1 2005-01-22 Richard Henderson <rth@redhat.com>
3 * config/i386/i386.md (smaxsf3): Fix mnemonic typo.
5 * config/i386/i386.c (ix86_prepare_fp_compare_args): Fix is_sse test.
7 2005-01-22 Richard Henderson <rth@redhat.com>
9 * genopinit.c (optabs): Use smin/smax for floating point too.
10 * doc/md.texi: Update to match. Clarify that floating point
11 results are undefined for +0/-0 and NaN.
12 * doc/rtl.texi: Likewise.
13 * rtl.def (SMIN, SMAX): Likewise
14 * tree.def (MIN_EXPR, MAX_EXPR): Likewise.
16 * config/alpha/alpha.md (smaxdf3, smindf3, smaxsf3, sminsf3): Add
17 leading 's' to the name.
18 * config/ia64/ia64.md (smaxsf3, sminsf3, smaxdf3, smindf3,
19 smaxxf3, sminxf3): Likewise.
20 * config/rs6000/rs6000.md (smaxdf3, smindf3, smaxsf3, sminsf3):
23 2005-01-22 Paul Brook <paul@codesourcery.com>
25 * config/arm/arm.md: Use "Uy" constraint for wcgr load/stores.
27 2005-01-21 Mark Dettinger <dettinge@de.ibm.com>
29 * config/s390/s390.c (struct processor_costs): New fields
30 dlgr, dlr, dr, dsgfr, dsgr.
31 (z900_cost, z990_cost): Values for new fields.
32 (s390_rtx_costs): New cases MEM und COMPARE in switch
33 statement. Modified handling of SIGN_EXTEND, ZERO_EXTEND,
36 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
38 * config/s390/s390.md ("doloop_si64"): Reload input value directly
39 into the register being decremented.
40 ("doloop_si31", "doloop_di"): Likewise.
41 ("*doloop_si_long"): Adapt pattern.
43 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
45 * config/s390/s390.h (HARD_REGNO_NREGS): Fix computation for
47 (CLASS_MAX_NREGS): Likewise.
49 2005-01-21 Daniel Berlin <dberlin@dberlin.org>
51 * doc/tree-ssa.texi (Statement Operands): Add example for new
52 must-def macro. Note deprecation of old operands interface.
54 2005-01-22 Richard Sandiford <rsandifo@redhat.com>
56 PR tree-optimization/19484
57 * tree-cfg.c (remove_fallthru_edge): New function.
58 (cleanup_control_flow): Remove fallthru edges from calls that are
59 now known not to return.
61 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
63 Fix PR tree-optimization/19038
64 * tree-ssa-dom.c (cprop_operand): Don't replace loop invaeriant
65 copies with loop variant ones.
67 2005-01-22 Kazu Hirata <kazu@cs.umass.edu>
69 * cfganal.c, real.h, reorg.c, timevar.def, tree-ssa-ccp.c,
70 config/alpha/alpha-protos.h, config/alpha/alpha.h,
71 config/alpha/alpha.md, config/alpha/predicates.md,
72 config/sparc/freebsd.h, config/sparc/netbsd-elf.h,
73 config/sparc/sol2.h: Update copyright.
75 * tree-cfg.c (remove_forwarder_block_with_phi): Look at the
76 first label to see if it is a nonlocal label.
78 2005-01-22 David Edelsohn <edelsohn@gnu.org>
79 Andrew Pinski <pinskia@physics.uc.edu>
82 * config/rs6000/rs6000.c (rs6000_va_start): Saturate n_gpr at
83 maximum number of GPRs. Saturate n_fpr at maximum number of FPRs.
85 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
87 * config/rs6000/rtems.h: Update copyright.
89 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
92 * config/rs6000/rtems.h: Resurrect cpp_os_rtems_spec from gcc < 3.4.
93 (CPP_OS_RTEMS_SPEC): New (From gcc-3.3's config/rs6000/sys4.h).
94 (SUBSUBTARGET_EXTRA_SPECS): Use CPP_OS_RTEMS_SPEC.
96 2005-01-22 Volker Reichelt <reichelt@igpm.rwth-aachen.de>
99 * c-typeck.c (convert_arguments): Check for error_mark_node.
101 2005-01-21 Ian Lance Taylor <ian@airs.com>
103 PR tree-optimization/13000
104 * tree-inline.c: Include "tree-flow.h".
105 (expand_call_inline): If warn_return_type, warn if non-void inline
106 function falls through.
107 * tree-cfg.c (execute_warn_function_return): Don't warn about
108 control reaching end if TREE_NO_WARNING is set. Set
110 * gimple-low.c (block_may_fallthru): Don't assume that SWITCH_EXPR
112 * gimplify.c (shortcut_cond_expr): Don't emit a jump over the else
113 branch if we don't need one.
114 * c-typeck.c: Include "tree-flow.h"
115 (c_finish_bc_stmt): Don't add a goto if the current statement
116 list doesn't fall through to the current point.
118 2005-01-21 Roger Sayle <roger@eyesopen.com>
120 PR rtl-optimization/576
121 * real.c (real_arithmetic): Change return type from void to bool
122 to return an indication that the result may be inexact.
123 * real.h (real_arithmeric): Update prototype.
124 * fold-const.c (const_binop): Don't constant fold floating
125 point expressions when the user specifies -frounding-math and
126 the result may depend upon the run-time rounding mode.
127 (fold_convert_const_real_from_real): Clean-up.
128 (fold_initializer): Ignore flag_rounding_math for initializers.
129 * simplify-rtx.c (simplify_binary_operation): Likewise, don't
130 constant fold FP operations with flag_rounding_math if the
131 result may depend upon the run-time rounding mode.
133 2005-01-21 Tom Tromey <tromey@redhat.com>
135 * c-cppbuiltin.c (define__GNUC__): Correct assertion.
137 2005-01-21 Kazu Hirata <kazu@cs.umass.edu>
139 * tree-cfg.c: Fix comment typos.
141 2005-01-21 J"orn Rennecke <joern.rennecke@st.com>
143 * passes.c (rest_of_handle_flow2): Close / open the flow2 dump file
144 around the call to rest_of_handle_branch_target_load_optimize.
146 2005-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
148 * config.gcc (sparc-*-netbsdelf*): Include dbxelf.h
149 (sparc64-*-openbsd*): Likewise.
150 (sparclite-*-elf*): Include tm-dwarf2.h.
151 (sparc86x-*-elf*): Likewise.
152 (sparc64-*-elf*): Likewise.
153 * config/sparc/linux64.h (DWARF2_DEBUGGING_INFO): Delete.
154 (DBX_DEBUGGING_INFO): Likewise.
155 * config/sparc/liteelf.h (DWARF2_DEBUGGING_INFO): Likewise.
156 (PREFERRED_DEBUGGING_TYPE): Likewise.
157 * config/sparc/netbsd-elf.h (PREFERRED_DEBUGGING_TYPE): Likewise.
158 * config/sparc/sp64-elf.h (DBX_DEBUGGING_INFO): Likewise.
159 (PREFERRED_DEBUGGING_TYPE): Likewise.
160 * config/sparc/sp86x-elf.h (DWARF2_DEBUGGING_INFO): Likewise.
161 (PREFERRED_DEBUGGING_TYPE): Likewise.
162 * config/sparc/sparc.h (DBX_DEBUGGING_INFO): Likewise.
164 2005-01-21 Hans-Peter Nilsson <hp@bitrange.com>
167 * combine.c (combine_simplify_rtx): Revert change of 2004-12-31.
169 2005-01-20 Paul Brook <paul@codesourcery.com>
171 * doc/rtl.texi: Document value extension requirements for CONST_INT.
173 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
175 * tree-cfg.c (tree_verify_flow_info): Check that a nonlocal
176 label is first in a sequence of labels.
178 * tree-cfg.c (tree_verify_flow_info): Fix a typo.
180 2005-01-20 Janis Johnson <janis187@us.ibm.com>
181 Giovanni Bajo <giovannibajo@gcc.gnu.org>
183 * doc/sourcebuild.texi (Test Directives): New.
185 2005-01-20 Roger Sayle <roger@eyesopen.com>
187 * real.c (real_floor): Don't leave the result uninitialized when
188 mode is VOIDMode, but return the unrounded intermediate value.
189 (real_ceil): Likewise.
191 2005-01-20 Paul Brook <paul@codesourcery.com>
193 * config/arm/arm.md (insv): Use gen_int_mode.
195 2005-01-20 Eric Botcazou <ebotcazou@libertysurf.fr>
197 * config/sparc/freebsd.h (ASM_OUTPUT_INTERNAL_LABELREF): Delete.
198 * config/sparc/linux.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
199 * config/sparc/linux64.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
200 * config/sparc/netbsd-elf.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
201 * config/sparc/sol2.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
203 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
206 * dwarf2out.c (concat_loc_descriptor): We don't know we can use
207 fbreg, so we have to assume we can't.
208 (loc_descriptor_from_tree_1): Ditto.
209 (containing_function_has_frame_base): New function.
210 (add_location_or_const_value_attribute): Use it.
211 Also try to generate a frame_base from a single element location
214 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
216 PR tree-optimization/15349
217 * timevar.def (TV_TREE_MERGE_PHI): New.
218 * tree-cfg.c (tree_forwarder_block_p): Add a new argument
220 (remove_forwarder_block, cleanup_forwarder_blocks): Adjust the
221 calls to tree_forwarder_block_p.
