1 2005-01-24 Paolo Bonzini <bonzini@gnu.org>
3 * rtlanal.c (non_rtx_starting_operands, for_each_rtx_1,
5 (for_each_rtx): Call for_each_rtx_1.
6 * rtl.h (init_rtlanal): Declare.
7 * toplev.c (backend_init): Call init_rtlanal.
9 2005-01-24 Jakub Jelinek <jakub@redhat.com>
11 * flow.c (propagate_one_insn): Formatting.
14 * flow.c (libcall_dead_p): Be more conservative if unsure.
15 If there are any instructions between insn and call, see if they are
16 all dead before saying the libcall is dead.
18 2005-01-24 Uros Bizjak <uros@kss-loka.si>
20 * config/i386/i386.md (*extendsfdf2_sse): Use "nonimmediate_operand"
21 operand constraint for operand 0.
23 2005-01-23 Richard Henderson <rth@redhat.com>
26 * tree-complex.c (expand_complex_div_wide): Set EDGE_FALLTHRU on
27 edges leading to join block.
29 2005-01-23 Roger Sayle <roger@eyesopen.com>
31 * expmed.c (expand_mult_highpart): Make static. Change type of
32 constant multiplier argument from unsigned HOST_WIDE_INT to rtx.
33 (expand_divmod): Updates calls to expand_mult_highpart by using
34 gen_int_mode to pass a CONST_INT rtx.
35 * rtl.h (expand_mult_highpart): Remove prototype.
37 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
39 * flow.c (regset_bytes, regset_size): Remove.
41 * cfganal.c: Fix a reference to Harvey's paper.
43 2005-01-23 Steven Bosscher <stevenb@suse.de>
45 PR rtl-optimization/19464
46 * tree-optimize.c (init_tree_optimization_passes): Add one more
47 copyrename pass just before out-of-ssa.
49 2005-01-23 Joseph S. Myers <joseph@codesourcery.com>
52 * recog.c (recog_memoized): Don't define if GENERATOR_FILE.
53 * ggc-none.c (ggc_free): Define.
55 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
57 * cse.c (max_reg, max_insn_uid): Remove.
58 (cse_main): Don't access max_reg or max_insn_uid.
60 * c-common.c (explicit_flag_signed_bitfields,
61 lang_statement_code_p, lang_expand_function_end): Remove.
62 * c-common.h: Remove the corresponding declarations.
63 * c-opts.c (c_common_handle_option): Remove all write access
64 to explicit_flag_signed_bitfields.
66 * cgraph.c (cgraph_varpool_n_nodes): Remove.
67 (cgraph_varpool_node): Don't access cgraph_varpool_n_nodes.
68 * cgraph.h: Remove the corresponding declaration.
70 * gcse.c (null_pointer_info): Remove.
72 2005-01-23 Roger Sayle <roger@eyesopen.com>
73 Eric Botcazou <ebotcazou@libertysurf.fr>
75 * combine.c (simplify_logical): Only simplify logical expressions
76 of the form ior(and(x,y),z) by the inverse distributive law if the
77 result is cheaper than the original.
79 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
81 * alias.c, c-common.h, c-incpath.c, c-incpath.h, expr.c,
82 fold-const.c, gimplify.c, params.h, tree-data-ref.c,
83 tree-if-conv.c, tree-nested.c, tree-outof-ssa.c,
84 tree-ssa-dom.c, tree-vectorizer.c, tree.def, config/darwin.c,
85 config/freebsd-spec.h, config/arm/arm.h,
86 config/h8300/h8300.md, config/i386/i386.md,
87 config/i386/predicates.md, config/i386/sse.md,
88 config/ia64/ia64.c, config/ip2k/ip2k.c, config/s390/s390.c,
89 config/vax/vax.md: Fix comment typos. Follow spelling
92 2005-01-23 Steven Bosscher <stevenb@suse.de>
94 * gengtype.c (walk_type): Produce `not equal to' compares for loop
95 tests, instead of `less than'.
97 2005-01-23 Richard Sandiford <rsandifo@redhat.com>
99 * function.c (assign_parm_setup_block): When creating a new stack slot
100 for a parameter, get its alignment from the parameter's DECL_ALIGN
101 rather than the type's TYPE_ALIGN. Make sure that the parameter
102 is at least word aligned.
104 2005-01-22 Roger Sayle <roger@eyesopen.com>
107 * config/avr/avr.c (avr_hard_regno_mode_ok): Rewrite.
109 2005-01-22 Richard Henderson <rth@redhat.com>
112 * config/i386/i386.md (movsfcc_1_sse_max): Use nonimmediate_operand
113 in both compare operands.
114 (movdfcc_1_sse_max): Likewise.
115 (movsfcc_1_sse): Likewise. Add earlyclobber for scratch.
116 (movdfcc_1_sse): Likewise.
117 * config/i386/i386.c (ix86_split_sse_movcc): Emit copies into the
118 scratch register as needed.
120 2005-01-22 Richard Henderson <rth@redhat.com>
122 * config/i386/i386.md (smaxsf3): Fix mnemonic typo.
124 * config/i386/i386.c (ix86_prepare_fp_compare_args): Fix is_sse test.
126 2005-01-22 Richard Henderson <rth@redhat.com>
128 * genopinit.c (optabs): Use smin/smax for floating point too.
129 * doc/md.texi: Update to match. Clarify that floating point
130 results are undefined for +0/-0 and NaN.
131 * doc/rtl.texi: Likewise.
132 * rtl.def (SMIN, SMAX): Likewise
133 * tree.def (MIN_EXPR, MAX_EXPR): Likewise.
135 * config/alpha/alpha.md (smaxdf3, smindf3, smaxsf3, sminsf3): Add
136 leading 's' to the name.
137 * config/ia64/ia64.md (smaxsf3, sminsf3, smaxdf3, smindf3,
138 smaxxf3, sminxf3): Likewise.
139 * config/rs6000/rs6000.md (smaxdf3, smindf3, smaxsf3, sminsf3):
142 2005-01-22 Paul Brook <paul@codesourcery.com>
144 * config/arm/arm.md: Use "Uy" constraint for wcgr load/stores.
146 2005-01-21 Mark Dettinger <dettinge@de.ibm.com>
148 * config/s390/s390.c (struct processor_costs): New fields
149 dlgr, dlr, dr, dsgfr, dsgr.
150 (z900_cost, z990_cost): Values for new fields.
151 (s390_rtx_costs): New cases MEM und COMPARE in switch
152 statement. Modified handling of SIGN_EXTEND, ZERO_EXTEND,
153 DIV, MOD, UDIV, UMOD.
155 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
157 * config/s390/s390.md ("doloop_si64"): Reload input value directly
158 into the register being decremented.
159 ("doloop_si31", "doloop_di"): Likewise.
160 ("*doloop_si_long"): Adapt pattern.
162 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
164 * config/s390/s390.h (HARD_REGNO_NREGS): Fix computation for
166 (CLASS_MAX_NREGS): Likewise.
168 2005-01-21 Daniel Berlin <dberlin@dberlin.org>
170 * doc/tree-ssa.texi (Statement Operands): Add example for new
171 must-def macro. Note deprecation of old operands interface.
173 2005-01-22 Richard Sandiford <rsandifo@redhat.com>
175 PR tree-optimization/19484
176 * tree-cfg.c (remove_fallthru_edge): New function.
177 (cleanup_control_flow): Remove fallthru edges from calls that are
178 now known not to return.
180 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
182 Fix PR tree-optimization/19038
183 * tree-ssa-dom.c (cprop_operand): Don't replace loop invaeriant
184 copies with loop variant ones.
186 2005-01-22 Kazu Hirata <kazu@cs.umass.edu>
188 * cfganal.c, real.h, reorg.c, timevar.def, tree-ssa-ccp.c,
189 config/alpha/alpha-protos.h, config/alpha/alpha.h,
190 config/alpha/alpha.md, config/alpha/predicates.md,
191 config/sparc/freebsd.h, config/sparc/netbsd-elf.h,
192 config/sparc/sol2.h: Update copyright.
194 * tree-cfg.c (remove_forwarder_block_with_phi): Look at the
195 first label to see if it is a nonlocal label.
197 2005-01-22 David Edelsohn <edelsohn@gnu.org>
198 Andrew Pinski <pinskia@physics.uc.edu>
201 * config/rs6000/rs6000.c (rs6000_va_start): Saturate n_gpr at
202 maximum number of GPRs. Saturate n_fpr at maximum number of FPRs.
204 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
206 * config/rs6000/rtems.h: Update copyright.
208 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
211 * config/rs6000/rtems.h: Resurrect cpp_os_rtems_spec from gcc < 3.4.
212 (CPP_OS_RTEMS_SPEC): New (From gcc-3.3's config/rs6000/sys4.h).
213 (SUBSUBTARGET_EXTRA_SPECS): Use CPP_OS_RTEMS_SPEC.
215 2005-01-22 Volker Reichelt <reichelt@igpm.rwth-aachen.de>
218 * c-typeck.c (convert_arguments): Check for error_mark_node.
220 2005-01-21 Ian Lance Taylor <ian@airs.com>
222 PR tree-optimization/13000
223 * tree-inline.c: Include "tree-flow.h".
224 (expand_call_inline): If warn_return_type, warn if non-void inline
225 function falls through.
226 * tree-cfg.c (execute_warn_function_return): Don't warn about
227 control reaching end if TREE_NO_WARNING is set. Set
229 * gimple-low.c (block_may_fallthru): Don't assume that SWITCH_EXPR
231 * gimplify.c (shortcut_cond_expr): Don't emit a jump over the else
232 branch if we don't need one.
233 * c-typeck.c: Include "tree-flow.h"
234 (c_finish_bc_stmt): Don't add a goto if the current statement
235 list doesn't fall through to the current point.
237 2005-01-21 Roger Sayle <roger@eyesopen.com>
239 PR rtl-optimization/576
240 * real.c (real_arithmetic): Change return type from void to bool
241 to return an indication that the result may be inexact.
242 * real.h (real_arithmeric): Update prototype.
