1 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
3 * flow.c (regset_bytes, regset_size): Remove.
5 2005-01-23 Steven Bosscher <stevenb@suse.de>
7 PR rtl-optimization/19464
8 * tree-optimize.c (init_tree_optimization_passes): Add one more
9 copyrename pass just before out-of-ssa.
11 2005-01-23 Joseph S. Myers <joseph@codesourcery.com>
14 * recog.c (recog_memoized): Don't define if GENERATOR_FILE.
15 * ggc-none.c (ggc_free): Define.
17 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
19 * cse.c (max_reg, max_insn_uid): Remove.
20 (cse_main): Don't access max_reg or max_insn_uid.
22 * c-common.c (explicit_flag_signed_bitfields,
23 lang_statement_code_p, lang_expand_function_end): Remove.
24 * c-common.h: Remove the corresponding declarations.
25 * c-opts.c (c_common_handle_option): Remove all write access
26 to explicit_flag_signed_bitfields.
28 * cgraph.c (cgraph_varpool_n_nodes): Remove.
29 (cgraph_varpool_node): Don't access cgraph_varpool_n_nodes.
30 * cgraph.h: Remove the corresponding declaration.
32 * gcse.c (null_pointer_info): Remove.
34 2005-01-23 Roger Sayle <roger@eyesopen.com>
35 Eric Botcazou <ebotcazou@libertysurf.fr>
37 * combine.c (simplify_logical): Only simplify logical expressions
38 of the form ior(and(x,y),z) by the inverse distributive law if the
39 result is cheaper than the original.
41 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
43 * alias.c, c-common.h, c-incpath.c, c-incpath.h, expr.c,
44 fold-const.c, gimplify.c, params.h, tree-data-ref.c,
45 tree-if-conv.c, tree-nested.c, tree-outof-ssa.c,
46 tree-ssa-dom.c, tree-vectorizer.c, tree.def, config/darwin.c,
47 config/freebsd-spec.h, config/arm/arm.h,
48 config/h8300/h8300.md, config/i386/i386.md,
49 config/i386/predicates.md, config/i386/sse.md,
50 config/ia64/ia64.c, config/ip2k/ip2k.c, config/s390/s390.c,
51 config/vax/vax.md: Fix comment typos. Follow spelling
54 2005-01-23 Steven Bosscher <stevenb@suse.de>
56 * gengtype.c (walk_type): Produce `not equal to' compares for loop
57 tests, instead of `less than'.
59 2005-01-23 Richard Sandiford <rsandifo@redhat.com>
61 * function.c (assign_parm_setup_block): When creating a new stack slot
62 for a parameter, get its alignment from the parameter's DECL_ALIGN
63 rather than the type's TYPE_ALIGN. Make sure that the parameter
64 is at least word aligned.
66 2005-01-22 Roger Sayle <roger@eyesopen.com>
69 * config/avr/avr.c (avr_hard_regno_mode_ok): Rewrite.
71 2005-01-22 Richard Henderson <rth@redhat.com>
74 * config/i386/i386.md (movsfcc_1_sse_max): Use nonimmediate_operand
75 in both compare operands.
76 (movdfcc_1_sse_max): Likewise.
77 (movsfcc_1_sse): Likewise. Add earlyclobber for scratch.
78 (movdfcc_1_sse): Likewise.
79 * config/i386/i386.c (ix86_split_sse_movcc): Emit copies into the
80 scratch register as needed.
82 2005-01-22 Richard Henderson <rth@redhat.com>
84 * config/i386/i386.md (smaxsf3): Fix mnemonic typo.
86 * config/i386/i386.c (ix86_prepare_fp_compare_args): Fix is_sse test.
88 2005-01-22 Richard Henderson <rth@redhat.com>
90 * genopinit.c (optabs): Use smin/smax for floating point too.
91 * doc/md.texi: Update to match. Clarify that floating point
92 results are undefined for +0/-0 and NaN.
93 * doc/rtl.texi: Likewise.
94 * rtl.def (SMIN, SMAX): Likewise
95 * tree.def (MIN_EXPR, MAX_EXPR): Likewise.
97 * config/alpha/alpha.md (smaxdf3, smindf3, smaxsf3, sminsf3): Add
98 leading 's' to the name.
99 * config/ia64/ia64.md (smaxsf3, sminsf3, smaxdf3, smindf3,
100 smaxxf3, sminxf3): Likewise.
101 * config/rs6000/rs6000.md (smaxdf3, smindf3, smaxsf3, sminsf3):
104 2005-01-22 Paul Brook <paul@codesourcery.com>
106 * config/arm/arm.md: Use "Uy" constraint for wcgr load/stores.
108 2005-01-21 Mark Dettinger <dettinge@de.ibm.com>
110 * config/s390/s390.c (struct processor_costs): New fields
111 dlgr, dlr, dr, dsgfr, dsgr.
112 (z900_cost, z990_cost): Values for new fields.
113 (s390_rtx_costs): New cases MEM und COMPARE in switch
114 statement. Modified handling of SIGN_EXTEND, ZERO_EXTEND,
115 DIV, MOD, UDIV, UMOD.
117 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
119 * config/s390/s390.md ("doloop_si64"): Reload input value directly
120 into the register being decremented.
121 ("doloop_si31", "doloop_di"): Likewise.
122 ("*doloop_si_long"): Adapt pattern.
124 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
126 * config/s390/s390.h (HARD_REGNO_NREGS): Fix computation for
128 (CLASS_MAX_NREGS): Likewise.
130 2005-01-21 Daniel Berlin <dberlin@dberlin.org>
132 * doc/tree-ssa.texi (Statement Operands): Add example for new
133 must-def macro. Note deprecation of old operands interface.
135 2005-01-22 Richard Sandiford <rsandifo@redhat.com>
137 PR tree-optimization/19484
138 * tree-cfg.c (remove_fallthru_edge): New function.
139 (cleanup_control_flow): Remove fallthru edges from calls that are
140 now known not to return.
142 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
144 Fix PR tree-optimization/19038
145 * tree-ssa-dom.c (cprop_operand): Don't replace loop invaeriant
146 copies with loop variant ones.
148 2005-01-22 Kazu Hirata <kazu@cs.umass.edu>
150 * cfganal.c, real.h, reorg.c, timevar.def, tree-ssa-ccp.c,
151 config/alpha/alpha-protos.h, config/alpha/alpha.h,
152 config/alpha/alpha.md, config/alpha/predicates.md,
153 config/sparc/freebsd.h, config/sparc/netbsd-elf.h,
154 config/sparc/sol2.h: Update copyright.
156 * tree-cfg.c (remove_forwarder_block_with_phi): Look at the
157 first label to see if it is a nonlocal label.
159 2005-01-22 David Edelsohn <edelsohn@gnu.org>
160 Andrew Pinski <pinskia@physics.uc.edu>
163 * config/rs6000/rs6000.c (rs6000_va_start): Saturate n_gpr at
164 maximum number of GPRs. Saturate n_fpr at maximum number of FPRs.
166 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
168 * config/rs6000/rtems.h: Update copyright.
170 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
173 * config/rs6000/rtems.h: Resurrect cpp_os_rtems_spec from gcc < 3.4.
174 (CPP_OS_RTEMS_SPEC): New (From gcc-3.3's config/rs6000/sys4.h).
175 (SUBSUBTARGET_EXTRA_SPECS): Use CPP_OS_RTEMS_SPEC.
177 2005-01-22 Volker Reichelt <reichelt@igpm.rwth-aachen.de>
180 * c-typeck.c (convert_arguments): Check for error_mark_node.
182 2005-01-21 Ian Lance Taylor <ian@airs.com>
184 PR tree-optimization/13000
185 * tree-inline.c: Include "tree-flow.h".
186 (expand_call_inline): If warn_return_type, warn if non-void inline
187 function falls through.
188 * tree-cfg.c (execute_warn_function_return): Don't warn about
189 control reaching end if TREE_NO_WARNING is set. Set
191 * gimple-low.c (block_may_fallthru): Don't assume that SWITCH_EXPR
193 * gimplify.c (shortcut_cond_expr): Don't emit a jump over the else
194 branch if we don't need one.
195 * c-typeck.c: Include "tree-flow.h"
196 (c_finish_bc_stmt): Don't add a goto if the current statement
197 list doesn't fall through to the current point.
199 2005-01-21 Roger Sayle <roger@eyesopen.com>
201 PR rtl-optimization/576
202 * real.c (real_arithmetic): Change return type from void to bool
203 to return an indication that the result may be inexact.
204 * real.h (real_arithmeric): Update prototype.
205 * fold-const.c (const_binop): Don't constant fold floating
206 point expressions when the user specifies -frounding-math and
207 the result may depend upon the run-time rounding mode.
208 (fold_convert_const_real_from_real): Clean-up.
209 (fold_initializer): Ignore flag_rounding_math for initializers.
210 * simplify-rtx.c (simplify_binary_operation): Likewise, don't
211 constant fold FP operations with flag_rounding_math if the
212 result may depend upon the run-time rounding mode.
214 2005-01-21 Tom Tromey <tromey@redhat.com>
216 * c-cppbuiltin.c (define__GNUC__): Correct assertion.
218 2005-01-21 Kazu Hirata <kazu@cs.umass.edu>
220 * tree-cfg.c: Fix comment typos.
222 2005-01-21 J"orn Rennecke <joern.rennecke@st.com>
224 * passes.c (rest_of_handle_flow2): Close / open the flow2 dump file
225 around the call to rest_of_handle_branch_target_load_optimize.
227 2005-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
229 * config.gcc (sparc-*-netbsdelf*): Include dbxelf.h
230 (sparc64-*-openbsd*): Likewise.
231 (sparclite-*-elf*): Include tm-dwarf2.h.
232 (sparc86x-*-elf*): Likewise.
233 (sparc64-*-elf*): Likewise.
