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 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/exec-all.h"
  26 #include "exec/helper-proto.h"
  27
  28 /* iwMMXt macros extracted from GNU gdb.  */
  29
  30 /* Set the SIMD wCASF flags for 8, 16, 32 or 64-bit operations.  */
  31 #define SIMD8_SET( v, n, b)     ((v != 0) << ((((b) + 1) * 4) + (n)))
  32 #define SIMD16_SET(v, n, h)     ((v != 0) << ((((h) + 1) * 8) + (n)))
  33 #define SIMD32_SET(v, n, w)     ((v != 0) << ((((w) + 1) * 16) + (n)))
  34 #define SIMD64_SET(v, n)        ((v != 0) << (32 + (n)))
  35 /* Flags to pass as "n" above.  */
  36 #define SIMD_NBIT       -1
  37 #define SIMD_ZBIT       -2
  38 #define SIMD_CBIT       -3
  39 #define SIMD_VBIT       -4
  40 /* Various status bit macros.  */
  41 #define NBIT8(x)        ((x) & 0x80)
  42 #define NBIT16(x)       ((x) & 0x8000)
  43 #define NBIT32(x)       ((x) & 0x80000000)
  44 #define NBIT64(x)       ((x) & 0x8000000000000000ULL)
  45 #define ZBIT8(x)        (((x) & 0xff) == 0)
  46 #define ZBIT16(x)       (((x) & 0xffff) == 0)
  47 #define ZBIT32(x)       (((x) & 0xffffffff) == 0)
  48 #define ZBIT64(x)       (x == 0)
  49 /* Sign extension macros.  */
  50 #define EXTEND8H(a)     ((uint16_t) (int8_t) (a))
  51 #define EXTEND8(a)      ((uint32_t) (int8_t) (a))
  52 #define EXTEND16(a)     ((uint32_t) (int16_t) (a))
  53 #define EXTEND16S(a)    ((int32_t) (int16_t) (a))
  54 #define EXTEND32(a)     ((uint64_t) (int32_t) (a))
  55
  56 uint64_t HELPER(iwmmxt_maddsq)(uint64_t a, uint64_t b)
  57 {
  58     a = ((
  59             EXTEND16S((a >> 0) & 0xffff) * EXTEND16S((b >> 0) & 0xffff) +
  60             EXTEND16S((a >> 16) & 0xffff) * EXTEND16S((b >> 16) & 0xffff)
  61         ) & 0xffffffff) | ((uint64_t) (
  62             EXTEND16S((a >> 32) & 0xffff) * EXTEND16S((b >> 32) & 0xffff) +
  63             EXTEND16S((a >> 48) & 0xffff) * EXTEND16S((b >> 48) & 0xffff)
  64         ) << 32);
  65     return a;
  66 }
  67
  68 uint64_t HELPER(iwmmxt_madduq)(uint64_t a, uint64_t b)
  69 {
  70     a = ((
  71             ((a >> 0) & 0xffff) * ((b >> 0) & 0xffff) +
  72             ((a >> 16) & 0xffff) * ((b >> 16) & 0xffff)
  73         ) & 0xffffffff) | ((
  74             ((a >> 32) & 0xffff) * ((b >> 32) & 0xffff) +
  75             ((a >> 48) & 0xffff) * ((b >> 48) & 0xffff)
  76         ) << 32);
  77     return a;
  78 }
  79
  80 uint64_t HELPER(iwmmxt_sadb)(uint64_t a, uint64_t b)
  81 {
  82 #define abs(x) (((x) >= 0) ? x : -x)
  83 #define SADB(SHR) abs((int) ((a >> SHR) & 0xff) - (int) ((b >> SHR) & 0xff))
  84     return
  85         SADB(0) + SADB(8) + SADB(16) + SADB(24) +
  86         SADB(32) + SADB(40) + SADB(48) + SADB(56);
  87 #undef SADB
  88 }
  89
  90 uint64_t HELPER(iwmmxt_sadw)(uint64_t a, uint64_t b)
  91 {
  92 #define SADW(SHR) \
  93     abs((int) ((a >> SHR) & 0xffff) - (int) ((b >> SHR) & 0xffff))
  94     return SADW(0) + SADW(16) + SADW(32) + SADW(48);
  95 #undef SADW
  96 }
  97
  98 uint64_t HELPER(iwmmxt_mulslw)(uint64_t a, uint64_t b)
  99 {
 100 #define MULS(SHR) ((uint64_t) ((( \
 101         EXTEND16S((a >> SHR) & 0xffff) * EXTEND16S((b >> SHR) & 0xffff) \
 102     ) >> 0) & 0xffff) << SHR)
 103     return MULS(0) | MULS(16) | MULS(32) | MULS(48);
 104 #undef MULS
 105 }
 106
 107 uint64_t HELPER(iwmmxt_mulshw)(uint64_t a, uint64_t b)
 108 {
 109 #define MULS(SHR) ((uint64_t) ((( \
 110         EXTEND16S((a >> SHR) & 0xffff) * EXTEND16S((b >> SHR) & 0xffff) \
 111     ) >> 16) & 0xffff) << SHR)
 112     return MULS(0) | MULS(16) | MULS(32) | MULS(48);
 113 #undef MULS
 114 }
 115
 116 uint64_t HELPER(iwmmxt_mululw)(uint64_t a, uint64_t b)
 117 {
 118 #define MULU(SHR) ((uint64_t) ((( \
 119         ((a >> SHR) & 0xffff) * ((b >> SHR) & 0xffff) \
 120     ) >> 0) & 0xffff) << SHR)
