1 2005-01-17 Steven Bosscher <stevenb@suse.de>
3 * basic-block.h: Document BB_* flags.
4 * regrename.c (copyprop_hardreg_forward): Don't use BB_VISITED,
5 use an sbitmap instead.
6 * sched-rgn.c (compute_trg_info): Likewise.
8 2005-01-17 Richard Sandiford <rsandifo@redhat.com>
10 * config.gcc (mips64*-*-linux*): Set the default abi to n32. Remove
11 redundant target_cpu_default and extra_parts lines. Use the default
13 * config/mips/linux64.h (DRIVER_DEFAULT_ABI_SELF_SPEC): Delete.
14 (SUBTARGET_EXTRA_SPECS): Delete.
15 (DRIVER_SELF_SPECS): Remove architecture lines. Set the ABI directly.
16 (SUBTARGET_ASM_SPEC): Remove -mabi=n32 and -mabi=64 mappings.
17 * config/mips/t-linux64 (MULTILIB_DIRNAMES): Use "32" for -mabi=32
18 and "n32" for -mabi=n32.
20 2005-01-17 Ian Lance Taylor <ian@airs.com>
23 * c-typeck.c (build_c_cast): Revert patch of 2003-02-16: don't
24 fold constant variables into initial values.
26 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
28 * tree-cfg.c (tree_can_merge_blocks_p): Reorder two checks.
30 * tree-cfg.c (tree_forwarder_block_p): Speed up by walking
31 through the statements backward.
33 2005-01-17 Ian Lance Taylor <ian@airs.com>
36 * tree-inline.c (expand_call_inline): Set TREE_NO_WARNING on
37 a variable set to the return value of the inlined function.
39 2005-01-17 Mark Dettinger <dettinge@de.ibm.com>
41 * rtlanal.c (rtx_cost): Assign cost of 0 to a SUBREG
42 when modes are tieable.
44 2005-01-17 Ranjit Mathew <rmathew@hotmail.com>
46 * gthr-posix.h (__gthread_active_p): Use pthread_cancel instead
47 of pthread_create to find out if threads are enabled.
48 * gthr-posix95.h (__gthread_active_p): Likewise.
50 2005-01-17 Ira Rosen <irar@il.ibm.com>
52 * fortran/f95-lang.c (gfc_init_builtin_functions): Call
53 targetm.init_builtins.
55 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
57 * common.opt (-fnew-ra): Remove.
59 * toplev.h (flag_new_regalloc): Remove.
60 * Makefile.in (ra*.*): Don't mention.
61 * passes.c (rest_of_handle_new_regalloc): Remove.
62 (rest_of_handle_combine, rest_of_compilation): Always consider
63 flag_new_regalloc as false.
64 * doc/invoke.texi: Don't document -fnew-ra.
66 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
68 * bb-reorder.c (fix_edges_for_rarely_executed_code): Remove
69 last parameter to reg_scan.
70 * loop.c (loop_optimize): Likewise.
71 * passes.c (rest_of_handle_tracer, rest_of_handle_if_conversion,
72 rest_of_handle_web, rest_of_handle_cfg, rest_of_handle_jump_bypass,
73 rest_of_handle_life, rest_of_handle_cse, rest_of_handle_cse2,
74 rest_of_handle_gcse, rest_of_handle_loop_optimize,
75 rest_of_handle_loop2, rest_of_handle_jump2): Likewise.
76 * regclass.c (reg_scan): Likewise, for the declaration.
77 * rtl.h (reg_scan): Likewise, for the prototype.
79 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
81 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Don't call
84 2005-01-16 Daniel Berlin <dberlin@dberlin.org>
86 * tree-ssa-pre.c (add_to_sets): s1 may be NULL.
87 (compute_avail): Uses don't go in tmp_gen.
89 2005-01-16 Steven Bosscher <stevenb@suse.de>
91 * ggc-page.c (ggc_alloc_stat): Use __builtin_ctzl instead of a
92 loop to look for a free slot in a page entry.
94 2005-01-16 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
97 * defaults.h (TARGET_DEFERRED_OUTPUT_DEFS): Provide default.
98 * toplev.c (compile_file): Call process_pending_assemble_output_defs
99 just before targetm.asm_out.file_end.
100 * tree.h (process_pending_assemble_output_defs): Declare.
101 * varasm.c (assemble_output_def, process_pending_assemble_output_defs):
103 (assemble_alias): Defer generation of assembly code for defines when
104 TARGET_DEFERRED_OUTPUT_DEFS is true.
105 * config/rs6000/aix41.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
106 * config/rs6000/aix43.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
107 * doc/tm.texi (TARGET_DEFERRED_OUTPUT_DEFS): document.
109 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
112 * pa.c (pa_scalar_mode_supported_p): New function.
113 (TARGET_SCALAR_MODE_SUPPORTED_P): Define.
114 * pa.h (MIN_UNITS_PER_WORD): Add comment.
116 * pa-protos.h (prefetch_operand): Delete.
117 (prefetch_cc_operand, prefetch_nocc_operand): New declations.
118 * pa.c (prefetch_operand): Delete.
119 (prefetch_cc_operand, prefetch_nocc_operand): New functions.
120 * pa.h (EXTRA_CONSTRAINT): Add `W' constraint.
121 (PREDICATE_CODES): Delete prefetch_operand. Add prefetch_cc_operand
122 and prefetch_nocc_operand.
123 * pa.md (prefetch): Rework to avoid reload problems handling short
124 displacements when a cache control completer needs to be provided.
125 (prefetch_32, prefetch_64): Delete.
126 (prefetch_cc, prefetch_nocc): New patterns.
128 2005-01-15 David Edelsohn <edelsohn@gnu.org>
130 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Revert previous change.
132 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
135 * pa.c (pa_asm_output_aligned_common): Warn if specified alignment
136 exceeds maximum alignment for global common data.
137 * pa64-hpux.h (MAX_OFILE_ALIGNMENT): Define to 32768.
