target/arm/iwmmxt_helper.c

   1 /*
   2  * iwMMXt micro operations for XScale.
   3  *
   4  * Copyright (c) 2007 OpenedHand, Ltd.
   5  * Written by Andrzej Zaborowski <andrew@openedhand.com>
   6  * Copyright (c) 2008 CodeSourcery
   7  *
   8  * This library is free software; you can redistribute it and/or
   9  * modify it under the terms of the GNU Lesser General Public
  10  * License as published by the Free Software Foundation; either
  11  * version 2.1 of the License, or (at your option) any later version.
  12  *
  13  * This library is distributed in the hope that it will be useful,
  14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  16  * Lesser General Public License for more details.
  17  *
  18  * You should have received a copy of the GNU Lesser General Public
  19  * License along with this library; if not, see <http://www.gnu.org/licenses/>.
  20  */
  21
  22 #include "qemu/osdep.h"
  23
  24 #include "cpu.h"
  25 #include "exec/helper-proto.h"
  26
  27 /* iwMMXt macros extracted from GNU gdb.  */
  28
  29 /* Set the SIMD wCASF flags for 8, 16, 32 or 64-bit operations.  */
  30 #define SIMD8_SET(v, n, b)      ((v != 0) << ((((b) + 1) * 4) + (n)))
  31 #define SIMD16_SET(v, n, h)     ((v != 0) << ((((h) + 1) * 8) + (n)))
  32 #define SIMD32_SET(v, n, w)     ((v != 0) << ((((w) + 1) * 16) + (n)))
  33 #define SIMD64_SET(v, n)        ((v != 0) << (32 + (n)))
  34 /* Flags to pass as "n" above.  */
  35 #define SIMD_NBIT       -1
  36 #define SIMD_ZBIT       -2
  37 #define SIMD_CBIT       -3
  38 #define SIMD_VBIT       -4
  39 /* Various status bit macros.  */
  40 #define NBIT8(x)        ((x) & 0x80)
  41 #define NBIT16(x)       ((x) & 0x8000)
  42 #define NBIT32(x)       ((x) & 0x80000000)
  43 #define NBIT64(x)       ((x) & 0x8000000000000000ULL)
  44 #define ZBIT8(x)        (((x) & 0xff) == 0)
  45 #define ZBIT16(x)       (((x) & 0xffff) == 0)
  46 #define ZBIT32(x)       (((x) & 0xffffffff) == 0)
  47 #define ZBIT64(x)       (x == 0)
  48 /* Sign extension macros.  */
  49 #define EXTEND8H(a)     ((uint16_t) (int8_t) (a))
  50 #define EXTEND8(a)      ((uint32_t) (int8_t) (a))
  51 #define EXTEND16(a)     ((uint32_t) (int16_t) (a))
  52 #define EXTEND16S(a)    ((int32_t) (int16_t) (a))
  53 #define EXTEND32(a)     ((uint64_t) (int32_t) (a))
  54
  55 uint64_t HELPER(iwmmxt_maddsq)(uint64_t a, uint64_t b)
  56 {
  57     a = ((
  58             EXTEND16S((a >> 0) & 0xffff) * EXTEND16S((b >> 0) & 0xffff) +
  59             EXTEND16S((a >> 16) & 0xffff) * EXTEND16S((b >> 16) & 0xffff)
  60         ) & 0xffffffff) | ((uint64_t) (
  61             EXTEND16S((a >> 32) & 0xffff) * EXTEND16S((b >> 32) & 0xffff) +
  62             EXTEND16S((a >> 48) & 0xffff) * EXTEND16S((b >> 48) & 0xffff)
  63         ) << 32);
  64     return a;
  65 }
  66
  67 uint64_t HELPER(iwmmxt_madduq)(uint64_t a, uint64_t b)
  68 {
  69     a = ((
  70             ((a >> 0) & 0xffff) * ((b >> 0) & 0xffff) +
  71             ((a >> 16) & 0xffff) * ((b >> 16) & 0xffff)
  72         ) & 0xffffffff) | ((
  73             ((a >> 32) & 0xffff) * ((b >> 32) & 0xffff) +
  74             ((a >> 48) & 0xffff) * ((b >> 48) & 0xffff)
  75         ) << 32);
  76     return a;
  77 }
  78
  79 uint64_t HELPER(iwmmxt_sadb)(uint64_t a, uint64_t b)
  80 {
  81 #define abs(x) (((x) >= 0) ? x : -x)
  82 #define SADB(SHR) abs((int) ((a >> SHR) & 0xff) - (int) ((b >> SHR) & 0xff))
  83     return
  84         SADB(0) + SADB(8) + SADB(16) + SADB(24) +
  85         SADB(32) + SADB(40) + SADB(48) + SADB(56);
  86 #undef SADB
  87 }
  88
  89 uint64_t HELPER(iwmmxt_sadw)(uint64_t a, uint64_t b)
  90 {
  91 #define SADW(SHR) \
  92     abs((int) ((a >> SHR) & 0xffff) - (int) ((b >> SHR) & 0xffff))
  93     return SADW(0) + SADW(16) + SADW(32) + SADW(48);
  94 #undef SADW
  95 }
  96
  97 uint64_t HELPER(iwmmxt_mulslw)(uint64_t a, uint64_t b)
  98 {
  99 #define MULS(SHR) ((uint64_t) ((( \
