1 2005-01-12 Nick Clifton <nickc@redhat.com>
3 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
4 an "in_delay_slot" attribute of "no" to prevent them being used in
5 delay slots. This is forbidden because they might generate
8 2005-01-12 Alan Modra <amodra@bigpond.net.au>
11 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
14 2005-01-12 Nick Clifton <nickc@redhat.com>
16 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
17 passed on to GAS) but do not define, as GAS no longer supports or
18 needs the -m2000 option.
20 2005-01-11 Roger Sayle <roger@eyesopen.com>
22 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
23 type as args of unprototyped builtins aren't automatically promoted.
24 (fold_builtin_abs): Likewise.
26 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
28 * Makefile.in: Set a `build-warn' variable.
30 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
32 PR tree-optimization/17949
33 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
34 (find_interesting_uses_address): Use it.
36 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
38 * regrename.c (kill_value): Handle subreg's that won't simplify.
40 2005-01-11 Richard Henderson <rth@redhat.com>
43 * config/i386/i386.h (enum ix86_builtins): Move ...
44 * config/i386/i386.c: ... here.
45 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
46 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
47 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
48 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
49 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
50 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
51 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
52 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
53 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
54 IX86_BUILTIN_LOADDDUP): Remove.
55 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
56 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
57 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
58 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
59 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
60 (ix86_init_builtins): Make static.
61 (ix86_init_mmx_sse_builtins): Update for changed builtins.
62 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
63 if all the modes match. Otherwise, fake it.
64 (get_element_number, ix86_expand_vec_init_builtin,
65 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
66 (ix86_expand_builtin): Make static. Update for changed builtins.
67 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
68 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
69 (ix86_expand_vector_init_duplicate): New.
70 (ix86_expand_vector_init_low_nonzero): New.
71 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
72 Split out from ix86_expand_vector_init; handle integer modes.
73 (ix86_expand_vector_init): Use them.
74 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
75 * config/i386/i386-protos.h: Update.
76 * config/i386/predicates.md (reg_or_0_operand): New.
77 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
78 (movv2sf_internal): Likewise. And a splitter to match them all.
79 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
80 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
81 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
82 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
83 (mmx_pinsrw): Fix operand ordering.
84 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
85 rather than sse_loadss expander.
86 (movv2df splitter): Similarly.
87 (sse_loadss, sse_loadlss): Remove.
88 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
89 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
90 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
91 (sse3_movddup): Rename with '*'.
92 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
93 (sse2_loadsd): Remove.
94 (vec_dupv2df_sse3): Rename from sse3_loadddup.
95 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
96 (sse2_pinsrw): Fix argument ordering.
97 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
98 (sse2_stored): Remove 'r' destination.
99 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
100 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
101 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
102 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
103 vec_extractv16qi, vec_initv16qi): New.
105 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
106 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
107 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
109 (_mm_load_pd, _mm_store_pd): Use plain dereference.
110 (_mm_load_si128, _mm_store_si128): Likewise.
111 (_mm_load1_pd): Use _mm_set1_pd.
112 (_mm_load_sd): Use _mm_set_sd.
113 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
114 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
115 (_mm_set_epi64): Use _mm_set_epi64x.
116 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
117 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
118 _mm_setr_epi8): Use _mm_set_foo form.
119 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
120 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
121 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
122 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
123 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
124 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
125 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
126 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
127 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
128 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
129 (_mm_movedup_pd): Use _mm_shuffle_pd.
130 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
131 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
132 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
133 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
135 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
136 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
137 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
138 (_mm_store_ps): Use raw dereference.
139 (_mm_store1_ps): Use _mm_storeu_ps.
140 (_mm_storer_ps): Use _mm_store_ps.
141 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
142 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
144 2005-01-11 Stan Shebs <shebs@apple.com>
146 * config/rs6000/rs6000.c (machopic_output_stub): Issue
147 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
149 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
152 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
153 Skip all DECLs except for FIELD_DECLs.
155 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
157 * config/s390/s390.c (override_options): Return error if
158 -mbackchain, -mpacked-stack and -mhard-float are used together.