138 * som.h (MAX_OFILE_ALIGNMENT): Likewise.
140 2004-01-15 Roger Sayle <roger@eyesopen.com>
142 * tree-ssa-dom.c (extract_range_from_cond): Correct condition.
144 2004-01-15 Roger Sayle <roger@eyesopen.com>
146 * harg-reg-set.h (reg_class_names): Prototype global array.
147 * regclass.c (reg_class_names): Declare here and initialize to
149 (dump_regclass): Remove local declaration of reg_class_names.
150 (regclass): Likewise.
151 * cfg.c (dump_flow_info): Likewise.
152 * ra-debug.c (reg_class_names): Likewise.
153 * regrename.c (reg_class_names): Likewise.
154 * reload.c (reg_class_names): Likewise.
155 * reload1.c (spill_failure): Likewise.
156 * config/m68hc11/m68hc11.c (reg_class_names): Likewise.
158 2005-01-15 Ulrich Weigand <uweigand@de.ibm.com>
160 * config/s390/s390.md ("reload_outti"): Remove predicate for
161 output operand. Abort if operand is not a MEM.
162 ("reload_outdi", "reload_outdf"): Likewise.
164 2005-01-15 Marc Espie <espie@openbsd.org>
166 * config.gcc (*-*-openbsd*): Set HAS_LIBC_R for OpenBSD <= 3.2.
167 (i[34567]86-*-openbsd*): Switch to ELF for OpenBSD >= 3.4.
168 * config/openbsd.h: Fix C++ includes for native configurations.
169 Add proper OS_CPP_BUILTINS. Fix libspec for recent OpenBSD.
170 Add trampoline support.
171 * config/i386/openbsdelf.h: New.
173 2005-01-15 Marc Espie <espie@openbsd.org>
175 * collect2.c (main): Explicitly parse -dynamic-linker option.
177 2005-01-15 Jakub Jelinek <jakub@redhat.com>
179 PR tree-optimization/19060
180 * tree-ssa-dom.c (extract_range_from_cond) <case LT_EXPR, GT_EXPR>:
181 Return 0 if op1 <= TYPE_MIN_VALUE () resp. op1 >= TYPE_MAX_VALUE ().
182 (simplify_cond_and_lookup_avail_expr): Add assert for dummy == 0
183 and handle extract_range_from_cond returning false.
184 * fold-const.c (fold): Optimize comparisons with min/max even for
185 width > HOST_BITS_PER_WIDE_INT.
187 2005-01-15 Ralf Corsepius <ralf.corsepius@rtems.org>
189 * config/mips/rtems.h (MIPS_DEFAULT_GVALUE): Set to 0.
190 * config/mips/t-rtems (MULTILIBS_DIRNAMES,MULTILIB_OPTIONS):
191 Remove little endian multilib variants.
192 Add mips32 multilib variant.
194 2005-01-15 Kazu Hirata <kazu@cs.umass.edu>
196 * explow.c (copy_all_reg, stabilize): Remove.
197 * expr.h: Remove the corresponding prototypes.
199 * rtlanal.c (regs_set_between_p): Remove.
200 * rtl.h: Remove the corresponding prototype.
202 2005-01-14 David Edelsohn <edelsohn@gnu.org>
204 * config/rs6000/rs6000.c (rs6k_nonimmediate_operand): Rename
205 to rs6000_nonimmediate_operand.
206 * config/rs6000/rs6000.h (PREDICATE_CODES): Rename
207 rs6k_nonimmediate_operand to rs6000_nonimmediate_operand.
208 * config/rs6000/rs6000.md (movsi_internal1): Rename predicate to
209 rs6000_nonimmediate_operand.
210 * config/rs6000/spe.md (movdf_e500_double): Rename predicate to
211 rs6000_nonimmediate_operand.
213 2005-01-14 Andrew Pinski <pinskia@physics.uc.edu>
215 * tree-ssa-alias.c (pass_may_alias): Add TODO_verify_stmts.
217 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
219 * config/rs6000/rs6000.h (CLASS_MAX_NREGS): DF goes in 1 register
221 (CANNOT_CHANGE_MODE_CLASS): Restrict DI mode changes on e500v2.
222 (PREDICATE_CODES): Add rs6k_nonimmediate_operand.
224 * config/rs6000/rs6000.c (invalid_e500_subreg): New.
225 (rs6k_nonimmediate_operand): New.
226 (rs6000_legitimate_offset_address_p): Handle DI modes on e500v2
228 (legitimate_lo_sum_address_p): Same.
229 (rs6000_legitimize_address): Same.
230 (rs6000_legitimize_reload_address): Same.
231 (rs6000_legitimate_address): Same.
232 (spe_build_register_parallel): Pass DF and DC modes in a DI
235 * config/rs6000/rs6000.md ("*movsi_internal1"): Change predicate
236 to rs6k_nonimmediate_operand.
238 * config/rs6000/spe.md ("*frob_df_di"): New.
239 ("*frob_di_df"): New.
240 ("*frob_di_df_2"): New.
241 ("*mov_sidf_e500_subreg0"): New.
242 ("*mov_sidf_e500_subreg4"): New.
243 ("*movdf_e500_double"): Change predicate to
244 rs6k_nonimmediate_operand.
246 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
248 * postreload.c (move2add_note_store): Only call
249 trunc_int_for_mode on scalar integers.
251 2005-01-14 Steven Bosscher <stevenb@suse.de>
253 * tree-ssa-dce.c (visited_control_parents): New sbitmap to
254 replace BB_VISITED uses.
255 (find_obviously_necessary_stmts): Don't clear BB_VISITED.
256 (propagate_necessity): Check the bitmap instead of BB_VISITED.
257 (tree_dce_done): Free visited_control_parents.
258 (perform_tree_ssa_dce): Allocate and clear it.
259 * tree-ssa-pre.c (compute_antic_aux): Make non-recursive.
260 (compute_antic): Iterate from here using a DFS. Use an sbitmap
261 instead of BB_VISITED.
