1 2005-01-13 Richard Henderson <rth@redhat.com>
6 * config/i386/i386.md (cmpdf, cmpsf, bunordered, bordered, buneq,
7 bunge, bungt, bunle, bunlt, bltgt): Enable for TARGET_SSE_MATH,
9 (cmpfp_i_387): Rename from cmpfp_i. Move after sse patterns.
10 (cmpfp_i_mixed): Rename from cmpfp_i_sse; use for TARGET_MIX_SSE_I387.
11 (cmpfp_i_sse): Rename from cmpfp_i_sse_only; use for TARGET_SSE_MATH.
12 (cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387): Similarly.
13 (fp_jcc_1_mixed, fp_jcc_1_sse, fp_jcc_1_387): Similarly.
14 (fp_jcc_2_mixed, fp_jcc_2_sse, fp_jcc_2_387): Similarly.
15 (fp_jcc_3_387, fp_jcc_4_387, fp_jcc_5_387, fp_jcc_6_387,
16 fp_jcc_7_387, fp_jcc_8_387): Rename from fp_jcc_N.
17 (movdicc_c_rex64): Rename with '*'.
18 (movsfcc, movdfcc): Add checks for 387 and sse math to condition.
19 (movsfcc_1_sse_min, movsfcc_1_sse_max, movsfcc_1_sse): New.
20 (movsfcc_1_387): Rename from movsfcc_1.
21 (movdfcc_1_sse_min, movdfcc_1_sse_max, movdfcc_1_sse): New.
22 (movdfcc_1, movdfcc_1_rex64): Add check for 387.
23 (sminsf3, smaxsf3, smindf3, smaxdf3): New.
24 (minsf3, minsf, minsf_nonieee, minsf_sse, mindf3, mindf,
25 mindf_nonieee, mindf_sse, maxsf3, maxsf, maxsf_nonieee, maxsf_sse,
26 maxdf3, maxdf, maxdf_nonieee, maxdf_sse, sse_movsfcc, sse_movsfcc_eq,
27 sse_movdfcc, sse_movdfcc_eq, sse_movsfcc_const0_1,
28 sse_movsfcc_const0_2, sse_movsfcc_const0_3, sse_movsfcc_const0_4,
29 sse_movdfcc_const0_1, sse_movdfcc_const0_2, sse_movdfcc_const0_3,
30 sse_movdfcc_const0_4): Remove.
31 * config/i386/i386.c (ix86_expand_fp_movcc): For TARGET_SSE_MATH,
32 recognize min/max early. Update for changed sse cmove patterns.
33 (ix86_split_sse_movcc): New.
34 * config/i386/i386-protos.h: Update.
36 2005-01-13 Steven Bosscher <stevenb@suse.de>
38 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
39 (fix_stmt_v_may_defs): Likewise.
41 2005-01-13 Richard Henderson <rth@redhat.com>
43 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
44 for converting the mask.
46 2005-01-13 David Edelsohn <edelsohn@gnu.org>
48 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
51 2005-01-13 Richard Henderson <rth@redhat.com>
53 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
54 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
55 IX86_BUILTIN_MOVDQ2Q): Remove.
56 (IX86_BUILTIN_VEC_EXT_V4SI): New.
57 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
58 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
60 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
61 (sse2_pextrw): Fix immediate constraint.
62 (sse2_loadq, sse2_loadq_rex64): Remove.
63 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
64 Use __builtin_ia32_vec_ext_<size>.
65 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
67 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
69 * function.c (assign_parm_setup_block): Look inside original
70 entry_parm when inspecting PARALLEL.
72 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
73 Joel Sherrill <joel@oarcorp.com>
76 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
77 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
78 rtems_gxx_recursive_mutex_init.
79 (__gthread_recursive_mutex_lock): New function.
80 (__gthread_recursive_mutex_trylock): Likewise.
81 (__gthread_recursive_mutex_unlock): Likewise.
83 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
85 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
87 2005-01-13 Jan Beulich <jbeulich@novell.com>
89 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
90 (ashrti3_internal): Indicate output is early clobber. Generate result
91 into output rather than first input. Use move for low word of output
92 if shift count is exactly 64.
