op_neon.h.svn-base

我们自己开发的一个OSEK操作系统！不知道可不可以？

/*
 * ARM NEON vector operations.
 *
 * Copyright (c) 2007 CodeSourcery.
 * Written by Paul Brook
 *
 * This code is licenced under the GPL.
 */
/* Note that for NEON an "l" prefix means it is a wide operation, unlike
   scalar arm ops where it means a word size operation.  */

/* ??? NEON ops should probably have their own float status.  */
#define NFS &env->vfp.fp_status
#define NEON_OP(name) void OPPROTO op_neon_##name (void)

NEON_OP(getreg_T0)
{
    T0 = *(uint32_t *)((char *) env + PARAM1);
}

NEON_OP(getreg_T1)
{
    T1 = *(uint32_t *)((char *) env + PARAM1);
}

NEON_OP(getreg_T2)
{
    T2 = *(uint32_t *)((char *) env + PARAM1);
}

NEON_OP(setreg_T0)
{
    *(uint32_t *)((char *) env + PARAM1) = T0;
}

NEON_OP(setreg_T1)
{
    *(uint32_t *)((char *) env + PARAM1) = T1;
}

NEON_OP(setreg_T2)
{
    *(uint32_t *)((char *) env + PARAM1) = T2;
}

#define NEON_TYPE1(name, type) \
typedef struct \
{ \
    type v1; \
} neon_##name;
#ifdef WORDS_BIGENDIAN
#define NEON_TYPE2(name, type) \
typedef struct \
{ \
    type v2; \
    type v1; \
} neon_##name;
#define NEON_TYPE4(name, type) \
typedef struct \
{ \
    type v4; \
    type v3; \
    type v2; \
    type v1; \
} neon_##name;
#else
#define NEON_TYPE2(name, type) \
typedef struct \
{ \
    type v1; \
    type v2; \
} neon_##name;
#define NEON_TYPE4(name, type) \
typedef struct \
{ \
    type v1; \
    type v2; \
    type v3; \
    type v4; \
} neon_##name;
#endif

NEON_TYPE4(s8, int8_t)
NEON_TYPE4(u8, uint8_t)
NEON_TYPE2(s16, int16_t)
NEON_TYPE2(u16, uint16_t)
NEON_TYPE1(s32, int32_t)
NEON_TYPE1(u32, uint32_t)
#undef NEON_TYPE4
#undef NEON_TYPE2
#undef NEON_TYPE1

/* Copy from a uint32_t to a vector structure type.  */
#define NEON_UNPACK(vtype, dest, val) do { \
    union { \
        vtype v; \
        uint32_t i; \
    } conv_u; \
    conv_u.i = (val); \
    dest = conv_u.v; \
    } while(0)

/* Copy from a vector structure type to a uint32_t.  */
#define NEON_PACK(vtype, dest, val) do { \
    union { \
        vtype v; \
        uint32_t i; \
    } conv_u; \
    conv_u.v = (val); \
    dest = conv_u.i; \
    } while(0)

#define NEON_DO1 \
    NEON_FN(vdest.v1, vsrc1.v1, vsrc2.v1);
#define NEON_DO2 \
    NEON_FN(vdest.v1, vsrc1.v1, vsrc2.v1); \
    NEON_FN(vdest.v2, vsrc1.v2, vsrc2.v2);
#define NEON_DO4 \
    NEON_FN(vdest.v1, vsrc1.v1, vsrc2.v1); \
    NEON_FN(vdest.v2, vsrc1.v2, vsrc2.v2); \
    NEON_FN(vdest.v3, vsrc1.v3, vsrc2.v3); \
    NEON_FN(vdest.v4, vsrc1.v4, vsrc2.v4);

#define NEON_VOP(name, vtype, n) \
NEON_OP(name) \
{ \
    vtype vsrc1; \
    vtype vsrc2; \
    vtype vdest; \
    NEON_UNPACK(vtype, vsrc1, T0); \
    NEON_UNPACK(vtype, vsrc2, T1); \
    NEON_DO##n; \
    NEON_PACK(vtype, T0, vdest); \
    FORCE_RET(); \
}

