1 2005-01-18 Jan Hubicka <jh@suse.cz>
4 * tree-inline.c (initialize_inlined_parameters): Copy
5 saved_static_chain_decl only.
7 2005-01-25 Richard Henderson <rth@redhat.com>
11 * config/i386/i386.c (x86_inter_unit_moves): Zero.
12 (ix86_preferred_reload_class): Rewrite fp-constant section, with
13 80387 enabled, to return a proper subclass. Return the subset that
14 overlaps with GENERAL_REGS for PLUS.
15 * config/i386/i386.md (movsi_1, movdi_2): Set type to "mmx" for pxor.
17 2005-01-25 Richard Henderson <rth@redhat.com>
20 * tree-complex.c (expand_complex_div_wide): Use the correct formulae.
22 2005-01-25 Roger Sayle <roger@eyesopen.com>
24 * expmed.c (choose_multiplier): Change interface to return the
25 multiplier lowpart as an rtx instead of an unsigned HOST_WIDE_INT.
26 (expand_divmod): Update calls to choose_multiplier.
28 2005-01-24 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
31 * expmed.c (extract_bit_field): Use adjust_address instead of
32 gen_lowpart when op0 is a MEM.
34 2005-01-24 Steven Bosscher <stevenb@suse.de>
36 * cfgexpand.c (expand_gimple_tailcall): Fix typo.
38 2005-01-24 Richard Sandiford <rsandifo@redhat.com>
40 * config/mips/irix-csr.c: New file.
41 * config/mips/t-iris6 (irix-csr.o): New rule to build it.
42 (EXTRA_MULTILIB_PARTS): Add irix-csr.o.
43 * config/mips/iris6.h (ENDFILE_SPEC): Include it in n32 and n64
46 2005-01-24 Eric Botcazou <ebotcazou@libertysurf.fr>
49 * config.gcc (sparc-*-elf*): Remove sol2.h, sparc/sol2.h and
50 sparc/elf.h, add sparc/sp-elf.h.
51 (sparc-*-rtems*): Likewise.
52 (sparclite-*-elf*): Remove sol2.h, sparc/sol2.h, sparc/elf.h and
53 tm-dwarf2.h, add sparc/sp-elf.h.
54 (sparc86x-*-elf): Likewise.
55 (sparc64-*-elf*): Remove sol2.h, sparc/sol2.h and tm-dwarf2.h.
56 * config/sparc/liteelf.h (TARGET_SUB_OS_CPP_BUILTINS): Rename into
57 TARGET_OS_CPP_BUILTINS.
58 * config/sparc/sp86x-elf (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
59 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
61 * config/sparc/openbsd64.h (NO_IMPLICIT_EXTERN_C): Undefine.
62 * config/sparc/sp64-elf.h (NO_IMPLICIT_EXTERN_C): New macro.
63 (SWITCH_TAKES_ARG): Likewise.
64 (LOCAL_LABEL_PREFIX): Likewise.
65 (ASM_GENERATE_INTERNAL_LABEL): Likewise.
66 (TARGET_N_FORMAT_TYPES): Delete.
67 (TARGET_FORMAT_TYPES): Likewise.
68 (ASM_DECLARE_FUNCTION_SIZE): Likewise.
69 * config/sparc/elf.h: Delete.
70 * config/sparc/sp-elf.h: New file.
72 2005-01-24 Kazu Hirata <kazu@cs.umass.edu>
74 * tree-flow-inline.h (phi_arg_from_edge): Remove.
75 * tree-flow.h: Remove the corresponding prototype.
76 * tree-ssa-operands.h, tree-ssa-threadupdate.c, tree-ssa.c,
77 tree-vectorizer.c: Use dest_idx instead of phi_arg_from_edge.
79 2005-01-24 Jakub Jelinek <jakub@redhat.com>
82 * config/rs6000/rs6000.c (rs6000_file_start): Create toc section
83 for AIX ABI or ELF -fPIC.
84 (rs6000_emit_load_toc_table): Don't create toc_section here.
85 (rs6000_xcoff_file_start): Nor here.
87 2005-01-24 J"orn Rennecke <joern.rennecke@st.com>
89 * sh.c (ra.h): Don't #include.
90 (hard_regs_intersect_p): New function, resurrected from ra.c.
92 * sh.c: Fix 1996 Copyright.
94 2005-01-24 Richard Henderson <rth@redhat.com>
95 Aldy Hernandez <aldyh@redhat.com>
97 * regrename.c (note_sets): Handle subregs.
99 2005-01-24 Joseph S. Myers <joseph@codesourcery.com>
102 * genconditions.c (write_header, write_conditions): Elide file if
105 2005-01-24 Kazu Hirata <kazu@cs.umass.edu>
107 * c-opts.c, expmed.c, gengtype.c, tree-complex.c,
108 config/arm/iwmmxt.md, config/avr/avr.c, config/s390/s390.h:
111 2005-01-24 Jakub Jelinek <jakub@redhat.com>
113 * builtins.c (get_pointer_alignment, gimplify_va_arg_expr,
114 expand_builtin_printf, expand_builtin_fprintf,
115 expand_builtin_sprintf): Use POINTER_TYPE_P instead TREE_CODE
116 checking against POINTER_TYPE.
117 (validate_arglist): Handle POINTER_TYPE code by checking
120 2005-01-24 Paolo Bonzini <bonzini@gnu.org>
122 * rtlanal.c (non_rtx_starting_operands, for_each_rtx_1,
124 (for_each_rtx): Call for_each_rtx_1.
125 * rtl.h (init_rtlanal): Declare.
126 * toplev.c (backend_init): Call init_rtlanal.
128 2005-01-24 Jakub Jelinek <jakub@redhat.com>
130 * flow.c (propagate_one_insn): Formatting.
133 * flow.c (libcall_dead_p): Be more conservative if unsure.
134 If there are any instructions between insn and call, see if they are
135 all dead before saying the libcall is dead.
137 2005-01-24 Uros Bizjak <uros@kss-loka.si>
139 * config/i386/i386.md (*extendsfdf2_sse): Use "nonimmediate_operand"
140 operand constraint for operand 0.
142 2005-01-23 Richard Henderson <rth@redhat.com>
145 * tree-complex.c (expand_complex_div_wide): Set EDGE_FALLTHRU on
146 edges leading to join block.
148 2005-01-23 Roger Sayle <roger@eyesopen.com>
150 * expmed.c (expand_mult_highpart): Make static. Change type of
151 constant multiplier argument from unsigned HOST_WIDE_INT to rtx.
152 (expand_divmod): Updates calls to expand_mult_highpart by using
153 gen_int_mode to pass a CONST_INT rtx.
154 * rtl.h (expand_mult_highpart): Remove prototype.
156 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
158 * flow.c (regset_bytes, regset_size): Remove.
160 * cfganal.c: Fix a reference to Harvey's paper.
162 2005-01-23 Steven Bosscher <stevenb@suse.de>
164 PR rtl-optimization/19464
165 * tree-optimize.c (init_tree_optimization_passes): Add one more
166 copyrename pass just before out-of-ssa.
168 2005-01-23 Joseph S. Myers <joseph@codesourcery.com>
171 * recog.c (recog_memoized): Don't define if GENERATOR_FILE.
172 * ggc-none.c (ggc_free): Define.
174 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
176 * cse.c (max_reg, max_insn_uid): Remove.
177 (cse_main): Don't access max_reg or max_insn_uid.
179 * c-common.c (explicit_flag_signed_bitfields,
180 lang_statement_code_p, lang_expand_function_end): Remove.
181 * c-common.h: Remove the corresponding declarations.
182 * c-opts.c (c_common_handle_option): Remove all write access
183 to explicit_flag_signed_bitfields.
185 * cgraph.c (cgraph_varpool_n_nodes): Remove.
186 (cgraph_varpool_node): Don't access cgraph_varpool_n_nodes.
187 * cgraph.h: Remove the corresponding declaration.
189 * gcse.c (null_pointer_info): Remove.
191 2005-01-23 Roger Sayle <roger@eyesopen.com>
192 Eric Botcazou <ebotcazou@libertysurf.fr>
194 * combine.c (simplify_logical): Only simplify logical expressions
195 of the form ior(and(x,y),z) by the inverse distributive law if the
196 result is cheaper than the original.
198 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
200 * alias.c, c-common.h, c-incpath.c, c-incpath.h, expr.c,
201 fold-const.c, gimplify.c, params.h, tree-data-ref.c,
202 tree-if-conv.c, tree-nested.c, tree-outof-ssa.c,
203 tree-ssa-dom.c, tree-vectorizer.c, tree.def, config/darwin.c,
204 config/freebsd-spec.h, config/arm/arm.h,
205 config/h8300/h8300.md, config/i386/i386.md,
206 config/i386/predicates.md, config/i386/sse.md,
207 config/ia64/ia64.c, config/ip2k/ip2k.c, config/s390/s390.c,
208 config/vax/vax.md: Fix comment typos. Follow spelling
211 2005-01-23 Steven Bosscher <stevenb@suse.de>
213 * gengtype.c (walk_type): Produce `not equal to' compares for loop
214 tests, instead of `less than'.
216 2005-01-23 Richard Sandiford <rsandifo@redhat.com>
218 * function.c (assign_parm_setup_block): When creating a new stack slot
219 for a parameter, get its alignment from the parameter's DECL_ALIGN
220 rather than the type's TYPE_ALIGN. Make sure that the parameter
221 is at least word aligned.
223 2005-01-22 Roger Sayle <roger@eyesopen.com>
226 * config/avr/avr.c (avr_hard_regno_mode_ok): Rewrite.
228 2005-01-22 Richard Henderson <rth@redhat.com>
231 * config/i386/i386.md (movsfcc_1_sse_max): Use nonimmediate_operand
232 in both compare operands.
233 (movdfcc_1_sse_max): Likewise.
234 (movsfcc_1_sse): Likewise. Add earlyclobber for scratch.
235 (movdfcc_1_sse): Likewise.
236 * config/i386/i386.c (ix86_split_sse_movcc): Emit copies into the
237 scratch register as needed.
239 2005-01-22 Richard Henderson <rth@redhat.com>
241 * config/i386/i386.md (smaxsf3): Fix mnemonic typo.
243 * config/i386/i386.c (ix86_prepare_fp_compare_args): Fix is_sse test.
245 2005-01-22 Richard Henderson <rth@redhat.com>
247 * genopinit.c (optabs): Use smin/smax for floating point too.
248 * doc/md.texi: Update to match. Clarify that floating point
249 results are undefined for +0/-0 and NaN.
250 * doc/rtl.texi: Likewise.
251 * rtl.def (SMIN, SMAX): Likewise
252 * tree.def (MIN_EXPR, MAX_EXPR): Likewise.
