//! ARMv7 NEON intrinsics

use crate::{core_arch::simd_llvm::*, mem::transmute};
#[cfg(test)]
use stdarch_test::assert_instr;

types! {
    /// ARM-specific 64-bit wide vector of eight packed `i8`.
    pub struct int8x8_t(i8, i8, i8, i8, i8, i8, i8, i8);
    /// ARM-specific 64-bit wide vector of eight packed `u8`.
    pub struct uint8x8_t(u8, u8, u8, u8, u8, u8, u8, u8);
    /// ARM-specific 64-bit wide polynomial vector of eight packed `u8`.
    pub struct poly8x8_t(u8, u8, u8, u8, u8, u8, u8, u8);
    /// ARM-specific 64-bit wide vector of four packed `i16`.
    pub struct int16x4_t(i16, i16, i16, i16);
    /// ARM-specific 64-bit wide vector of four packed `u16`.
    pub struct uint16x4_t(u16, u16, u16, u16);
    // FIXME: ARM-specific 64-bit wide vector of four packed `f16`.
    // pub struct float16x4_t(f16, f16, f16, f16);
    /// ARM-specific 64-bit wide vector of four packed `u16`.
    pub struct poly16x4_t(u16, u16, u16, u16);
    /// ARM-specific 64-bit wide vector of two packed `i32`.
    pub struct int32x2_t(i32, i32);
    /// ARM-specific 64-bit wide vector of two packed `u32`.
    pub struct uint32x2_t(u32, u32);
    /// ARM-specific 64-bit wide vector of two packed `f32`.
    pub struct float32x2_t(f32, f32);
    /// ARM-specific 64-bit wide vector of one packed `i64`.
    pub struct int64x1_t(i64);
    /// ARM-specific 64-bit wide vector of one packed `u64`.
    pub struct uint64x1_t(u64);

    /// ARM-specific 128-bit wide vector of sixteen packed `i8`.
    pub struct int8x16_t(
        i8, i8 ,i8, i8, i8, i8 ,i8, i8,
        i8, i8 ,i8, i8, i8, i8 ,i8, i8,
    );
    /// ARM-specific 128-bit wide vector of sixteen packed `u8`.
    pub struct uint8x16_t(
        u8, u8 ,u8, u8, u8, u8 ,u8, u8,
        u8, u8 ,u8, u8, u8, u8 ,u8, u8,
    );
    /// ARM-specific 128-bit wide vector of sixteen packed `u8`.
    pub struct poly8x16_t(
        u8, u8, u8, u8, u8, u8, u8, u8,
        u8, u8, u8, u8, u8, u8, u8, u8
    );
    /// ARM-specific 128-bit wide vector of eight packed `i16`.
    pub struct int16x8_t(i16, i16, i16, i16, i16, i16, i16, i16);
    /// ARM-specific 128-bit wide vector of eight packed `u16`.
    pub struct uint16x8_t(u16, u16, u16, u16, u16, u16, u16, u16);
    // FIXME: ARM-specific 128-bit wide vector of eight packed `f16`.
    // pub struct float16x8_t(f16, f16, f16, f16, f16, f16, f16);
    /// ARM-specific 128-bit wide vector of eight packed `u16`.
    pub struct poly16x8_t(u16, u16, u16, u16, u16, u16, u16, u16);
    /// ARM-specific 128-bit wide vector of four packed `i32`.
    pub struct int32x4_t(i32, i32, i32, i32);
    /// ARM-specific 128-bit wide vector of four packed `u32`.
    pub struct uint32x4_t(u32, u32, u32, u32);
    /// ARM-specific 128-bit wide vector of four packed `f32`.
    pub struct float32x4_t(f32, f32, f32, f32);
    /// ARM-specific 128-bit wide vector of two packed `i64`.
    pub struct int64x2_t(i64, i64);
    /// ARM-specific 128-bit wide vector of two packed `u64`.
    pub struct uint64x2_t(u64, u64);
}

/// ARM-specific type containing two `int8x8_t` vectors.
#[derive(Copy, Clone)]
pub struct int8x8x2_t(pub int8x8_t, pub int8x8_t);
/// ARM-specific type containing three `int8x8_t` vectors.
#[derive(Copy, Clone)]
pub struct int8x8x3_t(pub int8x8_t, pub int8x8_t, pub int8x8_t);
/// ARM-specific type containing four `int8x8_t` vectors.
#[derive(Copy, Clone)]
pub struct int8x8x4_t(pub int8x8_t, pub int8x8_t, pub int8x8_t, pub int8x8_t);

/// ARM-specific type containing two `uint8x8_t` vectors.
#[derive(Copy, Clone)]
pub struct uint8x8x2_t(pub uint8x8_t, pub uint8x8_t);
/// ARM-specific type containing three `uint8x8_t` vectors.
#[derive(Copy, Clone)]
pub struct uint8x8x3_t(pub uint8x8_t, pub uint8x8_t, pub uint8x8_t);
/// ARM-specific type containing four `uint8x8_t` vectors.
#[derive(Copy, Clone)]
pub struct uint8x8x4_t(pub uint8x8_t, pub uint8x8_t, pub uint8x8_t, pub uint8x8_t);

/// ARM-specific type containing two `poly8x8_t` vectors.
#[derive(Copy, Clone)]
pub struct poly8x8x2_t(pub poly8x8_t, pub poly8x8_t);
/// ARM-specific type containing three `poly8x8_t` vectors.
#[derive(Copy, Clone)]
pub struct poly8x8x3_t(pub poly8x8_t, pub poly8x8_t, pub poly8x8_t);
/// ARM-specific type containing four `poly8x8_t` vectors.
#[derive(Copy, Clone)]
pub struct poly8x8x4_t(pub poly8x8_t, pub poly8x8_t, pub poly8x8_t, pub poly8x8_t);

