qemu/target-arm/neon_helper.c

/*
 * ARM NEON vector operations.
 *
 * Copyright (c) 2007, 2008 CodeSourcery.
 * Written by Paul Brook
 *
 * This code is licenced under the GNU GPL v2.
 */
#include <stdlib.h>
#include <stdio.h>

#include "cpu.h"
#include "exec.h"
#include "helper.h"

#define SIGNBIT (uint32_t)0x80000000
#define SIGNBIT64 ((uint64_t)1 << 63)

#define SET_QC() env->vfp.xregs[ARM_VFP_FPSCR] = CPSR_Q

#define NFS (&env->vfp.standard_fp_status)

#define NEON_TYPE1(name, type) \
typedef struct \
{ \
    type v1; \
} neon_##name;
#ifdef HOST_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_BODY(vtype, n) \
{ \
    uint32_t res; \
    vtype vsrc1; \
    vtype vsrc2; \
    vtype vdest; \
    NEON_UNPACK(vtype, vsrc1, arg1); \
    NEON_UNPACK(vtype, vsrc2, arg2); \
    NEON_DO##n; \
    NEON_PACK(vtype, res, vdest); \
    return res; \
}

#define NEON_VOP(name, vtype, n) \
uint32_t HELPER(glue(neon_,name))(uint32_t arg1, uint32_t arg2) \
NEON_VOP_BODY(vtype, n)

/* 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) \
uint32_t HELPER(glue(neon_,name))(uint32_t arg1, uint32_t arg2) \
{ \
    uint32_t res; \
    vtype vsrc1; \
    vtype vsrc2; \
    vtype vdest; \
    NEON_UNPACK(vtype, vsrc1, arg1); \
    NEON_UNPACK(vtype, vsrc2, arg2); \
    NEON_PDO##n; \
    NEON_PACK(vtype, res, vdest); \
    return res; \
}

/* Unary operators.  */
#define NEON_VOP1(name, vtype, n) \
uint32_t HELPER(glue(neon_,name))(uint32_t arg) \
{ \
    vtype vsrc1; \
    vtype vdest; \
    NEON_UNPACK(vtype, vsrc1, arg); \
    NEON_DO##n; \
    NEON_PACK(vtype, arg, vdest); \
    return arg; \
}


#define NEON_USAT(dest, src1, src2, type) do { \
    uint32_t tmp = (uint32_t)src1 + (uint32_t)src2; \
    if (tmp != (type)tmp) { \
        SET_QC(); \
        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

uint32_t HELPER(neon_qadd_u32)(uint32_t a, uint32_t b)
{
    uint32_t res = a + b;
    if (res < a) {
        SET_QC();
        res = ~0;
    }
    return res;
}

uint64_t HELPER(neon_qadd_u64)(uint64_t src1, uint64_t src2)
{
    uint64_t res;

    res = src1 + src2;
    if (res < src1) {
        SET_QC();
        res = ~(uint64_t)0;
    }
    return res;
}

#define NEON_SSAT(dest, src1, src2, type) do { \
    int32_t tmp = (uint32_t)src1 + (uint32_t)src2; \
    if (tmp != (type)tmp) { \
        SET_QC(); \
        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

uint32_t HELPER(neon_qadd_s32)(uint32_t a, uint32_t b)
{
    uint32_t res = a + b;
    if (((res ^ a) & SIGNBIT) && !((a ^ b) & SIGNBIT)) {
        SET_QC();
        res = ~(((int32_t)a >> 31) ^ SIGNBIT);
    }
    return res;
}

uint64_t HELPER(neon_qadd_s64)(uint64_t src1, uint64_t src2)
{
    uint64_t res;

    res = src1 + src2;
    if (((res ^ src1) & SIGNBIT64) && !((src1 ^ src2) & SIGNBIT64)) {
        SET_QC();
        res = ((int64_t)src1 >> 63) ^ ~SIGNBIT64;
    }
    return res;
}

#define NEON_USAT(dest, src1, src2, type) do { \
    uint32_t tmp = (uint32_t)src1 - (uint32_t)src2; \
    if (tmp != (type)tmp) { \
        SET_QC(); \
        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

uint32_t HELPER(neon_qsub_u32)(uint32_t a, uint32_t b)
{
    uint32_t res = a - b;
    if (res > a) {
        SET_QC();
        res = 0;
    }
    return res;
}

uint64_t HELPER(neon_qsub_u64)(uint64_t src1, uint64_t src2)
{
    uint64_t res;

    if (src1 < src2) {
        SET_QC();
        res = 0;
    } else {
        res = src1 - src2;
    }
    return res;
}

#define NEON_SSAT(dest, src1, src2, type) do { \
    int32_t tmp = (uint32_t)src1 - (uint32_t)src2; \
    if (tmp != (type)tmp) { \
        SET_QC(); \
        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

uint32_t HELPER(neon_qsub_s32)(uint32_t a, uint32_t b)
{
    uint32_t res = a - b;
    if (((res ^ a) & SIGNBIT) && ((a ^ b) & SIGNBIT)) {
        SET_QC();
        res = ~(((int32_t)a >> 31) ^ SIGNBIT);
    }
    return res;
}

uint64_t HELPER(neon_qsub_s64)(uint64_t src1, uint64_t src2)
{
    uint64_t res;

    res = src1 - src2;
    if (((res ^ src1) & SIGNBIT64) && ((src1 ^ src2) & SIGNBIT64)) {
        SET_QC();
        res = ((int64_t)src1 >> 63) ^ ~SIGNBIT64;
    }
    return res;
}

#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

int32_t HELPER(neon_hadd_s32)(int32_t src1, int32_t src2)
{
    int32_t dest;

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

uint32_t HELPER(neon_hadd_u32)(uint32_t src1, uint32_t src2)
{
    uint32_t dest;

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

#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

int32_t HELPER(neon_rhadd_s32)(int32_t src1, int32_t src2)
{
    int32_t dest;

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

uint32_t HELPER(neon_rhadd_u32)(uint32_t src1, uint32_t src2)
{
    uint32_t dest;

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

#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

int32_t HELPER(neon_hsub_s32)(int32_t src1, int32_t src2)
{
    int32_t dest;

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

uint32_t HELPER(neon_hsub_u32)(uint32_t src1, uint32_t src2)
{
    uint32_t dest;

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

#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) dest = (src1 < src2) ? src1 : src2
NEON_VOP(min_s8, neon_s8, 4)
NEON_VOP(min_u8, neon_u8, 4)
NEON_VOP(min_s16, neon_s16, 2)
NEON_VOP(min_u16, neon_u16, 2)
NEON_VOP(min_s32, neon_s32, 1)
NEON_VOP(min_u32, neon_u32, 1)
NEON_POP(pmin_s8, neon_s8, 4)
NEON_POP(pmin_u8, neon_u8, 4)
NEON_POP(pmin_s16, neon_s16, 2)
NEON_POP(pmin_u16, neon_u16, 2)
#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)
NEON_POP(pmax_s8, neon_s8, 4)
NEON_POP(pmax_u8, neon_u8, 4)
NEON_POP(pmax_s16, neon_s16, 2)
NEON_POP(pmax_u16, neon_u16, 2)
#undef NEON_FN

#define NEON_FN(dest, src1, src2) \
    dest = (src1 > src2) ? (src1 - src2) : (src2 - src1)
NEON_VOP(abd_s8, neon_s8, 4)
NEON_VOP(abd_u8, neon_u8, 4)
NEON_VOP(abd_s16, neon_s16, 2)
NEON_VOP(abd_u16, neon_u16, 2)
NEON_VOP(abd_s32, neon_s32, 1)
NEON_VOP(abd_u32, neon_u32, 1)
#undef NEON_FN

#define NEON_FN(dest, src1, src2) do { \
    int8_t tmp; \
    tmp = (int8_t)src2; \
    if (tmp >= (ssize_t)sizeof(src1) * 8 || \
        tmp <= -(ssize_t)sizeof(src1) * 8) { \
        dest = 0; \
    } else if (tmp < 0) { \
        dest = src1 >> -tmp; \
    } else { \
        dest = src1 << tmp; \
    }} while (0)
NEON_VOP(shl_u8, neon_u8, 4)
NEON_VOP(shl_u16, neon_u16, 2)
NEON_VOP(shl_u32, neon_u32, 1)
#undef NEON_FN

uint64_t HELPER(neon_shl_u64)(uint64_t val, uint64_t shiftop)
{
    int8_t shift = (int8_t)shiftop;
    if (shift >= 64 || shift <= -64) {
        val = 0;
    } else if (shift < 0) {
        val >>= -shift;
    } else {
        val <<= shift;
    }
    return val;
}

