1 2005-01-23 Steven Bosscher <stevenb@suse.de>
3 * gengtype.c (walk_type): Produce `not equal to' compares for loop
4 tests, instead of `less than'.
6 2005-01-23 Richard Sandiford <rsandifo@redhat.com>
8 * function.c (assign_parm_setup_block): When creating a new stack slot
9 for a parameter, get its alignment from the parameter's DECL_ALIGN
10 rather than the type's TYPE_ALIGN. Make sure that the parameter
11 is at least word aligned.
13 2005-01-22 Roger Sayle <roger@eyesopen.com>
16 * config/avr/avr.c (avr_hard_regno_mode_ok): Rewrite.
18 2005-01-22 Richard Henderson <rth@redhat.com>
21 * config/i386/i386.md (movsfcc_1_sse_max): Use nonimmediate_operand
22 in both compare operands.
23 (movdfcc_1_sse_max): Likewise.
24 (movsfcc_1_sse): Likewise. Add earlyclobber for scratch.
25 (movdfcc_1_sse): Likewise.
26 * config/i386/i386.c (ix86_split_sse_movcc): Emit copies into the
27 scratch register as needed.
29 2005-01-22 Richard Henderson <rth@redhat.com>
31 * config/i386/i386.md (smaxsf3): Fix mnemonic typo.
33 * config/i386/i386.c (ix86_prepare_fp_compare_args): Fix is_sse test.
35 2005-01-22 Richard Henderson <rth@redhat.com>
37 * genopinit.c (optabs): Use smin/smax for floating point too.
38 * doc/md.texi: Update to match. Clarify that floating point
39 results are undefined for +0/-0 and NaN.
40 * doc/rtl.texi: Likewise.
41 * rtl.def (SMIN, SMAX): Likewise
42 * tree.def (MIN_EXPR, MAX_EXPR): Likewise.
44 * config/alpha/alpha.md (smaxdf3, smindf3, smaxsf3, sminsf3): Add
45 leading 's' to the name.
46 * config/ia64/ia64.md (smaxsf3, sminsf3, smaxdf3, smindf3,
47 smaxxf3, sminxf3): Likewise.
48 * config/rs6000/rs6000.md (smaxdf3, smindf3, smaxsf3, sminsf3):
51 2005-01-22 Paul Brook <paul@codesourcery.com>
53 * config/arm/arm.md: Use "Uy" constraint for wcgr load/stores.
55 2005-01-21 Mark Dettinger <dettinge@de.ibm.com>
57 * config/s390/s390.c (struct processor_costs): New fields
58 dlgr, dlr, dr, dsgfr, dsgr.
59 (z900_cost, z990_cost): Values for new fields.
60 (s390_rtx_costs): New cases MEM und COMPARE in switch
61 statement. Modified handling of SIGN_EXTEND, ZERO_EXTEND,
64 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
66 * config/s390/s390.md ("doloop_si64"): Reload input value directly
67 into the register being decremented.
68 ("doloop_si31", "doloop_di"): Likewise.
69 ("*doloop_si_long"): Adapt pattern.
71 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
73 * config/s390/s390.h (HARD_REGNO_NREGS): Fix computation for
75 (CLASS_MAX_NREGS): Likewise.
77 2005-01-21 Daniel Berlin <dberlin@dberlin.org>
79 * doc/tree-ssa.texi (Statement Operands): Add example for new
80 must-def macro. Note deprecation of old operands interface.
82 2005-01-22 Richard Sandiford <rsandifo@redhat.com>
84 PR tree-optimization/19484
85 * tree-cfg.c (remove_fallthru_edge): New function.
86 (cleanup_control_flow): Remove fallthru edges from calls that are
87 now known not to return.
89 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
91 Fix PR tree-optimization/19038
92 * tree-ssa-dom.c (cprop_operand): Don't replace loop invaeriant
93 copies with loop variant ones.
95 2005-01-22 Kazu Hirata <kazu@cs.umass.edu>
97 * cfganal.c, real.h, reorg.c, timevar.def, tree-ssa-ccp.c,
98 config/alpha/alpha-protos.h, config/alpha/alpha.h,
99 config/alpha/alpha.md, config/alpha/predicates.md,
100 config/sparc/freebsd.h, config/sparc/netbsd-elf.h,
101 config/sparc/sol2.h: Update copyright.
103 * tree-cfg.c (remove_forwarder_block_with_phi): Look at the
104 first label to see if it is a nonlocal label.
106 2005-01-22 David Edelsohn <edelsohn@gnu.org>
107 Andrew Pinski <pinskia@physics.uc.edu>
110 * config/rs6000/rs6000.c (rs6000_va_start): Saturate n_gpr at
111 maximum number of GPRs. Saturate n_fpr at maximum number of FPRs.
113 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
115 * config/rs6000/rtems.h: Update copyright.
117 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
120 * config/rs6000/rtems.h: Resurrect cpp_os_rtems_spec from gcc < 3.4.
121 (CPP_OS_RTEMS_SPEC): New (From gcc-3.3's config/rs6000/sys4.h).
122 (SUBSUBTARGET_EXTRA_SPECS): Use CPP_OS_RTEMS_SPEC.
124 2005-01-22 Volker Reichelt <reichelt@igpm.rwth-aachen.de>
127 * c-typeck.c (convert_arguments): Check for error_mark_node.
129 2005-01-21 Ian Lance Taylor <ian@airs.com>
131 PR tree-optimization/13000
132 * tree-inline.c: Include "tree-flow.h".
133 (expand_call_inline): If warn_return_type, warn if non-void inline
134 function falls through.
135 * tree-cfg.c (execute_warn_function_return): Don't warn about
136 control reaching end if TREE_NO_WARNING is set. Set
138 * gimple-low.c (block_may_fallthru): Don't assume that SWITCH_EXPR
140 * gimplify.c (shortcut_cond_expr): Don't emit a jump over the else
141 branch if we don't need one.
142 * c-typeck.c: Include "tree-flow.h"
143 (c_finish_bc_stmt): Don't add a goto if the current statement
144 list doesn't fall through to the current point.
146 2005-01-21 Roger Sayle <roger@eyesopen.com>
148 PR rtl-optimization/576
149 * real.c (real_arithmetic): Change return type from void to bool
150 to return an indication that the result may be inexact.
151 * real.h (real_arithmeric): Update prototype.
152 * fold-const.c (const_binop): Don't constant fold floating
153 point expressions when the user specifies -frounding-math and
154 the result may depend upon the run-time rounding mode.
155 (fold_convert_const_real_from_real): Clean-up.
156 (fold_initializer): Ignore flag_rounding_math for initializers.
157 * simplify-rtx.c (simplify_binary_operation): Likewise, don't
158 constant fold FP operations with flag_rounding_math if the
159 result may depend upon the run-time rounding mode.
161 2005-01-21 Tom Tromey <tromey@redhat.com>
163 * c-cppbuiltin.c (define__GNUC__): Correct assertion.
165 2005-01-21 Kazu Hirata <kazu@cs.umass.edu>
167 * tree-cfg.c: Fix comment typos.
169 2005-01-21 J"orn Rennecke <joern.rennecke@st.com>
171 * passes.c (rest_of_handle_flow2): Close / open the flow2 dump file
172 around the call to rest_of_handle_branch_target_load_optimize.
174 2005-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
176 * config.gcc (sparc-*-netbsdelf*): Include dbxelf.h
177 (sparc64-*-openbsd*): Likewise.
178 (sparclite-*-elf*): Include tm-dwarf2.h.
179 (sparc86x-*-elf*): Likewise.
180 (sparc64-*-elf*): Likewise.
181 * config/sparc/linux64.h (DWARF2_DEBUGGING_INFO): Delete.
182 (DBX_DEBUGGING_INFO): Likewise.
183 * config/sparc/liteelf.h (DWARF2_DEBUGGING_INFO): Likewise.
184 (PREFERRED_DEBUGGING_TYPE): Likewise.
185 * config/sparc/netbsd-elf.h (PREFERRED_DEBUGGING_TYPE): Likewise.
186 * config/sparc/sp64-elf.h (DBX_DEBUGGING_INFO): Likewise.
187 (PREFERRED_DEBUGGING_TYPE): Likewise.
188 * config/sparc/sp86x-elf.h (DWARF2_DEBUGGING_INFO): Likewise.
189 (PREFERRED_DEBUGGING_TYPE): Likewise.
190 * config/sparc/sparc.h (DBX_DEBUGGING_INFO): Likewise.
192 2005-01-21 Hans-Peter Nilsson <hp@bitrange.com>
195 * combine.c (combine_simplify_rtx): Revert change of 2004-12-31.
197 2005-01-20 Paul Brook <paul@codesourcery.com>
199 * doc/rtl.texi: Document value extension requirements for CONST_INT.
201 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
203 * tree-cfg.c (tree_verify_flow_info): Check that a nonlocal
204 label is first in a sequence of labels.
206 * tree-cfg.c (tree_verify_flow_info): Fix a typo.
208 2005-01-20 Janis Johnson <janis187@us.ibm.com>
209 Giovanni Bajo <giovannibajo@gcc.gnu.org>
211 * doc/sourcebuild.texi (Test Directives): New.
213 2005-01-20 Roger Sayle <roger@eyesopen.com>
215 * real.c (real_floor): Don't leave the result uninitialized when
216 mode is VOIDMode, but return the unrounded intermediate value.
217 (real_ceil): Likewise.
219 2005-01-20 Paul Brook <paul@codesourcery.com>
221 * config/arm/arm.md (insv): Use gen_int_mode.
