <kazu@cs.umass.edu>
119 * c-opts.c, expmed.c, gengtype.c, tree-complex.c,
120 config/arm/iwmmxt.md, config/avr/avr.c, config/s390/s390.h:
123 2005-01-24 Jakub Jelinek <jakub@redhat.com>
125 * builtins.c (get_pointer_alignment, gimplify_va_arg_expr,
126 expand_builtin_printf, expand_builtin_fprintf,
127 expand_builtin_sprintf): Use POINTER_TYPE_P instead TREE_CODE
128 checking against POINTER_TYPE.
129 (validate_arglist): Handle POINTER_TYPE code by checking
132 2005-01-24 Paolo Bonzini <bonzini@gnu.org>
134 * rtlanal.c (non_rtx_starting_operands, for_each_rtx_1,
136 (for_each_rtx): Call for_each_rtx_1.
137 * rtl.h (init_rtlanal): Declare.
138 * toplev.c (backend_init): Call init_rtlanal.
140 2005-01-24 Jakub Jelinek <jakub@redhat.com>
142 * flow.c (propagate_one_insn): Formatting.
145 * flow.c (libcall_dead_p): Be more conservative if unsure.
146 If there are any instructions between insn and call, see if they are
147 all dead before saying the libcall is dead.
149 2005-01-24 Uros Bizjak <uros@kss-loka.si>
151 * config/i386/i386.md (*extendsfdf2_sse): Use "nonimmediate_operand"
152 operand constraint for operand 0.
154 2005-01-23 Richard Henderson <rth@redhat.com>
157 * tree-complex.c (expand_complex_div_wide): Set EDGE_FALLTHRU on
158 edges leading to join block.
160 2005-01-23 Roger Sayle <roger@eyesopen.com>
162 * expmed.c (expand_mult_highpart): Make static. Change type of
163 constant multiplier argument from unsigned HOST_WIDE_INT to rtx.
164 (expand_divmod): Updates calls to expand_mult_highpart by using
165 gen_int_mode to pass a CONST_INT rtx.
166 * rtl.h (expand_mult_highpart): Remove prototype.
168 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
170 * flow.c (regset_bytes, regset_size): Remove.
172 * cfganal.c: Fix a reference to Harvey's paper.
174 2005-01-23 Steven Bosscher <stevenb@suse.de>
176 PR rtl-optimization/19464
177 * tree-optimize.c (init_tree_optimization_passes): Add one more
178 copyrename pass just before out-of-ssa.
180 2005-01-23 Joseph S. Myers <joseph@codesourcery.com>
183 * recog.c (recog_memoized): Don't define if GENERATOR_FILE.
184 * ggc-none.c (ggc_free): Define.
186 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
188 * cse.c (max_reg, max_insn_uid): Remove.
189 (cse_main): Don't access max_reg or max_insn_uid.
191 * c-common.c (explicit_flag_signed_bitfields,
192 lang_statement_code_p, lang_expand_function_end): Remove.
193 * c-common.h: Remove the corresponding declarations.
194 * c-opts.c (c_common_handle_option): Remove all write access
195 to explicit_flag_signed_bitfields.
197 * cgraph.c (cgraph_varpool_n_nodes): Remove.
198 (cgraph_varpool_node): Don't access cgraph_varpool_n_nodes.
199 * cgraph.h: Remove the corresponding declaration.
201 * gcse.c (null_pointer_info): Remove.
203 2005-01-23 Roger Sayle <roger@eyesopen.com>
204 Eric Botcazou <ebotcazou@libertysurf.fr>
206 * combine.c (simplify_logical): Only simplify logical expressions
207 of the form ior(and(x,y),z) by the inverse distributive law if the
208 result is cheaper than the original.
210 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
212 * alias.c, c-common.h, c-incpath.c, c-incpath.h, expr.c,
213 fold-const.c, gimplify.c, params.h, tree-data-ref.c,
214 tree-if-conv.c, tree-nested.c, tree-outof-ssa.c,
215 tree-ssa-dom.c, tree-vectorizer.c, tree.def, config/darwin.c,
216 config/freebsd-spec.h, config/arm/arm.h,
217 config/h8300/h8300.md, config/i386/i386.md,
218 config/i386/predicates.md, config/i386/sse.md,
219 config/ia64/ia64.c, config/ip2k/ip2k.c, config/s390/s390.c,
220 config/vax/vax.md: Fix comment typos. Follow spelling
223 2005-01-23 Steven Bosscher <stevenb@suse.de>
225 * gengtype.c (walk_type): Produce `not equal to' compares for loop
226 tests, instead of `less than'.
228 2005-01-23 Richard Sandiford <rsandifo@redhat.com>
230 * function.c (assign_parm_setup_block): When creating a new stack slot
231 for a parameter, get its alignment from the parameter's DECL_ALIGN
232 rather than the type's TYPE_ALIGN. Make sure that the parameter
233 is at least word aligned.
235 2005-01-22 Roger Sayle <roger@eyesopen.com>
238 * config/avr/avr.c (avr_hard_regno_mode_ok): Rewrite.
240 2005-01-22 Richard Henderson <rth@redhat.com>
243 * config/i386/i386.md (movsfcc_1_sse_max): Use nonimmediate_operand
244 in both compare operands.
245 (movdfcc_1_sse_max): Likewise.
246 (movsfcc_1_sse): Likewise. Add earlyclobber for scratch.
247 (movdfcc_1_sse): Likewise.
248 * config/i386/i386.c (ix86_split_sse_movcc): Emit copies into the
249 scratch register as needed.
251 2005-01-22 Richard Henderson <rth@redhat.com>
253 * config/i386/i386.md (smaxsf3): Fix mnemonic typo.
255 * config/i386/i386.c (ix86_prepare_fp_compare_args): Fix is_sse test.
257 2005-01-22 Richard Henderson <rth@redhat.com>
259 * genopinit.c (optabs): Use smin/smax for floating point too.
260 * doc/md.texi: Update to match. Clarify that floating point
261 results are undefined for +0/-0 and NaN.
262 * doc/rtl.texi: Likewise.
263 * rtl.def (SMIN, SMAX): Likewise
264 * tree.def (MIN_EXPR, MAX_EXPR): Likewise.
266 * config/alpha/alpha.md (smaxdf3, smindf3, smaxsf3, sminsf3): Add
267 leading 's' to the name.
268 * config/ia64/ia64.md (smaxsf3, sminsf3, smaxdf3, smindf3,
269 smaxxf3, sminxf3): Likewise.
270 * config/rs6000/rs6000.md (smaxdf3, smindf3, smaxsf3, sminsf3):
273 2005-01-22 Paul Brook <paul@codesourcery.com>
275 * config/arm/arm.md: Use "Uy" constraint for wcgr load/stores.
277 2005-01-21 Mark Dettinger <dettinge@de.ibm.com>
279 * config/s390/s390.c (struct processor_costs): New fields
280 dlgr, dlr, dr, dsgfr, dsgr.
281 (z900_cost, z990_cost): Values for new fields.
282 (s390_rtx_costs): New cases MEM und COMPARE in switch
283 statement. Modified handling of SIGN_EXTEND, ZERO_EXTEND,
284 DIV, MOD, UDIV, UMOD.
286 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
288 * config/s390/s390.md ("doloop_si64"): Reload input value directly
289 into the register being decremented.
290 ("doloop_si31", "doloop_di"): Likewise.
291 ("*doloop_si_long"): Adapt pattern.
293 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
295 * config/s390/s390.h (HARD_REGNO_NREGS): Fix computation for
297 (CLASS_MAX_NREGS): Likewise.
299 2005-01-21 Daniel Berlin <dberlin@dberlin.org>
301 * doc/tree-ssa.texi (Statement Operands): Add example for new
302 must-def macro. Note deprecation of old operands interface.
304 2005-01-22 Richard Sandiford <rsandifo@redhat.com>
306 PR tree-optimization/19484
307 * tree-cfg.c (remove_fallthru_edge): New function.
308 (cleanup_control_flow): Remove fallthru edges from calls that are
309 now known not to return.
311 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
313 Fix PR tree-optimization/19038
314 * tree-ssa-dom.c (cprop_operand): Don't replace loop invaeriant
315 copies with loop variant ones.
317 2005-01-22 Kazu Hirata <kazu@cs.umass.edu>
319 * cfganal.c, real.h, reorg.c, timevar.def, tree-ssa-ccp.c,
320 config/alpha/alpha-protos.h, config/alpha/alpha.h,
321 config/alpha/alpha.md, config/alpha/predicates.md,
322 config/sparc/freebsd.h, config/sparc/netbsd-elf.h,
323 config/sparc/sol2.h: Update copyright.
325 * tree-cfg.c (remove_forwarder_block_with_phi): Look at the
326 first label to see if it is a nonlocal label.
328 2005-01-22 David Edelsohn <edelsohn@gnu.org>
329 Andrew Pinski <pinskia@physics.uc.edu>
332 * config/rs6000/rs6000.c (rs6000_va_start): Saturate n_gpr at
333 maximum number of GPRs. Saturate n_fpr at maximum number of FPRs.
335 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
337 * config/rs6000/rtems.h: Update copyright.
339 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
342 * config/rs6000/rtems.h: Resurrect cpp_os_rtems_spec from gcc < 3.4.
343 (CPP_OS_RTEMS_SPEC): New (From gcc-3.3's config/rs6000/sys4.h).
344 (SUBSUBTARGET_EXTRA_SPECS): Use CPP_OS_RTEMS_SPEC.
346 2005-01-22 Volker Reichelt <reichelt@igpm.rwth-aachen.de>
349 * c-typeck.c (convert_arguments): Check for error_mark_node.
351 2005-01-21 Ian Lance Taylor <ian@airs.com>
353 PR tree-optimization/13000
354 * tree-inline.c: Include "tree-flow.h".
355 (expand_call_inline): If warn_return_type, warn if non-void inline
356 function falls through.
357 * tree-cfg.c (execute_warn_function_return): Don't warn about
358 control reaching end if TREE_NO_WARNING is set. Set
360 * gimple-low.c (block_may_fallthru): Don't assume that SWITCH_EXPR
362 * gimplify.c (shortcut_cond_expr): Don't emit a jump over the else
363 branch if we don't need one.