93 (lshrti3_internal): Likewise.
95 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
99 * reload1.c (delete_output_reload): Don't delete an output reload
100 if the pseudo lives longer than a single basic block. Adjust and
101 improve wording of comment.
103 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
106 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
107 of TYPE_SIZE where we mean the number of bytes.
109 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
111 * function.c (assign_parm_setup_block): Relax condition on
112 multi-register optimization.
114 2005-01-12 Nick Clifton <nickc@redhat.com>
116 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
117 an "in_delay_slot" attribute of "no" to prevent them being used in
118 delay slots. This is forbidden because they might generate
121 2005-01-12 Alan Modra <amodra@bigpond.net.au>
124 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
127 2005-01-12 Nick Clifton <nickc@redhat.com>
129 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
130 passed on to GAS) but do not define, as GAS no longer supports or
131 needs the -m2000 option.
133 2005-01-11 Roger Sayle <roger@eyesopen.com>
135 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
136 type as args of unprototyped builtins aren't automatically promoted.
137 (fold_builtin_abs): Likewise.
139 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
141 * Makefile.in: Set a `build-warn' variable.
143 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
145 PR tree-optimization/17949
146 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
147 (find_interesting_uses_address): Use it.
149 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
151 * regrename.c (kill_value): Handle subreg's that won't simplify.
153 2005-01-11 Richard Henderson <rth@redhat.com>
156 * config/i386/i386.h (enum ix86_builtins): Move ...
157 * config/i386/i386.c: ... here.
158 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
159 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
160 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
161 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
162 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
163 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
164 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
165 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
166 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
167 IX86_BUILTIN_LOADDDUP): Remove.
168 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
169 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
170 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
171 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
172 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
173 (ix86_init_builtins): Make static.
174 (ix86_init_mmx_sse_builtins): Update for changed builtins.
175 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
176 if all the modes match. Otherwise, fake it.
177 (get_element_number, ix86_expand_vec_init_builtin,
178 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
179 (ix86_expand_builtin): Make static. Update for changed builtins.
180 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
181 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
182 (ix86_expand_vector_init_duplicate): New.
183 (ix86_expand_vector_init_low_nonzero): New.
184 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
185 Split out from ix86_expand_vector_init; handle integer modes.
186 (ix86_expand_vector_init): Use them.
187 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
188 * config/i386/i386-protos.h: Update.
189 * config/i386/predicates.md (reg_or_0_operand): New.
190 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
191 (movv2sf_internal): Likewise. And a splitter to match them all.
192 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
193 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
194 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
195 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
196 (mmx_pinsrw): Fix operand ordering.
197 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
198 rather than sse_loadss expander.
199 (movv2df splitter): Similarly.
200 (sse_loadss, sse_loadlss): Remove.
201 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
202 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
203 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
204 (sse3_movddup): Rename with '*'.
205 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
206 (sse2_loadsd): Remove.
207 (vec_dupv2df_sse3): Rename from sse3_loadddup.
208 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
209 (sse2_pinsrw): Fix argument ordering.
210 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
211 (sse2_stored): Remove 'r' destination.
212 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
213 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
214 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
215 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
216 vec_extractv16qi, vec_initv16qi): New.
218 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
219 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
220 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
222 (_mm_load_pd, _mm_store_pd): Use plain dereference.
223 (_mm_load_si128, _mm_store_si128): Likewise.
224 (_mm_load1_pd): Use _mm_set1_pd.
225 (_mm_load_sd): Use _mm_set_sd.
226 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
227 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
228 (_mm_set_epi64): Use _mm_set_epi64x.
229 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
230 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
231 _mm_setr_epi8): Use _mm_set_foo form.
232 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
233 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
234 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
235 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
236 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
237 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
238 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
239 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
240 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
241 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
242 (_mm_movedup_pd): Use _mm_shuffle_pd.
243 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
244 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
245 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
246 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
248 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
249 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
250 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
251 (_mm_store_ps): Use raw dereference.
252 (_mm_store1_ps): Use _mm_storeu_ps.
