1 2005-01-12 Nick Clifton <nickc@redhat.com>
3 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
4 passed on to GAS) but do not define, as GAS no longer supports or
5 needs the -m2000 option.
7 2005-01-11 Roger Sayle <roger@eyesopen.com>
9 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
10 type as args of unprototyped builtins aren't automatically promoted.
11 (fold_builtin_abs): Likewise.
13 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
15 * Makefile.in: Set a `build-warn' variable.
17 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
19 PR tree-optimization/17949
20 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
21 (find_interesting_uses_address): Use it.
23 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
25 * regrename.c (kill_value): Handle subreg's that won't simplify.
27 2005-01-11 Richard Henderson <rth@redhat.com>
30 * config/i386/i386.h (enum ix86_builtins): Move ...
31 * config/i386/i386.c: ... here.
32 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
33 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
34 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
35 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
36 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
37 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
38 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
39 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
40 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
41 IX86_BUILTIN_LOADDDUP): Remove.
42 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
43 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
44 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
45 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
46 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
47 (ix86_init_builtins): Make static.
48 (ix86_init_mmx_sse_builtins): Update for changed builtins.
49 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
50 if all the modes match. Otherwise, fake it.
51 (get_element_number, ix86_expand_vec_init_builtin,
52 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
53 (ix86_expand_builtin): Make static. Update for changed builtins.
54 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
55 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
56 (ix86_expand_vector_init_duplicate): New.
57 (ix86_expand_vector_init_low_nonzero): New.
58 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
59 Split out from ix86_expand_vector_init; handle integer modes.
60 (ix86_expand_vector_init): Use them.
61 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
62 * config/i386/i386-protos.h: Update.
63 * config/i386/predicates.md (reg_or_0_operand): New.
64 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
65 (movv2sf_internal): Likewise. And a splitter to match them all.
66 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
67 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
68 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
69 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
70 (mmx_pinsrw): Fix operand ordering.
71 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
72 rather than sse_loadss expander.
73 (movv2df splitter): Similarly.
74 (sse_loadss, sse_loadlss): Remove.
75 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
76 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
77 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
78 (sse3_movddup): Rename with '*'.
79 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
80 (sse2_loadsd): Remove.
81 (vec_dupv2df_sse3): Rename from sse3_loadddup.
82 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
83 (sse2_pinsrw): Fix argument ordering.
84 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
85 (sse2_stored): Remove 'r' destination.
86 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
87 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
88 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
89 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
90 vec_extractv16qi, vec_initv16qi): New.
92 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
93 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
94 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
96 (_mm_load_pd, _mm_store_pd): Use plain dereference.
97 (_mm_load_si128, _mm_store_si128): Likewise.
98 (_mm_load1_pd): Use _mm_set1_pd.
99 (_mm_load_sd): Use _mm_set_sd.
100 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
101 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
102 (_mm_set_epi64): Use _mm_set_epi64x.
103 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
104 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
105 _mm_setr_epi8): Use _mm_set_foo form.
106 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
107 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
108 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
109 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
110 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
111 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
112 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
113 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
114 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
115 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
116 (_mm_movedup_pd): Use _mm_shuffle_pd.
117 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
118 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
119 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
120 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
122 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
123 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
124 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
125 (_mm_store_ps): Use raw dereference.
126 (_mm_store1_ps): Use _mm_storeu_ps.
127 (_mm_storer_ps): Use _mm_store_ps.
128 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
129 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
131 2005-01-11 Stan Shebs <shebs@apple.com>
133 * config/rs6000/rs6000.c (machopic_output_stub): Issue
134 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
136 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
139 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
140 Skip all DECLs except for FIELD_DECLs.
142 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
144 * config/s390/s390.c (override_options): Return error if
145 -mbackchain, -mpacked-stack and -mhard-float are used together.
146 (s390_va_start): Remove the backchain && packed-stack special case.
147 (s390_gimplify_va_arg): Likewise.
