1 2005-01-15 Ralf Corsepius <ralf.corsepius@rtems.org>
3 * config/mips/rtems.h (MIPS_DEFAULT_GVALUE): Set to 0.
4 * config/mips/t-rtems (MULTILIBS_DIRNAMES,MULTILIB_OPTIONS):
5 Remove little endian multilib variants.
6 Add mips32 multilib variant.
8 2005-01-15 Kazu Hirata <kazu@cs.umass.edu>
10 * explow.c (copy_all_reg, stabilize): Remove.
11 * expr.h: Remove the corresponding prototypes.
13 * rtlanal.c (regs_set_between_p): Remove.
14 * rtl.h: Remove the corresponding prototype.
16 2005-01-14 David Edelsohn <edelsohn@gnu.org>
18 * config/rs6000/rs6000.c (rs6k_nonimmediate_operand): Rename
19 to rs6000_nonimmediate_operand.
20 * config/rs6000/rs6000.h (PREDICATE_CODES): Rename
21 rs6k_nonimmediate_operand to rs6000_nonimmediate_operand.
22 * config/rs6000/rs6000.md (movsi_internal1): Rename predicate to
23 rs6000_nonimmediate_operand.
24 * config/rs6000/spe.md (movdf_e500_double): Rename predicate to
25 rs6000_nonimmediate_operand.
27 2005-01-14 Andrew Pinski <pinskia@physics.uc.edu>
29 * tree-ssa-alias.c (pass_may_alias): Add TODO_verify_stmts.
31 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
33 * config/rs6000/rs6000.h (CLASS_MAX_NREGS): DF goes in 1 register
35 (CANNOT_CHANGE_MODE_CLASS): Restrict DI mode changes on e500v2.
36 (PREDICATE_CODES): Add rs6k_nonimmediate_operand.
38 * config/rs6000/rs6000.c (invalid_e500_subreg): New.
39 (rs6k_nonimmediate_operand): New.
40 (rs6000_legitimate_offset_address_p): Handle DI modes on e500v2
42 (legitimate_lo_sum_address_p): Same.
43 (rs6000_legitimize_address): Same.
44 (rs6000_legitimize_reload_address): Same.
45 (rs6000_legitimate_address): Same.
46 (spe_build_register_parallel): Pass DF and DC modes in a DI
49 * config/rs6000/rs6000.md ("*movsi_internal1"): Change predicate
50 to rs6k_nonimmediate_operand.
52 * config/rs6000/spe.md ("*frob_df_di"): New.
54 ("*frob_di_df_2"): New.
55 ("*mov_sidf_e500_subreg0"): New.
56 ("*mov_sidf_e500_subreg4"): New.
57 ("*movdf_e500_double"): Change predicate to
58 rs6k_nonimmediate_operand.
60 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
62 * postreload.c (move2add_note_store): Only call
63 trunc_int_for_mode on scalar integers.
65 2005-01-14 Steven Bosscher <stevenb@suse.de>
67 * tree-ssa-dce.c (visited_control_parents): New sbitmap to
68 replace BB_VISITED uses.
69 (find_obviously_necessary_stmts): Don't clear BB_VISITED.
70 (propagate_necessity): Check the bitmap instead of BB_VISITED.
71 (tree_dce_done): Free visited_control_parents.
72 (perform_tree_ssa_dce): Allocate and clear it.
73 * tree-ssa-pre.c (compute_antic_aux): Make non-recursive.
74 (compute_antic): Iterate from here using a DFS. Use an sbitmap
75 instead of BB_VISITED.
77 2005-01-14 Kazu Hirata <kazu@cs.umass.edu>
79 * c-tree.h, coverage.h, langhooks-def.h, optabs.h, output.h,
80 rtl.h, targhooks.h, tree-flow.h, tree-mudflap.h, tree.h:
81 Remove unused prototypes.
83 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
85 * config/rs6000/rs6000.c (rs6000_stack_info): Fix caching of SPE
86 64-bit register usage.
87 (spe_func_has_64bit_regs_p): Add FIXME note.
89 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
91 * config/rs6000/rs6000.c (rs6000_generate_compare): Replace
92 flag_finite_math_only with flag_unsafe_math_optimizations.
94 2005-01-14 J. D. Johnston <jjohnst@us.ibm.com>
96 * config/s390/tpf-unwind.h (s390_fallback_frame_state): Correct
98 (__tpf_eh_return): Copy TPF private stack area from the module
99 boundary stack frame. Add check for when module addresses are
100 equal, but no stub address is found.
102 2005-01-14 Richard Earnshaw <rearnsha@arm.com>
105 * arm.h (struct machine_function): Add call_via field.
106 (thumb_call_via_label): Declare.
107 * arm.c (thumb_call_via_label): New variable.
108 (thumb_call_reg_needed): New variable.
109 (arm_output_function_epilogue): For Thumb code, output any per-function
110 call-indirect trampolines.