#define NEON_FN(dest, src1, src2) do { \
    int8_t tmp; \
    tmp = (int8_t)src2; \
    if (tmp >= (ssize_t)sizeof(src1) * 8) { \
        dest = 0; \
    } else if (tmp <= -(ssize_t)sizeof(src1) * 8) { \
        dest = src1 >> (sizeof(src1) * 8 - 1); \
    } else if (tmp < 0) { \
        dest = src1 >> -tmp; \
    } else { \
        dest = src1 << tmp; \
    }} while (0)
NEON_VOP(shl_s8, neon_s8, 4)
NEON_VOP(shl_s16, neon_s16, 2)
NEON_VOP(shl_s32, neon_s32, 1)
#undef NEON_FN

uint64_t HELPER(neon_shl_s64)(uint64_t valop, uint64_t shiftop)
{
    int8_t shift = (int8_t)shiftop;
    int64_t val = valop;
    if (shift >= 64) {
        val = 0;
    } else if (shift <= -64) {
        val >>= 63;
    } else if (shift < 0) {
        val >>= -shift;
    } else {
        val <<= shift;
    }
    return val;
}

#define NEON_FN(dest, src1, src2) do { \
    int8_t tmp; \
    tmp = (int8_t)src2; \
    if ((tmp >= (ssize_t)sizeof(src1) * 8) \
        || (tmp <= -(ssize_t)sizeof(src1) * 8)) { \
        dest = 0; \
    } else if (tmp < 0) { \
        dest = (src1 + (1 << (-1 - tmp))) >> -tmp; \
    } else { \
        dest = src1 << tmp; \
    }} while (0)
NEON_VOP(rshl_s8, neon_s8, 4)
NEON_VOP(rshl_s16, neon_s16, 2)
#undef NEON_FN

/* The addition of the rounding constant may overflow, so we use an
 * intermediate 64 bits accumulator.  */
uint32_t HELPER(neon_rshl_s32)(uint32_t valop, uint32_t shiftop)
{
    int32_t dest;
    int32_t val = (int32_t)valop;
    int8_t shift = (int8_t)shiftop;
    if ((shift >= 32) || (shift <= -32)) {
        dest = 0;
    } else if (shift < 0) {
        int64_t big_dest = ((int64_t)val + (1 << (-1 - shift)));
        dest = big_dest >> -shift;
    } else {
        dest = val << shift;
    }
    return dest;
}

/* Handling addition overflow with 64 bits inputs values is more
 * tricky than with 32 bits values.  */
uint64_t HELPER(neon_rshl_s64)(uint64_t valop, uint64_t shiftop)
{
    int8_t shift = (int8_t)shiftop;
    int64_t val = valop;
    if ((shift >= 64) || (shift <= -64)) {
        val = 0;
    } else if (shift < 0) {
        val >>= (-shift - 1);
        if (val == INT64_MAX) {
            /* In this case, it means that the rounding constant is 1,
             * and the addition would overflow. Return the actual
             * result directly.  */
            val = 0x4000000000000000LL;
        } else {
            val++;
            val >>= 1;
        }
    } else {
        val <<= shift;
    }
    return val;
}

#define NEON_FN(dest, src1, src2) do { \
    int8_t tmp; \
    tmp = (int8_t)src2; \
    if (tmp >= (ssize_t)sizeof(src1) * 8 || \
        tmp < -(ssize_t)sizeof(src1) * 8) { \
        dest = 0; \
    } else if (tmp == -(ssize_t)sizeof(src1) * 8) { \
        dest = src1 >> (-tmp - 1); \
    } else if (tmp < 0) { \
        dest = (src1 + (1 << (-1 - tmp))) >> -tmp; \
    } else { \
        dest = src1 << tmp; \
    }} while (0)
NEON_VOP(rshl_u8, neon_u8, 4)
NEON_VOP(rshl_u16, neon_u16, 2)
#undef NEON_FN

/* The addition of the rounding constant may overflow, so we use an
 * intermediate 64 bits accumulator.  */
uint32_t HELPER(neon_rshl_u32)(uint32_t val, uint32_t shiftop)
{
    uint32_t dest;
    int8_t shift = (int8_t)shiftop;
    if (shift >= 32 || shift < -32) {
        dest = 0;
    } else if (shift == -32) {
        dest = val >> 31;
    } else if (shift < 0) {
        uint64_t big_dest = ((uint64_t)val + (1 << (-1 - shift)));
        dest = big_dest >> -shift;
    } else {
        dest = val << shift;
    }
    return dest;
}

/* Handling addition overflow with 64 bits inputs values is more
 * tricky than with 32 bits values.  */
uint64_t HELPER(neon_rshl_u64)(uint64_t val, uint64_t shiftop)
{
    int8_t shift = (uint8_t)shiftop;
    if (shift >= 64 || shift < -64) {
        val = 0;
    } else if (shift == -64) {
        /* Rounding a 1-bit result just preserves that bit.  */
        val >>= 63;
    } else if (shift < 0) {
        val >>= (-shift - 1);
        if (val == UINT64_MAX) {
            /* In this case, it means that the rounding constant is 1,
             * and the addition would overflow. Return the actual
             * result directly.  */
            val = 0x8000000000000000ULL;
        } else {
            val++;
            val >>= 1;
        }
    } else {
        val <<= shift;
    }
    return val;
}

#define NEON_FN(dest, src1, src2) do { \
    int8_t tmp; \
    tmp = (int8_t)src2; \
    if (tmp >= (ssize_t)sizeof(src1) * 8) { \
        if (src1) { \
            SET_QC(); \
            dest = ~0; \
        } else { \
            dest = 0; \
        } \
    } else if (tmp <= -(ssize_t)sizeof(src1) * 8) { \
        dest = 0; \
    } else if (tmp < 0) { \
        dest = src1 >> -tmp; \
    } else { \
        dest = src1 << tmp; \
        if ((dest >> tmp) != src1) { \
            SET_QC(); \
            dest = ~0; \
        } \
    }} 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

uint64_t HELPER(neon_qshl_u64)(uint64_t val, uint64_t shiftop)
{
    int8_t shift = (int8_t)shiftop;
    if (shift >= 64) {
        if (val) {
            val = ~(uint64_t)0;
            SET_QC();
        }
    } else if (shift <= -64) {
        val = 0;
    } else if (shift < 0) {
        val >>= -shift;
    } else {
        uint64_t tmp = val;
        val <<= shift;
        if ((val >> shift) != tmp) {
            SET_QC();
            val = ~(uint64_t)0;
        }
    }
    return val;
}

#define NEON_FN(dest, src1, src2) do { \
    int8_t tmp; \
    tmp = (int8_t)src2; \
    if (tmp >= (ssize_t)sizeof(src1) * 8) { \
        if (src1) { \
            SET_QC(); \
            dest = (uint32_t)(1 << (sizeof(src1) * 8 - 1)); \
            if (src1 > 0) { \
                dest--; \
            } \
        } else { \
            dest = src1; \
        } \
    } else if (tmp <= -(ssize_t)sizeof(src1) * 8) { \
        dest = src1 >> 31; \
    } else if (tmp < 0) { \
        dest = src1 >> -tmp; \
    } else { \
        dest = src1 << tmp; \
        if ((dest >> tmp) != src1) { \
            SET_QC(); \
            dest = (uint32_t)(1 << (sizeof(src1) * 8 - 1)); \
            if (src1 > 0) { \
                dest--; \
            } \
        } \
    }} 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

uint64_t HELPER(neon_qshl_s64)(uint64_t valop, uint64_t shiftop)
{
    int8_t shift = (uint8_t)shiftop;
    int64_t val = valop;
    if (shift >= 64) {
        if (val) {
            SET_QC();
            val = (val >> 63) ^ ~SIGNBIT64;
        }
    } else if (shift <= -64) {
        val >>= 63;
    } else if (shift < 0) {
        val >>= -shift;
    } else {
        int64_t tmp = val;
        val <<= shift;
        if ((val >> shift) != tmp) {
            SET_QC();
            val = (tmp >> 63) ^ ~SIGNBIT64;
        }
    }
    return val;
}

