gcc/ChangeLog

   1 2005-01-10  David Edelsohn  <edelsohn@gnu.org>
   2
   3         * config/rs6000/t-aix43 (BOOT_LDFLAGS): Define.
   4
   5 2005-01-10  Jan Beulich  <jbeulich@novell.com>
   6
   7         * config/ia64/ia64.c (ia64_in_small_data_p): Also handle the section
   8         names resulting from -ffunction-sections/-fdata-sections and linkonce
   9         ones.
  10         * varasm.c (default_section_type_flags_1): Also set SECTION_SMALL
  11         based on the section name. Rearrange the section name comparison logic
  12         slightly so that each section name is compared against at most once.
  13
  14 2005-01-10  Ben Elliston  <bje@au.ibm.com>
  15
  16         * doc/invoke.texi (Code Gen Options): Add PowerPC to the list of
  17         targets for which -fPIC reduces limitations on the GOT size.
  18
  19 2005-01-09  Falk Hueffner  <falk@debian.org>
  20
  21         * fold-const.c (fold): Also handle EXACT_DIV_EXPR when folding
  22         X/C1 cmpop C2.
  23
  24 2005-01-09  David Edelsohn  <edelsohn@gnu.org>
  25
  26         PR target/18720
  27         * collect2.c (main): Set aixrtl_flag for -brtl option.
  28         (resolve_lib_name): Search for .so file extension before .a
  29         if aixrtl_flag set.
  30
  31 2005-01-09  Dorit Naishlos  <dorit@il.ibm.com>
  32
  33         * tree-vectorizer.c (vect_enhance_data_refs_alignment): Add dump prints.
  34         (vect_analyze_data_refs_alignment): Add dump prints.
  35
  36 2005-01-09  Ira Rosen  <irar@il.ibm.com>
  37
  38         * tree-vectorizer.c (vect_analyze_offset_expr): Use
  39         expr_invariant_in_loop_p.
  40         Initialize outputs first thing in the function.
  41         (vect_update_ivs_after_vectorizer): Call initial_condition_in_loop_num.
  42         (vect_is_simple_iv_evolution): Call initial_condition_in_loop_num.
  43         (vect_analyze_pointer_ref_access): Check that the initial condition of
  44         the access function is loop invariant.
  45
  46 2005-01-09  Richard Henderson  <rth@redhat.com>
  47
  48         * config/i386/i386.c (bdesc_2arg): Update names for mmx_ prefixes.
  49         (ix86_expand_builtin): Likewise.  Frob MASKMOVQ wrt the input mem
  50         just like MASKMOVDQU.  Return plain zero for MMX_ZERO.
  51         * config/i386/i386.md (MMXMODEI, mov<MMXMODEI>,
  52         mov<MMXMODEI>_internal_rex64, mov<MMXMODEI>_internal, movv2sf,
  53         movv2sf_internal_rex64, movv2sf_internal, MMXMODE,
  54         movmisalign<MMXMODE>, mmx_pmovmskb, mmx_maskmovq, mmx_maskmovq_rex,
  55         sse_movntdi, addv8qi3, addv4hi3, addv2si3, mmx_adddi3, ssaddv8qi3,
  56         ssaddv4hi3, usaddv8qi3, usaddv4hi3, subv8qi3, subv4hi3, subv2si3,
  57         mmx_subdi3, sssubv8qi3, sssubv4hi3, ussubv8qi3, ussubv4hi3,
  58         mulv4hi3, smulv4hi3_highpart, umulv4hi3_highpart, mmx_pmaddwd,
  59         sse2_umulsidi3, mmx_iordi3, mmx_xordi3, mmx_anddi3, mmx_nanddi3,
  60         mmx_uavgv8qi3, mmx_uavgv4hi3, mmx_psadbw, mmx_pinsrw, mmx_pinsrw,
  61         mmx_pextrw, mmx_pshufw, eqv8qi3, eqv4hi3, eqv2si3, gtv8qi3, gtv4hi3,
  62         gtv2si3, umaxv8qi3, smaxv4hi3, uminv8qi3, sminv4hi3, ashrv4hi3,
  63         ashrv2si3, lshrv4hi3, lshrv2si3, mmx_lshrdi3, ashlv4hi3, ashlv2si3,
  64         mmx_ashldi3, mmx_packsswb, mmx_packssdw, mmx_packuswb, mmx_punpckhbw,
  65         mmx_punpckhwd, mmx_punpckhdq, mmx_punpcklbw, mmx_punpcklwd,
  66         mmx_punpckldq, emms, addv2sf3, subv2sf3, subrv2sf3, gtv2sf3, gev2sf3,
  67         eqv2sf3, pfmaxv2sf3, pfminv2sf3, mulv2sf3, femms, pf2id, pf2iw,
  68         pfacc, pfnacc, pfpnacc, pi2fw, floatv2si2, pfrcpv2sf2, pfrcpit1v2sf3,
  69         pfrcpit2v2sf3, pfrsqrtv2sf2, pfrsqit1v2sf3, pmulhrwv4hi3, pswapdv2si2,
  70         pswapdv2sf2): Move to mmx.md; rename as necessary with leading
  71         mmx_ prefix.
  72         (mmx_clrdi, pavgusb): Remove.
  73         (ldmxcsr, stmxcsr, sfence, sfence_insn): Move to sse.md; rename
  74         with leading sse_ prefix.
  75         * config/i386/sse.md: Receive them.
  76         * config/i386/mmx.md: New file.
  77         (MMXMODE12, MMXMODE24, mmxvecsize): New.
  78         (subrv2sf3): Turn into expander for normal subtraction.
  79         (mmx_addv2sf3, mmx_mulv2sf3, mmx_smaxv2sf3, mmx_sminv2sf3,
  80         mmx_eqv2sf3, mmx_mulv4hi3, mmx_smulv4hi3_highpart,
  81         mmx_umulv4hi3_highpart, mmx_pmaddwd, mmx_pmulhrwv4hi3, sse2_umulsidi3,
  82         mmx_umaxv8qi3, mmx_smaxv4hi3, mmx_uminv8qi3, mmx_sminv4hi3): Mark
  83         commutative; use ix86_binary_operator_ok.