223 2005-01-20 Eric Botcazou <ebotcazou@libertysurf.fr>
225 * config/sparc/freebsd.h (ASM_OUTPUT_INTERNAL_LABELREF): Delete.
226 * config/sparc/linux.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
227 * config/sparc/linux64.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
228 * config/sparc/netbsd-elf.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
229 * config/sparc/sol2.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
231 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
234 * dwarf2out.c (concat_loc_descriptor): We don't know we can use
235 fbreg, so we have to assume we can't.
236 (loc_descriptor_from_tree_1): Ditto.
237 (containing_function_has_frame_base): New function.
238 (add_location_or_const_value_attribute): Use it.
239 Also try to generate a frame_base from a single element location
242 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
244 PR tree-optimization/15349
245 * timevar.def (TV_TREE_MERGE_PHI): New.
246 * tree-cfg.c (tree_forwarder_block_p): Add a new argument
248 (remove_forwarder_block, cleanup_forwarder_blocks): Adjust the
249 calls to tree_forwarder_block_p.
250 (remove_forwarder_block_with_phi, merge_phi_nodes,
251 gate_merge_phi, pass_merge_phi): New.
252 * tree-optimize.c (init_tree_optimization_passes): Add
254 * tree-pass.h: Add an extern for pass_merge_phi;
256 2005-01-20 Richard Henderson <rth@redhat.com>
259 * config/i386/emmintrin.h (_mm_castpd_ps, _mm_castpd_si128): New.
260 (_mm_castps_pd, _mm_castps_si128): New.
261 (_mm_castsi128_ps, _mm_castsi128_pd): New.
263 2005-01-20 Richard Henderson <rth@redhat.com>
266 * config/i386/mmintrin.h (_mm_cvtsi32_si64): Use
267 __builtin_ia32_vec_init_v2si.
268 (_mm_cvtsi64_si32): Use __builtin_ia32_vec_ext_v2si.
269 * config/i386/i386.c (IX86_BUILTIN_VEC_EXT_V2SI): New.
270 (ix86_init_mmx_sse_builtins): Create it.
271 (ix86_expand_builtin): Expand it.
272 (ix86_expand_vector_set): Handle V2SFmode and V2SImode.
273 * config/i386/mmx.md (vec_extractv2sf_0, vec_extractv2sf_1): New.
274 (vec_extractv2si_0, vec_extractv2si_1): New.
276 2005-01-20 Richard Henderson <rth@redhat.com>
279 * config/i386/i386.c (ix86_expand_push): New.
280 * config/i386/mmx.md (push<MMXMODE>1): New.
281 * config/i386/sse.md (push<SSEMODE>1): New.
282 * config/i386/i386-protos.h: Update.
284 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
286 * params.def, params.h, predict.c, tree-ssa-loop-im.c: Update
289 2005-01-20 Richard Henderson <rth@redhat.com>
291 * rtl.def (CONST_VECTOR): Use RTX_CONST_OBJ.
292 * rtl.h (CONSTANT_P): Don't special case CONST_VECTOR.
294 2005-01-19 Richard Henderson <rth@redhat.com>
297 * config/i386/i386.c (ix86_expand_vector_move_misalign): Convert
298 to V4SFmode in SSE1 fallback load path.
300 2005-01-19 Richard Henderson <rth@redhat.com>
302 * config/i386/i386.c (ix86_expand_vector_init_one_var): Fix typo
303 in QImode expansion to ix86_expand_vector_set.
304 (ix86_expand_vector_init_general): Fix typo in V8HImode recursive call.
306 2005-01-19 Richard Henderson <rth@redhat.com>
309 * config/i386/i386.c (ix86_preferred_reload_class): Return a proper
310 subclass of the input class.
311 (ix86_secondary_memory_needed): Always true for cross-MMX classes.
312 Always true for cross-SSE1 classes. Rationalize conditionals.
313 * config/i386/i386.h (SSE_CLASS_P, MMX_CLASS_P): Use straight equality.
314 * config/i386/i386.md (movsi_1): Add MMX/SSE zeros. Fix alternatives
315 for SSE1. Don't check TARGET_INTER_UNIT_MOVES.
316 (movdi_2): Add MMX/SSE zeros.
317 (movdi_1_rex64): Likewise. Don't check TARGET_INTER_UNIT_MOVES.
318 (movsf_1): Don't check TARGET_INTER_UNIT_MOVES.
319 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
320 (movsi_1_nointernunit, movdi_1_rex64_nointerunit): Remove.
321 (movsf_1_nointerunit, zero_extendsidi2_32_1): Remove.
322 (zero_extendsidi2_rex64_1): Remove.
323 (MOV0 peephole): Check GENERAL_REG_P.
325 2005-01-19 Richard Henderson <rth@redhat.com>
328 * config/i386/i386.c (ix86_expand_vector_set): Fix third and fourth
330 (ix86_expand_vector_extract): Likewise.
332 2005-01-19 Richard Henderson <rth@redhat.com>
335 * expr.c (emit_move_change_mode): New argument force; use
336 simplify_gen_subreg if true.
337 (emit_move_via_alt_mode): Merge into ...
338 (emit_move_via_integer): ... here.
339 (emit_move_ccmode): Use emit_move_change_mode directly.
341 2005-01-19 Richard Henderson <rth@redhat.com>
344 * config/alpha/alpha.c (alpha_rtx_costs): Handle HIGH.
345 (alpha_preferred_reload_class): Handle CONST_VECTOR.
346 (alpha_emit_set_const_1): Add no_output parameter; don't emit
348 (alpha_emit_set_const): Likewise. Make static.
349 (alpha_emit_set_long_const): Make static.
350 (alpha_extract_integer): Split out from alpha_expand_mov.
351 (alpha_split_const_mov): Likewise.
352 (alpha_expand_mov): Use them. Handle CONST_VECTOR.
353 (alpha_legitimate_constant_p): New.
354 * config/alpha/alpha-protos.h: Update.
355 * config/alpha/alpha.h (REGISTER_MOVE_COST): Correct fp<->gp cost.
356 (LEGITIMATE_CONSTANT_P): Re-implement with a function.
357 * config/alpha/alpha.md (movsi): Add n alternative.
358 (movsi_nt_vms, movdi_er_nofix, movdi_er_fix, movdi_fix): Likewise.
359 (mov<VEC>_fix, mov<VEC>_nofix): Add i alternative.
360 (splitters for all of the above): Use alpha_split_const_mov.
361 * config/alpha/predicates.md (non_add_const_operand): New.
362 (non_zero_const_operand): New.
363 (input_operand): Use alpha_legitimate_constant_p after reload.
365 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
367 PR tree-optimization/19038
368 * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): New function.
369 (add_candidate): Add ivs with increment in latch only if
370 allow_ip_end_pos_p is true.
371 (determine_iv_cost): Use empty_block_p.
373 2005-01-19 Daniel Berlin <dberlin@dberlin.org>
375 * cfganal.c (compute_dominance_frontiers_1): Replace with new algorithm
376 (compute_dominance_frontiers): Ditto.
378 2005-01-19 Ralf Corsepius <ralf.corsepius@rtems.org>
381 * config/sh/t-rtems: New.
382 * config.gcc (sh-*-rtems*): Reflect having added config/sh/t-rtems.
384 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
386 * tree-ssa-ccp.c (maybe_fold_offset_to_component_ref): Always subtract
387 the offset of the selected field.
389 2005-01-19 Kazu Hirata <kazu@cs.umass.edu>
391 * tree-cfg.c (remove_forwarder_block): Fix the check to
392 prevent a nonlocal label from appearing in the middle of a
395 2005-01-19 Hans-Peter Nilsson <hp@axis.com>
397 PR rtl-optimization/19462
398 * reorg.c (find_end_label): Create return insn only if
399 current_function_epilogue_delay_list is empty.
401 2005-01-19 Paolo Bonzini <bonzini@gnu.org>
403 * doc/passes.texi: Remove paragraph mentioning new-ra.
405 2005-01-19 Jakub Jelinek <jakub@redhat.com>
407 PR rtl-optimization/15139
408 * combine.c: Include params.h.
409 (count_rtxs): New function.
410 (record_value_for_reg): If replace_rtx would replace at least
411 2 occurrences of REG in VALUE and TEM is really large, replace REG with
412 (clobber (const_int 0)) instead of TEM.
413 * params.def (PARAM_MAX_LAST_VALUE_RTL): New.
414 * params.h (MAX_LAST_VALUE_RTL): New.
415 * Makefile.in (combine.o): Depend on $(PARAMS_H).
416 * doc/invoke.texi (--param max-last-value-rtl=N): Document.
419 * c-typeck.c (digest_init): Only call build_vector if all constructor
420 elements are *_CST nodes.
421 * gimplify.c (gimplify_init_constructor): Likewise.
424 * c-typeck.c (digest_init): Only call build_vector if inside_init
427 2005-01-18 Tobias Schl"uter <tobias.schlueter@physik.uni-muenchen.de>
429 * toplev.c (init_asm_output): Remove dead #ifdef.
431 2005-01-18 Kaz Kojima <kkojima@gcc.gnu.org>
433 * config/sh/linux.h (TARGET_C99_FUNCTIONS): Define.
435 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
437 * calls.c (expand_call): Check DECL_BUILT_IN_CLASS before
438 accessing DECL_FUNCTION_CODE.
439 * dojump.c (do_jump): Likewise.
440 * gimplify.c (gimplify_call_expr): Likewise.
441 * predict.c (expr_expected_value): Likewise.
442 (strip_builtin_expect): Likewise.
443 * tree-inline.c (estimate_num_insns_1): Likewise.
