1 2005-01-18 Richard Henderson <rth@redhat.com>
3 * config/i386/i386.c (ix86_expand_fp_movcc): Remove TARGET_IEEE_FP
5 * config/i386/predicates.md (sse_comparison_operator): Likewise.
7 2005-01-18 Dorit Naishlos <dorit@il.ibm.com>
9 * tree-inline.c (estimate_num_insns_1): Added cases for
10 ALIGN_INDIRECT_REF, MISALIGNED_INDIRECT_REF, and REALIGN_LOAD_EXPR.
12 2005-01-18 Eric Botcazou <ebotcazou@libertysurf.fr>
14 PR rtl-optimization/19296
15 * combine.c (simplify_comparison): Rewrite the condition under
16 which a non-paradoxical SUBREG of a PLUS can be lifted when
17 compared against a constant.
19 2005-01-18 Andi Kleen <ak@muc.de>
21 * c-typeck.c: (convert_for_assignment): Check warn_pointer_sign.
22 * c.opt (-Wpointer-sign): Add.
23 * doc/invoke.texi: (-Wpointer-sign): Add.
25 2005-01-18 Uros Bizjak <uros@kss-loka.si>
27 * config/i386/i386.c (override_options): Revert 2004-11-24 change.
28 * config/i386/i386.md (fmodsf3, dremsf3, *sinsf2, *cossf2, sincossf3,
29 *tansf3_1, tansf2, atan2sf3_1, atan2sf3, atansf2, asinsf2, acossf2,
30 logsf2, log10sf2, log2sf2, log1psf2, logbsf2, ilogbsf2, expsf2,
31 exp10sf2, exp2sf2, expm1sf2, rintsf2, floorsf2, ceilsf2, btruncsf2,
32 nearbyintsf2): Disable for TARGET_SSE_MATH. Leave patterns enabled
33 for TARGET_MIX_SSE_I387.
34 (fmoddf3, dremdf3, *sindf2, *sinextendsfdf2, *cosdf2, *cosextendsfdf2,
35 sincosdf3, *sincosextendsfdf3, *tandf3_1, tandf2, atan2df3_1,
36 atan2df3, atandf2, asindf2, acosdf2, logdf2, log10df2, log2df2,
37 log1pdf2, logbdf2, expdf2, exp10df2, exp2df2, expm1df2, rintdf2,
38 floordf2, ceildf2, btruncdf2, nearbyintdf2): Disable for
39 (TARGET_SSE2 && TARGET_SSE_MATH). Leave patterns enabled for
41 (atan2sf3, atan2df3, atan2xf3): Remove register constraints
44 2005-01-18 Uros Bizjak <uros@kss-loka.si>
47 * config/i386/mmx.md (*movv2sf_internal_rex64, *movv2sf_internal):
48 Add movaps alternative for xmm reg->reg move.
50 2005-01-17 Jeff Law <law@redhat.com>
52 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Clear SSA_NAME_VALUE
53 at the end of the main DOM loop rather than just before DOM exits.
55 2005-01-17 Diego Novillo <dnovillo@redhat.com>
57 PR tree-optimization/19121
58 * tree-ssa-alias.c (compute_flow_sensitive_aliasing): When
59 adding aliases to a name tag, also add them to the pointer's
61 * tree-ssa-copy.c (merge_alias_info): Do not merge flow
62 sensitive alias info at all. Only check that the two pointers
63 have compatible pointed-to sets.
64 * tree-ssa.c (verify_name_tags): Verify that the alias set of
65 a pointer's type tag is a superset of the alias set of the
68 2005-01-17 James E Wilson <wilson@specifixinc.com>
71 * config/ia64/ia64.md (movxf): Handle general register source. Adjust
72 comment to document why.
74 2005-01-17 Richard Henderson <rth@redhat.com>
76 * config/i386/sse.md (smaxv4sf3_finite, sse_vmsmaxv4sf3_finite,
77 sminv4sf3_finite, sse_vmsminv4sf3_finite, smaxv2df3_finite,
78 sse2_vmsmaxv2df3_finite, sminv2df3_finite,
79 sse2_vmsminv2df3_finite): New.
80 (smaxv4sf3, sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, smaxv2df3,
81 sse2_vmsmaxv2df3, sminv2df3, sse2_vmsminv2df3): Remove commutative.
82 Force op1 into register.
84 2005-01-17 Kelley Cook <kcook@gcc.gnu.org>
86 * Makefile.in (STAGEMOVESTUFF): Stage all the stamp files to prevent
87 bubblestrap from always rebuilding build directory.
89 2005-01-17 Ulrich Weigand <uweigand@de.ibm.com>
91 * config/s390/s390.md ("*clc"): Remove incorrect '='.
92 ("*cmpmem_short"): Likewise.
94 2005-01-17 Steven Bosscher <stevenb@suse.de>
96 * basic-block.h: Document BB_* flags.
97 * regrename.c (copyprop_hardreg_forward): Don't use BB_VISITED,
98 use an sbitmap instead.
99 * sched-rgn.c (compute_trg_info): Likewise.
101 2005-01-17 Richard Sandiford <rsandifo@redhat.com>
103 * config.gcc (mips64*-*-linux*): Set the default abi to n32. Remove
104 redundant target_cpu_default and extra_parts lines. Use the default
106 * config/mips/linux64.h (DRIVER_DEFAULT_ABI_SELF_SPEC): Delete.
107 (SUBTARGET_EXTRA_SPECS): Delete.
108 (DRIVER_SELF_SPECS): Remove architecture lines. Set the ABI directly.
109 (SUBTARGET_ASM_SPEC): Remove -mabi=n32 and -mabi=64 mappings.
110 * config/mips/t-linux64 (MULTILIB_DIRNAMES): Use "32" for -mabi=32
111 and "n32" for -mabi=n32.
113 2005-01-17 Ian Lance Taylor <ian@airs.com>
116 * c-typeck.c (build_c_cast): Revert patch of 2003-02-16: don't
117 fold constant variables into initial values.
119 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
121 * tree-cfg.c (tree_can_merge_blocks_p): Reorder two checks.
123 * tree-cfg.c (tree_forwarder_block_p): Speed up by walking
124 through the statements backward.
126 2005-01-17 Ian Lance Taylor <ian@airs.com>
129 * tree-inline.c (expand_call_inline): Set TREE_NO_WARNING on
130 a variable set to the return value of the inlined function.
132 2005-01-17 Mark Dettinger <dettinge@de.ibm.com>
134 * rtlanal.c (rtx_cost): Assign cost of 0 to a SUBREG
135 when modes are tieable.