222 (remove_forwarder_block_with_phi, merge_phi_nodes,
223 gate_merge_phi, pass_merge_phi): New.
224 * tree-optimize.c (init_tree_optimization_passes): Add
226 * tree-pass.h: Add an extern for pass_merge_phi;
228 2005-01-20 Richard Henderson <rth@redhat.com>
231 * config/i386/emmintrin.h (_mm_castpd_ps, _mm_castpd_si128): New.
232 (_mm_castps_pd, _mm_castps_si128): New.
233 (_mm_castsi128_ps, _mm_castsi128_pd): New.
235 2005-01-20 Richard Henderson <rth@redhat.com>
238 * config/i386/mmintrin.h (_mm_cvtsi32_si64): Use
239 __builtin_ia32_vec_init_v2si.
240 (_mm_cvtsi64_si32): Use __builtin_ia32_vec_ext_v2si.
241 * config/i386/i386.c (IX86_BUILTIN_VEC_EXT_V2SI): New.
242 (ix86_init_mmx_sse_builtins): Create it.
243 (ix86_expand_builtin): Expand it.
244 (ix86_expand_vector_set): Handle V2SFmode and V2SImode.
245 * config/i386/mmx.md (vec_extractv2sf_0, vec_extractv2sf_1): New.
246 (vec_extractv2si_0, vec_extractv2si_1): New.
248 2005-01-20 Richard Henderson <rth@redhat.com>
251 * config/i386/i386.c (ix86_expand_push): New.
252 * config/i386/mmx.md (push<MMXMODE>1): New.
253 * config/i386/sse.md (push<SSEMODE>1): New.
254 * config/i386/i386-protos.h: Update.
256 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
258 * params.def, params.h, predict.c, tree-ssa-loop-im.c: Update
261 2005-01-20 Richard Henderson <rth@redhat.com>
263 * rtl.def (CONST_VECTOR): Use RTX_CONST_OBJ.
264 * rtl.h (CONSTANT_P): Don't special case CONST_VECTOR.
266 2005-01-19 Richard Henderson <rth@redhat.com>
269 * config/i386/i386.c (ix86_expand_vector_move_misalign): Convert
270 to V4SFmode in SSE1 fallback load path.
272 2005-01-19 Richard Henderson <rth@redhat.com>
274 * config/i386/i386.c (ix86_expand_vector_init_one_var): Fix typo
275 in QImode expansion to ix86_expand_vector_set.
276 (ix86_expand_vector_init_general): Fix typo in V8HImode recursive call.
278 2005-01-19 Richard Henderson <rth@redhat.com>
281 * config/i386/i386.c (ix86_preferred_reload_class): Return a proper
282 subclass of the input class.
283 (ix86_secondary_memory_needed): Always true for cross-MMX classes.
284 Always true for cross-SSE1 classes. Rationalize conditionals.
285 * config/i386/i386.h (SSE_CLASS_P, MMX_CLASS_P): Use straight equality.
286 * config/i386/i386.md (movsi_1): Add MMX/SSE zeros. Fix alternatives
287 for SSE1. Don't check TARGET_INTER_UNIT_MOVES.
288 (movdi_2): Add MMX/SSE zeros.
289 (movdi_1_rex64): Likewise. Don't check TARGET_INTER_UNIT_MOVES.
290 (movsf_1): Don't check TARGET_INTER_UNIT_MOVES.
291 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
292 (movsi_1_nointernunit, movdi_1_rex64_nointerunit): Remove.
293 (movsf_1_nointerunit, zero_extendsidi2_32_1): Remove.
294 (zero_extendsidi2_rex64_1): Remove.
295 (MOV0 peephole): Check GENERAL_REG_P.
297 2005-01-19 Richard Henderson <rth@redhat.com>
300 * config/i386/i386.c (ix86_expand_vector_set): Fix third and fourth
302 (ix86_expand_vector_extract): Likewise.
304 2005-01-19 Richard Henderson <rth@redhat.com>
307 * expr.c (emit_move_change_mode): New argument force; use
308 simplify_gen_subreg if true.
309 (emit_move_via_alt_mode): Merge into ...
310 (emit_move_via_integer): ... here.
311 (emit_move_ccmode): Use emit_move_change_mode directly.
313 2005-01-19 Richard Henderson <rth@redhat.com>
316 * config/alpha/alpha.c (alpha_rtx_costs): Handle HIGH.
317 (alpha_preferred_reload_class): Handle CONST_VECTOR.
318 (alpha_emit_set_const_1): Add no_output parameter; don't emit
320 (alpha_emit_set_const): Likewise. Make static.
321 (alpha_emit_set_long_const): Make static.
322 (alpha_extract_integer): Split out from alpha_expand_mov.
323 (alpha_split_const_mov): Likewise.
324 (alpha_expand_mov): Use them. Handle CONST_VECTOR.
325 (alpha_legitimate_constant_p): New.
326 * config/alpha/alpha-protos.h: Update.
327 * config/alpha/alpha.h (REGISTER_MOVE_COST): Correct fp<->gp cost.
328 (LEGITIMATE_CONSTANT_P): Re-implement with a function.
329 * config/alpha/alpha.md (movsi): Add n alternative.
330 (movsi_nt_vms, movdi_er_nofix, movdi_er_fix, movdi_fix): Likewise.
331 (mov<VEC>_fix, mov<VEC>_nofix): Add i alternative.
332 (splitters for all of the above): Use alpha_split_const_mov.
333 * config/alpha/predicates.md (non_add_const_operand): New.
334 (non_zero_const_operand): New.
335 (input_operand): Use alpha_legitimate_constant_p after reload.
337 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
339 PR tree-optimization/19038
340 * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): New function.
341 (add_candidate): Add ivs with increment in latch only if
342 allow_ip_end_pos_p is true.
343 (determine_iv_cost): Use empty_block_p.
345 2005-01-19 Daniel Berlin <dberlin@dberlin.org>
347 * cfganal.c (compute_dominance_frontiers_1): Replace with new algorithm
348 (compute_dominance_frontiers): Ditto.
350 2005-01-19 Ralf Corsepius <ralf.corsepius@rtems.org>
353 * config/sh/t-rtems: New.
354 * config.gcc (sh-*-rtems*): Reflect having added config/sh/t-rtems.
356 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
358 * tree-ssa-ccp.c (maybe_fold_offset_to_component_ref): Always subtract
359 the offset of the selected field.
361 2005-01-19 Kazu Hirata <kazu@cs.umass.edu>
363 * tree-cfg.c (remove_forwarder_block): Fix the check to
364 prevent a nonlocal label from appearing in the middle of a
367 2005-01-19 Hans-Peter Nilsson <hp@axis.com>
369 PR rtl-optimization/19462
370 * reorg.c (find_end_label): Create return insn only if
371 current_function_epilogue_delay_list is empty.
373 2005-01-19 Paolo Bonzini <bonzini@gnu.org>
375 * doc/passes.texi: Remove paragraph mentioning new-ra.
377 2005-01-19 Jakub Jelinek <jakub@redhat.com>
379 PR rtl-optimization/15139
380 * combine.c: Include params.h.
381 (count_rtxs): New function.
382 (record_value_for_reg): If replace_rtx would replace at least
383 2 occurrences of REG in VALUE and TEM is really large, replace REG with
384 (clobber (const_int 0)) instead of TEM.
385 * params.def (PARAM_MAX_LAST_VALUE_RTL): New.
386 * params.h (MAX_LAST_VALUE_RTL): New.
387 * Makefile.in (combine.o): Depend on $(PARAMS_H).
388 * doc/invoke.texi (--param max-last-value-rtl=N): Document.
391 * c-typeck.c (digest_init): Only call build_vector if all constructor
392 elements are *_CST nodes.
393 * gimplify.c (gimplify_init_constructor): Likewise.
396 * c-typeck.c (digest_init): Only call build_vector if inside_init
399 2005-01-18 Tobias Schl"uter <tobias.schlueter@physik.uni-muenchen.de>
401 * toplev.c (init_asm_output): Remove dead #ifdef.
403 2005-01-18 Kaz Kojima <kkojima@gcc.gnu.org>
405 * config/sh/linux.h (TARGET_C99_FUNCTIONS): Define.
407 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
409 * calls.c (expand_call): Check DECL_BUILT_IN_CLASS before
410 accessing DECL_FUNCTION_CODE.
411 * dojump.c (do_jump): Likewise.
412 * gimplify.c (gimplify_call_expr): Likewise.
413 * predict.c (expr_expected_value): Likewise.
414 (strip_builtin_expect): Likewise.
415 * tree-inline.c (estimate_num_insns_1): Likewise.
416 * tree-ssa-loop-im.c (stmt_cost): Likewise
417 * fold-const.c (fold): Test for BUILT_IN_NORMAL.
418 (tree_expr_nonnegative_p): Likewise.
420 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
422 * config/sparc/sparc.c (load_pic_register): Emit the appropriate
423 variant of the load_pcrel_sym pattern.
424 * config/sparc/sparc.md (P macro): Move to the top.
425 (load_pcrel_sym): Macroize using P.
426 (save_register_window): Likewise.
428 2005-01-18 Aldy Hernandez <aldyh@redhat.com>
430 * config/rs6000/rs6000.md ("sunordered"): Disable for e500.
433 2005-01-18 Andrew Pinski <pinskia@physics.uc.edu>
436 * c-typeck.c (build_asm_expr): Strip nops off
437 input memory operands.
