1 2005-01-13 Richard Henderson <rth@redhat.com>
3 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
4 for converting the mask.
6 2005-01-13 David Edelsohn <edelsohn@gnu.org>
8 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
11 2005-01-13 Richard Henderson <rth@redhat.com>
13 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
14 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
15 IX86_BUILTIN_MOVDQ2Q): Remove.
16 (IX86_BUILTIN_VEC_EXT_V4SI): New.
17 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
18 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
20 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
21 (sse2_pextrw): Fix immediate constraint.
22 (sse2_loadq, sse2_loadq_rex64): Remove.
23 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
24 Use __builtin_ia32_vec_ext_<size>.
25 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
27 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
29 * function.c (assign_parm_setup_block): Look inside original
30 entry_parm when inspecting PARALLEL.
32 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
33 Joel Sherrill <joel@oarcorp.com>
36 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
37 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
38 rtems_gxx_recursive_mutex_init.
39 (__gthread_recursive_mutex_lock): New function.
40 (__gthread_recursive_mutex_trylock): Likewise.
41 (__gthread_recursive_mutex_unlock): Likewise.
43 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
45 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
47 2005-01-13 Jan Beulich <jbeulich@novell.com>
49 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
50 (ashrti3_internal): Indicate output is early clobber. Generate result
51 into output rather than first input. Use move for low word of output
52 if shift count is exactly 64.
53 (lshrti3_internal): Likewise.
55 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
59 * reload1.c (delete_output_reload): Don't delete an output reload
60 if the pseudo lives longer than a single basic block. Adjust and
61 improve wording of comment.
63 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
66 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
67 of TYPE_SIZE where we mean the number of bytes.
69 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
71 * function.c (assign_parm_setup_block): Relax condition on
72 multi-register optimization.
74 2005-01-12 Nick Clifton <nickc@redhat.com>
76 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
77 an "in_delay_slot" attribute of "no" to prevent them being used in
78 delay slots. This is forbidden because they might generate
81 2005-01-12 Alan Modra <amodra@bigpond.net.au>
84 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
87 2005-01-12 Nick Clifton <nickc@redhat.com>
89 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
90 passed on to GAS) but do not define, as GAS no longer supports or
91 needs the -m2000 option.
93 2005-01-11 Roger Sayle <roger@eyesopen.com>
95 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
96 type as args of unprototyped builtins aren't automatically promoted.
97 (fold_builtin_abs): Likewise.
99 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
101 * Makefile.in: Set a `build-warn' variable.
103 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
105 PR tree-optimization/17949
106 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
107 (find_interesting_uses_address): Use it.
109 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
111 * regrename.c (kill_value): Handle subreg's that won't simplify.
113 2005-01-11 Richard Henderson <rth@redhat.com>
116 * config/i386/i386.h (enum ix86_builtins): Move ...
117 * config/i386/i386.c: ... here.
118 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
119 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
120 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
121 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
122 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
123 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
124 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
125 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
126 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
127 IX86_BUILTIN_LOADDDUP): Remove.
128 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
129 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
130 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
131 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
132 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
133 (ix86_init_builtins): Make static.
134 (ix86_init_mmx_sse_builtins): Update for changed builtins.
135 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
136 if all the modes match. Otherwise, fake it.
137 (get_element_number, ix86_expand_vec_init_builtin,
138 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
139 (ix86_expand_builtin): Make static. Update for changed builtins.
140 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
141 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
142 (ix86_expand_vector_init_duplicate): New.
143 (ix86_expand_vector_init_low_nonzero): New.
144 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
145 Split out from ix86_expand_vector_init; handle integer modes.
146 (ix86_expand_vector_init): Use them.
147 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
148 * config/i386/i386-protos.h: Update.
149 * config/i386/predicates.md (reg_or_0_operand): New.
150 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
151 (movv2sf_internal): Likewise. And a splitter to match them all.
152 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
153 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
154 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
155 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
156 (mmx_pinsrw): Fix operand ordering.
157 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
158 rather than sse_loadss expander.
159 (movv2df splitter): Similarly.
160 (sse_loadss, sse_loadlss): Remove.
161 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
162 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
163 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
164 (sse3_movddup): Rename with '*'.
165 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
166 (sse2_loadsd): Remove.
