1 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
3 * pa-protos.h (prefetch_operand): Delete.
4 (prefetch_cc_operand, prefetch_nocc_operand): New declations.
5 * pa.c (prefetch_operand): Delete.
6 (prefetch_cc_operand, prefetch_nocc_operand): New functions.
7 * pa.h (EXTRA_CONSTRAINT): Add `W' constraint.
8 (PREDICATE_CODES): Delete prefetch_operand. Add prefetch_cc_operand
9 and prefetch_nocc_operand.
10 * pa.md (prefetch): Rework to avoid reload problems handling short
11 displacements when a cache control completer needs to be provided.
12 (prefetch_32, prefetch_64): Delete.
13 (prefetch_cc, prefetch_nocc): New patterns.
15 2005-01-15 David Edelsohn <edelsohn@gnu.org>
17 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Revert previous change.
19 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
22 * pa.c (pa_asm_output_aligned_common): Warn if specified alignment
23 exceeds maximum alignment for global common data.
24 * pa64-hpux.h (MAX_OFILE_ALIGNMENT): Define to 32768.
25 * som.h (MAX_OFILE_ALIGNMENT): Likewise.
27 2004-01-15 Roger Sayle <roger@eyesopen.com>
29 * tree-ssa-dom.c (extract_range_from_cond): Correct condition.
31 2004-01-15 Roger Sayle <roger@eyesopen.com>
33 * harg-reg-set.h (reg_class_names): Prototype global array.
34 * regclass.c (reg_class_names): Declare here and initialize to
36 (dump_regclass): Remove local declaration of reg_class_names.
38 * cfg.c (dump_flow_info): Likewise.
39 * ra-debug.c (reg_class_names): Likewise.
40 * regrename.c (reg_class_names): Likewise.
41 * reload.c (reg_class_names): Likewise.
42 * reload1.c (spill_failure): Likewise.
43 * config/m68hc11/m68hc11.c (reg_class_names): Likewise.
45 2005-01-15 Ulrich Weigand <uweigand@de.ibm.com>
47 * config/s390/s390.md ("reload_outti"): Remove predicate for
48 output operand. Abort if operand is not a MEM.
49 ("reload_outdi", "reload_outdf"): Likewise.
51 2005-01-15 Marc Espie <espie@openbsd.org>
53 * config.gcc (*-*-openbsd*): Set HAS_LIBC_R for OpenBSD <= 3.2.
54 (i[34567]86-*-openbsd*): Switch to ELF for OpenBSD >= 3.4.
55 * config/openbsd.h: Fix C++ includes for native configurations.
56 Add proper OS_CPP_BUILTINS. Fix libspec for recent OpenBSD.
57 Add trampoline support.
58 * config/i386/openbsdelf.h: New.
60 2005-01-15 Marc Espie <espie@openbsd.org>
62 * collect2.c (main): Explicitly parse -dynamic-linker option.
64 2005-01-15 Jakub Jelinek <jakub@redhat.com>
66 PR tree-optimization/19060
67 * tree-ssa-dom.c (extract_range_from_cond) <case LT_EXPR, GT_EXPR>:
68 Return 0 if op1 <= TYPE_MIN_VALUE () resp. op1 >= TYPE_MAX_VALUE ().
69 (simplify_cond_and_lookup_avail_expr): Add assert for dummy == 0
70 and handle extract_range_from_cond returning false.
71 * fold-const.c (fold): Optimize comparisons with min/max even for
72 width > HOST_BITS_PER_WIDE_INT.
74 2005-01-15 Ralf Corsepius <ralf.corsepius@rtems.org>
76 * config/mips/rtems.h (MIPS_DEFAULT_GVALUE): Set to 0.
77 * config/mips/t-rtems (MULTILIBS_DIRNAMES,MULTILIB_OPTIONS):
78 Remove little endian multilib variants.
79 Add mips32 multilib variant.
81 2005-01-15 Kazu Hirata <kazu@cs.umass.edu>
83 * explow.c (copy_all_reg, stabilize): Remove.
84 * expr.h: Remove the corresponding prototypes.
86 * rtlanal.c (regs_set_between_p): Remove.
87 * rtl.h: Remove the corresponding prototype.
89 2005-01-14 David Edelsohn <edelsohn@gnu.org>
91 * config/rs6000/rs6000.c (rs6k_nonimmediate_operand): Rename
92 to rs6000_nonimmediate_operand.
93 * config/rs6000/rs6000.h (PREDICATE_CODES): Rename
94 rs6k_nonimmediate_operand to rs6000_nonimmediate_operand.
95 * config/rs6000/rs6000.md (movsi_internal1): Rename predicate to
96 rs6000_nonimmediate_operand.
97 * config/rs6000/spe.md (movdf_e500_double): Rename predicate to
98 rs6000_nonimmediate_operand.
100 2005-01-14 Andrew Pinski <pinskia@physics.uc.edu>
102 * tree-ssa-alias.c (pass_may_alias): Add TODO_verify_stmts.
104 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
106 * config/rs6000/rs6000.h (CLASS_MAX_NREGS): DF goes in 1 register
108 (CANNOT_CHANGE_MODE_CLASS): Restrict DI mode changes on e500v2.
109 (PREDICATE_CODES): Add rs6k_nonimmediate_operand.
111 * config/rs6000/rs6000.c (invalid_e500_subreg): New.
