]> gcc.gnu.org Git - gcc.git/blob - gcc/ChangeLog
f7884397d8a409fe73a0335c440fcc6e4422f996
[gcc.git] / gcc / ChangeLog
1 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
2
3 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
4
5 2005-01-13 Jan Beulich <jbeulich@novell.com>
6
7 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
8 (ashrti3_internal): Indicate output is early clobber. Generate result
9 into output rather than first input. Use move for low word of output
10 if shift count is exactly 64.
11 (lshrti3_internal): Likewise.
12
13 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
14
15 PR target/18329
16 PR target/18330
17 * reload1.c (delete_output_reload): Don't delete an output reload
18 if the pseudo lives longer than a single basic block. Adjust and
19 improve wording of comment.
20
21 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
22
23 PR target/19334
24 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
25 of TYPE_SIZE where we mean the number of bytes.
26
27 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
28
29 * function.c (assign_parm_setup_block): Relax condition on
30 multi-register optimization.
31
32 2005-01-12 Nick Clifton <nickc@redhat.com>
33
34 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
35 an "in_delay_slot" attribute of "no" to prevent them being used in
36 delay slots. This is forbidden because they might generate
37 exceptions.
38
39 2005-01-12 Alan Modra <amodra@bigpond.net.au>
40
41 PR target/19389
42 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
43 with r->Y and Y->r.
44
45 2005-01-12 Nick Clifton <nickc@redhat.com>
46
47 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
48 passed on to GAS) but do not define, as GAS no longer supports or
49 needs the -m2000 option.
50
51 2005-01-11 Roger Sayle <roger@eyesopen.com>
52
53 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
54 type as args of unprototyped builtins aren't automatically promoted.
55 (fold_builtin_abs): Likewise.
56
57 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
58
59 * Makefile.in: Set a `build-warn' variable.
60
61 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
62
63 PR tree-optimization/17949
64 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
65 (find_interesting_uses_address): Use it.
66
67 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
68
69 * regrename.c (kill_value): Handle subreg's that won't simplify.
70
71 2005-01-11 Richard Henderson <rth@redhat.com>
72
73 PR target/13366
74 * config/i386/i386.h (enum ix86_builtins): Move ...
75 * config/i386/i386.c: ... here.
76 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
77 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
78 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
79 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
80 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
81 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
82 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
83 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
84 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
85 IX86_BUILTIN_LOADDDUP): Remove.
86 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
87 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
88 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
89 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
90 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
91 (ix86_init_builtins): Make static.
92 (ix86_init_mmx_sse_builtins): Update for changed builtins.
93 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
94 if all the modes match. Otherwise, fake it.
95 (get_element_number, ix86_expand_vec_init_builtin,
96 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
97 (ix86_expand_builtin): Make static. Update for changed builtins.
98 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
99 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
100 (ix86_expand_vector_init_duplicate): New.
101 (ix86_expand_vector_init_low_nonzero): New.
102 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
103 Split out from ix86_expand_vector_init; handle integer modes.
104 (ix86_expand_vector_init): Use them.
105 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
106 * config/i386/i386-protos.h: Update.
107 * config/i386/predicates.md (reg_or_0_operand): New.
108 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
109 (movv2sf_internal): Likewise. And a splitter to match them all.
110 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
111 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
112 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
113 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
114 (mmx_pinsrw): Fix operand ordering.
115 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
116 rather than sse_loadss expander.
117 (movv2df splitter): Similarly.
118 (sse_loadss, sse_loadlss): Remove.
119 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
120 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
121 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
122 (sse3_movddup): Rename with '*'.
123 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
124 (sse2_loadsd): Remove.
125 (vec_dupv2df_sse3): Rename from sse3_loadddup.
126 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
127 (sse2_pinsrw): Fix argument ordering.
128 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
129 (sse2_stored): Remove 'r' destination.
130 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
131 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
132 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
133 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
134 vec_extractv16qi, vec_initv16qi): New.
135
136 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
137 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
138 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
139 constructor form.