159 (s390_va_start): Remove the backchain && packed-stack special case.
160 (s390_gimplify_va_arg): Likewise.
161 * doc/invoke.texi: Remove the ABI incompatibility note.
163 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
165 * config/s390/s390.c (struct s390_frame_layout): Remove
167 (s390_frame_info, s390_emit_prologue): Replace occurrences of
168 save_backchain_p with TARGET_BACKCHAIN.
170 2005-01-11 Alan Modra <amodra@bigpond.net.au>
173 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
174 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
175 * function.c (locate_and_pad_parm): Set new field.
176 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
177 Tweak where_pad test to include "none". Always set mem align for
179 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
180 not sufficient for type.
181 (assign_parm_setup_block): If stack_parm is zero on entry, always
182 make a new stack local. Block move old stack parm if necessary
183 to new aligned stack local.
184 (assign_parm_setup_stack): Use a block move to handle
185 potentially misaligned entry_parm.
186 (assign_parms_unsplit_complex): Specify required alignment when
187 creating stack local.
188 * calls.c (compute_argument_addresses): Override alignment of stack
189 arg calculated from its type with the alignment given by
190 FUNCTION_ARG_BOUNDARY.
191 (store_one_arg): Likewise.
193 2005-01-11 Jan Beulich <jbeulich@novell.com>
195 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
196 Change respective itanium_class attribute to ialu.
197 (shladdp4_internal): New.
198 * config/ia64/predicates.md (shladd_log2_operand): New.
200 2005-01-11 Richard Henderson <rth@redhat.com>
202 * expr.c (store_constructor): Use rtvec_alloc instead of
203 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
205 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
207 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
209 2005-01-11 Alan Modra <amodra@bigpond.net.au>
211 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
212 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
213 (TARGET_RWRELOC): Define.
214 (ia64_rwreloc_section_type_flags): Delete.
215 (ia64_section_type_flags): New function.
216 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
217 (TARGET_RWRELOC): Define.
219 2005-01-10 David Mosberger <davidm@hpl.hp.com>
222 * config/ia64/ia64.c (process_set): For alloc insn, only call
223 process_epilogue is !frame_pointer_needed.
225 2005-01-10 Roger Sayle <roger@eyesopen.com>
228 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
229 unary operator and can't be treated as a binary/comparison operator.
231 2005-01-10 Richard Henderson <rth@redhat.com>
233 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
234 (ix86_return_in_memory): Likewise.
235 (function_arg_advance): Likewise. Mirror structure in function_arg
236 for choosing register to advance.
238 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
240 * tree-vectorizer.c, tree.def: Fix comment typos.
242 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
244 * config/s390/s390.c (struct processor_costs): 4 new fields:
245 ddbr, ddr, debr, der.
246 (s390_rtx_costs): More precise handling of divide instructions.
248 2005-01-10 David Edelsohn <edelsohn@gnu.org>
250 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
252 2005-01-10 Jan Beulich <jbeulich@novell.com>
254 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
255 names resulting from -ffunction-sections/-fdata-sections and linkonce
257 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
258 based on the section name. Rearrange the section name comparison logic
259 slightly so that each section name is compared against at most once.
261 2005-01-10 Ben Elliston <bje@au.ibm.com>
263 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
264 targets for which -fPIC reduces limitations on the GOT size.
266 2005-01-09 Falk Hueffner <falk@debian.org>
268 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
271 2005-01-09 David Edelsohn <edelsohn@gnu.org>
274 * collect2.c (main): Set aixrtl_flag for -brtl option.
275 (resolve_lib_name): Search for .so file extension before .a
278 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
280 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
281 (vect_analyze_data_refs_alignment): Add dump prints.
283 2005-01-09 Ira Rosen <irar@il.ibm.com>
285 * tree-vectorizer.c (vect_analyze_offset_expr): Use
286 expr_invariant_in_loop_p.