112 (rs6k_nonimmediate_operand): New.
113 (rs6000_legitimate_offset_address_p): Handle DI modes on e500v2
115 (legitimate_lo_sum_address_p): Same.
116 (rs6000_legitimize_address): Same.
117 (rs6000_legitimize_reload_address): Same.
118 (rs6000_legitimate_address): Same.
119 (spe_build_register_parallel): Pass DF and DC modes in a DI
122 * config/rs6000/rs6000.md ("*movsi_internal1"): Change predicate
123 to rs6k_nonimmediate_operand.
125 * config/rs6000/spe.md ("*frob_df_di"): New.
126 ("*frob_di_df"): New.
127 ("*frob_di_df_2"): New.
128 ("*mov_sidf_e500_subreg0"): New.
129 ("*mov_sidf_e500_subreg4"): New.
130 ("*movdf_e500_double"): Change predicate to
131 rs6k_nonimmediate_operand.
133 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
135 * postreload.c (move2add_note_store): Only call
136 trunc_int_for_mode on scalar integers.
138 2005-01-14 Steven Bosscher <stevenb@suse.de>
140 * tree-ssa-dce.c (visited_control_parents): New sbitmap to
141 replace BB_VISITED uses.
142 (find_obviously_necessary_stmts): Don't clear BB_VISITED.
143 (propagate_necessity): Check the bitmap instead of BB_VISITED.
144 (tree_dce_done): Free visited_control_parents.
145 (perform_tree_ssa_dce): Allocate and clear it.
146 * tree-ssa-pre.c (compute_antic_aux): Make non-recursive.
147 (compute_antic): Iterate from here using a DFS. Use an sbitmap
148 instead of BB_VISITED.
150 2005-01-14 Kazu Hirata <kazu@cs.umass.edu>
152 * c-tree.h, coverage.h, langhooks-def.h, optabs.h, output.h,
153 rtl.h, targhooks.h, tree-flow.h, tree-mudflap.h, tree.h:
154 Remove unused prototypes.
156 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
158 * config/rs6000/rs6000.c (rs6000_stack_info): Fix caching of SPE
159 64-bit register usage.
160 (spe_func_has_64bit_regs_p): Add FIXME note.
162 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
164 * config/rs6000/rs6000.c (rs6000_generate_compare): Replace
165 flag_finite_math_only with flag_unsafe_math_optimizations.
167 2005-01-14 J. D. Johnston <jjohnst@us.ibm.com>
169 * config/s390/tpf-unwind.h (s390_fallback_frame_state): Correct
171 (__tpf_eh_return): Copy TPF private stack area from the module
172 boundary stack frame. Add check for when module addresses are
173 equal, but no stub address is found.
175 2005-01-14 Richard Earnshaw <rearnsha@arm.com>
178 * arm.h (struct machine_function): Add call_via field.
179 (thumb_call_via_label): Declare.
180 * arm.c (thumb_call_via_label): New variable.
181 (thumb_call_reg_needed): New variable.
182 (arm_output_function_epilogue): For Thumb code, output any per-function
183 call-indirect trampolines.
184 (thumb_call_via_reg): New function.
185 (arm_file_end): New function.
186 (TARGET_ASM_FILE_END): Call arm_file_end.
187 (aof_file_end): Likewise.
188 * arm-protos.h (thumb_call_via_reg): Declare.
189 * arm.md (call_reg_thumb, call_value_reg_thumb): Call
190 thumb_call_via_reg in normal case.
192 2005-01-14 Jakub Jelinek <jakub@redhat.com>
195 PR rtl-optimization/19348
196 * recog.c (peephole2_optimize): Do global life update if some peephole
197 decides it doesn't need at least one of its inputs and that change
198 influences liveness at the start of the basic block.
200 * basic-block.h (EXECUTE_IF_AND_COMPL_IN_REG_SET): Needs 2 REGSET
201 arguments instead of 1.
203 2005-01-14 Eric Botcazou <ebotcazou@libertysurf.fr>
206 * varasm.c (initializer_constant_valid_p) <ADDR_EXPR>: Return
207 zero for nested functions needing a static chain or functions
208 with a non-constant address.
210 2005-01-13 Roger Sayle <roger@eyesopen.com>
212 * simplify-rtx.c (simplify_binary_operation) <AND>: Optimize
213 (and (sign_extend X) C) into (zero_extend (and X C)).
215 2005-01-13 David O'Brien <obrien@FreeBSD.org>
217 * config/freebsd-spec.h: Make KSE pthread lib logic the default.
219 2005-01-13 Richard Henderson <rth@redhat.com>
224 * config/i386/i386.md (cmpdf, cmpsf, bunordered, bordered, buneq,
225 bunge, bungt, bunle, bunlt, bltgt): Enable for TARGET_SSE_MATH,
227 (cmpfp_i_387): Rename from cmpfp_i. Move after sse patterns.
228 (cmpfp_i_mixed): Rename from cmpfp_i_sse; use for TARGET_MIX_SSE_I387.
229 (cmpfp_i_sse): Rename from cmpfp_i_sse_only; use for TARGET_SSE_MATH.
230 (cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387): Similarly.
231 (fp_jcc_1_mixed, fp_jcc_1_sse, fp_jcc_1_387): Similarly.
232 (fp_jcc_2_mixed, fp_jcc_2_sse, fp_jcc_2_387): Similarly.
