]> gcc.gnu.org Git - gcc.git/blob - gcc/ChangeLog
re PR tree-optimization/17949 (Tree loop optimization generates unaligned access...
[gcc.git] / gcc / ChangeLog
1 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
2
3 PR tree-optimization/17949
4 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
5 (find_interesting_uses_address): Use it.
6
7 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
8
9 * regrename.c (kill_value): Handle subreg's that won't simplify.
10
11 2005-01-11 Richard Henderson <rth@redhat.com>
12
13 PR target/13366
14 * config/i386/i386.h (enum ix86_builtins): Move ...
15 * config/i386/i386.c: ... here.
16 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
17 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
18 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
19 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
20 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
21 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
22 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
23 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
24 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
25 IX86_BUILTIN_LOADDDUP): Remove.
26 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
27 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
28 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
29 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
30 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
31 (ix86_init_builtins): Make static.
32 (ix86_init_mmx_sse_builtins): Update for changed builtins.
33 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
34 if all the modes match. Otherwise, fake it.
35 (get_element_number, ix86_expand_vec_init_builtin,
36 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
37 (ix86_expand_builtin): Make static. Update for changed builtins.
38 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
39 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
40 (ix86_expand_vector_init_duplicate): New.
41 (ix86_expand_vector_init_low_nonzero): New.
42 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
43 Split out from ix86_expand_vector_init; handle integer modes.
44 (ix86_expand_vector_init): Use them.
45 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
46 * config/i386/i386-protos.h: Update.
47 * config/i386/predicates.md (reg_or_0_operand): New.
48 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
49 (movv2sf_internal): Likewise. And a splitter to match them all.
50 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
51 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
52 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
53 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
54 (mmx_pinsrw): Fix operand ordering.
55 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
56 rather than sse_loadss expander.
57 (movv2df splitter): Similarly.
58 (sse_loadss, sse_loadlss): Remove.
59 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
60 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
61 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
62 (sse3_movddup): Rename with '*'.
63 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
64 (sse2_loadsd): Remove.
65 (vec_dupv2df_sse3): Rename from sse3_loadddup.
66 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
67 (sse2_pinsrw): Fix argument ordering.
68 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
69 (sse2_stored): Remove 'r' destination.
70 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
71 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
72 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
73 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
74 vec_extractv16qi, vec_initv16qi): New.
75
76 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
77 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
78 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
79 constructor form.
80 (_mm_load_pd, _mm_store_pd): Use plain dereference.
81 (_mm_load_si128, _mm_store_si128): Likewise.
82 (_mm_load1_pd): Use _mm_set1_pd.
83 (_mm_load_sd): Use _mm_set_sd.
84 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
85 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
86 (_mm_set_epi64): Use _mm_set_epi64x.
87 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
88 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
89 _mm_setr_epi8): Use _mm_set_foo form.
90 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
91 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
92 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
93 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
94 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
95 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
96 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
97 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
98 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
99 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
100 (_mm_movedup_pd): Use _mm_shuffle_pd.
101 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
102 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
103 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
104 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
105 Use _mm_setzero_ps.
106 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
107 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
108 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
109 (_mm_store_ps): Use raw dereference.
110 (_mm_store1_ps): Use _mm_storeu_ps.
111 (_mm_storer_ps): Use _mm_store_ps.
112 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
113 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
114
115 2005-01-11 Stan Shebs <shebs@apple.com>
116
117 * config/rs6000/rs6000.c (machopic_output_stub): Issue
118 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
119
120 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
121
122 PR target/18761
123 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
124 Skip all DECLs except for FIELD_DECLs.
125
126 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
127
128 * config/s390/s390.c (override_options): Return error if
129 -mbackchain, -mpacked-stack and -mhard-float are used together.
130 (s390_va_start): Remove the backchain && packed-stack special case.
131 (s390_gimplify_va_arg): Likewise.
132 * doc/invoke.texi: Remove the ABI incompatibility note.
133
134 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
135
136 * config/s390/s390.c (struct s390_frame_layout): Remove
137 save_backchain_p.
138 (s390_frame_info, s390_emit_prologue): Replace occurrences of
139 save_backchain_p with TARGET_BACKCHAIN.
140
141 2005-01-11 Alan Modra <amodra@bigpond.net.au>
142
143 PR target/18916
144 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
145 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
146 * function.c (locate_and_pad_parm): Set new field.
