1 2005-01-13 Steven Bosscher <stevenb@suse.de>
3 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
4 (fix_stmt_v_may_defs): Likewise.
6 2005-01-13 Richard Henderson <rth@redhat.com>
8 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
9 for converting the mask.
11 2005-01-13 David Edelsohn <edelsohn@gnu.org>
13 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
16 2005-01-13 Richard Henderson <rth@redhat.com>
18 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
19 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
20 IX86_BUILTIN_MOVDQ2Q): Remove.
21 (IX86_BUILTIN_VEC_EXT_V4SI): New.
22 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
23 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
25 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
26 (sse2_pextrw): Fix immediate constraint.
27 (sse2_loadq, sse2_loadq_rex64): Remove.
28 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
29 Use __builtin_ia32_vec_ext_<size>.
30 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
32 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
34 * function.c (assign_parm_setup_block): Look inside original
35 entry_parm when inspecting PARALLEL.
37 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
38 Joel Sherrill <joel@oarcorp.com>
41 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
42 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
43 rtems_gxx_recursive_mutex_init.
44 (__gthread_recursive_mutex_lock): New function.
45 (__gthread_recursive_mutex_trylock): Likewise.
46 (__gthread_recursive_mutex_unlock): Likewise.
48 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
50 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
52 2005-01-13 Jan Beulich <jbeulich@novell.com>
54 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
55 (ashrti3_internal): Indicate output is early clobber. Generate result
56 into output rather than first input. Use move for low word of output
57 if shift count is exactly 64.
58 (lshrti3_internal): Likewise.
60 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
64 * reload1.c (delete_output_reload): Don't delete an output reload
65 if the pseudo lives longer than a single basic block. Adjust and
66 improve wording of comment.
68 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
71 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
72 of TYPE_SIZE where we mean the number of bytes.
74 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
76 * function.c (assign_parm_setup_block): Relax condition on
77 multi-register optimization.
79 2005-01-12 Nick Clifton <nickc@redhat.com>
81 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
82 an "in_delay_slot" attribute of "no" to prevent them being used in
83 delay slots. This is forbidden because they might generate
86 2005-01-12 Alan Modra <amodra@bigpond.net.au>
89 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
92 2005-01-12 Nick Clifton <nickc@redhat.com>
94 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
95 passed on to GAS) but do not define, as GAS no longer supports or
96 needs the -m2000 option.
98 2005-01-11 Roger Sayle <roger@eyesopen.com>
100 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
101 type as args of unprototyped builtins aren't automatically promoted.
102 (fold_builtin_abs): Likewise.
104 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
106 * Makefile.in: Set a `build-warn' variable.
108 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
110 PR tree-optimization/17949
111 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
112 (find_interesting_uses_address): Use it.
114 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
116 * regrename.c (kill_value): Handle subreg's that won't simplify.
118 2005-01-11 Richard Henderson <rth@redhat.com>
121 * config/i386/i386.h (enum ix86_builtins): Move ...
122 * config/i386/i386.c: ... here.
123 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
124 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
125 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
126 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
127 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
128 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
129 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
130 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
131 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
132 IX86_BUILTIN_LOADDDUP): Remove.
133 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
134 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
135 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
136 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
137 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
138 (ix86_init_builtins): Make static.
139 (ix86_init_mmx_sse_builtins): Update for changed builtins.
140 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
141 if all the modes match. Otherwise, fake it.
142 (get_element_number, ix86_expand_vec_init_builtin,
143 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
144 (ix86_expand_builtin): Make static. Update for changed builtins.
145 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
146 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
147 (ix86_expand_vector_init_duplicate): New.
148 (ix86_expand_vector_init_low_nonzero): New.
149 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
150 Split out from ix86_expand_vector_init; handle integer modes.
151 (ix86_expand_vector_init): Use them.
152 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
153 * config/i386/i386-protos.h: Update.
154 * config/i386/predicates.md (reg_or_0_operand): New.
155 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
156 (movv2sf_internal): Likewise. And a splitter to match them all.
157 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
158 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
159 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
160 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
161 (mmx_pinsrw): Fix operand ordering.
162 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
163 rather than sse_loadss expander.
164 (movv2df splitter): Similarly.
165 (sse_loadss, sse_loadlss): Remove.
