1 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
3 * config/s390/s390.c (override_options): Return error if
4 -mbackchain, -mpacked-stack and -mhard-float are used together.
5 (s390_va_start): Remove the backchain && packed-stack special case.
6 (s390_gimplify_va_arg): Likewise.
7 * doc/invoke.texi: Remove the ABI incompatibility note.
9 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
11 * config/s390/s390.c (struct s390_frame_layout): Remove
13 (s390_frame_info, s390_emit_prologue): Replace occurrences of
14 save_backchain_p with TARGET_BACKCHAIN.
16 2005-01-11 Alan Modra <amodra@bigpond.net.au>
19 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
20 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
21 * function.c (locate_and_pad_parm): Set new field.
22 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
23 Tweak where_pad test to include "none". Always set mem align for
25 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
26 not sufficient for type.
27 (assign_parm_setup_block): If stack_parm is zero on entry, always
28 make a new stack local. Block move old stack parm if necessary
29 to new aligned stack local.
30 (assign_parm_setup_stack): Use a block move to handle
31 potentially misaligned entry_parm.
32 (assign_parms_unsplit_complex): Specify required alignment when
34 * calls.c (compute_argument_addresses): Override alignment of stack
35 arg calculated from its type with the alignment given by
36 FUNCTION_ARG_BOUNDARY.
37 (store_one_arg): Likewise.
39 2005-01-11 Jan Beulich <jbeulich@novell.com>
41 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
42 Change respective itanium_class attribute to ialu.
43 (shladdp4_internal): New.
44 * config/ia64/predicates.md (shladd_log2_operand): New.
46 2005-01-11 Richard Henderson <rth@redhat.com>
48 * expr.c (store_constructor): Use rtvec_alloc instead of
49 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
51 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
53 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
55 2005-01-11 Alan Modra <amodra@bigpond.net.au>
57 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
58 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
59 (TARGET_RWRELOC): Define.
60 (ia64_rwreloc_section_type_flags): Delete.
61 (ia64_section_type_flags): New function.
62 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
63 (TARGET_RWRELOC): Define.
65 2005-01-10 David Mosberger <davidm@hpl.hp.com>
68 * config/ia64/ia64.c (process_set): For alloc insn, only call
69 process_epilogue is !frame_pointer_needed.
71 2005-01-10 Roger Sayle <roger@eyesopen.com>
74 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
75 unary operator and can't be treated as a binary/comparison operator.
77 2005-01-10 Richard Henderson <rth@redhat.com>
79 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
80 (ix86_return_in_memory): Likewise.
81 (function_arg_advance): Likewise. Mirror structure in function_arg
82 for choosing register to advance.
84 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
86 * tree-vectorizer.c, tree.def: Fix comment typos.
88 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
90 * config/s390/s390.c (struct processor_costs): 4 new fields:
92 (s390_rtx_costs): More precise handling of divide instructions.
94 2005-01-10 David Edelsohn <edelsohn@gnu.org>
96 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
98 2005-01-10 Jan Beulich <jbeulich@novell.com>
100 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
101 names resulting from -ffunction-sections/-fdata-sections and linkonce
103 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
104 based on the section name. Rearrange the section name comparison logic
105 slightly so that each section name is compared against at most once.
107 2005-01-10 Ben Elliston <bje@au.ibm.com>
109 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
110 targets for which -fPIC reduces limitations on the GOT size.
112 2005-01-09 Falk Hueffner <falk@debian.org>
114 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
117 2005-01-09 David Edelsohn <edelsohn@gnu.org>
120 * collect2.c (main): Set aixrtl_flag for -brtl option.
121 (resolve_lib_name): Search for .so file extension before .a
124 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
126 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
127 (vect_analyze_data_refs_alignment): Add dump prints.
129 2005-01-09 Ira Rosen <irar@il.ibm.com>
131 * tree-vectorizer.c (vect_analyze_offset_expr): Use
132 expr_invariant_in_loop_p.