148 * doc/invoke.texi: Remove the ABI incompatibility note.
150 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
152 * config/s390/s390.c (struct s390_frame_layout): Remove
154 (s390_frame_info, s390_emit_prologue): Replace occurrences of
155 save_backchain_p with TARGET_BACKCHAIN.
157 2005-01-11 Alan Modra <amodra@bigpond.net.au>
160 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
161 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
162 * function.c (locate_and_pad_parm): Set new field.
163 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
164 Tweak where_pad test to include "none". Always set mem align for
166 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
167 not sufficient for type.
168 (assign_parm_setup_block): If stack_parm is zero on entry, always
169 make a new stack local. Block move old stack parm if necessary
170 to new aligned stack local.
171 (assign_parm_setup_stack): Use a block move to handle
172 potentially misaligned entry_parm.
173 (assign_parms_unsplit_complex): Specify required alignment when
174 creating stack local.
175 * calls.c (compute_argument_addresses): Override alignment of stack
176 arg calculated from its type with the alignment given by
177 FUNCTION_ARG_BOUNDARY.
178 (store_one_arg): Likewise.
180 2005-01-11 Jan Beulich <jbeulich@novell.com>
182 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
183 Change respective itanium_class attribute to ialu.
184 (shladdp4_internal): New.
185 * config/ia64/predicates.md (shladd_log2_operand): New.
187 2005-01-11 Richard Henderson <rth@redhat.com>
189 * expr.c (store_constructor): Use rtvec_alloc instead of
190 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
192 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
194 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
196 2005-01-11 Alan Modra <amodra@bigpond.net.au>
198 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
199 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
200 (TARGET_RWRELOC): Define.
201 (ia64_rwreloc_section_type_flags): Delete.
202 (ia64_section_type_flags): New function.
203 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
204 (TARGET_RWRELOC): Define.
206 2005-01-10 David Mosberger <davidm@hpl.hp.com>
209 * config/ia64/ia64.c (process_set): For alloc insn, only call
210 process_epilogue is !frame_pointer_needed.
212 2005-01-10 Roger Sayle <roger@eyesopen.com>
215 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
216 unary operator and can't be treated as a binary/comparison operator.
218 2005-01-10 Richard Henderson <rth@redhat.com>
220 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
221 (ix86_return_in_memory): Likewise.
222 (function_arg_advance): Likewise. Mirror structure in function_arg
223 for choosing register to advance.
225 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
227 * tree-vectorizer.c, tree.def: Fix comment typos.
229 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
231 * config/s390/s390.c (struct processor_costs): 4 new fields:
232 ddbr, ddr, debr, der.
233 (s390_rtx_costs): More precise handling of divide instructions.
235 2005-01-10 David Edelsohn <edelsohn@gnu.org>
237 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
239 2005-01-10 Jan Beulich <jbeulich@novell.com>
241 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
242 names resulting from -ffunction-sections/-fdata-sections and linkonce
244 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
245 based on the section name. Rearrange the section name comparison logic
246 slightly so that each section name is compared against at most once.
248 2005-01-10 Ben Elliston <bje@au.ibm.com>
250 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
251 targets for which -fPIC reduces limitations on the GOT size.
253 2005-01-09 Falk Hueffner <falk@debian.org>
255 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
258 2005-01-09 David Edelsohn <edelsohn@gnu.org>
261 * collect2.c (main): Set aixrtl_flag for -brtl option.
262 (resolve_lib_name): Search for .so file extension before .a
265 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
267 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
268 (vect_analyze_data_refs_alignment): Add dump prints.
270 2005-01-09 Ira Rosen <irar@il.ibm.com>
272 * tree-vectorizer.c (vect_analyze_offset_expr): Use
273 expr_invariant_in_loop_p.
274 Initialize outputs first thing in the function.
275 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
276 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
277 (vect_analyze_pointer_ref_access): Check that the initial condition of
278 the access function is loop invariant.