 121     return MULU(0) | MULU(16) | MULU(32) | MULU(48);
 122 #undef MULU
 123 }
 124
 125 uint64_t HELPER(iwmmxt_muluhw)(uint64_t a, uint64_t b)
 126 {
 127 #define MULU(SHR) ((uint64_t) ((( \
 128         ((a >> SHR) & 0xffff) * ((b >> SHR) & 0xffff) \
 129     ) >> 16) & 0xffff) << SHR)
 130     return MULU(0) | MULU(16) | MULU(32) | MULU(48);
 131 #undef MULU
 132 }
 133
 134 uint64_t HELPER(iwmmxt_macsw)(uint64_t a, uint64_t b)
 135 {
 136 #define MACS(SHR) ( \
 137         EXTEND16((a >> SHR) & 0xffff) * EXTEND16S((b >> SHR) & 0xffff))
 138     return (int64_t) (MACS(0) + MACS(16) + MACS(32) + MACS(48));
 139 #undef MACS
 140 }
 141
 142 uint64_t HELPER(iwmmxt_macuw)(uint64_t a, uint64_t b)
 143 {
 144 #define MACU(SHR) ( \
 145         (uint32_t) ((a >> SHR) & 0xffff) * \
 146         (uint32_t) ((b >> SHR) & 0xffff))
 147     return MACU(0) + MACU(16) + MACU(32) + MACU(48);
 148 #undef MACU
 149 }
 150
 151 #define NZBIT8(x, i) \
 152     SIMD8_SET(NBIT8((x) & 0xff), SIMD_NBIT, i) | \
 153     SIMD8_SET(ZBIT8((x) & 0xff), SIMD_ZBIT, i)
 154 #define NZBIT16(x, i) \
 155     SIMD16_SET(NBIT16((x) & 0xffff), SIMD_NBIT, i) | \
 156     SIMD16_SET(ZBIT16((x) & 0xffff), SIMD_ZBIT, i)
 157 #define NZBIT32(x, i) \
 158     SIMD32_SET(NBIT32((x) & 0xffffffff), SIMD_NBIT, i) | \
 159     SIMD32_SET(ZBIT32((x) & 0xffffffff), SIMD_ZBIT, i)
 160 #define NZBIT64(x) \
 161     SIMD64_SET(NBIT64(x), SIMD_NBIT) | \
 162     SIMD64_SET(ZBIT64(x), SIMD_ZBIT)
 163 #define IWMMXT_OP_UNPACK(S, SH0, SH1, SH2, SH3)                 \
 164 uint64_t HELPER(glue(iwmmxt_unpack, glue(S, b)))(CPUARMState *env, \
 165                                                  uint64_t a, uint64_t b) \
 166 {                                                               \
 167     a =                                                         \
 168         (((a >> SH0) & 0xff) << 0) | (((b >> SH0) & 0xff) << 8) |       \
 169         (((a >> SH1) & 0xff) << 16) | (((b >> SH1) & 0xff) << 24) |     \
 170         (((a >> SH2) & 0xff) << 32) | (((b >> SH2) & 0xff) << 40) |     \
 171         (((a >> SH3) & 0xff) << 48) | (((b >> SH3) & 0xff) << 56);      \
 172     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                       \
 173         NZBIT8(a >> 0, 0) | NZBIT8(a >> 8, 1) |                 \
 174         NZBIT8(a >> 16, 2) | NZBIT8(a >> 24, 3) |               \
 175         NZBIT8(a >> 32, 4) | NZBIT8(a >> 40, 5) |               \
 176         NZBIT8(a >> 48, 6) | NZBIT8(a >> 56, 7);                \
 177     return a;                                                   \
 178 }                                                               \
 179 uint64_t HELPER(glue(iwmmxt_unpack, glue(S, w)))(CPUARMState *env, \
 180                                         uint64_t a, uint64_t b) \
 181 {                                                               \
 182     a =                                                         \
 183         (((a >> SH0) & 0xffff) << 0) |                          \
 184         (((b >> SH0) & 0xffff) << 16) |                         \
 185         (((a >> SH2) & 0xffff) << 32) |                         \
 186         (((b >> SH2) & 0xffff) << 48);                          \
 187     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                       \
 188         NZBIT8(a >> 0, 0) | NZBIT8(a >> 16, 1) |                \
 189         NZBIT8(a >> 32, 2) | NZBIT8(a >> 48, 3);                \
 190     return a;                                                   \
 191 }                                                               \
 192 uint64_t HELPER(glue(iwmmxt_unpack, glue(S, l)))(CPUARMState *env, \
 193                                         uint64_t a, uint64_t b) \
 194 {                                                               \