147 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
148 Tweak where_pad test to include "none". Always set mem align for
149 stack_parm.
150 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
151 not sufficient for type.
152 (assign_parm_setup_block): If stack_parm is zero on entry, always
153 make a new stack local. Block move old stack parm if necessary
154 to new aligned stack local.
155 (assign_parm_setup_stack): Use a block move to handle
156 potentially misaligned entry_parm.
157 (assign_parms_unsplit_complex): Specify required alignment when
158 creating stack local.
159 * calls.c (compute_argument_addresses): Override alignment of stack
160 arg calculated from its type with the alignment given by
161 FUNCTION_ARG_BOUNDARY.
162 (store_one_arg): Likewise.
163
164 2005-01-11 Jan Beulich <jbeulich@novell.com>
165
166 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
167 Change respective itanium_class attribute to ialu.
168 (shladdp4_internal): New.
169 * config/ia64/predicates.md (shladd_log2_operand): New.
170
171 2005-01-11 Richard Henderson <rth@redhat.com>
172
173 * expr.c (store_constructor): Use rtvec_alloc instead of
174 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
175
176 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
177
178 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
179
180 2005-01-11 Alan Modra <amodra@bigpond.net.au>
181
182 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
183 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
184 (TARGET_RWRELOC): Define.
185 (ia64_rwreloc_section_type_flags): Delete.
186 (ia64_section_type_flags): New function.
187 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
188 (TARGET_RWRELOC): Define.
189
190 2005-01-10 David Mosberger <davidm@hpl.hp.com>
191
192 PR target/18987
193 * config/ia64/ia64.c (process_set): For alloc insn, only call
194 process_epilogue is !frame_pointer_needed.
195
196 2005-01-10 Roger Sayle <roger@eyesopen.com>
197
198 PR c++/19355
199 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
200 unary operator and can't be treated as a binary/comparison operator.
201
202 2005-01-10 Richard Henderson <rth@redhat.com>
203
204 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
205 (ix86_return_in_memory): Likewise.
206 (function_arg_advance): Likewise. Mirror structure in function_arg
207 for choosing register to advance.
208
209 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
210
211 * tree-vectorizer.c, tree.def: Fix comment typos.
212
213 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
214
215 * config/s390/s390.c (struct processor_costs): 4 new fields:
216 ddbr, ddr, debr, der.
217 (s390_rtx_costs): More precise handling of divide instructions.
218
219 2005-01-10 David Edelsohn <edelsohn@gnu.org>
220
221 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
222
223 2005-01-10 Jan Beulich <jbeulich@novell.com>
224
225 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
226 names resulting from -ffunction-sections/-fdata-sections and linkonce
227 ones.
228 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
229 based on the section name. Rearrange the section name comparison logic
230 slightly so that each section name is compared against at most once.
231
232 2005-01-10 Ben Elliston <bje@au.ibm.com>
233
234 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
235 targets for which -fPIC reduces limitations on the GOT size.
236
237 2005-01-09 Falk Hueffner <falk@debian.org>
238
239 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
240 X/C1 cmpop C2.
241
242 2005-01-09 David Edelsohn <edelsohn@gnu.org>
243
244 PR target/18720
245 * collect2.c (main): Set aixrtl_flag for -brtl option.
246 (resolve_lib_name): Search for .so file extension before .a
247 if aixrtl_flag set.
248
249 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
250
251 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
252 (vect_analyze_data_refs_alignment): Add dump prints.
253
254 2005-01-09 Ira Rosen <irar@il.ibm.com>
255
256 * tree-vectorizer.c (vect_analyze_offset_expr): Use
257 expr_invariant_in_loop_p.
258 Initialize outputs first thing in the function.
259 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
260 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
261 (vect_analyze_pointer_ref_access): Check that the initial condition of
262 the access function is loop invariant.
263
264 2005-01-09 Richard Henderson <rth@redhat.com>
265
266 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
267 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
268 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
269 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
270 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
271 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
272 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
273 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
274 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
275 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
276 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
277 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
278 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
279 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
280 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
281 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
282 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
283 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
284 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
285 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
286 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
287 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
288 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
289 mmx_ prefix.
290 (mmx_clrdi, pavgusb): Remove.
291 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
292 with leading sse_ prefix.