243 * fold-const.c (const_binop): Don't constant fold floating
244 point expressions when the user specifies -frounding-math and
245 the result may depend upon the run-time rounding mode.
246 (fold_convert_const_real_from_real): Clean-up.
247 (fold_initializer): Ignore flag_rounding_math for initializers.
248 * simplify-rtx.c (simplify_binary_operation): Likewise, don't
249 constant fold FP operations with flag_rounding_math if the
250 result may depend upon the run-time rounding mode.
252 2005-01-21 Tom Tromey <tromey@redhat.com>
254 * c-cppbuiltin.c (define__GNUC__): Correct assertion.
256 2005-01-21 Kazu Hirata <kazu@cs.umass.edu>
258 * tree-cfg.c: Fix comment typos.
260 2005-01-21 J"orn Rennecke <joern.rennecke@st.com>
262 * passes.c (rest_of_handle_flow2): Close / open the flow2 dump file
263 around the call to rest_of_handle_branch_target_load_optimize.
265 2005-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
267 * config.gcc (sparc-*-netbsdelf*): Include dbxelf.h
268 (sparc64-*-openbsd*): Likewise.
269 (sparclite-*-elf*): Include tm-dwarf2.h.
270 (sparc86x-*-elf*): Likewise.
271 (sparc64-*-elf*): Likewise.
272 * config/sparc/linux64.h (DWARF2_DEBUGGING_INFO): Delete.
273 (DBX_DEBUGGING_INFO): Likewise.
274 * config/sparc/liteelf.h (DWARF2_DEBUGGING_INFO): Likewise.
275 (PREFERRED_DEBUGGING_TYPE): Likewise.
276 * config/sparc/netbsd-elf.h (PREFERRED_DEBUGGING_TYPE): Likewise.
277 * config/sparc/sp64-elf.h (DBX_DEBUGGING_INFO): Likewise.
278 (PREFERRED_DEBUGGING_TYPE): Likewise.
279 * config/sparc/sp86x-elf.h (DWARF2_DEBUGGING_INFO): Likewise.
280 (PREFERRED_DEBUGGING_TYPE): Likewise.
281 * config/sparc/sparc.h (DBX_DEBUGGING_INFO): Likewise.
283 2005-01-21 Hans-Peter Nilsson <hp@bitrange.com>
286 * combine.c (combine_simplify_rtx): Revert change of 2004-12-31.
288 2005-01-20 Paul Brook <paul@codesourcery.com>
290 * doc/rtl.texi: Document value extension requirements for CONST_INT.
292 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
294 * tree-cfg.c (tree_verify_flow_info): Check that a nonlocal
295 label is first in a sequence of labels.
297 * tree-cfg.c (tree_verify_flow_info): Fix a typo.
299 2005-01-20 Janis Johnson <janis187@us.ibm.com>
300 Giovanni Bajo <giovannibajo@gcc.gnu.org>
302 * doc/sourcebuild.texi (Test Directives): New.
304 2005-01-20 Roger Sayle <roger@eyesopen.com>
306 * real.c (real_floor): Don't leave the result uninitialized when
307 mode is VOIDMode, but return the unrounded intermediate value.
308 (real_ceil): Likewise.
310 2005-01-20 Paul Brook <paul@codesourcery.com>
312 * config/arm/arm.md (insv): Use gen_int_mode.
314 2005-01-20 Eric Botcazou <ebotcazou@libertysurf.fr>
316 * config/sparc/freebsd.h (ASM_OUTPUT_INTERNAL_LABELREF): Delete.
317 * config/sparc/linux.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
318 * config/sparc/linux64.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
319 * config/sparc/netbsd-elf.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
320 * config/sparc/sol2.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
322 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
325 * dwarf2out.c (concat_loc_descriptor): We don't know we can use
326 fbreg, so we have to assume we can't.
327 (loc_descriptor_from_tree_1): Ditto.
328 (containing_function_has_frame_base): New function.
329 (add_location_or_const_value_attribute): Use it.
330 Also try to generate a frame_base from a single element location
333 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
335 PR tree-optimization/15349
336 * timevar.def (TV_TREE_MERGE_PHI): New.
337 * tree-cfg.c (tree_forwarder_block_p): Add a new argument
339 (remove_forwarder_block, cleanup_forwarder_blocks): Adjust the
340 calls to tree_forwarder_block_p.
341 (remove_forwarder_block_with_phi, merge_phi_nodes,
342 gate_merge_phi, pass_merge_phi): New.
343 * tree-optimize.c (init_tree_optimization_passes): Add
345 * tree-pass.h: Add an extern for pass_merge_phi;
347 2005-01-20 Richard Henderson <rth@redhat.com>
350 * config/i386/emmintrin.h (_mm_castpd_ps, _mm_castpd_si128): New.
351 (_mm_castps_pd, _mm_castps_si128): New.
352 (_mm_castsi128_ps, _mm_castsi128_pd): New.
354 2005-01-20 Richard Henderson <rth@redhat.com>
357 * config/i386/mmintrin.h (_mm_cvtsi32_si64): Use
358 __builtin_ia32_vec_init_v2si.
359 (_mm_cvtsi64_si32): Use __builtin_ia32_vec_ext_v2si.
360 * config/i386/i386.c (IX86_BUILTIN_VEC_EXT_V2SI): New.
361 (ix86_init_mmx_sse_builtins): Create it.
362 (ix86_expand_builtin): Expand it.
363 (ix86_expand_vector_set): Handle V2SFmode and V2SImode.
364 * config/i386/mmx.md (vec_extractv2sf_0, vec_extractv2sf_1): New.
365 (vec_extractv2si_0, vec_extractv2si_1): New.
367 2005-01-20 Richard Henderson <rth@redhat.com>
370 * config/i386/i386.c (ix86_expand_push): New.
371 * config/i386/mmx.md (push<MMXMODE>1): New.
372 * config/i386/sse.md (push<SSEMODE>1): New.
373 * config/i386/i386-protos.h: Update.
375 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
377 * params.def, params.h, predict.c, tree-ssa-loop-im.c: Update
380 2005-01-20 Richard Henderson <rth@redhat.com>
382 * rtl.def (CONST_VECTOR): Use RTX_CONST_OBJ.
383 * rtl.h (CONSTANT_P): Don't special case CONST_VECTOR.
385 2005-01-19 Richard Henderson <rth@redhat.com>
388 * config/i386/i386.c (ix86_expand_vector_move_misalign): Convert
389 to V4SFmode in SSE1 fallback load path.
391 2005-01-19 Richard Henderson <rth@redhat.com>
393 * config/i386/i386.c (ix86_expand_vector_init_one_var): Fix typo
394 in QImode expansion to ix86_expand_vector_set.
395 (ix86_expand_vector_init_general): Fix typo in V8HImode recursive call.
397 2005-01-19 Richard Henderson <rth@redhat.com>
400 * config/i386/i386.c (ix86_preferred_reload_class): Return a proper
401 subclass of the input class.
402 (ix86_secondary_memory_needed): Always true for cross-MMX classes.
403 Always true for cross-SSE1 classes. Rationalize conditionals.
404 * config/i386/i386.h (SSE_CLASS_P, MMX_CLASS_P): Use straight equality.
405 * config/i386/i386.md (movsi_1): Add MMX/SSE zeros. Fix alternatives
406 for SSE1. Don't check TARGET_INTER_UNIT_MOVES.
407 (movdi_2): Add MMX/SSE zeros.
408 (movdi_1_rex64): Likewise. Don't check TARGET_INTER_UNIT_MOVES.
409 (movsf_1): Don't check TARGET_INTER_UNIT_MOVES.
410 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
411 (movsi_1_nointernunit, movdi_1_rex64_nointerunit): Remove.
412 (movsf_1_nointerunit, zero_extendsidi2_32_1): Remove.
413 (zero_extendsidi2_rex64_1): Remove.
414 (MOV0 peephole): Check GENERAL_REG_P.
416 2005-01-19 Richard Henderson <rth@redhat.com>
419 * config/i386/i386.c (ix86_expand_vector_set): Fix third and fourth
421 (ix86_expand_vector_extract): Likewise.
423 2005-01-19 Richard Henderson <rth@redhat.com>
426 * expr.c (emit_move_change_mode): New argument force; use
427 simplify_gen_subreg if true.
428 (emit_move_via_alt_mode): Merge into ...
429 (emit_move_via_integer): ... here.
430 (emit_move_ccmode): Use emit_move_change_mode directly.
432 2005-01-19 Richard Henderson <rth@redhat.com>
435 * config/alpha/alpha.c (alpha_rtx_costs): Handle HIGH.
436 (alpha_preferred_reload_class): Handle CONST_VECTOR.
437 (alpha_emit_set_const_1): Add no_output parameter; don't emit
439 (alpha_emit_set_const): Likewise. Make static.
440 (alpha_emit_set_long_const): Make static.
441 (alpha_extract_integer): Split out from alpha_expand_mov.
442 (alpha_split_const_mov): Likewise.
443 (alpha_expand_mov): Use them. Handle CONST_VECTOR.
444 (alpha_legitimate_constant_p): New.
445 * config/alpha/alpha-protos.h: Update.
446 * config/alpha/alpha.h (REGISTER_MOVE_COST): Correct fp<->gp cost.
447 (LEGITIMATE_CONSTANT_P): Re-implement with a function.
448 * config/alpha/alpha.md (movsi): Add n alternative.
449 (movsi_nt_vms, movdi_er_nofix, movdi_er_fix, movdi_fix): Likewise.
450 (mov<VEC>_fix, mov<VEC>_nofix): Add i alternative.
451 (splitters for all of the above): Use alpha_split_const_mov.
452 * config/alpha/predicates.md (non_add_const_operand): New.
453 (non_zero_const_operand): New.
454 (input_operand): Use alpha_legitimate_constant_p after reload.
456 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
458 PR tree-optimization/19038
459 * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): New function.
460 (add_candidate): Add ivs with increment in latch only if
461 allow_ip_end_pos_p is true.
462 (determine_iv_cost): Use empty_block_p.
