1 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
5 * reload1.c (delete_output_reload): Don't delete an output reload
6 if the pseudo lives longer than a single basic block. Adjust and
7 improve wording of comment.
9 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
12 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
13 of TYPE_SIZE where we mean the number of bytes.
15 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
17 * function.c (assign_parm_setup_block): Relax condition on
18 multi-register optimization.
20 2005-01-12 Nick Clifton <nickc@redhat.com>
22 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
23 an "in_delay_slot" attribute of "no" to prevent them being used in
24 delay slots. This is forbidden because they might generate
27 2005-01-12 Alan Modra <amodra@bigpond.net.au>
30 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
33 2005-01-12 Nick Clifton <nickc@redhat.com>
35 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
36 passed on to GAS) but do not define, as GAS no longer supports or
37 needs the -m2000 option.
39 2005-01-11 Roger Sayle <roger@eyesopen.com>
41 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
42 type as args of unprototyped builtins aren't automatically promoted.
43 (fold_builtin_abs): Likewise.
45 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
47 * Makefile.in: Set a `build-warn' variable.
49 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
51 PR tree-optimization/17949
52 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
53 (find_interesting_uses_address): Use it.
55 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
57 * regrename.c (kill_value): Handle subreg's that won't simplify.
59 2005-01-11 Richard Henderson <rth@redhat.com>
62 * config/i386/i386.h (enum ix86_builtins): Move ...
63 * config/i386/i386.c: ... here.
64 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
65 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
66 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
67 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
68 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
69 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
70 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
71 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
72 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
73 IX86_BUILTIN_LOADDDUP): Remove.
74 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
75 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
76 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
77 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
78 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
79 (ix86_init_builtins): Make static.
80 (ix86_init_mmx_sse_builtins): Update for changed builtins.
81 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
82 if all the modes match. Otherwise, fake it.
83 (get_element_number, ix86_expand_vec_init_builtin,
84 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
85 (ix86_expand_builtin): Make static. Update for changed builtins.
86 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
87 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
88 (ix86_expand_vector_init_duplicate): New.
89 (ix86_expand_vector_init_low_nonzero): New.
90 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
91 Split out from ix86_expand_vector_init; handle integer modes.
92 (ix86_expand_vector_init): Use them.
93 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
94 * config/i386/i386-protos.h: Update.
95 * config/i386/predicates.md (reg_or_0_operand): New.
96 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
97 (movv2sf_internal): Likewise. And a splitter to match them all.
98 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
99 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
100 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
101 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
102 (mmx_pinsrw): Fix operand ordering.
103 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
104 rather than sse_loadss expander.
105 (movv2df splitter): Similarly.
106 (sse_loadss, sse_loadlss): Remove.
107 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
108 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
109 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
110 (sse3_movddup): Rename with '*'.
111 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
112 (sse2_loadsd): Remove.
113 (vec_dupv2df_sse3): Rename from sse3_loadddup.
114 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
115 (sse2_pinsrw): Fix argument ordering.
116 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
117 (sse2_stored): Remove 'r' destination.
118 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
119 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
120 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
121 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
122 vec_extractv16qi, vec_initv16qi): New.
124 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
125 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
126 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
128 (_mm_load_pd, _mm_store_pd): Use plain dereference.
129 (_mm_load_si128, _mm_store_si128): Likewise.
130 (_mm_load1_pd): Use _mm_set1_pd.
131 (_mm_load_sd): Use _mm_set_sd.
132 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
133 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
134 (_mm_set_epi64): Use _mm_set_epi64x.
135 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
136 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
137 _mm_setr_epi8): Use _mm_set_foo form.
138 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
139 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
140 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
141 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
142 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
143 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
144 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
145 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
146 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
147 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
148 (_mm_movedup_pd): Use _mm_shuffle_pd.
149 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
150 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
151 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
152 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
154 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
155 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
156 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
157 (_mm_store_ps): Use raw dereference.
158 (_mm_store1_ps): Use _mm_storeu_ps.
159 (_mm_storer_ps): Use _mm_store_ps.
