]> gcc.gnu.org Git - gcc.git/blob - gcc/ChangeLog
fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding X/C1 cmpop C2.
[gcc.git] / gcc / ChangeLog
1 2005-01-09 Falk Hueffner <falk@debian.org>
2
3 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
4 X/C1 cmpop C2.
5
6 2005-01-09 David Edelsohn <edelsohn@gnu.org>
7
8 PR target/18720
9 * collect2.c (main): Set aixrtl_flag for -brtl option.
10 (resolve_lib_name): Search for .so file extension before .a
11 if aixrtl_flag set.
12
13 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
14
15 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
16 (vect_analyze_data_refs_alignment): Add dump prints.
17
18 2005-01-09 Ira Rosen <irar@il.ibm.com>
19
20 * tree-vectorizer.c (vect_analyze_offset_expr): Use
21 expr_invariant_in_loop_p.
22 Initialize outputs first thing in the function.
23 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
24 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
25 (vect_analyze_pointer_ref_access): Check that the initial condition of
26 the access function is loop invariant.
27
28 2005-01-09 Richard Henderson <rth@redhat.com>
29
30 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
31 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
32 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
33 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
34 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
35 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
36 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
37 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
38 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
39 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
40 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
41 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
42 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
43 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
44 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
45 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
46 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
47 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
48 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
49 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
50 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
51 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
52 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
53 mmx_ prefix.
54 (mmx_clrdi, pavgusb): Remove.
55 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
56 with leading sse_ prefix.
57 * config/i386/sse.md: Receive them.
58 * config/i386/mmx.md: New file.
59 (MMXMODE12, MMXMODE24, mmxvecsize): New.
60 (subrv2sf3): Turn into expander for normal subtraction.
61 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
62 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
63 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
64 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
65 commutative; use ix86_binary_operator_ok.
66 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
67 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
68 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
69 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
70 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
71 Macroize from existing patterns; use ix86_binary_operator_ok.
72 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
73 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
74 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
75 with vec_select+vec_concat.
76 (mmx_pshufw, mmx_pshufw_1): Likewise.
77 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
78 (mmx_uavgv4hi3): Model correctly.
79 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
80 _mm_xor_si64): Remove casts.
81
82 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
83
84 PR tree-optimization/19224
85 * tree-scalar-evolution.c (get_instantiated_value,
86 set_instantiated_value): New functions.
87 (instantiate_parameters_1): Cache the results.
88 (instantiate_parameters, resolve_mixers): Initialize and free
89 the cache.
90
91 2005-01-08 David Edelsohn <edelsohn@gnu.org>
92
93 * config/i386/i386.md (addhi_4): Correct reference in comment.
94 (addqi_4): Same.
95
96 2005-01-08 Richard Henderson <rth@redhat.com>
97
98 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
99 (_mm_cvtsi128_si64x): Likewise.
100 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
101 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
102 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
103 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
104 * config/i386/i386-protos.h: Update.
105 * config/i386/i386.c (print_operand): Add 'H'.
106 (ix86_fixup_binary_operands): Split out from ...
107 (ix86_expand_binary_operator): ... here.
108 (ix86_fixup_binary_operands_no_copy): New.
109 (ix86_expand_fp_absneg_operator): Handle vector mode results.
110 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
111 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
112 (safe_vector_operand): Use CONST0_RTX.
113 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
114 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
115 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
116 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
117 special cases.
118 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
119 (IX86_BUILTIN_CMPNGESS): New.
120 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
121 (attr type): Add sselog1.
122 (attr unit, attr memory): Handle it.
123 (movti, movti_internal, movti_rex64): Move near other integer moves.
124 (movtf, movtf_internal): Move near other fp moves.
125 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
126 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
127 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
128 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
129 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
130 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
131 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
132 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
133 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
134 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
135 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
136 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
137 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
138 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
139 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
140 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
141 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
142 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
143 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
144 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
145 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
146 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
147 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
148 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
149 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
150 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
151 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
152 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
153 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
154 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
155 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
156 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
157 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
158 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
159 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
160 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
161 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
162 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
163 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
164 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
165 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
166 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
167 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
168 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
169 same time.
170 (SSEPUSH, push<SSEPUSH>): Remove.
171 (MMXPUSH, push<MMXPUSH>): Remove.
172 (sse_movaps, sse_movaps_1, sse_movups): Remove.
173 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
174 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
175 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
176 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
177 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
178 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
179 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
180 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
181 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
182 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
183 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
184 (ppro_sse_log_V4SF): Handle sselog1.
185 * config/i386/predicates.md (const_0_to_1_operand): New.