464 2005-01-19 Daniel Berlin <dberlin@dberlin.org>
466 * cfganal.c (compute_dominance_frontiers_1): Replace with new algorithm
467 (compute_dominance_frontiers): Ditto.
469 2005-01-19 Ralf Corsepius <ralf.corsepius@rtems.org>
472 * config/sh/t-rtems: New.
473 * config.gcc (sh-*-rtems*): Reflect having added config/sh/t-rtems.
475 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
477 * tree-ssa-ccp.c (maybe_fold_offset_to_component_ref): Always subtract
478 the offset of the selected field.
480 2005-01-19 Kazu Hirata <kazu@cs.umass.edu>
482 * tree-cfg.c (remove_forwarder_block): Fix the check to
483 prevent a nonlocal label from appearing in the middle of a
486 2005-01-19 Hans-Peter Nilsson <hp@axis.com>
488 PR rtl-optimization/19462
489 * reorg.c (find_end_label): Create return insn only if
490 current_function_epilogue_delay_list is empty.
492 2005-01-19 Paolo Bonzini <bonzini@gnu.org>
494 * doc/passes.texi: Remove paragraph mentioning new-ra.
496 2005-01-19 Jakub Jelinek <jakub@redhat.com>
498 PR rtl-optimization/15139
499 * combine.c: Include params.h.
500 (count_rtxs): New function.
501 (record_value_for_reg): If replace_rtx would replace at least
502 2 occurrences of REG in VALUE and TEM is really large, replace REG with
503 (clobber (const_int 0)) instead of TEM.
504 * params.def (PARAM_MAX_LAST_VALUE_RTL): New.
505 * params.h (MAX_LAST_VALUE_RTL): New.
506 * Makefile.in (combine.o): Depend on $(PARAMS_H).
507 * doc/invoke.texi (--param max-last-value-rtl=N): Document.
510 * c-typeck.c (digest_init): Only call build_vector if all constructor
511 elements are *_CST nodes.
512 * gimplify.c (gimplify_init_constructor): Likewise.
515 * c-typeck.c (digest_init): Only call build_vector if inside_init
518 2005-01-18 Tobias Schl"uter <tobias.schlueter@physik.uni-muenchen.de>
520 * toplev.c (init_asm_output): Remove dead #ifdef.
522 2005-01-18 Kaz Kojima <kkojima@gcc.gnu.org>
524 * config/sh/linux.h (TARGET_C99_FUNCTIONS): Define.
526 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
528 * calls.c (expand_call): Check DECL_BUILT_IN_CLASS before
529 accessing DECL_FUNCTION_CODE.
530 * dojump.c (do_jump): Likewise.
531 * gimplify.c (gimplify_call_expr): Likewise.
532 * predict.c (expr_expected_value): Likewise.
533 (strip_builtin_expect): Likewise.
534 * tree-inline.c (estimate_num_insns_1): Likewise.
535 * tree-ssa-loop-im.c (stmt_cost): Likewise
536 * fold-const.c (fold): Test for BUILT_IN_NORMAL.
537 (tree_expr_nonnegative_p): Likewise.
539 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
541 * config/sparc/sparc.c (load_pic_register): Emit the appropriate
542 variant of the load_pcrel_sym pattern.
543 * config/sparc/sparc.md (P macro): Move to the top.
544 (load_pcrel_sym): Macroize using P.
545 (save_register_window): Likewise.
547 2005-01-18 Aldy Hernandez <aldyh@redhat.com>
549 * config/rs6000/rs6000.md ("sunordered"): Disable for e500.
552 2005-01-18 Andrew Pinski <pinskia@physics.uc.edu>
555 * c-typeck.c (build_asm_expr): Strip nops off
556 input memory operands.
558 2005-01-18 David Edelsohn <edelsohn@gnu.org>
560 * config/rs6000/aix43.h (CPLUSPLUS_CPP_SPEC): Delete
561 _XOPEN_SOURCE, _XOPEN_SOURCE_EXTENDED, _LARGE_FILE_API.
562 * config/rs6000/aix51.h (CPLUSPLUS_CPP_SPEC): Same.
563 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Same.
565 2005-01-18 Joel Sherrill <joel.sherrill@oarcorp.com>
568 * config/i386/i386.c (override_options): If the 80387 is disabled,
569 then do not return FP values using FP registers.
571 2005-01-18 Kazu Hirata <kazu@cs.umass.edu>
573 * basic-block.h, c-common.c, c-cppbuiltin.c, c-lang.c,
574 c-tree.h, cfgbuild.c, cgraph.c, cgraph.h, collect2.c,
575 combine.c, config.gcc, coverage.h, cse.c, cselib.c,
576 defaults.h, df.c, dwarf2asm.c, dwarf2out.c, explow.c, expr.c,
577 flow.c, fold-const.c, gcse.c, ggc-page.c, gimple-low.c,
578 gimplify.c, ifcvt.c, langhooks-def.h, lcm.c, optabs.h,
579 output.h, postreload-gcse.c, postreload.c, recog.c,
580 resource.c, rtl.def, rtlanal.c, sched-deps.c, sched-rgn.c,
581 targhooks.h, toplev.c, tree-data-ref.c, tree-eh.c,
582 tree-flow.h, tree-loop-linear.c, tree-mudflap.h, tree-nrv.c,
583 tree-optimize.c, tree-outof-ssa.c, tree-pass.h,
584 tree-scalar-evolution.c, tree-ssa-copy.c, tree-ssa-dce.c,
585 tree-ssa-dse.c, tree-ssa-loop-ivopts.c, tree-ssa-operands.c,
586 tree-ssa-pre.c, tree-ssa.c, tree-vectorizer.c, tree.def,
587 unwind-dw2-fde-darwin.c, var-tracking.c: Update copyright.
589 * config/darwin.c, config/darwin.h, config/freebsd-spec.h,
590 config/alpha/alpha.c, config/arm/vfp.md,
591 config/i386/emmintrin.h, config/i386/i386-modes.def,
592 config/i386/i386-protos.h, config/i386/i386.h,
593 config/i386/pmmintrin.h, config/i386/ppro.md,
594 config/i386/predicates.md, config/i386/xmmintrin.h,
595 config/i860/i860.c, config/ia64/ia64-modes.def,
596 config/ia64/ia64-protos.h, config/ia64/ia64.h,
597 config/ia64/ia64.md, config/ia64/itanium1.md,
598 config/ia64/itanium2.md, config/ia64/predicates.md,
599 config/iq2000/iq2000.h, config/mips/linux64.h,
600 config/rs6000/aix41.h, config/rs6000/aix43.h,
601 config/rs6000/aix52.h, config/rs6000/darwin-fallback.c,
602 config/rs6000/rs6000.c, config/rs6000/rs6000.h,
603 config/rs6000/spe.md, config/sh/sh.md, config/sparc/linux.h,
604 config/sparc/linux64.h, config/sparc/litecoff.h,
605 config/sparc/sp64-elf.h, config/sparc/sparc.c,
606 config/sparc/sparc.h, config/sparc/sparc.md,
607 config/sparc/sysv4-only.h: Update copyright.
609 2005-01-18 Richard Henderson <rth@redhat.com>
612 * config/i386/i386.c (ix86_expand_fp_movcc): Fail for LTGT and UNEQ.
614 2005-01-18 Marc Espie <espie@openbsd.org>
615 * config/i386/openbsdelf.h: Typo.
617 2005-01-18 Richard Henderson <rth@redhat.com>
619 * config/i386/i386.c (ix86_expand_fp_movcc): Remove TARGET_IEEE_FP
620 special case for sse.
621 * config/i386/predicates.md (sse_comparison_operator): Likewise.
623 2005-01-18 Dorit Naishlos <dorit@il.ibm.com>
625 * tree-inline.c (estimate_num_insns_1): Added cases for
626 ALIGN_INDIRECT_REF, MISALIGNED_INDIRECT_REF, and REALIGN_LOAD_EXPR.
628 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
630 PR rtl-optimization/19296
631 * combine.c (simplify_comparison): Rewrite the condition under
632 which a non-paradoxical SUBREG of a PLUS can be lifted when
633 compared against a constant.
635 2005-01-18 Andi Kleen <ak@muc.de>
637 * c-typeck.c: (convert_for_assignment): Check warn_pointer_sign.
638 * c.opt (-Wpointer-sign): Add.
639 * doc/invoke.texi: (-Wpointer-sign): Add.
641 2005-01-18 Uros Bizjak <uros@kss-loka.si>
643 * config/i386/i386.c (override_options): Revert 2004-11-24 change.
644 * config/i386/i386.md (fmodsf3, dremsf3, *sinsf2, *cossf2, sincossf3,
645 *tansf3_1, tansf2, atan2sf3_1, atan2sf3, atansf2, asinsf2, acossf2,
646 logsf2, log10sf2, log2sf2, log1psf2, logbsf2, ilogbsf2, expsf2,
647 exp10sf2, exp2sf2, expm1sf2, rintsf2, floorsf2, ceilsf2, btruncsf2,
648 nearbyintsf2): Disable for TARGET_SSE_MATH. Leave patterns enabled
649 for TARGET_MIX_SSE_I387.
650 (fmoddf3, dremdf3, *sindf2, *sinextendsfdf2, *cosdf2, *cosextendsfdf2,
651 sincosdf3, *sincosextendsfdf3, *tandf3_1, tandf2, atan2df3_1,
652 atan2df3, atandf2, asindf2, acosdf2, logdf2, log10df2, log2df2,
653 log1pdf2, logbdf2, expdf2, exp10df2, exp2df2, expm1df2, rintdf2,
654 floordf2, ceildf2, btruncdf2, nearbyintdf2): Disable for
655 (TARGET_SSE2 && TARGET_SSE_MATH). Leave patterns enabled for
657 (atan2sf3, atan2df3, atan2xf3): Remove register constraints
660 2005-01-18 Uros Bizjak <uros@kss-loka.si>
663 * config/i386/mmx.md (*movv2sf_internal_rex64, *movv2sf_internal):
664 Add movaps alternative for xmm reg->reg move.
666 2005-01-17 Jeff Law <law@redhat.com>
668 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Clear SSA_NAME_VALUE
669 at the end of the main DOM loop rather than just before DOM exits.