  84         (mmx_add<MMXMODEI>3, mmx_ssadd<MMXMODE12>3, mmx_usadd<MMXMODE12>3,
  85         mmx_sub<MMXMODEI>3, mmx_sssub<MMXMODE12>3, mmx_ussub<MMXMODE12>3
  86         mmx_ashr<MMXMODE24>3, mmx_lshr<MMXMODE23>3, mmx_ashl<MMXMODE24>3
  87         mmx_eq<MMXMODEI>3, mmx_gt<MMXMODEI>3, mmx_and<MMXMODEI>3,
  88         mmx_nand<MMXMODEI>3, mmx_ior<MMXMODEI>3, mmx_xor<MMXMODEI>3):
  89         Macroize from existing patterns; use ix86_binary_operator_ok.
  90         (mmx_packsswb, mmx_packssdw, mmx_packuswb): Add memory alternative.
  91         (mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
  92         mmx_punpckhdq, mmx_punpckhdq, mmx_punpckldq): Likewise.  Model
  93         with vec_select+vec_concat.
  94         (mmx_pshufw, mmx_pshufw_1): Likewise.
  95         (mmx_uavgv8qi3): Merge pavgusb.  Model correcty.
  96         (mmx_uavgv4hi3): Model correctly.
  97         * config/i386/mmintrin.h (_mm_and_si64, _mm_andnot_si64, _mm_or_si64,
  98         _mm_xor_si64): Remove casts.
  99
 100 2005-01-09  Zdenek Dvorak  <dvorakz@suse.cz>
 101
 102         PR tree-optimization/19224
 103         * tree-scalar-evolution.c (get_instantiated_value,
 104         set_instantiated_value): New functions.
 105         (instantiate_parameters_1): Cache the results.
 106         (instantiate_parameters, resolve_mixers): Initialize and free
 107         the cache.
 108
 109 2005-01-08  David Edelsohn  <edelsohn@gnu.org>
 110
 111         * config/i386/i386.md (addhi_4): Correct reference in comment.
 112         (addqi_4): Same.
 113
 114 2005-01-08  Richard Henderson  <rth@redhat.com>
 115
 116         * config/i386/emmintrin.h (_mm_cvtsi128_si32): Move earlier.
 117         (_mm_cvtsi128_si64x): Likewise.
 118         (_mm_srl_epi64, _mm_srl_epi32, _mm_srl_epi16, _mm_sra_epi32,
 119         _mm_sra_epi16, _mm_sll_epi64, _mm_sll_epi32, _mm_sll_epi16): Use
 120         the _mm_{srl,sll}i_foo counterpart, and _mm_cvtsi128_si32.
 121         * config/i386/i386-modes.def: Add V16HI, V32QI, V4DF, V8SF.
 122         * config/i386/i386-protos.h: Update.
 123         * config/i386/i386.c (print_operand): Add 'H'.
 124         (ix86_fixup_binary_operands): Split out from ...
 125         (ix86_expand_binary_operator): ... here.
 126         (ix86_fixup_binary_operands_no_copy): New.
 127         (ix86_expand_fp_absneg_operator): Handle vector mode results.
 128         (bdesc_2arg): Update names for sse{,2,3}_ prefixes.
 129         (ix86_init_mmx_sse_builtins): Remove *maskncmp* special cases.
 130         (safe_vector_operand): Use CONST0_RTX.
 131         (ix86_expand_binop_builtin): Use ix86_fixup_binary_operands.
 132         (ix86_expand_builtin): Merge CODE_FOR_sse2_maskmovdqu_rex64 and
 133         CODE_FOR_sse2_maskmovdqu.  Special case SSE version of MASKMOVDQU
 134         expansion.  Update names for sse{,2,3}_ prefixes.  Remove *maskncmp*
 135         special cases.
 136         * config/i386/i386.h (IX86_BUILTIN_CMPNGTSS): New.
 137         (IX86_BUILTIN_CMPNGESS): New.
 138         * config/i386/i386.md (UNSPEC_FIX_NOTRUNC): New.
 139         (attr type): Add sselog1.
 140         (attr unit, attr memory): Handle it.
 141         (movti, movti_internal, movti_rex64): Move near other integer moves.
 142         (movtf, movtf_internal): Move near other fp moves.
 143         (SSEMODE, SSEMODEI, vec_setv2df, vec_extractv2df, vec_initv2df,
 144         vec_setv4sf, vec_extractv4sf, vec_initv4sf, movv4sf, movv4sf_internal,
 145         movv2df, movv2df_internal, mov<SSEMODEI>, mov<SSEMODEI>_internal,
 146         movmisalign<SSEMODE>, sse_movups_1, sse_movmskps, sse_movntv4sf,
 147         sse_movhlps, sse_movlhps, sse_storehps, sse_loadhps, sse_storelps,
 148         sse_loadlps, sse_loadss, sse_loadss_1, sse_movss, sse_storess,
 149         sse_shufps, addv4sf3, vmaddv4sf3, subv4sf3, vmsubv4sf3, negv4sf2,
 150         mulv4sf3, vmmulv4sf3, divv4sf3, vmdivv4sf3, rcpv4sf2, vmrcpv4sf2,
 151         rsqrtv4sf2, vmrsqrtv4sf2, sqrtv4sf2, vmsqrtv4sf2, sse_andv4sf3,
 152         sse_nandv4sf3, sse_iorv4sf3, sse_xorv4sf3, sse2_andv2df3,
 153         sse2_nandv2df3, sse2_iorv2df3, sse2_xorv2df3, sse2_andv2di3,
 154         sse2_nandv2di3, sse2_iorv2di3, sse2_xorv2di3, maskcmpv4sf3,
 155         vmmaskcmpv4sf3, sse_comi, sse_ucomi, sse_unpckhps, sse_unpcklps,
 156         smaxv4sf3, vmsmaxv4sf3, sminv4sf3, vmsminv4sf3, cvtpi2ps, cvtps2pi,
 157         cvttps2pi, cvtsi2ss, cvtsi2ssq, cvtss2si, cvtss2siq, cvttss2si,
 158         cvttss2siq, addv2df3, vmaddv2df3, subv2df3, vmsubv2df3, mulv2df3,