160 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
161 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
163 2005-01-11 Stan Shebs <shebs@apple.com>
165 * config/rs6000/rs6000.c (machopic_output_stub): Issue
166 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
168 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
171 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
172 Skip all DECLs except for FIELD_DECLs.
174 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
176 * config/s390/s390.c (override_options): Return error if
177 -mbackchain, -mpacked-stack and -mhard-float are used together.
178 (s390_va_start): Remove the backchain && packed-stack special case.
179 (s390_gimplify_va_arg): Likewise.
180 * doc/invoke.texi: Remove the ABI incompatibility note.
182 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
184 * config/s390/s390.c (struct s390_frame_layout): Remove
186 (s390_frame_info, s390_emit_prologue): Replace occurrences of
187 save_backchain_p with TARGET_BACKCHAIN.
189 2005-01-11 Alan Modra <amodra@bigpond.net.au>
192 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
193 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
194 * function.c (locate_and_pad_parm): Set new field.
195 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
196 Tweak where_pad test to include "none". Always set mem align for
198 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
199 not sufficient for type.
200 (assign_parm_setup_block): If stack_parm is zero on entry, always
201 make a new stack local. Block move old stack parm if necessary
202 to new aligned stack local.
203 (assign_parm_setup_stack): Use a block move to handle
204 potentially misaligned entry_parm.
205 (assign_parms_unsplit_complex): Specify required alignment when
206 creating stack local.
207 * calls.c (compute_argument_addresses): Override alignment of stack
208 arg calculated from its type with the alignment given by
209 FUNCTION_ARG_BOUNDARY.
210 (store_one_arg): Likewise.
212 2005-01-11 Jan Beulich <jbeulich@novell.com>
214 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
215 Change respective itanium_class attribute to ialu.
216 (shladdp4_internal): New.
217 * config/ia64/predicates.md (shladd_log2_operand): New.
219 2005-01-11 Richard Henderson <rth@redhat.com>
221 * expr.c (store_constructor): Use rtvec_alloc instead of
222 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
224 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
226 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
228 2005-01-11 Alan Modra <amodra@bigpond.net.au>
230 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
231 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
232 (TARGET_RWRELOC): Define.
233 (ia64_rwreloc_section_type_flags): Delete.
234 (ia64_section_type_flags): New function.
235 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
236 (TARGET_RWRELOC): Define.
238 2005-01-10 David Mosberger <davidm@hpl.hp.com>
241 * config/ia64/ia64.c (process_set): For alloc insn, only call
242 process_epilogue is !frame_pointer_needed.
244 2005-01-10 Roger Sayle <roger@eyesopen.com>
247 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
248 unary operator and can't be treated as a binary/comparison operator.
250 2005-01-10 Richard Henderson <rth@redhat.com>
252 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
253 (ix86_return_in_memory): Likewise.
254 (function_arg_advance): Likewise. Mirror structure in function_arg
255 for choosing register to advance.
257 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
259 * tree-vectorizer.c, tree.def: Fix comment typos.
261 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
263 * config/s390/s390.c (struct processor_costs): 4 new fields:
264 ddbr, ddr, debr, der.
265 (s390_rtx_costs): More precise handling of divide instructions.
267 2005-01-10 David Edelsohn <edelsohn@gnu.org>
269 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
271 2005-01-10 Jan Beulich <jbeulich@novell.com>
273 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
274 names resulting from -ffunction-sections/-fdata-sections and linkonce
276 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
277 based on the section name. Rearrange the section name comparison logic
278 slightly so that each section name is compared against at most once.
280 2005-01-10 Ben Elliston <bje@au.ibm.com>
282 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
283 targets for which -fPIC reduces limitations on the GOT size.
285 2005-01-09 Falk Hueffner <falk@debian.org>
287 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
290 2005-01-09 David Edelsohn <edelsohn@gnu.org>
293 * collect2.c (main): Set aixrtl_flag for -brtl option.
294 (resolve_lib_name): Search for .so file extension before .a
297 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
299 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
300 (vect_analyze_data_refs_alignment): Add dump prints.