133 Initialize outputs first thing in the function.
134 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
135 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
136 (vect_analyze_pointer_ref_access): Check that the initial condition of
137 the access function is loop invariant.
139 2005-01-09 Richard Henderson <rth@redhat.com>
141 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
142 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
143 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
144 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
145 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
146 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
147 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
148 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
149 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
150 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
151 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
152 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
153 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
154 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
155 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
156 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
157 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
158 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
159 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
160 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
161 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
162 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
163 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
165 (mmx_clrdi, pavgusb): Remove.
166 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
167 with leading sse_ prefix.
168 * config/i386/sse.md: Receive them.
169 * config/i386/mmx.md: New file.
170 (MMXMODE12, MMXMODE24, mmxvecsize): New.
171 (subrv2sf3): Turn into expander for normal subtraction.
172 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
173 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
174 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
175 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
176 commutative; use ix86_binary_operator_ok.
177 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
178 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
179 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
180 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
181 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
182 Macroize from existing patterns; use ix86_binary_operator_ok.
183 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
184 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
185 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
186 with vec_select+vec_concat.
187 (mmx_pshufw, mmx_pshufw_1): Likewise.
188 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
189 (mmx_uavgv4hi3): Model correctly.
190 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
191 _mm_xor_si64): Remove casts.
193 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
195 PR tree-optimization/19224
196 * tree-scalar-evolution.c (get_instantiated_value,
197 set_instantiated_value): New functions.
198 (instantiate_parameters_1): Cache the results.
199 (instantiate_parameters, resolve_mixers): Initialize and free
202 2005-01-08 David Edelsohn <edelsohn@gnu.org>
204 * config/i386/i386.md (addhi_4): Correct reference in comment.
207 2005-01-08 Richard Henderson <rth@redhat.com>
209 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
210 (_mm_cvtsi128_si64x): Likewise.
211 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
212 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
213 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
214 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
215 * config/i386/i386-protos.h: Update.
216 * config/i386/i386.c (print_operand): Add 'H'.
217 (ix86_fixup_binary_operands): Split out from ...
218 (ix86_expand_binary_operator): ... here.
219 (ix86_fixup_binary_operands_no_copy): New.
220 (ix86_expand_fp_absneg_operator): Handle vector mode results.
221 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
222 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
223 (safe_vector_operand): Use CONST0_RTX.
224 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
225 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
226 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
227 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
229 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
230 (IX86_BUILTIN_CMPNGESS): New.
231 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
232 (attr type): Add sselog1.
233 (attr unit, attr memory): Handle it.
234 (movti, movti_internal, movti_rex64): Move near other integer moves.
235 (movtf, movtf_internal): Move near other fp moves.
236 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
237 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
238 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
239 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
240 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
241 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
242 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
243 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
244 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
245 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
246 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
247 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
248 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
249 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
250 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
251 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
252 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
253 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
254 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
255 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
256 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
257 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
258 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
259 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
260 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
261 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
262 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
263 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
264 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
265 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
266 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
267 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
268 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
269 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
270 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
271 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
272 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
273 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
274 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
275 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
276 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
277 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
278 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
279 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
281 (SSEPUSH, push<SSEPUSH>): Remove.
282 (MMXPUSH, push<MMXPUSH>): Remove.
283 (sse_movaps, sse_movaps_1, sse_movups): Remove.
284 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
285 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
286 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
287 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
288 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
289 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
290 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
291 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
292 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
293 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
294 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
295 (ppro_sse_log_V4SF): Handle sselog1.
296 * config/i386/predicates.md (const_0_to_1_operand): New.
297 (const_0_to_255_mul_8_operand): New.
298 (const_1_to_31_operand): Rename from const_int_1_31_operand.
299 (const_2_to_3_operand, const_4_to_7_operand): New.
300 * config/i386/sse.md: New file.
301 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
302 (sse_movups): Rename from sse_movups_1.
303 (sse_loadlss): Rename from sse_loadss_1.
304 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
306 (negv4sf2): Use ix86_expand_fp_absneg_operator.
