]> gcc.gnu.org Git - gcc.git/blob - gcc/ChangeLog
i386.c (bdesc_2arg): Update names for mmx_ prefixes.
[gcc.git] / gcc / ChangeLog
1 2005-01-09 Richard Henderson <rth@redhat.com>
2
3 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
4 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
5 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
6 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
7 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
8 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
9 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
10 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
11 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
12 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
13 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
14 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
15 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
16 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
17 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
18 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
19 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
20 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
21 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
22 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
23 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
24 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
25 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
26 mmx_ prefix.
27 (mmx_clrdi, pavgusb): Remove.
28 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
29 with leading sse_ prefix.
30 * config/i386/sse.md: Receive them.
31 * config/i386/mmx.md: New file.
32 (MMXMODE12, MMXMODE24, mmxvecsize): New.
33 (subrv2sf3): Turn into expander for normal subtraction.
34 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
35 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
36 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
37 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
38 commutative; use ix86_binary_operator_ok.
39 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
40 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
41 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
42 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
43 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
44 Macroize from existing patterns; use ix86_binary_operator_ok.
45 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
46 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
47 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
48 with vec_select+vec_concat.
49 (mmx_pshufw, mmx_pshufw_1): Likewise.
50 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
51 (mmx_uavgv4hi3): Model correctly.
52 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
53 _mm_xor_si64): Remove casts.
54
55 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
56
57 PR tree-optimization/19224
58 * tree-scalar-evolution.c (get_instantiated_value,
59 set_instantiated_value): New functions.
60 (instantiate_parameters_1): Cache the results.
61 (instantiate_parameters, resolve_mixers): Initialize and free
62 the cache.
63
64 2005-01-08 David Edelsohn <edelsohn@gnu.org>
65
66 * config/i386/i386.md (addhi_4): Correct reference in comment.
67 (addqi_4): Same.
68
69 2005-01-08 Richard Henderson <rth@redhat.com>
70
71 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
72 (_mm_cvtsi128_si64x): Likewise.
73 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
74 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
75 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
76 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
77 * config/i386/i386-protos.h: Update.
78 * config/i386/i386.c (print_operand): Add 'H'.
79 (ix86_fixup_binary_operands): Split out from ...
80 (ix86_expand_binary_operator): ... here.
81 (ix86_fixup_binary_operands_no_copy): New.
82 (ix86_expand_fp_absneg_operator): Handle vector mode results.
83 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
84 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
85 (safe_vector_operand): Use CONST0_RTX.
86 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
87 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
88 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
89 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
90 special cases.
91 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
92 (IX86_BUILTIN_CMPNGESS): New.
93 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
94 (attr type): Add sselog1.
95 (attr unit, attr memory): Handle it.
96 (movti, movti_internal, movti_rex64): Move near other integer moves.
97 (movtf, movtf_internal): Move near other fp moves.
98 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
99 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
100 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
101 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
102 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
103 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
104 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
105 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
106 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
107 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
108 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
109 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
110 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
111 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
112 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
113 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
114 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
115 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
116 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
117 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
118 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
119 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
120 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
121 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
122 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
123 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
124 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
125 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
126 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
127 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
128 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
129 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
130 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
131 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
132 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
133 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
134 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
135 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
136 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
137 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
138 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
139 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
140 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
141 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
142 same time.
143 (SSEPUSH, push<SSEPUSH>): Remove.
144 (MMXPUSH, push<MMXPUSH>): Remove.
145 (sse_movaps, sse_movaps_1, sse_movups): Remove.
146 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
147 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
148 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
149 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
150 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
151 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
152 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
153 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
154 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
155 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
156 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
157 (ppro_sse_log_V4SF): Handle sselog1.
158 * config/i386/predicates.md (const_0_to_1_operand): New.
159 (const_0_to_255_mul_8_operand): New.
160 (const_1_to_31_operand): Rename from const_int_1_31_operand.
161 (const_2_to_3_operand, const_4_to_7_operand): New.
162 * config/i386/sse.md: New file.
163 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
164 (sse_movups): Rename from sse_movups_1.
165 (sse_loadlss): Rename from sse_loadss_1.
166 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
167 from the name.
168 (negv4sf2): Use ix86_expand_fp_absneg_operator.
169 (absv4sf2, negv2df, absv2df): New.
170 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
171 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
172 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
173 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
174 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
175 (sse3_addsubv4sf3): Model correctly.
