1 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
3 * tree-vectorizer.c, tree.def: Fix comment typos.
5 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
7 * config/s390/s390.c (struct processor_costs): 4 new fields:
9 (s390_rtx_costs): More precise handling of divide instructions.
11 2005-01-10 David Edelsohn <edelsohn@gnu.org>
13 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
15 2005-01-10 Jan Beulich <jbeulich@novell.com>
17 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
18 names resulting from -ffunction-sections/-fdata-sections and linkonce
20 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
21 based on the section name. Rearrange the section name comparison logic
22 slightly so that each section name is compared against at most once.
24 2005-01-10 Ben Elliston <bje@au.ibm.com>
26 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
27 targets for which -fPIC reduces limitations on the GOT size.
29 2005-01-09 Falk Hueffner <falk@debian.org>
31 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
34 2005-01-09 David Edelsohn <edelsohn@gnu.org>
37 * collect2.c (main): Set aixrtl_flag for -brtl option.
38 (resolve_lib_name): Search for .so file extension before .a
41 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
43 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
44 (vect_analyze_data_refs_alignment): Add dump prints.
46 2005-01-09 Ira Rosen <irar@il.ibm.com>
48 * tree-vectorizer.c (vect_analyze_offset_expr): Use
49 expr_invariant_in_loop_p.
50 Initialize outputs first thing in the function.
51 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
52 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
53 (vect_analyze_pointer_ref_access): Check that the initial condition of
54 the access function is loop invariant.
56 2005-01-09 Richard Henderson <rth@redhat.com>
58 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
59 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
60 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
61 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
62 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
63 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
64 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
65 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
66 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
67 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
68 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
69 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
70 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
71 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
72 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
73 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
74 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
75 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
76 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
77 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
78 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
79 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
80 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
82 (mmx_clrdi, pavgusb): Remove.
83 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
84 with leading sse_ prefix.
85 * config/i386/sse.md: Receive them.
86 * config/i386/mmx.md: New file.
87 (MMXMODE12, MMXMODE24, mmxvecsize): New.
88 (subrv2sf3): Turn into expander for normal subtraction.
89 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
90 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
91 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
92 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
93 commutative; use ix86_binary_operator_ok.
94 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
95 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
96 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
97 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
98 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
99 Macroize from existing patterns; use ix86_binary_operator_ok.
100 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
101 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
102 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
103 with vec_select+vec_concat.
104 (mmx_pshufw, mmx_pshufw_1): Likewise.
105 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
106 (mmx_uavgv4hi3): Model correctly.
107 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
108 _mm_xor_si64): Remove casts.
110 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
112 PR tree-optimization/19224
113 * tree-scalar-evolution.c (get_instantiated_value,
114 set_instantiated_value): New functions.
115 (instantiate_parameters_1): Cache the results.
116 (instantiate_parameters, resolve_mixers): Initialize and free
119 2005-01-08 David Edelsohn <edelsohn@gnu.org>
121 * config/i386/i386.md (addhi_4): Correct reference in comment.
124 2005-01-08 Richard Henderson <rth@redhat.com>
126 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
127 (_mm_cvtsi128_si64x): Likewise.
128 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
129 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
130 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
131 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
132 * config/i386/i386-protos.h: Update.
133 * config/i386/i386.c (print_operand): Add 'H'.
134 (ix86_fixup_binary_operands): Split out from ...
135 (ix86_expand_binary_operator): ... here.
136 (ix86_fixup_binary_operands_no_copy): New.
137 (ix86_expand_fp_absneg_operator): Handle vector mode results.
138 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
139 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
140 (safe_vector_operand): Use CONST0_RTX.
141 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
142 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
143 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
144 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
146 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
147 (IX86_BUILTIN_CMPNGESS): New.
148 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
149 (attr type): Add sselog1.
150 (attr unit, attr memory): Handle it.
151 (movti, movti_internal, movti_rex64): Move near other integer moves.
152 (movtf, movtf_internal): Move near other fp moves.
153 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
154 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
155 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
156 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
157 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
158 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
159 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
160 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
161 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
162 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
163 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
164 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
165 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
166 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
167 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
168 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
169 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
170 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
171 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
172 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
173 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
174 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
175 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
176 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
177 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
178 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
179 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
180 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
181 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
182 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
183 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
184 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
185 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
186 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
187 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
188 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
189 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
190 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
191 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
192 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
193 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
194 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
195 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
196 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
198 (SSEPUSH, push<SSEPUSH>): Remove.