280 2005-01-09 Richard Henderson <rth@redhat.com>
282 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
283 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
284 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
285 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
286 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
287 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
288 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
289 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
290 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
291 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
292 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
293 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
294 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
295 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
296 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
297 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
298 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
299 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
300 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
301 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
302 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
303 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
304 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
306 (mmx_clrdi, pavgusb): Remove.
307 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
308 with leading sse_ prefix.
309 * config/i386/sse.md: Receive them.
310 * config/i386/mmx.md: New file.
311 (MMXMODE12, MMXMODE24, mmxvecsize): New.
312 (subrv2sf3): Turn into expander for normal subtraction.
313 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
314 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
315 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
316 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
317 commutative; use ix86_binary_operator_ok.
318 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
319 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
320 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
321 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
322 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
323 Macroize from existing patterns; use ix86_binary_operator_ok.
324 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
325 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
326 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
327 with vec_select+vec_concat.
328 (mmx_pshufw, mmx_pshufw_1): Likewise.
329 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
330 (mmx_uavgv4hi3): Model correctly.
331 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
332 _mm_xor_si64): Remove casts.
334 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
336 PR tree-optimization/19224
337 * tree-scalar-evolution.c (get_instantiated_value,
338 set_instantiated_value): New functions.
339 (instantiate_parameters_1): Cache the results.
340 (instantiate_parameters, resolve_mixers): Initialize and free
343 2005-01-08 David Edelsohn <edelsohn@gnu.org>
345 * config/i386/i386.md (addhi_4): Correct reference in comment.
348 2005-01-08 Richard Henderson <rth@redhat.com>
350 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
351 (_mm_cvtsi128_si64x): Likewise.
352 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
353 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
354 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
355 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
356 * config/i386/i386-protos.h: Update.
357 * config/i386/i386.c (print_operand): Add 'H'.
358 (ix86_fixup_binary_operands): Split out from ...
359 (ix86_expand_binary_operator): ... here.
360 (ix86_fixup_binary_operands_no_copy): New.
361 (ix86_expand_fp_absneg_operator): Handle vector mode results.
362 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
363 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
364 (safe_vector_operand): Use CONST0_RTX.
365 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
366 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
367 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
368 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
370 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
371 (IX86_BUILTIN_CMPNGESS): New.
372 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
373 (attr type): Add sselog1.
374 (attr unit, attr memory): Handle it.
375 (movti, movti_internal, movti_rex64): Move near other integer moves.
376 (movtf, movtf_internal): Move near other fp moves.
377 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
378 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
379 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
380 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
381 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
382 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
383 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
384 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
385 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
386 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
387 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
388 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
389 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
390 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
391 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
392 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
393 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
394 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
395 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
396 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
397 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
398 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
399 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
400 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
401 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
402 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
403 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
404 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
405 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
406 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
407 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
408 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
409 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
410 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
411 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
412 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
413 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
414 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
415 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
416 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
417 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
418 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
419 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
420 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
422 (SSEPUSH, push<SSEPUSH>): Remove.
423 (MMXPUSH, push<MMXPUSH>): Remove.
424 (sse_movaps, sse_movaps_1, sse_movups): Remove.
425 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
426 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
427 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
428 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
429 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
430 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
431 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
432 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
433 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
434 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
435 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
436 (ppro_sse_log_V4SF): Handle sselog1.
437 * config/i386/predicates.md (const_0_to_1_operand): New.
438 (const_0_to_255_mul_8_operand): New.
439 (const_1_to_31_operand): Rename from const_int_1_31_operand.
440 (const_2_to_3_operand, const_4_to_7_operand): New.
441 * config/i386/sse.md: New file.
442 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
443 (sse_movups): Rename from sse_movups_1.
444 (sse_loadlss): Rename from sse_loadss_1.
445 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
447 (negv4sf2): Use ix86_expand_fp_absneg_operator.
448 (absv4sf2, negv2df, absv2df): New.
449 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
450 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
451 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
452 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
453 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
454 (sse3_addsubv4sf3): Model correctly.