176 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
177 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
178 (sse_movhlps): Model with vec_select+vec_concat.
179 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
180 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
181 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
182 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
183 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
184 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
185 sse2_pshufhw_1): Likewise.
186 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
187 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
188 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
189 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
190 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
191 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
192 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
193 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
194 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
195 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
196 sminv8hi3): Mark commutative
197 operands. Use ix86_binary_operator_ok.
198 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
199 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
200 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
201 sse2_punpcklqdq): Allow operand2 in memory.
202 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
203 sse2_movsd): Add memory alternatives.
204 (sse_storelps): Turn expander into an insn; split after reload.
205 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
206 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
207
208 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
209
210 * configure.ac (DWARF-2 debug_line): Use objdump.
211 * configure: Regenerate.
212
213 2005-01-08 Jeff Law <law@redhat.com>
214 Diego Novillo <dnovillo@redhat.com>
215
216 PR tree-optimization/18241
217 * tree-nrv.c (tree_nrv): Ignore volatile return values.
218 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
219 statements with volatile operands.
220 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
221 operands after marking a statement with has_volatile_ops.
222
223 2005-01-08 Roger Sayle <roger@eyesopen.com>
224
225 * tree.c (int_fits_type_p): Always honor integer constant
226 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
227
228 2005-01-08 Roger Sayle <roger@eyesopen.com>
229
230 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
231 the latent bug in rtl_delete_block has been resolved.
232
233 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
234
235 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
236 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
237
238 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
239
240 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
241 library as @multilib_dir@/$(SHLIB_SONAME).
242
243 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
244
245 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
246 * configure: Regenerate.
247
248 2005-01-07 Jakub Jelinek <jakub@redhat.com>
249
250 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
251 TYPE_MODE.
252
253 2005-01-07 David Edelsohn <edelsohn@gnu.org>
254
255 PR target/13674
256 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
257 Convert non-word aligned offset address using ld/std into
258 indirect address.
259
260 2005-01-07 Richard Henderson <rth@redhat.com>
261
262 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
263
264 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
265
266 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
267
268 2005-01-07 Jakub Jelinek <jakub@redhat.com>
269
270 PR tree-optimization/19283
271 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
272 is not shorter than the original type.
273
274 PR rtl-optimization/19012
275 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
276
277 PR rtl-optimization/18861
278 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
279 (FULL_STATE): Define.
280 (mark_tablejump_edge): New function.
281 (purge_dead_tablejump_edges): New function.
282 (find_bb_boundaries): Use it.
283
284 PR tree-optimization/18828
285 * builtins.c (expand_builtin_next_arg): Remove argument and all
286 the argument checking.
287 (expand_builtin): Adjust caller.
288 (expand_builtin_va_start): Likewise. Remove error for too many
289 arguments.
290 (fold_builtin_next_arg): Issue error for too many arguments.
291 After checking arguments, replace them with magic arguments that
292 prevent further checking of the args.
293
294 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
295
296 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
297
298 2005-01-06 Roger Sayle <roger@eyesopen.com>
299
300 PR target/6077
301 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
302 and %{.S:...} (and their negative variants) to test whether the
303 input file is assembler or pre-processed-assembler independent of
304 the actual filename extension.
305
306 2005-01-06 Roger Sayle <roger@eyesopen.com>
307
308 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
309 of sign or zero extended values.
310
311 2005-01-06 Geoffrey Keating <geoffk@apple.com>
312
313 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
314 for IBM long double format correctly.
315
316 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
317
318 Fix PR tree-optimization/18792
319
320 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
321 to first_loop_depth, and use loop depth instead of loop number.
322 (build_classic_dir_vector): Ditto.
323 (compute_data_dependences_for_loop): Use depth, not loop number.
324 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
325 number. Pass in loops, instead of loop numbers.
326 (gather_interchange_stats): Ditto.
327 (linear_transform_loops): Ditto.
328
329 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
330
331 PR rtl-opt/13299
332 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
333 biv_fits_mode_p, extension_within_bounds_p): New functions.
334 (check_ext_dependent_givs): Use them.
335
336 2005-01-06 Roger Sayle <roger@eyesopen.com>
337
338 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
339 more than one barrier, in which case we should delete them all.
340
341 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
342
343 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
344
345 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
346
347 Fix PR middle-end/19286
348 Fix PR debug/19267
349 * dwarf2out.c (gen_subprogram_die): If we've already tried to
350 output this subprogram, simply ignore this attempt to do it again.