254 * config/alpha/alpha.md (smaxdf3, smindf3, smaxsf3, sminsf3): Add
255 leading 's' to the name.
256 * config/ia64/ia64.md (smaxsf3, sminsf3, smaxdf3, smindf3,
257 smaxxf3, sminxf3): Likewise.
258 * config/rs6000/rs6000.md (smaxdf3, smindf3, smaxsf3, sminsf3):
261 2005-01-22 Paul Brook <paul@codesourcery.com>
263 * config/arm/arm.md: Use "Uy" constraint for wcgr load/stores.
265 2005-01-21 Mark Dettinger <dettinge@de.ibm.com>
267 * config/s390/s390.c (struct processor_costs): New fields
268 dlgr, dlr, dr, dsgfr, dsgr.
269 (z900_cost, z990_cost): Values for new fields.
270 (s390_rtx_costs): New cases MEM und COMPARE in switch
271 statement. Modified handling of SIGN_EXTEND, ZERO_EXTEND,
272 DIV, MOD, UDIV, UMOD.
274 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
276 * config/s390/s390.md ("doloop_si64"): Reload input value directly
277 into the register being decremented.
278 ("doloop_si31", "doloop_di"): Likewise.
279 ("*doloop_si_long"): Adapt pattern.
281 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
283 * config/s390/s390.h (HARD_REGNO_NREGS): Fix computation for
285 (CLASS_MAX_NREGS): Likewise.
287 2005-01-21 Daniel Berlin <dberlin@dberlin.org>
289 * doc/tree-ssa.texi (Statement Operands): Add example for new
290 must-def macro. Note deprecation of old operands interface.
292 2005-01-22 Richard Sandiford <rsandifo@redhat.com>
294 PR tree-optimization/19484
295 * tree-cfg.c (remove_fallthru_edge): New function.
296 (cleanup_control_flow): Remove fallthru edges from calls that are
297 now known not to return.
299 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
301 Fix PR tree-optimization/19038
302 * tree-ssa-dom.c (cprop_operand): Don't replace loop invaeriant
303 copies with loop variant ones.
305 2005-01-22 Kazu Hirata <kazu@cs.umass.edu>
307 * cfganal.c, real.h, reorg.c, timevar.def, tree-ssa-ccp.c,
308 config/alpha/alpha-protos.h, config/alpha/alpha.h,
309 config/alpha/alpha.md, config/alpha/predicates.md,
310 config/sparc/freebsd.h, config/sparc/netbsd-elf.h,
311 config/sparc/sol2.h: Update copyright.
313 * tree-cfg.c (remove_forwarder_block_with_phi): Look at the
314 first label to see if it is a nonlocal label.
316 2005-01-22 David Edelsohn <edelsohn@gnu.org>
317 Andrew Pinski <pinskia@physics.uc.edu>
320 * config/rs6000/rs6000.c (rs6000_va_start): Saturate n_gpr at
321 maximum number of GPRs. Saturate n_fpr at maximum number of FPRs.
323 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
325 * config/rs6000/rtems.h: Update copyright.
327 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
330 * config/rs6000/rtems.h: Resurrect cpp_os_rtems_spec from gcc < 3.4.
331 (CPP_OS_RTEMS_SPEC): New (From gcc-3.3's config/rs6000/sys4.h).
332 (SUBSUBTARGET_EXTRA_SPECS): Use CPP_OS_RTEMS_SPEC.
334 2005-01-22 Volker Reichelt <reichelt@igpm.rwth-aachen.de>
337 * c-typeck.c (convert_arguments): Check for error_mark_node.
339 2005-01-21 Ian Lance Taylor <ian@airs.com>
341 PR tree-optimization/13000
342 * tree-inline.c: Include "tree-flow.h".
343 (expand_call_inline): If warn_return_type, warn if non-void inline
344 function falls through.
345 * tree-cfg.c (execute_warn_function_return): Don't warn about
346 control reaching end if TREE_NO_WARNING is set. Set
348 * gimple-low.c (block_may_fallthru): Don't assume that SWITCH_EXPR
350 * gimplify.c (shortcut_cond_expr): Don't emit a jump over the else
351 branch if we don't need one.
352 * c-typeck.c: Include "tree-flow.h"
353 (c_finish_bc_stmt): Don't add a goto if the current statement
354 list doesn't fall through to the current point.
356 2005-01-21 Roger Sayle <roger@eyesopen.com>
358 PR rtl-optimization/576
359 * real.c (real_arithmetic): Change return type from void to bool
360 to return an indication that the result may be inexact.
361 * real.h (real_arithmeric): Update prototype.
362 * fold-const.c (const_binop): Don't constant fold floating
363 point expressions when the user specifies -frounding-math and
364 the result may depend upon the run-time rounding mode.
365 (fold_convert_const_real_from_real): Clean-up.
366 (fold_initializer): Ignore flag_rounding_math for initializers.
367 * simplify-rtx.c (simplify_binary_operation): Likewise, don't
368 constant fold FP operations with flag_rounding_math if the
369 result may depend upon the run-time rounding mode.
371 2005-01-21 Tom Tromey <tromey@redhat.com>
373 * c-cppbuiltin.c (define__GNUC__): Correct assertion.
375 2005-01-21 Kazu Hirata <kazu@cs.umass.edu>
377 * tree-cfg.c: Fix comment typos.
379 2005-01-21 J"orn Rennecke <joern.rennecke@st.com>
381 * passes.c (rest_of_handle_flow2): Close / open the flow2 dump file
382 around the call to rest_of_handle_branch_target_load_optimize.
384 2005-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
386 * config.gcc (sparc-*-netbsdelf*): Include dbxelf.h
387 (sparc64-*-openbsd*): Likewise.
388 (sparclite-*-elf*): Include tm-dwarf2.h.
389 (sparc86x-*-elf*): Likewise.
390 (sparc64-*-elf*): Likewise.
391 * config/sparc/linux64.h (DWARF2_DEBUGGING_INFO): Delete.
392 (DBX_DEBUGGING_INFO): Likewise.
393 * config/sparc/liteelf.h (DWARF2_DEBUGGING_INFO): Likewise.
394 (PREFERRED_DEBUGGING_TYPE): Likewise.
395 * config/sparc/netbsd-elf.h (PREFERRED_DEBUGGING_TYPE): Likewise.
396 * config/sparc/sp64-elf.h (DBX_DEBUGGING_INFO): Likewise.
397 (PREFERRED_DEBUGGING_TYPE): Likewise.
398 * config/sparc/sp86x-elf.h (DWARF2_DEBUGGING_INFO): Likewise.
399 (PREFERRED_DEBUGGING_TYPE): Likewise.
400 * config/sparc/sparc.h (DBX_DEBUGGING_INFO): Likewise.
402 2005-01-21 Hans-Peter Nilsson <hp@bitrange.com>
405 * combine.c (combine_simplify_rtx): Revert change of 2004-12-31.
407 2005-01-20 Paul Brook <paul@codesourcery.com>
409 * doc/rtl.texi: Document value extension requirements for CONST_INT.
411 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
413 * tree-cfg.c (tree_verify_flow_info): Check that a nonlocal
414 label is first in a sequence of labels.
416 * tree-cfg.c (tree_verify_flow_info): Fix a typo.
418 2005-01-20 Janis Johnson <janis187@us.ibm.com>
419 Giovanni Bajo <giovannibajo@gcc.gnu.org>
421 * doc/sourcebuild.texi (Test Directives): New.
423 2005-01-20 Roger Sayle <roger@eyesopen.com>
425 * real.c (real_floor): Don't leave the result uninitialized when
426 mode is VOIDMode, but return the unrounded intermediate value.
427 (real_ceil): Likewise.
429 2005-01-20 Paul Brook <paul@codesourcery.com>
431 * config/arm/arm.md (insv): Use gen_int_mode.
433 2005-01-20 Eric Botcazou <ebotcazou@libertysurf.fr>
435 * config/sparc/freebsd.h (ASM_OUTPUT_INTERNAL_LABELREF): Delete.
436 * config/sparc/linux.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
437 * config/sparc/linux64.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
438 * config/sparc/netbsd-elf.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
439 * config/sparc/sol2.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
441 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
444 * dwarf2out.c (concat_loc_descriptor): We don't know we can use
445 fbreg, so we have to assume we can't.
446 (loc_descriptor_from_tree_1): Ditto.
447 (containing_function_has_frame_base): New function.
448 (add_location_or_const_value_attribute): Use it.
449 Also try to generate a frame_base from a single element location
452 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
454 PR tree-optimization/15349
455 * timevar.def (TV_TREE_MERGE_PHI): New.
456 * tree-cfg.c (tree_forwarder_block_p): Add a new argument
458 (remove_forwarder_block, cleanup_forwarder_blocks): Adjust the
459 calls to tree_forwarder_block_p.
460 (remove_forwarder_block_with_phi, merge_phi_nodes,
461 gate_merge_phi, pass_merge_phi): New.
462 * tree-optimize.c (init_tree_optimization_passes): Add
464 * tree-pass.h: Add an extern for pass_merge_phi;
466 2005-01-20 Richard Henderson <rth@redhat.com>
469 * config/i386/emmintrin.h (_mm_castpd_ps, _mm_castpd_si128): New.
470 (_mm_castps_pd, _mm_castps_si128): New.
471 (_mm_castsi128_ps, _mm_castsi128_pd): New.
473 2005-01-20 Richard Henderson <rth@redhat.com>
476 * config/i386/mmintrin.h (_mm_cvtsi32_si64): Use
477 __builtin_ia32_vec_init_v2si.
478 (_mm_cvtsi64_si32): Use __builtin_ia32_vec_ext_v2si.
479 * config/i386/i386.c (IX86_BUILTIN_VEC_EXT_V2SI): New.
480 (ix86_init_mmx_sse_builtins): Create it.
481 (ix86_expand_builtin): Expand it.
482 (ix86_expand_vector_set): Handle V2SFmode and V2SImode.
483 * config/i386/mmx.md (vec_extractv2sf_0, vec_extractv2sf_1): New.
484 (vec_extractv2si_0, vec_extractv2si_1): New.
486 2005-01-20 Richard Henderson <rth@redhat.com>
489 * config/i386/i386.c (ix86_expand_push): New.
490 * config/i386/mmx.md (push<MMXMODE>1): New.
491 * config/i386/sse.md (push<SSEMODE>1): New.
492 * config/i386/i386-protos.h: Update.
494 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
496 * params.def, params.h, predict.c, tree-ssa-loop-im.c: Update
499 2005-01-20 Richard Henderson <rth@redhat.com>
501 * rtl.def (CONST_VECTOR): Use RTX_CONST_OBJ.
502 * rtl.h (CONSTANT_P): Don't special case CONST_VECTOR.