263 2005-01-14 Kazu Hirata <kazu@cs.umass.edu>
265 * c-tree.h, coverage.h, langhooks-def.h, optabs.h, output.h,
266 rtl.h, targhooks.h, tree-flow.h, tree-mudflap.h, tree.h:
267 Remove unused prototypes.
269 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
271 * config/rs6000/rs6000.c (rs6000_stack_info): Fix caching of SPE
272 64-bit register usage.
273 (spe_func_has_64bit_regs_p): Add FIXME note.
275 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
277 * config/rs6000/rs6000.c (rs6000_generate_compare): Replace
278 flag_finite_math_only with flag_unsafe_math_optimizations.
280 2005-01-14 J. D. Johnston <jjohnst@us.ibm.com>
282 * config/s390/tpf-unwind.h (s390_fallback_frame_state): Correct
284 (__tpf_eh_return): Copy TPF private stack area from the module
285 boundary stack frame. Add check for when module addresses are
286 equal, but no stub address is found.
288 2005-01-14 Richard Earnshaw <rearnsha@arm.com>
291 * arm.h (struct machine_function): Add call_via field.
292 (thumb_call_via_label): Declare.
293 * arm.c (thumb_call_via_label): New variable.
294 (thumb_call_reg_needed): New variable.
295 (arm_output_function_epilogue): For Thumb code, output any per-function
296 call-indirect trampolines.
297 (thumb_call_via_reg): New function.
298 (arm_file_end): New function.
299 (TARGET_ASM_FILE_END): Call arm_file_end.
300 (aof_file_end): Likewise.
301 * arm-protos.h (thumb_call_via_reg): Declare.
302 * arm.md (call_reg_thumb, call_value_reg_thumb): Call
303 thumb_call_via_reg in normal case.
305 2005-01-14 Jakub Jelinek <jakub@redhat.com>
308 PR rtl-optimization/19348
309 * recog.c (peephole2_optimize): Do global life update if some peephole
310 decides it doesn't need at least one of its inputs and that change
311 influences liveness at the start of the basic block.
313 * basic-block.h (EXECUTE_IF_AND_COMPL_IN_REG_SET): Needs 2 REGSET
314 arguments instead of 1.
316 2005-01-14 Eric Botcazou <ebotcazou@libertysurf.fr>
319 * varasm.c (initializer_constant_valid_p) <ADDR_EXPR>: Return
320 zero for nested functions needing a static chain or functions
321 with a non-constant address.
323 2005-01-13 Roger Sayle <roger@eyesopen.com>
325 * simplify-rtx.c (simplify_binary_operation) <AND>: Optimize
326 (and (sign_extend X) C) into (zero_extend (and X C)).
328 2005-01-13 David O'Brien <obrien@FreeBSD.org>
330 * config/freebsd-spec.h: Make KSE pthread lib logic the default.
332 2005-01-13 Richard Henderson <rth@redhat.com>
337 * config/i386/i386.md (cmpdf, cmpsf, bunordered, bordered, buneq,
338 bunge, bungt, bunle, bunlt, bltgt): Enable for TARGET_SSE_MATH,
340 (cmpfp_i_387): Rename from cmpfp_i. Move after sse patterns.
341 (cmpfp_i_mixed): Rename from cmpfp_i_sse; use for TARGET_MIX_SSE_I387.
342 (cmpfp_i_sse): Rename from cmpfp_i_sse_only; use for TARGET_SSE_MATH.
343 (cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387): Similarly.
344 (fp_jcc_1_mixed, fp_jcc_1_sse, fp_jcc_1_387): Similarly.
345 (fp_jcc_2_mixed, fp_jcc_2_sse, fp_jcc_2_387): Similarly.
346 (fp_jcc_3_387, fp_jcc_4_387, fp_jcc_5_387, fp_jcc_6_387,
347 fp_jcc_7_387, fp_jcc_8_387): Rename from fp_jcc_N.
348 (movdicc_c_rex64): Rename with '*'.
349 (movsfcc, movdfcc): Add checks for 387 and sse math to condition.
350 (movsfcc_1_sse_min, movsfcc_1_sse_max, movsfcc_1_sse): New.
351 (movsfcc_1_387): Rename from movsfcc_1.
352 (movdfcc_1_sse_min, movdfcc_1_sse_max, movdfcc_1_sse): New.
353 (movdfcc_1, movdfcc_1_rex64): Add check for 387.
354 (sminsf3, smaxsf3, smindf3, smaxdf3): New.
355 (minsf3, minsf, minsf_nonieee, minsf_sse, mindf3, mindf,
356 mindf_nonieee, mindf_sse, maxsf3, maxsf, maxsf_nonieee, maxsf_sse,
357 maxdf3, maxdf, maxdf_nonieee, maxdf_sse, sse_movsfcc, sse_movsfcc_eq,
358 sse_movdfcc, sse_movdfcc_eq, sse_movsfcc_const0_1,
359 sse_movsfcc_const0_2, sse_movsfcc_const0_3, sse_movsfcc_const0_4,
360 sse_movdfcc_const0_1, sse_movdfcc_const0_2, sse_movdfcc_const0_3,
361 sse_movdfcc_const0_4): Remove.
362 * config/i386/i386.c (ix86_expand_fp_movcc): For TARGET_SSE_MATH,
363 recognize min/max early. Update for changed sse cmove patterns.
364 (ix86_split_sse_movcc): New.
365 * config/i386/i386-protos.h: Update.
367 2005-01-13 Steven Bosscher <stevenb@suse.de>
369 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
370 (fix_stmt_v_may_defs): Likewise.
372 2005-01-13 Richard Henderson <rth@redhat.com>
374 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
375 for converting the mask.
377 2005-01-13 David Edelsohn <edelsohn@gnu.org>
379 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
382 2005-01-13 Richard Henderson <rth@redhat.com>
384 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
385 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
386 IX86_BUILTIN_MOVDQ2Q): Remove.
387 (IX86_BUILTIN_VEC_EXT_V4SI): New.