455 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
456 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
457 (sse_movhlps): Model with vec_select+vec_concat.
458 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
459 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
460 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
461 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
462 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
463 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
464 sse2_pshufhw_1): Likewise.
465 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
466 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
467 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
468 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
469 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
470 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
471 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
472 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
473 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
474 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
475 sminv8hi3): Mark commutative
476 operands. Use ix86_binary_operator_ok.
477 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
478 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
479 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
480 sse2_punpcklqdq): Allow operand2 in memory.
481 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
482 sse2_movsd): Add memory alternatives.
483 (sse_storelps): Turn expander into an insn; split after reload.
484 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
485 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
487 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
489 * configure.ac (DWARF-2 debug_line): Use objdump.
490 * configure: Regenerate.
492 2005-01-08 Jeff Law <law@redhat.com>
493 Diego Novillo <dnovillo@redhat.com>
495 PR tree-optimization/18241
496 * tree-nrv.c (tree_nrv): Ignore volatile return values.
497 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
498 statements with volatile operands.
499 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
500 operands after marking a statement with has_volatile_ops.
502 2005-01-08 Roger Sayle <roger@eyesopen.com>
504 * tree.c (int_fits_type_p): Always honor integer constant
505 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
507 2005-01-08 Roger Sayle <roger@eyesopen.com>
509 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
510 the latent bug in rtl_delete_block has been resolved.
512 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
514 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
515 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
517 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
519 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
520 library as @multilib_dir@/$(SHLIB_SONAME).
522 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
524 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
525 * configure: Regenerate.
527 2005-01-07 Jakub Jelinek <jakub@redhat.com>
529 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
532 2005-01-07 David Edelsohn <edelsohn@gnu.org>
535 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
536 Convert non-word aligned offset address using ld/std into
539 2005-01-07 Richard Henderson <rth@redhat.com>
541 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
543 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
545 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
547 2005-01-07 Jakub Jelinek <jakub@redhat.com>
549 PR tree-optimization/19283
550 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
551 is not shorter than the original type.
553 PR rtl-optimization/19012
554 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
556 PR rtl-optimization/18861
557 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
558 (FULL_STATE): Define.
559 (mark_tablejump_edge): New function.
560 (purge_dead_tablejump_edges): New function.
561 (find_bb_boundaries): Use it.
563 PR tree-optimization/18828
564 * builtins.c (expand_builtin_next_arg): Remove argument and all
565 the argument checking.
566 (expand_builtin): Adjust caller.
567 (expand_builtin_va_start): Likewise. Remove error for too many
569 (fold_builtin_next_arg): Issue error for too many arguments.
570 After checking arguments, replace them with magic arguments that
571 prevent further checking of the args.
573 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
575 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
577 2005-01-06 Roger Sayle <roger@eyesopen.com>
580 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
581 and %{.S:...} (and their negative variants) to test whether the
582 input file is assembler or pre-processed-assembler independent of
583 the actual filename extension.
585 2005-01-06 Roger Sayle <roger@eyesopen.com>
587 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
588 of sign or zero extended values.
590 2005-01-06 Geoffrey Keating <geoffk@apple.com>
592 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
593 for IBM long double format correctly.
595 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
597 Fix PR tree-optimization/18792
599 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
600 to first_loop_depth, and use loop depth instead of loop number.
601 (build_classic_dir_vector): Ditto.
602 (compute_data_dependences_for_loop): Use depth, not loop number.
603 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
604 number. Pass in loops, instead of loop numbers.
605 (gather_interchange_stats): Ditto.
606 (linear_transform_loops): Ditto.
608 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
611 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
612 biv_fits_mode_p, extension_within_bounds_p): New functions.
613 (check_ext_dependent_givs): Use them.
615 2005-01-06 Roger Sayle <roger@eyesopen.com>
617 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
618 more than one barrier, in which case we should delete them all.