504 2005-01-19 Richard Henderson <rth@redhat.com>
507 * config/i386/i386.c (ix86_expand_vector_move_misalign): Convert
508 to V4SFmode in SSE1 fallback load path.
510 2005-01-19 Richard Henderson <rth@redhat.com>
512 * config/i386/i386.c (ix86_expand_vector_init_one_var): Fix typo
513 in QImode expansion to ix86_expand_vector_set.
514 (ix86_expand_vector_init_general): Fix typo in V8HImode recursive call.
516 2005-01-19 Richard Henderson <rth@redhat.com>
519 * config/i386/i386.c (ix86_preferred_reload_class): Return a proper
520 subclass of the input class.
521 (ix86_secondary_memory_needed): Always true for cross-MMX classes.
522 Always true for cross-SSE1 classes. Rationalize conditionals.
523 * config/i386/i386.h (SSE_CLASS_P, MMX_CLASS_P): Use straight equality.
524 * config/i386/i386.md (movsi_1): Add MMX/SSE zeros. Fix alternatives
525 for SSE1. Don't check TARGET_INTER_UNIT_MOVES.
526 (movdi_2): Add MMX/SSE zeros.
527 (movdi_1_rex64): Likewise. Don't check TARGET_INTER_UNIT_MOVES.
528 (movsf_1): Don't check TARGET_INTER_UNIT_MOVES.
529 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
530 (movsi_1_nointernunit, movdi_1_rex64_nointerunit): Remove.
531 (movsf_1_nointerunit, zero_extendsidi2_32_1): Remove.
532 (zero_extendsidi2_rex64_1): Remove.
533 (MOV0 peephole): Check GENERAL_REG_P.
535 2005-01-19 Richard Henderson <rth@redhat.com>
538 * config/i386/i386.c (ix86_expand_vector_set): Fix third and fourth
540 (ix86_expand_vector_extract): Likewise.
542 2005-01-19 Richard Henderson <rth@redhat.com>
545 * expr.c (emit_move_change_mode): New argument force; use
546 simplify_gen_subreg if true.
547 (emit_move_via_alt_mode): Merge into ...
548 (emit_move_via_integer): ... here.
549 (emit_move_ccmode): Use emit_move_change_mode directly.
551 2005-01-19 Richard Henderson <rth@redhat.com>
554 * config/alpha/alpha.c (alpha_rtx_costs): Handle HIGH.
555 (alpha_preferred_reload_class): Handle CONST_VECTOR.
556 (alpha_emit_set_const_1): Add no_output parameter; don't emit
558 (alpha_emit_set_const): Likewise. Make static.
559 (alpha_emit_set_long_const): Make static.
560 (alpha_extract_integer): Split out from alpha_expand_mov.
561 (alpha_split_const_mov): Likewise.
562 (alpha_expand_mov): Use them. Handle CONST_VECTOR.
563 (alpha_legitimate_constant_p): New.
564 * config/alpha/alpha-protos.h: Update.
565 * config/alpha/alpha.h (REGISTER_MOVE_COST): Correct fp<->gp cost.
566 (LEGITIMATE_CONSTANT_P): Re-implement with a function.
567 * config/alpha/alpha.md (movsi): Add n alternative.
568 (movsi_nt_vms, movdi_er_nofix, movdi_er_fix, movdi_fix): Likewise.
569 (mov<VEC>_fix, mov<VEC>_nofix): Add i alternative.
570 (splitters for all of the above): Use alpha_split_const_mov.
571 * config/alpha/predicates.md (non_add_const_operand): New.
572 (non_zero_const_operand): New.
573 (input_operand): Use alpha_legitimate_constant_p after reload.
575 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
577 PR tree-optimization/19038
578 * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): New function.
579 (add_candidate): Add ivs with increment in latch only if
580 allow_ip_end_pos_p is true.
581 (determine_iv_cost): Use empty_block_p.
583 2005-01-19 Daniel Berlin <dberlin@dberlin.org>
585 * cfganal.c (compute_dominance_frontiers_1): Replace with new algorithm
586 (compute_dominance_frontiers): Ditto.
588 2005-01-19 Ralf Corsepius <ralf.corsepius@rtems.org>
591 * config/sh/t-rtems: New.
592 * config.gcc (sh-*-rtems*): Reflect having added config/sh/t-rtems.
594 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
596 * tree-ssa-ccp.c (maybe_fold_offset_to_component_ref): Always subtract
597 the offset of the selected field.
599 2005-01-19 Kazu Hirata <kazu@cs.umass.edu>
601 * tree-cfg.c (remove_forwarder_block): Fix the check to
602 prevent a nonlocal label from appearing in the middle of a
605 2005-01-19 Hans-Peter Nilsson <hp@axis.com>
607 PR rtl-optimization/19462
608 * reorg.c (find_end_label): Create return insn only if
609 current_function_epilogue_delay_list is empty.
611 2005-01-19 Paolo Bonzini <bonzini@gnu.org>
613 * doc/passes.texi: Remove paragraph mentioning new-ra.
615 2005-01-19 Jakub Jelinek <jakub@redhat.com>
617 PR rtl-optimization/15139
618 * combine.c: Include params.h.
619 (count_rtxs): New function.
620 (record_value_for_reg): If replace_rtx would replace at least
621 2 occurrences of REG in VALUE and TEM is really large, replace REG with
622 (clobber (const_int 0)) instead of TEM.
623 * params.def (PARAM_MAX_LAST_VALUE_RTL): New.
624 * params.h (MAX_LAST_VALUE_RTL): New.
625 * Makefile.in (combine.o): Depend on $(PARAMS_H).
626 * doc/invoke.texi (--param max-last-value-rtl=N): Document.
629 * c-typeck.c (digest_init): Only call build_vector if all constructor
630 elements are *_CST nodes.
631 * gimplify.c (gimplify_init_constructor): Likewise.
634 * c-typeck.c (digest_init): Only call build_vector if inside_init
637 2005-01-18 Tobias Schl"uter <tobias.schlueter@physik.uni-muenchen.de>
639 * toplev.c (init_asm_output): Remove dead #ifdef.
641 2005-01-18 Kaz Kojima <kkojima@gcc.gnu.org>
643 * config/sh/linux.h (TARGET_C99_FUNCTIONS): Define.
645 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
647 * calls.c (expand_call): Check DECL_BUILT_IN_CLASS before
648 accessing DECL_FUNCTION_CODE.
649 * dojump.c (do_jump): Likewise.
650 * gimplify.c (gimplify_call_expr): Likewise.
651 * predict.c (expr_expected_value): Likewise.
652 (strip_builtin_expect): Likewise.
653 * tree-inline.c (estimate_num_insns_1): Likewise.
654 * tree-ssa-loop-im.c (stmt_cost): Likewise
655 * fold-const.c (fold): Test for BUILT_IN_NORMAL.
656 (tree_expr_nonnegative_p): Likewise.
658 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
660 * config/sparc/sparc.c (load_pic_register): Emit the appropriate
661 variant of the load_pcrel_sym pattern.
662 * config/sparc/sparc.md (P macro): Move to the top.
663 (load_pcrel_sym): Macroize using P.
664 (save_register_window): Likewise.
666 2005-01-18 Aldy Hernandez <aldyh@redhat.com>
668 * config/rs6000/rs6000.md ("sunordered"): Disable for e500.
671 2005-01-18 Andrew Pinski <pinskia@physics.uc.edu>
674 * c-typeck.c (build_asm_expr): Strip nops off
675 input memory operands.
677 2005-01-18 David Edelsohn <edelsohn@gnu.org>
679 * config/rs6000/aix43.h (CPLUSPLUS_CPP_SPEC): Delete
680 _XOPEN_SOURCE, _XOPEN_SOURCE_EXTENDED, _LARGE_FILE_API.
681 * config/rs6000/aix51.h (CPLUSPLUS_CPP_SPEC): Same.
682 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Same.
684 2005-01-18 Joel Sherrill <joel.sherrill@oarcorp.com>
687 * config/i386/i386.c (override_options): If the 80387 is disabled,
688 then do not return FP values using FP registers.
690 2005-01-18 Kazu Hirata <kazu@cs.umass.edu>
692 * basic-block.h, c-common.c, c-cppbuiltin.c, c-lang.c,
693 c-tree.h, cfgbuild.c, cgraph.c, cgraph.h, collect2.c,
694 combine.c, config.gcc, coverage.h, cse.c, cselib.c,
695 defaults.h, df.c, dwarf2asm.c, dwarf2out.c, explow.c, expr.c,
696 flow.c, fold-const.c, gcse.c, ggc-page.c, gimple-low.c,
697 gimplify.c, ifcvt.c, langhooks-def.h, lcm.c, optabs.h,
698 output.h, postreload-gcse.c, postreload.c, recog.c,
699 resource.c, rtl.def, rtlanal.c, sched-deps.c, sched-rgn.c,
700 targhooks.h, toplev.c, tree-data-ref.c, tree-eh.c,
701 tree-flow.h, tree-loop-linear.c, tree-mudflap.h, tree-nrv.c,
702 tree-optimize.c, tree-outof-ssa.c, tree-pass.h,
703 tree-scalar-evolution.c, tree-ssa-copy.c, tree-ssa-dce.c,
704 tree-ssa-dse.c, tree-ssa-loop-ivopts.c, tree-ssa-operands.c,
705 tree-ssa-pre.c, tree-ssa.c, tree-vectorizer.c, tree.def,
706 unwind-dw2-fde-darwin.c, var-tracking.c: Update copyright.
708 * config/darwin.c, config/darwin.h, config/freebsd-spec.h,
709 config/alpha/alpha.c, config/arm/vfp.md,
710 config/i386/emmintrin.h, config/i386/i386-modes.def,
711 config/i386/i386-protos.h, config/i386/i386.h,
712 config/i386/pmmintrin.h, config/i386/ppro.md,
713 config/i386/predicates.md, config/i386/xmmintrin.h,
714 config/i860/i860.c, config/ia64/ia64-modes.def,
715 config/ia64/ia64-protos.h, config/ia64/ia64.h,
716 config/ia64/ia64.md, config/ia64/itanium1.md,
717 config/ia64/itanium2.md, config/ia64/predicates.md,
718 config/iq2000/iq2000.h, config/mips/linux64.h,
719 config/rs6000/aix41.h, config/rs6000/aix43.h,
720 config/rs6000/aix52.h, config/rs6000/darwin-fallback.c,
721 config/rs6000/rs6000.c, config/rs6000/rs6000.h,
722 config/rs6000/spe.md, config/sh/sh.md, config/sparc/linux.h,
723 config/sparc/linux64.h, config/sparc/litecoff.h,
724 config/sparc/sp64-elf.h, config/sparc/sparc.c,
725 config/sparc/sparc.h, config/sparc/sparc.md,
726 config/sparc/sysv4-only.h: Update copyright.