302 2005-01-09 Ira Rosen <irar@il.ibm.com>
304 * tree-vectorizer.c (vect_analyze_offset_expr): Use
305 expr_invariant_in_loop_p.
306 Initialize outputs first thing in the function.
307 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
308 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
309 (vect_analyze_pointer_ref_access): Check that the initial condition of
310 the access function is loop invariant.
312 2005-01-09 Richard Henderson <rth@redhat.com>
314 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
315 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
316 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
317 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
318 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
319 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
320 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
321 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
322 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
323 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
324 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
325 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
326 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
327 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
328 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
329 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
330 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
331 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
332 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
333 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
334 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
335 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
336 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
338 (mmx_clrdi, pavgusb): Remove.
339 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
340 with leading sse_ prefix.
341 * config/i386/sse.md: Receive them.
342 * config/i386/mmx.md: New file.
343 (MMXMODE12, MMXMODE24, mmxvecsize): New.
344 (subrv2sf3): Turn into expander for normal subtraction.
345 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
346 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
347 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
348 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
349 commutative; use ix86_binary_operator_ok.
350 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
351 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
352 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
353 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
354 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
355 Macroize from existing patterns; use ix86_binary_operator_ok.
356 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
357 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
358 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
359 with vec_select+vec_concat.
360 (mmx_pshufw, mmx_pshufw_1): Likewise.
361 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
362 (mmx_uavgv4hi3): Model correctly.
363 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
364 _mm_xor_si64): Remove casts.
366 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
368 PR tree-optimization/19224
369 * tree-scalar-evolution.c (get_instantiated_value,
370 set_instantiated_value): New functions.
371 (instantiate_parameters_1): Cache the results.
372 (instantiate_parameters, resolve_mixers): Initialize and free
375 2005-01-08 David Edelsohn <edelsohn@gnu.org>
377 * config/i386/i386.md (addhi_4): Correct reference in comment.
380 2005-01-08 Richard Henderson <rth@redhat.com>
382 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
383 (_mm_cvtsi128_si64x): Likewise.
384 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
385 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
386 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
387 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
388 * config/i386/i386-protos.h: Update.
389 * config/i386/i386.c (print_operand): Add 'H'.
390 (ix86_fixup_binary_operands): Split out from ...
391 (ix86_expand_binary_operator): ... here.
392 (ix86_fixup_binary_operands_no_copy): New.
393 (ix86_expand_fp_absneg_operator): Handle vector mode results.
394 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
395 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
396 (safe_vector_operand): Use CONST0_RTX.
397 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
398 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
399 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
400 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
402 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
403 (IX86_BUILTIN_CMPNGESS): New.
404 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
405 (attr type): Add sselog1.
406 (attr unit, attr memory): Handle it.
407 (movti, movti_internal, movti_rex64): Move near other integer moves.
408 (movtf, movtf_internal): Move near other fp moves.
409 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
410 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
411 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
412 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
413 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
414 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
415 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
416 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
417 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
418 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
419 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
420 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
421 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
422 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
423 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
424 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
425 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
426 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
427 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
428 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
429 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
430 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
431 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
432 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
433 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
434 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
435 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
436 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
437 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
438 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
439 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
440 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
441 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
442 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
443 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
444 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
445 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
446 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
447 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
448 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
449 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
450 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
451 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
452 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
454 (SSEPUSH, push<SSEPUSH>): Remove.
455 (MMXPUSH, push<MMXPUSH>): Remove.
456 (sse_movaps, sse_movaps_1, sse_movups): Remove.
457 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
458 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
459 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
460 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
461 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
462 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
463 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
464 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
465 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
466 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
467 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
468 (ppro_sse_log_V4SF): Handle sselog1.
469 * config/i386/predicates.md (const_0_to_1_operand): New.
470 (const_0_to_255_mul_8_operand): New.
471 (const_1_to_31_operand): Rename from const_int_1_31_operand.
472 (const_2_to_3_operand, const_4_to_7_operand): New.
473 * config/i386/sse.md: New file.
474 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
475 (sse_movups): Rename from sse_movups_1.