111 (thumb_call_via_reg): New function.
112 (arm_file_end): New function.
113 (TARGET_ASM_FILE_END): Call arm_file_end.
114 (aof_file_end): Likewise.
115 * arm-protos.h (thumb_call_via_reg): Declare.
116 * arm.md (call_reg_thumb, call_value_reg_thumb): Call
117 thumb_call_via_reg in normal case.
119 2005-01-14 Jakub Jelinek <jakub@redhat.com>
122 PR rtl-optimization/19348
123 * recog.c (peephole2_optimize): Do global life update if some peephole
124 decides it doesn't need at least one of its inputs and that change
125 influences liveness at the start of the basic block.
127 * basic-block.h (EXECUTE_IF_AND_COMPL_IN_REG_SET): Needs 2 REGSET
128 arguments instead of 1.
130 2005-01-14 Eric Botcazou <ebotcazou@libertysurf.fr>
133 * varasm.c (initializer_constant_valid_p) <ADDR_EXPR>: Return
134 zero for nested functions needing a static chain or functions
135 with a non-constant address.
137 2005-01-13 Roger Sayle <roger@eyesopen.com>
139 * simplify-rtx.c (simplify_binary_operation) <AND>: Optimize
140 (and (sign_extend X) C) into (zero_extend (and X C)).
142 2005-01-13 David O'Brien <obrien@FreeBSD.org>
144 * config/freebsd-spec.h: Make KSE pthread lib logic the default.
146 2005-01-13 Richard Henderson <rth@redhat.com>
151 * config/i386/i386.md (cmpdf, cmpsf, bunordered, bordered, buneq,
152 bunge, bungt, bunle, bunlt, bltgt): Enable for TARGET_SSE_MATH,
154 (cmpfp_i_387): Rename from cmpfp_i. Move after sse patterns.
155 (cmpfp_i_mixed): Rename from cmpfp_i_sse; use for TARGET_MIX_SSE_I387.
156 (cmpfp_i_sse): Rename from cmpfp_i_sse_only; use for TARGET_SSE_MATH.
157 (cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387): Similarly.
158 (fp_jcc_1_mixed, fp_jcc_1_sse, fp_jcc_1_387): Similarly.
159 (fp_jcc_2_mixed, fp_jcc_2_sse, fp_jcc_2_387): Similarly.
160 (fp_jcc_3_387, fp_jcc_4_387, fp_jcc_5_387, fp_jcc_6_387,
161 fp_jcc_7_387, fp_jcc_8_387): Rename from fp_jcc_N.
162 (movdicc_c_rex64): Rename with '*'.
163 (movsfcc, movdfcc): Add checks for 387 and sse math to condition.
164 (movsfcc_1_sse_min, movsfcc_1_sse_max, movsfcc_1_sse): New.
165 (movsfcc_1_387): Rename from movsfcc_1.
166 (movdfcc_1_sse_min, movdfcc_1_sse_max, movdfcc_1_sse): New.
167 (movdfcc_1, movdfcc_1_rex64): Add check for 387.
168 (sminsf3, smaxsf3, smindf3, smaxdf3): New.
169 (minsf3, minsf, minsf_nonieee, minsf_sse, mindf3, mindf,
170 mindf_nonieee, mindf_sse, maxsf3, maxsf, maxsf_nonieee, maxsf_sse,
171 maxdf3, maxdf, maxdf_nonieee, maxdf_sse, sse_movsfcc, sse_movsfcc_eq,
172 sse_movdfcc, sse_movdfcc_eq, sse_movsfcc_const0_1,
173 sse_movsfcc_const0_2, sse_movsfcc_const0_3, sse_movsfcc_const0_4,
174 sse_movdfcc_const0_1, sse_movdfcc_const0_2, sse_movdfcc_const0_3,
175 sse_movdfcc_const0_4): Remove.
176 * config/i386/i386.c (ix86_expand_fp_movcc): For TARGET_SSE_MATH,
177 recognize min/max early. Update for changed sse cmove patterns.
178 (ix86_split_sse_movcc): New.
179 * config/i386/i386-protos.h: Update.
181 2005-01-13 Steven Bosscher <stevenb@suse.de>
183 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
184 (fix_stmt_v_may_defs): Likewise.
186 2005-01-13 Richard Henderson <rth@redhat.com>
188 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
189 for converting the mask.
191 2005-01-13 David Edelsohn <edelsohn@gnu.org>
193 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
196 2005-01-13 Richard Henderson <rth@redhat.com>
198 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
199 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
200 IX86_BUILTIN_MOVDQ2Q): Remove.
201 (IX86_BUILTIN_VEC_EXT_V4SI): New.
202 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
203 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
205 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
206 (sse2_pextrw): Fix immediate constraint.
207 (sse2_loadq, sse2_loadq_rex64): Remove.
208 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
209 Use __builtin_ia32_vec_ext_<size>.
210 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
212 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
214 * function.c (assign_parm_setup_block): Look inside original
215 entry_parm when inspecting PARALLEL.