 159         vmmulv2df3, divv2df3, vmdivv2df3, smaxv2df3, vmsmaxv2df3, sminv2df3,
 160         vmsminv2df3, sqrtv2df2, vmsqrtv2df2, maskcmpv2df3, vmmaskcmpv2df3,
 161         sse2_comi, sse2_ucomi, sse2_movmskpd, sse2_pmovmskb, sse2_maskmovdqu,
 162         sse2_maskmovdqu_rex64, sse2_movntv2df, sse2_movntv2di, sse2_movntsi,
 163         cvtdq2ps, cvtps2dq, cvttps2dq, cvtdq2pd, cvtpd2dq, cvttpd2dq,
 164         cvtpd2pi, cvttpd2pi, cvtpi2pd, cvtsd2si, cvtsd2siq, cvttsd2si,
 165         cvttsd2siq, cvtsi2sd, cvtsi2sdq, cvtsd2ss, cvtss2sd, cvtpd2ps,
 166         cvtps2pd, addv16qi3, addv8hi3, addv4si3, addv2di3, ssaddv16qi3,
 167         ssaddv8hi3, usaddv16qi3, usaddv8hi3, subv16qi3, subv8hi3, subv4si3,
 168         subv2di3, sssubv16qi3, sssubv8hi3, ussubv16qi3, ussubv8hi3, mulv8hi3,
 169         smulv8hi3_highpart, umulv8hi3_highpart, sse2_umulsidi3,
 170         sse2_umulv2siv2di3, sse2_pmaddwd, sse2_uavgv16qi3, sse2_uavgv8hi3,
 171         sse2_psadbw, sse2_pinsrw, sse2_pextrw, sse2_pshufd, sse2_pshuflw,
 172         sse2_pshufhw, eqv16qi3, eqv8hi3, eqv4si3, gtv16qi3, gtv8hi3,
 173         gtv4si3, umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3, ashrv8hi3,
 174         ashrv4si3, lshrv8hi3, lshrv4si3, lshrv2di3, ashlv8hi3, ashlv4si3,
 175         ashlv2di3, sse2_ashlti3, sse2_lshrti3, sse2_unpckhpd, sse2_unpcklpd,
 176         sse2_packsswb, sse2_packssdw, sse2_packuswb, sse2_punpckhbw,
 177         sse2_punpckhwd, sse2_punpckhdq, sse2_punpcklbw, sse2_punpcklwd,
 178         sse2_punpckldq, sse2_punpcklqdq, sse2_punpckhqdq, sse2_movupd,
 179         sse2_movdqu, sse2_movdq2q, sse2_movdq2q_rex64, sse2_movq2dq,
 180         sse2_movq2dq_rex64, sse2_loadd, sse2_stored, sse2_storehpd,
 181         sse2_loadhpd, sse2_storelpd, sse2_loadlpd, sse2_movsd, sse2_loadsd,
 182         sse2_loadsd_1, sse2_storesd, sse2_shufpd, sse2_clflush, sse2_mfence,
 183         mfence_insn, sse2_lfence, lfence_insn, mwait, monitor, addsubv4sf3,
 184         addsubv2df3, haddv4sf3, haddv2df3, hsubv4sf3, hsubv2df3, movshdup,
 185         movsldup, lddqu, loadddup, movddup): Move to sse.md.  Any with
 186         non-optabs meanings renamed with an "sse{,2,3}_" prefix at the
 187         same time.
 188         (SSEPUSH, push<SSEPUSH>): Remove.
 189         (MMXPUSH, push<MMXPUSH>): Remove.
 190         (sse_movaps, sse_movaps_1, sse_movups): Remove.
 191         (sse2_movapd, sse2_movdqa, sse2_movq): Remove.
 192         (sse2_andti3, sse2_nandti3, sse2_iorti3, sse2_xorti3): Remove.
 193         (sse_clrv4sf, sse_clrv2df, sse2_clrti): Remove.
 194         (maskncmpv4sf3, vmmaskncmpv4sf3): Remove.
 195         (maskncmpv2df3, vmmaskncmpv2df3): Remove.
 196         (ashrv8hi3_ti, ashrv4si3_ti, lshrv8hi3_ti, lshrv4si3_ti): Remove.
 197         (lshrv2di3_ti, ashlv8hi3_ti, ashlv4si3_ti, ashlv2di3_ti): Remove.
 198         * config/i386/athlon.md (athlon_sselog_load): Handle sselog1.
 199         (athlon_sselog_load_k8, athlon_sselog, athlon_sselog_k8): Likewise.
 200         * config/i386/ppro.md (ppro_sse_div_V4SF_load): Fix memory attr.
 201         (ppro_sse_log_V4SF_load): Similarly.  Handle sselog1.
 202         (ppro_sse_log_V4SF): Handle sselog1.
 203         * config/i386/predicates.md (const_0_to_1_operand): New.
 204         (const_0_to_255_mul_8_operand): New.
 205         (const_1_to_31_operand): Rename from const_int_1_31_operand.
 206         (const_2_to_3_operand, const_4_to_7_operand): New.
 207         * config/i386/sse.md: New file.
 208         (SSEMODE12, SSEMODE24, SSEMODE124, SSEMODE248, ssevecsize): New.
 209         (sse_movups): Rename from sse_movups_1.
 210         (sse_loadlss): Rename from sse_loadss_1.
 211         (andv4sf3, iorv4sf3, xorv4sf3, andv2df3): Remove the sse prefix
 212         from the name.
 213         (negv4sf2): Use ix86_expand_fp_absneg_operator.
 214         (absv4sf2, negv2df, absv2df): New.
 215         (addv4sf3): Add expander to call ix86_fixup_binary_operands_no_copy.
 216         (subv4sf3, mulv4sf3, divv4sf3, smaxv4sf3, sminv4sf3, andv4sf3,
 217         iorv4sf3, xorv4sf3, addv2df3, subv2df3, mulv2df3, divv2df3,
 218         smaxv2df3, sminv2df3, andv2df3, iorv2df3, xorv2df3, mulv8hi3,
 219         umaxv16qi3, smaxv8hi3, uminv16qi3, sminv8hi3): Likewise.
 220         (sse3_addsubv4sf3): Model correctly.
 221         sse3_haddv4sf3, sse3_hsubv4sf3, sse3_addsubv2df3, sse3_haddv2df3,
 222         sse3_hsubv2df3, sse2_ashlti3, sse2_lshrti3): Likewise.
 223         (sse_movhlps): Model with vec_select+vec_concat.