293 * config/i386/sse.md: Receive them.
294 * config/i386/mmx.md: New file.
295 (MMXMODE12, MMXMODE24, mmxvecsize): New.
296 (subrv2sf3): Turn into expander for normal subtraction.
297 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
298 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
299 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
300 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
301 commutative; use ix86_binary_operator_ok.
302 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
303 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
304 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
305 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
306 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
307 Macroize from existing patterns; use ix86_binary_operator_ok.
308 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
309 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
310 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
311 with vec_select+vec_concat.
312 (mmx_pshufw, mmx_pshufw_1): Likewise.
313 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
314 (mmx_uavgv4hi3): Model correctly.
315 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
316 _mm_xor_si64): Remove casts.
317
318 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
319
320 PR tree-optimization/19224
321 * tree-scalar-evolution.c (get_instantiated_value,
322 set_instantiated_value): New functions.
323 (instantiate_parameters_1): Cache the results.
324 (instantiate_parameters, resolve_mixers): Initialize and free
325 the cache.
326
327 2005-01-08 David Edelsohn <edelsohn@gnu.org>
328
329 * config/i386/i386.md (addhi_4): Correct reference in comment.
330 (addqi_4): Same.
331
332 2005-01-08 Richard Henderson <rth@redhat.com>
333
334 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
335 (_mm_cvtsi128_si64x): Likewise.
336 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
337 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
338 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
339 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
340 * config/i386/i386-protos.h: Update.
341 * config/i386/i386.c (print_operand): Add 'H'.
342 (ix86_fixup_binary_operands): Split out from ...
343 (ix86_expand_binary_operator): ... here.
344 (ix86_fixup_binary_operands_no_copy): New.
345 (ix86_expand_fp_absneg_operator): Handle vector mode results.
346 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
347 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
348 (safe_vector_operand): Use CONST0_RTX.
349 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
350 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
351 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
352 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
353 special cases.
354 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
355 (IX86_BUILTIN_CMPNGESS): New.
356 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
357 (attr type): Add sselog1.
358 (attr unit, attr memory): Handle it.
359 (movti, movti_internal, movti_rex64): Move near other integer moves.
360 (movtf, movtf_internal): Move near other fp moves.
361 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
362 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
363 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
364 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
365 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
366 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
367 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
368 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
369 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
370 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
371 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
372 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
373 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
374 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
375 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
376 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
377 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
378 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
379 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
380 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
381 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
382 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
383 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
384 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
385 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
386 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
387 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
388 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
389 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
390 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
391 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
392 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
393 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
394 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
395 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
396 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
397 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
398 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
399 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
400 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
401 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
402 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
403 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
404 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
405 same time.
406 (SSEPUSH, push<SSEPUSH>): Remove.
407 (MMXPUSH, push<MMXPUSH>): Remove.
408 (sse_movaps, sse_movaps_1, sse_movups): Remove.
409 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
410 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
411 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
412 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
413 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
414 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
415 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
416 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
417 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
418 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
419 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
420 (ppro_sse_log_V4SF): Handle sselog1.
421 * config/i386/predicates.md (const_0_to_1_operand): New.
422 (const_0_to_255_mul_8_operand): New.
423 (const_1_to_31_operand): Rename from const_int_1_31_operand.
424 (const_2_to_3_operand, const_4_to_7_operand): New.
425 * config/i386/sse.md: New file.
426 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
427 (sse_movups): Rename from sse_movups_1.
428 (sse_loadlss): Rename from sse_loadss_1.
429 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
430 from the name.
431 (negv4sf2): Use ix86_expand_fp_absneg_operator.
432 (absv4sf2, negv2df, absv2df): New.
433 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
434 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
435 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
436 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
437 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
438 (sse3_addsubv4sf3): Model correctly.
439 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
440 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
441 (sse_movhlps): Model with vec_select+vec_concat.
442 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
443 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
444 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
445 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
446 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
447 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
448 sse2_pshufhw_1): Likewise.
449 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
450 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
451 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
452 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
453 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
454 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
455 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
456 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
457 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
458 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
459 sminv8hi3): Mark commutative
460 operands. Use ix86_binary_operator_ok.
461 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
462 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
463 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
464 sse2_punpcklqdq): Allow operand2 in memory.
465 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
466 sse2_movsd): Add memory alternatives.
