1 2005-01-14 Andrew Pinski <pinskia@physics.uc.edu>
3 * tree-ssa-alias.c (pass_may_alias): Add TODO_verify_stmts.
5 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
7 * config/rs6000/rs6000.h (CLASS_MAX_NREGS): DF goes in 1 register
9 (CANNOT_CHANGE_MODE_CLASS): Restrict DI mode changes on e500v2.
10 (PREDICATE_CODES): Add rs6k_nonimmediate_operand.
12 * config/rs6000/rs6000.c (invalid_e500_subreg): New.
13 (rs6k_nonimmediate_operand): New.
14 (rs6000_legitimate_offset_address_p): Handle DI modes on e500v2
16 (legitimate_lo_sum_address_p): Same.
17 (rs6000_legitimize_address): Same.
18 (rs6000_legitimize_reload_address): Same.
19 (rs6000_legitimate_address): Same.
20 (spe_build_register_parallel): Pass DF and DC modes in a DI
23 * config/rs6000/rs6000.md ("*movsi_internal1"): Change predicate
24 to rs6k_nonimmediate_operand.
26 * config/rs6000/spe.md ("*frob_df_di"): New.
28 ("*frob_di_df_2"): New.
29 ("*mov_sidf_e500_subreg0"): New.
30 ("*mov_sidf_e500_subreg4"): New.
31 ("*movdf_e500_double"): Change predicate to
32 rs6k_nonimmediate_operand.
34 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
36 * postreload.c (move2add_note_store): Only call
37 trunc_int_for_mode on scalar integers.
39 2005-01-14 Steven Bosscher <stevenb@suse.de>
41 * tree-ssa-dce.c (visited_control_parents): New sbitmap to
42 replace BB_VISITED uses.
43 (find_obviously_necessary_stmts): Don't clear BB_VISITED.
44 (propagate_necessity): Check the bitmap instead of BB_VISITED.
45 (tree_dce_done): Free visited_control_parents.
46 (perform_tree_ssa_dce): Allocate and clear it.
47 * tree-ssa-pre.c (compute_antic_aux): Make non-recursive.
48 (compute_antic): Iterate from here using a DFS. Use an sbitmap
49 instead of BB_VISITED.
51 2005-01-14 Kazu Hirata <kazu@cs.umass.edu>
53 * c-tree.h, coverage.h, langhooks-def.h, optabs.h, output.h,
54 rtl.h, targhooks.h, tree-flow.h, tree-mudflap.h, tree.h:
55 Remove unused prototypes.
57 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
59 * config/rs6000/rs6000.c (rs6000_stack_info): Fix caching of SPE
60 64-bit register usage.
61 (spe_func_has_64bit_regs_p): Add FIXME note.
63 2005-01-14 Aldy Hernandez <aldyh@redhat.com>
65 * config/rs6000/rs6000.c (rs6000_generate_compare): Replace
66 flag_finite_math_only with flag_unsafe_math_optimizations.
68 2005-01-14 J. D. Johnston <jjohnst@us.ibm.com>
70 * config/s390/tpf-unwind.h (s390_fallback_frame_state): Correct
72 (__tpf_eh_return): Copy TPF private stack area from the module
73 boundary stack frame. Add check for when module addresses are
74 equal, but no stub address is found.
76 2005-01-14 Richard Earnshaw <rearnsha@arm.com>
79 * arm.h (struct machine_function): Add call_via field.
80 (thumb_call_via_label): Declare.
81 * arm.c (thumb_call_via_label): New variable.
82 (thumb_call_reg_needed): New variable.
83 (arm_output_function_epilogue): For Thumb code, output any per-function
84 call-indirect trampolines.
85 (thumb_call_via_reg): New function.
86 (arm_file_end): New function.
87 (TARGET_ASM_FILE_END): Call arm_file_end.
88 (aof_file_end): Likewise.
89 * arm-protos.h (thumb_call_via_reg): Declare.
90 * arm.md (call_reg_thumb, call_value_reg_thumb): Call
91 thumb_call_via_reg in normal case.
93 2005-01-14 Jakub Jelinek <jakub@redhat.com>
96 PR rtl-optimization/19348
97 * recog.c (peephole2_optimize): Do global life update if some peephole
98 decides it doesn't need at least one of its inputs and that change
99 influences liveness at the start of the basic block.
101 * basic-block.h (EXECUTE_IF_AND_COMPL_IN_REG_SET): Needs 2 REGSET
102 arguments instead of 1.
104 2005-01-14 Eric Botcazou <ebotcazou@libertysurf.fr>
107 * varasm.c (initializer_constant_valid_p) <ADDR_EXPR>: Return
108 zero for nested functions needing a static chain or functions
109 with a non-constant address.
111 2005-01-13 Roger Sayle <roger@eyesopen.com>
113 * simplify-rtx.c (simplify_binary_operation) <AND>: Optimize
114 (and (sign_extend X) C) into (zero_extend (and X C)).
116 2005-01-13 David O'Brien <obrien@FreeBSD.org>
118 * config/freebsd-spec.h: Make KSE pthread lib logic the default.
120 2005-01-13 Richard Henderson <rth@redhat.com>
125 * config/i386/i386.md (cmpdf, cmpsf, bunordered, bordered, buneq,
126 bunge, bungt, bunle, bunlt, bltgt): Enable for TARGET_SSE_MATH,
128 (cmpfp_i_387): Rename from cmpfp_i. Move after sse patterns.
129 (cmpfp_i_mixed): Rename from cmpfp_i_sse; use for TARGET_MIX_SSE_I387.
130 (cmpfp_i_sse): Rename from cmpfp_i_sse_only; use for TARGET_SSE_MATH.
131 (cmpfp_iu_mixed, cmpfp_iu_sse, cmpfp_iu_387): Similarly.
132 (fp_jcc_1_mixed, fp_jcc_1_sse, fp_jcc_1_387): Similarly.
133 (fp_jcc_2_mixed, fp_jcc_2_sse, fp_jcc_2_387): Similarly.
134 (fp_jcc_3_387, fp_jcc_4_387, fp_jcc_5_387, fp_jcc_6_387,
135 fp_jcc_7_387, fp_jcc_8_387): Rename from fp_jcc_N.
136 (movdicc_c_rex64): Rename with '*'.
137 (movsfcc, movdfcc): Add checks for 387 and sse math to condition.
138 (movsfcc_1_sse_min, movsfcc_1_sse_max, movsfcc_1_sse): New.