140 (_mm_load_pd, _mm_store_pd): Use plain dereference.
141 (_mm_load_si128, _mm_store_si128): Likewise.
142 (_mm_load1_pd): Use _mm_set1_pd.
143 (_mm_load_sd): Use _mm_set_sd.
144 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
145 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
146 (_mm_set_epi64): Use _mm_set_epi64x.
147 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
148 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
149 _mm_setr_epi8): Use _mm_set_foo form.
150 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
151 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
152 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
153 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
154 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
155 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
156 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
157 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
158 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
159 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
160 (_mm_movedup_pd): Use _mm_shuffle_pd.
161 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
162 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
163 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
164 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
165 Use _mm_setzero_ps.
166 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
167 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
168 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
169 (_mm_store_ps): Use raw dereference.
170 (_mm_store1_ps): Use _mm_storeu_ps.
171 (_mm_storer_ps): Use _mm_store_ps.
172 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
173 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
174
175 2005-01-11 Stan Shebs <shebs@apple.com>
176
177 * config/rs6000/rs6000.c (machopic_output_stub): Issue
178 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
179
180 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
181
182 PR target/18761
183 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
184 Skip all DECLs except for FIELD_DECLs.
185
186 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
187
188 * config/s390/s390.c (override_options): Return error if
189 -mbackchain, -mpacked-stack and -mhard-float are used together.
190 (s390_va_start): Remove the backchain && packed-stack special case.
191 (s390_gimplify_va_arg): Likewise.
192 * doc/invoke.texi: Remove the ABI incompatibility note.
193
194 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
195
196 * config/s390/s390.c (struct s390_frame_layout): Remove
197 save_backchain_p.
198 (s390_frame_info, s390_emit_prologue): Replace occurrences of
199 save_backchain_p with TARGET_BACKCHAIN.
200
201 2005-01-11 Alan Modra <amodra@bigpond.net.au>
202
203 PR target/18916
204 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
205 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
206 * function.c (locate_and_pad_parm): Set new field.
207 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
208 Tweak where_pad test to include "none". Always set mem align for
209 stack_parm.
210 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
211 not sufficient for type.
212 (assign_parm_setup_block): If stack_parm is zero on entry, always
213 make a new stack local. Block move old stack parm if necessary
214 to new aligned stack local.
215 (assign_parm_setup_stack): Use a block move to handle
216 potentially misaligned entry_parm.
217 (assign_parms_unsplit_complex): Specify required alignment when
218 creating stack local.
219 * calls.c (compute_argument_addresses): Override alignment of stack
220 arg calculated from its type with the alignment given by
221 FUNCTION_ARG_BOUNDARY.
222 (store_one_arg): Likewise.
223
224 2005-01-11 Jan Beulich <jbeulich@novell.com>
225
226 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
227 Change respective itanium_class attribute to ialu.
228 (shladdp4_internal): New.
229 * config/ia64/predicates.md (shladd_log2_operand): New.
230
231 2005-01-11 Richard Henderson <rth@redhat.com>
232
233 * expr.c (store_constructor): Use rtvec_alloc instead of
234 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
235
236 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
237
238 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
239
240 2005-01-11 Alan Modra <amodra@bigpond.net.au>
241
242 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
243 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
244 (TARGET_RWRELOC): Define.
245 (ia64_rwreloc_section_type_flags): Delete.
246 (ia64_section_type_flags): New function.
247 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
248 (TARGET_RWRELOC): Define.
249
250 2005-01-10 David Mosberger <davidm@hpl.hp.com>
251
252 PR target/18987
253 * config/ia64/ia64.c (process_set): For alloc insn, only call
254 process_epilogue is !frame_pointer_needed.
255
256 2005-01-10 Roger Sayle <roger@eyesopen.com>
257
258 PR c++/19355
259 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
260 unary operator and can't be treated as a binary/comparison operator.
261
262 2005-01-10 Richard Henderson <rth@redhat.com>
263
264 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
265 (ix86_return_in_memory): Likewise.
