]> gcc.gnu.org Git - gcc.git/blob - gcc/ChangeLog
f95-lang.c (gfc_init_builtin_functions): Call targetm.init_builtins.
[gcc.git] / gcc / ChangeLog
1 2005-01-17 Ira Rosen <irar@il.ibm.com>
2
3 * fortran/f95-lang.c (gfc_init_builtin_functions): Call
4 targetm.init_builtins.
5
6 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
7
8 * common.opt (-fnew-ra): Remove.
9 * ra*.*: Remove.
10 * toplev.h (flag_new_regalloc): Remove.
11 * Makefile.in (ra*.*): Don't mention.
12 * passes.c (rest_of_handle_new_regalloc): Remove.
13 (rest_of_handle_combine, rest_of_compilation): Always consider
14 flag_new_regalloc as false.
15 * doc/invoke.texi: Don't document -fnew-ra.
16
17 2005-01-17 Paolo Bonzini <bonzini@gnu.org>
18
19 * bb-reorder.c (fix_edges_for_rarely_executed_code): Remove
20 last parameter to reg_scan.
21 * loop.c (loop_optimize): Likewise.
22 * passes.c (rest_of_handle_tracer, rest_of_handle_if_conversion,
23 rest_of_handle_web, rest_of_handle_cfg, rest_of_handle_jump_bypass,
24 rest_of_handle_life, rest_of_handle_cse, rest_of_handle_cse2,
25 rest_of_handle_gcse, rest_of_handle_loop_optimize,
26 rest_of_handle_loop2, rest_of_handle_jump2): Likewise.
27 * regclass.c (reg_scan): Likewise, for the declaration.
28 * rtl.h (reg_scan): Likewise, for the prototype.
29
30 2005-01-17 Kazu Hirata <kazu@cs.umass.edu>
31
32 * tree-ssa-dom.c (tree_ssa_dominator_optimize): Don't call
33 mark_dfs_back_edges.
34
35 2005-01-16 Daniel Berlin <dberlin@dberlin.org>
36
37 * tree-ssa-pre.c (add_to_sets): s1 may be NULL.
38 (compute_avail): Uses don't go in tmp_gen.
39
40 2005-01-16 Steven Bosscher <stevenb@suse.de>
41
42 * ggc-page.c (ggc_alloc_stat): Use __builtin_ctzl instead of a
43 loop to look for a free slot in a page entry.
44
45 2005-01-16 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
46
47 PR target/16304
48 * defaults.h (TARGET_DEFERRED_OUTPUT_DEFS): Provide default.
49 * toplev.c (compile_file): Call process_pending_assemble_output_defs
50 just before targetm.asm_out.file_end.
51 * tree.h (process_pending_assemble_output_defs): Declare.
52 * varasm.c (assemble_output_def, process_pending_assemble_output_defs):
53 New functions.
54 (assemble_alias): Defer generation of assembly code for defines when
55 TARGET_DEFERRED_OUTPUT_DEFS is true.
56 * config/rs6000/aix41.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
57 * config/rs6000/aix43.h (TARGET_DEFERRED_OUTPUT_DEFS): Define.
58 * doc/tm.texi (TARGET_DEFERRED_OUTPUT_DEFS): document.
59
60 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
61
62 PR target/19336
63 * pa.c (pa_scalar_mode_supported_p): New function.
64 (TARGET_SCALAR_MODE_SUPPORTED_P): Define.
65 * pa.h (MIN_UNITS_PER_WORD): Add comment.
66
67 * pa-protos.h (prefetch_operand): Delete.
68 (prefetch_cc_operand, prefetch_nocc_operand): New declations.
69 * pa.c (prefetch_operand): Delete.
70 (prefetch_cc_operand, prefetch_nocc_operand): New functions.
71 * pa.h (EXTRA_CONSTRAINT): Add `W' constraint.
72 (PREDICATE_CODES): Delete prefetch_operand. Add prefetch_cc_operand
73 and prefetch_nocc_operand.
74 * pa.md (prefetch): Rework to avoid reload problems handling short
75 displacements when a cache control completer needs to be provided.
76 (prefetch_32, prefetch_64): Delete.
77 (prefetch_cc, prefetch_nocc): New patterns.
78
79 2005-01-15 David Edelsohn <edelsohn@gnu.org>
80
81 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Revert previous change.
82
83 2005-01-15 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
84
85 PR ada/19388
86 * pa.c (pa_asm_output_aligned_common): Warn if specified alignment
87 exceeds maximum alignment for global common data.
88 * pa64-hpux.h (MAX_OFILE_ALIGNMENT): Define to 32768.
89 * som.h (MAX_OFILE_ALIGNMENT): Likewise.
90
91 2004-01-15 Roger Sayle <roger@eyesopen.com>
92
93 * tree-ssa-dom.c (extract_range_from_cond): Correct condition.
94
95 2004-01-15 Roger Sayle <roger@eyesopen.com>
96
97 * harg-reg-set.h (reg_class_names): Prototype global array.
98 * regclass.c (reg_class_names): Declare here and initialize to
99 REG_CLASS_NAMES.
100 (dump_regclass): Remove local declaration of reg_class_names.
101 (regclass): Likewise.
102 * cfg.c (dump_flow_info): Likewise.
103 * ra-debug.c (reg_class_names): Likewise.
104 * regrename.c (reg_class_names): Likewise.
105 * reload.c (reg_class_names): Likewise.
106 * reload1.c (spill_failure): Likewise.
107 * config/m68hc11/m68hc11.c (reg_class_names): Likewise.
108
109 2005-01-15 Ulrich Weigand <uweigand@de.ibm.com>
110
111 * config/s390/s390.md ("reload_outti"): Remove predicate for
112 output operand. Abort if operand is not a MEM.
113 ("reload_outdi", "reload_outdf"): Likewise.
114
115 2005-01-15 Marc Espie <espie@openbsd.org>
116
117 * config.gcc (*-*-openbsd*): Set HAS_LIBC_R for OpenBSD <= 3.2.
118 (i[34567]86-*-openbsd*): Switch to ELF for OpenBSD >= 3.4.
119 * config/openbsd.h: Fix C++ includes for native configurations.
120 Add proper OS_CPP_BUILTINS. Fix libspec for recent OpenBSD.
121 Add trampoline support.
122 * config/i386/openbsdelf.h: New.