287 Initialize outputs first thing in the function.
288 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
289 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
290 (vect_analyze_pointer_ref_access): Check that the initial condition of
291 the access function is loop invariant.
293 2005-01-09 Richard Henderson <rth@redhat.com>
295 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
296 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
297 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
298 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
299 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
300 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
301 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
302 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
303 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
304 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
305 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
306 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
307 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
308 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
309 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
310 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
311 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
312 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
313 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
314 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
315 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
316 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
317 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
319 (mmx_clrdi, pavgusb): Remove.
320 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
321 with leading sse_ prefix.
322 * config/i386/sse.md: Receive them.
323 * config/i386/mmx.md: New file.
324 (MMXMODE12, MMXMODE24, mmxvecsize): New.
325 (subrv2sf3): Turn into expander for normal subtraction.
326 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
327 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
328 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
329 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
330 commutative; use ix86_binary_operator_ok.
331 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
332 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
333 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
334 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
335 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
336 Macroize from existing patterns; use ix86_binary_operator_ok.
337 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
338 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
339 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
340 with vec_select+vec_concat.
341 (mmx_pshufw, mmx_pshufw_1): Likewise.
342 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
343 (mmx_uavgv4hi3): Model correctly.
344 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
345 _mm_xor_si64): Remove casts.
347 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
349 PR tree-optimization/19224
350 * tree-scalar-evolution.c (get_instantiated_value,
351 set_instantiated_value): New functions.
352 (instantiate_parameters_1): Cache the results.
353 (instantiate_parameters, resolve_mixers): Initialize and free
356 2005-01-08 David Edelsohn <edelsohn@gnu.org>
358 * config/i386/i386.md (addhi_4): Correct reference in comment.
361 2005-01-08 Richard Henderson <rth@redhat.com>
363 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
364 (_mm_cvtsi128_si64x): Likewise.
365 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
366 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
367 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
368 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
369 * config/i386/i386-protos.h: Update.
370 * config/i386/i386.c (print_operand): Add 'H'.
371 (ix86_fixup_binary_operands): Split out from ...
372 (ix86_expand_binary_operator): ... here.
373 (ix86_fixup_binary_operands_no_copy): New.
374 (ix86_expand_fp_absneg_operator): Handle vector mode results.
375 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
376 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
377 (safe_vector_operand): Use CONST0_RTX.
378 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
379 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
380 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
381 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
383 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
384 (IX86_BUILTIN_CMPNGESS): New.
385 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
386 (attr type): Add sselog1.
387 (attr unit, attr memory): Handle it.
388 (movti, movti_internal, movti_rex64): Move near other integer moves.
389 (movtf, movtf_internal): Move near other fp moves.
390 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
391 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
392 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
393 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
394 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
395 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
396 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
397 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
398 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
399 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
400 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
401 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
402 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
403 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
404 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
405 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
406 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
407 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
408 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
409 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
410 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
411 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
412 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
413 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
414 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
415 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
416 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
417 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
418 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
419 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
420 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
421 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
422 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
423 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
424 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
425 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
426 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
427 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
428 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
429 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
430 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
431 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
432 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
433 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
435 (SSEPUSH, push<SSEPUSH>): Remove.
436 (MMXPUSH, push<MMXPUSH>): Remove.
437 (sse_movaps, sse_movaps_1, sse_movups): Remove.
438 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
439 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
440 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
441 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
442 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
443 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
444 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
445 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
446 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
447 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
448 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
449 (ppro_sse_log_V4SF): Handle sselog1.
450 * config/i386/predicates.md (const_0_to_1_operand): New.
451 (const_0_to_255_mul_8_operand): New.
452 (const_1_to_31_operand): Rename from const_int_1_31_operand.
453 (const_2_to_3_operand, const_4_to_7_operand): New.
454 * config/i386/sse.md: New file.
455 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
456 (sse_movups): Rename from sse_movups_1.
457 (sse_loadlss): Rename from sse_loadss_1.
458 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
460 (negv4sf2): Use ix86_expand_fp_absneg_operator.