166 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
167 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
168 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
169 (sse3_movddup): Rename with '*'.
170 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
171 (sse2_loadsd): Remove.
172 (vec_dupv2df_sse3): Rename from sse3_loadddup.
173 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
174 (sse2_pinsrw): Fix argument ordering.
175 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
176 (sse2_stored): Remove 'r' destination.
177 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
178 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
179 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
180 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
181 vec_extractv16qi, vec_initv16qi): New.
183 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
184 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
185 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
187 (_mm_load_pd, _mm_store_pd): Use plain dereference.
188 (_mm_load_si128, _mm_store_si128): Likewise.
189 (_mm_load1_pd): Use _mm_set1_pd.
190 (_mm_load_sd): Use _mm_set_sd.
191 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
192 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
193 (_mm_set_epi64): Use _mm_set_epi64x.
194 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
195 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
196 _mm_setr_epi8): Use _mm_set_foo form.
197 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
198 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
199 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
200 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
201 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
202 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
203 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
204 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
205 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
206 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
207 (_mm_movedup_pd): Use _mm_shuffle_pd.
208 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
209 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
210 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
211 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
213 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
214 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
215 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
216 (_mm_store_ps): Use raw dereference.
217 (_mm_store1_ps): Use _mm_storeu_ps.
218 (_mm_storer_ps): Use _mm_store_ps.
219 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
220 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
222 2005-01-11 Stan Shebs <shebs@apple.com>
224 * config/rs6000/rs6000.c (machopic_output_stub): Issue
225 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
227 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
230 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
231 Skip all DECLs except for FIELD_DECLs.
233 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
235 * config/s390/s390.c (override_options): Return error if
236 -mbackchain, -mpacked-stack and -mhard-float are used together.
237 (s390_va_start): Remove the backchain && packed-stack special case.
238 (s390_gimplify_va_arg): Likewise.
239 * doc/invoke.texi: Remove the ABI incompatibility note.
241 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
243 * config/s390/s390.c (struct s390_frame_layout): Remove
245 (s390_frame_info, s390_emit_prologue): Replace occurrences of
246 save_backchain_p with TARGET_BACKCHAIN.
248 2005-01-11 Alan Modra <amodra@bigpond.net.au>
251 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
252 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
253 * function.c (locate_and_pad_parm): Set new field.
254 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
255 Tweak where_pad test to include "none". Always set mem align for
257 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
258 not sufficient for type.
259 (assign_parm_setup_block): If stack_parm is zero on entry, always
260 make a new stack local. Block move old stack parm if necessary
261 to new aligned stack local.
262 (assign_parm_setup_stack): Use a block move to handle
263 potentially misaligned entry_parm.
264 (assign_parms_unsplit_complex): Specify required alignment when
265 creating stack local.
266 * calls.c (compute_argument_addresses): Override alignment of stack
267 arg calculated from its type with the alignment given by
268 FUNCTION_ARG_BOUNDARY.
269 (store_one_arg): Likewise.
271 2005-01-11 Jan Beulich <jbeulich@novell.com>
273 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
274 Change respective itanium_class attribute to ialu.
275 (shladdp4_internal): New.
276 * config/ia64/predicates.md (shladd_log2_operand): New.
278 2005-01-11 Richard Henderson <rth@redhat.com>
280 * expr.c (store_constructor): Use rtvec_alloc instead of
281 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
283 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
285 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
287 2005-01-11 Alan Modra <amodra@bigpond.net.au>
289 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
290 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
291 (TARGET_RWRELOC): Define.
292 (ia64_rwreloc_section_type_flags): Delete.
293 (ia64_section_type_flags): New function.
294 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
295 (TARGET_RWRELOC): Define.
297 2005-01-10 David Mosberger <davidm@hpl.hp.com>
300 * config/ia64/ia64.c (process_set): For alloc insn, only call
301 process_epilogue is !frame_pointer_needed.
303 2005-01-10 Roger Sayle <roger@eyesopen.com>
306 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
307 unary operator and can't be treated as a binary/comparison operator.
309 2005-01-10 Richard Henderson <rth@redhat.com>
311 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
312 (ix86_return_in_memory): Likewise.
313 (function_arg_advance): Likewise. Mirror structure in function_arg
314 for choosing register to advance.
316 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
318 * tree-vectorizer.c, tree.def: Fix comment typos.