234 * config/sparc/linux64.h (DWARF2_DEBUGGING_INFO): Delete.
235 (DBX_DEBUGGING_INFO): Likewise.
236 * config/sparc/liteelf.h (DWARF2_DEBUGGING_INFO): Likewise.
237 (PREFERRED_DEBUGGING_TYPE): Likewise.
238 * config/sparc/netbsd-elf.h (PREFERRED_DEBUGGING_TYPE): Likewise.
239 * config/sparc/sp64-elf.h (DBX_DEBUGGING_INFO): Likewise.
240 (PREFERRED_DEBUGGING_TYPE): Likewise.
241 * config/sparc/sp86x-elf.h (DWARF2_DEBUGGING_INFO): Likewise.
242 (PREFERRED_DEBUGGING_TYPE): Likewise.
243 * config/sparc/sparc.h (DBX_DEBUGGING_INFO): Likewise.
245 2005-01-21 Hans-Peter Nilsson <hp@bitrange.com>
248 * combine.c (combine_simplify_rtx): Revert change of 2004-12-31.
250 2005-01-20 Paul Brook <paul@codesourcery.com>
252 * doc/rtl.texi: Document value extension requirements for CONST_INT.
254 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
256 * tree-cfg.c (tree_verify_flow_info): Check that a nonlocal
257 label is first in a sequence of labels.
259 * tree-cfg.c (tree_verify_flow_info): Fix a typo.
261 2005-01-20 Janis Johnson <janis187@us.ibm.com>
262 Giovanni Bajo <giovannibajo@gcc.gnu.org>
264 * doc/sourcebuild.texi (Test Directives): New.
266 2005-01-20 Roger Sayle <roger@eyesopen.com>
268 * real.c (real_floor): Don't leave the result uninitialized when
269 mode is VOIDMode, but return the unrounded intermediate value.
270 (real_ceil): Likewise.
272 2005-01-20 Paul Brook <paul@codesourcery.com>
274 * config/arm/arm.md (insv): Use gen_int_mode.
276 2005-01-20 Eric Botcazou <ebotcazou@libertysurf.fr>
278 * config/sparc/freebsd.h (ASM_OUTPUT_INTERNAL_LABELREF): Delete.
279 * config/sparc/linux.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
280 * config/sparc/linux64.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
281 * config/sparc/netbsd-elf.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
282 * config/sparc/sol2.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
284 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
287 * dwarf2out.c (concat_loc_descriptor): We don't know we can use
288 fbreg, so we have to assume we can't.
289 (loc_descriptor_from_tree_1): Ditto.
290 (containing_function_has_frame_base): New function.
291 (add_location_or_const_value_attribute): Use it.
292 Also try to generate a frame_base from a single element location
295 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
297 PR tree-optimization/15349
298 * timevar.def (TV_TREE_MERGE_PHI): New.
299 * tree-cfg.c (tree_forwarder_block_p): Add a new argument
301 (remove_forwarder_block, cleanup_forwarder_blocks): Adjust the
302 calls to tree_forwarder_block_p.
303 (remove_forwarder_block_with_phi, merge_phi_nodes,
304 gate_merge_phi, pass_merge_phi): New.
305 * tree-optimize.c (init_tree_optimization_passes): Add
307 * tree-pass.h: Add an extern for pass_merge_phi;
309 2005-01-20 Richard Henderson <rth@redhat.com>
312 * config/i386/emmintrin.h (_mm_castpd_ps, _mm_castpd_si128): New.
313 (_mm_castps_pd, _mm_castps_si128): New.
314 (_mm_castsi128_ps, _mm_castsi128_pd): New.
316 2005-01-20 Richard Henderson <rth@redhat.com>
319 * config/i386/mmintrin.h (_mm_cvtsi32_si64): Use
320 __builtin_ia32_vec_init_v2si.
321 (_mm_cvtsi64_si32): Use __builtin_ia32_vec_ext_v2si.
322 * config/i386/i386.c (IX86_BUILTIN_VEC_EXT_V2SI): New.
323 (ix86_init_mmx_sse_builtins): Create it.
324 (ix86_expand_builtin): Expand it.
325 (ix86_expand_vector_set): Handle V2SFmode and V2SImode.
326 * config/i386/mmx.md (vec_extractv2sf_0, vec_extractv2sf_1): New.
327 (vec_extractv2si_0, vec_extractv2si_1): New.
329 2005-01-20 Richard Henderson <rth@redhat.com>
332 * config/i386/i386.c (ix86_expand_push): New.
333 * config/i386/mmx.md (push<MMXMODE>1): New.
334 * config/i386/sse.md (push<SSEMODE>1): New.
335 * config/i386/i386-protos.h: Update.
337 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
339 * params.def, params.h, predict.c, tree-ssa-loop-im.c: Update
342 2005-01-20 Richard Henderson <rth@redhat.com>
344 * rtl.def (CONST_VECTOR): Use RTX_CONST_OBJ.
345 * rtl.h (CONSTANT_P): Don't special case CONST_VECTOR.
347 2005-01-19 Richard Henderson <rth@redhat.com>
350 * config/i386/i386.c (ix86_expand_vector_move_misalign): Convert
351 to V4SFmode in SSE1 fallback load path.
353 2005-01-19 Richard Henderson <rth@redhat.com>
355 * config/i386/i386.c (ix86_expand_vector_init_one_var): Fix typo
356 in QImode expansion to ix86_expand_vector_set.
357 (ix86_expand_vector_init_general): Fix typo in V8HImode recursive call.
359 2005-01-19 Richard Henderson <rth@redhat.com>
362 * config/i386/i386.c (ix86_preferred_reload_class): Return a proper
363 subclass of the input class.
364 (ix86_secondary_memory_needed): Always true for cross-MMX classes.
365 Always true for cross-SSE1 classes. Rationalize conditionals.
366 * config/i386/i386.h (SSE_CLASS_P, MMX_CLASS_P): Use straight equality.
367 * config/i386/i386.md (movsi_1): Add MMX/SSE zeros. Fix alternatives
368 for SSE1. Don't check TARGET_INTER_UNIT_MOVES.
369 (movdi_2): Add MMX/SSE zeros.
370 (movdi_1_rex64): Likewise. Don't check TARGET_INTER_UNIT_MOVES.
371 (movsf_1): Don't check TARGET_INTER_UNIT_MOVES.
372 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
373 (movsi_1_nointernunit, movdi_1_rex64_nointerunit): Remove.
374 (movsf_1_nointerunit, zero_extendsidi2_32_1): Remove.
375 (zero_extendsidi2_rex64_1): Remove.
376 (MOV0 peephole): Check GENERAL_REG_P.
378 2005-01-19 Richard Henderson <rth@redhat.com>
381 * config/i386/i386.c (ix86_expand_vector_set): Fix third and fourth
383 (ix86_expand_vector_extract): Likewise.
385 2005-01-19 Richard Henderson <rth@redhat.com>
388 * expr.c (emit_move_change_mode): New argument force; use
389 simplify_gen_subreg if true.
390 (emit_move_via_alt_mode): Merge into ...
391 (emit_move_via_integer): ... here.
392 (emit_move_ccmode): Use emit_move_change_mode directly.
394 2005-01-19 Richard Henderson <rth@redhat.com>
397 * config/alpha/alpha.c (alpha_rtx_costs): Handle HIGH.
398 (alpha_preferred_reload_class): Handle CONST_VECTOR.
399 (alpha_emit_set_const_1): Add no_output parameter; don't emit
401 (alpha_emit_set_const): Likewise. Make static.
402 (alpha_emit_set_long_const): Make static.
403 (alpha_extract_integer): Split out from alpha_expand_mov.
404 (alpha_split_const_mov): Likewise.
405 (alpha_expand_mov): Use them. Handle CONST_VECTOR.
406 (alpha_legitimate_constant_p): New.
407 * config/alpha/alpha-protos.h: Update.
408 * config/alpha/alpha.h (REGISTER_MOVE_COST): Correct fp<->gp cost.
409 (LEGITIMATE_CONSTANT_P): Re-implement with a function.
410 * config/alpha/alpha.md (movsi): Add n alternative.
411 (movsi_nt_vms, movdi_er_nofix, movdi_er_fix, movdi_fix): Likewise.
412 (mov<VEC>_fix, mov<VEC>_nofix): Add i alternative.
413 (splitters for all of the above): Use alpha_split_const_mov.
414 * config/alpha/predicates.md (non_add_const_operand): New.
415 (non_zero_const_operand): New.
416 (input_operand): Use alpha_legitimate_constant_p after reload.
418 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
420 PR tree-optimization/19038
421 * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): New function.
422 (add_candidate): Add ivs with increment in latch only if
423 allow_ip_end_pos_p is true.
424 (determine_iv_cost): Use empty_block_p.
426 2005-01-19 Daniel Berlin <dberlin@dberlin.org>
428 * cfganal.c (compute_dominance_frontiers_1): Replace with new algorithm
429 (compute_dominance_frontiers): Ditto.
431 2005-01-19 Ralf Corsepius <ralf.corsepius@rtems.org>
434 * config/sh/t-rtems: New.
435 * config.gcc (sh-*-rtems*): Reflect having added config/sh/t-rtems.
437 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
439 * tree-ssa-ccp.c (maybe_fold_offset_to_component_ref): Always subtract
440 the offset of the selected field.
442 2005-01-19 Kazu Hirata <kazu@cs.umass.edu>
444 * tree-cfg.c (remove_forwarder_block): Fix the check to
445 prevent a nonlocal label from appearing in the middle of a
448 2005-01-19 Hans-Peter Nilsson <hp@axis.com>
450 PR rtl-optimization/19462
451 * reorg.c (find_end_label): Create return insn only if
452 current_function_epilogue_delay_list is empty.
454 2005-01-19 Paolo Bonzini <bonzini@gnu.org>
456 * doc/passes.texi: Remove paragraph mentioning new-ra.