#define NEON_FN(dest, src1, src2) do { \
    if (src1 & (1 << (sizeof(src1) * 8 - 1))) { \
        SET_QC(); \
        dest = 0; \
    } else { \
        int8_t tmp; \
        tmp = (int8_t)src2; \
        if (tmp >= (ssize_t)sizeof(src1) * 8) { \
            if (src1) { \
                SET_QC(); \
                dest = ~0; \
            } else { \
                dest = 0; \
            } \
        } else if (tmp <= -(ssize_t)sizeof(src1) * 8) { \
            dest = 0; \
        } else if (tmp < 0) { \
            dest = src1 >> -tmp; \
        } else { \
            dest = src1 << tmp; \
            if ((dest >> tmp) != src1) { \
                SET_QC(); \
                dest = ~0; \
            } \
        } \
    }} while (0)
NEON_VOP(qshlu_s8, neon_u8, 4)
NEON_VOP(qshlu_s16, neon_u16, 2)
#undef NEON_FN

uint32_t HELPER(neon_qshlu_s32)(uint32_t valop, uint32_t shiftop)
{
    if ((int32_t)valop < 0) {
        SET_QC();
        return 0;
    }
    return helper_neon_qshl_u32(valop, shiftop);
}

uint64_t HELPER(neon_qshlu_s64)(uint64_t valop, uint64_t shiftop)
{
    if ((int64_t)valop < 0) {
        SET_QC();
        return 0;
    }
    return helper_neon_qshl_u64(valop, shiftop);
}

/* FIXME: This is wrong.  */
#define NEON_FN(dest, src1, src2) do { \
    int8_t tmp; \
    tmp = (int8_t)src2; \
    if (tmp >= (ssize_t)sizeof(src1) * 8) { \
        if (src1) { \
            SET_QC(); \
            dest = ~0; \
        } else { \
            dest = 0; \
        } \
    } else if (tmp < -(ssize_t)sizeof(src1) * 8) { \
        dest = 0; \
    } else if (tmp == -(ssize_t)sizeof(src1) * 8) { \
        dest = src1 >> (sizeof(src1) * 8 - 1); \
    } else if (tmp < 0) { \
        dest = (src1 + (1 << (-1 - tmp))) >> -tmp; \
    } else { \
        dest = src1 << tmp; \
        if ((dest >> tmp) != src1) { \
            SET_QC(); \
            dest = ~0; \
        } \
    }} while (0)
NEON_VOP(qrshl_u8, neon_u8, 4)
NEON_VOP(qrshl_u16, neon_u16, 2)
#undef NEON_FN

/* The addition of the rounding constant may overflow, so we use an
 * intermediate 64 bits accumulator.  */
uint32_t HELPER(neon_qrshl_u32)(uint32_t val, uint32_t shiftop)
{
    uint32_t dest;
    int8_t shift = (int8_t)shiftop;
    if (shift >= 32) {
        if (val) {
            SET_QC();
            dest = ~0;
        } else {
            dest = 0;
        }
    } else if (shift < -32) {
        dest = 0;
    } else if (shift == -32) {
        dest = val >> 31;
    } else if (shift < 0) {
        uint64_t big_dest = ((uint64_t)val + (1 << (-1 - shift)));
        dest = big_dest >> -shift;
    } else {
        dest = val << shift;
        if ((dest >> shift) != val) {
            SET_QC();
            dest = ~0;
        }
    }
    return dest;
}

/* Handling addition overflow with 64 bits inputs values is more
 * tricky than with 32 bits values.  */
uint64_t HELPER(neon_qrshl_u64)(uint64_t val, uint64_t shiftop)
{
    int8_t shift = (int8_t)shiftop;
    if (shift >= 64) {
        if (val) {
            SET_QC();
            val = ~0;
        }
    } else if (shift < -64) {
        val = 0;
    } else if (shift == -64) {
        val >>= 63;
    } else if (shift < 0) {
        val >>= (-shift - 1);
        if (val == UINT64_MAX) {
            /* In this case, it means that the rounding constant is 1,
             * and the addition would overflow. Return the actual
             * result directly.  */
            val = 0x8000000000000000ULL;
        } else {
            val++;
            val >>= 1;
        }
    } else { \
        uint64_t tmp = val;
        val <<= shift;
        if ((val >> shift) != tmp) {
            SET_QC();
            val = ~0;
        }
    }
    return val;
}

#define NEON_FN(dest, src1, src2) do { \
    int8_t tmp; \
    tmp = (int8_t)src2; \
    if (tmp >= (ssize_t)sizeof(src1) * 8) { \
        if (src1) { \
            SET_QC(); \
            dest = (1 << (sizeof(src1) * 8 - 1)); \
            if (src1 > 0) { \
                dest--; \
            } \
        } else { \
            dest = 0; \
        } \
    } else if (tmp <= -(ssize_t)sizeof(src1) * 8) { \
        dest = 0; \
    } else if (tmp < 0) { \
        dest = (src1 + (1 << (-1 - tmp))) >> -tmp; \
    } else { \
        dest = src1 << tmp; \
        if ((dest >> tmp) != src1) { \
            SET_QC(); \
            dest = (uint32_t)(1 << (sizeof(src1) * 8 - 1)); \
            if (src1 > 0) { \
                dest--; \
            } \
        } \
    }} while (0)
NEON_VOP(qrshl_s8, neon_s8, 4)
NEON_VOP(qrshl_s16, neon_s16, 2)
#undef NEON_FN

/* The addition of the rounding constant may overflow, so we use an
 * intermediate 64 bits accumulator.  */
uint32_t HELPER(neon_qrshl_s32)(uint32_t valop, uint32_t shiftop)
{
    int32_t dest;
    int32_t val = (int32_t)valop;
    int8_t shift = (int8_t)shiftop;
    if (shift >= 32) {
        if (val) {
            SET_QC();
            dest = (val >> 31) ^ ~SIGNBIT;
        } else {
            dest = 0;
        }
    } else if (shift <= -32) {
        dest = 0;
    } else if (shift < 0) {
        int64_t big_dest = ((int64_t)val + (1 << (-1 - shift)));
        dest = big_dest >> -shift;
    } else {
        dest = val << shift;
        if ((dest >> shift) != val) {
            SET_QC();
            dest = (val >> 31) ^ ~SIGNBIT;
        }
    }
    return dest;
}

/* Handling addition overflow with 64 bits inputs values is more
 * tricky than with 32 bits values.  */
uint64_t HELPER(neon_qrshl_s64)(uint64_t valop, uint64_t shiftop)
{
    int8_t shift = (uint8_t)shiftop;
    int64_t val = valop;

    if (shift >= 64) {
        if (val) {
            SET_QC();
            val = (val >> 63) ^ ~SIGNBIT64;
        }
    } else if (shift <= -64) {
        val = 0;
    } else if (shift < 0) {
        val >>= (-shift - 1);
        if (val == INT64_MAX) {
            /* In this case, it means that the rounding constant is 1,
             * and the addition would overflow. Return the actual
             * result directly.  */
            val = 0x4000000000000000ULL;
        } else {
            val++;
            val >>= 1;
        }
    } else {
        int64_t tmp = val;
        val <<= shift;
        if ((val >> shift) != tmp) {
            SET_QC();
            val = (tmp >> 63) ^ ~SIGNBIT64;
        }
    }
    return val;
}

uint32_t HELPER(neon_add_u8)(uint32_t a, uint32_t b)
{
    uint32_t mask;
    mask = (a ^ b) & 0x80808080u;
    a &= ~0x80808080u;
    b &= ~0x80808080u;
    return (a + b) ^ mask;
}

uint32_t HELPER(neon_add_u16)(uint32_t a, uint32_t b)
{
    uint32_t mask;
    mask = (a ^ b) & 0x80008000u;
    a &= ~0x80008000u;
    b &= ~0x80008000u;
    return (a + b) ^ mask;
}

#define NEON_FN(dest, src1, src2) dest = src1 + src2
NEON_POP(padd_u8, neon_u8, 4)
NEON_POP(padd_u16, neon_u16, 2)
#undef NEON_FN