 224         (sse_movlhps, sse_unpckhps, sse_unpcklps, sse3_movshdup,
 225         sse3_movsldup, sse_shufps, sse_shufps_1, sse2_unpckhpd, sse3_movddup,
 226         sse2_unpcklpd, sse2_shufpd, sse2_shufpd_1, sse2_punpckhbw,
 227         sse2_punpcklbw, sse2_punpckhwd, sse2_punpcklwd, sse2_punpckhdq,
 228         sse2_punpckldq, sse2_punpckhqdq, sse2_punpcklqdq, sse2_pshufd,
 229         sse2_pshufd_1, sse2_pshuflw, sse2_pshuflw_1, sse2_pshufhw,
 230         sse2_pshufhw_1): Likewise.
 231         (neg<SSEMODEI>2, one_cmpl<SSEMODEI>2): New.
 232         (add<SSEMODEI>3, sse2_ssadd<SSEMODE12>3, sse2_usadd<SSEMODE12>3,
 233         sub<SSEMODEI>3, sse2_sssub<SSEMODE12>3, sse2_ussub<SSEMODE12>3,
 234         ashr<SSEMODE24>3, lshr<SSEMODE248>3, sse2_eq<SSEMODE124>3,
 235         sse2_gt<SSEMODDE124>3, and<SSEMODEI>3, sse_nand<SSEMODEI>3,
 236         ior<SSEMODEI>3, xor<SSEMODEI>3): Macroize from existing patterns.
 237         (addv4sf3, sse_vmaddv4sf3, mulv4sf3, sse_vmmulv4sf3, smaxv4sf3,
 238         sse_vmsmaxv4sf3, sminv4sf3, sse_vmsminv4sf3, addv2df3, sse2_vmaddv2df3,
 239         mulv2df3, sse2_vmmulv2df3, smaxv2df3, sse2_vmsmaxv2df3, sminv2df3,
 240         sse2_vmsminv2df3, umaxv16qi3, smaxv8hi3, uminv16qi3
 241         sminv8hi3): Mark commutative
 242         operands.  Use ix86_binary_operator_ok.
 243         (sse_unpckhps, sse_unpcklps, sse2_packsswb, sse2_packssdw,
 244         sse2_packuswb, sse2_punpckhbw, sse2_punpcklbw, sse2_punpckhwd,
 245         sse2_punpcklwd, sse2_punpckhdq, sse2_punpckldq, sse2_punpckhqdq,
 246         sse2_punpcklqdq): Allow operand2 in memory.
 247         (sse_movhlps, sse_movlhps, sse2_unpckhpd, sse2_unpcklpd
 248         sse2_movsd): Add memory alternatives.
 249         (sse_storelps): Turn expander into an insn; split after reload.
 250         (sse_storess, sse2_loadhpd, sse2_loadlpd): Add non-xmm inputs.
 251         (sse2_storehpd, sse2_storelpd): Add non-xmm outputs.
 252
 253 2005-01-08  Eric Botcazou  <ebotcazou@libertysurf.fr>
 254
 255         * configure.ac (DWARF-2 debug_line): Use objdump.
 256         * configure: Regenerate.
 257
 258 2005-01-08  Jeff Law  <law@redhat.com>
 259             Diego Novillo  <dnovillo@redhat.com>
 260
 261         PR tree-optimization/18241
 262         * tree-nrv.c (tree_nrv): Ignore volatile return values.
 263         * tree-ssa-dse.c (dse_optimize_stmt): Do not optimize
 264         statements with volatile operands.
 265         * tree-ssa-operands.c (add_stmt_operand): Do add volatile
 266         operands after marking a statement with has_volatile_ops.
 267
 268 2005-01-08  Roger Sayle  <roger@eyesopen.com>
 269
 270         * tree.c (int_fits_type_p): Always honor integer constant
 271         TYPE_MIN_VALUE and TYPE_MAX_VALUE if they exist.
 272
 273 2005-01-08  Roger Sayle  <roger@eyesopen.com>
 274
 275         * ifcvt.c (find_if_case_1): Reinstate 2005-01-04 change, now that
 276         the latent bug in rtl_delete_block has been resolved.
 277
 278 2005-01-08  Richard Sandiford  <rsandifo@redhat.com>
 279
 280         * config/mips/t-iris6 (MULTILIB_DIRNAMES): Use -mabi argument values.
 281         (MULTILIB_OSDIRNAMES): Use the standard lib, lib32 and lib64.
 282
 283 2005-01-08  Richard Sandiford  <rsandifo@redhat.com>
 284
 285         * config/mips/t-slibgcc-irix (SHLIB_LINK): Install a copy of the
 286         library as @multilib_dir@/$(SHLIB_SONAME).
 287
 288 2005-01-07  Eric Botcazou  <ebotcazou@libertysurf.fr>
 289
 290         * configure.ac (HAVE_AS_OFFSETABLE_LO10): Fix typo.
 291         * configure: Regenerate.
 292
 293 2005-01-07  Jakub Jelinek  <jakub@redhat.com>
 294
 295         * c-common.c (handle_mode_attribute): For ENUMERAL_TYPE, also copy
 296         TYPE_MODE.
 297
 298 2005-01-07  David Edelsohn  <edelsohn@gnu.org>
 299
 300         PR target/13674
 301         * config/rs6000/rs6000.c (rs6000_legitimize_reload_address):
 302         Convert non-word aligned offset address using ld/std into
 303         indirect address.
 304
 305 2005-01-07  Richard Henderson  <rth@redhat.com>
 306
 307         * config/i386/i386.md (sse_loadhps splitter): Fix operand number typo.
 308
 309 2005-01-07  Richard SAndiford  <rsandifo@redhat.com>
 310
 311         * tree.h (IS_EXPR_CODE_CLASS): Use a straight-forward range check.
 312
 313 2005-01-07  Jakub Jelinek  <jakub@redhat.com>
 314
 315         PR tree-optimization/19283
 316         * fold-const.c (fold_widened_comparison): Return NULL if shorter_type
 317         is not shorter than the original type.
 318
 319         PR rtl-optimization/19012
 320         * config/i386/i386.md (addqi_1_slp): Set memory attribute.
 321
 322         PR rtl-optimization/18861
 323         * cfgbuild.c (BLOCK_USED_BY_TABLEJUMP): Define.
 324         (FULL_STATE): Define.
 325         (mark_tablejump_edge): New function.