#define NEON_VOP1(name, vtype, n) \
NEON_OP(name) \
{ \
    vtype vsrc1; \
    vtype vdest; \
    NEON_UNPACK(vtype, vsrc1, T0); \
    NEON_DO##n; \
    NEON_PACK(vtype, T0, vdest); \
    FORCE_RET(); \
}

/* Pairwise operations.  */
/* For 32-bit elements each segment only contains a single element, so
   the elementwise and pairwise operations are the same.  */
#define NEON_PDO2 \
    NEON_FN(vdest.v1, vsrc1.v1, vsrc1.v2); \
    NEON_FN(vdest.v2, vsrc2.v1, vsrc2.v2);
#define NEON_PDO4 \
    NEON_FN(vdest.v1, vsrc1.v1, vsrc1.v2); \
    NEON_FN(vdest.v2, vsrc1.v3, vsrc1.v4); \
    NEON_FN(vdest.v3, vsrc2.v1, vsrc2.v2); \
    NEON_FN(vdest.v4, vsrc2.v3, vsrc2.v4); \

#define NEON_POP(name, vtype, n) \
NEON_OP(name) \
{ \
    vtype vsrc1; \
    vtype vsrc2; \
    vtype vdest; \
    NEON_UNPACK(vtype, vsrc1, T0); \
    NEON_UNPACK(vtype, vsrc2, T1); \
    NEON_PDO##n; \
    NEON_PACK(vtype, T0, vdest); \
    FORCE_RET(); \
}

#define NEON_FN(dest, src1, src2) dest = (src1 + src2) >> 1
NEON_VOP(hadd_s8, neon_s8, 4)
NEON_VOP(hadd_u8, neon_u8, 4)
NEON_VOP(hadd_s16, neon_s16, 2)
NEON_VOP(hadd_u16, neon_u16, 2)
#undef NEON_FN

NEON_OP(hadd_s32)
{
    int32_t src1 = T0;
    int32_t src2 = T1;
    int32_t dest;

    dest = (src1 >> 1) + (src2 >> 1);
    if (src1 & src2 & 1)
        dest++;
    T0 = dest;
    FORCE_RET();
}

NEON_OP(hadd_u32)
{
    uint32_t src1 = T0;
    uint32_t src2 = T1;
    uint32_t dest;

    dest = (src1 >> 1) + (src2 >> 1);
    if (src1 & src2 & 1)
        dest++;
    T0 = dest;
    FORCE_RET();
}

#define NEON_FN(dest, src1, src2) dest = (src1 + src2 + 1) >> 1
NEON_VOP(rhadd_s8, neon_s8, 4)
NEON_VOP(rhadd_u8, neon_u8, 4)
NEON_VOP(rhadd_s16, neon_s16, 2)
NEON_VOP(rhadd_u16, neon_u16, 2)
#undef NEON_FN

NEON_OP(rhadd_s32)
{
    int32_t src1 = T0;
    int32_t src2 = T1;
    int32_t dest;

    dest = (src1 >> 1) + (src2 >> 1);
    if ((src1 | src2) & 1)
        dest++;
    T0 = dest;
    FORCE_RET();
}

NEON_OP(rhadd_u32)
{
    uint32_t src1 = T0;
    uint32_t src2 = T1;
    uint32_t dest;

    dest = (src1 >> 1) + (src2 >> 1);
    if ((src1 | src2) & 1)
        dest++;
    T0 = dest;
    FORCE_RET();
}

#define NEON_FN(dest, src1, src2) dest = (src1 - src2) >> 1
NEON_VOP(hsub_s8, neon_s8, 4)
NEON_VOP(hsub_u8, neon_u8, 4)
NEON_VOP(hsub_s16, neon_s16, 2)
NEON_VOP(hsub_u16, neon_u16, 2)
#undef NEON_FN