266 (function_arg_advance): Likewise. Mirror structure in function_arg
267 for choosing register to advance.
268
269 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
270
271 * tree-vectorizer.c, tree.def: Fix comment typos.
272
273 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
274
275 * config/s390/s390.c (struct processor_costs): 4 new fields:
276 ddbr, ddr, debr, der.
277 (s390_rtx_costs): More precise handling of divide instructions.
278
279 2005-01-10 David Edelsohn <edelsohn@gnu.org>
280
281 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
282
283 2005-01-10 Jan Beulich <jbeulich@novell.com>
284
285 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
286 names resulting from -ffunction-sections/-fdata-sections and linkonce
287 ones.
288 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
289 based on the section name. Rearrange the section name comparison logic
290 slightly so that each section name is compared against at most once.
291
292 2005-01-10 Ben Elliston <bje@au.ibm.com>
293
294 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
295 targets for which -fPIC reduces limitations on the GOT size.
296
297 2005-01-09 Falk Hueffner <falk@debian.org>
298
299 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
300 X/C1 cmpop C2.
301
302 2005-01-09 David Edelsohn <edelsohn@gnu.org>
303
304 PR target/18720
305 * collect2.c (main): Set aixrtl_flag for -brtl option.
306 (resolve_lib_name): Search for .so file extension before .a
307 if aixrtl_flag set.
308
309 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
310
311 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
312 (vect_analyze_data_refs_alignment): Add dump prints.
313
314 2005-01-09 Ira Rosen <irar@il.ibm.com>
315
316 * tree-vectorizer.c (vect_analyze_offset_expr): Use
317 expr_invariant_in_loop_p.
318 Initialize outputs first thing in the function.
319 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
320 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
321 (vect_analyze_pointer_ref_access): Check that the initial condition of
322 the access function is loop invariant.
323
324 2005-01-09 Richard Henderson <rth@redhat.com>
325
326 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
327 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
328 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
329 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
330 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
331 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
332 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
333 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
334 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
335 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
336 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
337 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
338 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
339 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
340 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
341 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
342 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
343 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
344 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
345 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
346 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
347 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
348 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
349 mmx_ prefix.
350 (mmx_clrdi, pavgusb): Remove.
351 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
352 with leading sse_ prefix.
353 * config/i386/sse.md: Receive them.
354 * config/i386/mmx.md: New file.
355 (MMXMODE12, MMXMODE24, mmxvecsize): New.
356 (subrv2sf3): Turn into expander for normal subtraction.
357 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
358 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
359 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
360 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
361 commutative; use ix86_binary_operator_ok.
362 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
363 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
364 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
365 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
366 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
367 Macroize from existing patterns; use ix86_binary_operator_ok.
368 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
369 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
370 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
371 with vec_select+vec_concat.
372 (mmx_pshufw, mmx_pshufw_1): Likewise.
373 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
374 (mmx_uavgv4hi3): Model correctly.
375 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
376 _mm_xor_si64): Remove casts.
377
378 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
379
380 PR tree-optimization/19224
381 * tree-scalar-evolution.c (get_instantiated_value,
382 set_instantiated_value): New functions.
383 (instantiate_parameters_1): Cache the results.
384 (instantiate_parameters, resolve_mixers): Initialize and free
385 the cache.
386
387 2005-01-08 David Edelsohn <edelsohn@gnu.org>
388
389 * config/i386/i386.md (addhi_4): Correct reference in comment.
390 (addqi_4): Same.
391
392 2005-01-08 Richard Henderson <rth@redhat.com>
393
394 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
395 (_mm_cvtsi128_si64x): Likewise.
396 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
397 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
398 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
399 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
400 * config/i386/i386-protos.h: Update.
401 * config/i386/i386.c (print_operand): Add 'H'.
402 (ix86_fixup_binary_operands): Split out from ...
403 (ix86_expand_binary_operator): ... here.
404 (ix86_fixup_binary_operands_no_copy): New.
405 (ix86_expand_fp_absneg_operator): Handle vector mode results.