476 (sse_loadlss): Rename from sse_loadss_1.
477 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
479 (negv4sf2): Use ix86_expand_fp_absneg_operator.
480 (absv4sf2, negv2df, absv2df): New.
481 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
482 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
483 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
484 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
485 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
486 (sse3_addsubv4sf3): Model correctly.
487 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
488 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
489 (sse_movhlps): Model with vec_select+vec_concat.
490 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
491 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
492 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
493 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
494 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
495 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
496 sse2_pshufhw_1): Likewise.
497 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
498 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
499 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
500 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
501 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
502 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
503 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
504 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
505 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
506 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
507 sminv8hi3): Mark commutative
508 operands. Use ix86_binary_operator_ok.
509 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
510 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
511 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
512 sse2_punpcklqdq): Allow operand2 in memory.
513 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
514 sse2_movsd): Add memory alternatives.
515 (sse_storelps): Turn expander into an insn; split after reload.
516 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
517 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
519 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
521 * configure.ac (DWARF-2 debug_line): Use objdump.
522 * configure: Regenerate.
524 2005-01-08 Jeff Law <law@redhat.com>
525 Diego Novillo <dnovillo@redhat.com>
527 PR tree-optimization/18241
528 * tree-nrv.c (tree_nrv): Ignore volatile return values.
529 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
530 statements with volatile operands.
531 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
532 operands after marking a statement with has_volatile_ops.
534 2005-01-08 Roger Sayle <roger@eyesopen.com>
536 * tree.c (int_fits_type_p): Always honor integer constant
537 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
539 2005-01-08 Roger Sayle <roger@eyesopen.com>
541 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
542 the latent bug in rtl_delete_block has been resolved.
544 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
546 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
547 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
549 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
551 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
552 library as @multilib_dir@/$(SHLIB_SONAME).
554 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
556 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
557 * configure: Regenerate.
559 2005-01-07 Jakub Jelinek <jakub@redhat.com>
561 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
564 2005-01-07 David Edelsohn <edelsohn@gnu.org>
567 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
568 Convert non-word aligned offset address using ld/std into
571 2005-01-07 Richard Henderson <rth@redhat.com>
573 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
575 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
577 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
579 2005-01-07 Jakub Jelinek <jakub@redhat.com>
581 PR tree-optimization/19283
582 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
583 is not shorter than the original type.
585 PR rtl-optimization/19012
586 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
588 PR rtl-optimization/18861
589 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
590 (FULL_STATE): Define.
591 (mark_tablejump_edge): New function.
592 (purge_dead_tablejump_edges): New function.
593 (find_bb_boundaries): Use it.
595 PR tree-optimization/18828
596 * builtins.c (expand_builtin_next_arg): Remove argument and all
597 the argument checking.
598 (expand_builtin): Adjust caller.
599 (expand_builtin_va_start): Likewise. Remove error for too many
601 (fold_builtin_next_arg): Issue error for too many arguments.
602 After checking arguments, replace them with magic arguments that
603 prevent further checking of the args.
605 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
607 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
609 2005-01-06 Roger Sayle <roger@eyesopen.com>
612 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
613 and %{.S:...} (and their negative variants) to test whether the
614 input file is assembler or pre-processed-assembler independent of
615 the actual filename extension.
617 2005-01-06 Roger Sayle <roger@eyesopen.com>
619 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
620 of sign or zero extended values.
622 2005-01-06 Geoffrey Keating <geoffk@apple.com>
624 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
625 for IBM long double format correctly.
627 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
629 Fix PR tree-optimization/18792
631 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
632 to first_loop_depth, and use loop depth instead of loop number.
633 (build_classic_dir_vector): Ditto.
634 (compute_data_dependences_for_loop): Use depth, not loop number.
635 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
636 number. Pass in loops, instead of loop numbers.
637 (gather_interchange_stats): Ditto.
638 (linear_transform_loops): Ditto.
640 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
643 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
644 biv_fits_mode_p, extension_within_bounds_p): New functions.
645 (check_ext_dependent_givs): Use them.