388 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
389 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
391 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
392 (sse2_pextrw): Fix immediate constraint.
393 (sse2_loadq, sse2_loadq_rex64): Remove.
394 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
395 Use __builtin_ia32_vec_ext_<size>.
396 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
398 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
400 * function.c (assign_parm_setup_block): Look inside original
401 entry_parm when inspecting PARALLEL.
403 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
404 Joel Sherrill <joel@oarcorp.com>
407 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
408 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
409 rtems_gxx_recursive_mutex_init.
410 (__gthread_recursive_mutex_lock): New function.
411 (__gthread_recursive_mutex_trylock): Likewise.
412 (__gthread_recursive_mutex_unlock): Likewise.
414 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
416 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
418 2005-01-13 Jan Beulich <jbeulich@novell.com>
420 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
421 (ashrti3_internal): Indicate output is early clobber. Generate result
422 into output rather than first input. Use move for low word of output
423 if shift count is exactly 64.
424 (lshrti3_internal): Likewise.
426 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
430 * reload1.c (delete_output_reload): Don't delete an output reload
431 if the pseudo lives longer than a single basic block. Adjust and
432 improve wording of comment.
434 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
437 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
438 of TYPE_SIZE where we mean the number of bytes.
440 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
442 * function.c (assign_parm_setup_block): Relax condition on
443 multi-register optimization.
445 2005-01-12 Nick Clifton <nickc@redhat.com>
447 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
448 an "in_delay_slot" attribute of "no" to prevent them being used in
449 delay slots. This is forbidden because they might generate
452 2005-01-12 Alan Modra <amodra@bigpond.net.au>
455 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
458 2005-01-12 Nick Clifton <nickc@redhat.com>
460 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
461 passed on to GAS) but do not define, as GAS no longer supports or
462 needs the -m2000 option.
464 2005-01-11 Roger Sayle <roger@eyesopen.com>
466 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
467 type as args of unprototyped builtins aren't automatically promoted.
468 (fold_builtin_abs): Likewise.
470 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
472 * Makefile.in: Set a `build-warn' variable.
474 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
476 PR tree-optimization/17949
477 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
478 (find_interesting_uses_address): Use it.
480 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
482 * regrename.c (kill_value): Handle subreg's that won't simplify.
484 2005-01-11 Richard Henderson <rth@redhat.com>
487 * config/i386/i386.h (enum ix86_builtins): Move ...
488 * config/i386/i386.c: ... here.
489 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
490 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
491 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
492 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
493 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
494 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
495 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
496 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
497 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
498 IX86_BUILTIN_LOADDDUP): Remove.
499 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
500 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
501 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
502 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
503 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
504 (ix86_init_builtins): Make static.
505 (ix86_init_mmx_sse_builtins): Update for changed builtins.
506 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
507 if all the modes match. Otherwise, fake it.
508 (get_element_number, ix86_expand_vec_init_builtin,
509 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
510 (ix86_expand_builtin): Make static. Update for changed builtins.
511 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
512 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
513 (ix86_expand_vector_init_duplicate): New.
514 (ix86_expand_vector_init_low_nonzero): New.
515 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
516 Split out from ix86_expand_vector_init; handle integer modes.
517 (ix86_expand_vector_init): Use them.
518 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
519 * config/i386/i386-protos.h: Update.
520 * config/i386/predicates.md (reg_or_0_operand): New.
521 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
522 (movv2sf_internal): Likewise. And a splitter to match them all.
523 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
524 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
525 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
526 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
527 (mmx_pinsrw): Fix operand ordering.
528 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
529 rather than sse_loadss expander.
530 (movv2df splitter): Similarly.
531 (sse_loadss, sse_loadlss): Remove.
532 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
533 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
534 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
535 (sse3_movddup): Rename with '*'.
536 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
537 (sse2_loadsd): Remove.
538 (vec_dupv2df_sse3): Rename from sse3_loadddup.
539 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
540 (sse2_pinsrw): Fix argument ordering.
541 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
542 (sse2_stored): Remove 'r' destination.
543 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
544 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
545 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
546 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
547 vec_extractv16qi, vec_initv16qi): New.
549 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
550 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
551 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
553 (_mm_load_pd, _mm_store_pd): Use plain dereference.
554 (_mm_load_si128, _mm_store_si128): Likewise.
555 (_mm_load1_pd): Use _mm_set1_pd.
556 (_mm_load_sd): Use _mm_set_sd.
557 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
558 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
559 (_mm_set_epi64): Use _mm_set_epi64x.
560 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
561 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
562 _mm_setr_epi8): Use _mm_set_foo form.
563 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
564 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
565 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
566 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
567 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
568 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
569 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
570 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
571 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
572 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
573 (_mm_movedup_pd): Use _mm_shuffle_pd.
574 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
575 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
576 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
577 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
579 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
580 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
581 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
582 (_mm_store_ps): Use raw dereference.
583 (_mm_store1_ps): Use _mm_storeu_ps.
584 (_mm_storer_ps): Use _mm_store_ps.
585 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
586 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
588 2005-01-11 Stan Shebs <shebs@apple.com>
590 * config/rs6000/rs6000.c (machopic_output_stub): Issue
591 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
593 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
596 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
597 Skip all DECLs except for FIELD_DECLs.
599 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
601 * config/s390/s390.c (override_options): Return error if
602 -mbackchain, -mpacked-stack and -mhard-float are used together.
603 (s390_va_start): Remove the backchain && packed-stack special case.
604 (s390_gimplify_va_arg): Likewise.
605 * doc/invoke.texi: Remove the ABI incompatibility note.
607 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
609 * config/s390/s390.c (struct s390_frame_layout): Remove
611 (s390_frame_info, s390_emit_prologue): Replace occurrences of
612 save_backchain_p with TARGET_BACKCHAIN.
614 2005-01-11 Alan Modra <amodra@bigpond.net.au>
617 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
618 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
619 * function.c (locate_and_pad_parm): Set new field.