620 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
622 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
624 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
626 Fix PR middle-end/19286
628 * dwarf2out.c (gen_subprogram_die): If we've already tried to
629 output this subprogram, simply ignore this attempt to do it again.
630 (add_abstract_origin_attribute): Don't abort trying to add the abstract
631 origin attribute if it's not possible.
632 (gen_block_die): Don't ignore subblocks of "unused" blocks.
633 (decls_for_scope): Ditto.
634 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
635 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
637 2005-01-05 Richard Henderson <rth@redhat.com>
640 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
641 (bdesc_2arg): Use it.
642 (ix86_expand_binop_builtin): Force operands into registers
644 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
645 ix86_expand_sse_compare, ix86_expand_sse_comi,
646 ix86_expand_builtin): Likewise.
648 2005-01-05 Richard Henderson <rth@redhat.com>
650 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
652 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
653 (mulv4hi3): Set itanium_class mmmul.
654 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
655 (fpack): Rename from fpack_sfsf.
657 2005-01-05 Richard Henderson <rth@redhat.com>
660 * reload1.c (do_input_reload): Restrict the optimization deleteing
661 a previous output reload to RELOAD_FOR_INPUT.
663 2005-01-05 Steven Bosscher <stevenb@suse.de>
665 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
666 comment that we fall through after case.
667 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
668 (distribute_links): Likewise.
669 * cse.c (cse_insn): Likewise.
670 * cselib.c (cselib_invalidate_mem): Likewise.
671 * df.c: Update comments at the top of the file.
672 (read_modify_subreg_p): Update comments here too.
673 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
674 * flow.c (mark_set_1): Likewise.
675 (mark_used_regs): Likewise.
676 * gcse.c (mems_conflict_for_gcse_p): Likewise.
677 (canon_list_insert): Likewise.
678 (mark_set): Likewise.
679 (try_replace_reg): Likewise.
680 (store_killed_in_insn): Likewise.
681 * loop.c (count_one_set): Likewise.
682 (basic_induction_var): Likewise.
683 * postreload-gcse.c (find_mem_conflicts): Likewise.
684 * postreload.c (reload_combine_note_store): Likewise.
685 (move2add_note_store): Likewise.
686 * reload.c (find_equiv_reg): Likewise.
687 (mark_referenced_resources): Likewise.
688 * rtlanal.c (set_noop_p): Likewise.
689 (note_stores): Likewise.
690 (note_uses): Likewise.
691 * sched-deps.c (sched_analyze_1): Likewise.
692 * sched-rgn.c (check_live_1): Likewise.
693 (update_live_1): Likewise.
694 * config/i860/i860.c: Likewise.
696 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
698 (ZERO_EXTRACT): Mention that this one can be an lvalue.
700 * doc/rtl.texi: Update documentation for bit-fields and SET.
702 2005-01-05 Roger Sayle <roger@eyesopen.com>
704 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
706 2005-01-05 Jan Hubicka <jh@suse.cz>
707 Richard Henderson <rth@redhat.com>
710 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
713 2005-01-05 Richard Henderson <rth@redhat.com>
716 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
717 (sse_shufps): Change operand 3 to const_int_operand.
718 (sse2_storelps): Fix typo in template.
719 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
720 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
721 (ix86_expand_builtin): Likewise.
723 2005-01-05 Julian Brown <julian@codesourcery.com>
725 * config/arm/arm.c (arm_return_in_memory): Treat complex types
726 as aggregates for AAPCS ABIs.
728 2005-01-05 Stan Shebs <shebs@apple.com>
730 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
731 Mach-O getters if ppc64.
732 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
734 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
735 signed four-byte field for global code case.
736 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
738 2005-01-05 Roger Sayle <roger@eyesopen.com>
741 * c-common.c: Include real.h.
742 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
743 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
744 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
745 * Makefile.in (c-common.o): Update dependencies.
747 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
749 * c-parse.in (asm_string): Add trailing semicolon.
751 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
753 * c-parse.in (asm_string): New. Don't allow wide strings in
755 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
756 asm_string instead of STRING.