461 (absv4sf2, negv2df, absv2df): New.
462 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
463 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
464 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
465 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
466 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
467 (sse3_addsubv4sf3): Model correctly.
468 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
469 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
470 (sse_movhlps): Model with vec_select+vec_concat.
471 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
472 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
473 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
474 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
475 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
476 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
477 sse2_pshufhw_1): Likewise.
478 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
479 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
480 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
481 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
482 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
483 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
484 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
485 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
486 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
487 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
488 sminv8hi3): Mark commutative
489 operands. Use ix86_binary_operator_ok.
490 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
491 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
492 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
493 sse2_punpcklqdq): Allow operand2 in memory.
494 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
495 sse2_movsd): Add memory alternatives.
496 (sse_storelps): Turn expander into an insn; split after reload.
497 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
498 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
500 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
502 * configure.ac (DWARF-2 debug_line): Use objdump.
503 * configure: Regenerate.
505 2005-01-08 Jeff Law <law@redhat.com>
506 Diego Novillo <dnovillo@redhat.com>
508 PR tree-optimization/18241
509 * tree-nrv.c (tree_nrv): Ignore volatile return values.
510 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
511 statements with volatile operands.
512 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
513 operands after marking a statement with has_volatile_ops.
515 2005-01-08 Roger Sayle <roger@eyesopen.com>
517 * tree.c (int_fits_type_p): Always honor integer constant
518 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
520 2005-01-08 Roger Sayle <roger@eyesopen.com>
522 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
523 the latent bug in rtl_delete_block has been resolved.
525 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
527 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
528 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
530 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
532 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
533 library as @multilib_dir@/$(SHLIB_SONAME).
535 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
537 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
538 * configure: Regenerate.
540 2005-01-07 Jakub Jelinek <jakub@redhat.com>
542 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
545 2005-01-07 David Edelsohn <edelsohn@gnu.org>
548 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
549 Convert non-word aligned offset address using ld/std into
552 2005-01-07 Richard Henderson <rth@redhat.com>
554 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
556 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
558 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
560 2005-01-07 Jakub Jelinek <jakub@redhat.com>
562 PR tree-optimization/19283
563 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
564 is not shorter than the original type.
566 PR rtl-optimization/19012
567 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
569 PR rtl-optimization/18861
570 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
571 (FULL_STATE): Define.
572 (mark_tablejump_edge): New function.
573 (purge_dead_tablejump_edges): New function.
574 (find_bb_boundaries): Use it.
576 PR tree-optimization/18828
577 * builtins.c (expand_builtin_next_arg): Remove argument and all
578 the argument checking.
579 (expand_builtin): Adjust caller.
580 (expand_builtin_va_start): Likewise. Remove error for too many
582 (fold_builtin_next_arg): Issue error for too many arguments.
583 After checking arguments, replace them with magic arguments that
584 prevent further checking of the args.
586 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
588 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
590 2005-01-06 Roger Sayle <roger@eyesopen.com>
593 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
594 and %{.S:...} (and their negative variants) to test whether the
595 input file is assembler or pre-processed-assembler independent of
596 the actual filename extension.
598 2005-01-06 Roger Sayle <roger@eyesopen.com>
600 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
601 of sign or zero extended values.
603 2005-01-06 Geoffrey Keating <geoffk@apple.com>
605 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
606 for IBM long double format correctly.
608 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
610 Fix PR tree-optimization/18792
612 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
613 to first_loop_depth, and use loop depth instead of loop number.
614 (build_classic_dir_vector): Ditto.
615 (compute_data_dependences_for_loop): Use depth, not loop number.
616 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
617 number. Pass in loops, instead of loop numbers.
618 (gather_interchange_stats): Ditto.
619 (linear_transform_loops): Ditto.
621 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
624 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
625 biv_fits_mode_p, extension_within_bounds_p): New functions.
626 (check_ext_dependent_givs): Use them.
628 2005-01-06 Roger Sayle <roger@eyesopen.com>
630 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
631 more than one barrier, in which case we should delete them all.