647 2005-01-06 Roger Sayle <roger@eyesopen.com>
649 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
650 more than one barrier, in which case we should delete them all.
652 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
654 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
656 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
658 Fix PR middle-end/19286
660 * dwarf2out.c (gen_subprogram_die): If we've already tried to
661 output this subprogram, simply ignore this attempt to do it again.
662 (add_abstract_origin_attribute): Don't abort trying to add the abstract
663 origin attribute if it's not possible.
664 (gen_block_die): Don't ignore subblocks of "unused" blocks.
665 (decls_for_scope): Ditto.
666 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
667 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
669 2005-01-05 Richard Henderson <rth@redhat.com>
672 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
673 (bdesc_2arg): Use it.
674 (ix86_expand_binop_builtin): Force operands into registers
676 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
677 ix86_expand_sse_compare, ix86_expand_sse_comi,
678 ix86_expand_builtin): Likewise.
680 2005-01-05 Richard Henderson <rth@redhat.com>
682 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
684 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
685 (mulv4hi3): Set itanium_class mmmul.
686 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
687 (fpack): Rename from fpack_sfsf.
689 2005-01-05 Richard Henderson <rth@redhat.com>
692 * reload1.c (do_input_reload): Restrict the optimization deleteing
693 a previous output reload to RELOAD_FOR_INPUT.
695 2005-01-05 Steven Bosscher <stevenb@suse.de>
697 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
698 comment that we fall through after case.
699 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
700 (distribute_links): Likewise.
701 * cse.c (cse_insn): Likewise.
702 * cselib.c (cselib_invalidate_mem): Likewise.
703 * df.c: Update comments at the top of the file.
704 (read_modify_subreg_p): Update comments here too.
705 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
706 * flow.c (mark_set_1): Likewise.
707 (mark_used_regs): Likewise.
708 * gcse.c (mems_conflict_for_gcse_p): Likewise.
709 (canon_list_insert): Likewise.
710 (mark_set): Likewise.
711 (try_replace_reg): Likewise.
712 (store_killed_in_insn): Likewise.
713 * loop.c (count_one_set): Likewise.
714 (basic_induction_var): Likewise.
715 * postreload-gcse.c (find_mem_conflicts): Likewise.
716 * postreload.c (reload_combine_note_store): Likewise.
717 (move2add_note_store): Likewise.
718 * reload.c (find_equiv_reg): Likewise.
719 (mark_referenced_resources): Likewise.
720 * rtlanal.c (set_noop_p): Likewise.
721 (note_stores): Likewise.
722 (note_uses): Likewise.
723 * sched-deps.c (sched_analyze_1): Likewise.
724 * sched-rgn.c (check_live_1): Likewise.
725 (update_live_1): Likewise.
726 * config/i860/i860.c: Likewise.
728 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
730 (ZERO_EXTRACT): Mention that this one can be an lvalue.
732 * doc/rtl.texi: Update documentation for bit-fields and SET.
734 2005-01-05 Roger Sayle <roger@eyesopen.com>
736 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
738 2005-01-05 Jan Hubicka <jh@suse.cz>
739 Richard Henderson <rth@redhat.com>
742 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
745 2005-01-05 Richard Henderson <rth@redhat.com>
748 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
749 (sse_shufps): Change operand 3 to const_int_operand.
750 (sse2_storelps): Fix typo in template.
751 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
752 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
753 (ix86_expand_builtin): Likewise.
755 2005-01-05 Julian Brown <julian@codesourcery.com>
757 * config/arm/arm.c (arm_return_in_memory): Treat complex types
758 as aggregates for AAPCS ABIs.
760 2005-01-05 Stan Shebs <shebs@apple.com>
762 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
763 Mach-O getters if ppc64.
764 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
766 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
767 signed four-byte field for global code case.
768 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
770 2005-01-05 Roger Sayle <roger@eyesopen.com>
773 * c-common.c: Include real.h.
774 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
775 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
776 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
777 * Makefile.in (c-common.o): Update dependencies.
779 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
781 * c-parse.in (asm_string): Add trailing semicolon.