#define NEON_FN(dest, src1, src2) dest = src1 - src2
NEON_VOP(sub_u8, neon_u8, 4)
NEON_VOP(sub_u16, neon_u16, 2)
#undef NEON_FN

#define NEON_FN(dest, src1, src2) dest = src1 * src2
NEON_VOP(mul_u8, neon_u8, 4)
NEON_VOP(mul_u16, neon_u16, 2)
#undef NEON_FN

/* Polynomial multiplication is like integer multiplication except the
   partial products are XORed, not added.  */
uint32_t HELPER(neon_mul_p8)(uint32_t op1, uint32_t op2)
{
    uint32_t mask;
    uint32_t result;
    result = 0;
    while (op1) {
        mask = 0;
        if (op1 & 1)
            mask |= 0xff;
        if (op1 & (1 << 8))
            mask |= (0xff << 8);
        if (op1 & (1 << 16))
            mask |= (0xff << 16);
        if (op1 & (1 << 24))
            mask |= (0xff << 24);
        result ^= op2 & mask;
        op1 = (op1 >> 1) & 0x7f7f7f7f;
        op2 = (op2 << 1) & 0xfefefefe;
    }
    return result;
}

uint64_t HELPER(neon_mull_p8)(uint32_t op1, uint32_t op2)
{
    uint64_t result = 0;
    uint64_t mask;
    uint64_t op2ex = op2;
    op2ex = (op2ex & 0xff) |
        ((op2ex & 0xff00) << 8) |
        ((op2ex & 0xff0000) << 16) |
        ((op2ex & 0xff000000) << 24);
    while (op1) {
        mask = 0;
        if (op1 & 1) {
            mask |= 0xffff;
        }
        if (op1 & (1 << 8)) {
            mask |= (0xffffU << 16);
        }
        if (op1 & (1 << 16)) {
            mask |= (0xffffULL << 32);
        }
        if (op1 & (1 << 24)) {
            mask |= (0xffffULL << 48);
        }
        result ^= op2ex & mask;
        op1 = (op1 >> 1) & 0x7f7f7f7f;
        op2ex <<= 1;
    }
    return result;
}

#define NEON_FN(dest, src1, src2) dest = (src1 & src2) ? -1 : 0
NEON_VOP(tst_u8, neon_u8, 4)
NEON_VOP(tst_u16, neon_u16, 2)
NEON_VOP(tst_u32, neon_u32, 1)
#undef NEON_FN

#define NEON_FN(dest, src1, src2) dest = (src1 == src2) ? -1 : 0
NEON_VOP(ceq_u8, neon_u8, 4)
NEON_VOP(ceq_u16, neon_u16, 2)
NEON_VOP(ceq_u32, neon_u32, 1)
#undef NEON_FN

#define NEON_FN(dest, src, dummy) dest = (src < 0) ? -src : src
NEON_VOP1(abs_s8, neon_s8, 4)
NEON_VOP1(abs_s16, neon_s16, 2)
#undef NEON_FN

/* Count Leading Sign/Zero Bits.  */
static inline int do_clz8(uint8_t x)
{
    int n;
    for (n = 8; x; n--)
        x >>= 1;
    return n;
}

static inline int do_clz16(uint16_t x)
{
    int n;
    for (n = 16; x; n--)
        x >>= 1;
    return n;
}

#define NEON_FN(dest, src, dummy) dest = do_clz8(src)
NEON_VOP1(clz_u8, neon_u8, 4)
#undef NEON_FN

#define NEON_FN(dest, src, dummy) dest = do_clz16(src)
NEON_VOP1(clz_u16, neon_u16, 2)
#undef NEON_FN

#define NEON_FN(dest, src, dummy) dest = do_clz8((src < 0) ? ~src : src) - 1
NEON_VOP1(cls_s8, neon_s8, 4)
#undef NEON_FN

#define NEON_FN(dest, src, dummy) dest = do_clz16((src < 0) ? ~src : src) - 1
NEON_VOP1(cls_s16, neon_s16, 2)
#undef NEON_FN

uint32_t HELPER(neon_cls_s32)(uint32_t x)
{
    int count;
    if ((int32_t)x < 0)
        x = ~x;
    for (count = 32; x; count--)
        x = x >> 1;
    return count - 1;
}

/* Bit count.  */
uint32_t HELPER(neon_cnt_u8)(uint32_t x)
{
    x = (x & 0x55555555) + ((x >>  1) & 0x55555555);
    x = (x & 0x33333333) + ((x >>  2) & 0x33333333);
    x = (x & 0x0f0f0f0f) + ((x >>  4) & 0x0f0f0f0f);
    return x;
}

#define NEON_QDMULH16(dest, src1, src2, round) do { \
    uint32_t tmp = (int32_t)(int16_t) src1 * (int16_t) src2; \
    if ((tmp ^ (tmp << 1)) & SIGNBIT) { \
        SET_QC(); \
        tmp = (tmp >> 31) ^ ~SIGNBIT; \
    } else { \
        tmp <<= 1; \
    } \
    if (round) { \
        int32_t old = tmp; \
        tmp += 1 << 15; \
        if ((int32_t)tmp < old) { \
            SET_QC(); \
            tmp = SIGNBIT - 1; \
        } \
    } \
    dest = tmp >> 16; \
    } while(0)
#define NEON_FN(dest, src1, src2) NEON_QDMULH16(dest, src1, src2, 0)
NEON_VOP(qdmulh_s16, neon_s16, 2)
#undef NEON_FN
#define NEON_FN(dest, src1, src2) NEON_QDMULH16(dest, src1, src2, 1)
NEON_VOP(qrdmulh_s16, neon_s16, 2)
#undef NEON_FN
#undef NEON_QDMULH16

#define NEON_QDMULH32(dest, src1, src2, round) do { \
    uint64_t tmp = (int64_t)(int32_t) src1 * (int32_t) src2; \
    if ((tmp ^ (tmp << 1)) & SIGNBIT64) { \
        SET_QC(); \
        tmp = (tmp >> 63) ^ ~SIGNBIT64; \
    } else { \
        tmp <<= 1; \
    } \
    if (round) { \
        int64_t old = tmp; \
        tmp += (int64_t)1 << 31; \
        if ((int64_t)tmp < old) { \
            SET_QC(); \
            tmp = SIGNBIT64 - 1; \
        } \
    } \
    dest = tmp >> 32; \
    } while(0)
#define NEON_FN(dest, src1, src2) NEON_QDMULH32(dest, src1, src2, 0)
NEON_VOP(qdmulh_s32, neon_s32, 1)
#undef NEON_FN
#define NEON_FN(dest, src1, src2) NEON_QDMULH32(dest, src1, src2, 1)
NEON_VOP(qrdmulh_s32, neon_s32, 1)
#undef NEON_FN
#undef NEON_QDMULH32

uint32_t HELPER(neon_narrow_u8)(uint64_t x)
{
    return (x & 0xffu) | ((x >> 8) & 0xff00u) | ((x >> 16) & 0xff0000u)
           | ((x >> 24) & 0xff000000u);
}

uint32_t HELPER(neon_narrow_u16)(uint64_t x)
{
    return (x & 0xffffu) | ((x >> 16) & 0xffff0000u);
}

uint32_t HELPER(neon_narrow_high_u8)(uint64_t x)
{
    return ((x >> 8) & 0xff) | ((x >> 16) & 0xff00)
            | ((x >> 24) & 0xff0000) | ((x >> 32) & 0xff000000);
}

uint32_t HELPER(neon_narrow_high_u16)(uint64_t x)
{
    return ((x >> 16) & 0xffff) | ((x >> 32) & 0xffff0000);
}

uint32_t HELPER(neon_narrow_round_high_u8)(uint64_t x)
{
    x &= 0xff80ff80ff80ff80ull;
    x += 0x0080008000800080ull;
    return ((x >> 8) & 0xff) | ((x >> 16) & 0xff00)
            | ((x >> 24) & 0xff0000) | ((x >> 32) & 0xff000000);
}

uint32_t HELPER(neon_narrow_round_high_u16)(uint64_t x)
{
    x &= 0xffff8000ffff8000ull;
    x += 0x0000800000008000ull;
    return ((x >> 16) & 0xffff) | ((x >> 32) & 0xffff0000);
}

uint32_t HELPER(neon_unarrow_sat8)(uint64_t x)
{
    uint16_t s;
    uint8_t d;
    uint32_t res = 0;
#define SAT8(n) \
    s = x >> n; \
    if (s & 0x8000) { \
        SET_QC(); \
    } else { \
        if (s > 0xff) { \
            d = 0xff; \
            SET_QC(); \
        } else  { \
            d = s; \
        } \
        res |= (uint32_t)d << (n / 2); \
    }