458 2005-01-19 Jakub Jelinek <jakub@redhat.com>
460 PR rtl-optimization/15139
461 * combine.c: Include params.h.
462 (count_rtxs): New function.
463 (record_value_for_reg): If replace_rtx would replace at least
464 2 occurrences of REG in VALUE and TEM is really large, replace REG with
465 (clobber (const_int 0)) instead of TEM.
466 * params.def (PARAM_MAX_LAST_VALUE_RTL): New.
467 * params.h (MAX_LAST_VALUE_RTL): New.
468 * Makefile.in (combine.o): Depend on $(PARAMS_H).
469 * doc/invoke.texi (--param max-last-value-rtl=N): Document.
472 * c-typeck.c (digest_init): Only call build_vector if all constructor
473 elements are *_CST nodes.
474 * gimplify.c (gimplify_init_constructor): Likewise.
477 * c-typeck.c (digest_init): Only call build_vector if inside_init
480 2005-01-18 Tobias Schl"uter <tobias.schlueter@physik.uni-muenchen.de>
482 * toplev.c (init_asm_output): Remove dead #ifdef.
484 2005-01-18 Kaz Kojima <kkojima@gcc.gnu.org>
486 * config/sh/linux.h (TARGET_C99_FUNCTIONS): Define.
488 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
490 * calls.c (expand_call): Check DECL_BUILT_IN_CLASS before
491 accessing DECL_FUNCTION_CODE.
492 * dojump.c (do_jump): Likewise.
493 * gimplify.c (gimplify_call_expr): Likewise.
494 * predict.c (expr_expected_value): Likewise.
495 (strip_builtin_expect): Likewise.
496 * tree-inline.c (estimate_num_insns_1): Likewise.
497 * tree-ssa-loop-im.c (stmt_cost): Likewise
498 * fold-const.c (fold): Test for BUILT_IN_NORMAL.
499 (tree_expr_nonnegative_p): Likewise.
501 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
503 * config/sparc/sparc.c (load_pic_register): Emit the appropriate
504 variant of the load_pcrel_sym pattern.
505 * config/sparc/sparc.md (P macro): Move to the top.
506 (load_pcrel_sym): Macroize using P.
507 (save_register_window): Likewise.
509 2005-01-18 Aldy Hernandez <aldyh@redhat.com>
511 * config/rs6000/rs6000.md ("sunordered"): Disable for e500.
514 2005-01-18 Andrew Pinski <pinskia@physics.uc.edu>
517 * c-typeck.c (build_asm_expr): Strip nops off
518 input memory operands.
520 2005-01-18 David Edelsohn <edelsohn@gnu.org>
522 * config/rs6000/aix43.h (CPLUSPLUS_CPP_SPEC): Delete
523 _XOPEN_SOURCE, _XOPEN_SOURCE_EXTENDED, _LARGE_FILE_API.
524 * config/rs6000/aix51.h (CPLUSPLUS_CPP_SPEC): Same.
525 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Same.
527 2005-01-18 Joel Sherrill <joel.sherrill@oarcorp.com>
530 * config/i386/i386.c (override_options): If the 80387 is disabled,
531 then do not return FP values using FP registers.
533 2005-01-18 Kazu Hirata <kazu@cs.umass.edu>
535 * basic-block.h, c-common.c, c-cppbuiltin.c, c-lang.c,
536 c-tree.h, cfgbuild.c, cgraph.c, cgraph.h, collect2.c,
537 combine.c, config.gcc, coverage.h, cse.c, cselib.c,
538 defaults.h, df.c, dwarf2asm.c, dwarf2out.c, explow.c, expr.c,
539 flow.c, fold-const.c, gcse.c, ggc-page.c, gimple-low.c,
540 gimplify.c, ifcvt.c, langhooks-def.h, lcm.c, optabs.h,
541 output.h, postreload-gcse.c, postreload.c, recog.c,
542 resource.c, rtl.def, rtlanal.c, sched-deps.c, sched-rgn.c,
543 targhooks.h, toplev.c, tree-data-ref.c, tree-eh.c,
544 tree-flow.h, tree-loop-linear.c, tree-mudflap.h, tree-nrv.c,
545 tree-optimize.c, tree-outof-ssa.c, tree-pass.h,
546 tree-scalar-evolution.c, tree-ssa-copy.c, tree-ssa-dce.c,
547 tree-ssa-dse.c, tree-ssa-loop-ivopts.c, tree-ssa-operands.c,
548 tree-ssa-pre.c, tree-ssa.c, tree-vectorizer.c, tree.def,
549 unwind-dw2-fde-darwin.c, var-tracking.c: Update copyright.
551 * config/darwin.c, config/darwin.h, config/freebsd-spec.h,
552 config/alpha/alpha.c, config/arm/vfp.md,
553 config/i386/emmintrin.h, config/i386/i386-modes.def,
554 config/i386/i386-protos.h, config/i386/i386.h,
555 config/i386/pmmintrin.h, config/i386/ppro.md,
556 config/i386/predicates.md, config/i386/xmmintrin.h,
557 config/i860/i860.c, config/ia64/ia64-modes.def,
558 config/ia64/ia64-protos.h, config/ia64/ia64.h,
559 config/ia64/ia64.md, config/ia64/itanium1.md,
560 config/ia64/itanium2.md, config/ia64/predicates.md,
561 config/iq2000/iq2000.h, config/mips/linux64.h,
562 config/rs6000/aix41.h, config/rs6000/aix43.h,
563 config/rs6000/aix52.h, config/rs6000/darwin-fallback.c,
564 config/rs6000/rs6000.c, config/rs6000/rs6000.h,
565 config/rs6000/spe.md, config/sh/sh.md, config/sparc/linux.h,
566 config/sparc/linux64.h, config/sparc/litecoff.h,
567 config/sparc/sp64-elf.h, config/sparc/sparc.c,
568 config/sparc/sparc.h, config/sparc/sparc.md,
569 config/sparc/sysv4-only.h: Update copyright.
571 2005-01-18 Richard Henderson <rth@redhat.com>
574 * config/i386/i386.c (ix86_expand_fp_movcc): Fail for LTGT and UNEQ.
576 2005-01-18 Marc Espie <espie@openbsd.org>
577 * config/i386/openbsdelf.h: Typo.
579 2005-01-18 Richard Henderson <rth@redhat.com>
581 * config/i386/i386.c (ix86_expand_fp_movcc): Remove TARGET_IEEE_FP
582 special case for sse.
583 * config/i386/predicates.md (sse_comparison_operator): Likewise.
585 2005-01-18 Dorit Naishlos <dorit@il.ibm.com>
587 * tree-inline.c (estimate_num_insns_1): Added cases for
588 ALIGN_INDIRECT_REF, MISALIGNED_INDIRECT_REF, and REALIGN_LOAD_EXPR.
590 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
592 PR rtl-optimization/19296
593 * combine.c (simplify_comparison): Rewrite the condition under
594 which a non-paradoxical SUBREG of a PLUS can be lifted when
595 compared against a constant.
597 2005-01-18 Andi Kleen <ak@muc.de>
599 * c-typeck.c: (convert_for_assignment): Check warn_pointer_sign.
600 * c.opt (-Wpointer-sign): Add.
601 * doc/invoke.texi: (-Wpointer-sign): Add.
603 2005-01-18 Uros Bizjak <uros@kss-loka.si>
605 * config/i386/i386.c (override_options): Revert 2004-11-24 change.
606 * config/i386/i386.md (fmodsf3, dremsf3, *sinsf2, *cossf2, sincossf3,
607 *tansf3_1, tansf2, atan2sf3_1, atan2sf3, atansf2, asinsf2, acossf2,
608 logsf2, log10sf2, log2sf2, log1psf2, logbsf2, ilogbsf2, expsf2,
609 exp10sf2, exp2sf2, expm1sf2, rintsf2, floorsf2, ceilsf2, btruncsf2,
610 nearbyintsf2): Disable for TARGET_SSE_MATH. Leave patterns enabled
611 for TARGET_MIX_SSE_I387.
612 (fmoddf3, dremdf3, *sindf2, *sinextendsfdf2, *cosdf2, *cosextendsfdf2,
613 sincosdf3, *sincosextendsfdf3, *tandf3_1, tandf2, atan2df3_1,
614 atan2df3, atandf2, asindf2, acosdf2, logdf2, log10df2, log2df2,
615 log1pdf2, logbdf2, expdf2, exp10df2, exp2df2, expm1df2, rintdf2,
616 floordf2, ceildf2, btruncdf2, nearbyintdf2): Disable for
617 (TARGET_SSE2 && TARGET_SSE_MATH). Leave patterns enabled for
619 (atan2sf3, atan2df3, atan2xf3): Remove register constraints
622 2005-01-18 Uros Bizjak <uros@kss-loka.si>
625 * config/i386/mmx.md (*movv2sf_internal_rex64, *movv2sf_internal):
626 Add movaps alternative for xmm reg->reg move.
628 2005-01-17 Jeff Law <law@redhat.com>
630 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Clear SSA_NAME_VALUE
631 at the end of the main DOM loop rather than just before DOM exits.
633 2005-01-17 Diego Novillo <dnovillo@redhat.com>
635 PR tree-optimization/19121
636 * tree-ssa-alias.c (compute_flow_sensitive_aliasing): When
637 adding aliases to a name tag, also add them to the pointer's
639 * tree-ssa-copy.c (merge_alias_info): Do not merge flow
640 sensitive alias info at all. Only check that the two pointers
641 have compatible pointed-to sets.
642 * tree-ssa.c (verify_name_tags): Verify that the alias set of
643 a pointer's type tag is a superset of the alias set of the
646 2005-01-17 James E Wilson <wilson@specifixinc.com>
649 * config/ia64/ia64.md (movxf): Handle general register source. Adjust
650 comment to document why.