217 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
218 Joel Sherrill <joel@oarcorp.com>
221 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
222 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
223 rtems_gxx_recursive_mutex_init.
224 (__gthread_recursive_mutex_lock): New function.
225 (__gthread_recursive_mutex_trylock): Likewise.
226 (__gthread_recursive_mutex_unlock): Likewise.
228 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
230 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
232 2005-01-13 Jan Beulich <jbeulich@novell.com>
234 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
235 (ashrti3_internal): Indicate output is early clobber. Generate result
236 into output rather than first input. Use move for low word of output
237 if shift count is exactly 64.
238 (lshrti3_internal): Likewise.
240 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
244 * reload1.c (delete_output_reload): Don't delete an output reload
245 if the pseudo lives longer than a single basic block. Adjust and
246 improve wording of comment.
248 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
251 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
252 of TYPE_SIZE where we mean the number of bytes.
254 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
256 * function.c (assign_parm_setup_block): Relax condition on
257 multi-register optimization.
259 2005-01-12 Nick Clifton <nickc@redhat.com>
261 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
262 an "in_delay_slot" attribute of "no" to prevent them being used in
263 delay slots. This is forbidden because they might generate
266 2005-01-12 Alan Modra <amodra@bigpond.net.au>
269 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
272 2005-01-12 Nick Clifton <nickc@redhat.com>
274 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
275 passed on to GAS) but do not define, as GAS no longer supports or
276 needs the -m2000 option.
278 2005-01-11 Roger Sayle <roger@eyesopen.com>
280 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
281 type as args of unprototyped builtins aren't automatically promoted.
282 (fold_builtin_abs): Likewise.
284 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
286 * Makefile.in: Set a `build-warn' variable.
288 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
290 PR tree-optimization/17949
291 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
292 (find_interesting_uses_address): Use it.
294 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
296 * regrename.c (kill_value): Handle subreg's that won't simplify.
298 2005-01-11 Richard Henderson <rth@redhat.com>
301 * config/i386/i386.h (enum ix86_builtins): Move ...
302 * config/i386/i386.c: ... here.
303 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
304 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
305 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
306 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
307 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
308 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
309 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
310 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
311 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
312 IX86_BUILTIN_LOADDDUP): Remove.
313 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
314 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
315 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
316 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
317 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
318 (ix86_init_builtins): Make static.
319 (ix86_init_mmx_sse_builtins): Update for changed builtins.
320 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
321 if all the modes match. Otherwise, fake it.
322 (get_element_number, ix86_expand_vec_init_builtin,
323 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
324 (ix86_expand_builtin): Make static. Update for changed builtins.
325 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
326 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
327 (ix86_expand_vector_init_duplicate): New.
328 (ix86_expand_vector_init_low_nonzero): New.
329 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
330 Split out from ix86_expand_vector_init; handle integer modes.
331 (ix86_expand_vector_init): Use them.
332 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
333 * config/i386/i386-protos.h: Update.
334 * config/i386/predicates.md (reg_or_0_operand): New.
335 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
336 (movv2sf_internal): Likewise. And a splitter to match them all.
337 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
338 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
339 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
340 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
341 (mmx_pinsrw): Fix operand ordering.
342 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
343 rather than sse_loadss expander.
344 (movv2df splitter): Similarly.
345 (sse_loadss, sse_loadlss): Remove.
346 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
347 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
348 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
349 (sse3_movddup): Rename with '*'.
350 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
351 (sse2_loadsd): Remove.
352 (vec_dupv2df_sse3): Rename from sse3_loadddup.
353 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
354 (sse2_pinsrw): Fix argument ordering.
355 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
356 (sse2_stored): Remove 'r' destination.
357 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
358 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
359 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
360 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
361 vec_extractv16qi, vec_initv16qi): New.
363 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
364 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
365 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
367 (_mm_load_pd, _mm_store_pd): Use plain dereference.
368 (_mm_load_si128, _mm_store_si128): Likewise.
369 (_mm_load1_pd): Use _mm_set1_pd.
370 (_mm_load_sd): Use _mm_set_sd.
371 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
372 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
373 (_mm_set_epi64): Use _mm_set_epi64x.
374 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
375 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
376 _mm_setr_epi8): Use _mm_set_foo form.
377 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
378 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
379 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
380 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
381 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
382 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
383 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
384 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
385 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
386 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
387 (_mm_movedup_pd): Use _mm_shuffle_pd.
388 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
389 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
390 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
391 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
393 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
394 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
395 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
396 (_mm_store_ps): Use raw dereference.
397 (_mm_store1_ps): Use _mm_storeu_ps.
398 (_mm_storer_ps): Use _mm_store_ps.