NEON_OP(hsub_s32)
{
    int32_t src1 = T0;
    int32_t src2 = T1;
    int32_t dest;

    dest = (src1 >> 1) - (src2 >> 1);
    if ((~src1) & src2 & 1)
        dest--;
    T0 = dest;
    FORCE_RET();
}

NEON_OP(hsub_u32)
{
    uint32_t src1 = T0;
    uint32_t src2 = T1;
    uint32_t dest;

    dest = (src1 >> 1) - (src2 >> 1);
    if ((~src1) & src2 & 1)
        dest--;
    T0 = dest;
    FORCE_RET();
}

/* ??? bsl, bif and bit are all the same op, just with the oparands in a
   differnet order.  It's currently easier to have 3 differnt ops than
   rearange the operands.  */

/* Bitwise Select.  */
NEON_OP(bsl)
{
    T0 = (T0 & T2) | (T1 & ~T2);
}

/* Bitwise Insert If True.  */
NEON_OP(bit)
{
    T0 = (T0 & T1) | (T2 & ~T1);
}

/* Bitwise Insert If False.  */
NEON_OP(bif)
{
    T0 = (T2 & T1) | (T0 & ~T1);
}

#define NEON_USAT(dest, src1, src2, type) do { \
    uint32_t tmp = (uint32_t)src1 + (uint32_t)src2; \
    if (tmp != (type)tmp) { \
        env->QF = 1; \
        dest = ~0; \
    } else { \
        dest = tmp; \
    }} while(0)
#define NEON_FN(dest, src1, src2) NEON_USAT(dest, src1, src2, uint8_t)
NEON_VOP(qadd_u8, neon_u8, 4)
#undef NEON_FN
#define NEON_FN(dest, src1, src2) NEON_USAT(dest, src1, src2, uint16_t)
NEON_VOP(qadd_u16, neon_u16, 2)
#undef NEON_FN
#undef NEON_USAT

#define NEON_SSAT(dest, src1, src2, type) do { \
    int32_t tmp = (uint32_t)src1 + (uint32_t)src2; \
    if (tmp != (type)tmp) { \
        env->QF = 1; \
        if (src2 > 0) { \
            tmp = (1 << (sizeof(type) * 8 - 1)) - 1; \
        } else { \
            tmp = 1 << (sizeof(type) * 8 - 1); \
        } \
    } \
    dest = tmp; \
    } while(0)
#define NEON_FN(dest, src1, src2) NEON_SSAT(dest, src1, src2, int8_t)
NEON_VOP(qadd_s8, neon_s8, 4)
#undef NEON_FN
#define NEON_FN(dest, src1, src2) NEON_SSAT(dest, src1, src2, int16_t)
NEON_VOP(qadd_s16, neon_s16, 2)
#undef NEON_FN
#undef NEON_SSAT

#define NEON_USAT(dest, src1, src2, type) do { \
    uint32_t tmp = (uint32_t)src1 - (uint32_t)src2; \
    if (tmp != (type)tmp) { \
        env->QF = 1; \
        dest = 0; \
    } else { \
        dest = tmp; \
    }} while(0)
#define NEON_FN(dest, src1, src2) NEON_USAT(dest, src1, src2, uint8_t)
NEON_VOP(qsub_u8, neon_u8, 4)
#undef NEON_FN
#define NEON_FN(dest, src1, src2) NEON_USAT(dest, src1, src2, uint16_t)
NEON_VOP(qsub_u16, neon_u16, 2)
#undef NEON_FN
#undef NEON_USAT

#define NEON_SSAT(dest, src1, src2, type) do { \
    int32_t tmp = (uint32_t)src1 - (uint32_t)src2; \
    if (tmp != (type)tmp) { \
        env->QF = 1; \
        if (src2 < 0) { \
            tmp = (1 << (sizeof(type) * 8 - 1)) - 1; \
        } else { \
            tmp = 1 << (sizeof(type) * 8 - 1); \
        } \
    } \
    dest = tmp; \
    } while(0)
#define NEON_FN(dest, src1, src2) NEON_SSAT(dest, src1, src2, int8_t)
NEON_VOP(qsub_s8, neon_s8, 4)
#undef NEON_FN
#define NEON_FN(dest, src1, src2) NEON_SSAT(dest, src1, src2, int16_t)
NEON_VOP(qsub_s16, neon_s16, 2)
#undef NEON_FN
#undef NEON_SSAT