123
124 2005-01-15 Marc Espie <espie@openbsd.org>
125
126 * collect2.c (main): Explicitly parse -dynamic-linker option.
127
128 2005-01-15 Jakub Jelinek <jakub@redhat.com>
129
130 PR tree-optimization/19060
131 * tree-ssa-dom.c (extract_range_from_cond) <case LT_EXPR, GT_EXPR>:
132 Return 0 if op1 <= TYPE_MIN_VALUE () resp. op1 >= TYPE_MAX_VALUE ().
133 (simplify_cond_and_lookup_avail_expr): Add assert for dummy == 0
134 and handle extract_range_from_cond returning false.
135 * fold-const.c (fold): Optimize comparisons with min/max even for
136 width > HOST_BITS_PER_WIDE_INT.
137
138 2005-01-15 Ralf Corsepius <ralf.corsepius@rtems.org>
139
140 * config/mips/rtems.h (MIPS_DEFAULT_GVALUE): Set to 0.
141 * config/mips/t-rtems (MULTILIBS_DIRNAMES,MULTILIB_OPTIONS):
142 Remove little endian multilib variants.
143 Add mips32 multilib variant.
144
145 2005-01-15 Kazu Hirata <kazu@cs.umass.edu>
146
147 * explow.c (copy_all_reg, stabilize): Remove.
148 * expr.h: Remove the corresponding prototypes.
149
150 * rtlanal.c (regs_set_between_p): Remove.
151 * rtl.h: Remove the corresponding prototype.
152
153 2005-01-14 David Edelsohn <edelsohn@gnu.org>
154
155 * config/rs6000/rs6000.c (rs6k_nonimmediate_operand): Rename
156 to rs6000_nonimmediate_operand.
157 * config/rs6000/rs6000.h (PREDICATE_CODES): Rename
158 rs6k_nonimmediate_operand to rs6000_nonimmediate_operand.
159 * config/rs6000/rs6000.md (movsi_internal1): Rename predicate to
160 rs6000_nonimmediate_operand.
161 * config/rs6000/spe.md (movdf_e500_double): Rename predicate to
162 rs6000_nonimmediate_operand.
163
164 2005-01-14 Andrew Pinski <pinskia@physics.uc.edu>
165
166 * tree-ssa-alias.c (pass_may_alias): Add TODO_verify_stmts.
167
168 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
169
170 * config/rs6000/rs6000.h (CLASS_MAX_NREGS): DF goes in 1 register
171 on e500v2.
172 (CANNOT_CHANGE_MODE_CLASS): Restrict DI mode changes on e500v2.
173 (PREDICATE_CODES): Add rs6k_nonimmediate_operand.
174
175 * config/rs6000/rs6000.c (invalid_e500_subreg): New.
176 (rs6k_nonimmediate_operand): New.
177 (rs6000_legitimate_offset_address_p): Handle DI modes on e500v2
178 correctly.
179 (legitimate_lo_sum_address_p): Same.
180 (rs6000_legitimize_address): Same.
181 (rs6000_legitimize_reload_address): Same.
182 (rs6000_legitimate_address): Same.
183 (spe_build_register_parallel): Pass DF and DC modes in a DI
184 register.
185
186 * config/rs6000/rs6000.md ("*movsi_internal1"): Change predicate
187 to rs6k_nonimmediate_operand.
188
189 * config/rs6000/spe.md ("*frob_df_di"): New.
190 ("*frob_di_df"): New.
191 ("*frob_di_df_2"): New.
192 ("*mov_sidf_e500_subreg0"): New.
193 ("*mov_sidf_e500_subreg4"): New.
194 ("*movdf_e500_double"): Change predicate to
195 rs6k_nonimmediate_operand.
196
197 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
198
199 * postreload.c (move2add_note_store): Only call
200 trunc_int_for_mode on scalar integers.
201
202 2005-01-14 Steven Bosscher <stevenb@suse.de>
203
204 * tree-ssa-dce.c (visited_control_parents): New sbitmap to
205 replace BB_VISITED uses.
206 (find_obviously_necessary_stmts): Don't clear BB_VISITED.
207 (propagate_necessity): Check the bitmap instead of BB_VISITED.
208 (tree_dce_done): Free visited_control_parents.
209 (perform_tree_ssa_dce): Allocate and clear it.
210 * tree-ssa-pre.c (compute_antic_aux): Make non-recursive.
211 (compute_antic): Iterate from here using a DFS. Use an sbitmap
212 instead of BB_VISITED.
213
214 2005-01-14 Kazu Hirata <kazu@cs.umass.edu>
215
216 * c-tree.h, coverage.h, langhooks-def.h, optabs.h, output.h,
217 rtl.h, targhooks.h, tree-flow.h, tree-mudflap.h, tree.h:
218 Remove unused prototypes.
219
220 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
221
222 * config/rs6000/rs6000.c (rs6000_stack_info): Fix caching of SPE
223 64-bit register usage.
224 (spe_func_has_64bit_regs_p): Add FIXME note.
225
226 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
227
228 * config/rs6000/rs6000.c (rs6000_generate_compare): Replace
229 flag_finite_math_only with flag_unsafe_math_optimizations.
230
231 2005-01-14 J. D. Johnston <jjohnst@us.ibm.com>
232
233 * config/s390/tpf-unwind.h (s390_fallback_frame_state): Correct
234 end-of-stack check.
235 (__tpf_eh_return): Copy TPF private stack area from the module
236 boundary stack frame. Add check for when module addresses are
237 equal, but no stub address is found.
238
239 2005-01-14 Richard Earnshaw <rearnsha@arm.com>
240
241 PR target/7525
242 * arm.h (struct machine_function): Add call_via field.
243 (thumb_call_via_label): Declare.
244 * arm.c (thumb_call_via_label): New variable.
245 (thumb_call_reg_needed): New variable.
246 (arm_output_function_epilogue): For Thumb code, output any per-function
247 call-indirect trampolines.
248 (thumb_call_via_reg): New function.
249 (arm_file_end): New function.
250 (TARGET_ASM_FILE_END): Call arm_file_end.
251 (aof_file_end): Likewise.
252 * arm-protos.h (thumb_call_via_reg): Declare.