728 2005-01-18 Richard Henderson <rth@redhat.com>
731 * config/i386/i386.c (ix86_expand_fp_movcc): Fail for LTGT and UNEQ.
733 2005-01-18 Marc Espie <espie@openbsd.org>
734 * config/i386/openbsdelf.h: Typo.
736 2005-01-18 Richard Henderson <rth@redhat.com>
738 * config/i386/i386.c (ix86_expand_fp_movcc): Remove TARGET_IEEE_FP
739 special case for sse.
740 * config/i386/predicates.md (sse_comparison_operator): Likewise.
742 2005-01-18 Dorit Naishlos <dorit@il.ibm.com>
744 * tree-inline.c (estimate_num_insns_1): Added cases for
745 ALIGN_INDIRECT_REF, MISALIGNED_INDIRECT_REF, and REALIGN_LOAD_EXPR.
747 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
749 PR rtl-optimization/19296
750 * combine.c (simplify_comparison): Rewrite the condition under
751 which a non-paradoxical SUBREG of a PLUS can be lifted when
752 compared against a constant.
754 2005-01-18 Andi Kleen <ak@muc.de>
756 * c-typeck.c: (convert_for_assignment): Check warn_pointer_sign.
757 * c.opt (-Wpointer-sign): Add.
758 * doc/invoke.texi: (-Wpointer-sign): Add.
760 2005-01-18 Uros Bizjak <uros@kss-loka.si>
762 * config/i386/i386.c (override_options): Revert 2004-11-24 change.
763 * config/i386/i386.md (fmodsf3, dremsf3, *sinsf2, *cossf2, sincossf3,
764 *tansf3_1, tansf2, atan2sf3_1, atan2sf3, atansf2, asinsf2, acossf2,
765 logsf2, log10sf2, log2sf2, log1psf2, logbsf2, ilogbsf2, expsf2,
766 exp10sf2, exp2sf2, expm1sf2, rintsf2, floorsf2, ceilsf2, btruncsf2,
767 nearbyintsf2): Disable for TARGET_SSE_MATH. Leave patterns enabled
768 for TARGET_MIX_SSE_I387.
769 (fmoddf3, dremdf3, *sindf2, *sinextendsfdf2, *cosdf2, *cosextendsfdf2,
770 sincosdf3, *sincosextendsfdf3, *tandf3_1, tandf2, atan2df3_1,
771 atan2df3, atandf2, asindf2, acosdf2, logdf2, log10df2, log2df2,
772 log1pdf2, logbdf2, expdf2, exp10df2, exp2df2, expm1df2, rintdf2,
773 floordf2, ceildf2, btruncdf2, nearbyintdf2): Disable for
774 (TARGET_SSE2 && TARGET_SSE_MATH). Leave patterns enabled for
776 (atan2sf3, atan2df3, atan2xf3): Remove register constraints
779 2005-01-18 Uros Bizjak <uros@kss-loka.si>
782 * config/i386/mmx.md (*movv2sf_internal_rex64, *movv2sf_internal):
783 Add movaps alternative for xmm reg->reg move.
785 2005-01-17 Jeff Law <law@redhat.com>
787 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Clear SSA_NAME_VALUE
788 at the end of the main DOM loop rather than just before DOM exits.
790 2005-01-17 Diego Novillo <dnovillo@redhat.com>
792 PR tree-optimization/19121
793 * tree-ssa-alias.c (compute_flow_sensitive_aliasing): When
794 adding aliases to a name tag, also add them to the pointer's
796 * tree-ssa-copy.c (merge_alias_info): Do not merge flow
797 sensitive alias info at all. Only check that the two pointers
798 have compatible pointed-to sets.
799 * tree-ssa.c (verify_name_tags): Verify that the alias set of
800 a pointer's type tag is a superset of the alias set of the
803 2005-01-17 James E Wilson <wilson@specifixinc.com>
806 * config/ia64/ia64.md (movxf): Handle general register source. Adjust
807 comment to document why.
809 2005-01-17 Richard Henderson <rth@redhat.com>
811 * config/i386/sse.md (smaxv4sf3_finite, sse_vmsmaxv4sf3_finite,
812 sminv4sf3_finite, sse_vmsminv4sf3_finite, smaxv2df3_finite,
813 sse2_vmsmaxv2df3_finite, sminv2df3_finite,
814 sse2_vmsminv2df3_finite): New.
815 (smaxv4sf3, sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, smaxv2df3,
816 sse2_vmsmaxv2df3, sminv2df3, sse2_vmsminv2df3): Remove commutative.
817 Force op1 into register.
819 2005-01-17 Kelley Cook <kcook@gcc.gnu.org>
821 * Makefile.in (STAGEMOVESTUFF): Stage all the stamp files to prevent
822 bubblestrap from always rebuilding build directory.
824 2005-01-17 Ulrich Weigand <uweigand@de.ibm.com>
826 * config/s390/s390.md ("*clc"): Remove incorrect '='.
827 ("*cmpmem_short"): Likewise.
829 2005-01-17 Steven Bosscher <stevenb@suse.de>
831 * basic-block.h: Document BB_* flags.
832 * regrename.c (copyprop_hardreg_forward): Don't use BB_VISITED,
833 use an sbitmap instead.
834 * sched-rgn.c (compute_trg_info): Likewise.
836 2005-01-17 Richard Sandiford <rsandifo@redhat.com>
838 * config.gcc (mips64*-*-linux*): Set the default abi to n32. Remove
839 redundant target_cpu_default and extra_parts lines. Use the default
841 * config/mips/linux64.h (DRIVER_DEFAULT_ABI_SELF_SPEC): Delete.
842 (SUBTARGET_EXTRA_SPECS): Delete.
843 (DRIVER_SELF_SPECS): Remove architecture lines. Set the ABI directly.
844 (SUBTARGET_ASM_SPEC): Remove -mabi=n32 and -mabi=64 mappings.
845 * config/mips/t-linux64 (MULTILIB_DIRNAMES): Use "32" for -mabi=32
846 and "n32" for -mabi=n32.
848 2005-01-17 Ian Lance Taylor <ian@airs.com>
851 * c-typeck.c (build_c_cast): Revert patch of 2003-02-16: don't
852 fold constant variables into initial values.
854 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
856 * tree-cfg.c (tree_can_merge_blocks_p): Reorder two checks.
858 * tree-cfg.c (tree_forwarder_block_p): Speed up by walking
859 through the statements backward.
861 2005-01-17 Ian Lance Taylor <ian@airs.com>
864 * tree-inline.c (expand_call_inline): Set TREE_NO_WARNING on
865 a variable set to the return value of the inlined function.
867 2005-01-17 Mark Dettinger <dettinge@de.ibm.com>
869 * rtlanal.c (rtx_cost): Assign cost of 0 to a SUBREG
870 when modes are tieable.
872 2005-01-17 Ranjit Mathew <rmathew@hotmail.com>
874 * gthr-posix.h (__gthread_active_p): Use pthread_cancel instead
875 of pthread_create to find out if threads are enabled.
876 * gthr-posix95.h (__gthread_active_p): Likewise.
878 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
880 * common.opt (-fnew-ra): Remove.
882 * toplev.h (flag_new_regalloc): Remove.
883 * Makefile.in (ra*.*): Don't mention.
884 * passes.c (rest_of_handle_new_regalloc): Remove.
885 (rest_of_handle_combine, rest_of_compilation): Always consider
886 flag_new_regalloc as false.
887 * doc/invoke.texi: Don't document -fnew-ra.
889 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
891 * bb-reorder.c (fix_edges_for_rarely_executed_code): Remove
892 last parameter to reg_scan.
893 * loop.c (loop_optimize): Likewise.
894 * passes.c (rest_of_handle_tracer, rest_of_handle_if_conversion,
895 rest_of_handle_web, rest_of_handle_cfg, rest_of_handle_jump_bypass,
896 rest_of_handle_life, rest_of_handle_cse, rest_of_handle_cse2,
897 rest_of_handle_gcse, rest_of_handle_loop_optimize,
898 rest_of_handle_loop2, rest_of_handle_jump2): Likewise.
899 * regclass.c (reg_scan): Likewise, for the declaration.
900 * rtl.h (reg_scan): Likewise, for the prototype.
902 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
904 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Don't call
907 2005-01-16 Daniel Berlin <dberlin@dberlin.org>
909 * tree-ssa-pre.c (add_to_sets): s1 may be NULL.
910 (compute_avail): Uses don't go in tmp_gen.
912 2005-01-16 Steven Bosscher <stevenb@suse.de>
914 * ggc-page.c (ggc_alloc_stat): Use __builtin_ctzl instead of a
915 loop to look for a free slot in a page entry.
917 2005-01-16 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
920 * defaults.h (TARGET_DEFERRED_OUTPUT_DEFS): Provide default.
921 * toplev.c (compile_file): Call process_pending_assemble_output_defs
922 just before targetm.asm_out.file_end.
923 * tree.h (process_pending_assemble_output_defs): Declare.
924 * varasm.c (assemble_output_def, process_pending_assemble_output_defs):
926 (assemble_alias): Defer generation of assembly code for defines when
927 TARGET_DEFERRED_OUTPUT_DEFS is true.
928 * config/rs6000/aix41.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
929 * config/rs6000/aix43.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
930 * doc/tm.texi (TARGET_DEFERRED_OUTPUT_DEFS): document.
932 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
935 * pa.c (pa_scalar_mode_supported_p): New function.
936 (TARGET_SCALAR_MODE_SUPPORTED_P): Define.
937 * pa.h (MIN_UNITS_PER_WORD): Add comment.
939 * pa-protos.h (prefetch_operand): Delete.
940 (prefetch_cc_operand, prefetch_nocc_operand): New declations.
941 * pa.c (prefetch_operand): Delete.
942 (prefetch_cc_operand, prefetch_nocc_operand): New functions.
943 * pa.h (EXTRA_CONSTRAINT): Add `W' constraint.
944 (PREDICATE_CODES): Delete prefetch_operand. Add prefetch_cc_operand
945 and prefetch_nocc_operand.
946 * pa.md (prefetch): Rework to avoid reload problems handling short
947 displacements when a cache control completer needs to be provided.
948 (prefetch_32, prefetch_64): Delete.
949 (prefetch_cc, prefetch_nocc): New patterns.
951 2005-01-15 David Edelsohn <edelsohn@gnu.org>
953 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Revert previous change.
955 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
958 * pa.c (pa_asm_output_aligned_common): Warn if specified alignment
959 exceeds maximum alignment for global common data.
960 * pa64-hpux.h (MAX_OFILE_ALIGNMENT): Define to 32768.