#[allow(improper_ctypes)]
extern "C" {
    #[cfg_attr(target_arch = "aarch64", link_name = "llvm.aarch64.neon.frsqrte.v2f32")]
    #[cfg_attr(target_arch = "arm", link_name = "llvm.arm.neon.vrsqrte.v2f32")]
    fn frsqrte_v2f32(a: float32x2_t) -> float32x2_t;

    #[cfg_attr(target_arch = "arm", link_name = "llvm.arm.neon.vpmins.v8i8")]
    #[cfg_attr(target_arch = "aarch64", link_name = "llvm.aarch64.neon.sminp.v8i8")]
    fn vpmins_v8i8(a: int8x8_t, b: int8x8_t) -> int8x8_t;
    #[cfg_attr(target_arch = "arm", link_name = "llvm.arm.neon.vpmins.v4i16")]
    #[cfg_attr(target_arch = "aarch64", link_name = "llvm.aarch64.neon.sminp.v4i16")]
    fn vpmins_v4i16(a: int16x4_t, b: int16x4_t) -> int16x4_t;
    #[cfg_attr(target_arch = "arm", link_name = "llvm.arm.neon.vpmins.v2i32")]
    #[cfg_attr(target_arch = "aarch64", link_name = "llvm.aarch64.neon.sminp.v2i32")]
    fn vpmins_v2i32(a: int32x2_t, b: int32x2_t) -> int32x2_t;
    #[cfg_attr(target_arch = "arm", link_name = "llvm.arm.neon.vpminu.v8i8")]
    #[cfg_attr(target_arch = "aarch64", link_name = "llvm.aarch64.neon.uminp.v8i8")]
    fn vpminu_v8i8(a: uint8x8_t, b: uint8x8_t) -> uint8x8_t;
    #[cfg_attr(target_arch = "arm", link_name = "llvm.arm.neon.vpminu.v4i16")]
    #[cfg_attr(target_arch = "aarch64", link_name = "llvm.aarch64.neon.uminp.v4i16")]
    fn vpminu_v4i16(a: uint16x4_t, b: uint16x4_t) -> uint16x4_t;
    #[cfg_attr(target_arch = "arm", link_name = "llvm.arm.neon.vpminu.v2i32")]
    #[cfg_attr(target_arch = "aarch64", link_name = "llvm.aarch64.neon.uminp.v2i32")]
    fn vpminu_v2i32(a: uint32x2_t, b: uint32x2_t) -> uint32x2_t;
    #[cfg_attr(target_arch = "arm", link_name = "llvm.arm.neon.vpmins.v2f32")]
    #[cfg_attr(target_arch = "aarch64", link_name = "llvm.aarch64.neon.fminp.v2f32")]
    fn vpminf_v2f32(a: float32x2_t, b: float32x2_t) -> float32x2_t;

    #[cfg_attr(target_arch = "arm", link_name = "llvm.arm.neon.vpmaxs.v8i8")]
    #[cfg_attr(target_arch = "aarch64", link_name = "llvm.aarch64.neon.smaxp.v8i8")]
    fn vpmaxs_v8i8(a: int8x8_t, b: int8x8_t) -> int8x8_t;
    #[cfg_attr(target_arch = "arm", link_name = "llvm.arm.neon.vpmaxs.v4i16")]
    #[cfg_attr(target_arch = "aarch64", link_name = "llvm.aarch64.neon.smaxp.v4i16")]
    fn vpmaxs_v4i16(a: int16x4_t, b: int16x4_t) -> int16x4_t;
    #[cfg_attr(target_arch = "arm", link_name = "llvm.arm.neon.vpmaxs.v2i32")]
    #[cfg_attr(target_arch = "aarch64", link_name = "llvm.aarch64.neon.smaxp.v2i32")]
    fn vpmaxs_v2i32(a: int32x2_t, b: int32x2_t) -> int32x2_t;
    #[cfg_attr(target_arch = "arm", link_name = "llvm.arm.neon.vpmaxu.v8i8")]
    #[cfg_attr(target_arch = "aarch64", link_name = "llvm.aarch64.neon.umaxp.v8i8")]
    fn vpmaxu_v8i8(a: uint8x8_t, b: uint8x8_t) -> uint8x8_t;
    #[cfg_attr(target_arch = "arm", link_name = "llvm.arm.neon.vpmaxu.v4i16")]
    #[cfg_attr(target_arch = "aarch64", link_name = "llvm.aarch64.neon.umaxp.v4i16")]
    fn vpmaxu_v4i16(a: uint16x4_t, b: uint16x4_t) -> uint16x4_t;
    #[cfg_attr(target_arch = "arm", link_name = "llvm.arm.neon.vpmaxu.v2i32")]
    #[cfg_attr(target_arch = "aarch64", link_name = "llvm.aarch64.neon.umaxp.v2i32")]
    fn vpmaxu_v2i32(a: uint32x2_t, b: uint32x2_t) -> uint32x2_t;
    #[cfg_attr(target_arch = "arm", link_name = "llvm.arm.neon.vpmaxs.v2f32")]
    #[cfg_attr(target_arch = "aarch64", link_name = "llvm.aarch64.neon.fmaxp.v2f32")]
    fn vpmaxf_v2f32(a: float32x2_t, b: float32x2_t) -> float32x2_t;
}