199 (MMXPUSH, push<MMXPUSH>): Remove.
200 (sse_movaps, sse_movaps_1, sse_movups): Remove.
201 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
202 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
203 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
204 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
205 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
206 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
207 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
208 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
209 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
210 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
211 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
212 (ppro_sse_log_V4SF): Handle sselog1.
213 * config/i386/predicates.md (const_0_to_1_operand): New.
214 (const_0_to_255_mul_8_operand): New.
215 (const_1_to_31_operand): Rename from const_int_1_31_operand.
216 (const_2_to_3_operand, const_4_to_7_operand): New.
217 * config/i386/sse.md: New file.
218 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
219 (sse_movups): Rename from sse_movups_1.
220 (sse_loadlss): Rename from sse_loadss_1.
221 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
223 (negv4sf2): Use ix86_expand_fp_absneg_operator.
224 (absv4sf2, negv2df, absv2df): New.
225 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
226 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
227 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
228 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
229 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
230 (sse3_addsubv4sf3): Model correctly.
231 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
232 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
233 (sse_movhlps): Model with vec_select+vec_concat.
234 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
235 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
236 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
237 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
238 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
239 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
240 sse2_pshufhw_1): Likewise.
241 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
242 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
243 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
244 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
245 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
246 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
247 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
248 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
249 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
250 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
251 sminv8hi3): Mark commutative
252 operands. Use ix86_binary_operator_ok.
253 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
254 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
255 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
256 sse2_punpcklqdq): Allow operand2 in memory.
257 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
258 sse2_movsd): Add memory alternatives.
259 (sse_storelps): Turn expander into an insn; split after reload.
260 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
261 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
263 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
265 * configure.ac (DWARF-2 debug_line): Use objdump.
266 * configure: Regenerate.
268 2005-01-08 Jeff Law <law@redhat.com>
269 Diego Novillo <dnovillo@redhat.com>
271 PR tree-optimization/18241
272 * tree-nrv.c (tree_nrv): Ignore volatile return values.
273 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
274 statements with volatile operands.
275 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
276 operands after marking a statement with has_volatile_ops.
278 2005-01-08 Roger Sayle <roger@eyesopen.com>
280 * tree.c (int_fits_type_p): Always honor integer constant
281 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
283 2005-01-08 Roger Sayle <roger@eyesopen.com>
285 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
286 the latent bug in rtl_delete_block has been resolved.
288 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
290 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
291 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
293 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
295 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
296 library as @multilib_dir@/$(SHLIB_SONAME).
298 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
300 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
301 * configure: Regenerate.
303 2005-01-07 Jakub Jelinek <jakub@redhat.com>
305 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
308 2005-01-07 David Edelsohn <edelsohn@gnu.org>
311 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
312 Convert non-word aligned offset address using ld/std into
315 2005-01-07 Richard Henderson <rth@redhat.com>
317 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
319 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
321 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
323 2005-01-07 Jakub Jelinek <jakub@redhat.com>
325 PR tree-optimization/19283
326 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
327 is not shorter than the original type.
329 PR rtl-optimization/19012
330 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
332 PR rtl-optimization/18861
333 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
334 (FULL_STATE): Define.
335 (mark_tablejump_edge): New function.
336 (purge_dead_tablejump_edges): New function.
337 (find_bb_boundaries): Use it.
339 PR tree-optimization/18828
340 * builtins.c (expand_builtin_next_arg): Remove argument and all
341 the argument checking.
342 (expand_builtin): Adjust caller.
343 (expand_builtin_va_start): Likewise. Remove error for too many
345 (fold_builtin_next_arg): Issue error for too many arguments.
346 After checking arguments, replace them with magic arguments that
347 prevent further checking of the args.
349 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
351 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
353 2005-01-06 Roger Sayle <roger@eyesopen.com>
356 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
357 and %{.S:...} (and their negative variants) to test whether the
358 input file is assembler or pre-processed-assembler independent of
359 the actual filename extension.