467 (sse_storelps): Turn expander into an insn; split after reload.
468 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
469 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
470
471 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
472
473 * configure.ac (DWARF-2 debug_line): Use objdump.
474 * configure: Regenerate.
475
476 2005-01-08 Jeff Law <law@redhat.com>
477 Diego Novillo <dnovillo@redhat.com>
478
479 PR tree-optimization/18241
480 * tree-nrv.c (tree_nrv): Ignore volatile return values.
481 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
482 statements with volatile operands.
483 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
484 operands after marking a statement with has_volatile_ops.
485
486 2005-01-08 Roger Sayle <roger@eyesopen.com>
487
488 * tree.c (int_fits_type_p): Always honor integer constant
489 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
490
491 2005-01-08 Roger Sayle <roger@eyesopen.com>
492
493 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
494 the latent bug in rtl_delete_block has been resolved.
495
496 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
497
498 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
499 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
500
501 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
502
503 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
504 library as @multilib_dir@/$(SHLIB_SONAME).
505
506 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
507
508 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
509 * configure: Regenerate.
510
511 2005-01-07 Jakub Jelinek <jakub@redhat.com>
512
513 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
514 TYPE_MODE.
515
516 2005-01-07 David Edelsohn <edelsohn@gnu.org>
517
518 PR target/13674
519 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
520 Convert non-word aligned offset address using ld/std into
521 indirect address.
522
523 2005-01-07 Richard Henderson <rth@redhat.com>
524
525 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
526
527 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
528
529 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
530
531 2005-01-07 Jakub Jelinek <jakub@redhat.com>
532
533 PR tree-optimization/19283
534 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
535 is not shorter than the original type.
536
537 PR rtl-optimization/19012
538 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
539
540 PR rtl-optimization/18861
541 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
542 (FULL_STATE): Define.
543 (mark_tablejump_edge): New function.
544 (purge_dead_tablejump_edges): New function.
545 (find_bb_boundaries): Use it.
546
547 PR tree-optimization/18828
548 * builtins.c (expand_builtin_next_arg): Remove argument and all
549 the argument checking.
550 (expand_builtin): Adjust caller.
551 (expand_builtin_va_start): Likewise. Remove error for too many
552 arguments.
553 (fold_builtin_next_arg): Issue error for too many arguments.
554 After checking arguments, replace them with magic arguments that
555 prevent further checking of the args.
556
557 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
558
559 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
560
561 2005-01-06 Roger Sayle <roger@eyesopen.com>
562
563 PR target/6077
564 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
565 and %{.S:...} (and their negative variants) to test whether the
566 input file is assembler or pre-processed-assembler independent of
567 the actual filename extension.
568
569 2005-01-06 Roger Sayle <roger@eyesopen.com>
570
571 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
572 of sign or zero extended values.
573
574 2005-01-06 Geoffrey Keating <geoffk@apple.com>
575
576 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
577 for IBM long double format correctly.
578
579 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
580
581 Fix PR tree-optimization/18792
582
583 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
584 to first_loop_depth, and use loop depth instead of loop number.
585 (build_classic_dir_vector): Ditto.
586 (compute_data_dependences_for_loop): Use depth, not loop number.
587 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
588 number. Pass in loops, instead of loop numbers.
589 (gather_interchange_stats): Ditto.
590 (linear_transform_loops): Ditto.
591
592 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
593
594 PR rtl-opt/13299
595 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
596 biv_fits_mode_p, extension_within_bounds_p): New functions.
597 (check_ext_dependent_givs): Use them.
598
599 2005-01-06 Roger Sayle <roger@eyesopen.com>
600
601 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
602 more than one barrier, in which case we should delete them all.
603
604 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
605
606 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
607
608 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
609
610 Fix PR middle-end/19286
611 Fix PR debug/19267
612 * dwarf2out.c (gen_subprogram_die): If we've already tried to
613 output this subprogram, simply ignore this attempt to do it again.
614 (add_abstract_origin_attribute): Don't abort trying to add the abstract
615 origin attribute if it's not possible.
616 (gen_block_die): Don't ignore subblocks of "unused" blocks.
617 (decls_for_scope): Ditto.
618 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
619 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
620
621 2005-01-05 Richard Henderson <rth@redhat.com>
622
623 PR target/11327
624 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
625 (bdesc_2arg): Use it.