139 (movsfcc_1_387): Rename from movsfcc_1.
140 (movdfcc_1_sse_min, movdfcc_1_sse_max, movdfcc_1_sse): New.
141 (movdfcc_1, movdfcc_1_rex64): Add check for 387.
142 (sminsf3, smaxsf3, smindf3, smaxdf3): New.
143 (minsf3, minsf, minsf_nonieee, minsf_sse, mindf3, mindf,
144 mindf_nonieee, mindf_sse, maxsf3, maxsf, maxsf_nonieee, maxsf_sse,
145 maxdf3, maxdf, maxdf_nonieee, maxdf_sse, sse_movsfcc, sse_movsfcc_eq,
146 sse_movdfcc, sse_movdfcc_eq, sse_movsfcc_const0_1,
147 sse_movsfcc_const0_2, sse_movsfcc_const0_3, sse_movsfcc_const0_4,
148 sse_movdfcc_const0_1, sse_movdfcc_const0_2, sse_movdfcc_const0_3,
149 sse_movdfcc_const0_4): Remove.
150 * config/i386/i386.c (ix86_expand_fp_movcc): For TARGET_SSE_MATH,
151 recognize min/max early. Update for changed sse cmove patterns.
152 (ix86_split_sse_movcc): New.
153 * config/i386/i386-protos.h: Update.
155 2005-01-13 Steven Bosscher <stevenb@suse.de>
157 * tree-ssa-dse.c (fix_phi_uses): Use SSA operand iterators.
158 (fix_stmt_v_may_defs): Likewise.
160 2005-01-13 Richard Henderson <rth@redhat.com>
162 * config/i386/i386.c (ix86_expand_fp_absneg_operator): Use elt_mode
163 for converting the mask.
165 2005-01-13 David Edelsohn <edelsohn@gnu.org>
167 * config/rs6000/aix52.h (CPLUSPLUS_CPP_SPEC): Change _XOPEN_SOURCE
170 2005-01-13 Richard Henderson <rth@redhat.com>
172 * config/i386/i386.c (IX86_BUILTIN_MOVQ, IX86_BUILTIN_LOADD,
173 IX86_BUILTIN_STORED, IX86_BUILTIN_MOVQ2DQ,
174 IX86_BUILTIN_MOVDQ2Q): Remove.
175 (IX86_BUILTIN_VEC_EXT_V4SI): New.
176 (ix86_init_mmx_sse_builtins, ix86_expand_builtin): Update to match.
177 (ix86_expand_vector_extract): For V4S[FI], extract element 0 after
179 * config/i386/sse.md (sse_concatv2sf): Accept zero operand 2.
180 (sse2_pextrw): Fix immediate constraint.
181 (sse2_loadq, sse2_loadq_rex64): Remove.
182 * config/i386/emmintrin.h (_mm_cvtsi128_si32, _mm_cvtsi128_si64x):
183 Use __builtin_ia32_vec_ext_<size>.
184 (_mm_cvtsi32_si128, _mm_cvtsi64x_si128): Use _mm_set_epi<size>.
186 2005-01-13 Aldy Hernandez <aldyh@redhat.com>
188 * function.c (assign_parm_setup_block): Look inside original
189 entry_parm when inspecting PARALLEL.
191 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
192 Joel Sherrill <joel@oarcorp.com>
195 * gthr-rtems.h (__gthread_recursive_mutex_t): New type.
196 (__GTHREAD_RECURSIVE_MUTEX_INIT_FUNCTION): Define to
197 rtems_gxx_recursive_mutex_init.
198 (__gthread_recursive_mutex_lock): New function.
199 (__gthread_recursive_mutex_trylock): Likewise.
200 (__gthread_recursive_mutex_unlock): Likewise.
202 2005-01-13 Ralf Corsepius <ralf.corsepius@rtems.org>
204 * config/i386/t-rtems-i386: Multilib on -mtune instead of -mcpu.
206 2005-01-13 Jan Beulich <jbeulich@novell.com>
208 * config/ia64/ia64.md (ashlti3, ashlti3_internal): New.
209 (ashrti3_internal): Indicate output is early clobber. Generate result
210 into output rather than first input. Use move for low word of output
211 if shift count is exactly 64.
212 (lshrti3_internal): Likewise.
214 2005-01-13 Hans-Peter Nilsson <hp@bitrange.com>
218 * reload1.c (delete_output_reload): Don't delete an output reload
219 if the pseudo lives longer than a single basic block. Adjust and
220 improve wording of comment.
222 2005-01-12 Andrew Pinski <pinskia@physics.uc.edu>
225 * config/darwin.c (machopic_select_section): Use TYPE_SIZE_UNIT instead
226 of TYPE_SIZE where we mean the number of bytes.
228 2005-01-12 Aldy Hernandez <aldyh@redhat.com>
230 * function.c (assign_parm_setup_block): Relax condition on
231 multi-register optimization.
233 2005-01-12 Nick Clifton <nickc@redhat.com>
235 * config/sh/sh.md (udivsi3_sh2a, divsi3_sh2a): Give these patterns
236 an "in_delay_slot" attribute of "no" to prevent them being used in
237 delay slots. This is forbidden because they might generate
240 2005-01-12 Alan Modra <amodra@bigpond.net.au>
243 * config/rs6000/rs6000.md (movtf_internal): Replace r->o and m->r
246 2005-01-12 Nick Clifton <nickc@redhat.com>
248 * config/iq2000/iq2000.h (ASM_SPEC): Undefine (to stop -Qy being
249 passed on to GAS) but do not define, as GAS no longer supports or
250 needs the -m2000 option.
252 2005-01-11 Roger Sayle <roger@eyesopen.com>
254 * builtins.c (fold_builtin_fabs): Convert argument to the appropriate
255 type as args of unprototyped builtins aren't automatically promoted.
256 (fold_builtin_abs): Likewise.
258 2005-01-11 Kaveh R. Ghazi <ghazi@caip.rutgers.edu>
260 * Makefile.in: Set a `build-warn' variable.
262 2005-01-11 Zdenek Dvorak <dvorakz@suse.cz>
264 PR tree-optimization/17949
265 * tree-ssa-loop-ivopts.c (may_be_unaligned_p): New function.
266 (find_interesting_uses_address): Use it.
268 2005-01-11 Aldy Hernandez <aldyh@redhat.com>
270 * regrename.c (kill_value): Handle subreg's that won't simplify.