307 (absv4sf2, negv2df, absv2df): New.
308 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
309 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
310 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
311 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
312 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
313 (sse3_addsubv4sf3): Model correctly.
314 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
315 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
316 (sse_movhlps): Model with vec_select+vec_concat.
317 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
318 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
319 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
320 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
321 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
322 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
323 sse2_pshufhw_1): Likewise.
324 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
325 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
326 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
327 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
328 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
329 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
330 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
331 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
332 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
333 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
334 sminv8hi3): Mark commutative
335 operands. Use ix86_binary_operator_ok.
336 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
337 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
338 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
339 sse2_punpcklqdq): Allow operand2 in memory.
340 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
341 sse2_movsd): Add memory alternatives.
342 (sse_storelps): Turn expander into an insn; split after reload.
343 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
344 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
346 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
348 * configure.ac (DWARF-2 debug_line): Use objdump.
349 * configure: Regenerate.
351 2005-01-08 Jeff Law <law@redhat.com>
352 Diego Novillo <dnovillo@redhat.com>
354 PR tree-optimization/18241
355 * tree-nrv.c (tree_nrv): Ignore volatile return values.
356 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
357 statements with volatile operands.
358 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
359 operands after marking a statement with has_volatile_ops.
361 2005-01-08 Roger Sayle <roger@eyesopen.com>
363 * tree.c (int_fits_type_p): Always honor integer constant
364 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
366 2005-01-08 Roger Sayle <roger@eyesopen.com>
368 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
369 the latent bug in rtl_delete_block has been resolved.
371 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
373 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
374 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
376 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
378 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
379 library as @multilib_dir@/$(SHLIB_SONAME).
381 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
383 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
384 * configure: Regenerate.
386 2005-01-07 Jakub Jelinek <jakub@redhat.com>
388 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
391 2005-01-07 David Edelsohn <edelsohn@gnu.org>
394 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
395 Convert non-word aligned offset address using ld/std into
398 2005-01-07 Richard Henderson <rth@redhat.com>
400 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
402 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
404 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
406 2005-01-07 Jakub Jelinek <jakub@redhat.com>
408 PR tree-optimization/19283
409 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
410 is not shorter than the original type.
412 PR rtl-optimization/19012
413 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
415 PR rtl-optimization/18861
416 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
417 (FULL_STATE): Define.
418 (mark_tablejump_edge): New function.
419 (purge_dead_tablejump_edges): New function.
420 (find_bb_boundaries): Use it.
422 PR tree-optimization/18828
423 * builtins.c (expand_builtin_next_arg): Remove argument and all
424 the argument checking.
425 (expand_builtin): Adjust caller.
426 (expand_builtin_va_start): Likewise. Remove error for too many
428 (fold_builtin_next_arg): Issue error for too many arguments.
429 After checking arguments, replace them with magic arguments that
430 prevent further checking of the args.
432 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
434 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
436 2005-01-06 Roger Sayle <roger@eyesopen.com>
439 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
440 and %{.S:...} (and their negative variants) to test whether the
441 input file is assembler or pre-processed-assembler independent of
442 the actual filename extension.
444 2005-01-06 Roger Sayle <roger@eyesopen.com>
446 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
447 of sign or zero extended values.
449 2005-01-06 Geoffrey Keating <geoffk@apple.com>
451 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
452 for IBM long double format correctly.
454 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
456 Fix PR tree-optimization/18792
458 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
459 to first_loop_depth, and use loop depth instead of loop number.
460 (build_classic_dir_vector): Ditto.
461 (compute_data_dependences_for_loop): Use depth, not loop number.
462 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
463 number. Pass in loops, instead of loop numbers.
464 (gather_interchange_stats): Ditto.
465 (linear_transform_loops): Ditto.
467 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
470 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
471 biv_fits_mode_p, extension_within_bounds_p): New functions.
472 (check_ext_dependent_givs): Use them.
474 2005-01-06 Roger Sayle <roger@eyesopen.com>
476 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
477 more than one barrier, in which case we should delete them all.