439 2005-01-18 David Edelsohn <edelsohn@gnu.org>
441 * config/rs6000/aix43.h (CPLUSPLUS_CPP_SPEC): Delete
442 _XOPEN_SOURCE, _XOPEN_SOURCE_EXTENDED, _LARGE_FILE_API.
443 * config/rs6000/aix51.h (CPLUSPLUS_CPP_SPEC): Same.
444 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Same.
446 2005-01-18 Joel Sherrill <joel.sherrill@oarcorp.com>
449 * config/i386/i386.c (override_options): If the 80387 is disabled,
450 then do not return FP values using FP registers.
452 2005-01-18 Kazu Hirata <kazu@cs.umass.edu>
454 * basic-block.h, c-common.c, c-cppbuiltin.c, c-lang.c,
455 c-tree.h, cfgbuild.c, cgraph.c, cgraph.h, collect2.c,
456 combine.c, config.gcc, coverage.h, cse.c, cselib.c,
457 defaults.h, df.c, dwarf2asm.c, dwarf2out.c, explow.c, expr.c,
458 flow.c, fold-const.c, gcse.c, ggc-page.c, gimple-low.c,
459 gimplify.c, ifcvt.c, langhooks-def.h, lcm.c, optabs.h,
460 output.h, postreload-gcse.c, postreload.c, recog.c,
461 resource.c, rtl.def, rtlanal.c, sched-deps.c, sched-rgn.c,
462 targhooks.h, toplev.c, tree-data-ref.c, tree-eh.c,
463 tree-flow.h, tree-loop-linear.c, tree-mudflap.h, tree-nrv.c,
464 tree-optimize.c, tree-outof-ssa.c, tree-pass.h,
465 tree-scalar-evolution.c, tree-ssa-copy.c, tree-ssa-dce.c,
466 tree-ssa-dse.c, tree-ssa-loop-ivopts.c, tree-ssa-operands.c,
467 tree-ssa-pre.c, tree-ssa.c, tree-vectorizer.c, tree.def,
468 unwind-dw2-fde-darwin.c, var-tracking.c: Update copyright.
470 * config/darwin.c, config/darwin.h, config/freebsd-spec.h,
471 config/alpha/alpha.c, config/arm/vfp.md,
472 config/i386/emmintrin.h, config/i386/i386-modes.def,
473 config/i386/i386-protos.h, config/i386/i386.h,
474 config/i386/pmmintrin.h, config/i386/ppro.md,
475 config/i386/predicates.md, config/i386/xmmintrin.h,
476 config/i860/i860.c, config/ia64/ia64-modes.def,
477 config/ia64/ia64-protos.h, config/ia64/ia64.h,
478 config/ia64/ia64.md, config/ia64/itanium1.md,
479 config/ia64/itanium2.md, config/ia64/predicates.md,
480 config/iq2000/iq2000.h, config/mips/linux64.h,
481 config/rs6000/aix41.h, config/rs6000/aix43.h,
482 config/rs6000/aix52.h, config/rs6000/darwin-fallback.c,
483 config/rs6000/rs6000.c, config/rs6000/rs6000.h,
484 config/rs6000/spe.md, config/sh/sh.md, config/sparc/linux.h,
485 config/sparc/linux64.h, config/sparc/litecoff.h,
486 config/sparc/sp64-elf.h, config/sparc/sparc.c,
487 config/sparc/sparc.h, config/sparc/sparc.md,
488 config/sparc/sysv4-only.h: Update copyright.
490 2005-01-18 Richard Henderson <rth@redhat.com>
493 * config/i386/i386.c (ix86_expand_fp_movcc): Fail for LTGT and UNEQ.
495 2005-01-18 Marc Espie <espie@openbsd.org>
496 * config/i386/openbsdelf.h: Typo.
498 2005-01-18 Richard Henderson <rth@redhat.com>
500 * config/i386/i386.c (ix86_expand_fp_movcc): Remove TARGET_IEEE_FP
501 special case for sse.
502 * config/i386/predicates.md (sse_comparison_operator): Likewise.
504 2005-01-18 Dorit Naishlos <dorit@il.ibm.com>
506 * tree-inline.c (estimate_num_insns_1): Added cases for
507 ALIGN_INDIRECT_REF, MISALIGNED_INDIRECT_REF, and REALIGN_LOAD_EXPR.
509 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
511 PR rtl-optimization/19296
512 * combine.c (simplify_comparison): Rewrite the condition under
513 which a non-paradoxical SUBREG of a PLUS can be lifted when
514 compared against a constant.
516 2005-01-18 Andi Kleen <ak@muc.de>
518 * c-typeck.c: (convert_for_assignment): Check warn_pointer_sign.
519 * c.opt (-Wpointer-sign): Add.
520 * doc/invoke.texi: (-Wpointer-sign): Add.
522 2005-01-18 Uros Bizjak <uros@kss-loka.si>
524 * config/i386/i386.c (override_options): Revert 2004-11-24 change.
525 * config/i386/i386.md (fmodsf3, dremsf3, *sinsf2, *cossf2, sincossf3,
526 *tansf3_1, tansf2, atan2sf3_1, atan2sf3, atansf2, asinsf2, acossf2,
527 logsf2, log10sf2, log2sf2, log1psf2, logbsf2, ilogbsf2, expsf2,
528 exp10sf2, exp2sf2, expm1sf2, rintsf2, floorsf2, ceilsf2, btruncsf2,
529 nearbyintsf2): Disable for TARGET_SSE_MATH. Leave patterns enabled
530 for TARGET_MIX_SSE_I387.
531 (fmoddf3, dremdf3, *sindf2, *sinextendsfdf2, *cosdf2, *cosextendsfdf2,
532 sincosdf3, *sincosextendsfdf3, *tandf3_1, tandf2, atan2df3_1,
533 atan2df3, atandf2, asindf2, acosdf2, logdf2, log10df2, log2df2,
534 log1pdf2, logbdf2, expdf2, exp10df2, exp2df2, expm1df2, rintdf2,
535 floordf2, ceildf2, btruncdf2, nearbyintdf2): Disable for
536 (TARGET_SSE2 && TARGET_SSE_MATH). Leave patterns enabled for
538 (atan2sf3, atan2df3, atan2xf3): Remove register constraints
541 2005-01-18 Uros Bizjak <uros@kss-loka.si>
544 * config/i386/mmx.md (*movv2sf_internal_rex64, *movv2sf_internal):
545 Add movaps alternative for xmm reg->reg move.
547 2005-01-17 Jeff Law <law@redhat.com>
549 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Clear SSA_NAME_VALUE
550 at the end of the main DOM loop rather than just before DOM exits.
552 2005-01-17 Diego Novillo <dnovillo@redhat.com>
554 PR tree-optimization/19121
555 * tree-ssa-alias.c (compute_flow_sensitive_aliasing): When
556 adding aliases to a name tag, also add them to the pointer's
558 * tree-ssa-copy.c (merge_alias_info): Do not merge flow
559 sensitive alias info at all. Only check that the two pointers
560 have compatible pointed-to sets.
561 * tree-ssa.c (verify_name_tags): Verify that the alias set of
562 a pointer's type tag is a superset of the alias set of the
565 2005-01-17 James E Wilson <wilson@specifixinc.com>
568 * config/ia64/ia64.md (movxf): Handle general register source. Adjust
569 comment to document why.
571 2005-01-17 Richard Henderson <rth@redhat.com>
573 * config/i386/sse.md (smaxv4sf3_finite, sse_vmsmaxv4sf3_finite,
574 sminv4sf3_finite, sse_vmsminv4sf3_finite, smaxv2df3_finite,
575 sse2_vmsmaxv2df3_finite, sminv2df3_finite,
576 sse2_vmsminv2df3_finite): New.
577 (smaxv4sf3, sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, smaxv2df3,
578 sse2_vmsmaxv2df3, sminv2df3, sse2_vmsminv2df3): Remove commutative.
579 Force op1 into register.
581 2005-01-17 Kelley Cook <kcook@gcc.gnu.org>
583 * Makefile.in (STAGEMOVESTUFF): Stage all the stamp files to prevent
584 bubblestrap from always rebuilding build directory.
586 2005-01-17 Ulrich Weigand <uweigand@de.ibm.com>
588 * config/s390/s390.md ("*clc"): Remove incorrect '='.
589 ("*cmpmem_short"): Likewise.
591 2005-01-17 Steven Bosscher <stevenb@suse.de>
593 * basic-block.h: Document BB_* flags.
594 * regrename.c (copyprop_hardreg_forward): Don't use BB_VISITED,
595 use an sbitmap instead.
596 * sched-rgn.c (compute_trg_info): Likewise.
598 2005-01-17 Richard Sandiford <rsandifo@redhat.com>
600 * config.gcc (mips64*-*-linux*): Set the default abi to n32. Remove
601 redundant target_cpu_default and extra_parts lines. Use the default
603 * config/mips/linux64.h (DRIVER_DEFAULT_ABI_SELF_SPEC): Delete.
604 (SUBTARGET_EXTRA_SPECS): Delete.
605 (DRIVER_SELF_SPECS): Remove architecture lines. Set the ABI directly.
606 (SUBTARGET_ASM_SPEC): Remove -mabi=n32 and -mabi=64 mappings.
607 * config/mips/t-linux64 (MULTILIB_DIRNAMES): Use "32" for -mabi=32
608 and "n32" for -mabi=n32.