361 2005-01-06 Roger Sayle <roger@eyesopen.com>
363 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
364 of sign or zero extended values.
366 2005-01-06 Geoffrey Keating <geoffk@apple.com>
368 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
369 for IBM long double format correctly.
371 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
373 Fix PR tree-optimization/18792
375 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
376 to first_loop_depth, and use loop depth instead of loop number.
377 (build_classic_dir_vector): Ditto.
378 (compute_data_dependences_for_loop): Use depth, not loop number.
379 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
380 number. Pass in loops, instead of loop numbers.
381 (gather_interchange_stats): Ditto.
382 (linear_transform_loops): Ditto.
384 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
387 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
388 biv_fits_mode_p, extension_within_bounds_p): New functions.
389 (check_ext_dependent_givs): Use them.
391 2005-01-06 Roger Sayle <roger@eyesopen.com>
393 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
394 more than one barrier, in which case we should delete them all.
396 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
398 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
400 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
402 Fix PR middle-end/19286
404 * dwarf2out.c (gen_subprogram_die): If we've already tried to
405 output this subprogram, simply ignore this attempt to do it again.
406 (add_abstract_origin_attribute): Don't abort trying to add the abstract
407 origin attribute if it's not possible.
408 (gen_block_die): Don't ignore subblocks of "unused" blocks.
409 (decls_for_scope): Ditto.
410 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
411 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
413 2005-01-05 Richard Henderson <rth@redhat.com>
416 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
417 (bdesc_2arg): Use it.
418 (ix86_expand_binop_builtin): Force operands into registers
420 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
421 ix86_expand_sse_compare, ix86_expand_sse_comi,
422 ix86_expand_builtin): Likewise.
424 2005-01-05 Richard Henderson <rth@redhat.com>
426 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
428 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
429 (mulv4hi3): Set itanium_class mmmul.
430 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
431 (fpack): Rename from fpack_sfsf.
433 2005-01-05 Richard Henderson <rth@redhat.com>
436 * reload1.c (do_input_reload): Restrict the optimization deleteing
437 a previous output reload to RELOAD_FOR_INPUT.
439 2005-01-05 Steven Bosscher <stevenb@suse.de>
441 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
442 comment that we fall through after case.
443 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
444 (distribute_links): Likewise.
445 * cse.c (cse_insn): Likewise.
446 * cselib.c (cselib_invalidate_mem): Likewise.
447 * df.c: Update comments at the top of the file.
448 (read_modify_subreg_p): Update comments here too.
449 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
450 * flow.c (mark_set_1): Likewise.
451 (mark_used_regs): Likewise.
452 * gcse.c (mems_conflict_for_gcse_p): Likewise.
453 (canon_list_insert): Likewise.
454 (mark_set): Likewise.
455 (try_replace_reg): Likewise.
456 (store_killed_in_insn): Likewise.
457 * loop.c (count_one_set): Likewise.
458 (basic_induction_var): Likewise.
459 * postreload-gcse.c (find_mem_conflicts): Likewise.
460 * postreload.c (reload_combine_note_store): Likewise.
461 (move2add_note_store): Likewise.
462 * reload.c (find_equiv_reg): Likewise.
463 (mark_referenced_resources): Likewise.
464 * rtlanal.c (set_noop_p): Likewise.
465 (note_stores): Likewise.
466 (note_uses): Likewise.
467 * sched-deps.c (sched_analyze_1): Likewise.
468 * sched-rgn.c (check_live_1): Likewise.
469 (update_live_1): Likewise.
470 * config/i860/i860.c: Likewise.
472 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
474 (ZERO_EXTRACT): Mention that this one can be an lvalue.
476 * doc/rtl.texi: Update documentation for bit-fields and SET.
478 2005-01-05 Roger Sayle <roger@eyesopen.com>
480 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
482 2005-01-05 Jan Hubicka <jh@suse.cz>
483 Richard Henderson <rth@redhat.com>
486 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
489 2005-01-05 Richard Henderson <rth@redhat.com>
492 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
493 (sse_shufps): Change operand 3 to const_int_operand.
494 (sse2_storelps): Fix typo in template.
495 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
496 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
497 (ix86_expand_builtin): Likewise.
499 2005-01-05 Julian Brown <julian@codesourcery.com>
501 * config/arm/arm.c (arm_return_in_memory): Treat complex types
502 as aggregates for AAPCS ABIs.