671 2005-01-17 Diego Novillo <dnovillo@redhat.com>
673 PR tree-optimization/19121
674 * tree-ssa-alias.c (compute_flow_sensitive_aliasing): When
675 adding aliases to a name tag, also add them to the pointer's
677 * tree-ssa-copy.c (merge_alias_info): Do not merge flow
678 sensitive alias info at all. Only check that the two pointers
679 have compatible pointed-to sets.
680 * tree-ssa.c (verify_name_tags): Verify that the alias set of
681 a pointer's type tag is a superset of the alias set of the
684 2005-01-17 James E Wilson <wilson@specifixinc.com>
687 * config/ia64/ia64.md (movxf): Handle general register source. Adjust
688 comment to document why.
690 2005-01-17 Richard Henderson <rth@redhat.com>
692 * config/i386/sse.md (smaxv4sf3_finite, sse_vmsmaxv4sf3_finite,
693 sminv4sf3_finite, sse_vmsminv4sf3_finite, smaxv2df3_finite,
694 sse2_vmsmaxv2df3_finite, sminv2df3_finite,
695 sse2_vmsminv2df3_finite): New.
696 (smaxv4sf3, sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, smaxv2df3,
697 sse2_vmsmaxv2df3, sminv2df3, sse2_vmsminv2df3): Remove commutative.
698 Force op1 into register.
700 2005-01-17 Kelley Cook <kcook@gcc.gnu.org>
702 * Makefile.in (STAGEMOVESTUFF): Stage all the stamp files to prevent
703 bubblestrap from always rebuilding build directory.
705 2005-01-17 Ulrich Weigand <uweigand@de.ibm.com>
707 * config/s390/s390.md ("*clc"): Remove incorrect '='.
708 ("*cmpmem_short"): Likewise.
710 2005-01-17 Steven Bosscher <stevenb@suse.de>
712 * basic-block.h: Document BB_* flags.
713 * regrename.c (copyprop_hardreg_forward): Don't use BB_VISITED,
714 use an sbitmap instead.
715 * sched-rgn.c (compute_trg_info): Likewise.
717 2005-01-17 Richard Sandiford <rsandifo@redhat.com>
719 * config.gcc (mips64*-*-linux*): Set the default abi to n32. Remove
720 redundant target_cpu_default and extra_parts lines. Use the default
722 * config/mips/linux64.h (DRIVER_DEFAULT_ABI_SELF_SPEC): Delete.
723 (SUBTARGET_EXTRA_SPECS): Delete.
724 (DRIVER_SELF_SPECS): Remove architecture lines. Set the ABI directly.
725 (SUBTARGET_ASM_SPEC): Remove -mabi=n32 and -mabi=64 mappings.
726 * config/mips/t-linux64 (MULTILIB_DIRNAMES): Use "32" for -mabi=32
727 and "n32" for -mabi=n32.
729 2005-01-17 Ian Lance Taylor <ian@airs.com>
732 * c-typeck.c (build_c_cast): Revert patch of 2003-02-16: don't
733 fold constant variables into initial values.
735 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
737 * tree-cfg.c (tree_can_merge_blocks_p): Reorder two checks.
739 * tree-cfg.c (tree_forwarder_block_p): Speed up by walking
740 through the statements backward.
742 2005-01-17 Ian Lance Taylor <ian@airs.com>
745 * tree-inline.c (expand_call_inline): Set TREE_NO_WARNING on
746 a variable set to the return value of the inlined function.
748 2005-01-17 Mark Dettinger <dettinge@de.ibm.com>
750 * rtlanal.c (rtx_cost): Assign cost of 0 to a SUBREG
751 when modes are tieable.
753 2005-01-17 Ranjit Mathew <rmathew@hotmail.com>
755 * gthr-posix.h (__gthread_active_p): Use pthread_cancel instead
756 of pthread_create to find out if threads are enabled.
757 * gthr-posix95.h (__gthread_active_p): Likewise.
759 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
761 * common.opt (-fnew-ra): Remove.
763 * toplev.h (flag_new_regalloc): Remove.
764 * Makefile.in (ra*.*): Don't mention.
765 * passes.c (rest_of_handle_new_regalloc): Remove.
766 (rest_of_handle_combine, rest_of_compilation): Always consider
767 flag_new_regalloc as false.
768 * doc/invoke.texi: Don't document -fnew-ra.
770 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
772 * bb-reorder.c (fix_edges_for_rarely_executed_code): Remove
773 last parameter to reg_scan.
774 * loop.c (loop_optimize): Likewise.
775 * passes.c (rest_of_handle_tracer, rest_of_handle_if_conversion,
776 rest_of_handle_web, rest_of_handle_cfg, rest_of_handle_jump_bypass,
777 rest_of_handle_life, rest_of_handle_cse, rest_of_handle_cse2,
778 rest_of_handle_gcse, rest_of_handle_loop_optimize,
779 rest_of_handle_loop2, rest_of_handle_jump2): Likewise.
780 * regclass.c (reg_scan): Likewise, for the declaration.
781 * rtl.h (reg_scan): Likewise, for the prototype.
783 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
785 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Don't call
788 2005-01-16 Daniel Berlin <dberlin@dberlin.org>
790 * tree-ssa-pre.c (add_to_sets): s1 may be NULL.
791 (compute_avail): Uses don't go in tmp_gen.
793 2005-01-16 Steven Bosscher <stevenb@suse.de>
795 * ggc-page.c (ggc_alloc_stat): Use __builtin_ctzl instead of a
796 loop to look for a free slot in a page entry.
798 2005-01-16 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
801 * defaults.h (TARGET_DEFERRED_OUTPUT_DEFS): Provide default.
802 * toplev.c (compile_file): Call process_pending_assemble_output_defs
803 just before targetm.asm_out.file_end.
804 * tree.h (process_pending_assemble_output_defs): Declare.
805 * varasm.c (assemble_output_def, process_pending_assemble_output_defs):
807 (assemble_alias): Defer generation of assembly code for defines when
808 TARGET_DEFERRED_OUTPUT_DEFS is true.
809 * config/rs6000/aix41.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
810 * config/rs6000/aix43.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
811 * doc/tm.texi (TARGET_DEFERRED_OUTPUT_DEFS): document.
813 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
816 * pa.c (pa_scalar_mode_supported_p): New function.
817 (TARGET_SCALAR_MODE_SUPPORTED_P): Define.
818 * pa.h (MIN_UNITS_PER_WORD): Add comment.
820 * pa-protos.h (prefetch_operand): Delete.
821 (prefetch_cc_operand, prefetch_nocc_operand): New declations.
822 * pa.c (prefetch_operand): Delete.
823 (prefetch_cc_operand, prefetch_nocc_operand): New functions.
824 * pa.h (EXTRA_CONSTRAINT): Add `W' constraint.
825 (PREDICATE_CODES): Delete prefetch_operand. Add prefetch_cc_operand
826 and prefetch_nocc_operand.
827 * pa.md (prefetch): Rework to avoid reload problems handling short
828 displacements when a cache control completer needs to be provided.
829 (prefetch_32, prefetch_64): Delete.
830 (prefetch_cc, prefetch_nocc): New patterns.
832 2005-01-15 David Edelsohn <edelsohn@gnu.org>
834 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Revert previous change.
836 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
839 * pa.c (pa_asm_output_aligned_common): Warn if specified alignment
840 exceeds maximum alignment for global common data.
841 * pa64-hpux.h (MAX_OFILE_ALIGNMENT): Define to 32768.
842 * som.h (MAX_OFILE_ALIGNMENT): Likewise.
844 2004-01-15 Roger Sayle <roger@eyesopen.com>
846 * tree-ssa-dom.c (extract_range_from_cond): Correct condition.
848 2004-01-15 Roger Sayle <roger@eyesopen.com>
850 * harg-reg-set.h (reg_class_names): Prototype global array.
851 * regclass.c (reg_class_names): Declare here and initialize to
853 (dump_regclass): Remove local declaration of reg_class_names.
854 (regclass): Likewise.
855 * cfg.c (dump_flow_info): Likewise.
856 * ra-debug.c (reg_class_names): Likewise.
857 * regrename.c (reg_class_names): Likewise.
858 * reload.c (reg_class_names): Likewise.
859 * reload1.c (spill_failure): Likewise.
860 * config/m68hc11/m68hc11.c (reg_class_names): Likewise.
862 2005-01-15 Ulrich Weigand <uweigand@de.ibm.com>
864 * config/s390/s390.md ("reload_outti"): Remove predicate for
865 output operand. Abort if operand is not a MEM.
866 ("reload_outdi", "reload_outdf"): Likewise.
868 2005-01-15 Marc Espie <espie@openbsd.org>
870 * config.gcc (*-*-openbsd*): Set HAS_LIBC_R for OpenBSD <= 3.2.
871 (i[34567]86-*-openbsd*): Switch to ELF for OpenBSD >= 3.4.
872 * config/openbsd.h: Fix C++ includes for native configurations.
873 Add proper OS_CPP_BUILTINS. Fix libspec for recent OpenBSD.
874 Add trampoline support.
875 * config/i386/openbsdelf.h: New.
877 2005-01-15 Marc Espie <espie@openbsd.org>
879 * collect2.c (main): Explicitly parse -dynamic-linker option.
881 2005-01-15 Jakub Jelinek <jakub@redhat.com>
883 PR tree-optimization/19060
884 * tree-ssa-dom.c (extract_range_from_cond) <case LT_EXPR, GT_EXPR>:
885 Return 0 if op1 <= TYPE_MIN_VALUE () resp. op1 >= TYPE_MAX_VALUE ().
886 (simplify_cond_and_lookup_avail_expr): Add assert for dummy == 0
887 and handle extract_range_from_cond returning false.
888 * fold-const.c (fold): Optimize comparisons with min/max even for
889 width > HOST_BITS_PER_WIDE_INT.