610 2005-01-17 Ian Lance Taylor <ian@airs.com>
613 * c-typeck.c (build_c_cast): Revert patch of 2003-02-16: don't
614 fold constant variables into initial values.
616 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
618 * tree-cfg.c (tree_can_merge_blocks_p): Reorder two checks.
620 * tree-cfg.c (tree_forwarder_block_p): Speed up by walking
621 through the statements backward.
623 2005-01-17 Ian Lance Taylor <ian@airs.com>
626 * tree-inline.c (expand_call_inline): Set TREE_NO_WARNING on
627 a variable set to the return value of the inlined function.
629 2005-01-17 Mark Dettinger <dettinge@de.ibm.com>
631 * rtlanal.c (rtx_cost): Assign cost of 0 to a SUBREG
632 when modes are tieable.
634 2005-01-17 Ranjit Mathew <rmathew@hotmail.com>
636 * gthr-posix.h (__gthread_active_p): Use pthread_cancel instead
637 of pthread_create to find out if threads are enabled.
638 * gthr-posix95.h (__gthread_active_p): Likewise.
640 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
642 * common.opt (-fnew-ra): Remove.
644 * toplev.h (flag_new_regalloc): Remove.
645 * Makefile.in (ra*.*): Don't mention.
646 * passes.c (rest_of_handle_new_regalloc): Remove.
647 (rest_of_handle_combine, rest_of_compilation): Always consider
648 flag_new_regalloc as false.
649 * doc/invoke.texi: Don't document -fnew-ra.
651 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
653 * bb-reorder.c (fix_edges_for_rarely_executed_code): Remove
654 last parameter to reg_scan.
655 * loop.c (loop_optimize): Likewise.
656 * passes.c (rest_of_handle_tracer, rest_of_handle_if_conversion,
657 rest_of_handle_web, rest_of_handle_cfg, rest_of_handle_jump_bypass,
658 rest_of_handle_life, rest_of_handle_cse, rest_of_handle_cse2,
659 rest_of_handle_gcse, rest_of_handle_loop_optimize,
660 rest_of_handle_loop2, rest_of_handle_jump2): Likewise.
661 * regclass.c (reg_scan): Likewise, for the declaration.
662 * rtl.h (reg_scan): Likewise, for the prototype.
664 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
666 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Don't call
669 2005-01-16 Daniel Berlin <dberlin@dberlin.org>
671 * tree-ssa-pre.c (add_to_sets): s1 may be NULL.
672 (compute_avail): Uses don't go in tmp_gen.
674 2005-01-16 Steven Bosscher <stevenb@suse.de>
676 * ggc-page.c (ggc_alloc_stat): Use __builtin_ctzl instead of a
677 loop to look for a free slot in a page entry.
679 2005-01-16 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
682 * defaults.h (TARGET_DEFERRED_OUTPUT_DEFS): Provide default.
683 * toplev.c (compile_file): Call process_pending_assemble_output_defs
684 just before targetm.asm_out.file_end.
685 * tree.h (process_pending_assemble_output_defs): Declare.
686 * varasm.c (assemble_output_def, process_pending_assemble_output_defs):
688 (assemble_alias): Defer generation of assembly code for defines when
689 TARGET_DEFERRED_OUTPUT_DEFS is true.
690 * config/rs6000/aix41.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
691 * config/rs6000/aix43.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
692 * doc/tm.texi (TARGET_DEFERRED_OUTPUT_DEFS): document.
694 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
697 * pa.c (pa_scalar_mode_supported_p): New function.
698 (TARGET_SCALAR_MODE_SUPPORTED_P): Define.
699 * pa.h (MIN_UNITS_PER_WORD): Add comment.
701 * pa-protos.h (prefetch_operand): Delete.
702 (prefetch_cc_operand, prefetch_nocc_operand): New declations.
703 * pa.c (prefetch_operand): Delete.
704 (prefetch_cc_operand, prefetch_nocc_operand): New functions.
705 * pa.h (EXTRA_CONSTRAINT): Add `W' constraint.
706 (PREDICATE_CODES): Delete prefetch_operand. Add prefetch_cc_operand
707 and prefetch_nocc_operand.
708 * pa.md (prefetch): Rework to avoid reload problems handling short
709 displacements when a cache control completer needs to be provided.
710 (prefetch_32, prefetch_64): Delete.
711 (prefetch_cc, prefetch_nocc): New patterns.
713 2005-01-15 David Edelsohn <edelsohn@gnu.org>
715 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Revert previous change.
717 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
720 * pa.c (pa_asm_output_aligned_common): Warn if specified alignment
721 exceeds maximum alignment for global common data.
722 * pa64-hpux.h (MAX_OFILE_ALIGNMENT): Define to 32768.
723 * som.h (MAX_OFILE_ALIGNMENT): Likewise.
725 2004-01-15 Roger Sayle <roger@eyesopen.com>
727 * tree-ssa-dom.c (extract_range_from_cond): Correct condition.
729 2004-01-15 Roger Sayle <roger@eyesopen.com>
731 * harg-reg-set.h (reg_class_names): Prototype global array.
732 * regclass.c (reg_class_names): Declare here and initialize to
734 (dump_regclass): Remove local declaration of reg_class_names.
735 (regclass): Likewise.
736 * cfg.c (dump_flow_info): Likewise.
737 * ra-debug.c (reg_class_names): Likewise.
738 * regrename.c (reg_class_names): Likewise.
739 * reload.c (reg_class_names): Likewise.
740 * reload1.c (spill_failure): Likewise.
741 * config/m68hc11/m68hc11.c (reg_class_names): Likewise.
743 2005-01-15 Ulrich Weigand <uweigand@de.ibm.com>
745 * config/s390/s390.md ("reload_outti"): Remove predicate for
746 output operand. Abort if operand is not a MEM.
747 ("reload_outdi", "reload_outdf"): Likewise.
749 2005-01-15 Marc Espie <espie@openbsd.org>
751 * config.gcc (*-*-openbsd*): Set HAS_LIBC_R for OpenBSD <= 3.2.
752 (i[34567]86-*-openbsd*): Switch to ELF for OpenBSD >= 3.4.
753 * config/openbsd.h: Fix C++ includes for native configurations.
754 Add proper OS_CPP_BUILTINS. Fix libspec for recent OpenBSD.
755 Add trampoline support.
756 * config/i386/openbsdelf.h: New.
758 2005-01-15 Marc Espie <espie@openbsd.org>
760 * collect2.c (main): Explicitly parse -dynamic-linker option.
762 2005-01-15 Jakub Jelinek <jakub@redhat.com>
764 PR tree-optimization/19060
765 * tree-ssa-dom.c (extract_range_from_cond) <case LT_EXPR, GT_EXPR>:
766 Return 0 if op1 <= TYPE_MIN_VALUE () resp. op1 >= TYPE_MAX_VALUE ().
767 (simplify_cond_and_lookup_avail_expr): Add assert for dummy == 0
768 and handle extract_range_from_cond returning false.
769 * fold-const.c (fold): Optimize comparisons with min/max even for
770 width > HOST_BITS_PER_WIDE_INT.
772 2005-01-15 Ralf Corsepius <ralf.corsepius@rtems.org>
774 * config/mips/rtems.h (MIPS_DEFAULT_GVALUE): Set to 0.
775 * config/mips/t-rtems (MULTILIBS_DIRNAMES,MULTILIB_OPTIONS):
776 Remove little endian multilib variants.
777 Add mips32 multilib variant.
779 2005-01-15 Kazu Hirata <kazu@cs.umass.edu>
781 * explow.c (copy_all_reg, stabilize): Remove.
782 * expr.h: Remove the corresponding prototypes.
784 * rtlanal.c (regs_set_between_p): Remove.
785 * rtl.h: Remove the corresponding prototype.
787 2005-01-14 David Edelsohn <edelsohn@gnu.org>
789 * config/rs6000/rs6000.c (rs6k_nonimmediate_operand): Rename
790 to rs6000_nonimmediate_operand.
791 * config/rs6000/rs6000.h (PREDICATE_CODES): Rename
792 rs6k_nonimmediate_operand to rs6000_nonimmediate_operand.
793 * config/rs6000/rs6000.md (movsi_internal1): Rename predicate to
794 rs6000_nonimmediate_operand.
795 * config/rs6000/spe.md (movdf_e500_double): Rename predicate to
796 rs6000_nonimmediate_operand.
798 2005-01-14 Andrew Pinski <pinskia@physics.uc.edu>
800 * tree-ssa-alias.c (pass_may_alias): Add TODO_verify_stmts.
802 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
804 * config/rs6000/rs6000.h (CLASS_MAX_NREGS): DF goes in 1 register
806 (CANNOT_CHANGE_MODE_CLASS): Restrict DI mode changes on e500v2.
807 (PREDICATE_CODES): Add rs6k_nonimmediate_operand.
809 * config/rs6000/rs6000.c (invalid_e500_subreg): New.
810 (rs6k_nonimmediate_operand): New.
811 (rs6000_legitimate_offset_address_p): Handle DI modes on e500v2
813 (legitimate_lo_sum_address_p): Same.
814 (rs6000_legitimize_address): Same.
815 (rs6000_legitimize_reload_address): Same.
816 (rs6000_legitimate_address): Same.
817 (spe_build_register_parallel): Pass DF and DC modes in a DI
820 * config/rs6000/rs6000.md ("*movsi_internal1"): Change predicate
821 to rs6k_nonimmediate_operand.