351 (add_abstract_origin_attribute): Don't abort trying to add the abstract
352 origin attribute if it's not possible.
353 (gen_block_die): Don't ignore subblocks of "unused" blocks.
354 (decls_for_scope): Ditto.
355 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
356 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
357
358 2005-01-05 Richard Henderson <rth@redhat.com>
359
360 PR target/11327
361 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
362 (bdesc_2arg): Use it.
363 (ix86_expand_binop_builtin): Force operands into registers
364 when optimizing.
365 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
366 ix86_expand_sse_compare, ix86_expand_sse_comi,
367 ix86_expand_builtin): Likewise.
368
369 2005-01-05 Richard Henderson <rth@redhat.com>
370
371 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
372 and VEC_SELECT.
373 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
374 (mulv4hi3): Set itanium_class mmmul.
375 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
376 (fpack): Rename from fpack_sfsf.
377
378 2005-01-05 Richard Henderson <rth@redhat.com>
379
380 PR rtl-opt/10692
381 * reload1.c (do_input_reload): Restrict the optimization deleteing
382 a previous output reload to RELOAD_FOR_INPUT.
383
384 2005-01-05 Steven Bosscher <stevenb@suse.de>
385
386 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
387 comment that we fall through after case.
388 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
389 (distribute_links): Likewise.
390 * cse.c (cse_insn): Likewise.
391 * cselib.c (cselib_invalidate_mem): Likewise.
392 * df.c: Update comments at the top of the file.
393 (read_modify_subreg_p): Update comments here too.
394 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
395 * flow.c (mark_set_1): Likewise.
396 (mark_used_regs): Likewise.
397 * gcse.c (mems_conflict_for_gcse_p): Likewise.
398 (canon_list_insert): Likewise.
399 (mark_set): Likewise.
400 (try_replace_reg): Likewise.
401 (store_killed_in_insn): Likewise.
402 * loop.c (count_one_set): Likewise.
403 (basic_induction_var): Likewise.
404 * postreload-gcse.c (find_mem_conflicts): Likewise.
405 * postreload.c (reload_combine_note_store): Likewise.
406 (move2add_note_store): Likewise.
407 * reload.c (find_equiv_reg): Likewise.
408 (mark_referenced_resources): Likewise.
409 * rtlanal.c (set_noop_p): Likewise.
410 (note_stores): Likewise.
411 (note_uses): Likewise.
412 * sched-deps.c (sched_analyze_1): Likewise.
413 * sched-rgn.c (check_live_1): Likewise.
414 (update_live_1): Likewise.
415 * config/i860/i860.c: Likewise.
416
417 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
418 an lvalue.
419 (ZERO_EXTRACT): Mention that this one can be an lvalue.
420
421 * doc/rtl.texi: Update documentation for bit-fields and SET.
422
423 2005-01-05 Roger Sayle <roger@eyesopen.com>
424
425 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
426
427 2005-01-05 Jan Hubicka <jh@suse.cz>
428 Richard Henderson <rth@redhat.com>
429
430 PR target/18910
431 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
432 with an offset.
433
434 2005-01-05 Richard Henderson <rth@redhat.com>
435
436 PR target/12902
437 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
438 (sse_shufps): Change operand 3 to const_int_operand.
439 (sse2_storelps): Fix typo in template.
440 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
441 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
442 (ix86_expand_builtin): Likewise.
443
444 2005-01-05 Julian Brown <julian@codesourcery.com>
445
446 * config/arm/arm.c (arm_return_in_memory): Treat complex types
447 as aggregates for AAPCS ABIs.
448
449 2005-01-05 Stan Shebs <shebs@apple.com>
450
451 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
452 Mach-O getters if ppc64.
453 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
454 size argument.
455 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
456 signed four-byte field for global code case.
457 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
458
459 2005-01-05 Roger Sayle <roger@eyesopen.com>
460
461 PR middle-end/19100
462 * c-common.c: Include real.h.
463 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
464 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
465 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
466 * Makefile.in (c-common.o): Update dependencies.
467
468 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
469
470 * c-parse.in (asm_string): Add trailing semicolon.
471
472 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
473
474 * c-parse.in (asm_string): New. Don't allow wide strings in
475 'asm'.
476 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
477 asm_string instead of STRING.
478
479 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
480
481 * c-typeck.c (constructor_no_implicit): Remove.
