1 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
3 * cse.c (max_reg, max_insn_uid): Remove.
4 (cse_main): Don't access max_reg or max_insn_uid.
6 * c-common.c (explicit_flag_signed_bitfields,
7 lang_statement_code_p, lang_expand_function_end): Remove.
8 * c-common.h: Remove the corresponding declarations.
9 * c-opts.c (c_common_handle_option): Remove all write access
10 to explicit_flag_signed_bitfields.
12 2005-01-23 Roger Sayle <roger@eyesopen.com>
13 Eric Botcazou <ebotcazou@libertysurf.fr>
15 * combine.c (simplify_logical): Only simplify logical expressions
16 of the form ior(and(x,y),z) by the inverse distributive law if the
17 result is cheaper than the original.
19 2005-01-23 Kazu Hirata <kazu@cs.umass.edu>
21 * alias.c, c-common.h, c-incpath.c, c-incpath.h, expr.c,
22 fold-const.c, gimplify.c, params.h, tree-data-ref.c,
23 tree-if-conv.c, tree-nested.c, tree-outof-ssa.c,
24 tree-ssa-dom.c, tree-vectorizer.c, tree.def, config/darwin.c,
25 config/freebsd-spec.h, config/arm/arm.h,
26 config/h8300/h8300.md, config/i386/i386.md,
27 config/i386/predicates.md, config/i386/sse.md,
28 config/ia64/ia64.c, config/ip2k/ip2k.c, config/s390/s390.c,
29 config/vax/vax.md: Fix comment typos. Follow spelling
32 2005-01-23 Steven Bosscher <stevenb@suse.de>
34 * gengtype.c (walk_type): Produce `not equal to' compares for loop
35 tests, instead of `less than'.
37 2005-01-23 Richard Sandiford <rsandifo@redhat.com>
39 * function.c (assign_parm_setup_block): When creating a new stack slot
40 for a parameter, get its alignment from the parameter's DECL_ALIGN
41 rather than the type's TYPE_ALIGN. Make sure that the parameter
42 is at least word aligned.
44 2005-01-22 Roger Sayle <roger@eyesopen.com>
47 * config/avr/avr.c (avr_hard_regno_mode_ok): Rewrite.
49 2005-01-22 Richard Henderson <rth@redhat.com>
52 * config/i386/i386.md (movsfcc_1_sse_max): Use nonimmediate_operand
53 in both compare operands.
54 (movdfcc_1_sse_max): Likewise.
55 (movsfcc_1_sse): Likewise. Add earlyclobber for scratch.
56 (movdfcc_1_sse): Likewise.
57 * config/i386/i386.c (ix86_split_sse_movcc): Emit copies into the
58 scratch register as needed.
60 2005-01-22 Richard Henderson <rth@redhat.com>
62 * config/i386/i386.md (smaxsf3): Fix mnemonic typo.
64 * config/i386/i386.c (ix86_prepare_fp_compare_args): Fix is_sse test.
66 2005-01-22 Richard Henderson <rth@redhat.com>
68 * genopinit.c (optabs): Use smin/smax for floating point too.
69 * doc/md.texi: Update to match. Clarify that floating point
70 results are undefined for +0/-0 and NaN.
71 * doc/rtl.texi: Likewise.
72 * rtl.def (SMIN, SMAX): Likewise
73 * tree.def (MIN_EXPR, MAX_EXPR): Likewise.
75 * config/alpha/alpha.md (smaxdf3, smindf3, smaxsf3, sminsf3): Add
76 leading 's' to the name.
77 * config/ia64/ia64.md (smaxsf3, sminsf3, smaxdf3, smindf3,
78 smaxxf3, sminxf3): Likewise.
79 * config/rs6000/rs6000.md (smaxdf3, smindf3, smaxsf3, sminsf3):
82 2005-01-22 Paul Brook <paul@codesourcery.com>
84 * config/arm/arm.md: Use "Uy" constraint for wcgr load/stores.
86 2005-01-21 Mark Dettinger <dettinge@de.ibm.com>
88 * config/s390/s390.c (struct processor_costs): New fields
89 dlgr, dlr, dr, dsgfr, dsgr.
90 (z900_cost, z990_cost): Values for new fields.
91 (s390_rtx_costs): New cases MEM und COMPARE in switch
92 statement. Modified handling of SIGN_EXTEND, ZERO_EXTEND,
95 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
97 * config/s390/s390.md ("doloop_si64"): Reload input value directly
98 into the register being decremented.
99 ("doloop_si31", "doloop_di"): Likewise.
100 ("*doloop_si_long"): Adapt pattern.
102 2005-01-21 Ulrich Weigand <uweigand@de.ibm.com>
104 * config/s390/s390.h (HARD_REGNO_NREGS): Fix computation for
106 (CLASS_MAX_NREGS): Likewise.
108 2005-01-21 Daniel Berlin <dberlin@dberlin.org>
110 * doc/tree-ssa.texi (Statement Operands): Add example for new
111 must-def macro. Note deprecation of old operands interface.
113 2005-01-22 Richard Sandiford <rsandifo@redhat.com>
115 PR tree-optimization/19484
116 * tree-cfg.c (remove_fallthru_edge): New function.
117 (cleanup_control_flow): Remove fallthru edges from calls that are
118 now known not to return.
120 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
122 Fix PR tree-optimization/19038
123 * tree-ssa-dom.c (cprop_operand): Don't replace loop invaeriant
124 copies with loop variant ones.
126 2005-01-22 Kazu Hirata <kazu@cs.umass.edu>
128 * cfganal.c, real.h, reorg.c, timevar.def, tree-ssa-ccp.c,
129 config/alpha/alpha-protos.h, config/alpha/alpha.h,
130 config/alpha/alpha.md, config/alpha/predicates.md,
131 config/sparc/freebsd.h, config/sparc/netbsd-elf.h,
132 config/sparc/sol2.h: Update copyright.
134 * tree-cfg.c (remove_forwarder_block_with_phi): Look at the
135 first label to see if it is a nonlocal label.
137 2005-01-22 David Edelsohn <edelsohn@gnu.org>
138 Andrew Pinski <pinskia@physics.uc.edu>
141 * config/rs6000/rs6000.c (rs6000_va_start): Saturate n_gpr at
142 maximum number of GPRs. Saturate n_fpr at maximum number of FPRs.
144 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
146 * config/rs6000/rtems.h: Update copyright.
148 2005-01-22 Ralf Corsepius <ralf.corsepius@rtems.org>
151 * config/rs6000/rtems.h: Resurrect cpp_os_rtems_spec from gcc < 3.4.
152 (CPP_OS_RTEMS_SPEC): New (From gcc-3.3's config/rs6000/sys4.h).
153 (SUBSUBTARGET_EXTRA_SPECS): Use CPP_OS_RTEMS_SPEC.
155 2005-01-22 Volker Reichelt <reichelt@igpm.rwth-aachen.de>
158 * c-typeck.c (convert_arguments): Check for error_mark_node.
160 2005-01-21 Ian Lance Taylor <ian@airs.com>
162 PR tree-optimization/13000
163 * tree-inline.c: Include "tree-flow.h".
164 (expand_call_inline): If warn_return_type, warn if non-void inline
165 function falls through.
166 * tree-cfg.c (execute_warn_function_return): Don't warn about
167 control reaching end if TREE_NO_WARNING is set. Set
169 * gimple-low.c (block_may_fallthru): Don't assume that SWITCH_EXPR
171 * gimplify.c (shortcut_cond_expr): Don't emit a jump over the else
172 branch if we don't need one.
173 * c-typeck.c: Include "tree-flow.h"
174 (c_finish_bc_stmt): Don't add a goto if the current statement
175 list doesn't fall through to the current point.
177 2005-01-21 Roger Sayle <roger@eyesopen.com>
179 PR rtl-optimization/576
180 * real.c (real_arithmetic): Change return type from void to bool
181 to return an indication that the result may be inexact.
182 * real.h (real_arithmeric): Update prototype.
183 * fold-const.c (const_binop): Don't constant fold floating
184 point expressions when the user specifies -frounding-math and
185 the result may depend upon the run-time rounding mode.
186 (fold_convert_const_real_from_real): Clean-up.
187 (fold_initializer): Ignore flag_rounding_math for initializers.
188 * simplify-rtx.c (simplify_binary_operation): Likewise, don't
189 constant fold FP operations with flag_rounding_math if the
190 result may depend upon the run-time rounding mode.
192 2005-01-21 Tom Tromey <tromey@redhat.com>
194 * c-cppbuiltin.c (define__GNUC__): Correct assertion.
196 2005-01-21 Kazu Hirata <kazu@cs.umass.edu>
198 * tree-cfg.c: Fix comment typos.
200 2005-01-21 J"orn Rennecke <joern.rennecke@st.com>
202 * passes.c (rest_of_handle_flow2): Close / open the flow2 dump file
203 around the call to rest_of_handle_branch_target_load_optimize.
205 2005-01-21 Eric Botcazou <ebotcazou@libertysurf.fr>
207 * config.gcc (sparc-*-netbsdelf*): Include dbxelf.h
208 (sparc64-*-openbsd*): Likewise.
209 (sparclite-*-elf*): Include tm-dwarf2.h.
210 (sparc86x-*-elf*): Likewise.
211 (sparc64-*-elf*): Likewise.
212 * config/sparc/linux64.h (DWARF2_DEBUGGING_INFO): Delete.
213 (DBX_DEBUGGING_INFO): Likewise.
214 * config/sparc/liteelf.h (DWARF2_DEBUGGING_INFO): Likewise.
215 (PREFERRED_DEBUGGING_TYPE): Likewise.
216 * config/sparc/netbsd-elf.h (PREFERRED_DEBUGGING_TYPE): Likewise.
217 * config/sparc/sp64-elf.h (DBX_DEBUGGING_INFO): Likewise.