823 * config/rs6000/spe.md ("*frob_df_di"): New.
824 ("*frob_di_df"): New.
825 ("*frob_di_df_2"): New.
826 ("*mov_sidf_e500_subreg0"): New.
827 ("*mov_sidf_e500_subreg4"): New.
828 ("*movdf_e500_double"): Change predicate to
829 rs6k_nonimmediate_operand.
831 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
833 * postreload.c (move2add_note_store): Only call
834 trunc_int_for_mode on scalar integers.
836 2005-01-14 Steven Bosscher <stevenb@suse.de>
838 * tree-ssa-dce.c (visited_control_parents): New sbitmap to
839 replace BB_VISITED uses.
840 (find_obviously_necessary_stmts): Don't clear BB_VISITED.
841 (propagate_necessity): Check the bitmap instead of BB_VISITED.
842 (tree_dce_done): Free visited_control_parents.
843 (perform_tree_ssa_dce): Allocate and clear it.
844 * tree-ssa-pre.c (compute_antic_aux): Make non-recursive.
845 (compute_antic): Iterate from here using a DFS. Use an sbitmap
846 instead of BB_VISITED.
848 2005-01-14 Kazu Hirata <kazu@cs.umass.edu>
850 * c-tree.h, coverage.h, langhooks-def.h, optabs.h, output.h,
851 rtl.h, targhooks.h, tree-flow.h, tree-mudflap.h, tree.h:
852 Remove unused prototypes.
854 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
856 * config/rs6000/rs6000.c (rs6000_stack_info): Fix caching of SPE
857 64-bit register usage.
858 (spe_func_has_64bit_regs_p): Add FIXME note.
860 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
862 * config/rs6000/rs6000.c (rs6000_generate_compare): Replace
863 flag_finite_math_only with flag_unsafe_math_optimizations.
865 2005-01-14 J. D. Johnston <jjohnst@us.ibm.com>
867 * config/s390/tpf-unwind.h (s390_fallback_frame_state): Correct
869 (__tpf_eh_return): Copy TPF private stack area from the module
870 boundary stack frame. Add check for when module addresses are
871 equal, but no stub address is found.
873 2005-01-14 Richard Earnshaw <rearnsha@arm.com>
876 * arm.h (struct machine_function): Add call_via field.
877 (thumb_call_via_label): Declare.
878 * arm.c (thumb_call_via_label): New variable.
879 (thumb_call_reg_needed): New variable.
880 (arm_output_function_epilogue): For Thumb code, output any per-function
881 call-indirect trampolines.
882 (thumb_call_via_reg): New function.
883 (arm_file_end): New function.
884 (TARGET_ASM_FILE_END): Call arm_file_end.
885 (aof_file_end): Likewise.
886 * arm-protos.h (thumb_call_via_reg): Declare.
887 * arm.md (call_reg_thumb, call_value_reg_thumb): Call
888 thumb_call_via_reg in normal case.
890 2005-01-14 Jakub Jelinek <jakub@redhat.com>
893 PR rtl-optimization/19348
894 * recog.c (peephole2_optimize): Do global life update if some peephole
895 decides it doesn't need at least one of its inputs and that change
896 influences liveness at the start of the basic block.
898 * basic-block.h (EXECUTE_IF_AND_COMPL_IN_REG_SET): Needs 2 REGSET
899 arguments instead of 1.
901 2005-01-14 Eric Botcazou <ebotcazou@libertysurf.fr>
904 * varasm.c (initializer_constant_valid_p) <ADDR_EXPR>: Return
905 zero for nested functions needing a static chain or functions
906 with a non-constant address.
908 2005-01-13 Roger Sayle <roger@eyesopen.com>
910 * simplify-rtx.c (simplify_binary_operation) <AND>: Optimize
911 (and (sign_extend X) C) into (zero_extend (and X C)).
913 2005-01-13 David O'Brien <obrien@FreeBSD.org>
915 * config/freebsd-spec.h: Make KSE pthread lib logic the default.
917 2005-01-13 Richard Henderson <rth@redhat.com>
922 * config/i386/i386.md (cmpdf, cmpsf, bunordered, bordered, buneq,
923 bunge, bungt, bunle, bunlt, bltgt): Enable for TARGET_SSE_MATH,
925 (cmpfp_i_387): Rename from cmpfp_i. Move after sse patterns.
926 (cmpfp_i_mixed): Rename from cmpfp_i_sse; use for TARGET_MIX_SSE_I387.
927 (cmpfp_i_sse): Rename from cmpfp_i_sse_only; use for TARGET_SSE_MATH.
928 (cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387): Similarly.
929 (fp_jcc_1_mixed, fp_jcc_1_sse, fp_jcc_1_387): Similarly.
930 (fp_jcc_2_mixed, fp_jcc_2_sse, fp_jcc_2_387): Similarly.
931 (fp_jcc_3_387, fp_jcc_4_387, fp_jcc_5_387, fp_jcc_6_387,
932 fp_jcc_7_387, fp_jcc_8_387): Rename from fp_jcc_N.
933 (movdicc_c_rex64): Rename with '*'.
934 (movsfcc, movdfcc): Add checks for 387 and sse math to condition.
935 (movsfcc_1_sse_min, movsfcc_1_sse_max, movsfcc_1_sse): New.
936 (movsfcc_1_387): Rename from movsfcc_1.
937 (movdfcc_1_sse_min, movdfcc_1_sse_max, movdfcc_1_sse): New.
938 (movdfcc_1, movdfcc_1_rex64): Add check for 387.
939 (sminsf3, smaxsf3, smindf3, smaxdf3): New.
940 (minsf3, minsf, minsf_nonieee, minsf_sse, mindf3, mindf,
941 mindf_nonieee, mindf_sse, maxsf3, maxsf, maxsf_nonieee, maxsf_sse,
942 maxdf3, maxdf, maxdf_nonieee, maxdf_sse, sse_movsfcc, sse_movsfcc_eq,
943 sse_movdfcc, sse_movdfcc_eq, sse_movsfcc_const0_1,
944 sse_movsfcc_const0_2, sse_movsfcc_const0_3, sse_movsfcc_const0_4,
945 sse_movdfcc_const0_1, sse_movdfcc_const0_2, sse_movdfcc_const0_3,
946 sse_movdfcc_const0_4): Remove.
947 * config/i386/i386.c (ix86_expand_fp_movcc): For TARGET_SSE_MATH,
948 recognize min/max early. Update for changed sse cmove patterns.
949 (ix86_split_sse_movcc): New.
950 * config/i386/i386-protos.h: Update.
952 2005-01-13 Steven Bosscher <stevenb@suse.de>
954 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
955 (fix_stmt_v_may_defs): Likewise.
957 2005-01-13 Richard Henderson <rth@redhat.com>
959 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
960 for converting the mask.
962 2005-01-13 David Edelsohn <edelsohn@gnu.org>
964 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
967 2005-01-13 Richard Henderson <rth@redhat.com>
969 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
970 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
971 IX86_BUILTIN_MOVDQ2Q): Remove.
972 (IX86_BUILTIN_VEC_EXT_V4SI): New.
973 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
974 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
976 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
977 (sse2_pextrw): Fix immediate constraint.
978 (sse2_loadq, sse2_loadq_rex64): Remove.
979 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
980 Use __builtin_ia32_vec_ext_<size>.
981 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
983 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
985 * function.c (assign_parm_setup_block): Look inside original
986 entry_parm when inspecting PARALLEL.
988 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
989 Joel Sherrill <joel@oarcorp.com>
992 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
993 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
994 rtems_gxx_recursive_mutex_init.
995 (__gthread_recursive_mutex_lock): New function.
996 (__gthread_recursive_mutex_trylock): Likewise.
997 (__gthread_recursive_mutex_unlock): Likewise.
999 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
1001 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
1003 2005-01-13 Jan Beulich <jbeulich@novell.com>
1005 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
1006 (ashrti3_internal): Indicate output is early clobber. Generate result
1007 into output rather than first input. Use move for low word of output
1008 if shift count is exactly 64.
1009 (lshrti3_internal): Likewise.
1011 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
1015 * reload1.c (delete_output_reload): Don't delete an output reload
1016 if the pseudo lives longer than a single basic block. Adjust and
1017 improve wording of comment.
1019 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
1022 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
1023 of TYPE_SIZE where we mean the number of bytes.
1025 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
1027 * function.c (assign_parm_setup_block): Relax condition on
1028 multi-register optimization.
1030 2005-01-12 Nick Clifton <nickc@redhat.com>
1032 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
1033 an "in_delay_slot" attribute of "no" to prevent them being used in
1034 delay slots. This is forbidden because they might generate
1037 2005-01-12 Alan Modra <amodra@bigpond.net.au>
1040 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
1043 2005-01-12 Nick Clifton <nickc@redhat.com>
1045 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
1046 passed on to GAS) but do not define, as GAS no longer supports or
1047 needs the -m2000 option.
1049 2005-01-11 Roger Sayle <roger@eyesopen.com>
1051 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
1052 type as args of unprototyped builtins aren't automatically promoted.
1053 (fold_builtin_abs): Likewise.
1055 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1057 * Makefile.in: Set a `build-warn' variable.
1059 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
1061 PR tree-optimization/17949
1062 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
1063 (find_interesting_uses_address): Use it.
1065 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
1067 * regrename.c (kill_value): Handle subreg's that won't simplify.
1069 2005-01-11 Richard Henderson <rth@redhat.com>
1072 * config/i386/i386.h (enum ix86_builtins): Move ...