652 2005-01-17 Richard Henderson <rth@redhat.com>
654 * config/i386/sse.md (smaxv4sf3_finite, sse_vmsmaxv4sf3_finite,
655 sminv4sf3_finite, sse_vmsminv4sf3_finite, smaxv2df3_finite,
656 sse2_vmsmaxv2df3_finite, sminv2df3_finite,
657 sse2_vmsminv2df3_finite): New.
658 (smaxv4sf3, sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, smaxv2df3,
659 sse2_vmsmaxv2df3, sminv2df3, sse2_vmsminv2df3): Remove commutative.
660 Force op1 into register.
662 2005-01-17 Kelley Cook <kcook@gcc.gnu.org>
664 * Makefile.in (STAGEMOVESTUFF): Stage all the stamp files to prevent
665 bubblestrap from always rebuilding build directory.
667 2005-01-17 Ulrich Weigand <uweigand@de.ibm.com>
669 * config/s390/s390.md ("*clc"): Remove incorrect '='.
670 ("*cmpmem_short"): Likewise.
672 2005-01-17 Steven Bosscher <stevenb@suse.de>
674 * basic-block.h: Document BB_* flags.
675 * regrename.c (copyprop_hardreg_forward): Don't use BB_VISITED,
676 use an sbitmap instead.
677 * sched-rgn.c (compute_trg_info): Likewise.
679 2005-01-17 Richard Sandiford <rsandifo@redhat.com>
681 * config.gcc (mips64*-*-linux*): Set the default abi to n32. Remove
682 redundant target_cpu_default and extra_parts lines. Use the default
684 * config/mips/linux64.h (DRIVER_DEFAULT_ABI_SELF_SPEC): Delete.
685 (SUBTARGET_EXTRA_SPECS): Delete.
686 (DRIVER_SELF_SPECS): Remove architecture lines. Set the ABI directly.
687 (SUBTARGET_ASM_SPEC): Remove -mabi=n32 and -mabi=64 mappings.
688 * config/mips/t-linux64 (MULTILIB_DIRNAMES): Use "32" for -mabi=32
689 and "n32" for -mabi=n32.
691 2005-01-17 Ian Lance Taylor <ian@airs.com>
694 * c-typeck.c (build_c_cast): Revert patch of 2003-02-16: don't
695 fold constant variables into initial values.
697 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
699 * tree-cfg.c (tree_can_merge_blocks_p): Reorder two checks.
701 * tree-cfg.c (tree_forwarder_block_p): Speed up by walking
702 through the statements backward.
704 2005-01-17 Ian Lance Taylor <ian@airs.com>
707 * tree-inline.c (expand_call_inline): Set TREE_NO_WARNING on
708 a variable set to the return value of the inlined function.
710 2005-01-17 Mark Dettinger <dettinge@de.ibm.com>
712 * rtlanal.c (rtx_cost): Assign cost of 0 to a SUBREG
713 when modes are tieable.
715 2005-01-17 Ranjit Mathew <rmathew@hotmail.com>
717 * gthr-posix.h (__gthread_active_p): Use pthread_cancel instead
718 of pthread_create to find out if threads are enabled.
719 * gthr-posix95.h (__gthread_active_p): Likewise.
721 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
723 * common.opt (-fnew-ra): Remove.
725 * toplev.h (flag_new_regalloc): Remove.
726 * Makefile.in (ra*.*): Don't mention.
727 * passes.c (rest_of_handle_new_regalloc): Remove.
728 (rest_of_handle_combine, rest_of_compilation): Always consider
729 flag_new_regalloc as false.
730 * doc/invoke.texi: Don't document -fnew-ra.
732 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
734 * bb-reorder.c (fix_edges_for_rarely_executed_code): Remove
735 last parameter to reg_scan.
736 * loop.c (loop_optimize): Likewise.
737 * passes.c (rest_of_handle_tracer, rest_of_handle_if_conversion,
738 rest_of_handle_web, rest_of_handle_cfg, rest_of_handle_jump_bypass,
739 rest_of_handle_life, rest_of_handle_cse, rest_of_handle_cse2,
740 rest_of_handle_gcse, rest_of_handle_loop_optimize,
741 rest_of_handle_loop2, rest_of_handle_jump2): Likewise.
742 * regclass.c (reg_scan): Likewise, for the declaration.
743 * rtl.h (reg_scan): Likewise, for the prototype.
745 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
747 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Don't call
750 2005-01-16 Daniel Berlin <dberlin@dberlin.org>
752 * tree-ssa-pre.c (add_to_sets): s1 may be NULL.
753 (compute_avail): Uses don't go in tmp_gen.
755 2005-01-16 Steven Bosscher <stevenb@suse.de>
757 * ggc-page.c (ggc_alloc_stat): Use __builtin_ctzl instead of a
758 loop to look for a free slot in a page entry.
760 2005-01-16 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
763 * defaults.h (TARGET_DEFERRED_OUTPUT_DEFS): Provide default.
764 * toplev.c (compile_file): Call process_pending_assemble_output_defs
765 just before targetm.asm_out.file_end.
766 * tree.h (process_pending_assemble_output_defs): Declare.
767 * varasm.c (assemble_output_def, process_pending_assemble_output_defs):
769 (assemble_alias): Defer generation of assembly code for defines when
770 TARGET_DEFERRED_OUTPUT_DEFS is true.
771 * config/rs6000/aix41.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
772 * config/rs6000/aix43.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
773 * doc/tm.texi (TARGET_DEFERRED_OUTPUT_DEFS): document.
775 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
778 * pa.c (pa_scalar_mode_supported_p): New function.
779 (TARGET_SCALAR_MODE_SUPPORTED_P): Define.
780 * pa.h (MIN_UNITS_PER_WORD): Add comment.
782 * pa-protos.h (prefetch_operand): Delete.
783 (prefetch_cc_operand, prefetch_nocc_operand): New declations.
784 * pa.c (prefetch_operand): Delete.
785 (prefetch_cc_operand, prefetch_nocc_operand): New functions.
786 * pa.h (EXTRA_CONSTRAINT): Add `W' constraint.
787 (PREDICATE_CODES): Delete prefetch_operand. Add prefetch_cc_operand
788 and prefetch_nocc_operand.
789 * pa.md (prefetch): Rework to avoid reload problems handling short
790 displacements when a cache control completer needs to be provided.
791 (prefetch_32, prefetch_64): Delete.
792 (prefetch_cc, prefetch_nocc): New patterns.
794 2005-01-15 David Edelsohn <edelsohn@gnu.org>
796 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Revert previous change.
798 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
801 * pa.c (pa_asm_output_aligned_common): Warn if specified alignment
802 exceeds maximum alignment for global common data.
803 * pa64-hpux.h (MAX_OFILE_ALIGNMENT): Define to 32768.
804 * som.h (MAX_OFILE_ALIGNMENT): Likewise.
806 2004-01-15 Roger Sayle <roger@eyesopen.com>
808 * tree-ssa-dom.c (extract_range_from_cond): Correct condition.
810 2004-01-15 Roger Sayle <roger@eyesopen.com>
812 * harg-reg-set.h (reg_class_names): Prototype global array.
813 * regclass.c (reg_class_names): Declare here and initialize to
815 (dump_regclass): Remove local declaration of reg_class_names.
816 (regclass): Likewise.
817 * cfg.c (dump_flow_info): Likewise.
818 * ra-debug.c (reg_class_names): Likewise.
819 * regrename.c (reg_class_names): Likewise.
820 * reload.c (reg_class_names): Likewise.
821 * reload1.c (spill_failure): Likewise.
822 * config/m68hc11/m68hc11.c (reg_class_names): Likewise.
824 2005-01-15 Ulrich Weigand <uweigand@de.ibm.com>
826 * config/s390/s390.md ("reload_outti"): Remove predicate for
827 output operand. Abort if operand is not a MEM.
828 ("reload_outdi", "reload_outdf"): Likewise.
830 2005-01-15 Marc Espie <espie@openbsd.org>
832 * config.gcc (*-*-openbsd*): Set HAS_LIBC_R for OpenBSD <= 3.2.
833 (i[34567]86-*-openbsd*): Switch to ELF for OpenBSD >= 3.4.
834 * config/openbsd.h: Fix C++ includes for native configurations.
835 Add proper OS_CPP_BUILTINS. Fix libspec for recent OpenBSD.
836 Add trampoline support.
837 * config/i386/openbsdelf.h: New.
839 2005-01-15 Marc Espie <espie@openbsd.org>
841 * collect2.c (main): Explicitly parse -dynamic-linker option.
843 2005-01-15 Jakub Jelinek <jakub@redhat.com>
845 PR tree-optimization/19060
846 * tree-ssa-dom.c (extract_range_from_cond) <case LT_EXPR, GT_EXPR>:
847 Return 0 if op1 <= TYPE_MIN_VALUE () resp. op1 >= TYPE_MAX_VALUE ().
848 (simplify_cond_and_lookup_avail_expr): Add assert for dummy == 0
849 and handle extract_range_from_cond returning false.
850 * fold-const.c (fold): Optimize comparisons with min/max even for
851 width > HOST_BITS_PER_WIDE_INT.
853 2005-01-15 Ralf Corsepius <ralf.corsepius@rtems.org>
855 * config/mips/rtems.h (MIPS_DEFAULT_GVALUE): Set to 0.
856 * config/mips/t-rtems (MULTILIBS_DIRNAMES,MULTILIB_OPTIONS):
857 Remove little endian multilib variants.
858 Add mips32 multilib variant.
860 2005-01-15 Kazu Hirata <kazu@cs.umass.edu>
862 * explow.c (copy_all_reg, stabilize): Remove.
863 * expr.h: Remove the corresponding prototypes.
865 * rtlanal.c (regs_set_between_p): Remove.
866 * rtl.h: Remove the corresponding prototype.
868 2005-01-14 David Edelsohn <edelsohn@gnu.org>
870 * config/rs6000/rs6000.c (rs6k_nonimmediate_operand): Rename
871 to rs6000_nonimmediate_operand.