399 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
400 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
402 2005-01-11 Stan Shebs <shebs@apple.com>
404 * config/rs6000/rs6000.c (machopic_output_stub): Issue
405 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
407 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
410 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
411 Skip all DECLs except for FIELD_DECLs.
413 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
415 * config/s390/s390.c (override_options): Return error if
416 -mbackchain, -mpacked-stack and -mhard-float are used together.
417 (s390_va_start): Remove the backchain && packed-stack special case.
418 (s390_gimplify_va_arg): Likewise.
419 * doc/invoke.texi: Remove the ABI incompatibility note.
421 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
423 * config/s390/s390.c (struct s390_frame_layout): Remove
425 (s390_frame_info, s390_emit_prologue): Replace occurrences of
426 save_backchain_p with TARGET_BACKCHAIN.
428 2005-01-11 Alan Modra <amodra@bigpond.net.au>
431 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
432 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
433 * function.c (locate_and_pad_parm): Set new field.
434 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
435 Tweak where_pad test to include "none". Always set mem align for
437 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
438 not sufficient for type.
439 (assign_parm_setup_block): If stack_parm is zero on entry, always
440 make a new stack local. Block move old stack parm if necessary
441 to new aligned stack local.
442 (assign_parm_setup_stack): Use a block move to handle
443 potentially misaligned entry_parm.
444 (assign_parms_unsplit_complex): Specify required alignment when
445 creating stack local.
446 * calls.c (compute_argument_addresses): Override alignment of stack
447 arg calculated from its type with the alignment given by
448 FUNCTION_ARG_BOUNDARY.
449 (store_one_arg): Likewise.
451 2005-01-11 Jan Beulich <jbeulich@novell.com>
453 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
454 Change respective itanium_class attribute to ialu.
455 (shladdp4_internal): New.
456 * config/ia64/predicates.md (shladd_log2_operand): New.
458 2005-01-11 Richard Henderson <rth@redhat.com>
460 * expr.c (store_constructor): Use rtvec_alloc instead of
461 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
463 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
465 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
467 2005-01-11 Alan Modra <amodra@bigpond.net.au>
469 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
470 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
471 (TARGET_RWRELOC): Define.
472 (ia64_rwreloc_section_type_flags): Delete.
473 (ia64_section_type_flags): New function.
474 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
475 (TARGET_RWRELOC): Define.
477 2005-01-10 David Mosberger <davidm@hpl.hp.com>
480 * config/ia64/ia64.c (process_set): For alloc insn, only call
481 process_epilogue is !frame_pointer_needed.
483 2005-01-10 Roger Sayle <roger@eyesopen.com>
486 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
487 unary operator and can't be treated as a binary/comparison operator.
489 2005-01-10 Richard Henderson <rth@redhat.com>
491 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
492 (ix86_return_in_memory): Likewise.
493 (function_arg_advance): Likewise. Mirror structure in function_arg
494 for choosing register to advance.
496 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
498 * tree-vectorizer.c, tree.def: Fix comment typos.
500 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
502 * config/s390/s390.c (struct processor_costs): 4 new fields:
503 ddbr, ddr, debr, der.
504 (s390_rtx_costs): More precise handling of divide instructions.
506 2005-01-10 David Edelsohn <edelsohn@gnu.org>
508 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
510 2005-01-10 Jan Beulich <jbeulich@novell.com>
512 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
513 names resulting from -ffunction-sections/-fdata-sections and linkonce
515 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
516 based on the section name. Rearrange the section name comparison logic
517 slightly so that each section name is compared against at most once.
519 2005-01-10 Ben Elliston <bje@au.ibm.com>
521 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
522 targets for which -fPIC reduces limitations on the GOT size.
524 2005-01-09 Falk Hueffner <falk@debian.org>
526 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
529 2005-01-09 David Edelsohn <edelsohn@gnu.org>
532 * collect2.c (main): Set aixrtl_flag for -brtl option.
533 (resolve_lib_name): Search for .so file extension before .a
536 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
538 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
539 (vect_analyze_data_refs_alignment): Add dump prints.
541 2005-01-09 Ira Rosen <irar@il.ibm.com>
543 * tree-vectorizer.c (vect_analyze_offset_expr): Use
544 expr_invariant_in_loop_p.
545 Initialize outputs first thing in the function.
546 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
547 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
548 (vect_analyze_pointer_ref_access): Check that the initial condition of
549 the access function is loop invariant.
551 2005-01-09 Richard Henderson <rth@redhat.com>
553 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
554 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
555 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
556 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
557 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
558 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
559 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
560 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
561 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
562 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
563 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
564 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
565 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
566 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
567 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
568 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
569 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
570 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
571 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
572 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
573 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
574 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
575 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
577 (mmx_clrdi, pavgusb): Remove.
578 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
579 with leading sse_ prefix.
580 * config/i386/sse.md: Receive them.
581 * config/i386/mmx.md: New file.
582 (MMXMODE12, MMXMODE24, mmxvecsize): New.
583 (subrv2sf3): Turn into expander for normal subtraction.