137 2005-01-17 Ranjit Mathew <rmathew@hotmail.com>
139 * gthr-posix.h (__gthread_active_p): Use pthread_cancel instead
140 of pthread_create to find out if threads are enabled.
141 * gthr-posix95.h (__gthread_active_p): Likewise.
143 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
145 * common.opt (-fnew-ra): Remove.
147 * toplev.h (flag_new_regalloc): Remove.
148 * Makefile.in (ra*.*): Don't mention.
149 * passes.c (rest_of_handle_new_regalloc): Remove.
150 (rest_of_handle_combine, rest_of_compilation): Always consider
151 flag_new_regalloc as false.
152 * doc/invoke.texi: Don't document -fnew-ra.
154 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
156 * bb-reorder.c (fix_edges_for_rarely_executed_code): Remove
157 last parameter to reg_scan.
158 * loop.c (loop_optimize): Likewise.
159 * passes.c (rest_of_handle_tracer, rest_of_handle_if_conversion,
160 rest_of_handle_web, rest_of_handle_cfg, rest_of_handle_jump_bypass,
161 rest_of_handle_life, rest_of_handle_cse, rest_of_handle_cse2,
162 rest_of_handle_gcse, rest_of_handle_loop_optimize,
163 rest_of_handle_loop2, rest_of_handle_jump2): Likewise.
164 * regclass.c (reg_scan): Likewise, for the declaration.
165 * rtl.h (reg_scan): Likewise, for the prototype.
167 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
169 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Don't call
172 2005-01-16 Daniel Berlin <dberlin@dberlin.org>
174 * tree-ssa-pre.c (add_to_sets): s1 may be NULL.
175 (compute_avail): Uses don't go in tmp_gen.
177 2005-01-16 Steven Bosscher <stevenb@suse.de>
179 * ggc-page.c (ggc_alloc_stat): Use __builtin_ctzl instead of a
180 loop to look for a free slot in a page entry.
182 2005-01-16 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
185 * defaults.h (TARGET_DEFERRED_OUTPUT_DEFS): Provide default.
186 * toplev.c (compile_file): Call process_pending_assemble_output_defs
187 just before targetm.asm_out.file_end.
188 * tree.h (process_pending_assemble_output_defs): Declare.
189 * varasm.c (assemble_output_def, process_pending_assemble_output_defs):
191 (assemble_alias): Defer generation of assembly code for defines when
192 TARGET_DEFERRED_OUTPUT_DEFS is true.
193 * config/rs6000/aix41.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
194 * config/rs6000/aix43.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
195 * doc/tm.texi (TARGET_DEFERRED_OUTPUT_DEFS): document.
197 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
200 * pa.c (pa_scalar_mode_supported_p): New function.
201 (TARGET_SCALAR_MODE_SUPPORTED_P): Define.
202 * pa.h (MIN_UNITS_PER_WORD): Add comment.
204 * pa-protos.h (prefetch_operand): Delete.
205 (prefetch_cc_operand, prefetch_nocc_operand): New declations.
206 * pa.c (prefetch_operand): Delete.
207 (prefetch_cc_operand, prefetch_nocc_operand): New functions.
208 * pa.h (EXTRA_CONSTRAINT): Add `W' constraint.
209 (PREDICATE_CODES): Delete prefetch_operand. Add prefetch_cc_operand
210 and prefetch_nocc_operand.
211 * pa.md (prefetch): Rework to avoid reload problems handling short
212 displacements when a cache control completer needs to be provided.
213 (prefetch_32, prefetch_64): Delete.
214 (prefetch_cc, prefetch_nocc): New patterns.
216 2005-01-15 David Edelsohn <edelsohn@gnu.org>
218 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Revert previous change.
220 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
223 * pa.c (pa_asm_output_aligned_common): Warn if specified alignment
224 exceeds maximum alignment for global common data.
225 * pa64-hpux.h (MAX_OFILE_ALIGNMENT): Define to 32768.
226 * som.h (MAX_OFILE_ALIGNMENT): Likewise.
228 2004-01-15 Roger Sayle <roger@eyesopen.com>
230 * tree-ssa-dom.c (extract_range_from_cond): Correct condition.
232 2004-01-15 Roger Sayle <roger@eyesopen.com>
234 * harg-reg-set.h (reg_class_names): Prototype global array.
235 * regclass.c (reg_class_names): Declare here and initialize to
237 (dump_regclass): Remove local declaration of reg_class_names.
238 (regclass): Likewise.
239 * cfg.c (dump_flow_info): Likewise.
240 * ra-debug.c (reg_class_names): Likewise.
241 * regrename.c (reg_class_names): Likewise.
242 * reload.c (reg_class_names): Likewise.
243 * reload1.c (spill_failure): Likewise.
244 * config/m68hc11/m68hc11.c (reg_class_names): Likewise.
246 2005-01-15 Ulrich Weigand <uweigand@de.ibm.com>
248 * config/s390/s390.md ("reload_outti"): Remove predicate for
249 output operand. Abort if operand is not a MEM.
250 ("reload_outdi", "reload_outdf"): Likewise.
252 2005-01-15 Marc Espie <espie@openbsd.org>
254 * config.gcc (*-*-openbsd*): Set HAS_LIBC_R for OpenBSD <= 3.2.
255 (i[34567]86-*-openbsd*): Switch to ELF for OpenBSD >= 3.4.
256 * config/openbsd.h: Fix C++ includes for native configurations.
257 Add proper OS_CPP_BUILTINS. Fix libspec for recent OpenBSD.
258 Add trampoline support.
259 * config/i386/openbsdelf.h: New.
261 2005-01-15 Marc Espie <espie@openbsd.org>
263 * collect2.c (main): Explicitly parse -dynamic-linker option.
265 2005-01-15 Jakub Jelinek <jakub@redhat.com>
267 PR tree-optimization/19060
268 * tree-ssa-dom.c (extract_range_from_cond) <case LT_EXPR, GT_EXPR>:
269 Return 0 if op1 <= TYPE_MIN_VALUE () resp. op1 >= TYPE_MAX_VALUE ().
270 (simplify_cond_and_lookup_avail_expr): Add assert for dummy == 0
271 and handle extract_range_from_cond returning false.
272 * fold-const.c (fold): Optimize comparisons with min/max even for
273 width > HOST_BITS_PER_WIDE_INT.
275 2005-01-15 Ralf Corsepius <ralf.corsepius@rtems.org>
277 * config/mips/rtems.h (MIPS_DEFAULT_GVALUE): Set to 0.