872 * config/rs6000/rs6000.h (PREDICATE_CODES): Rename
873 rs6k_nonimmediate_operand to rs6000_nonimmediate_operand.
874 * config/rs6000/rs6000.md (movsi_internal1): Rename predicate to
875 rs6000_nonimmediate_operand.
876 * config/rs6000/spe.md (movdf_e500_double): Rename predicate to
877 rs6000_nonimmediate_operand.
879 2005-01-14 Andrew Pinski <pinskia@physics.uc.edu>
881 * tree-ssa-alias.c (pass_may_alias): Add TODO_verify_stmts.
883 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
885 * config/rs6000/rs6000.h (CLASS_MAX_NREGS): DF goes in 1 register
887 (CANNOT_CHANGE_MODE_CLASS): Restrict DI mode changes on e500v2.
888 (PREDICATE_CODES): Add rs6k_nonimmediate_operand.
890 * config/rs6000/rs6000.c (invalid_e500_subreg): New.
891 (rs6k_nonimmediate_operand): New.
892 (rs6000_legitimate_offset_address_p): Handle DI modes on e500v2
894 (legitimate_lo_sum_address_p): Same.
895 (rs6000_legitimize_address): Same.
896 (rs6000_legitimize_reload_address): Same.
897 (rs6000_legitimate_address): Same.
898 (spe_build_register_parallel): Pass DF and DC modes in a DI
901 * config/rs6000/rs6000.md ("*movsi_internal1"): Change predicate
902 to rs6k_nonimmediate_operand.
904 * config/rs6000/spe.md ("*frob_df_di"): New.
905 ("*frob_di_df"): New.
906 ("*frob_di_df_2"): New.
907 ("*mov_sidf_e500_subreg0"): New.
908 ("*mov_sidf_e500_subreg4"): New.
909 ("*movdf_e500_double"): Change predicate to
910 rs6k_nonimmediate_operand.
912 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
914 * postreload.c (move2add_note_store): Only call
915 trunc_int_for_mode on scalar integers.
917 2005-01-14 Steven Bosscher <stevenb@suse.de>
919 * tree-ssa-dce.c (visited_control_parents): New sbitmap to
920 replace BB_VISITED uses.
921 (find_obviously_necessary_stmts): Don't clear BB_VISITED.
922 (propagate_necessity): Check the bitmap instead of BB_VISITED.
923 (tree_dce_done): Free visited_control_parents.
924 (perform_tree_ssa_dce): Allocate and clear it.
925 * tree-ssa-pre.c (compute_antic_aux): Make non-recursive.
926 (compute_antic): Iterate from here using a DFS. Use an sbitmap
927 instead of BB_VISITED.
929 2005-01-14 Kazu Hirata <kazu@cs.umass.edu>
931 * c-tree.h, coverage.h, langhooks-def.h, optabs.h, output.h,
932 rtl.h, targhooks.h, tree-flow.h, tree-mudflap.h, tree.h:
933 Remove unused prototypes.
935 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
937 * config/rs6000/rs6000.c (rs6000_stack_info): Fix caching of SPE
938 64-bit register usage.
939 (spe_func_has_64bit_regs_p): Add FIXME note.
941 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
943 * config/rs6000/rs6000.c (rs6000_generate_compare): Replace
944 flag_finite_math_only with flag_unsafe_math_optimizations.
946 2005-01-14 J. D. Johnston <jjohnst@us.ibm.com>
948 * config/s390/tpf-unwind.h (s390_fallback_frame_state): Correct
950 (__tpf_eh_return): Copy TPF private stack area from the module
951 boundary stack frame. Add check for when module addresses are
952 equal, but no stub address is found.
954 2005-01-14 Richard Earnshaw <rearnsha@arm.com>
957 * arm.h (struct machine_function): Add call_via field.
958 (thumb_call_via_label): Declare.
959 * arm.c (thumb_call_via_label): New variable.
960 (thumb_call_reg_needed): New variable.
961 (arm_output_function_epilogue): For Thumb code, output any per-function
962 call-indirect trampolines.
963 (thumb_call_via_reg): New function.
964 (arm_file_end): New function.
965 (TARGET_ASM_FILE_END): Call arm_file_end.
966 (aof_file_end): Likewise.
967 * arm-protos.h (thumb_call_via_reg): Declare.
968 * arm.md (call_reg_thumb, call_value_reg_thumb): Call
969 thumb_call_via_reg in normal case.
971 2005-01-14 Jakub Jelinek <jakub@redhat.com>
974 PR rtl-optimization/19348
975 * recog.c (peephole2_optimize): Do global life update if some peephole
976 decides it doesn't need at least one of its inputs and that change
977 influences liveness at the start of the basic block.
979 * basic-block.h (EXECUTE_IF_AND_COMPL_IN_REG_SET): Needs 2 REGSET
980 arguments instead of 1.
982 2005-01-14 Eric Botcazou <ebotcazou@libertysurf.fr>
985 * varasm.c (initializer_constant_valid_p) <ADDR_EXPR>: Return
986 zero for nested functions needing a static chain or functions
987 with a non-constant address.
989 2005-01-13 Roger Sayle <roger@eyesopen.com>
991 * simplify-rtx.c (simplify_binary_operation) <AND>: Optimize
992 (and (sign_extend X) C) into (zero_extend (and X C)).
994 2005-01-13 David O'Brien <obrien@FreeBSD.org>
996 * config/freebsd-spec.h: Make KSE pthread lib logic the default.
998 2005-01-13 Richard Henderson <rth@redhat.com>
1003 * config/i386/i386.md (cmpdf, cmpsf, bunordered, bordered, buneq,
1004 bunge, bungt, bunle, bunlt, bltgt): Enable for TARGET_SSE_MATH,
1005 not just TARGET_SSE.
1006 (cmpfp_i_387): Rename from cmpfp_i. Move after sse patterns.
1007 (cmpfp_i_mixed): Rename from cmpfp_i_sse; use for TARGET_MIX_SSE_I387.
1008 (cmpfp_i_sse): Rename from cmpfp_i_sse_only; use for TARGET_SSE_MATH.
1009 (cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387): Similarly.
1010 (fp_jcc_1_mixed, fp_jcc_1_sse, fp_jcc_1_387): Similarly.
1011 (fp_jcc_2_mixed, fp_jcc_2_sse, fp_jcc_2_387): Similarly.
1012 (fp_jcc_3_387, fp_jcc_4_387, fp_jcc_5_387, fp_jcc_6_387,
1013 fp_jcc_7_387, fp_jcc_8_387): Rename from fp_jcc_N.
1014 (movdicc_c_rex64): Rename with '*'.
1015 (movsfcc, movdfcc): Add checks for 387 and sse math to condition.
1016 (movsfcc_1_sse_min, movsfcc_1_sse_max, movsfcc_1_sse): New.
1017 (movsfcc_1_387): Rename from movsfcc_1.
1018 (movdfcc_1_sse_min, movdfcc_1_sse_max, movdfcc_1_sse): New.
1019 (movdfcc_1, movdfcc_1_rex64): Add check for 387.
1020 (sminsf3, smaxsf3, smindf3, smaxdf3): New.
1021 (minsf3, minsf, minsf_nonieee, minsf_sse, mindf3, mindf,
1022 mindf_nonieee, mindf_sse, maxsf3, maxsf, maxsf_nonieee, maxsf_sse,
1023 maxdf3, maxdf, maxdf_nonieee, maxdf_sse, sse_movsfcc, sse_movsfcc_eq,
1024 sse_movdfcc, sse_movdfcc_eq, sse_movsfcc_const0_1,
1025 sse_movsfcc_const0_2, sse_movsfcc_const0_3, sse_movsfcc_const0_4,
1026 sse_movdfcc_const0_1, sse_movdfcc_const0_2, sse_movdfcc_const0_3,
1027 sse_movdfcc_const0_4): Remove.
1028 * config/i386/i386.c (ix86_expand_fp_movcc): For TARGET_SSE_MATH,
1029 recognize min/max early. Update for changed sse cmove patterns.
1030 (ix86_split_sse_movcc): New.
1031 * config/i386/i386-protos.h: Update.
1033 2005-01-13 Steven Bosscher <stevenb@suse.de>
1035 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
1036 (fix_stmt_v_may_defs): Likewise.
1038 2005-01-13 Richard Henderson <rth@redhat.com>
1040 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
1041 for converting the mask.
1043 2005-01-13 David Edelsohn <edelsohn@gnu.org>
1045 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
1048 2005-01-13 Richard Henderson <rth@redhat.com>
1050 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
1051 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
1052 IX86_BUILTIN_MOVDQ2Q): Remove.
1053 (IX86_BUILTIN_VEC_EXT_V4SI): New.
1054 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
1055 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
1057 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
1058 (sse2_pextrw): Fix immediate constraint.
1059 (sse2_loadq, sse2_loadq_rex64): Remove.
1060 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
1061 Use __builtin_ia32_vec_ext_<size>.
1062 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
1064 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
1066 * function.c (assign_parm_setup_block): Look inside original
1067 entry_parm when inspecting PARALLEL.
1069 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
1070 Joel Sherrill <joel@oarcorp.com>
1073 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
1074 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
1075 rtems_gxx_recursive_mutex_init.
1076 (__gthread_recursive_mutex_lock): New function.
1077 (__gthread_recursive_mutex_trylock): Likewise.
1078 (__gthread_recursive_mutex_unlock): Likewise.
1080 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
1082 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
1084 2005-01-13 Jan Beulich <jbeulich@novell.com>
1086 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
1087 (ashrti3_internal): Indicate output is early clobber. Generate result
1088 into output rather than first input. Use move for low word of output
1089 if shift count is exactly 64.
1090 (lshrti3_internal): Likewise.