253 (_mm_storer_ps): Use _mm_store_ps.
254 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
255 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
257 2005-01-11 Stan Shebs <shebs@apple.com>
259 * config/rs6000/rs6000.c (machopic_output_stub): Issue
260 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
262 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
265 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
266 Skip all DECLs except for FIELD_DECLs.
268 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
270 * config/s390/s390.c (override_options): Return error if
271 -mbackchain, -mpacked-stack and -mhard-float are used together.
272 (s390_va_start): Remove the backchain && packed-stack special case.
273 (s390_gimplify_va_arg): Likewise.
274 * doc/invoke.texi: Remove the ABI incompatibility note.
276 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
278 * config/s390/s390.c (struct s390_frame_layout): Remove
280 (s390_frame_info, s390_emit_prologue): Replace occurrences of
281 save_backchain_p with TARGET_BACKCHAIN.
283 2005-01-11 Alan Modra <amodra@bigpond.net.au>
286 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
287 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
288 * function.c (locate_and_pad_parm): Set new field.
289 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
290 Tweak where_pad test to include "none". Always set mem align for
292 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
293 not sufficient for type.
294 (assign_parm_setup_block): If stack_parm is zero on entry, always
295 make a new stack local. Block move old stack parm if necessary
296 to new aligned stack local.
297 (assign_parm_setup_stack): Use a block move to handle
298 potentially misaligned entry_parm.
299 (assign_parms_unsplit_complex): Specify required alignment when
300 creating stack local.
301 * calls.c (compute_argument_addresses): Override alignment of stack
302 arg calculated from its type with the alignment given by
303 FUNCTION_ARG_BOUNDARY.
304 (store_one_arg): Likewise.
306 2005-01-11 Jan Beulich <jbeulich@novell.com>
308 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
309 Change respective itanium_class attribute to ialu.
310 (shladdp4_internal): New.
311 * config/ia64/predicates.md (shladd_log2_operand): New.
313 2005-01-11 Richard Henderson <rth@redhat.com>
315 * expr.c (store_constructor): Use rtvec_alloc instead of
316 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
318 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
320 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
322 2005-01-11 Alan Modra <amodra@bigpond.net.au>
324 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
325 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
326 (TARGET_RWRELOC): Define.
327 (ia64_rwreloc_section_type_flags): Delete.
328 (ia64_section_type_flags): New function.
329 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
330 (TARGET_RWRELOC): Define.
332 2005-01-10 David Mosberger <davidm@hpl.hp.com>
335 * config/ia64/ia64.c (process_set): For alloc insn, only call
336 process_epilogue is !frame_pointer_needed.
338 2005-01-10 Roger Sayle <roger@eyesopen.com>
341 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
342 unary operator and can't be treated as a binary/comparison operator.
344 2005-01-10 Richard Henderson <rth@redhat.com>
346 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
347 (ix86_return_in_memory): Likewise.
348 (function_arg_advance): Likewise. Mirror structure in function_arg
349 for choosing register to advance.
351 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
353 * tree-vectorizer.c, tree.def: Fix comment typos.
355 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
357 * config/s390/s390.c (struct processor_costs): 4 new fields:
358 ddbr, ddr, debr, der.
359 (s390_rtx_costs): More precise handling of divide instructions.
361 2005-01-10 David Edelsohn <edelsohn@gnu.org>
363 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
365 2005-01-10 Jan Beulich <jbeulich@novell.com>
367 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
368 names resulting from -ffunction-sections/-fdata-sections and linkonce
370 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
371 based on the section name. Rearrange the section name comparison logic
372 slightly so that each section name is compared against at most once.
374 2005-01-10 Ben Elliston <bje@au.ibm.com>
376 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
377 targets for which -fPIC reduces limitations on the GOT size.
379 2005-01-09 Falk Hueffner <falk@debian.org>
381 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
384 2005-01-09 David Edelsohn <edelsohn@gnu.org>
387 * collect2.c (main): Set aixrtl_flag for -brtl option.
388 (resolve_lib_name): Search for .so file extension before .a
391 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
393 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
394 (vect_analyze_data_refs_alignment): Add dump prints.