 195     a =                                                         \
 196         (((a >> SH0) & 0xffffffff) << 0) |                      \
 197         (((b >> SH0) & 0xffffffff) << 32);                      \
 198     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                       \
 199         NZBIT32(a >> 0, 0) | NZBIT32(a >> 32, 1);               \
 200     return a;                                                   \
 201 }                                                               \
 202 uint64_t HELPER(glue(iwmmxt_unpack, glue(S, ub)))(CPUARMState *env, \
 203                                                   uint64_t x)   \
 204 {                                                               \
 205     x =                                                         \
 206         (((x >> SH0) & 0xff) << 0) |                            \
 207         (((x >> SH1) & 0xff) << 16) |                           \
 208         (((x >> SH2) & 0xff) << 32) |                           \
 209         (((x >> SH3) & 0xff) << 48);                            \
 210     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                       \
 211         NZBIT16(x >> 0, 0) | NZBIT16(x >> 16, 1) |              \
 212         NZBIT16(x >> 32, 2) | NZBIT16(x >> 48, 3);              \
 213     return x;                                                   \
 214 }                                                               \
 215 uint64_t HELPER(glue(iwmmxt_unpack, glue(S, uw)))(CPUARMState *env, \
 216                                                   uint64_t x)   \
 217 {                                                               \
 218     x =                                                         \
 219         (((x >> SH0) & 0xffff) << 0) |                          \
 220         (((x >> SH2) & 0xffff) << 32);                          \
 221     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                       \
 222         NZBIT32(x >> 0, 0) | NZBIT32(x >> 32, 1);               \
 223     return x;                                                   \
 224 }                                                               \
 225 uint64_t HELPER(glue(iwmmxt_unpack, glue(S, ul)))(CPUARMState *env, \
 226                                                   uint64_t x)   \
 227 {                                                               \
 228     x = (((x >> SH0) & 0xffffffff) << 0);                       \
 229     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] = NZBIT64(x >> 0);      \
 230     return x;                                                   \
 231 }                                                               \
 232 uint64_t HELPER(glue(iwmmxt_unpack, glue(S, sb)))(CPUARMState *env, \
 233                                                   uint64_t x)   \
 234 {                                                               \
 235     x =                                                         \
 236         ((uint64_t) EXTEND8H((x >> SH0) & 0xff) << 0) |         \
 237         ((uint64_t) EXTEND8H((x >> SH1) & 0xff) << 16) |        \
 238         ((uint64_t) EXTEND8H((x >> SH2) & 0xff) << 32) |        \
 239         ((uint64_t) EXTEND8H((x >> SH3) & 0xff) << 48);         \
 240     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                       \
 241         NZBIT16(x >> 0, 0) | NZBIT16(x >> 16, 1) |              \
 242         NZBIT16(x >> 32, 2) | NZBIT16(x >> 48, 3);              \
 243     return x;                                                   \
 244 }                                                               \
 245 uint64_t HELPER(glue(iwmmxt_unpack, glue(S, sw)))(CPUARMState *env, \
 246                                                   uint64_t x)   \