    SAT8(0);
    SAT8(16);
    SAT8(32);
    SAT8(48);
#undef SAT8
    return res;
}

uint32_t HELPER(neon_narrow_sat_u8)(uint64_t x)
{
    uint16_t s;
    uint8_t d;
    uint32_t res = 0;
#define SAT8(n) \
    s = x >> n; \
    if (s > 0xff) { \
        d = 0xff; \
        SET_QC(); \
    } else  { \
        d = s; \
    } \
    res |= (uint32_t)d << (n / 2);

    SAT8(0);
    SAT8(16);
    SAT8(32);
    SAT8(48);
#undef SAT8
    return res;
}

uint32_t HELPER(neon_narrow_sat_s8)(uint64_t x)
{
    int16_t s;
    uint8_t d;
    uint32_t res = 0;
#define SAT8(n) \
    s = x >> n; \
    if (s != (int8_t)s) { \
        d = (s >> 15) ^ 0x7f; \
        SET_QC(); \
    } else  { \
        d = s; \
    } \
    res |= (uint32_t)d << (n / 2);

    SAT8(0);
    SAT8(16);
    SAT8(32);
    SAT8(48);
#undef SAT8
    return res;
}

uint32_t HELPER(neon_unarrow_sat16)(uint64_t x)
{
    uint32_t high;
    uint32_t low;
    low = x;
    if (low & 0x80000000) {
        low = 0;
        SET_QC();
    } else if (low > 0xffff) {
        low = 0xffff;
        SET_QC();
    }
    high = x >> 32;
    if (high & 0x80000000) {
        high = 0;
        SET_QC();
    } else if (high > 0xffff) {
        high = 0xffff;
        SET_QC();
    }
    return low | (high << 16);
}

uint32_t HELPER(neon_narrow_sat_u16)(uint64_t x)
{
    uint32_t high;
    uint32_t low;
    low = x;
    if (low > 0xffff) {
        low = 0xffff;
        SET_QC();
    }
    high = x >> 32;
    if (high > 0xffff) {
        high = 0xffff;
        SET_QC();
    }
    return low | (high << 16);
}

uint32_t HELPER(neon_narrow_sat_s16)(uint64_t x)
{
    int32_t low;
    int32_t high;
    low = x;
    if (low != (int16_t)low) {
        low = (low >> 31) ^ 0x7fff;
        SET_QC();
    }
    high = x >> 32;
    if (high != (int16_t)high) {
        high = (high >> 31) ^ 0x7fff;
        SET_QC();
    }
    return (uint16_t)low | (high << 16);
}

uint32_t HELPER(neon_unarrow_sat32)(uint64_t x)
{
    if (x & 0x8000000000000000ull) {
        SET_QC();
        return 0;
    }
    if (x > 0xffffffffu) {
        SET_QC();
        return 0xffffffffu;
    }
    return x;
}

uint32_t HELPER(neon_narrow_sat_u32)(uint64_t x)
{
    if (x > 0xffffffffu) {
        SET_QC();
        return 0xffffffffu;
    }
    return x;
}

uint32_t HELPER(neon_narrow_sat_s32)(uint64_t x)
{
    if ((int64_t)x != (int32_t)x) {
        SET_QC();
        return ((int64_t)x >> 63) ^ 0x7fffffff;
    }
    return x;
}

uint64_t HELPER(neon_widen_u8)(uint32_t x)
{
    uint64_t tmp;
    uint64_t ret;
    ret = (uint8_t)x;
    tmp = (uint8_t)(x >> 8);
    ret |= tmp << 16;
    tmp = (uint8_t)(x >> 16);
    ret |= tmp << 32;
    tmp = (uint8_t)(x >> 24);
    ret |= tmp << 48;
    return ret;
}

uint64_t HELPER(neon_widen_s8)(uint32_t x)
{
    uint64_t tmp;
    uint64_t ret;
    ret = (uint16_t)(int8_t)x;
    tmp = (uint16_t)(int8_t)(x >> 8);
    ret |= tmp << 16;
    tmp = (uint16_t)(int8_t)(x >> 16);
    ret |= tmp << 32;
    tmp = (uint16_t)(int8_t)(x >> 24);
    ret |= tmp << 48;
    return ret;
}

uint64_t HELPER(neon_widen_u16)(uint32_t x)
{
    uint64_t high = (uint16_t)(x >> 16);
    return ((uint16_t)x) | (high << 32);
}

uint64_t HELPER(neon_widen_s16)(uint32_t x)
{
    uint64_t high = (int16_t)(x >> 16);
    return ((uint32_t)(int16_t)x) | (high << 32);
}

uint64_t HELPER(neon_addl_u16)(uint64_t a, uint64_t b)
{
    uint64_t mask;
    mask = (a ^ b) & 0x8000800080008000ull;
    a &= ~0x8000800080008000ull;
    b &= ~0x8000800080008000ull;
    return (a + b) ^ mask;
}

uint64_t HELPER(neon_addl_u32)(uint64_t a, uint64_t b)
{
    uint64_t mask;
    mask = (a ^ b) & 0x8000000080000000ull;
    a &= ~0x8000000080000000ull;
    b &= ~0x8000000080000000ull;
    return (a + b) ^ mask;
}

uint64_t HELPER(neon_paddl_u16)(uint64_t a, uint64_t b)
{
    uint64_t tmp;
    uint64_t tmp2;

    tmp = a & 0x0000ffff0000ffffull;
    tmp += (a >> 16) & 0x0000ffff0000ffffull;
    tmp2 = b & 0xffff0000ffff0000ull;
    tmp2 += (b << 16) & 0xffff0000ffff0000ull;
    return    ( tmp         & 0xffff)
            | ((tmp  >> 16) & 0xffff0000ull)
            | ((tmp2 << 16) & 0xffff00000000ull)
            | ( tmp2        & 0xffff000000000000ull);
}

uint64_t HELPER(neon_paddl_u32)(uint64_t a, uint64_t b)
{
    uint32_t low = a + (a >> 32);
    uint32_t high = b + (b >> 32);
    return low + ((uint64_t)high << 32);
}

uint64_t HELPER(neon_subl_u16)(uint64_t a, uint64_t b)
{
    uint64_t mask;
    mask = (a ^ ~b) & 0x8000800080008000ull;
    a |= 0x8000800080008000ull;
    b &= ~0x8000800080008000ull;
    return (a - b) ^ mask;
}

uint64_t HELPER(neon_subl_u32)(uint64_t a, uint64_t b)
{
    uint64_t mask;
    mask = (a ^ ~b) & 0x8000000080000000ull;
    a |= 0x8000000080000000ull;
    b &= ~0x8000000080000000ull;
    return (a - b) ^ mask;
}

uint64_t HELPER(neon_addl_saturate_s32)(uint64_t a, uint64_t b)
{
    uint32_t x, y;
    uint32_t low, high;

    x = a;
    y = b;
    low = x + y;
    if (((low ^ x) & SIGNBIT) && !((x ^ y) & SIGNBIT)) {
        SET_QC();
        low = ((int32_t)x >> 31) ^ ~SIGNBIT;
    }
    x = a >> 32;
    y = b >> 32;
    high = x + y;
    if (((high ^ x) & SIGNBIT) && !((x ^ y) & SIGNBIT)) {
        SET_QC();
        high = ((int32_t)x >> 31) ^ ~SIGNBIT;
    }
    return low | ((uint64_t)high << 32);
}

uint64_t HELPER(neon_addl_saturate_s64)(uint64_t a, uint64_t b)
{
    uint64_t result;

    result = a + b;
    if (((result ^ a) & SIGNBIT64) && !((a ^ b) & SIGNBIT64)) {
        SET_QC();
        result = ((int64_t)a >> 63) ^ ~SIGNBIT64;
    }
    return result;
}