 326         (purge_dead_tablejump_edges): New function.
 327         (find_bb_boundaries): Use it.
 328
 329         PR tree-optimization/18828
 330         * builtins.c (expand_builtin_next_arg): Remove argument and all
 331         the argument checking.
 332         (expand_builtin): Adjust caller.
 333         (expand_builtin_va_start): Likewise.  Remove error for too many
 334         arguments.
 335         (fold_builtin_next_arg): Issue error for too many arguments.
 336         After checking arguments, replace them with magic arguments that
 337         prevent further checking of the args.
 338
 339 2005-01-06  John David Anglin  <dave.anglin@nrc-cnrc.gc.ca>
 340
 341         * pa64-hpux.h (STARTFILE_SPEC): Fix typo in spec.
 342
 343 2005-01-06  Roger Sayle  <roger@eyesopen.com>
 344
 345         PR target/6077
 346         * gcc.c (input_suffix_matches): Tweak the semantics of %{.s:...}
 347         and %{.S:...} (and their negative variants) to test whether the
 348         input file is assembler or pre-processed-assembler independent of
 349         the actual filename extension.
 350
 351 2005-01-06  Roger Sayle  <roger@eyesopen.com>
 352
 353         * simplify-rtx.c (simplify_subreg): Simplify truncations of shifts
 354         of sign or zero extended values.
 355
 356 2005-01-06  Geoffrey Keating  <geoffk@apple.com>
 357
 358         * c-cppbuiltin.c (builtin_define_float_constants): Set __*_EPSILON__
 359         for IBM long double format correctly.
 360
 361 2005-01-06  Daniel Berlin <dberlin@dberlin.org>
 362
 363         Fix PR tree-optimization/18792
 364
 365         * tree-data-ref.c (build_classic_dist_vector): Change first_loop
 366         to first_loop_depth, and use loop depth instead of loop number.
 367         (build_classic_dir_vector): Ditto.
 368         (compute_data_dependences_for_loop): Use depth, not loop number.
 369         * tree-loop-linear.c (try_interchange_loops): Use loop depth, not loop
 370         number. Pass in loops, instead of loop numbers.
 371         (gather_interchange_stats): Ditto.
 372         (linear_transform_loops): Ditto.
 373
 374 2005-01-06  Richard Sandiford  <rsandifo@redhat.com>
 375
 376         PR rtl-opt/13299
 377         * loop.c (get_monotonic_increment, biased_biv_fits_mode_p,
 378         biv_fits_mode_p, extension_within_bounds_p): New functions.
 379         (check_ext_dependent_givs): Use them.
 380
 381 2005-01-06  Roger Sayle  <roger@eyesopen.com>
 382
 383         * cfgrtl.c (rtl_delete_block): A basic block may be followed by
 384         more than one barrier, in which case we should delete them all.
 385
 386 2005-01-06  Tobias Schlueter  <tobias.schlueter@physik.uni-muenchen.de>
 387
 388         * gcc.c (process_command): Change year in 'gcc --version' to 2005.
 389
 390 2005-01-05  Daniel Berlin  <dberlin@dberlin.org>
 391
 392         Fix PR middle-end/19286
 393         Fix PR debug/19267
 394         * dwarf2out.c (gen_subprogram_die): If we've already tried to
 395         output this subprogram, simply ignore this attempt to do it again.
 396         (add_abstract_origin_attribute): Don't abort trying to add the abstract
 397         origin attribute if it's not possible.
 398         (gen_block_die): Don't ignore subblocks of "unused" blocks.
 399         (decls_for_scope): Ditto.
 400         * gimple-low.c (mark_blocks_with_used_subblocks): Remove.
 401         (mark_used_blocks): Don't call mark_blocks_with_used_subblocks.
 402
 403 2005-01-05  Richard Henderson  <rth@redhat.com>
 404
 405         PR target/11327
 406         * config/i386/i386.c (BUILTIN_DESC_SWAP_OPERANDS): New.
 407         (bdesc_2arg): Use it.
 408         (ix86_expand_binop_builtin): Force operands into registers
 409         when optimizing.
 410         (ix86_expand_unop_builtin, ix86_expand_unop1_builtin,
 411         ix86_expand_sse_compare, ix86_expand_sse_comi,
 412         ix86_expand_builtin): Likewise.
 413
 414 2005-01-05  Richard Henderson  <rth@redhat.com>
 415
 416         * config/ia64/ia64.c (rtx_needs_barrier): Handle CONST_VECTOR
 417         and VEC_SELECT.
 418         * config/ia64/vect.md (mulv8qi3): Re-implement with mix_[rl].
 419         (mulv4hi3): Set itanium_class mmmul.
 420         (fpack_sfxf, fpack_xfsf, fpack_xfxf): Remove.
 421         (fpack): Rename from fpack_sfsf.
 422
 423 2005-01-05  Richard Henderson  <rth@redhat.com>
 424
 425         PR rtl-opt/10692
 426         * reload1.c (do_input_reload): Restrict the optimization deleteing
 427         a previous output reload to RELOAD_FOR_INPUT.
 428
 429 2005-01-05  Steven Bosscher  <stevenb@suse.de>
 430
 431         * combine.c (expand_compound_operation)  <ZERO_EXTRACT>: Add
 432         comment that we fall through after case.
 433         (mark_used_regs_combine): Don't expect a SIGN_EXTRACT in a SET_DEST.
 434         (distribute_links): Likewise.
 435         * cse.c (cse_insn): Likewise.
 436         * cselib.c (cselib_invalidate_mem): Likewise.
 437         * df.c: Update comments at the top of the file.
 438         (read_modify_subreg_p): Update comments here too.
 439         (df_def_record_1): Don't expect a SIGN_EXTRACT in a SET_DEST.
 440         * flow.c (mark_set_1): Likewise.
 441         (mark_used_regs): Likewise.
 442         * gcse.c (mems_conflict_for_gcse_p): Likewise.
 443         (canon_list_insert): Likewise.
 444         (mark_set): Likewise.
 445         (try_replace_reg): Likewise.
 446         (store_killed_in_insn): Likewise.
 447         * loop.c (count_one_set): Likewise.
 448         (basic_induction_var): Likewise.
 449         * postreload-gcse.c (find_mem_conflicts): Likewise.
 450         * postreload.c (reload_combine_note_store): Likewise.