253 * arm.md (call_reg_thumb, call_value_reg_thumb): Call
254 thumb_call_via_reg in normal case.
255
256 2005-01-14 Jakub Jelinek <jakub@redhat.com>
257
258 PR middle-end/19084
259 PR rtl-optimization/19348
260 * recog.c (peephole2_optimize): Do global life update if some peephole
261 decides it doesn't need at least one of its inputs and that change
262 influences liveness at the start of the basic block.
263
264 * basic-block.h (EXECUTE_IF_AND_COMPL_IN_REG_SET): Needs 2 REGSET
265 arguments instead of 1.
266
267 2005-01-14 Eric Botcazou <ebotcazou@libertysurf.fr>
268
269 PR middle-end/18820
270 * varasm.c (initializer_constant_valid_p) <ADDR_EXPR>: Return
271 zero for nested functions needing a static chain or functions
272 with a non-constant address.
273
274 2005-01-13 Roger Sayle <roger@eyesopen.com>
275
276 * simplify-rtx.c (simplify_binary_operation) <AND>: Optimize
277 (and (sign_extend X) C) into (zero_extend (and X C)).
278
279 2005-01-13 David O'Brien <obrien@FreeBSD.org>
280
281 * config/freebsd-spec.h: Make KSE pthread lib logic the default.
282
283 2005-01-13 Richard Henderson <rth@redhat.com>
284
285 PR target/19009
286 PR target/19250
287 PR target/19252
288 * config/i386/i386.md (cmpdf, cmpsf, bunordered, bordered, buneq,
289 bunge, bungt, bunle, bunlt, bltgt): Enable for TARGET_SSE_MATH,
290 not just TARGET_SSE.
291 (cmpfp_i_387): Rename from cmpfp_i. Move after sse patterns.
292 (cmpfp_i_mixed): Rename from cmpfp_i_sse; use for TARGET_MIX_SSE_I387.
293 (cmpfp_i_sse): Rename from cmpfp_i_sse_only; use for TARGET_SSE_MATH.
294 (cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387): Similarly.
295 (fp_jcc_1_mixed, fp_jcc_1_sse, fp_jcc_1_387): Similarly.
296 (fp_jcc_2_mixed, fp_jcc_2_sse, fp_jcc_2_387): Similarly.
297 (fp_jcc_3_387, fp_jcc_4_387, fp_jcc_5_387, fp_jcc_6_387,
298 fp_jcc_7_387, fp_jcc_8_387): Rename from fp_jcc_N.
299 (movdicc_c_rex64): Rename with '*'.
300 (movsfcc, movdfcc): Add checks for 387 and sse math to condition.
301 (movsfcc_1_sse_min, movsfcc_1_sse_max, movsfcc_1_sse): New.
302 (movsfcc_1_387): Rename from movsfcc_1.
303 (movdfcc_1_sse_min, movdfcc_1_sse_max, movdfcc_1_sse): New.
304 (movdfcc_1, movdfcc_1_rex64): Add check for 387.
305 (sminsf3, smaxsf3, smindf3, smaxdf3): New.
306 (minsf3, minsf, minsf_nonieee, minsf_sse, mindf3, mindf,
307 mindf_nonieee, mindf_sse, maxsf3, maxsf, maxsf_nonieee, maxsf_sse,
308 maxdf3, maxdf, maxdf_nonieee, maxdf_sse, sse_movsfcc, sse_movsfcc_eq,
309 sse_movdfcc, sse_movdfcc_eq, sse_movsfcc_const0_1,
310 sse_movsfcc_const0_2, sse_movsfcc_const0_3, sse_movsfcc_const0_4,
311 sse_movdfcc_const0_1, sse_movdfcc_const0_2, sse_movdfcc_const0_3,
312 sse_movdfcc_const0_4): Remove.
313 * config/i386/i386.c (ix86_expand_fp_movcc): For TARGET_SSE_MATH,
314 recognize min/max early. Update for changed sse cmove patterns.
315 (ix86_split_sse_movcc): New.
316 * config/i386/i386-protos.h: Update.
317
318 2005-01-13 Steven Bosscher <stevenb@suse.de>
319
320 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
321 (fix_stmt_v_may_defs): Likewise.
322
323 2005-01-13 Richard Henderson <rth@redhat.com>
324
325 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
326 for converting the mask.
327
328 2005-01-13 David Edelsohn <edelsohn@gnu.org>
329
330 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
331 definition to 600.
332
333 2005-01-13 Richard Henderson <rth@redhat.com>
334
335 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
336 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
337 IX86_BUILTIN_MOVDQ2Q): Remove.
338 (IX86_BUILTIN_VEC_EXT_V4SI): New.
339 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
340 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
341 shuffling.
342 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
343 (sse2_pextrw): Fix immediate constraint.
344 (sse2_loadq, sse2_loadq_rex64): Remove.
345 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
346 Use __builtin_ia32_vec_ext_<size>.
347 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
348
349 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
350
351 * function.c (assign_parm_setup_block): Look inside original
352 entry_parm when inspecting PARALLEL.
353
354 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
355 Joel Sherrill <joel@oarcorp.com>
356
357 PR target/19399
358 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
359 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
360 rtems_gxx_recursive_mutex_init.
361 (__gthread_recursive_mutex_lock): New function.
362 (__gthread_recursive_mutex_trylock): Likewise.
363 (__gthread_recursive_mutex_unlock): Likewise.
364
365 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
366
367 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
368
369 2005-01-13 Jan Beulich <jbeulich@novell.com>
370
371 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
372 (ashrti3_internal): Indicate output is early clobber. Generate result
373 into output rather than first input. Use move for low word of output
374 if shift count is exactly 64.
375 (lshrti3_internal): Likewise.
376
377 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
378
379 PR target/18329
380 PR target/18330
381 * reload1.c (delete_output_reload): Don't delete an output reload
382 if the pseudo lives longer than a single basic block. Adjust and
383 improve wording of comment.
384
385 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
386
387 PR target/19334
388 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
389 of TYPE_SIZE where we mean the number of bytes.
390
391 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
392
393 * function.c (assign_parm_setup_block): Relax condition on
394 multi-register optimization.
395
396 2005-01-12 Nick Clifton <nickc@redhat.com>
397
398 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
399 an "in_delay_slot" attribute of "no" to prevent them being used in
400 delay slots. This is forbidden because they might generate
401 exceptions.