961 * som.h (MAX_OFILE_ALIGNMENT): Likewise.
963 2004-01-15 Roger Sayle <roger@eyesopen.com>
965 * tree-ssa-dom.c (extract_range_from_cond): Correct condition.
967 2004-01-15 Roger Sayle <roger@eyesopen.com>
969 * harg-reg-set.h (reg_class_names): Prototype global array.
970 * regclass.c (reg_class_names): Declare here and initialize to
972 (dump_regclass): Remove local declaration of reg_class_names.
973 (regclass): Likewise.
974 * cfg.c (dump_flow_info): Likewise.
975 * ra-debug.c (reg_class_names): Likewise.
976 * regrename.c (reg_class_names): Likewise.
977 * reload.c (reg_class_names): Likewise.
978 * reload1.c (spill_failure): Likewise.
979 * config/m68hc11/m68hc11.c (reg_class_names): Likewise.
981 2005-01-15 Ulrich Weigand <uweigand@de.ibm.com>
983 * config/s390/s390.md ("reload_outti"): Remove predicate for
984 output operand. Abort if operand is not a MEM.
985 ("reload_outdi", "reload_outdf"): Likewise.
987 2005-01-15 Marc Espie <espie@openbsd.org>
989 * config.gcc (*-*-openbsd*): Set HAS_LIBC_R for OpenBSD <= 3.2.
990 (i[34567]86-*-openbsd*): Switch to ELF for OpenBSD >= 3.4.
991 * config/openbsd.h: Fix C++ includes for native configurations.
992 Add proper OS_CPP_BUILTINS. Fix libspec for recent OpenBSD.
993 Add trampoline support.
994 * config/i386/openbsdelf.h: New.
996 2005-01-15 Marc Espie <espie@openbsd.org>
998 * collect2.c (main): Explicitly parse -dynamic-linker option.
1000 2005-01-15 Jakub Jelinek <jakub@redhat.com>
1002 PR tree-optimization/19060
1003 * tree-ssa-dom.c (extract_range_from_cond) <case LT_EXPR, GT_EXPR>:
1004 Return 0 if op1 <= TYPE_MIN_VALUE () resp. op1 >= TYPE_MAX_VALUE ().
1005 (simplify_cond_and_lookup_avail_expr): Add assert for dummy == 0
1006 and handle extract_range_from_cond returning false.
1007 * fold-const.c (fold): Optimize comparisons with min/max even for
1008 width > HOST_BITS_PER_WIDE_INT.
1010 2005-01-15 Ralf Corsepius <ralf.corsepius@rtems.org>
1012 * config/mips/rtems.h (MIPS_DEFAULT_GVALUE): Set to 0.
1013 * config/mips/t-rtems (MULTILIBS_DIRNAMES,MULTILIB_OPTIONS):
1014 Remove little endian multilib variants.
1015 Add mips32 multilib variant.
1017 2005-01-15 Kazu Hirata <kazu@cs.umass.edu>
1019 * explow.c (copy_all_reg, stabilize): Remove.
1020 * expr.h: Remove the corresponding prototypes.
1022 * rtlanal.c (regs_set_between_p): Remove.
1023 * rtl.h: Remove the corresponding prototype.
1025 2005-01-14 David Edelsohn <edelsohn@gnu.org>
1027 * config/rs6000/rs6000.c (rs6k_nonimmediate_operand): Rename
1028 to rs6000_nonimmediate_operand.
1029 * config/rs6000/rs6000.h (PREDICATE_CODES): Rename
1030 rs6k_nonimmediate_operand to rs6000_nonimmediate_operand.
1031 * config/rs6000/rs6000.md (movsi_internal1): Rename predicate to
1032 rs6000_nonimmediate_operand.
1033 * config/rs6000/spe.md (movdf_e500_double): Rename predicate to
1034 rs6000_nonimmediate_operand.
1036 2005-01-14 Andrew Pinski <pinskia@physics.uc.edu>
1038 * tree-ssa-alias.c (pass_may_alias): Add TODO_verify_stmts.
1040 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1042 * config/rs6000/rs6000.h (CLASS_MAX_NREGS): DF goes in 1 register
1044 (CANNOT_CHANGE_MODE_CLASS): Restrict DI mode changes on e500v2.
1045 (PREDICATE_CODES): Add rs6k_nonimmediate_operand.
1047 * config/rs6000/rs6000.c (invalid_e500_subreg): New.
1048 (rs6k_nonimmediate_operand): New.
1049 (rs6000_legitimate_offset_address_p): Handle DI modes on e500v2
1051 (legitimate_lo_sum_address_p): Same.
1052 (rs6000_legitimize_address): Same.
1053 (rs6000_legitimize_reload_address): Same.
1054 (rs6000_legitimate_address): Same.
1055 (spe_build_register_parallel): Pass DF and DC modes in a DI
1058 * config/rs6000/rs6000.md ("*movsi_internal1"): Change predicate
1059 to rs6k_nonimmediate_operand.
1061 * config/rs6000/spe.md ("*frob_df_di"): New.
1062 ("*frob_di_df"): New.
1063 ("*frob_di_df_2"): New.
1064 ("*mov_sidf_e500_subreg0"): New.
1065 ("*mov_sidf_e500_subreg4"): New.
1066 ("*movdf_e500_double"): Change predicate to
1067 rs6k_nonimmediate_operand.
1069 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1071 * postreload.c (move2add_note_store): Only call
1072 trunc_int_for_mode on scalar integers.
1074 2005-01-14 Steven Bosscher <stevenb@suse.de>
1076 * tree-ssa-dce.c (visited_control_parents): New sbitmap to
1077 replace BB_VISITED uses.
1078 (find_obviously_necessary_stmts): Don't clear BB_VISITED.
1079 (propagate_necessity): Check the bitmap instead of BB_VISITED.
1080 (tree_dce_done): Free visited_control_parents.
1081 (perform_tree_ssa_dce): Allocate and clear it.
1082 * tree-ssa-pre.c (compute_antic_aux): Make non-recursive.
1083 (compute_antic): Iterate from here using a DFS. Use an sbitmap
1084 instead of BB_VISITED.
1086 2005-01-14 Kazu Hirata <kazu@cs.umass.edu>
1088 * c-tree.h, coverage.h, langhooks-def.h, optabs.h, output.h,
1089 rtl.h, targhooks.h, tree-flow.h, tree-mudflap.h, tree.h:
1090 Remove unused prototypes.
1092 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1094 * config/rs6000/rs6000.c (rs6000_stack_info): Fix caching of SPE
1095 64-bit register usage.
1096 (spe_func_has_64bit_regs_p): Add FIXME note.
1098 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
1100 * config/rs6000/rs6000.c (rs6000_generate_compare): Replace
1101 flag_finite_math_only with flag_unsafe_math_optimizations.
1103 2005-01-14 J. D. Johnston <jjohnst@us.ibm.com>
1105 * config/s390/tpf-unwind.h (s390_fallback_frame_state): Correct
1107 (__tpf_eh_return): Copy TPF private stack area from the module
1108 boundary stack frame. Add check for when module addresses are
1109 equal, but no stub address is found.
1111 2005-01-14 Richard Earnshaw <rearnsha@arm.com>
1114 * arm.h (struct machine_function): Add call_via field.
1115 (thumb_call_via_label): Declare.
1116 * arm.c (thumb_call_via_label): New variable.
1117 (thumb_call_reg_needed): New variable.
1118 (arm_output_function_epilogue): For Thumb code, output any per-function
1119 call-indirect trampolines.
1120 (thumb_call_via_reg): New function.
1121 (arm_file_end): New function.
1122 (TARGET_ASM_FILE_END): Call arm_file_end.
1123 (aof_file_end): Likewise.
1124 * arm-protos.h (thumb_call_via_reg): Declare.
1125 * arm.md (call_reg_thumb, call_value_reg_thumb): Call
1126 thumb_call_via_reg in normal case.
1128 2005-01-14 Jakub Jelinek <jakub@redhat.com>
1131 PR rtl-optimization/19348
1132 * recog.c (peephole2_optimize): Do global life update if some peephole
1133 decides it doesn't need at least one of its inputs and that change
1134 influences liveness at the start of the basic block.
1136 * basic-block.h (EXECUTE_IF_AND_COMPL_IN_REG_SET): Needs 2 REGSET
1137 arguments instead of 1.
1139 2005-01-14 Eric Botcazou <ebotcazou@libertysurf.fr>
1142 * varasm.c (initializer_constant_valid_p) <ADDR_EXPR>: Return
1143 zero for nested functions needing a static chain or functions
1144 with a non-constant address.
1146 2005-01-13 Roger Sayle <roger@eyesopen.com>
1148 * simplify-rtx.c (simplify_binary_operation) <AND>: Optimize
1149 (and (sign_extend X) C) into (zero_extend (and X C)).
1151 2005-01-13 David O'Brien <obrien@FreeBSD.org>
1153 * config/freebsd-spec.h: Make KSE pthread lib logic the default.
1155 2005-01-13 Richard Henderson <rth@redhat.com>
1160 * config/i386/i386.md (cmpdf, cmpsf, bunordered, bordered, buneq,
1161 bunge, bungt, bunle, bunlt, bltgt): Enable for TARGET_SSE_MATH,
1162 not just TARGET_SSE.
1163 (cmpfp_i_387): Rename from cmpfp_i. Move after sse patterns.
1164 (cmpfp_i_mixed): Rename from cmpfp_i_sse; use for TARGET_MIX_SSE_I387.
1165 (cmpfp_i_sse): Rename from cmpfp_i_sse_only; use for TARGET_SSE_MATH.
1166 (cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387): Similarly.
1167 (fp_jcc_1_mixed, fp_jcc_1_sse, fp_jcc_1_387): Similarly.
1168 (fp_jcc_2_mixed, fp_jcc_2_sse, fp_jcc_2_387): Similarly.
1169 (fp_jcc_3_387, fp_jcc_4_387, fp_jcc_5_387, fp_jcc_6_387,
1170 fp_jcc_7_387, fp_jcc_8_387): Rename from fp_jcc_N.
1171 (movdicc_c_rex64): Rename with '*'.
1172 (movsfcc, movdfcc): Add checks for 387 and sse math to condition.
1173 (movsfcc_1_sse_min, movsfcc_1_sse_max, movsfcc_1_sse): New.
1174 (movsfcc_1_387): Rename from movsfcc_1.
1175 (movdfcc_1_sse_min, movdfcc_1_sse_max, movdfcc_1_sse): New.
1176 (movdfcc_1, movdfcc_1_rex64): Add check for 387.
1177 (sminsf3, smaxsf3, smindf3, smaxdf3): New.