482 (set_designator, process_init_element): Don't check
483 constructor_no_implicit.
484
485 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
486 Kaz Kojima <kkojima@gcc.gnu.org>
487
488 PR target/16482
489 * lcm.c (create_pre_exit): New.
490 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
491 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
492
493 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
494
495 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
496 -msoft-float in target_float_switch.
497 * arm.c (arm_override_options): Fix processing of target_float_switch.
498
499 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
500
501 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
502 preferencing.
503
504 2004-01-05 Uros Bizjak <uros@kss-loka.si>
505
506 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
507 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
508
509 2005-01-04 Roger Sayle <roger@eyesopen.com>
510
511 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
512 if deleting the then-block allows the test-block to fallthru to the
513 else-block.
514
515 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
516
517 PR c/19152
518 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
519 after the full declaration if the are in two different TUs.
520
521 2005-01-04 Richard Henderson <rth@redhat.com>
522
523 PR tree-opt/19158
524 * tree-sra.c (generate_one_element_init): Just
525 call gimplify_and_add.
526 (generate_element_init): Record the
527 new referenced variables and mark them for renaming
528 and split out to ...
529 (generate_element_init_1): This.
530 (scalarize_init): Don't call push_gimplify_context/
531 pop_gimplify_context.
532
533 2005-01-04 Geoffrey Keating <geoffk@apple.com>
534
535 * toplev.c (get_src_pwd): Handle failure of getpwd().
536
537 2005-01-04 Roger Sayle <roger@eyesopen.com>
538
539 * fold-const.c (fold_single_bit_test): Delete unreachable handling
540 of TRUTH_NOT_EXPR.
541 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
542 all the cases handled by it are inverted by invert_truthvalue.
543
544 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
545
546 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
547 (save_register_windowsi): Likewise.
548
549 2005-01-04 Richard Henderson <rth@redhat.com>
550
551 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
552
553 2005-01-04 Uros Bizjak <uros@kss-loka.si>
554
555 PR middle-end/17767
556 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
557 any simplifications of vector mode comparison operators.
558 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
559
560 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
561 Devang Patel <dpatel@apple.com>
562
563 PR tree-optimization/18308
564 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
565 the operands before creating a new expression.
566 * dojump.c (do_jump): Make drop_through_label available
567 for all cases. Add expansion of COND_EXPR.
568
569 2005-01-04 Ira Rosen <irar@il.ibm.com>
570
571 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
572 INTEGER_CST instead of TREE_CONSTANT.
573 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
574 instead of TREE_CONSTANT.
575 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
576 instead of TREE_CONSTANT.
577
578 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
579
580 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
581
582 2005-01-04 Uros Bizjak <uros@kss-loka.si>
583
584 PR target/19240
585 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
586 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
587
588 2005-01-03 Richard Henderson <rth@redhat.com>
589
590 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
591 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
592 fold_convert_const_real_from_real): Split out from ...
593 (fold_convert_const): ... here.
594
595 2005-01-03 Richard Henderson <rth@redhat.com>
596
597 PR target/19235
598 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
599 (mov<MMXMODEI>_internal): Likewise.
600 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
601 Add V2SF case; use it for SSE1; don't use TI.
602 (movdf_integer): Likewise.
603 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
604
605 2005-01-03 Ira Rosen <irar@il.ibm.com>
606
607 * tree-vectorizer.c (vect_strip_conversions): New function.
608 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
609 check for binary class.
610
611 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
612
613 Fix PR debug/17924
614 Fix PR debug/19191
615 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
616 is a decl.
617 * gimple-low.c (mark_blocks_with_used_vars): New function.
618 (mark_blocks_with_used_subblocks): Ditto.
619 (mark_used_blocks): Ditto.
620 (pass_mark_used_blocks): New pass.
621 * tree-inline.c: Include debug.h.
622 (expand_call_inline): Call outlining_inline_function here.
623 * tree-optimize.c (init_tree_optimization_passes): Add
624 pass_mark_used_blocks.
625 * tree-pass.h (pass_mark_used_blocks): New.
626 * Makefile.in (tree-inline.o): Add debug.h dependency.
627
628 2005-01-03 Geoffrey Keating <geoffk@apple.com>
629
630 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
631 VAR_DECLs to have weak_import attribute.
632
633 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
634 'struct ucontext'.
635
636 2004-01-03 Steven Bosscher <stevenb@suse.de>
637
638 * passes.c (rest_of_compilation): Don't run regmove if only
639 flag_expensive_optimizations, require flag_regmove instead.