 247 {                                                               \
 248     x =                                                         \
 249         ((uint64_t) EXTEND16((x >> SH0) & 0xffff) << 0) |       \
 250         ((uint64_t) EXTEND16((x >> SH2) & 0xffff) << 32);       \
 251     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                       \
 252         NZBIT32(x >> 0, 0) | NZBIT32(x >> 32, 1);               \
 253     return x;                                                   \
 254 }                                                               \
 255 uint64_t HELPER(glue(iwmmxt_unpack, glue(S, sl)))(CPUARMState *env, \
 256                                                   uint64_t x)   \
 257 {                                                               \
 258     x = EXTEND32((x >> SH0) & 0xffffffff);                      \
 259     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] = NZBIT64(x >> 0);      \
 260     return x;                                                   \
 261 }
 262 IWMMXT_OP_UNPACK(l, 0, 8, 16, 24)
 263 IWMMXT_OP_UNPACK(h, 32, 40, 48, 56)
 264
 265 #define IWMMXT_OP_CMP(SUFF, Tb, Tw, Tl, O)                      \
 266 uint64_t HELPER(glue(iwmmxt_, glue(SUFF, b)))(CPUARMState *env,    \
 267                                         uint64_t a, uint64_t b) \
 268 {                                                               \
 269     a =                                                         \
 270         CMP(0, Tb, O, 0xff) | CMP(8, Tb, O, 0xff) |             \
 271         CMP(16, Tb, O, 0xff) | CMP(24, Tb, O, 0xff) |           \
 272         CMP(32, Tb, O, 0xff) | CMP(40, Tb, O, 0xff) |           \
 273         CMP(48, Tb, O, 0xff) | CMP(56, Tb, O, 0xff);            \
 274     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                       \
 275         NZBIT8(a >> 0, 0) | NZBIT8(a >> 8, 1) |                 \
 276         NZBIT8(a >> 16, 2) | NZBIT8(a >> 24, 3) |               \
 277         NZBIT8(a >> 32, 4) | NZBIT8(a >> 40, 5) |               \
 278         NZBIT8(a >> 48, 6) | NZBIT8(a >> 56, 7);                \
 279     return a;                                                   \
 280 }                                                               \
 281 uint64_t HELPER(glue(iwmmxt_, glue(SUFF, w)))(CPUARMState *env,    \
 282                                         uint64_t a, uint64_t b) \
 283 {                                                               \
 284     a = CMP(0, Tw, O, 0xffff) | CMP(16, Tw, O, 0xffff) |        \
 285         CMP(32, Tw, O, 0xffff) | CMP(48, Tw, O, 0xffff);        \
 286     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                       \
 287         NZBIT16(a >> 0, 0) | NZBIT16(a >> 16, 1) |              \
 288         NZBIT16(a >> 32, 2) | NZBIT16(a >> 48, 3);              \
 289     return a;                                                   \
 290 }                                                               \
 291 uint64_t HELPER(glue(iwmmxt_, glue(SUFF, l)))(CPUARMState *env,    \
 292                                         uint64_t a, uint64_t b) \
 293 {                                                               \
 294     a = CMP(0, Tl, O, 0xffffffff) |                             \
 295         CMP(32, Tl, O, 0xffffffff);                             \
 296     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                       \
 297         NZBIT32(a >> 0, 0) | NZBIT32(a >> 32, 1);               \
 298     return a;                                                   \
 299 }
 300 #define CMP(SHR, TYPE, OPER, MASK) ((((TYPE) ((a >> SHR) & MASK) OPER \
 301             (TYPE) ((b >> SHR) & MASK)) ? (uint64_t) MASK : 0) << SHR)
 302 IWMMXT_OP_CMP(cmpeq, uint8_t, uint16_t, uint32_t, ==)