#define NEON_FN(dest, src1, src2) dest = (src1 > src2) ? ~0 : 0
NEON_VOP(cgt_s8, neon_s8, 4)
NEON_VOP(cgt_u8, neon_u8, 4)
NEON_VOP(cgt_s16, neon_s16, 2)
NEON_VOP(cgt_u16, neon_u16, 2)
NEON_VOP(cgt_s32, neon_s32, 1)
NEON_VOP(cgt_u32, neon_u32, 1)
#undef NEON_FN

#define NEON_FN(dest, src1, src2) dest = (src1 >= src2) ? ~0 : 0
NEON_VOP(cge_s8, neon_s8, 4)
NEON_VOP(cge_u8, neon_u8, 4)
NEON_VOP(cge_s16, neon_s16, 2)
NEON_VOP(cge_u16, neon_u16, 2)
NEON_VOP(cge_s32, neon_s32, 1)
NEON_VOP(cge_u32, neon_u32, 1)
#undef NEON_FN

#define NEON_FN(dest, src1, src2) do { \
    int8_t tmp; \
    tmp = (int8_t)src2; \
    if (tmp < 0) { \
        dest = src1 >> -tmp; \
    } else { \
        dest = src1 << tmp; \
    }} while (0)
NEON_VOP(shl_s8, neon_s8, 4)
NEON_VOP(shl_u8, neon_u8, 4)
NEON_VOP(shl_s16, neon_s16, 2)
NEON_VOP(shl_u16, neon_u16, 2)
NEON_VOP(shl_s32, neon_s32, 1)
NEON_VOP(shl_u32, neon_u32, 1)
#undef NEON_FN

NEON_OP(shl_u64)
{
    int8_t shift = T2;
    uint64_t val = T0 | ((uint64_t)T1 << 32);
    if (shift < 0) {
        val >>= -shift;
    } else {
        val <<= shift;
    }
    T0 = val;
    T1 = val >> 32;
    FORCE_RET();
}

NEON_OP(shl_s64)
{
    int8_t shift = T2;
    int64_t val = T0 | ((uint64_t)T1 << 32);
    if (shift < 0) {
        val >>= -shift;
    } else {
        val <<= shift;
    }
    T0 = val;
    T1 = val >> 32;
    FORCE_RET();
}

#define NEON_FN(dest, src1, src2) do { \
    int8_t tmp; \
    tmp = (int8_t)src1; \
    if (tmp < 0) { \
        dest = (src2 + (1 << (-1 - tmp))) >> -tmp; \
    } else { \
        dest = src2 << tmp; \
    }} while (0)

NEON_VOP(rshl_s8, neon_s8, 4)
NEON_VOP(rshl_u8, neon_u8, 4)
NEON_VOP(rshl_s16, neon_s16, 2)
NEON_VOP(rshl_u16, neon_u16, 2)
NEON_VOP(rshl_s32, neon_s32, 1)
NEON_VOP(rshl_u32, neon_u32, 1)
#undef NEON_FN

NEON_OP(rshl_u64)
{
    int8_t shift = T2;
    uint64_t val = T0 | ((uint64_t)T1 << 32);
    if (shift < 0) {
        val = (val + ((uint64_t)1 << (-1 - shift))) >> -shift;
        val >>= -shift;
    } else {
        val <<= shift;
    }
    T0 = val;
    T1 = val >> 32;
    FORCE_RET();
}

NEON_OP(rshl_s64)
{
    int8_t shift = T2;
    int64_t val = T0 | ((uint64_t)T1 << 32);
    if (shift < 0) {
        val = (val + ((int64_t)1 << (-1 - shift))) >> -shift;
    } else {
        val <<= shift;
    }
    T0 = val;
    T1 = val >> 32;
    FORCE_RET();
}