278 * config/mips/t-rtems (MULTILIBS_DIRNAMES,MULTILIB_OPTIONS):
279 Remove little endian multilib variants.
280 Add mips32 multilib variant.
282 2005-01-15 Kazu Hirata <kazu@cs.umass.edu>
284 * explow.c (copy_all_reg, stabilize): Remove.
285 * expr.h: Remove the corresponding prototypes.
287 * rtlanal.c (regs_set_between_p): Remove.
288 * rtl.h: Remove the corresponding prototype.
290 2005-01-14 David Edelsohn <edelsohn@gnu.org>
292 * config/rs6000/rs6000.c (rs6k_nonimmediate_operand): Rename
293 to rs6000_nonimmediate_operand.
294 * config/rs6000/rs6000.h (PREDICATE_CODES): Rename
295 rs6k_nonimmediate_operand to rs6000_nonimmediate_operand.
296 * config/rs6000/rs6000.md (movsi_internal1): Rename predicate to
297 rs6000_nonimmediate_operand.
298 * config/rs6000/spe.md (movdf_e500_double): Rename predicate to
299 rs6000_nonimmediate_operand.
301 2005-01-14 Andrew Pinski <pinskia@physics.uc.edu>
303 * tree-ssa-alias.c (pass_may_alias): Add TODO_verify_stmts.
305 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
307 * config/rs6000/rs6000.h (CLASS_MAX_NREGS): DF goes in 1 register
309 (CANNOT_CHANGE_MODE_CLASS): Restrict DI mode changes on e500v2.
310 (PREDICATE_CODES): Add rs6k_nonimmediate_operand.
312 * config/rs6000/rs6000.c (invalid_e500_subreg): New.
313 (rs6k_nonimmediate_operand): New.
314 (rs6000_legitimate_offset_address_p): Handle DI modes on e500v2
316 (legitimate_lo_sum_address_p): Same.
317 (rs6000_legitimize_address): Same.
318 (rs6000_legitimize_reload_address): Same.
319 (rs6000_legitimate_address): Same.
320 (spe_build_register_parallel): Pass DF and DC modes in a DI
323 * config/rs6000/rs6000.md ("*movsi_internal1"): Change predicate
324 to rs6k_nonimmediate_operand.
326 * config/rs6000/spe.md ("*frob_df_di"): New.
327 ("*frob_di_df"): New.
328 ("*frob_di_df_2"): New.
329 ("*mov_sidf_e500_subreg0"): New.
330 ("*mov_sidf_e500_subreg4"): New.
331 ("*movdf_e500_double"): Change predicate to
332 rs6k_nonimmediate_operand.
334 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
336 * postreload.c (move2add_note_store): Only call
337 trunc_int_for_mode on scalar integers.
339 2005-01-14 Steven Bosscher <stevenb@suse.de>
341 * tree-ssa-dce.c (visited_control_parents): New sbitmap to
342 replace BB_VISITED uses.
343 (find_obviously_necessary_stmts): Don't clear BB_VISITED.
344 (propagate_necessity): Check the bitmap instead of BB_VISITED.
345 (tree_dce_done): Free visited_control_parents.
346 (perform_tree_ssa_dce): Allocate and clear it.
347 * tree-ssa-pre.c (compute_antic_aux): Make non-recursive.
348 (compute_antic): Iterate from here using a DFS. Use an sbitmap
349 instead of BB_VISITED.
351 2005-01-14 Kazu Hirata <kazu@cs.umass.edu>
353 * c-tree.h, coverage.h, langhooks-def.h, optabs.h, output.h,
354 rtl.h, targhooks.h, tree-flow.h, tree-mudflap.h, tree.h:
355 Remove unused prototypes.
357 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
359 * config/rs6000/rs6000.c (rs6000_stack_info): Fix caching of SPE
360 64-bit register usage.
361 (spe_func_has_64bit_regs_p): Add FIXME note.
363 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
365 * config/rs6000/rs6000.c (rs6000_generate_compare): Replace
366 flag_finite_math_only with flag_unsafe_math_optimizations.
368 2005-01-14 J. D. Johnston <jjohnst@us.ibm.com>
370 * config/s390/tpf-unwind.h (s390_fallback_frame_state): Correct
372 (__tpf_eh_return): Copy TPF private stack area from the module
373 boundary stack frame. Add check for when module addresses are
374 equal, but no stub address is found.
376 2005-01-14 Richard Earnshaw <rearnsha@arm.com>
379 * arm.h (struct machine_function): Add call_via field.
380 (thumb_call_via_label): Declare.
381 * arm.c (thumb_call_via_label): New variable.
382 (thumb_call_reg_needed): New variable.
383 (arm_output_function_epilogue): For Thumb code, output any per-function
384 call-indirect trampolines.
385 (thumb_call_via_reg): New function.
386 (arm_file_end): New function.
387 (TARGET_ASM_FILE_END): Call arm_file_end.
388 (aof_file_end): Likewise.
389 * arm-protos.h (thumb_call_via_reg): Declare.
390 * arm.md (call_reg_thumb, call_value_reg_thumb): Call
391 thumb_call_via_reg in normal case.
393 2005-01-14 Jakub Jelinek <jakub@redhat.com>
396 PR rtl-optimization/19348
397 * recog.c (peephole2_optimize): Do global life update if some peephole
398 decides it doesn't need at least one of its inputs and that change
399 influences liveness at the start of the basic block.
401 * basic-block.h (EXECUTE_IF_AND_COMPL_IN_REG_SET): Needs 2 REGSET
402 arguments instead of 1.
404 2005-01-14 Eric Botcazou <ebotcazou@libertysurf.fr>
407 * varasm.c (initializer_constant_valid_p) <ADDR_EXPR>: Return
408 zero for nested functions needing a static chain or functions
409 with a non-constant address.
411 2005-01-13 Roger Sayle <roger@eyesopen.com>
413 * simplify-rtx.c (simplify_binary_operation) <AND>: Optimize
414 (and (sign_extend X) C) into (zero_extend (and X C)).
416 2005-01-13 David O'Brien <obrien@FreeBSD.org>
418 * config/freebsd-spec.h: Make KSE pthread lib logic the default.
420 2005-01-13 Richard Henderson <rth@redhat.com>
425 * config/i386/i386.md (cmpdf, cmpsf, bunordered, bordered, buneq,
426 bunge, bungt, bunle, bunlt, bltgt): Enable for TARGET_SSE_MATH,
428 (cmpfp_i_387): Rename from cmpfp_i. Move after sse patterns.