444 * tree-ssa-loop-im.c (stmt_cost): Likewise
445 * fold-const.c (fold): Test for BUILT_IN_NORMAL.
446 (tree_expr_nonnegative_p): Likewise.
448 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
450 * config/sparc/sparc.c (load_pic_register): Emit the appropriate
451 variant of the load_pcrel_sym pattern.
452 * config/sparc/sparc.md (P macro): Move to the top.
453 (load_pcrel_sym): Macroize using P.
454 (save_register_window): Likewise.
456 2005-01-18 Aldy Hernandez <aldyh@redhat.com>
458 * config/rs6000/rs6000.md ("sunordered"): Disable for e500.
461 2005-01-18 Andrew Pinski <pinskia@physics.uc.edu>
464 * c-typeck.c (build_asm_expr): Strip nops off
465 input memory operands.
467 2005-01-18 David Edelsohn <edelsohn@gnu.org>
469 * config/rs6000/aix43.h (CPLUSPLUS_CPP_SPEC): Delete
470 _XOPEN_SOURCE, _XOPEN_SOURCE_EXTENDED, _LARGE_FILE_API.
471 * config/rs6000/aix51.h (CPLUSPLUS_CPP_SPEC): Same.
472 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Same.
474 2005-01-18 Joel Sherrill <joel.sherrill@oarcorp.com>
477 * config/i386/i386.c (override_options): If the 80387 is disabled,
478 then do not return FP values using FP registers.
480 2005-01-18 Kazu Hirata <kazu@cs.umass.edu>
482 * basic-block.h, c-common.c, c-cppbuiltin.c, c-lang.c,
483 c-tree.h, cfgbuild.c, cgraph.c, cgraph.h, collect2.c,
484 combine.c, config.gcc, coverage.h, cse.c, cselib.c,
485 defaults.h, df.c, dwarf2asm.c, dwarf2out.c, explow.c, expr.c,
486 flow.c, fold-const.c, gcse.c, ggc-page.c, gimple-low.c,
487 gimplify.c, ifcvt.c, langhooks-def.h, lcm.c, optabs.h,
488 output.h, postreload-gcse.c, postreload.c, recog.c,
489 resource.c, rtl.def, rtlanal.c, sched-deps.c, sched-rgn.c,
490 targhooks.h, toplev.c, tree-data-ref.c, tree-eh.c,
491 tree-flow.h, tree-loop-linear.c, tree-mudflap.h, tree-nrv.c,
492 tree-optimize.c, tree-outof-ssa.c, tree-pass.h,
493 tree-scalar-evolution.c, tree-ssa-copy.c, tree-ssa-dce.c,
494 tree-ssa-dse.c, tree-ssa-loop-ivopts.c, tree-ssa-operands.c,
495 tree-ssa-pre.c, tree-ssa.c, tree-vectorizer.c, tree.def,
496 unwind-dw2-fde-darwin.c, var-tracking.c: Update copyright.
498 * config/darwin.c, config/darwin.h, config/freebsd-spec.h,
499 config/alpha/alpha.c, config/arm/vfp.md,
500 config/i386/emmintrin.h, config/i386/i386-modes.def,
501 config/i386/i386-protos.h, config/i386/i386.h,
502 config/i386/pmmintrin.h, config/i386/ppro.md,
503 config/i386/predicates.md, config/i386/xmmintrin.h,
504 config/i860/i860.c, config/ia64/ia64-modes.def,
505 config/ia64/ia64-protos.h, config/ia64/ia64.h,
506 config/ia64/ia64.md, config/ia64/itanium1.md,
507 config/ia64/itanium2.md, config/ia64/predicates.md,
508 config/iq2000/iq2000.h, config/mips/linux64.h,
509 config/rs6000/aix41.h, config/rs6000/aix43.h,
510 config/rs6000/aix52.h, config/rs6000/darwin-fallback.c,
511 config/rs6000/rs6000.c, config/rs6000/rs6000.h,
512 config/rs6000/spe.md, config/sh/sh.md, config/sparc/linux.h,
513 config/sparc/linux64.h, config/sparc/litecoff.h,
514 config/sparc/sp64-elf.h, config/sparc/sparc.c,
515 config/sparc/sparc.h, config/sparc/sparc.md,
516 config/sparc/sysv4-only.h: Update copyright.
518 2005-01-18 Richard Henderson <rth@redhat.com>
521 * config/i386/i386.c (ix86_expand_fp_movcc): Fail for LTGT and UNEQ.
523 2005-01-18 Marc Espie <espie@openbsd.org>
524 * config/i386/openbsdelf.h: Typo.
526 2005-01-18 Richard Henderson <rth@redhat.com>
528 * config/i386/i386.c (ix86_expand_fp_movcc): Remove TARGET_IEEE_FP
529 special case for sse.
530 * config/i386/predicates.md (sse_comparison_operator): Likewise.
532 2005-01-18 Dorit Naishlos <dorit@il.ibm.com>
534 * tree-inline.c (estimate_num_insns_1): Added cases for
535 ALIGN_INDIRECT_REF, MISALIGNED_INDIRECT_REF, and REALIGN_LOAD_EXPR.
537 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
539 PR rtl-optimization/19296
540 * combine.c (simplify_comparison): Rewrite the condition under
541 which a non-paradoxical SUBREG of a PLUS can be lifted when
542 compared against a constant.
544 2005-01-18 Andi Kleen <ak@muc.de>
546 * c-typeck.c: (convert_for_assignment): Check warn_pointer_sign.
547 * c.opt (-Wpointer-sign): Add.
548 * doc/invoke.texi: (-Wpointer-sign): Add.
550 2005-01-18 Uros Bizjak <uros@kss-loka.si>
552 * config/i386/i386.c (override_options): Revert 2004-11-24 change.
553 * config/i386/i386.md (fmodsf3, dremsf3, *sinsf2, *cossf2, sincossf3,
554 *tansf3_1, tansf2, atan2sf3_1, atan2sf3, atansf2, asinsf2, acossf2,
555 logsf2, log10sf2, log2sf2, log1psf2, logbsf2, ilogbsf2, expsf2,
556 exp10sf2, exp2sf2, expm1sf2, rintsf2, floorsf2, ceilsf2, btruncsf2,
557 nearbyintsf2): Disable for TARGET_SSE_MATH. Leave patterns enabled
558 for TARGET_MIX_SSE_I387.
559 (fmoddf3, dremdf3, *sindf2, *sinextendsfdf2, *cosdf2, *cosextendsfdf2,
560 sincosdf3, *sincosextendsfdf3, *tandf3_1, tandf2, atan2df3_1,
561 atan2df3, atandf2, asindf2, acosdf2, logdf2, log10df2, log2df2,
562 log1pdf2, logbdf2, expdf2, exp10df2, exp2df2, expm1df2, rintdf2,
563 floordf2, ceildf2, btruncdf2, nearbyintdf2): Disable for
564 (TARGET_SSE2 && TARGET_SSE_MATH). Leave patterns enabled for
566 (atan2sf3, atan2df3, atan2xf3): Remove register constraints
569 2005-01-18 Uros Bizjak <uros@kss-loka.si>
572 * config/i386/mmx.md (*movv2sf_internal_rex64, *movv2sf_internal):
573 Add movaps alternative for xmm reg->reg move.
575 2005-01-17 Jeff Law <law@redhat.com>
577 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Clear SSA_NAME_VALUE
578 at the end of the main DOM loop rather than just before DOM exits.
580 2005-01-17 Diego Novillo <dnovillo@redhat.com>
582 PR tree-optimization/19121
583 * tree-ssa-alias.c (compute_flow_sensitive_aliasing): When
584 adding aliases to a name tag, also add them to the pointer's
586 * tree-ssa-copy.c (merge_alias_info): Do not merge flow
587 sensitive alias info at all. Only check that the two pointers
588 have compatible pointed-to sets.
589 * tree-ssa.c (verify_name_tags): Verify that the alias set of
590 a pointer's type tag is a superset of the alias set of the
593 2005-01-17 James E Wilson <wilson@specifixinc.com>
596 * config/ia64/ia64.md (movxf): Handle general register source. Adjust
597 comment to document why.
599 2005-01-17 Richard Henderson <rth@redhat.com>
601 * config/i386/sse.md (smaxv4sf3_finite, sse_vmsmaxv4sf3_finite,
602 sminv4sf3_finite, sse_vmsminv4sf3_finite, smaxv2df3_finite,
603 sse2_vmsmaxv2df3_finite, sminv2df3_finite,
604 sse2_vmsminv2df3_finite): New.
605 (smaxv4sf3, sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, smaxv2df3,
606 sse2_vmsmaxv2df3, sminv2df3, sse2_vmsminv2df3): Remove commutative.
607 Force op1 into register.
609 2005-01-17 Kelley Cook <kcook@gcc.gnu.org>
611 * Makefile.in (STAGEMOVESTUFF): Stage all the stamp files to prevent
612 bubblestrap from always rebuilding build directory.
614 2005-01-17 Ulrich Weigand <uweigand@de.ibm.com>
616 * config/s390/s390.md ("*clc"): Remove incorrect '='.
617 ("*cmpmem_short"): Likewise.
619 2005-01-17 Steven Bosscher <stevenb@suse.de>
621 * basic-block.h: Document BB_* flags.
622 * regrename.c (copyprop_hardreg_forward): Don't use BB_VISITED,
623 use an sbitmap instead.
624 * sched-rgn.c (compute_trg_info): Likewise.