626 (ix86_expand_binop_builtin): Force operands into registers
627 when optimizing.
628 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
629 ix86_expand_sse_compare, ix86_expand_sse_comi,
630 ix86_expand_builtin): Likewise.
631
632 2005-01-05 Richard Henderson <rth@redhat.com>
633
634 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
635 and VEC_SELECT.
636 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
637 (mulv4hi3): Set itanium_class mmmul.
638 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
639 (fpack): Rename from fpack_sfsf.
640
641 2005-01-05 Richard Henderson <rth@redhat.com>
642
643 PR rtl-opt/10692
644 * reload1.c (do_input_reload): Restrict the optimization deleteing
645 a previous output reload to RELOAD_FOR_INPUT.
646
647 2005-01-05 Steven Bosscher <stevenb@suse.de>
648
649 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
650 comment that we fall through after case.
651 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
652 (distribute_links): Likewise.
653 * cse.c (cse_insn): Likewise.
654 * cselib.c (cselib_invalidate_mem): Likewise.
655 * df.c: Update comments at the top of the file.
656 (read_modify_subreg_p): Update comments here too.
657 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
658 * flow.c (mark_set_1): Likewise.
659 (mark_used_regs): Likewise.
660 * gcse.c (mems_conflict_for_gcse_p): Likewise.
661 (canon_list_insert): Likewise.
662 (mark_set): Likewise.
663 (try_replace_reg): Likewise.
664 (store_killed_in_insn): Likewise.
665 * loop.c (count_one_set): Likewise.
666 (basic_induction_var): Likewise.
667 * postreload-gcse.c (find_mem_conflicts): Likewise.
668 * postreload.c (reload_combine_note_store): Likewise.
669 (move2add_note_store): Likewise.
670 * reload.c (find_equiv_reg): Likewise.
671 (mark_referenced_resources): Likewise.
672 * rtlanal.c (set_noop_p): Likewise.
673 (note_stores): Likewise.
674 (note_uses): Likewise.
675 * sched-deps.c (sched_analyze_1): Likewise.
676 * sched-rgn.c (check_live_1): Likewise.
677 (update_live_1): Likewise.
678 * config/i860/i860.c: Likewise.
679
680 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
681 an lvalue.
682 (ZERO_EXTRACT): Mention that this one can be an lvalue.
683
684 * doc/rtl.texi: Update documentation for bit-fields and SET.
685
686 2005-01-05 Roger Sayle <roger@eyesopen.com>
687
688 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
689
690 2005-01-05 Jan Hubicka <jh@suse.cz>
691 Richard Henderson <rth@redhat.com>
692
693 PR target/18910
694 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
695 with an offset.
696
697 2005-01-05 Richard Henderson <rth@redhat.com>
698
699 PR target/12902
700 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
701 (sse_shufps): Change operand 3 to const_int_operand.
702 (sse2_storelps): Fix typo in template.
703 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
704 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
705 (ix86_expand_builtin): Likewise.
706
707 2005-01-05 Julian Brown <julian@codesourcery.com>
708
709 * config/arm/arm.c (arm_return_in_memory): Treat complex types
710 as aggregates for AAPCS ABIs.
711
712 2005-01-05 Stan Shebs <shebs@apple.com>
713
714 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
715 Mach-O getters if ppc64.
716 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
717 size argument.
718 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
719 signed four-byte field for global code case.
720 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
721
722 2005-01-05 Roger Sayle <roger@eyesopen.com>
723
724 PR middle-end/19100
725 * c-common.c: Include real.h.
726 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
727 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
728 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
729 * Makefile.in (c-common.o): Update dependencies.
730
731 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
732
733 * c-parse.in (asm_string): Add trailing semicolon.
734
735 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
736
737 * c-parse.in (asm_string): New. Don't allow wide strings in
738 'asm'.
739 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
740 asm_string instead of STRING.
741
742 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
743
744 * c-typeck.c (constructor_no_implicit): Remove.
745 (set_designator, process_init_element): Don't check
746 constructor_no_implicit.
747
748 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
749 Kaz Kojima <kkojima@gcc.gnu.org>
750
751 PR target/16482
752 * lcm.c (create_pre_exit): New.
753 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
754 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
755
756 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
757
758 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
759 -msoft-float in target_float_switch.