402
403 2005-01-12 Alan Modra <amodra@bigpond.net.au>
404
405 PR target/19389
406 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
407 with r->Y and Y->r.
408
409 2005-01-12 Nick Clifton <nickc@redhat.com>
410
411 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
412 passed on to GAS) but do not define, as GAS no longer supports or
413 needs the -m2000 option.
414
415 2005-01-11 Roger Sayle <roger@eyesopen.com>
416
417 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
418 type as args of unprototyped builtins aren't automatically promoted.
419 (fold_builtin_abs): Likewise.
420
421 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
422
423 * Makefile.in: Set a `build-warn' variable.
424
425 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
426
427 PR tree-optimization/17949
428 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
429 (find_interesting_uses_address): Use it.
430
431 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
432
433 * regrename.c (kill_value): Handle subreg's that won't simplify.
434
435 2005-01-11 Richard Henderson <rth@redhat.com>
436
437 PR target/13366
438 * config/i386/i386.h (enum ix86_builtins): Move ...
439 * config/i386/i386.c: ... here.
440 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
441 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
442 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
443 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
444 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
445 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
446 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
447 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
448 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
449 IX86_BUILTIN_LOADDDUP): Remove.
450 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
451 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
452 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
453 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
454 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
455 (ix86_init_builtins): Make static.
456 (ix86_init_mmx_sse_builtins): Update for changed builtins.
457 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
458 if all the modes match. Otherwise, fake it.
459 (get_element_number, ix86_expand_vec_init_builtin,
460 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
461 (ix86_expand_builtin): Make static. Update for changed builtins.
462 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
463 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
464 (ix86_expand_vector_init_duplicate): New.
465 (ix86_expand_vector_init_low_nonzero): New.
466 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
467 Split out from ix86_expand_vector_init; handle integer modes.
468 (ix86_expand_vector_init): Use them.
469 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
470 * config/i386/i386-protos.h: Update.
471 * config/i386/predicates.md (reg_or_0_operand): New.
472 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
473 (movv2sf_internal): Likewise. And a splitter to match them all.
474 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
475 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
476 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
477 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
478 (mmx_pinsrw): Fix operand ordering.
479 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
480 rather than sse_loadss expander.
481 (movv2df splitter): Similarly.
482 (sse_loadss, sse_loadlss): Remove.
483 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
484 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
485 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
486 (sse3_movddup): Rename with '*'.
487 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
488 (sse2_loadsd): Remove.
489 (vec_dupv2df_sse3): Rename from sse3_loadddup.
490 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
491 (sse2_pinsrw): Fix argument ordering.
492 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
493 (sse2_stored): Remove 'r' destination.
494 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
495 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
496 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
497 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
498 vec_extractv16qi, vec_initv16qi): New.
499
500 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
501 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
502 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
503 constructor form.
504 (_mm_load_pd, _mm_store_pd): Use plain dereference.
505 (_mm_load_si128, _mm_store_si128): Likewise.
506 (_mm_load1_pd): Use _mm_set1_pd.
507 (_mm_load_sd): Use _mm_set_sd.
508 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
509 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
510 (_mm_set_epi64): Use _mm_set_epi64x.
511 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
512 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
513 _mm_setr_epi8): Use _mm_set_foo form.
514 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
515 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
516 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
517 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
518 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
519 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
520 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
521 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
522 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
523 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
524 (_mm_movedup_pd): Use _mm_shuffle_pd.
525 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
526 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
527 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
528 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
529 Use _mm_setzero_ps.
530 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
531 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
532 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
533 (_mm_store_ps): Use raw dereference.
534 (_mm_store1_ps): Use _mm_storeu_ps.
535 (_mm_storer_ps): Use _mm_store_ps.
536 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
537 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
538
539 2005-01-11 Stan Shebs <shebs@apple.com>
540
541 * config/rs6000/rs6000.c (machopic_output_stub): Issue
542 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
543
544 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
545
546 PR target/18761
547 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
548 Skip all DECLs except for FIELD_DECLs.
549
550 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
551
552 * config/s390/s390.c (override_options): Return error if
553 -mbackchain, -mpacked-stack and -mhard-float are used together.
554 (s390_va_start): Remove the backchain && packed-stack special case.
555 (s390_gimplify_va_arg): Likewise.
556 * doc/invoke.texi: Remove the ABI incompatibility note.
557
558 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
559
560 * config/s390/s390.c (struct s390_frame_layout): Remove
561 save_backchain_p.
562 (s390_frame_info, s390_emit_prologue): Replace occurrences of
563 save_backchain_p with TARGET_BACKCHAIN.
564
565 2005-01-11 Alan Modra <amodra@bigpond.net.au>
566
567 PR target/18916
568 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
569 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
570 * function.c (locate_and_pad_parm): Set new field.
571 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
572 Tweak where_pad test to include "none". Always set mem align for
573 stack_parm.
574 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
575 not sufficient for type.
576 (assign_parm_setup_block): If stack_parm is zero on entry, always
577 make a new stack local. Block move old stack parm if necessary
578 to new aligned stack local.
579 (assign_parm_setup_stack): Use a block move to handle
580 potentially misaligned entry_parm.
581 (assign_parms_unsplit_complex): Specify required alignment when
582 creating stack local.
583 * calls.c (compute_argument_addresses): Override alignment of stack
584 arg calculated from its type with the alignment given by
585 FUNCTION_ARG_BOUNDARY.
586 (store_one_arg): Likewise.
587
588 2005-01-11 Jan Beulich <jbeulich@novell.com>
589
590 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
591 Change respective itanium_class attribute to ialu.
592 (shladdp4_internal): New.
593 * config/ia64/predicates.md (shladd_log2_operand): New.
594
595 2005-01-11 Richard Henderson <rth@redhat.com>
596
597 * expr.c (store_constructor): Use rtvec_alloc instead of
598 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
599
600 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
601
602 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
603
604 2005-01-11 Alan Modra <amodra@bigpond.net.au>
605
606 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
607 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
608 (TARGET_RWRELOC): Define.
609 (ia64_rwreloc_section_type_flags): Delete.
610 (ia64_section_type_flags): New function.
611 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
612 (TARGET_RWRELOC): Define.