167 (vec_dupv2df_sse3): Rename from sse3_loadddup.
168 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
169 (sse2_pinsrw): Fix argument ordering.
170 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
171 (sse2_stored): Remove 'r' destination.
172 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
173 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
174 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
175 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
176 vec_extractv16qi, vec_initv16qi): New.
178 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
179 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
180 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
182 (_mm_load_pd, _mm_store_pd): Use plain dereference.
183 (_mm_load_si128, _mm_store_si128): Likewise.
184 (_mm_load1_pd): Use _mm_set1_pd.
185 (_mm_load_sd): Use _mm_set_sd.
186 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
187 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
188 (_mm_set_epi64): Use _mm_set_epi64x.
189 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
190 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
191 _mm_setr_epi8): Use _mm_set_foo form.
192 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
193 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
194 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
195 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
196 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
197 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
198 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
199 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
200 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
201 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
202 (_mm_movedup_pd): Use _mm_shuffle_pd.
203 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
204 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
205 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
206 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
208 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
209 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
210 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
211 (_mm_store_ps): Use raw dereference.
212 (_mm_store1_ps): Use _mm_storeu_ps.
213 (_mm_storer_ps): Use _mm_store_ps.
214 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
215 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
217 2005-01-11 Stan Shebs <shebs@apple.com>
219 * config/rs6000/rs6000.c (machopic_output_stub): Issue
220 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
222 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
225 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
226 Skip all DECLs except for FIELD_DECLs.
228 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
230 * config/s390/s390.c (override_options): Return error if
231 -mbackchain, -mpacked-stack and -mhard-float are used together.
232 (s390_va_start): Remove the backchain && packed-stack special case.
233 (s390_gimplify_va_arg): Likewise.
234 * doc/invoke.texi: Remove the ABI incompatibility note.
236 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
238 * config/s390/s390.c (struct s390_frame_layout): Remove
240 (s390_frame_info, s390_emit_prologue): Replace occurrences of
241 save_backchain_p with TARGET_BACKCHAIN.
243 2005-01-11 Alan Modra <amodra@bigpond.net.au>
246 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
247 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
248 * function.c (locate_and_pad_parm): Set new field.
249 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
250 Tweak where_pad test to include "none". Always set mem align for
252 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
253 not sufficient for type.
254 (assign_parm_setup_block): If stack_parm is zero on entry, always
255 make a new stack local. Block move old stack parm if necessary
256 to new aligned stack local.
257 (assign_parm_setup_stack): Use a block move to handle
258 potentially misaligned entry_parm.
259 (assign_parms_unsplit_complex): Specify required alignment when
260 creating stack local.
261 * calls.c (compute_argument_addresses): Override alignment of stack
262 arg calculated from its type with the alignment given by
263 FUNCTION_ARG_BOUNDARY.
264 (store_one_arg): Likewise.
266 2005-01-11 Jan Beulich <jbeulich@novell.com>
268 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
269 Change respective itanium_class attribute to ialu.
270 (shladdp4_internal): New.
271 * config/ia64/predicates.md (shladd_log2_operand): New.
273 2005-01-11 Richard Henderson <rth@redhat.com>
275 * expr.c (store_constructor): Use rtvec_alloc instead of
276 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
278 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
280 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
282 2005-01-11 Alan Modra <amodra@bigpond.net.au>
284 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
285 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
286 (TARGET_RWRELOC): Define.
287 (ia64_rwreloc_section_type_flags): Delete.
288 (ia64_section_type_flags): New function.
289 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
290 (TARGET_RWRELOC): Define.
292 2005-01-10 David Mosberger <davidm@hpl.hp.com>
295 * config/ia64/ia64.c (process_set): For alloc insn, only call
296 process_epilogue is !frame_pointer_needed.
298 2005-01-10 Roger Sayle <roger@eyesopen.com>
301 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
302 unary operator and can't be treated as a binary/comparison operator.
304 2005-01-10 Richard Henderson <rth@redhat.com>
306 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
307 (ix86_return_in_memory): Likewise.
308 (function_arg_advance): Likewise. Mirror structure in function_arg
309 for choosing register to advance.
311 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
313 * tree-vectorizer.c, tree.def: Fix comment typos.
315 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
317 * config/s390/s390.c (struct processor_costs): 4 new fields:
318 ddbr, ddr, debr, der.