429 (cmpfp_i_mixed): Rename from cmpfp_i_sse; use for TARGET_MIX_SSE_I387.
430 (cmpfp_i_sse): Rename from cmpfp_i_sse_only; use for TARGET_SSE_MATH.
431 (cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387): Similarly.
432 (fp_jcc_1_mixed, fp_jcc_1_sse, fp_jcc_1_387): Similarly.
433 (fp_jcc_2_mixed, fp_jcc_2_sse, fp_jcc_2_387): Similarly.
434 (fp_jcc_3_387, fp_jcc_4_387, fp_jcc_5_387, fp_jcc_6_387,
435 fp_jcc_7_387, fp_jcc_8_387): Rename from fp_jcc_N.
436 (movdicc_c_rex64): Rename with '*'.
437 (movsfcc, movdfcc): Add checks for 387 and sse math to condition.
438 (movsfcc_1_sse_min, movsfcc_1_sse_max, movsfcc_1_sse): New.
439 (movsfcc_1_387): Rename from movsfcc_1.
440 (movdfcc_1_sse_min, movdfcc_1_sse_max, movdfcc_1_sse): New.
441 (movdfcc_1, movdfcc_1_rex64): Add check for 387.
442 (sminsf3, smaxsf3, smindf3, smaxdf3): New.
443 (minsf3, minsf, minsf_nonieee, minsf_sse, mindf3, mindf,
444 mindf_nonieee, mindf_sse, maxsf3, maxsf, maxsf_nonieee, maxsf_sse,
445 maxdf3, maxdf, maxdf_nonieee, maxdf_sse, sse_movsfcc, sse_movsfcc_eq,
446 sse_movdfcc, sse_movdfcc_eq, sse_movsfcc_const0_1,
447 sse_movsfcc_const0_2, sse_movsfcc_const0_3, sse_movsfcc_const0_4,
448 sse_movdfcc_const0_1, sse_movdfcc_const0_2, sse_movdfcc_const0_3,
449 sse_movdfcc_const0_4): Remove.
450 * config/i386/i386.c (ix86_expand_fp_movcc): For TARGET_SSE_MATH,
451 recognize min/max early. Update for changed sse cmove patterns.
452 (ix86_split_sse_movcc): New.
453 * config/i386/i386-protos.h: Update.
455 2005-01-13 Steven Bosscher <stevenb@suse.de>
457 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
458 (fix_stmt_v_may_defs): Likewise.
460 2005-01-13 Richard Henderson <rth@redhat.com>
462 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
463 for converting the mask.
465 2005-01-13 David Edelsohn <edelsohn@gnu.org>
467 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
470 2005-01-13 Richard Henderson <rth@redhat.com>
472 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
473 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
474 IX86_BUILTIN_MOVDQ2Q): Remove.
475 (IX86_BUILTIN_VEC_EXT_V4SI): New.
476 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
477 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
479 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
480 (sse2_pextrw): Fix immediate constraint.
481 (sse2_loadq, sse2_loadq_rex64): Remove.
482 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
483 Use __builtin_ia32_vec_ext_<size>.
484 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
486 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
488 * function.c (assign_parm_setup_block): Look inside original
489 entry_parm when inspecting PARALLEL.
491 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
492 Joel Sherrill <joel@oarcorp.com>
495 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
496 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
497 rtems_gxx_recursive_mutex_init.
498 (__gthread_recursive_mutex_lock): New function.
499 (__gthread_recursive_mutex_trylock): Likewise.
500 (__gthread_recursive_mutex_unlock): Likewise.
502 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
504 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
506 2005-01-13 Jan Beulich <jbeulich@novell.com>
508 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
509 (ashrti3_internal): Indicate output is early clobber. Generate result
510 into output rather than first input. Use move for low word of output
511 if shift count is exactly 64.
512 (lshrti3_internal): Likewise.
514 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
518 * reload1.c (delete_output_reload): Don't delete an output reload
519 if the pseudo lives longer than a single basic block. Adjust and
520 improve wording of comment.
522 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
525 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
526 of TYPE_SIZE where we mean the number of bytes.
528 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
530 * function.c (assign_parm_setup_block): Relax condition on
531 multi-register optimization.
533 2005-01-12 Nick Clifton <nickc@redhat.com>
535 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
536 an "in_delay_slot" attribute of "no" to prevent them being used in
537 delay slots. This is forbidden because they might generate
540 2005-01-12 Alan Modra <amodra@bigpond.net.au>
543 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
546 2005-01-12 Nick Clifton <nickc@redhat.com>
548 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
549 passed on to GAS) but do not define, as GAS no longer supports or
550 needs the -m2000 option.
552 2005-01-11 Roger Sayle <roger@eyesopen.com>
554 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
555 type as args of unprototyped builtins aren't automatically promoted.
556 (fold_builtin_abs): Likewise.
558 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
560 * Makefile.in: Set a `build-warn' variable.
562 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
564 PR tree-optimization/17949
565 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
566 (find_interesting_uses_address): Use it.
568 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
570 * regrename.c (kill_value): Handle subreg's that won't simplify.
572 2005-01-11 Richard Henderson <rth@redhat.com>
575 * config/i386/i386.h (enum ix86_builtins): Move ...
576 * config/i386/i386.c: ... here.
577 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
578 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
579 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
580 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
581 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
582 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
583 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
584 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
585 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
586 IX86_BUILTIN_LOADDDUP): Remove.
587 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
588 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
589 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
590 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
591 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
592 (ix86_init_builtins): Make static.
593 (ix86_init_mmx_sse_builtins): Update for changed builtins.
594 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
595 if all the modes match. Otherwise, fake it.
596 (get_element_number, ix86_expand_vec_init_builtin,
597 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
598 (ix86_expand_builtin): Make static. Update for changed builtins.
599 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
600 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
601 (ix86_expand_vector_init_duplicate): New.
602 (ix86_expand_vector_init_low_nonzero): New.
603 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
604 Split out from ix86_expand_vector_init; handle integer modes.
605 (ix86_expand_vector_init): Use them.
606 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
607 * config/i386/i386-protos.h: Update.
608 * config/i386/predicates.md (reg_or_0_operand): New.
609 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
610 (movv2sf_internal): Likewise. And a splitter to match them all.
611 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
612 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
613 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
614 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
615 (mmx_pinsrw): Fix operand ordering.
616 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
617 rather than sse_loadss expander.
618 (movv2df splitter): Similarly.
619 (sse_loadss, sse_loadlss): Remove.
620 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
621 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
622 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
623 (sse3_movddup): Rename with '*'.
624 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
625 (sse2_loadsd): Remove.