1073 * config/i386/i386.c: ... here.
1074 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
1075 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
1076 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
1077 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
1078 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
1079 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
1080 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
1081 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
1082 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
1083 IX86_BUILTIN_LOADDDUP): Remove.
1084 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
1085 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
1086 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
1087 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
1088 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
1089 (ix86_init_builtins): Make static.
1090 (ix86_init_mmx_sse_builtins): Update for changed builtins.
1091 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
1092 if all the modes match. Otherwise, fake it.
1093 (get_element_number, ix86_expand_vec_init_builtin,
1094 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
1095 (ix86_expand_builtin): Make static. Update for changed builtins.
1096 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
1097 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
1098 (ix86_expand_vector_init_duplicate): New.
1099 (ix86_expand_vector_init_low_nonzero): New.
1100 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
1101 Split out from ix86_expand_vector_init; handle integer modes.
1102 (ix86_expand_vector_init): Use them.
1103 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
1104 * config/i386/i386-protos.h: Update.
1105 * config/i386/predicates.md (reg_or_0_operand): New.
1106 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
1107 (movv2sf_internal): Likewise. And a splitter to match them all.
1108 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
1109 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
1110 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
1111 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
1112 (mmx_pinsrw): Fix operand ordering.
1113 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
1114 rather than sse_loadss expander.
1115 (movv2df splitter): Similarly.
1116 (sse_loadss, sse_loadlss): Remove.
1117 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
1118 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
1119 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
1120 (sse3_movddup): Rename with '*'.
1121 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
1122 (sse2_loadsd): Remove.
1123 (vec_dupv2df_sse3): Rename from sse3_loadddup.
1124 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
1125 (sse2_pinsrw): Fix argument ordering.
1126 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
1127 (sse2_stored): Remove 'r' destination.
1128 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
1129 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
1130 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
1131 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
1132 vec_extractv16qi, vec_initv16qi): New.
1134 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
1135 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
1136 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
1138 (_mm_load_pd, _mm_store_pd): Use plain dereference.
1139 (_mm_load_si128, _mm_store_si128): Likewise.
1140 (_mm_load1_pd): Use _mm_set1_pd.
1141 (_mm_load_sd): Use _mm_set_sd.
1142 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
1143 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
1144 (_mm_set_epi64): Use _mm_set_epi64x.
1145 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
1146 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
1147 _mm_setr_epi8): Use _mm_set_foo form.
1148 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
1149 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
1150 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
1151 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
1152 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
1153 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
1154 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
1155 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
1156 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
1157 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
1158 (_mm_movedup_pd): Use _mm_shuffle_pd.
1159 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
1160 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
1161 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
1162 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
1164 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
1165 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
1166 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
1167 (_mm_store_ps): Use raw dereference.
1168 (_mm_store1_ps): Use _mm_storeu_ps.
1169 (_mm_storer_ps): Use _mm_store_ps.
1170 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
1171 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
1173 2005-01-11 Stan Shebs <shebs@apple.com>
1175 * config/rs6000/rs6000.c (machopic_output_stub): Issue
1176 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
1178 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
1181 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
1182 Skip all DECLs except for FIELD_DECLs.
1184 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
1186 * config/s390/s390.c (override_options): Return error if
1187 -mbackchain, -mpacked-stack and -mhard-float are used together.
1188 (s390_va_start): Remove the backchain && packed-stack special case.
1189 (s390_gimplify_va_arg): Likewise.
1190 * doc/invoke.texi: Remove the ABI incompatibility note.
1192 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
1194 * config/s390/s390.c (struct s390_frame_layout): Remove
1196 (s390_frame_info, s390_emit_prologue): Replace occurrences of
1197 save_backchain_p with TARGET_BACKCHAIN.
1199 2005-01-11 Alan Modra <amodra@bigpond.net.au>
1202 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
1203 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
1204 * function.c (locate_and_pad_parm): Set new field.
1205 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
1206 Tweak where_pad test to include "none". Always set mem align for
1208 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
1209 not sufficient for type.
1210 (assign_parm_setup_block): If stack_parm is zero on entry, always
1211 make a new stack local. Block move old stack parm if necessary
1212 to new aligned stack local.
1213 (assign_parm_setup_stack): Use a block move to handle
1214 potentially misaligned entry_parm.
1215 (assign_parms_unsplit_complex): Specify required alignment when
1216 creating stack local.
1217 * calls.c (compute_argument_addresses): Override alignment of stack
1218 arg calculated from its type with the alignment given by
1219 FUNCTION_ARG_BOUNDARY.
1220 (store_one_arg): Likewise.
1222 2005-01-11 Jan Beulich <jbeulich@novell.com>
1224 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
1225 Change respective itanium_class attribute to ialu.
1226 (shladdp4_internal): New.
1227 * config/ia64/predicates.md (shladd_log2_operand): New.
1229 2005-01-11 Richard Henderson <rth@redhat.com>
1231 * expr.c (store_constructor): Use rtvec_alloc instead of
1232 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
1234 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
1236 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
1238 2005-01-11 Alan Modra <amodra@bigpond.net.au>
1240 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
1241 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
1242 (TARGET_RWRELOC): Define.
1243 (ia64_rwreloc_section_type_flags): Delete.
1244 (ia64_section_type_flags): New function.
1245 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
1246 (TARGET_RWRELOC): Define.
1248 2005-01-10 David Mosberger <davidm@hpl.hp.com>
1251 * config/ia64/ia64.c (process_set): For alloc insn, only call
1252 process_epilogue is !frame_pointer_needed.
1254 2005-01-10 Roger Sayle <roger@eyesopen.com>
1257 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
1258 unary operator and can't be treated as a binary/comparison operator.
1260 2005-01-10 Richard Henderson <rth@redhat.com>
1262 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
1263 (ix86_return_in_memory): Likewise.
1264 (function_arg_advance): Likewise. Mirror structure in function_arg
1265 for choosing register to advance.
1267 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
1269 * tree-vectorizer.c, tree.def: Fix comment typos.
1271 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
1273 * config/s390/s390.c (struct processor_costs): 4 new fields:
1274 ddbr, ddr, debr, der.
1275 (s390_rtx_costs): More precise handling of divide instructions.
1277 2005-01-10 David Edelsohn <edelsohn@gnu.org>
1279 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
1281 2005-01-10 Jan Beulich <jbeulich@novell.com>
1283 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
1284 names resulting from -ffunction-sections/-fdata-sections and linkonce
1286 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
1287 based on the section name. Rearrange the section name comparison logic
1288 slightly so that each section name is compared against at most once.
1290 2005-01-10 Ben Elliston <bje@au.ibm.com>
1292 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
1293 targets for which -fPIC reduces limitations on the GOT size.
1295 2005-01-09 Falk Hueffner <falk@debian.org>
1297 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
1300 2005-01-09 David Edelsohn <edelsohn@gnu.org>
1303 * collect2.c (main): Set aixrtl_flag for -brtl option.
1304 (resolve_lib_name): Search for .so file extension before .a
1307 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
1309 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
1310 (vect_analyze_data_refs_alignment): Add dump prints.
1312 2005-01-09 Ira Rosen <irar@il.ibm.com>
1314 * tree-vectorizer.c (vect_analyze_offset_expr): Use
1315 expr_invariant_in_loop_p.
1316 Initialize outputs first thing in the function.
1317 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
1318 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
1319 (vect_analyze_pointer_ref_access): Check that the initial condition of
1320 the access function is loop invariant.
1322 2005-01-09 Richard Henderson <rth@redhat.com>
1324 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
1325 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
1326 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
1327 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
1328 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
1329 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
1330 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
1331 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
1332 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
1333 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
1334 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
1335 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
1336 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
1337 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
1338 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
1339 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
1340 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
1341 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
1342 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
1343 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
1344 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
1345 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
1346 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
1348 (mmx_clrdi, pavgusb): Remove.
1349 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
1350 with leading sse_ prefix.
1351 * config/i386/sse.md: Receive them.
1352 * config/i386/mmx.md: New file.
1353 (MMXMODE12, MMXMODE24, mmxvecsize): New.
1354 (subrv2sf3): Turn into expander for normal subtraction.
1355 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
1356 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
1357 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
1358 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
1359 commutative; use ix86_binary_operator_ok.
1360 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
1361 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
1362 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
1363 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
1364 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
1365 Macroize from existing patterns; use ix86_binary_operator_ok.
1366 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
1367 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
1368 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
1369 with vec_select+vec_concat.
1370 (mmx_pshufw, mmx_pshufw_1): Likewise.
1371 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
1372 (mmx_uavgv4hi3): Model correctly.
1373 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
1374 _mm_xor_si64): Remove casts.
1376 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
1378 PR tree-optimization/19224
1379 * tree-scalar-evolution.c (get_instantiated_value,
1380 set_instantiated_value): New functions.
1381 (instantiate_parameters_1): Cache the results.
1382 (instantiate_parameters, resolve_mixers): Initialize and free
1385 2005-01-08 David Edelsohn <edelsohn@gnu.org>
1387 * config/i386/i386.md (addhi_4): Correct reference in comment.
1390 2005-01-08 Richard Henderson <rth@redhat.com>
1392 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
1393 (_mm_cvtsi128_si64x): Likewise.
1394 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
1395 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
1396 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
1397 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
1398 * config/i386/i386-protos.h: Update.
1399 * config/i386/i386.c (print_operand): Add 'H'.