760 * arm.c (arm_override_options): Fix processing of target_float_switch.
761
762 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
763
764 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
765 preferencing.
766
767 2004-01-05 Uros Bizjak <uros@kss-loka.si>
768
769 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
770 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
771
772 2005-01-04 Roger Sayle <roger@eyesopen.com>
773
774 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
775 if deleting the then-block allows the test-block to fallthru to the
776 else-block.
777
778 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
779
780 PR c/19152
781 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
782 after the full declaration if the are in two different TUs.
783
784 2005-01-04 Richard Henderson <rth@redhat.com>
785
786 PR tree-opt/19158
787 * tree-sra.c (generate_one_element_init): Just
788 call gimplify_and_add.
789 (generate_element_init): Record the
790 new referenced variables and mark them for renaming
791 and split out to ...
792 (generate_element_init_1): This.
793 (scalarize_init): Don't call push_gimplify_context/
794 pop_gimplify_context.
795
796 2005-01-04 Geoffrey Keating <geoffk@apple.com>
797
798 * toplev.c (get_src_pwd): Handle failure of getpwd().
799
800 2005-01-04 Roger Sayle <roger@eyesopen.com>
801
802 * fold-const.c (fold_single_bit_test): Delete unreachable handling
803 of TRUTH_NOT_EXPR.
804 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
805 all the cases handled by it are inverted by invert_truthvalue.
806
807 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
808
809 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
810 (save_register_windowsi): Likewise.
811
812 2005-01-04 Richard Henderson <rth@redhat.com>
813
814 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
815
816 2005-01-04 Uros Bizjak <uros@kss-loka.si>
817
818 PR middle-end/17767
819 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
820 any simplifications of vector mode comparison operators.
821 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
822
823 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
824 Devang Patel <dpatel@apple.com>
825
826 PR tree-optimization/18308
827 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
828 the operands before creating a new expression.
829 * dojump.c (do_jump): Make drop_through_label available
830 for all cases. Add expansion of COND_EXPR.
831
832 2005-01-04 Ira Rosen <irar@il.ibm.com>
833
834 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
835 INTEGER_CST instead of TREE_CONSTANT.
836 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
837 instead of TREE_CONSTANT.
838 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
839 instead of TREE_CONSTANT.
840
841 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
842
843 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
844
845 2005-01-04 Uros Bizjak <uros@kss-loka.si>
846
847 PR target/19240
848 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
849 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
850
851 2005-01-03 Richard Henderson <rth@redhat.com>
852
853 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
854 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
855 fold_convert_const_real_from_real): Split out from ...
856 (fold_convert_const): ... here.
857
858 2005-01-03 Richard Henderson <rth@redhat.com>
859
860 PR target/19235
861 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
862 (mov<MMXMODEI>_internal): Likewise.
863 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
864 Add V2SF case; use it for SSE1; don't use TI.
865 (movdf_integer): Likewise.
866 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
867
868 2005-01-03 Ira Rosen <irar@il.ibm.com>
869
870 * tree-vectorizer.c (vect_strip_conversions): New function.
871 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
872 check for binary class.
873
874 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
875
876 Fix PR debug/17924
877 Fix PR debug/19191
878 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
879 is a decl.
880 * gimple-low.c (mark_blocks_with_used_vars): New function.
881 (mark_blocks_with_used_subblocks): Ditto.
882 (mark_used_blocks): Ditto.
883 (pass_mark_used_blocks): New pass.
884 * tree-inline.c: Include debug.h.
885 (expand_call_inline): Call outlining_inline_function here.
886 * tree-optimize.c (init_tree_optimization_passes): Add
887 pass_mark_used_blocks.
888 * tree-pass.h (pass_mark_used_blocks): New.
889 * Makefile.in (tree-inline.o): Add debug.h dependency.
890
891 2005-01-03 Geoffrey Keating <geoffk@apple.com>
892
893 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
894 VAR_DECLs to have weak_import attribute.
895
896 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
897 'struct ucontext'.
898
899 2004-01-03 Steven Bosscher <stevenb@suse.de>
900
901 * passes.c (rest_of_compilation): Don't run regmove if only
902 flag_expensive_optimizations, require flag_regmove instead.
903
904 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
905
906 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
907 register "sparc".