 100         EXTEND16S((a >> SHR) & 0xffff) * EXTEND16S((b >> SHR) & 0xffff) \
 101     ) >> 0) & 0xffff) << SHR)
 102     return MULS(0) | MULS(16) | MULS(32) | MULS(48);
 103 #undef MULS
 104 }
 105
 106 uint64_t HELPER(iwmmxt_mulshw)(uint64_t a, uint64_t b)
 107 {
 108 #define MULS(SHR) ((uint64_t) ((( \
 109         EXTEND16S((a >> SHR) & 0xffff) * EXTEND16S((b >> SHR) & 0xffff) \
 110     ) >> 16) & 0xffff) << SHR)
 111     return MULS(0) | MULS(16) | MULS(32) | MULS(48);
 112 #undef MULS
 113 }
 114
 115 uint64_t HELPER(iwmmxt_mululw)(uint64_t a, uint64_t b)
 116 {
 117 #define MULU(SHR) ((uint64_t) ((( \
 118         ((a >> SHR) & 0xffff) * ((b >> SHR) & 0xffff) \
 119     ) >> 0) & 0xffff) << SHR)
 120     return MULU(0) | MULU(16) | MULU(32) | MULU(48);
 121 #undef MULU
 122 }
 123
 124 uint64_t HELPER(iwmmxt_muluhw)(uint64_t a, uint64_t b)
 125 {
 126 #define MULU(SHR) ((uint64_t) ((( \
 127         ((a >> SHR) & 0xffff) * ((b >> SHR) & 0xffff) \
 128     ) >> 16) & 0xffff) << SHR)
 129     return MULU(0) | MULU(16) | MULU(32) | MULU(48);
 130 #undef MULU
 131 }
 132
 133 uint64_t HELPER(iwmmxt_macsw)(uint64_t a, uint64_t b)
 134 {
 135 #define MACS(SHR) ( \
 136         EXTEND16((a >> SHR) & 0xffff) * EXTEND16S((b >> SHR) & 0xffff))
 137     return (int64_t) (MACS(0) + MACS(16) + MACS(32) + MACS(48));
 138 #undef MACS
 139 }
 140
 141 uint64_t HELPER(iwmmxt_macuw)(uint64_t a, uint64_t b)
 142 {
 143 #define MACU(SHR) ( \
 144         (uint32_t) ((a >> SHR) & 0xffff) * \
 145         (uint32_t) ((b >> SHR) & 0xffff))
 146     return MACU(0) + MACU(16) + MACU(32) + MACU(48);
 147 #undef MACU
 148 }
 149
 150 #define NZBIT8(x, i) \
 151     SIMD8_SET(NBIT8((x) & 0xff), SIMD_NBIT, i) | \
 152     SIMD8_SET(ZBIT8((x) & 0xff), SIMD_ZBIT, i)
 153 #define NZBIT16(x, i) \
 154     SIMD16_SET(NBIT16((x) & 0xffff), SIMD_NBIT, i) | \
 155     SIMD16_SET(ZBIT16((x) & 0xffff), SIMD_ZBIT, i)
 156 #define NZBIT32(x, i) \
 157     SIMD32_SET(NBIT32((x) & 0xffffffff), SIMD_NBIT, i) | \
 158     SIMD32_SET(ZBIT32((x) & 0xffffffff), SIMD_ZBIT, i)
 159 #define NZBIT64(x) \
 160     SIMD64_SET(NBIT64(x), SIMD_NBIT) | \
 161     SIMD64_SET(ZBIT64(x), SIMD_ZBIT)
 162 #define IWMMXT_OP_UNPACK(S, SH0, SH1, SH2, SH3)                         \
 163 uint64_t HELPER(glue(iwmmxt_unpack, glue(S, b)))(CPUARMState *env, \
 164                                                  uint64_t a, uint64_t b) \
 165 {                                                               \
 166     a =                                                                 \
 167         (((a >> SH0) & 0xff) << 0) | (((b >> SH0) & 0xff) << 8) |       \
 168         (((a >> SH1) & 0xff) << 16) | (((b >> SH1) & 0xff) << 24) |     \
 169         (((a >> SH2) & 0xff) << 32) | (((b >> SH2) & 0xff) << 40) |     \
 170         (((a >> SH3) & 0xff) << 48) | (((b >> SH3) & 0xff) << 56);      \
 171     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                       \
 172         NZBIT8(a >> 0, 0) | NZBIT8(a >> 8, 1) |                         \
 173         NZBIT8(a >> 16, 2) | NZBIT8(a >> 24, 3) |               \
 174         NZBIT8(a >> 32, 4) | NZBIT8(a >> 40, 5) |               \
 175         NZBIT8(a >> 48, 6) | NZBIT8(a >> 56, 7);                \
 176     return a;                                                   \
 177 }                                                               \
 178 uint64_t HELPER(glue(iwmmxt_unpack, glue(S, w)))(CPUARMState *env, \
 179                                         uint64_t a, uint64_t b) \
 180 {                                                               \
 181     a =                                                                 \
 182         (((a >> SH0) & 0xffff) << 0) |                          \
 183         (((b >> SH0) & 0xffff) << 16) |                                 \
 184         (((a >> SH2) & 0xffff) << 32) |                                 \