218 (PREFERRED_DEBUGGING_TYPE): Likewise.
219 * config/sparc/sp86x-elf.h (DWARF2_DEBUGGING_INFO): Likewise.
220 (PREFERRED_DEBUGGING_TYPE): Likewise.
221 * config/sparc/sparc.h (DBX_DEBUGGING_INFO): Likewise.
223 2005-01-21 Hans-Peter Nilsson <hp@bitrange.com>
226 * combine.c (combine_simplify_rtx): Revert change of 2004-12-31.
228 2005-01-20 Paul Brook <paul@codesourcery.com>
230 * doc/rtl.texi: Document value extension requirements for CONST_INT.
232 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
234 * tree-cfg.c (tree_verify_flow_info): Check that a nonlocal
235 label is first in a sequence of labels.
237 * tree-cfg.c (tree_verify_flow_info): Fix a typo.
239 2005-01-20 Janis Johnson <janis187@us.ibm.com>
240 Giovanni Bajo <giovannibajo@gcc.gnu.org>
242 * doc/sourcebuild.texi (Test Directives): New.
244 2005-01-20 Roger Sayle <roger@eyesopen.com>
246 * real.c (real_floor): Don't leave the result uninitialized when
247 mode is VOIDMode, but return the unrounded intermediate value.
248 (real_ceil): Likewise.
250 2005-01-20 Paul Brook <paul@codesourcery.com>
252 * config/arm/arm.md (insv): Use gen_int_mode.
254 2005-01-20 Eric Botcazou <ebotcazou@libertysurf.fr>
256 * config/sparc/freebsd.h (ASM_OUTPUT_INTERNAL_LABELREF): Delete.
257 * config/sparc/linux.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
258 * config/sparc/linux64.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
259 * config/sparc/netbsd-elf.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
260 * config/sparc/sol2.h (ASM_OUTPUT_INTERNAL_LABELREF): Likewise.
262 2005-01-20 Daniel Berlin <dberlin@dberlin.org>
265 * dwarf2out.c (concat_loc_descriptor): We don't know we can use
266 fbreg, so we have to assume we can't.
267 (loc_descriptor_from_tree_1): Ditto.
268 (containing_function_has_frame_base): New function.
269 (add_location_or_const_value_attribute): Use it.
270 Also try to generate a frame_base from a single element location
273 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
275 PR tree-optimization/15349
276 * timevar.def (TV_TREE_MERGE_PHI): New.
277 * tree-cfg.c (tree_forwarder_block_p): Add a new argument
279 (remove_forwarder_block, cleanup_forwarder_blocks): Adjust the
280 calls to tree_forwarder_block_p.
281 (remove_forwarder_block_with_phi, merge_phi_nodes,
282 gate_merge_phi, pass_merge_phi): New.
283 * tree-optimize.c (init_tree_optimization_passes): Add
285 * tree-pass.h: Add an extern for pass_merge_phi;
287 2005-01-20 Richard Henderson <rth@redhat.com>
290 * config/i386/emmintrin.h (_mm_castpd_ps, _mm_castpd_si128): New.
291 (_mm_castps_pd, _mm_castps_si128): New.
292 (_mm_castsi128_ps, _mm_castsi128_pd): New.
294 2005-01-20 Richard Henderson <rth@redhat.com>
297 * config/i386/mmintrin.h (_mm_cvtsi32_si64): Use
298 __builtin_ia32_vec_init_v2si.
299 (_mm_cvtsi64_si32): Use __builtin_ia32_vec_ext_v2si.
300 * config/i386/i386.c (IX86_BUILTIN_VEC_EXT_V2SI): New.
301 (ix86_init_mmx_sse_builtins): Create it.
302 (ix86_expand_builtin): Expand it.
303 (ix86_expand_vector_set): Handle V2SFmode and V2SImode.
304 * config/i386/mmx.md (vec_extractv2sf_0, vec_extractv2sf_1): New.
305 (vec_extractv2si_0, vec_extractv2si_1): New.
307 2005-01-20 Richard Henderson <rth@redhat.com>
310 * config/i386/i386.c (ix86_expand_push): New.
311 * config/i386/mmx.md (push<MMXMODE>1): New.
312 * config/i386/sse.md (push<SSEMODE>1): New.
313 * config/i386/i386-protos.h: Update.
315 2005-01-20 Kazu Hirata <kazu@cs.umass.edu>
317 * params.def, params.h, predict.c, tree-ssa-loop-im.c: Update
320 2005-01-20 Richard Henderson <rth@redhat.com>
322 * rtl.def (CONST_VECTOR): Use RTX_CONST_OBJ.
323 * rtl.h (CONSTANT_P): Don't special case CONST_VECTOR.
325 2005-01-19 Richard Henderson <rth@redhat.com>
328 * config/i386/i386.c (ix86_expand_vector_move_misalign): Convert
329 to V4SFmode in SSE1 fallback load path.
331 2005-01-19 Richard Henderson <rth@redhat.com>
333 * config/i386/i386.c (ix86_expand_vector_init_one_var): Fix typo
334 in QImode expansion to ix86_expand_vector_set.
335 (ix86_expand_vector_init_general): Fix typo in V8HImode recursive call.
337 2005-01-19 Richard Henderson <rth@redhat.com>
340 * config/i386/i386.c (ix86_preferred_reload_class): Return a proper
341 subclass of the input class.
342 (ix86_secondary_memory_needed): Always true for cross-MMX classes.
343 Always true for cross-SSE1 classes. Rationalize conditionals.
344 * config/i386/i386.h (SSE_CLASS_P, MMX_CLASS_P): Use straight equality.
345 * config/i386/i386.md (movsi_1): Add MMX/SSE zeros. Fix alternatives
346 for SSE1. Don't check TARGET_INTER_UNIT_MOVES.
347 (movdi_2): Add MMX/SSE zeros.
348 (movdi_1_rex64): Likewise. Don't check TARGET_INTER_UNIT_MOVES.
349 (movsf_1): Don't check TARGET_INTER_UNIT_MOVES.
350 (zero_extendsidi2_32, zero_extendsidi2_rex64): Likewise.
351 (movsi_1_nointernunit, movdi_1_rex64_nointerunit): Remove.
352 (movsf_1_nointerunit, zero_extendsidi2_32_1): Remove.
353 (zero_extendsidi2_rex64_1): Remove.
354 (MOV0 peephole): Check GENERAL_REG_P.
356 2005-01-19 Richard Henderson <rth@redhat.com>
359 * config/i386/i386.c (ix86_expand_vector_set): Fix third and fourth
361 (ix86_expand_vector_extract): Likewise.
363 2005-01-19 Richard Henderson <rth@redhat.com>
366 * expr.c (emit_move_change_mode): New argument force; use
367 simplify_gen_subreg if true.
368 (emit_move_via_alt_mode): Merge into ...
369 (emit_move_via_integer): ... here.
370 (emit_move_ccmode): Use emit_move_change_mode directly.
372 2005-01-19 Richard Henderson <rth@redhat.com>
375 * config/alpha/alpha.c (alpha_rtx_costs): Handle HIGH.
376 (alpha_preferred_reload_class): Handle CONST_VECTOR.
377 (alpha_emit_set_const_1): Add no_output parameter; don't emit
379 (alpha_emit_set_const): Likewise. Make static.
380 (alpha_emit_set_long_const): Make static.
381 (alpha_extract_integer): Split out from alpha_expand_mov.
382 (alpha_split_const_mov): Likewise.
383 (alpha_expand_mov): Use them. Handle CONST_VECTOR.
384 (alpha_legitimate_constant_p): New.
385 * config/alpha/alpha-protos.h: Update.
386 * config/alpha/alpha.h (REGISTER_MOVE_COST): Correct fp<->gp cost.
387 (LEGITIMATE_CONSTANT_P): Re-implement with a function.
388 * config/alpha/alpha.md (movsi): Add n alternative.
389 (movsi_nt_vms, movdi_er_nofix, movdi_er_fix, movdi_fix): Likewise.
390 (mov<VEC>_fix, mov<VEC>_nofix): Add i alternative.
391 (splitters for all of the above): Use alpha_split_const_mov.
392 * config/alpha/predicates.md (non_add_const_operand): New.
393 (non_zero_const_operand): New.
394 (input_operand): Use alpha_legitimate_constant_p after reload.
396 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
398 PR tree-optimization/19038
399 * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): New function.
400 (add_candidate): Add ivs with increment in latch only if
401 allow_ip_end_pos_p is true.
402 (determine_iv_cost): Use empty_block_p.
404 2005-01-19 Daniel Berlin <dberlin@dberlin.org>
406 * cfganal.c (compute_dominance_frontiers_1): Replace with new algorithm
407 (compute_dominance_frontiers): Ditto.
409 2005-01-19 Ralf Corsepius <ralf.corsepius@rtems.org>
412 * config/sh/t-rtems: New.
413 * config.gcc (sh-*-rtems*): Reflect having added config/sh/t-rtems.
415 2005-01-19 Zdenek Dvorak <dvorakz@suse.cz>
417 * tree-ssa-ccp.c (maybe_fold_offset_to_component_ref): Always subtract
418 the offset of the selected field.
420 2005-01-19 Kazu Hirata <kazu@cs.umass.edu>
422 * tree-cfg.c (remove_forwarder_block): Fix the check to
423 prevent a nonlocal label from appearing in the middle of a
426 2005-01-19 Hans-Peter Nilsson <hp@axis.com>
428 PR rtl-optimization/19462
429 * reorg.c (find_end_label): Create return insn only if
430 current_function_epilogue_delay_list is empty.
432 2005-01-19 Paolo Bonzini <bonzini@gnu.org>
434 * doc/passes.texi: Remove paragraph mentioning new-ra.
436 2005-01-19 Jakub Jelinek <jakub@redhat.com>
438 PR rtl-optimization/15139
439 * combine.c: Include params.h.