1178 (minsf3, minsf, minsf_nonieee, minsf_sse, mindf3, mindf,
1179 mindf_nonieee, mindf_sse, maxsf3, maxsf, maxsf_nonieee, maxsf_sse,
1180 maxdf3, maxdf, maxdf_nonieee, maxdf_sse, sse_movsfcc, sse_movsfcc_eq,
1181 sse_movdfcc, sse_movdfcc_eq, sse_movsfcc_const0_1,
1182 sse_movsfcc_const0_2, sse_movsfcc_const0_3, sse_movsfcc_const0_4,
1183 sse_movdfcc_const0_1, sse_movdfcc_const0_2, sse_movdfcc_const0_3,
1184 sse_movdfcc_const0_4): Remove.
1185 * config/i386/i386.c (ix86_expand_fp_movcc): For TARGET_SSE_MATH,
1186 recognize min/max early. Update for changed sse cmove patterns.
1187 (ix86_split_sse_movcc): New.
1188 * config/i386/i386-protos.h: Update.
1190 2005-01-13 Steven Bosscher <stevenb@suse.de>
1192 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
1193 (fix_stmt_v_may_defs): Likewise.
1195 2005-01-13 Richard Henderson <rth@redhat.com>
1197 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
1198 for converting the mask.
1200 2005-01-13 David Edelsohn <edelsohn@gnu.org>
1202 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
1205 2005-01-13 Richard Henderson <rth@redhat.com>
1207 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
1208 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
1209 IX86_BUILTIN_MOVDQ2Q): Remove.
1210 (IX86_BUILTIN_VEC_EXT_V4SI): New.
1211 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
1212 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
1214 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
1215 (sse2_pextrw): Fix immediate constraint.
1216 (sse2_loadq, sse2_loadq_rex64): Remove.
1217 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
1218 Use __builtin_ia32_vec_ext_<size>.
1219 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
1221 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
1223 * function.c (assign_parm_setup_block): Look inside original
1224 entry_parm when inspecting PARALLEL.
1226 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
1227 Joel Sherrill <joel@oarcorp.com>
1230 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
1231 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
1232 rtems_gxx_recursive_mutex_init.
1233 (__gthread_recursive_mutex_lock): New function.
1234 (__gthread_recursive_mutex_trylock): Likewise.
1235 (__gthread_recursive_mutex_unlock): Likewise.
1237 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
1239 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
1241 2005-01-13 Jan Beulich <jbeulich@novell.com>
1243 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
1244 (ashrti3_internal): Indicate output is early clobber. Generate result
1245 into output rather than first input. Use move for low word of output
1246 if shift count is exactly 64.
1247 (lshrti3_internal): Likewise.
1249 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
1253 * reload1.c (delete_output_reload): Don't delete an output reload
1254 if the pseudo lives longer than a single basic block. Adjust and
1255 improve wording of comment.
1257 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
1260 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
1261 of TYPE_SIZE where we mean the number of bytes.
1263 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
1265 * function.c (assign_parm_setup_block): Relax condition on
1266 multi-register optimization.
1268 2005-01-12 Nick Clifton <nickc@redhat.com>
1270 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
1271 an "in_delay_slot" attribute of "no" to prevent them being used in
1272 delay slots. This is forbidden because they might generate
1275 2005-01-12 Alan Modra <amodra@bigpond.net.au>
1278 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
1281 2005-01-12 Nick Clifton <nickc@redhat.com>
1283 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
1284 passed on to GAS) but do not define, as GAS no longer supports or
1285 needs the -m2000 option.
1287 2005-01-11 Roger Sayle <roger@eyesopen.com>
1289 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
1290 type as args of unprototyped builtins aren't automatically promoted.
1291 (fold_builtin_abs): Likewise.
1293 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1295 * Makefile.in: Set a `build-warn' variable.
1297 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
1299 PR tree-optimization/17949
1300 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
1301 (find_interesting_uses_address): Use it.
1303 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
1305 * regrename.c (kill_value): Handle subreg's that won't simplify.
1307 2005-01-11 Richard Henderson <rth@redhat.com>
1310 * config/i386/i386.h (enum ix86_builtins): Move ...
1311 * config/i386/i386.c: ... here.
1312 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
1313 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
1314 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
1315 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
1316 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
1317 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
1318 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
1319 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
1320 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
1321 IX86_BUILTIN_LOADDDUP): Remove.
1322 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
1323 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
1324 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
1325 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
1326 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
1327 (ix86_init_builtins): Make static.
1328 (ix86_init_mmx_sse_builtins): Update for changed builtins.
1329 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
1330 if all the modes match. Otherwise, fake it.
1331 (get_element_number, ix86_expand_vec_init_builtin,
1332 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
1333 (ix86_expand_builtin): Make static. Update for changed builtins.
1334 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
1335 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
1336 (ix86_expand_vector_init_duplicate): New.
1337 (ix86_expand_vector_init_low_nonzero): New.
1338 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
1339 Split out from ix86_expand_vector_init; handle integer modes.
1340 (ix86_expand_vector_init): Use them.
1341 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
1342 * config/i386/i386-protos.h: Update.
1343 * config/i386/predicates.md (reg_or_0_operand): New.
1344 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
1345 (movv2sf_internal): Likewise. And a splitter to match them all.
1346 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
1347 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
1348 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
1349 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
1350 (mmx_pinsrw): Fix operand ordering.
1351 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
1352 rather than sse_loadss expander.
1353 (movv2df splitter): Similarly.
1354 (sse_loadss, sse_loadlss): Remove.
1355 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
1356 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
1357 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
1358 (sse3_movddup): Rename with '*'.
1359 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
1360 (sse2_loadsd): Remove.
1361 (vec_dupv2df_sse3): Rename from sse3_loadddup.
1362 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
1363 (sse2_pinsrw): Fix argument ordering.
1364 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
1365 (sse2_stored): Remove 'r' destination.
1366 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
1367 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
1368 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
1369 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
1370 vec_extractv16qi, vec_initv16qi): New.
1372 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
1373 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
1374 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
1376 (_mm_load_pd, _mm_store_pd): Use plain dereference.
1377 (_mm_load_si128, _mm_store_si128): Likewise.
1378 (_mm_load1_pd): Use _mm_set1_pd.
1379 (_mm_load_sd): Use _mm_set_sd.
1380 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
1381 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
1382 (_mm_set_epi64): Use _mm_set_epi64x.
1383 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
1384 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
1385 _mm_setr_epi8): Use _mm_set_foo form.
1386 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
1387 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
1388 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
1389 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
1390 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
1391 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
1392 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
1393 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
1394 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
1395 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
1396 (_mm_movedup_pd): Use _mm_shuffle_pd.
1397 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
1398 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
1399 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
1400 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
1402 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
1403 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
1404 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
1405 (_mm_store_ps): Use raw dereference.
1406 (_mm_store1_ps): Use _mm_storeu_ps.
1407 (_mm_storer_ps): Use _mm_store_ps.
1408 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
1409 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
1411 2005-01-11 Stan Shebs <shebs@apple.com>
1413 * config/rs6000/rs6000.c (machopic_output_stub): Issue
1414 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
1416 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
1419 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
1420 Skip all DECLs except for FIELD_DECLs.
1422 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
1424 * config/s390/s390.c (override_options): Return error if
1425 -mbackchain, -mpacked-stack and -mhard-float are used together.
1426 (s390_va_start): Remove the backchain && packed-stack special case.
1427 (s390_gimplify_va_arg): Likewise.
1428 * doc/invoke.texi: Remove the ABI incompatibility note.
1430 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
1432 * config/s390/s390.c (struct s390_frame_layout): Remove
1434 (s390_frame_info, s390_emit_prologue): Replace occurrences of
1435 save_backchain_p with TARGET_BACKCHAIN.
1437 2005-01-11 Alan Modra <amodra@bigpond.net.au>
1440 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
1441 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
1442 * function.c (locate_and_pad_parm): Set new field.
1443 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
1444 Tweak where_pad test to include "none". Always set mem align for
1446 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
1447 not sufficient for type.
1448 (assign_parm_setup_block): If stack_parm is zero on entry, always
1449 make a new stack local. Block move old stack parm if necessary
1450 to new aligned stack local.
1451 (assign_parm_setup_stack): Use a block move to handle
1452 potentially misaligned entry_parm.
1453 (assign_parms_unsplit_complex): Specify required alignment when
1454 creating stack local.
1455 * calls.c (compute_argument_addresses): Override alignment of stack
1456 arg calculated from its type with the alignment given by
1457 FUNCTION_ARG_BOUNDARY.
1458 (store_one_arg): Likewise.
1460 2005-01-11 Jan Beulich <jbeulich@novell.com>
1462 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
1463 Change respective itanium_class attribute to ialu.
1464 (shladdp4_internal): New.
1465 * config/ia64/predicates.md (shladd_log2_operand): New.
1467 2005-01-11 Richard Henderson <rth@redhat.com>
1469 * expr.c (store_constructor): Use rtvec_alloc instead of
1470 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
1472 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
1474 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
1476 2005-01-11 Alan Modra <amodra@bigpond.net.au>
1478 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
1479 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
1480 (TARGET_RWRELOC): Define.
1481 (ia64_rwreloc_section_type_flags): Delete.
1482 (ia64_section_type_flags): New function.
1483 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
1484 (TARGET_RWRELOC): Define.
1486 2005-01-10 David Mosberger <davidm@hpl.hp.com>
1489 * config/ia64/ia64.c (process_set): For alloc insn, only call
1490 process_epilogue is !frame_pointer_needed.
1492 2005-01-10 Roger Sayle <roger@eyesopen.com>
1495 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
1496 unary operator and can't be treated as a binary/comparison operator.
1498 2005-01-10 Richard Henderson <rth@redhat.com>
1500 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
1501 (ix86_return_in_memory): Likewise.
1502 (function_arg_advance): Likewise. Mirror structure in function_arg
1503 for choosing register to advance.
1505 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
1507 * tree-vectorizer.c, tree.def: Fix comment typos.
1509 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
1511 * config/s390/s390.c (struct processor_costs): 4 new fields:
1512 ddbr, ddr, debr, der.
1513 (s390_rtx_costs): More precise handling of divide instructions.
1515 2005-01-10 David Edelsohn <edelsohn@gnu.org>
1517 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
1519 2005-01-10 Jan Beulich <jbeulich@novell.com>
1521 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
1522 names resulting from -ffunction-sections/-fdata-sections and linkonce
1524 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
1525 based on the section name. Rearrange the section name comparison logic
1526 slightly so that each section name is compared against at most once.
1528 2005-01-10 Ben Elliston <bje@au.ibm.com>
1530 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
1531 targets for which -fPIC reduces limitations on the GOT size.