#[cfg(target_arch = "arm")]
#[allow(improper_ctypes)]
extern "C" {
    #[link_name = "llvm.arm.neon.vtbl1"]
    fn vtbl1(a: int8x8_t, b: int8x8_t) -> int8x8_t;
    #[link_name = "llvm.arm.neon.vtbl2"]
    fn vtbl2(a: int8x8_t, b: int8x8_t, b: int8x8_t) -> int8x8_t;
    #[link_name = "llvm.arm.neon.vtbl3"]
    fn vtbl3(a: int8x8_t, b: int8x8_t, b: int8x8_t, c: int8x8_t) -> int8x8_t;
    #[link_name = "llvm.arm.neon.vtbl4"]
    fn vtbl4(a: int8x8_t, b: int8x8_t, b: int8x8_t, c: int8x8_t, d: int8x8_t) -> int8x8_t;

    #[link_name = "llvm.arm.neon.vtbx1"]
    fn vtbx1(a: int8x8_t, b: int8x8_t, b: int8x8_t) -> int8x8_t;
    #[link_name = "llvm.arm.neon.vtbx2"]
    fn vtbx2(a: int8x8_t, b: int8x8_t, b: int8x8_t, c: int8x8_t) -> int8x8_t;
    #[link_name = "llvm.arm.neon.vtbx3"]
    fn vtbx3(a: int8x8_t, b: int8x8_t, b: int8x8_t, c: int8x8_t, d: int8x8_t) -> int8x8_t;
    #[link_name = "llvm.arm.neon.vtbx4"]
    fn vtbx4(
        a: int8x8_t,
        b: int8x8_t,
        b: int8x8_t,
        c: int8x8_t,
        d: int8x8_t,
        e: int8x8_t,
    ) -> int8x8_t;
}

/// Vector add.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vadd))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(add))]
pub unsafe fn vadd_s8(a: int8x8_t, b: int8x8_t) -> int8x8_t {
    simd_add(a, b)
}

/// Vector add.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vadd))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(add))]
pub unsafe fn vaddq_s8(a: int8x16_t, b: int8x16_t) -> int8x16_t {
    simd_add(a, b)
}

/// Vector add.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vadd))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(add))]
pub unsafe fn vadd_s16(a: int16x4_t, b: int16x4_t) -> int16x4_t {
    simd_add(a, b)
}

/// Vector add.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vadd))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(add))]
pub unsafe fn vaddq_s16(a: int16x8_t, b: int16x8_t) -> int16x8_t {
    simd_add(a, b)
}

/// Vector add.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vadd))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(add))]
pub unsafe fn vadd_s32(a: int32x2_t, b: int32x2_t) -> int32x2_t {
    simd_add(a, b)
}

/// Vector add.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vadd))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(add))]
pub unsafe fn vaddq_s32(a: int32x4_t, b: int32x4_t) -> int32x4_t {
    simd_add(a, b)
}

/// Vector add.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vadd))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(add))]
pub unsafe fn vaddq_s64(a: int64x2_t, b: int64x2_t) -> int64x2_t {
    simd_add(a, b)
}

/// Vector add.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vadd))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(add))]
pub unsafe fn vadd_u8(a: uint8x8_t, b: uint8x8_t) -> uint8x8_t {
    simd_add(a, b)
}

/// Vector add.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vadd))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(add))]
pub unsafe fn vaddq_u8(a: uint8x16_t, b: uint8x16_t) -> uint8x16_t {
    simd_add(a, b)
}

/// Vector add.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vadd))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(add))]
pub unsafe fn vadd_u16(a: uint16x4_t, b: uint16x4_t) -> uint16x4_t {
    simd_add(a, b)
}

/// Vector add.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vadd))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(add))]
pub unsafe fn vaddq_u16(a: uint16x8_t, b: uint16x8_t) -> uint16x8_t {
    simd_add(a, b)
}

/// Vector add.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vadd))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(add))]
pub unsafe fn vadd_u32(a: uint32x2_t, b: uint32x2_t) -> uint32x2_t {
    simd_add(a, b)
}

/// Vector add.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vadd))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(add))]
pub unsafe fn vaddq_u32(a: uint32x4_t, b: uint32x4_t) -> uint32x4_t {
    simd_add(a, b)
}

/// Vector add.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vadd))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(add))]
pub unsafe fn vaddq_u64(a: uint64x2_t, b: uint64x2_t) -> uint64x2_t {
    simd_add(a, b)
}

/// Vector add.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vadd))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(fadd))]
pub unsafe fn vadd_f32(a: float32x2_t, b: float32x2_t) -> float32x2_t {
    simd_add(a, b)
}

/// Vector add.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vadd))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(fadd))]
pub unsafe fn vaddq_f32(a: float32x4_t, b: float32x4_t) -> float32x4_t {
    simd_add(a, b)
}

/// Vector long add.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vaddl))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(saddl))]
pub unsafe fn vaddl_s8(a: int8x8_t, b: int8x8_t) -> int16x8_t {
    let a: int16x8_t = simd_cast(a);
    let b: int16x8_t = simd_cast(b);
    simd_add(a, b)
}

/// Vector long add.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vaddl))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(saddl))]
pub unsafe fn vaddl_s16(a: int16x4_t, b: int16x4_t) -> int32x4_t {
    let a: int32x4_t = simd_cast(a);
    let b: int32x4_t = simd_cast(b);
    simd_add(a, b)
}

/// Vector long add.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vaddl))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(saddl))]
pub unsafe fn vaddl_s32(a: int32x2_t, b: int32x2_t) -> int64x2_t {
    let a: int64x2_t = simd_cast(a);
    let b: int64x2_t = simd_cast(b);
    simd_add(a, b)
}

/// Vector long add.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vaddl))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(uaddl))]
pub unsafe fn vaddl_u8(a: uint8x8_t, b: uint8x8_t) -> uint16x8_t {
    let a: uint16x8_t = simd_cast(a);
    let b: uint16x8_t = simd_cast(b);
    simd_add(a, b)
}