319 (s390_rtx_costs): More precise handling of divide instructions.
321 2005-01-10 David Edelsohn <edelsohn@gnu.org>
323 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
325 2005-01-10 Jan Beulich <jbeulich@novell.com>
327 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
328 names resulting from -ffunction-sections/-fdata-sections and linkonce
330 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
331 based on the section name. Rearrange the section name comparison logic
332 slightly so that each section name is compared against at most once.
334 2005-01-10 Ben Elliston <bje@au.ibm.com>
336 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
337 targets for which -fPIC reduces limitations on the GOT size.
339 2005-01-09 Falk Hueffner <falk@debian.org>
341 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
344 2005-01-09 David Edelsohn <edelsohn@gnu.org>
347 * collect2.c (main): Set aixrtl_flag for -brtl option.
348 (resolve_lib_name): Search for .so file extension before .a
351 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
353 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
354 (vect_analyze_data_refs_alignment): Add dump prints.
356 2005-01-09 Ira Rosen <irar@il.ibm.com>
358 * tree-vectorizer.c (vect_analyze_offset_expr): Use
359 expr_invariant_in_loop_p.
360 Initialize outputs first thing in the function.
361 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
362 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
363 (vect_analyze_pointer_ref_access): Check that the initial condition of
364 the access function is loop invariant.
366 2005-01-09 Richard Henderson <rth@redhat.com>
368 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
369 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
370 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
371 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
372 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
373 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
374 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
375 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
376 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
377 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
378 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
379 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
380 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
381 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
382 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
383 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
384 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
385 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
386 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
387 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
388 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
389 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
390 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
392 (mmx_clrdi, pavgusb): Remove.
393 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
394 with leading sse_ prefix.
395 * config/i386/sse.md: Receive them.
396 * config/i386/mmx.md: New file.
397 (MMXMODE12, MMXMODE24, mmxvecsize): New.
398 (subrv2sf3): Turn into expander for normal subtraction.
399 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
400 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
401 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
402 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
403 commutative; use ix86_binary_operator_ok.
404 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
405 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
406 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
407 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
408 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
409 Macroize from existing patterns; use ix86_binary_operator_ok.
410 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
411 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
412 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
413 with vec_select+vec_concat.
414 (mmx_pshufw, mmx_pshufw_1): Likewise.
415 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
416 (mmx_uavgv4hi3): Model correctly.
417 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
418 _mm_xor_si64): Remove casts.
420 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
422 PR tree-optimization/19224
423 * tree-scalar-evolution.c (get_instantiated_value,
424 set_instantiated_value): New functions.
425 (instantiate_parameters_1): Cache the results.
426 (instantiate_parameters, resolve_mixers): Initialize and free
429 2005-01-08 David Edelsohn <edelsohn@gnu.org>
431 * config/i386/i386.md (addhi_4): Correct reference in comment.
434 2005-01-08 Richard Henderson <rth@redhat.com>
436 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
437 (_mm_cvtsi128_si64x): Likewise.
438 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
439 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
440 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
441 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
442 * config/i386/i386-protos.h: Update.
443 * config/i386/i386.c (print_operand): Add 'H'.
444 (ix86_fixup_binary_operands): Split out from ...
445 (ix86_expand_binary_operator): ... here.
446 (ix86_fixup_binary_operands_no_copy): New.
447 (ix86_expand_fp_absneg_operator): Handle vector mode results.
448 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
449 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
450 (safe_vector_operand): Use CONST0_RTX.
451 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
452 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
453 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
454 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
456 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
457 (IX86_BUILTIN_CMPNGESS): New.
458 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
459 (attr type): Add sselog1.
460 (attr unit, attr memory): Handle it.
461 (movti, movti_internal, movti_rex64): Move near other integer moves.
462 (movtf, movtf_internal): Move near other fp moves.
463 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
464 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
465 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
466 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
467 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
468 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
469 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
470 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
471 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
472 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
473 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
474 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
475 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
476 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
477 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
478 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
479 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
480 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
481 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
482 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
483 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
484 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
485 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
486 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
487 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
488 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
489 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
490 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
491 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
492 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
493 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
494 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
495 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
496 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
497 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
498 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
499 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
500 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
501 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
502 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
503 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
504 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
505 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
506 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
508 (SSEPUSH, push<SSEPUSH>): Remove.