233 (fp_jcc_3_387, fp_jcc_4_387, fp_jcc_5_387, fp_jcc_6_387,
234 fp_jcc_7_387, fp_jcc_8_387): Rename from fp_jcc_N.
235 (movdicc_c_rex64): Rename with '*'.
236 (movsfcc, movdfcc): Add checks for 387 and sse math to condition.
237 (movsfcc_1_sse_min, movsfcc_1_sse_max, movsfcc_1_sse): New.
238 (movsfcc_1_387): Rename from movsfcc_1.
239 (movdfcc_1_sse_min, movdfcc_1_sse_max, movdfcc_1_sse): New.
240 (movdfcc_1, movdfcc_1_rex64): Add check for 387.
241 (sminsf3, smaxsf3, smindf3, smaxdf3): New.
242 (minsf3, minsf, minsf_nonieee, minsf_sse, mindf3, mindf,
243 mindf_nonieee, mindf_sse, maxsf3, maxsf, maxsf_nonieee, maxsf_sse,
244 maxdf3, maxdf, maxdf_nonieee, maxdf_sse, sse_movsfcc, sse_movsfcc_eq,
245 sse_movdfcc, sse_movdfcc_eq, sse_movsfcc_const0_1,
246 sse_movsfcc_const0_2, sse_movsfcc_const0_3, sse_movsfcc_const0_4,
247 sse_movdfcc_const0_1, sse_movdfcc_const0_2, sse_movdfcc_const0_3,
248 sse_movdfcc_const0_4): Remove.
249 * config/i386/i386.c (ix86_expand_fp_movcc): For TARGET_SSE_MATH,
250 recognize min/max early. Update for changed sse cmove patterns.
251 (ix86_split_sse_movcc): New.
252 * config/i386/i386-protos.h: Update.
254 2005-01-13 Steven Bosscher <stevenb@suse.de>
256 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
257 (fix_stmt_v_may_defs): Likewise.
259 2005-01-13 Richard Henderson <rth@redhat.com>
261 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
262 for converting the mask.
264 2005-01-13 David Edelsohn <edelsohn@gnu.org>
266 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
269 2005-01-13 Richard Henderson <rth@redhat.com>
271 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
272 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
273 IX86_BUILTIN_MOVDQ2Q): Remove.
274 (IX86_BUILTIN_VEC_EXT_V4SI): New.
275 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
276 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
278 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
279 (sse2_pextrw): Fix immediate constraint.
280 (sse2_loadq, sse2_loadq_rex64): Remove.
281 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
282 Use __builtin_ia32_vec_ext_<size>.
283 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
285 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
287 * function.c (assign_parm_setup_block): Look inside original
288 entry_parm when inspecting PARALLEL.
290 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
291 Joel Sherrill <joel@oarcorp.com>
294 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
295 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
296 rtems_gxx_recursive_mutex_init.
297 (__gthread_recursive_mutex_lock): New function.
298 (__gthread_recursive_mutex_trylock): Likewise.
299 (__gthread_recursive_mutex_unlock): Likewise.
301 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
303 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
305 2005-01-13 Jan Beulich <jbeulich@novell.com>
307 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
308 (ashrti3_internal): Indicate output is early clobber. Generate result
309 into output rather than first input. Use move for low word of output
310 if shift count is exactly 64.
311 (lshrti3_internal): Likewise.
313 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
317 * reload1.c (delete_output_reload): Don't delete an output reload
318 if the pseudo lives longer than a single basic block. Adjust and
319 improve wording of comment.
321 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
324 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
325 of TYPE_SIZE where we mean the number of bytes.
327 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
329 * function.c (assign_parm_setup_block): Relax condition on
330 multi-register optimization.
332 2005-01-12 Nick Clifton <nickc@redhat.com>
334 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
335 an "in_delay_slot" attribute of "no" to prevent them being used in
336 delay slots. This is forbidden because they might generate
339 2005-01-12 Alan Modra <amodra@bigpond.net.au>
342 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
345 2005-01-12 Nick Clifton <nickc@redhat.com>
347 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
348 passed on to GAS) but do not define, as GAS no longer supports or
349 needs the -m2000 option.
351 2005-01-11 Roger Sayle <roger@eyesopen.com>
353 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
354 type as args of unprototyped builtins aren't automatically promoted.
355 (fold_builtin_abs): Likewise.
357 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
359 * Makefile.in: Set a `build-warn' variable.
361 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
363 PR tree-optimization/17949
364 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
365 (find_interesting_uses_address): Use it.
367 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
369 * regrename.c (kill_value): Handle subreg's that won't simplify.
371 2005-01-11 Richard Henderson <rth@redhat.com>
374 * config/i386/i386.h (enum ix86_builtins): Move ...
375 * config/i386/i386.c: ... here.
376 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
377 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
378 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
379 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
380 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
381 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
382 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
383 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
384 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
385 IX86_BUILTIN_LOADDDUP): Remove.
386 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
387 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
388 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
389 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
390 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
391 (ix86_init_builtins): Make static.
392 (ix86_init_mmx_sse_builtins): Update for changed builtins.
393 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
394 if all the modes match. Otherwise, fake it.
395 (get_element_number, ix86_expand_vec_init_builtin,
396 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
397 (ix86_expand_builtin): Make static. Update for changed builtins.
398 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
399 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
400 (ix86_expand_vector_init_duplicate): New.