/// Vector long add.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vaddl))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(uaddl))]
pub unsafe fn vaddl_u16(a: uint16x4_t, b: uint16x4_t) -> uint32x4_t {
    let a: uint32x4_t = simd_cast(a);
    let b: uint32x4_t = simd_cast(b);
    simd_add(a, b)
}

/// Vector long add.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vaddl))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(uaddl))]
pub unsafe fn vaddl_u32(a: uint32x2_t, b: uint32x2_t) -> uint64x2_t {
    let a: uint64x2_t = simd_cast(a);
    let b: uint64x2_t = simd_cast(b);
    simd_add(a, b)
}

/// Vector narrow integer.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmovn))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(xtn))]
pub unsafe fn vmovn_s16(a: int16x8_t) -> int8x8_t {
    simd_cast(a)
}

/// Vector narrow integer.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmovn))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(xtn))]
pub unsafe fn vmovn_s32(a: int32x4_t) -> int16x4_t {
    simd_cast(a)
}

/// Vector narrow integer.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmovn))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(xtn))]
pub unsafe fn vmovn_s64(a: int64x2_t) -> int32x2_t {
    simd_cast(a)
}

/// Vector narrow integer.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmovn))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(xtn))]
pub unsafe fn vmovn_u16(a: uint16x8_t) -> uint8x8_t {
    simd_cast(a)
}

/// Vector narrow integer.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmovn))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(xtn))]
pub unsafe fn vmovn_u32(a: uint32x4_t) -> uint16x4_t {
    simd_cast(a)
}

/// Vector narrow integer.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmovn))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(xtn))]
pub unsafe fn vmovn_u64(a: uint64x2_t) -> uint32x2_t {
    simd_cast(a)
}

/// Vector long move.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmovl))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(sxtl))]
pub unsafe fn vmovl_s8(a: int8x8_t) -> int16x8_t {
    simd_cast(a)
}

/// Vector long move.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmovl))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(sxtl))]
pub unsafe fn vmovl_s16(a: int16x4_t) -> int32x4_t {
    simd_cast(a)
}

/// Vector long move.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmovl))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(sxtl))]
pub unsafe fn vmovl_s32(a: int32x2_t) -> int64x2_t {
    simd_cast(a)
}

/// Vector long move.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmovl))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(uxtl))]
pub unsafe fn vmovl_u8(a: uint8x8_t) -> uint16x8_t {
    simd_cast(a)
}

/// Vector long move.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmovl))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(uxtl))]
pub unsafe fn vmovl_u16(a: uint16x4_t) -> uint32x4_t {
    simd_cast(a)
}

/// Vector long move.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmovl))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(uxtl))]
pub unsafe fn vmovl_u32(a: uint32x2_t) -> uint64x2_t {
    simd_cast(a)
}

/// Reciprocal square-root estimate.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(frsqrte))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vrsqrte))]
pub unsafe fn vrsqrte_f32(a: float32x2_t) -> float32x2_t {
    frsqrte_v2f32(a)
}

/// Vector bitwise not.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmvn))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(mvn))]
pub unsafe fn vmvn_s8(a: int8x8_t) -> int8x8_t {
    let b = int8x8_t(-1, -1, -1, -1, -1, -1, -1, -1);
    simd_xor(a, b)
}

/// Vector bitwise not.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmvn))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(mvn))]
pub unsafe fn vmvnq_s8(a: int8x16_t) -> int8x16_t {
    let b = int8x16_t(
        -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1,
    );
    simd_xor(a, b)
}

/// Vector bitwise not.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmvn))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(mvn))]
pub unsafe fn vmvn_s16(a: int16x4_t) -> int16x4_t {
    let b = int16x4_t(-1, -1, -1, -1);
    simd_xor(a, b)
}

/// Vector bitwise not.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmvn))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(mvn))]
pub unsafe fn vmvnq_s16(a: int16x8_t) -> int16x8_t {
    let b = int16x8_t(-1, -1, -1, -1, -1, -1, -1, -1);
    simd_xor(a, b)
}

/// Vector bitwise not.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmvn))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(mvn))]
pub unsafe fn vmvn_s32(a: int32x2_t) -> int32x2_t {
    let b = int32x2_t(-1, -1);
    simd_xor(a, b)
}

/// Vector bitwise not.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmvn))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(mvn))]
pub unsafe fn vmvnq_s32(a: int32x4_t) -> int32x4_t {
    let b = int32x4_t(-1, -1, -1, -1);
    simd_xor(a, b)
}

/// Vector bitwise not.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmvn))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(mvn))]
pub unsafe fn vmvn_u8(a: uint8x8_t) -> uint8x8_t {
    let b = uint8x8_t(255, 255, 255, 255, 255, 255, 255, 255);
    simd_xor(a, b)
}

/// Vector bitwise not.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmvn))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(mvn))]
pub unsafe fn vmvnq_u8(a: uint8x16_t) -> uint8x16_t {
    let b = uint8x16_t(
        255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255,
    );
    simd_xor(a, b)
}

/// Vector bitwise not.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmvn))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(mvn))]
pub unsafe fn vmvn_u16(a: uint16x4_t) -> uint16x4_t {
    let b = uint16x4_t(65_535, 65_535, 65_535, 65_535);
    simd_xor(a, b)
}

/// Vector bitwise not.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmvn))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(mvn))]
pub unsafe fn vmvnq_u16(a: uint16x8_t) -> uint16x8_t {
    let b = uint16x8_t(
        65_535, 65_535, 65_535, 65_535, 65_535, 65_535, 65_535, 65_535,
    );
    simd_xor(a, b)
}