479 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
481 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
483 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
485 Fix PR middle-end/19286
487 * dwarf2out.c (gen_subprogram_die): If we've already tried to
488 output this subprogram, simply ignore this attempt to do it again.
489 (add_abstract_origin_attribute): Don't abort trying to add the abstract
490 origin attribute if it's not possible.
491 (gen_block_die): Don't ignore subblocks of "unused" blocks.
492 (decls_for_scope): Ditto.
493 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
494 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
496 2005-01-05 Richard Henderson <rth@redhat.com>
499 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
500 (bdesc_2arg): Use it.
501 (ix86_expand_binop_builtin): Force operands into registers
503 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
504 ix86_expand_sse_compare, ix86_expand_sse_comi,
505 ix86_expand_builtin): Likewise.
507 2005-01-05 Richard Henderson <rth@redhat.com>
509 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
511 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
512 (mulv4hi3): Set itanium_class mmmul.
513 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
514 (fpack): Rename from fpack_sfsf.
516 2005-01-05 Richard Henderson <rth@redhat.com>
519 * reload1.c (do_input_reload): Restrict the optimization deleteing
520 a previous output reload to RELOAD_FOR_INPUT.
522 2005-01-05 Steven Bosscher <stevenb@suse.de>
524 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
525 comment that we fall through after case.
526 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
527 (distribute_links): Likewise.
528 * cse.c (cse_insn): Likewise.
529 * cselib.c (cselib_invalidate_mem): Likewise.
530 * df.c: Update comments at the top of the file.
531 (read_modify_subreg_p): Update comments here too.
532 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
533 * flow.c (mark_set_1): Likewise.
534 (mark_used_regs): Likewise.
535 * gcse.c (mems_conflict_for_gcse_p): Likewise.
536 (canon_list_insert): Likewise.
537 (mark_set): Likewise.
538 (try_replace_reg): Likewise.
539 (store_killed_in_insn): Likewise.
540 * loop.c (count_one_set): Likewise.
541 (basic_induction_var): Likewise.
542 * postreload-gcse.c (find_mem_conflicts): Likewise.
543 * postreload.c (reload_combine_note_store): Likewise.
544 (move2add_note_store): Likewise.
545 * reload.c (find_equiv_reg): Likewise.
546 (mark_referenced_resources): Likewise.
547 * rtlanal.c (set_noop_p): Likewise.
548 (note_stores): Likewise.
549 (note_uses): Likewise.
550 * sched-deps.c (sched_analyze_1): Likewise.
551 * sched-rgn.c (check_live_1): Likewise.
552 (update_live_1): Likewise.
553 * config/i860/i860.c: Likewise.
555 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
557 (ZERO_EXTRACT): Mention that this one can be an lvalue.
559 * doc/rtl.texi: Update documentation for bit-fields and SET.
561 2005-01-05 Roger Sayle <roger@eyesopen.com>
563 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
565 2005-01-05 Jan Hubicka <jh@suse.cz>
566 Richard Henderson <rth@redhat.com>
569 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
572 2005-01-05 Richard Henderson <rth@redhat.com>
575 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
576 (sse_shufps): Change operand 3 to const_int_operand.
577 (sse2_storelps): Fix typo in template.
578 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
579 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
580 (ix86_expand_builtin): Likewise.
582 2005-01-05 Julian Brown <julian@codesourcery.com>
584 * config/arm/arm.c (arm_return_in_memory): Treat complex types
585 as aggregates for AAPCS ABIs.
587 2005-01-05 Stan Shebs <shebs@apple.com>
589 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
590 Mach-O getters if ppc64.
591 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
593 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
594 signed four-byte field for global code case.
595 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
597 2005-01-05 Roger Sayle <roger@eyesopen.com>
600 * c-common.c: Include real.h.
601 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
602 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
603 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
604 * Makefile.in (c-common.o): Update dependencies.
606 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
608 * c-parse.in (asm_string): Add trailing semicolon.