633 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
635 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
637 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
639 Fix PR middle-end/19286
641 * dwarf2out.c (gen_subprogram_die): If we've already tried to
642 output this subprogram, simply ignore this attempt to do it again.
643 (add_abstract_origin_attribute): Don't abort trying to add the abstract
644 origin attribute if it's not possible.
645 (gen_block_die): Don't ignore subblocks of "unused" blocks.
646 (decls_for_scope): Ditto.
647 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
648 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
650 2005-01-05 Richard Henderson <rth@redhat.com>
653 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
654 (bdesc_2arg): Use it.
655 (ix86_expand_binop_builtin): Force operands into registers
657 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
658 ix86_expand_sse_compare, ix86_expand_sse_comi,
659 ix86_expand_builtin): Likewise.
661 2005-01-05 Richard Henderson <rth@redhat.com>
663 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
665 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
666 (mulv4hi3): Set itanium_class mmmul.
667 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
668 (fpack): Rename from fpack_sfsf.
670 2005-01-05 Richard Henderson <rth@redhat.com>
673 * reload1.c (do_input_reload): Restrict the optimization deleteing
674 a previous output reload to RELOAD_FOR_INPUT.
676 2005-01-05 Steven Bosscher <stevenb@suse.de>
678 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
679 comment that we fall through after case.
680 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
681 (distribute_links): Likewise.
682 * cse.c (cse_insn): Likewise.
683 * cselib.c (cselib_invalidate_mem): Likewise.
684 * df.c: Update comments at the top of the file.
685 (read_modify_subreg_p): Update comments here too.
686 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
687 * flow.c (mark_set_1): Likewise.
688 (mark_used_regs): Likewise.
689 * gcse.c (mems_conflict_for_gcse_p): Likewise.
690 (canon_list_insert): Likewise.
691 (mark_set): Likewise.
692 (try_replace_reg): Likewise.
693 (store_killed_in_insn): Likewise.
694 * loop.c (count_one_set): Likewise.
695 (basic_induction_var): Likewise.
696 * postreload-gcse.c (find_mem_conflicts): Likewise.
697 * postreload.c (reload_combine_note_store): Likewise.
698 (move2add_note_store): Likewise.
699 * reload.c (find_equiv_reg): Likewise.
700 (mark_referenced_resources): Likewise.
701 * rtlanal.c (set_noop_p): Likewise.
702 (note_stores): Likewise.
703 (note_uses): Likewise.
704 * sched-deps.c (sched_analyze_1): Likewise.
705 * sched-rgn.c (check_live_1): Likewise.
706 (update_live_1): Likewise.
707 * config/i860/i860.c: Likewise.
709 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
711 (ZERO_EXTRACT): Mention that this one can be an lvalue.
713 * doc/rtl.texi: Update documentation for bit-fields and SET.
715 2005-01-05 Roger Sayle <roger@eyesopen.com>
717 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
719 2005-01-05 Jan Hubicka <jh@suse.cz>
720 Richard Henderson <rth@redhat.com>
723 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
726 2005-01-05 Richard Henderson <rth@redhat.com>
729 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
730 (sse_shufps): Change operand 3 to const_int_operand.
731 (sse2_storelps): Fix typo in template.
732 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
733 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
734 (ix86_expand_builtin): Likewise.
736 2005-01-05 Julian Brown <julian@codesourcery.com>
738 * config/arm/arm.c (arm_return_in_memory): Treat complex types
739 as aggregates for AAPCS ABIs.
741 2005-01-05 Stan Shebs <shebs@apple.com>
743 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
744 Mach-O getters if ppc64.
745 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
747 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
748 signed four-byte field for global code case.
749 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
751 2005-01-05 Roger Sayle <roger@eyesopen.com>
754 * c-common.c: Include real.h.
755 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
756 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
757 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
758 * Makefile.in (c-common.o): Update dependencies.
760 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
762 * c-parse.in (asm_string): Add trailing semicolon.