 185         (((b >> SH2) & 0xffff) << 48);                          \
 186     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                       \
 187         NZBIT8(a >> 0, 0) | NZBIT8(a >> 16, 1) |                \
 188         NZBIT8(a >> 32, 2) | NZBIT8(a >> 48, 3);                \
 189     return a;                                                   \
 190 }                                                               \
 191 uint64_t HELPER(glue(iwmmxt_unpack, glue(S, l)))(CPUARMState *env, \
 192                                         uint64_t a, uint64_t b) \
 193 {                                                               \
 194     a =                                                                 \
 195         (((a >> SH0) & 0xffffffff) << 0) |                      \
 196         (((b >> SH0) & 0xffffffff) << 32);                      \
 197     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                       \
 198         NZBIT32(a >> 0, 0) | NZBIT32(a >> 32, 1);               \
 199     return a;                                                   \
 200 }                                                               \
 201 uint64_t HELPER(glue(iwmmxt_unpack, glue(S, ub)))(CPUARMState *env, \
 202                                                   uint64_t x)   \
 203 {                                                               \
 204     x =                                                                 \
 205         (((x >> SH0) & 0xff) << 0) |                            \
 206         (((x >> SH1) & 0xff) << 16) |                           \
 207         (((x >> SH2) & 0xff) << 32) |                           \
 208         (((x >> SH3) & 0xff) << 48);                            \
 209     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                       \
 210         NZBIT16(x >> 0, 0) | NZBIT16(x >> 16, 1) |              \
 211         NZBIT16(x >> 32, 2) | NZBIT16(x >> 48, 3);              \
 212     return x;                                                   \
 213 }                                                               \
 214 uint64_t HELPER(glue(iwmmxt_unpack, glue(S, uw)))(CPUARMState *env, \
 215                                                   uint64_t x)   \
 216 {                                                               \
 217     x =                                                                 \
 218         (((x >> SH0) & 0xffff) << 0) |                          \
 219         (((x >> SH2) & 0xffff) << 32);                          \
 220     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                       \
 221         NZBIT32(x >> 0, 0) | NZBIT32(x >> 32, 1);               \
 222     return x;                                                   \
 223 }                                                               \
 224 uint64_t HELPER(glue(iwmmxt_unpack, glue(S, ul)))(CPUARMState *env, \
 225                                                   uint64_t x)   \
 226 {                                                               \
 227     x = (((x >> SH0) & 0xffffffff) << 0);                       \
 228     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] = NZBIT64(x >> 0);      \
 229     return x;                                                   \
 230 }                                                               \
 231 uint64_t HELPER(glue(iwmmxt_unpack, glue(S, sb)))(CPUARMState *env, \
 232                                                   uint64_t x)   \
 233 {                                                               \
 234     x =                                                                 \
 235         ((uint64_t) EXTEND8H((x >> SH0) & 0xff) << 0) |                 \
 236         ((uint64_t) EXTEND8H((x >> SH1) & 0xff) << 16) |        \
 237         ((uint64_t) EXTEND8H((x >> SH2) & 0xff) << 32) |        \