610 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
612 * c-parse.in (asm_string): New. Don't allow wide strings in
614 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
615 asm_string instead of STRING.
617 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
619 * c-typeck.c (constructor_no_implicit): Remove.
620 (set_designator, process_init_element): Don't check
621 constructor_no_implicit.
623 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
624 Kaz Kojima <kkojima@gcc.gnu.org>
627 * lcm.c (create_pre_exit): New.
628 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
629 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
631 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
633 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
634 -msoft-float in target_float_switch.
635 * arm.c (arm_override_options): Fix processing of target_float_switch.
637 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
639 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
642 2004-01-05 Uros Bizjak <uros@kss-loka.si>
644 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
645 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
647 2005-01-04 Roger Sayle <roger@eyesopen.com>
649 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
650 if deleting the then-block allows the test-block to fallthru to the
653 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
656 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
657 after the full declaration if the are in two different TUs.
659 2005-01-04 Richard Henderson <rth@redhat.com>
662 * tree-sra.c (generate_one_element_init): Just
663 call gimplify_and_add.
664 (generate_element_init): Record the
665 new referenced variables and mark them for renaming
667 (generate_element_init_1): This.
668 (scalarize_init): Don't call push_gimplify_context/
669 pop_gimplify_context.
671 2005-01-04 Geoffrey Keating <geoffk@apple.com>
673 * toplev.c (get_src_pwd): Handle failure of getpwd().
675 2005-01-04 Roger Sayle <roger@eyesopen.com>
677 * fold-const.c (fold_single_bit_test): Delete unreachable handling
679 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
680 all the cases handled by it are inverted by invert_truthvalue.
682 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
684 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
685 (save_register_windowsi): Likewise.
687 2005-01-04 Richard Henderson <rth@redhat.com>
689 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
691 2005-01-04 Uros Bizjak <uros@kss-loka.si>
694 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
695 any simplifications of vector mode comparison operators.
696 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
698 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
699 Devang Patel <dpatel@apple.com>
701 PR tree-optimization/18308
702 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
703 the operands before creating a new expression.
704 * dojump.c (do_jump): Make drop_through_label available
705 for all cases. Add expansion of COND_EXPR.
707 2005-01-04 Ira Rosen <irar@il.ibm.com>
709 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
710 INTEGER_CST instead of TREE_CONSTANT.
711 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
712 instead of TREE_CONSTANT.
713 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
714 instead of TREE_CONSTANT.
716 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
718 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
720 2005-01-04 Uros Bizjak <uros@kss-loka.si>
723 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
724 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
726 2005-01-03 Richard Henderson <rth@redhat.com>
728 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
729 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
730 fold_convert_const_real_from_real): Split out from ...
731 (fold_convert_const): ... here.
733 2005-01-03 Richard Henderson <rth@redhat.com>
736 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
737 (mov<MMXMODEI>_internal): Likewise.
738 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
739 Add V2SF case; use it for SSE1; don't use TI.
740 (movdf_integer): Likewise.
741 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
743 2005-01-03 Ira Rosen <irar@il.ibm.com>
745 * tree-vectorizer.c (vect_strip_conversions): New function.
746 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
747 check for binary class.
749 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
753 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
755 * gimple-low.c (mark_blocks_with_used_vars): New function.
756 (mark_blocks_with_used_subblocks): Ditto.
757 (mark_used_blocks): Ditto.
758 (pass_mark_used_blocks): New pass.
759 * tree-inline.c: Include debug.h.
760 (expand_call_inline): Call outlining_inline_function here.
761 * tree-optimize.c (init_tree_optimization_passes): Add
762 pass_mark_used_blocks.
763 * tree-pass.h (pass_mark_used_blocks): New.
764 * Makefile.in (tree-inline.o): Add debug.h dependency.
766 2005-01-03 Geoffrey Keating <geoffk@apple.com>
768 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
769 VAR_DECLs to have weak_import attribute.
771 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
774 2004-01-03 Steven Bosscher <stevenb@suse.de>
776 * passes.c (rest_of_compilation): Don't run regmove if only
777 flag_expensive_optimizations, require flag_regmove instead.