186 (const_0_to_255_mul_8_operand): New.
187 (const_1_to_31_operand): Rename from const_int_1_31_operand.
188 (const_2_to_3_operand, const_4_to_7_operand): New.
189 * config/i386/sse.md: New file.
190 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
191 (sse_movups): Rename from sse_movups_1.
192 (sse_loadlss): Rename from sse_loadss_1.
193 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
194 from the name.
195 (negv4sf2): Use ix86_expand_fp_absneg_operator.
196 (absv4sf2, negv2df, absv2df): New.
197 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
198 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
199 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
200 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
201 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
202 (sse3_addsubv4sf3): Model correctly.
203 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
204 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
205 (sse_movhlps): Model with vec_select+vec_concat.
206 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
207 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
208 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
209 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
210 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
211 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
212 sse2_pshufhw_1): Likewise.
213 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
214 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
215 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
216 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
217 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
218 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
219 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
220 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
221 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
222 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
223 sminv8hi3): Mark commutative
224 operands. Use ix86_binary_operator_ok.
225 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
226 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
227 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
228 sse2_punpcklqdq): Allow operand2 in memory.
229 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
230 sse2_movsd): Add memory alternatives.
231 (sse_storelps): Turn expander into an insn; split after reload.
232 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
233 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
234
235 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
236
237 * configure.ac (DWARF-2 debug_line): Use objdump.
238 * configure: Regenerate.
239
240 2005-01-08 Jeff Law <law@redhat.com>
241 Diego Novillo <dnovillo@redhat.com>
242
243 PR tree-optimization/18241
244 * tree-nrv.c (tree_nrv): Ignore volatile return values.
245 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
246 statements with volatile operands.
247 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
248 operands after marking a statement with has_volatile_ops.
249
250 2005-01-08 Roger Sayle <roger@eyesopen.com>
251
252 * tree.c (int_fits_type_p): Always honor integer constant
253 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
254
255 2005-01-08 Roger Sayle <roger@eyesopen.com>
256
257 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
258 the latent bug in rtl_delete_block has been resolved.
259
260 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
261
262 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
263 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
264
265 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
266
267 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
268 library as @multilib_dir@/$(SHLIB_SONAME).
269
270 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
271
272 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
273 * configure: Regenerate.
274
275 2005-01-07 Jakub Jelinek <jakub@redhat.com>
276
277 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
278 TYPE_MODE.
279
280 2005-01-07 David Edelsohn <edelsohn@gnu.org>
281
282 PR target/13674
283 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
284 Convert non-word aligned offset address using ld/std into
285 indirect address.
286
287 2005-01-07 Richard Henderson <rth@redhat.com>
288
289 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
290
291 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
292
293 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
294
295 2005-01-07 Jakub Jelinek <jakub@redhat.com>
296
297 PR tree-optimization/19283
298 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
299 is not shorter than the original type.
300
301 PR rtl-optimization/19012
302 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
303
304 PR rtl-optimization/18861
305 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
306 (FULL_STATE): Define.
307 (mark_tablejump_edge): New function.
308 (purge_dead_tablejump_edges): New function.
309 (find_bb_boundaries): Use it.
310
311 PR tree-optimization/18828
312 * builtins.c (expand_builtin_next_arg): Remove argument and all
313 the argument checking.
314 (expand_builtin): Adjust caller.
315 (expand_builtin_va_start): Likewise. Remove error for too many
316 arguments.
317 (fold_builtin_next_arg): Issue error for too many arguments.
318 After checking arguments, replace them with magic arguments that
319 prevent further checking of the args.
320
321 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
322
323 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
324
325 2005-01-06 Roger Sayle <roger@eyesopen.com>
326
327 PR target/6077
328 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
329 and %{.S:...} (and their negative variants) to test whether the
330 input file is assembler or pre-processed-assembler independent of
331 the actual filename extension.
332
333 2005-01-06 Roger Sayle <roger@eyesopen.com>
334
335 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
336 of sign or zero extended values.
337
338 2005-01-06 Geoffrey Keating <geoffk@apple.com>
339
340 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
341 for IBM long double format correctly.
342
343 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
344
345 Fix PR tree-optimization/18792
346
347 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
348 to first_loop_depth, and use loop depth instead of loop number.
349 (build_classic_dir_vector): Ditto.
350 (compute_data_dependences_for_loop): Use depth, not loop number.
351 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
352 number. Pass in loops, instead of loop numbers.
353 (gather_interchange_stats): Ditto.
354 (linear_transform_loops): Ditto.