 238         ((uint64_t) EXTEND8H((x >> SH3) & 0xff) << 48);                 \
 239     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                       \
 240         NZBIT16(x >> 0, 0) | NZBIT16(x >> 16, 1) |              \
 241         NZBIT16(x >> 32, 2) | NZBIT16(x >> 48, 3);              \
 242     return x;                                                   \
 243 }                                                               \
 244 uint64_t HELPER(glue(iwmmxt_unpack, glue(S, sw)))(CPUARMState *env, \
 245                                                   uint64_t x)   \
 246 {                                                               \
 247     x =                                                                 \
 248         ((uint64_t) EXTEND16((x >> SH0) & 0xffff) << 0) |       \
 249         ((uint64_t) EXTEND16((x >> SH2) & 0xffff) << 32);       \
 250     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                       \
 251         NZBIT32(x >> 0, 0) | NZBIT32(x >> 32, 1);               \
 252     return x;                                                   \
 253 }                                                               \
 254 uint64_t HELPER(glue(iwmmxt_unpack, glue(S, sl)))(CPUARMState *env, \
 255                                                   uint64_t x)   \
 256 {                                                               \
 257     x = EXTEND32((x >> SH0) & 0xffffffff);                      \
 258     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] = NZBIT64(x >> 0);      \
 259     return x;                                                   \
 260 }
 261 IWMMXT_OP_UNPACK(l, 0, 8, 16, 24)
 262 IWMMXT_OP_UNPACK(h, 32, 40, 48, 56)
 263
 264 #define IWMMXT_OP_CMP(SUFF, Tb, Tw, Tl, O)                      \
 265 uint64_t HELPER(glue(iwmmxt_, glue(SUFF, b)))(CPUARMState *env,    \
 266                                         uint64_t a, uint64_t b) \
 267 {                                                               \
 268     a =                                                                 \
 269         CMP(0, Tb, O, 0xff) | CMP(8, Tb, O, 0xff) |             \
 270         CMP(16, Tb, O, 0xff) | CMP(24, Tb, O, 0xff) |           \
 271         CMP(32, Tb, O, 0xff) | CMP(40, Tb, O, 0xff) |           \
 272         CMP(48, Tb, O, 0xff) | CMP(56, Tb, O, 0xff);            \
 273     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                       \
 274         NZBIT8(a >> 0, 0) | NZBIT8(a >> 8, 1) |                         \
 275         NZBIT8(a >> 16, 2) | NZBIT8(a >> 24, 3) |               \
 276         NZBIT8(a >> 32, 4) | NZBIT8(a >> 40, 5) |               \
 277         NZBIT8(a >> 48, 6) | NZBIT8(a >> 56, 7);                \
 278     return a;                                                   \
 279 }                                                               \
 280 uint64_t HELPER(glue(iwmmxt_, glue(SUFF, w)))(CPUARMState *env,    \
 281                                         uint64_t a, uint64_t b) \
 282 {                                                               \
 283     a = CMP(0, Tw, O, 0xffff) | CMP(16, Tw, O, 0xffff) |        \
 284         CMP(32, Tw, O, 0xffff) | CMP(48, Tw, O, 0xffff);        \
 285     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                       \
 286         NZBIT16(a >> 0, 0) | NZBIT16(a >> 16, 1) |              \
 287         NZBIT16(a >> 32, 2) | NZBIT16(a >> 48, 3);              \
 288     return a;                                                   \
 289 }                                                               \
 290 uint64_t HELPER(glue(iwmmxt_, glue(SUFF, l)))(CPUARMState *env,    \
 291                                         uint64_t a, uint64_t b) \
 292 {                                                               \
 293     a = CMP(0, Tl, O, 0xffffffff) |                             \
 294         CMP(32, Tl, O, 0xffffffff);                             \