/// Vector bitwise not.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmvn))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(mvn))]
pub unsafe fn vmvn_u32(a: uint32x2_t) -> uint32x2_t {
    let b = uint32x2_t(4_294_967_295, 4_294_967_295);
    simd_xor(a, b)
}

/// Vector bitwise not.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmvn))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(mvn))]
pub unsafe fn vmvnq_u32(a: uint32x4_t) -> uint32x4_t {
    let b = uint32x4_t(4_294_967_295, 4_294_967_295, 4_294_967_295, 4_294_967_295);
    simd_xor(a, b)
}

/// Vector bitwise not.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmvn))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(mvn))]
pub unsafe fn vmvn_p8(a: poly8x8_t) -> poly8x8_t {
    let b = poly8x8_t(255, 255, 255, 255, 255, 255, 255, 255);
    simd_xor(a, b)
}

/// Vector bitwise not.
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vmvn))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(mvn))]
pub unsafe fn vmvnq_p8(a: poly8x16_t) -> poly8x16_t {
    let b = poly8x16_t(
        255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255,
    );
    simd_xor(a, b)
}

/// Folding minimum of adjacent pairs
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vpmin))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(sminp))]
pub unsafe fn vpmin_s8(a: int8x8_t, b: int8x8_t) -> int8x8_t {
    vpmins_v8i8(a, b)
}

/// Folding minimum of adjacent pairs
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vpmin))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(sminp))]
pub unsafe fn vpmin_s16(a: int16x4_t, b: int16x4_t) -> int16x4_t {
    vpmins_v4i16(a, b)
}

/// Folding minimum of adjacent pairs
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vpmin))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(sminp))]
pub unsafe fn vpmin_s32(a: int32x2_t, b: int32x2_t) -> int32x2_t {
    vpmins_v2i32(a, b)
}

/// Folding minimum of adjacent pairs
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vpmin))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(uminp))]
pub unsafe fn vpmin_u8(a: uint8x8_t, b: uint8x8_t) -> uint8x8_t {
    vpminu_v8i8(a, b)
}

/// Folding minimum of adjacent pairs
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vpmin))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(uminp))]
pub unsafe fn vpmin_u16(a: uint16x4_t, b: uint16x4_t) -> uint16x4_t {
    vpminu_v4i16(a, b)
}

/// Folding minimum of adjacent pairs
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vpmin))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(uminp))]
pub unsafe fn vpmin_u32(a: uint32x2_t, b: uint32x2_t) -> uint32x2_t {
    vpminu_v2i32(a, b)
}

/// Folding minimum of adjacent pairs
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vpmin))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(fminp))]
pub unsafe fn vpmin_f32(a: float32x2_t, b: float32x2_t) -> float32x2_t {
    vpminf_v2f32(a, b)
}

/// Folding maximum of adjacent pairs
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vpmax))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(smaxp))]
pub unsafe fn vpmax_s8(a: int8x8_t, b: int8x8_t) -> int8x8_t {
    vpmaxs_v8i8(a, b)
}

/// Folding maximum of adjacent pairs
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vpmax))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(smaxp))]
pub unsafe fn vpmax_s16(a: int16x4_t, b: int16x4_t) -> int16x4_t {
    vpmaxs_v4i16(a, b)
}

/// Folding maximum of adjacent pairs
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vpmax))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(smaxp))]
pub unsafe fn vpmax_s32(a: int32x2_t, b: int32x2_t) -> int32x2_t {
    vpmaxs_v2i32(a, b)
}

/// Folding maximum of adjacent pairs
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vpmax))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(umaxp))]
pub unsafe fn vpmax_u8(a: uint8x8_t, b: uint8x8_t) -> uint8x8_t {
    vpmaxu_v8i8(a, b)
}

/// Folding maximum of adjacent pairs
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vpmax))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(umaxp))]
pub unsafe fn vpmax_u16(a: uint16x4_t, b: uint16x4_t) -> uint16x4_t {
    vpmaxu_v4i16(a, b)
}

/// Folding maximum of adjacent pairs
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vpmax))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(umaxp))]
pub unsafe fn vpmax_u32(a: uint32x2_t, b: uint32x2_t) -> uint32x2_t {
    vpmaxu_v2i32(a, b)
}

/// Folding maximum of adjacent pairs
#[inline]
#[target_feature(enable = "neon")]
#[cfg_attr(target_arch = "arm", target_feature(enable = "v7"))]
#[cfg_attr(all(test, target_arch = "arm"), assert_instr(vpmax))]
#[cfg_attr(all(test, target_arch = "aarch64"), assert_instr(fmaxp))]
pub unsafe fn vpmax_f32(a: float32x2_t, b: float32x2_t) -> float32x2_t {
    vpmaxf_v2f32(a, b)
}

/// Table look-up
#[inline]
#[cfg(target_arch = "arm")]
#[cfg(target_endian = "little")]
#[target_feature(enable = "neon,v7")]
#[cfg_attr(test, assert_instr(vtbl))]
pub unsafe fn vtbl1_s8(a: int8x8_t, b: int8x8_t) -> int8x8_t {
    vtbl1(a, b)
}

/// Table look-up
#[inline]
#[cfg(target_arch = "arm")]
#[cfg(target_endian = "little")]
#[target_feature(enable = "neon,v7")]
#[cfg_attr(test, assert_instr(vtbl))]
pub unsafe fn vtbl1_u8(a: uint8x8_t, b: uint8x8_t) -> uint8x8_t {
    transmute(vtbl1(transmute(a), transmute(b)))
}

/// Table look-up
#[inline]
#[cfg(target_arch = "arm")]
#[cfg(target_endian = "little")]
#[target_feature(enable = "neon,v7")]
#[cfg_attr(test, assert_instr(vtbl))]
pub unsafe fn vtbl1_p8(a: poly8x8_t, b: uint8x8_t) -> poly8x8_t {
    transmute(vtbl1(transmute(a), transmute(b)))
}