396 2005-01-09 Ira Rosen <irar@il.ibm.com>
398 * tree-vectorizer.c (vect_analyze_offset_expr): Use
399 expr_invariant_in_loop_p.
400 Initialize outputs first thing in the function.
401 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
402 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
403 (vect_analyze_pointer_ref_access): Check that the initial condition of
404 the access function is loop invariant.
406 2005-01-09 Richard Henderson <rth@redhat.com>
408 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
409 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
410 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
411 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
412 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
413 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
414 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
415 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
416 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
417 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
418 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
419 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
420 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
421 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
422 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
423 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
424 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
425 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
426 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
427 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
428 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
429 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
430 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
432 (mmx_clrdi, pavgusb): Remove.
433 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
434 with leading sse_ prefix.
435 * config/i386/sse.md: Receive them.
436 * config/i386/mmx.md: New file.
437 (MMXMODE12, MMXMODE24, mmxvecsize): New.
438 (subrv2sf3): Turn into expander for normal subtraction.
439 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
440 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
441 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
442 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
443 commutative; use ix86_binary_operator_ok.
444 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
445 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
446 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
447 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
448 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
449 Macroize from existing patterns; use ix86_binary_operator_ok.
450 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
451 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
452 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
453 with vec_select+vec_concat.
454 (mmx_pshufw, mmx_pshufw_1): Likewise.
455 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
456 (mmx_uavgv4hi3): Model correctly.
457 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
458 _mm_xor_si64): Remove casts.
460 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
462 PR tree-optimization/19224
463 * tree-scalar-evolution.c (get_instantiated_value,
464 set_instantiated_value): New functions.
465 (instantiate_parameters_1): Cache the results.
466 (instantiate_parameters, resolve_mixers): Initialize and free
469 2005-01-08 David Edelsohn <edelsohn@gnu.org>
471 * config/i386/i386.md (addhi_4): Correct reference in comment.
474 2005-01-08 Richard Henderson <rth@redhat.com>
476 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
477 (_mm_cvtsi128_si64x): Likewise.
478 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
479 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
480 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
481 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
482 * config/i386/i386-protos.h: Update.
483 * config/i386/i386.c (print_operand): Add 'H'.
484 (ix86_fixup_binary_operands): Split out from ...
485 (ix86_expand_binary_operator): ... here.
486 (ix86_fixup_binary_operands_no_copy): New.
487 (ix86_expand_fp_absneg_operator): Handle vector mode results.
488 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
489 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
490 (safe_vector_operand): Use CONST0_RTX.
491 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
492 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
493 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
494 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
496 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
497 (IX86_BUILTIN_CMPNGESS): New.
498 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
499 (attr type): Add sselog1.
500 (attr unit, attr memory): Handle it.
501 (movti, movti_internal, movti_rex64): Move near other integer moves.
502 (movtf, movtf_internal): Move near other fp moves.
503 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
504 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
505 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
506 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
507 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
508 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
509 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
510 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
511 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
512 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
513 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
514 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
515 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
516 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
517 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
518 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
519 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
520 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
521 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
522 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
523 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
524 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
525 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
526 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
527 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
528 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
529 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
530 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
531 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
532 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
533 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
534 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
535 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
536 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
537 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
538 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
539 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
540 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
541 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
542 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
543 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
544 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
545 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
546 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
548 (SSEPUSH, push<SSEPUSH>): Remove.
549 (MMXPUSH, push<MMXPUSH>): Remove.
550 (sse_movaps, sse_movaps_1, sse_movups): Remove.
551 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
552 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
553 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
554 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
555 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
556 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
557 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
558 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
559 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
560 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
561 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
562 (ppro_sse_log_V4SF): Handle sselog1.
563 * config/i386/predicates.md (const_0_to_1_operand): New.
564 (const_0_to_255_mul_8_operand): New.
565 (const_1_to_31_operand): Rename from const_int_1_31_operand.
566 (const_2_to_3_operand, const_4_to_7_operand): New.
567 * config/i386/sse.md: New file.
568 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
569 (sse_movups): Rename from sse_movups_1.