#define NEON_FN(dest, src1, src2) do { \
    int8_t tmp; \
    tmp = (int8_t)src1; \
    if (tmp < 0) { \
        dest = src2 >> -tmp; \
    } else { \
        dest = src2 << tmp; \
        if ((dest >> tmp) != src2) { \
            env->QF = 1; \
            dest = ~0; \
        } \
    }} while (0)
NEON_VOP(qshl_s8, neon_s8, 4)
NEON_VOP(qshl_s16, neon_s16, 2)
NEON_VOP(qshl_s32, neon_s32, 1)
#undef NEON_FN

NEON_OP(qshl_s64)
{
    int8_t shift = T2;
    int64_t val = T0 | ((uint64_t)T1 << 32);
    if (shift < 0) {
        val >>= -shift;
    } else {
        int64_t tmp = val;
        val <<= shift;
        if ((val >> shift) != tmp) {
            env->QF = 1;
            val = (tmp >> 63) ^ 0x7fffffffffffffffULL;
        }
    }
    T0 = val;
    T1 = val >> 32;
    FORCE_RET();
}

#define NEON_FN(dest, src1, src2) do { \
    int8_t tmp; \
    tmp = (int8_t)src1; \
    if (tmp < 0) { \
        dest = src2 >> -tmp; \
    } else { \
        dest = src2 << tmp; \
        if ((dest >> tmp) != src2) { \
            env->QF = 1; \
            dest = src2 >> 31; \
        } \
    }} while (0)
NEON_VOP(qshl_u8, neon_u8, 4)
NEON_VOP(qshl_u16, neon_u16, 2)
NEON_VOP(qshl_u32, neon_u32, 1)
#undef NEON_FN

NEON_OP(qshl_u64)
{
    int8_t shift = T2;
    uint64_t val = T0 | ((uint64_t)T1 << 32);
    if (shift < 0) {
        val >>= -shift;
    } else {
        uint64_t tmp = val;
        val <<= shift;
        if ((val >> shift) != tmp) {
            env->QF = 1;
            val = ~(uint64_t)0;
        }
    }
    T0 = val;
    T1 = val >> 32;
    FORCE_RET();
}

#define NEON_FN(dest, src1, src2) do { \
    int8_t tmp; \
    tmp = (int8_t)src1; \
    if (tmp < 0) { \
        dest = (src2 + (1 << (-1 - tmp))) >> -tmp; \
    } else { \
        dest = src2 << tmp; \
        if ((dest >> tmp) != src2) { \
            dest = ~0; \
        } \
    }} while (0)
NEON_VOP(qrshl_s8, neon_s8, 4)
NEON_VOP(qrshl_s16, neon_s16, 2)
NEON_VOP(qrshl_s32, neon_s32, 1)
#undef NEON_FN

#define NEON_FN(dest, src1, src2) do { \
    int8_t tmp; \
    tmp = (int8_t)src1; \
    if (tmp < 0) { \
        dest = (src2 + (1 << (-1 - tmp))) >> -tmp; \
    } else { \
        dest = src2 << tmp; \
        if ((dest >> tmp) != src2) { \
            env->QF = 1; \
            dest = src2 >> 31; \
        } \
    }} while (0)
NEON_VOP(qrshl_u8, neon_u8, 4)
NEON_VOP(qrshl_u16, neon_u16, 2)
NEON_VOP(qrshl_u32, neon_u32, 1)
#undef NEON_FN

#define NEON_FN(dest, src1, src2) dest = (src1 > src2) ? src1 : src2
NEON_VOP(max_s8, neon_s8, 4)
NEON_VOP(max_u8, neon_u8, 4)
NEON_VOP(max_s16, neon_s16, 2)
NEON_VOP(max_u16, neon_u16, 2)
NEON_VOP(max_s32, neon_s32, 1)
NEON_VOP(max_u32, neon_u32, 1)