440 (count_rtxs): New function.
441 (record_value_for_reg): If replace_rtx would replace at least
442 2 occurrences of REG in VALUE and TEM is really large, replace REG with
443 (clobber (const_int 0)) instead of TEM.
444 * params.def (PARAM_MAX_LAST_VALUE_RTL): New.
445 * params.h (MAX_LAST_VALUE_RTL): New.
446 * Makefile.in (combine.o): Depend on $(PARAMS_H).
447 * doc/invoke.texi (--param max-last-value-rtl=N): Document.
450 * c-typeck.c (digest_init): Only call build_vector if all constructor
451 elements are *_CST nodes.
452 * gimplify.c (gimplify_init_constructor): Likewise.
455 * c-typeck.c (digest_init): Only call build_vector if inside_init
458 2005-01-18 Tobias Schl"uter <tobias.schlueter@physik.uni-muenchen.de>
460 * toplev.c (init_asm_output): Remove dead #ifdef.
462 2005-01-18 Kaz Kojima <kkojima@gcc.gnu.org>
464 * config/sh/linux.h (TARGET_C99_FUNCTIONS): Define.
466 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
468 * calls.c (expand_call): Check DECL_BUILT_IN_CLASS before
469 accessing DECL_FUNCTION_CODE.
470 * dojump.c (do_jump): Likewise.
471 * gimplify.c (gimplify_call_expr): Likewise.
472 * predict.c (expr_expected_value): Likewise.
473 (strip_builtin_expect): Likewise.
474 * tree-inline.c (estimate_num_insns_1): Likewise.
475 * tree-ssa-loop-im.c (stmt_cost): Likewise
476 * fold-const.c (fold): Test for BUILT_IN_NORMAL.
477 (tree_expr_nonnegative_p): Likewise.
479 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
481 * config/sparc/sparc.c (load_pic_register): Emit the appropriate
482 variant of the load_pcrel_sym pattern.
483 * config/sparc/sparc.md (P macro): Move to the top.
484 (load_pcrel_sym): Macroize using P.
485 (save_register_window): Likewise.
487 2005-01-18 Aldy Hernandez <aldyh@redhat.com>
489 * config/rs6000/rs6000.md ("sunordered"): Disable for e500.
492 2005-01-18 Andrew Pinski <pinskia@physics.uc.edu>
495 * c-typeck.c (build_asm_expr): Strip nops off
496 input memory operands.
498 2005-01-18 David Edelsohn <edelsohn@gnu.org>
500 * config/rs6000/aix43.h (CPLUSPLUS_CPP_SPEC): Delete
501 _XOPEN_SOURCE, _XOPEN_SOURCE_EXTENDED, _LARGE_FILE_API.
502 * config/rs6000/aix51.h (CPLUSPLUS_CPP_SPEC): Same.
503 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Same.
505 2005-01-18 Joel Sherrill <joel.sherrill@oarcorp.com>
508 * config/i386/i386.c (override_options): If the 80387 is disabled,
509 then do not return FP values using FP registers.
511 2005-01-18 Kazu Hirata <kazu@cs.umass.edu>
513 * basic-block.h, c-common.c, c-cppbuiltin.c, c-lang.c,
514 c-tree.h, cfgbuild.c, cgraph.c, cgraph.h, collect2.c,
515 combine.c, config.gcc, coverage.h, cse.c, cselib.c,
516 defaults.h, df.c, dwarf2asm.c, dwarf2out.c, explow.c, expr.c,
517 flow.c, fold-const.c, gcse.c, ggc-page.c, gimple-low.c,
518 gimplify.c, ifcvt.c, langhooks-def.h, lcm.c, optabs.h,
519 output.h, postreload-gcse.c, postreload.c, recog.c,
520 resource.c, rtl.def, rtlanal.c, sched-deps.c, sched-rgn.c,
521 targhooks.h, toplev.c, tree-data-ref.c, tree-eh.c,
522 tree-flow.h, tree-loop-linear.c, tree-mudflap.h, tree-nrv.c,
523 tree-optimize.c, tree-outof-ssa.c, tree-pass.h,
524 tree-scalar-evolution.c, tree-ssa-copy.c, tree-ssa-dce.c,
525 tree-ssa-dse.c, tree-ssa-loop-ivopts.c, tree-ssa-operands.c,
526 tree-ssa-pre.c, tree-ssa.c, tree-vectorizer.c, tree.def,
527 unwind-dw2-fde-darwin.c, var-tracking.c: Update copyright.
529 * config/darwin.c, config/darwin.h, config/freebsd-spec.h,
530 config/alpha/alpha.c, config/arm/vfp.md,
531 config/i386/emmintrin.h, config/i386/i386-modes.def,
532 config/i386/i386-protos.h, config/i386/i386.h,
533 config/i386/pmmintrin.h, config/i386/ppro.md,
534 config/i386/predicates.md, config/i386/xmmintrin.h,
535 config/i860/i860.c, config/ia64/ia64-modes.def,
536 config/ia64/ia64-protos.h, config/ia64/ia64.h,
537 config/ia64/ia64.md, config/ia64/itanium1.md,
538 config/ia64/itanium2.md, config/ia64/predicates.md,
539 config/iq2000/iq2000.h, config/mips/linux64.h,
540 config/rs6000/aix41.h, config/rs6000/aix43.h,
541 config/rs6000/aix52.h, config/rs6000/darwin-fallback.c,
542 config/rs6000/rs6000.c, config/rs6000/rs6000.h,
543 config/rs6000/spe.md, config/sh/sh.md, config/sparc/linux.h,
544 config/sparc/linux64.h, config/sparc/litecoff.h,
545 config/sparc/sp64-elf.h, config/sparc/sparc.c,
546 config/sparc/sparc.h, config/sparc/sparc.md,
547 config/sparc/sysv4-only.h: Update copyright.
549 2005-01-18 Richard Henderson <rth@redhat.com>
552 * config/i386/i386.c (ix86_expand_fp_movcc): Fail for LTGT and UNEQ.
554 2005-01-18 Marc Espie <espie@openbsd.org>
555 * config/i386/openbsdelf.h: Typo.
557 2005-01-18 Richard Henderson <rth@redhat.com>
559 * config/i386/i386.c (ix86_expand_fp_movcc): Remove TARGET_IEEE_FP
560 special case for sse.
561 * config/i386/predicates.md (sse_comparison_operator): Likewise.
563 2005-01-18 Dorit Naishlos <dorit@il.ibm.com>
565 * tree-inline.c (estimate_num_insns_1): Added cases for
566 ALIGN_INDIRECT_REF, MISALIGNED_INDIRECT_REF, and REALIGN_LOAD_EXPR.
568 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
570 PR rtl-optimization/19296
571 * combine.c (simplify_comparison): Rewrite the condition under
572 which a non-paradoxical SUBREG of a PLUS can be lifted when
573 compared against a constant.
575 2005-01-18 Andi Kleen <ak@muc.de>
577 * c-typeck.c: (convert_for_assignment): Check warn_pointer_sign.
578 * c.opt (-Wpointer-sign): Add.
579 * doc/invoke.texi: (-Wpointer-sign): Add.
581 2005-01-18 Uros Bizjak <uros@kss-loka.si>
583 * config/i386/i386.c (override_options): Revert 2004-11-24 change.
584 * config/i386/i386.md (fmodsf3, dremsf3, *sinsf2, *cossf2, sincossf3,
585 *tansf3_1, tansf2, atan2sf3_1, atan2sf3, atansf2, asinsf2, acossf2,
586 logsf2, log10sf2, log2sf2, log1psf2, logbsf2, ilogbsf2, expsf2,
587 exp10sf2, exp2sf2, expm1sf2, rintsf2, floorsf2, ceilsf2, btruncsf2,
588 nearbyintsf2): Disable for TARGET_SSE_MATH. Leave patterns enabled
589 for TARGET_MIX_SSE_I387.
590 (fmoddf3, dremdf3, *sindf2, *sinextendsfdf2, *cosdf2, *cosextendsfdf2,
591 sincosdf3, *sincosextendsfdf3, *tandf3_1, tandf2, atan2df3_1,
592 atan2df3, atandf2, asindf2, acosdf2, logdf2, log10df2, log2df2,
593 log1pdf2, logbdf2, expdf2, exp10df2, exp2df2, expm1df2, rintdf2,
594 floordf2, ceildf2, btruncdf2, nearbyintdf2): Disable for
595 (TARGET_SSE2 && TARGET_SSE_MATH). Leave patterns enabled for
597 (atan2sf3, atan2df3, atan2xf3): Remove register constraints
600 2005-01-18 Uros Bizjak <uros@kss-loka.si>
603 * config/i386/mmx.md (*movv2sf_internal_rex64, *movv2sf_internal):
604 Add movaps alternative for xmm reg->reg move.
606 2005-01-17 Jeff Law <law@redhat.com>
608 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Clear SSA_NAME_VALUE
609 at the end of the main DOM loop rather than just before DOM exits.
611 2005-01-17 Diego Novillo <dnovillo@redhat.com>
613 PR tree-optimization/19121
614 * tree-ssa-alias.c (compute_flow_sensitive_aliasing): When
615 adding aliases to a name tag, also add them to the pointer's
617 * tree-ssa-copy.c (merge_alias_info): Do not merge flow
618 sensitive alias info at all. Only check that the two pointers
619 have compatible pointed-to sets.
620 * tree-ssa.c (verify_name_tags): Verify that the alias set of
621 a pointer's type tag is a superset of the alias set of the
624 2005-01-17 James E Wilson <wilson@specifixinc.com>
627 * config/ia64/ia64.md (movxf): Handle general register source. Adjust
628 comment to document why.