272 2005-01-11 Richard Henderson <rth@redhat.com>
275 * config/i386/i386.h (enum ix86_builtins): Move ...
276 * config/i386/i386.c: ... here.
277 (IX86_BUILTIN_MOVDDUP, IX86_BUILTIN_MMX_ZERO, IX86_BUILTIN_PEXTRW,
278 IX86_BUILTIN_PINSRW, IX86_BUILTIN_LOADAPS, IX86_BUILTIN_LOADSS,
279 IX86_BUILTIN_STORESS, IX86_BUILTIN_SSE_ZERO, IX86_BUILTIN_PEXTRW128,
280 IX86_BUILTIN_PINSRW128, IX86_BUILTIN_LOADAPD, IX86_BUILTIN_LOADSD,
281 IX86_BUILTIN_STOREAPD, IX86_BUILTIN_STORESD, IX86_BUILTIN_STOREHPD,
282 IX86_BUILTIN_STORELPD, IX86_BUILTIN_SETPD1, IX86_BUILTIN_SETPD,
283 IX86_BUILTIN_CLRPD, IX86_BUILTIN_LOADPD1, IX86_BUILTIN_LOADRPD,
284 IX86_BUILTIN_STOREPD1, IX86_BUILTIN_STORERPD, IX86_BUILTIN_LOADDQA,
285 IX86_BUILTIN_STOREDQA, IX86_BUILTIN_CLRTI,
286 IX86_BUILTIN_LOADDDUP): Remove.
287 (IX86_BUILTIN_VEC_INIT_V2SI, IX86_BUILTIN_VEC_INIT_V4HI,
288 IX86_BUILTIN_VEC_INIT_V8QI, IX86_BUILTIN_VEC_EXT_V2DF,
289 IX86_BUILTIN_VEC_EXT_V2DI, IX86_BUILTIN_VEC_EXT_V4SF,
290 IX86_BUILTIN_VEC_EXT_V8HI, IX86_BUILTIN_VEC_EXT_V4HI,
291 IX86_BUILTIN_VEC_SET_V8HI, IX86_BUILTIN_VEC_SET_V4HI): New.
292 (ix86_init_builtins): Make static.
293 (ix86_init_mmx_sse_builtins): Update for changed builtins.
294 (ix86_expand_binop_builtin): Only use ix86_fixup_binary_operands
295 if all the modes match. Otherwise, fake it.
296 (get_element_number, ix86_expand_vec_init_builtin,
297 ix86_expand_vec_ext_builtin, ix86_expand_vec_set_builtin): New.
298 (ix86_expand_builtin): Make static. Update for changed builtins.
299 (ix86_expand_vector_move_misalign): Use sse2_loadlpd with zero
300 operand instead of sse2_loadsd. Cast sse1 fallback to V4SFmode.
301 (ix86_expand_vector_init_duplicate): New.
302 (ix86_expand_vector_init_low_nonzero): New.
303 (ix86_expand_vector_init_one_var, ix86_expand_vector_init_general):
304 Split out from ix86_expand_vector_init; handle integer modes.
305 (ix86_expand_vector_init): Use them.
306 (ix86_expand_vector_set, ix86_expand_vector_extract): New.
307 * config/i386/i386-protos.h: Update.
308 * config/i386/predicates.md (reg_or_0_operand): New.
309 * config/i386/mmx.md (mov<MMXMODEI>_internal): Add 'r' variants.
310 (movv2sf_internal): Likewise. And a splitter to match them all.
311 (vec_dupv2sf, mmx_concatv2sf, vec_setv2sf, vec_extractv2sf,
312 vec_initv2sf, vec_dupv4hi, vec_dupv2si, mmx_concatv2si, vec_setv2si,
313 vec_extractv2si, vec_initv2si, vec_setv4hi, vec_extractv4hi,
314 vec_initv4hi, vec_setv8qi, vec_extractv8qi, vec_initv8qi): New.
315 (mmx_pinsrw): Fix operand ordering.
316 * config/i386/sse.md (movv4sf splitter): Use direct pattern,
317 rather than sse_loadss expander.
318 (movv2df splitter): Similarly.
319 (sse_loadss, sse_loadlss): Remove.
320 (vec_dupv4sf, sse_concatv2sf, sse_concatv4sf, vec_extractv4sf_0): New.
321 (vec_setv4sf, vec_setv2df): Use ix86_expand_vector_set.
322 (vec_extractv4sf, vec_extractv2df): Use ix86_expand_vector_extract.
323 (sse3_movddup): Rename with '*'.
324 (sse3_movddup splitter): Use gen_rtx_REG instead of gen_lowpart.
325 (sse2_loadsd): Remove.
326 (vec_dupv2df_sse3): Rename from sse3_loadddup.
327 (vec_dupv2df, vec_concatv2df_sse3, vec_concatv2df): New.
328 (sse2_pinsrw): Fix argument ordering.
329 (sse2_loadld, sse2_loadq): Add sse1 alternatives.
330 (sse2_stored): Remove 'r' destination.
331 (vec_dupv4si, vec_dupv2di, sse2_concatv2si, sse1_concatv2si,
332 vec_concatv4si_1, vec_concatv2di, vec_setv2di, vec_extractv2di,
333 vec_initv2di, vec_setv4si, vec_extractv4si, vec_initv4si,
334 vec_setv8hi, vec_extractv8hi, vec_initv8hi, vec_setv16qi,
335 vec_extractv16qi, vec_initv16qi): New.
337 * config/i386/emmintrin.h (__m128i, __m128d): Use typedef, not define.
338 (_mm_set_sd, _mm_set1_pd, _mm_setzero_pd, _mm_set_epi64x,
339 _mm_set_epi32, _mm_set_epi16, _mm_set_epi8, _mm_setzero_si128): Use
341 (_mm_load_pd, _mm_store_pd): Use plain dereference.
342 (_mm_load_si128, _mm_store_si128): Likewise.
343 (_mm_load1_pd): Use _mm_set1_pd.
344 (_mm_load_sd): Use _mm_set_sd.
345 (_mm_store_sd, _mm_storeh_pd): Use __builtin_ia32_vec_ext_v2df.
346 (_mm_store1_pd, _mm_storer_pd): Use _mm_store_pd.
347 (_mm_set_epi64): Use _mm_set_epi64x.