509 (MMXPUSH, push<MMXPUSH>): Remove.
510 (sse_movaps, sse_movaps_1, sse_movups): Remove.
511 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
512 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
513 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
514 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
515 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
516 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
517 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
518 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
519 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
520 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
521 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
522 (ppro_sse_log_V4SF): Handle sselog1.
523 * config/i386/predicates.md (const_0_to_1_operand): New.
524 (const_0_to_255_mul_8_operand): New.
525 (const_1_to_31_operand): Rename from const_int_1_31_operand.
526 (const_2_to_3_operand, const_4_to_7_operand): New.
527 * config/i386/sse.md: New file.
528 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
529 (sse_movups): Rename from sse_movups_1.
530 (sse_loadlss): Rename from sse_loadss_1.
531 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
533 (negv4sf2): Use ix86_expand_fp_absneg_operator.
534 (absv4sf2, negv2df, absv2df): New.
535 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
536 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
537 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
538 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
539 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
540 (sse3_addsubv4sf3): Model correctly.
541 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
542 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
543 (sse_movhlps): Model with vec_select+vec_concat.
544 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
545 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
546 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
547 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
548 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
549 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
550 sse2_pshufhw_1): Likewise.
551 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
552 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
553 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
554 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
555 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
556 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
557 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
558 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
559 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
560 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
561 sminv8hi3): Mark commutative
562 operands. Use ix86_binary_operator_ok.
563 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
564 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
565 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
566 sse2_punpcklqdq): Allow operand2 in memory.
567 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
568 sse2_movsd): Add memory alternatives.
569 (sse_storelps): Turn expander into an insn; split after reload.
570 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
571 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
573 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
575 * configure.ac (DWARF-2 debug_line): Use objdump.
576 * configure: Regenerate.
578 2005-01-08 Jeff Law <law@redhat.com>
579 Diego Novillo <dnovillo@redhat.com>
581 PR tree-optimization/18241
582 * tree-nrv.c (tree_nrv): Ignore volatile return values.
583 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
584 statements with volatile operands.
585 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
586 operands after marking a statement with has_volatile_ops.
588 2005-01-08 Roger Sayle <roger@eyesopen.com>
590 * tree.c (int_fits_type_p): Always honor integer constant
591 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
593 2005-01-08 Roger Sayle <roger@eyesopen.com>
595 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
596 the latent bug in rtl_delete_block has been resolved.
598 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
600 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
601 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
603 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
605 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
606 library as @multilib_dir@/$(SHLIB_SONAME).
608 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
610 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
611 * configure: Regenerate.
613 2005-01-07 Jakub Jelinek <jakub@redhat.com>
615 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
618 2005-01-07 David Edelsohn <edelsohn@gnu.org>
621 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
622 Convert non-word aligned offset address using ld/std into
625 2005-01-07 Richard Henderson <rth@redhat.com>
627 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
629 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
631 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
633 2005-01-07 Jakub Jelinek <jakub@redhat.com>
635 PR tree-optimization/19283
636 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
637 is not shorter than the original type.
639 PR rtl-optimization/19012
640 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
642 PR rtl-optimization/18861
643 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
644 (FULL_STATE): Define.
645 (mark_tablejump_edge): New function.
646 (purge_dead_tablejump_edges): New function.
647 (find_bb_boundaries): Use it.
649 PR tree-optimization/18828
650 * builtins.c (expand_builtin_next_arg): Remove argument and all
651 the argument checking.
652 (expand_builtin): Adjust caller.
653 (expand_builtin_va_start): Likewise. Remove error for too many
655 (fold_builtin_next_arg): Issue error for too many arguments.
656 After checking arguments, replace them with magic arguments that
657 prevent further checking of the args.
659 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
661 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
663 2005-01-06 Roger Sayle <roger@eyesopen.com>
666 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
667 and %{.S:...} (and their negative variants) to test whether the
668 input file is assembler or pre-processed-assembler independent of
669 the actual filename extension.
671 2005-01-06 Roger Sayle <roger@eyesopen.com>
673 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
674 of sign or zero extended values.
676 2005-01-06 Geoffrey Keating <geoffk@apple.com>
678 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
679 for IBM long double format correctly.