630 2005-01-17 Richard Henderson <rth@redhat.com>
632 * config/i386/sse.md (smaxv4sf3_finite, sse_vmsmaxv4sf3_finite,
633 sminv4sf3_finite, sse_vmsminv4sf3_finite, smaxv2df3_finite,
634 sse2_vmsmaxv2df3_finite, sminv2df3_finite,
635 sse2_vmsminv2df3_finite): New.
636 (smaxv4sf3, sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, smaxv2df3,
637 sse2_vmsmaxv2df3, sminv2df3, sse2_vmsminv2df3): Remove commutative.
638 Force op1 into register.
640 2005-01-17 Kelley Cook <kcook@gcc.gnu.org>
642 * Makefile.in (STAGEMOVESTUFF): Stage all the stamp files to prevent
643 bubblestrap from always rebuilding build directory.
645 2005-01-17 Ulrich Weigand <uweigand@de.ibm.com>
647 * config/s390/s390.md ("*clc"): Remove incorrect '='.
648 ("*cmpmem_short"): Likewise.
650 2005-01-17 Steven Bosscher <stevenb@suse.de>
652 * basic-block.h: Document BB_* flags.
653 * regrename.c (copyprop_hardreg_forward): Don't use BB_VISITED,
654 use an sbitmap instead.
655 * sched-rgn.c (compute_trg_info): Likewise.
657 2005-01-17 Richard Sandiford <rsandifo@redhat.com>
659 * config.gcc (mips64*-*-linux*): Set the default abi to n32. Remove
660 redundant target_cpu_default and extra_parts lines. Use the default
662 * config/mips/linux64.h (DRIVER_DEFAULT_ABI_SELF_SPEC): Delete.
663 (SUBTARGET_EXTRA_SPECS): Delete.
664 (DRIVER_SELF_SPECS): Remove architecture lines. Set the ABI directly.
665 (SUBTARGET_ASM_SPEC): Remove -mabi=n32 and -mabi=64 mappings.
666 * config/mips/t-linux64 (MULTILIB_DIRNAMES): Use "32" for -mabi=32
667 and "n32" for -mabi=n32.
669 2005-01-17 Ian Lance Taylor <ian@airs.com>
672 * c-typeck.c (build_c_cast): Revert patch of 2003-02-16: don't
673 fold constant variables into initial values.
675 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
677 * tree-cfg.c (tree_can_merge_blocks_p): Reorder two checks.
679 * tree-cfg.c (tree_forwarder_block_p): Speed up by walking
680 through the statements backward.
682 2005-01-17 Ian Lance Taylor <ian@airs.com>
685 * tree-inline.c (expand_call_inline): Set TREE_NO_WARNING on
686 a variable set to the return value of the inlined function.
688 2005-01-17 Mark Dettinger <dettinge@de.ibm.com>
690 * rtlanal.c (rtx_cost): Assign cost of 0 to a SUBREG
691 when modes are tieable.
693 2005-01-17 Ranjit Mathew <rmathew@hotmail.com>
695 * gthr-posix.h (__gthread_active_p): Use pthread_cancel instead
696 of pthread_create to find out if threads are enabled.
697 * gthr-posix95.h (__gthread_active_p): Likewise.
699 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
701 * common.opt (-fnew-ra): Remove.
703 * toplev.h (flag_new_regalloc): Remove.
704 * Makefile.in (ra*.*): Don't mention.
705 * passes.c (rest_of_handle_new_regalloc): Remove.
706 (rest_of_handle_combine, rest_of_compilation): Always consider
707 flag_new_regalloc as false.
708 * doc/invoke.texi: Don't document -fnew-ra.
710 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
712 * bb-reorder.c (fix_edges_for_rarely_executed_code): Remove
713 last parameter to reg_scan.
714 * loop.c (loop_optimize): Likewise.
715 * passes.c (rest_of_handle_tracer, rest_of_handle_if_conversion,
716 rest_of_handle_web, rest_of_handle_cfg, rest_of_handle_jump_bypass,
717 rest_of_handle_life, rest_of_handle_cse, rest_of_handle_cse2,
718 rest_of_handle_gcse, rest_of_handle_loop_optimize,
719 rest_of_handle_loop2, rest_of_handle_jump2): Likewise.
720 * regclass.c (reg_scan): Likewise, for the declaration.
721 * rtl.h (reg_scan): Likewise, for the prototype.
723 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
725 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Don't call
728 2005-01-16 Daniel Berlin <dberlin@dberlin.org>
730 * tree-ssa-pre.c (add_to_sets): s1 may be NULL.
731 (compute_avail): Uses don't go in tmp_gen.
733 2005-01-16 Steven Bosscher <stevenb@suse.de>
735 * ggc-page.c (ggc_alloc_stat): Use __builtin_ctzl instead of a
736 loop to look for a free slot in a page entry.
738 2005-01-16 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
741 * defaults.h (TARGET_DEFERRED_OUTPUT_DEFS): Provide default.
742 * toplev.c (compile_file): Call process_pending_assemble_output_defs
743 just before targetm.asm_out.file_end.
744 * tree.h (process_pending_assemble_output_defs): Declare.
745 * varasm.c (assemble_output_def, process_pending_assemble_output_defs):
747 (assemble_alias): Defer generation of assembly code for defines when
748 TARGET_DEFERRED_OUTPUT_DEFS is true.
749 * config/rs6000/aix41.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
750 * config/rs6000/aix43.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
751 * doc/tm.texi (TARGET_DEFERRED_OUTPUT_DEFS): document.
753 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
756 * pa.c (pa_scalar_mode_supported_p): New function.
757 (TARGET_SCALAR_MODE_SUPPORTED_P): Define.
758 * pa.h (MIN_UNITS_PER_WORD): Add comment.
760 * pa-protos.h (prefetch_operand): Delete.
761 (prefetch_cc_operand, prefetch_nocc_operand): New declations.
762 * pa.c (prefetch_operand): Delete.
763 (prefetch_cc_operand, prefetch_nocc_operand): New functions.
764 * pa.h (EXTRA_CONSTRAINT): Add `W' constraint.
765 (PREDICATE_CODES): Delete prefetch_operand. Add prefetch_cc_operand
766 and prefetch_nocc_operand.
767 * pa.md (prefetch): Rework to avoid reload problems handling short
768 displacements when a cache control completer needs to be provided.
769 (prefetch_32, prefetch_64): Delete.
770 (prefetch_cc, prefetch_nocc): New patterns.
772 2005-01-15 David Edelsohn <edelsohn@gnu.org>
774 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Revert previous change.
776 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
779 * pa.c (pa_asm_output_aligned_common): Warn if specified alignment
780 exceeds maximum alignment for global common data.
781 * pa64-hpux.h (MAX_OFILE_ALIGNMENT): Define to 32768.
782 * som.h (MAX_OFILE_ALIGNMENT): Likewise.
784 2004-01-15 Roger Sayle <roger@eyesopen.com>
786 * tree-ssa-dom.c (extract_range_from_cond): Correct condition.
788 2004-01-15 Roger Sayle <roger@eyesopen.com>
790 * harg-reg-set.h (reg_class_names): Prototype global array.
791 * regclass.c (reg_class_names): Declare here and initialize to
793 (dump_regclass): Remove local declaration of reg_class_names.
794 (regclass): Likewise.
795 * cfg.c (dump_flow_info): Likewise.
796 * ra-debug.c (reg_class_names): Likewise.
797 * regrename.c (reg_class_names): Likewise.
798 * reload.c (reg_class_names): Likewise.
799 * reload1.c (spill_failure): Likewise.
800 * config/m68hc11/m68hc11.c (reg_class_names): Likewise.
802 2005-01-15 Ulrich Weigand <uweigand@de.ibm.com>
804 * config/s390/s390.md ("reload_outti"): Remove predicate for
805 output operand. Abort if operand is not a MEM.
806 ("reload_outdi", "reload_outdf"): Likewise.
808 2005-01-15 Marc Espie <espie@openbsd.org>
810 * config.gcc (*-*-openbsd*): Set HAS_LIBC_R for OpenBSD <= 3.2.
811 (i[34567]86-*-openbsd*): Switch to ELF for OpenBSD >= 3.4.
812 * config/openbsd.h: Fix C++ includes for native configurations.
813 Add proper OS_CPP_BUILTINS. Fix libspec for recent OpenBSD.
814 Add trampoline support.
815 * config/i386/openbsdelf.h: New.
817 2005-01-15 Marc Espie <espie@openbsd.org>
819 * collect2.c (main): Explicitly parse -dynamic-linker option.
821 2005-01-15 Jakub Jelinek <jakub@redhat.com>
823 PR tree-optimization/19060
824 * tree-ssa-dom.c (extract_range_from_cond) <case LT_EXPR, GT_EXPR>:
825 Return 0 if op1 <= TYPE_MIN_VALUE () resp. op1 >= TYPE_MAX_VALUE ().
826 (simplify_cond_and_lookup_avail_expr): Add assert for dummy == 0
827 and handle extract_range_from_cond returning false.
828 * fold-const.c (fold): Optimize comparisons with min/max even for
829 width > HOST_BITS_PER_WIDE_INT.
831 2005-01-15 Ralf Corsepius <ralf.corsepius@rtems.org>
833 * config/mips/rtems.h (MIPS_DEFAULT_GVALUE): Set to 0.
834 * config/mips/t-rtems (MULTILIBS_DIRNAMES,MULTILIB_OPTIONS):
835 Remove little endian multilib variants.
836 Add mips32 multilib variant.
838 2005-01-15 Kazu Hirata <kazu@cs.umass.edu>
840 * explow.c (copy_all_reg, stabilize): Remove.
841 * expr.h: Remove the corresponding prototypes.
843 * rtlanal.c (regs_set_between_p): Remove.
844 * rtl.h: Remove the corresponding prototype.
846 2005-01-14 David Edelsohn <edelsohn@gnu.org>
848 * config/rs6000/rs6000.c (rs6k_nonimmediate_operand): Rename
849 to rs6000_nonimmediate_operand.