406 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
407 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
408 (safe_vector_operand): Use CONST0_RTX.
409 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
410 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
411 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
412 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
413 special cases.
414 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
415 (IX86_BUILTIN_CMPNGESS): New.
416 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
417 (attr type): Add sselog1.
418 (attr unit, attr memory): Handle it.
419 (movti, movti_internal, movti_rex64): Move near other integer moves.
420 (movtf, movtf_internal): Move near other fp moves.
421 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
422 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
423 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
424 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
425 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
426 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
427 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
428 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
429 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
430 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
431 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
432 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
433 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
434 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
435 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
436 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
437 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
438 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
439 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
440 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
441 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
442 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
443 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
444 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
445 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
446 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
447 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
448 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
449 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
450 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
451 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
452 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
453 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
454 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
455 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
456 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
457 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
458 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
459 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
460 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
461 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
462 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
463 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
464 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
465 same time.
466 (SSEPUSH, push<SSEPUSH>): Remove.
467 (MMXPUSH, push<MMXPUSH>): Remove.
468 (sse_movaps, sse_movaps_1, sse_movups): Remove.
469 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
470 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
471 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
472 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
473 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
474 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
475 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
476 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
477 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
478 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
479 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
480 (ppro_sse_log_V4SF): Handle sselog1.
481 * config/i386/predicates.md (const_0_to_1_operand): New.
482 (const_0_to_255_mul_8_operand): New.
483 (const_1_to_31_operand): Rename from const_int_1_31_operand.
484 (const_2_to_3_operand, const_4_to_7_operand): New.
485 * config/i386/sse.md: New file.
486 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
487 (sse_movups): Rename from sse_movups_1.
488 (sse_loadlss): Rename from sse_loadss_1.
489 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
490 from the name.
491 (negv4sf2): Use ix86_expand_fp_absneg_operator.
492 (absv4sf2, negv2df, absv2df): New.
493 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
494 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
495 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
496 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
497 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
498 (sse3_addsubv4sf3): Model correctly.
499 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
500 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
501 (sse_movhlps): Model with vec_select+vec_concat.
502 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
503 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
504 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
505 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
506 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
507 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
508 sse2_pshufhw_1): Likewise.
509 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
510 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
511 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
512 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
513 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
514 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
515 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
516 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
517 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
518 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
519 sminv8hi3): Mark commutative
520 operands. Use ix86_binary_operator_ok.
521 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
522 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
523 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
524 sse2_punpcklqdq): Allow operand2 in memory.
525 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
526 sse2_movsd): Add memory alternatives.
527 (sse_storelps): Turn expander into an insn; split after reload.
528 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
529 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
530
531 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
532
533 * configure.ac (DWARF-2 debug_line): Use objdump.
534 * configure: Regenerate.
535
536 2005-01-08 Jeff Law <law@redhat.com>
537 Diego Novillo <dnovillo@redhat.com>
538
539 PR tree-optimization/18241
540 * tree-nrv.c (tree_nrv): Ignore volatile return values.
541 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
542 statements with volatile operands.
543 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
544 operands after marking a statement with has_volatile_ops.
545
546 2005-01-08 Roger Sayle <roger@eyesopen.com>
547
548 * tree.c (int_fits_type_p): Always honor integer constant
549 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
550
551 2005-01-08 Roger Sayle <roger@eyesopen.com>
552
553 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
554 the latent bug in rtl_delete_block has been resolved.
555
556 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
557
558 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
559 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
560
561 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
562
563 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
564 library as @multilib_dir@/$(SHLIB_SONAME).
565
566 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
567
568 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
569 * configure: Regenerate.
570
571 2005-01-07 Jakub Jelinek <jakub@redhat.com>
572
573 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
574 TYPE_MODE.
575
576 2005-01-07 David Edelsohn <edelsohn@gnu.org>
577
578 PR target/13674
579 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
580 Convert non-word aligned offset address using ld/std into
581 indirect address.