626 2005-01-17 Richard Sandiford <rsandifo@redhat.com>
628 * config.gcc (mips64*-*-linux*): Set the default abi to n32. Remove
629 redundant target_cpu_default and extra_parts lines. Use the default
631 * config/mips/linux64.h (DRIVER_DEFAULT_ABI_SELF_SPEC): Delete.
632 (SUBTARGET_EXTRA_SPECS): Delete.
633 (DRIVER_SELF_SPECS): Remove architecture lines. Set the ABI directly.
634 (SUBTARGET_ASM_SPEC): Remove -mabi=n32 and -mabi=64 mappings.
635 * config/mips/t-linux64 (MULTILIB_DIRNAMES): Use "32" for -mabi=32
636 and "n32" for -mabi=n32.
638 2005-01-17 Ian Lance Taylor <ian@airs.com>
641 * c-typeck.c (build_c_cast): Revert patch of 2003-02-16: don't
642 fold constant variables into initial values.
644 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
646 * tree-cfg.c (tree_can_merge_blocks_p): Reorder two checks.
648 * tree-cfg.c (tree_forwarder_block_p): Speed up by walking
649 through the statements backward.
651 2005-01-17 Ian Lance Taylor <ian@airs.com>
654 * tree-inline.c (expand_call_inline): Set TREE_NO_WARNING on
655 a variable set to the return value of the inlined function.
657 2005-01-17 Mark Dettinger <dettinge@de.ibm.com>
659 * rtlanal.c (rtx_cost): Assign cost of 0 to a SUBREG
660 when modes are tieable.
662 2005-01-17 Ranjit Mathew <rmathew@hotmail.com>
664 * gthr-posix.h (__gthread_active_p): Use pthread_cancel instead
665 of pthread_create to find out if threads are enabled.
666 * gthr-posix95.h (__gthread_active_p): Likewise.
668 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
670 * common.opt (-fnew-ra): Remove.
672 * toplev.h (flag_new_regalloc): Remove.
673 * Makefile.in (ra*.*): Don't mention.
674 * passes.c (rest_of_handle_new_regalloc): Remove.
675 (rest_of_handle_combine, rest_of_compilation): Always consider
676 flag_new_regalloc as false.
677 * doc/invoke.texi: Don't document -fnew-ra.
679 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
681 * bb-reorder.c (fix_edges_for_rarely_executed_code): Remove
682 last parameter to reg_scan.
683 * loop.c (loop_optimize): Likewise.
684 * passes.c (rest_of_handle_tracer, rest_of_handle_if_conversion,
685 rest_of_handle_web, rest_of_handle_cfg, rest_of_handle_jump_bypass,
686 rest_of_handle_life, rest_of_handle_cse, rest_of_handle_cse2,
687 rest_of_handle_gcse, rest_of_handle_loop_optimize,
688 rest_of_handle_loop2, rest_of_handle_jump2): Likewise.
689 * regclass.c (reg_scan): Likewise, for the declaration.
690 * rtl.h (reg_scan): Likewise, for the prototype.
692 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
694 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Don't call
697 2005-01-16 Daniel Berlin <dberlin@dberlin.org>
699 * tree-ssa-pre.c (add_to_sets): s1 may be NULL.
700 (compute_avail): Uses don't go in tmp_gen.
702 2005-01-16 Steven Bosscher <stevenb@suse.de>
704 * ggc-page.c (ggc_alloc_stat): Use __builtin_ctzl instead of a
705 loop to look for a free slot in a page entry.
707 2005-01-16 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
710 * defaults.h (TARGET_DEFERRED_OUTPUT_DEFS): Provide default.
711 * toplev.c (compile_file): Call process_pending_assemble_output_defs
712 just before targetm.asm_out.file_end.
713 * tree.h (process_pending_assemble_output_defs): Declare.
714 * varasm.c (assemble_output_def, process_pending_assemble_output_defs):
716 (assemble_alias): Defer generation of assembly code for defines when
717 TARGET_DEFERRED_OUTPUT_DEFS is true.
718 * config/rs6000/aix41.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
719 * config/rs6000/aix43.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
720 * doc/tm.texi (TARGET_DEFERRED_OUTPUT_DEFS): document.
722 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
725 * pa.c (pa_scalar_mode_supported_p): New function.
726 (TARGET_SCALAR_MODE_SUPPORTED_P): Define.
727 * pa.h (MIN_UNITS_PER_WORD): Add comment.
729 * pa-protos.h (prefetch_operand): Delete.
730 (prefetch_cc_operand, prefetch_nocc_operand): New declations.
731 * pa.c (prefetch_operand): Delete.
732 (prefetch_cc_operand, prefetch_nocc_operand): New functions.
733 * pa.h (EXTRA_CONSTRAINT): Add `W' constraint.
734 (PREDICATE_CODES): Delete prefetch_operand. Add prefetch_cc_operand
735 and prefetch_nocc_operand.
736 * pa.md (prefetch): Rework to avoid reload problems handling short
737 displacements when a cache control completer needs to be provided.
738 (prefetch_32, prefetch_64): Delete.
739 (prefetch_cc, prefetch_nocc): New patterns.
741 2005-01-15 David Edelsohn <edelsohn@gnu.org>
743 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Revert previous change.
745 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
748 * pa.c (pa_asm_output_aligned_common): Warn if specified alignment
749 exceeds maximum alignment for global common data.
750 * pa64-hpux.h (MAX_OFILE_ALIGNMENT): Define to 32768.
751 * som.h (MAX_OFILE_ALIGNMENT): Likewise.
753 2004-01-15 Roger Sayle <roger@eyesopen.com>
755 * tree-ssa-dom.c (extract_range_from_cond): Correct condition.
757 2004-01-15 Roger Sayle <roger@eyesopen.com>
759 * harg-reg-set.h (reg_class_names): Prototype global array.
760 * regclass.c (reg_class_names): Declare here and initialize to
762 (dump_regclass): Remove local declaration of reg_class_names.
763 (regclass): Likewise.
764 * cfg.c (dump_flow_info): Likewise.
765 * ra-debug.c (reg_class_names): Likewise.
766 * regrename.c (reg_class_names): Likewise.
767 * reload.c (reg_class_names): Likewise.
768 * reload1.c (spill_failure): Likewise.
769 * config/m68hc11/m68hc11.c (reg_class_names): Likewise.
771 2005-01-15 Ulrich Weigand <uweigand@de.ibm.com>
773 * config/s390/s390.md ("reload_outti"): Remove predicate for
774 output operand. Abort if operand is not a MEM.
775 ("reload_outdi", "reload_outdf"): Likewise.
777 2005-01-15 Marc Espie <espie@openbsd.org>
779 * config.gcc (*-*-openbsd*): Set HAS_LIBC_R for OpenBSD <= 3.2.
780 (i[34567]86-*-openbsd*): Switch to ELF for OpenBSD >= 3.4.
781 * config/openbsd.h: Fix C++ includes for native configurations.
782 Add proper OS_CPP_BUILTINS. Fix libspec for recent OpenBSD.
783 Add trampoline support.
784 * config/i386/openbsdelf.h: New.
786 2005-01-15 Marc Espie <espie@openbsd.org>
788 * collect2.c (main): Explicitly parse -dynamic-linker option.
790 2005-01-15 Jakub Jelinek <jakub@redhat.com>
792 PR tree-optimization/19060
793 * tree-ssa-dom.c (extract_range_from_cond) <case LT_EXPR, GT_EXPR>:
794 Return 0 if op1 <= TYPE_MIN_VALUE () resp. op1 >= TYPE_MAX_VALUE ().
795 (simplify_cond_and_lookup_avail_expr): Add assert for dummy == 0
796 and handle extract_range_from_cond returning false.
797 * fold-const.c (fold): Optimize comparisons with min/max even for
798 width > HOST_BITS_PER_WIDE_INT.
800 2005-01-15 Ralf Corsepius <ralf.corsepius@rtems.org>
802 * config/mips/rtems.h (MIPS_DEFAULT_GVALUE): Set to 0.
803 * config/mips/t-rtems (MULTILIBS_DIRNAMES,MULTILIB_OPTIONS):
804 Remove little endian multilib variants.
805 Add mips32 multilib variant.
807 2005-01-15 Kazu Hirata <kazu@cs.umass.edu>
809 * explow.c (copy_all_reg, stabilize): Remove.
810 * expr.h: Remove the corresponding prototypes.
812 * rtlanal.c (regs_set_between_p): Remove.
813 * rtl.h: Remove the corresponding prototype.
815 2005-01-14 David Edelsohn <edelsohn@gnu.org>
817 * config/rs6000/rs6000.c (rs6k_nonimmediate_operand): Rename
818 to rs6000_nonimmediate_operand.
819 * config/rs6000/rs6000.h (PREDICATE_CODES): Rename
820 rs6k_nonimmediate_operand to rs6000_nonimmediate_operand.
821 * config/rs6000/rs6000.md (movsi_internal1): Rename predicate to
822 rs6000_nonimmediate_operand.
823 * config/rs6000/spe.md (movdf_e500_double): Rename predicate to
824 rs6000_nonimmediate_operand.
826 2005-01-14 Andrew Pinski <pinskia@physics.uc.edu>
828 * tree-ssa-alias.c (pass_may_alias): Add TODO_verify_stmts.
830 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
832 * config/rs6000/rs6000.h (CLASS_MAX_NREGS): DF goes in 1 register
834 (CANNOT_CHANGE_MODE_CLASS): Restrict DI mode changes on e500v2.
835 (PREDICATE_CODES): Add rs6k_nonimmediate_operand.
837 * config/rs6000/rs6000.c (invalid_e500_subreg): New.
838 (rs6k_nonimmediate_operand): New.