 303 IWMMXT_OP_CMP(cmpgts, int8_t, int16_t, int32_t, >)
 304 IWMMXT_OP_CMP(cmpgtu, uint8_t, uint16_t, uint32_t, >)
 305 #undef CMP
 306 #define CMP(SHR, TYPE, OPER, MASK) ((((TYPE) ((a >> SHR) & MASK) OPER \
 307             (TYPE) ((b >> SHR) & MASK)) ? a : b) & ((uint64_t) MASK << SHR))
 308 IWMMXT_OP_CMP(mins, int8_t, int16_t, int32_t, <)
 309 IWMMXT_OP_CMP(minu, uint8_t, uint16_t, uint32_t, <)
 310 IWMMXT_OP_CMP(maxs, int8_t, int16_t, int32_t, >)
 311 IWMMXT_OP_CMP(maxu, uint8_t, uint16_t, uint32_t, >)
 312 #undef CMP
 313 #define CMP(SHR, TYPE, OPER, MASK) ((uint64_t) (((TYPE) ((a >> SHR) & MASK) \
 314             OPER (TYPE) ((b >> SHR) & MASK)) & MASK) << SHR)
 315 IWMMXT_OP_CMP(subn, uint8_t, uint16_t, uint32_t, -)
 316 IWMMXT_OP_CMP(addn, uint8_t, uint16_t, uint32_t, +)
 317 #undef CMP
 318 /* TODO Signed- and Unsigned-Saturation */
 319 #define CMP(SHR, TYPE, OPER, MASK) ((uint64_t) (((TYPE) ((a >> SHR) & MASK) \
 320             OPER (TYPE) ((b >> SHR) & MASK)) & MASK) << SHR)
 321 IWMMXT_OP_CMP(subu, uint8_t, uint16_t, uint32_t, -)
 322 IWMMXT_OP_CMP(addu, uint8_t, uint16_t, uint32_t, +)
 323 IWMMXT_OP_CMP(subs, int8_t, int16_t, int32_t, -)
 324 IWMMXT_OP_CMP(adds, int8_t, int16_t, int32_t, +)
 325 #undef CMP
 326 #undef IWMMXT_OP_CMP
 327
 328 #define AVGB(SHR) ((( \
 329         ((a >> SHR) & 0xff) + ((b >> SHR) & 0xff) + round) >> 1) << SHR)
 330 #define IWMMXT_OP_AVGB(r)                                                 \
 331 uint64_t HELPER(iwmmxt_avgb##r)(CPUARMState *env, uint64_t a, uint64_t b)    \
 332 {                                                                         \
 333     const int round = r;                                                  \
 334     a = AVGB(0) | AVGB(8) | AVGB(16) | AVGB(24) |                         \
 335         AVGB(32) | AVGB(40) | AVGB(48) | AVGB(56);                        \
 336     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                                 \
 337         SIMD8_SET(ZBIT8((a >> 0) & 0xff), SIMD_ZBIT, 0) |                 \
 338         SIMD8_SET(ZBIT8((a >> 8) & 0xff), SIMD_ZBIT, 1) |                 \
 339         SIMD8_SET(ZBIT8((a >> 16) & 0xff), SIMD_ZBIT, 2) |                \
 340         SIMD8_SET(ZBIT8((a >> 24) & 0xff), SIMD_ZBIT, 3) |                \
 341         SIMD8_SET(ZBIT8((a >> 32) & 0xff), SIMD_ZBIT, 4) |                \
 342         SIMD8_SET(ZBIT8((a >> 40) & 0xff), SIMD_ZBIT, 5) |                \
 343         SIMD8_SET(ZBIT8((a >> 48) & 0xff), SIMD_ZBIT, 6) |                \
 344         SIMD8_SET(ZBIT8((a >> 56) & 0xff), SIMD_ZBIT, 7);                 \
 345     return a;                                                             \
 346 }
 347 IWMMXT_OP_AVGB(0)
 348 IWMMXT_OP_AVGB(1)
 349 #undef IWMMXT_OP_AVGB
 350 #undef AVGB
 351
 352 #define AVGW(SHR) ((( \
 353         ((a >> SHR) & 0xffff) + ((b >> SHR) & 0xffff) + round) >> 1) << SHR)
 354 #define IWMMXT_OP_AVGW(r)                                               \
 355 uint64_t HELPER(iwmmxt_avgw##r)(CPUARMState *env, uint64_t a, uint64_t b)  \
 356 {                                                                       \
 357     const int round = r;                                                \
 358     a = AVGW(0) | AVGW(16) | AVGW(32) | AVGW(48);                       \
 359     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =                               \
 360         SIMD16_SET(ZBIT16((a >> 0) & 0xffff), SIMD_ZBIT, 0) |           \
 361         SIMD16_SET(ZBIT16((a >> 16) & 0xffff), SIMD_ZBIT, 1) |          \
 362         SIMD16_SET(ZBIT16((a >> 32) & 0xffff), SIMD_ZBIT, 2) |          \
 363         SIMD16_SET(ZBIT16((a >> 48) & 0xffff), SIMD_ZBIT, 3);           \
 364     return a;                                                           \
 365 }