582
583 2005-01-07 Richard Henderson <rth@redhat.com>
584
585 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
586
587 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
588
589 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
590
591 2005-01-07 Jakub Jelinek <jakub@redhat.com>
592
593 PR tree-optimization/19283
594 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
595 is not shorter than the original type.
596
597 PR rtl-optimization/19012
598 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
599
600 PR rtl-optimization/18861
601 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
602 (FULL_STATE): Define.
603 (mark_tablejump_edge): New function.
604 (purge_dead_tablejump_edges): New function.
605 (find_bb_boundaries): Use it.
606
607 PR tree-optimization/18828
608 * builtins.c (expand_builtin_next_arg): Remove argument and all
609 the argument checking.
610 (expand_builtin): Adjust caller.
611 (expand_builtin_va_start): Likewise. Remove error for too many
612 arguments.
613 (fold_builtin_next_arg): Issue error for too many arguments.
614 After checking arguments, replace them with magic arguments that
615 prevent further checking of the args.
616
617 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
618
619 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
620
621 2005-01-06 Roger Sayle <roger@eyesopen.com>
622
623 PR target/6077
624 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
625 and %{.S:...} (and their negative variants) to test whether the
626 input file is assembler or pre-processed-assembler independent of
627 the actual filename extension.
628
629 2005-01-06 Roger Sayle <roger@eyesopen.com>
630
631 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
632 of sign or zero extended values.
633
634 2005-01-06 Geoffrey Keating <geoffk@apple.com>
635
636 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
637 for IBM long double format correctly.
638
639 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
640
641 Fix PR tree-optimization/18792
642
643 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
644 to first_loop_depth, and use loop depth instead of loop number.
645 (build_classic_dir_vector): Ditto.
646 (compute_data_dependences_for_loop): Use depth, not loop number.
647 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
648 number. Pass in loops, instead of loop numbers.
649 (gather_interchange_stats): Ditto.
650 (linear_transform_loops): Ditto.
651
652 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
653
654 PR rtl-opt/13299
655 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
656 biv_fits_mode_p, extension_within_bounds_p): New functions.
657 (check_ext_dependent_givs): Use them.
658
659 2005-01-06 Roger Sayle <roger@eyesopen.com>
660
661 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
662 more than one barrier, in which case we should delete them all.
663
664 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
665
666 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
667
668 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
669
670 Fix PR middle-end/19286
671 Fix PR debug/19267
672 * dwarf2out.c (gen_subprogram_die): If we've already tried to
673 output this subprogram, simply ignore this attempt to do it again.
674 (add_abstract_origin_attribute): Don't abort trying to add the abstract
675 origin attribute if it's not possible.
676 (gen_block_die): Don't ignore subblocks of "unused" blocks.
677 (decls_for_scope): Ditto.
678 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
679 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
680
681 2005-01-05 Richard Henderson <rth@redhat.com>
682
683 PR target/11327
684 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
685 (bdesc_2arg): Use it.
686 (ix86_expand_binop_builtin): Force operands into registers
687 when optimizing.
688 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
689 ix86_expand_sse_compare, ix86_expand_sse_comi,
690 ix86_expand_builtin): Likewise.
691
692 2005-01-05 Richard Henderson <rth@redhat.com>
693
694 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
695 and VEC_SELECT.
696 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
697 (mulv4hi3): Set itanium_class mmmul.
698 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
699 (fpack): Rename from fpack_sfsf.
700
701 2005-01-05 Richard Henderson <rth@redhat.com>
702
703 PR rtl-opt/10692
704 * reload1.c (do_input_reload): Restrict the optimization deleteing
705 a previous output reload to RELOAD_FOR_INPUT.
706
707 2005-01-05 Steven Bosscher <stevenb@suse.de>
708
709 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
710 comment that we fall through after case.
711 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
712 (distribute_links): Likewise.
713 * cse.c (cse_insn): Likewise.
714 * cselib.c (cselib_invalidate_mem): Likewise.
715 * df.c: Update comments at the top of the file.
716 (read_modify_subreg_p): Update comments here too.