850 * config/rs6000/rs6000.h (PREDICATE_CODES): Rename
851 rs6k_nonimmediate_operand to rs6000_nonimmediate_operand.
852 * config/rs6000/rs6000.md (movsi_internal1): Rename predicate to
853 rs6000_nonimmediate_operand.
854 * config/rs6000/spe.md (movdf_e500_double): Rename predicate to
855 rs6000_nonimmediate_operand.
857 2005-01-14 Andrew Pinski <pinskia@physics.uc.edu>
859 * tree-ssa-alias.c (pass_may_alias): Add TODO_verify_stmts.
861 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
863 * config/rs6000/rs6000.h (CLASS_MAX_NREGS): DF goes in 1 register
865 (CANNOT_CHANGE_MODE_CLASS): Restrict DI mode changes on e500v2.
866 (PREDICATE_CODES): Add rs6k_nonimmediate_operand.
868 * config/rs6000/rs6000.c (invalid_e500_subreg): New.
869 (rs6k_nonimmediate_operand): New.
870 (rs6000_legitimate_offset_address_p): Handle DI modes on e500v2
872 (legitimate_lo_sum_address_p): Same.
873 (rs6000_legitimize_address): Same.
874 (rs6000_legitimize_reload_address): Same.
875 (rs6000_legitimate_address): Same.
876 (spe_build_register_parallel): Pass DF and DC modes in a DI
879 * config/rs6000/rs6000.md ("*movsi_internal1"): Change predicate
880 to rs6k_nonimmediate_operand.
882 * config/rs6000/spe.md ("*frob_df_di"): New.
883 ("*frob_di_df"): New.
884 ("*frob_di_df_2"): New.
885 ("*mov_sidf_e500_subreg0"): New.
886 ("*mov_sidf_e500_subreg4"): New.
887 ("*movdf_e500_double"): Change predicate to
888 rs6k_nonimmediate_operand.
890 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
892 * postreload.c (move2add_note_store): Only call
893 trunc_int_for_mode on scalar integers.
895 2005-01-14 Steven Bosscher <stevenb@suse.de>
897 * tree-ssa-dce.c (visited_control_parents): New sbitmap to
898 replace BB_VISITED uses.
899 (find_obviously_necessary_stmts): Don't clear BB_VISITED.
900 (propagate_necessity): Check the bitmap instead of BB_VISITED.
901 (tree_dce_done): Free visited_control_parents.
902 (perform_tree_ssa_dce): Allocate and clear it.
903 * tree-ssa-pre.c (compute_antic_aux): Make non-recursive.
904 (compute_antic): Iterate from here using a DFS. Use an sbitmap
905 instead of BB_VISITED.
907 2005-01-14 Kazu Hirata <kazu@cs.umass.edu>
909 * c-tree.h, coverage.h, langhooks-def.h, optabs.h, output.h,
910 rtl.h, targhooks.h, tree-flow.h, tree-mudflap.h, tree.h:
911 Remove unused prototypes.
913 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
915 * config/rs6000/rs6000.c (rs6000_stack_info): Fix caching of SPE
916 64-bit register usage.
917 (spe_func_has_64bit_regs_p): Add FIXME note.
919 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
921 * config/rs6000/rs6000.c (rs6000_generate_compare): Replace
922 flag_finite_math_only with flag_unsafe_math_optimizations.
924 2005-01-14 J. D. Johnston <jjohnst@us.ibm.com>
926 * config/s390/tpf-unwind.h (s390_fallback_frame_state): Correct
928 (__tpf_eh_return): Copy TPF private stack area from the module
929 boundary stack frame. Add check for when module addresses are
930 equal, but no stub address is found.
932 2005-01-14 Richard Earnshaw <rearnsha@arm.com>
935 * arm.h (struct machine_function): Add call_via field.
936 (thumb_call_via_label): Declare.
937 * arm.c (thumb_call_via_label): New variable.
938 (thumb_call_reg_needed): New variable.
939 (arm_output_function_epilogue): For Thumb code, output any per-function
940 call-indirect trampolines.
941 (thumb_call_via_reg): New function.
942 (arm_file_end): New function.
943 (TARGET_ASM_FILE_END): Call arm_file_end.
944 (aof_file_end): Likewise.
945 * arm-protos.h (thumb_call_via_reg): Declare.
946 * arm.md (call_reg_thumb, call_value_reg_thumb): Call
947 thumb_call_via_reg in normal case.
949 2005-01-14 Jakub Jelinek <jakub@redhat.com>
952 PR rtl-optimization/19348
953 * recog.c (peephole2_optimize): Do global life update if some peephole
954 decides it doesn't need at least one of its inputs and that change
955 influences liveness at the start of the basic block.
957 * basic-block.h (EXECUTE_IF_AND_COMPL_IN_REG_SET): Needs 2 REGSET
958 arguments instead of 1.
960 2005-01-14 Eric Botcazou <ebotcazou@libertysurf.fr>
963 * varasm.c (initializer_constant_valid_p) <ADDR_EXPR>: Return
964 zero for nested functions needing a static chain or functions
965 with a non-constant address.
967 2005-01-13 Roger Sayle <roger@eyesopen.com>
969 * simplify-rtx.c (simplify_binary_operation) <AND>: Optimize
970 (and (sign_extend X) C) into (zero_extend (and X C)).
972 2005-01-13 David O'Brien <obrien@FreeBSD.org>
974 * config/freebsd-spec.h: Make KSE pthread lib logic the default.
976 2005-01-13 Richard Henderson <rth@redhat.com>
981 * config/i386/i386.md (cmpdf, cmpsf, bunordered, bordered, buneq,
982 bunge, bungt, bunle, bunlt, bltgt): Enable for TARGET_SSE_MATH,
984 (cmpfp_i_387): Rename from cmpfp_i. Move after sse patterns.
985 (cmpfp_i_mixed): Rename from cmpfp_i_sse; use for TARGET_MIX_SSE_I387.
986 (cmpfp_i_sse): Rename from cmpfp_i_sse_only; use for TARGET_SSE_MATH.
987 (cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387): Similarly.
988 (fp_jcc_1_mixed, fp_jcc_1_sse, fp_jcc_1_387): Similarly.
989 (fp_jcc_2_mixed, fp_jcc_2_sse, fp_jcc_2_387): Similarly.
990 (fp_jcc_3_387, fp_jcc_4_387, fp_jcc_5_387, fp_jcc_6_387,
991 fp_jcc_7_387, fp_jcc_8_387): Rename from fp_jcc_N.
992 (movdicc_c_rex64): Rename with '*'.
993 (movsfcc, movdfcc): Add checks for 387 and sse math to condition.
994 (movsfcc_1_sse_min, movsfcc_1_sse_max, movsfcc_1_sse): New.
995 (movsfcc_1_387): Rename from movsfcc_1.
996 (movdfcc_1_sse_min, movdfcc_1_sse_max, movdfcc_1_sse): New.
997 (movdfcc_1, movdfcc_1_rex64): Add check for 387.
998 (sminsf3, smaxsf3, smindf3, smaxdf3): New.
999 (minsf3, minsf, minsf_nonieee, minsf_sse, mindf3, mindf,
1000 mindf_nonieee, mindf_sse, maxsf3, maxsf, maxsf_nonieee, maxsf_sse,
1001 maxdf3, maxdf, maxdf_nonieee, maxdf_sse, sse_movsfcc, sse_movsfcc_eq,
1002 sse_movdfcc, sse_movdfcc_eq, sse_movsfcc_const0_1,
1003 sse_movsfcc_const0_2, sse_movsfcc_const0_3, sse_movsfcc_const0_4,
1004 sse_movdfcc_const0_1, sse_movdfcc_const0_2, sse_movdfcc_const0_3,
1005 sse_movdfcc_const0_4): Remove.
1006 * config/i386/i386.c (ix86_expand_fp_movcc): For TARGET_SSE_MATH,
1007 recognize min/max early. Update for changed sse cmove patterns.
1008 (ix86_split_sse_movcc): New.
1009 * config/i386/i386-protos.h: Update.
1011 2005-01-13 Steven Bosscher <stevenb@suse.de>
1013 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
1014 (fix_stmt_v_may_defs): Likewise.
1016 2005-01-13 Richard Henderson <rth@redhat.com>
1018 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
1019 for converting the mask.
1021 2005-01-13 David Edelsohn <edelsohn@gnu.org>
1023 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
1026 2005-01-13 Richard Henderson <rth@redhat.com>
1028 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
1029 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
1030 IX86_BUILTIN_MOVDQ2Q): Remove.
1031 (IX86_BUILTIN_VEC_EXT_V4SI): New.
1032 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
1033 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
1035 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
1036 (sse2_pextrw): Fix immediate constraint.
1037 (sse2_loadq, sse2_loadq_rex64): Remove.
1038 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
1039 Use __builtin_ia32_vec_ext_<size>.
1040 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
1042 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
1044 * function.c (assign_parm_setup_block): Look inside original
1045 entry_parm when inspecting PARALLEL.
1047 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
1048 Joel Sherrill <joel@oarcorp.com>
1051 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
1052 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
1053 rtems_gxx_recursive_mutex_init.
1054 (__gthread_recursive_mutex_lock): New function.
1055 (__gthread_recursive_mutex_trylock): Likewise.
1056 (__gthread_recursive_mutex_unlock): Likewise.
1058 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
1060 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
1062 2005-01-13 Jan Beulich <jbeulich@novell.com>
1064 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
1065 (ashrti3_internal): Indicate output is early clobber. Generate result
1066 into output rather than first input. Use move for low word of output
1067 if shift count is exactly 64.
1068 (lshrti3_internal): Likewise.
1070 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
1074 * reload1.c (delete_output_reload): Don't delete an output reload
1075 if the pseudo lives longer than a single basic block. Adjust and
1076 improve wording of comment.