570 (sse_loadlss): Rename from sse_loadss_1.
571 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
573 (negv4sf2): Use ix86_expand_fp_absneg_operator.
574 (absv4sf2, negv2df, absv2df): New.
575 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
576 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
577 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
578 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
579 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
580 (sse3_addsubv4sf3): Model correctly.
581 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
582 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
583 (sse_movhlps): Model with vec_select+vec_concat.
584 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
585 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
586 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
587 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
588 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
589 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
590 sse2_pshufhw_1): Likewise.
591 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
592 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
593 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
594 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
595 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
596 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
597 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
598 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
599 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
600 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
601 sminv8hi3): Mark commutative
602 operands. Use ix86_binary_operator_ok.
603 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
604 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
605 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
606 sse2_punpcklqdq): Allow operand2 in memory.
607 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
608 sse2_movsd): Add memory alternatives.
609 (sse_storelps): Turn expander into an insn; split after reload.
610 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
611 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
613 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
615 * configure.ac (DWARF-2 debug_line): Use objdump.
616 * configure: Regenerate.
618 2005-01-08 Jeff Law <law@redhat.com>
619 Diego Novillo <dnovillo@redhat.com>
621 PR tree-optimization/18241
622 * tree-nrv.c (tree_nrv): Ignore volatile return values.
623 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
624 statements with volatile operands.
625 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
626 operands after marking a statement with has_volatile_ops.
628 2005-01-08 Roger Sayle <roger@eyesopen.com>
630 * tree.c (int_fits_type_p): Always honor integer constant
631 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
633 2005-01-08 Roger Sayle <roger@eyesopen.com>
635 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
636 the latent bug in rtl_delete_block has been resolved.
638 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
640 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
641 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
643 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
645 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
646 library as @multilib_dir@/$(SHLIB_SONAME).
648 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
650 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
651 * configure: Regenerate.
653 2005-01-07 Jakub Jelinek <jakub@redhat.com>
655 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
658 2005-01-07 David Edelsohn <edelsohn@gnu.org>
661 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
662 Convert non-word aligned offset address using ld/std into
665 2005-01-07 Richard Henderson <rth@redhat.com>
667 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
669 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
671 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
673 2005-01-07 Jakub Jelinek <jakub@redhat.com>
675 PR tree-optimization/19283
676 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
677 is not shorter than the original type.
679 PR rtl-optimization/19012
680 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
682 PR rtl-optimization/18861
683 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
684 (FULL_STATE): Define.
685 (mark_tablejump_edge): New function.
686 (purge_dead_tablejump_edges): New function.
687 (find_bb_boundaries): Use it.
689 PR tree-optimization/18828
690 * builtins.c (expand_builtin_next_arg): Remove argument and all
691 the argument checking.
692 (expand_builtin): Adjust caller.
693 (expand_builtin_va_start): Likewise. Remove error for too many
695 (fold_builtin_next_arg): Issue error for too many arguments.
696 After checking arguments, replace them with magic arguments that
697 prevent further checking of the args.
699 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
701 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
703 2005-01-06 Roger Sayle <roger@eyesopen.com>
706 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
707 and %{.S:...} (and their negative variants) to test whether the
708 input file is assembler or pre-processed-assembler independent of
709 the actual filename extension.
711 2005-01-06 Roger Sayle <roger@eyesopen.com>
713 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
714 of sign or zero extended values.
716 2005-01-06 Geoffrey Keating <geoffk@apple.com>
718 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
719 for IBM long double format correctly.
721 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
723 Fix PR tree-optimization/18792
725 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
726 to first_loop_depth, and use loop depth instead of loop number.
727 (build_classic_dir_vector): Ditto.
728 (compute_data_dependences_for_loop): Use depth, not loop number.
729 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
730 number. Pass in loops, instead of loop numbers.
731 (gather_interchange_stats): Ditto.
732 (linear_transform_loops): Ditto.
734 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
737 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
738 biv_fits_mode_p, extension_within_bounds_p): New functions.
739 (check_ext_dependent_givs): Use them.