 451         (move2add_note_store): Likewise.
 452         * reload.c (find_equiv_reg): Likewise.
 453         (mark_referenced_resources): Likewise.
 454         * rtlanal.c (set_noop_p): Likewise.
 455         (note_stores): Likewise.
 456         (note_uses): Likewise.
 457         * sched-deps.c (sched_analyze_1): Likewise.
 458         * sched-rgn.c (check_live_1): Likewise.
 459         (update_live_1): Likewise.
 460         * config/i860/i860.c: Likewise.
 461
 462         * rtl.dec (SIGN_EXTRACT): Document that this cannot appear as
 463         an lvalue.
 464         (ZERO_EXTRACT): Mention that this one can be an lvalue.
 465
 466         * doc/rtl.texi: Update documentation for bit-fields and SET.
 467
 468 2005-01-05  Roger Sayle  <roger@eyesopen.com>
 469
 470         * ifcvt.c (find_if_case_1): Revert 2005-01-04 change.
 471
 472 2005-01-05  Jan Hubicka  <jh@suse.cz>
 473             Richard Henderson  <rth@redhat.com>
 474
 475         PR target/18910
 476         * config/i386/i386.c (ix86_expand_move): Handle tls symbols
 477         with an offset.
 478
 479 2005-01-05  Richard Henderson  <rth@redhat.com>
 480
 481         PR target/12902
 482         * config/i386/i386.md (sse_movhps, sse_movlps): Remove.
 483         (sse_shufps): Change operand 3 to const_int_operand.
 484         (sse2_storelps): Fix typo in template.
 485         (sse_storehps, sse_loadhps, sse_storelps, sse_loadlps): New.
 486         * config/i386/i386.c (ix86_expand_vector_move_misalign): Use them.
 487         (ix86_expand_builtin): Likewise.
 488
 489 2005-01-05  Julian Brown  <julian@codesourcery.com>
 490
 491         * config/arm/arm.c (arm_return_in_memory): Treat complex types
 492         as aggregates for AAPCS ABIs.
 493
 494 2005-01-05  Stan Shebs  <shebs@apple.com>
 495
 496         * unwind-dw2-fde-darwin.c (examine_objects): Use 64-bit
 497         Mach-O getters if ppc64.
 498         * config/darwin.c (darwin_asm_output_dwarf_delta): Obey
 499         size argument.
 500         * config/darwin.h (ASM_PREFERRED_EH_DATA_FORMAT): Use
 501         signed four-byte field for global code case.
 502         (STARTFILE_SPEC): Avoid crt2.o for 64-bit compilation.
 503
 504 2005-01-05  Roger Sayle  <roger@eyesopen.com>
 505
 506         PR middle-end/19100
 507         * c-common.c: Include real.h.
 508         (c_common_truthvalue_conversion): Avoid destructively modifying expr.
 509         Correctly handle TREE_CONSTANT_OVERFLOW for INTEGER_CST.
 510         Correctly handle TREE_CONSTANT_OVERFLOW and NaNs for REAL_CST.
 511         * Makefile.in (c-common.o): Update dependencies.
 512
 513 2005-01-05  Joseph S. Myers  <joseph@codesourcery.com>
 514
 515         * c-parse.in (asm_string): Add trailing semicolon.
 516
 517 2005-01-05  Joseph S. Myers  <joseph@codesourcery.com>
 518
 519         * c-parse.in (asm_string): New.  Don't allow wide strings in
 520         'asm'.
 521         (simple_asm_expr, asm_argument, asm_operand, asm_clobbers): Use
 522         asm_string instead of STRING.
 523
 524 2005-01-05  Joseph S. Myers  <joseph@codesourcery.com>
 525
 526         * c-typeck.c (constructor_no_implicit): Remove.
 527         (set_designator, process_init_element): Don't check
 528         constructor_no_implicit.
 529
 530 2005-01-05  J"orn Rennecke <joern.rennecke@st.com>
 531             Kaz Kojima  <kkojima@gcc.gnu.org>
 532
 533         PR target/16482
 534         * lcm.c (create_pre_exit): New.
 535         (optimize_mode_switching): In MODE_ENTRY / MODE_EXIT case, set
 536         ENTRY_EXIT_EXTRA to 3.  Use create_pre_exit.
 537
 538 2004-01-05  Richard Earnshaw  <rearnsha@arm.com>
 539
 540         * arm.h (TARGET_OPTIONS): Correctly record -mhard-float and
 541         -msoft-float in target_float_switch.
 542         * arm.c (arm_override_options): Fix processing of target_float_switch.
 543
 544 2004-01-05  Richard Earnshaw  <rearnsha@arm.com>
 545
 546         * arm/vfp.md (arm_movsi_vfp): Hide VFP register classes from register
 547         preferencing.
 548
 549 2004-01-05  Uros Bizjak  <uros@kss-loka.si>
 550
 551         * doc/invoke.texi (Intel 386 and AMD x86-64 Options):
 552         Replace i387 with 'i386 compiler' in -mfpmath=sse option.
 553
 554 2005-01-04  Roger Sayle  <roger@eyesopen.com>
 555
 556         * ifcvt.c (find_if_case_1): Avoid creating an empty forwarder block,
 557         if deleting the then-block allows the test-block to fallthru to the
 558         else-block.
 559
 560 2005-01-04  Andrew Pinski  <pinskia@physics.uc.edu>
 561
 562         PR c/19152
 563         * c-decl.c (diagnose_mismatched_decls): Accept "extern inline" declared
 564         after the full declaration if the are in two different TUs.
 565
 566 2005-01-04  Richard Henderson  <rth@redhat.com>
 567
 568         PR tree-opt/19158
 569         * tree-sra.c (generate_one_element_init): Just
 570         call gimplify_and_add.
 571         (generate_element_init): Record the
 572         new referenced variables and mark them for renaming
 573         and split out to ...
 574         (generate_element_init_1): This.
 575         (scalarize_init): Don't call push_gimplify_context/
 576         pop_gimplify_context.
 577
 578 2005-01-04  Geoffrey Keating  <geoffk@apple.com>
 579
 580         * toplev.c (get_src_pwd): Handle failure of getpwd().
 581
 582 2005-01-04  Roger Sayle  <roger@eyesopen.com>
 583
 584         * fold-const.c (fold_single_bit_test): Delete unreachable handling
 585         of TRUTH_NOT_EXPR.