 295     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                       \
 296         NZBIT32(a >> 0, 0) | NZBIT32(a >> 32, 1);               \
 297     return a;                                                   \
 298 }
 299 #define CMP(SHR, TYPE, OPER, MASK) ((((TYPE) ((a >> SHR) & MASK) OPER \
 300             (TYPE) ((b >> SHR) & MASK)) ? (uint64_t) MASK : 0) << SHR)
 301 IWMMXT_OP_CMP(cmpeq, uint8_t, uint16_t, uint32_t, ==)
 302 IWMMXT_OP_CMP(cmpgts, int8_t, int16_t, int32_t, >)
 303 IWMMXT_OP_CMP(cmpgtu, uint8_t, uint16_t, uint32_t, >)
 304 #undef CMP
 305 #define CMP(SHR, TYPE, OPER, MASK) ((((TYPE) ((a >> SHR) & MASK) OPER \
 306             (TYPE) ((b >> SHR) & MASK)) ? a : b) & ((uint64_t) MASK << SHR))
 307 IWMMXT_OP_CMP(mins, int8_t, int16_t, int32_t, <)
 308 IWMMXT_OP_CMP(minu, uint8_t, uint16_t, uint32_t, <)
 309 IWMMXT_OP_CMP(maxs, int8_t, int16_t, int32_t, >)
 310 IWMMXT_OP_CMP(maxu, uint8_t, uint16_t, uint32_t, >)
 311 #undef CMP
 312 #define CMP(SHR, TYPE, OPER, MASK) ((uint64_t) (((TYPE) ((a >> SHR) & MASK) \
 313             OPER (TYPE) ((b >> SHR) & MASK)) & MASK) << SHR)
 314 IWMMXT_OP_CMP(subn, uint8_t, uint16_t, uint32_t, -)
 315 IWMMXT_OP_CMP(addn, uint8_t, uint16_t, uint32_t, +)
 316 #undef CMP
 317 /* TODO Signed- and Unsigned-Saturation */
 318 #define CMP(SHR, TYPE, OPER, MASK) ((uint64_t) (((TYPE) ((a >> SHR) & MASK) \
 319             OPER (TYPE) ((b >> SHR) & MASK)) & MASK) << SHR)
 320 IWMMXT_OP_CMP(subu, uint8_t, uint16_t, uint32_t, -)
 321 IWMMXT_OP_CMP(addu, uint8_t, uint16_t, uint32_t, +)
 322 IWMMXT_OP_CMP(subs, int8_t, int16_t, int32_t, -)
 323 IWMMXT_OP_CMP(adds, int8_t, int16_t, int32_t, +)
 324 #undef CMP
 325 #undef IWMMXT_OP_CMP
 326
 327 #define AVGB(SHR) ((( \
 328         ((a >> SHR) & 0xff) + ((b >> SHR) & 0xff) + round) >> 1) << SHR)
 329 #define IWMMXT_OP_AVGB(r)                                                 \
 330 uint64_t HELPER(iwmmxt_avgb##r)(CPUARMState *env, uint64_t a, uint64_t b)    \
 331 {                                                                         \
 332     const int round = r;                                                  \
 333     a = AVGB(0) | AVGB(8) | AVGB(16) | AVGB(24) |                         \
 334         AVGB(32) | AVGB(40) | AVGB(48) | AVGB(56);                        \
 335     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                                 \
 336         SIMD8_SET(ZBIT8((a >> 0) & 0xff), SIMD_ZBIT, 0) |                 \
 337         SIMD8_SET(ZBIT8((a >> 8) & 0xff), SIMD_ZBIT, 1) |                 \
 338         SIMD8_SET(ZBIT8((a >> 16) & 0xff), SIMD_ZBIT, 2) |                \
 339         SIMD8_SET(ZBIT8((a >> 24) & 0xff), SIMD_ZBIT, 3) |                \
 340         SIMD8_SET(ZBIT8((a >> 32) & 0xff), SIMD_ZBIT, 4) |                \
 341         SIMD8_SET(ZBIT8((a >> 40) & 0xff), SIMD_ZBIT, 5) |                \
 342         SIMD8_SET(ZBIT8((a >> 48) & 0xff), SIMD_ZBIT, 6) |                \
 343         SIMD8_SET(ZBIT8((a >> 56) & 0xff), SIMD_ZBIT, 7);                 \
 344     return a;                                                             \
 345 }
 346 IWMMXT_OP_AVGB(0)
 347 IWMMXT_OP_AVGB(1)
 348 #undef IWMMXT_OP_AVGB
 349 #undef AVGB
 350
 351 #define AVGW(SHR) ((( \
 352         ((a >> SHR) & 0xffff) + ((b >> SHR) & 0xffff) + round) >> 1) << SHR)
 353 #define IWMMXT_OP_AVGW(r)                                               \
 354 uint64_t HELPER(iwmmxt_avgw##r)(CPUARMState *env, uint64_t a, uint64_t b)  \
 355 {                                                                       \
 356     const int round = r;                                                \
 357     a = AVGW(0) | AVGW(16) | AVGW(32) | AVGW(48);                       \
 358     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                               \
 359         SIMD16_SET(ZBIT16((a >> 0) & 0xffff), SIMD_ZBIT, 0) |           \