741 2005-01-06 Roger Sayle <roger@eyesopen.com>
743 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
744 more than one barrier, in which case we should delete them all.
746 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
748 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
750 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
752 Fix PR middle-end/19286
754 * dwarf2out.c (gen_subprogram_die): If we've already tried to
755 output this subprogram, simply ignore this attempt to do it again.
756 (add_abstract_origin_attribute): Don't abort trying to add the abstract
757 origin attribute if it's not possible.
758 (gen_block_die): Don't ignore subblocks of "unused" blocks.
759 (decls_for_scope): Ditto.
760 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
761 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
763 2005-01-05 Richard Henderson <rth@redhat.com>
766 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
767 (bdesc_2arg): Use it.
768 (ix86_expand_binop_builtin): Force operands into registers
770 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
771 ix86_expand_sse_compare, ix86_expand_sse_comi,
772 ix86_expand_builtin): Likewise.
774 2005-01-05 Richard Henderson <rth@redhat.com>
776 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
778 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
779 (mulv4hi3): Set itanium_class mmmul.
780 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
781 (fpack): Rename from fpack_sfsf.
783 2005-01-05 Richard Henderson <rth@redhat.com>
786 * reload1.c (do_input_reload): Restrict the optimization deleteing
787 a previous output reload to RELOAD_FOR_INPUT.
789 2005-01-05 Steven Bosscher <stevenb@suse.de>
791 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
792 comment that we fall through after case.
793 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
794 (distribute_links): Likewise.
795 * cse.c (cse_insn): Likewise.
796 * cselib.c (cselib_invalidate_mem): Likewise.
797 * df.c: Update comments at the top of the file.
798 (read_modify_subreg_p): Update comments here too.
799 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
800 * flow.c (mark_set_1): Likewise.
801 (mark_used_regs): Likewise.
802 * gcse.c (mems_conflict_for_gcse_p): Likewise.
803 (canon_list_insert): Likewise.
804 (mark_set): Likewise.
805 (try_replace_reg): Likewise.
806 (store_killed_in_insn): Likewise.
807 * loop.c (count_one_set): Likewise.
808 (basic_induction_var): Likewise.
809 * postreload-gcse.c (find_mem_conflicts): Likewise.
810 * postreload.c (reload_combine_note_store): Likewise.
811 (move2add_note_store): Likewise.
812 * reload.c (find_equiv_reg): Likewise.
813 (mark_referenced_resources): Likewise.
814 * rtlanal.c (set_noop_p): Likewise.
815 (note_stores): Likewise.
816 (note_uses): Likewise.
817 * sched-deps.c (sched_analyze_1): Likewise.
818 * sched-rgn.c (check_live_1): Likewise.
819 (update_live_1): Likewise.
820 * config/i860/i860.c: Likewise.
822 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
824 (ZERO_EXTRACT): Mention that this one can be an lvalue.
826 * doc/rtl.texi: Update documentation for bit-fields and SET.
828 2005-01-05 Roger Sayle <roger@eyesopen.com>
830 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
832 2005-01-05 Jan Hubicka <jh@suse.cz>
833 Richard Henderson <rth@redhat.com>
836 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
839 2005-01-05 Richard Henderson <rth@redhat.com>
842 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
843 (sse_shufps): Change operand 3 to const_int_operand.
844 (sse2_storelps): Fix typo in template.
845 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
846 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
847 (ix86_expand_builtin): Likewise.
849 2005-01-05 Julian Brown <julian@codesourcery.com>
851 * config/arm/arm.c (arm_return_in_memory): Treat complex types
852 as aggregates for AAPCS ABIs.
854 2005-01-05 Stan Shebs <shebs@apple.com>
856 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
857 Mach-O getters if ppc64.
858 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
860 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
861 signed four-byte field for global code case.
862 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
864 2005-01-05 Roger Sayle <roger@eyesopen.com>
867 * c-common.c: Include real.h.
868 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
869 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
870 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
871 * Makefile.in (c-common.o): Update dependencies.
873 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
875 * c-parse.in (asm_string): Add trailing semicolon.