758 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
760 * c-typeck.c (constructor_no_implicit): Remove.
761 (set_designator, process_init_element): Don't check
762 constructor_no_implicit.
764 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
765 Kaz Kojima <kkojima@gcc.gnu.org>
768 * lcm.c (create_pre_exit): New.
769 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
770 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
772 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
774 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
775 -msoft-float in target_float_switch.
776 * arm.c (arm_override_options): Fix processing of target_float_switch.
778 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
780 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
783 2004-01-05 Uros Bizjak <uros@kss-loka.si>
785 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
786 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
788 2005-01-04 Roger Sayle <roger@eyesopen.com>
790 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
791 if deleting the then-block allows the test-block to fallthru to the
794 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
797 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
798 after the full declaration if the are in two different TUs.
800 2005-01-04 Richard Henderson <rth@redhat.com>
803 * tree-sra.c (generate_one_element_init): Just
804 call gimplify_and_add.
805 (generate_element_init): Record the
806 new referenced variables and mark them for renaming
808 (generate_element_init_1): This.
809 (scalarize_init): Don't call push_gimplify_context/
810 pop_gimplify_context.
812 2005-01-04 Geoffrey Keating <geoffk@apple.com>
814 * toplev.c (get_src_pwd): Handle failure of getpwd().
816 2005-01-04 Roger Sayle <roger@eyesopen.com>
818 * fold-const.c (fold_single_bit_test): Delete unreachable handling
820 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
821 all the cases handled by it are inverted by invert_truthvalue.
823 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
825 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
826 (save_register_windowsi): Likewise.
828 2005-01-04 Richard Henderson <rth@redhat.com>
830 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
832 2005-01-04 Uros Bizjak <uros@kss-loka.si>
835 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
836 any simplifications of vector mode comparison operators.
837 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
839 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
840 Devang Patel <dpatel@apple.com>
842 PR tree-optimization/18308
843 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
844 the operands before creating a new expression.
845 * dojump.c (do_jump): Make drop_through_label available
846 for all cases. Add expansion of COND_EXPR.
848 2005-01-04 Ira Rosen <irar@il.ibm.com>
850 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
851 INTEGER_CST instead of TREE_CONSTANT.
852 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
853 instead of TREE_CONSTANT.
854 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
855 instead of TREE_CONSTANT.
857 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
859 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
861 2005-01-04 Uros Bizjak <uros@kss-loka.si>
864 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
865 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
867 2005-01-03 Richard Henderson <rth@redhat.com>
869 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
870 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
871 fold_convert_const_real_from_real): Split out from ...
872 (fold_convert_const): ... here.
874 2005-01-03 Richard Henderson <rth@redhat.com>
877 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
878 (mov<MMXMODEI>_internal): Likewise.
879 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
880 Add V2SF case; use it for SSE1; don't use TI.
881 (movdf_integer): Likewise.
882 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
884 2005-01-03 Ira Rosen <irar@il.ibm.com>
886 * tree-vectorizer.c (vect_strip_conversions): New function.
887 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
888 check for binary class.
890 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
894 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
896 * gimple-low.c (mark_blocks_with_used_vars): New function.
897 (mark_blocks_with_used_subblocks): Ditto.
898 (mark_used_blocks): Ditto.
899 (pass_mark_used_blocks): New pass.
900 * tree-inline.c: Include debug.h.
901 (expand_call_inline): Call outlining_inline_function here.
902 * tree-optimize.c (init_tree_optimization_passes): Add
903 pass_mark_used_blocks.
904 * tree-pass.h (pass_mark_used_blocks): New.
905 * Makefile.in (tree-inline.o): Add debug.h dependency.
907 2005-01-03 Geoffrey Keating <geoffk@apple.com>
909 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
910 VAR_DECLs to have weak_import attribute.
912 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
915 2004-01-03 Steven Bosscher <stevenb@suse.de>
917 * passes.c (rest_of_compilation): Don't run regmove if only
918 flag_expensive_optimizations, require flag_regmove instead.