626 (vec_dupv2df_sse3): Rename from sse3_loadddup.
627 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
628 (sse2_pinsrw): Fix argument ordering.
629 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
630 (sse2_stored): Remove 'r' destination.
631 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
632 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
633 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
634 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
635 vec_extractv16qi, vec_initv16qi): New.
637 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
638 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
639 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
641 (_mm_load_pd, _mm_store_pd): Use plain dereference.
642 (_mm_load_si128, _mm_store_si128): Likewise.
643 (_mm_load1_pd): Use _mm_set1_pd.
644 (_mm_load_sd): Use _mm_set_sd.
645 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
646 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
647 (_mm_set_epi64): Use _mm_set_epi64x.
648 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
649 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
650 _mm_setr_epi8): Use _mm_set_foo form.
651 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
652 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
653 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
654 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
655 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
656 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
657 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
658 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
659 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
660 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
661 (_mm_movedup_pd): Use _mm_shuffle_pd.
662 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
663 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
664 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
665 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
667 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
668 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
669 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
670 (_mm_store_ps): Use raw dereference.
671 (_mm_store1_ps): Use _mm_storeu_ps.
672 (_mm_storer_ps): Use _mm_store_ps.
673 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
674 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
676 2005-01-11 Stan Shebs <shebs@apple.com>
678 * config/rs6000/rs6000.c (machopic_output_stub): Issue
679 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
681 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
684 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
685 Skip all DECLs except for FIELD_DECLs.
687 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
689 * config/s390/s390.c (override_options): Return error if
690 -mbackchain, -mpacked-stack and -mhard-float are used together.
691 (s390_va_start): Remove the backchain && packed-stack special case.
692 (s390_gimplify_va_arg): Likewise.
693 * doc/invoke.texi: Remove the ABI incompatibility note.
695 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
697 * config/s390/s390.c (struct s390_frame_layout): Remove
699 (s390_frame_info, s390_emit_prologue): Replace occurrences of
700 save_backchain_p with TARGET_BACKCHAIN.
702 2005-01-11 Alan Modra <amodra@bigpond.net.au>
705 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
706 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
707 * function.c (locate_and_pad_parm): Set new field.
708 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
709 Tweak where_pad test to include "none". Always set mem align for
711 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
712 not sufficient for type.
713 (assign_parm_setup_block): If stack_parm is zero on entry, always
714 make a new stack local. Block move old stack parm if necessary
715 to new aligned stack local.
716 (assign_parm_setup_stack): Use a block move to handle
717 potentially misaligned entry_parm.
718 (assign_parms_unsplit_complex): Specify required alignment when
719 creating stack local.
720 * calls.c (compute_argument_addresses): Override alignment of stack
721 arg calculated from its type with the alignment given by
722 FUNCTION_ARG_BOUNDARY.
723 (store_one_arg): Likewise.
725 2005-01-11 Jan Beulich <jbeulich@novell.com>
727 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
728 Change respective itanium_class attribute to ialu.
729 (shladdp4_internal): New.
730 * config/ia64/predicates.md (shladd_log2_operand): New.
732 2005-01-11 Richard Henderson <rth@redhat.com>
734 * expr.c (store_constructor): Use rtvec_alloc instead of
735 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
737 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
739 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
741 2005-01-11 Alan Modra <amodra@bigpond.net.au>
743 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
744 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
745 (TARGET_RWRELOC): Define.
746 (ia64_rwreloc_section_type_flags): Delete.
747 (ia64_section_type_flags): New function.
748 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
749 (TARGET_RWRELOC): Define.
751 2005-01-10 David Mosberger <davidm@hpl.hp.com>
754 * config/ia64/ia64.c (process_set): For alloc insn, only call
755 process_epilogue is !frame_pointer_needed.
757 2005-01-10 Roger Sayle <roger@eyesopen.com>
760 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
761 unary operator and can't be treated as a binary/comparison operator.
763 2005-01-10 Richard Henderson <rth@redhat.com>
765 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
766 (ix86_return_in_memory): Likewise.
767 (function_arg_advance): Likewise. Mirror structure in function_arg
768 for choosing register to advance.
770 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
772 * tree-vectorizer.c, tree.def: Fix comment typos.
774 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
776 * config/s390/s390.c (struct processor_costs): 4 new fields:
777 ddbr, ddr, debr, der.
778 (s390_rtx_costs): More precise handling of divide instructions.
780 2005-01-10 David Edelsohn <edelsohn@gnu.org>
782 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
784 2005-01-10 Jan Beulich <jbeulich@novell.com>
786 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
787 names resulting from -ffunction-sections/-fdata-sections and linkonce
789 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
790 based on the section name. Rearrange the section name comparison logic
791 slightly so that each section name is compared against at most once.
793 2005-01-10 Ben Elliston <bje@au.ibm.com>
795 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
796 targets for which -fPIC reduces limitations on the GOT size.
798 2005-01-09 Falk Hueffner <falk@debian.org>
800 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
803 2005-01-09 David Edelsohn <edelsohn@gnu.org>
806 * collect2.c (main): Set aixrtl_flag for -brtl option.
807 (resolve_lib_name): Search for .so file extension before .a
810 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
812 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
813 (vect_analyze_data_refs_alignment): Add dump prints.
815 2005-01-09 Ira Rosen <irar@il.ibm.com>
817 * tree-vectorizer.c (vect_analyze_offset_expr): Use
818 expr_invariant_in_loop_p.
819 Initialize outputs first thing in the function.
820 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
821 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
822 (vect_analyze_pointer_ref_access): Check that the initial condition of
823 the access function is loop invariant.
825 2005-01-09 Richard Henderson <rth@redhat.com>
827 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
828 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
829 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
830 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
831 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
832 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
833 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
834 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
835 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
836 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
837 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
838 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
839 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
840 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
841 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
842 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
843 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
844 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
845 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
846 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
847 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
848 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
849 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
851 (mmx_clrdi, pavgusb): Remove.
852 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
853 with leading sse_ prefix.
854 * config/i386/sse.md: Receive them.
855 * config/i386/mmx.md: New file.
856 (MMXMODE12, MMXMODE24, mmxvecsize): New.
857 (subrv2sf3): Turn into expander for normal subtraction.
858 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
859 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
860 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
861 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
862 commutative; use ix86_binary_operator_ok.
863 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
864 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
865 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
866 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
867 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
868 Macroize from existing patterns; use ix86_binary_operator_ok.
869 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
870 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
871 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
872 with vec_select+vec_concat.
873 (mmx_pshufw, mmx_pshufw_1): Likewise.