717 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
718 * flow.c (mark_set_1): Likewise.
719 (mark_used_regs): Likewise.
720 * gcse.c (mems_conflict_for_gcse_p): Likewise.
721 (canon_list_insert): Likewise.
722 (mark_set): Likewise.
723 (try_replace_reg): Likewise.
724 (store_killed_in_insn): Likewise.
725 * loop.c (count_one_set): Likewise.
726 (basic_induction_var): Likewise.
727 * postreload-gcse.c (find_mem_conflicts): Likewise.
728 * postreload.c (reload_combine_note_store): Likewise.
729 (move2add_note_store): Likewise.
730 * reload.c (find_equiv_reg): Likewise.
731 (mark_referenced_resources): Likewise.
732 * rtlanal.c (set_noop_p): Likewise.
733 (note_stores): Likewise.
734 (note_uses): Likewise.
735 * sched-deps.c (sched_analyze_1): Likewise.
736 * sched-rgn.c (check_live_1): Likewise.
737 (update_live_1): Likewise.
738 * config/i860/i860.c: Likewise.
739
740 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
741 an lvalue.
742 (ZERO_EXTRACT): Mention that this one can be an lvalue.
743
744 * doc/rtl.texi: Update documentation for bit-fields and SET.
745
746 2005-01-05 Roger Sayle <roger@eyesopen.com>
747
748 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
749
750 2005-01-05 Jan Hubicka <jh@suse.cz>
751 Richard Henderson <rth@redhat.com>
752
753 PR target/18910
754 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
755 with an offset.
756
757 2005-01-05 Richard Henderson <rth@redhat.com>
758
759 PR target/12902
760 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
761 (sse_shufps): Change operand 3 to const_int_operand.
762 (sse2_storelps): Fix typo in template.
763 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
764 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
765 (ix86_expand_builtin): Likewise.
766
767 2005-01-05 Julian Brown <julian@codesourcery.com>
768
769 * config/arm/arm.c (arm_return_in_memory): Treat complex types
770 as aggregates for AAPCS ABIs.
771
772 2005-01-05 Stan Shebs <shebs@apple.com>
773
774 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
775 Mach-O getters if ppc64.
776 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
777 size argument.
778 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
779 signed four-byte field for global code case.
780 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
781
782 2005-01-05 Roger Sayle <roger@eyesopen.com>
783
784 PR middle-end/19100
785 * c-common.c: Include real.h.
786 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
787 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
788 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
789 * Makefile.in (c-common.o): Update dependencies.
790
791 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
792
793 * c-parse.in (asm_string): Add trailing semicolon.
794
795 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
796
797 * c-parse.in (asm_string): New. Don't allow wide strings in
798 'asm'.
799 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
800 asm_string instead of STRING.
801
802 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
803
804 * c-typeck.c (constructor_no_implicit): Remove.
805 (set_designator, process_init_element): Don't check
806 constructor_no_implicit.
807
808 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
809 Kaz Kojima <kkojima@gcc.gnu.org>
810
811 PR target/16482
812 * lcm.c (create_pre_exit): New.
813 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
814 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
815
816 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
817
818 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
819 -msoft-float in target_float_switch.
820 * arm.c (arm_override_options): Fix processing of target_float_switch.
821
822 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
823
824 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
825 preferencing.
826
827 2004-01-05 Uros Bizjak <uros@kss-loka.si>
828
829 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
830 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
831
832 2005-01-04 Roger Sayle <roger@eyesopen.com>
833
834 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
835 if deleting the then-block allows the test-block to fallthru to the
836 else-block.
837
838 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
839
840 PR c/19152
841 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
842 after the full declaration if the are in two different TUs.
843
844 2005-01-04 Richard Henderson <rth@redhat.com>
845
846 PR tree-opt/19158
847 * tree-sra.c (generate_one_element_init): Just
848 call gimplify_and_add.
849 (generate_element_init): Record the
850 new referenced variables and mark them for renaming
851 and split out to ...