364 * c-typeck.c: Include "tree-flow.h"
365 (c_finish_bc_stmt): Don't add a goto if the current statement
366 list doesn't fall through to the current point.
368 2005-01-21 Roger Sayle <roger@eyesopen.com>
370 PR rtl-optimization/576
371 * real.c (real_arithmetic): Change return type from void to bool
372 to return an indication that the result may be inexact.
373 * real.h (real_arithmeric): Update prototype.
374 * fold-const.c (const_binop): Don't constant fold floating
375 point expressions when the user specifies -frounding-math and
376 the result may depend upon the run-time rounding mode.
377 (fold_convert_const_real_from_real): Clean-up.
378 (fold_initializer): Ignore flag_rounding_math for initializers.
379 * simplify-rtx.c (simplify_binary_operation): Likewise, don't
380 constant fold FP operations with flag_rounding_math if the
381 result may depend upon the run-time rounding mode.
383 2005-01-21 Tom Tromey <tromey@redhat.com>
385 * c-cppbuiltin.c (define__GNUC__): Correct assertion.
387 2005-01-21 Kazu Hirata <kazu@cs.umass.edu>
389 * tree-cfg.c: Fix comment typos.
391 2005-01-21 J"orn Rennecke <joern.rennecke@st.com>
393 * passes.c (rest_of_handle_flow2): Close / open the flow2 dump file
394 around the call to rest_of_handle_branch_target_load_optimize.
396 2005-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
398 * config.gcc (sparc-*-netbsdelf*): Include dbxelf.h
399 (sparc64-*-openbsd*): Likewise.
400 (sparclite-*-elf*): Include tm-dwarf2.h.
401 (sparc86x-*-elf*): Likewise.
402 (sparc64-*-elf*): Likewise.
403 * config/sparc/linux64.h (DWARF2_DEBUGGING_INFO): Delete.
404 (DBX_DEBUGGING_INFO): Likewise.
405 * config/sparc/liteelf.h (DWARF2_DEBUGGING_INFO): Likewise.
406 (PREFERRED_DEBUGGING_TYPE): Likewise.
407 * config/sparc/netbsd-elf.h (PREFERRED_DEBUGGING_TYPE): Likewise.
408 * config/sparc/sp64-elf.h (DBX_DEBUGGING_INFO): Likewise.
409 (PREFERRED_DEBUGGING_TYPE): Likewise.
410 * config/sparc/sp86x-elf.h (DWARF2_DEBUGGING_INFO): Likewise.
411 (PREFERRED_DEBUGGING_TYPE): Likewise.
412 * config/sparc/sparc.h (DBX_DEBUGGING_INFO): Likewise.
414 2005-01-21 Hans-Peter Nilsson <hp@bitrange.com>
417 * combine.c (combine_simplify_rtx): Revert change of 2004-12-31.
419 2005-01-20 Paul Brook <paul@codesourcery.com>
421 * doc/rtl.texi: Document value extension requirements for CONST_INT.
423 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
425 * tree-cfg.c (tree_verify_flow_info): Check that a nonlocal
426 label is first in a sequence of labels.
428 * tree-cfg.c (tree_verify_flow_info): Fix a typo.
430 2005-01-20 Janis Johnson <janis187@us.ibm.com>
431 Giovanni Bajo <giovannibajo@gcc.gnu.org>
433 * doc/sourcebuild.texi (Test Directives): New.
435 2005-01-20 Roger Sayle <roger@eyesopen.com>
437 * real.c (real_floor): Don't leave the result uninitialized when
438 mode is VOIDMode, but return the unrounded intermediate value.
439 (real_ceil): Likewise.
441 2005-01-20 Paul Brook <paul@codesourcery.com>
443 * config/arm/arm.md (insv): Use gen_int_mode.
445 2005-01-20 Eric Botcazou <ebotcazou@libertysurf.fr>
447 * config/sparc/freebsd.h (ASM_OUTPUT_INTERNAL_LABELREF): Delete.
448 * config/sparc/linux.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
449 * config/sparc/linux64.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
450 * config/sparc/netbsd-elf.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
451 * config/sparc/sol2.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
453 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
456 * dwarf2out.c (concat_loc_descriptor): We don't know we can use
457 fbreg, so we have to assume we can't.
458 (loc_descriptor_from_tree_1): Ditto.
459 (containing_function_has_frame_base): New function.
460 (add_location_or_const_value_attribute): Use it.
461 Also try to generate a frame_base from a single element location
464 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
466 PR tree-optimization/15349
467 * timevar.def (TV_TREE_MERGE_PHI): New.
468 * tree-cfg.c (tree_forwarder_block_p): Add a new argument
470 (remove_forwarder_block, cleanup_forwarder_blocks): Adjust the
471 calls to tree_forwarder_block_p.
472 (remove_forwarder_block_with_phi, merge_phi_nodes,
473 gate_merge_phi, pass_merge_phi): New.
474 * tree-optimize.c (init_tree_optimization_passes): Add
476 * tree-pass.h: Add an extern for pass_merge_phi;
478 2005-01-20 Richard Henderson <rth@redhat.com>
481 * config/i386/emmintrin.h (_mm_castpd_ps, _mm_castpd_si128): New.
482 (_mm_castps_pd, _mm_castps_si128): New.
483 (_mm_castsi128_ps, _mm_castsi128_pd): New.
485 2005-01-20 Richard Henderson <rth@redhat.com>
488 * config/i386/mmintrin.h (_mm_cvtsi32_si64): Use
489 __builtin_ia32_vec_init_v2si.
490 (_mm_cvtsi64_si32): Use __builtin_ia32_vec_ext_v2si.
491 * config/i386/i386.c (IX86_BUILTIN_VEC_EXT_V2SI): New.
492 (ix86_init_mmx_sse_builtins): Create it.
493 (ix86_expand_builtin): Expand it.
494 (ix86_expand_vector_set): Handle V2SFmode and V2SImode.
495 * config/i386/mmx.md (vec_extractv2sf_0, vec_extractv2sf_1): New.
496 (vec_extractv2si_0, vec_extractv2si_1): New.
498 2005-01-20 Richard Henderson <rth@redhat.com>
501 * config/i386/i386.c (ix86_expand_push): New.
502 * config/i386/mmx.md (push<MMXMODE>1): New.
503 * config/i386/sse.md (push<SSEMODE>1): New.
504 * config/i386/i386-protos.h: Update.
506 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
508 * params.def, params.h, predict.c, tree-ssa-loop-im.c: Update
511 2005-01-20 Richard Henderson <rth@redhat.com>
513 * rtl.def (CONST_VECTOR): Use RTX_CONST_OBJ.
514 * rtl.h (CONSTANT_P): Don't special case CONST_VECTOR.
516 2005-01-19 Richard Henderson <rth@redhat.com>
519 * config/i386/i386.c (ix86_expand_vector_move_misalign): Convert
520 to V4SFmode in SSE1 fallback load path.
522 2005-01-19 Richard Henderson <rth@redhat.com>
524 * config/i386/i386.c (ix86_expand_vector_init_one_var): Fix typo
525 in QImode expansion to ix86_expand_vector_set.
526 (ix86_expand_vector_init_general): Fix typo in V8HImode recursive call.
528 2005-01-19 Richard Henderson <rth@redhat.com>
531 * config/i386/i386.c (ix86_preferred_reload_class): Return a proper
532 subclass of the input class.
533 (ix86_secondary_memory_needed): Always true for cross-MMX classes.
534 Always true for cross-SSE1 classes. Rationalize conditionals.
535 * config/i386/i386.h (SSE_CLASS_P, MMX_CLASS_P): Use straight equality.
536 * config/i386/i386.md (movsi_1): Add MMX/SSE zeros. Fix alternatives
537 for SSE1. Don't check TARGET_INTER_UNIT_MOVES.
538 (movdi_2): Add MMX/SSE zeros.
539 (movdi_1_rex64): Likewise. Don't check TARGET_INTER_UNIT_MOVES.
540 (movsf_1): Don't check TARGET_INTER_UNIT_MOVES.
541 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
542 (movsi_1_nointernunit, movdi_1_rex64_nointerunit): Remove.
543 (movsf_1_nointerunit, zero_extendsidi2_32_1): Remove.
544 (zero_extendsidi2_rex64_1): Remove.
545 (MOV0 peephole): Check GENERAL_REG_P.
547 2005-01-19 Richard Henderson <rth@redhat.com>
550 * config/i386/i386.c (ix86_expand_vector_set): Fix third and fourth
552 (ix86_expand_vector_extract): Likewise.
554 2005-01-19 Richard Henderson <rth@redhat.com>
557 * expr.c (emit_move_change_mode): New argument force; use
558 simplify_gen_subreg if true.
559 (emit_move_via_alt_mode): Merge into ...
560 (emit_move_via_integer): ... here.
561 (emit_move_ccmode): Use emit_move_change_mode directly.
563 2005-01-19 Richard Henderson <rth@redhat.com>
566 * config/alpha/alpha.c (alpha_rtx_costs): Handle HIGH.
567 (alpha_preferred_reload_class): Handle CONST_VECTOR.
568 (alpha_emit_set_const_1): Add no_output parameter; don't emit
570 (alpha_emit_set_const): Likewise. Make static.
571 (alpha_emit_set_long_const): Make static.
572 (alpha_extract_integer): Split out from alpha_expand_mov.
573 (alpha_split_const_mov): Likewise.
574 (alpha_expand_mov): Use them. Handle CONST_VECTOR.
575 (alpha_legitimate_constant_p): New.
576 * config/alpha/alpha-protos.h: Update.
577 * config/alpha/alpha.h (REGISTER_MOVE_COST): Correct fp<->gp cost.
578 (LEGITIMATE_CONSTANT_P): Re-implement with a function.
579 * config/alpha/alpha.md (movsi): Add n alternative.
580 (movsi_nt_vms, movdi_er_nofix, movdi_er_fix, movdi_fix): Likewise.
581 (mov<VEC>_fix, mov<VEC>_nofix): Add i alternative.
582 (splitters for all of the above): Use alpha_split_const_mov.