 360         SIMD16_SET(ZBIT16((a >> 16) & 0xffff), SIMD_ZBIT, 1) |          \
 361         SIMD16_SET(ZBIT16((a >> 32) & 0xffff), SIMD_ZBIT, 2) |          \
 362         SIMD16_SET(ZBIT16((a >> 48) & 0xffff), SIMD_ZBIT, 3);           \
 363     return a;                                                           \
 364 }
 365 IWMMXT_OP_AVGW(0)
 366 IWMMXT_OP_AVGW(1)
 367 #undef IWMMXT_OP_AVGW
 368 #undef AVGW
 369
 370 uint64_t HELPER(iwmmxt_align)(uint64_t a, uint64_t b, uint32_t n)
 371 {
 372     a >>= n << 3;
 373     a |= b << (64 - (n << 3));
 374     return a;
 375 }
 376
 377 uint64_t HELPER(iwmmxt_insr)(uint64_t x, uint32_t a, uint32_t b, uint32_t n)
 378 {
 379     x &= ~((uint64_t) b << n);
 380     x |= (uint64_t) (a & b) << n;
 381     return x;
 382 }
 383
 384 uint32_t HELPER(iwmmxt_setpsr_nz)(uint64_t x)
 385 {
 386     return SIMD64_SET((x == 0), SIMD_ZBIT) |
 387            SIMD64_SET((x & (1ULL << 63)), SIMD_NBIT);
 388 }
 389
 390 uint64_t HELPER(iwmmxt_bcstb)(uint32_t arg)
 391 {
 392     arg &= 0xff;
 393     return
 394         ((uint64_t) arg << 0 ) | ((uint64_t) arg << 8 ) |
 395         ((uint64_t) arg << 16) | ((uint64_t) arg << 24) |
 396         ((uint64_t) arg << 32) | ((uint64_t) arg << 40) |
 397         ((uint64_t) arg << 48) | ((uint64_t) arg << 56);
 398 }
 399
 400 uint64_t HELPER(iwmmxt_bcstw)(uint32_t arg)
 401 {
 402     arg &= 0xffff;
 403     return
 404         ((uint64_t) arg << 0 ) | ((uint64_t) arg << 16) |
 405         ((uint64_t) arg << 32) | ((uint64_t) arg << 48);
 406 }
 407
 408 uint64_t HELPER(iwmmxt_bcstl)(uint32_t arg)
 409 {
 410     return arg | ((uint64_t) arg << 32);
 411 }
 412
 413 uint64_t HELPER(iwmmxt_addcb)(uint64_t x)
 414 {
 415     return
 416         ((x >> 0) & 0xff) + ((x >> 8) & 0xff) +
 417         ((x >> 16) & 0xff) + ((x >> 24) & 0xff) +
 418         ((x >> 32) & 0xff) + ((x >> 40) & 0xff) +
 419         ((x >> 48) & 0xff) + ((x >> 56) & 0xff);
 420 }
 421
 422 uint64_t HELPER(iwmmxt_addcw)(uint64_t x)
 423 {
 424     return
 425         ((x >> 0) & 0xffff) + ((x >> 16) & 0xffff) +
 426         ((x >> 32) & 0xffff) + ((x >> 48) & 0xffff);
 427 }
 428
 429 uint64_t HELPER(iwmmxt_addcl)(uint64_t x)
 430 {
 431     return (x & 0xffffffff) + (x >> 32);
 432 }
 433
 434 uint32_t HELPER(iwmmxt_msbb)(uint64_t x)
 435 {
 436     return
 437         ((x >> 7) & 0x01) | ((x >> 14) & 0x02) |
 438         ((x >> 21) & 0x04) | ((x >> 28) & 0x08) |
 439         ((x >> 35) & 0x10) | ((x >> 42) & 0x20) |
 440         ((x >> 49) & 0x40) | ((x >> 56) & 0x80);
 441 }
 442
 443 uint32_t HELPER(iwmmxt_msbw)(uint64_t x)
 444 {
 445     return
 446         ((x >> 15) & 0x01) | ((x >> 30) & 0x02) |
 447         ((x >> 45) & 0x04) | ((x >> 52) & 0x08);
 448 }
 449
 450 uint32_t HELPER(iwmmxt_msbl)(uint64_t x)
 451 {
 452     return ((x >> 31) & 0x01) | ((x >> 62) & 0x02);
 453 }
 454
 455 /* FIXME: Split wCASF setting into a separate op to avoid env use.  */
 456 uint64_t HELPER(iwmmxt_srlw)(CPUARMState *env, uint64_t x, uint32_t n)
 457 {
 458     x = (((x & (0xffffll << 0)) >> n) & (0xffffll << 0)) |
 459         (((x & (0xffffll << 16)) >> n) & (0xffffll << 16)) |
 460         (((x & (0xffffll << 32)) >> n) & (0xffffll << 32)) |
 461         (((x & (0xffffll << 48)) >> n) & (0xffffll << 48));
 462     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 463         NZBIT16(x >> 0, 0) | NZBIT16(x >> 16, 1) |
 464         NZBIT16(x >> 32, 2) | NZBIT16(x >> 48, 3);
 465     return x;
 466 }
 467
 468 uint64_t HELPER(iwmmxt_srll)(CPUARMState *env, uint64_t x, uint32_t n)
 469 {
 470     x = ((x & (0xffffffffll << 0)) >> n) |
 471         ((x >> n) & (0xffffffffll << 32));
 472     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 473         NZBIT32(x >> 0, 0) | NZBIT32(x >> 32, 1);
 474     return x;
 475 }
 476
 477 uint64_t HELPER(iwmmxt_srlq)(CPUARMState *env, uint64_t x, uint32_t n)
 478 {