681 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
683 Fix PR tree-optimization/18792
685 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
686 to first_loop_depth, and use loop depth instead of loop number.
687 (build_classic_dir_vector): Ditto.
688 (compute_data_dependences_for_loop): Use depth, not loop number.
689 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
690 number. Pass in loops, instead of loop numbers.
691 (gather_interchange_stats): Ditto.
692 (linear_transform_loops): Ditto.
694 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
697 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
698 biv_fits_mode_p, extension_within_bounds_p): New functions.
699 (check_ext_dependent_givs): Use them.
701 2005-01-06 Roger Sayle <roger@eyesopen.com>
703 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
704 more than one barrier, in which case we should delete them all.
706 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
708 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
710 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
712 Fix PR middle-end/19286
714 * dwarf2out.c (gen_subprogram_die): If we've already tried to
715 output this subprogram, simply ignore this attempt to do it again.
716 (add_abstract_origin_attribute): Don't abort trying to add the abstract
717 origin attribute if it's not possible.
718 (gen_block_die): Don't ignore subblocks of "unused" blocks.
719 (decls_for_scope): Ditto.
720 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
721 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
723 2005-01-05 Richard Henderson <rth@redhat.com>
726 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
727 (bdesc_2arg): Use it.
728 (ix86_expand_binop_builtin): Force operands into registers
730 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
731 ix86_expand_sse_compare, ix86_expand_sse_comi,
732 ix86_expand_builtin): Likewise.
734 2005-01-05 Richard Henderson <rth@redhat.com>
736 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
738 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
739 (mulv4hi3): Set itanium_class mmmul.
740 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
741 (fpack): Rename from fpack_sfsf.
743 2005-01-05 Richard Henderson <rth@redhat.com>
746 * reload1.c (do_input_reload): Restrict the optimization deleteing
747 a previous output reload to RELOAD_FOR_INPUT.
749 2005-01-05 Steven Bosscher <stevenb@suse.de>
751 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
752 comment that we fall through after case.
753 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
754 (distribute_links): Likewise.
755 * cse.c (cse_insn): Likewise.
756 * cselib.c (cselib_invalidate_mem): Likewise.
757 * df.c: Update comments at the top of the file.
758 (read_modify_subreg_p): Update comments here too.
759 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
760 * flow.c (mark_set_1): Likewise.
761 (mark_used_regs): Likewise.
762 * gcse.c (mems_conflict_for_gcse_p): Likewise.
763 (canon_list_insert): Likewise.
764 (mark_set): Likewise.
765 (try_replace_reg): Likewise.
766 (store_killed_in_insn): Likewise.
767 * loop.c (count_one_set): Likewise.
768 (basic_induction_var): Likewise.
769 * postreload-gcse.c (find_mem_conflicts): Likewise.
770 * postreload.c (reload_combine_note_store): Likewise.
771 (move2add_note_store): Likewise.
772 * reload.c (find_equiv_reg): Likewise.
773 (mark_referenced_resources): Likewise.
774 * rtlanal.c (set_noop_p): Likewise.
775 (note_stores): Likewise.
776 (note_uses): Likewise.
777 * sched-deps.c (sched_analyze_1): Likewise.
778 * sched-rgn.c (check_live_1): Likewise.
779 (update_live_1): Likewise.
780 * config/i860/i860.c: Likewise.
782 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
784 (ZERO_EXTRACT): Mention that this one can be an lvalue.
786 * doc/rtl.texi: Update documentation for bit-fields and SET.
788 2005-01-05 Roger Sayle <roger@eyesopen.com>
790 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
792 2005-01-05 Jan Hubicka <jh@suse.cz>
793 Richard Henderson <rth@redhat.com>
796 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
799 2005-01-05 Richard Henderson <rth@redhat.com>
802 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
803 (sse_shufps): Change operand 3 to const_int_operand.
804 (sse2_storelps): Fix typo in template.
805 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
806 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
807 (ix86_expand_builtin): Likewise.
809 2005-01-05 Julian Brown <julian@codesourcery.com>
811 * config/arm/arm.c (arm_return_in_memory): Treat complex types
812 as aggregates for AAPCS ABIs.
814 2005-01-05 Stan Shebs <shebs@apple.com>
816 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
817 Mach-O getters if ppc64.
818 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
820 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
821 signed four-byte field for global code case.