640
641 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
642
643 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
644 register "sparc".
645 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
646 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
647 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
648 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
649
650 2005-01-03 Richard Henderson <rth@redhat.com>
651
652 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
653 (ia64_const_ok_for_letter_p): New.
654 (ia64_const_double_ok_for_letter_p): New.
655 (ia64_extra_constraint): New.
656 (ia64_expand_vecint_compare): New.
657 (ia64_expand_vcondu_v2si): New.
658 (ia64_expand_vecint_cmov): New.
659 (ia64_expand_vecint_minmax): New.
660 (ia64_print_operand): Add 'v'.
661 (ia64_preferred_reload_class): New.
662 (ia64_vector_mode_supported_p): New.
663 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
664 (PREFERRED_RELOAD_CLASS): Move to function.
665 (CONST_OK_FOR_LETTER_P): Move to function.
666 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
667 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
668 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
669 (EXTRA_CONSTRAINT): Move to function.
670 * config/ia64/ia64.md: Include vect.md.
671 (itanium_class): Add mmalua.
672 (type): Handle it.
673 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
674 (1b_mmalua): New.
675 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
676 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
677 CONST0_RTX.
678 (const_int_2bit_operand): New.
679 (fr_reg_or_0_operand): New.
680 * config/ia64/ia64-modes.def: Add vector modes.
681 * config/ia64/ia64-protos.h: Update.
682 * config/ia64/vect.md: New file.
683
684 2005-01-03 Richard Henderson <rth@redhat.com>
685
686 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
687
688 2005-01-03 Uros Bizjak <uros@kss-loka.si>
689
690 PR target/19236
691 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
692 to SFmode.
693 (log1pdf2): Change mode of operands[0,1] to DFmode.
694
695 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
696
697 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
698 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
699 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
700 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
701 SPARC_RELAXED_ORDERING.
702
703 2005-01-03 Richard Henderson <rth@redhat.com>
704 Uros Bizjak <uros@kss-loka.si>
705
706 PR target/14631
707 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
708 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
709 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
710 handling.
711 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
712 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
713 const_pow2_1_to_128_operand): New predicates.
714
715 2005-01-02 Greg McGary <greg@mcgary.org>
716
717 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
718 anonymous decls.
719 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
720 object. [ptr_type]: Remove unused variable.
721 Remember nearest addressable array-element or record-component
722 when checking bitfield components. Tolerate empty BIND_EXPR.
723
724 2005-01-01 Richard Henderson <rth@redhat.com>
725
726 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
727 that don't narrow the value. Fail for other conversions.
728
729 2005-01-01 Richard Henderson <rth@redhat.com>
730
731 PR c/19031
732 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
733 * c-lang.c (finish_file): Don't do it here.
734 * objc/objc-act.c (objc_finish_file): Likewise.
735
736 * cgraph.c (decl_assembler_name_equal): New.
737 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
738 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
739 * cgraph.h (struct cgraph_varpool_node): Add next.
740 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
741 * varasm.c (assemble_alias): Mark the target as needed.
742
743 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
744
745 PR middle-end/19221
746 * function.c (get_arg_pointer_save_area): Use entry_of_function
747 instead of get_insns.
748
749 2005-01-01 Roger Sayle <roger@eyesopen.com>
750 Andrew Pinski <pinskia@physics.uc.edu>
751 James E. Wilson <wilson@specifixinc.com>
752
753 PR rtl-optimization/12092
754 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
755 is zero.
756
757 2005-01-01 Roger Sayle <roger@eyesopen.com>
758 Olivier Hainque <hainque@act-europe.fr>
759
760 * tree.c (int_fits_type_p): A narrower type always fits in a
761 wider one, except for negative values into unsigned types.
762
763 2005-01-01 Roger Sayle <roger@eyesopen.com>
764
765 * tree.c (int_fits_type_p): Compare the result of force_fit_type
766 with the original constant rather than require TREE_OVERFLOW.
767
768 2005-01-01 Steven Bosscher <stevenb@suse.de>
769
770 PR middle-end/17544
771 * c-decl.c (finish_function): If compiling C99, annotate the
772 compiler generated return with the current file name and line 0.
773 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
774 the source line is greater than 0.
775 (remove_bb): Likewise.
776
777 See ChangeLog.12 for earlier changes.
This page took 0.075307 seconds and 6 git commands to generate.