320 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
322 * config/s390/s390.c (struct processor_costs): 4 new fields:
323 ddbr, ddr, debr, der.
324 (s390_rtx_costs): More precise handling of divide instructions.
326 2005-01-10 David Edelsohn <edelsohn@gnu.org>
328 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
330 2005-01-10 Jan Beulich <jbeulich@novell.com>
332 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
333 names resulting from -ffunction-sections/-fdata-sections and linkonce
335 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
336 based on the section name. Rearrange the section name comparison logic
337 slightly so that each section name is compared against at most once.
339 2005-01-10 Ben Elliston <bje@au.ibm.com>
341 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
342 targets for which -fPIC reduces limitations on the GOT size.
344 2005-01-09 Falk Hueffner <falk@debian.org>
346 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
349 2005-01-09 David Edelsohn <edelsohn@gnu.org>
352 * collect2.c (main): Set aixrtl_flag for -brtl option.
353 (resolve_lib_name): Search for .so file extension before .a
356 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
358 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
359 (vect_analyze_data_refs_alignment): Add dump prints.
361 2005-01-09 Ira Rosen <irar@il.ibm.com>
363 * tree-vectorizer.c (vect_analyze_offset_expr): Use
364 expr_invariant_in_loop_p.
365 Initialize outputs first thing in the function.
366 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
367 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
368 (vect_analyze_pointer_ref_access): Check that the initial condition of
369 the access function is loop invariant.
371 2005-01-09 Richard Henderson <rth@redhat.com>
373 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
374 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
375 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
376 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
377 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
378 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
379 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
380 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
381 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
382 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
383 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
384 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
385 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
386 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
387 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
388 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
389 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
390 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
391 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
392 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
393 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
394 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
395 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
397 (mmx_clrdi, pavgusb): Remove.
398 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
399 with leading sse_ prefix.
400 * config/i386/sse.md: Receive them.
401 * config/i386/mmx.md: New file.
402 (MMXMODE12, MMXMODE24, mmxvecsize): New.
403 (subrv2sf3): Turn into expander for normal subtraction.
404 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
405 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
406 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
407 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
408 commutative; use ix86_binary_operator_ok.
409 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
410 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
411 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
412 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
413 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
414 Macroize from existing patterns; use ix86_binary_operator_ok.
415 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
416 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
417 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
418 with vec_select+vec_concat.
419 (mmx_pshufw, mmx_pshufw_1): Likewise.
420 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
421 (mmx_uavgv4hi3): Model correctly.
422 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
423 _mm_xor_si64): Remove casts.
425 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
427 PR tree-optimization/19224
428 * tree-scalar-evolution.c (get_instantiated_value,
429 set_instantiated_value): New functions.
430 (instantiate_parameters_1): Cache the results.
431 (instantiate_parameters, resolve_mixers): Initialize and free
434 2005-01-08 David Edelsohn <edelsohn@gnu.org>
436 * config/i386/i386.md (addhi_4): Correct reference in comment.
439 2005-01-08 Richard Henderson <rth@redhat.com>
441 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
442 (_mm_cvtsi128_si64x): Likewise.
443 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
444 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
445 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
446 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
447 * config/i386/i386-protos.h: Update.
448 * config/i386/i386.c (print_operand): Add 'H'.
449 (ix86_fixup_binary_operands): Split out from ...
450 (ix86_expand_binary_operator): ... here.
451 (ix86_fixup_binary_operands_no_copy): New.
452 (ix86_expand_fp_absneg_operator): Handle vector mode results.
453 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
454 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
455 (safe_vector_operand): Use CONST0_RTX.
456 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
457 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
458 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
459 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
461 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
462 (IX86_BUILTIN_CMPNGESS): New.
463 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
464 (attr type): Add sselog1.
465 (attr unit, attr memory): Handle it.
466 (movti, movti_internal, movti_rex64): Move near other integer moves.
467 (movtf, movtf_internal): Move near other fp moves.
468 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
469 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
470 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
471 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
472 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
473 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
474 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
475 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
476 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
477 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
478 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
479 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
480 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
481 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
482 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
483 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
484 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
485 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
486 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
487 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
488 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
489 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
490 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
491 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
492 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
493 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
494 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
495 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
496 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
497 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
498 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
499 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
500 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
501 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
502 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
503 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
504 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
505 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
506 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
507 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
508 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
509 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
510 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
511 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
513 (SSEPUSH, push<SSEPUSH>): Remove.