/* We have to do the arithmetic in a larger type than
 * the input type, because for example with a signed 32 bit
 * op the absolute difference can overflow a signed 32 bit value.
 */
#define DO_ABD(dest, x, y, intype, arithtype) do {            \
    arithtype tmp_x = (intype)(x);                            \
    arithtype tmp_y = (intype)(y);                            \
    dest = ((tmp_x > tmp_y) ? tmp_x - tmp_y : tmp_y - tmp_x); \
    } while(0)

uint64_t HELPER(neon_abdl_u16)(uint32_t a, uint32_t b)
{
    uint64_t tmp;
    uint64_t result;
    DO_ABD(result, a, b, uint8_t, uint32_t);
    DO_ABD(tmp, a >> 8, b >> 8, uint8_t, uint32_t);
    result |= tmp << 16;
    DO_ABD(tmp, a >> 16, b >> 16, uint8_t, uint32_t);
    result |= tmp << 32;
    DO_ABD(tmp, a >> 24, b >> 24, uint8_t, uint32_t);
    result |= tmp << 48;
    return result;
}

uint64_t HELPER(neon_abdl_s16)(uint32_t a, uint32_t b)
{
    uint64_t tmp;
    uint64_t result;
    DO_ABD(result, a, b, int8_t, int32_t);
    DO_ABD(tmp, a >> 8, b >> 8, int8_t, int32_t);
    result |= tmp << 16;
    DO_ABD(tmp, a >> 16, b >> 16, int8_t, int32_t);
    result |= tmp << 32;
    DO_ABD(tmp, a >> 24, b >> 24, int8_t, int32_t);
    result |= tmp << 48;
    return result;
}

uint64_t HELPER(neon_abdl_u32)(uint32_t a, uint32_t b)
{
    uint64_t tmp;
    uint64_t result;
    DO_ABD(result, a, b, uint16_t, uint32_t);
    DO_ABD(tmp, a >> 16, b >> 16, uint16_t, uint32_t);
    return result | (tmp << 32);
}

uint64_t HELPER(neon_abdl_s32)(uint32_t a, uint32_t b)
{
    uint64_t tmp;
    uint64_t result;
    DO_ABD(result, a, b, int16_t, int32_t);
    DO_ABD(tmp, a >> 16, b >> 16, int16_t, int32_t);
    return result | (tmp << 32);
}

uint64_t HELPER(neon_abdl_u64)(uint32_t a, uint32_t b)
{
    uint64_t result;
    DO_ABD(result, a, b, uint32_t, uint64_t);
    return result;
}

uint64_t HELPER(neon_abdl_s64)(uint32_t a, uint32_t b)
{
    uint64_t result;
    DO_ABD(result, a, b, int32_t, int64_t);
    return result;
}
#undef DO_ABD

/* Widening multiply. Named type is the source type.  */
#define DO_MULL(dest, x, y, type1, type2) do { \
    type1 tmp_x = x; \
    type1 tmp_y = y; \
    dest = (type2)((type2)tmp_x * (type2)tmp_y); \
    } while(0)

uint64_t HELPER(neon_mull_u8)(uint32_t a, uint32_t b)
{
    uint64_t tmp;
    uint64_t result;

    DO_MULL(result, a, b, uint8_t, uint16_t);
    DO_MULL(tmp, a >> 8, b >> 8, uint8_t, uint16_t);
    result |= tmp << 16;
    DO_MULL(tmp, a >> 16, b >> 16, uint8_t, uint16_t);
    result |= tmp << 32;
    DO_MULL(tmp, a >> 24, b >> 24, uint8_t, uint16_t);
    result |= tmp << 48;
    return result;
}

uint64_t HELPER(neon_mull_s8)(uint32_t a, uint32_t b)
{
    uint64_t tmp;
    uint64_t result;

    DO_MULL(result, a, b, int8_t, uint16_t);
    DO_MULL(tmp, a >> 8, b >> 8, int8_t, uint16_t);
    result |= tmp << 16;
    DO_MULL(tmp, a >> 16, b >> 16, int8_t, uint16_t);
    result |= tmp << 32;
    DO_MULL(tmp, a >> 24, b >> 24, int8_t, uint16_t);
    result |= tmp << 48;
    return result;
}

uint64_t HELPER(neon_mull_u16)(uint32_t a, uint32_t b)
{
    uint64_t tmp;
    uint64_t result;

    DO_MULL(result, a, b, uint16_t, uint32_t);
    DO_MULL(tmp, a >> 16, b >> 16, uint16_t, uint32_t);
    return result | (tmp << 32);
}

uint64_t HELPER(neon_mull_s16)(uint32_t a, uint32_t b)
{
    uint64_t tmp;
    uint64_t result;

    DO_MULL(result, a, b, int16_t, uint32_t);
    DO_MULL(tmp, a >> 16, b >> 16, int16_t, uint32_t);
    return result | (tmp << 32);
}

uint64_t HELPER(neon_negl_u16)(uint64_t x)
{
    uint16_t tmp;
    uint64_t result;
    result = (uint16_t)-x;
    tmp = -(x >> 16);
    result |= (uint64_t)tmp << 16;
    tmp = -(x >> 32);
    result |= (uint64_t)tmp << 32;
    tmp = -(x >> 48);
    result |= (uint64_t)tmp << 48;
    return result;
}

uint64_t HELPER(neon_negl_u32)(uint64_t x)
{
    uint32_t low = -x;
    uint32_t high = -(x >> 32);
    return low | ((uint64_t)high << 32);
}

/* FIXME:  There should be a native op for this.  */
uint64_t HELPER(neon_negl_u64)(uint64_t x)
{
    return -x;
}

/* Saturnating sign manuipulation.  */
/* ??? Make these use NEON_VOP1 */
#define DO_QABS8(x) do { \
    if (x == (int8_t)0x80) { \
        x = 0x7f; \
        SET_QC(); \
    } else if (x < 0) { \
        x = -x; \
    }} while (0)
uint32_t HELPER(neon_qabs_s8)(uint32_t x)
{
    neon_s8 vec;
    NEON_UNPACK(neon_s8, vec, x);
    DO_QABS8(vec.v1);
    DO_QABS8(vec.v2);
    DO_QABS8(vec.v3);
    DO_QABS8(vec.v4);
    NEON_PACK(neon_s8, x, vec);
    return x;
}
#undef DO_QABS8

#define DO_QNEG8(x) do { \
    if (x == (int8_t)0x80) { \
        x = 0x7f; \
        SET_QC(); \
    } else { \
        x = -x; \
    }} while (0)
uint32_t HELPER(neon_qneg_s8)(uint32_t x)
{
    neon_s8 vec;
    NEON_UNPACK(neon_s8, vec, x);
    DO_QNEG8(vec.v1);
    DO_QNEG8(vec.v2);
    DO_QNEG8(vec.v3);
    DO_QNEG8(vec.v4);
    NEON_PACK(neon_s8, x, vec);
    return x;
}
#undef DO_QNEG8

#define DO_QABS16(x) do { \
    if (x == (int16_t)0x8000) { \
        x = 0x7fff; \
        SET_QC(); \
    } else if (x < 0) { \
        x = -x; \
    }} while (0)
uint32_t HELPER(neon_qabs_s16)(uint32_t x)
{
    neon_s16 vec;
    NEON_UNPACK(neon_s16, vec, x);
    DO_QABS16(vec.v1);
    DO_QABS16(vec.v2);
    NEON_PACK(neon_s16, x, vec);
    return x;
}
#undef DO_QABS16

#define DO_QNEG16(x) do { \
    if (x == (int16_t)0x8000) { \
        x = 0x7fff; \
        SET_QC(); \
    } else { \
        x = -x; \
    }} while (0)
uint32_t HELPER(neon_qneg_s16)(uint32_t x)
{
    neon_s16 vec;
    NEON_UNPACK(neon_s16, vec, x);
    DO_QNEG16(vec.v1);
    DO_QNEG16(vec.v2);
    NEON_PACK(neon_s16, x, vec);
    return x;
}
#undef DO_QNEG16

uint32_t HELPER(neon_qabs_s32)(uint32_t x)
{
    if (x == SIGNBIT) {
        SET_QC();
        x = ~SIGNBIT;
    } else if ((int32_t)x < 0) {
        x = -x;
    }
    return x;
}

uint32_t HELPER(neon_qneg_s32)(uint32_t x)
{
    if (x == SIGNBIT) {
        SET_QC();
        x = ~SIGNBIT;
    } else {
        x = -x;
    }
    return x;
}