839 (rs6000_legitimate_offset_address_p): Handle DI modes on e500v2
841 (legitimate_lo_sum_address_p): Same.
842 (rs6000_legitimize_address): Same.
843 (rs6000_legitimize_reload_address): Same.
844 (rs6000_legitimate_address): Same.
845 (spe_build_register_parallel): Pass DF and DC modes in a DI
848 * config/rs6000/rs6000.md ("*movsi_internal1"): Change predicate
849 to rs6k_nonimmediate_operand.
851 * config/rs6000/spe.md ("*frob_df_di"): New.
852 ("*frob_di_df"): New.
853 ("*frob_di_df_2"): New.
854 ("*mov_sidf_e500_subreg0"): New.
855 ("*mov_sidf_e500_subreg4"): New.
856 ("*movdf_e500_double"): Change predicate to
857 rs6k_nonimmediate_operand.
859 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
861 * postreload.c (move2add_note_store): Only call
862 trunc_int_for_mode on scalar integers.
864 2005-01-14 Steven Bosscher <stevenb@suse.de>
866 * tree-ssa-dce.c (visited_control_parents): New sbitmap to
867 replace BB_VISITED uses.
868 (find_obviously_necessary_stmts): Don't clear BB_VISITED.
869 (propagate_necessity): Check the bitmap instead of BB_VISITED.
870 (tree_dce_done): Free visited_control_parents.
871 (perform_tree_ssa_dce): Allocate and clear it.
872 * tree-ssa-pre.c (compute_antic_aux): Make non-recursive.
873 (compute_antic): Iterate from here using a DFS. Use an sbitmap
874 instead of BB_VISITED.
876 2005-01-14 Kazu Hirata <kazu@cs.umass.edu>
878 * c-tree.h, coverage.h, langhooks-def.h, optabs.h, output.h,
879 rtl.h, targhooks.h, tree-flow.h, tree-mudflap.h, tree.h:
880 Remove unused prototypes.
882 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
884 * config/rs6000/rs6000.c (rs6000_stack_info): Fix caching of SPE
885 64-bit register usage.
886 (spe_func_has_64bit_regs_p): Add FIXME note.
888 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
890 * config/rs6000/rs6000.c (rs6000_generate_compare): Replace
891 flag_finite_math_only with flag_unsafe_math_optimizations.
893 2005-01-14 J. D. Johnston <jjohnst@us.ibm.com>
895 * config/s390/tpf-unwind.h (s390_fallback_frame_state): Correct
897 (__tpf_eh_return): Copy TPF private stack area from the module
898 boundary stack frame. Add check for when module addresses are
899 equal, but no stub address is found.
901 2005-01-14 Richard Earnshaw <rearnsha@arm.com>
904 * arm.h (struct machine_function): Add call_via field.
905 (thumb_call_via_label): Declare.
906 * arm.c (thumb_call_via_label): New variable.
907 (thumb_call_reg_needed): New variable.
908 (arm_output_function_epilogue): For Thumb code, output any per-function
909 call-indirect trampolines.
910 (thumb_call_via_reg): New function.
911 (arm_file_end): New function.
912 (TARGET_ASM_FILE_END): Call arm_file_end.
913 (aof_file_end): Likewise.
914 * arm-protos.h (thumb_call_via_reg): Declare.
915 * arm.md (call_reg_thumb, call_value_reg_thumb): Call
916 thumb_call_via_reg in normal case.
918 2005-01-14 Jakub Jelinek <jakub@redhat.com>
921 PR rtl-optimization/19348
922 * recog.c (peephole2_optimize): Do global life update if some peephole
923 decides it doesn't need at least one of its inputs and that change
924 influences liveness at the start of the basic block.
926 * basic-block.h (EXECUTE_IF_AND_COMPL_IN_REG_SET): Needs 2 REGSET
927 arguments instead of 1.
929 2005-01-14 Eric Botcazou <ebotcazou@libertysurf.fr>
932 * varasm.c (initializer_constant_valid_p) <ADDR_EXPR>: Return
933 zero for nested functions needing a static chain or functions
934 with a non-constant address.
936 2005-01-13 Roger Sayle <roger@eyesopen.com>
938 * simplify-rtx.c (simplify_binary_operation) <AND>: Optimize
939 (and (sign_extend X) C) into (zero_extend (and X C)).
941 2005-01-13 David O'Brien <obrien@FreeBSD.org>
943 * config/freebsd-spec.h: Make KSE pthread lib logic the default.
945 2005-01-13 Richard Henderson <rth@redhat.com>
950 * config/i386/i386.md (cmpdf, cmpsf, bunordered, bordered, buneq,
951 bunge, bungt, bunle, bunlt, bltgt): Enable for TARGET_SSE_MATH,
953 (cmpfp_i_387): Rename from cmpfp_i. Move after sse patterns.
954 (cmpfp_i_mixed): Rename from cmpfp_i_sse; use for TARGET_MIX_SSE_I387.
955 (cmpfp_i_sse): Rename from cmpfp_i_sse_only; use for TARGET_SSE_MATH.
956 (cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387): Similarly.
957 (fp_jcc_1_mixed, fp_jcc_1_sse, fp_jcc_1_387): Similarly.
958 (fp_jcc_2_mixed, fp_jcc_2_sse, fp_jcc_2_387): Similarly.
959 (fp_jcc_3_387, fp_jcc_4_387, fp_jcc_5_387, fp_jcc_6_387,
960 fp_jcc_7_387, fp_jcc_8_387): Rename from fp_jcc_N.
961 (movdicc_c_rex64): Rename with '*'.
962 (movsfcc, movdfcc): Add checks for 387 and sse math to condition.
963 (movsfcc_1_sse_min, movsfcc_1_sse_max, movsfcc_1_sse): New.
964 (movsfcc_1_387): Rename from movsfcc_1.
965 (movdfcc_1_sse_min, movdfcc_1_sse_max, movdfcc_1_sse): New.
966 (movdfcc_1, movdfcc_1_rex64): Add check for 387.
967 (sminsf3, smaxsf3, smindf3, smaxdf3): New.
968 (minsf3, minsf, minsf_nonieee, minsf_sse, mindf3, mindf,
969 mindf_nonieee, mindf_sse, maxsf3, maxsf, maxsf_nonieee, maxsf_sse,
970 maxdf3, maxdf, maxdf_nonieee, maxdf_sse, sse_movsfcc, sse_movsfcc_eq,
971 sse_movdfcc, sse_movdfcc_eq, sse_movsfcc_const0_1,
972 sse_movsfcc_const0_2, sse_movsfcc_const0_3, sse_movsfcc_const0_4,
973 sse_movdfcc_const0_1, sse_movdfcc_const0_2, sse_movdfcc_const0_3,
974 sse_movdfcc_const0_4): Remove.
975 * config/i386/i386.c (ix86_expand_fp_movcc): For TARGET_SSE_MATH,
976 recognize min/max early. Update for changed sse cmove patterns.
977 (ix86_split_sse_movcc): New.
978 * config/i386/i386-protos.h: Update.
980 2005-01-13 Steven Bosscher <stevenb@suse.de>
982 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
983 (fix_stmt_v_may_defs): Likewise.
985 2005-01-13 Richard Henderson <rth@redhat.com>
987 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
988 for converting the mask.
990 2005-01-13 David Edelsohn <edelsohn@gnu.org>
992 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
995 2005-01-13 Richard Henderson <rth@redhat.com>
997 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
998 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
999 IX86_BUILTIN_MOVDQ2Q): Remove.
1000 (IX86_BUILTIN_VEC_EXT_V4SI): New.
1001 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
1002 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
1004 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
1005 (sse2_pextrw): Fix immediate constraint.
1006 (sse2_loadq, sse2_loadq_rex64): Remove.
1007 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
1008 Use __builtin_ia32_vec_ext_<size>.
1009 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
1011 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
1013 * function.c (assign_parm_setup_block): Look inside original
1014 entry_parm when inspecting PARALLEL.
1016 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
1017 Joel Sherrill <joel@oarcorp.com>
1020 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
1021 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
1022 rtems_gxx_recursive_mutex_init.
1023 (__gthread_recursive_mutex_lock): New function.
1024 (__gthread_recursive_mutex_trylock): Likewise.
1025 (__gthread_recursive_mutex_unlock): Likewise.
1027 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
1029 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
1031 2005-01-13 Jan Beulich <jbeulich@novell.com>
1033 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
1034 (ashrti3_internal): Indicate output is early clobber. Generate result
1035 into output rather than first input. Use move for low word of output
1036 if shift count is exactly 64.
1037 (lshrti3_internal): Likewise.
1039 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
1043 * reload1.c (delete_output_reload): Don't delete an output reload
1044 if the pseudo lives longer than a single basic block. Adjust and
1045 improve wording of comment.
1047 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
1050 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
1051 of TYPE_SIZE where we mean the number of bytes.
1053 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
1055 * function.c (assign_parm_setup_block): Relax condition on
1056 multi-register optimization.
1058 2005-01-12 Nick Clifton <nickc@redhat.com>
1060 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
1061 an "in_delay_slot" attribute of "no" to prevent them being used in
1062 delay slots. This is forbidden because they might generate
1065 2005-01-12 Alan Modra <amodra@bigpond.net.au>
1068 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
1071 2005-01-12 Nick Clifton <nickc@redhat.com>
1073 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
1074 passed on to GAS) but do not define, as GAS no longer supports or
1075 needs the -m2000 option.
1077 2005-01-11 Roger Sayle <roger@eyesopen.com>
1079 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
1080 type as args of unprototyped builtins aren't automatically promoted.