584 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
585 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
586 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
587 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
588 commutative; use ix86_binary_operator_ok.
589 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
590 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
591 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
592 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
593 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
594 Macroize from existing patterns; use ix86_binary_operator_ok.
595 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
596 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
597 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
598 with vec_select+vec_concat.
599 (mmx_pshufw, mmx_pshufw_1): Likewise.
600 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
601 (mmx_uavgv4hi3): Model correctly.
602 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
603 _mm_xor_si64): Remove casts.
605 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
607 PR tree-optimization/19224
608 * tree-scalar-evolution.c (get_instantiated_value,
609 set_instantiated_value): New functions.
610 (instantiate_parameters_1): Cache the results.
611 (instantiate_parameters, resolve_mixers): Initialize and free
614 2005-01-08 David Edelsohn <edelsohn@gnu.org>
616 * config/i386/i386.md (addhi_4): Correct reference in comment.
619 2005-01-08 Richard Henderson <rth@redhat.com>
621 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
622 (_mm_cvtsi128_si64x): Likewise.
623 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
624 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
625 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
626 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
627 * config/i386/i386-protos.h: Update.
628 * config/i386/i386.c (print_operand): Add 'H'.
629 (ix86_fixup_binary_operands): Split out from ...
630 (ix86_expand_binary_operator): ... here.
631 (ix86_fixup_binary_operands_no_copy): New.
632 (ix86_expand_fp_absneg_operator): Handle vector mode results.
633 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
634 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
635 (safe_vector_operand): Use CONST0_RTX.
636 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
637 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
638 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
639 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
641 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
642 (IX86_BUILTIN_CMPNGESS): New.
643 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
644 (attr type): Add sselog1.
645 (attr unit, attr memory): Handle it.
646 (movti, movti_internal, movti_rex64): Move near other integer moves.
647 (movtf, movtf_internal): Move near other fp moves.
648 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
649 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
650 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
651 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
652 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
653 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
654 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
655 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
656 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
657 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
658 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
659 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
660 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
661 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
662 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
663 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
664 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
665 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
666 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
667 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
668 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
669 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
670 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
671 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
672 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
673 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
674 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
675 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
676 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
677 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
678 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
679 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
680 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
681 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
682 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
683 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
684 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
685 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
686 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
687 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
688 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
689 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
690 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
691 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
693 (SSEPUSH, push<SSEPUSH>): Remove.
694 (MMXPUSH, push<MMXPUSH>): Remove.
695 (sse_movaps, sse_movaps_1, sse_movups): Remove.
696 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
697 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
698 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
699 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
700 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
701 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
702 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
703 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
704 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
705 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
706 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
707 (ppro_sse_log_V4SF): Handle sselog1.
708 * config/i386/predicates.md (const_0_to_1_operand): New.
709 (const_0_to_255_mul_8_operand): New.
710 (const_1_to_31_operand): Rename from const_int_1_31_operand.
711 (const_2_to_3_operand, const_4_to_7_operand): New.
712 * config/i386/sse.md: New file.
713 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
714 (sse_movups): Rename from sse_movups_1.
715 (sse_loadlss): Rename from sse_loadss_1.
716 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
718 (negv4sf2): Use ix86_expand_fp_absneg_operator.
719 (absv4sf2, negv2df, absv2df): New.
720 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
721 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
722 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
723 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
724 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
725 (sse3_addsubv4sf3): Model correctly.
726 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
727 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
728 (sse_movhlps): Model with vec_select+vec_concat.
729 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
730 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
731 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
732 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
733 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
734 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
735 sse2_pshufhw_1): Likewise.
736 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
737 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
738 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
739 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
740 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
741 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
742 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
743 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
744 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
745 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
746 sminv8hi3): Mark commutative
747 operands. Use ix86_binary_operator_ok.
748 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
749 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
750 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
751 sse2_punpcklqdq): Allow operand2 in memory.
752 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
753 sse2_movsd): Add memory alternatives.
754 (sse_storelps): Turn expander into an insn; split after reload.
755 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
756 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
758 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
760 * configure.ac (DWARF-2 debug_line): Use objdump.
761 * configure: Regenerate.
763 2005-01-08 Jeff Law <law@redhat.com>
764 Diego Novillo <dnovillo@redhat.com>
766 PR tree-optimization/18241
767 * tree-nrv.c (tree_nrv): Ignore volatile return values.
768 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
769 statements with volatile operands.
770 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
771 operands after marking a statement with has_volatile_ops.
773 2005-01-08 Roger Sayle <roger@eyesopen.com>
775 * tree.c (int_fits_type_p): Always honor integer constant
776 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
778 2005-01-08 Roger Sayle <roger@eyesopen.com>
780 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
781 the latent bug in rtl_delete_block has been resolved.
783 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
785 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
786 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
788 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
790 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
791 library as @multilib_dir@/$(SHLIB_SONAME).
793 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
795 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
796 * configure: Regenerate.