764 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
766 * c-parse.in (asm_string): New. Don't allow wide strings in
768 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
769 asm_string instead of STRING.
771 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
773 * c-typeck.c (constructor_no_implicit): Remove.
774 (set_designator, process_init_element): Don't check
775 constructor_no_implicit.
777 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
778 Kaz Kojima <kkojima@gcc.gnu.org>
781 * lcm.c (create_pre_exit): New.
782 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
783 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
785 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
787 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
788 -msoft-float in target_float_switch.
789 * arm.c (arm_override_options): Fix processing of target_float_switch.
791 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
793 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
796 2004-01-05 Uros Bizjak <uros@kss-loka.si>
798 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
799 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
801 2005-01-04 Roger Sayle <roger@eyesopen.com>
803 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
804 if deleting the then-block allows the test-block to fallthru to the
807 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
810 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
811 after the full declaration if the are in two different TUs.
813 2005-01-04 Richard Henderson <rth@redhat.com>
816 * tree-sra.c (generate_one_element_init): Just
817 call gimplify_and_add.
818 (generate_element_init): Record the
819 new referenced variables and mark them for renaming
821 (generate_element_init_1): This.
822 (scalarize_init): Don't call push_gimplify_context/
823 pop_gimplify_context.
825 2005-01-04 Geoffrey Keating <geoffk@apple.com>
827 * toplev.c (get_src_pwd): Handle failure of getpwd().
829 2005-01-04 Roger Sayle <roger@eyesopen.com>
831 * fold-const.c (fold_single_bit_test): Delete unreachable handling
833 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
834 all the cases handled by it are inverted by invert_truthvalue.
836 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
838 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
839 (save_register_windowsi): Likewise.
841 2005-01-04 Richard Henderson <rth@redhat.com>
843 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
845 2005-01-04 Uros Bizjak <uros@kss-loka.si>
848 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
849 any simplifications of vector mode comparison operators.
850 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
852 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
853 Devang Patel <dpatel@apple.com>
855 PR tree-optimization/18308
856 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
857 the operands before creating a new expression.
858 * dojump.c (do_jump): Make drop_through_label available
859 for all cases. Add expansion of COND_EXPR.
861 2005-01-04 Ira Rosen <irar@il.ibm.com>
863 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
864 INTEGER_CST instead of TREE_CONSTANT.
865 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
866 instead of TREE_CONSTANT.
867 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
868 instead of TREE_CONSTANT.
870 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
872 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
874 2005-01-04 Uros Bizjak <uros@kss-loka.si>
877 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
878 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
880 2005-01-03 Richard Henderson <rth@redhat.com>
882 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
883 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
884 fold_convert_const_real_from_real): Split out from ...
885 (fold_convert_const): ... here.
887 2005-01-03 Richard Henderson <rth@redhat.com>
890 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
891 (mov<MMXMODEI>_internal): Likewise.
892 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
893 Add V2SF case; use it for SSE1; don't use TI.
894 (movdf_integer): Likewise.
895 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
897 2005-01-03 Ira Rosen <irar@il.ibm.com>
899 * tree-vectorizer.c (vect_strip_conversions): New function.
900 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
901 check for binary class.
903 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
907 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
909 * gimple-low.c (mark_blocks_with_used_vars): New function.
910 (mark_blocks_with_used_subblocks): Ditto.
911 (mark_used_blocks): Ditto.
912 (pass_mark_used_blocks): New pass.
913 * tree-inline.c: Include debug.h.
914 (expand_call_inline): Call outlining_inline_function here.
915 * tree-optimize.c (init_tree_optimization_passes): Add
916 pass_mark_used_blocks.
917 * tree-pass.h (pass_mark_used_blocks): New.
918 * Makefile.in (tree-inline.o): Add debug.h dependency.
920 2005-01-03 Geoffrey Keating <geoffk@apple.com>
922 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
923 VAR_DECLs to have weak_import attribute.
925 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
928 2004-01-03 Steven Bosscher <stevenb@suse.de>
930 * passes.c (rest_of_compilation): Don't run regmove if only
931 flag_expensive_optimizations, require flag_regmove instead.