514 (MMXPUSH, push<MMXPUSH>): Remove.
515 (sse_movaps, sse_movaps_1, sse_movups): Remove.
516 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
517 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
518 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
519 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
520 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
521 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
522 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
523 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
524 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
525 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
526 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
527 (ppro_sse_log_V4SF): Handle sselog1.
528 * config/i386/predicates.md (const_0_to_1_operand): New.
529 (const_0_to_255_mul_8_operand): New.
530 (const_1_to_31_operand): Rename from const_int_1_31_operand.
531 (const_2_to_3_operand, const_4_to_7_operand): New.
532 * config/i386/sse.md: New file.
533 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
534 (sse_movups): Rename from sse_movups_1.
535 (sse_loadlss): Rename from sse_loadss_1.
536 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
538 (negv4sf2): Use ix86_expand_fp_absneg_operator.
539 (absv4sf2, negv2df, absv2df): New.
540 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
541 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
542 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
543 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
544 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
545 (sse3_addsubv4sf3): Model correctly.
546 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
547 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
548 (sse_movhlps): Model with vec_select+vec_concat.
549 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
550 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
551 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
552 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
553 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
554 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
555 sse2_pshufhw_1): Likewise.
556 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
557 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
558 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
559 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
560 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
561 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
562 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
563 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
564 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
565 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
566 sminv8hi3): Mark commutative
567 operands. Use ix86_binary_operator_ok.
568 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
569 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
570 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
571 sse2_punpcklqdq): Allow operand2 in memory.
572 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
573 sse2_movsd): Add memory alternatives.
574 (sse_storelps): Turn expander into an insn; split after reload.
575 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
576 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
578 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
580 * configure.ac (DWARF-2 debug_line): Use objdump.
581 * configure: Regenerate.
583 2005-01-08 Jeff Law <law@redhat.com>
584 Diego Novillo <dnovillo@redhat.com>
586 PR tree-optimization/18241
587 * tree-nrv.c (tree_nrv): Ignore volatile return values.
588 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
589 statements with volatile operands.
590 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
591 operands after marking a statement with has_volatile_ops.
593 2005-01-08 Roger Sayle <roger@eyesopen.com>
595 * tree.c (int_fits_type_p): Always honor integer constant
596 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
598 2005-01-08 Roger Sayle <roger@eyesopen.com>
600 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
601 the latent bug in rtl_delete_block has been resolved.
603 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
605 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
606 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
608 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
610 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
611 library as @multilib_dir@/$(SHLIB_SONAME).
613 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
615 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
616 * configure: Regenerate.
618 2005-01-07 Jakub Jelinek <jakub@redhat.com>
620 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
623 2005-01-07 David Edelsohn <edelsohn@gnu.org>
626 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
627 Convert non-word aligned offset address using ld/std into
630 2005-01-07 Richard Henderson <rth@redhat.com>
632 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
634 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
636 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
638 2005-01-07 Jakub Jelinek <jakub@redhat.com>
640 PR tree-optimization/19283
641 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
642 is not shorter than the original type.
644 PR rtl-optimization/19012
645 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
647 PR rtl-optimization/18861
648 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
649 (FULL_STATE): Define.
650 (mark_tablejump_edge): New function.
651 (purge_dead_tablejump_edges): New function.
652 (find_bb_boundaries): Use it.
654 PR tree-optimization/18828
655 * builtins.c (expand_builtin_next_arg): Remove argument and all
656 the argument checking.
657 (expand_builtin): Adjust caller.
658 (expand_builtin_va_start): Likewise. Remove error for too many
660 (fold_builtin_next_arg): Issue error for too many arguments.
661 After checking arguments, replace them with magic arguments that
662 prevent further checking of the args.
664 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
666 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
668 2005-01-06 Roger Sayle <roger@eyesopen.com>
671 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
672 and %{.S:...} (and their negative variants) to test whether the
673 input file is assembler or pre-processed-assembler independent of
674 the actual filename extension.
676 2005-01-06 Roger Sayle <roger@eyesopen.com>
678 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
679 of sign or zero extended values.