783 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
785 * c-parse.in (asm_string): New. Don't allow wide strings in
787 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
788 asm_string instead of STRING.
790 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
792 * c-typeck.c (constructor_no_implicit): Remove.
793 (set_designator, process_init_element): Don't check
794 constructor_no_implicit.
796 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
797 Kaz Kojima <kkojima@gcc.gnu.org>
800 * lcm.c (create_pre_exit): New.
801 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
802 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
804 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
806 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
807 -msoft-float in target_float_switch.
808 * arm.c (arm_override_options): Fix processing of target_float_switch.
810 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
812 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
815 2004-01-05 Uros Bizjak <uros@kss-loka.si>
817 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
818 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
820 2005-01-04 Roger Sayle <roger@eyesopen.com>
822 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
823 if deleting the then-block allows the test-block to fallthru to the
826 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
829 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
830 after the full declaration if the are in two different TUs.
832 2005-01-04 Richard Henderson <rth@redhat.com>
835 * tree-sra.c (generate_one_element_init): Just
836 call gimplify_and_add.
837 (generate_element_init): Record the
838 new referenced variables and mark them for renaming
840 (generate_element_init_1): This.
841 (scalarize_init): Don't call push_gimplify_context/
842 pop_gimplify_context.
844 2005-01-04 Geoffrey Keating <geoffk@apple.com>
846 * toplev.c (get_src_pwd): Handle failure of getpwd().
848 2005-01-04 Roger Sayle <roger@eyesopen.com>
850 * fold-const.c (fold_single_bit_test): Delete unreachable handling
852 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
853 all the cases handled by it are inverted by invert_truthvalue.
855 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
857 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
858 (save_register_windowsi): Likewise.
860 2005-01-04 Richard Henderson <rth@redhat.com>
862 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
864 2005-01-04 Uros Bizjak <uros@kss-loka.si>
867 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
868 any simplifications of vector mode comparison operators.
869 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
871 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
872 Devang Patel <dpatel@apple.com>
874 PR tree-optimization/18308
875 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
876 the operands before creating a new expression.
877 * dojump.c (do_jump): Make drop_through_label available
878 for all cases. Add expansion of COND_EXPR.
880 2005-01-04 Ira Rosen <irar@il.ibm.com>
882 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
883 INTEGER_CST instead of TREE_CONSTANT.
884 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
885 instead of TREE_CONSTANT.
886 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
887 instead of TREE_CONSTANT.
889 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
891 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
893 2005-01-04 Uros Bizjak <uros@kss-loka.si>
896 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
897 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
899 2005-01-03 Richard Henderson <rth@redhat.com>
901 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
902 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
903 fold_convert_const_real_from_real): Split out from ...
904 (fold_convert_const): ... here.
906 2005-01-03 Richard Henderson <rth@redhat.com>
909 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
910 (mov<MMXMODEI>_internal): Likewise.
911 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
912 Add V2SF case; use it for SSE1; don't use TI.
913 (movdf_integer): Likewise.
914 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
916 2005-01-03 Ira Rosen <irar@il.ibm.com>
918 * tree-vectorizer.c (vect_strip_conversions): New function.
919 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
920 check for binary class.
922 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
926 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
928 * gimple-low.c (mark_blocks_with_used_vars): New function.
929 (mark_blocks_with_used_subblocks): Ditto.
930 (mark_used_blocks): Ditto.
931 (pass_mark_used_blocks): New pass.
932 * tree-inline.c: Include debug.h.
933 (expand_call_inline): Call outlining_inline_function here.
934 * tree-optimize.c (init_tree_optimization_passes): Add
935 pass_mark_used_blocks.
936 * tree-pass.h (pass_mark_used_blocks): New.
937 * Makefile.in (tree-inline.o): Add debug.h dependency.
939 2005-01-03 Geoffrey Keating <geoffk@apple.com>
941 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
942 VAR_DECLs to have weak_import attribute.
944 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
947 2004-01-03 Steven Bosscher <stevenb@suse.de>
949 * passes.c (rest_of_compilation): Don't run regmove if only
950 flag_expensive_optimizations, require flag_regmove instead.