891 2005-01-15 Ralf Corsepius <ralf.corsepius@rtems.org>
893 * config/mips/rtems.h (MIPS_DEFAULT_GVALUE): Set to 0.
894 * config/mips/t-rtems (MULTILIBS_DIRNAMES,MULTILIB_OPTIONS):
895 Remove little endian multilib variants.
896 Add mips32 multilib variant.
898 2005-01-15 Kazu Hirata <kazu@cs.umass.edu>
900 * explow.c (copy_all_reg, stabilize): Remove.
901 * expr.h: Remove the corresponding prototypes.
903 * rtlanal.c (regs_set_between_p): Remove.
904 * rtl.h: Remove the corresponding prototype.
906 2005-01-14 David Edelsohn <edelsohn@gnu.org>
908 * config/rs6000/rs6000.c (rs6k_nonimmediate_operand): Rename
909 to rs6000_nonimmediate_operand.
910 * config/rs6000/rs6000.h (PREDICATE_CODES): Rename
911 rs6k_nonimmediate_operand to rs6000_nonimmediate_operand.
912 * config/rs6000/rs6000.md (movsi_internal1): Rename predicate to
913 rs6000_nonimmediate_operand.
914 * config/rs6000/spe.md (movdf_e500_double): Rename predicate to
915 rs6000_nonimmediate_operand.
917 2005-01-14 Andrew Pinski <pinskia@physics.uc.edu>
919 * tree-ssa-alias.c (pass_may_alias): Add TODO_verify_stmts.
921 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
923 * config/rs6000/rs6000.h (CLASS_MAX_NREGS): DF goes in 1 register
925 (CANNOT_CHANGE_MODE_CLASS): Restrict DI mode changes on e500v2.
926 (PREDICATE_CODES): Add rs6k_nonimmediate_operand.
928 * config/rs6000/rs6000.c (invalid_e500_subreg): New.
929 (rs6k_nonimmediate_operand): New.
930 (rs6000_legitimate_offset_address_p): Handle DI modes on e500v2
932 (legitimate_lo_sum_address_p): Same.
933 (rs6000_legitimize_address): Same.
934 (rs6000_legitimize_reload_address): Same.
935 (rs6000_legitimate_address): Same.
936 (spe_build_register_parallel): Pass DF and DC modes in a DI
939 * config/rs6000/rs6000.md ("*movsi_internal1"): Change predicate
940 to rs6k_nonimmediate_operand.
942 * config/rs6000/spe.md ("*frob_df_di"): New.
943 ("*frob_di_df"): New.
944 ("*frob_di_df_2"): New.
945 ("*mov_sidf_e500_subreg0"): New.
946 ("*mov_sidf_e500_subreg4"): New.
947 ("*movdf_e500_double"): Change predicate to
948 rs6k_nonimmediate_operand.
950 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
952 * postreload.c (move2add_note_store): Only call
953 trunc_int_for_mode on scalar integers.
955 2005-01-14 Steven Bosscher <stevenb@suse.de>
957 * tree-ssa-dce.c (visited_control_parents): New sbitmap to
958 replace BB_VISITED uses.
959 (find_obviously_necessary_stmts): Don't clear BB_VISITED.
960 (propagate_necessity): Check the bitmap instead of BB_VISITED.
961 (tree_dce_done): Free visited_control_parents.
962 (perform_tree_ssa_dce): Allocate and clear it.
963 * tree-ssa-pre.c (compute_antic_aux): Make non-recursive.
964 (compute_antic): Iterate from here using a DFS. Use an sbitmap
965 instead of BB_VISITED.
967 2005-01-14 Kazu Hirata <kazu@cs.umass.edu>
969 * c-tree.h, coverage.h, langhooks-def.h, optabs.h, output.h,
970 rtl.h, targhooks.h, tree-flow.h, tree-mudflap.h, tree.h:
971 Remove unused prototypes.
973 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
975 * config/rs6000/rs6000.c (rs6000_stack_info): Fix caching of SPE
976 64-bit register usage.
977 (spe_func_has_64bit_regs_p): Add FIXME note.
979 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
981 * config/rs6000/rs6000.c (rs6000_generate_compare): Replace
982 flag_finite_math_only with flag_unsafe_math_optimizations.
984 2005-01-14 J. D. Johnston <jjohnst@us.ibm.com>
986 * config/s390/tpf-unwind.h (s390_fallback_frame_state): Correct
988 (__tpf_eh_return): Copy TPF private stack area from the module
989 boundary stack frame. Add check for when module addresses are
990 equal, but no stub address is found.
992 2005-01-14 Richard Earnshaw <rearnsha@arm.com>
995 * arm.h (struct machine_function): Add call_via field.
996 (thumb_call_via_label): Declare.
997 * arm.c (thumb_call_via_label): New variable.
998 (thumb_call_reg_needed): New variable.
999 (arm_output_function_epilogue): For Thumb code, output any per-function
1000 call-indirect trampolines.
1001 (thumb_call_via_reg): New function.
1002 (arm_file_end): New function.
1003 (TARGET_ASM_FILE_END): Call arm_file_end.
1004 (aof_file_end): Likewise.
1005 * arm-protos.h (thumb_call_via_reg): Declare.
1006 * arm.md (call_reg_thumb, call_value_reg_thumb): Call
1007 thumb_call_via_reg in normal case.
1009 2005-01-14 Jakub Jelinek <jakub@redhat.com>
1012 PR rtl-optimization/19348
1013 * recog.c (peephole2_optimize): Do global life update if some peephole
1014 decides it doesn't need at least one of its inputs and that change
1015 influences liveness at the start of the basic block.
1017 * basic-block.h (EXECUTE_IF_AND_COMPL_IN_REG_SET): Needs 2 REGSET
1018 arguments instead of 1.
1020 2005-01-14 Eric Botcazou <ebotcazou@libertysurf.fr>
1023 * varasm.c (initializer_constant_valid_p) <ADDR_EXPR>: Return
1024 zero for nested functions needing a static chain or functions
1025 with a non-constant address.
1027 2005-01-13 Roger Sayle <roger@eyesopen.com>
1029 * simplify-rtx.c (simplify_binary_operation) <AND>: Optimize
1030 (and (sign_extend X) C) into (zero_extend (and X C)).
1032 2005-01-13 David O'Brien <obrien@FreeBSD.org>
1034 * config/freebsd-spec.h: Make KSE pthread lib logic the default.
1036 2005-01-13 Richard Henderson <rth@redhat.com>
1041 * config/i386/i386.md (cmpdf, cmpsf, bunordered, bordered, buneq,
1042 bunge, bungt, bunle, bunlt, bltgt): Enable for TARGET_SSE_MATH,
1043 not just TARGET_SSE.
1044 (cmpfp_i_387): Rename from cmpfp_i. Move after sse patterns.
1045 (cmpfp_i_mixed): Rename from cmpfp_i_sse; use for TARGET_MIX_SSE_I387.
1046 (cmpfp_i_sse): Rename from cmpfp_i_sse_only; use for TARGET_SSE_MATH.
1047 (cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387): Similarly.
1048 (fp_jcc_1_mixed, fp_jcc_1_sse, fp_jcc_1_387): Similarly.
1049 (fp_jcc_2_mixed, fp_jcc_2_sse, fp_jcc_2_387): Similarly.
1050 (fp_jcc_3_387, fp_jcc_4_387, fp_jcc_5_387, fp_jcc_6_387,
1051 fp_jcc_7_387, fp_jcc_8_387): Rename from fp_jcc_N.
1052 (movdicc_c_rex64): Rename with '*'.
1053 (movsfcc, movdfcc): Add checks for 387 and sse math to condition.
1054 (movsfcc_1_sse_min, movsfcc_1_sse_max, movsfcc_1_sse): New.
1055 (movsfcc_1_387): Rename from movsfcc_1.
1056 (movdfcc_1_sse_min, movdfcc_1_sse_max, movdfcc_1_sse): New.
1057 (movdfcc_1, movdfcc_1_rex64): Add check for 387.
1058 (sminsf3, smaxsf3, smindf3, smaxdf3): New.
1059 (minsf3, minsf, minsf_nonieee, minsf_sse, mindf3, mindf,
1060 mindf_nonieee, mindf_sse, maxsf3, maxsf, maxsf_nonieee, maxsf_sse,
1061 maxdf3, maxdf, maxdf_nonieee, maxdf_sse, sse_movsfcc, sse_movsfcc_eq,
1062 sse_movdfcc, sse_movdfcc_eq, sse_movsfcc_const0_1,
1063 sse_movsfcc_const0_2, sse_movsfcc_const0_3, sse_movsfcc_const0_4,
1064 sse_movdfcc_const0_1, sse_movdfcc_const0_2, sse_movdfcc_const0_3,
1065 sse_movdfcc_const0_4): Remove.
1066 * config/i386/i386.c (ix86_expand_fp_movcc): For TARGET_SSE_MATH,
1067 recognize min/max early. Update for changed sse cmove patterns.
1068 (ix86_split_sse_movcc): New.
1069 * config/i386/i386-protos.h: Update.
1071 2005-01-13 Steven Bosscher <stevenb@suse.de>
1073 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
1074 (fix_stmt_v_may_defs): Likewise.
1076 2005-01-13 Richard Henderson <rth@redhat.com>
1078 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
1079 for converting the mask.
1081 2005-01-13 David Edelsohn <edelsohn@gnu.org>
1083 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
1086 2005-01-13 Richard Henderson <rth@redhat.com>
1088 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
1089 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
1090 IX86_BUILTIN_MOVDQ2Q): Remove.
1091 (IX86_BUILTIN_VEC_EXT_V4SI): New.
1092 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
1093 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
1095 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
1096 (sse2_pextrw): Fix immediate constraint.
1097 (sse2_loadq, sse2_loadq_rex64): Remove.
1098 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
1099 Use __builtin_ia32_vec_ext_<size>.
1100 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
1102 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
1104 * function.c (assign_parm_setup_block): Look inside original
1105 entry_parm when inspecting PARALLEL.
1107 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
1108 Joel Sherrill <joel@oarcorp.com>
1111 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
1112 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
1113 rtems_gxx_recursive_mutex_init.