401 (ix86_expand_vector_init_low_nonzero): New.
402 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
403 Split out from ix86_expand_vector_init; handle integer modes.
404 (ix86_expand_vector_init): Use them.
405 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
406 * config/i386/i386-protos.h: Update.
407 * config/i386/predicates.md (reg_or_0_operand): New.
408 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
409 (movv2sf_internal): Likewise. And a splitter to match them all.
410 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
411 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
412 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
413 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
414 (mmx_pinsrw): Fix operand ordering.
415 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
416 rather than sse_loadss expander.
417 (movv2df splitter): Similarly.
418 (sse_loadss, sse_loadlss): Remove.
419 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
420 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
421 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
422 (sse3_movddup): Rename with '*'.
423 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
424 (sse2_loadsd): Remove.
425 (vec_dupv2df_sse3): Rename from sse3_loadddup.
426 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
427 (sse2_pinsrw): Fix argument ordering.
428 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
429 (sse2_stored): Remove 'r' destination.
430 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
431 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
432 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
433 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
434 vec_extractv16qi, vec_initv16qi): New.
436 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
437 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
438 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
440 (_mm_load_pd, _mm_store_pd): Use plain dereference.
441 (_mm_load_si128, _mm_store_si128): Likewise.
442 (_mm_load1_pd): Use _mm_set1_pd.
443 (_mm_load_sd): Use _mm_set_sd.
444 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
445 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
446 (_mm_set_epi64): Use _mm_set_epi64x.
447 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
448 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
449 _mm_setr_epi8): Use _mm_set_foo form.
450 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
451 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
452 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
453 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
454 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
455 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
456 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
457 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
458 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
459 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
460 (_mm_movedup_pd): Use _mm_shuffle_pd.
461 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
462 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
463 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
464 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
466 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
467 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
468 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
469 (_mm_store_ps): Use raw dereference.
470 (_mm_store1_ps): Use _mm_storeu_ps.
471 (_mm_storer_ps): Use _mm_store_ps.
472 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
473 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
475 2005-01-11 Stan Shebs <shebs@apple.com>
477 * config/rs6000/rs6000.c (machopic_output_stub): Issue
478 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
480 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
483 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
484 Skip all DECLs except for FIELD_DECLs.
486 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
488 * config/s390/s390.c (override_options): Return error if
489 -mbackchain, -mpacked-stack and -mhard-float are used together.
490 (s390_va_start): Remove the backchain && packed-stack special case.
491 (s390_gimplify_va_arg): Likewise.
492 * doc/invoke.texi: Remove the ABI incompatibility note.
494 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
496 * config/s390/s390.c (struct s390_frame_layout): Remove
498 (s390_frame_info, s390_emit_prologue): Replace occurrences of
499 save_backchain_p with TARGET_BACKCHAIN.
501 2005-01-11 Alan Modra <amodra@bigpond.net.au>
504 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
505 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
506 * function.c (locate_and_pad_parm): Set new field.
507 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
508 Tweak where_pad test to include "none". Always set mem align for
510 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
511 not sufficient for type.
512 (assign_parm_setup_block): If stack_parm is zero on entry, always
513 make a new stack local. Block move old stack parm if necessary
514 to new aligned stack local.
515 (assign_parm_setup_stack): Use a block move to handle
516 potentially misaligned entry_parm.
517 (assign_parms_unsplit_complex): Specify required alignment when
518 creating stack local.
519 * calls.c (compute_argument_addresses): Override alignment of stack
520 arg calculated from its type with the alignment given by
521 FUNCTION_ARG_BOUNDARY.
522 (store_one_arg): Likewise.
524 2005-01-11 Jan Beulich <jbeulich@novell.com>
526 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
527 Change respective itanium_class attribute to ialu.
528 (shladdp4_internal): New.
529 * config/ia64/predicates.md (shladd_log2_operand): New.
531 2005-01-11 Richard Henderson <rth@redhat.com>
533 * expr.c (store_constructor): Use rtvec_alloc instead of
534 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
536 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
538 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
540 2005-01-11 Alan Modra <amodra@bigpond.net.au>
542 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
543 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
544 (TARGET_RWRELOC): Define.
545 (ia64_rwreloc_section_type_flags): Delete.
546 (ia64_section_type_flags): New function.
547 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
548 (TARGET_RWRELOC): Define.
550 2005-01-10 David Mosberger <davidm@hpl.hp.com>
553 * config/ia64/ia64.c (process_set): For alloc insn, only call
554 process_epilogue is !frame_pointer_needed.
556 2005-01-10 Roger Sayle <roger@eyesopen.com>
559 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
560 unary operator and can't be treated as a binary/comparison operator.
562 2005-01-10 Richard Henderson <rth@redhat.com>
564 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
565 (ix86_return_in_memory): Likewise.
566 (function_arg_advance): Likewise. Mirror structure in function_arg
567 for choosing register to advance.
569 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
571 * tree-vectorizer.c, tree.def: Fix comment typos.
573 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
575 * config/s390/s390.c (struct processor_costs): 4 new fields:
576 ddbr, ddr, debr, der.
577 (s390_rtx_costs): More precise handling of divide instructions.