613
614 2005-01-10 David Mosberger <davidm@hpl.hp.com>
615
616 PR target/18987
617 * config/ia64/ia64.c (process_set): For alloc insn, only call
618 process_epilogue is !frame_pointer_needed.
619
620 2005-01-10 Roger Sayle <roger@eyesopen.com>
621
622 PR c++/19355
623 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
624 unary operator and can't be treated as a binary/comparison operator.
625
626 2005-01-10 Richard Henderson <rth@redhat.com>
627
628 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
629 (ix86_return_in_memory): Likewise.
630 (function_arg_advance): Likewise. Mirror structure in function_arg
631 for choosing register to advance.
632
633 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
634
635 * tree-vectorizer.c, tree.def: Fix comment typos.
636
637 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
638
639 * config/s390/s390.c (struct processor_costs): 4 new fields:
640 ddbr, ddr, debr, der.
641 (s390_rtx_costs): More precise handling of divide instructions.
642
643 2005-01-10 David Edelsohn <edelsohn@gnu.org>
644
645 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
646
647 2005-01-10 Jan Beulich <jbeulich@novell.com>
648
649 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
650 names resulting from -ffunction-sections/-fdata-sections and linkonce
651 ones.
652 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
653 based on the section name. Rearrange the section name comparison logic
654 slightly so that each section name is compared against at most once.
655
656 2005-01-10 Ben Elliston <bje@au.ibm.com>
657
658 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
659 targets for which -fPIC reduces limitations on the GOT size.
660
661 2005-01-09 Falk Hueffner <falk@debian.org>
662
663 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
664 X/C1 cmpop C2.
665
666 2005-01-09 David Edelsohn <edelsohn@gnu.org>
667
668 PR target/18720
669 * collect2.c (main): Set aixrtl_flag for -brtl option.
670 (resolve_lib_name): Search for .so file extension before .a
671 if aixrtl_flag set.
672
673 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
674
675 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
676 (vect_analyze_data_refs_alignment): Add dump prints.
677
678 2005-01-09 Ira Rosen <irar@il.ibm.com>
679
680 * tree-vectorizer.c (vect_analyze_offset_expr): Use
681 expr_invariant_in_loop_p.
682 Initialize outputs first thing in the function.
683 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
684 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
685 (vect_analyze_pointer_ref_access): Check that the initial condition of
686 the access function is loop invariant.
687
688 2005-01-09 Richard Henderson <rth@redhat.com>
689
690 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
691 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
692 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
693 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
694 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
695 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
696 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
697 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
698 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
699 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
700 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
701 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
702 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
703 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
704 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
705 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
706 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
707 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
708 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
709 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
710 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
711 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
712 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
713 mmx_ prefix.
714 (mmx_clrdi, pavgusb): Remove.
715 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
716 with leading sse_ prefix.
717 * config/i386/sse.md: Receive them.
718 * config/i386/mmx.md: New file.
719 (MMXMODE12, MMXMODE24, mmxvecsize): New.
720 (subrv2sf3): Turn into expander for normal subtraction.
721 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
722 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
723 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
724 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
725 commutative; use ix86_binary_operator_ok.
726 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
727 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
728 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
729 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
730 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
731 Macroize from existing patterns; use ix86_binary_operator_ok.
732 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
733 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
734 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
735 with vec_select+vec_concat.
736 (mmx_pshufw, mmx_pshufw_1): Likewise.
737 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
738 (mmx_uavgv4hi3): Model correctly.
739 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
740 _mm_xor_si64): Remove casts.
741
742 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
743
744 PR tree-optimization/19224
745 * tree-scalar-evolution.c (get_instantiated_value,
746 set_instantiated_value): New functions.
747 (instantiate_parameters_1): Cache the results.
748 (instantiate_parameters, resolve_mixers): Initialize and free
749 the cache.
750
751 2005-01-08 David Edelsohn <edelsohn@gnu.org>
752
753 * config/i386/i386.md (addhi_4): Correct reference in comment.
754 (addqi_4): Same.
755
756 2005-01-08 Richard Henderson <rth@redhat.com>
757
758 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
759 (_mm_cvtsi128_si64x): Likewise.
760 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
761 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
762 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
763 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
764 * config/i386/i386-protos.h: Update.
765 * config/i386/i386.c (print_operand): Add 'H'.
766 (ix86_fixup_binary_operands): Split out from ...
767 (ix86_expand_binary_operator): ... here.
768 (ix86_fixup_binary_operands_no_copy): New.
769 (ix86_expand_fp_absneg_operator): Handle vector mode results.
770 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
771 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
772 (safe_vector_operand): Use CONST0_RTX.
773 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
774 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
775 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
776 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
777 special cases.
778 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
779 (IX86_BUILTIN_CMPNGESS): New.
780 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
781 (attr type): Add sselog1.
782 (attr unit, attr memory): Handle it.
783 (movti, movti_internal, movti_rex64): Move near other integer moves.
784 (movtf, movtf_internal): Move near other fp moves.
785 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
786 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
787 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
788 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
789 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
790 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
791 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
792 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
793 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
794 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
795 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
796 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
797 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
798 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
799 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
800 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
801 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
802 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
803 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
804 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
805 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
806 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
807 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
808 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
809 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
810 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
811 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
812 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
813 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
814 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
815 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
816 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
817 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
818 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
819 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
820 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
821 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
822 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
823 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
824 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
825 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
826 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
827 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
828 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
829 same time.
830 (SSEPUSH, push<SSEPUSH>): Remove.
831 (MMXPUSH, push<MMXPUSH>): Remove.
832 (sse_movaps, sse_movaps_1, sse_movups): Remove.
833 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
834 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
835 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
836 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
837 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
838 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
839 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
840 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
841 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
842 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
843 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
844 (ppro_sse_log_V4SF): Handle sselog1.
845 * config/i386/predicates.md (const_0_to_1_operand): New.
846 (const_0_to_255_mul_8_operand): New.
847 (const_1_to_31_operand): Rename from const_int_1_31_operand.
848 (const_2_to_3_operand, const_4_to_7_operand): New.
849 * config/i386/sse.md: New file.
850 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
851 (sse_movups): Rename from sse_movups_1.
852 (sse_loadlss): Rename from sse_loadss_1.
853 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
854 from the name.
855 (negv4sf2): Use ix86_expand_fp_absneg_operator.