877 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
879 * c-parse.in (asm_string): New. Don't allow wide strings in
881 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
882 asm_string instead of STRING.
884 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
886 * c-typeck.c (constructor_no_implicit): Remove.
887 (set_designator, process_init_element): Don't check
888 constructor_no_implicit.
890 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
891 Kaz Kojima <kkojima@gcc.gnu.org>
894 * lcm.c (create_pre_exit): New.
895 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
896 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
898 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
900 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
901 -msoft-float in target_float_switch.
902 * arm.c (arm_override_options): Fix processing of target_float_switch.
904 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
906 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
909 2004-01-05 Uros Bizjak <uros@kss-loka.si>
911 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
912 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
914 2005-01-04 Roger Sayle <roger@eyesopen.com>
916 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
917 if deleting the then-block allows the test-block to fallthru to the
920 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
923 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
924 after the full declaration if the are in two different TUs.
926 2005-01-04 Richard Henderson <rth@redhat.com>
929 * tree-sra.c (generate_one_element_init): Just
930 call gimplify_and_add.
931 (generate_element_init): Record the
932 new referenced variables and mark them for renaming
934 (generate_element_init_1): This.
935 (scalarize_init): Don't call push_gimplify_context/
936 pop_gimplify_context.
938 2005-01-04 Geoffrey Keating <geoffk@apple.com>
940 * toplev.c (get_src_pwd): Handle failure of getpwd().
942 2005-01-04 Roger Sayle <roger@eyesopen.com>
944 * fold-const.c (fold_single_bit_test): Delete unreachable handling
946 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
947 all the cases handled by it are inverted by invert_truthvalue.
949 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
951 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
952 (save_register_windowsi): Likewise.
954 2005-01-04 Richard Henderson <rth@redhat.com>
956 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
958 2005-01-04 Uros Bizjak <uros@kss-loka.si>
961 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
962 any simplifications of vector mode comparison operators.
963 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
965 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
966 Devang Patel <dpatel@apple.com>
968 PR tree-optimization/18308
969 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
970 the operands before creating a new expression.
971 * dojump.c (do_jump): Make drop_through_label available
972 for all cases. Add expansion of COND_EXPR.
974 2005-01-04 Ira Rosen <irar@il.ibm.com>
976 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
977 INTEGER_CST instead of TREE_CONSTANT.
978 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
979 instead of TREE_CONSTANT.
980 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
981 instead of TREE_CONSTANT.
983 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
985 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
987 2005-01-04 Uros Bizjak <uros@kss-loka.si>
990 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
991 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
993 2005-01-03 Richard Henderson <rth@redhat.com>
995 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
996 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
997 fold_convert_const_real_from_real): Split out from ...
998 (fold_convert_const): ... here.
1000 2005-01-03 Richard Henderson <rth@redhat.com>
1003 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
1004 (mov<MMXMODEI>_internal): Likewise.
1005 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
1006 Add V2SF case; use it for SSE1; don't use TI.
1007 (movdf_integer): Likewise.
1008 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
1010 2005-01-03 Ira Rosen <irar@il.ibm.com>
1012 * tree-vectorizer.c (vect_strip_conversions): New function.
1013 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
1014 check for binary class.
1016 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
1020 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
1022 * gimple-low.c (mark_blocks_with_used_vars): New function.
1023 (mark_blocks_with_used_subblocks): Ditto.
1024 (mark_used_blocks): Ditto.
1025 (pass_mark_used_blocks): New pass.
1026 * tree-inline.c: Include debug.h.
1027 (expand_call_inline): Call outlining_inline_function here.
1028 * tree-optimize.c (init_tree_optimization_passes): Add
1029 pass_mark_used_blocks.
1030 * tree-pass.h (pass_mark_used_blocks): New.
1031 * Makefile.in (tree-inline.o): Add debug.h dependency.
1033 2005-01-03 Geoffrey Keating <geoffk@apple.com>
1035 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
1036 VAR_DECLs to have weak_import attribute.
1038 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
1041 2004-01-03 Steven Bosscher <stevenb@suse.de>
1043 * passes.c (rest_of_compilation): Don't run regmove if only
1044 flag_expensive_optimizations, require flag_regmove instead.