579 2005-01-10 David Edelsohn <edelsohn@gnu.org>
581 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
583 2005-01-10 Jan Beulich <jbeulich@novell.com>
585 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
586 names resulting from -ffunction-sections/-fdata-sections and linkonce
588 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
589 based on the section name. Rearrange the section name comparison logic
590 slightly so that each section name is compared against at most once.
592 2005-01-10 Ben Elliston <bje@au.ibm.com>
594 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
595 targets for which -fPIC reduces limitations on the GOT size.
597 2005-01-09 Falk Hueffner <falk@debian.org>
599 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
602 2005-01-09 David Edelsohn <edelsohn@gnu.org>
605 * collect2.c (main): Set aixrtl_flag for -brtl option.
606 (resolve_lib_name): Search for .so file extension before .a
609 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
611 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
612 (vect_analyze_data_refs_alignment): Add dump prints.
614 2005-01-09 Ira Rosen <irar@il.ibm.com>
616 * tree-vectorizer.c (vect_analyze_offset_expr): Use
617 expr_invariant_in_loop_p.
618 Initialize outputs first thing in the function.
619 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
620 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
621 (vect_analyze_pointer_ref_access): Check that the initial condition of
622 the access function is loop invariant.
624 2005-01-09 Richard Henderson <rth@redhat.com>
626 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
627 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
628 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
629 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
630 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
631 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
632 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
633 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
634 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
635 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
636 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
637 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
638 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
639 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
640 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
641 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
642 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
643 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
644 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
645 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
646 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
647 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
648 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
650 (mmx_clrdi, pavgusb): Remove.
651 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
652 with leading sse_ prefix.
653 * config/i386/sse.md: Receive them.
654 * config/i386/mmx.md: New file.
655 (MMXMODE12, MMXMODE24, mmxvecsize): New.
656 (subrv2sf3): Turn into expander for normal subtraction.
657 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
658 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
659 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
660 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
661 commutative; use ix86_binary_operator_ok.
662 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
663 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
664 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
665 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
666 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
667 Macroize from existing patterns; use ix86_binary_operator_ok.
668 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
669 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
670 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
671 with vec_select+vec_concat.
672 (mmx_pshufw, mmx_pshufw_1): Likewise.
673 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
674 (mmx_uavgv4hi3): Model correctly.
675 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
676 _mm_xor_si64): Remove casts.
678 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
680 PR tree-optimization/19224
681 * tree-scalar-evolution.c (get_instantiated_value,
682 set_instantiated_value): New functions.
683 (instantiate_parameters_1): Cache the results.
684 (instantiate_parameters, resolve_mixers): Initialize and free
687 2005-01-08 David Edelsohn <edelsohn@gnu.org>
689 * config/i386/i386.md (addhi_4): Correct reference in comment.
692 2005-01-08 Richard Henderson <rth@redhat.com>
694 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
695 (_mm_cvtsi128_si64x): Likewise.
696 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
697 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
698 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
699 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
700 * config/i386/i386-protos.h: Update.
701 * config/i386/i386.c (print_operand): Add 'H'.
702 (ix86_fixup_binary_operands): Split out from ...
703 (ix86_expand_binary_operator): ... here.
704 (ix86_fixup_binary_operands_no_copy): New.
705 (ix86_expand_fp_absneg_operator): Handle vector mode results.
706 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
707 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
708 (safe_vector_operand): Use CONST0_RTX.
709 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
710 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
711 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
712 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
714 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
715 (IX86_BUILTIN_CMPNGESS): New.
716 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
717 (attr type): Add sselog1.
718 (attr unit, attr memory): Handle it.
719 (movti, movti_internal, movti_rex64): Move near other integer moves.
720 (movtf, movtf_internal): Move near other fp moves.
721 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
722 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
723 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
724 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
725 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
726 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
727 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
728 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
729 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
730 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
731 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
732 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
733 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
734 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
735 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
736 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
737 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
738 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
739 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
740 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
741 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
742 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
743 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
744 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
745 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
746 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
747 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
748 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
749 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
750 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
751 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
752 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
753 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
754 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
755 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
756 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
757 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
758 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
759 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
760 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
761 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
762 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
763 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
764 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
766 (SSEPUSH, push<SSEPUSH>): Remove.
767 (MMXPUSH, push<MMXPUSH>): Remove.
768 (sse_movaps, sse_movaps_1, sse_movups): Remove.
769 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
770 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
771 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
772 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
773 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
774 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
775 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
776 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
777 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
778 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
779 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
780 (ppro_sse_log_V4SF): Handle sselog1.
781 * config/i386/predicates.md (const_0_to_1_operand): New.
782 (const_0_to_255_mul_8_operand): New.
783 (const_1_to_31_operand): Rename from const_int_1_31_operand.
784 (const_2_to_3_operand, const_4_to_7_operand): New.
785 * config/i386/sse.md: New file.
786 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
787 (sse_movups): Rename from sse_movups_1.
788 (sse_loadlss): Rename from sse_loadss_1.
789 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
791 (negv4sf2): Use ix86_expand_fp_absneg_operator.
792 (absv4sf2, negv2df, absv2df): New.
793 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
794 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
795 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
796 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
797 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
798 (sse3_addsubv4sf3): Model correctly.
799 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
800 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
801 (sse_movhlps): Model with vec_select+vec_concat.
802 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
803 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
804 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
805 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
806 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
807 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
808 sse2_pshufhw_1): Likewise.