348 (_mm_set1_epi64x, _mm_set1_epi64, _mm_set1_epi32, _mm_set_epi16,
349 _mm_set1_epi8, _mm_setr_epi64, _mm_setr_epi32, _mm_setr_epi16,
350 _mm_setr_epi8): Use _mm_set_foo form.
351 (_mm_loadl_epi64, _mm_movpi64_epi64, _mm_move_epi64): Use _mm_set_epi64.
352 (_mm_storel_epi64, _mm_movepi64_pi64): Use __builtin_ia32_vec_ext_v2di.
353 (_mm_extract_epi16): Use __builtin_ia32_vec_ext_v8hi.
354 (_mm_insert_epi16): Use __builtin_ia32_vec_set_v8hi.
355 * config/i386/mmintrin.h (_mm_setzero_si64): Use plain cast.
356 (_mm_set_pi32): Use __builtin_ia32_vec_init_v2si.
357 (_mm_set_pi16): Use __builtin_ia32_vec_init_v4hi.
358 (_mm_set_pi8): Use __builtin_ia32_vec_init_v8qi.
359 (_mm_set1_pi16, _mm_set1_pi8): Use _mm_set_piN variant.
360 * config/i386/pmmintrin.h (_mm_loaddup_pd): Use _mm_load1_pd.
361 (_mm_movedup_pd): Use _mm_shuffle_pd.
362 * config/i386/xmmintrin.h (_mm_setzero_ps, _mm_set_ss,
363 _mm_set1_ps, _mm_set_ps, _mm_setr_ps): Use constructor form.
364 (_mm_cvtpi16_ps, _mm_cvtpu16_ps, _mm_cvtpi8_ps, _mm_cvtpu8_ps,
365 _mm_cvtps_pi8, _mm_cvtpi32x2_ps): Avoid __builtin_ia32_mmx_zero;
367 (_mm_load_ss, _mm_load1_ps): Use _mm_set* form.
368 (_mm_load_ps, _mm_loadr_ps): Use raw dereference.
369 (_mm_store_ss): Use __builtin_ia32_vec_ext_v4sf.
370 (_mm_store_ps): Use raw dereference.
371 (_mm_store1_ps): Use _mm_storeu_ps.
372 (_mm_storer_ps): Use _mm_store_ps.
373 (_mm_extract_pi16): Use __builtin_ia32_vec_ext_v4hi.
374 (_mm_insert_pi16): Use __builtin_ia32_vec_set_v4hi.
376 2005-01-11 Stan Shebs <shebs@apple.com>
378 * config/rs6000/rs6000.c (machopic_output_stub): Issue
379 ldu instead of lwzu if 64-bit -mdynamic-no-pic.
381 2005-01-11 Andrew Pinski <pinskia@physics.uc.edu>
384 * config/rs6000/rs6000.c (rs6000_special_round_type_align):
385 Skip all DECLs except for FIELD_DECLs.
387 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
389 * config/s390/s390.c (override_options): Return error if
390 -mbackchain, -mpacked-stack and -mhard-float are used together.
391 (s390_va_start): Remove the backchain && packed-stack special case.
392 (s390_gimplify_va_arg): Likewise.
393 * doc/invoke.texi: Remove the ABI incompatibility note.
395 2005-01-11 Andreas Krebbel <krebbel1@de.ibm.com>
397 * config/s390/s390.c (struct s390_frame_layout): Remove
399 (s390_frame_info, s390_emit_prologue): Replace occurrences of
400 save_backchain_p with TARGET_BACKCHAIN.
402 2005-01-11 Alan Modra <amodra@bigpond.net.au>
405 * builtins.c (std_gimplify_va_arg_expr): Adjust alignment of *ap.
406 * expr.h (struct locate_and_pad_arg_data): Add "boundary".
407 * function.c (locate_and_pad_parm): Set new field.
408 (assign_parm_find_stack_rtl): Use it instead of FUNCTION_ARG_BOUNDARY.
409 Tweak where_pad test to include "none". Always set mem align for
411 (assign_parm_adjust_stack_rtl): Discard stack_parm if alignment
412 not sufficient for type.
413 (assign_parm_setup_block): If stack_parm is zero on entry, always
414 make a new stack local. Block move old stack parm if necessary
415 to new aligned stack local.
416 (assign_parm_setup_stack): Use a block move to handle
417 potentially misaligned entry_parm.
418 (assign_parms_unsplit_complex): Specify required alignment when
419 creating stack local.
420 * calls.c (compute_argument_addresses): Override alignment of stack
421 arg calculated from its type with the alignment given by
422 FUNCTION_ARG_BOUNDARY.
423 (store_one_arg): Likewise.
425 2005-01-11 Jan Beulich <jbeulich@novell.com>
427 * config/ia64/ia64.md (zero_extendsidi2): Replace zxt4 by addp4.
428 Change respective itanium_class attribute to ialu.
429 (shladdp4_internal): New.
430 * config/ia64/predicates.md (shladd_log2_operand): New.
432 2005-01-11 Richard Henderson <rth@redhat.com>
434 * expr.c (store_constructor): Use rtvec_alloc instead of
435 alloca+gen_rtvec_v, and an incorrect number passed to alloca.
437 2005-01-11 Kazu Hirata <kazu@cs.umass.edu>
439 * config/alpha/alpha.c, config/i386/mmx.md: Fix comment typos.
441 2005-01-11 Alan Modra <amodra@bigpond.net.au>
443 * varasm.c (default_section_type_flags_1): Don't set SECTION_SMALL.
444 * config/ia64/ia64.c (TARGET_SECTION_TYPE_FLAGS): Define.
445 (TARGET_RWRELOC): Define.
446 (ia64_rwreloc_section_type_flags): Delete.
447 (ia64_section_type_flags): New function.
448 * config/ia64/hpux.h (TARGET_SECTION_TYPE_FLAGS): Don't define.
449 (TARGET_RWRELOC): Define.
451 2005-01-10 David Mosberger <davidm@hpl.hp.com>
454 * config/ia64/ia64.c (process_set): For alloc insn, only call
455 process_epilogue is !frame_pointer_needed.
457 2005-01-10 Roger Sayle <roger@eyesopen.com>
460 * c-common.c (c_common_truthvalue_conversion): TRUTH_NOT_EXPR is a
461 unary operator and can't be treated as a binary/comparison operator.
463 2005-01-10 Richard Henderson <rth@redhat.com>
465 * config/i386/i386.c (ix86_function_value): Use type_natural_mode.
466 (ix86_return_in_memory): Likewise.