920 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
922 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
924 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
925 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
926 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
927 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
929 2005-01-03 Richard Henderson <rth@redhat.com>
931 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
932 (ia64_const_ok_for_letter_p): New.
933 (ia64_const_double_ok_for_letter_p): New.
934 (ia64_extra_constraint): New.
935 (ia64_expand_vecint_compare): New.
936 (ia64_expand_vcondu_v2si): New.
937 (ia64_expand_vecint_cmov): New.
938 (ia64_expand_vecint_minmax): New.
939 (ia64_print_operand): Add 'v'.
940 (ia64_preferred_reload_class): New.
941 (ia64_vector_mode_supported_p): New.
942 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
943 (PREFERRED_RELOAD_CLASS): Move to function.
944 (CONST_OK_FOR_LETTER_P): Move to function.
945 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
946 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
947 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
948 (EXTRA_CONSTRAINT): Move to function.
949 * config/ia64/ia64.md: Include vect.md.
950 (itanium_class): Add mmalua.
952 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
954 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
955 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
957 (const_int_2bit_operand): New.
958 (fr_reg_or_0_operand): New.
959 * config/ia64/ia64-modes.def: Add vector modes.
960 * config/ia64/ia64-protos.h: Update.
961 * config/ia64/vect.md: New file.
963 2005-01-03 Richard Henderson <rth@redhat.com>
965 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
967 2005-01-03 Uros Bizjak <uros@kss-loka.si>
970 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
972 (log1pdf2): Change mode of operands[0,1] to DFmode.
974 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
976 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
977 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
978 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
979 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
980 SPARC_RELAXED_ORDERING.
982 2005-01-03 Richard Henderson <rth@redhat.com>
983 Uros Bizjak <uros@kss-loka.si>
986 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
987 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
988 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
990 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
991 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
992 const_pow2_1_to_128_operand): New predicates.
994 2005-01-02 Greg McGary <greg@mcgary.org>
996 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
998 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
999 object. [ptr_type]: Remove unused variable.
1000 Remember nearest addressable array-element or record-component
1001 when checking bitfield components. Tolerate empty BIND_EXPR.
1003 2005-01-01 Richard Henderson <rth@redhat.com>
1005 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
1006 that don't narrow the value. Fail for other conversions.
1008 2005-01-01 Richard Henderson <rth@redhat.com>
1011 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
1012 * c-lang.c (finish_file): Don't do it here.
1013 * objc/objc-act.c (objc_finish_file): Likewise.
1015 * cgraph.c (decl_assembler_name_equal): New.
1016 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
1017 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
1018 * cgraph.h (struct cgraph_varpool_node): Add next.
1019 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
1020 * varasm.c (assemble_alias): Mark the target as needed.
1022 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
1025 * function.c (get_arg_pointer_save_area): Use entry_of_function
1026 instead of get_insns.
1028 2005-01-01 Roger Sayle <roger@eyesopen.com>
1029 Andrew Pinski <pinskia@physics.uc.edu>
1030 James E. Wilson <wilson@specifixinc.com>
1032 PR rtl-optimization/12092
1033 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
1036 2005-01-01 Roger Sayle <roger@eyesopen.com>
1037 Olivier Hainque <hainque@act-europe.fr>
1039 * tree.c (int_fits_type_p): A narrower type always fits in a
1040 wider one, except for negative values into unsigned types.
1042 2005-01-01 Roger Sayle <roger@eyesopen.com>
1044 * tree.c (int_fits_type_p): Compare the result of force_fit_type
1045 with the original constant rather than require TREE_OVERFLOW.
1047 2005-01-01 Steven Bosscher <stevenb@suse.de>
1050 * c-decl.c (finish_function): If compiling C99, annotate the
1051 compiler generated return with the current file name and line 0.
1052 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
1053 the source line is greater than 0.
1054 (remove_bb): Likewise.
1056 See ChangeLog.12 for earlier changes.