1092 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
1096 * reload1.c (delete_output_reload): Don't delete an output reload
1097 if the pseudo lives longer than a single basic block. Adjust and
1098 improve wording of comment.
1100 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
1103 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
1104 of TYPE_SIZE where we mean the number of bytes.
1106 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
1108 * function.c (assign_parm_setup_block): Relax condition on
1109 multi-register optimization.
1111 2005-01-12 Nick Clifton <nickc@redhat.com>
1113 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
1114 an "in_delay_slot" attribute of "no" to prevent them being used in
1115 delay slots. This is forbidden because they might generate
1118 2005-01-12 Alan Modra <amodra@bigpond.net.au>
1121 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
1124 2005-01-12 Nick Clifton <nickc@redhat.com>
1126 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
1127 passed on to GAS) but do not define, as GAS no longer supports or
1128 needs the -m2000 option.
1130 2005-01-11 Roger Sayle <roger@eyesopen.com>
1132 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
1133 type as args of unprototyped builtins aren't automatically promoted.
1134 (fold_builtin_abs): Likewise.
1136 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1138 * Makefile.in: Set a `build-warn' variable.
1140 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
1142 PR tree-optimization/17949
1143 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
1144 (find_interesting_uses_address): Use it.
1146 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
1148 * regrename.c (kill_value): Handle subreg's that won't simplify.
1150 2005-01-11 Richard Henderson <rth@redhat.com>
1153 * config/i386/i386.h (enum ix86_builtins): Move ...
1154 * config/i386/i386.c: ... here.
1155 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
1156 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
1157 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
1158 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
1159 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
1160 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
1161 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
1162 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
1163 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
1164 IX86_BUILTIN_LOADDDUP): Remove.
1165 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
1166 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
1167 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
1168 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
1169 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
1170 (ix86_init_builtins): Make static.
1171 (ix86_init_mmx_sse_builtins): Update for changed builtins.
1172 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
1173 if all the modes match. Otherwise, fake it.
1174 (get_element_number, ix86_expand_vec_init_builtin,
1175 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
1176 (ix86_expand_builtin): Make static. Update for changed builtins.
1177 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
1178 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
1179 (ix86_expand_vector_init_duplicate): New.
1180 (ix86_expand_vector_init_low_nonzero): New.
1181 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
1182 Split out from ix86_expand_vector_init; handle integer modes.
1183 (ix86_expand_vector_init): Use them.
1184 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
1185 * config/i386/i386-protos.h: Update.
1186 * config/i386/predicates.md (reg_or_0_operand): New.
1187 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
1188 (movv2sf_internal): Likewise. And a splitter to match them all.
1189 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
1190 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
1191 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
1192 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
1193 (mmx_pinsrw): Fix operand ordering.
1194 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
1195 rather than sse_loadss expander.
1196 (movv2df splitter): Similarly.
1197 (sse_loadss, sse_loadlss): Remove.
1198 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
1199 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
1200 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
1201 (sse3_movddup): Rename with '*'.
1202 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
1203 (sse2_loadsd): Remove.
1204 (vec_dupv2df_sse3): Rename from sse3_loadddup.
1205 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
1206 (sse2_pinsrw): Fix argument ordering.
1207 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
1208 (sse2_stored): Remove 'r' destination.
1209 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
1210 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
1211 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
1212 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
1213 vec_extractv16qi, vec_initv16qi): New.
1215 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
1216 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
1217 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
1219 (_mm_load_pd, _mm_store_pd): Use plain dereference.
1220 (_mm_load_si128, _mm_store_si128): Likewise.
1221 (_mm_load1_pd): Use _mm_set1_pd.
1222 (_mm_load_sd): Use _mm_set_sd.
1223 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
1224 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
1225 (_mm_set_epi64): Use _mm_set_epi64x.
1226 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
1227 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
1228 _mm_setr_epi8): Use _mm_set_foo form.
1229 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
1230 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
1231 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
1232 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
1233 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
1234 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
1235 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
1236 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
1237 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
1238 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
1239 (_mm_movedup_pd): Use _mm_shuffle_pd.
1240 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
1241 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
1242 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
1243 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
1245 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
1246 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
1247 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
1248 (_mm_store_ps): Use raw dereference.
1249 (_mm_store1_ps): Use _mm_storeu_ps.
1250 (_mm_storer_ps): Use _mm_store_ps.
1251 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
1252 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
1254 2005-01-11 Stan Shebs <shebs@apple.com>
1256 * config/rs6000/rs6000.c (machopic_output_stub): Issue
1257 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
1259 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
1262 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
1263 Skip all DECLs except for FIELD_DECLs.
1265 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
1267 * config/s390/s390.c (override_options): Return error if
1268 -mbackchain, -mpacked-stack and -mhard-float are used together.
1269 (s390_va_start): Remove the backchain && packed-stack special case.
1270 (s390_gimplify_va_arg): Likewise.
1271 * doc/invoke.texi: Remove the ABI incompatibility note.
1273 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
1275 * config/s390/s390.c (struct s390_frame_layout): Remove
1277 (s390_frame_info, s390_emit_prologue): Replace occurrences of
1278 save_backchain_p with TARGET_BACKCHAIN.
1280 2005-01-11 Alan Modra <amodra@bigpond.net.au>
1283 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
1284 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
1285 * function.c (locate_and_pad_parm): Set new field.
1286 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
1287 Tweak where_pad test to include "none". Always set mem align for
1289 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
1290 not sufficient for type.
1291 (assign_parm_setup_block): If stack_parm is zero on entry, always
1292 make a new stack local. Block move old stack parm if necessary
1293 to new aligned stack local.
1294 (assign_parm_setup_stack): Use a block move to handle
1295 potentially misaligned entry_parm.
1296 (assign_parms_unsplit_complex): Specify required alignment when
1297 creating stack local.
1298 * calls.c (compute_argument_addresses): Override alignment of stack
1299 arg calculated from its type with the alignment given by
1300 FUNCTION_ARG_BOUNDARY.
1301 (store_one_arg): Likewise.
1303 2005-01-11 Jan Beulich <jbeulich@novell.com>
1305 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
1306 Change respective itanium_class attribute to ialu.
1307 (shladdp4_internal): New.
1308 * config/ia64/predicates.md (shladd_log2_operand): New.
1310 2005-01-11 Richard Henderson <rth@redhat.com>
1312 * expr.c (store_constructor): Use rtvec_alloc instead of
1313 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
1315 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
1317 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
1319 2005-01-11 Alan Modra <amodra@bigpond.net.au>
1321 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
1322 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
1323 (TARGET_RWRELOC): Define.
1324 (ia64_rwreloc_section_type_flags): Delete.
1325 (ia64_section_type_flags): New function.
1326 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
1327 (TARGET_RWRELOC): Define.
1329 2005-01-10 David Mosberger <davidm@hpl.hp.com>
1332 * config/ia64/ia64.c (process_set): For alloc insn, only call
1333 process_epilogue is !frame_pointer_needed.
1335 2005-01-10 Roger Sayle <roger@eyesopen.com>
1338 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
1339 unary operator and can't be treated as a binary/comparison operator.
1341 2005-01-10 Richard Henderson <rth@redhat.com>
1343 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
1344 (ix86_return_in_memory): Likewise.
1345 (function_arg_advance): Likewise. Mirror structure in function_arg
1346 for choosing register to advance.
1348 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
1350 * tree-vectorizer.c, tree.def: Fix comment typos.
1352 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
1354 * config/s390/s390.c (struct processor_costs): 4 new fields:
1355 ddbr, ddr, debr, der.
1356 (s390_rtx_costs): More precise handling of divide instructions.
1358 2005-01-10 David Edelsohn <edelsohn@gnu.org>
1360 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
1362 2005-01-10 Jan Beulich <jbeulich@novell.com>
1364 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
1365 names resulting from -ffunction-sections/-fdata-sections and linkonce
1367 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
1368 based on the section name. Rearrange the section name comparison logic
1369 slightly so that each section name is compared against at most once.
1371 2005-01-10 Ben Elliston <bje@au.ibm.com>
1373 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
1374 targets for which -fPIC reduces limitations on the GOT size.
1376 2005-01-09 Falk Hueffner <falk@debian.org>
1378 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
1381 2005-01-09 David Edelsohn <edelsohn@gnu.org>
1384 * collect2.c (main): Set aixrtl_flag for -brtl option.
1385 (resolve_lib_name): Search for .so file extension before .a
1388 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
1390 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
1391 (vect_analyze_data_refs_alignment): Add dump prints.
1393 2005-01-09 Ira Rosen <irar@il.ibm.com>
1395 * tree-vectorizer.c (vect_analyze_offset_expr): Use
1396 expr_invariant_in_loop_p.
1397 Initialize outputs first thing in the function.
1398 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
1399 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
1400 (vect_analyze_pointer_ref_access): Check that the initial condition of
1401 the access function is loop invariant.
1403 2005-01-09 Richard Henderson <rth@redhat.com>
1405 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
1406 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
1407 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
1408 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
1409 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
1410 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
1411 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
1412 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
1413 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
1414 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
1415 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
1416 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
1417 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
1418 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
1419 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
1420 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
1421 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
1422 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
1423 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
1424 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
1425 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
1426 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
1427 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
1429 (mmx_clrdi, pavgusb): Remove.
1430 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
1431 with leading sse_ prefix.
1432 * config/i386/sse.md: Receive them.
1433 * config/i386/mmx.md: New file.
1434 (MMXMODE12, MMXMODE24, mmxvecsize): New.
1435 (subrv2sf3): Turn into expander for normal subtraction.
1436 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
1437 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
1438 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
1439 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
1440 commutative; use ix86_binary_operator_ok.
1441 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
1442 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
1443 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
1444 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
1445 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
1446 Macroize from existing patterns; use ix86_binary_operator_ok.
1447 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
1448 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
1449 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
1450 with vec_select+vec_concat.
1451 (mmx_pshufw, mmx_pshufw_1): Likewise.