 366 IWMMXT_OP_AVGW(0)
 367 IWMMXT_OP_AVGW(1)
 368 #undef IWMMXT_OP_AVGW
 369 #undef AVGW
 370
 371 uint64_t HELPER(iwmmxt_align)(uint64_t a, uint64_t b, uint32_t n)
 372 {
 373     a >>= n << 3;
 374     a |= b << (64 - (n << 3));
 375     return a;
 376 }
 377
 378 uint64_t HELPER(iwmmxt_insr)(uint64_t x, uint32_t a, uint32_t b, uint32_t n)
 379 {
 380     x &= ~((uint64_t) b << n);
 381     x |= (uint64_t) (a & b) << n;
 382     return x;
 383 }
 384
 385 uint32_t HELPER(iwmmxt_setpsr_nz)(uint64_t x)
 386 {
 387     return SIMD64_SET((x == 0), SIMD_ZBIT) |
 388            SIMD64_SET((x & (1ULL << 63)), SIMD_NBIT);
 389 }
 390
 391 uint64_t HELPER(iwmmxt_bcstb)(uint32_t arg)
 392 {
 393     arg &= 0xff;
 394     return
 395         ((uint64_t) arg << 0 ) | ((uint64_t) arg << 8 ) |
 396         ((uint64_t) arg << 16) | ((uint64_t) arg << 24) |
 397         ((uint64_t) arg << 32) | ((uint64_t) arg << 40) |
 398         ((uint64_t) arg << 48) | ((uint64_t) arg << 56);
 399 }
 400
 401 uint64_t HELPER(iwmmxt_bcstw)(uint32_t arg)
 402 {
 403     arg &= 0xffff;
 404     return
 405         ((uint64_t) arg << 0 ) | ((uint64_t) arg << 16) |
 406         ((uint64_t) arg << 32) | ((uint64_t) arg << 48);
 407 }
 408
 409 uint64_t HELPER(iwmmxt_bcstl)(uint32_t arg)
 410 {
 411     return arg | ((uint64_t) arg << 32);
 412 }
 413
 414 uint64_t HELPER(iwmmxt_addcb)(uint64_t x)
 415 {
 416     return
 417         ((x >> 0) & 0xff) + ((x >> 8) & 0xff) +
 418         ((x >> 16) & 0xff) + ((x >> 24) & 0xff) +
 419         ((x >> 32) & 0xff) + ((x >> 40) & 0xff) +
 420         ((x >> 48) & 0xff) + ((x >> 56) & 0xff);
 421 }
 422
 423 uint64_t HELPER(iwmmxt_addcw)(uint64_t x)
 424 {
 425     return
 426         ((x >> 0) & 0xffff) + ((x >> 16) & 0xffff) +
 427         ((x >> 32) & 0xffff) + ((x >> 48) & 0xffff);
 428 }
 429
 430 uint64_t HELPER(iwmmxt_addcl)(uint64_t x)
 431 {
 432     return (x & 0xffffffff) + (x >> 32);
 433 }
 434
 435 uint32_t HELPER(iwmmxt_msbb)(uint64_t x)
 436 {
 437     return
 438         ((x >> 7) & 0x01) | ((x >> 14) & 0x02) |
 439         ((x >> 21) & 0x04) | ((x >> 28) & 0x08) |
 440         ((x >> 35) & 0x10) | ((x >> 42) & 0x20) |
 441         ((x >> 49) & 0x40) | ((x >> 56) & 0x80);
 442 }
 443
 444 uint32_t HELPER(iwmmxt_msbw)(uint64_t x)
 445 {
 446     return
 447         ((x >> 15) & 0x01) | ((x >> 30) & 0x02) |
 448         ((x >> 45) & 0x04) | ((x >> 52) & 0x08);
 449 }
 450
 451 uint32_t HELPER(iwmmxt_msbl)(uint64_t x)
 452 {
 453     return ((x >> 31) & 0x01) | ((x >> 62) & 0x02);
 454 }
 455
 456 /* FIXME: Split wCASF setting into a separate op to avoid env use.  */
 457 uint64_t HELPER(iwmmxt_srlw)(CPUARMState *env, uint64_t x, uint32_t n)
 458 {
 459     x = (((x & (0xffffll << 0)) >> n) & (0xffffll << 0)) |
 460         (((x & (0xffffll << 16)) >> n) & (0xffffll << 16)) |
 461         (((x & (0xffffll << 32)) >> n) & (0xffffll << 32)) |
 462         (((x & (0xffffll << 48)) >> n) & (0xffffll << 48));
 463     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 464         NZBIT16(x >> 0, 0) | NZBIT16(x >> 16, 1) |
 465         NZBIT16(x >> 32, 2) | NZBIT16(x >> 48, 3);
 466     return x;
 467 }
 468
 469 uint64_t HELPER(iwmmxt_srll)(CPUARMState *env, uint64_t x, uint32_t n)
 470 {
 471     x = ((x & (0xffffffffll << 0)) >> n) |
 472         ((x >> n) & (0xffffffffll << 32));
 473     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 474         NZBIT32(x >> 0, 0) | NZBIT32(x >> 32, 1);
 475     return x;
 476 }
 477
 478 uint64_t HELPER(iwmmxt_srlq)(CPUARMState *env, uint64_t x, uint32_t n)
 479 {
 480     x >>= n;
 481     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] = NZBIT64(x);
 482     return x;
 483 }
 484
 485 uint64_t HELPER(iwmmxt_sllw)(CPUARMState *env, uint64_t x, uint32_t n)
 486 {