852 (generate_element_init_1): This.
853 (scalarize_init): Don't call push_gimplify_context/
854 pop_gimplify_context.
855
856 2005-01-04 Geoffrey Keating <geoffk@apple.com>
857
858 * toplev.c (get_src_pwd): Handle failure of getpwd().
859
860 2005-01-04 Roger Sayle <roger@eyesopen.com>
861
862 * fold-const.c (fold_single_bit_test): Delete unreachable handling
863 of TRUTH_NOT_EXPR.
864 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
865 all the cases handled by it are inverted by invert_truthvalue.
866
867 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
868
869 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
870 (save_register_windowsi): Likewise.
871
872 2005-01-04 Richard Henderson <rth@redhat.com>
873
874 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
875
876 2005-01-04 Uros Bizjak <uros@kss-loka.si>
877
878 PR middle-end/17767
879 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
880 any simplifications of vector mode comparison operators.
881 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
882
883 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
884 Devang Patel <dpatel@apple.com>
885
886 PR tree-optimization/18308
887 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
888 the operands before creating a new expression.
889 * dojump.c (do_jump): Make drop_through_label available
890 for all cases. Add expansion of COND_EXPR.
891
892 2005-01-04 Ira Rosen <irar@il.ibm.com>
893
894 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
895 INTEGER_CST instead of TREE_CONSTANT.
896 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
897 instead of TREE_CONSTANT.
898 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
899 instead of TREE_CONSTANT.
900
901 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
902
903 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
904
905 2005-01-04 Uros Bizjak <uros@kss-loka.si>
906
907 PR target/19240
908 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
909 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
910
911 2005-01-03 Richard Henderson <rth@redhat.com>
912
913 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
914 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
915 fold_convert_const_real_from_real): Split out from ...
916 (fold_convert_const): ... here.
917
918 2005-01-03 Richard Henderson <rth@redhat.com>
919
920 PR target/19235
921 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
922 (mov<MMXMODEI>_internal): Likewise.
923 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
924 Add V2SF case; use it for SSE1; don't use TI.
925 (movdf_integer): Likewise.
926 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
927
928 2005-01-03 Ira Rosen <irar@il.ibm.com>
929
930 * tree-vectorizer.c (vect_strip_conversions): New function.
931 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
932 check for binary class.
933
934 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
935
936 Fix PR debug/17924
937 Fix PR debug/19191
938 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
939 is a decl.
940 * gimple-low.c (mark_blocks_with_used_vars): New function.
941 (mark_blocks_with_used_subblocks): Ditto.
942 (mark_used_blocks): Ditto.
943 (pass_mark_used_blocks): New pass.
944 * tree-inline.c: Include debug.h.
945 (expand_call_inline): Call outlining_inline_function here.
946 * tree-optimize.c (init_tree_optimization_passes): Add
947 pass_mark_used_blocks.
948 * tree-pass.h (pass_mark_used_blocks): New.
949 * Makefile.in (tree-inline.o): Add debug.h dependency.
950
951 2005-01-03 Geoffrey Keating <geoffk@apple.com>
952
953 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
954 VAR_DECLs to have weak_import attribute.
955
956 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
957 'struct ucontext'.
958
959 2004-01-03 Steven Bosscher <stevenb@suse.de>
960
961 * passes.c (rest_of_compilation): Don't run regmove if only
962 flag_expensive_optimizations, require flag_regmove instead.
963
964 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
965
966 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
967 register "sparc".
968 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
969 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
970 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
971 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
972
973 2005-01-03 Richard Henderson <rth@redhat.com>
974
975 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
976 (ia64_const_ok_for_letter_p): New.
977 (ia64_const_double_ok_for_letter_p): New.
978 (ia64_extra_constraint): New.
979 (ia64_expand_vecint_compare): New.
980 (ia64_expand_vcondu_v2si): New.
981 (ia64_expand_vecint_cmov): New.
982 (ia64_expand_vecint_minmax): New.
983 (ia64_print_operand): Add 'v'.