 479     x >>= n;
 480     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] = NZBIT64(x);
 481     return x;
 482 }
 483
 484 uint64_t HELPER(iwmmxt_sllw)(CPUARMState *env, uint64_t x, uint32_t n)
 485 {
 486     x = (((x & (0xffffll << 0)) << n) & (0xffffll << 0)) |
 487         (((x & (0xffffll << 16)) << n) & (0xffffll << 16)) |
 488         (((x & (0xffffll << 32)) << n) & (0xffffll << 32)) |
 489         (((x & (0xffffll << 48)) << n) & (0xffffll << 48));
 490     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 491         NZBIT16(x >> 0, 0) | NZBIT16(x >> 16, 1) |
 492         NZBIT16(x >> 32, 2) | NZBIT16(x >> 48, 3);
 493     return x;
 494 }
 495
 496 uint64_t HELPER(iwmmxt_slll)(CPUARMState *env, uint64_t x, uint32_t n)
 497 {
 498     x = ((x << n) & (0xffffffffll << 0)) |
 499         ((x & (0xffffffffll << 32)) << n);
 500     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 501         NZBIT32(x >> 0, 0) | NZBIT32(x >> 32, 1);
 502     return x;
 503 }
 504
 505 uint64_t HELPER(iwmmxt_sllq)(CPUARMState *env, uint64_t x, uint32_t n)
 506 {
 507     x <<= n;
 508     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] = NZBIT64(x);
 509     return x;
 510 }
 511
 512 uint64_t HELPER(iwmmxt_sraw)(CPUARMState *env, uint64_t x, uint32_t n)
 513 {
 514     x = ((uint64_t) ((EXTEND16(x >> 0) >> n) & 0xffff) << 0) |
 515         ((uint64_t) ((EXTEND16(x >> 16) >> n) & 0xffff) << 16) |
 516         ((uint64_t) ((EXTEND16(x >> 32) >> n) & 0xffff) << 32) |
 517         ((uint64_t) ((EXTEND16(x >> 48) >> n) & 0xffff) << 48);
 518     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 519         NZBIT16(x >> 0, 0) | NZBIT16(x >> 16, 1) |
 520         NZBIT16(x >> 32, 2) | NZBIT16(x >> 48, 3);
 521     return x;
 522 }
 523
 524 uint64_t HELPER(iwmmxt_sral)(CPUARMState *env, uint64_t x, uint32_t n)
 525 {
 526     x = (((EXTEND32(x >> 0) >> n) & 0xffffffff) << 0) |
 527         (((EXTEND32(x >> 32) >> n) & 0xffffffff) << 32);
 528     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 529         NZBIT32(x >> 0, 0) | NZBIT32(x >> 32, 1);
 530     return x;
 531 }
 532
 533 uint64_t HELPER(iwmmxt_sraq)(CPUARMState *env, uint64_t x, uint32_t n)
 534 {
 535     x = (int64_t) x >> n;
 536     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] = NZBIT64(x);
 537     return x;
 538 }
 539
 540 uint64_t HELPER(iwmmxt_rorw)(CPUARMState *env, uint64_t x, uint32_t n)
 541 {
 542     x = ((((x & (0xffffll << 0)) >> n) |
 543           ((x & (0xffffll << 0)) << (16 - n))) & (0xffffll << 0)) |
 544         ((((x & (0xffffll << 16)) >> n) |
 545           ((x & (0xffffll << 16)) << (16 - n))) & (0xffffll << 16)) |
 546         ((((x & (0xffffll << 32)) >> n) |
 547           ((x & (0xffffll << 32)) << (16 - n))) & (0xffffll << 32)) |
 548         ((((x & (0xffffll << 48)) >> n) |
 549           ((x & (0xffffll << 48)) << (16 - n))) & (0xffffll << 48));
 550     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 551         NZBIT16(x >> 0, 0) | NZBIT16(x >> 16, 1) |
 552         NZBIT16(x >> 32, 2) | NZBIT16(x >> 48, 3);
 553     return x;
 554 }
 555
 556 uint64_t HELPER(iwmmxt_rorl)(CPUARMState *env, uint64_t x, uint32_t n)
 557 {
 558     x = ((x & (0xffffffffll << 0)) >> n) |
 559         ((x >> n) & (0xffffffffll << 32)) |
 560         ((x << (32 - n)) & (0xffffffffll << 0)) |
 561         ((x & (0xffffffffll << 32)) << (32 - n));
 562     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 563         NZBIT32(x >> 0, 0) | NZBIT32(x >> 32, 1);
 564     return x;
 565 }
 566
 567 uint64_t HELPER(iwmmxt_rorq)(CPUARMState *env, uint64_t x, uint32_t n)
 568 {
 569     x = ror64(x, n);
 570     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] = NZBIT64(x);
 571     return x;
 572 }
 573
 574 uint64_t HELPER(iwmmxt_shufh)(CPUARMState *env, uint64_t x, uint32_t n)
 575 {
 576     x = (((x >> ((n << 4) & 0x30)) & 0xffff) << 0) |