1114 (__gthread_recursive_mutex_lock): New function.
1115 (__gthread_recursive_mutex_trylock): Likewise.
1116 (__gthread_recursive_mutex_unlock): Likewise.
1118 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
1120 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
1122 2005-01-13 Jan Beulich <jbeulich@novell.com>
1124 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
1125 (ashrti3_internal): Indicate output is early clobber. Generate result
1126 into output rather than first input. Use move for low word of output
1127 if shift count is exactly 64.
1128 (lshrti3_internal): Likewise.
1130 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
1134 * reload1.c (delete_output_reload): Don't delete an output reload
1135 if the pseudo lives longer than a single basic block. Adjust and
1136 improve wording of comment.
1138 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
1141 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
1142 of TYPE_SIZE where we mean the number of bytes.
1144 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
1146 * function.c (assign_parm_setup_block): Relax condition on
1147 multi-register optimization.
1149 2005-01-12 Nick Clifton <nickc@redhat.com>
1151 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
1152 an "in_delay_slot" attribute of "no" to prevent them being used in
1153 delay slots. This is forbidden because they might generate
1156 2005-01-12 Alan Modra <amodra@bigpond.net.au>
1159 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
1162 2005-01-12 Nick Clifton <nickc@redhat.com>
1164 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
1165 passed on to GAS) but do not define, as GAS no longer supports or
1166 needs the -m2000 option.
1168 2005-01-11 Roger Sayle <roger@eyesopen.com>
1170 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
1171 type as args of unprototyped builtins aren't automatically promoted.
1172 (fold_builtin_abs): Likewise.
1174 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1176 * Makefile.in: Set a `build-warn' variable.
1178 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
1180 PR tree-optimization/17949
1181 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
1182 (find_interesting_uses_address): Use it.
1184 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
1186 * regrename.c (kill_value): Handle subreg's that won't simplify.
1188 2005-01-11 Richard Henderson <rth@redhat.com>
1191 * config/i386/i386.h (enum ix86_builtins): Move ...
1192 * config/i386/i386.c: ... here.
1193 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
1194 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
1195 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
1196 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
1197 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
1198 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
1199 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
1200 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
1201 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
1202 IX86_BUILTIN_LOADDDUP): Remove.
1203 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
1204 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
1205 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
1206 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
1207 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
1208 (ix86_init_builtins): Make static.
1209 (ix86_init_mmx_sse_builtins): Update for changed builtins.
1210 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
1211 if all the modes match. Otherwise, fake it.
1212 (get_element_number, ix86_expand_vec_init_builtin,
1213 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
1214 (ix86_expand_builtin): Make static. Update for changed builtins.
1215 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
1216 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
1217 (ix86_expand_vector_init_duplicate): New.
1218 (ix86_expand_vector_init_low_nonzero): New.
1219 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
1220 Split out from ix86_expand_vector_init; handle integer modes.
1221 (ix86_expand_vector_init): Use them.
1222 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
1223 * config/i386/i386-protos.h: Update.
1224 * config/i386/predicates.md (reg_or_0_operand): New.
1225 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
1226 (movv2sf_internal): Likewise. And a splitter to match them all.
1227 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
1228 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
1229 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
1230 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
1231 (mmx_pinsrw): Fix operand ordering.
1232 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
1233 rather than sse_loadss expander.
1234 (movv2df splitter): Similarly.
1235 (sse_loadss, sse_loadlss): Remove.
1236 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
1237 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
1238 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
1239 (sse3_movddup): Rename with '*'.
1240 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
1241 (sse2_loadsd): Remove.
1242 (vec_dupv2df_sse3): Rename from sse3_loadddup.
1243 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
1244 (sse2_pinsrw): Fix argument ordering.
1245 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
1246 (sse2_stored): Remove 'r' destination.
1247 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
1248 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
1249 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
1250 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
1251 vec_extractv16qi, vec_initv16qi): New.
1253 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
1254 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
1255 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
1257 (_mm_load_pd, _mm_store_pd): Use plain dereference.
1258 (_mm_load_si128, _mm_store_si128): Likewise.
1259 (_mm_load1_pd): Use _mm_set1_pd.
1260 (_mm_load_sd): Use _mm_set_sd.
1261 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
1262 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
1263 (_mm_set_epi64): Use _mm_set_epi64x.
1264 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
1265 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
1266 _mm_setr_epi8): Use _mm_set_foo form.
1267 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
1268 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
1269 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
1270 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
1271 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
1272 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
1273 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
1274 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
1275 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
1276 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
1277 (_mm_movedup_pd): Use _mm_shuffle_pd.
1278 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
1279 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
1280 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
1281 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
1283 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
1284 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
1285 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
1286 (_mm_store_ps): Use raw dereference.
1287 (_mm_store1_ps): Use _mm_storeu_ps.
1288 (_mm_storer_ps): Use _mm_store_ps.
1289 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
1290 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
1292 2005-01-11 Stan Shebs <shebs@apple.com>
1294 * config/rs6000/rs6000.c (machopic_output_stub): Issue
1295 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
1297 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
1300 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
1301 Skip all DECLs except for FIELD_DECLs.
1303 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
1305 * config/s390/s390.c (override_options): Return error if
1306 -mbackchain, -mpacked-stack and -mhard-float are used together.
1307 (s390_va_start): Remove the backchain && packed-stack special case.
1308 (s390_gimplify_va_arg): Likewise.
1309 * doc/invoke.texi: Remove the ABI incompatibility note.
1311 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
1313 * config/s390/s390.c (struct s390_frame_layout): Remove
1315 (s390_frame_info, s390_emit_prologue): Replace occurrences of
1316 save_backchain_p with TARGET_BACKCHAIN.
1318 2005-01-11 Alan Modra <amodra@bigpond.net.au>
1321 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
1322 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
1323 * function.c (locate_and_pad_parm): Set new field.
1324 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
1325 Tweak where_pad test to include "none". Always set mem align for
1327 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
1328 not sufficient for type.
1329 (assign_parm_setup_block): If stack_parm is zero on entry, always
1330 make a new stack local. Block move old stack parm if necessary
1331 to new aligned stack local.
1332 (assign_parm_setup_stack): Use a block move to handle
1333 potentially misaligned entry_parm.
1334 (assign_parms_unsplit_complex): Specify required alignment when
1335 creating stack local.
1336 * calls.c (compute_argument_addresses): Override alignment of stack
1337 arg calculated from its type with the alignment given by
1338 FUNCTION_ARG_BOUNDARY.
1339 (store_one_arg): Likewise.
1341 2005-01-11 Jan Beulich <jbeulich@novell.com>
1343 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
1344 Change respective itanium_class attribute to ialu.
1345 (shladdp4_internal): New.
1346 * config/ia64/predicates.md (shladd_log2_operand): New.
1348 2005-01-11 Richard Henderson <rth@redhat.com>
1350 * expr.c (store_constructor): Use rtvec_alloc instead of
1351 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
1353 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
1355 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
1357 2005-01-11 Alan Modra <amodra@bigpond.net.au>
1359 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
1360 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
1361 (TARGET_RWRELOC): Define.
1362 (ia64_rwreloc_section_type_flags): Delete.
1363 (ia64_section_type_flags): New function.
1364 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
1365 (TARGET_RWRELOC): Define.
1367 2005-01-10 David Mosberger <davidm@hpl.hp.com>
1370 * config/ia64/ia64.c (process_set): For alloc insn, only call
1371 process_epilogue is !frame_pointer_needed.
1373 2005-01-10 Roger Sayle <roger@eyesopen.com>
1376 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
1377 unary operator and can't be treated as a binary/comparison operator.
1379 2005-01-10 Richard Henderson <rth@redhat.com>
1381 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
1382 (ix86_return_in_memory): Likewise.
1383 (function_arg_advance): Likewise. Mirror structure in function_arg
1384 for choosing register to advance.
1386 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
1388 * tree-vectorizer.c, tree.def: Fix comment typos.
1390 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
1392 * config/s390/s390.c (struct processor_costs): 4 new fields:
1393 ddbr, ddr, debr, der.
1394 (s390_rtx_costs): More precise handling of divide instructions.
1396 2005-01-10 David Edelsohn <edelsohn@gnu.org>
1398 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
1400 2005-01-10 Jan Beulich <jbeulich@novell.com>
1402 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
1403 names resulting from -ffunction-sections/-fdata-sections and linkonce
1405 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
1406 based on the section name. Rearrange the section name comparison logic
1407 slightly so that each section name is compared against at most once.
1409 2005-01-10 Ben Elliston <bje@au.ibm.com>
1411 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
1412 targets for which -fPIC reduces limitations on the GOT size.
1414 2005-01-09 Falk Hueffner <falk@debian.org>
1416 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
1419 2005-01-09 David Edelsohn <edelsohn@gnu.org>
1422 * collect2.c (main): Set aixrtl_flag for -brtl option.
1423 (resolve_lib_name): Search for .so file extension before .a
1426 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
1428 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
1429 (vect_analyze_data_refs_alignment): Add dump prints.
1431 2005-01-09 Ira Rosen <irar@il.ibm.com>
1433 * tree-vectorizer.c (vect_analyze_offset_expr): Use
1434 expr_invariant_in_loop_p.
1435 Initialize outputs first thing in the function.
1436 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
1437 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
1438 (vect_analyze_pointer_ref_access): Check that the initial condition of
1439 the access function is loop invariant.
1441 2005-01-09 Richard Henderson <rth@redhat.com>
1443 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
1444 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
1445 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
1446 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
1447 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
1448 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
1449 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
1450 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
1451 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
1452 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
1453 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
1454 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
1455 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
1456 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
1457 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
1458 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
1459 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
1460 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
1461 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
1462 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
1463 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
1464 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
1465 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
1467 (mmx_clrdi, pavgusb): Remove.
1468 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
1469 with leading sse_ prefix.
1470 * config/i386/sse.md: Receive them.
1471 * config/i386/mmx.md: New file.
1472 (MMXMODE12, MMXMODE24, mmxvecsize): New.
1473 (subrv2sf3): Turn into expander for normal subtraction.