1081 (fold_builtin_abs): Likewise.
1083 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1085 * Makefile.in: Set a `build-warn' variable.
1087 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
1089 PR tree-optimization/17949
1090 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
1091 (find_interesting_uses_address): Use it.
1093 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
1095 * regrename.c (kill_value): Handle subreg's that won't simplify.
1097 2005-01-11 Richard Henderson <rth@redhat.com>
1100 * config/i386/i386.h (enum ix86_builtins): Move ...
1101 * config/i386/i386.c: ... here.
1102 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
1103 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
1104 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
1105 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
1106 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
1107 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
1108 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
1109 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
1110 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
1111 IX86_BUILTIN_LOADDDUP): Remove.
1112 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
1113 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
1114 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
1115 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
1116 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
1117 (ix86_init_builtins): Make static.
1118 (ix86_init_mmx_sse_builtins): Update for changed builtins.
1119 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
1120 if all the modes match. Otherwise, fake it.
1121 (get_element_number, ix86_expand_vec_init_builtin,
1122 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
1123 (ix86_expand_builtin): Make static. Update for changed builtins.
1124 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
1125 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
1126 (ix86_expand_vector_init_duplicate): New.
1127 (ix86_expand_vector_init_low_nonzero): New.
1128 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
1129 Split out from ix86_expand_vector_init; handle integer modes.
1130 (ix86_expand_vector_init): Use them.
1131 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
1132 * config/i386/i386-protos.h: Update.
1133 * config/i386/predicates.md (reg_or_0_operand): New.
1134 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
1135 (movv2sf_internal): Likewise. And a splitter to match them all.
1136 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
1137 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
1138 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
1139 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
1140 (mmx_pinsrw): Fix operand ordering.
1141 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
1142 rather than sse_loadss expander.
1143 (movv2df splitter): Similarly.
1144 (sse_loadss, sse_loadlss): Remove.
1145 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
1146 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
1147 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
1148 (sse3_movddup): Rename with '*'.
1149 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
1150 (sse2_loadsd): Remove.
1151 (vec_dupv2df_sse3): Rename from sse3_loadddup.
1152 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
1153 (sse2_pinsrw): Fix argument ordering.
1154 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
1155 (sse2_stored): Remove 'r' destination.
1156 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
1157 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
1158 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
1159 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
1160 vec_extractv16qi, vec_initv16qi): New.
1162 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
1163 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
1164 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
1166 (_mm_load_pd, _mm_store_pd): Use plain dereference.
1167 (_mm_load_si128, _mm_store_si128): Likewise.
1168 (_mm_load1_pd): Use _mm_set1_pd.
1169 (_mm_load_sd): Use _mm_set_sd.
1170 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
1171 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
1172 (_mm_set_epi64): Use _mm_set_epi64x.
1173 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
1174 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
1175 _mm_setr_epi8): Use _mm_set_foo form.
1176 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
1177 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
1178 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
1179 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
1180 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
1181 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
1182 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
1183 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
1184 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
1185 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
1186 (_mm_movedup_pd): Use _mm_shuffle_pd.
1187 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
1188 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
1189 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
1190 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
1192 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
1193 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
1194 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
1195 (_mm_store_ps): Use raw dereference.
1196 (_mm_store1_ps): Use _mm_storeu_ps.
1197 (_mm_storer_ps): Use _mm_store_ps.
1198 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
1199 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
1201 2005-01-11 Stan Shebs <shebs@apple.com>
1203 * config/rs6000/rs6000.c (machopic_output_stub): Issue
1204 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
1206 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
1209 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
1210 Skip all DECLs except for FIELD_DECLs.
1212 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
1214 * config/s390/s390.c (override_options): Return error if
1215 -mbackchain, -mpacked-stack and -mhard-float are used together.
1216 (s390_va_start): Remove the backchain && packed-stack special case.
1217 (s390_gimplify_va_arg): Likewise.
1218 * doc/invoke.texi: Remove the ABI incompatibility note.
1220 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
1222 * config/s390/s390.c (struct s390_frame_layout): Remove
1224 (s390_frame_info, s390_emit_prologue): Replace occurrences of
1225 save_backchain_p with TARGET_BACKCHAIN.
1227 2005-01-11 Alan Modra <amodra@bigpond.net.au>
1230 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
1231 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
1232 * function.c (locate_and_pad_parm): Set new field.
1233 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
1234 Tweak where_pad test to include "none". Always set mem align for
1236 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
1237 not sufficient for type.
1238 (assign_parm_setup_block): If stack_parm is zero on entry, always
1239 make a new stack local. Block move old stack parm if necessary
1240 to new aligned stack local.
1241 (assign_parm_setup_stack): Use a block move to handle
1242 potentially misaligned entry_parm.
1243 (assign_parms_unsplit_complex): Specify required alignment when
1244 creating stack local.
1245 * calls.c (compute_argument_addresses): Override alignment of stack
1246 arg calculated from its type with the alignment given by
1247 FUNCTION_ARG_BOUNDARY.
1248 (store_one_arg): Likewise.
1250 2005-01-11 Jan Beulich <jbeulich@novell.com>
1252 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
1253 Change respective itanium_class attribute to ialu.
1254 (shladdp4_internal): New.
1255 * config/ia64/predicates.md (shladd_log2_operand): New.
1257 2005-01-11 Richard Henderson <rth@redhat.com>
1259 * expr.c (store_constructor): Use rtvec_alloc instead of
1260 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
1262 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
1264 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
1266 2005-01-11 Alan Modra <amodra@bigpond.net.au>
1268 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
1269 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
1270 (TARGET_RWRELOC): Define.
1271 (ia64_rwreloc_section_type_flags): Delete.
1272 (ia64_section_type_flags): New function.
1273 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
1274 (TARGET_RWRELOC): Define.
1276 2005-01-10 David Mosberger <davidm@hpl.hp.com>
1279 * config/ia64/ia64.c (process_set): For alloc insn, only call
1280 process_epilogue is !frame_pointer_needed.
1282 2005-01-10 Roger Sayle <roger@eyesopen.com>
1285 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
1286 unary operator and can't be treated as a binary/comparison operator.
1288 2005-01-10 Richard Henderson <rth@redhat.com>
1290 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
1291 (ix86_return_in_memory): Likewise.
1292 (function_arg_advance): Likewise. Mirror structure in function_arg
1293 for choosing register to advance.
1295 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
1297 * tree-vectorizer.c, tree.def: Fix comment typos.
1299 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
1301 * config/s390/s390.c (struct processor_costs): 4 new fields:
1302 ddbr, ddr, debr, der.
1303 (s390_rtx_costs): More precise handling of divide instructions.
1305 2005-01-10 David Edelsohn <edelsohn@gnu.org>
1307 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
1309 2005-01-10 Jan Beulich <jbeulich@novell.com>
1311 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
1312 names resulting from -ffunction-sections/-fdata-sections and linkonce
1314 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
1315 based on the section name. Rearrange the section name comparison logic
1316 slightly so that each section name is compared against at most once.
1318 2005-01-10 Ben Elliston <bje@au.ibm.com>
1320 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
1321 targets for which -fPIC reduces limitations on the GOT size.
1323 2005-01-09 Falk Hueffner <falk@debian.org>
1325 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
1328 2005-01-09 David Edelsohn <edelsohn@gnu.org>
1331 * collect2.c (main): Set aixrtl_flag for -brtl option.
1332 (resolve_lib_name): Search for .so file extension before .a
1335 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
1337 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
1338 (vect_analyze_data_refs_alignment): Add dump prints.
1340 2005-01-09 Ira Rosen <irar@il.ibm.com>
1342 * tree-vectorizer.c (vect_analyze_offset_expr): Use
1343 expr_invariant_in_loop_p.
1344 Initialize outputs first thing in the function.
1345 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
1346 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
1347 (vect_analyze_pointer_ref_access): Check that the initial condition of
1348 the access function is loop invariant.
1350 2005-01-09 Richard Henderson <rth@redhat.com>
1352 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
1353 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
1354 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
1355 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
1356 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
1357 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
1358 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
1359 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
1360 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
1361 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
1362 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
1363 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
1364 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
1365 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
1366 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
1367 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
1368 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
1369 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
1370 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
1371 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
1372 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
1373 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
1374 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
1376 (mmx_clrdi, pavgusb): Remove.
1377 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
1378 with leading sse_ prefix.
1379 * config/i386/sse.md: Receive them.
1380 * config/i386/mmx.md: New file.
1381 (MMXMODE12, MMXMODE24, mmxvecsize): New.
1382 (subrv2sf3): Turn into expander for normal subtraction.
1383 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
1384 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
1385 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
1386 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
1387 commutative; use ix86_binary_operator_ok.
1388 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
1389 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
1390 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
1391 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
1392 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
1393 Macroize from existing patterns; use ix86_binary_operator_ok.
1394 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
1395 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
1396 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
1397 with vec_select+vec_concat.
1398 (mmx_pshufw, mmx_pshufw_1): Likewise.
1399 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
1400 (mmx_uavgv4hi3): Model correctly.
1401 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
1402 _mm_xor_si64): Remove casts.
1404 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
1406 PR tree-optimization/19224
1407 * tree-scalar-evolution.c (get_instantiated_value,
1408 set_instantiated_value): New functions.
1409 (instantiate_parameters_1): Cache the results.
1410 (instantiate_parameters, resolve_mixers): Initialize and free
1413 2005-01-08 David Edelsohn <edelsohn@gnu.org>
1415 * config/i386/i386.md (addhi_4): Correct reference in comment.