1400 (ix86_fixup_binary_operands): Split out from ...
1401 (ix86_expand_binary_operator): ... here.
1402 (ix86_fixup_binary_operands_no_copy): New.
1403 (ix86_expand_fp_absneg_operator): Handle vector mode results.
1404 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
1405 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
1406 (safe_vector_operand): Use CONST0_RTX.
1407 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
1408 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
1409 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
1410 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
1412 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
1413 (IX86_BUILTIN_CMPNGESS): New.
1414 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
1415 (attr type): Add sselog1.
1416 (attr unit, attr memory): Handle it.
1417 (movti, movti_internal, movti_rex64): Move near other integer moves.
1418 (movtf, movtf_internal): Move near other fp moves.
1419 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
1420 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
1421 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
1422 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
1423 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
1424 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
1425 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
1426 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
1427 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
1428 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
1429 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
1430 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
1431 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
1432 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
1433 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
1434 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
1435 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
1436 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
1437 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
1438 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
1439 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
1440 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
1441 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
1442 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
1443 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
1444 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
1445 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
1446 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
1447 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
1448 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
1449 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
1450 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
1451 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
1452 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
1453 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
1454 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
1455 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
1456 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
1457 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
1458 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
1459 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
1460 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
1461 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
1462 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
1464 (SSEPUSH, push<SSEPUSH>): Remove.
1465 (MMXPUSH, push<MMXPUSH>): Remove.
1466 (sse_movaps, sse_movaps_1, sse_movups): Remove.
1467 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
1468 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
1469 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
1470 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
1471 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
1472 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
1473 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
1474 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
1475 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
1476 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
1477 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
1478 (ppro_sse_log_V4SF): Handle sselog1.
1479 * config/i386/predicates.md (const_0_to_1_operand): New.
1480 (const_0_to_255_mul_8_operand): New.
1481 (const_1_to_31_operand): Rename from const_int_1_31_operand.
1482 (const_2_to_3_operand, const_4_to_7_operand): New.
1483 * config/i386/sse.md: New file.
1484 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
1485 (sse_movups): Rename from sse_movups_1.
1486 (sse_loadlss): Rename from sse_loadss_1.
1487 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
1489 (negv4sf2): Use ix86_expand_fp_absneg_operator.
1490 (absv4sf2, negv2df, absv2df): New.
1491 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
1492 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
1493 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
1494 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
1495 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
1496 (sse3_addsubv4sf3): Model correctly.
1497 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
1498 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
1499 (sse_movhlps): Model with vec_select+vec_concat.
1500 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
1501 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
1502 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
1503 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
1504 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
1505 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
1506 sse2_pshufhw_1): Likewise.
1507 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
1508 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
1509 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
1510 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
1511 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
1512 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
1513 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
1514 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
1515 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
1516 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
1517 sminv8hi3): Mark commutative
1518 operands. Use ix86_binary_operator_ok.
1519 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
1520 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
1521 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
1522 sse2_punpcklqdq): Allow operand2 in memory.
1523 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
1524 sse2_movsd): Add memory alternatives.
1525 (sse_storelps): Turn expander into an insn; split after reload.
1526 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
1527 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
1529 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
1531 * configure.ac (DWARF-2 debug_line): Use objdump.
1532 * configure: Regenerate.
1534 2005-01-08 Jeff Law <law@redhat.com>
1535 Diego Novillo <dnovillo@redhat.com>
1537 PR tree-optimization/18241
1538 * tree-nrv.c (tree_nrv): Ignore volatile return values.
1539 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
1540 statements with volatile operands.
1541 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
1542 operands after marking a statement with has_volatile_ops.
1544 2005-01-08 Roger Sayle <roger@eyesopen.com>
1546 * tree.c (int_fits_type_p): Always honor integer constant
1547 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
1549 2005-01-08 Roger Sayle <roger@eyesopen.com>
1551 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
1552 the latent bug in rtl_delete_block has been resolved.
1554 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
1556 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
1557 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
1559 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
1561 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
1562 library as @multilib_dir@/$(SHLIB_SONAME).
1564 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
1566 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
1567 * configure: Regenerate.
1569 2005-01-07 Jakub Jelinek <jakub@redhat.com>
1571 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
1574 2005-01-07 David Edelsohn <edelsohn@gnu.org>
1577 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
1578 Convert non-word aligned offset address using ld/std into
1581 2005-01-07 Richard Henderson <rth@redhat.com>
1583 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
1585 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
1587 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
1589 2005-01-07 Jakub Jelinek <jakub@redhat.com>
1591 PR tree-optimization/19283
1592 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
1593 is not shorter than the original type.
1595 PR rtl-optimization/19012
1596 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
1598 PR rtl-optimization/18861
1599 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
1600 (FULL_STATE): Define.
1601 (mark_tablejump_edge): New function.
1602 (purge_dead_tablejump_edges): New function.
1603 (find_bb_boundaries): Use it.
1605 PR tree-optimization/18828
1606 * builtins.c (expand_builtin_next_arg): Remove argument and all
1607 the argument checking.
1608 (expand_builtin): Adjust caller.
1609 (expand_builtin_va_start): Likewise. Remove error for too many
1611 (fold_builtin_next_arg): Issue error for too many arguments.
1612 After checking arguments, replace them with magic arguments that
1613 prevent further checking of the args.
1615 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1617 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
1619 2005-01-06 Roger Sayle <roger@eyesopen.com>
1622 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
1623 and %{.S:...} (and their negative variants) to test whether the
1624 input file is assembler or pre-processed-assembler independent of
1625 the actual filename extension.
1627 2005-01-06 Roger Sayle <roger@eyesopen.com>
1629 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
1630 of sign or zero extended values.
1632 2005-01-06 Geoffrey Keating <geoffk@apple.com>
1634 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
1635 for IBM long double format correctly.
1637 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
1639 Fix PR tree-optimization/18792
1641 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
1642 to first_loop_depth, and use loop depth instead of loop number.
1643 (build_classic_dir_vector): Ditto.
1644 (compute_data_dependences_for_loop): Use depth, not loop number.
1645 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
1646 number. Pass in loops, instead of loop numbers.
1647 (gather_interchange_stats): Ditto.
1648 (linear_transform_loops): Ditto.
1650 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
1653 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
1654 biv_fits_mode_p, extension_within_bounds_p): New functions.
1655 (check_ext_dependent_givs): Use them.
1657 2005-01-06 Roger Sayle <roger@eyesopen.com>
1659 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
1660 more than one barrier, in which case we should delete them all.
1662 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
1664 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
1666 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
1668 Fix PR middle-end/19286
1670 * dwarf2out.c (gen_subprogram_die): If we've already tried to
1671 output this subprogram, simply ignore this attempt to do it again.
1672 (add_abstract_origin_attribute): Don't abort trying to add the abstract
1673 origin attribute if it's not possible.
1674 (gen_block_die): Don't ignore subblocks of "unused" blocks.
1675 (decls_for_scope): Ditto.
1676 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
1677 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
1679 2005-01-05 Richard Henderson <rth@redhat.com>
1682 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
1683 (bdesc_2arg): Use it.
1684 (ix86_expand_binop_builtin): Force operands into registers
1686 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
1687 ix86_expand_sse_compare, ix86_expand_sse_comi,
1688 ix86_expand_builtin): Likewise.
1690 2005-01-05 Richard Henderson <rth@redhat.com>
1692 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
1694 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
1695 (mulv4hi3): Set itanium_class mmmul.
1696 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
1697 (fpack): Rename from fpack_sfsf.
1699 2005-01-05 Richard Henderson <rth@redhat.com>
1702 * reload1.c (do_input_reload): Restrict the optimization deleteing
1703 a previous output reload to RELOAD_FOR_INPUT.
1705 2005-01-05 Steven Bosscher <stevenb@suse.de>
1707 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
1708 comment that we fall through after case.
1709 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
1710 (distribute_links): Likewise.
1711 * cse.c (cse_insn): Likewise.
1712 * cselib.c (cselib_invalidate_mem): Likewise.
1713 * df.c: Update comments at the top of the file.
1714 (read_modify_subreg_p): Update comments here too.
1715 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
1716 * flow.c (mark_set_1): Likewise.
1717 (mark_used_regs): Likewise.
1718 * gcse.c (mems_conflict_for_gcse_p): Likewise.
1719 (canon_list_insert): Likewise.
1720 (mark_set): Likewise.
1721 (try_replace_reg): Likewise.
1722 (store_killed_in_insn): Likewise.
1723 * loop.c (count_one_set): Likewise.
1724 (basic_induction_var): Likewise.
1725 * postreload-gcse.c (find_mem_conflicts): Likewise.
1726 * postreload.c (reload_combine_note_store): Likewise.
1727 (move2add_note_store): Likewise.
1728 * reload.c (find_equiv_reg): Likewise.
1729 (mark_referenced_resources): Likewise.
1730 * rtlanal.c (set_noop_p): Likewise.
1731 (note_stores): Likewise.
1732 (note_uses): Likewise.
1733 * sched-deps.c (sched_analyze_1): Likewise.
1734 * sched-rgn.c (check_live_1): Likewise.
1735 (update_live_1): Likewise.
1736 * config/i860/i860.c: Likewise.
1738 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
1740 (ZERO_EXTRACT): Mention that this one can be an lvalue.
1742 * doc/rtl.texi: Update documentation for bit-fields and SET.