681 2005-01-06 Geoffrey Keating <geoffk@apple.com>
683 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
684 for IBM long double format correctly.
686 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
688 Fix PR tree-optimization/18792
690 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
691 to first_loop_depth, and use loop depth instead of loop number.
692 (build_classic_dir_vector): Ditto.
693 (compute_data_dependences_for_loop): Use depth, not loop number.
694 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
695 number. Pass in loops, instead of loop numbers.
696 (gather_interchange_stats): Ditto.
697 (linear_transform_loops): Ditto.
699 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
702 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
703 biv_fits_mode_p, extension_within_bounds_p): New functions.
704 (check_ext_dependent_givs): Use them.
706 2005-01-06 Roger Sayle <roger@eyesopen.com>
708 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
709 more than one barrier, in which case we should delete them all.
711 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
713 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
715 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
717 Fix PR middle-end/19286
719 * dwarf2out.c (gen_subprogram_die): If we've already tried to
720 output this subprogram, simply ignore this attempt to do it again.
721 (add_abstract_origin_attribute): Don't abort trying to add the abstract
722 origin attribute if it's not possible.
723 (gen_block_die): Don't ignore subblocks of "unused" blocks.
724 (decls_for_scope): Ditto.
725 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
726 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
728 2005-01-05 Richard Henderson <rth@redhat.com>
731 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
732 (bdesc_2arg): Use it.
733 (ix86_expand_binop_builtin): Force operands into registers
735 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
736 ix86_expand_sse_compare, ix86_expand_sse_comi,
737 ix86_expand_builtin): Likewise.
739 2005-01-05 Richard Henderson <rth@redhat.com>
741 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
743 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
744 (mulv4hi3): Set itanium_class mmmul.
745 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
746 (fpack): Rename from fpack_sfsf.
748 2005-01-05 Richard Henderson <rth@redhat.com>
751 * reload1.c (do_input_reload): Restrict the optimization deleteing
752 a previous output reload to RELOAD_FOR_INPUT.
754 2005-01-05 Steven Bosscher <stevenb@suse.de>
756 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
757 comment that we fall through after case.
758 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
759 (distribute_links): Likewise.
760 * cse.c (cse_insn): Likewise.
761 * cselib.c (cselib_invalidate_mem): Likewise.
762 * df.c: Update comments at the top of the file.
763 (read_modify_subreg_p): Update comments here too.
764 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
765 * flow.c (mark_set_1): Likewise.
766 (mark_used_regs): Likewise.
767 * gcse.c (mems_conflict_for_gcse_p): Likewise.
768 (canon_list_insert): Likewise.
769 (mark_set): Likewise.
770 (try_replace_reg): Likewise.
771 (store_killed_in_insn): Likewise.
772 * loop.c (count_one_set): Likewise.
773 (basic_induction_var): Likewise.
774 * postreload-gcse.c (find_mem_conflicts): Likewise.
775 * postreload.c (reload_combine_note_store): Likewise.
776 (move2add_note_store): Likewise.
777 * reload.c (find_equiv_reg): Likewise.
778 (mark_referenced_resources): Likewise.
779 * rtlanal.c (set_noop_p): Likewise.
780 (note_stores): Likewise.
781 (note_uses): Likewise.
782 * sched-deps.c (sched_analyze_1): Likewise.
783 * sched-rgn.c (check_live_1): Likewise.
784 (update_live_1): Likewise.
785 * config/i860/i860.c: Likewise.
787 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
789 (ZERO_EXTRACT): Mention that this one can be an lvalue.
791 * doc/rtl.texi: Update documentation for bit-fields and SET.
793 2005-01-05 Roger Sayle <roger@eyesopen.com>
795 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
797 2005-01-05 Jan Hubicka <jh@suse.cz>
798 Richard Henderson <rth@redhat.com>
801 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
804 2005-01-05 Richard Henderson <rth@redhat.com>
807 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
808 (sse_shufps): Change operand 3 to const_int_operand.
809 (sse2_storelps): Fix typo in template.
810 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
811 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
812 (ix86_expand_builtin): Likewise.
814 2005-01-05 Julian Brown <julian@codesourcery.com>
816 * config/arm/arm.c (arm_return_in_memory): Treat complex types
817 as aggregates for AAPCS ABIs.