1046 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
1048 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
1050 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
1051 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1052 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
1053 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1055 2005-01-03 Richard Henderson <rth@redhat.com>
1057 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
1058 (ia64_const_ok_for_letter_p): New.
1059 (ia64_const_double_ok_for_letter_p): New.
1060 (ia64_extra_constraint): New.
1061 (ia64_expand_vecint_compare): New.
1062 (ia64_expand_vcondu_v2si): New.
1063 (ia64_expand_vecint_cmov): New.
1064 (ia64_expand_vecint_minmax): New.
1065 (ia64_print_operand): Add 'v'.
1066 (ia64_preferred_reload_class): New.
1067 (ia64_vector_mode_supported_p): New.
1068 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
1069 (PREFERRED_RELOAD_CLASS): Move to function.
1070 (CONST_OK_FOR_LETTER_P): Move to function.
1071 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
1072 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
1073 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
1074 (EXTRA_CONSTRAINT): Move to function.
1075 * config/ia64/ia64.md: Include vect.md.
1076 (itanium_class): Add mmalua.
1078 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
1080 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
1081 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
1083 (const_int_2bit_operand): New.
1084 (fr_reg_or_0_operand): New.
1085 * config/ia64/ia64-modes.def: Add vector modes.
1086 * config/ia64/ia64-protos.h: Update.
1087 * config/ia64/vect.md: New file.
1089 2005-01-03 Richard Henderson <rth@redhat.com>
1091 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
1093 2005-01-03 Uros Bizjak <uros@kss-loka.si>
1096 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
1098 (log1pdf2): Change mode of operands[0,1] to DFmode.
1100 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
1102 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
1103 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
1104 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
1105 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
1106 SPARC_RELAXED_ORDERING.
1108 2005-01-03 Richard Henderson <rth@redhat.com>
1109 Uros Bizjak <uros@kss-loka.si>
1112 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
1113 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
1114 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
1116 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
1117 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
1118 const_pow2_1_to_128_operand): New predicates.
1120 2005-01-02 Greg McGary <greg@mcgary.org>
1122 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
1124 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
1125 object. [ptr_type]: Remove unused variable.
1126 Remember nearest addressable array-element or record-component
1127 when checking bitfield components. Tolerate empty BIND_EXPR.
1129 2005-01-01 Richard Henderson <rth@redhat.com>
1131 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
1132 that don't narrow the value. Fail for other conversions.
1134 2005-01-01 Richard Henderson <rth@redhat.com>
1137 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
1138 * c-lang.c (finish_file): Don't do it here.
1139 * objc/objc-act.c (objc_finish_file): Likewise.
1141 * cgraph.c (decl_assembler_name_equal): New.
1142 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
1143 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
1144 * cgraph.h (struct cgraph_varpool_node): Add next.
1145 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
1146 * varasm.c (assemble_alias): Mark the target as needed.
1148 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
1151 * function.c (get_arg_pointer_save_area): Use entry_of_function
1152 instead of get_insns.
1154 2005-01-01 Roger Sayle <roger@eyesopen.com>
1155 Andrew Pinski <pinskia@physics.uc.edu>
1156 James E. Wilson <wilson@specifixinc.com>
1158 PR rtl-optimization/12092
1159 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
1162 2005-01-01 Roger Sayle <roger@eyesopen.com>
1163 Olivier Hainque <hainque@act-europe.fr>
1165 * tree.c (int_fits_type_p): A narrower type always fits in a
1166 wider one, except for negative values into unsigned types.
1168 2005-01-01 Roger Sayle <roger@eyesopen.com>
1170 * tree.c (int_fits_type_p): Compare the result of force_fit_type
1171 with the original constant rather than require TREE_OVERFLOW.
1173 2005-01-01 Steven Bosscher <stevenb@suse.de>
1176 * c-decl.c (finish_function): If compiling C99, annotate the
1177 compiler generated return with the current file name and line 0.
1178 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
1179 the source line is greater than 0.
1180 (remove_bb): Likewise.
1182 See ChangeLog.12 for earlier changes.