1474 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
1475 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
1476 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
1477 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
1478 commutative; use ix86_binary_operator_ok.
1479 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
1480 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
1481 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
1482 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
1483 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
1484 Macroize from existing patterns; use ix86_binary_operator_ok.
1485 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
1486 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
1487 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
1488 with vec_select+vec_concat.
1489 (mmx_pshufw, mmx_pshufw_1): Likewise.
1490 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
1491 (mmx_uavgv4hi3): Model correctly.
1492 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
1493 _mm_xor_si64): Remove casts.
1495 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
1497 PR tree-optimization/19224
1498 * tree-scalar-evolution.c (get_instantiated_value,
1499 set_instantiated_value): New functions.
1500 (instantiate_parameters_1): Cache the results.
1501 (instantiate_parameters, resolve_mixers): Initialize and free
1504 2005-01-08 David Edelsohn <edelsohn@gnu.org>
1506 * config/i386/i386.md (addhi_4): Correct reference in comment.
1509 2005-01-08 Richard Henderson <rth@redhat.com>
1511 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
1512 (_mm_cvtsi128_si64x): Likewise.
1513 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
1514 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
1515 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
1516 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
1517 * config/i386/i386-protos.h: Update.
1518 * config/i386/i386.c (print_operand): Add 'H'.
1519 (ix86_fixup_binary_operands): Split out from ...
1520 (ix86_expand_binary_operator): ... here.
1521 (ix86_fixup_binary_operands_no_copy): New.
1522 (ix86_expand_fp_absneg_operator): Handle vector mode results.
1523 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
1524 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
1525 (safe_vector_operand): Use CONST0_RTX.
1526 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
1527 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
1528 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
1529 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
1531 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
1532 (IX86_BUILTIN_CMPNGESS): New.
1533 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
1534 (attr type): Add sselog1.
1535 (attr unit, attr memory): Handle it.
1536 (movti, movti_internal, movti_rex64): Move near other integer moves.
1537 (movtf, movtf_internal): Move near other fp moves.
1538 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
1539 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
1540 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
1541 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
1542 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
1543 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
1544 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
1545 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
1546 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
1547 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
1548 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
1549 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
1550 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
1551 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
1552 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
1553 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
1554 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
1555 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
1556 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
1557 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
1558 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
1559 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
1560 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
1561 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
1562 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
1563 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
1564 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
1565 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
1566 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
1567 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
1568 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
1569 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
1570 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
1571 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
1572 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
1573 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
1574 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
1575 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
1576 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
1577 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
1578 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
1579 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
1580 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
1581 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
1583 (SSEPUSH, push<SSEPUSH>): Remove.
1584 (MMXPUSH, push<MMXPUSH>): Remove.
1585 (sse_movaps, sse_movaps_1, sse_movups): Remove.
1586 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
1587 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
1588 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
1589 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
1590 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
1591 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
1592 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
1593 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
1594 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
1595 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
1596 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
1597 (ppro_sse_log_V4SF): Handle sselog1.
1598 * config/i386/predicates.md (const_0_to_1_operand): New.
1599 (const_0_to_255_mul_8_operand): New.
1600 (const_1_to_31_operand): Rename from const_int_1_31_operand.
1601 (const_2_to_3_operand, const_4_to_7_operand): New.
1602 * config/i386/sse.md: New file.
1603 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
1604 (sse_movups): Rename from sse_movups_1.
1605 (sse_loadlss): Rename from sse_loadss_1.
1606 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
1608 (negv4sf2): Use ix86_expand_fp_absneg_operator.
1609 (absv4sf2, negv2df, absv2df): New.
1610 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
1611 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
1612 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
1613 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
1614 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
1615 (sse3_addsubv4sf3): Model correctly.
1616 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
1617 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
1618 (sse_movhlps): Model with vec_select+vec_concat.
1619 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
1620 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
1621 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
1622 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
1623 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
1624 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
1625 sse2_pshufhw_1): Likewise.
1626 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
1627 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
1628 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
1629 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
1630 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
1631 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
1632 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
1633 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
1634 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
1635 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
1636 sminv8hi3): Mark commutative
1637 operands. Use ix86_binary_operator_ok.
1638 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
1639 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
1640 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
1641 sse2_punpcklqdq): Allow operand2 in memory.
1642 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
1643 sse2_movsd): Add memory alternatives.
1644 (sse_storelps): Turn expander into an insn; split after reload.
1645 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
1646 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
1648 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
1650 * configure.ac (DWARF-2 debug_line): Use objdump.
1651 * configure: Regenerate.
1653 2005-01-08 Jeff Law <law@redhat.com>
1654 Diego Novillo <dnovillo@redhat.com>
1656 PR tree-optimization/18241
1657 * tree-nrv.c (tree_nrv): Ignore volatile return values.
1658 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
1659 statements with volatile operands.
1660 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
1661 operands after marking a statement with has_volatile_ops.
1663 2005-01-08 Roger Sayle <roger@eyesopen.com>
1665 * tree.c (int_fits_type_p): Always honor integer constant
1666 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
1668 2005-01-08 Roger Sayle <roger@eyesopen.com>
1670 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
1671 the latent bug in rtl_delete_block has been resolved.
1673 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
1675 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
1676 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
1678 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
1680 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
1681 library as @multilib_dir@/$(SHLIB_SONAME).
1683 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
1685 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
1686 * configure: Regenerate.
1688 2005-01-07 Jakub Jelinek <jakub@redhat.com>
1690 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
1693 2005-01-07 David Edelsohn <edelsohn@gnu.org>
1696 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
1697 Convert non-word aligned offset address using ld/std into
1700 2005-01-07 Richard Henderson <rth@redhat.com>
1702 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
1704 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
1706 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
1708 2005-01-07 Jakub Jelinek <jakub@redhat.com>
1710 PR tree-optimization/19283
1711 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
1712 is not shorter than the original type.
1714 PR rtl-optimization/19012
1715 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
1717 PR rtl-optimization/18861
1718 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
1719 (FULL_STATE): Define.
1720 (mark_tablejump_edge): New function.
1721 (purge_dead_tablejump_edges): New function.
1722 (find_bb_boundaries): Use it.
1724 PR tree-optimization/18828
1725 * builtins.c (expand_builtin_next_arg): Remove argument and all
1726 the argument checking.
1727 (expand_builtin): Adjust caller.
1728 (expand_builtin_va_start): Likewise. Remove error for too many
1730 (fold_builtin_next_arg): Issue error for too many arguments.
1731 After checking arguments, replace them with magic arguments that
1732 prevent further checking of the args.
1734 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1736 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
1738 2005-01-06 Roger Sayle <roger@eyesopen.com>
1741 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
1742 and %{.S:...} (and their negative variants) to test whether the
1743 input file is assembler or pre-processed-assembler independent of
1744 the actual filename extension.
1746 2005-01-06 Roger Sayle <roger@eyesopen.com>
1748 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
1749 of sign or zero extended values.
1751 2005-01-06 Geoffrey Keating <geoffk@apple.com>
1753 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
1754 for IBM long double format correctly.
1756 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
1758 Fix PR tree-optimization/18792
1760 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
1761 to first_loop_depth, and use loop depth instead of loop number.
1762 (build_classic_dir_vector): Ditto.
1763 (compute_data_dependences_for_loop): Use depth, not loop number.
1764 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
1765 number. Pass in loops, instead of loop numbers.
1766 (gather_interchange_stats): Ditto.
1767 (linear_transform_loops): Ditto.
1769 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
1772 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
1773 biv_fits_mode_p, extension_within_bounds_p): New functions.
1774 (check_ext_dependent_givs): Use them.
1776 2005-01-06 Roger Sayle <roger@eyesopen.com>
1778 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
1779 more than one barrier, in which case we should delete them all.
1781 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
1783 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
1785 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
1787 Fix PR middle-end/19286
1789 * dwarf2out.c (gen_subprogram_die): If we've already tried to
1790 output this subprogram, simply ignore this attempt to do it again.
1791 (add_abstract_origin_attribute): Don't abort trying to add the abstract
1792 origin attribute if it's not possible.
1793 (gen_block_die): Don't ignore subblocks of "unused" blocks.
1794 (decls_for_scope): Ditto.
1795 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
1796 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
1798 2005-01-05 Richard Henderson <rth@redhat.com>
1801 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
1802 (bdesc_2arg): Use it.
1803 (ix86_expand_binop_builtin): Force operands into registers
1805 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
1806 ix86_expand_sse_compare, ix86_expand_sse_comi,
1807 ix86_expand_builtin): Likewise.
1809 2005-01-05 Richard Henderson <rth@redhat.com>
1811 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
1813 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
1814 (mulv4hi3): Set itanium_class mmmul.
1815 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
1816 (fpack): Rename from fpack_sfsf.
1818 2005-01-05 Richard Henderson <rth@redhat.com>
1821 * reload1.c (do_input_reload): Restrict the optimization deleteing
1822 a previous output reload to RELOAD_FOR_INPUT.
1824 2005-01-05 Steven Bosscher <stevenb@suse.de>
1826 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
1827 comment that we fall through after case.
1828 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
1829 (distribute_links): Likewise.
1830 * cse.c (cse_insn): Likewise.
1831 * cselib.c (cselib_invalidate_mem): Likewise.
1832 * df.c: Update comments at the top of the file.
1833 (read_modify_subreg_p): Update comments here too.
1834 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
1835 * flow.c (mark_set_1): Likewise.
1836 (mark_used_regs): Likewise.
1837 * gcse.c (mems_conflict_for_gcse_p): Likewise.
1838 (canon_list_insert): Likewise.
1839 (mark_set): Likewise.
1840 (try_replace_reg): Likewise.
1841 (store_killed_in_insn): Likewise.
1842 * loop.c (count_one_set): Likewise.
1843 (basic_induction_var): Likewise.
1844 * postreload-gcse.c (find_mem_conflicts): Likewise.
1845 * postreload.c (reload_combine_note_store): Likewise.
1846 (move2add_note_store): Likewise.
1847 * reload.c (find_equiv_reg): Likewise.