467 (function_arg_advance): Likewise. Mirror structure in function_arg
468 for choosing register to advance.
470 2005-01-10 Kazu Hirata <kazu@cs.umass.edu>
472 * tree-vectorizer.c, tree.def: Fix comment typos.
474 2005-01-10 Mark Dettinger <dettinge@de.ibm.com>
476 * config/s390/s390.c (struct processor_costs): 4 new fields:
477 ddbr, ddr, debr, der.
478 (s390_rtx_costs): More precise handling of divide instructions.
480 2005-01-10 David Edelsohn <edelsohn@gnu.org>
482 * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
484 2005-01-10 Jan Beulich <jbeulich@novell.com>
486 * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
487 names resulting from -ffunction-sections/-fdata-sections and linkonce
489 * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
490 based on the section name. Rearrange the section name comparison logic
491 slightly so that each section name is compared against at most once.
493 2005-01-10 Ben Elliston <bje@au.ibm.com>
495 * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
496 targets for which -fPIC reduces limitations on the GOT size.
498 2005-01-09 Falk Hueffner <falk@debian.org>
500 * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
503 2005-01-09 David Edelsohn <edelsohn@gnu.org>
506 * collect2.c (main): Set aixrtl_flag for -brtl option.
507 (resolve_lib_name): Search for .so file extension before .a
510 2005-01-09 Dorit Naishlos <dorit@il.ibm.com>
512 * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
513 (vect_analyze_data_refs_alignment): Add dump prints.
515 2005-01-09 Ira Rosen <irar@il.ibm.com>
517 * tree-vectorizer.c (vect_analyze_offset_expr): Use
518 expr_invariant_in_loop_p.
519 Initialize outputs first thing in the function.
520 (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
521 (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
522 (vect_analyze_pointer_ref_access): Check that the initial condition of
523 the access function is loop invariant.
525 2005-01-09 Richard Henderson <rth@redhat.com>
527 * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
528 (ix86_expand_builtin): Likewise. Frob MASKMOVQ wrt the input mem
529 just like MASKMOVDQU. Return plain zero for MMX_ZERO.
530 * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
531 mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
532 movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
533 movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
534 sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
535 ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
536 mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
537 mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
538 sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
539 mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
540 mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
541 gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
542 ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
543 mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
544 mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
545 mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
546 eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
547 pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
548 pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
549 pswapdv2sf2): Move to mmx.md; rename as necessary with leading
551 (mmx_clrdi, pavgusb): Remove.
552 (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
553 with leading sse_ prefix.
554 * config/i386/sse.md: Receive them.
555 * config/i386/mmx.md: New file.
556 (MMXMODE12, MMXMODE24, mmxvecsize): New.
557 (subrv2sf3): Turn into expander for normal subtraction.
558 (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
559 mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
560 mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
561 mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
562 commutative; use ix86_binary_operator_ok.
563 (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
564 mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
565 mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
566 mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
567 mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
568 Macroize from existing patterns; use ix86_binary_operator_ok.
569 (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
570 (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
571 mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise. Model
572 with vec_select+vec_concat.
573 (mmx_pshufw, mmx_pshufw_1): Likewise.
574 (mmx_uavgv8qi3): Merge pavgusb. Model correcty.
575 (mmx_uavgv4hi3): Model correctly.
576 * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
577 _mm_xor_si64): Remove casts.
579 2005-01-09 Zdenek Dvorak <dvorakz@suse.cz>
581 PR tree-optimization/19224
582 * tree-scalar-evolution.c (get_instantiated_value,
583 set_instantiated_value): New functions.
584 (instantiate_parameters_1): Cache the results.
585 (instantiate_parameters, resolve_mixers): Initialize and free
588 2005-01-08 David Edelsohn <edelsohn@gnu.org>
590 * config/i386/i386.md (addhi_4): Correct reference in comment.
593 2005-01-08 Richard Henderson <rth@redhat.com>
595 * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
596 (_mm_cvtsi128_si64x): Likewise.
597 (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
598 _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
599 the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
600 * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
601 * config/i386/i386-protos.h: Update.
602 * config/i386/i386.c (print_operand): Add 'H'.
603 (ix86_fixup_binary_operands): Split out from ...
604 (ix86_expand_binary_operator): ... here.
605 (ix86_fixup_binary_operands_no_copy): New.
606 (ix86_expand_fp_absneg_operator): Handle vector mode results.
607 (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
608 (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
609 (safe_vector_operand): Use CONST0_RTX.
610 (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
611 (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
612 CODE_FOR_sse2_maskmovdqu. Special case SSE version of MASKMOVDQU
613 expansion. Update names for sse{,2,3}_ prefixes. Remove *maskncmp*
615 * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
616 (IX86_BUILTIN_CMPNGESS): New.
617 * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
618 (attr type): Add sselog1.
619 (attr unit, attr memory): Handle it.
620 (movti, movti_internal, movti_rex64): Move near other integer moves.
621 (movtf, movtf_internal): Move near other fp moves.
622 (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
623 vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
624 movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
625 movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
626 sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
627 sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
628 sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
629 mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
630 rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
631 sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
632 sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
633 sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
634 vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
635 smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
636 cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
637 cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
638 vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
639 vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
640 sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
641 sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
642 cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
643 cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
644 cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
645 cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
646 ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
647 subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
648 smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
649 sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
650 sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
651 sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
652 gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
653 ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
654 ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
655 sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
656 sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
657 sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
658 sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
659 sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
660 sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
661 sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
662 mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
663 addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
664 movsldup, lddqu, loadddup, movddup): Move to sse.md. Any with
665 non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
667 (SSEPUSH, push<SSEPUSH>): Remove.
668 (MMXPUSH, push<MMXPUSH>): Remove.
669 (sse_movaps, sse_movaps_1, sse_movups): Remove.
670 (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
671 (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
672 (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
673 (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
674 (maskncmpv2df3, vmmaskncmpv2df3): Remove.
675 (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
676 (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
677 * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
678 (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
679 * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
680 (ppro_sse_log_V4SF_load): Similarly. Handle sselog1.
681 (ppro_sse_log_V4SF): Handle sselog1.
682 * config/i386/predicates.md (const_0_to_1_operand): New.
683 (const_0_to_255_mul_8_operand): New.
684 (const_1_to_31_operand): Rename from const_int_1_31_operand.
685 (const_2_to_3_operand, const_4_to_7_operand): New.