620 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
621 Tweak where_pad test to include "none". Always set mem align for
623 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
624 not sufficient for type.
625 (assign_parm_setup_block): If stack_parm is zero on entry, always
626 make a new stack local. Block move old stack parm if necessary
627 to new aligned stack local.
628 (assign_parm_setup_stack): Use a block move to handle
629 potentially misaligned entry_parm.
630 (assign_parms_unsplit_complex): Specify required alignment when
631 creating stack local.
632 * calls.c (compute_argument_addresses): Override alignment of stack
633 arg calculated from its type with the alignment given by
634 FUNCTION_ARG_BOUNDARY.
635 (store_one_arg): Likewise.
637 2005-01-11 Jan Beulich <jbeulich@novell.com>
639 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
640 Change respective itanium_class attribute to ialu.
641 (shladdp4_internal): New.
642 * config/ia64/predicates.md (shladd_log2_operand): New.
644 2005-01-11 Richard Henderson <rth@redhat.com>
646 * expr.c (store_constructor): Use rtvec_alloc instead of
647 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
649 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
651 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
653 2005-01-11 Alan Modra <amodra@bigpond.net.au>
655 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
656 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
657 (TARGET_RWRELOC): Define.
658 (ia64_rwreloc_section_type_flags): Delete.
659 (ia64_section_type_flags): New function.
660 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
661 (TARGET_RWRELOC): Define.
663 2005-01-10 David Mosberger <davidm@hpl.hp.com>
666 * config/ia64/ia64.c (process_set): For alloc insn, only call
667 process_epilogue is !frame_pointer_needed.
669 2005-01-10 Roger Sayle <roger@eyesopen.com>
672 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
673 unary operator and can't be treated as a binary/comparison operator.
675 2005-01-10 Richard Henderson <rth@redhat.com>
677 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
678 (ix86_return_in_memory): Likewise.
679 (function_arg_advance): Likewise. Mirror structure in function_arg
680 for choosing register to advance.
682 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
684 * tree-vectorizer.c, tree.def: Fix comment typos.
686 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
688 * config/s390/s390.c (struct processor_costs): 4 new fields:
689 ddbr, ddr, debr, der.
690 (s390_rtx_costs): More precise handling of divide instructions.
692 2005-01-10 David Edelsohn <edelsohn@gnu.org>
694 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
696 2005-01-10 Jan Beulich <jbeulich@novell.com>
698 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
699 names resulting from -ffunction-sections/-fdata-sections and linkonce
701 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
702 based on the section name. Rearrange the section name comparison logic
703 slightly so that each section name is compared against at most once.
705 2005-01-10 Ben Elliston <bje@au.ibm.com>
707 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
708 targets for which -fPIC reduces limitations on the GOT size.
710 2005-01-09 Falk Hueffner <falk@debian.org>
712 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
715 2005-01-09 David Edelsohn <edelsohn@gnu.org>
718 * collect2.c (main): Set aixrtl_flag for -brtl option.
719 (resolve_lib_name): Search for .so file extension before .a
722 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
724 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
725 (vect_analyze_data_refs_alignment): Add dump prints.
727 2005-01-09 Ira Rosen <irar@il.ibm.com>
729 * tree-vectorizer.c (vect_analyze_offset_expr): Use
730 expr_invariant_in_loop_p.
731 Initialize outputs first thing in the function.
732 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
733 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
734 (vect_analyze_pointer_ref_access): Check that the initial condition of
735 the access function is loop invariant.
737 2005-01-09 Richard Henderson <rth@redhat.com>
739 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
740 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
741 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
742 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
743 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
744 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
745 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
746 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
747 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
748 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
749 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
750 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
751 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
752 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
753 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
754 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
755 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
756 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
757 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
758 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
759 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
760 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
761 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
763 (mmx_clrdi, pavgusb): Remove.
764 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
765 with leading sse_ prefix.
766 * config/i386/sse.md: Receive them.
767 * config/i386/mmx.md: New file.
768 (MMXMODE12, MMXMODE24, mmxvecsize): New.
769 (subrv2sf3): Turn into expander for normal subtraction.
770 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
771 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
772 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
773 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
774 commutative; use ix86_binary_operator_ok.
775 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
776 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
777 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
778 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
779 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
780 Macroize from existing patterns; use ix86_binary_operator_ok.
781 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
782 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
783 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
784 with vec_select+vec_concat.
785 (mmx_pshufw, mmx_pshufw_1): Likewise.
786 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
787 (mmx_uavgv4hi3): Model correctly.
788 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
789 _mm_xor_si64): Remove casts.
791 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
793 PR tree-optimization/19224
794 * tree-scalar-evolution.c (get_instantiated_value,
795 set_instantiated_value): New functions.
796 (instantiate_parameters_1): Cache the results.
797 (instantiate_parameters, resolve_mixers): Initialize and free
800 2005-01-08 David Edelsohn <edelsohn@gnu.org>
802 * config/i386/i386.md (addhi_4): Correct reference in comment.
805 2005-01-08 Richard Henderson <rth@redhat.com>
807 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
808 (_mm_cvtsi128_si64x): Likewise.
809 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
810 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
811 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
812 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
813 * config/i386/i386-protos.h: Update.
814 * config/i386/i386.c (print_operand): Add 'H'.
815 (ix86_fixup_binary_operands): Split out from ...
816 (ix86_expand_binary_operator): ... here.
817 (ix86_fixup_binary_operands_no_copy): New.
818 (ix86_expand_fp_absneg_operator): Handle vector mode results.
819 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
820 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
821 (safe_vector_operand): Use CONST0_RTX.
822 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
823 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
824 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
825 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
827 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
828 (IX86_BUILTIN_CMPNGESS): New.
829 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
830 (attr type): Add sselog1.
831 (attr unit, attr memory): Handle it.
832 (movti, movti_internal, movti_rex64): Move near other integer moves.
833 (movtf, movtf_internal): Move near other fp moves.