583 * config/alpha/predicates.md (non_add_const_operand): New.
584 (non_zero_const_operand): New.
585 (input_operand): Use alpha_legitimate_constant_p after reload.
587 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
589 PR tree-optimization/19038
590 * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): New function.
591 (add_candidate): Add ivs with increment in latch only if
592 allow_ip_end_pos_p is true.
593 (determine_iv_cost): Use empty_block_p.
595 2005-01-19 Daniel Berlin <dberlin@dberlin.org>
597 * cfganal.c (compute_dominance_frontiers_1): Replace with new algorithm
598 (compute_dominance_frontiers): Ditto.
600 2005-01-19 Ralf Corsepius <ralf.corsepius@rtems.org>
603 * config/sh/t-rtems: New.
604 * config.gcc (sh-*-rtems*): Reflect having added config/sh/t-rtems.
606 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
608 * tree-ssa-ccp.c (maybe_fold_offset_to_component_ref): Always subtract
609 the offset of the selected field.
611 2005-01-19 Kazu Hirata <kazu@cs.umass.edu>
613 * tree-cfg.c (remove_forwarder_block): Fix the check to
614 prevent a nonlocal label from appearing in the middle of a
617 2005-01-19 Hans-Peter Nilsson <hp@axis.com>
619 PR rtl-optimization/19462
620 * reorg.c (find_end_label): Create return insn only if
621 current_function_epilogue_delay_list is empty.
623 2005-01-19 Paolo Bonzini <bonzini@gnu.org>
625 * doc/passes.texi: Remove paragraph mentioning new-ra.
627 2005-01-19 Jakub Jelinek <jakub@redhat.com>
629 PR rtl-optimization/15139
630 * combine.c: Include params.h.
631 (count_rtxs): New function.
632 (record_value_for_reg): If replace_rtx would replace at least
633 2 occurrences of REG in VALUE and TEM is really large, replace REG with
634 (clobber (const_int 0)) instead of TEM.
635 * params.def (PARAM_MAX_LAST_VALUE_RTL): New.
636 * params.h (MAX_LAST_VALUE_RTL): New.
637 * Makefile.in (combine.o): Depend on $(PARAMS_H).
638 * doc/invoke.texi (--param max-last-value-rtl=N): Document.
641 * c-typeck.c (digest_init): Only call build_vector if all constructor
642 elements are *_CST nodes.
643 * gimplify.c (gimplify_init_constructor): Likewise.
646 * c-typeck.c (digest_init): Only call build_vector if inside_init
649 2005-01-18 Tobias Schl"uter <tobias.schlueter@physik.uni-muenchen.de>
651 * toplev.c (init_asm_output): Remove dead #ifdef.
653 2005-01-18 Kaz Kojima <kkojima@gcc.gnu.org>
655 * config/sh/linux.h (TARGET_C99_FUNCTIONS): Define.
657 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
659 * calls.c (expand_call): Check DECL_BUILT_IN_CLASS before
660 accessing DECL_FUNCTION_CODE.
661 * dojump.c (do_jump): Likewise.
662 * gimplify.c (gimplify_call_expr): Likewise.
663 * predict.c (expr_expected_value): Likewise.
664 (strip_builtin_expect): Likewise.
665 * tree-inline.c (estimate_num_insns_1): Likewise.
666 * tree-ssa-loop-im.c (stmt_cost): Likewise
667 * fold-const.c (fold): Test for BUILT_IN_NORMAL.
668 (tree_expr_nonnegative_p): Likewise.
670 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
672 * config/sparc/sparc.c (load_pic_register): Emit the appropriate
673 variant of the load_pcrel_sym pattern.
674 * config/sparc/sparc.md (P macro): Move to the top.
675 (load_pcrel_sym): Macroize using P.
676 (save_register_window): Likewise.
678 2005-01-18 Aldy Hernandez <aldyh@redhat.com>
680 * config/rs6000/rs6000.md ("sunordered"): Disable for e500.
683 2005-01-18 Andrew Pinski <pinskia@physics.uc.edu>
686 * c-typeck.c (build_asm_expr): Strip nops off
687 input memory operands.
689 2005-01-18 David Edelsohn <edelsohn@gnu.org>
691 * config/rs6000/aix43.h (CPLUSPLUS_CPP_SPEC): Delete
692 _XOPEN_SOURCE, _XOPEN_SOURCE_EXTENDED, _LARGE_FILE_API.
693 * config/rs6000/aix51.h (CPLUSPLUS_CPP_SPEC): Same.
694 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Same.
696 2005-01-18 Joel Sherrill <joel.sherrill@oarcorp.com>
699 * config/i386/i386.c (override_options): If the 80387 is disabled,
700 then do not return FP values using FP registers.
702 2005-01-18 Kazu Hirata <kazu@cs.umass.edu>
704 * basic-block.h, c-common.c, c-cppbuiltin.c, c-lang.c,
705 c-tree.h, cfgbuild.c, cgraph.c, cgraph.h, collect2.c,
706 combine.c, config.gcc, coverage.h, cse.c, cselib.c,
707 defaults.h, df.c, dwarf2asm.c, dwarf2out.c, explow.c, expr.c,
708 flow.c, fold-const.c, gcse.c, ggc-page.c, gimple-low.c,
709 gimplify.c, ifcvt.c, langhooks-def.h, lcm.c, optabs.h,
710 output.h, postreload-gcse.c, postreload.c, recog.c,
711 resource.c, rtl.def, rtlanal.c, sched-deps.c, sched-rgn.c,
712 targhooks.h, toplev.c, tree-data-ref.c, tree-eh.c,
713 tree-flow.h, tree-loop-linear.c, tree-mudflap.h, tree-nrv.c,
714 tree-optimize.c, tree-outof-ssa.c, tree-pass.h,
715 tree-scalar-evolution.c, tree-ssa-copy.c, tree-ssa-dce.c,
716 tree-ssa-dse.c, tree-ssa-loop-ivopts.c, tree-ssa-operands.c,
717 tree-ssa-pre.c, tree-ssa.c, tree-vectorizer.c, tree.def,
718 unwind-dw2-fde-darwin.c, var-tracking.c: Update copyright.
720 * config/darwin.c, config/darwin.h, config/freebsd-spec.h,
721 config/alpha/alpha.c, config/arm/vfp.md,
722 config/i386/emmintrin.h, config/i386/i386-modes.def,
723 config/i386/i386-protos.h, config/i386/i386.h,
724 config/i386/pmmintrin.h, config/i386/ppro.md,
725 config/i386/predicates.md, config/i386/xmmintrin.h,
726 config/i860/i860.c, config/ia64/ia64-modes.def,
727 config/ia64/ia64-protos.h, config/ia64/ia64.h,
728 config/ia64/ia64.md, config/ia64/itanium1.md,
729 config/ia64/itanium2.md, config/ia64/predicates.md,
730 config/iq2000/iq2000.h, config/mips/linux64.h,
731 config/rs6000/aix41.h, config/rs6000/aix43.h,
732 config/rs6000/aix52.h, config/rs6000/darwin-fallback.c,
733 config/rs6000/rs6000.c, config/rs6000/rs6000.h,
734 config/rs6000/spe.md, config/sh/sh.md, config/sparc/linux.h,
735 config/sparc/linux64.h, config/sparc/litecoff.h,
736 config/sparc/sp64-elf.h, config/sparc/sparc.c,
737 config/sparc/sparc.h, config/sparc/sparc.md,
738 config/sparc/sysv4-only.h: Update copyright.
740 2005-01-18 Richard Henderson <rth@redhat.com>
743 * config/i386/i386.c (ix86_expand_fp_movcc): Fail for LTGT and UNEQ.
745 2005-01-18 Marc Espie <espie@openbsd.org>
746 * config/i386/openbsdelf.h: Typo.
748 2005-01-18 Richard Henderson <rth@redhat.com>
750 * config/i386/i386.c (ix86_expand_fp_movcc): Remove TARGET_IEEE_FP
751 special case for sse.
752 * config/i386/predicates.md (sse_comparison_operator): Likewise.
754 2005-01-18 Dorit Naishlos <dorit@il.ibm.com>
756 * tree-inline.c (estimate_num_insns_1): Added cases for
757 ALIGN_INDIRECT_REF, MISALIGNED_INDIRECT_REF, and REALIGN_LOAD_EXPR.
759 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
761 PR rtl-optimization/19296
762 * combine.c (simplify_comparison): Rewrite the condition under
763 which a non-paradoxical SUBREG of a PLUS can be lifted when
764 compared against a constant.
766 2005-01-18 Andi Kleen <ak@muc.de>
768 * c-typeck.c: (convert_for_assignment): Check warn_pointer_sign.
769 * c.opt (-Wpointer-sign): Add.
770 * doc/invoke.texi: (-Wpointer-sign): Add.
772 2005-01-18 Uros Bizjak <uros@kss-loka.si>
774 * config/i386/i386.c (override_options): Revert 2004-11-24 change.
775 * config/i386/i386.md (fmodsf3, dremsf3, *sinsf2, *cossf2, sincossf3,
776 *tansf3_1, tansf2, atan2sf3_1, atan2sf3, atansf2, asinsf2, acossf2,
777 logsf2, log10sf2, log2sf2, log1psf2, logbsf2, ilogbsf2, expsf2,
778 exp10sf2, exp2sf2, expm1sf2, rintsf2, floorsf2, ceilsf2, btruncsf2,
779 nearbyintsf2): Disable for TARGET_SSE_MATH. Leave patterns enabled
780 for TARGET_MIX_SSE_I387.
781 (fmoddf3, dremdf3, *sindf2, *sinextendsfdf2, *cosdf2, *cosextendsfdf2,
782 sincosdf3, *sincosextendsfdf3, *tandf3_1, tandf2, atan2df3_1,
783 atan2df3, atandf2, asindf2, acosdf2, logdf2, log10df2, log2df2,
784 log1pdf2, logbdf2, expdf2, exp10df2, exp2df2, expm1df2, rintdf2,
785 floordf2, ceildf2, btruncdf2, nearbyintdf2): Disable for
786 (TARGET_SSE2 && TARGET_SSE_MATH). Leave patterns enabled for
788 (atan2sf3, atan2df3, atan2xf3): Remove register constraints
791 2005-01-18 Uros Bizjak <uros@kss-loka.si>
794 * config/i386/mmx.md (*movv2sf_internal_rex64, *movv2sf_internal):
795 Add movaps alternative for xmm reg->reg move.
797 2005-01-17 Jeff Law <law@redhat.com>
799 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Clear SSA_NAME_VALUE
800 at the end of the main DOM loop rather than just before DOM exits.