798 2005-01-07 Jakub Jelinek <jakub@redhat.com>
800 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
803 2005-01-07 David Edelsohn <edelsohn@gnu.org>
806 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
807 Convert non-word aligned offset address using ld/std into
810 2005-01-07 Richard Henderson <rth@redhat.com>
812 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
814 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
816 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
818 2005-01-07 Jakub Jelinek <jakub@redhat.com>
820 PR tree-optimization/19283
821 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
822 is not shorter than the original type.
824 PR rtl-optimization/19012
825 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
827 PR rtl-optimization/18861
828 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
829 (FULL_STATE): Define.
830 (mark_tablejump_edge): New function.
831 (purge_dead_tablejump_edges): New function.
832 (find_bb_boundaries): Use it.
834 PR tree-optimization/18828
835 * builtins.c (expand_builtin_next_arg): Remove argument and all
836 the argument checking.
837 (expand_builtin): Adjust caller.
838 (expand_builtin_va_start): Likewise. Remove error for too many
840 (fold_builtin_next_arg): Issue error for too many arguments.
841 After checking arguments, replace them with magic arguments that
842 prevent further checking of the args.
844 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
846 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
848 2005-01-06 Roger Sayle <roger@eyesopen.com>
851 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
852 and %{.S:...} (and their negative variants) to test whether the
853 input file is assembler or pre-processed-assembler independent of
854 the actual filename extension.
856 2005-01-06 Roger Sayle <roger@eyesopen.com>
858 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
859 of sign or zero extended values.
861 2005-01-06 Geoffrey Keating <geoffk@apple.com>
863 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
864 for IBM long double format correctly.
866 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
868 Fix PR tree-optimization/18792
870 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
871 to first_loop_depth, and use loop depth instead of loop number.
872 (build_classic_dir_vector): Ditto.
873 (compute_data_dependences_for_loop): Use depth, not loop number.
874 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
875 number. Pass in loops, instead of loop numbers.
876 (gather_interchange_stats): Ditto.
877 (linear_transform_loops): Ditto.
879 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
882 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
883 biv_fits_mode_p, extension_within_bounds_p): New functions.
884 (check_ext_dependent_givs): Use them.
886 2005-01-06 Roger Sayle <roger@eyesopen.com>
888 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
889 more than one barrier, in which case we should delete them all.
891 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
893 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
895 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
897 Fix PR middle-end/19286
899 * dwarf2out.c (gen_subprogram_die): If we've already tried to
900 output this subprogram, simply ignore this attempt to do it again.
901 (add_abstract_origin_attribute): Don't abort trying to add the abstract
902 origin attribute if it's not possible.
903 (gen_block_die): Don't ignore subblocks of "unused" blocks.
904 (decls_for_scope): Ditto.
905 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
906 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
908 2005-01-05 Richard Henderson <rth@redhat.com>
911 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
912 (bdesc_2arg): Use it.
913 (ix86_expand_binop_builtin): Force operands into registers
915 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
916 ix86_expand_sse_compare, ix86_expand_sse_comi,
917 ix86_expand_builtin): Likewise.
919 2005-01-05 Richard Henderson <rth@redhat.com>
921 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
923 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
924 (mulv4hi3): Set itanium_class mmmul.
925 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
926 (fpack): Rename from fpack_sfsf.
928 2005-01-05 Richard Henderson <rth@redhat.com>
931 * reload1.c (do_input_reload): Restrict the optimization deleteing
932 a previous output reload to RELOAD_FOR_INPUT.
934 2005-01-05 Steven Bosscher <stevenb@suse.de>
936 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
937 comment that we fall through after case.
938 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
939 (distribute_links): Likewise.
940 * cse.c (cse_insn): Likewise.
941 * cselib.c (cselib_invalidate_mem): Likewise.
942 * df.c: Update comments at the top of the file.
943 (read_modify_subreg_p): Update comments here too.
944 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
945 * flow.c (mark_set_1): Likewise.
946 (mark_used_regs): Likewise.
947 * gcse.c (mems_conflict_for_gcse_p): Likewise.
948 (canon_list_insert): Likewise.
949 (mark_set): Likewise.
950 (try_replace_reg): Likewise.
951 (store_killed_in_insn): Likewise.
952 * loop.c (count_one_set): Likewise.
953 (basic_induction_var): Likewise.
954 * postreload-gcse.c (find_mem_conflicts): Likewise.
955 * postreload.c (reload_combine_note_store): Likewise.
956 (move2add_note_store): Likewise.
957 * reload.c (find_equiv_reg): Likewise.
958 (mark_referenced_resources): Likewise.
959 * rtlanal.c (set_noop_p): Likewise.
960 (note_stores): Likewise.
961 (note_uses): Likewise.
962 * sched-deps.c (sched_analyze_1): Likewise.
963 * sched-rgn.c (check_live_1): Likewise.
964 (update_live_1): Likewise.