834 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
835 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
836 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
837 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
838 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
839 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
840 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
841 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
842 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
843 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
844 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
845 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
846 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
847 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
848 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
849 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
850 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
851 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
852 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
853 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
854 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
855 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
856 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
857 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
858 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
859 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
860 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
861 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
862 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
863 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
864 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
865 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
866 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
867 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
868 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
869 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
870 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
871 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
872 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
873 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
874 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
875 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
876 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
877 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
879 (SSEPUSH, push<SSEPUSH>): Remove.
880 (MMXPUSH, push<MMXPUSH>): Remove.
881 (sse_movaps, sse_movaps_1, sse_movups): Remove.
882 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
883 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
884 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
885 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
886 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
887 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
888 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
889 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
890 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
891 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
892 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
893 (ppro_sse_log_V4SF): Handle sselog1.
894 * config/i386/predicates.md (const_0_to_1_operand): New.
895 (const_0_to_255_mul_8_operand): New.
896 (const_1_to_31_operand): Rename from const_int_1_31_operand.
897 (const_2_to_3_operand, const_4_to_7_operand): New.
898 * config/i386/sse.md: New file.
899 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
900 (sse_movups): Rename from sse_movups_1.
901 (sse_loadlss): Rename from sse_loadss_1.
902 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
904 (negv4sf2): Use ix86_expand_fp_absneg_operator.
905 (absv4sf2, negv2df, absv2df): New.
906 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
907 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
908 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
909 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
910 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
911 (sse3_addsubv4sf3): Model correctly.
912 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
913 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
914 (sse_movhlps): Model with vec_select+vec_concat.
915 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
916 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
917 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
918 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
919 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
920 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
921 sse2_pshufhw_1): Likewise.
922 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
923 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
924 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
925 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
926 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
927 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
928 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
929 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
930 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
931 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
932 sminv8hi3): Mark commutative
933 operands. Use ix86_binary_operator_ok.
934 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
935 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
936 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
937 sse2_punpcklqdq): Allow operand2 in memory.
938 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
939 sse2_movsd): Add memory alternatives.
940 (sse_storelps): Turn expander into an insn; split after reload.
941 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
942 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
944 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
946 * configure.ac (DWARF-2 debug_line): Use objdump.
947 * configure: Regenerate.
949 2005-01-08 Jeff Law <law@redhat.com>
950 Diego Novillo <dnovillo@redhat.com>
952 PR tree-optimization/18241
953 * tree-nrv.c (tree_nrv): Ignore volatile return values.
954 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
955 statements with volatile operands.
956 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
957 operands after marking a statement with has_volatile_ops.
959 2005-01-08 Roger Sayle <roger@eyesopen.com>
961 * tree.c (int_fits_type_p): Always honor integer constant
962 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
964 2005-01-08 Roger Sayle <roger@eyesopen.com>
966 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
967 the latent bug in rtl_delete_block has been resolved.
969 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
971 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
972 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
974 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
976 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
977 library as @multilib_dir@/$(SHLIB_SONAME).
979 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
981 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
982 * configure: Regenerate.
984 2005-01-07 Jakub Jelinek <jakub@redhat.com>
986 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
989 2005-01-07 David Edelsohn <edelsohn@gnu.org>
992 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
993 Convert non-word aligned offset address using ld/std into
996 2005-01-07 Richard Henderson <rth@redhat.com>
998 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
1000 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
1002 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
1004 2005-01-07 Jakub Jelinek <jakub@redhat.com>
1006 PR tree-optimization/19283
1007 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
1008 is not shorter than the original type.
1010 PR rtl-optimization/19012
1011 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
1013 PR rtl-optimization/18861
1014 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
1015 (FULL_STATE): Define.
1016 (mark_tablejump_edge): New function.
1017 (purge_dead_tablejump_edges): New function.
1018 (find_bb_boundaries): Use it.
1020 PR tree-optimization/18828
1021 * builtins.c (expand_builtin_next_arg): Remove argument and all
1022 the argument checking.
1023 (expand_builtin): Adjust caller.
1024 (expand_builtin_va_start): Likewise. Remove error for too many
1026 (fold_builtin_next_arg): Issue error for too many arguments.
1027 After checking arguments, replace them with magic arguments that
1028 prevent further checking of the args.
1030 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
1032 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
1034 2005-01-06 Roger Sayle <roger@eyesopen.com>
1037 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
1038 and %{.S:...} (and their negative variants) to test whether the
1039 input file is assembler or pre-processed-assembler independent of
1040 the actual filename extension.
1042 2005-01-06 Roger Sayle <roger@eyesopen.com>
1044 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
1045 of sign or zero extended values.
1047 2005-01-06 Geoffrey Keating <geoffk@apple.com>
1049 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
1050 for IBM long double format correctly.
1052 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
1054 Fix PR tree-optimization/18792
1056 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
1057 to first_loop_depth, and use loop depth instead of loop number.
1058 (build_classic_dir_vector): Ditto.
1059 (compute_data_dependences_for_loop): Use depth, not loop number.
1060 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
1061 number. Pass in loops, instead of loop numbers.
1062 (gather_interchange_stats): Ditto.
1063 (linear_transform_loops): Ditto.
1065 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
1068 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
1069 biv_fits_mode_p, extension_within_bounds_p): New functions.
1070 (check_ext_dependent_givs): Use them.
1072 2005-01-06 Roger Sayle <roger@eyesopen.com>
1074 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
1075 more than one barrier, in which case we should delete them all.
1077 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
1079 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
1081 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
1083 Fix PR middle-end/19286
1085 * dwarf2out.c (gen_subprogram_die): If we've already tried to
1086 output this subprogram, simply ignore this attempt to do it again.
1087 (add_abstract_origin_attribute): Don't abort trying to add the abstract
1088 origin attribute if it's not possible.
1089 (gen_block_die): Don't ignore subblocks of "unused" blocks.
1090 (decls_for_scope): Ditto.
1091 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
1092 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
1094 2005-01-05 Richard Henderson <rth@redhat.com>
1097 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
1098 (bdesc_2arg): Use it.