355
356 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
357
358 PR rtl-opt/13299
359 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
360 biv_fits_mode_p, extension_within_bounds_p): New functions.
361 (check_ext_dependent_givs): Use them.
362
363 2005-01-06 Roger Sayle <roger@eyesopen.com>
364
365 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
366 more than one barrier, in which case we should delete them all.
367
368 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
369
370 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
371
372 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
373
374 Fix PR middle-end/19286
375 Fix PR debug/19267
376 * dwarf2out.c (gen_subprogram_die): If we've already tried to
377 output this subprogram, simply ignore this attempt to do it again.
378 (add_abstract_origin_attribute): Don't abort trying to add the abstract
379 origin attribute if it's not possible.
380 (gen_block_die): Don't ignore subblocks of "unused" blocks.
381 (decls_for_scope): Ditto.
382 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
383 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
384
385 2005-01-05 Richard Henderson <rth@redhat.com>
386
387 PR target/11327
388 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
389 (bdesc_2arg): Use it.
390 (ix86_expand_binop_builtin): Force operands into registers
391 when optimizing.
392 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
393 ix86_expand_sse_compare, ix86_expand_sse_comi,
394 ix86_expand_builtin): Likewise.
395
396 2005-01-05 Richard Henderson <rth@redhat.com>
397
398 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
399 and VEC_SELECT.
400 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
401 (mulv4hi3): Set itanium_class mmmul.
402 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
403 (fpack): Rename from fpack_sfsf.
404
405 2005-01-05 Richard Henderson <rth@redhat.com>
406
407 PR rtl-opt/10692
408 * reload1.c (do_input_reload): Restrict the optimization deleteing
409 a previous output reload to RELOAD_FOR_INPUT.
410
411 2005-01-05 Steven Bosscher <stevenb@suse.de>
412
413 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
414 comment that we fall through after case.
415 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
416 (distribute_links): Likewise.
417 * cse.c (cse_insn): Likewise.
418 * cselib.c (cselib_invalidate_mem): Likewise.
419 * df.c: Update comments at the top of the file.
420 (read_modify_subreg_p): Update comments here too.
421 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
422 * flow.c (mark_set_1): Likewise.
423 (mark_used_regs): Likewise.
424 * gcse.c (mems_conflict_for_gcse_p): Likewise.
425 (canon_list_insert): Likewise.
426 (mark_set): Likewise.
427 (try_replace_reg): Likewise.
428 (store_killed_in_insn): Likewise.
429 * loop.c (count_one_set): Likewise.
430 (basic_induction_var): Likewise.
431 * postreload-gcse.c (find_mem_conflicts): Likewise.
432 * postreload.c (reload_combine_note_store): Likewise.
433 (move2add_note_store): Likewise.
434 * reload.c (find_equiv_reg): Likewise.
435 (mark_referenced_resources): Likewise.
436 * rtlanal.c (set_noop_p): Likewise.
437 (note_stores): Likewise.
438 (note_uses): Likewise.
439 * sched-deps.c (sched_analyze_1): Likewise.
440 * sched-rgn.c (check_live_1): Likewise.
441 (update_live_1): Likewise.
442 * config/i860/i860.c: Likewise.
443
444 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
445 an lvalue.
446 (ZERO_EXTRACT): Mention that this one can be an lvalue.
447
448 * doc/rtl.texi: Update documentation for bit-fields and SET.
449
450 2005-01-05 Roger Sayle <roger@eyesopen.com>
451
452 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
453
454 2005-01-05 Jan Hubicka <jh@suse.cz>
455 Richard Henderson <rth@redhat.com>
456
457 PR target/18910
458 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
459 with an offset.
460
461 2005-01-05 Richard Henderson <rth@redhat.com>
462
463 PR target/12902
464 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
465 (sse_shufps): Change operand 3 to const_int_operand.
466 (sse2_storelps): Fix typo in template.
467 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
468 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
469 (ix86_expand_builtin): Likewise.
470
471 2005-01-05 Julian Brown <julian@codesourcery.com>
472
473 * config/arm/arm.c (arm_return_in_memory): Treat complex types
474 as aggregates for AAPCS ABIs.
475
476 2005-01-05 Stan Shebs <shebs@apple.com>
477
478 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
479 Mach-O getters if ppc64.
480 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
481 size argument.
482 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
483 signed four-byte field for global code case.
484 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
485
486 2005-01-05 Roger Sayle <roger@eyesopen.com>
487
488 PR middle-end/19100
489 * c-common.c: Include real.h.
490 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
491 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
492 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
493 * Makefile.in (c-common.o): Update dependencies.