504 2005-01-05 Stan Shebs <shebs@apple.com>
506 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
507 Mach-O getters if ppc64.
508 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
510 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
511 signed four-byte field for global code case.
512 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
514 2005-01-05 Roger Sayle <roger@eyesopen.com>
517 * c-common.c: Include real.h.
518 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
519 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
520 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
521 * Makefile.in (c-common.o): Update dependencies.
523 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
525 * c-parse.in (asm_string): Add trailing semicolon.
527 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
529 * c-parse.in (asm_string): New. Don't allow wide strings in
531 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
532 asm_string instead of STRING.
534 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
536 * c-typeck.c (constructor_no_implicit): Remove.
537 (set_designator, process_init_element): Don't check
538 constructor_no_implicit.
540 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
541 Kaz Kojima <kkojima@gcc.gnu.org>
544 * lcm.c (create_pre_exit): New.
545 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
546 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
548 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
550 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
551 -msoft-float in target_float_switch.
552 * arm.c (arm_override_options): Fix processing of target_float_switch.
554 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
556 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
559 2004-01-05 Uros Bizjak <uros@kss-loka.si>
561 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
562 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
564 2005-01-04 Roger Sayle <roger@eyesopen.com>
566 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
567 if deleting the then-block allows the test-block to fallthru to the
570 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
573 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
574 after the full declaration if the are in two different TUs.
576 2005-01-04 Richard Henderson <rth@redhat.com>
579 * tree-sra.c (generate_one_element_init): Just
580 call gimplify_and_add.
581 (generate_element_init): Record the
582 new referenced variables and mark them for renaming
584 (generate_element_init_1): This.
585 (scalarize_init): Don't call push_gimplify_context/
586 pop_gimplify_context.
588 2005-01-04 Geoffrey Keating <geoffk@apple.com>
590 * toplev.c (get_src_pwd): Handle failure of getpwd().
592 2005-01-04 Roger Sayle <roger@eyesopen.com>
594 * fold-const.c (fold_single_bit_test): Delete unreachable handling
596 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
597 all the cases handled by it are inverted by invert_truthvalue.
599 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
601 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
602 (save_register_windowsi): Likewise.
604 2005-01-04 Richard Henderson <rth@redhat.com>
606 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
608 2005-01-04 Uros Bizjak <uros@kss-loka.si>
611 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
612 any simplifications of vector mode comparison operators.
613 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
615 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
616 Devang Patel <dpatel@apple.com>
618 PR tree-optimization/18308
619 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
620 the operands before creating a new expression.
621 * dojump.c (do_jump): Make drop_through_label available
622 for all cases. Add expansion of COND_EXPR.
624 2005-01-04 Ira Rosen <irar@il.ibm.com>
626 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
627 INTEGER_CST instead of TREE_CONSTANT.
628 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
629 instead of TREE_CONSTANT.
630 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
631 instead of TREE_CONSTANT.
633 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
635 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
637 2005-01-04 Uros Bizjak <uros@kss-loka.si>
640 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
641 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
643 2005-01-03 Richard Henderson <rth@redhat.com>
645 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
646 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
647 fold_convert_const_real_from_real): Split out from ...
648 (fold_convert_const): ... here.
650 2005-01-03 Richard Henderson <rth@redhat.com>
653 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
654 (mov<MMXMODEI>_internal): Likewise.
655 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
656 Add V2SF case; use it for SSE1; don't use TI.
657 (movdf_integer): Likewise.
658 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
660 2005-01-03 Ira Rosen <irar@il.ibm.com>
662 * tree-vectorizer.c (vect_strip_conversions): New function.
663 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
664 check for binary class.
666 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
670 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
672 * gimple-low.c (mark_blocks_with_used_vars): New function.
673 (mark_blocks_with_used_subblocks): Ditto.
674 (mark_used_blocks): Ditto.
675 (pass_mark_used_blocks): New pass.
676 * tree-inline.c: Include debug.h.
677 (expand_call_inline): Call outlining_inline_function here.
678 * tree-optimize.c (init_tree_optimization_passes): Add
679 pass_mark_used_blocks.
680 * tree-pass.h (pass_mark_used_blocks): New.