856 (absv4sf2, negv2df, absv2df): New.
857 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
858 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
859 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
860 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
861 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
862 (sse3_addsubv4sf3): Model correctly.
863 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
864 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
865 (sse_movhlps): Model with vec_select+vec_concat.
866 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
867 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
868 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
869 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
870 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
871 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
872 sse2_pshufhw_1): Likewise.
873 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
874 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
875 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
876 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
877 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
878 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
879 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
880 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
881 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
882 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
883 sminv8hi3): Mark commutative
884 operands. Use ix86_binary_operator_ok.
885 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
886 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
887 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
888 sse2_punpcklqdq): Allow operand2 in memory.
889 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
890 sse2_movsd): Add memory alternatives.
891 (sse_storelps): Turn expander into an insn; split after reload.
892 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
893 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
894
895 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
896
897 * configure.ac (DWARF-2 debug_line): Use objdump.
898 * configure: Regenerate.
899
900 2005-01-08 Jeff Law <law@redhat.com>
901 Diego Novillo <dnovillo@redhat.com>
902
903 PR tree-optimization/18241
904 * tree-nrv.c (tree_nrv): Ignore volatile return values.
905 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
906 statements with volatile operands.
907 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
908 operands after marking a statement with has_volatile_ops.
909
910 2005-01-08 Roger Sayle <roger@eyesopen.com>
911
912 * tree.c (int_fits_type_p): Always honor integer constant
913 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
914
915 2005-01-08 Roger Sayle <roger@eyesopen.com>
916
917 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
918 the latent bug in rtl_delete_block has been resolved.
919
920 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
921
922 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
923 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
924
925 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
926
927 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
928 library as @multilib_dir@/$(SHLIB_SONAME).
929
930 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
931
932 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
933 * configure: Regenerate.
934
935 2005-01-07 Jakub Jelinek <jakub@redhat.com>
936
937 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
938 TYPE_MODE.
939
940 2005-01-07 David Edelsohn <edelsohn@gnu.org>
941
942 PR target/13674
943 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
944 Convert non-word aligned offset address using ld/std into
945 indirect address.
946
947 2005-01-07 Richard Henderson <rth@redhat.com>
948
949 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
950
951 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
952
953 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
954
955 2005-01-07 Jakub Jelinek <jakub@redhat.com>
956
957 PR tree-optimization/19283
958 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
959 is not shorter than the original type.
960
961 PR rtl-optimization/19012
962 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
963
964 PR rtl-optimization/18861
965 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
966 (FULL_STATE): Define.
967 (mark_tablejump_edge): New function.
968 (purge_dead_tablejump_edges): New function.
969 (find_bb_boundaries): Use it.
970
971 PR tree-optimization/18828
972 * builtins.c (expand_builtin_next_arg): Remove argument and all
973 the argument checking.
974 (expand_builtin): Adjust caller.
975 (expand_builtin_va_start): Likewise. Remove error for too many
976 arguments.
977 (fold_builtin_next_arg): Issue error for too many arguments.
978 After checking arguments, replace them with magic arguments that
979 prevent further checking of the args.
980
981 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
982
983 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
984
985 2005-01-06 Roger Sayle <roger@eyesopen.com>
986
987 PR target/6077
988 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
989 and %{.S:...} (and their negative variants) to test whether the
990 input file is assembler or pre-processed-assembler independent of
991 the actual filename extension.
992
993 2005-01-06 Roger Sayle <roger@eyesopen.com>
994
995 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
996 of sign or zero extended values.
997
998 2005-01-06 Geoffrey Keating <geoffk@apple.com>
999
1000 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
1001 for IBM long double format correctly.
1002
1003 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
1004
1005 Fix PR tree-optimization/18792
1006
1007 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
1008 to first_loop_depth, and use loop depth instead of loop number.
1009 (build_classic_dir_vector): Ditto.
1010 (compute_data_dependences_for_loop): Use depth, not loop number.
1011 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
1012 number. Pass in loops, instead of loop numbers.
1013 (gather_interchange_stats): Ditto.
1014 (linear_transform_loops): Ditto.
1015
1016 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
1017
1018 PR rtl-opt/13299
1019 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
1020 biv_fits_mode_p, extension_within_bounds_p): New functions.
1021 (check_ext_dependent_givs): Use them.
1022
1023 2005-01-06 Roger Sayle <roger@eyesopen.com>
1024
1025 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
1026 more than one barrier, in which case we should delete them all.
1027
1028 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
1029
1030 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
1031
1032 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
1033
1034 Fix PR middle-end/19286
1035 Fix PR debug/19267
1036 * dwarf2out.c (gen_subprogram_die): If we've already tried to
1037 output this subprogram, simply ignore this attempt to do it again.
1038 (add_abstract_origin_attribute): Don't abort trying to add the abstract
1039 origin attribute if it's not possible.
1040 (gen_block_die): Don't ignore subblocks of "unused" blocks.
1041 (decls_for_scope): Ditto.
1042 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
1043 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
1044
1045 2005-01-05 Richard Henderson <rth@redhat.com>
1046
1047 PR target/11327
1048 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
1049 (bdesc_2arg): Use it.
1050 (ix86_expand_binop_builtin): Force operands into registers
1051 when optimizing.
1052 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
1053 ix86_expand_sse_compare, ix86_expand_sse_comi,
1054 ix86_expand_builtin): Likewise.
1055
1056 2005-01-05 Richard Henderson <rth@redhat.com>
1057
1058 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
1059 and VEC_SELECT.
1060 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
1061 (mulv4hi3): Set itanium_class mmmul.
1062 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
1063 (fpack): Rename from fpack_sfsf.
1064
1065 2005-01-05 Richard Henderson <rth@redhat.com>
1066
1067 PR rtl-opt/10692
1068 * reload1.c (do_input_reload): Restrict the optimization deleteing
1069 a previous output reload to RELOAD_FOR_INPUT.
1070
1071 2005-01-05 Steven Bosscher <stevenb@suse.de>
1072
1073 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
1074 comment that we fall through after case.
1075 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
1076 (distribute_links): Likewise.
1077 * cse.c (cse_insn): Likewise.
1078 * cselib.c (cselib_invalidate_mem): Likewise.
1079 * df.c: Update comments at the top of the file.
1080 (read_modify_subreg_p): Update comments here too.