 487     x = (((x & (0xffffll << 0)) << n) & (0xffffll << 0)) |
 488         (((x & (0xffffll << 16)) << n) & (0xffffll << 16)) |
 489         (((x & (0xffffll << 32)) << n) & (0xffffll << 32)) |
 490         (((x & (0xffffll << 48)) << n) & (0xffffll << 48));
 491     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 492         NZBIT16(x >> 0, 0) | NZBIT16(x >> 16, 1) |
 493         NZBIT16(x >> 32, 2) | NZBIT16(x >> 48, 3);
 494     return x;
 495 }
 496
 497 uint64_t HELPER(iwmmxt_slll)(CPUARMState *env, uint64_t x, uint32_t n)
 498 {
 499     x = ((x << n) & (0xffffffffll << 0)) |
 500         ((x & (0xffffffffll << 32)) << n);
 501     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 502         NZBIT32(x >> 0, 0) | NZBIT32(x >> 32, 1);
 503     return x;
 504 }
 505
 506 uint64_t HELPER(iwmmxt_sllq)(CPUARMState *env, uint64_t x, uint32_t n)
 507 {
 508     x <<= n;
 509     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] = NZBIT64(x);
 510     return x;
 511 }
 512
 513 uint64_t HELPER(iwmmxt_sraw)(CPUARMState *env, uint64_t x, uint32_t n)
 514 {
 515     x = ((uint64_t) ((EXTEND16(x >> 0) >> n) & 0xffff) << 0) |
 516         ((uint64_t) ((EXTEND16(x >> 16) >> n) & 0xffff) << 16) |
 517         ((uint64_t) ((EXTEND16(x >> 32) >> n) & 0xffff) << 32) |
 518         ((uint64_t) ((EXTEND16(x >> 48) >> n) & 0xffff) << 48);
 519     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 520         NZBIT16(x >> 0, 0) | NZBIT16(x >> 16, 1) |
 521         NZBIT16(x >> 32, 2) | NZBIT16(x >> 48, 3);
 522     return x;
 523 }
 524
 525 uint64_t HELPER(iwmmxt_sral)(CPUARMState *env, uint64_t x, uint32_t n)
 526 {
 527     x = (((EXTEND32(x >> 0) >> n) & 0xffffffff) << 0) |
 528         (((EXTEND32(x >> 32) >> n) & 0xffffffff) << 32);
 529     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 530         NZBIT32(x >> 0, 0) | NZBIT32(x >> 32, 1);
 531     return x;
 532 }
 533
 534 uint64_t HELPER(iwmmxt_sraq)(CPUARMState *env, uint64_t x, uint32_t n)
 535 {
 536     x = (int64_t) x >> n;
 537     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] = NZBIT64(x);
 538     return x;
 539 }
 540
 541 uint64_t HELPER(iwmmxt_rorw)(CPUARMState *env, uint64_t x, uint32_t n)
 542 {
 543     x = ((((x & (0xffffll << 0)) >> n) |
 544           ((x & (0xffffll << 0)) << (16 - n))) & (0xffffll << 0)) |
 545         ((((x & (0xffffll << 16)) >> n) |
 546           ((x & (0xffffll << 16)) << (16 - n))) & (0xffffll << 16)) |
 547         ((((x & (0xffffll << 32)) >> n) |
 548           ((x & (0xffffll << 32)) << (16 - n))) & (0xffffll << 32)) |
 549         ((((x & (0xffffll << 48)) >> n) |
 550           ((x & (0xffffll << 48)) << (16 - n))) & (0xffffll << 48));
 551     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 552         NZBIT16(x >> 0, 0) | NZBIT16(x >> 16, 1) |
 553         NZBIT16(x >> 32, 2) | NZBIT16(x >> 48, 3);
 554     return x;
 555 }
 556
 557 uint64_t HELPER(iwmmxt_rorl)(CPUARMState *env, uint64_t x, uint32_t n)
 558 {
 559     x = ((x & (0xffffffffll << 0)) >> n) |
 560         ((x >> n) & (0xffffffffll << 32)) |
 561         ((x << (32 - n)) & (0xffffffffll << 0)) |
 562         ((x & (0xffffffffll << 32)) << (32 - n));
 563     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 564         NZBIT32(x >> 0, 0) | NZBIT32(x >> 32, 1);
 565     return x;
 566 }
 567
 568 uint64_t HELPER(iwmmxt_rorq)(CPUARMState *env, uint64_t x, uint32_t n)
 569 {
 570     x = ror64(x, n);
 571     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] = NZBIT64(x);
 572     return x;
 573 }
 574
 575 uint64_t HELPER(iwmmxt_shufh)(CPUARMState *env, uint64_t x, uint32_t n)
 576 {
 577     x = (((x >> ((n << 4) & 0x30)) & 0xffff) << 0) |
 578         (((x >> ((n << 2) & 0x30)) & 0xffff) << 16) |
 579         (((x >> ((n << 0) & 0x30)) & 0xffff) << 32) |