802 2005-01-17 Diego Novillo <dnovillo@redhat.com>
804 PR tree-optimization/19121
805 * tree-ssa-alias.c (compute_flow_sensitive_aliasing): When
806 adding aliases to a name tag, also add them to the pointer's
808 * tree-ssa-copy.c (merge_alias_info): Do not merge flow
809 sensitive alias info at all. Only check that the two pointers
810 have compatible pointed-to sets.
811 * tree-ssa.c (verify_name_tags): Verify that the alias set of
812 a pointer's type tag is a superset of the alias set of the
815 2005-01-17 James E Wilson <wilson@specifixinc.com>
818 * config/ia64/ia64.md (movxf): Handle general register source. Adjust
819 comment to document why.
821 2005-01-17 Richard Henderson <rth@redhat.com>
823 * config/i386/sse.md (smaxv4sf3_finite, sse_vmsmaxv4sf3_finite,
824 sminv4sf3_finite, sse_vmsminv4sf3_finite, smaxv2df3_finite,
825 sse2_vmsmaxv2df3_finite, sminv2df3_finite,
826 sse2_vmsminv2df3_finite): New.
827 (smaxv4sf3, sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, smaxv2df3,
828 sse2_vmsmaxv2df3, sminv2df3, sse2_vmsminv2df3): Remove commutative.
829 Force op1 into register.
831 2005-01-17 Kelley Cook <kcook@gcc.gnu.org>
833 * Makefile.in (STAGEMOVESTUFF): Stage all the stamp files to prevent
834 bubblestrap from always rebuilding build directory.
836 2005-01-17 Ulrich Weigand <uweigand@de.ibm.com>
838 * config/s390/s390.md ("*clc"): Remove incorrect '='.
839 ("*cmpmem_short"): Likewise.
841 2005-01-17 Steven Bosscher <stevenb@suse.de>
843 * basic-block.h: Document BB_* flags.
844 * regrename.c (copyprop_hardreg_forward): Don't use BB_VISITED,
845 use an sbitmap instead.
846 * sched-rgn.c (compute_trg_info): Likewise.
848 2005-01-17 Richard Sandiford <rsandifo@redhat.com>
850 * config.gcc (mips64*-*-linux*): Set the default abi to n32. Remove
851 redundant target_cpu_default and extra_parts lines. Use the default
853 * config/mips/linux64.h (DRIVER_DEFAULT_ABI_SELF_SPEC): Delete.
854 (SUBTARGET_EXTRA_SPECS): Delete.
855 (DRIVER_SELF_SPECS): Remove architecture lines. Set the ABI directly.
856 (SUBTARGET_ASM_SPEC): Remove -mabi=n32 and -mabi=64 mappings.
857 * config/mips/t-linux64 (MULTILIB_DIRNAMES): Use "32" for -mabi=32
858 and "n32" for -mabi=n32.
860 2005-01-17 Ian Lance Taylor <ian@airs.com>
863 * c-typeck.c (build_c_cast): Revert patch of 2003-02-16: don't
864 fold constant variables into initial values.
866 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
868 * tree-cfg.c (tree_can_merge_blocks_p): Reorder two checks.
870 * tree-cfg.c (tree_forwarder_block_p): Speed up by walking
871 through the statements backward.
873 2005-01-17 Ian Lance Taylor <ian@airs.com>
876 * tree-inline.c (expand_call_inline): Set TREE_NO_WARNING on
877 a variable set to the return value of the inlined function.
879 2005-01-17 Mark Dettinger <dettinge@de.ibm.com>
881 * rtlanal.c (rtx_cost): Assign cost of 0 to a SUBREG
882 when modes are tieable.
884 2005-01-17 Ranjit Mathew <rmathew@hotmail.com>
886 * gthr-posix.h (__gthread_active_p): Use pthread_cancel instead
887 of pthread_create to find out if threads are enabled.
888 * gthr-posix95.h (__gthread_active_p): Likewise.
890 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
892 * common.opt (-fnew-ra): Remove.
894 * toplev.h (flag_new_regalloc): Remove.
895 * Makefile.in (ra*.*): Don't mention.
896 * passes.c (rest_of_handle_new_regalloc): Remove.
897 (rest_of_handle_combine, rest_of_compilation): Always consider
898 flag_new_regalloc as false.
899 * doc/invoke.texi: Don't document -fnew-ra.
901 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
903 * bb-reorder.c (fix_edges_for_rarely_executed_code): Remove
904 last parameter to reg_scan.
905 * loop.c (loop_optimize): Likewise.
906 * passes.c (rest_of_handle_tracer, rest_of_handle_if_conversion,
907 rest_of_handle_web, rest_of_handle_cfg, rest_of_handle_jump_bypass,
908 rest_of_handle_life, rest_of_handle_cse, rest_of_handle_cse2,
909 rest_of_handle_gcse, rest_of_handle_loop_optimize,
910 rest_of_handle_loop2, rest_of_handle_jump2): Likewise.
911 * regclass.c (reg_scan): Likewise, for the declaration.
912 * rtl.h (reg_scan): Likewise, for the prototype.
914 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
916 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Don't call
919 2005-01-16 Daniel Berlin <dberlin@dberlin.org>
921 * tree-ssa-pre.c (add_to_sets): s1 may be NULL.
922 (compute_avail): Uses don't go in tmp_gen.
924 2005-01-16 Steven Bosscher <stevenb@suse.de>
926 * ggc-page.c (ggc_alloc_stat): Use __builtin_ctzl instead of a
927 loop to look for a free slot in a page entry.
929 2005-01-16 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
932 * defaults.h (TARGET_DEFERRED_OUTPUT_DEFS): Provide default.
933 * toplev.c (compile_file): Call process_pending_assemble_output_defs
934 just before targetm.asm_out.file_end.
935 * tree.h (process_pending_assemble_output_defs): Declare.
936 * varasm.c (assemble_output_def, process_pending_assemble_output_defs):
938 (assemble_alias): Defer generation of assembly code for defines when
939 TARGET_DEFERRED_OUTPUT_DEFS is true.
940 * config/rs6000/aix41.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
941 * config/rs6000/aix43.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
942 * doc/tm.texi (TARGET_DEFERRED_OUTPUT_DEFS): document.
944 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
947 * pa.c (pa_scalar_mode_supported_p): New function.
948 (TARGET_SCALAR_MODE_SUPPORTED_P): Define.
949 * pa.h (MIN_UNITS_PER_WORD): Add comment.
951 * pa-protos.h (prefetch_operand): Delete.
952 (prefetch_cc_operand, prefetch_nocc_operand): New declations.
953 * pa.c (prefetch_operand): Delete.
954 (prefetch_cc_operand, prefetch_nocc_operand): New functions.
955 * pa.h (EXTRA_CONSTRAINT): Add `W' constraint.
956 (PREDICATE_CODES): Delete prefetch_operand. Add prefetch_cc_operand
957 and prefetch_nocc_operand.
958 * pa.md (prefetch): Rework to avoid reload problems handling short
959 displacements when a cache control completer needs to be provided.
960 (prefetch_32, prefetch_64): Delete.
961 (prefetch_cc, prefetch_nocc): New patterns.
963 2005-01-15 David Edelsohn <edelsohn@gnu.org>
965 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Revert previous change.
967 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
970 * pa.c (pa_asm_output_aligned_common): Warn if specified alignment
971 exceeds maximum alignment for global common data.
972 * pa64-hpux.h (MAX_OFILE_ALIGNMENT): Define to 32768.
973 * som.h (MAX_OFILE_ALIGNMENT): Likewise.
975 2004-01-15 Roger Sayle <roger@eyesopen.com>
977 * tree-ssa-dom.c (extract_range_from_cond): Correct condition.
979 2004-01-15 Roger Sayle <roger@eyesopen.com>
981 * harg-reg-set.h (reg_class_names): Prototype global array.
982 * regclass.c (reg_class_names): Declare here and initialize to
984 (dump_regclass): Remove local declaration of reg_class_names.
985 (regclass): Likewise.
986 * cfg.c (dump_flow_info): Likewise.
987 * ra-debug.c (reg_class_names): Likewise.
988 * regrename.c (reg_class_names): Likewise.
989 * reload.c (reg_class_names): Likewise.
990 * reload1.c (spill_failure): Likewise.
991 * config/m68hc11/m68hc11.c (reg_class_names): Likewise.
993 2005-01-15 Ulrich Weigand <uweigand@de.ibm.com>
995 * config/s390/s390.md ("reload_outti"): Remove predicate for
996 output operand. Abort if operand is not a MEM.
997 ("reload_outdi", "reload_outdf"): Likewise.
999 2005-01-15 Marc Espie <espie@openbsd.org>
1001 * config.gcc (*-*-openbsd*): Set HAS_LIBC_R for OpenBSD <= 3.2.
1002 (i[34567]86-*-openbsd*): Switch to ELF for OpenBSD >= 3.4.
1003 * config/openbsd.h: Fix C++ includes for native configurations.
1004 Add proper OS_CPP_BUILTINS. Fix libspec for recent OpenBSD.
1005 Add trampoline support.
1006 * config/i386/openbsdelf.h: New.
1008 2005-01-15 Marc Espie <espie@openbsd.org>
1010 * collect2.c (main): Explicitly parse -dynamic-linker option.
1012 2005-01-15 Jakub Jelinek <jakub@redhat.com>
1014 PR tree-optimization/19060
1015 * tree-ssa-dom.c (extract_range_from_cond) <case LT_EXPR, GT_EXPR>:
1016 Return 0 if op1 <= TYPE_MIN_VALUE () resp. op1 >= TYPE_MAX_VALUE ().
1017 (simplify_cond_and_lookup_avail_expr): Add assert for dummy == 0
1018 and handle extract_range_from_cond returning false.