952 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
954 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
956 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
957 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
958 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
959 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
961 2005-01-03 Richard Henderson <rth@redhat.com>
963 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
964 (ia64_const_ok_for_letter_p): New.
965 (ia64_const_double_ok_for_letter_p): New.
966 (ia64_extra_constraint): New.
967 (ia64_expand_vecint_compare): New.
968 (ia64_expand_vcondu_v2si): New.
969 (ia64_expand_vecint_cmov): New.
970 (ia64_expand_vecint_minmax): New.
971 (ia64_print_operand): Add 'v'.
972 (ia64_preferred_reload_class): New.
973 (ia64_vector_mode_supported_p): New.
974 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
975 (PREFERRED_RELOAD_CLASS): Move to function.
976 (CONST_OK_FOR_LETTER_P): Move to function.
977 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
978 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
979 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
980 (EXTRA_CONSTRAINT): Move to function.
981 * config/ia64/ia64.md: Include vect.md.
982 (itanium_class): Add mmalua.
984 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
986 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
987 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
989 (const_int_2bit_operand): New.
990 (fr_reg_or_0_operand): New.
991 * config/ia64/ia64-modes.def: Add vector modes.
992 * config/ia64/ia64-protos.h: Update.
993 * config/ia64/vect.md: New file.
995 2005-01-03 Richard Henderson <rth@redhat.com>
997 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
999 2005-01-03 Uros Bizjak <uros@kss-loka.si>
1002 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
1004 (log1pdf2): Change mode of operands[0,1] to DFmode.
1006 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
1008 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
1009 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
1010 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
1011 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
1012 SPARC_RELAXED_ORDERING.
1014 2005-01-03 Richard Henderson <rth@redhat.com>
1015 Uros Bizjak <uros@kss-loka.si>
1018 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
1019 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
1020 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
1022 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
1023 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
1024 const_pow2_1_to_128_operand): New predicates.
1026 2005-01-02 Greg McGary <greg@mcgary.org>
1028 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
1030 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
1031 object. [ptr_type]: Remove unused variable.
1032 Remember nearest addressable array-element or record-component
1033 when checking bitfield components. Tolerate empty BIND_EXPR.
1035 2005-01-01 Richard Henderson <rth@redhat.com>
1037 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
1038 that don't narrow the value. Fail for other conversions.
1040 2005-01-01 Richard Henderson <rth@redhat.com>
1043 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
1044 * c-lang.c (finish_file): Don't do it here.
1045 * objc/objc-act.c (objc_finish_file): Likewise.
1047 * cgraph.c (decl_assembler_name_equal): New.
1048 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
1049 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
1050 * cgraph.h (struct cgraph_varpool_node): Add next.
1051 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
1052 * varasm.c (assemble_alias): Mark the target as needed.
1054 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
1057 * function.c (get_arg_pointer_save_area): Use entry_of_function
1058 instead of get_insns.
1060 2005-01-01 Roger Sayle <roger@eyesopen.com>
1061 Andrew Pinski <pinskia@physics.uc.edu>
1062 James E. Wilson <wilson@specifixinc.com>
1064 PR rtl-optimization/12092
1065 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
1068 2005-01-01 Roger Sayle <roger@eyesopen.com>
1069 Olivier Hainque <hainque@act-europe.fr>
1071 * tree.c (int_fits_type_p): A narrower type always fits in a
1072 wider one, except for negative values into unsigned types.
1074 2005-01-01 Roger Sayle <roger@eyesopen.com>
1076 * tree.c (int_fits_type_p): Compare the result of force_fit_type
1077 with the original constant rather than require TREE_OVERFLOW.
1079 2005-01-01 Steven Bosscher <stevenb@suse.de>
1082 * c-decl.c (finish_function): If compiling C99, annotate the
1083 compiler generated return with the current file name and line 0.
1084 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
1085 the source line is greater than 0.
1086 (remove_bb): Likewise.
1088 See ChangeLog.12 for earlier changes.