874 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
875 (mmx_uavgv4hi3): Model correctly.
876 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
877 _mm_xor_si64): Remove casts.
879 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
881 PR tree-optimization/19224
882 * tree-scalar-evolution.c (get_instantiated_value,
883 set_instantiated_value): New functions.
884 (instantiate_parameters_1): Cache the results.
885 (instantiate_parameters, resolve_mixers): Initialize and free
888 2005-01-08 David Edelsohn <edelsohn@gnu.org>
890 * config/i386/i386.md (addhi_4): Correct reference in comment.
893 2005-01-08 Richard Henderson <rth@redhat.com>
895 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
896 (_mm_cvtsi128_si64x): Likewise.
897 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
898 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
899 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
900 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
901 * config/i386/i386-protos.h: Update.
902 * config/i386/i386.c (print_operand): Add 'H'.
903 (ix86_fixup_binary_operands): Split out from ...
904 (ix86_expand_binary_operator): ... here.
905 (ix86_fixup_binary_operands_no_copy): New.
906 (ix86_expand_fp_absneg_operator): Handle vector mode results.
907 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
908 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
909 (safe_vector_operand): Use CONST0_RTX.
910 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
911 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
912 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
913 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
915 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
916 (IX86_BUILTIN_CMPNGESS): New.
917 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
918 (attr type): Add sselog1.
919 (attr unit, attr memory): Handle it.
920 (movti, movti_internal, movti_rex64): Move near other integer moves.
921 (movtf, movtf_internal): Move near other fp moves.
922 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
923 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
924 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
925 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
926 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
927 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
928 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
929 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
930 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
931 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
932 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
933 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
934 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
935 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
936 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
937 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
938 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
939 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
940 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
941 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
942 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
943 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
944 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
945 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
946 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
947 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
948 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
949 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
950 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
951 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
952 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
953 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
954 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
955 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
956 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
957 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
958 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
959 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
960 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
961 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
962 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
963 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
964 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
965 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
967 (SSEPUSH, push<SSEPUSH>): Remove.
968 (MMXPUSH, push<MMXPUSH>): Remove.
969 (sse_movaps, sse_movaps_1, sse_movups): Remove.
970 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
971 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
972 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
973 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
974 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
975 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
976 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
977 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
978 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
979 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
980 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
981 (ppro_sse_log_V4SF): Handle sselog1.
982 * config/i386/predicates.md (const_0_to_1_operand): New.
983 (const_0_to_255_mul_8_operand): New.
984 (const_1_to_31_operand): Rename from const_int_1_31_operand.
985 (const_2_to_3_operand, const_4_to_7_operand): New.
986 * config/i386/sse.md: New file.
987 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
988 (sse_movups): Rename from sse_movups_1.
989 (sse_loadlss): Rename from sse_loadss_1.
990 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
992 (negv4sf2): Use ix86_expand_fp_absneg_operator.
993 (absv4sf2, negv2df, absv2df): New.
994 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
995 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
996 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
997 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
998 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
999 (sse3_addsubv4sf3): Model correctly.
1000 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
1001 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
1002 (sse_movhlps): Model with vec_select+vec_concat.
1003 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
1004 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
1005 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
1006 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
1007 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
1008 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
1009 sse2_pshufhw_1): Likewise.
1010 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
1011 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
1012 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
1013 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
1014 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
1015 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
1016 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
1017 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
1018 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
1019 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
1020 sminv8hi3): Mark commutative
1021 operands. Use ix86_binary_operator_ok.
1022 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
1023 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
1024 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
1025 sse2_punpcklqdq): Allow operand2 in memory.
1026 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
1027 sse2_movsd): Add memory alternatives.
1028 (sse_storelps): Turn expander into an insn; split after reload.
1029 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
1030 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
1032 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
1034 * configure.ac (DWARF-2 debug_line): Use objdump.
1035 * configure: Regenerate.
1037 2005-01-08 Jeff Law <law@redhat.com>
1038 Diego Novillo <dnovillo@redhat.com>
1040 PR tree-optimization/18241
1041 * tree-nrv.c (tree_nrv): Ignore volatile return values.
1042 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
1043 statements with volatile operands.
1044 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
1045 operands after marking a statement with has_volatile_ops.
1047 2005-01-08 Roger Sayle <roger@eyesopen.com>
1049 * tree.c (int_fits_type_p): Always honor integer constant
1050 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
1052 2005-01-08 Roger Sayle <roger@eyesopen.com>
1054 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
1055 the latent bug in rtl_delete_block has been resolved.
1057 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
1059 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
1060 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
1062 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
1064 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
1065 library as @multilib_dir@/$(SHLIB_SONAME).
1067 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
1069 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
1070 * configure: Regenerate.
1072 2005-01-07 Jakub Jelinek <jakub@redhat.com>
1074 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
1077 2005-01-07 David Edelsohn <edelsohn@gnu.org>
1080 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
1081 Convert non-word aligned offset address using ld/std into
1084 2005-01-07 Richard Henderson <rth@redhat.com>
1086 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
1088 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
1090 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
1092 2005-01-07 Jakub Jelinek <jakub@redhat.com>
1094 PR tree-optimization/19283
1095 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
1096 is not shorter than the original type.
1098 PR rtl-optimization/19012
1099 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
1101 PR rtl-optimization/18861
1102 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
1103 (FULL_STATE): Define.
1104 (mark_tablejump_edge): New function.
1105 (purge_dead_tablejump_edges): New function.
1106 (find_bb_boundaries): Use it.
1108 PR tree-optimization/18828
1109 * builtins.c (expand_builtin_next_arg): Remove argument and all
1110 the argument checking.
1111 (expand_builtin): Adjust caller.
1112 (expand_builtin_va_start): Likewise. Remove error for too many
1114 (fold_builtin_next_arg): Issue error for too many arguments.
1115 After checking arguments, replace them with magic arguments that
1116 prevent further checking of the args.
1118 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1120 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
1122 2005-01-06 Roger Sayle <roger@eyesopen.com>
1125 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
1126 and %{.S:...} (and their negative variants) to test whether the
1127 input file is assembler or pre-processed-assembler independent of
1128 the actual filename extension.
1130 2005-01-06 Roger Sayle <roger@eyesopen.com>
1132 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
1133 of sign or zero extended values.
1135 2005-01-06 Geoffrey Keating <geoffk@apple.com>
1137 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
1138 for IBM long double format correctly.
1140 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
1142 Fix PR tree-optimization/18792
1144 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
1145 to first_loop_depth, and use loop depth instead of loop number.