1078 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
1081 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
1082 of TYPE_SIZE where we mean the number of bytes.
1084 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
1086 * function.c (assign_parm_setup_block): Relax condition on
1087 multi-register optimization.
1089 2005-01-12 Nick Clifton <nickc@redhat.com>
1091 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
1092 an "in_delay_slot" attribute of "no" to prevent them being used in
1093 delay slots. This is forbidden because they might generate
1096 2005-01-12 Alan Modra <amodra@bigpond.net.au>
1099 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
1102 2005-01-12 Nick Clifton <nickc@redhat.com>
1104 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
1105 passed on to GAS) but do not define, as GAS no longer supports or
1106 needs the -m2000 option.
1108 2005-01-11 Roger Sayle <roger@eyesopen.com>
1110 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
1111 type as args of unprototyped builtins aren't automatically promoted.
1112 (fold_builtin_abs): Likewise.
1114 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
1116 * Makefile.in: Set a `build-warn' variable.
1118 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
1120 PR tree-optimization/17949
1121 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
1122 (find_interesting_uses_address): Use it.
1124 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
1126 * regrename.c (kill_value): Handle subreg's that won't simplify.
1128 2005-01-11 Richard Henderson <rth@redhat.com>
1131 * config/i386/i386.h (enum ix86_builtins): Move ...
1132 * config/i386/i386.c: ... here.
1133 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
1134 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
1135 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
1136 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
1137 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
1138 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
1139 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
1140 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
1141 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
1142 IX86_BUILTIN_LOADDDUP): Remove.
1143 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
1144 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
1145 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
1146 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
1147 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
1148 (ix86_init_builtins): Make static.
1149 (ix86_init_mmx_sse_builtins): Update for changed builtins.
1150 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
1151 if all the modes match. Otherwise, fake it.
1152 (get_element_number, ix86_expand_vec_init_builtin,
1153 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
1154 (ix86_expand_builtin): Make static. Update for changed builtins.
1155 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
1156 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
1157 (ix86_expand_vector_init_duplicate): New.
1158 (ix86_expand_vector_init_low_nonzero): New.
1159 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
1160 Split out from ix86_expand_vector_init; handle integer modes.
1161 (ix86_expand_vector_init): Use them.
1162 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
1163 * config/i386/i386-protos.h: Update.
1164 * config/i386/predicates.md (reg_or_0_operand): New.
1165 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
1166 (movv2sf_internal): Likewise. And a splitter to match them all.
1167 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
1168 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
1169 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
1170 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
1171 (mmx_pinsrw): Fix operand ordering.
1172 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
1173 rather than sse_loadss expander.
1174 (movv2df splitter): Similarly.
1175 (sse_loadss, sse_loadlss): Remove.
1176 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
1177 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
1178 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
1179 (sse3_movddup): Rename with '*'.
1180 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
1181 (sse2_loadsd): Remove.
1182 (vec_dupv2df_sse3): Rename from sse3_loadddup.
1183 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
1184 (sse2_pinsrw): Fix argument ordering.
1185 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
1186 (sse2_stored): Remove 'r' destination.
1187 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
1188 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
1189 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
1190 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
1191 vec_extractv16qi, vec_initv16qi): New.
1193 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
1194 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
1195 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
1197 (_mm_load_pd, _mm_store_pd): Use plain dereference.
1198 (_mm_load_si128, _mm_store_si128): Likewise.
1199 (_mm_load1_pd): Use _mm_set1_pd.
1200 (_mm_load_sd): Use _mm_set_sd.
1201 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
1202 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
1203 (_mm_set_epi64): Use _mm_set_epi64x.
1204 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
1205 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
1206 _mm_setr_epi8): Use _mm_set_foo form.
1207 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
1208 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
1209 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
1210 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
1211 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
1212 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
1213 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
1214 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
1215 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
1216 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
1217 (_mm_movedup_pd): Use _mm_shuffle_pd.
1218 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
1219 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
1220 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
1221 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
1223 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
1224 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
1225 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
1226 (_mm_store_ps): Use raw dereference.
1227 (_mm_store1_ps): Use _mm_storeu_ps.
1228 (_mm_storer_ps): Use _mm_store_ps.
1229 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
1230 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
1232 2005-01-11 Stan Shebs <shebs@apple.com>
1234 * config/rs6000/rs6000.c (machopic_output_stub): Issue
1235 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
1237 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
1240 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
1241 Skip all DECLs except for FIELD_DECLs.
1243 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
1245 * config/s390/s390.c (override_options): Return error if
1246 -mbackchain, -mpacked-stack and -mhard-float are used together.
1247 (s390_va_start): Remove the backchain && packed-stack special case.
1248 (s390_gimplify_va_arg): Likewise.
1249 * doc/invoke.texi: Remove the ABI incompatibility note.
1251 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
1253 * config/s390/s390.c (struct s390_frame_layout): Remove
1255 (s390_frame_info, s390_emit_prologue): Replace occurrences of
1256 save_backchain_p with TARGET_BACKCHAIN.
1258 2005-01-11 Alan Modra <amodra@bigpond.net.au>
1261 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
1262 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
1263 * function.c (locate_and_pad_parm): Set new field.
1264 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
1265 Tweak where_pad test to include "none". Always set mem align for
1267 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
1268 not sufficient for type.
1269 (assign_parm_setup_block): If stack_parm is zero on entry, always
1270 make a new stack local. Block move old stack parm if necessary
1271 to new aligned stack local.
1272 (assign_parm_setup_stack): Use a block move to handle
1273 potentially misaligned entry_parm.
1274 (assign_parms_unsplit_complex): Specify required alignment when
1275 creating stack local.
1276 * calls.c (compute_argument_addresses): Override alignment of stack
1277 arg calculated from its type with the alignment given by
1278 FUNCTION_ARG_BOUNDARY.
1279 (store_one_arg): Likewise.
1281 2005-01-11 Jan Beulich <jbeulich@novell.com>
1283 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
1284 Change respective itanium_class attribute to ialu.
1285 (shladdp4_internal): New.
1286 * config/ia64/predicates.md (shladd_log2_operand): New.
1288 2005-01-11 Richard Henderson <rth@redhat.com>
1290 * expr.c (store_constructor): Use rtvec_alloc instead of
1291 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
1293 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
1295 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
1297 2005-01-11 Alan Modra <amodra@bigpond.net.au>
1299 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
1300 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
1301 (TARGET_RWRELOC): Define.
1302 (ia64_rwreloc_section_type_flags): Delete.
1303 (ia64_section_type_flags): New function.
1304 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
1305 (TARGET_RWRELOC): Define.
1307 2005-01-10 David Mosberger <davidm@hpl.hp.com>
1310 * config/ia64/ia64.c (process_set): For alloc insn, only call
1311 process_epilogue is !frame_pointer_needed.
1313 2005-01-10 Roger Sayle <roger@eyesopen.com>
1316 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
1317 unary operator and can't be treated as a binary/comparison operator.
1319 2005-01-10 Richard Henderson <rth@redhat.com>
1321 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
1322 (ix86_return_in_memory): Likewise.
1323 (function_arg_advance): Likewise. Mirror structure in function_arg
1324 for choosing register to advance.
1326 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
1328 * tree-vectorizer.c, tree.def: Fix comment typos.
1330 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
1332 * config/s390/s390.c (struct processor_costs): 4 new fields:
1333 ddbr, ddr, debr, der.
1334 (s390_rtx_costs): More precise handling of divide instructions.
1336 2005-01-10 David Edelsohn <edelsohn@gnu.org>
1338 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
1340 2005-01-10 Jan Beulich <jbeulich@novell.com>
1342 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
1343 names resulting from -ffunction-sections/-fdata-sections and linkonce
1345 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
1346 based on the section name. Rearrange the section name comparison logic
1347 slightly so that each section name is compared against at most once.
1349 2005-01-10 Ben Elliston <bje@au.ibm.com>
1351 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
1352 targets for which -fPIC reduces limitations on the GOT size.
1354 2005-01-09 Falk Hueffner <falk@debian.org>
1356 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
1359 2005-01-09 David Edelsohn <edelsohn@gnu.org>
1362 * collect2.c (main): Set aixrtl_flag for -brtl option.
1363 (resolve_lib_name): Search for .so file extension before .a
1366 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
1368 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
1369 (vect_analyze_data_refs_alignment): Add dump prints.
1371 2005-01-09 Ira Rosen <irar@il.ibm.com>
1373 * tree-vectorizer.c (vect_analyze_offset_expr): Use
1374 expr_invariant_in_loop_p.
1375 Initialize outputs first thing in the function.
1376 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
1377 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
1378 (vect_analyze_pointer_ref_access): Check that the initial condition of
1379 the access function is loop invariant.
1381 2005-01-09 Richard Henderson <rth@redhat.com>
1383 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
1384 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
1385 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
1386 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
1387 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
1388 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
1389 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
1390 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
1391 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
1392 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
1393 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
1394 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
1395 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
1396 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
1397 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
1398 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
1399 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
1400 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
1401 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
1402 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
1403 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
1404 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
1405 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
1407 (mmx_clrdi, pavgusb): Remove.
1408 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
1409 with leading sse_ prefix.
1410 * config/i386/sse.md: Receive them.
1411 * config/i386/mmx.md: New file.
1412 (MMXMODE12, MMXMODE24, mmxvecsize): New.
1413 (subrv2sf3): Turn into expander for normal subtraction.
1414 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
1415 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
1416 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
1417 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
1418 commutative; use ix86_binary_operator_ok.
1419 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
1420 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
1421 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
1422 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
1423 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
1424 Macroize from existing patterns; use ix86_binary_operator_ok.
1425 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
1426 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
1427 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
1428 with vec_select+vec_concat.
1429 (mmx_pshufw, mmx_pshufw_1): Likewise.
1430 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
1431 (mmx_uavgv4hi3): Model correctly.