1533 2005-01-09 Falk Hueffner <falk@debian.org>
1535 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
1538 2005-01-09 David Edelsohn <edelsohn@gnu.org>
1541 * collect2.c (main): Set aixrtl_flag for -brtl option.
1542 (resolve_lib_name): Search for .so file extension before .a
1545 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
1547 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
1548 (vect_analyze_data_refs_alignment): Add dump prints.
1550 2005-01-09 Ira Rosen <irar@il.ibm.com>
1552 * tree-vectorizer.c (vect_analyze_offset_expr): Use
1553 expr_invariant_in_loop_p.
1554 Initialize outputs first thing in the function.
1555 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
1556 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
1557 (vect_analyze_pointer_ref_access): Check that the initial condition of
1558 the access function is loop invariant.
1560 2005-01-09 Richard Henderson <rth@redhat.com>
1562 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
1563 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
1564 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
1565 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
1566 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
1567 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
1568 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
1569 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
1570 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
1571 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
1572 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
1573 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
1574 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
1575 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
1576 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
1577 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
1578 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
1579 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
1580 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
1581 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
1582 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
1583 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
1584 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
1586 (mmx_clrdi, pavgusb): Remove.
1587 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
1588 with leading sse_ prefix.
1589 * config/i386/sse.md: Receive them.
1590 * config/i386/mmx.md: New file.
1591 (MMXMODE12, MMXMODE24, mmxvecsize): New.
1592 (subrv2sf3): Turn into expander for normal subtraction.
1593 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
1594 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
1595 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
1596 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
1597 commutative; use ix86_binary_operator_ok.
1598 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
1599 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
1600 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
1601 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
1602 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
1603 Macroize from existing patterns; use ix86_binary_operator_ok.
1604 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
1605 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
1606 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
1607 with vec_select+vec_concat.
1608 (mmx_pshufw, mmx_pshufw_1): Likewise.
1609 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
1610 (mmx_uavgv4hi3): Model correctly.
1611 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
1612 _mm_xor_si64): Remove casts.
1614 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
1616 PR tree-optimization/19224
1617 * tree-scalar-evolution.c (get_instantiated_value,
1618 set_instantiated_value): New functions.
1619 (instantiate_parameters_1): Cache the results.
1620 (instantiate_parameters, resolve_mixers): Initialize and free
1623 2005-01-08 David Edelsohn <edelsohn@gnu.org>
1625 * config/i386/i386.md (addhi_4): Correct reference in comment.
1628 2005-01-08 Richard Henderson <rth@redhat.com>
1630 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
1631 (_mm_cvtsi128_si64x): Likewise.
1632 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
1633 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
1634 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
1635 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
1636 * config/i386/i386-protos.h: Update.
1637 * config/i386/i386.c (print_operand): Add 'H'.
1638 (ix86_fixup_binary_operands): Split out from ...
1639 (ix86_expand_binary_operator): ... here.
1640 (ix86_fixup_binary_operands_no_copy): New.
1641 (ix86_expand_fp_absneg_operator): Handle vector mode results.
1642 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
1643 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
1644 (safe_vector_operand): Use CONST0_RTX.
1645 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
1646 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
1647 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
1648 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
1650 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
1651 (IX86_BUILTIN_CMPNGESS): New.
1652 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
1653 (attr type): Add sselog1.
1654 (attr unit, attr memory): Handle it.
1655 (movti, movti_internal, movti_rex64): Move near other integer moves.
1656 (movtf, movtf_internal): Move near other fp moves.
1657 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
1658 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
1659 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
1660 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
1661 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
1662 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
1663 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
1664 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
1665 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
1666 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
1667 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
1668 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
1669 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
1670 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
1671 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
1672 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
1673 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
1674 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
1675 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
1676 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
1677 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
1678 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
1679 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
1680 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
1681 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
1682 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
1683 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
1684 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
1685 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
1686 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
1687 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
1688 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
1689 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
1690 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
1691 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
1692 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
1693 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
1694 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
1695 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
1696 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
1697 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
1698 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
1699 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
1700 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
1702 (SSEPUSH, push<SSEPUSH>): Remove.
1703 (MMXPUSH, push<MMXPUSH>): Remove.
1704 (sse_movaps, sse_movaps_1, sse_movups): Remove.
1705 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
1706 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
1707 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
1708 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
1709 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
1710 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
1711 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
1712 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
1713 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
1714 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
1715 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
1716 (ppro_sse_log_V4SF): Handle sselog1.
1717 * config/i386/predicates.md (const_0_to_1_operand): New.
1718 (const_0_to_255_mul_8_operand): New.
1719 (const_1_to_31_operand): Rename from const_int_1_31_operand.
1720 (const_2_to_3_operand, const_4_to_7_operand): New.
1721 * config/i386/sse.md: New file.
1722 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
1723 (sse_movups): Rename from sse_movups_1.
1724 (sse_loadlss): Rename from sse_loadss_1.
1725 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
1727 (negv4sf2): Use ix86_expand_fp_absneg_operator.
1728 (absv4sf2, negv2df, absv2df): New.
1729 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
1730 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
1731 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
1732 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
1733 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
1734 (sse3_addsubv4sf3): Model correctly.
1735 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
1736 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
1737 (sse_movhlps): Model with vec_select+vec_concat.
1738 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
1739 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
1740 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
1741 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
1742 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
1743 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
1744 sse2_pshufhw_1): Likewise.
1745 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
1746 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
1747 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
1748 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
1749 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
1750 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
1751 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
1752 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
1753 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
1754 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
1755 sminv8hi3): Mark commutative
1756 operands. Use ix86_binary_operator_ok.
1757 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
1758 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
1759 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
1760 sse2_punpcklqdq): Allow operand2 in memory.
1761 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
1762 sse2_movsd): Add memory alternatives.
1763 (sse_storelps): Turn expander into an insn; split after reload.
1764 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
1765 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
1767 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
1769 * configure.ac (DWARF-2 debug_line): Use objdump.
1770 * configure: Regenerate.
1772 2005-01-08 Jeff Law <law@redhat.com>
1773 Diego Novillo <dnovillo@redhat.com>
1775 PR tree-optimization/18241
1776 * tree-nrv.c (tree_nrv): Ignore volatile return values.
1777 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
1778 statements with volatile operands.
1779 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
1780 operands after marking a statement with has_volatile_ops.
1782 2005-01-08 Roger Sayle <roger@eyesopen.com>
1784 * tree.c (int_fits_type_p): Always honor integer constant
1785 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
1787 2005-01-08 Roger Sayle <roger@eyesopen.com>
1789 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
1790 the latent bug in rtl_delete_block has been resolved.
1792 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
1794 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
1795 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
1797 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
1799 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
1800 library as @multilib_dir@/$(SHLIB_SONAME).
1802 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
1804 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
1805 * configure: Regenerate.
1807 2005-01-07 Jakub Jelinek <jakub@redhat.com>
1809 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
1812 2005-01-07 David Edelsohn <edelsohn@gnu.org>
1815 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
1816 Convert non-word aligned offset address using ld/std into
1819 2005-01-07 Richard Henderson <rth@redhat.com>
1821 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
1823 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
1825 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
1827 2005-01-07 Jakub Jelinek <jakub@redhat.com>
1829 PR tree-optimization/19283
1830 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
1831 is not shorter than the original type.
1833 PR rtl-optimization/19012
1834 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
1836 PR rtl-optimization/18861
1837 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
1838 (FULL_STATE): Define.
1839 (mark_tablejump_edge): New function.
1840 (purge_dead_tablejump_edges): New function.
1841 (find_bb_boundaries): Use it.
1843 PR tree-optimization/18828
1844 * builtins.c (expand_builtin_next_arg): Remove argument and all
1845 the argument checking.
1846 (expand_builtin): Adjust caller.
1847 (expand_builtin_va_start): Likewise. Remove error for too many
1849 (fold_builtin_next_arg): Issue error for too many arguments.
1850 After checking arguments, replace them with magic arguments that
1851 prevent further checking of the args.
1853 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1855 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
1857 2005-01-06 Roger Sayle <roger@eyesopen.com>
1860 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
1861 and %{.S:...} (and their negative variants) to test whether the
1862 input file is assembler or pre-processed-assembler independent of
1863 the actual filename extension.
1865 2005-01-06 Roger Sayle <roger@eyesopen.com>
1867 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
1868 of sign or zero extended values.
1870 2005-01-06 Geoffrey Keating <geoffk@apple.com>
1872 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
1873 for IBM long double format correctly.
1875 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
1877 Fix PR tree-optimization/18792
1879 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
1880 to first_loop_depth, and use loop depth instead of loop number.
1881 (build_classic_dir_vector): Ditto.
1882 (compute_data_dependences_for_loop): Use depth, not loop number.
1883 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
1884 number. Pass in loops, instead of loop numbers.
1885 (gather_interchange_stats): Ditto.
1886 (linear_transform_loops): Ditto.
1888 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
1891 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
1892 biv_fits_mode_p, extension_within_bounds_p): New functions.
1893 (check_ext_dependent_givs): Use them.
1895 2005-01-06 Roger Sayle <roger@eyesopen.com>
1897 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
1898 more than one barrier, in which case we should delete them all.
1900 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
1902 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
1904 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
1906 Fix PR middle-end/19286
1908 * dwarf2out.c (gen_subprogram_die): If we've already tried to
1909 output this subprogram, simply ignore this attempt to do it again.
1910 (add_abstract_origin_attribute): Don't abort trying to add the abstract
1911 origin attribute if it's not possible.
1912 (gen_block_die): Don't ignore subblocks of "unused" blocks.
1913 (decls_for_scope): Ditto.
1914 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
1915 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
1917 2005-01-05 Richard Henderson <rth@redhat.com>
1920 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
1921 (bdesc_2arg): Use it.
1922 (ix86_expand_binop_builtin): Force operands into registers
1924 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
1925 ix86_expand_sse_compare, ix86_expand_sse_comi,
1926 ix86_expand_builtin): Likewise.
1928 2005-01-05 Richard Henderson <rth@redhat.com>
1930 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
1932 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
1933 (mulv4hi3): Set itanium_class mmmul.
1934 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
1935 (fpack): Rename from fpack_sfsf.
1937 2005-01-05 Richard Henderson <rth@redhat.com>
1940 * reload1.c (do_input_reload): Restrict the optimization deleteing
1941 a previous output reload to RELOAD_FOR_INPUT.
1943 2005-01-05 Steven Bosscher <stevenb@suse.de>
1945 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
1946 comment that we fall through after case.
1947 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
1948 (distribute_links): Likewise.
1949 * cse.c (cse_insn): Likewise.
1950 * cselib.c (cselib_invalidate_mem): Likewise.
1951 * df.c: Update comments at the top of the file.
1952 (read_modify_subreg_p): Update comments here too.