1019 * fold-const.c (fold): Optimize comparisons with min/max even for
1020 width > HOST_BITS_PER_WIDE_INT.
1022 2005-01-15 Ralf Corsepius <ralf.corsepius@rtems.org>
1024 * config/mips/rtems.h (MIPS_DEFAULT_GVALUE): Set to 0.
1025 * config/mips/t-rtems (MULTILIBS_DIRNAMES,MULTILIB_OPTIONS):
1026 Remove little endian multilib variants.
1027 Add mips32 multilib variant.
1029 2005-01-15 Kazu Hirata <kazu@cs.umass.edu>
1031 * explow.c (copy_all_reg, stabilize): Remove.
1032 * expr.h: Remove the corresponding prototypes.
1034 * rtlanal.c (regs_set_between_p): Remove.
1035 * rtl.h: Remove the corresponding prototype.
1037 2005-01-14 David Edelsohn <edelsohn@gnu.org>
1039 * config/rs6000/rs6000.c (rs6k_nonimmediate_operand): Rename
1040 to rs6000_nonimmediate_operand.
1041 * config/rs6000/rs6000.h (PREDICATE_CODES): Rename
1042 rs6k_nonimmediate_operand to rs6000_nonimmediate_operand.
1043 * config/rs6000/rs6000.md (movsi_internal1): Rename predicate to
1044 rs6000_nonimmediate_operand.
1045 * config/rs6000/spe.md (movdf_e500_double): Rename predicate to
1046 rs6000_nonimmediate_operand.
1048 2005-01-14 Andrew Pinski <pinskia@physics.uc.edu>
1050 * tree-ssa-alias.c (pass_may_alias): Add TODO_verify_stmts.
1052 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1054 * config/rs6000/rs6000.h (CLASS_MAX_NREGS): DF goes in 1 register
1056 (CANNOT_CHANGE_MODE_CLASS): Restrict DI mode changes on e500v2.
1057 (PREDICATE_CODES): Add rs6k_nonimmediate_operand.
1059 * config/rs6000/rs6000.c (invalid_e500_subreg): New.
1060 (rs6k_nonimmediate_operand): New.
1061 (rs6000_legitimate_offset_address_p): Handle DI modes on e500v2
1063 (legitimate_lo_sum_address_p): Same.
1064 (rs6000_legitimize_address): Same.
1065 (rs6000_legitimize_reload_address): Same.
1066 (rs6000_legitimate_address): Same.
1067 (spe_build_register_parallel): Pass DF and DC modes in a DI
1070 * config/rs6000/rs6000.md ("*movsi_internal1"): Change predicate
1071 to rs6k_nonimmediate_operand.
1073 * config/rs6000/spe.md ("*frob_df_di"): New.
1074 ("*frob_di_df"): New.
1075 ("*frob_di_df_2"): New.
1076 ("*mov_sidf_e500_subreg0"): New.
1077 ("*mov_sidf_e500_subreg4"): New.
1078 ("*movdf_e500_double"): Change predicate to
1079 rs6k_nonimmediate_operand.
1081 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1083 * postreload.c (move2add_note_store): Only call
1084 trunc_int_for_mode on scalar integers.
1086 2005-01-14 Steven Bosscher <stevenb@suse.de>
1088 * tree-ssa-dce.c (visited_control_parents): New sbitmap to
1089 replace BB_VISITED uses.
1090 (find_obviously_necessary_stmts): Don't clear BB_VISITED.
1091 (propagate_necessity): Check the bitmap instead of BB_VISITED.
1092 (tree_dce_done): Free visited_control_parents.
1093 (perform_tree_ssa_dce): Allocate and clear it.
1094 * tree-ssa-pre.c (compute_antic_aux): Make non-recursive.
1095 (compute_antic): Iterate from here using a DFS. Use an sbitmap
1096 instead of BB_VISITED.
1098 2005-01-14 Kazu Hirata <kazu@cs.umass.edu>
1100 * c-tree.h, coverage.h, langhooks-def.h, optabs.h, output.h,
1101 rtl.h, targhooks.h, tree-flow.h, tree-mudflap.h, tree.h:
1102 Remove unused prototypes.
1104 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1106 * config/rs6000/rs6000.c (rs6000_stack_info): Fix caching of SPE
1107 64-bit register usage.
1108 (spe_func_has_64bit_regs_p): Add FIXME note.
1110 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1112 * config/rs6000/rs6000.c (rs6000_generate_compare): Replace
1113 flag_finite_math_only with flag_unsafe_math_optimizations.
1115 2005-01-14 J. D. Johnston <jjohnst@us.ibm.com>
1117 * config/s390/tpf-unwind.h (s390_fallback_frame_state): Correct
1119 (__tpf_eh_return): Copy TPF private stack area from the module
1120 boundary stack frame. Add check for when module addresses are
1121 equal, but no stub address is found.
1123 2005-01-14 Richard Earnshaw <rearnsha@arm.com>
1126 * arm.h (struct machine_function): Add call_via field.
1127 (thumb_call_via_label): Declare.
1128 * arm.c (thumb_call_via_label): New variable.
1129 (thumb_call_reg_needed): New variable.
1130 (arm_output_function_epilogue): For Thumb code, output any per-function
1131 call-indirect trampolines.
1132 (thumb_call_via_reg): New function.
1133 (arm_file_end): New function.
1134 (TARGET_ASM_FILE_END): Call arm_file_end.
1135 (aof_file_end): Likewise.
1136 * arm-protos.h (thumb_call_via_reg): Declare.
1137 * arm.md (call_reg_thumb, call_value_reg_thumb): Call
1138 thumb_call_via_reg in normal case.
1140 2005-01-14 Jakub Jelinek <jakub@redhat.com>
1143 PR rtl-optimization/19348
1144 * recog.c (peephole2_optimize): Do global life update if some peephole
1145 decides it doesn't need at least one of its inputs and that change
1146 influences liveness at the start of the basic block.
1148 * basic-block.h (EXECUTE_IF_AND_COMPL_IN_REG_SET): Needs 2 REGSET
1149 arguments instead of 1.
1151 2005-01-14 Eric Botcazou <ebotcazou@libertysurf.fr>
1154 * varasm.c (initializer_constant_valid_p) <ADDR_EXPR>: Return
1155 zero for nested functions needing a static chain or functions
1156 with a non-constant address.
1158 2005-01-13 Roger Sayle <roger@eyesopen.com>
1160 * simplify-rtx.c (simplify_binary_operation) <AND>: Optimize
1161 (and (sign_extend X) C) into (zero_extend (and X C)).
1163 2005-01-13 David O'Brien <obrien@FreeBSD.org>
1165 * config/freebsd-spec.h: Make KSE pthread lib logic the default.
1167 2005-01-13 Richard Henderson <rth@redhat.com>
1172 * config/i386/i386.md (cmpdf, cmpsf, bunordered, bordered, buneq,
1173 bunge, bungt, bunle, bunlt, bltgt): Enable for TARGET_SSE_MATH,
1174 not just TARGET_SSE.
1175 (cmpfp_i_387): Rename from cmpfp_i. Move after sse patterns.
1176 (cmpfp_i_mixed): Rename from cmpfp_i_sse; use for TARGET_MIX_SSE_I387.
1177 (cmpfp_i_sse): Rename from cmpfp_i_sse_only; use for TARGET_SSE_MATH.
1178 (cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387): Similarly.
1179 (fp_jcc_1_mixed, fp_jcc_1_sse, fp_jcc_1_387): Similarly.
1180 (fp_jcc_2_mixed, fp_jcc_2_sse, fp_jcc_2_387): Similarly.
1181 (fp_jcc_3_387, fp_jcc_4_387, fp_jcc_5_387, fp_jcc_6_387,
1182 fp_jcc_7_387, fp_jcc_8_387): Rename from fp_jcc_N.
1183 (movdicc_c_rex64): Rename with '*'.
1184 (movsfcc, movdfcc): Add checks for 387 and sse math to condition.
1185 (movsfcc_1_sse_min, movsfcc_1_sse_max, movsfcc_1_sse): New.
1186 (movsfcc_1_387): Rename from movsfcc_1.
1187 (movdfcc_1_sse_min, movdfcc_1_sse_max, movdfcc_1_sse): New.
1188 (movdfcc_1, movdfcc_1_rex64): Add check for 387.
1189 (sminsf3, smaxsf3, smindf3, smaxdf3): New.
1190 (minsf3, minsf, minsf_nonieee, minsf_sse, mindf3, mindf,
1191 mindf_nonieee, mindf_sse, maxsf3, maxsf, maxsf_nonieee, maxsf_sse,
1192 maxdf3, maxdf, maxdf_nonieee, maxdf_sse, sse_movsfcc, sse_movsfcc_eq,
1193 sse_movdfcc, sse_movdfcc_eq, sse_movsfcc_const0_1,
1194 sse_movsfcc_const0_2, sse_movsfcc_const0_3, sse_movsfcc_const0_4,
1195 sse_movdfcc_const0_1, sse_movdfcc_const0_2, sse_movdfcc_const0_3,
1196 sse_movdfcc_const0_4): Remove.
1197 * config/i386/i386.c (ix86_expand_fp_movcc): For TARGET_SSE_MATH,
1198 recognize min/max early. Update for changed sse cmove patterns.
1199 (ix86_split_sse_movcc): New.
1200 * config/i386/i386-protos.h: Update.
1202 2005-01-13 Steven Bosscher <stevenb@suse.de>
1204 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
1205 (fix_stmt_v_may_defs): Likewise.
1207 2005-01-13 Richard Henderson <rth@redhat.com>
1209 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
1210 for converting the mask.
1212 2005-01-13 David Edelsohn <edelsohn@gnu.org>
1214 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
1217 2005-01-13 Richard Henderson <rth@redhat.com>
1219 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
1220 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
1221 IX86_BUILTIN_MOVDQ2Q): Remove.
1222 (IX86_BUILTIN_VEC_EXT_V4SI): New.
1223 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
1224 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
1226 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
1227 (sse2_pextrw): Fix immediate constraint.
1228 (sse2_loadq, sse2_loadq_rex64): Remove.