1432 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
1433 _mm_xor_si64): Remove casts.
1435 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
1437 PR tree-optimization/19224
1438 * tree-scalar-evolution.c (get_instantiated_value,
1439 set_instantiated_value): New functions.
1440 (instantiate_parameters_1): Cache the results.
1441 (instantiate_parameters, resolve_mixers): Initialize and free
1444 2005-01-08 David Edelsohn <edelsohn@gnu.org>
1446 * config/i386/i386.md (addhi_4): Correct reference in comment.
1449 2005-01-08 Richard Henderson <rth@redhat.com>
1451 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
1452 (_mm_cvtsi128_si64x): Likewise.
1453 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
1454 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
1455 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
1456 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
1457 * config/i386/i386-protos.h: Update.
1458 * config/i386/i386.c (print_operand): Add 'H'.
1459 (ix86_fixup_binary_operands): Split out from ...
1460 (ix86_expand_binary_operator): ... here.
1461 (ix86_fixup_binary_operands_no_copy): New.
1462 (ix86_expand_fp_absneg_operator): Handle vector mode results.
1463 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
1464 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
1465 (safe_vector_operand): Use CONST0_RTX.
1466 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
1467 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
1468 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
1469 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
1471 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
1472 (IX86_BUILTIN_CMPNGESS): New.
1473 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
1474 (attr type): Add sselog1.
1475 (attr unit, attr memory): Handle it.
1476 (movti, movti_internal, movti_rex64): Move near other integer moves.
1477 (movtf, movtf_internal): Move near other fp moves.
1478 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
1479 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
1480 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
1481 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
1482 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
1483 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
1484 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
1485 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
1486 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
1487 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
1488 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
1489 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
1490 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
1491 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
1492 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
1493 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
1494 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
1495 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
1496 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
1497 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
1498 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
1499 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
1500 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
1501 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
1502 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
1503 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
1504 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
1505 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
1506 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
1507 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
1508 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
1509 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
1510 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
1511 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
1512 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
1513 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
1514 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
1515 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
1516 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
1517 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
1518 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
1519 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
1520 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
1521 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
1523 (SSEPUSH, push<SSEPUSH>): Remove.
1524 (MMXPUSH, push<MMXPUSH>): Remove.
1525 (sse_movaps, sse_movaps_1, sse_movups): Remove.
1526 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
1527 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
1528 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
1529 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
1530 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
1531 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
1532 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
1533 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
1534 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
1535 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
1536 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
1537 (ppro_sse_log_V4SF): Handle sselog1.
1538 * config/i386/predicates.md (const_0_to_1_operand): New.
1539 (const_0_to_255_mul_8_operand): New.
1540 (const_1_to_31_operand): Rename from const_int_1_31_operand.
1541 (const_2_to_3_operand, const_4_to_7_operand): New.
1542 * config/i386/sse.md: New file.
1543 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
1544 (sse_movups): Rename from sse_movups_1.
1545 (sse_loadlss): Rename from sse_loadss_1.
1546 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
1548 (negv4sf2): Use ix86_expand_fp_absneg_operator.
1549 (absv4sf2, negv2df, absv2df): New.
1550 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
1551 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
1552 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
1553 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
1554 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
1555 (sse3_addsubv4sf3): Model correctly.
1556 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
1557 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
1558 (sse_movhlps): Model with vec_select+vec_concat.
1559 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
1560 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
1561 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
1562 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
1563 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
1564 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
1565 sse2_pshufhw_1): Likewise.
1566 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
1567 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
1568 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
1569 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
1570 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
1571 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
1572 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
1573 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
1574 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
1575 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
1576 sminv8hi3): Mark commutative
1577 operands. Use ix86_binary_operator_ok.
1578 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
1579 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
1580 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
1581 sse2_punpcklqdq): Allow operand2 in memory.
1582 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
1583 sse2_movsd): Add memory alternatives.
1584 (sse_storelps): Turn expander into an insn; split after reload.
1585 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
1586 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
1588 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
1590 * configure.ac (DWARF-2 debug_line): Use objdump.
1591 * configure: Regenerate.
1593 2005-01-08 Jeff Law <law@redhat.com>
1594 Diego Novillo <dnovillo@redhat.com>
1596 PR tree-optimization/18241
1597 * tree-nrv.c (tree_nrv): Ignore volatile return values.
1598 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
1599 statements with volatile operands.
1600 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
1601 operands after marking a statement with has_volatile_ops.
1603 2005-01-08 Roger Sayle <roger@eyesopen.com>
1605 * tree.c (int_fits_type_p): Always honor integer constant
1606 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
1608 2005-01-08 Roger Sayle <roger@eyesopen.com>
1610 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
1611 the latent bug in rtl_delete_block has been resolved.
1613 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
1615 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
1616 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
1618 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
1620 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
1621 library as @multilib_dir@/$(SHLIB_SONAME).
1623 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
1625 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
1626 * configure: Regenerate.
1628 2005-01-07 Jakub Jelinek <jakub@redhat.com>
1630 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
1633 2005-01-07 David Edelsohn <edelsohn@gnu.org>
1636 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
1637 Convert non-word aligned offset address using ld/std into
1640 2005-01-07 Richard Henderson <rth@redhat.com>
1642 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
1644 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
1646 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
1648 2005-01-07 Jakub Jelinek <jakub@redhat.com>
1650 PR tree-optimization/19283
1651 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
1652 is not shorter than the original type.
1654 PR rtl-optimization/19012
1655 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
1657 PR rtl-optimization/18861
1658 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
1659 (FULL_STATE): Define.
1660 (mark_tablejump_edge): New function.
1661 (purge_dead_tablejump_edges): New function.
1662 (find_bb_boundaries): Use it.
1664 PR tree-optimization/18828
1665 * builtins.c (expand_builtin_next_arg): Remove argument and all
1666 the argument checking.
1667 (expand_builtin): Adjust caller.
1668 (expand_builtin_va_start): Likewise. Remove error for too many
1670 (fold_builtin_next_arg): Issue error for too many arguments.
1671 After checking arguments, replace them with magic arguments that
1672 prevent further checking of the args.
1674 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1676 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
1678 2005-01-06 Roger Sayle <roger@eyesopen.com>
1681 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
1682 and %{.S:...} (and their negative variants) to test whether the
1683 input file is assembler or pre-processed-assembler independent of
1684 the actual filename extension.
1686 2005-01-06 Roger Sayle <roger@eyesopen.com>
1688 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
1689 of sign or zero extended values.
1691 2005-01-06 Geoffrey Keating <geoffk@apple.com>
1693 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
1694 for IBM long double format correctly.
1696 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
1698 Fix PR tree-optimization/18792
1700 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
1701 to first_loop_depth, and use loop depth instead of loop number.
1702 (build_classic_dir_vector): Ditto.
1703 (compute_data_dependences_for_loop): Use depth, not loop number.
1704 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
1705 number. Pass in loops, instead of loop numbers.
1706 (gather_interchange_stats): Ditto.
1707 (linear_transform_loops): Ditto.
1709 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
1712 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
1713 biv_fits_mode_p, extension_within_bounds_p): New functions.
1714 (check_ext_dependent_givs): Use them.
1716 2005-01-06 Roger Sayle <roger@eyesopen.com>
1718 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
1719 more than one barrier, in which case we should delete them all.
1721 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
1723 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
1725 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
1727 Fix PR middle-end/19286
1729 * dwarf2out.c (gen_subprogram_die): If we've already tried to
1730 output this subprogram, simply ignore this attempt to do it again.
1731 (add_abstract_origin_attribute): Don't abort trying to add the abstract
1732 origin attribute if it's not possible.
1733 (gen_block_die): Don't ignore subblocks of "unused" blocks.
1734 (decls_for_scope): Ditto.
1735 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
1736 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
1738 2005-01-05 Richard Henderson <rth@redhat.com>
1741 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
1742 (bdesc_2arg): Use it.
1743 (ix86_expand_binop_builtin): Force operands into registers
1745 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
1746 ix86_expand_sse_compare, ix86_expand_sse_comi,
1747 ix86_expand_builtin): Likewise.
1749 2005-01-05 Richard Henderson <rth@redhat.com>
1751 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
1753 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
1754 (mulv4hi3): Set itanium_class mmmul.
1755 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
1756 (fpack): Rename from fpack_sfsf.
1758 2005-01-05 Richard Henderson <rth@redhat.com>
1761 * reload1.c (do_input_reload): Restrict the optimization deleteing
1762 a previous output reload to RELOAD_FOR_INPUT.
1764 2005-01-05 Steven Bosscher <stevenb@suse.de>
1766 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
1767 comment that we fall through after case.
1768 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
1769 (distribute_links): Likewise.
1770 * cse.c (cse_insn): Likewise.
1771 * cselib.c (cselib_invalidate_mem): Likewise.
1772 * df.c: Update comments at the top of the file.
1773 (read_modify_subreg_p): Update comments here too.
1774 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
1775 * flow.c (mark_set_1): Likewise.
1776 (mark_used_regs): Likewise.
1777 * gcse.c (mems_conflict_for_gcse_p): Likewise.
1778 (canon_list_insert): Likewise.
1779 (mark_set): Likewise.
1780 (try_replace_reg): Likewise.
1781 (store_killed_in_insn): Likewise.
1782 * loop.c (count_one_set): Likewise.
1783 (basic_induction_var): Likewise.
1784 * postreload-gcse.c (find_mem_conflicts): Likewise.
1785 * postreload.c (reload_combine_note_store): Likewise.
1786 (move2add_note_store): Likewise.
1787 * reload.c (find_equiv_reg): Likewise.
1788 (mark_referenced_resources): Likewise.
1789 * rtlanal.c (set_noop_p): Likewise.
1790 (note_stores): Likewise.
1791 (note_uses): Likewise.