1953 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
1954 * flow.c (mark_set_1): Likewise.
1955 (mark_used_regs): Likewise.
1956 * gcse.c (mems_conflict_for_gcse_p): Likewise.
1957 (canon_list_insert): Likewise.
1958 (mark_set): Likewise.
1959 (try_replace_reg): Likewise.
1960 (store_killed_in_insn): Likewise.
1961 * loop.c (count_one_set): Likewise.
1962 (basic_induction_var): Likewise.
1963 * postreload-gcse.c (find_mem_conflicts): Likewise.
1964 * postreload.c (reload_combine_note_store): Likewise.
1965 (move2add_note_store): Likewise.
1966 * reload.c (find_equiv_reg): Likewise.
1967 (mark_referenced_resources): Likewise.
1968 * rtlanal.c (set_noop_p): Likewise.
1969 (note_stores): Likewise.
1970 (note_uses): Likewise.
1971 * sched-deps.c (sched_analyze_1): Likewise.
1972 * sched-rgn.c (check_live_1): Likewise.
1973 (update_live_1): Likewise.
1974 * config/i860/i860.c: Likewise.
1976 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
1978 (ZERO_EXTRACT): Mention that this one can be an lvalue.
1980 * doc/rtl.texi: Update documentation for bit-fields and SET.
1982 2005-01-05 Roger Sayle <roger@eyesopen.com>
1984 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
1986 2005-01-05 Jan Hubicka <jh@suse.cz>
1987 Richard Henderson <rth@redhat.com>
1990 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
1993 2005-01-05 Richard Henderson <rth@redhat.com>
1996 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
1997 (sse_shufps): Change operand 3 to const_int_operand.
1998 (sse2_storelps): Fix typo in template.
1999 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
2000 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
2001 (ix86_expand_builtin): Likewise.
2003 2005-01-05 Julian Brown <julian@codesourcery.com>
2005 * config/arm/arm.c (arm_return_in_memory): Treat complex types
2006 as aggregates for AAPCS ABIs.
2008 2005-01-05 Stan Shebs <shebs@apple.com>
2010 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
2011 Mach-O getters if ppc64.
2012 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
2014 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
2015 signed four-byte field for global code case.
2016 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
2018 2005-01-05 Roger Sayle <roger@eyesopen.com>
2021 * c-common.c: Include real.h.
2022 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
2023 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
2024 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
2025 * Makefile.in (c-common.o): Update dependencies.
2027 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
2029 * c-parse.in (asm_string): Add trailing semicolon.
2031 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
2033 * c-parse.in (asm_string): New. Don't allow wide strings in
2035 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
2036 asm_string instead of STRING.
2038 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
2040 * c-typeck.c (constructor_no_implicit): Remove.
2041 (set_designator, process_init_element): Don't check
2042 constructor_no_implicit.
2044 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
2045 Kaz Kojima <kkojima@gcc.gnu.org>
2048 * lcm.c (create_pre_exit): New.
2049 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
2050 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
2052 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
2054 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
2055 -msoft-float in target_float_switch.
2056 * arm.c (arm_override_options): Fix processing of target_float_switch.
2058 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
2060 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
2063 2004-01-05 Uros Bizjak <uros@kss-loka.si>
2065 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
2066 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
2068 2005-01-04 Roger Sayle <roger@eyesopen.com>
2070 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
2071 if deleting the then-block allows the test-block to fallthru to the
2074 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
2077 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
2078 after the full declaration if the are in two different TUs.
2080 2005-01-04 Richard Henderson <rth@redhat.com>
2083 * tree-sra.c (generate_one_element_init): Just
2084 call gimplify_and_add.
2085 (generate_element_init): Record the
2086 new referenced variables and mark them for renaming
2087 and split out to ...
2088 (generate_element_init_1): This.
2089 (scalarize_init): Don't call push_gimplify_context/
2090 pop_gimplify_context.
2092 2005-01-04 Geoffrey Keating <geoffk@apple.com>
2094 * toplev.c (get_src_pwd): Handle failure of getpwd().
2096 2005-01-04 Roger Sayle <roger@eyesopen.com>
2098 * fold-const.c (fold_single_bit_test): Delete unreachable handling
2100 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
2101 all the cases handled by it are inverted by invert_truthvalue.
2103 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
2105 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
2106 (save_register_windowsi): Likewise.
2108 2005-01-04 Richard Henderson <rth@redhat.com>
2110 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
2112 2005-01-04 Uros Bizjak <uros@kss-loka.si>
2115 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
2116 any simplifications of vector mode comparison operators.
2117 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
2119 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
2120 Devang Patel <dpatel@apple.com>
2122 PR tree-optimization/18308
2123 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
2124 the operands before creating a new expression.
2125 * dojump.c (do_jump): Make drop_through_label available
2126 for all cases. Add expansion of COND_EXPR.
2128 2005-01-04 Ira Rosen <irar@il.ibm.com>
2130 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
2131 INTEGER_CST instead of TREE_CONSTANT.
2132 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
2133 instead of TREE_CONSTANT.
2134 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
2135 instead of TREE_CONSTANT.
2137 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
2139 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
2141 2005-01-04 Uros Bizjak <uros@kss-loka.si>
2144 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
2145 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
2147 2005-01-03 Richard Henderson <rth@redhat.com>
2149 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
2150 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
2151 fold_convert_const_real_from_real): Split out from ...
2152 (fold_convert_const): ... here.
2154 2005-01-03 Richard Henderson <rth@redhat.com>
2157 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
2158 (mov<MMXMODEI>_internal): Likewise.
2159 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
2160 Add V2SF case; use it for SSE1; don't use TI.
2161 (movdf_integer): Likewise.
2162 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
2164 2005-01-03 Ira Rosen <irar@il.ibm.com>
2166 * tree-vectorizer.c (vect_strip_conversions): New function.
2167 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
2168 check for binary class.
2170 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
2174 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
2176 * gimple-low.c (mark_blocks_with_used_vars): New function.
2177 (mark_blocks_with_used_subblocks): Ditto.
2178 (mark_used_blocks): Ditto.
2179 (pass_mark_used_blocks): New pass.
2180 * tree-inline.c: Include debug.h.
2181 (expand_call_inline): Call outlining_inline_function here.
2182 * tree-optimize.c (init_tree_optimization_passes): Add
2183 pass_mark_used_blocks.
2184 * tree-pass.h (pass_mark_used_blocks): New.
2185 * Makefile.in (tree-inline.o): Add debug.h dependency.
2187 2005-01-03 Geoffrey Keating <geoffk@apple.com>
2189 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
2190 VAR_DECLs to have weak_import attribute.
2192 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
2195 2004-01-03 Steven Bosscher <stevenb@suse.de>
2197 * passes.c (rest_of_compilation): Don't run regmove if only
2198 flag_expensive_optimizations, require flag_regmove instead.
2200 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
2202 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
2204 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
2205 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
2206 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
2207 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
2209 2005-01-03 Richard Henderson <rth@redhat.com>
2211 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
2212 (ia64_const_ok_for_letter_p): New.
2213 (ia64_const_double_ok_for_letter_p): New.
2214 (ia64_extra_constraint): New.
2215 (ia64_expand_vecint_compare): New.
2216 (ia64_expand_vcondu_v2si): New.
2217 (ia64_expand_vecint_cmov): New.
2218 (ia64_expand_vecint_minmax): New.
2219 (ia64_print_operand): Add 'v'.
2220 (ia64_preferred_reload_class): New.
2221 (ia64_vector_mode_supported_p): New.
2222 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
2223 (PREFERRED_RELOAD_CLASS): Move to function.
2224 (CONST_OK_FOR_LETTER_P): Move to function.
2225 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
2226 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
2227 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
2228 (EXTRA_CONSTRAINT): Move to function.
2229 * config/ia64/ia64.md: Include vect.md.
2230 (itanium_class): Add mmalua.
2232 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
2234 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
2235 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
2237 (const_int_2bit_operand): New.
2238 (fr_reg_or_0_operand): New.
2239 * config/ia64/ia64-modes.def: Add vector modes.
2240 * config/ia64/ia64-protos.h: Update.
2241 * config/ia64/vect.md: New file.
2243 2005-01-03 Richard Henderson <rth@redhat.com>
2245 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
2247 2005-01-03 Uros Bizjak <uros@kss-loka.si>
2250 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
2252 (log1pdf2): Change mode of operands[0,1] to DFmode.
2254 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
2256 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
2257 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
2258 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
2259 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
2260 SPARC_RELAXED_ORDERING.
2262 2005-01-03 Richard Henderson <rth@redhat.com>
2263 Uros Bizjak <uros@kss-loka.si>
2266 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
2267 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
2268 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
2270 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
2271 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
2272 const_pow2_1_to_128_operand): New predicates.
2274 2005-01-02 Greg McGary <greg@mcgary.org>
2276 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
2278 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
2279 object. [ptr_type]: Remove unused variable.
2280 Remember nearest addressable array-element or record-component
2281 when checking bitfield components. Tolerate empty BIND_EXPR.
2283 2005-01-01 Richard Henderson <rth@redhat.com>
2285 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
2286 that don't narrow the value. Fail for other conversions.
2288 2005-01-01 Richard Henderson <rth@redhat.com>
2291 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
2292 * c-lang.c (finish_file): Don't do it here.
2293 * objc/objc-act.c (objc_finish_file): Likewise.
2295 * cgraph.c (decl_assembler_name_equal): New.
2296 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
2297 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
2298 * cgraph.h (struct cgraph_varpool_node): Add next.
2299 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
2300 * varasm.c (assemble_alias): Mark the target as needed.
2302 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
2305 * function.c (get_arg_pointer_save_area): Use entry_of_function
2306 instead of get_insns.
2308 2005-01-01 Roger Sayle <roger@eyesopen.com>
2309 Andrew Pinski <pinskia@physics.uc.edu>
2310 James E. Wilson <wilson@specifixinc.com>
2312 PR rtl-optimization/12092
2313 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
2316 2005-01-01 Roger Sayle <roger@eyesopen.com>
2317 Olivier Hainque <hainque@act-europe.fr>
2319 * tree.c (int_fits_type_p): A narrower type always fits in a
2320 wider one, except for negative values into unsigned types.
2322 2005-01-01 Roger Sayle <roger@eyesopen.com>
2324 * tree.c (int_fits_type_p): Compare the result of force_fit_type
2325 with the original constant rather than require TREE_OVERFLOW.
2327 2005-01-01 Steven Bosscher <stevenb@suse.de>
2330 * c-decl.c (finish_function): If compiling C99, annotate the
2331 compiler generated return with the current file name and line 0.
2332 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
2333 the source line is greater than 0.
2334 (remove_bb): Likewise.
2336 See ChangeLog.12 for earlier changes.