1099 (ix86_expand_binop_builtin): Force operands into registers
1101 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
1102 ix86_expand_sse_compare, ix86_expand_sse_comi,
1103 ix86_expand_builtin): Likewise.
1105 2005-01-05 Richard Henderson <rth@redhat.com>
1107 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
1109 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
1110 (mulv4hi3): Set itanium_class mmmul.
1111 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
1112 (fpack): Rename from fpack_sfsf.
1114 2005-01-05 Richard Henderson <rth@redhat.com>
1117 * reload1.c (do_input_reload): Restrict the optimization deleteing
1118 a previous output reload to RELOAD_FOR_INPUT.
1120 2005-01-05 Steven Bosscher <stevenb@suse.de>
1122 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
1123 comment that we fall through after case.
1124 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
1125 (distribute_links): Likewise.
1126 * cse.c (cse_insn): Likewise.
1127 * cselib.c (cselib_invalidate_mem): Likewise.
1128 * df.c: Update comments at the top of the file.
1129 (read_modify_subreg_p): Update comments here too.
1130 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
1131 * flow.c (mark_set_1): Likewise.
1132 (mark_used_regs): Likewise.
1133 * gcse.c (mems_conflict_for_gcse_p): Likewise.
1134 (canon_list_insert): Likewise.
1135 (mark_set): Likewise.
1136 (try_replace_reg): Likewise.
1137 (store_killed_in_insn): Likewise.
1138 * loop.c (count_one_set): Likewise.
1139 (basic_induction_var): Likewise.
1140 * postreload-gcse.c (find_mem_conflicts): Likewise.
1141 * postreload.c (reload_combine_note_store): Likewise.
1142 (move2add_note_store): Likewise.
1143 * reload.c (find_equiv_reg): Likewise.
1144 (mark_referenced_resources): Likewise.
1145 * rtlanal.c (set_noop_p): Likewise.
1146 (note_stores): Likewise.
1147 (note_uses): Likewise.
1148 * sched-deps.c (sched_analyze_1): Likewise.
1149 * sched-rgn.c (check_live_1): Likewise.
1150 (update_live_1): Likewise.
1151 * config/i860/i860.c: Likewise.
1153 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
1155 (ZERO_EXTRACT): Mention that this one can be an lvalue.
1157 * doc/rtl.texi: Update documentation for bit-fields and SET.
1159 2005-01-05 Roger Sayle <roger@eyesopen.com>
1161 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
1163 2005-01-05 Jan Hubicka <jh@suse.cz>
1164 Richard Henderson <rth@redhat.com>
1167 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
1170 2005-01-05 Richard Henderson <rth@redhat.com>
1173 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
1174 (sse_shufps): Change operand 3 to const_int_operand.
1175 (sse2_storelps): Fix typo in template.
1176 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
1177 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
1178 (ix86_expand_builtin): Likewise.
1180 2005-01-05 Julian Brown <julian@codesourcery.com>
1182 * config/arm/arm.c (arm_return_in_memory): Treat complex types
1183 as aggregates for AAPCS ABIs.
1185 2005-01-05 Stan Shebs <shebs@apple.com>
1187 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
1188 Mach-O getters if ppc64.
1189 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
1191 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
1192 signed four-byte field for global code case.
1193 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
1195 2005-01-05 Roger Sayle <roger@eyesopen.com>
1198 * c-common.c: Include real.h.
1199 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
1200 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
1201 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
1202 * Makefile.in (c-common.o): Update dependencies.
1204 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1206 * c-parse.in (asm_string): Add trailing semicolon.
1208 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1210 * c-parse.in (asm_string): New. Don't allow wide strings in
1212 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
1213 asm_string instead of STRING.
1215 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1217 * c-typeck.c (constructor_no_implicit): Remove.
1218 (set_designator, process_init_element): Don't check
1219 constructor_no_implicit.
1221 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
1222 Kaz Kojima <kkojima@gcc.gnu.org>
1225 * lcm.c (create_pre_exit): New.
1226 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
1227 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
1229 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
1231 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
1232 -msoft-float in target_float_switch.
1233 * arm.c (arm_override_options): Fix processing of target_float_switch.
1235 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
1237 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
1240 2004-01-05 Uros Bizjak <uros@kss-loka.si>
1242 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
1243 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
1245 2005-01-04 Roger Sayle <roger@eyesopen.com>
1247 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
1248 if deleting the then-block allows the test-block to fallthru to the
1251 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
1254 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
1255 after the full declaration if the are in two different TUs.
1257 2005-01-04 Richard Henderson <rth@redhat.com>
1260 * tree-sra.c (generate_one_element_init): Just
1261 call gimplify_and_add.
1262 (generate_element_init): Record the
1263 new referenced variables and mark them for renaming
1264 and split out to ...
1265 (generate_element_init_1): This.
1266 (scalarize_init): Don't call push_gimplify_context/
1267 pop_gimplify_context.
1269 2005-01-04 Geoffrey Keating <geoffk@apple.com>
1271 * toplev.c (get_src_pwd): Handle failure of getpwd().
1273 2005-01-04 Roger Sayle <roger@eyesopen.com>
1275 * fold-const.c (fold_single_bit_test): Delete unreachable handling
1277 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
1278 all the cases handled by it are inverted by invert_truthvalue.
1280 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
1282 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
1283 (save_register_windowsi): Likewise.
1285 2005-01-04 Richard Henderson <rth@redhat.com>
1287 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
1289 2005-01-04 Uros Bizjak <uros@kss-loka.si>
1292 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
1293 any simplifications of vector mode comparison operators.
1294 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
1296 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
1297 Devang Patel <dpatel@apple.com>
1299 PR tree-optimization/18308
1300 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
1301 the operands before creating a new expression.
1302 * dojump.c (do_jump): Make drop_through_label available
1303 for all cases. Add expansion of COND_EXPR.