 577         (((x >> ((n << 2) & 0x30)) & 0xffff) << 16) |
 578         (((x >> ((n << 0) & 0x30)) & 0xffff) << 32) |
 579         (((x >> ((n >> 2) & 0x30)) & 0xffff) << 48);
 580     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 581         NZBIT16(x >> 0, 0) | NZBIT16(x >> 16, 1) |
 582         NZBIT16(x >> 32, 2) | NZBIT16(x >> 48, 3);
 583     return x;
 584 }
 585
 586 /* TODO: Unsigned-Saturation */
 587 uint64_t HELPER(iwmmxt_packuw)(CPUARMState *env, uint64_t a, uint64_t b)
 588 {
 589     a = (((a >> 0) & 0xff) << 0) | (((a >> 16) & 0xff) << 8) |
 590         (((a >> 32) & 0xff) << 16) | (((a >> 48) & 0xff) << 24) |
 591         (((b >> 0) & 0xff) << 32) | (((b >> 16) & 0xff) << 40) |
 592         (((b >> 32) & 0xff) << 48) | (((b >> 48) & 0xff) << 56);
 593     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 594         NZBIT8(a >> 0, 0) | NZBIT8(a >> 8, 1) |
 595         NZBIT8(a >> 16, 2) | NZBIT8(a >> 24, 3) |
 596         NZBIT8(a >> 32, 4) | NZBIT8(a >> 40, 5) |
 597         NZBIT8(a >> 48, 6) | NZBIT8(a >> 56, 7);
 598     return a;
 599 }
 600
 601 uint64_t HELPER(iwmmxt_packul)(CPUARMState *env, uint64_t a, uint64_t b)
 602 {
 603     a = (((a >> 0) & 0xffff) << 0) | (((a >> 32) & 0xffff) << 16) |
 604         (((b >> 0) & 0xffff) << 32) | (((b >> 32) & 0xffff) << 48);
 605     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 606         NZBIT16(a >> 0, 0) | NZBIT16(a >> 16, 1) |
 607         NZBIT16(a >> 32, 2) | NZBIT16(a >> 48, 3);
 608     return a;
 609 }
 610
 611 uint64_t HELPER(iwmmxt_packuq)(CPUARMState *env, uint64_t a, uint64_t b)
 612 {
 613     a = (a & 0xffffffff) | ((b & 0xffffffff) << 32);
 614     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 615         NZBIT32(a >> 0, 0) | NZBIT32(a >> 32, 1);
 616     return a;
 617 }
 618
 619 /* TODO: Signed-Saturation */
 620 uint64_t HELPER(iwmmxt_packsw)(CPUARMState *env, uint64_t a, uint64_t b)
 621 {
 622     a = (((a >> 0) & 0xff) << 0) | (((a >> 16) & 0xff) << 8) |
 623         (((a >> 32) & 0xff) << 16) | (((a >> 48) & 0xff) << 24) |
 624         (((b >> 0) & 0xff) << 32) | (((b >> 16) & 0xff) << 40) |
 625         (((b >> 32) & 0xff) << 48) | (((b >> 48) & 0xff) << 56);
 626     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 627         NZBIT8(a >> 0, 0) | NZBIT8(a >> 8, 1) |
 628         NZBIT8(a >> 16, 2) | NZBIT8(a >> 24, 3) |
 629         NZBIT8(a >> 32, 4) | NZBIT8(a >> 40, 5) |
 630         NZBIT8(a >> 48, 6) | NZBIT8(a >> 56, 7);
 631     return a;
 632 }
 633
 634 uint64_t HELPER(iwmmxt_packsl)(CPUARMState *env, uint64_t a, uint64_t b)
 635 {
 636     a = (((a >> 0) & 0xffff) << 0) | (((a >> 32) & 0xffff) << 16) |
 637         (((b >> 0) & 0xffff) << 32) | (((b >> 32) & 0xffff) << 48);
 638     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 639         NZBIT16(a >> 0, 0) | NZBIT16(a >> 16, 1) |
 640         NZBIT16(a >> 32, 2) | NZBIT16(a >> 48, 3);
 641     return a;
 642 }
 643
 644 uint64_t HELPER(iwmmxt_packsq)(CPUARMState *env, uint64_t a, uint64_t b)
 645 {
 646     a = (a & 0xffffffff) | ((b & 0xffffffff) << 32);
 647     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 648         NZBIT32(a >> 0, 0) | NZBIT32(a >> 32, 1);
 649     return a;
 650 }
 651
 652 uint64_t HELPER(iwmmxt_muladdsl)(uint64_t c, uint32_t a, uint32_t b)
 653 {
 654     return c + ((int32_t) EXTEND32(a) * (int32_t) EXTEND32(b));
 655 }
 656
 657 uint64_t HELPER(iwmmxt_muladdsw)(uint64_t c, uint32_t a, uint32_t b)
 658 {
 659     c += EXTEND32(EXTEND16S((a >> 0) & 0xffff) *
 660                   EXTEND16S((b >> 0) & 0xffff));
 661     c += EXTEND32(EXTEND16S((a >> 16) & 0xffff) *
 662                   EXTEND16S((b >> 16) & 0xffff));
 663     return c;
 664 }
 665
 666 uint64_t HELPER(iwmmxt_muladdswl)(uint64_t c, uint32_t a, uint32_t b)
 667 {
 668     return c + (EXTEND32(EXTEND16S(a & 0xffff) *
 669                          EXTEND16S(b & 0xffff)));
 670 }