809 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
810 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
811 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
812 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
813 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
814 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
815 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
816 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
817 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
818 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
819 sminv8hi3): Mark commutative
820 operands. Use ix86_binary_operator_ok.
821 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
822 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
823 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
824 sse2_punpcklqdq): Allow operand2 in memory.
825 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
826 sse2_movsd): Add memory alternatives.
827 (sse_storelps): Turn expander into an insn; split after reload.
828 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
829 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
831 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
833 * configure.ac (DWARF-2 debug_line): Use objdump.
834 * configure: Regenerate.
836 2005-01-08 Jeff Law <law@redhat.com>
837 Diego Novillo <dnovillo@redhat.com>
839 PR tree-optimization/18241
840 * tree-nrv.c (tree_nrv): Ignore volatile return values.
841 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
842 statements with volatile operands.
843 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
844 operands after marking a statement with has_volatile_ops.
846 2005-01-08 Roger Sayle <roger@eyesopen.com>
848 * tree.c (int_fits_type_p): Always honor integer constant
849 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
851 2005-01-08 Roger Sayle <roger@eyesopen.com>
853 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
854 the latent bug in rtl_delete_block has been resolved.
856 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
858 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
859 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
861 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
863 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
864 library as @multilib_dir@/$(SHLIB_SONAME).
866 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
868 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
869 * configure: Regenerate.
871 2005-01-07 Jakub Jelinek <jakub@redhat.com>
873 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
876 2005-01-07 David Edelsohn <edelsohn@gnu.org>
879 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
880 Convert non-word aligned offset address using ld/std into
883 2005-01-07 Richard Henderson <rth@redhat.com>
885 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
887 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
889 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
891 2005-01-07 Jakub Jelinek <jakub@redhat.com>
893 PR tree-optimization/19283
894 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
895 is not shorter than the original type.
897 PR rtl-optimization/19012
898 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
900 PR rtl-optimization/18861
901 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
902 (FULL_STATE): Define.
903 (mark_tablejump_edge): New function.
904 (purge_dead_tablejump_edges): New function.
905 (find_bb_boundaries): Use it.
907 PR tree-optimization/18828
908 * builtins.c (expand_builtin_next_arg): Remove argument and all
909 the argument checking.
910 (expand_builtin): Adjust caller.
911 (expand_builtin_va_start): Likewise. Remove error for too many
913 (fold_builtin_next_arg): Issue error for too many arguments.
914 After checking arguments, replace them with magic arguments that
915 prevent further checking of the args.
917 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
919 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
921 2005-01-06 Roger Sayle <roger@eyesopen.com>
924 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
925 and %{.S:...} (and their negative variants) to test whether the
926 input file is assembler or pre-processed-assembler independent of
927 the actual filename extension.
929 2005-01-06 Roger Sayle <roger@eyesopen.com>
931 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
932 of sign or zero extended values.
934 2005-01-06 Geoffrey Keating <geoffk@apple.com>
936 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
937 for IBM long double format correctly.
939 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
941 Fix PR tree-optimization/18792
943 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
944 to first_loop_depth, and use loop depth instead of loop number.
945 (build_classic_dir_vector): Ditto.
946 (compute_data_dependences_for_loop): Use depth, not loop number.
947 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
948 number. Pass in loops, instead of loop numbers.
949 (gather_interchange_stats): Ditto.
950 (linear_transform_loops): Ditto.
952 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
955 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
956 biv_fits_mode_p, extension_within_bounds_p): New functions.
957 (check_ext_dependent_givs): Use them.
959 2005-01-06 Roger Sayle <roger@eyesopen.com>
961 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
962 more than one barrier, in which case we should delete them all.
964 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
966 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
968 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
970 Fix PR middle-end/19286
972 * dwarf2out.c (gen_subprogram_die): If we've already tried to
973 output this subprogram, simply ignore this attempt to do it again.
974 (add_abstract_origin_attribute): Don't abort trying to add the abstract
975 origin attribute if it's not possible.
976 (gen_block_die): Don't ignore subblocks of "unused" blocks.
977 (decls_for_scope): Ditto.
978 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
979 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
981 2005-01-05 Richard Henderson <rth@redhat.com>
984 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
985 (bdesc_2arg): Use it.
986 (ix86_expand_binop_builtin): Force operands into registers
988 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
989 ix86_expand_sse_compare, ix86_expand_sse_comi,
990 ix86_expand_builtin): Likewise.
992 2005-01-05 Richard Henderson <rth@redhat.com>
994 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
996 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
997 (mulv4hi3): Set itanium_class mmmul.
998 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
999 (fpack): Rename from fpack_sfsf.
1001 2005-01-05 Richard Henderson <rth@redhat.com>
1004 * reload1.c (do_input_reload): Restrict the optimization deleteing
1005 a previous output reload to RELOAD_FOR_INPUT.
1007 2005-01-05 Steven Bosscher <stevenb@suse.de>
1009 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
1010 comment that we fall through after case.
1011 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
1012 (distribute_links): Likewise.
1013 * cse.c (cse_insn): Likewise.
1014 * cselib.c (cselib_invalidate_mem): Likewise.
1015 * df.c: Update comments at the top of the file.
1016 (read_modify_subreg_p): Update comments here too.
1017 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
1018 * flow.c (mark_set_1): Likewise.
1019 (mark_used_regs): Likewise.
1020 * gcse.c (mems_conflict_for_gcse_p): Likewise.