965 * config/i860/i860.c: Likewise.
967 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
969 (ZERO_EXTRACT): Mention that this one can be an lvalue.
971 * doc/rtl.texi: Update documentation for bit-fields and SET.
973 2005-01-05 Roger Sayle <roger@eyesopen.com>
975 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
977 2005-01-05 Jan Hubicka <jh@suse.cz>
978 Richard Henderson <rth@redhat.com>
981 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
984 2005-01-05 Richard Henderson <rth@redhat.com>
987 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
988 (sse_shufps): Change operand 3 to const_int_operand.
989 (sse2_storelps): Fix typo in template.
990 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
991 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
992 (ix86_expand_builtin): Likewise.
994 2005-01-05 Julian Brown <julian@codesourcery.com>
996 * config/arm/arm.c (arm_return_in_memory): Treat complex types
997 as aggregates for AAPCS ABIs.
999 2005-01-05 Stan Shebs <shebs@apple.com>
1001 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
1002 Mach-O getters if ppc64.
1003 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
1005 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
1006 signed four-byte field for global code case.
1007 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
1009 2005-01-05 Roger Sayle <roger@eyesopen.com>
1012 * c-common.c: Include real.h.
1013 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
1014 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
1015 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
1016 * Makefile.in (c-common.o): Update dependencies.
1018 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1020 * c-parse.in (asm_string): Add trailing semicolon.
1022 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1024 * c-parse.in (asm_string): New. Don't allow wide strings in
1026 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
1027 asm_string instead of STRING.
1029 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1031 * c-typeck.c (constructor_no_implicit): Remove.
1032 (set_designator, process_init_element): Don't check
1033 constructor_no_implicit.
1035 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
1036 Kaz Kojima <kkojima@gcc.gnu.org>
1039 * lcm.c (create_pre_exit): New.
1040 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
1041 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
1043 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
1045 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
1046 -msoft-float in target_float_switch.
1047 * arm.c (arm_override_options): Fix processing of target_float_switch.
1049 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
1051 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
1054 2004-01-05 Uros Bizjak <uros@kss-loka.si>
1056 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
1057 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
1059 2005-01-04 Roger Sayle <roger@eyesopen.com>
1061 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
1062 if deleting the then-block allows the test-block to fallthru to the
1065 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
1068 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
1069 after the full declaration if the are in two different TUs.
1071 2005-01-04 Richard Henderson <rth@redhat.com>
1074 * tree-sra.c (generate_one_element_init): Just
1075 call gimplify_and_add.
1076 (generate_element_init): Record the
1077 new referenced variables and mark them for renaming
1078 and split out to ...
1079 (generate_element_init_1): This.
1080 (scalarize_init): Don't call push_gimplify_context/
1081 pop_gimplify_context.
1083 2005-01-04 Geoffrey Keating <geoffk@apple.com>
1085 * toplev.c (get_src_pwd): Handle failure of getpwd().
1087 2005-01-04 Roger Sayle <roger@eyesopen.com>
1089 * fold-const.c (fold_single_bit_test): Delete unreachable handling
1091 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
1092 all the cases handled by it are inverted by invert_truthvalue.
1094 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
1096 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
1097 (save_register_windowsi): Likewise.
1099 2005-01-04 Richard Henderson <rth@redhat.com>
1101 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
1103 2005-01-04 Uros Bizjak <uros@kss-loka.si>
1106 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
1107 any simplifications of vector mode comparison operators.
1108 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
1110 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
1111 Devang Patel <dpatel@apple.com>
1113 PR tree-optimization/18308
1114 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
1115 the operands before creating a new expression.
1116 * dojump.c (do_jump): Make drop_through_label available
1117 for all cases. Add expansion of COND_EXPR.
1119 2005-01-04 Ira Rosen <irar@il.ibm.com>
1121 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
1122 INTEGER_CST instead of TREE_CONSTANT.
1123 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
1124 instead of TREE_CONSTANT.
1125 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
1126 instead of TREE_CONSTANT.
1128 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
1130 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
1132 2005-01-04 Uros Bizjak <uros@kss-loka.si>
1135 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
1136 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
1138 2005-01-03 Richard Henderson <rth@redhat.com>
1140 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
1141 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
1142 fold_convert_const_real_from_real): Split out from ...
1143 (fold_convert_const): ... here.
1145 2005-01-03 Richard Henderson <rth@redhat.com>
1148 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
1149 (mov<MMXMODEI>_internal): Likewise.
1150 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
1151 Add V2SF case; use it for SSE1; don't use TI.
1152 (movdf_integer): Likewise.
1153 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
1155 2005-01-03 Ira Rosen <irar@il.ibm.com>
1157 * tree-vectorizer.c (vect_strip_conversions): New function.
1158 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
1159 check for binary class.
1161 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
1165 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
1167 * gimple-low.c (mark_blocks_with_used_vars): New function.