822 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
824 2005-01-05 Roger Sayle <roger@eyesopen.com>
827 * c-common.c: Include real.h.
828 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
829 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
830 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
831 * Makefile.in (c-common.o): Update dependencies.
833 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
835 * c-parse.in (asm_string): Add trailing semicolon.
837 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
839 * c-parse.in (asm_string): New. Don't allow wide strings in
841 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
842 asm_string instead of STRING.
844 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
846 * c-typeck.c (constructor_no_implicit): Remove.
847 (set_designator, process_init_element): Don't check
848 constructor_no_implicit.
850 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
851 Kaz Kojima <kkojima@gcc.gnu.org>
854 * lcm.c (create_pre_exit): New.
855 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
856 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
858 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
860 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
861 -msoft-float in target_float_switch.
862 * arm.c (arm_override_options): Fix processing of target_float_switch.
864 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
866 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
869 2004-01-05 Uros Bizjak <uros@kss-loka.si>
871 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
872 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
874 2005-01-04 Roger Sayle <roger@eyesopen.com>
876 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
877 if deleting the then-block allows the test-block to fallthru to the
880 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
883 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
884 after the full declaration if the are in two different TUs.
886 2005-01-04 Richard Henderson <rth@redhat.com>
889 * tree-sra.c (generate_one_element_init): Just
890 call gimplify_and_add.
891 (generate_element_init): Record the
892 new referenced variables and mark them for renaming
894 (generate_element_init_1): This.
895 (scalarize_init): Don't call push_gimplify_context/
896 pop_gimplify_context.
898 2005-01-04 Geoffrey Keating <geoffk@apple.com>
900 * toplev.c (get_src_pwd): Handle failure of getpwd().
902 2005-01-04 Roger Sayle <roger@eyesopen.com>
904 * fold-const.c (fold_single_bit_test): Delete unreachable handling
906 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
907 all the cases handled by it are inverted by invert_truthvalue.
909 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
911 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
912 (save_register_windowsi): Likewise.
914 2005-01-04 Richard Henderson <rth@redhat.com>
916 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
918 2005-01-04 Uros Bizjak <uros@kss-loka.si>
921 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
922 any simplifications of vector mode comparison operators.
923 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
925 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
926 Devang Patel <dpatel@apple.com>
928 PR tree-optimization/18308
929 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
930 the operands before creating a new expression.
931 * dojump.c (do_jump): Make drop_through_label available
932 for all cases. Add expansion of COND_EXPR.
934 2005-01-04 Ira Rosen <irar@il.ibm.com>
936 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
937 INTEGER_CST instead of TREE_CONSTANT.
938 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
939 instead of TREE_CONSTANT.
940 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
941 instead of TREE_CONSTANT.
943 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
945 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
947 2005-01-04 Uros Bizjak <uros@kss-loka.si>
950 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
951 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
953 2005-01-03 Richard Henderson <rth@redhat.com>
955 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
956 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
957 fold_convert_const_real_from_real): Split out from ...
958 (fold_convert_const): ... here.
960 2005-01-03 Richard Henderson <rth@redhat.com>
963 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
964 (mov<MMXMODEI>_internal): Likewise.
965 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
966 Add V2SF case; use it for SSE1; don't use TI.
967 (movdf_integer): Likewise.
968 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
970 2005-01-03 Ira Rosen <irar@il.ibm.com>
972 * tree-vectorizer.c (vect_strip_conversions): New function.
973 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
974 check for binary class.
976 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
980 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
982 * gimple-low.c (mark_blocks_with_used_vars): New function.
983 (mark_blocks_with_used_subblocks): Ditto.
984 (mark_used_blocks): Ditto.
985 (pass_mark_used_blocks): New pass.
986 * tree-inline.c: Include debug.h.
987 (expand_call_inline): Call outlining_inline_function here.
988 * tree-optimize.c (init_tree_optimization_passes): Add
989 pass_mark_used_blocks.
990 * tree-pass.h (pass_mark_used_blocks): New.
991 * Makefile.in (tree-inline.o): Add debug.h dependency.