494
495 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
496
497 * c-parse.in (asm_string): Add trailing semicolon.
498
499 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
500
501 * c-parse.in (asm_string): New. Don't allow wide strings in
502 'asm'.
503 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
504 asm_string instead of STRING.
505
506 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
507
508 * c-typeck.c (constructor_no_implicit): Remove.
509 (set_designator, process_init_element): Don't check
510 constructor_no_implicit.
511
512 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
513 Kaz Kojima <kkojima@gcc.gnu.org>
514
515 PR target/16482
516 * lcm.c (create_pre_exit): New.
517 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
518 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
519
520 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
521
522 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
523 -msoft-float in target_float_switch.
524 * arm.c (arm_override_options): Fix processing of target_float_switch.
525
526 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
527
528 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
529 preferencing.
530
531 2004-01-05 Uros Bizjak <uros@kss-loka.si>
532
533 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
534 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
535
536 2005-01-04 Roger Sayle <roger@eyesopen.com>
537
538 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
539 if deleting the then-block allows the test-block to fallthru to the
540 else-block.
541
542 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
543
544 PR c/19152
545 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
546 after the full declaration if the are in two different TUs.
547
548 2005-01-04 Richard Henderson <rth@redhat.com>
549
550 PR tree-opt/19158
551 * tree-sra.c (generate_one_element_init): Just
552 call gimplify_and_add.
553 (generate_element_init): Record the
554 new referenced variables and mark them for renaming
555 and split out to ...
556 (generate_element_init_1): This.
557 (scalarize_init): Don't call push_gimplify_context/
558 pop_gimplify_context.
559
560 2005-01-04 Geoffrey Keating <geoffk@apple.com>
561
562 * toplev.c (get_src_pwd): Handle failure of getpwd().
563
564 2005-01-04 Roger Sayle <roger@eyesopen.com>
565
566 * fold-const.c (fold_single_bit_test): Delete unreachable handling
567 of TRUTH_NOT_EXPR.
568 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
569 all the cases handled by it are inverted by invert_truthvalue.
570
571 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
572
573 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
574 (save_register_windowsi): Likewise.
575
576 2005-01-04 Richard Henderson <rth@redhat.com>
577
578 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
579
580 2005-01-04 Uros Bizjak <uros@kss-loka.si>
581
582 PR middle-end/17767
583 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
584 any simplifications of vector mode comparison operators.
585 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
586
587 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
588 Devang Patel <dpatel@apple.com>
589
590 PR tree-optimization/18308
591 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
592 the operands before creating a new expression.
593 * dojump.c (do_jump): Make drop_through_label available
594 for all cases. Add expansion of COND_EXPR.
595
596 2005-01-04 Ira Rosen <irar@il.ibm.com>
597
598 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
599 INTEGER_CST instead of TREE_CONSTANT.
600 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
601 instead of TREE_CONSTANT.
602 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
603 instead of TREE_CONSTANT.
604
605 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
606
607 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
608
609 2005-01-04 Uros Bizjak <uros@kss-loka.si>
610
611 PR target/19240
612 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
613 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
614
615 2005-01-03 Richard Henderson <rth@redhat.com>
616
617 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
618 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
619 fold_convert_const_real_from_real): Split out from ...
620 (fold_convert_const): ... here.
621
622 2005-01-03 Richard Henderson <rth@redhat.com>
623
624 PR target/19235
625 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
626 (mov<MMXMODEI>_internal): Likewise.
627 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
628 Add V2SF case; use it for SSE1; don't use TI.
629 (movdf_integer): Likewise.
630 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
631
632 2005-01-03 Ira Rosen <irar@il.ibm.com>
633
634 * tree-vectorizer.c (vect_strip_conversions): New function.
635 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
636 check for binary class.
637
638 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
639
640 Fix PR debug/17924
641 Fix PR debug/19191
642 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
643 is a decl.
644 * gimple-low.c (mark_blocks_with_used_vars): New function.
645 (mark_blocks_with_used_subblocks): Ditto.
646 (mark_used_blocks): Ditto.
647 (pass_mark_used_blocks): New pass.
648 * tree-inline.c: Include debug.h.
649 (expand_call_inline): Call outlining_inline_function here.
650 * tree-optimize.c (init_tree_optimization_passes): Add
651 pass_mark_used_blocks.
652 * tree-pass.h (pass_mark_used_blocks): New.
653 * Makefile.in (tree-inline.o): Add debug.h dependency.
654
655 2005-01-03 Geoffrey Keating <geoffk@apple.com>
656
657 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
658 VAR_DECLs to have weak_import attribute.
659
660 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
661 'struct ucontext'.