 586         (fold): Don't call fold_single_bit_test with a TRUTH_NOT_EXPR, as
 587         all the cases handled by it are inverted by invert_truthvalue.
 588
 589 2005-01-04  Eric Botcazou  <ebotcazou@libertysurf.fr>
 590
 591         * config/sparc/sparc.md (save_register_windowdi): Add missing mode.
 592         (save_register_windowsi): Likewise.
 593
 594 2005-01-04  Richard Henderson  <rth@redhat.com>
 595
 596         * tree-mudflap.c (mf_varname_tree): Fix thinko setting declname.
 597
 598 2005-01-04  Uros Bizjak  <uros@kss-loka.si>
 599
 600         PR middle-end/17767
 601         * cse.c (fold_rtx) [RTX_COMPARE, RTX_COMM_COMPARE]: Don't attempt
 602         any simplifications of vector mode comparison operators.
 603         * simplify-rtx.c (simplify_relational_operation): Fix variable name.
 604
 605 2005-01-04  Paolo Bonzini  <bonzini@gnu.org>
 606             Devang Patel  <dpatel@apple.com>
 607
 608         PR tree-optimization/18308
 609         * tree-if-conv.c (add_to_dst_predicate_list): Gimplify
 610         the operands before creating a new expression.
 611         * dojump.c (do_jump): Make drop_through_label available
 612         for all cases.  Add expansion of COND_EXPR.
 613
 614 2005-01-04  Ira Rosen  <irar@il.ibm.com>
 615
 616         * tree-vectorizer.c (vect_analyze_offset_expr): Test for
 617         INTEGER_CST instead of TREE_CONSTANT.
 618         (vect_gen_niters_for_prolog_loop): Test for INTEGER_CST
 619         instead of TREE_CONSTANT.
 620         (vect_analyze_pointer_ref_access): Test for INTEGER_CST
 621         instead of TREE_CONSTANT.
 622
 623 2005-01-04  Eric Botcazou  <ebotcazou@libertysurf.fr>
 624
 625         * tree-eh.c (replace_goto_queue): Return early if the queue is empty.
 626
 627 2005-01-04  Uros Bizjak  <uros@kss-loka.si>
 628
 629         PR target/19240
 630         * config/i386/i386.md (*fop_df_1_i387): Disable for TARGET_SSE_MATH.
 631         (*fop_df_1_i387): Disable for (TARGET_SSE2 && TARGET_SSE_MATH).
 632
 633 2005-01-03  Richard Henderson  <rth@redhat.com>
 634
 635         * fold-const.c (force_fit_type): Cope with types larger than 2 HWI.
 636         (fold_convert_const_int_from_int, fold_convert_const_int_from_real,
 637         fold_convert_const_real_from_real): Split out from ...
 638         (fold_convert_const): ... here.
 639
 640 2005-01-03  Richard Henderson  <rth@redhat.com>
 641
 642         PR target/19235
 643         * config/i386/i386.md (movdi_2): Separate SSE1 and SSE2 alternatives.
 644         (mov<MMXMODEI>_internal): Likewise.
 645         (movdf_nointeger): Prefer Y while not preferring, but allowing, x.
 646         Add V2SF case; use it for SSE1; don't use TI.
 647         (movdf_integer): Likewise.
 648         (mov<SSEMODEI>_internal, movti_internal): Force V4SF for SSE1.
 649
 650 2005-01-03  Ira Rosen  <irar@il.ibm.com>
 651
 652         * tree-vectorizer.c (vect_strip_conversions): New function.
 653         (vect_analyze_offset_expr): Call vect_strip_conversions. Add
 654         check for binary class.
 655
 656 2005-01-03  Daniel Berlin  <dberlin@dberlin.org>
 657
 658         Fix PR debug/17924
 659         Fix PR debug/19191
 660         * dwarf2out.c (block_ultimate_origin): Follow decl origin if origin
 661         is a decl.
 662         * gimple-low.c (mark_blocks_with_used_vars): New function.
 663         (mark_blocks_with_used_subblocks): Ditto.
 664         (mark_used_blocks): Ditto.
 665         (pass_mark_used_blocks): New pass.
 666         * tree-inline.c: Include debug.h.
 667         (expand_call_inline): Call outlining_inline_function here.
 668         * tree-optimize.c (init_tree_optimization_passes): Add
 669         pass_mark_used_blocks.
 670         * tree-pass.h (pass_mark_used_blocks): New.
 671         * Makefile.in (tree-inline.o): Add debug.h dependency.
 672
 673 2005-01-03  Geoffrey Keating  <geoffk@apple.com>
 674
 675         * config/darwin.c (darwin_handle_weak_import_attribute): Permit
 676         VAR_DECLs to have weak_import attribute.
 677
 678         * config/rs6000/darwin-fallback.c: Use 'ucontext_t' rather than
 679         'struct ucontext'.
 680
 681 2004-01-03  Steven Bosscher  <stevenb@suse.de>
 682
 683         * passes.c (rest_of_compilation): Don't run regmove if only
 684         flag_expensive_optimizations, require flag_regmove instead.
 685
 686 2005-01-03  Eric Botcazou  <ebotcazou@libertysurf.fr>
 687
 688         * config/sparc/litecoff.h (TARGET_OS_CPP_BUILTINS): Do not
 689         register "sparc".
 690         * config/sparc/sysv4-only.h (TARGET_OS_CPP_BUILTINS): Likewise.
 691         * config/sparc/rtemself.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
 692         * config/sparc/sol2-64.h (TARGET_SUB_OS_CPP_BUILTINS): Delete.
 693         * config/sparc/sp64-elf.h (TARGET_SUB_OS_CPP_BUILTINS): Likewise.
 694
 695 2005-01-03  Richard Henderson  <rth@redhat.com>
 696
 697         * config/ia64/ia64.c (TARGET_VECTOR_MODE_SUPPORTED_P): New.
 698         (ia64_const_ok_for_letter_p): New.
 699         (ia64_const_double_ok_for_letter_p): New.
 700         (ia64_extra_constraint): New.
 701         (ia64_expand_vecint_compare): New.
 702         (ia64_expand_vcondu_v2si): New.
 703         (ia64_expand_vecint_cmov): New.
 704         (ia64_expand_vecint_minmax): New.