993 2005-01-03 Geoffrey Keating <geoffk@apple.com>
995 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
996 VAR_DECLs to have weak_import attribute.
998 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
1001 2004-01-03 Steven Bosscher <stevenb@suse.de>
1003 * passes.c (rest_of_compilation): Don't run regmove if only
1004 flag_expensive_optimizations, require flag_regmove instead.
1006 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
1008 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
1010 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
1011 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1012 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
1013 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1015 2005-01-03 Richard Henderson <rth@redhat.com>
1017 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
1018 (ia64_const_ok_for_letter_p): New.
1019 (ia64_const_double_ok_for_letter_p): New.
1020 (ia64_extra_constraint): New.
1021 (ia64_expand_vecint_compare): New.
1022 (ia64_expand_vcondu_v2si): New.
1023 (ia64_expand_vecint_cmov): New.
1024 (ia64_expand_vecint_minmax): New.
1025 (ia64_print_operand): Add 'v'.
1026 (ia64_preferred_reload_class): New.
1027 (ia64_vector_mode_supported_p): New.
1028 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
1029 (PREFERRED_RELOAD_CLASS): Move to function.
1030 (CONST_OK_FOR_LETTER_P): Move to function.
1031 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
1032 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
1033 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
1034 (EXTRA_CONSTRAINT): Move to function.
1035 * config/ia64/ia64.md: Include vect.md.
1036 (itanium_class): Add mmalua.
1038 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
1040 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
1041 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
1043 (const_int_2bit_operand): New.
1044 (fr_reg_or_0_operand): New.
1045 * config/ia64/ia64-modes.def: Add vector modes.
1046 * config/ia64/ia64-protos.h: Update.
1047 * config/ia64/vect.md: New file.
1049 2005-01-03 Richard Henderson <rth@redhat.com>
1051 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
1053 2005-01-03 Uros Bizjak <uros@kss-loka.si>
1056 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
1058 (log1pdf2): Change mode of operands[0,1] to DFmode.
1060 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
1062 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
1063 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
1064 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
1065 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
1066 SPARC_RELAXED_ORDERING.
1068 2005-01-03 Richard Henderson <rth@redhat.com>
1069 Uros Bizjak <uros@kss-loka.si>
1072 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
1073 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
1074 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
1076 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
1077 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
1078 const_pow2_1_to_128_operand): New predicates.
1080 2005-01-02 Greg McGary <greg@mcgary.org>
1082 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
1084 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
1085 object. [ptr_type]: Remove unused variable.
1086 Remember nearest addressable array-element or record-component
1087 when checking bitfield components. Tolerate empty BIND_EXPR.
1089 2005-01-01 Richard Henderson <rth@redhat.com>
1091 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
1092 that don't narrow the value. Fail for other conversions.
1094 2005-01-01 Richard Henderson <rth@redhat.com>
1097 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
1098 * c-lang.c (finish_file): Don't do it here.
1099 * objc/objc-act.c (objc_finish_file): Likewise.
1101 * cgraph.c (decl_assembler_name_equal): New.
1102 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
1103 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
1104 * cgraph.h (struct cgraph_varpool_node): Add next.
1105 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
1106 * varasm.c (assemble_alias): Mark the target as needed.
1108 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
1111 * function.c (get_arg_pointer_save_area): Use entry_of_function
1112 instead of get_insns.
1114 2005-01-01 Roger Sayle <roger@eyesopen.com>
1115 Andrew Pinski <pinskia@physics.uc.edu>
1116 James E. Wilson <wilson@specifixinc.com>
1118 PR rtl-optimization/12092
1119 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
1122 2005-01-01 Roger Sayle <roger@eyesopen.com>
1123 Olivier Hainque <hainque@act-europe.fr>
1125 * tree.c (int_fits_type_p): A narrower type always fits in a
1126 wider one, except for negative values into unsigned types.
1128 2005-01-01 Roger Sayle <roger@eyesopen.com>
1130 * tree.c (int_fits_type_p): Compare the result of force_fit_type
1131 with the original constant rather than require TREE_OVERFLOW.
1133 2005-01-01 Steven Bosscher <stevenb@suse.de>
1136 * c-decl.c (finish_function): If compiling C99, annotate the
1137 compiler generated return with the current file name and line 0.
1138 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
1139 the source line is greater than 0.
1140 (remove_bb): Likewise.
1142 See ChangeLog.12 for earlier changes.