1081 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
1082 * flow.c (mark_set_1): Likewise.
1083 (mark_used_regs): Likewise.
1084 * gcse.c (mems_conflict_for_gcse_p): Likewise.
1085 (canon_list_insert): Likewise.
1086 (mark_set): Likewise.
1087 (try_replace_reg): Likewise.
1088 (store_killed_in_insn): Likewise.
1089 * loop.c (count_one_set): Likewise.
1090 (basic_induction_var): Likewise.
1091 * postreload-gcse.c (find_mem_conflicts): Likewise.
1092 * postreload.c (reload_combine_note_store): Likewise.
1093 (move2add_note_store): Likewise.
1094 * reload.c (find_equiv_reg): Likewise.
1095 (mark_referenced_resources): Likewise.
1096 * rtlanal.c (set_noop_p): Likewise.
1097 (note_stores): Likewise.
1098 (note_uses): Likewise.
1099 * sched-deps.c (sched_analyze_1): Likewise.
1100 * sched-rgn.c (check_live_1): Likewise.
1101 (update_live_1): Likewise.
1102 * config/i860/i860.c: Likewise.
1103
1104 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
1105 an lvalue.
1106 (ZERO_EXTRACT): Mention that this one can be an lvalue.
1107
1108 * doc/rtl.texi: Update documentation for bit-fields and SET.
1109
1110 2005-01-05 Roger Sayle <roger@eyesopen.com>
1111
1112 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
1113
1114 2005-01-05 Jan Hubicka <jh@suse.cz>
1115 Richard Henderson <rth@redhat.com>
1116
1117 PR target/18910
1118 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
1119 with an offset.
1120
1121 2005-01-05 Richard Henderson <rth@redhat.com>
1122
1123 PR target/12902
1124 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
1125 (sse_shufps): Change operand 3 to const_int_operand.
1126 (sse2_storelps): Fix typo in template.
1127 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
1128 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
1129 (ix86_expand_builtin): Likewise.
1130
1131 2005-01-05 Julian Brown <julian@codesourcery.com>
1132
1133 * config/arm/arm.c (arm_return_in_memory): Treat complex types
1134 as aggregates for AAPCS ABIs.
1135
1136 2005-01-05 Stan Shebs <shebs@apple.com>
1137
1138 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
1139 Mach-O getters if ppc64.
1140 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
1141 size argument.
1142 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
1143 signed four-byte field for global code case.
1144 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
1145
1146 2005-01-05 Roger Sayle <roger@eyesopen.com>
1147
1148 PR middle-end/19100
1149 * c-common.c: Include real.h.
1150 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
1151 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
1152 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
1153 * Makefile.in (c-common.o): Update dependencies.
1154
1155 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1156
1157 * c-parse.in (asm_string): Add trailing semicolon.
1158
1159 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1160
1161 * c-parse.in (asm_string): New. Don't allow wide strings in
1162 'asm'.
1163 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
1164 asm_string instead of STRING.
1165
1166 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1167
1168 * c-typeck.c (constructor_no_implicit): Remove.
1169 (set_designator, process_init_element): Don't check
1170 constructor_no_implicit.
1171
1172 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
1173 Kaz Kojima <kkojima@gcc.gnu.org>
1174
1175 PR target/16482
1176 * lcm.c (create_pre_exit): New.
1177 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
1178 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
1179
1180 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
1181
1182 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
1183 -msoft-float in target_float_switch.
1184 * arm.c (arm_override_options): Fix processing of target_float_switch.
1185
1186 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
1187
1188 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
1189 preferencing.
1190
1191 2004-01-05 Uros Bizjak <uros@kss-loka.si>
1192
1193 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
1194 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
1195
1196 2005-01-04 Roger Sayle <roger@eyesopen.com>
1197
1198 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
1199 if deleting the then-block allows the test-block to fallthru to the
1200 else-block.
1201
1202 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
1203
1204 PR c/19152
1205 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
1206 after the full declaration if the are in two different TUs.
1207
1208 2005-01-04 Richard Henderson <rth@redhat.com>
1209
1210 PR tree-opt/19158
1211 * tree-sra.c (generate_one_element_init): Just
1212 call gimplify_and_add.
1213 (generate_element_init): Record the
1214 new referenced variables and mark them for renaming
1215 and split out to ...
1216 (generate_element_init_1): This.
1217 (scalarize_init): Don't call push_gimplify_context/
1218 pop_gimplify_context.
1219
1220 2005-01-04 Geoffrey Keating <geoffk@apple.com>
1221
1222 * toplev.c (get_src_pwd): Handle failure of getpwd().
1223
1224 2005-01-04 Roger Sayle <roger@eyesopen.com>
1225
1226 * fold-const.c (fold_single_bit_test): Delete unreachable handling
1227 of TRUTH_NOT_EXPR.
1228 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
1229 all the cases handled by it are inverted by invert_truthvalue.
1230
1231 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
1232
1233 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
1234 (save_register_windowsi): Likewise.
1235
1236 2005-01-04 Richard Henderson <rth@redhat.com>
1237
1238 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
1239
1240 2005-01-04 Uros Bizjak <uros@kss-loka.si>
1241
1242 PR middle-end/17767
1243 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
1244 any simplifications of vector mode comparison operators.
1245 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
1246
1247 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
1248 Devang Patel <dpatel@apple.com>
1249
1250 PR tree-optimization/18308
1251 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
1252 the operands before creating a new expression.
1253 * dojump.c (do_jump): Make drop_through_label available
1254 for all cases. Add expansion of COND_EXPR.
1255
1256 2005-01-04 Ira Rosen <irar@il.ibm.com>
1257
1258 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
1259 INTEGER_CST instead of TREE_CONSTANT.
1260 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
1261 instead of TREE_CONSTANT.
1262 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
1263 instead of TREE_CONSTANT.
1264
1265 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
1266
1267 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
1268
1269 2005-01-04 Uros Bizjak <uros@kss-loka.si>
1270
1271 PR target/19240
1272 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
1273 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
1274
1275 2005-01-03 Richard Henderson <rth@redhat.com>
1276
1277 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
1278 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
1279 fold_convert_const_real_from_real): Split out from ...
1280 (fold_convert_const): ... here.