/* NEON Float helpers.  */
uint32_t HELPER(neon_min_f32)(uint32_t a, uint32_t b)
{
    return float32_val(float32_min(make_float32(a), make_float32(b), NFS));
}

uint32_t HELPER(neon_max_f32)(uint32_t a, uint32_t b)
{
    return float32_val(float32_max(make_float32(a), make_float32(b), NFS));
}

uint32_t HELPER(neon_abd_f32)(uint32_t a, uint32_t b)
{
    float32 f0 = make_float32(a);
    float32 f1 = make_float32(b);
    return float32_val(float32_abs(float32_sub(f0, f1, NFS)));
}

uint32_t HELPER(neon_add_f32)(uint32_t a, uint32_t b)
{
    return float32_val(float32_add(make_float32(a), make_float32(b), NFS));
}

uint32_t HELPER(neon_sub_f32)(uint32_t a, uint32_t b)
{
    return float32_val(float32_sub(make_float32(a), make_float32(b), NFS));
}

uint32_t HELPER(neon_mul_f32)(uint32_t a, uint32_t b)
{
    return float32_val(float32_mul(make_float32(a), make_float32(b), NFS));
}

/* Floating point comparisons produce an integer result.
 * Note that EQ doesn't signal InvalidOp for QNaNs but GE and GT do.
 * Softfloat routines return 0/1, which we convert to the 0/-1 Neon requires.
 */
uint32_t HELPER(neon_ceq_f32)(uint32_t a, uint32_t b)
{
    return -float32_eq_quiet(make_float32(a), make_float32(b), NFS);
}

uint32_t HELPER(neon_cge_f32)(uint32_t a, uint32_t b)
{
    return -float32_le(make_float32(b), make_float32(a), NFS);
}

uint32_t HELPER(neon_cgt_f32)(uint32_t a, uint32_t b)
{
    return -float32_lt(make_float32(b), make_float32(a), NFS);
}

uint32_t HELPER(neon_acge_f32)(uint32_t a, uint32_t b)
{
    float32 f0 = float32_abs(make_float32(a));
    float32 f1 = float32_abs(make_float32(b));
    return -float32_le(f1, f0, NFS);
}

uint32_t HELPER(neon_acgt_f32)(uint32_t a, uint32_t b)
{
    float32 f0 = float32_abs(make_float32(a));
    float32 f1 = float32_abs(make_float32(b));
    return -float32_lt(f1, f0, NFS);
}

#define ELEM(V, N, SIZE) (((V) >> ((N) * (SIZE))) & ((1ull << (SIZE)) - 1))

void HELPER(neon_qunzip8)(uint32_t rd, uint32_t rm)
{
    uint64_t zm0 = float64_val(env->vfp.regs[rm]);
    uint64_t zm1 = float64_val(env->vfp.regs[rm + 1]);
    uint64_t zd0 = float64_val(env->vfp.regs[rd]);
    uint64_t zd1 = float64_val(env->vfp.regs[rd + 1]);
    uint64_t d0 = ELEM(zd0, 0, 8) | (ELEM(zd0, 2, 8) << 8)
        | (ELEM(zd0, 4, 8) << 16) | (ELEM(zd0, 6, 8) << 24)
        | (ELEM(zd1, 0, 8) << 32) | (ELEM(zd1, 2, 8) << 40)
        | (ELEM(zd1, 4, 8) << 48) | (ELEM(zd1, 6, 8) << 56);
    uint64_t d1 = ELEM(zm0, 0, 8) | (ELEM(zm0, 2, 8) << 8)
        | (ELEM(zm0, 4, 8) << 16) | (ELEM(zm0, 6, 8) << 24)
        | (ELEM(zm1, 0, 8) << 32) | (ELEM(zm1, 2, 8) << 40)
        | (ELEM(zm1, 4, 8) << 48) | (ELEM(zm1, 6, 8) << 56);
    uint64_t m0 = ELEM(zd0, 1, 8) | (ELEM(zd0, 3, 8) << 8)
        | (ELEM(zd0, 5, 8) << 16) | (ELEM(zd0, 7, 8) << 24)
        | (ELEM(zd1, 1, 8) << 32) | (ELEM(zd1, 3, 8) << 40)
        | (ELEM(zd1, 5, 8) << 48) | (ELEM(zd1, 7, 8) << 56);
    uint64_t m1 = ELEM(zm0, 1, 8) | (ELEM(zm0, 3, 8) << 8)
        | (ELEM(zm0, 5, 8) << 16) | (ELEM(zm0, 7, 8) << 24)
        | (ELEM(zm1, 1, 8) << 32) | (ELEM(zm1, 3, 8) << 40)
        | (ELEM(zm1, 5, 8) << 48) | (ELEM(zm1, 7, 8) << 56);
    env->vfp.regs[rm] = make_float64(m0);
    env->vfp.regs[rm + 1] = make_float64(m1);
    env->vfp.regs[rd] = make_float64(d0);
    env->vfp.regs[rd + 1] = make_float64(d1);
}

void HELPER(neon_qunzip16)(uint32_t rd, uint32_t rm)
{
    uint64_t zm0 = float64_val(env->vfp.regs[rm]);
    uint64_t zm1 = float64_val(env->vfp.regs[rm + 1]);
    uint64_t zd0 = float64_val(env->vfp.regs[rd]);
    uint64_t zd1 = float64_val(env->vfp.regs[rd + 1]);
    uint64_t d0 = ELEM(zd0, 0, 16) | (ELEM(zd0, 2, 16) << 16)
        | (ELEM(zd1, 0, 16) << 32) | (ELEM(zd1, 2, 16) << 48);
    uint64_t d1 = ELEM(zm0, 0, 16) | (ELEM(zm0, 2, 16) << 16)
        | (ELEM(zm1, 0, 16) << 32) | (ELEM(zm1, 2, 16) << 48);
    uint64_t m0 = ELEM(zd0, 1, 16) | (ELEM(zd0, 3, 16) << 16)
        | (ELEM(zd1, 1, 16) << 32) | (ELEM(zd1, 3, 16) << 48);
    uint64_t m1 = ELEM(zm0, 1, 16) | (ELEM(zm0, 3, 16) << 16)
        | (ELEM(zm1, 1, 16) << 32) | (ELEM(zm1, 3, 16) << 48);
    env->vfp.regs[rm] = make_float64(m0);
    env->vfp.regs[rm + 1] = make_float64(m1);
    env->vfp.regs[rd] = make_float64(d0);
    env->vfp.regs[rd + 1] = make_float64(d1);
}

void HELPER(neon_qunzip32)(uint32_t rd, uint32_t rm)
{
    uint64_t zm0 = float64_val(env->vfp.regs[rm]);
    uint64_t zm1 = float64_val(env->vfp.regs[rm + 1]);
    uint64_t zd0 = float64_val(env->vfp.regs[rd]);
    uint64_t zd1 = float64_val(env->vfp.regs[rd + 1]);
    uint64_t d0 = ELEM(zd0, 0, 32) | (ELEM(zd1, 0, 32) << 32);
    uint64_t d1 = ELEM(zm0, 0, 32) | (ELEM(zm1, 0, 32) << 32);
    uint64_t m0 = ELEM(zd0, 1, 32) | (ELEM(zd1, 1, 32) << 32);
    uint64_t m1 = ELEM(zm0, 1, 32) | (ELEM(zm1, 1, 32) << 32);
    env->vfp.regs[rm] = make_float64(m0);
    env->vfp.regs[rm + 1] = make_float64(m1);
    env->vfp.regs[rd] = make_float64(d0);
    env->vfp.regs[rd + 1] = make_float64(d1);
}

void HELPER(neon_unzip8)(uint32_t rd, uint32_t rm)
{
    uint64_t zm = float64_val(env->vfp.regs[rm]);
    uint64_t zd = float64_val(env->vfp.regs[rd]);
    uint64_t d0 = ELEM(zd, 0, 8) | (ELEM(zd, 2, 8) << 8)
        | (ELEM(zd, 4, 8) << 16) | (ELEM(zd, 6, 8) << 24)
        | (ELEM(zm, 0, 8) << 32) | (ELEM(zm, 2, 8) << 40)
        | (ELEM(zm, 4, 8) << 48) | (ELEM(zm, 6, 8) << 56);
    uint64_t m0 = ELEM(zd, 1, 8) | (ELEM(zd, 3, 8) << 8)
        | (ELEM(zd, 5, 8) << 16) | (ELEM(zd, 7, 8) << 24)
        | (ELEM(zm, 1, 8) << 32) | (ELEM(zm, 3, 8) << 40)
        | (ELEM(zm, 5, 8) << 48) | (ELEM(zm, 7, 8) << 56);
    env->vfp.regs[rm] = make_float64(m0);
    env->vfp.regs[rd] = make_float64(d0);
}