933 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
935 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
937 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
938 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
939 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
940 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
942 2005-01-03 Richard Henderson <rth@redhat.com>
944 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
945 (ia64_const_ok_for_letter_p): New.
946 (ia64_const_double_ok_for_letter_p): New.
947 (ia64_extra_constraint): New.
948 (ia64_expand_vecint_compare): New.
949 (ia64_expand_vcondu_v2si): New.
950 (ia64_expand_vecint_cmov): New.
951 (ia64_expand_vecint_minmax): New.
952 (ia64_print_operand): Add 'v'.
953 (ia64_preferred_reload_class): New.
954 (ia64_vector_mode_supported_p): New.
955 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
956 (PREFERRED_RELOAD_CLASS): Move to function.
957 (CONST_OK_FOR_LETTER_P): Move to function.
958 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
959 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
960 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
961 (EXTRA_CONSTRAINT): Move to function.
962 * config/ia64/ia64.md: Include vect.md.
963 (itanium_class): Add mmalua.
965 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
967 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
968 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
970 (const_int_2bit_operand): New.
971 (fr_reg_or_0_operand): New.
972 * config/ia64/ia64-modes.def: Add vector modes.
973 * config/ia64/ia64-protos.h: Update.
974 * config/ia64/vect.md: New file.
976 2005-01-03 Richard Henderson <rth@redhat.com>
978 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
980 2005-01-03 Uros Bizjak <uros@kss-loka.si>
983 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
985 (log1pdf2): Change mode of operands[0,1] to DFmode.
987 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
989 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
990 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
991 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
992 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
993 SPARC_RELAXED_ORDERING.
995 2005-01-03 Richard Henderson <rth@redhat.com>
996 Uros Bizjak <uros@kss-loka.si>
999 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
1000 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
1001 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
1003 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
1004 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
1005 const_pow2_1_to_128_operand): New predicates.
1007 2005-01-02 Greg McGary <greg@mcgary.org>
1009 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
1011 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
1012 object. [ptr_type]: Remove unused variable.
1013 Remember nearest addressable array-element or record-component
1014 when checking bitfield components. Tolerate empty BIND_EXPR.
1016 2005-01-01 Richard Henderson <rth@redhat.com>
1018 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
1019 that don't narrow the value. Fail for other conversions.
1021 2005-01-01 Richard Henderson <rth@redhat.com>
1024 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
1025 * c-lang.c (finish_file): Don't do it here.
1026 * objc/objc-act.c (objc_finish_file): Likewise.
1028 * cgraph.c (decl_assembler_name_equal): New.
1029 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
1030 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
1031 * cgraph.h (struct cgraph_varpool_node): Add next.
1032 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
1033 * varasm.c (assemble_alias): Mark the target as needed.
1035 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
1038 * function.c (get_arg_pointer_save_area): Use entry_of_function
1039 instead of get_insns.
1041 2005-01-01 Roger Sayle <roger@eyesopen.com>
1042 Andrew Pinski <pinskia@physics.uc.edu>
1043 James E. Wilson <wilson@specifixinc.com>
1045 PR rtl-optimization/12092
1046 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
1049 2005-01-01 Roger Sayle <roger@eyesopen.com>
1050 Olivier Hainque <hainque@act-europe.fr>
1052 * tree.c (int_fits_type_p): A narrower type always fits in a
1053 wider one, except for negative values into unsigned types.
1055 2005-01-01 Roger Sayle <roger@eyesopen.com>
1057 * tree.c (int_fits_type_p): Compare the result of force_fit_type
1058 with the original constant rather than require TREE_OVERFLOW.
1060 2005-01-01 Steven Bosscher <stevenb@suse.de>
1063 * c-decl.c (finish_function): If compiling C99, annotate the
1064 compiler generated return with the current file name and line 0.
1065 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
1066 the source line is greater than 0.
1067 (remove_bb): Likewise.
1069 See ChangeLog.12 for earlier changes.