1792 * sched-deps.c (sched_analyze_1): Likewise.
1793 * sched-rgn.c (check_live_1): Likewise.
1794 (update_live_1): Likewise.
1795 * config/i860/i860.c: Likewise.
1797 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
1799 (ZERO_EXTRACT): Mention that this one can be an lvalue.
1801 * doc/rtl.texi: Update documentation for bit-fields and SET.
1803 2005-01-05 Roger Sayle <roger@eyesopen.com>
1805 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
1807 2005-01-05 Jan Hubicka <jh@suse.cz>
1808 Richard Henderson <rth@redhat.com>
1811 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
1814 2005-01-05 Richard Henderson <rth@redhat.com>
1817 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
1818 (sse_shufps): Change operand 3 to const_int_operand.
1819 (sse2_storelps): Fix typo in template.
1820 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
1821 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
1822 (ix86_expand_builtin): Likewise.
1824 2005-01-05 Julian Brown <julian@codesourcery.com>
1826 * config/arm/arm.c (arm_return_in_memory): Treat complex types
1827 as aggregates for AAPCS ABIs.
1829 2005-01-05 Stan Shebs <shebs@apple.com>
1831 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
1832 Mach-O getters if ppc64.
1833 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
1835 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
1836 signed four-byte field for global code case.
1837 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
1839 2005-01-05 Roger Sayle <roger@eyesopen.com>
1842 * c-common.c: Include real.h.
1843 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
1844 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
1845 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
1846 * Makefile.in (c-common.o): Update dependencies.
1848 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1850 * c-parse.in (asm_string): Add trailing semicolon.
1852 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1854 * c-parse.in (asm_string): New. Don't allow wide strings in
1856 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
1857 asm_string instead of STRING.
1859 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1861 * c-typeck.c (constructor_no_implicit): Remove.
1862 (set_designator, process_init_element): Don't check
1863 constructor_no_implicit.
1865 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
1866 Kaz Kojima <kkojima@gcc.gnu.org>
1869 * lcm.c (create_pre_exit): New.
1870 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
1871 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
1873 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
1875 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
1876 -msoft-float in target_float_switch.
1877 * arm.c (arm_override_options): Fix processing of target_float_switch.
1879 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
1881 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
1884 2004-01-05 Uros Bizjak <uros@kss-loka.si>
1886 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
1887 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
1889 2005-01-04 Roger Sayle <roger@eyesopen.com>
1891 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
1892 if deleting the then-block allows the test-block to fallthru to the
1895 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
1898 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
1899 after the full declaration if the are in two different TUs.
1901 2005-01-04 Richard Henderson <rth@redhat.com>
1904 * tree-sra.c (generate_one_element_init): Just
1905 call gimplify_and_add.
1906 (generate_element_init): Record the
1907 new referenced variables and mark them for renaming
1908 and split out to ...
1909 (generate_element_init_1): This.
1910 (scalarize_init): Don't call push_gimplify_context/
1911 pop_gimplify_context.
1913 2005-01-04 Geoffrey Keating <geoffk@apple.com>
1915 * toplev.c (get_src_pwd): Handle failure of getpwd().
1917 2005-01-04 Roger Sayle <roger@eyesopen.com>
1919 * fold-const.c (fold_single_bit_test): Delete unreachable handling
1921 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
1922 all the cases handled by it are inverted by invert_truthvalue.
1924 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
1926 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
1927 (save_register_windowsi): Likewise.
1929 2005-01-04 Richard Henderson <rth@redhat.com>
1931 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
1933 2005-01-04 Uros Bizjak <uros@kss-loka.si>
1936 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
1937 any simplifications of vector mode comparison operators.
1938 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
1940 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
1941 Devang Patel <dpatel@apple.com>
1943 PR tree-optimization/18308
1944 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
1945 the operands before creating a new expression.
1946 * dojump.c (do_jump): Make drop_through_label available
1947 for all cases. Add expansion of COND_EXPR.
1949 2005-01-04 Ira Rosen <irar@il.ibm.com>
1951 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
1952 INTEGER_CST instead of TREE_CONSTANT.
1953 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
1954 instead of TREE_CONSTANT.
1955 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
1956 instead of TREE_CONSTANT.
1958 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
1960 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
1962 2005-01-04 Uros Bizjak <uros@kss-loka.si>
1965 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
1966 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
1968 2005-01-03 Richard Henderson <rth@redhat.com>
1970 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
1971 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
1972 fold_convert_const_real_from_real): Split out from ...
1973 (fold_convert_const): ... here.
1975 2005-01-03 Richard Henderson <rth@redhat.com>
1978 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
1979 (mov<MMXMODEI>_internal): Likewise.
1980 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
1981 Add V2SF case; use it for SSE1; don't use TI.
1982 (movdf_integer): Likewise.
1983 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
1985 2005-01-03 Ira Rosen <irar@il.ibm.com>
1987 * tree-vectorizer.c (vect_strip_conversions): New function.
1988 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
1989 check for binary class.
1991 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
1995 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
1997 * gimple-low.c (mark_blocks_with_used_vars): New function.
1998 (mark_blocks_with_used_subblocks): Ditto.
1999 (mark_used_blocks): Ditto.
2000 (pass_mark_used_blocks): New pass.
2001 * tree-inline.c: Include debug.h.
2002 (expand_call_inline): Call outlining_inline_function here.
2003 * tree-optimize.c (init_tree_optimization_passes): Add
2004 pass_mark_used_blocks.
2005 * tree-pass.h (pass_mark_used_blocks): New.
2006 * Makefile.in (tree-inline.o): Add debug.h dependency.
2008 2005-01-03 Geoffrey Keating <geoffk@apple.com>
2010 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
2011 VAR_DECLs to have weak_import attribute.
2013 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
2016 2004-01-03 Steven Bosscher <stevenb@suse.de>
2018 * passes.c (rest_of_compilation): Don't run regmove if only
2019 flag_expensive_optimizations, require flag_regmove instead.
2021 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
2023 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
2025 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
2026 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
2027 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
2028 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
2030 2005-01-03 Richard Henderson <rth@redhat.com>
2032 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
2033 (ia64_const_ok_for_letter_p): New.
2034 (ia64_const_double_ok_for_letter_p): New.
2035 (ia64_extra_constraint): New.
2036 (ia64_expand_vecint_compare): New.
2037 (ia64_expand_vcondu_v2si): New.
2038 (ia64_expand_vecint_cmov): New.
2039 (ia64_expand_vecint_minmax): New.
2040 (ia64_print_operand): Add 'v'.
2041 (ia64_preferred_reload_class): New.
2042 (ia64_vector_mode_supported_p): New.
2043 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
2044 (PREFERRED_RELOAD_CLASS): Move to function.
2045 (CONST_OK_FOR_LETTER_P): Move to function.
2046 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
2047 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
2048 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
2049 (EXTRA_CONSTRAINT): Move to function.
2050 * config/ia64/ia64.md: Include vect.md.
2051 (itanium_class): Add mmalua.
2053 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
2055 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
2056 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
2058 (const_int_2bit_operand): New.
2059 (fr_reg_or_0_operand): New.
2060 * config/ia64/ia64-modes.def: Add vector modes.
2061 * config/ia64/ia64-protos.h: Update.
2062 * config/ia64/vect.md: New file.
2064 2005-01-03 Richard Henderson <rth@redhat.com>
2066 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
2068 2005-01-03 Uros Bizjak <uros@kss-loka.si>
2071 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
2073 (log1pdf2): Change mode of operands[0,1] to DFmode.
2075 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
2077 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
2078 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
2079 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
2080 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
2081 SPARC_RELAXED_ORDERING.
2083 2005-01-03 Richard Henderson <rth@redhat.com>
2084 Uros Bizjak <uros@kss-loka.si>
2087 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
2088 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
2089 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
2091 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
2092 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
2093 const_pow2_1_to_128_operand): New predicates.
2095 2005-01-02 Greg McGary <greg@mcgary.org>
2097 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
2099 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
2100 object. [ptr_type]: Remove unused variable.
2101 Remember nearest addressable array-element or record-component
2102 when checking bitfield components. Tolerate empty BIND_EXPR.
2104 2005-01-01 Richard Henderson <rth@redhat.com>
2106 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
2107 that don't narrow the value. Fail for other conversions.
2109 2005-01-01 Richard Henderson <rth@redhat.com>
2112 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
2113 * c-lang.c (finish_file): Don't do it here.
2114 * objc/objc-act.c (objc_finish_file): Likewise.
2116 * cgraph.c (decl_assembler_name_equal): New.
2117 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
2118 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
2119 * cgraph.h (struct cgraph_varpool_node): Add next.
2120 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
2121 * varasm.c (assemble_alias): Mark the target as needed.
2123 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
2126 * function.c (get_arg_pointer_save_area): Use entry_of_function
2127 instead of get_insns.
2129 2005-01-01 Roger Sayle <roger@eyesopen.com>
2130 Andrew Pinski <pinskia@physics.uc.edu>
2131 James E. Wilson <wilson@specifixinc.com>
2133 PR rtl-optimization/12092
2134 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
2137 2005-01-01 Roger Sayle <roger@eyesopen.com>
2138 Olivier Hainque <hainque@act-europe.fr>
2140 * tree.c (int_fits_type_p): A narrower type always fits in a
2141 wider one, except for negative values into unsigned types.
2143 2005-01-01 Roger Sayle <roger@eyesopen.com>
2145 * tree.c (int_fits_type_p): Compare the result of force_fit_type
2146 with the original constant rather than require TREE_OVERFLOW.
2148 2005-01-01 Steven Bosscher <stevenb@suse.de>
2151 * c-decl.c (finish_function): If compiling C99, annotate the
2152 compiler generated return with the current file name and line 0.
2153 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
2154 the source line is greater than 0.
2155 (remove_bb): Likewise.
2157 See ChangeLog.12 for earlier changes.