908 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
909 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
910 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
911 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
912
913 2005-01-03 Richard Henderson <rth@redhat.com>
914
915 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
916 (ia64_const_ok_for_letter_p): New.
917 (ia64_const_double_ok_for_letter_p): New.
918 (ia64_extra_constraint): New.
919 (ia64_expand_vecint_compare): New.
920 (ia64_expand_vcondu_v2si): New.
921 (ia64_expand_vecint_cmov): New.
922 (ia64_expand_vecint_minmax): New.
923 (ia64_print_operand): Add 'v'.
924 (ia64_preferred_reload_class): New.
925 (ia64_vector_mode_supported_p): New.
926 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
927 (PREFERRED_RELOAD_CLASS): Move to function.
928 (CONST_OK_FOR_LETTER_P): Move to function.
929 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
930 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
931 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
932 (EXTRA_CONSTRAINT): Move to function.
933 * config/ia64/ia64.md: Include vect.md.
934 (itanium_class): Add mmalua.
935 (type): Handle it.
936 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
937 (1b_mmalua): New.
938 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
939 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
940 CONST0_RTX.
941 (const_int_2bit_operand): New.
942 (fr_reg_or_0_operand): New.
943 * config/ia64/ia64-modes.def: Add vector modes.
944 * config/ia64/ia64-protos.h: Update.
945 * config/ia64/vect.md: New file.
946
947 2005-01-03 Richard Henderson <rth@redhat.com>
948
949 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
950
951 2005-01-03 Uros Bizjak <uros@kss-loka.si>
952
953 PR target/19236
954 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
955 to SFmode.
956 (log1pdf2): Change mode of operands[0,1] to DFmode.
957
958 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
959
960 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
961 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
962 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
963 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
964 SPARC_RELAXED_ORDERING.
965
966 2005-01-03 Richard Henderson <rth@redhat.com>
967 Uros Bizjak <uros@kss-loka.si>
968
969 PR target/14631
970 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
971 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
972 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
973 handling.
974 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
975 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
976 const_pow2_1_to_128_operand): New predicates.
977
978 2005-01-02 Greg McGary <greg@mcgary.org>
979
980 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
981 anonymous decls.
982 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
983 object. [ptr_type]: Remove unused variable.
984 Remember nearest addressable array-element or record-component
985 when checking bitfield components. Tolerate empty BIND_EXPR.
986
987 2005-01-01 Richard Henderson <rth@redhat.com>
988
989 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
990 that don't narrow the value. Fail for other conversions.
991
992 2005-01-01 Richard Henderson <rth@redhat.com>
993
994 PR c/19031
995 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
996 * c-lang.c (finish_file): Don't do it here.
997 * objc/objc-act.c (objc_finish_file): Likewise.
998
999 * cgraph.c (decl_assembler_name_equal): New.
1000 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
1001 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
1002 * cgraph.h (struct cgraph_varpool_node): Add next.
1003 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
1004 * varasm.c (assemble_alias): Mark the target as needed.
1005
1006 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
1007
1008 PR middle-end/19221
1009 * function.c (get_arg_pointer_save_area): Use entry_of_function
1010 instead of get_insns.
1011
1012 2005-01-01 Roger Sayle <roger@eyesopen.com>
1013 Andrew Pinski <pinskia@physics.uc.edu>
1014 James E. Wilson <wilson@specifixinc.com>
1015
1016 PR rtl-optimization/12092
1017 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
1018 is zero.
1019
1020 2005-01-01 Roger Sayle <roger@eyesopen.com>
1021 Olivier Hainque <hainque@act-europe.fr>
1022
1023 * tree.c (int_fits_type_p): A narrower type always fits in a
1024 wider one, except for negative values into unsigned types.
1025
1026 2005-01-01 Roger Sayle <roger@eyesopen.com>
1027
1028 * tree.c (int_fits_type_p): Compare the result of force_fit_type
1029 with the original constant rather than require TREE_OVERFLOW.
1030
1031 2005-01-01 Steven Bosscher <stevenb@suse.de>
1032
1033 PR middle-end/17544
1034 * c-decl.c (finish_function): If compiling C99, annotate the
1035 compiler generated return with the current file name and line 0.
1036 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
1037 the source line is greater than 0.
1038 (remove_bb): Likewise.
1039
1040 See ChangeLog.12 for earlier changes.
This page took 0.089295 seconds and 6 git commands to generate.