1848 (mark_referenced_resources): Likewise.
1849 * rtlanal.c (set_noop_p): Likewise.
1850 (note_stores): Likewise.
1851 (note_uses): Likewise.
1852 * sched-deps.c (sched_analyze_1): Likewise.
1853 * sched-rgn.c (check_live_1): Likewise.
1854 (update_live_1): Likewise.
1855 * config/i860/i860.c: Likewise.
1857 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
1859 (ZERO_EXTRACT): Mention that this one can be an lvalue.
1861 * doc/rtl.texi: Update documentation for bit-fields and SET.
1863 2005-01-05 Roger Sayle <roger@eyesopen.com>
1865 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
1867 2005-01-05 Jan Hubicka <jh@suse.cz>
1868 Richard Henderson <rth@redhat.com>
1871 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
1874 2005-01-05 Richard Henderson <rth@redhat.com>
1877 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
1878 (sse_shufps): Change operand 3 to const_int_operand.
1879 (sse2_storelps): Fix typo in template.
1880 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
1881 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
1882 (ix86_expand_builtin): Likewise.
1884 2005-01-05 Julian Brown <julian@codesourcery.com>
1886 * config/arm/arm.c (arm_return_in_memory): Treat complex types
1887 as aggregates for AAPCS ABIs.
1889 2005-01-05 Stan Shebs <shebs@apple.com>
1891 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
1892 Mach-O getters if ppc64.
1893 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
1895 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
1896 signed four-byte field for global code case.
1897 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
1899 2005-01-05 Roger Sayle <roger@eyesopen.com>
1902 * c-common.c: Include real.h.
1903 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
1904 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
1905 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
1906 * Makefile.in (c-common.o): Update dependencies.
1908 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1910 * c-parse.in (asm_string): Add trailing semicolon.
1912 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1914 * c-parse.in (asm_string): New. Don't allow wide strings in
1916 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
1917 asm_string instead of STRING.
1919 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1921 * c-typeck.c (constructor_no_implicit): Remove.
1922 (set_designator, process_init_element): Don't check
1923 constructor_no_implicit.
1925 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
1926 Kaz Kojima <kkojima@gcc.gnu.org>
1929 * lcm.c (create_pre_exit): New.
1930 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
1931 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
1933 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
1935 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
1936 -msoft-float in target_float_switch.
1937 * arm.c (arm_override_options): Fix processing of target_float_switch.
1939 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
1941 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
1944 2004-01-05 Uros Bizjak <uros@kss-loka.si>
1946 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
1947 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
1949 2005-01-04 Roger Sayle <roger@eyesopen.com>
1951 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
1952 if deleting the then-block allows the test-block to fallthru to the
1955 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
1958 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
1959 after the full declaration if the are in two different TUs.
1961 2005-01-04 Richard Henderson <rth@redhat.com>
1964 * tree-sra.c (generate_one_element_init): Just
1965 call gimplify_and_add.
1966 (generate_element_init): Record the
1967 new referenced variables and mark them for renaming
1968 and split out to ...
1969 (generate_element_init_1): This.
1970 (scalarize_init): Don't call push_gimplify_context/
1971 pop_gimplify_context.
1973 2005-01-04 Geoffrey Keating <geoffk@apple.com>
1975 * toplev.c (get_src_pwd): Handle failure of getpwd().
1977 2005-01-04 Roger Sayle <roger@eyesopen.com>
1979 * fold-const.c (fold_single_bit_test): Delete unreachable handling
1981 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
1982 all the cases handled by it are inverted by invert_truthvalue.
1984 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
1986 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
1987 (save_register_windowsi): Likewise.
1989 2005-01-04 Richard Henderson <rth@redhat.com>
1991 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
1993 2005-01-04 Uros Bizjak <uros@kss-loka.si>
1996 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
1997 any simplifications of vector mode comparison operators.
1998 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
2000 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
2001 Devang Patel <dpatel@apple.com>
2003 PR tree-optimization/18308
2004 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
2005 the operands before creating a new expression.
2006 * dojump.c (do_jump): Make drop_through_label available
2007 for all cases. Add expansion of COND_EXPR.
2009 2005-01-04 Ira Rosen <irar@il.ibm.com>
2011 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
2012 INTEGER_CST instead of TREE_CONSTANT.
2013 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
2014 instead of TREE_CONSTANT.
2015 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
2016 instead of TREE_CONSTANT.
2018 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
2020 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
2022 2005-01-04 Uros Bizjak <uros@kss-loka.si>
2025 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
2026 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
2028 2005-01-03 Richard Henderson <rth@redhat.com>
2030 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
2031 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
2032 fold_convert_const_real_from_real): Split out from ...
2033 (fold_convert_const): ... here.
2035 2005-01-03 Richard Henderson <rth@redhat.com>
2038 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
2039 (mov<MMXMODEI>_internal): Likewise.
2040 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
2041 Add V2SF case; use it for SSE1; don't use TI.
2042 (movdf_integer): Likewise.
2043 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
2045 2005-01-03 Ira Rosen <irar@il.ibm.com>
2047 * tree-vectorizer.c (vect_strip_conversions): New function.
2048 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
2049 check for binary class.
2051 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
2055 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
2057 * gimple-low.c (mark_blocks_with_used_vars): New function.
2058 (mark_blocks_with_used_subblocks): Ditto.
2059 (mark_used_blocks): Ditto.
2060 (pass_mark_used_blocks): New pass.
2061 * tree-inline.c: Include debug.h.
2062 (expand_call_inline): Call outlining_inline_function here.
2063 * tree-optimize.c (init_tree_optimization_passes): Add
2064 pass_mark_used_blocks.
2065 * tree-pass.h (pass_mark_used_blocks): New.
2066 * Makefile.in (tree-inline.o): Add debug.h dependency.
2068 2005-01-03 Geoffrey Keating <geoffk@apple.com>
2070 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
2071 VAR_DECLs to have weak_import attribute.
2073 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
2076 2004-01-03 Steven Bosscher <stevenb@suse.de>
2078 * passes.c (rest_of_compilation): Don't run regmove if only
2079 flag_expensive_optimizations, require flag_regmove instead.
2081 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
2083 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
2085 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
2086 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
2087 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
2088 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
2090 2005-01-03 Richard Henderson <rth@redhat.com>
2092 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
2093 (ia64_const_ok_for_letter_p): New.
2094 (ia64_const_double_ok_for_letter_p): New.
2095 (ia64_extra_constraint): New.
2096 (ia64_expand_vecint_compare): New.
2097 (ia64_expand_vcondu_v2si): New.
2098 (ia64_expand_vecint_cmov): New.
2099 (ia64_expand_vecint_minmax): New.
2100 (ia64_print_operand): Add 'v'.
2101 (ia64_preferred_reload_class): New.
2102 (ia64_vector_mode_supported_p): New.
2103 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
2104 (PREFERRED_RELOAD_CLASS): Move to function.
2105 (CONST_OK_FOR_LETTER_P): Move to function.
2106 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
2107 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
2108 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
2109 (EXTRA_CONSTRAINT): Move to function.
2110 * config/ia64/ia64.md: Include vect.md.
2111 (itanium_class): Add mmalua.
2113 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
2115 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
2116 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
2118 (const_int_2bit_operand): New.
2119 (fr_reg_or_0_operand): New.
2120 * config/ia64/ia64-modes.def: Add vector modes.
2121 * config/ia64/ia64-protos.h: Update.
2122 * config/ia64/vect.md: New file.
2124 2005-01-03 Richard Henderson <rth@redhat.com>
2126 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
2128 2005-01-03 Uros Bizjak <uros@kss-loka.si>
2131 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
2133 (log1pdf2): Change mode of operands[0,1] to DFmode.
2135 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
2137 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
2138 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
2139 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
2140 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
2141 SPARC_RELAXED_ORDERING.
2143 2005-01-03 Richard Henderson <rth@redhat.com>
2144 Uros Bizjak <uros@kss-loka.si>
2147 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
2148 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
2149 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
2151 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
2152 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
2153 const_pow2_1_to_128_operand): New predicates.
2155 2005-01-02 Greg McGary <greg@mcgary.org>
2157 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
2159 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
2160 object. [ptr_type]: Remove unused variable.
2161 Remember nearest addressable array-element or record-component
2162 when checking bitfield components. Tolerate empty BIND_EXPR.
2164 2005-01-01 Richard Henderson <rth@redhat.com>
2166 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
2167 that don't narrow the value. Fail for other conversions.
2169 2005-01-01 Richard Henderson <rth@redhat.com>
2172 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
2173 * c-lang.c (finish_file): Don't do it here.
2174 * objc/objc-act.c (objc_finish_file): Likewise.
2176 * cgraph.c (decl_assembler_name_equal): New.
2177 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
2178 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
2179 * cgraph.h (struct cgraph_varpool_node): Add next.
2180 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
2181 * varasm.c (assemble_alias): Mark the target as needed.
2183 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
2186 * function.c (get_arg_pointer_save_area): Use entry_of_function
2187 instead of get_insns.
2189 2005-01-01 Roger Sayle <roger@eyesopen.com>
2190 Andrew Pinski <pinskia@physics.uc.edu>
2191 James E. Wilson <wilson@specifixinc.com>
2193 PR rtl-optimization/12092
2194 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
2197 2005-01-01 Roger Sayle <roger@eyesopen.com>
2198 Olivier Hainque <hainque@act-europe.fr>
2200 * tree.c (int_fits_type_p): A narrower type always fits in a
2201 wider one, except for negative values into unsigned types.
2203 2005-01-01 Roger Sayle <roger@eyesopen.com>
2205 * tree.c (int_fits_type_p): Compare the result of force_fit_type
2206 with the original constant rather than require TREE_OVERFLOW.
2208 2005-01-01 Steven Bosscher <stevenb@suse.de>
2211 * c-decl.c (finish_function): If compiling C99, annotate the
2212 compiler generated return with the current file name and line 0.
2213 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
2214 the source line is greater than 0.
2215 (remove_bb): Likewise.
2217 See ChangeLog.12 for earlier changes.