1418 2005-01-08 Richard Henderson <rth@redhat.com>
1420 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
1421 (_mm_cvtsi128_si64x): Likewise.
1422 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
1423 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
1424 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
1425 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
1426 * config/i386/i386-protos.h: Update.
1427 * config/i386/i386.c (print_operand): Add 'H'.
1428 (ix86_fixup_binary_operands): Split out from ...
1429 (ix86_expand_binary_operator): ... here.
1430 (ix86_fixup_binary_operands_no_copy): New.
1431 (ix86_expand_fp_absneg_operator): Handle vector mode results.
1432 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
1433 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
1434 (safe_vector_operand): Use CONST0_RTX.
1435 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
1436 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
1437 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
1438 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
1440 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
1441 (IX86_BUILTIN_CMPNGESS): New.
1442 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
1443 (attr type): Add sselog1.
1444 (attr unit, attr memory): Handle it.
1445 (movti, movti_internal, movti_rex64): Move near other integer moves.
1446 (movtf, movtf_internal): Move near other fp moves.
1447 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
1448 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
1449 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
1450 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
1451 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
1452 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
1453 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
1454 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
1455 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
1456 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
1457 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
1458 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
1459 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
1460 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
1461 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
1462 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
1463 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
1464 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
1465 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
1466 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
1467 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
1468 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
1469 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
1470 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
1471 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
1472 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
1473 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
1474 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
1475 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
1476 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
1477 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
1478 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
1479 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
1480 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
1481 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
1482 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
1483 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
1484 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
1485 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
1486 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
1487 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
1488 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
1489 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
1490 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
1492 (SSEPUSH, push<SSEPUSH>): Remove.
1493 (MMXPUSH, push<MMXPUSH>): Remove.
1494 (sse_movaps, sse_movaps_1, sse_movups): Remove.
1495 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
1496 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
1497 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
1498 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
1499 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
1500 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
1501 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
1502 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
1503 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
1504 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
1505 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
1506 (ppro_sse_log_V4SF): Handle sselog1.
1507 * config/i386/predicates.md (const_0_to_1_operand): New.
1508 (const_0_to_255_mul_8_operand): New.
1509 (const_1_to_31_operand): Rename from const_int_1_31_operand.
1510 (const_2_to_3_operand, const_4_to_7_operand): New.
1511 * config/i386/sse.md: New file.
1512 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
1513 (sse_movups): Rename from sse_movups_1.
1514 (sse_loadlss): Rename from sse_loadss_1.
1515 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
1517 (negv4sf2): Use ix86_expand_fp_absneg_operator.
1518 (absv4sf2, negv2df, absv2df): New.
1519 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
1520 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
1521 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
1522 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
1523 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
1524 (sse3_addsubv4sf3): Model correctly.
1525 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
1526 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
1527 (sse_movhlps): Model with vec_select+vec_concat.
1528 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
1529 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
1530 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
1531 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
1532 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
1533 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
1534 sse2_pshufhw_1): Likewise.
1535 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
1536 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
1537 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
1538 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
1539 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
1540 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
1541 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
1542 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
1543 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
1544 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
1545 sminv8hi3): Mark commutative
1546 operands. Use ix86_binary_operator_ok.
1547 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
1548 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
1549 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
1550 sse2_punpcklqdq): Allow operand2 in memory.
1551 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
1552 sse2_movsd): Add memory alternatives.
1553 (sse_storelps): Turn expander into an insn; split after reload.
1554 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
1555 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
1557 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
1559 * configure.ac (DWARF-2 debug_line): Use objdump.
1560 * configure: Regenerate.
1562 2005-01-08 Jeff Law <law@redhat.com>
1563 Diego Novillo <dnovillo@redhat.com>
1565 PR tree-optimization/18241
1566 * tree-nrv.c (tree_nrv): Ignore volatile return values.
1567 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
1568 statements with volatile operands.
1569 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
1570 operands after marking a statement with has_volatile_ops.
1572 2005-01-08 Roger Sayle <roger@eyesopen.com>
1574 * tree.c (int_fits_type_p): Always honor integer constant
1575 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
1577 2005-01-08 Roger Sayle <roger@eyesopen.com>
1579 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
1580 the latent bug in rtl_delete_block has been resolved.
1582 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
1584 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
1585 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
1587 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
1589 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
1590 library as @multilib_dir@/$(SHLIB_SONAME).
1592 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
1594 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
1595 * configure: Regenerate.
1597 2005-01-07 Jakub Jelinek <jakub@redhat.com>
1599 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
1602 2005-01-07 David Edelsohn <edelsohn@gnu.org>
1605 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
1606 Convert non-word aligned offset address using ld/std into
1609 2005-01-07 Richard Henderson <rth@redhat.com>
1611 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
1613 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
1615 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
1617 2005-01-07 Jakub Jelinek <jakub@redhat.com>
1619 PR tree-optimization/19283
1620 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
1621 is not shorter than the original type.
1623 PR rtl-optimization/19012
1624 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
1626 PR rtl-optimization/18861
1627 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
1628 (FULL_STATE): Define.
1629 (mark_tablejump_edge): New function.
1630 (purge_dead_tablejump_edges): New function.
1631 (find_bb_boundaries): Use it.
1633 PR tree-optimization/18828
1634 * builtins.c (expand_builtin_next_arg): Remove argument and all
1635 the argument checking.
1636 (expand_builtin): Adjust caller.
1637 (expand_builtin_va_start): Likewise. Remove error for too many
1639 (fold_builtin_next_arg): Issue error for too many arguments.
1640 After checking arguments, replace them with magic arguments that
1641 prevent further checking of the args.
1643 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1645 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
1647 2005-01-06 Roger Sayle <roger@eyesopen.com>
1650 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
1651 and %{.S:...} (and their negative variants) to test whether the
1652 input file is assembler or pre-processed-assembler independent of
1653 the actual filename extension.
1655 2005-01-06 Roger Sayle <roger@eyesopen.com>
1657 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
1658 of sign or zero extended values.
1660 2005-01-06 Geoffrey Keating <geoffk@apple.com>
1662 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
1663 for IBM long double format correctly.
1665 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
1667 Fix PR tree-optimization/18792
1669 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
1670 to first_loop_depth, and use loop depth instead of loop number.
1671 (build_classic_dir_vector): Ditto.
1672 (compute_data_dependences_for_loop): Use depth, not loop number.
1673 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
1674 number. Pass in loops, instead of loop numbers.
1675 (gather_interchange_stats): Ditto.
1676 (linear_transform_loops): Ditto.
1678 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
1681 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
1682 biv_fits_mode_p, extension_within_bounds_p): New functions.
1683 (check_ext_dependent_givs): Use them.
1685 2005-01-06 Roger Sayle <roger@eyesopen.com>
1687 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
1688 more than one barrier, in which case we should delete them all.
1690 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
1692 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
1694 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
1696 Fix PR middle-end/19286
1698 * dwarf2out.c (gen_subprogram_die): If we've already tried to
1699 output this subprogram, simply ignore this attempt to do it again.
1700 (add_abstract_origin_attribute): Don't abort trying to add the abstract
1701 origin attribute if it's not possible.
1702 (gen_block_die): Don't ignore subblocks of "unused" blocks.
1703 (decls_for_scope): Ditto.
1704 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
1705 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
1707 2005-01-05 Richard Henderson <rth@redhat.com>
1710 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
1711 (bdesc_2arg): Use it.
1712 (ix86_expand_binop_builtin): Force operands into registers
1714 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
1715 ix86_expand_sse_compare, ix86_expand_sse_comi,
1716 ix86_expand_builtin): Likewise.
1718 2005-01-05 Richard Henderson <rth@redhat.com>
1720 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
1722 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
1723 (mulv4hi3): Set itanium_class mmmul.
1724 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
1725 (fpack): Rename from fpack_sfsf.
1727 2005-01-05 Richard Henderson <rth@redhat.com>
1730 * reload1.c (do_input_reload): Restrict the optimization deleteing
1731 a previous output reload to RELOAD_FOR_INPUT.
1733 2005-01-05 Steven Bosscher <stevenb@suse.de>
1735 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
1736 comment that we fall through after case.
1737 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
1738 (distribute_links): Likewise.
1739 * cse.c (cse_insn): Likewise.
1740 * cselib.c (cselib_invalidate_mem): Likewise.
1741 * df.c: Update comments at the top of the file.
1742 (read_modify_subreg_p): Update comments here too.
1743 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
1744 * flow.c (mark_set_1): Likewise.
1745 (mark_used_regs): Likewise.
1746 * gcse.c (mems_conflict_for_gcse_p): Likewise.
1747 (canon_list_insert): Likewise.
1748 (mark_set): Likewise.
1749 (try_replace_reg): Likewise.
1750 (store_killed_in_insn): Likewise.
1751 * loop.c (count_one_set): Likewise.
1752 (basic_induction_var): Likewise.
1753 * postreload-gcse.c (find_mem_conflicts): Likewise.
1754 * postreload.c (reload_combine_note_store): Likewise.
1755 (move2add_note_store): Likewise.
1756 * reload.c (find_equiv_reg): Likewise.
1757 (mark_referenced_resources): Likewise.
1758 * rtlanal.c (set_noop_p): Likewise.
1759 (note_stores): Likewise.
1760 (note_uses): Likewise.
1761 * sched-deps.c (sched_analyze_1): Likewise.
1762 * sched-rgn.c (check_live_1): Likewise.
1763 (update_live_1): Likewise.
1764 * config/i860/i860.c: Likewise.
1766 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
1768 (ZERO_EXTRACT): Mention that this one can be an lvalue.