686 * config/i386/sse.md: New file.
687 (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
688 (sse_movups): Rename from sse_movups_1.
689 (sse_loadlss): Rename from sse_loadss_1.
690 (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
692 (negv4sf2): Use ix86_expand_fp_absneg_operator.
693 (absv4sf2, negv2df, absv2df): New.
694 (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
695 (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
696 iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
697 smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
698 umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
699 (sse3_addsubv4sf3): Model correctly.
700 sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
701 sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
702 (sse_movhlps): Model with vec_select+vec_concat.
703 (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
704 sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
705 sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
706 sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
707 sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
708 sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
709 sse2_pshufhw_1): Likewise.
710 (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
711 (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
712 sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
713 ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
714 sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
715 ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
716 (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
717 sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
718 mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
719 sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
720 sminv8hi3): Mark commutative
721 operands. Use ix86_binary_operator_ok.
722 (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
723 sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
724 sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
725 sse2_punpcklqdq): Allow operand2 in memory.
726 (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
727 sse2_movsd): Add memory alternatives.
728 (sse_storelps): Turn expander into an insn; split after reload.
729 (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
730 (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
732 2005-01-08 Eric Botcazou <ebotcazou@libertysurf.fr>
734 * configure.ac (DWARF-2 debug_line): Use objdump.
735 * configure: Regenerate.
737 2005-01-08 Jeff Law <law@redhat.com>
738 Diego Novillo <dnovillo@redhat.com>
740 PR tree-optimization/18241
741 * tree-nrv.c (tree_nrv): Ignore volatile return values.
742 * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
743 statements with volatile operands.
744 * tree-ssa-operands.c (add_stmt_operand): Do add volatile
745 operands after marking a statement with has_volatile_ops.
747 2005-01-08 Roger Sayle <roger@eyesopen.com>
749 * tree.c (int_fits_type_p): Always honor integer constant
750 TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
752 2005-01-08 Roger Sayle <roger@eyesopen.com>
754 * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
755 the latent bug in rtl_delete_block has been resolved.
757 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
759 * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
760 (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
762 2005-01-08 Richard Sandiford <rsandifo@redhat.com>
764 * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
765 library as @multilib_dir@/$(SHLIB_SONAME).
767 2005-01-07 Eric Botcazou <ebotcazou@libertysurf.fr>
769 * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
770 * configure: Regenerate.
772 2005-01-07 Jakub Jelinek <jakub@redhat.com>
774 * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
777 2005-01-07 David Edelsohn <edelsohn@gnu.org>
780 * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
781 Convert non-word aligned offset address using ld/std into
784 2005-01-07 Richard Henderson <rth@redhat.com>
786 * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
788 2005-01-07 Richard SAndiford <rsandifo@redhat.com>
790 * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
792 2005-01-07 Jakub Jelinek <jakub@redhat.com>
794 PR tree-optimization/19283
795 * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
796 is not shorter than the original type.
798 PR rtl-optimization/19012
799 * config/i386/i386.md (addqi_1_slp): Set memory attribute.
801 PR rtl-optimization/18861
802 * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
803 (FULL_STATE): Define.
804 (mark_tablejump_edge): New function.
805 (purge_dead_tablejump_edges): New function.
806 (find_bb_boundaries): Use it.
808 PR tree-optimization/18828
809 * builtins.c (expand_builtin_next_arg): Remove argument and all
810 the argument checking.
811 (expand_builtin): Adjust caller.
812 (expand_builtin_va_start): Likewise. Remove error for too many
814 (fold_builtin_next_arg): Issue error for too many arguments.
815 After checking arguments, replace them with magic arguments that
816 prevent further checking of the args.
818 2005-01-06 John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
820 * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
822 2005-01-06 Roger Sayle <roger@eyesopen.com>
825 * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
826 and %{.S:...} (and their negative variants) to test whether the
827 input file is assembler or pre-processed-assembler independent of
828 the actual filename extension.
830 2005-01-06 Roger Sayle <roger@eyesopen.com>
832 * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
833 of sign or zero extended values.
835 2005-01-06 Geoffrey Keating <geoffk@apple.com>
837 * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
838 for IBM long double format correctly.
840 2005-01-06 Daniel Berlin <dberlin@dberlin.org>
842 Fix PR tree-optimization/18792
844 * tree-data-ref.c (build_classic_dist_vector): Change first_loop
845 to first_loop_depth, and use loop depth instead of loop number.
846 (build_classic_dir_vector): Ditto.
847 (compute_data_dependences_for_loop): Use depth, not loop number.
848 * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
849 number. Pass in loops, instead of loop numbers.
850 (gather_interchange_stats): Ditto.
851 (linear_transform_loops): Ditto.
853 2005-01-06 Richard Sandiford <rsandifo@redhat.com>
856 * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
857 biv_fits_mode_p, extension_within_bounds_p): New functions.
858 (check_ext_dependent_givs): Use them.
860 2005-01-06 Roger Sayle <roger@eyesopen.com>
862 * cfgrtl.c (rtl_delete_block): A basic block may be followed by
863 more than one barrier, in which case we should delete them all.
865 2005-01-06 Tobias Schlueter <tobias.schlueter@physik.uni-muenchen.de>
867 * gcc.c (process_command): Change year in 'gcc --version' to 2005.
869 2005-01-05 Daniel Berlin <dberlin@dberlin.org>
871 Fix PR middle-end/19286
873 * dwarf2out.c (gen_subprogram_die): If we've already tried to
874 output this subprogram, simply ignore this attempt to do it again.
875 (add_abstract_origin_attribute): Don't abort trying to add the abstract
876 origin attribute if it's not possible.
877 (gen_block_die): Don't ignore subblocks of "unused" blocks.
878 (decls_for_scope): Ditto.
879 * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
880 (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
882 2005-01-05 Richard Henderson <rth@redhat.com>
885 * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
886 (bdesc_2arg): Use it.
887 (ix86_expand_binop_builtin): Force operands into registers
889 (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
890 ix86_expand_sse_compare, ix86_expand_sse_comi,
891 ix86_expand_builtin): Likewise.
893 2005-01-05 Richard Henderson <rth@redhat.com>
895 * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
897 * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
898 (mulv4hi3): Set itanium_class mmmul.
899 (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
900 (fpack): Rename from fpack_sfsf.
902 2005-01-05 Richard Henderson <rth@redhat.com>
905 * reload1.c (do_input_reload): Restrict the optimization deleteing
906 a previous output reload to RELOAD_FOR_INPUT.