/// Table look-up
#[inline]
#[cfg(target_arch = "arm")]
#[cfg(target_endian = "little")]
#[target_feature(enable = "neon,v7")]
#[cfg_attr(test, assert_instr(vtbl))]
pub unsafe fn vtbl2_s8(a: int8x8x2_t, b: int8x8_t) -> int8x8_t {
    vtbl2(a.0, a.1, b)
}

/// Table look-up
#[inline]
#[cfg(target_arch = "arm")]
#[cfg(target_endian = "little")]
#[target_feature(enable = "neon,v7")]
#[cfg_attr(test, assert_instr(vtbl))]
pub unsafe fn vtbl2_u8(a: uint8x8x2_t, b: uint8x8_t) -> uint8x8_t {
    transmute(vtbl2(transmute(a.0), transmute(a.1), transmute(b)))
}

/// Table look-up
#[inline]
#[cfg(target_arch = "arm")]
#[cfg(target_endian = "little")]
#[target_feature(enable = "neon,v7")]
#[cfg_attr(test, assert_instr(vtbl))]
pub unsafe fn vtbl2_p8(a: poly8x8x2_t, b: uint8x8_t) -> poly8x8_t {
    transmute(vtbl2(transmute(a.0), transmute(a.1), transmute(b)))
}

/// Table look-up
#[inline]
#[cfg(target_arch = "arm")]
#[cfg(target_endian = "little")]
#[target_feature(enable = "neon,v7")]
#[cfg_attr(test, assert_instr(vtbl))]
pub unsafe fn vtbl3_s8(a: int8x8x3_t, b: int8x8_t) -> int8x8_t {
    vtbl3(a.0, a.1, a.2, b)
}

/// Table look-up
#[inline]
#[cfg(target_arch = "arm")]
#[cfg(target_endian = "little")]
#[target_feature(enable = "neon,v7")]
#[cfg_attr(test, assert_instr(vtbl))]
pub unsafe fn vtbl3_u8(a: uint8x8x3_t, b: uint8x8_t) -> uint8x8_t {
    transmute(vtbl3(
        transmute(a.0),
        transmute(a.1),
        transmute(a.2),
        transmute(b),
    ))
}

/// Table look-up
#[inline]
#[cfg(target_arch = "arm")]
#[cfg(target_endian = "little")]
#[target_feature(enable = "neon,v7")]
#[cfg_attr(test, assert_instr(vtbl))]
pub unsafe fn vtbl3_p8(a: poly8x8x3_t, b: uint8x8_t) -> poly8x8_t {
    transmute(vtbl3(
        transmute(a.0),
        transmute(a.1),
        transmute(a.2),
        transmute(b),
    ))
}

/// Table look-up
#[inline]
#[cfg(target_arch = "arm")]
#[cfg(target_endian = "little")]
#[target_feature(enable = "neon,v7")]
#[cfg_attr(test, assert_instr(vtbl))]
pub unsafe fn vtbl4_s8(a: int8x8x4_t, b: int8x8_t) -> int8x8_t {
    vtbl4(a.0, a.1, a.2, a.3, b)
}

/// Table look-up
#[inline]
#[cfg(target_arch = "arm")]
#[cfg(target_endian = "little")]
#[target_feature(enable = "neon,v7")]
#[cfg_attr(test, assert_instr(vtbl))]
pub unsafe fn vtbl4_u8(a: uint8x8x4_t, b: uint8x8_t) -> uint8x8_t {
    transmute(vtbl4(
        transmute(a.0),
        transmute(a.1),
        transmute(a.2),
        transmute(a.3),
        transmute(b),
    ))
}

/// Table look-up
#[inline]
#[cfg(target_arch = "arm")]
#[cfg(target_endian = "little")]
#[target_feature(enable = "neon,v7")]
#[cfg_attr(test, assert_instr(vtbl))]
pub unsafe fn vtbl4_p8(a: poly8x8x4_t, b: uint8x8_t) -> poly8x8_t {
    transmute(vtbl4(
        transmute(a.0),
        transmute(a.1),
        transmute(a.2),
        transmute(a.3),
        transmute(b),
    ))
}

/// Extended table look-up
#[inline]
#[cfg(target_arch = "arm")]
#[cfg(target_endian = "little")]
#[target_feature(enable = "neon,v7")]
#[cfg_attr(test, assert_instr(vtbx))]
pub unsafe fn vtbx1_s8(a: int8x8_t, b: int8x8_t, c: int8x8_t) -> int8x8_t {
    vtbx1(a, b, c)
}

/// Extended table look-up
#[inline]
#[cfg(target_arch = "arm")]
#[cfg(target_endian = "little")]
#[target_feature(enable = "neon,v7")]
#[cfg_attr(test, assert_instr(vtbx))]
pub unsafe fn vtbx1_u8(a: uint8x8_t, b: uint8x8_t, c: uint8x8_t) -> uint8x8_t {
    transmute(vtbx1(transmute(a), transmute(b), transmute(c)))
}

/// Extended table look-up
#[inline]
#[cfg(target_arch = "arm")]
#[cfg(target_endian = "little")]
#[target_feature(enable = "neon,v7")]
#[cfg_attr(test, assert_instr(vtbx))]
pub unsafe fn vtbx1_p8(a: poly8x8_t, b: poly8x8_t, c: uint8x8_t) -> poly8x8_t {
    transmute(vtbx1(transmute(a), transmute(b), transmute(c)))
}