819 2005-01-05 Stan Shebs <shebs@apple.com>
821 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
822 Mach-O getters if ppc64.
823 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
825 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
826 signed four-byte field for global code case.
827 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
829 2005-01-05 Roger Sayle <roger@eyesopen.com>
832 * c-common.c: Include real.h.
833 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
834 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
835 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
836 * Makefile.in (c-common.o): Update dependencies.
838 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
840 * c-parse.in (asm_string): Add trailing semicolon.
842 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
844 * c-parse.in (asm_string): New. Don't allow wide strings in
846 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
847 asm_string instead of STRING.
849 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
851 * c-typeck.c (constructor_no_implicit): Remove.
852 (set_designator, process_init_element): Don't check
853 constructor_no_implicit.
855 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
856 Kaz Kojima <kkojima@gcc.gnu.org>
859 * lcm.c (create_pre_exit): New.
860 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
861 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
863 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
865 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
866 -msoft-float in target_float_switch.
867 * arm.c (arm_override_options): Fix processing of target_float_switch.
869 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
871 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
874 2004-01-05 Uros Bizjak <uros@kss-loka.si>
876 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
877 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
879 2005-01-04 Roger Sayle <roger@eyesopen.com>
881 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
882 if deleting the then-block allows the test-block to fallthru to the
885 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
888 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
889 after the full declaration if the are in two different TUs.
891 2005-01-04 Richard Henderson <rth@redhat.com>
894 * tree-sra.c (generate_one_element_init): Just
895 call gimplify_and_add.
896 (generate_element_init): Record the
897 new referenced variables and mark them for renaming
899 (generate_element_init_1): This.
900 (scalarize_init): Don't call push_gimplify_context/
901 pop_gimplify_context.
903 2005-01-04 Geoffrey Keating <geoffk@apple.com>
905 * toplev.c (get_src_pwd): Handle failure of getpwd().
907 2005-01-04 Roger Sayle <roger@eyesopen.com>
909 * fold-const.c (fold_single_bit_test): Delete unreachable handling
911 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
912 all the cases handled by it are inverted by invert_truthvalue.
914 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
916 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
917 (save_register_windowsi): Likewise.
919 2005-01-04 Richard Henderson <rth@redhat.com>
921 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
923 2005-01-04 Uros Bizjak <uros@kss-loka.si>
926 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
927 any simplifications of vector mode comparison operators.
928 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
930 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
931 Devang Patel <dpatel@apple.com>
933 PR tree-optimization/18308
934 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
935 the operands before creating a new expression.
936 * dojump.c (do_jump): Make drop_through_label available
937 for all cases. Add expansion of COND_EXPR.
939 2005-01-04 Ira Rosen <irar@il.ibm.com>
941 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
942 INTEGER_CST instead of TREE_CONSTANT.
943 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
944 instead of TREE_CONSTANT.
945 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
946 instead of TREE_CONSTANT.
948 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
950 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
952 2005-01-04 Uros Bizjak <uros@kss-loka.si>
955 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
956 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
958 2005-01-03 Richard Henderson <rth@redhat.com>
960 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
961 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
962 fold_convert_const_real_from_real): Split out from ...
963 (fold_convert_const): ... here.
965 2005-01-03 Richard Henderson <rth@redhat.com>
968 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
969 (mov<MMXMODEI>_internal): Likewise.
970 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
971 Add V2SF case; use it for SSE1; don't use TI.
972 (movdf_integer): Likewise.
973 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
975 2005-01-03 Ira Rosen <irar@il.ibm.com>
977 * tree-vectorizer.c (vect_strip_conversions): New function.
978 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
979 check for binary class.
981 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
985 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
987 * gimple-low.c (mark_blocks_with_used_vars): New function.
988 (mark_blocks_with_used_subblocks): Ditto.
989 (mark_used_blocks): Ditto.
990 (pass_mark_used_blocks): New pass.
991 * tree-inline.c: Include debug.h.
992 (expand_call_inline): Call outlining_inline_function here.
993 * tree-optimize.c (init_tree_optimization_passes): Add
994 pass_mark_used_blocks.
995 * tree-pass.h (pass_mark_used_blocks): New.
996 * Makefile.in (tree-inline.o): Add debug.h dependency.