 705         (ia64_print_operand): Add 'v'.
 706         (ia64_preferred_reload_class): New.
 707         (ia64_vector_mode_supported_p): New.
 708         * config/ia64/ia64.h (UNITS_PER_SIMD_WORD): New.
 709         (PREFERRED_RELOAD_CLASS): Move to function.
 710         (CONST_OK_FOR_LETTER_P): Move to function.
 711         (CONST_DOUBLE_OK_FOR_LETTER_P): Move to function.
 712         (CONSTRAINT_OK_FOR_Q, CONSTRAINT_OK_FOR_R): Remove.
 713         (CONSTRAINT_OK_FOR_S, CONSTRAINT_OK_FOR_T): Remove.
 714         (EXTRA_CONSTRAINT): Move to function.
 715         * config/ia64/ia64.md: Include vect.md.
 716         (itanium_class): Add mmalua.
 717         (type): Handle it.
 718         * config/ia64/itanium1.md (1_mmalua): New.  Add it to bypasses.
 719         (1b_mmalua): New.
 720         * config/ia64/itanium2.md (2_mmalua, 2b_mmalua): Similarly.
 721         * config/ia64/predicates.md (gr_reg_or_0_operand): Accept any
 722         CONST0_RTX.
 723         (const_int_2bit_operand): New.
 724         (fr_reg_or_0_operand): New.
 725         * config/ia64/ia64-modes.def: Add vector modes.
 726         * config/ia64/ia64-protos.h: Update.
 727         * config/ia64/vect.md: New file.
 728
 729 2005-01-03  Richard Henderson  <rth@redhat.com>
 730
 731         * simplify-rtx.c (simplify_binary_operation): Handle VEC_CONCAT.
 732
 733 2005-01-03  Uros Bizjak  <uros@kss-loka.si>
 734
 735         PR target/19236
 736         * config/i386/i386.c (log1psf2): Change mode of operands[0,1]
 737         to SFmode.
 738         (log1pdf2): Change mode of operands[0,1] to DFmode.
 739
 740 2005-01-03  Eric Botcazou  <ebotcazou@libertysurf.fr>
 741
 742         * config/sparc/sparc.h (SPARC_RELAXED_ORDERING): Define to false.
 743         * config/sparc/linux.h (SPARC_RELAXED_ORDERING): Define to true.
 744         * config/sparc/linux64.h (SPARC_RELAXED_ORDERING): Likewise.
 745         * config/sparc/sparc.c (TARGET_RELAXED_ORDERING): Define to
 746         SPARC_RELAXED_ORDERING.
 747
 748 2005-01-03  Richard Henderson  <rth@redhat.com>
 749             Uros Bizjak  <uros@kss-loka.si>
 750
 751         PR target/14631
 752         * config/i386/i386.c (ix86_expand_builtin): [IX86_BUILTIN_PINSRW,
 753         IX86_BUILTIN_PINSRW128]: Fix wrong selector range in error message.
 754         * config/i386/i386.md (mmx_pinsrw, sse2_pinsrw): Fix selector
 755         handling.
 756         (*mmx_pinsrw, *sse2_pinsrw): New patterns.
 757         * config/i386/i386/predicates.md (const_pow2_1_to_8_operand,
 758         const_pow2_1_to_128_operand): New predicates.
 759
 760 2005-01-02  Greg McGary  <greg@mcgary.org>
 761
 762         * tree-mudflap.c (mf_varname_tree): decl_printable_name handles
 763         anonymous decls.
 764         (mf_xform_derefs_1) [RESULT_DECL, STRING_CST]: Handle as innermost
 765         object.  [ptr_type]: Remove unused variable.
 766         Remember nearest addressable array-element or record-component
 767         when checking bitfield components.  Tolerate empty BIND_EXPR.
 768
 769 2005-01-01  Richard Henderson  <rth@redhat.com>
 770
 771         * tree-vectorizer.c (vect_analyze_offset_expr): Strip conversions
 772         that don't narrow the value.  Fail for other conversions.
 773
 774 2005-01-01  Richard Henderson  <rth@redhat.com>
 775
 776         PR c/19031
 777         * c-decl.c (pop_file_scope): Call maybe_apply_pending_pragma_weaks.
 778         * c-lang.c (finish_file): Don't do it here.
 779         * objc/objc-act.c (objc_finish_file): Likewise.
 780
 781         * cgraph.c (decl_assembler_name_equal): New.
 782         (cgraph_node_for_asm, cgraph_varpool_node_for_asm): New.
 783         (cgraph_varpool_node): Actually link up cgraph_varpool_nodes.
 784         * cgraph.h (struct cgraph_varpool_node): Add next.
 785         (cgraph_node_for_asm, cgraph_varpool_node_for_asm): Declare.
 786         * varasm.c (assemble_alias): Mark the target as needed.
 787
 788 2005-01-01  Andrew Pinski  <pinskia@physics.uc.edu>
 789
 790         PR middle-end/19221
 791         * function.c (get_arg_pointer_save_area): Use entry_of_function
 792         instead of get_insns.
 793
 794 2005-01-01  Roger Sayle  <roger@eyesopen.com>
 795             Andrew Pinski  <pinskia@physics.uc.edu>
 796             James E. Wilson  <wilson@specifixinc.com>
 797
 798         PR rtl-optimization/12092
 799         * loop.c (emit_prefetch_instructions): Do nothing if PREFETCH_BLOCK
 800         is zero.
 801
 802 2005-01-01  Roger Sayle  <roger@eyesopen.com>
 803             Olivier Hainque  <hainque@act-europe.fr>
 804
 805         * tree.c (int_fits_type_p): A narrower type always fits in a
 806         wider one, except for negative values into unsigned types.
 807
 808 2005-01-01  Roger Sayle  <roger@eyesopen.com>
 809
 810         * tree.c (int_fits_type_p): Compare the result of force_fit_type
 811         with the original constant rather than require TREE_OVERFLOW.
 812
 813 2005-01-01  Steven Bosscher  <stevenb@suse.de>
 814
 815         PR middle-end/17544
 816         * c-decl.c (finish_function): If compiling C99, annotate the
 817         compiler generated return with the current file name and line 0.
 818         * tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
 819         the source line is greater than 0.
 820         (remove_bb): Likewise.
 821
 822 See ChangeLog.12 for earlier changes.