779 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
781 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
783 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
784 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
785 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
786 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
788 2005-01-03 Richard Henderson <rth@redhat.com>
790 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
791 (ia64_const_ok_for_letter_p): New.
792 (ia64_const_double_ok_for_letter_p): New.
793 (ia64_extra_constraint): New.
794 (ia64_expand_vecint_compare): New.
795 (ia64_expand_vcondu_v2si): New.
796 (ia64_expand_vecint_cmov): New.
797 (ia64_expand_vecint_minmax): New.
798 (ia64_print_operand): Add 'v'.
799 (ia64_preferred_reload_class): New.
800 (ia64_vector_mode_supported_p): New.
801 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
802 (PREFERRED_RELOAD_CLASS): Move to function.
803 (CONST_OK_FOR_LETTER_P): Move to function.
804 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
805 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
806 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
807 (EXTRA_CONSTRAINT): Move to function.
808 * config/ia64/ia64.md: Include vect.md.
809 (itanium_class): Add mmalua.
811 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
813 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
814 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
816 (const_int_2bit_operand): New.
817 (fr_reg_or_0_operand): New.
818 * config/ia64/ia64-modes.def: Add vector modes.
819 * config/ia64/ia64-protos.h: Update.
820 * config/ia64/vect.md: New file.
822 2005-01-03 Richard Henderson <rth@redhat.com>
824 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
826 2005-01-03 Uros Bizjak <uros@kss-loka.si>
829 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
831 (log1pdf2): Change mode of operands[0,1] to DFmode.
833 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
835 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
836 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
837 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
838 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
839 SPARC_RELAXED_ORDERING.
841 2005-01-03 Richard Henderson <rth@redhat.com>
842 Uros Bizjak <uros@kss-loka.si>
845 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
846 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
847 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
849 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
850 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
851 const_pow2_1_to_128_operand): New predicates.
853 2005-01-02 Greg McGary <greg@mcgary.org>
855 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
857 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
858 object. [ptr_type]: Remove unused variable.
859 Remember nearest addressable array-element or record-component
860 when checking bitfield components. Tolerate empty BIND_EXPR.
862 2005-01-01 Richard Henderson <rth@redhat.com>
864 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
865 that don't narrow the value. Fail for other conversions.
867 2005-01-01 Richard Henderson <rth@redhat.com>
870 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
871 * c-lang.c (finish_file): Don't do it here.
872 * objc/objc-act.c (objc_finish_file): Likewise.
874 * cgraph.c (decl_assembler_name_equal): New.
875 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
876 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
877 * cgraph.h (struct cgraph_varpool_node): Add next.
878 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
879 * varasm.c (assemble_alias): Mark the target as needed.
881 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
884 * function.c (get_arg_pointer_save_area): Use entry_of_function
885 instead of get_insns.
887 2005-01-01 Roger Sayle <roger@eyesopen.com>
888 Andrew Pinski <pinskia@physics.uc.edu>
889 James E. Wilson <wilson@specifixinc.com>
891 PR rtl-optimization/12092
892 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
895 2005-01-01 Roger Sayle <roger@eyesopen.com>
896 Olivier Hainque <hainque@act-europe.fr>
898 * tree.c (int_fits_type_p): A narrower type always fits in a
899 wider one, except for negative values into unsigned types.
901 2005-01-01 Roger Sayle <roger@eyesopen.com>
903 * tree.c (int_fits_type_p): Compare the result of force_fit_type
904 with the original constant rather than require TREE_OVERFLOW.
906 2005-01-01 Steven Bosscher <stevenb@suse.de>
909 * c-decl.c (finish_function): If compiling C99, annotate the
910 compiler generated return with the current file name and line 0.
911 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
912 the source line is greater than 0.
913 (remove_bb): Likewise.
915 See ChangeLog.12 for earlier changes.