/// Extended table look-up
#[inline]
#[cfg(target_arch = "arm")]
#[cfg(target_endian = "little")]
#[target_feature(enable = "neon,v7")]
#[cfg_attr(test, assert_instr(vtbx))]
pub unsafe fn vtbx2_s8(a: int8x8_t, b: int8x8x2_t, c: int8x8_t) -> int8x8_t {
    vtbx2(a, b.0, b.1, c)
}

/// Extended table look-up
#[inline]
#[cfg(target_arch = "arm")]
#[cfg(target_endian = "little")]
#[target_feature(enable = "neon,v7")]
#[cfg_attr(test, assert_instr(vtbx))]
pub unsafe fn vtbx2_u8(a: uint8x8_t, b: uint8x8x2_t, c: uint8x8_t) -> uint8x8_t {
    transmute(vtbx2(
        transmute(a),
        transmute(b.0),
        transmute(b.1),
        transmute(c),
    ))
}

/// Extended table look-up
#[inline]
#[cfg(target_arch = "arm")]
#[cfg(target_endian = "little")]
#[target_feature(enable = "neon,v7")]
#[cfg_attr(test, assert_instr(vtbx))]
pub unsafe fn vtbx2_p8(a: poly8x8_t, b: poly8x8x2_t, c: uint8x8_t) -> poly8x8_t {
    transmute(vtbx2(
        transmute(a),
        transmute(b.0),
        transmute(b.1),
        transmute(c),
    ))
}

/// Extended table look-up
#[inline]
#[cfg(target_arch = "arm")]
#[cfg(target_endian = "little")]
#[target_feature(enable = "neon,v7")]
#[cfg_attr(test, assert_instr(vtbx))]
pub unsafe fn vtbx3_s8(a: int8x8_t, b: int8x8x3_t, c: int8x8_t) -> int8x8_t {
    vtbx3(a, b.0, b.1, b.2, c)
}

/// Extended table look-up
#[inline]
#[cfg(target_arch = "arm")]
#[cfg(target_endian = "little")]
#[target_feature(enable = "neon,v7")]
#[cfg_attr(test, assert_instr(vtbx))]
pub unsafe fn vtbx3_u8(a: uint8x8_t, b: uint8x8x3_t, c: uint8x8_t) -> uint8x8_t {
    transmute(vtbx3(
        transmute(a),
        transmute(b.0),
        transmute(b.1),
        transmute(b.2),
        transmute(c),
    ))
}

/// Extended table look-up
#[inline]
#[cfg(target_arch = "arm")]
#[cfg(target_endian = "little")]
#[target_feature(enable = "neon,v7")]
#[cfg_attr(test, assert_instr(vtbx))]
pub unsafe fn vtbx3_p8(a: poly8x8_t, b: poly8x8x3_t, c: uint8x8_t) -> poly8x8_t {
    transmute(vtbx3(
        transmute(a),
        transmute(b.0),
        transmute(b.1),
        transmute(b.2),
        transmute(c),
    ))
}

/// Extended table look-up
#[inline]
#[cfg(target_arch = "arm")]
#[cfg(target_endian = "little")]
#[target_feature(enable = "neon,v7")]
#[cfg_attr(test, assert_instr(vtbx))]
pub unsafe fn vtbx4_s8(a: int8x8_t, b: int8x8x4_t, c: int8x8_t) -> int8x8_t {
    vtbx4(a, b.0, b.1, b.2, b.3, c)
}

/// Extended table look-up
#[inline]
#[cfg(target_arch = "arm")]
#[cfg(target_endian = "little")]
#[target_feature(enable = "neon,v7")]
#[cfg_attr(test, assert_instr(vtbx))]
pub unsafe fn vtbx4_u8(a: uint8x8_t, b: uint8x8x4_t, c: uint8x8_t) -> uint8x8_t {
    transmute(vtbx4(
        transmute(a),
        transmute(b.0),
        transmute(b.1),
        transmute(b.2),
        transmute(b.3),
        transmute(c),
    ))
}

/// Extended table look-up
#[inline]
#[cfg(target_arch = "arm")]
#[cfg(target_endian = "little")]
#[target_feature(enable = "neon,v7")]
#[cfg_attr(test, assert_instr(vtbx))]
pub unsafe fn vtbx4_p8(a: poly8x8_t, b: poly8x8x4_t, c: uint8x8_t) -> poly8x8_t {
    transmute(vtbx4(
        transmute(a),
        transmute(b.0),
        transmute(b.1),
        transmute(b.2),
        transmute(b.3),
        transmute(c),
    ))
}

#[cfg(test)]
mod tests {
    use crate::core_arch::{arm::*, simd::*};
    use std::{i16, i32, i8, mem::transmute, u16, u32, u8};
    use stdarch_test::simd_test;