1021 (canon_list_insert): Likewise.
1022 (mark_set): Likewise.
1023 (try_replace_reg): Likewise.
1024 (store_killed_in_insn): Likewise.
1025 * loop.c (count_one_set): Likewise.
1026 (basic_induction_var): Likewise.
1027 * postreload-gcse.c (find_mem_conflicts): Likewise.
1028 * postreload.c (reload_combine_note_store): Likewise.
1029 (move2add_note_store): Likewise.
1030 * reload.c (find_equiv_reg): Likewise.
1031 (mark_referenced_resources): Likewise.
1032 * rtlanal.c (set_noop_p): Likewise.
1033 (note_stores): Likewise.
1034 (note_uses): Likewise.
1035 * sched-deps.c (sched_analyze_1): Likewise.
1036 * sched-rgn.c (check_live_1): Likewise.
1037 (update_live_1): Likewise.
1038 * config/i860/i860.c: Likewise.
1040 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
1042 (ZERO_EXTRACT): Mention that this one can be an lvalue.
1044 * doc/rtl.texi: Update documentation for bit-fields and SET.
1046 2005-01-05 Roger Sayle <roger@eyesopen.com>
1048 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
1050 2005-01-05 Jan Hubicka <jh@suse.cz>
1051 Richard Henderson <rth@redhat.com>
1054 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
1057 2005-01-05 Richard Henderson <rth@redhat.com>
1060 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
1061 (sse_shufps): Change operand 3 to const_int_operand.
1062 (sse2_storelps): Fix typo in template.
1063 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
1064 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
1065 (ix86_expand_builtin): Likewise.
1067 2005-01-05 Julian Brown <julian@codesourcery.com>
1069 * config/arm/arm.c (arm_return_in_memory): Treat complex types
1070 as aggregates for AAPCS ABIs.
1072 2005-01-05 Stan Shebs <shebs@apple.com>
1074 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
1075 Mach-O getters if ppc64.
1076 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
1078 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
1079 signed four-byte field for global code case.
1080 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
1082 2005-01-05 Roger Sayle <roger@eyesopen.com>
1085 * c-common.c: Include real.h.
1086 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
1087 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
1088 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
1089 * Makefile.in (c-common.o): Update dependencies.
1091 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1093 * c-parse.in (asm_string): Add trailing semicolon.
1095 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1097 * c-parse.in (asm_string): New. Don't allow wide strings in
1099 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
1100 asm_string instead of STRING.
1102 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1104 * c-typeck.c (constructor_no_implicit): Remove.
1105 (set_designator, process_init_element): Don't check
1106 constructor_no_implicit.
1108 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
1109 Kaz Kojima <kkojima@gcc.gnu.org>
1112 * lcm.c (create_pre_exit): New.
1113 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
1114 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
1116 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
1118 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
1119 -msoft-float in target_float_switch.
1120 * arm.c (arm_override_options): Fix processing of target_float_switch.
1122 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
1124 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
1127 2004-01-05 Uros Bizjak <uros@kss-loka.si>
1129 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
1130 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
1132 2005-01-04 Roger Sayle <roger@eyesopen.com>
1134 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
1135 if deleting the then-block allows the test-block to fallthru to the
1138 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
1141 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
1142 after the full declaration if the are in two different TUs.
1144 2005-01-04 Richard Henderson <rth@redhat.com>
1147 * tree-sra.c (generate_one_element_init): Just
1148 call gimplify_and_add.
1149 (generate_element_init): Record the
1150 new referenced variables and mark them for renaming
1151 and split out to ...
1152 (generate_element_init_1): This.
1153 (scalarize_init): Don't call push_gimplify_context/
1154 pop_gimplify_context.
1156 2005-01-04 Geoffrey Keating <geoffk@apple.com>
1158 * toplev.c (get_src_pwd): Handle failure of getpwd().
1160 2005-01-04 Roger Sayle <roger@eyesopen.com>
1162 * fold-const.c (fold_single_bit_test): Delete unreachable handling
1164 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
1165 all the cases handled by it are inverted by invert_truthvalue.
1167 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
1169 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
1170 (save_register_windowsi): Likewise.
1172 2005-01-04 Richard Henderson <rth@redhat.com>
1174 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
1176 2005-01-04 Uros Bizjak <uros@kss-loka.si>
1179 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
1180 any simplifications of vector mode comparison operators.
1181 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
1183 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
1184 Devang Patel <dpatel@apple.com>
1186 PR tree-optimization/18308
1187 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
1188 the operands before creating a new expression.
1189 * dojump.c (do_jump): Make drop_through_label available
1190 for all cases. Add expansion of COND_EXPR.
1192 2005-01-04 Ira Rosen <irar@il.ibm.com>
1194 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
1195 INTEGER_CST instead of TREE_CONSTANT.
1196 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
1197 instead of TREE_CONSTANT.
1198 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
1199 instead of TREE_CONSTANT.
1201 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
1203 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
1205 2005-01-04 Uros Bizjak <uros@kss-loka.si>
1208 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
1209 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
1211 2005-01-03 Richard Henderson <rth@redhat.com>
1213 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
1214 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
1215 fold_convert_const_real_from_real): Split out from ...
1216 (fold_convert_const): ... here.
1218 2005-01-03 Richard Henderson <rth@redhat.com>
1221 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
1222 (mov<MMXMODEI>_internal): Likewise.