1452 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
1453 (mmx_uavgv4hi3): Model correctly.
1454 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
1455 _mm_xor_si64): Remove casts.
1457 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
1459 PR tree-optimization/19224
1460 * tree-scalar-evolution.c (get_instantiated_value,
1461 set_instantiated_value): New functions.
1462 (instantiate_parameters_1): Cache the results.
1463 (instantiate_parameters, resolve_mixers): Initialize and free
1466 2005-01-08 David Edelsohn <edelsohn@gnu.org>
1468 * config/i386/i386.md (addhi_4): Correct reference in comment.
1471 2005-01-08 Richard Henderson <rth@redhat.com>
1473 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
1474 (_mm_cvtsi128_si64x): Likewise.
1475 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
1476 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
1477 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
1478 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
1479 * config/i386/i386-protos.h: Update.
1480 * config/i386/i386.c (print_operand): Add 'H'.
1481 (ix86_fixup_binary_operands): Split out from ...
1482 (ix86_expand_binary_operator): ... here.
1483 (ix86_fixup_binary_operands_no_copy): New.
1484 (ix86_expand_fp_absneg_operator): Handle vector mode results.
1485 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
1486 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
1487 (safe_vector_operand): Use CONST0_RTX.
1488 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
1489 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
1490 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
1491 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
1493 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
1494 (IX86_BUILTIN_CMPNGESS): New.
1495 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
1496 (attr type): Add sselog1.
1497 (attr unit, attr memory): Handle it.
1498 (movti, movti_internal, movti_rex64): Move near other integer moves.
1499 (movtf, movtf_internal): Move near other fp moves.
1500 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
1501 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
1502 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
1503 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
1504 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
1505 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
1506 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
1507 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
1508 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
1509 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
1510 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
1511 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
1512 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
1513 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
1514 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
1515 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
1516 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
1517 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
1518 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
1519 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
1520 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
1521 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
1522 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
1523 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
1524 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
1525 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
1526 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
1527 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
1528 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
1529 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
1530 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
1531 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
1532 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
1533 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
1534 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
1535 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
1536 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
1537 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
1538 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
1539 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
1540 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
1541 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
1542 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
1543 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
1545 (SSEPUSH, push<SSEPUSH>): Remove.
1546 (MMXPUSH, push<MMXPUSH>): Remove.
1547 (sse_movaps, sse_movaps_1, sse_movups): Remove.
1548 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
1549 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
1550 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
1551 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
1552 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
1553 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
1554 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
1555 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
1556 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
1557 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
1558 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
1559 (ppro_sse_log_V4SF): Handle sselog1.
1560 * config/i386/predicates.md (const_0_to_1_operand): New.
1561 (const_0_to_255_mul_8_operand): New.
1562 (const_1_to_31_operand): Rename from const_int_1_31_operand.
1563 (const_2_to_3_operand, const_4_to_7_operand): New.
1564 * config/i386/sse.md: New file.
1565 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
1566 (sse_movups): Rename from sse_movups_1.
1567 (sse_loadlss): Rename from sse_loadss_1.
1568 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
1570 (negv4sf2): Use ix86_expand_fp_absneg_operator.
1571 (absv4sf2, negv2df, absv2df): New.
1572 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
1573 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
1574 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
1575 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
1576 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
1577 (sse3_addsubv4sf3): Model correctly.
1578 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
1579 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
1580 (sse_movhlps): Model with vec_select+vec_concat.
1581 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
1582 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
1583 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
1584 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
1585 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
1586 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
1587 sse2_pshufhw_1): Likewise.
1588 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
1589 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
1590 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
1591 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
1592 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
1593 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
1594 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
1595 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
1596 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
1597 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
1598 sminv8hi3): Mark commutative
1599 operands. Use ix86_binary_operator_ok.
1600 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
1601 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
1602 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
1603 sse2_punpcklqdq): Allow operand2 in memory.
1604 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
1605 sse2_movsd): Add memory alternatives.
1606 (sse_storelps): Turn expander into an insn; split after reload.
1607 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
1608 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
1610 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
1612 * configure.ac (DWARF-2 debug_line): Use objdump.
1613 * configure: Regenerate.
1615 2005-01-08 Jeff Law <law@redhat.com>
1616 Diego Novillo <dnovillo@redhat.com>
1618 PR tree-optimization/18241
1619 * tree-nrv.c (tree_nrv): Ignore volatile return values.
1620 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
1621 statements with volatile operands.
1622 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
1623 operands after marking a statement with has_volatile_ops.
1625 2005-01-08 Roger Sayle <roger@eyesopen.com>
1627 * tree.c (int_fits_type_p): Always honor integer constant
1628 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
1630 2005-01-08 Roger Sayle <roger@eyesopen.com>
1632 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
1633 the latent bug in rtl_delete_block has been resolved.
1635 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
1637 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
1638 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
1640 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
1642 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
1643 library as @multilib_dir@/$(SHLIB_SONAME).
1645 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
1647 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
1648 * configure: Regenerate.
1650 2005-01-07 Jakub Jelinek <jakub@redhat.com>
1652 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
1655 2005-01-07 David Edelsohn <edelsohn@gnu.org>
1658 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
1659 Convert non-word aligned offset address using ld/std into
1662 2005-01-07 Richard Henderson <rth@redhat.com>
1664 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
1666 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
1668 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
1670 2005-01-07 Jakub Jelinek <jakub@redhat.com>
1672 PR tree-optimization/19283
1673 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
1674 is not shorter than the original type.
1676 PR rtl-optimization/19012
1677 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
1679 PR rtl-optimization/18861
1680 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
1681 (FULL_STATE): Define.
1682 (mark_tablejump_edge): New function.
1683 (purge_dead_tablejump_edges): New function.
1684 (find_bb_boundaries): Use it.
1686 PR tree-optimization/18828
1687 * builtins.c (expand_builtin_next_arg): Remove argument and all
1688 the argument checking.
1689 (expand_builtin): Adjust caller.
1690 (expand_builtin_va_start): Likewise. Remove error for too many
1692 (fold_builtin_next_arg): Issue error for too many arguments.
1693 After checking arguments, replace them with magic arguments that
1694 prevent further checking of the args.
1696 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1698 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
1700 2005-01-06 Roger Sayle <roger@eyesopen.com>
1703 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
1704 and %{.S:...} (and their negative variants) to test whether the
1705 input file is assembler or pre-processed-assembler independent of
1706 the actual filename extension.
1708 2005-01-06 Roger Sayle <roger@eyesopen.com>
1710 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
1711 of sign or zero extended values.
1713 2005-01-06 Geoffrey Keating <geoffk@apple.com>
1715 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
1716 for IBM long double format correctly.
1718 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
1720 Fix PR tree-optimization/18792
1722 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
1723 to first_loop_depth, and use loop depth instead of loop number.
1724 (build_classic_dir_vector): Ditto.
1725 (compute_data_dependences_for_loop): Use depth, not loop number.
1726 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
1727 number. Pass in loops, instead of loop numbers.
1728 (gather_interchange_stats): Ditto.
1729 (linear_transform_loops): Ditto.
1731 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
1734 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
1735 biv_fits_mode_p, extension_within_bounds_p): New functions.
1736 (check_ext_dependent_givs): Use them.
1738 2005-01-06 Roger Sayle <roger@eyesopen.com>
1740 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
1741 more than one barrier, in which case we should delete them all.
1743 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
1745 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
1747 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
1749 Fix PR middle-end/19286
1751 * dwarf2out.c (gen_subprogram_die): If we've already tried to
1752 output this subprogram, simply ignore this attempt to do it again.
1753 (add_abstract_origin_attribute): Don't abort trying to add the abstract
1754 origin attribute if it's not possible.
1755 (gen_block_die): Don't ignore subblocks of "unused" blocks.
1756 (decls_for_scope): Ditto.
1757 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
1758 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
1760 2005-01-05 Richard Henderson <rth@redhat.com>
1763 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
1764 (bdesc_2arg): Use it.
1765 (ix86_expand_binop_builtin): Force operands into registers
1767 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
1768 ix86_expand_sse_compare, ix86_expand_sse_comi,
1769 ix86_expand_builtin): Likewise.
1771 2005-01-05 Richard Henderson <rth@redhat.com>
1773 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
1775 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
1776 (mulv4hi3): Set itanium_class mmmul.
1777 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
1778 (fpack): Rename from fpack_sfsf.
1780 2005-01-05 Richard Henderson <rth@redhat.com>
1783 * reload1.c (do_input_reload): Restrict the optimization deleteing
1784 a previous output reload to RELOAD_FOR_INPUT.
1786 2005-01-05 Steven Bosscher <stevenb@suse.de>
1788 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
1789 comment that we fall through after case.
1790 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
1791 (distribute_links): Likewise.
1792 * cse.c (cse_insn): Likewise.
1793 * cselib.c (cselib_invalidate_mem): Likewise.
1794 * df.c: Update comments at the top of the file.
1795 (read_modify_subreg_p): Update comments here too.
1796 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
1797 * flow.c (mark_set_1): Likewise.
1798 (mark_used_regs): Likewise.
1799 * gcse.c (mems_conflict_for_gcse_p): Likewise.
1800 (canon_list_insert): Likewise.
1801 (mark_set): Likewise.
1802 (try_replace_reg): Likewise.
1803 (store_killed_in_insn): Likewise.
1804 * loop.c (count_one_set): Likewise.
1805 (basic_induction_var): Likewise.
1806 * postreload-gcse.c (find_mem_conflicts): Likewise.
1807 * postreload.c (reload_combine_note_store): Likewise.
1808 (move2add_note_store): Likewise.
1809 * reload.c (find_equiv_reg): Likewise.
1810 (mark_referenced_resources): Likewise.
1811 * rtlanal.c (set_noop_p): Likewise.
1812 (note_stores): Likewise.
1813 (note_uses): Likewise.