1168 (mark_blocks_with_used_subblocks): Ditto.
1169 (mark_used_blocks): Ditto.
1170 (pass_mark_used_blocks): New pass.
1171 * tree-inline.c: Include debug.h.
1172 (expand_call_inline): Call outlining_inline_function here.
1173 * tree-optimize.c (init_tree_optimization_passes): Add
1174 pass_mark_used_blocks.
1175 * tree-pass.h (pass_mark_used_blocks): New.
1176 * Makefile.in (tree-inline.o): Add debug.h dependency.
1178 2005-01-03 Geoffrey Keating <geoffk@apple.com>
1180 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
1181 VAR_DECLs to have weak_import attribute.
1183 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
1186 2004-01-03 Steven Bosscher <stevenb@suse.de>
1188 * passes.c (rest_of_compilation): Don't run regmove if only
1189 flag_expensive_optimizations, require flag_regmove instead.
1191 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
1193 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
1195 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
1196 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1197 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
1198 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1200 2005-01-03 Richard Henderson <rth@redhat.com>
1202 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
1203 (ia64_const_ok_for_letter_p): New.
1204 (ia64_const_double_ok_for_letter_p): New.
1205 (ia64_extra_constraint): New.
1206 (ia64_expand_vecint_compare): New.
1207 (ia64_expand_vcondu_v2si): New.
1208 (ia64_expand_vecint_cmov): New.
1209 (ia64_expand_vecint_minmax): New.
1210 (ia64_print_operand): Add 'v'.
1211 (ia64_preferred_reload_class): New.
1212 (ia64_vector_mode_supported_p): New.
1213 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
1214 (PREFERRED_RELOAD_CLASS): Move to function.
1215 (CONST_OK_FOR_LETTER_P): Move to function.
1216 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
1217 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
1218 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
1219 (EXTRA_CONSTRAINT): Move to function.
1220 * config/ia64/ia64.md: Include vect.md.
1221 (itanium_class): Add mmalua.
1223 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
1225 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
1226 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
1228 (const_int_2bit_operand): New.
1229 (fr_reg_or_0_operand): New.
1230 * config/ia64/ia64-modes.def: Add vector modes.
1231 * config/ia64/ia64-protos.h: Update.
1232 * config/ia64/vect.md: New file.
1234 2005-01-03 Richard Henderson <rth@redhat.com>
1236 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
1238 2005-01-03 Uros Bizjak <uros@kss-loka.si>
1241 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
1243 (log1pdf2): Change mode of operands[0,1] to DFmode.
1245 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
1247 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
1248 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
1249 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
1250 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
1251 SPARC_RELAXED_ORDERING.
1253 2005-01-03 Richard Henderson <rth@redhat.com>
1254 Uros Bizjak <uros@kss-loka.si>
1257 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
1258 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
1259 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
1261 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
1262 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
1263 const_pow2_1_to_128_operand): New predicates.
1265 2005-01-02 Greg McGary <greg@mcgary.org>
1267 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
1269 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
1270 object. [ptr_type]: Remove unused variable.
1271 Remember nearest addressable array-element or record-component
1272 when checking bitfield components. Tolerate empty BIND_EXPR.
1274 2005-01-01 Richard Henderson <rth@redhat.com>
1276 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
1277 that don't narrow the value. Fail for other conversions.
1279 2005-01-01 Richard Henderson <rth@redhat.com>
1282 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
1283 * c-lang.c (finish_file): Don't do it here.
1284 * objc/objc-act.c (objc_finish_file): Likewise.
1286 * cgraph.c (decl_assembler_name_equal): New.
1287 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
1288 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
1289 * cgraph.h (struct cgraph_varpool_node): Add next.
1290 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
1291 * varasm.c (assemble_alias): Mark the target as needed.
1293 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
1296 * function.c (get_arg_pointer_save_area): Use entry_of_function
1297 instead of get_insns.
1299 2005-01-01 Roger Sayle <roger@eyesopen.com>
1300 Andrew Pinski <pinskia@physics.uc.edu>
1301 James E. Wilson <wilson@specifixinc.com>
1303 PR rtl-optimization/12092
1304 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
1307 2005-01-01 Roger Sayle <roger@eyesopen.com>
1308 Olivier Hainque <hainque@act-europe.fr>
1310 * tree.c (int_fits_type_p): A narrower type always fits in a
1311 wider one, except for negative values into unsigned types.
1313 2005-01-01 Roger Sayle <roger@eyesopen.com>
1315 * tree.c (int_fits_type_p): Compare the result of force_fit_type
1316 with the original constant rather than require TREE_OVERFLOW.
1318 2005-01-01 Steven Bosscher <stevenb@suse.de>
1321 * c-decl.c (finish_function): If compiling C99, annotate the
1322 compiler generated return with the current file name and line 0.
1323 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
1324 the source line is greater than 0.
1325 (remove_bb): Likewise.
1327 See ChangeLog.12 for earlier changes.