1744 2005-01-05 Roger Sayle <roger@eyesopen.com>
1746 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
1748 2005-01-05 Jan Hubicka <jh@suse.cz>
1749 Richard Henderson <rth@redhat.com>
1752 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
1755 2005-01-05 Richard Henderson <rth@redhat.com>
1758 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
1759 (sse_shufps): Change operand 3 to const_int_operand.
1760 (sse2_storelps): Fix typo in template.
1761 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
1762 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
1763 (ix86_expand_builtin): Likewise.
1765 2005-01-05 Julian Brown <julian@codesourcery.com>
1767 * config/arm/arm.c (arm_return_in_memory): Treat complex types
1768 as aggregates for AAPCS ABIs.
1770 2005-01-05 Stan Shebs <shebs@apple.com>
1772 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
1773 Mach-O getters if ppc64.
1774 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
1776 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
1777 signed four-byte field for global code case.
1778 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
1780 2005-01-05 Roger Sayle <roger@eyesopen.com>
1783 * c-common.c: Include real.h.
1784 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
1785 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
1786 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
1787 * Makefile.in (c-common.o): Update dependencies.
1789 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1791 * c-parse.in (asm_string): Add trailing semicolon.
1793 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1795 * c-parse.in (asm_string): New. Don't allow wide strings in
1797 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
1798 asm_string instead of STRING.
1800 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1802 * c-typeck.c (constructor_no_implicit): Remove.
1803 (set_designator, process_init_element): Don't check
1804 constructor_no_implicit.
1806 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
1807 Kaz Kojima <kkojima@gcc.gnu.org>
1810 * lcm.c (create_pre_exit): New.
1811 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
1812 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
1814 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
1816 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
1817 -msoft-float in target_float_switch.
1818 * arm.c (arm_override_options): Fix processing of target_float_switch.
1820 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
1822 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
1825 2004-01-05 Uros Bizjak <uros@kss-loka.si>
1827 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
1828 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
1830 2005-01-04 Roger Sayle <roger@eyesopen.com>
1832 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
1833 if deleting the then-block allows the test-block to fallthru to the
1836 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
1839 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
1840 after the full declaration if the are in two different TUs.
1842 2005-01-04 Richard Henderson <rth@redhat.com>
1845 * tree-sra.c (generate_one_element_init): Just
1846 call gimplify_and_add.
1847 (generate_element_init): Record the
1848 new referenced variables and mark them for renaming
1849 and split out to ...
1850 (generate_element_init_1): This.
1851 (scalarize_init): Don't call push_gimplify_context/
1852 pop_gimplify_context.
1854 2005-01-04 Geoffrey Keating <geoffk@apple.com>
1856 * toplev.c (get_src_pwd): Handle failure of getpwd().
1858 2005-01-04 Roger Sayle <roger@eyesopen.com>
1860 * fold-const.c (fold_single_bit_test): Delete unreachable handling
1862 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
1863 all the cases handled by it are inverted by invert_truthvalue.
1865 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
1867 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
1868 (save_register_windowsi): Likewise.
1870 2005-01-04 Richard Henderson <rth@redhat.com>
1872 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
1874 2005-01-04 Uros Bizjak <uros@kss-loka.si>
1877 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
1878 any simplifications of vector mode comparison operators.
1879 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
1881 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
1882 Devang Patel <dpatel@apple.com>
1884 PR tree-optimization/18308
1885 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
1886 the operands before creating a new expression.
1887 * dojump.c (do_jump): Make drop_through_label available
1888 for all cases. Add expansion of COND_EXPR.
1890 2005-01-04 Ira Rosen <irar@il.ibm.com>
1892 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
1893 INTEGER_CST instead of TREE_CONSTANT.
1894 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
1895 instead of TREE_CONSTANT.
1896 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
1897 instead of TREE_CONSTANT.
1899 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
1901 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
1903 2005-01-04 Uros Bizjak <uros@kss-loka.si>
1906 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
1907 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
1909 2005-01-03 Richard Henderson <rth@redhat.com>
1911 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
1912 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
1913 fold_convert_const_real_from_real): Split out from ...
1914 (fold_convert_const): ... here.
1916 2005-01-03 Richard Henderson <rth@redhat.com>
1919 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
1920 (mov<MMXMODEI>_internal): Likewise.
1921 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
1922 Add V2SF case; use it for SSE1; don't use TI.
1923 (movdf_integer): Likewise.
1924 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
1926 2005-01-03 Ira Rosen <irar@il.ibm.com>
1928 * tree-vectorizer.c (vect_strip_conversions): New function.
1929 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
1930 check for binary class.
1932 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
1936 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
1938 * gimple-low.c (mark_blocks_with_used_vars): New function.
1939 (mark_blocks_with_used_subblocks): Ditto.
1940 (mark_used_blocks): Ditto.
1941 (pass_mark_used_blocks): New pass.
1942 * tree-inline.c: Include debug.h.
1943 (expand_call_inline): Call outlining_inline_function here.
1944 * tree-optimize.c (init_tree_optimization_passes): Add
1945 pass_mark_used_blocks.
1946 * tree-pass.h (pass_mark_used_blocks): New.
1947 * Makefile.in (tree-inline.o): Add debug.h dependency.
1949 2005-01-03 Geoffrey Keating <geoffk@apple.com>
1951 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
1952 VAR_DECLs to have weak_import attribute.
1954 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
1957 2004-01-03 Steven Bosscher <stevenb@suse.de>
1959 * passes.c (rest_of_compilation): Don't run regmove if only
1960 flag_expensive_optimizations, require flag_regmove instead.
1962 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
1964 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
1966 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
1967 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1968 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
1969 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1971 2005-01-03 Richard Henderson <rth@redhat.com>
1973 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
1974 (ia64_const_ok_for_letter_p): New.
1975 (ia64_const_double_ok_for_letter_p): New.
1976 (ia64_extra_constraint): New.
1977 (ia64_expand_vecint_compare): New.
1978 (ia64_expand_vcondu_v2si): New.
1979 (ia64_expand_vecint_cmov): New.
1980 (ia64_expand_vecint_minmax): New.
1981 (ia64_print_operand): Add 'v'.
1982 (ia64_preferred_reload_class): New.
1983 (ia64_vector_mode_supported_p): New.
1984 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
1985 (PREFERRED_RELOAD_CLASS): Move to function.
1986 (CONST_OK_FOR_LETTER_P): Move to function.
1987 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
1988 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
1989 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
1990 (EXTRA_CONSTRAINT): Move to function.
1991 * config/ia64/ia64.md: Include vect.md.
1992 (itanium_class): Add mmalua.
1994 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
1996 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
1997 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
1999 (const_int_2bit_operand): New.
2000 (fr_reg_or_0_operand): New.
2001 * config/ia64/ia64-modes.def: Add vector modes.
2002 * config/ia64/ia64-protos.h: Update.
2003 * config/ia64/vect.md: New file.
2005 2005-01-03 Richard Henderson <rth@redhat.com>
2007 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
2009 2005-01-03 Uros Bizjak <uros@kss-loka.si>
2012 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
2014 (log1pdf2): Change mode of operands[0,1] to DFmode.
2016 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
2018 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
2019 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
2020 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
2021 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
2022 SPARC_RELAXED_ORDERING.
2024 2005-01-03 Richard Henderson <rth@redhat.com>
2025 Uros Bizjak <uros@kss-loka.si>
2028 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
2029 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
2030 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
2032 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
2033 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
2034 const_pow2_1_to_128_operand): New predicates.
2036 2005-01-02 Greg McGary <greg@mcgary.org>
2038 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
2040 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
2041 object. [ptr_type]: Remove unused variable.
2042 Remember nearest addressable array-element or record-component
2043 when checking bitfield components. Tolerate empty BIND_EXPR.
2045 2005-01-01 Richard Henderson <rth@redhat.com>
2047 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
2048 that don't narrow the value. Fail for other conversions.
2050 2005-01-01 Richard Henderson <rth@redhat.com>
2053 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
2054 * c-lang.c (finish_file): Don't do it here.
2055 * objc/objc-act.c (objc_finish_file): Likewise.
2057 * cgraph.c (decl_assembler_name_equal): New.
2058 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
2059 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
2060 * cgraph.h (struct cgraph_varpool_node): Add next.
2061 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
2062 * varasm.c (assemble_alias): Mark the target as needed.
2064 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
2067 * function.c (get_arg_pointer_save_area): Use entry_of_function
2068 instead of get_insns.
2070 2005-01-01 Roger Sayle <roger@eyesopen.com>
2071 Andrew Pinski <pinskia@physics.uc.edu>
2072 James E. Wilson <wilson@specifixinc.com>
2074 PR rtl-optimization/12092
2075 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
2078 2005-01-01 Roger Sayle <roger@eyesopen.com>
2079 Olivier Hainque <hainque@act-europe.fr>
2081 * tree.c (int_fits_type_p): A narrower type always fits in a
2082 wider one, except for negative values into unsigned types.
2084 2005-01-01 Roger Sayle <roger@eyesopen.com>
2086 * tree.c (int_fits_type_p): Compare the result of force_fit_type
2087 with the original constant rather than require TREE_OVERFLOW.
2089 2005-01-01 Steven Bosscher <stevenb@suse.de>
2092 * c-decl.c (finish_function): If compiling C99, annotate the
2093 compiler generated return with the current file name and line 0.
2094 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
2095 the source line is greater than 0.
2096 (remove_bb): Likewise.
2098 See ChangeLog.12 for earlier changes.