1229 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
1230 Use __builtin_ia32_vec_ext_<size>.
1231 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
1233 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
1235 * function.c (assign_parm_setup_block): Look inside original
1236 entry_parm when inspecting PARALLEL.
1238 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
1239 Joel Sherrill <joel@oarcorp.com>
1242 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
1243 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
1244 rtems_gxx_recursive_mutex_init.
1245 (__gthread_recursive_mutex_lock): New function.
1246 (__gthread_recursive_mutex_trylock): Likewise.
1247 (__gthread_recursive_mutex_unlock): Likewise.
1249 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
1251 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
1253 2005-01-13 Jan Beulich <jbeulich@novell.com>
1255 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
1256 (ashrti3_internal): Indicate output is early clobber. Generate result
1257 into output rather than first input. Use move for low word of output
1258 if shift count is exactly 64.
1259 (lshrti3_internal): Likewise.
1261 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
1265 * reload1.c (delete_output_reload): Don't delete an output reload
1266 if the pseudo lives longer than a single basic block. Adjust and
1267 improve wording of comment.
1269 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
1272 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
1273 of TYPE_SIZE where we mean the number of bytes.
1275 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
1277 * function.c (assign_parm_setup_block): Relax condition on
1278 multi-register optimization.
1280 2005-01-12 Nick Clifton <nickc@redhat.com>
1282 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
1283 an "in_delay_slot" attribute of "no" to prevent them being used in
1284 delay slots. This is forbidden because they might generate
1287 2005-01-12 Alan Modra <amodra@bigpond.net.au>
1290 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
1293 2005-01-12 Nick Clifton <nickc@redhat.com>
1295 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
1296 passed on to GAS) but do not define, as GAS no longer supports or
1297 needs the -m2000 option.
1299 2005-01-11 Roger Sayle <roger@eyesopen.com>
1301 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
1302 type as args of unprototyped builtins aren't automatically promoted.
1303 (fold_builtin_abs): Likewise.
1305 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1307 * Makefile.in: Set a `build-warn' variable.
1309 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
1311 PR tree-optimization/17949
1312 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
1313 (find_interesting_uses_address): Use it.
1315 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
1317 * regrename.c (kill_value): Handle subreg's that won't simplify.
1319 2005-01-11 Richard Henderson <rth@redhat.com>
1322 * config/i386/i386.h (enum ix86_builtins): Move ...
1323 * config/i386/i386.c: ... here.
1324 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
1325 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
1326 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
1327 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
1328 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
1329 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
1330 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
1331 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
1332 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
1333 IX86_BUILTIN_LOADDDUP): Remove.
1334 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
1335 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
1336 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
1337 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
1338 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
1339 (ix86_init_builtins): Make static.
1340 (ix86_init_mmx_sse_builtins): Update for changed builtins.
1341 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
1342 if all the modes match. Otherwise, fake it.
1343 (get_element_number, ix86_expand_vec_init_builtin,
1344 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
1345 (ix86_expand_builtin): Make static. Update for changed builtins.
1346 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
1347 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
1348 (ix86_expand_vector_init_duplicate): New.
1349 (ix86_expand_vector_init_low_nonzero): New.
1350 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
1351 Split out from ix86_expand_vector_init; handle integer modes.
1352 (ix86_expand_vector_init): Use them.
1353 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
1354 * config/i386/i386-protos.h: Update.
1355 * config/i386/predicates.md (reg_or_0_operand): New.
1356 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
1357 (movv2sf_internal): Likewise. And a splitter to match them all.
1358 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
1359 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
1360 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
1361 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
1362 (mmx_pinsrw): Fix operand ordering.
1363 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
1364 rather than sse_loadss expander.
1365 (movv2df splitter): Similarly.
1366 (sse_loadss, sse_loadlss): Remove.
1367 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
1368 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
1369 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
1370 (sse3_movddup): Rename with '*'.
1371 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
1372 (sse2_loadsd): Remove.
1373 (vec_dupv2df_sse3): Rename from sse3_loadddup.
1374 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
1375 (sse2_pinsrw): Fix argument ordering.
1376 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
1377 (sse2_stored): Remove 'r' destination.
1378 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
1379 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
1380 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
1381 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
1382 vec_extractv16qi, vec_initv16qi): New.
1384 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
1385 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
1386 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
1388 (_mm_load_pd, _mm_store_pd): Use plain dereference.
1389 (_mm_load_si128, _mm_store_si128): Likewise.
1390 (_mm_load1_pd): Use _mm_set1_pd.
1391 (_mm_load_sd): Use _mm_set_sd.
1392 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
1393 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
1394 (_mm_set_epi64): Use _mm_set_epi64x.
1395 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
1396 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
1397 _mm_setr_epi8): Use _mm_set_foo form.
1398 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
1399 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
1400 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
1401 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
1402 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
1403 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
1404 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
1405 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
1406 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
1407 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
1408 (_mm_movedup_pd): Use _mm_shuffle_pd.
1409 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
1410 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
1411 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
1412 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
1414 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
1415 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
1416 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
1417 (_mm_store_ps): Use raw dereference.
1418 (_mm_store1_ps): Use _mm_storeu_ps.
1419 (_mm_storer_ps): Use _mm_store_ps.
1420 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
1421 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
1423 2005-01-11 Stan Shebs <shebs@apple.com>
1425 * config/rs6000/rs6000.c (machopic_output_stub): Issue
1426 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
1428 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
1431 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
1432 Skip all DECLs except for FIELD_DECLs.
1434 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
1436 * config/s390/s390.c (override_options): Return error if
1437 -mbackchain, -mpacked-stack and -mhard-float are used together.
1438 (s390_va_start): Remove the backchain && packed-stack special case.
1439 (s390_gimplify_va_arg): Likewise.
1440 * doc/invoke.texi: Remove the ABI incompatibility note.
1442 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
1444 * config/s390/s390.c (struct s390_frame_layout): Remove
1446 (s390_frame_info, s390_emit_prologue): Replace occurrences of
1447 save_backchain_p with TARGET_BACKCHAIN.
1449 2005-01-11 Alan Modra <amodra@bigpond.net.au>
1452 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
1453 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
1454 * function.c (locate_and_pad_parm): Set new field.
1455 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
1456 Tweak where_pad test to include "none". Always set mem align for
1458 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
1459 not sufficient for type.
1460 (assign_parm_setup_block): If stack_parm is zero on entry, always
1461 make a new stack local. Block move old stack parm if necessary
1462 to new aligned stack local.
1463 (assign_parm_setup_stack): Use a block move to handle
1464 potentially misaligned entry_parm.
1465 (assign_parms_unsplit_complex): Specify required alignment when
1466 creating stack local.
1467 * calls.c (compute_argument_addresses): Override alignment of stack
1468 arg calculated from its type with the alignment given by
1469 FUNCTION_ARG_BOUNDARY.
1470 (store_one_arg): Likewise.
1472 2005-01-11 Jan Beulich <jbeulich@novell.com>
1474 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
1475 Change respective itanium_class attribute to ialu.
1476 (shladdp4_internal): New.
1477 * config/ia64/predicates.md (shladd_log2_operand): New.
1479 2005-01-11 Richard Henderson <rth@redhat.com>
1481 * expr.c (store_constructor): Use rtvec_alloc instead of
1482 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
1484 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
1486 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
1488 2005-01-11 Alan Modra <amodra@bigpond.net.au>
1490 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
1491 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
1492 (TARGET_RWRELOC): Define.
1493 (ia64_rwreloc_section_type_flags): Delete.
1494 (ia64_section_type_flags): New function.
1495 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
1496 (TARGET_RWRELOC): Define.
1498 2005-01-10 David Mosberger <davidm@hpl.hp.com>
1501 * config/ia64/ia64.c (process_set): For alloc insn, only call
1502 process_epilogue is !frame_pointer_needed.
1504 2005-01-10 Roger Sayle <roger@eyesopen.com>
1507 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
1508 unary operator and can't be treated as a binary/comparison operator.
1510 2005-01-10 Richard Henderson <rth@redhat.com>
1512 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
1513 (ix86_return_in_memory): Likewise.
1514 (function_arg_advance): Likewise. Mirror structure in function_arg
1515 for choosing register to advance.
1517 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
1519 * tree-vectorizer.c, tree.def: Fix comment typos.
1521 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
1523 * config/s390/s390.c (struct processor_costs): 4 new fields:
1524 ddbr, ddr, debr, der.
1525 (s390_rtx_costs): More precise handling of divide instructions.
1527 2005-01-10 David Edelsohn <edelsohn@gnu.org>
1529 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
1531 2005-01-10 Jan Beulich <jbeulich@novell.com>
1533 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
1534 names resulting from -ffunction-sections/-fdata-sections and linkonce
1536 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
1537 based on the section name. Rearrange the section name comparison logic
1538 slightly so that each section name is compared against at most once.
1540 2005-01-10 Ben Elliston <bje@au.ibm.com>
1542 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
1543 targets for which -fPIC reduces limitations on the GOT size.
1545 2005-01-09 Falk Hueffner <falk@debian.org>
1547 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
1550 2005-01-09 David Edelsohn <edelsohn@gnu.org>
1553 * collect2.c (main): Set aixrtl_flag for -brtl option.
1554 (resolve_lib_name): Search for .so file extension before .a
1557 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
1559 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
1560 (vect_analyze_data_refs_alignment): Add dump prints.
1562 2005-01-09 Ira Rosen <irar@il.ibm.com>
1564 * tree-vectorizer.c (vect_analyze_offset_expr): Use
1565 expr_invariant_in_loop_p.
1566 Initialize outputs first thing in the function.
1567 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
1568 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
1569 (vect_analyze_pointer_ref_access): Check that the initial condition of
1570 the access function is loop invariant.
1572 2005-01-09 Richard Henderson <rth@redhat.com>
1574 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
1575 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
1576 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
1577 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
1578 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
1579 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
1580 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
1581 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
1582 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
1583 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
1584 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
1585 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
1586 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
1587 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
1588 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
1589 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
1590 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
1591 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
1592 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
1593 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
1594 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
1595 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
1596 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
1598 (mmx_clrdi, pavgusb): Remove.
1599 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
1600 with leading sse_ prefix.