1223 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
1224 Add V2SF case; use it for SSE1; don't use TI.
1225 (movdf_integer): Likewise.
1226 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
1228 2005-01-03 Ira Rosen <irar@il.ibm.com>
1230 * tree-vectorizer.c (vect_strip_conversions): New function.
1231 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
1232 check for binary class.
1234 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
1238 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
1240 * gimple-low.c (mark_blocks_with_used_vars): New function.
1241 (mark_blocks_with_used_subblocks): Ditto.
1242 (mark_used_blocks): Ditto.
1243 (pass_mark_used_blocks): New pass.
1244 * tree-inline.c: Include debug.h.
1245 (expand_call_inline): Call outlining_inline_function here.
1246 * tree-optimize.c (init_tree_optimization_passes): Add
1247 pass_mark_used_blocks.
1248 * tree-pass.h (pass_mark_used_blocks): New.
1249 * Makefile.in (tree-inline.o): Add debug.h dependency.
1251 2005-01-03 Geoffrey Keating <geoffk@apple.com>
1253 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
1254 VAR_DECLs to have weak_import attribute.
1256 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
1259 2004-01-03 Steven Bosscher <stevenb@suse.de>
1261 * passes.c (rest_of_compilation): Don't run regmove if only
1262 flag_expensive_optimizations, require flag_regmove instead.
1264 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
1266 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
1268 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
1269 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1270 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
1271 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1273 2005-01-03 Richard Henderson <rth@redhat.com>
1275 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
1276 (ia64_const_ok_for_letter_p): New.
1277 (ia64_const_double_ok_for_letter_p): New.
1278 (ia64_extra_constraint): New.
1279 (ia64_expand_vecint_compare): New.
1280 (ia64_expand_vcondu_v2si): New.
1281 (ia64_expand_vecint_cmov): New.
1282 (ia64_expand_vecint_minmax): New.
1283 (ia64_print_operand): Add 'v'.
1284 (ia64_preferred_reload_class): New.
1285 (ia64_vector_mode_supported_p): New.
1286 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
1287 (PREFERRED_RELOAD_CLASS): Move to function.
1288 (CONST_OK_FOR_LETTER_P): Move to function.
1289 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
1290 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
1291 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
1292 (EXTRA_CONSTRAINT): Move to function.
1293 * config/ia64/ia64.md: Include vect.md.
1294 (itanium_class): Add mmalua.
1296 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
1298 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
1299 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
1301 (const_int_2bit_operand): New.
1302 (fr_reg_or_0_operand): New.
1303 * config/ia64/ia64-modes.def: Add vector modes.
1304 * config/ia64/ia64-protos.h: Update.
1305 * config/ia64/vect.md: New file.
1307 2005-01-03 Richard Henderson <rth@redhat.com>
1309 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
1311 2005-01-03 Uros Bizjak <uros@kss-loka.si>
1314 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
1316 (log1pdf2): Change mode of operands[0,1] to DFmode.
1318 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
1320 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
1321 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
1322 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
1323 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
1324 SPARC_RELAXED_ORDERING.
1326 2005-01-03 Richard Henderson <rth@redhat.com>
1327 Uros Bizjak <uros@kss-loka.si>
1330 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
1331 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
1332 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
1334 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
1335 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
1336 const_pow2_1_to_128_operand): New predicates.
1338 2005-01-02 Greg McGary <greg@mcgary.org>
1340 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
1342 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
1343 object. [ptr_type]: Remove unused variable.
1344 Remember nearest addressable array-element or record-component
1345 when checking bitfield components. Tolerate empty BIND_EXPR.
1347 2005-01-01 Richard Henderson <rth@redhat.com>
1349 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
1350 that don't narrow the value. Fail for other conversions.
1352 2005-01-01 Richard Henderson <rth@redhat.com>
1355 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
1356 * c-lang.c (finish_file): Don't do it here.
1357 * objc/objc-act.c (objc_finish_file): Likewise.
1359 * cgraph.c (decl_assembler_name_equal): New.
1360 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
1361 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
1362 * cgraph.h (struct cgraph_varpool_node): Add next.
1363 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
1364 * varasm.c (assemble_alias): Mark the target as needed.
1366 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
1369 * function.c (get_arg_pointer_save_area): Use entry_of_function
1370 instead of get_insns.
1372 2005-01-01 Roger Sayle <roger@eyesopen.com>
1373 Andrew Pinski <pinskia@physics.uc.edu>
1374 James E. Wilson <wilson@specifixinc.com>
1376 PR rtl-optimization/12092
1377 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
1380 2005-01-01 Roger Sayle <roger@eyesopen.com>
1381 Olivier Hainque <hainque@act-europe.fr>
1383 * tree.c (int_fits_type_p): A narrower type always fits in a
1384 wider one, except for negative values into unsigned types.
1386 2005-01-01 Roger Sayle <roger@eyesopen.com>
1388 * tree.c (int_fits_type_p): Compare the result of force_fit_type
1389 with the original constant rather than require TREE_OVERFLOW.
1391 2005-01-01 Steven Bosscher <stevenb@suse.de>
1394 * c-decl.c (finish_function): If compiling C99, annotate the
1395 compiler generated return with the current file name and line 0.
1396 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
1397 the source line is greater than 0.
1398 (remove_bb): Likewise.
1400 See ChangeLog.12 for earlier changes.