1305 2005-01-04 Ira Rosen <irar@il.ibm.com>
1307 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
1308 INTEGER_CST instead of TREE_CONSTANT.
1309 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
1310 instead of TREE_CONSTANT.
1311 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
1312 instead of TREE_CONSTANT.
1314 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
1316 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
1318 2005-01-04 Uros Bizjak <uros@kss-loka.si>
1321 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
1322 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
1324 2005-01-03 Richard Henderson <rth@redhat.com>
1326 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
1327 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
1328 fold_convert_const_real_from_real): Split out from ...
1329 (fold_convert_const): ... here.
1331 2005-01-03 Richard Henderson <rth@redhat.com>
1334 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
1335 (mov<MMXMODEI>_internal): Likewise.
1336 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
1337 Add V2SF case; use it for SSE1; don't use TI.
1338 (movdf_integer): Likewise.
1339 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
1341 2005-01-03 Ira Rosen <irar@il.ibm.com>
1343 * tree-vectorizer.c (vect_strip_conversions): New function.
1344 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
1345 check for binary class.
1347 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
1351 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
1353 * gimple-low.c (mark_blocks_with_used_vars): New function.
1354 (mark_blocks_with_used_subblocks): Ditto.
1355 (mark_used_blocks): Ditto.
1356 (pass_mark_used_blocks): New pass.
1357 * tree-inline.c: Include debug.h.
1358 (expand_call_inline): Call outlining_inline_function here.
1359 * tree-optimize.c (init_tree_optimization_passes): Add
1360 pass_mark_used_blocks.
1361 * tree-pass.h (pass_mark_used_blocks): New.
1362 * Makefile.in (tree-inline.o): Add debug.h dependency.
1364 2005-01-03 Geoffrey Keating <geoffk@apple.com>
1366 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
1367 VAR_DECLs to have weak_import attribute.
1369 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
1372 2004-01-03 Steven Bosscher <stevenb@suse.de>
1374 * passes.c (rest_of_compilation): Don't run regmove if only
1375 flag_expensive_optimizations, require flag_regmove instead.
1377 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
1379 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
1381 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
1382 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1383 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
1384 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1386 2005-01-03 Richard Henderson <rth@redhat.com>
1388 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
1389 (ia64_const_ok_for_letter_p): New.
1390 (ia64_const_double_ok_for_letter_p): New.
1391 (ia64_extra_constraint): New.
1392 (ia64_expand_vecint_compare): New.
1393 (ia64_expand_vcondu_v2si): New.
1394 (ia64_expand_vecint_cmov): New.
1395 (ia64_expand_vecint_minmax): New.
1396 (ia64_print_operand): Add 'v'.
1397 (ia64_preferred_reload_class): New.
1398 (ia64_vector_mode_supported_p): New.
1399 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
1400 (PREFERRED_RELOAD_CLASS): Move to function.
1401 (CONST_OK_FOR_LETTER_P): Move to function.
1402 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
1403 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
1404 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
1405 (EXTRA_CONSTRAINT): Move to function.
1406 * config/ia64/ia64.md: Include vect.md.
1407 (itanium_class): Add mmalua.
1409 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
1411 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
1412 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
1414 (const_int_2bit_operand): New.
1415 (fr_reg_or_0_operand): New.
1416 * config/ia64/ia64-modes.def: Add vector modes.
1417 * config/ia64/ia64-protos.h: Update.
1418 * config/ia64/vect.md: New file.
1420 2005-01-03 Richard Henderson <rth@redhat.com>
1422 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
1424 2005-01-03 Uros Bizjak <uros@kss-loka.si>
1427 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
1429 (log1pdf2): Change mode of operands[0,1] to DFmode.
1431 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
1433 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
1434 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
1435 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
1436 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
1437 SPARC_RELAXED_ORDERING.
1439 2005-01-03 Richard Henderson <rth@redhat.com>
1440 Uros Bizjak <uros@kss-loka.si>
1443 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
1444 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
1445 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
1447 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
1448 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
1449 const_pow2_1_to_128_operand): New predicates.
1451 2005-01-02 Greg McGary <greg@mcgary.org>
1453 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
1455 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
1456 object. [ptr_type]: Remove unused variable.
1457 Remember nearest addressable array-element or record-component
1458 when checking bitfield components. Tolerate empty BIND_EXPR.
1460 2005-01-01 Richard Henderson <rth@redhat.com>
1462 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
1463 that don't narrow the value. Fail for other conversions.
1465 2005-01-01 Richard Henderson <rth@redhat.com>
1468 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
1469 * c-lang.c (finish_file): Don't do it here.
1470 * objc/objc-act.c (objc_finish_file): Likewise.
1472 * cgraph.c (decl_assembler_name_equal): New.
1473 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
1474 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
1475 * cgraph.h (struct cgraph_varpool_node): Add next.
1476 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
1477 * varasm.c (assemble_alias): Mark the target as needed.
1479 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
1482 * function.c (get_arg_pointer_save_area): Use entry_of_function
1483 instead of get_insns.
1485 2005-01-01 Roger Sayle <roger@eyesopen.com>
1486 Andrew Pinski <pinskia@physics.uc.edu>
1487 James E. Wilson <wilson@specifixinc.com>
1489 PR rtl-optimization/12092
1490 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
1493 2005-01-01 Roger Sayle <roger@eyesopen.com>
1494 Olivier Hainque <hainque@act-europe.fr>
1496 * tree.c (int_fits_type_p): A narrower type always fits in a
1497 wider one, except for negative values into unsigned types.
1499 2005-01-01 Roger Sayle <roger@eyesopen.com>
1501 * tree.c (int_fits_type_p): Compare the result of force_fit_type
1502 with the original constant rather than require TREE_OVERFLOW.
1504 2005-01-01 Steven Bosscher <stevenb@suse.de>
1507 * c-decl.c (finish_function): If compiling C99, annotate the
1508 compiler generated return with the current file name and line 0.
1509 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
1510 the source line is greater than 0.
1511 (remove_bb): Likewise.
1513 See ChangeLog.12 for earlier changes.