998 2005-01-03 Geoffrey Keating <geoffk@apple.com>
1000 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
1001 VAR_DECLs to have weak_import attribute.
1003 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
1006 2004-01-03 Steven Bosscher <stevenb@suse.de>
1008 * passes.c (rest_of_compilation): Don't run regmove if only
1009 flag_expensive_optimizations, require flag_regmove instead.
1011 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
1013 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
1015 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
1016 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1017 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
1018 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1020 2005-01-03 Richard Henderson <rth@redhat.com>
1022 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
1023 (ia64_const_ok_for_letter_p): New.
1024 (ia64_const_double_ok_for_letter_p): New.
1025 (ia64_extra_constraint): New.
1026 (ia64_expand_vecint_compare): New.
1027 (ia64_expand_vcondu_v2si): New.
1028 (ia64_expand_vecint_cmov): New.
1029 (ia64_expand_vecint_minmax): New.
1030 (ia64_print_operand): Add 'v'.
1031 (ia64_preferred_reload_class): New.
1032 (ia64_vector_mode_supported_p): New.
1033 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
1034 (PREFERRED_RELOAD_CLASS): Move to function.
1035 (CONST_OK_FOR_LETTER_P): Move to function.
1036 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
1037 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
1038 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
1039 (EXTRA_CONSTRAINT): Move to function.
1040 * config/ia64/ia64.md: Include vect.md.
1041 (itanium_class): Add mmalua.
1043 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
1045 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
1046 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
1048 (const_int_2bit_operand): New.
1049 (fr_reg_or_0_operand): New.
1050 * config/ia64/ia64-modes.def: Add vector modes.
1051 * config/ia64/ia64-protos.h: Update.
1052 * config/ia64/vect.md: New file.
1054 2005-01-03 Richard Henderson <rth@redhat.com>
1056 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
1058 2005-01-03 Uros Bizjak <uros@kss-loka.si>
1061 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
1063 (log1pdf2): Change mode of operands[0,1] to DFmode.
1065 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
1067 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
1068 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
1069 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
1070 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
1071 SPARC_RELAXED_ORDERING.
1073 2005-01-03 Richard Henderson <rth@redhat.com>
1074 Uros Bizjak <uros@kss-loka.si>
1077 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
1078 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
1079 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
1081 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
1082 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
1083 const_pow2_1_to_128_operand): New predicates.
1085 2005-01-02 Greg McGary <greg@mcgary.org>
1087 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
1089 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
1090 object. [ptr_type]: Remove unused variable.
1091 Remember nearest addressable array-element or record-component
1092 when checking bitfield components. Tolerate empty BIND_EXPR.
1094 2005-01-01 Richard Henderson <rth@redhat.com>
1096 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
1097 that don't narrow the value. Fail for other conversions.
1099 2005-01-01 Richard Henderson <rth@redhat.com>
1102 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
1103 * c-lang.c (finish_file): Don't do it here.
1104 * objc/objc-act.c (objc_finish_file): Likewise.
1106 * cgraph.c (decl_assembler_name_equal): New.
1107 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
1108 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
1109 * cgraph.h (struct cgraph_varpool_node): Add next.
1110 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
1111 * varasm.c (assemble_alias): Mark the target as needed.
1113 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
1116 * function.c (get_arg_pointer_save_area): Use entry_of_function
1117 instead of get_insns.
1119 2005-01-01 Roger Sayle <roger@eyesopen.com>
1120 Andrew Pinski <pinskia@physics.uc.edu>
1121 James E. Wilson <wilson@specifixinc.com>
1123 PR rtl-optimization/12092
1124 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
1127 2005-01-01 Roger Sayle <roger@eyesopen.com>
1128 Olivier Hainque <hainque@act-europe.fr>
1130 * tree.c (int_fits_type_p): A narrower type always fits in a
1131 wider one, except for negative values into unsigned types.
1133 2005-01-01 Roger Sayle <roger@eyesopen.com>
1135 * tree.c (int_fits_type_p): Compare the result of force_fit_type
1136 with the original constant rather than require TREE_OVERFLOW.
1138 2005-01-01 Steven Bosscher <stevenb@suse.de>
1141 * c-decl.c (finish_function): If compiling C99, annotate the
1142 compiler generated return with the current file name and line 0.
1143 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
1144 the source line is greater than 0.
1145 (remove_bb): Likewise.
1147 See ChangeLog.12 for earlier changes.