1601 * config/i386/sse.md: Receive them.
1602 * config/i386/mmx.md: New file.
1603 (MMXMODE12, MMXMODE24, mmxvecsize): New.
1604 (subrv2sf3): Turn into expander for normal subtraction.
1605 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
1606 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
1607 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
1608 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
1609 commutative; use ix86_binary_operator_ok.
1610 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
1611 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
1612 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
1613 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
1614 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
1615 Macroize from existing patterns; use ix86_binary_operator_ok.
1616 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
1617 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
1618 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
1619 with vec_select+vec_concat.
1620 (mmx_pshufw, mmx_pshufw_1): Likewise.
1621 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
1622 (mmx_uavgv4hi3): Model correctly.
1623 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
1624 _mm_xor_si64): Remove casts.
1626 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
1628 PR tree-optimization/19224
1629 * tree-scalar-evolution.c (get_instantiated_value,
1630 set_instantiated_value): New functions.
1631 (instantiate_parameters_1): Cache the results.
1632 (instantiate_parameters, resolve_mixers): Initialize and free
1635 2005-01-08 David Edelsohn <edelsohn@gnu.org>
1637 * config/i386/i386.md (addhi_4): Correct reference in comment.
1640 2005-01-08 Richard Henderson <rth@redhat.com>
1642 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
1643 (_mm_cvtsi128_si64x): Likewise.
1644 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
1645 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
1646 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
1647 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
1648 * config/i386/i386-protos.h: Update.
1649 * config/i386/i386.c (print_operand): Add 'H'.
1650 (ix86_fixup_binary_operands): Split out from ...
1651 (ix86_expand_binary_operator): ... here.
1652 (ix86_fixup_binary_operands_no_copy): New.
1653 (ix86_expand_fp_absneg_operator): Handle vector mode results.
1654 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
1655 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
1656 (safe_vector_operand): Use CONST0_RTX.
1657 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
1658 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
1659 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
1660 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
1662 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
1663 (IX86_BUILTIN_CMPNGESS): New.
1664 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
1665 (attr type): Add sselog1.
1666 (attr unit, attr memory): Handle it.
1667 (movti, movti_internal, movti_rex64): Move near other integer moves.
1668 (movtf, movtf_internal): Move near other fp moves.
1669 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
1670 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
1671 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
1672 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
1673 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
1674 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
1675 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
1676 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
1677 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
1678 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
1679 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
1680 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
1681 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
1682 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
1683 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
1684 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
1685 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
1686 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
1687 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
1688 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
1689 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
1690 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
1691 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
1692 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
1693 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
1694 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
1695 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
1696 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
1697 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
1698 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
1699 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
1700 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
1701 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
1702 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
1703 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
1704 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
1705 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
1706 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
1707 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
1708 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
1709 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
1710 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
1711 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
1712 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
1714 (SSEPUSH, push<SSEPUSH>): Remove.
1715 (MMXPUSH, push<MMXPUSH>): Remove.
1716 (sse_movaps, sse_movaps_1, sse_movups): Remove.
1717 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
1718 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
1719 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
1720 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
1721 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
1722 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
1723 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
1724 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
1725 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
1726 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
1727 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
1728 (ppro_sse_log_V4SF): Handle sselog1.
1729 * config/i386/predicates.md (const_0_to_1_operand): New.
1730 (const_0_to_255_mul_8_operand): New.
1731 (const_1_to_31_operand): Rename from const_int_1_31_operand.
1732 (const_2_to_3_operand, const_4_to_7_operand): New.
1733 * config/i386/sse.md: New file.
1734 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
1735 (sse_movups): Rename from sse_movups_1.
1736 (sse_loadlss): Rename from sse_loadss_1.
1737 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
1739 (negv4sf2): Use ix86_expand_fp_absneg_operator.
1740 (absv4sf2, negv2df, absv2df): New.
1741 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
1742 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
1743 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
1744 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
1745 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
1746 (sse3_addsubv4sf3): Model correctly.
1747 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
1748 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
1749 (sse_movhlps): Model with vec_select+vec_concat.
1750 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
1751 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
1752 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
1753 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
1754 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
1755 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
1756 sse2_pshufhw_1): Likewise.
1757 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
1758 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
1759 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
1760 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
1761 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
1762 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
1763 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
1764 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
1765 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
1766 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
1767 sminv8hi3): Mark commutative
1768 operands. Use ix86_binary_operator_ok.
1769 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
1770 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
1771 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
1772 sse2_punpcklqdq): Allow operand2 in memory.
1773 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
1774 sse2_movsd): Add memory alternatives.
1775 (sse_storelps): Turn expander into an insn; split after reload.
1776 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
1777 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
1779 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
1781 * configure.ac (DWARF-2 debug_line): Use objdump.
1782 * configure: Regenerate.
1784 2005-01-08 Jeff Law <law@redhat.com>
1785 Diego Novillo <dnovillo@redhat.com>
1787 PR tree-optimization/18241
1788 * tree-nrv.c (tree_nrv): Ignore volatile return values.
1789 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
1790 statements with volatile operands.
1791 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
1792 operands after marking a statement with has_volatile_ops.
1794 2005-01-08 Roger Sayle <roger@eyesopen.com>
1796 * tree.c (int_fits_type_p): Always honor integer constant
1797 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
1799 2005-01-08 Roger Sayle <roger@eyesopen.com>
1801 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
1802 the latent bug in rtl_delete_block has been resolved.
1804 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
1806 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
1807 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
1809 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
1811 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
1812 library as @multilib_dir@/$(SHLIB_SONAME).
1814 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
1816 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
1817 * configure: Regenerate.
1819 2005-01-07 Jakub Jelinek <jakub@redhat.com>
1821 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
1824 2005-01-07 David Edelsohn <edelsohn@gnu.org>
1827 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
1828 Convert non-word aligned offset address using ld/std into
1831 2005-01-07 Richard Henderson <rth@redhat.com>
1833 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
1835 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
1837 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
1839 2005-01-07 Jakub Jelinek <jakub@redhat.com>
1841 PR tree-optimization/19283
1842 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
1843 is not shorter than the original type.
1845 PR rtl-optimization/19012
1846 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
1848 PR rtl-optimization/18861
1849 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
1850 (FULL_STATE): Define.
1851 (mark_tablejump_edge): New function.
1852 (purge_dead_tablejump_edges): New function.
1853 (find_bb_boundaries): Use it.
1855 PR tree-optimization/18828
1856 * builtins.c (expand_builtin_next_arg): Remove argument and all
1857 the argument checking.
1858 (expand_builtin): Adjust caller.
1859 (expand_builtin_va_start): Likewise. Remove error for too many
1861 (fold_builtin_next_arg): Issue error for too many arguments.
1862 After checking arguments, replace them with magic arguments that
1863 prevent further checking of the args.
1865 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1867 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
1869 2005-01-06 Roger Sayle <roger@eyesopen.com>
1872 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
1873 and %{.S:...} (and their negative variants) to test whether the
1874 input file is assembler or pre-processed-assembler independent of
1875 the actual filename extension.
1877 2005-01-06 Roger Sayle <roger@eyesopen.com>
1879 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
1880 of sign or zero extended values.
1882 2005-01-06 Geoffrey Keating <geoffk@apple.com>
1884 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
1885 for IBM long double format correctly.
1887 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
1889 Fix PR tree-optimization/18792
1891 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
1892 to first_loop_depth, and use loop depth instead of loop number.
1893 (build_classic_dir_vector): Ditto.
1894 (compute_data_dependences_for_loop): Use depth, not loop number.
1895 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
1896 number. Pass in loops, instead of loop numbers.
1897 (gather_interchange_stats): Ditto.
1898 (linear_transform_loops): Ditto.
1900 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
1903 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
1904 biv_fits_mode_p, extension_within_bounds_p): New functions.
1905 (check_ext_dependent_givs): Use them.
1907 2005-01-06 Roger Sayle <roger@eyesopen.com>
1909 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
1910 more than one barrier, in which case we should delete them all.
1912 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
1914 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
1916 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
1918 Fix PR middle-end/19286
1920 * dwarf2out.c (gen_subprogram_die): If we've already tried to
1921 output this subprogram, simply ignore this attempt to do it again.
1922 (add_abstract_origin_attribute): Don't abort trying to add the abstract
1923 origin attribute if it's not possible.
1924 (gen_block_die): Don't ignore subblocks of "unused" blocks.
1925 (decls_for_scope): Ditto.
1926 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
1927 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
1929 2005-01-05 Richard Henderson <rth@redhat.com>
1932 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
1933 (bdesc_2arg): Use it.
1934 (ix86_expand_binop_builtin): Force operands into registers
1936 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
1937 ix86_expand_sse_compare, ix86_expand_sse_comi,
1938 ix86_expand_builtin): Likewise.
1940 2005-01-05 Richard Henderson <rth@redhat.com>
1942 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
1944 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
1945 (mulv4hi3): Set itanium_class mmmul.
1946 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
1947 (fpack): Rename from fpack_sfsf.
1949 2005-01-05 Richard Henderson <rth@redhat.com>
1952 * reload1.c (do_input_reload): Restrict the optimization deleteing
1953 a previous output reload to RELOAD_FOR_INPUT.
1955 2005-01-05 Steven Bosscher <stevenb@suse.de>
1957 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
1958 comment that we fall through after case.
1959 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
1960 (distribute_links): Likewise.
1961 * cse.c (cse_insn): Likewise.
1962 * cselib.c (cselib_invalidate_mem): Likewise.
1963 * df.c: Update comments at the top of the file.
1964 (read_modify_subreg_p): Update comments here too.
1965 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
1966 * flow.c (mark_set_1): Likewise.
1967 (mark_used_regs): Likewise.
1968 * gcse.c (mems_conflict_for_gcse_p): Likewise.
1969 (canon_list_insert): Likewise.
1970 (mark_set): Likewise.
1971 (try_replace_reg): Likewise.
1972 (store_killed_in_insn): Likewise.
1973 * loop.c (count_one_set): Likewise.
1974 (basic_induction_var): Likewise.
1975 * postreload-gcse.c (find_mem_conflicts): Likewise.
1976 * postreload.c (reload_combine_note_store): Likewise.