 580         (((x >> ((n >> 2) & 0x30)) & 0xffff) << 48);
 581     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 582         NZBIT16(x >> 0, 0) | NZBIT16(x >> 16, 1) |
 583         NZBIT16(x >> 32, 2) | NZBIT16(x >> 48, 3);
 584     return x;
 585 }
 586
 587 /* TODO: Unsigned-Saturation */
 588 uint64_t HELPER(iwmmxt_packuw)(CPUARMState *env, uint64_t a, uint64_t b)
 589 {
 590     a = (((a >> 0) & 0xff) << 0) | (((a >> 16) & 0xff) << 8) |
 591         (((a >> 32) & 0xff) << 16) | (((a >> 48) & 0xff) << 24) |
 592         (((b >> 0) & 0xff) << 32) | (((b >> 16) & 0xff) << 40) |
 593         (((b >> 32) & 0xff) << 48) | (((b >> 48) & 0xff) << 56);
 594     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 595         NZBIT8(a >> 0, 0) | NZBIT8(a >> 8, 1) |
 596         NZBIT8(a >> 16, 2) | NZBIT8(a >> 24, 3) |
 597         NZBIT8(a >> 32, 4) | NZBIT8(a >> 40, 5) |
 598         NZBIT8(a >> 48, 6) | NZBIT8(a >> 56, 7);
 599     return a;
 600 }
 601
 602 uint64_t HELPER(iwmmxt_packul)(CPUARMState *env, uint64_t a, uint64_t b)
 603 {
 604     a = (((a >> 0) & 0xffff) << 0) | (((a >> 32) & 0xffff) << 16) |
 605         (((b >> 0) & 0xffff) << 32) | (((b >> 32) & 0xffff) << 48);
 606     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 607         NZBIT16(a >> 0, 0) | NZBIT16(a >> 16, 1) |
 608         NZBIT16(a >> 32, 2) | NZBIT16(a >> 48, 3);
 609     return a;
 610 }
 611
 612 uint64_t HELPER(iwmmxt_packuq)(CPUARMState *env, uint64_t a, uint64_t b)
 613 {
 614     a = (a & 0xffffffff) | ((b & 0xffffffff) << 32);
 615     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 616         NZBIT32(a >> 0, 0) | NZBIT32(a >> 32, 1);
 617     return a;
 618 }
 619
 620 /* TODO: Signed-Saturation */
 621 uint64_t HELPER(iwmmxt_packsw)(CPUARMState *env, uint64_t a, uint64_t b)
 622 {
 623     a = (((a >> 0) & 0xff) << 0) | (((a >> 16) & 0xff) << 8) |
 624         (((a >> 32) & 0xff) << 16) | (((a >> 48) & 0xff) << 24) |
 625         (((b >> 0) & 0xff) << 32) | (((b >> 16) & 0xff) << 40) |
 626         (((b >> 32) & 0xff) << 48) | (((b >> 48) & 0xff) << 56);
 627     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 628         NZBIT8(a >> 0, 0) | NZBIT8(a >> 8, 1) |
 629         NZBIT8(a >> 16, 2) | NZBIT8(a >> 24, 3) |
 630         NZBIT8(a >> 32, 4) | NZBIT8(a >> 40, 5) |
 631         NZBIT8(a >> 48, 6) | NZBIT8(a >> 56, 7);
 632     return a;
 633 }
 634
 635 uint64_t HELPER(iwmmxt_packsl)(CPUARMState *env, uint64_t a, uint64_t b)
 636 {
 637     a = (((a >> 0) & 0xffff) << 0) | (((a >> 32) & 0xffff) << 16) |
 638         (((b >> 0) & 0xffff) << 32) | (((b >> 32) & 0xffff) << 48);
 639     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 640         NZBIT16(a >> 0, 0) | NZBIT16(a >> 16, 1) |
 641         NZBIT16(a >> 32, 2) | NZBIT16(a >> 48, 3);
 642     return a;
 643 }
 644
 645 uint64_t HELPER(iwmmxt_packsq)(CPUARMState *env, uint64_t a, uint64_t b)
 646 {
 647     a = (a & 0xffffffff) | ((b & 0xffffffff) << 32);
 648     env->iwmmxt.cregs[ARM_IWMMXT_wCASF] =
 649         NZBIT32(a >> 0, 0) | NZBIT32(a >> 32, 1);
 650     return a;
 651 }
 652
 653 uint64_t HELPER(iwmmxt_muladdsl)(uint64_t c, uint32_t a, uint32_t b)
 654 {
 655     return c + ((int32_t) EXTEND32(a) * (int32_t) EXTEND32(b));
 656 }
 657
 658 uint64_t HELPER(iwmmxt_muladdsw)(uint64_t c, uint32_t a, uint32_t b)
 659 {
 660     c += EXTEND32(EXTEND16S((a >> 0) & 0xffff) *
 661                   EXTEND16S((b >> 0) & 0xffff));
 662     c += EXTEND32(EXTEND16S((a >> 16) & 0xffff) *
 663                   EXTEND16S((b >> 16) & 0xffff));
 664     return c;
 665 }
 666
 667 uint64_t HELPER(iwmmxt_muladdswl)(uint64_t c, uint32_t a, uint32_t b)
 668 {
 669     return c + (EXTEND32(EXTEND16S(a & 0xffff) *
 670                          EXTEND16S(b & 0xffff)));
 671 }