1281
1282 2005-01-03 Richard Henderson <rth@redhat.com>
1283
1284 PR target/19235
1285 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
1286 (mov<MMXMODEI>_internal): Likewise.
1287 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
1288 Add V2SF case; use it for SSE1; don't use TI.
1289 (movdf_integer): Likewise.
1290 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
1291
1292 2005-01-03 Ira Rosen <irar@il.ibm.com>
1293
1294 * tree-vectorizer.c (vect_strip_conversions): New function.
1295 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
1296 check for binary class.
1297
1298 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
1299
1300 Fix PR debug/17924
1301 Fix PR debug/19191
1302 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
1303 is a decl.
1304 * gimple-low.c (mark_blocks_with_used_vars): New function.
1305 (mark_blocks_with_used_subblocks): Ditto.
1306 (mark_used_blocks): Ditto.
1307 (pass_mark_used_blocks): New pass.
1308 * tree-inline.c: Include debug.h.
1309 (expand_call_inline): Call outlining_inline_function here.
1310 * tree-optimize.c (init_tree_optimization_passes): Add
1311 pass_mark_used_blocks.
1312 * tree-pass.h (pass_mark_used_blocks): New.
1313 * Makefile.in (tree-inline.o): Add debug.h dependency.
1314
1315 2005-01-03 Geoffrey Keating <geoffk@apple.com>
1316
1317 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
1318 VAR_DECLs to have weak_import attribute.
1319
1320 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
1321 'struct ucontext'.
1322
1323 2004-01-03 Steven Bosscher <stevenb@suse.de>
1324
1325 * passes.c (rest_of_compilation): Don't run regmove if only
1326 flag_expensive_optimizations, require flag_regmove instead.
1327
1328 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
1329
1330 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
1331 register "sparc".
1332 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
1333 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1334 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
1335 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1336
1337 2005-01-03 Richard Henderson <rth@redhat.com>
1338
1339 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
1340 (ia64_const_ok_for_letter_p): New.
1341 (ia64_const_double_ok_for_letter_p): New.
1342 (ia64_extra_constraint): New.
1343 (ia64_expand_vecint_compare): New.
1344 (ia64_expand_vcondu_v2si): New.
1345 (ia64_expand_vecint_cmov): New.
1346 (ia64_expand_vecint_minmax): New.
1347 (ia64_print_operand): Add 'v'.
1348 (ia64_preferred_reload_class): New.
1349 (ia64_vector_mode_supported_p): New.
1350 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
1351 (PREFERRED_RELOAD_CLASS): Move to function.
1352 (CONST_OK_FOR_LETTER_P): Move to function.
1353 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
1354 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
1355 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
1356 (EXTRA_CONSTRAINT): Move to function.
1357 * config/ia64/ia64.md: Include vect.md.
1358 (itanium_class): Add mmalua.
1359 (type): Handle it.
1360 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
1361 (1b_mmalua): New.
1362 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
1363 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
1364 CONST0_RTX.
1365 (const_int_2bit_operand): New.
1366 (fr_reg_or_0_operand): New.
1367 * config/ia64/ia64-modes.def: Add vector modes.
1368 * config/ia64/ia64-protos.h: Update.
1369 * config/ia64/vect.md: New file.
1370
1371 2005-01-03 Richard Henderson <rth@redhat.com>
1372
1373 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
1374
1375 2005-01-03 Uros Bizjak <uros@kss-loka.si>
1376
1377 PR target/19236
1378 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
1379 to SFmode.
1380 (log1pdf2): Change mode of operands[0,1] to DFmode.
1381
1382 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
1383
1384 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
1385 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
1386 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
1387 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
1388 SPARC_RELAXED_ORDERING.
1389
1390 2005-01-03 Richard Henderson <rth@redhat.com>
1391 Uros Bizjak <uros@kss-loka.si>
1392
1393 PR target/14631
1394 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
1395 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
1396 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
1397 handling.
1398 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
1399 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
1400 const_pow2_1_to_128_operand): New predicates.
1401
1402 2005-01-02 Greg McGary <greg@mcgary.org>
1403
1404 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
1405 anonymous decls.
1406 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
1407 object. [ptr_type]: Remove unused variable.
1408 Remember nearest addressable array-element or record-component
1409 when checking bitfield components. Tolerate empty BIND_EXPR.
1410
1411 2005-01-01 Richard Henderson <rth@redhat.com>
1412
1413 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
1414 that don't narrow the value. Fail for other conversions.
1415
1416 2005-01-01 Richard Henderson <rth@redhat.com>
1417
1418 PR c/19031
1419 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
1420 * c-lang.c (finish_file): Don't do it here.
1421 * objc/objc-act.c (objc_finish_file): Likewise.
1422
1423 * cgraph.c (decl_assembler_name_equal): New.
1424 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
1425 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
1426 * cgraph.h (struct cgraph_varpool_node): Add next.
1427 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
1428 * varasm.c (assemble_alias): Mark the target as needed.
1429
1430 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
1431
1432 PR middle-end/19221
1433 * function.c (get_arg_pointer_save_area): Use entry_of_function
1434 instead of get_insns.
1435
1436 2005-01-01 Roger Sayle <roger@eyesopen.com>
1437 Andrew Pinski <pinskia@physics.uc.edu>
1438 James E. Wilson <wilson@specifixinc.com>
1439
1440 PR rtl-optimization/12092
1441 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
1442 is zero.
1443
1444 2005-01-01 Roger Sayle <roger@eyesopen.com>
1445 Olivier Hainque <hainque@act-europe.fr>
1446
1447 * tree.c (int_fits_type_p): A narrower type always fits in a
1448 wider one, except for negative values into unsigned types.
1449
1450 2005-01-01 Roger Sayle <roger@eyesopen.com>
1451
1452 * tree.c (int_fits_type_p): Compare the result of force_fit_type
1453 with the original constant rather than require TREE_OVERFLOW.
1454
1455 2005-01-01 Steven Bosscher <stevenb@suse.de>
1456
1457 PR middle-end/17544
1458 * c-decl.c (finish_function): If compiling C99, annotate the
1459 compiler generated return with the current file name and line 0.
1460 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
1461 the source line is greater than 0.
1462 (remove_bb): Likewise.
1463
1464 See ChangeLog.12 for earlier changes.
This page took 0.100469 seconds and 6 git commands to generate.