void HELPER(neon_unzip16)(uint32_t rd, uint32_t rm)
{
    uint64_t zm = float64_val(env->vfp.regs[rm]);
    uint64_t zd = float64_val(env->vfp.regs[rd]);
    uint64_t d0 = ELEM(zd, 0, 16) | (ELEM(zd, 2, 16) << 16)
        | (ELEM(zm, 0, 16) << 32) | (ELEM(zm, 2, 16) << 48);
    uint64_t m0 = ELEM(zd, 1, 16) | (ELEM(zd, 3, 16) << 16)
        | (ELEM(zm, 1, 16) << 32) | (ELEM(zm, 3, 16) << 48);
    env->vfp.regs[rm] = make_float64(m0);
    env->vfp.regs[rd] = make_float64(d0);
}

void HELPER(neon_qzip8)(uint32_t rd, uint32_t rm)
{
    uint64_t zm0 = float64_val(env->vfp.regs[rm]);
    uint64_t zm1 = float64_val(env->vfp.regs[rm + 1]);
    uint64_t zd0 = float64_val(env->vfp.regs[rd]);
    uint64_t zd1 = float64_val(env->vfp.regs[rd + 1]);
    uint64_t d0 = ELEM(zd0, 0, 8) | (ELEM(zm0, 0, 8) << 8)
        | (ELEM(zd0, 1, 8) << 16) | (ELEM(zm0, 1, 8) << 24)
        | (ELEM(zd0, 2, 8) << 32) | (ELEM(zm0, 2, 8) << 40)
        | (ELEM(zd0, 3, 8) << 48) | (ELEM(zm0, 3, 8) << 56);
    uint64_t d1 = ELEM(zd0, 4, 8) | (ELEM(zm0, 4, 8) << 8)
        | (ELEM(zd0, 5, 8) << 16) | (ELEM(zm0, 5, 8) << 24)
        | (ELEM(zd0, 6, 8) << 32) | (ELEM(zm0, 6, 8) << 40)
        | (ELEM(zd0, 7, 8) << 48) | (ELEM(zm0, 7, 8) << 56);
    uint64_t m0 = ELEM(zd1, 0, 8) | (ELEM(zm1, 0, 8) << 8)
        | (ELEM(zd1, 1, 8) << 16) | (ELEM(zm1, 1, 8) << 24)
        | (ELEM(zd1, 2, 8) << 32) | (ELEM(zm1, 2, 8) << 40)
        | (ELEM(zd1, 3, 8) << 48) | (ELEM(zm1, 3, 8) << 56);
    uint64_t m1 = ELEM(zd1, 4, 8) | (ELEM(zm1, 4, 8) << 8)
        | (ELEM(zd1, 5, 8) << 16) | (ELEM(zm1, 5, 8) << 24)
        | (ELEM(zd1, 6, 8) << 32) | (ELEM(zm1, 6, 8) << 40)
        | (ELEM(zd1, 7, 8) << 48) | (ELEM(zm1, 7, 8) << 56);
    env->vfp.regs[rm] = make_float64(m0);
    env->vfp.regs[rm + 1] = make_float64(m1);
    env->vfp.regs[rd] = make_float64(d0);
    env->vfp.regs[rd + 1] = make_float64(d1);
}

void HELPER(neon_qzip16)(uint32_t rd, uint32_t rm)
{
    uint64_t zm0 = float64_val(env->vfp.regs[rm]);
    uint64_t zm1 = float64_val(env->vfp.regs[rm + 1]);
    uint64_t zd0 = float64_val(env->vfp.regs[rd]);
    uint64_t zd1 = float64_val(env->vfp.regs[rd + 1]);
    uint64_t d0 = ELEM(zd0, 0, 16) | (ELEM(zm0, 0, 16) << 16)
        | (ELEM(zd0, 1, 16) << 32) | (ELEM(zm0, 1, 16) << 48);
    uint64_t d1 = ELEM(zd0, 2, 16) | (ELEM(zm0, 2, 16) << 16)
        | (ELEM(zd0, 3, 16) << 32) | (ELEM(zm0, 3, 16) << 48);
    uint64_t m0 = ELEM(zd1, 0, 16) | (ELEM(zm1, 0, 16) << 16)
        | (ELEM(zd1, 1, 16) << 32) | (ELEM(zm1, 1, 16) << 48);
    uint64_t m1 = ELEM(zd1, 2, 16) | (ELEM(zm1, 2, 16) << 16)
        | (ELEM(zd1, 3, 16) << 32) | (ELEM(zm1, 3, 16) << 48);
    env->vfp.regs[rm] = make_float64(m0);
    env->vfp.regs[rm + 1] = make_float64(m1);
    env->vfp.regs[rd] = make_float64(d0);
    env->vfp.regs[rd + 1] = make_float64(d1);
}

void HELPER(neon_qzip32)(uint32_t rd, uint32_t rm)
{
    uint64_t zm0 = float64_val(env->vfp.regs[rm]);
    uint64_t zm1 = float64_val(env->vfp.regs[rm + 1]);
    uint64_t zd0 = float64_val(env->vfp.regs[rd]);
    uint64_t zd1 = float64_val(env->vfp.regs[rd + 1]);
    uint64_t d0 = ELEM(zd0, 0, 32) | (ELEM(zm0, 0, 32) << 32);
    uint64_t d1 = ELEM(zd0, 1, 32) | (ELEM(zm0, 1, 32) << 32);
    uint64_t m0 = ELEM(zd1, 0, 32) | (ELEM(zm1, 0, 32) << 32);
    uint64_t m1 = ELEM(zd1, 1, 32) | (ELEM(zm1, 1, 32) << 32);
    env->vfp.regs[rm] = make_float64(m0);
    env->vfp.regs[rm + 1] = make_float64(m1);
    env->vfp.regs[rd] = make_float64(d0);
    env->vfp.regs[rd + 1] = make_float64(d1);
}

void HELPER(neon_zip8)(uint32_t rd, uint32_t rm)
{
    uint64_t zm = float64_val(env->vfp.regs[rm]);
    uint64_t zd = float64_val(env->vfp.regs[rd]);
    uint64_t d0 = ELEM(zd, 0, 8) | (ELEM(zm, 0, 8) << 8)
        | (ELEM(zd, 1, 8) << 16) | (ELEM(zm, 1, 8) << 24)
        | (ELEM(zd, 2, 8) << 32) | (ELEM(zm, 2, 8) << 40)
        | (ELEM(zd, 3, 8) << 48) | (ELEM(zm, 3, 8) << 56);
    uint64_t m0 = ELEM(zd, 4, 8) | (ELEM(zm, 4, 8) << 8)
        | (ELEM(zd, 5, 8) << 16) | (ELEM(zm, 5, 8) << 24)
        | (ELEM(zd, 6, 8) << 32) | (ELEM(zm, 6, 8) << 40)
        | (ELEM(zd, 7, 8) << 48) | (ELEM(zm, 7, 8) << 56);
    env->vfp.regs[rm] = make_float64(m0);
    env->vfp.regs[rd] = make_float64(d0);
}

void HELPER(neon_zip16)(uint32_t rd, uint32_t rm)
{
    uint64_t zm = float64_val(env->vfp.regs[rm]);
    uint64_t zd = float64_val(env->vfp.regs[rd]);
    uint64_t d0 = ELEM(zd, 0, 16) | (ELEM(zm, 0, 16) << 16)
        | (ELEM(zd, 1, 16) << 32) | (ELEM(zm, 1, 16) << 48);
    uint64_t m0 = ELEM(zd, 2, 16) | (ELEM(zm, 2, 16) << 16)
        | (ELEM(zd, 3, 16) << 32) | (ELEM(zm, 3, 16) << 48);
    env->vfp.regs[rm] = make_float64(m0);
    env->vfp.regs[rd] = make_float64(d0);
}