1977 (move2add_note_store): Likewise.
1978 * reload.c (find_equiv_reg): Likewise.
1979 (mark_referenced_resources): Likewise.
1980 * rtlanal.c (set_noop_p): Likewise.
1981 (note_stores): Likewise.
1982 (note_uses): Likewise.
1983 * sched-deps.c (sched_analyze_1): Likewise.
1984 * sched-rgn.c (check_live_1): Likewise.
1985 (update_live_1): Likewise.
1986 * config/i860/i860.c: Likewise.
1988 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
1990 (ZERO_EXTRACT): Mention that this one can be an lvalue.
1992 * doc/rtl.texi: Update documentation for bit-fields and SET.
1994 2005-01-05 Roger Sayle <roger@eyesopen.com>
1996 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
1998 2005-01-05 Jan Hubicka <jh@suse.cz>
1999 Richard Henderson <rth@redhat.com>
2002 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
2005 2005-01-05 Richard Henderson <rth@redhat.com>
2008 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
2009 (sse_shufps): Change operand 3 to const_int_operand.
2010 (sse2_storelps): Fix typo in template.
2011 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
2012 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
2013 (ix86_expand_builtin): Likewise.
2015 2005-01-05 Julian Brown <julian@codesourcery.com>
2017 * config/arm/arm.c (arm_return_in_memory): Treat complex types
2018 as aggregates for AAPCS ABIs.
2020 2005-01-05 Stan Shebs <shebs@apple.com>
2022 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
2023 Mach-O getters if ppc64.
2024 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
2026 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
2027 signed four-byte field for global code case.
2028 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
2030 2005-01-05 Roger Sayle <roger@eyesopen.com>
2033 * c-common.c: Include real.h.
2034 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
2035 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
2036 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
2037 * Makefile.in (c-common.o): Update dependencies.
2039 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
2041 * c-parse.in (asm_string): Add trailing semicolon.
2043 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
2045 * c-parse.in (asm_string): New. Don't allow wide strings in
2047 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
2048 asm_string instead of STRING.
2050 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
2052 * c-typeck.c (constructor_no_implicit): Remove.
2053 (set_designator, process_init_element): Don't check
2054 constructor_no_implicit.
2056 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
2057 Kaz Kojima <kkojima@gcc.gnu.org>
2060 * lcm.c (create_pre_exit): New.
2061 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
2062 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
2064 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
2066 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
2067 -msoft-float in target_float_switch.
2068 * arm.c (arm_override_options): Fix processing of target_float_switch.
2070 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
2072 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
2075 2004-01-05 Uros Bizjak <uros@kss-loka.si>
2077 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
2078 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
2080 2005-01-04 Roger Sayle <roger@eyesopen.com>
2082 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
2083 if deleting the then-block allows the test-block to fallthru to the
2086 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
2089 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
2090 after the full declaration if the are in two different TUs.
2092 2005-01-04 Richard Henderson <rth@redhat.com>
2095 * tree-sra.c (generate_one_element_init): Just
2096 call gimplify_and_add.
2097 (generate_element_init): Record the
2098 new referenced variables and mark them for renaming
2099 and split out to ...
2100 (generate_element_init_1): This.
2101 (scalarize_init): Don't call push_gimplify_context/
2102 pop_gimplify_context.
2104 2005-01-04 Geoffrey Keating <geoffk@apple.com>
2106 * toplev.c (get_src_pwd): Handle failure of getpwd().
2108 2005-01-04 Roger Sayle <roger@eyesopen.com>
2110 * fold-const.c (fold_single_bit_test): Delete unreachable handling
2112 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
2113 all the cases handled by it are inverted by invert_truthvalue.
2115 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
2117 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
2118 (save_register_windowsi): Likewise.
2120 2005-01-04 Richard Henderson <rth@redhat.com>
2122 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
2124 2005-01-04 Uros Bizjak <uros@kss-loka.si>
2127 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
2128 any simplifications of vector mode comparison operators.
2129 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
2131 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
2132 Devang Patel <dpatel@apple.com>
2134 PR tree-optimization/18308
2135 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
2136 the operands before creating a new expression.
2137 * dojump.c (do_jump): Make drop_through_label available
2138 for all cases. Add expansion of COND_EXPR.
2140 2005-01-04 Ira Rosen <irar@il.ibm.com>
2142 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
2143 INTEGER_CST instead of TREE_CONSTANT.
2144 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
2145 instead of TREE_CONSTANT.
2146 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
2147 instead of TREE_CONSTANT.
2149 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
2151 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
2153 2005-01-04 Uros Bizjak <uros@kss-loka.si>
2156 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
2157 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
2159 2005-01-03 Richard Henderson <rth@redhat.com>
2161 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
2162 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
2163 fold_convert_const_real_from_real): Split out from ...
2164 (fold_convert_const): ... here.
2166 2005-01-03 Richard Henderson <rth@redhat.com>
2169 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
2170 (mov<MMXMODEI>_internal): Likewise.
2171 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
2172 Add V2SF case; use it for SSE1; don't use TI.
2173 (movdf_integer): Likewise.
2174 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
2176 2005-01-03 Ira Rosen <irar@il.ibm.com>
2178 * tree-vectorizer.c (vect_strip_conversions): New function.
2179 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
2180 check for binary class.
2182 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
2186 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
2188 * gimple-low.c (mark_blocks_with_used_vars): New function.
2189 (mark_blocks_with_used_subblocks): Ditto.
2190 (mark_used_blocks): Ditto.
2191 (pass_mark_used_blocks): New pass.
2192 * tree-inline.c: Include debug.h.
2193 (expand_call_inline): Call outlining_inline_function here.
2194 * tree-optimize.c (init_tree_optimization_passes): Add
2195 pass_mark_used_blocks.
2196 * tree-pass.h (pass_mark_used_blocks): New.
2197 * Makefile.in (tree-inline.o): Add debug.h dependency.
2199 2005-01-03 Geoffrey Keating <geoffk@apple.com>
2201 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
2202 VAR_DECLs to have weak_import attribute.
2204 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
2207 2004-01-03 Steven Bosscher <stevenb@suse.de>
2209 * passes.c (rest_of_compilation): Don't run regmove if only
2210 flag_expensive_optimizations, require flag_regmove instead.
2212 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
2214 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
2216 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
2217 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
2218 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
2219 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
2221 2005-01-03 Richard Henderson <rth@redhat.com>
2223 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
2224 (ia64_const_ok_for_letter_p): New.
2225 (ia64_const_double_ok_for_letter_p): New.
2226 (ia64_extra_constraint): New.
2227 (ia64_expand_vecint_compare): New.
2228 (ia64_expand_vcondu_v2si): New.
2229 (ia64_expand_vecint_cmov): New.
2230 (ia64_expand_vecint_minmax): New.
2231 (ia64_print_operand): Add 'v'.
2232 (ia64_preferred_reload_class): New.
2233 (ia64_vector_mode_supported_p): New.
2234 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
2235 (PREFERRED_RELOAD_CLASS): Move to function.
2236 (CONST_OK_FOR_LETTER_P): Move to function.
2237 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
2238 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
2239 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
2240 (EXTRA_CONSTRAINT): Move to function.
2241 * config/ia64/ia64.md: Include vect.md.
2242 (itanium_class): Add mmalua.
2244 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
2246 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
2247 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
2249 (const_int_2bit_operand): New.
2250 (fr_reg_or_0_operand): New.
2251 * config/ia64/ia64-modes.def: Add vector modes.
2252 * config/ia64/ia64-protos.h: Update.
2253 * config/ia64/vect.md: New file.
2255 2005-01-03 Richard Henderson <rth@redhat.com>
2257 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
2259 2005-01-03 Uros Bizjak <uros@kss-loka.si>
2262 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
2264 (log1pdf2): Change mode of operands[0,1] to DFmode.
2266 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
2268 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
2269 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
2270 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
2271 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
2272 SPARC_RELAXED_ORDERING.
2274 2005-01-03 Richard Henderson <rth@redhat.com>
2275 Uros Bizjak <uros@kss-loka.si>
2278 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
2279 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
2280 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
2282 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
2283 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
2284 const_pow2_1_to_128_operand): New predicates.
2286 2005-01-02 Greg McGary <greg@mcgary.org>
2288 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
2290 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
2291 object. [ptr_type]: Remove unused variable.
2292 Remember nearest addressable array-element or record-component
2293 when checking bitfield components. Tolerate empty BIND_EXPR.
2295 2005-01-01 Richard Henderson <rth@redhat.com>
2297 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
2298 that don't narrow the value. Fail for other conversions.
2300 2005-01-01 Richard Henderson <rth@redhat.com>
2303 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
2304 * c-lang.c (finish_file): Don't do it here.
2305 * objc/objc-act.c (objc_finish_file): Likewise.
2307 * cgraph.c (decl_assembler_name_equal): New.
2308 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
2309 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
2310 * cgraph.h (struct cgraph_varpool_node): Add next.
2311 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
2312 * varasm.c (assemble_alias): Mark the target as needed.
2314 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
2317 * function.c (get_arg_pointer_save_area): Use entry_of_function
2318 instead of get_insns.
2320 2005-01-01 Roger Sayle <roger@eyesopen.com>
2321 Andrew Pinski <pinskia@physics.uc.edu>
2322 James E. Wilson <wilson@specifixinc.com>
2324 PR rtl-optimization/12092
2325 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
2328 2005-01-01 Roger Sayle <roger@eyesopen.com>
2329 Olivier Hainque <hainque@act-europe.fr>
2331 * tree.c (int_fits_type_p): A narrower type always fits in a
2332 wider one, except for negative values into unsigned types.
2334 2005-01-01 Roger Sayle <roger@eyesopen.com>
2336 * tree.c (int_fits_type_p): Compare the result of force_fit_type
2337 with the original constant rather than require TREE_OVERFLOW.
2339 2005-01-01 Steven Bosscher <stevenb@suse.de>
2342 * c-decl.c (finish_function): If compiling C99, annotate the
2343 compiler generated return with the current file name and line 0.
2344 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
2345 the source line is greater than 0.
2346 (remove_bb): Likewise.
2348 See ChangeLog.12 for earlier changes.