908 2005-01-05 Steven Bosscher <stevenb@suse.de>
910 * combine.c (expand_compound_operation) <ZERO_EXTRACT>: Add
911 comment that we fall through after case.
912 (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
913 (distribute_links): Likewise.
914 * cse.c (cse_insn): Likewise.
915 * cselib.c (cselib_invalidate_mem): Likewise.
916 * df.c: Update comments at the top of the file.
917 (read_modify_subreg_p): Update comments here too.
918 (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
919 * flow.c (mark_set_1): Likewise.
920 (mark_used_regs): Likewise.
921 * gcse.c (mems_conflict_for_gcse_p): Likewise.
922 (canon_list_insert): Likewise.
923 (mark_set): Likewise.
924 (try_replace_reg): Likewise.
925 (store_killed_in_insn): Likewise.
926 * loop.c (count_one_set): Likewise.
927 (basic_induction_var): Likewise.
928 * postreload-gcse.c (find_mem_conflicts): Likewise.
929 * postreload.c (reload_combine_note_store): Likewise.
930 (move2add_note_store): Likewise.
931 * reload.c (find_equiv_reg): Likewise.
932 (mark_referenced_resources): Likewise.
933 * rtlanal.c (set_noop_p): Likewise.
934 (note_stores): Likewise.
935 (note_uses): Likewise.
936 * sched-deps.c (sched_analyze_1): Likewise.
937 * sched-rgn.c (check_live_1): Likewise.
938 (update_live_1): Likewise.
939 * config/i860/i860.c: Likewise.
941 * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
943 (ZERO_EXTRACT): Mention that this one can be an lvalue.
945 * doc/rtl.texi: Update documentation for bit-fields and SET.
947 2005-01-05 Roger Sayle <roger@eyesopen.com>
949 * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
951 2005-01-05 Jan Hubicka <jh@suse.cz>
952 Richard Henderson <rth@redhat.com>
955 * config/i386/i386.c (ix86_expand_move): Handle tls symbols
958 2005-01-05 Richard Henderson <rth@redhat.com>
961 * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
962 (sse_shufps): Change operand 3 to const_int_operand.
963 (sse2_storelps): Fix typo in template.
964 (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
965 * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
966 (ix86_expand_builtin): Likewise.
968 2005-01-05 Julian Brown <julian@codesourcery.com>
970 * config/arm/arm.c (arm_return_in_memory): Treat complex types
971 as aggregates for AAPCS ABIs.
973 2005-01-05 Stan Shebs <shebs@apple.com>
975 * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
976 Mach-O getters if ppc64.
977 * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
979 * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
980 signed four-byte field for global code case.
981 (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
983 2005-01-05 Roger Sayle <roger@eyesopen.com>
986 * c-common.c: Include real.h.
987 (c_common_truthvalue_conversion): Avoid destructively modifying expr.
988 Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
989 Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
990 * Makefile.in (c-common.o): Update dependencies.
992 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
994 * c-parse.in (asm_string): Add trailing semicolon.
996 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
998 * c-parse.in (asm_string): New. Don't allow wide strings in
1000 (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
1001 asm_string instead of STRING.
1003 2005-01-05 Joseph S. Myers <joseph@codesourcery.com>
1005 * c-typeck.c (constructor_no_implicit): Remove.
1006 (set_designator, process_init_element): Don't check
1007 constructor_no_implicit.
1009 2005-01-05 J"orn Rennecke <joern.rennecke@st.com>
1010 Kaz Kojima <kkojima@gcc.gnu.org>
1013 * lcm.c (create_pre_exit): New.
1014 (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
1015 ENTRY_EXIT_EXTRA to 3. Use create_pre_exit.
1017 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
1019 * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
1020 -msoft-float in target_float_switch.
1021 * arm.c (arm_override_options): Fix processing of target_float_switch.
1023 2004-01-05 Richard Earnshaw <rearnsha@arm.com>
1025 * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
1028 2004-01-05 Uros Bizjak <uros@kss-loka.si>
1030 * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
1031 Replace i387 with 'i386 compiler' in -mfpmath=sse option.
1033 2005-01-04 Roger Sayle <roger@eyesopen.com>
1035 * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
1036 if deleting the then-block allows the test-block to fallthru to the
1039 2005-01-04 Andrew Pinski <pinskia@physics.uc.edu>
1042 * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
1043 after the full declaration if the are in two different TUs.
1045 2005-01-04 Richard Henderson <rth@redhat.com>
1048 * tree-sra.c (generate_one_element_init): Just
1049 call gimplify_and_add.
1050 (generate_element_init): Record the
1051 new referenced variables and mark them for renaming
1052 and split out to ...
1053 (generate_element_init_1): This.
1054 (scalarize_init): Don't call push_gimplify_context/
1055 pop_gimplify_context.
1057 2005-01-04 Geoffrey Keating <geoffk@apple.com>
1059 * toplev.c (get_src_pwd): Handle failure of getpwd().
1061 2005-01-04 Roger Sayle <roger@eyesopen.com>
1063 * fold-const.c (fold_single_bit_test): Delete unreachable handling
1065 (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
1066 all the cases handled by it are inverted by invert_truthvalue.
1068 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
1070 * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
1071 (save_register_windowsi): Likewise.
1073 2005-01-04 Richard Henderson <rth@redhat.com>
1075 * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
1077 2005-01-04 Uros Bizjak <uros@kss-loka.si>
1080 * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
1081 any simplifications of vector mode comparison operators.
1082 * simplify-rtx.c (simplify_relational_operation): Fix variable name.
1084 2005-01-04 Paolo Bonzini <bonzini@gnu.org>
1085 Devang Patel <dpatel@apple.com>
1087 PR tree-optimization/18308
1088 * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
1089 the operands before creating a new expression.
1090 * dojump.c (do_jump): Make drop_through_label available
1091 for all cases. Add expansion of COND_EXPR.
1093 2005-01-04 Ira Rosen <irar@il.ibm.com>
1095 * tree-vectorizer.c (vect_analyze_offset_expr): Test for
1096 INTEGER_CST instead of TREE_CONSTANT.
1097 (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
1098 instead of TREE_CONSTANT.
1099 (vect_analyze_pointer_ref_access): Test for INTEGER_CST
1100 instead of TREE_CONSTANT.