1146 (build_classic_dir_vector): Ditto.
1147 (compute_data_dependences_for_loop): Use depth, not loop number.
1148 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
1149 number. Pass in loops, instead of loop numbers.
1150 (gather_interchange_stats): Ditto.
1151 (linear_transform_loops): Ditto.
1153 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
1156 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
1157 biv_fits_mode_p, extension_within_bounds_p): New functions.
1158 (check_ext_dependent_givs): Use them.
1160 2005-01-06 Roger Sayle <roger@eyesopen.com>
1162 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
1163 more than one barrier, in which case we should delete them all.
1165 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
1167 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
1169 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
1171 Fix PR middle-end/19286
1173 * dwarf2out.c (gen_subprogram_die): If we've already tried to
1174 output this subprogram, simply ignore this attempt to do it again.
1175 (add_abstract_origin_attribute): Don't abort trying to add the abstract
1176 origin attribute if it's not possible.
1177 (gen_block_die): Don't ignore subblocks of "unused" blocks.
1178 (decls_for_scope): Ditto.
1179 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
1180 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
1182 2005-01-05 Richard Henderson <rth@redhat.com>
1185 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
1186 (bdesc_2arg): Use it.
1187 (ix86_expand_binop_builtin): Force operands into registers
1189 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
1190 ix86_expand_sse_compare, ix86_expand_sse_comi,
1191 ix86_expand_builtin): Likewise.
1193 2005-01-05 Richard Henderson <rth@redhat.com>
1195 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
1197 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
1198 (mulv4hi3): Set itanium_class mmmul.
1199 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
1200 (fpack): Rename from fpack_sfsf.
1202 2005-01-05 Richard Henderson <rth@redhat.com>
1205 * reload1.c (do_input_reload): Restrict the optimization deleteing
1206 a previous output reload to RELOAD_FOR_INPUT.
1208 2005-01-05 Steven Bosscher <stevenb@suse.de>
1210 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
1211 comment that we fall through after case.
1212 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
1213 (distribute_links): Likewise.
1214 * cse.c (cse_insn): Likewise.
1215 * cselib.c (cselib_invalidate_mem): Likewise.
1216 * df.c: Update comments at the top of the file.
1217 (read_modify_subreg_p): Update comments here too.
1218 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
1219 * flow.c (mark_set_1): Likewise.
1220 (mark_used_regs): Likewise.
1221 * gcse.c (mems_conflict_for_gcse_p): Likewise.
1222 (canon_list_insert): Likewise.
1223 (mark_set): Likewise.
1224 (try_replace_reg): Likewise.
1225 (store_killed_in_insn): Likewise.
1226 * loop.c (count_one_set): Likewise.
1227 (basic_induction_var): Likewise.
1228 * postreload-gcse.c (find_mem_conflicts): Likewise.
1229 * postreload.c (reload_combine_note_store): Likewise.
1230 (move2add_note_store): Likewise.
1231 * reload.c (find_equiv_reg): Likewise.
1232 (mark_referenced_resources): Likewise.
1233 * rtlanal.c (set_noop_p): Likewise.
1234 (note_stores): Likewise.
1235 (note_uses): Likewise.
1236 * sched-deps.c (sched_analyze_1): Likewise.
1237 * sched-rgn.c (check_live_1): Likewise.
1238 (update_live_1): Likewise.
1239 * config/i860/i860.c: Likewise.
1241 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
1243 (ZERO_EXTRACT): Mention that this one can be an lvalue.
1245 * doc/rtl.texi: Update documentation for bit-fields and SET.
1247 2005-01-05 Roger Sayle <roger@eyesopen.com>
1249 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
1251 2005-01-05 Jan Hubicka <jh@suse.cz>
1252 Richard Henderson <rth@redhat.com>
1255 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
1258 2005-01-05 Richard Henderson <rth@redhat.com>
1261 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
1262 (sse_shufps): Change operand 3 to const_int_operand.
1263 (sse2_storelps): Fix typo in template.
1264 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
1265 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
1266 (ix86_expand_builtin): Likewise.
1268 2005-01-05 Julian Brown <julian@codesourcery.com>
1270 * config/arm/arm.c (arm_return_in_memory): Treat complex types
1271 as aggregates for AAPCS ABIs.
1273 2005-01-05 Stan Shebs <shebs@apple.com>
1275 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
1276 Mach-O getters if ppc64.
1277 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
1279 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
1280 signed four-byte field for global code case.
1281 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
1283 2005-01-05 Roger Sayle <roger@eyesopen.com>
1286 * c-common.c: Include real.h.
1287 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
1288 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
1289 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
1290 * Makefile.in (c-common.o): Update dependencies.
1292 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1294 * c-parse.in (asm_string): Add trailing semicolon.
1296 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1298 * c-parse.in (asm_string): New. Don't allow wide strings in
1300 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
1301 asm_string instead of STRING.
1303 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1305 * c-typeck.c (constructor_no_implicit): Remove.
1306 (set_designator, process_init_element): Don't check
1307 constructor_no_implicit.
1309 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
1310 Kaz Kojima <kkojima@gcc.gnu.org>
1313 * lcm.c (create_pre_exit): New.
1314 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
1315 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
1317 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
1319 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
1320 -msoft-float in target_float_switch.
1321 * arm.c (arm_override_options): Fix processing of target_float_switch.
1323 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
1325 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
1328 2004-01-05 Uros Bizjak <uros@kss-loka.si>
1330 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
1331 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
1333 2005-01-04 Roger Sayle <roger@eyesopen.com>
1335 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
1336 if deleting the then-block allows the test-block to fallthru to the
1339 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
1342 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
1343 after the full declaration if the are in two different TUs.
1345 2005-01-04 Richard Henderson <rth@redhat.com>
1348 * tree-sra.c (generate_one_element_init): Just
1349 call gimplify_and_add.
1350 (generate_element_init): Record the
1351 new referenced variables and mark them for renaming
1352 and split out to ...
1353 (generate_element_init_1): This.
1354 (scalarize_init): Don't call push_gimplify_context/
1355 pop_gimplify_context.
1357 2005-01-04 Geoffrey Keating <geoffk@apple.com>
1359 * toplev.c (get_src_pwd): Handle failure of getpwd().
1361 2005-01-04 Roger Sayle <roger@eyesopen.com>
1363 * fold-const.c (fold_single_bit_test): Delete unreachable handling
1365 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
1366 all the cases handled by it are inverted by invert_truthvalue.
1368 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
1370 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
1371 (save_register_windowsi): Likewise.