681 * Makefile.in (tree-inline.o): Add debug.h dependency.
683 2005-01-03 Geoffrey Keating <geoffk@apple.com>
685 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
686 VAR_DECLs to have weak_import attribute.
688 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
691 2004-01-03 Steven Bosscher <stevenb@suse.de>
693 * passes.c (rest_of_compilation): Don't run regmove if only
694 flag_expensive_optimizations, require flag_regmove instead.
696 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
698 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
700 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
701 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
702 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
703 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
705 2005-01-03 Richard Henderson <rth@redhat.com>
707 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
708 (ia64_const_ok_for_letter_p): New.
709 (ia64_const_double_ok_for_letter_p): New.
710 (ia64_extra_constraint): New.
711 (ia64_expand_vecint_compare): New.
712 (ia64_expand_vcondu_v2si): New.
713 (ia64_expand_vecint_cmov): New.
714 (ia64_expand_vecint_minmax): New.
715 (ia64_print_operand): Add 'v'.
716 (ia64_preferred_reload_class): New.
717 (ia64_vector_mode_supported_p): New.
718 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
719 (PREFERRED_RELOAD_CLASS): Move to function.
720 (CONST_OK_FOR_LETTER_P): Move to function.
721 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
722 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
723 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
724 (EXTRA_CONSTRAINT): Move to function.
725 * config/ia64/ia64.md: Include vect.md.
726 (itanium_class): Add mmalua.
728 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
730 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
731 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
733 (const_int_2bit_operand): New.
734 (fr_reg_or_0_operand): New.
735 * config/ia64/ia64-modes.def: Add vector modes.
736 * config/ia64/ia64-protos.h: Update.
737 * config/ia64/vect.md: New file.
739 2005-01-03 Richard Henderson <rth@redhat.com>
741 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
743 2005-01-03 Uros Bizjak <uros@kss-loka.si>
746 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
748 (log1pdf2): Change mode of operands[0,1] to DFmode.
750 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
752 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
753 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
754 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
755 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
756 SPARC_RELAXED_ORDERING.
758 2005-01-03 Richard Henderson <rth@redhat.com>
759 Uros Bizjak <uros@kss-loka.si>
762 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
763 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
764 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
766 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
767 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
768 const_pow2_1_to_128_operand): New predicates.
770 2005-01-02 Greg McGary <greg@mcgary.org>
772 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
774 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
775 object. [ptr_type]: Remove unused variable.
776 Remember nearest addressable array-element or record-component
777 when checking bitfield components. Tolerate empty BIND_EXPR.
779 2005-01-01 Richard Henderson <rth@redhat.com>
781 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
782 that don't narrow the value. Fail for other conversions.
784 2005-01-01 Richard Henderson <rth@redhat.com>
787 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
788 * c-lang.c (finish_file): Don't do it here.
789 * objc/objc-act.c (objc_finish_file): Likewise.
791 * cgraph.c (decl_assembler_name_equal): New.
792 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
793 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
794 * cgraph.h (struct cgraph_varpool_node): Add next.
795 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
796 * varasm.c (assemble_alias): Mark the target as needed.
798 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
801 * function.c (get_arg_pointer_save_area): Use entry_of_function
802 instead of get_insns.
804 2005-01-01 Roger Sayle <roger@eyesopen.com>
805 Andrew Pinski <pinskia@physics.uc.edu>
806 James E. Wilson <wilson@specifixinc.com>
808 PR rtl-optimization/12092
809 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
812 2005-01-01 Roger Sayle <roger@eyesopen.com>
813 Olivier Hainque <hainque@act-europe.fr>
815 * tree.c (int_fits_type_p): A narrower type always fits in a
816 wider one, except for negative values into unsigned types.
818 2005-01-01 Roger Sayle <roger@eyesopen.com>
820 * tree.c (int_fits_type_p): Compare the result of force_fit_type
821 with the original constant rather than require TREE_OVERFLOW.
823 2005-01-01 Steven Bosscher <stevenb@suse.de>
826 * c-decl.c (finish_function): If compiling C99, annotate the
827 compiler generated return with the current file name and line 0.
828 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
829 the source line is greater than 0.
830 (remove_bb): Likewise.
832 See ChangeLog.12 for earlier changes.