1102 2005-01-04 Eric Botcazou <ebotcazou@libertysurf.fr>
1104 * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
1106 2005-01-04 Uros Bizjak <uros@kss-loka.si>
1109 * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
1110 (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
1112 2005-01-03 Richard Henderson <rth@redhat.com>
1114 * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
1115 (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
1116 fold_convert_const_real_from_real): Split out from ...
1117 (fold_convert_const): ... here.
1119 2005-01-03 Richard Henderson <rth@redhat.com>
1122 * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
1123 (mov<MMXMODEI>_internal): Likewise.
1124 (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
1125 Add V2SF case; use it for SSE1; don't use TI.
1126 (movdf_integer): Likewise.
1127 (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
1129 2005-01-03 Ira Rosen <irar@il.ibm.com>
1131 * tree-vectorizer.c (vect_strip_conversions): New function.
1132 (vect_analyze_offset_expr): Call vect_strip_conversions. Add
1133 check for binary class.
1135 2005-01-03 Daniel Berlin <dberlin@dberlin.org>
1139 * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
1141 * gimple-low.c (mark_blocks_with_used_vars): New function.
1142 (mark_blocks_with_used_subblocks): Ditto.
1143 (mark_used_blocks): Ditto.
1144 (pass_mark_used_blocks): New pass.
1145 * tree-inline.c: Include debug.h.
1146 (expand_call_inline): Call outlining_inline_function here.
1147 * tree-optimize.c (init_tree_optimization_passes): Add
1148 pass_mark_used_blocks.
1149 * tree-pass.h (pass_mark_used_blocks): New.
1150 * Makefile.in (tree-inline.o): Add debug.h dependency.
1152 2005-01-03 Geoffrey Keating <geoffk@apple.com>
1154 * config/darwin.c (darwin_handle_weak_import_attribute): Permit
1155 VAR_DECLs to have weak_import attribute.
1157 * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
1160 2004-01-03 Steven Bosscher <stevenb@suse.de>
1162 * passes.c (rest_of_compilation): Don't run regmove if only
1163 flag_expensive_optimizations, require flag_regmove instead.
1165 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
1167 * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
1169 * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
1170 * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1171 * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
1172 * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
1174 2005-01-03 Richard Henderson <rth@redhat.com>
1176 * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
1177 (ia64_const_ok_for_letter_p): New.
1178 (ia64_const_double_ok_for_letter_p): New.
1179 (ia64_extra_constraint): New.
1180 (ia64_expand_vecint_compare): New.
1181 (ia64_expand_vcondu_v2si): New.
1182 (ia64_expand_vecint_cmov): New.
1183 (ia64_expand_vecint_minmax): New.
1184 (ia64_print_operand): Add 'v'.
1185 (ia64_preferred_reload_class): New.
1186 (ia64_vector_mode_supported_p): New.
1187 * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
1188 (PREFERRED_RELOAD_CLASS): Move to function.
1189 (CONST_OK_FOR_LETTER_P): Move to function.
1190 (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
1191 (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
1192 (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
1193 (EXTRA_CONSTRAINT): Move to function.
1194 * config/ia64/ia64.md: Include vect.md.
1195 (itanium_class): Add mmalua.
1197 * config/ia64/itanium1.md (1_mmalua): New. Add it to bypasses.
1199 * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
1200 * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
1202 (const_int_2bit_operand): New.
1203 (fr_reg_or_0_operand): New.
1204 * config/ia64/ia64-modes.def: Add vector modes.
1205 * config/ia64/ia64-protos.h: Update.
1206 * config/ia64/vect.md: New file.
1208 2005-01-03 Richard Henderson <rth@redhat.com>
1210 * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
1212 2005-01-03 Uros Bizjak <uros@kss-loka.si>
1215 * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
1217 (log1pdf2): Change mode of operands[0,1] to DFmode.
1219 2005-01-03 Eric Botcazou <ebotcazou@libertysurf.fr>
1221 * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
1222 * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
1223 * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
1224 * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
1225 SPARC_RELAXED_ORDERING.
1227 2005-01-03 Richard Henderson <rth@redhat.com>
1228 Uros Bizjak <uros@kss-loka.si>
1231 * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
1232 IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
1233 * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
1235 (*mmx_pinsrw, *sse2_pinsrw): New patterns.
1236 * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
1237 const_pow2_1_to_128_operand): New predicates.
1239 2005-01-02 Greg McGary <greg@mcgary.org>
1241 * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
1243 (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
1244 object. [ptr_type]: Remove unused variable.
1245 Remember nearest addressable array-element or record-component
1246 when checking bitfield components. Tolerate empty BIND_EXPR.
1248 2005-01-01 Richard Henderson <rth@redhat.com>
1250 * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
1251 that don't narrow the value. Fail for other conversions.
1253 2005-01-01 Richard Henderson <rth@redhat.com>
1256 * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
1257 * c-lang.c (finish_file): Don't do it here.
1258 * objc/objc-act.c (objc_finish_file): Likewise.
1260 * cgraph.c (decl_assembler_name_equal): New.
1261 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
1262 (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
1263 * cgraph.h (struct cgraph_varpool_node): Add next.
1264 (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
1265 * varasm.c (assemble_alias): Mark the target as needed.
1267 2005-01-01 Andrew Pinski <pinskia@physics.uc.edu>
1270 * function.c (get_arg_pointer_save_area): Use entry_of_function
1271 instead of get_insns.
1273 2005-01-01 Roger Sayle <roger@eyesopen.com>
1274 Andrew Pinski <pinskia@physics.uc.edu>
1275 James E. Wilson <wilson@specifixinc.com>
1277 PR rtl-optimization/12092
1278 * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
1281 2005-01-01 Roger Sayle <roger@eyesopen.com>
1282 Olivier Hainque <hainque@act-europe.fr>
1284 * tree.c (int_fits_type_p): A narrower type always fits in a
1285 wider one, except for negative values into unsigned types.
1287 2005-01-01 Roger Sayle <roger@eyesopen.com>
1289 * tree.c (int_fits_type_p): Compare the result of force_fit_type
1290 with the original constant rather than require TREE_OVERFLOW.
1292 2005-01-01 Steven Bosscher <stevenb@suse.de>
1295 * c-decl.c (finish_function): If compiling C99, annotate the
1296 compiler generated return with the current file name and line 0.
1297 * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
1298 the source line is greater than 0.
1299 (remove_bb): Likewise.
1301 See ChangeLog.12 for earlier changes.