662
663 2004-01-03 Steven Bosscher <stevenb@suse.de>
664
665 * passes.c (rest_of_compilation): Don't run regmove if only
666 flag_expensive_optimizations, require flag_regmove instead.
667
668 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
669
670 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
671 register "sparc".
672 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
673 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
674 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
675 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
676
677 2005-01-03 Richard Henderson <rth@redhat.com>
678
679 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
680 (ia64_const_ok_for_letter_p): New.
681 (ia64_const_double_ok_for_letter_p): New.
682 (ia64_extra_constraint): New.
683 (ia64_expand_vecint_compare): New.
684 (ia64_expand_vcondu_v2si): New.
685 (ia64_expand_vecint_cmov): New.
686 (ia64_expand_vecint_minmax): New.
687 (ia64_print_operand): Add 'v'.
688 (ia64_preferred_reload_class): New.
689 (ia64_vector_mode_supported_p): New.
690 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
691 (PREFERRED_RELOAD_CLASS): Move to function.
692 (CONST_OK_FOR_LETTER_P): Move to function.
693 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
694 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
695 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
696 (EXTRA_CONSTRAINT): Move to function.
697 * config/ia64/ia64.md: Include vect.md.
698 (itanium_class): Add mmalua.
699 (type): Handle it.
700 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
701 (1b_mmalua): New.
702 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
703 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
704 CONST0_RTX.
705 (const_int_2bit_operand): New.
706 (fr_reg_or_0_operand): New.
707 * config/ia64/ia64-modes.def: Add vector modes.
708 * config/ia64/ia64-protos.h: Update.
709 * config/ia64/vect.md: New file.
710
711 2005-01-03 Richard Henderson <rth@redhat.com>
712
713 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
714
715 2005-01-03 Uros Bizjak <uros@kss-loka.si>
716
717 PR target/19236
718 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
719 to SFmode.
720 (log1pdf2): Change mode of operands[0,1] to DFmode.
721
722 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
723
724 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
725 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
726 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
727 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
728 SPARC_RELAXED_ORDERING.
729
730 2005-01-03 Richard Henderson <rth@redhat.com>
731 Uros Bizjak <uros@kss-loka.si>
732
733 PR target/14631
734 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
735 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
736 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
737 handling.
738 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
739 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
740 const_pow2_1_to_128_operand): New predicates.
741
742 2005-01-02 Greg McGary <greg@mcgary.org>
743
744 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
745 anonymous decls.
746 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
747 object. [ptr_type]: Remove unused variable.
748 Remember nearest addressable array-element or record-component
749 when checking bitfield components. Tolerate empty BIND_EXPR.
750
751 2005-01-01 Richard Henderson <rth@redhat.com>
752
753 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
754 that don't narrow the value. Fail for other conversions.
755
756 2005-01-01 Richard Henderson <rth@redhat.com>
757
758 PR c/19031
759 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
760 * c-lang.c (finish_file): Don't do it here.
761 * objc/objc-act.c (objc_finish_file): Likewise.
762
763 * cgraph.c (decl_assembler_name_equal): New.
764 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
765 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
766 * cgraph.h (struct cgraph_varpool_node): Add next.
767 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
768 * varasm.c (assemble_alias): Mark the target as needed.
769
770 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
771
772 PR middle-end/19221
773 * function.c (get_arg_pointer_save_area): Use entry_of_function
774 instead of get_insns.
775
776 2005-01-01 Roger Sayle <roger@eyesopen.com>
777 Andrew Pinski <pinskia@physics.uc.edu>
778 James E. Wilson <wilson@specifixinc.com>
779
780 PR rtl-optimization/12092
781 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
782 is zero.
783
784 2005-01-01 Roger Sayle <roger@eyesopen.com>
785 Olivier Hainque <hainque@act-europe.fr>
786
787 * tree.c (int_fits_type_p): A narrower type always fits in a
788 wider one, except for negative values into unsigned types.
789
790 2005-01-01 Roger Sayle <roger@eyesopen.com>
791
792 * tree.c (int_fits_type_p): Compare the result of force_fit_type
793 with the original constant rather than require TREE_OVERFLOW.
794
795 2005-01-01 Steven Bosscher <stevenb@suse.de>
796
797 PR middle-end/17544
798 * c-decl.c (finish_function): If compiling C99, annotate the
799 compiler generated return with the current file name and line 0.
800 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
801 the source line is greater than 0.
802 (remove_bb): Likewise.
803
804 See ChangeLog.12 for earlier changes.
This page took 0.076134 seconds and 6 git commands to generate.