1770 * doc/rtl.texi: Update documentation for bit-fields and SET.
1772 2005-01-05 Roger Sayle <roger@eyesopen.com>
1774 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
1776 2005-01-05 Jan Hubicka <jh@suse.cz>
1777 Richard Henderson <rth@redhat.com>
1780 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
1783 2005-01-05 Richard Henderson <rth@redhat.com>
1786 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
1787 (sse_shufps): Change operand 3 to const_int_operand.
1788 (sse2_storelps): Fix typo in template.
1789 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
1790 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
1791 (ix86_expand_builtin): Likewise.
1793 2005-01-05 Julian Brown <julian@codesourcery.com>
1795 * config/arm/arm.c (arm_return_in_memory): Treat complex types
1796 as aggregates for AAPCS ABIs.
1798 2005-01-05 Stan Shebs <shebs@apple.com>
1800 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
1801 Mach-O getters if ppc64.
1802 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
1804 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
1805 signed four-byte field for global code case.
1806 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
1808 2005-01-05 Roger Sayle <roger@eyesopen.com>
1811 * c-common.c: Include real.h.
1812 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
1813 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
1814 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
1815 * Makefile.in (c-common.o): Update dependencies.
1817 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1819 * c-parse.in (asm_string): Add trailing semicolon.
1821 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1823 * c-parse.in (asm_string): New. Don't allow wide strings in
1825 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
1826 asm_string instead of STRING.
1828 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1830 * c-typeck.c (constructor_no_implicit): Remove.
1831 (set_designator, process_init_element): Don't check
1832 constructor_no_implicit.
1834 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
1835 Kaz Kojima <kkojima@gcc.gnu.org>
1838 * lcm.c (create_pre_exit): New.
1839 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
1840 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
1842 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
1844 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
1845 -msoft-float in target_float_switch.
1846 * arm.c (arm_override_options): Fix processing of target_float_switch.
1848 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
1850 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
1853 2004-01-05 Uros Bizjak <uros@kss-loka.si>
1855 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
1856 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
1858 2005-01-04 Roger Sayle <roger@eyesopen.com>
1860 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
1861 if deleting the then-block allows the test-block to fallthru to the
1864 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
1867 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
1868 after the full declaration if the are in two different TUs.
1870 2005-01-04 Richard Henderson <rth@redhat.com>
1873 * tree-sra.c (generate_one_element_init): Just
1874 call gimplify_and_add.
1875 (generate_element_init): Record the
1876 new referenced variables and mark them for renaming
1877 and split out to ...
1878 (generate_element_init_1): This.
1879 (scalarize_init): Don't call push_gimplify_context/
1880 pop_gimplify_context.
1882 2005-01-04 Geoffrey Keating <geoffk@apple.com>
1884 * toplev.c (get_src_pwd): Handle failure of getpwd().
1886 2005-01-04 Roger Sayle <roger@eyesopen.com>
1888 * fold-const.c (fold_single_bit_test): Delete unreachable handling
1890 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
1891 all the cases handled by it are inverted by invert_truthvalue.
1893 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
1895 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
1896 (save_register_windowsi): Likewise.
1898 2005-01-04 Richard Henderson <rth@redhat.com>
1900 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
1902 2005-01-04 Uros Bizjak <uros@kss-loka.si>
1905 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
1906 any simplifications of vector mode comparison operators.
1907 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
1909 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
1910 Devang Patel <dpatel@apple.com>
1912 PR tree-optimization/18308
1913 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
1914 the operands before creating a new expression.
1915 * dojump.c (do_jump): Make drop_through_label available
1916 for all cases. Add expansion of COND_EXPR.
1918 2005-01-04 Ira Rosen <irar@il.ibm.com>
1920 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
1921 INTEGER_CST instead of TREE_CONSTANT.
1922 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
1923 instead of TREE_CONSTANT.
1924 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
1925 instead of TREE_CONSTANT.
1927 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
1929 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
1931 2005-01-04 Uros Bizjak <uros@kss-loka.si>
1934 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
1935 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
1937 2005-01-03 Richard Henderson <rth@redhat.com>
1939 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
1940 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
1941 fold_convert_const_real_from_real): Split out from ...
1942 (fold_convert_const): ... here.
1944 2005-01-03 Richard Henderson <rth@redhat.com>
1947 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
1948 (mov<MMXMODEI>_internal): Likewise.
1949 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
1950 Add V2SF case; use it for SSE1; don't use TI.
1951 (movdf_integer): Likewise.
1952 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
1954 2005-01-03 Ira Rosen <irar@il.ibm.com>
1956 * tree-vectorizer.c (vect_strip_conversions): New function.
1957 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
1958 check for binary class.
1960 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
1964 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
1966 * gimple-low.c (mark_blocks_with_used_vars): New function.
1967 (mark_blocks_with_used_subblocks): Ditto.
1968 (mark_used_blocks): Ditto.
1969 (pass_mark_used_blocks): New pass.
1970 * tree-inline.c: Include debug.h.
1971 (expand_call_inline): Call outlining_inline_function here.
1972 * tree-optimize.c (init_tree_optimization_passes): Add
1973 pass_mark_used_blocks.
1974 * tree-pass.h (pass_mark_used_blocks): New.
1975 * Makefile.in (tree-inline.o): Add debug.h dependency.
1977 2005-01-03 Geoffrey Keating <geoffk@apple.com>
1979 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
1980 VAR_DECLs to have weak_import attribute.
1982 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
1985 2004-01-03 Steven Bosscher <stevenb@suse.de>
1987 * passes.c (rest_of_compilation): Don't run regmove if only
1988 flag_expensive_optimizations, require flag_regmove instead.
1990 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
1992 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
1994 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
1995 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1996 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
1997 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1999 2005-01-03 Richard Henderson <rth@redhat.com>
2001 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
2002 (ia64_const_ok_for_letter_p): New.
2003 (ia64_const_double_ok_for_letter_p): New.
2004 (ia64_extra_constraint): New.
2005 (ia64_expand_vecint_compare): New.
2006 (ia64_expand_vcondu_v2si): New.
2007 (ia64_expand_vecint_cmov): New.
2008 (ia64_expand_vecint_minmax): New.
2009 (ia64_print_operand): Add 'v'.
2010 (ia64_preferred_reload_class): New.
2011 (ia64_vector_mode_supported_p): New.
2012 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
2013 (PREFERRED_RELOAD_CLASS): Move to function.
2014 (CONST_OK_FOR_LETTER_P): Move to function.
2015 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
2016 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
2017 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
2018 (EXTRA_CONSTRAINT): Move to function.
2019 * config/ia64/ia64.md: Include vect.md.
2020 (itanium_class): Add mmalua.
2022 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
2024 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
2025 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
2027 (const_int_2bit_operand): New.
2028 (fr_reg_or_0_operand): New.
2029 * config/ia64/ia64-modes.def: Add vector modes.
2030 * config/ia64/ia64-protos.h: Update.
2031 * config/ia64/vect.md: New file.
2033 2005-01-03 Richard Henderson <rth@redhat.com>
2035 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
2037 2005-01-03 Uros Bizjak <uros@kss-loka.si>
2040 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
2042 (log1pdf2): Change mode of operands[0,1] to DFmode.
2044 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
2046 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
2047 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
2048 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
2049 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
2050 SPARC_RELAXED_ORDERING.
2052 2005-01-03 Richard Henderson <rth@redhat.com>
2053 Uros Bizjak <uros@kss-loka.si>
2056 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
2057 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
2058 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
2060 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
2061 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
2062 const_pow2_1_to_128_operand): New predicates.
2064 2005-01-02 Greg McGary <greg@mcgary.org>
2066 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
2068 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
2069 object. [ptr_type]: Remove unused variable.
2070 Remember nearest addressable array-element or record-component
2071 when checking bitfield components. Tolerate empty BIND_EXPR.
2073 2005-01-01 Richard Henderson <rth@redhat.com>
2075 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
2076 that don't narrow the value. Fail for other conversions.
2078 2005-01-01 Richard Henderson <rth@redhat.com>
2081 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
2082 * c-lang.c (finish_file): Don't do it here.
2083 * objc/objc-act.c (objc_finish_file): Likewise.
2085 * cgraph.c (decl_assembler_name_equal): New.
2086 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
2087 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
2088 * cgraph.h (struct cgraph_varpool_node): Add next.
2089 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
2090 * varasm.c (assemble_alias): Mark the target as needed.
2092 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
2095 * function.c (get_arg_pointer_save_area): Use entry_of_function
2096 instead of get_insns.
2098 2005-01-01 Roger Sayle <roger@eyesopen.com>
2099 Andrew Pinski <pinskia@physics.uc.edu>
2100 James E. Wilson <wilson@specifixinc.com>
2102 PR rtl-optimization/12092
2103 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
2106 2005-01-01 Roger Sayle <roger@eyesopen.com>
2107 Olivier Hainque <hainque@act-europe.fr>
2109 * tree.c (int_fits_type_p): A narrower type always fits in a
2110 wider one, except for negative values into unsigned types.
2112 2005-01-01 Roger Sayle <roger@eyesopen.com>
2114 * tree.c (int_fits_type_p): Compare the result of force_fit_type
2115 with the original constant rather than require TREE_OVERFLOW.
2117 2005-01-01 Steven Bosscher <stevenb@suse.de>
2120 * c-decl.c (finish_function): If compiling C99, annotate the
2121 compiler generated return with the current file name and line 0.
2122 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
2123 the source line is greater than 0.
2124 (remove_bb): Likewise.
2126 See ChangeLog.12 for earlier changes.