984 (ia64_preferred_reload_class): New.
985 (ia64_vector_mode_supported_p): New.
986 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
987 (PREFERRED_RELOAD_CLASS): Move to function.
988 (CONST_OK_FOR_LETTER_P): Move to function.
989 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
990 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
991 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
992 (EXTRA_CONSTRAINT): Move to function.
993 * config/ia64/ia64.md: Include vect.md.
994 (itanium_class): Add mmalua.
995 (type): Handle it.
996 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
997 (1b_mmalua): New.
998 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
999 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
1000 CONST0_RTX.
1001 (const_int_2bit_operand): New.
1002 (fr_reg_or_0_operand): New.
1003 * config/ia64/ia64-modes.def: Add vector modes.
1004 * config/ia64/ia64-protos.h: Update.
1005 * config/ia64/vect.md: New file.
1006
1007 2005-01-03 Richard Henderson <rth@redhat.com>
1008
1009 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
1010
1011 2005-01-03 Uros Bizjak <uros@kss-loka.si>
1012
1013 PR target/19236
1014 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
1015 to SFmode.
1016 (log1pdf2): Change mode of operands[0,1] to DFmode.
1017
1018 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
1019
1020 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
1021 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
1022 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
1023 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
1024 SPARC_RELAXED_ORDERING.
1025
1026 2005-01-03 Richard Henderson <rth@redhat.com>
1027 Uros Bizjak <uros@kss-loka.si>
1028
1029 PR target/14631
1030 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
1031 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
1032 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
1033 handling.
1034 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
1035 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
1036 const_pow2_1_to_128_operand): New predicates.
1037
1038 2005-01-02 Greg McGary <greg@mcgary.org>
1039
1040 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
1041 anonymous decls.
1042 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
1043 object. [ptr_type]: Remove unused variable.
1044 Remember nearest addressable array-element or record-component
1045 when checking bitfield components. Tolerate empty BIND_EXPR.
1046
1047 2005-01-01 Richard Henderson <rth@redhat.com>
1048
1049 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
1050 that don't narrow the value. Fail for other conversions.
1051
1052 2005-01-01 Richard Henderson <rth@redhat.com>
1053
1054 PR c/19031
1055 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
1056 * c-lang.c (finish_file): Don't do it here.
1057 * objc/objc-act.c (objc_finish_file): Likewise.
1058
1059 * cgraph.c (decl_assembler_name_equal): New.
1060 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
1061 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
1062 * cgraph.h (struct cgraph_varpool_node): Add next.
1063 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
1064 * varasm.c (assemble_alias): Mark the target as needed.
1065
1066 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
1067
1068 PR middle-end/19221
1069 * function.c (get_arg_pointer_save_area): Use entry_of_function
1070 instead of get_insns.
1071
1072 2005-01-01 Roger Sayle <roger@eyesopen.com>
1073 Andrew Pinski <pinskia@physics.uc.edu>
1074 James E. Wilson <wilson@specifixinc.com>
1075
1076 PR rtl-optimization/12092
1077 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
1078 is zero.
1079
1080 2005-01-01 Roger Sayle <roger@eyesopen.com>
1081 Olivier Hainque <hainque@act-europe.fr>
1082
1083 * tree.c (int_fits_type_p): A narrower type always fits in a
1084 wider one, except for negative values into unsigned types.
1085
1086 2005-01-01 Roger Sayle <roger@eyesopen.com>
1087
1088 * tree.c (int_fits_type_p): Compare the result of force_fit_type
1089 with the original constant rather than require TREE_OVERFLOW.
1090
1091 2005-01-01 Steven Bosscher <stevenb@suse.de>
1092
1093 PR middle-end/17544
1094 * c-decl.c (finish_function): If compiling C99, annotate the
1095 compiler generated return with the current file name and line 0.
1096 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
1097 the source line is greater than 0.
1098 (remove_bb): Likewise.
1099
1100 See ChangeLog.12 for earlier changes.
This page took 0.084541 seconds and 5 git commands to generate.