1373 2005-01-04 Richard Henderson <rth@redhat.com>
1375 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
1377 2005-01-04 Uros Bizjak <uros@kss-loka.si>
1380 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
1381 any simplifications of vector mode comparison operators.
1382 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
1384 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
1385 Devang Patel <dpatel@apple.com>
1387 PR tree-optimization/18308
1388 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
1389 the operands before creating a new expression.
1390 * dojump.c (do_jump): Make drop_through_label available
1391 for all cases. Add expansion of COND_EXPR.
1393 2005-01-04 Ira Rosen <irar@il.ibm.com>
1395 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
1396 INTEGER_CST instead of TREE_CONSTANT.
1397 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
1398 instead of TREE_CONSTANT.
1399 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
1400 instead of TREE_CONSTANT.
1402 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
1404 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
1406 2005-01-04 Uros Bizjak <uros@kss-loka.si>
1409 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
1410 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
1412 2005-01-03 Richard Henderson <rth@redhat.com>
1414 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
1415 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
1416 fold_convert_const_real_from_real): Split out from ...
1417 (fold_convert_const): ... here.
1419 2005-01-03 Richard Henderson <rth@redhat.com>
1422 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
1423 (mov<MMXMODEI>_internal): Likewise.
1424 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
1425 Add V2SF case; use it for SSE1; don't use TI.
1426 (movdf_integer): Likewise.
1427 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
1429 2005-01-03 Ira Rosen <irar@il.ibm.com>
1431 * tree-vectorizer.c (vect_strip_conversions): New function.
1432 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
1433 check for binary class.
1435 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
1439 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
1441 * gimple-low.c (mark_blocks_with_used_vars): New function.
1442 (mark_blocks_with_used_subblocks): Ditto.
1443 (mark_used_blocks): Ditto.
1444 (pass_mark_used_blocks): New pass.
1445 * tree-inline.c: Include debug.h.
1446 (expand_call_inline): Call outlining_inline_function here.
1447 * tree-optimize.c (init_tree_optimization_passes): Add
1448 pass_mark_used_blocks.
1449 * tree-pass.h (pass_mark_used_blocks): New.
1450 * Makefile.in (tree-inline.o): Add debug.h dependency.
1452 2005-01-03 Geoffrey Keating <geoffk@apple.com>
1454 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
1455 VAR_DECLs to have weak_import attribute.
1457 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
1460 2004-01-03 Steven Bosscher <stevenb@suse.de>
1462 * passes.c (rest_of_compilation): Don't run regmove if only
1463 flag_expensive_optimizations, require flag_regmove instead.
1465 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
1467 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
1469 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
1470 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1471 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
1472 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1474 2005-01-03 Richard Henderson <rth@redhat.com>
1476 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
1477 (ia64_const_ok_for_letter_p): New.
1478 (ia64_const_double_ok_for_letter_p): New.
1479 (ia64_extra_constraint): New.
1480 (ia64_expand_vecint_compare): New.
1481 (ia64_expand_vcondu_v2si): New.
1482 (ia64_expand_vecint_cmov): New.
1483 (ia64_expand_vecint_minmax): New.
1484 (ia64_print_operand): Add 'v'.
1485 (ia64_preferred_reload_class): New.
1486 (ia64_vector_mode_supported_p): New.
1487 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
1488 (PREFERRED_RELOAD_CLASS): Move to function.
1489 (CONST_OK_FOR_LETTER_P): Move to function.
1490 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
1491 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
1492 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
1493 (EXTRA_CONSTRAINT): Move to function.
1494 * config/ia64/ia64.md: Include vect.md.
1495 (itanium_class): Add mmalua.
1497 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
1499 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
1500 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
1502 (const_int_2bit_operand): New.
1503 (fr_reg_or_0_operand): New.
1504 * config/ia64/ia64-modes.def: Add vector modes.
1505 * config/ia64/ia64-protos.h: Update.
1506 * config/ia64/vect.md: New file.
1508 2005-01-03 Richard Henderson <rth@redhat.com>
1510 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
1512 2005-01-03 Uros Bizjak <uros@kss-loka.si>
1515 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
1517 (log1pdf2): Change mode of operands[0,1] to DFmode.
1519 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
1521 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
1522 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
1523 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
1524 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
1525 SPARC_RELAXED_ORDERING.
1527 2005-01-03 Richard Henderson <rth@redhat.com>
1528 Uros Bizjak <uros@kss-loka.si>
1531 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
1532 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
1533 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
1535 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
1536 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
1537 const_pow2_1_to_128_operand): New predicates.
1539 2005-01-02 Greg McGary <greg@mcgary.org>
1541 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
1543 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
1544 object. [ptr_type]: Remove unused variable.
1545 Remember nearest addressable array-element or record-component
1546 when checking bitfield components. Tolerate empty BIND_EXPR.
1548 2005-01-01 Richard Henderson <rth@redhat.com>
1550 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
1551 that don't narrow the value. Fail for other conversions.
1553 2005-01-01 Richard Henderson <rth@redhat.com>
1556 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
1557 * c-lang.c (finish_file): Don't do it here.
1558 * objc/objc-act.c (objc_finish_file): Likewise.
1560 * cgraph.c (decl_assembler_name_equal): New.
1561 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
1562 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
1563 * cgraph.h (struct cgraph_varpool_node): Add next.
1564 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
1565 * varasm.c (assemble_alias): Mark the target as needed.
1567 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
1570 * function.c (get_arg_pointer_save_area): Use entry_of_function
1571 instead of get_insns.
1573 2005-01-01 Roger Sayle <roger@eyesopen.com>
1574 Andrew Pinski <pinskia@physics.uc.edu>
1575 James E. Wilson <wilson@specifixinc.com>
1577 PR rtl-optimization/12092
1578 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
1581 2005-01-01 Roger Sayle <roger@eyesopen.com>
1582 Olivier Hainque <hainque@act-europe.fr>
1584 * tree.c (int_fits_type_p): A narrower type always fits in a
1585 wider one, except for negative values into unsigned types.
1587 2005-01-01 Roger Sayle <roger@eyesopen.com>
1589 * tree.c (int_fits_type_p): Compare the result of force_fit_type
1590 with the original constant rather than require TREE_OVERFLOW.
1592 2005-01-01 Steven Bosscher <stevenb@suse.de>
1595 * c-decl.c (finish_function): If compiling C99, annotate the
1596 compiler generated return with the current file name and line 0.
1597 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
1598 the source line is greater than 0.
1599 (remove_bb): Likewise.
1601 See ChangeLog.12 for earlier changes.