1814 * sched-deps.c (sched_analyze_1): Likewise.
1815 * sched-rgn.c (check_live_1): Likewise.
1816 (update_live_1): Likewise.
1817 * config/i860/i860.c: Likewise.
1819 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
1821 (ZERO_EXTRACT): Mention that this one can be an lvalue.
1823 * doc/rtl.texi: Update documentation for bit-fields and SET.
1825 2005-01-05 Roger Sayle <roger@eyesopen.com>
1827 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
1829 2005-01-05 Jan Hubicka <jh@suse.cz>
1830 Richard Henderson <rth@redhat.com>
1833 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
1836 2005-01-05 Richard Henderson <rth@redhat.com>
1839 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
1840 (sse_shufps): Change operand 3 to const_int_operand.
1841 (sse2_storelps): Fix typo in template.
1842 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
1843 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
1844 (ix86_expand_builtin): Likewise.
1846 2005-01-05 Julian Brown <julian@codesourcery.com>
1848 * config/arm/arm.c (arm_return_in_memory): Treat complex types
1849 as aggregates for AAPCS ABIs.
1851 2005-01-05 Stan Shebs <shebs@apple.com>
1853 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
1854 Mach-O getters if ppc64.
1855 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
1857 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
1858 signed four-byte field for global code case.
1859 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
1861 2005-01-05 Roger Sayle <roger@eyesopen.com>
1864 * c-common.c: Include real.h.
1865 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
1866 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
1867 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
1868 * Makefile.in (c-common.o): Update dependencies.
1870 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1872 * c-parse.in (asm_string): Add trailing semicolon.
1874 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1876 * c-parse.in (asm_string): New. Don't allow wide strings in
1878 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
1879 asm_string instead of STRING.
1881 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1883 * c-typeck.c (constructor_no_implicit): Remove.
1884 (set_designator, process_init_element): Don't check
1885 constructor_no_implicit.
1887 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
1888 Kaz Kojima <kkojima@gcc.gnu.org>
1891 * lcm.c (create_pre_exit): New.
1892 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
1893 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
1895 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
1897 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
1898 -msoft-float in target_float_switch.
1899 * arm.c (arm_override_options): Fix processing of target_float_switch.
1901 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
1903 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
1906 2004-01-05 Uros Bizjak <uros@kss-loka.si>
1908 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
1909 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
1911 2005-01-04 Roger Sayle <roger@eyesopen.com>
1913 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
1914 if deleting the then-block allows the test-block to fallthru to the
1917 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
1920 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
1921 after the full declaration if the are in two different TUs.
1923 2005-01-04 Richard Henderson <rth@redhat.com>
1926 * tree-sra.c (generate_one_element_init): Just
1927 call gimplify_and_add.
1928 (generate_element_init): Record the
1929 new referenced variables and mark them for renaming
1930 and split out to ...
1931 (generate_element_init_1): This.
1932 (scalarize_init): Don't call push_gimplify_context/
1933 pop_gimplify_context.
1935 2005-01-04 Geoffrey Keating <geoffk@apple.com>
1937 * toplev.c (get_src_pwd): Handle failure of getpwd().
1939 2005-01-04 Roger Sayle <roger@eyesopen.com>
1941 * fold-const.c (fold_single_bit_test): Delete unreachable handling
1943 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
1944 all the cases handled by it are inverted by invert_truthvalue.
1946 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
1948 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
1949 (save_register_windowsi): Likewise.
1951 2005-01-04 Richard Henderson <rth@redhat.com>
1953 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
1955 2005-01-04 Uros Bizjak <uros@kss-loka.si>
1958 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
1959 any simplifications of vector mode comparison operators.
1960 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
1962 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
1963 Devang Patel <dpatel@apple.com>
1965 PR tree-optimization/18308
1966 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
1967 the operands before creating a new expression.
1968 * dojump.c (do_jump): Make drop_through_label available
1969 for all cases. Add expansion of COND_EXPR.
1971 2005-01-04 Ira Rosen <irar@il.ibm.com>
1973 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
1974 INTEGER_CST instead of TREE_CONSTANT.
1975 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
1976 instead of TREE_CONSTANT.
1977 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
1978 instead of TREE_CONSTANT.
1980 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
1982 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
1984 2005-01-04 Uros Bizjak <uros@kss-loka.si>
1987 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
1988 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
1990 2005-01-03 Richard Henderson <rth@redhat.com>
1992 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
1993 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
1994 fold_convert_const_real_from_real): Split out from ...
1995 (fold_convert_const): ... here.
1997 2005-01-03 Richard Henderson <rth@redhat.com>
2000 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
2001 (mov<MMXMODEI>_internal): Likewise.
2002 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
2003 Add V2SF case; use it for SSE1; don't use TI.
2004 (movdf_integer): Likewise.
2005 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
2007 2005-01-03 Ira Rosen <irar@il.ibm.com>
2009 * tree-vectorizer.c (vect_strip_conversions): New function.
2010 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
2011 check for binary class.
2013 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
2017 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
2019 * gimple-low.c (mark_blocks_with_used_vars): New function.
2020 (mark_blocks_with_used_subblocks): Ditto.
2021 (mark_used_blocks): Ditto.
2022 (pass_mark_used_blocks): New pass.
2023 * tree-inline.c: Include debug.h.
2024 (expand_call_inline): Call outlining_inline_function here.
2025 * tree-optimize.c (init_tree_optimization_passes): Add
2026 pass_mark_used_blocks.
2027 * tree-pass.h (pass_mark_used_blocks): New.
2028 * Makefile.in (tree-inline.o): Add debug.h dependency.
2030 2005-01-03 Geoffrey Keating <geoffk@apple.com>
2032 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
2033 VAR_DECLs to have weak_import attribute.
2035 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
2038 2004-01-03 Steven Bosscher <stevenb@suse.de>
2040 * passes.c (rest_of_compilation): Don't run regmove if only
2041 flag_expensive_optimizations, require flag_regmove instead.
2043 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
2045 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
2047 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
2048 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
2049 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
2050 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
2052 2005-01-03 Richard Henderson <rth@redhat.com>
2054 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
2055 (ia64_const_ok_for_letter_p): New.
2056 (ia64_const_double_ok_for_letter_p): New.
2057 (ia64_extra_constraint): New.
2058 (ia64_expand_vecint_compare): New.
2059 (ia64_expand_vcondu_v2si): New.
2060 (ia64_expand_vecint_cmov): New.
2061 (ia64_expand_vecint_minmax): New.
2062 (ia64_print_operand): Add 'v'.
2063 (ia64_preferred_reload_class): New.
2064 (ia64_vector_mode_supported_p): New.
2065 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
2066 (PREFERRED_RELOAD_CLASS): Move to function.
2067 (CONST_OK_FOR_LETTER_P): Move to function.
2068 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
2069 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
2070 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
2071 (EXTRA_CONSTRAINT): Move to function.
2072 * config/ia64/ia64.md: Include vect.md.
2073 (itanium_class): Add mmalua.
2075 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
2077 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
2078 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
2080 (const_int_2bit_operand): New.
2081 (fr_reg_or_0_operand): New.
2082 * config/ia64/ia64-modes.def: Add vector modes.
2083 * config/ia64/ia64-protos.h: Update.
2084 * config/ia64/vect.md: New file.
2086 2005-01-03 Richard Henderson <rth@redhat.com>
2088 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
2090 2005-01-03 Uros Bizjak <uros@kss-loka.si>
2093 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
2095 (log1pdf2): Change mode of operands[0,1] to DFmode.
2097 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
2099 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
2100 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
2101 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
2102 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
2103 SPARC_RELAXED_ORDERING.
2105 2005-01-03 Richard Henderson <rth@redhat.com>
2106 Uros Bizjak <uros@kss-loka.si>
2109 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
2110 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
2111 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
2113 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
2114 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
2115 const_pow2_1_to_128_operand): New predicates.
2117 2005-01-02 Greg McGary <greg@mcgary.org>
2119 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
2121 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
2122 object. [ptr_type]: Remove unused variable.
2123 Remember nearest addressable array-element or record-component
2124 when checking bitfield components. Tolerate empty BIND_EXPR.
2126 2005-01-01 Richard Henderson <rth@redhat.com>
2128 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
2129 that don't narrow the value. Fail for other conversions.
2131 2005-01-01 Richard Henderson <rth@redhat.com>
2134 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
2135 * c-lang.c (finish_file): Don't do it here.
2136 * objc/objc-act.c (objc_finish_file): Likewise.
2138 * cgraph.c (decl_assembler_name_equal): New.
2139 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
2140 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
2141 * cgraph.h (struct cgraph_varpool_node): Add next.
2142 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
2143 * varasm.c (assemble_alias): Mark the target as needed.
2145 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
2148 * function.c (get_arg_pointer_save_area): Use entry_of_function
2149 instead of get_insns.
2151 2005-01-01 Roger Sayle <roger@eyesopen.com>
2152 Andrew Pinski <pinskia@physics.uc.edu>
2153 James E. Wilson <wilson@specifixinc.com>
2155 PR rtl-optimization/12092
2156 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
2159 2005-01-01 Roger Sayle <roger@eyesopen.com>
2160 Olivier Hainque <hainque@act-europe.fr>
2162 * tree.c (int_fits_type_p): A narrower type always fits in a
2163 wider one, except for negative values into unsigned types.
2165 2005-01-01 Roger Sayle <roger@eyesopen.com>
2167 * tree.c (int_fits_type_p): Compare the result of force_fit_type
2168 with the original constant rather than require TREE_OVERFLOW.
2170 2005-01-01 Steven Bosscher <stevenb@suse.de>
2173 * c-decl.c (finish_function): If compiling C99, annotate the
2174 compiler generated return with the current file name and line 0.
2175 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
2176 the source line is greater than 0.
2177 (remove_bb): Likewise.
2179 See ChangeLog.12 for earlier changes.