    #[simd_test(enable = "neon")]
    unsafe fn test_vadd_s8() {
        let a = i8x8::new(1, 2, 3, 4, 5, 6, 7, 8);
        let b = i8x8::new(8, 7, 6, 5, 4, 3, 2, 1);
        let e = i8x8::new(9, 9, 9, 9, 9, 9, 9, 9);
        let r: i8x8 = transmute(vadd_s8(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vaddq_s8() {
        let a = i8x16::new(1, 2, 3, 4, 5, 6, 7, 8, 1, 2, 3, 4, 5, 6, 7, 8);
        let b = i8x16::new(8, 7, 6, 5, 4, 3, 2, 1, 8, 7, 6, 5, 4, 3, 2, 1);
        let e = i8x16::new(9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9);
        let r: i8x16 = transmute(vaddq_s8(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vadd_s16() {
        let a = i16x4::new(1, 2, 3, 4);
        let b = i16x4::new(8, 7, 6, 5);
        let e = i16x4::new(9, 9, 9, 9);
        let r: i16x4 = transmute(vadd_s16(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vaddq_s16() {
        let a = i16x8::new(1, 2, 3, 4, 5, 6, 7, 8);
        let b = i16x8::new(8, 7, 6, 5, 4, 3, 2, 1);
        let e = i16x8::new(9, 9, 9, 9, 9, 9, 9, 9);
        let r: i16x8 = transmute(vaddq_s16(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vadd_s32() {
        let a = i32x2::new(1, 2);
        let b = i32x2::new(8, 7);
        let e = i32x2::new(9, 9);
        let r: i32x2 = transmute(vadd_s32(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vaddq_s32() {
        let a = i32x4::new(1, 2, 3, 4);
        let b = i32x4::new(8, 7, 6, 5);
        let e = i32x4::new(9, 9, 9, 9);
        let r: i32x4 = transmute(vaddq_s32(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vadd_u8() {
        let a = u8x8::new(1, 2, 3, 4, 5, 6, 7, 8);
        let b = u8x8::new(8, 7, 6, 5, 4, 3, 2, 1);
        let e = u8x8::new(9, 9, 9, 9, 9, 9, 9, 9);
        let r: u8x8 = transmute(vadd_u8(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vaddq_u8() {
        let a = u8x16::new(1, 2, 3, 4, 5, 6, 7, 8, 1, 2, 3, 4, 5, 6, 7, 8);
        let b = u8x16::new(8, 7, 6, 5, 4, 3, 2, 1, 8, 7, 6, 5, 4, 3, 2, 1);
        let e = u8x16::new(9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9);
        let r: u8x16 = transmute(vaddq_u8(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vadd_u16() {
        let a = u16x4::new(1, 2, 3, 4);
        let b = u16x4::new(8, 7, 6, 5);
        let e = u16x4::new(9, 9, 9, 9);
        let r: u16x4 = transmute(vadd_u16(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vaddq_u16() {
        let a = u16x8::new(1, 2, 3, 4, 5, 6, 7, 8);
        let b = u16x8::new(8, 7, 6, 5, 4, 3, 2, 1);
        let e = u16x8::new(9, 9, 9, 9, 9, 9, 9, 9);
        let r: u16x8 = transmute(vaddq_u16(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vadd_u32() {
        let a = u32x2::new(1, 2);
        let b = u32x2::new(8, 7);
        let e = u32x2::new(9, 9);
        let r: u32x2 = transmute(vadd_u32(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vaddq_u32() {
        let a = u32x4::new(1, 2, 3, 4);
        let b = u32x4::new(8, 7, 6, 5);
        let e = u32x4::new(9, 9, 9, 9);
        let r: u32x4 = transmute(vaddq_u32(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vadd_f32() {
        let a = f32x2::new(1., 2.);
        let b = f32x2::new(8., 7.);
        let e = f32x2::new(9., 9.);
        let r: f32x2 = transmute(vadd_f32(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vaddq_f32() {
        let a = f32x4::new(1., 2., 3., 4.);
        let b = f32x4::new(8., 7., 6., 5.);
        let e = f32x4::new(9., 9., 9., 9.);
        let r: f32x4 = transmute(vaddq_f32(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vaddl_s8() {
        let v = i8::MAX;
        let a = i8x8::new(v, v, v, v, v, v, v, v);
        let v = 2 * (v as i16);
        let e = i16x8::new(v, v, v, v, v, v, v, v);
        let r: i16x8 = transmute(vaddl_s8(transmute(a), transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vaddl_s16() {
        let v = i16::MAX;
        let a = i16x4::new(v, v, v, v);
        let v = 2 * (v as i32);
        let e = i32x4::new(v, v, v, v);
        let r: i32x4 = transmute(vaddl_s16(transmute(a), transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vaddl_s32() {
        let v = i32::MAX;
        let a = i32x2::new(v, v);
        let v = 2 * (v as i64);
        let e = i64x2::new(v, v);
        let r: i64x2 = transmute(vaddl_s32(transmute(a), transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vaddl_u8() {
        let v = u8::MAX;
        let a = u8x8::new(v, v, v, v, v, v, v, v);
        let v = 2 * (v as u16);
        let e = u16x8::new(v, v, v, v, v, v, v, v);
        let r: u16x8 = transmute(vaddl_u8(transmute(a), transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vaddl_u16() {
        let v = u16::MAX;
        let a = u16x4::new(v, v, v, v);
        let v = 2 * (v as u32);
        let e = u32x4::new(v, v, v, v);
        let r: u32x4 = transmute(vaddl_u16(transmute(a), transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vaddl_u32() {
        let v = u32::MAX;
        let a = u32x2::new(v, v);
        let v = 2 * (v as u64);
        let e = u64x2::new(v, v);
        let r: u64x2 = transmute(vaddl_u32(transmute(a), transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmvn_s8() {
        let a = i8x8::new(0, 1, 2, 3, 4, 5, 6, 7);
        let e = i8x8::new(-1, -2, -3, -4, -5, -6, -7, -8);
        let r: i8x8 = transmute(vmvn_s8(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmvnq_s8() {
        let a = i8x16::new(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15);
        let e = i8x16::new(
            -1, -2, -3, -4, -5, -6, -7, -8, -9, -10, -11, -12, -13, -14, -15, -16,
        );
        let r: i8x16 = transmute(vmvnq_s8(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmvn_s16() {
        let a = i16x4::new(0, 1, 2, 3);
        let e = i16x4::new(-1, -2, -3, -4);
        let r: i16x4 = transmute(vmvn_s16(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmvnq_s16() {
        let a = i16x8::new(0, 1, 2, 3, 4, 5, 6, 7);
        let e = i16x8::new(-1, -2, -3, -4, -5, -6, -7, -8);
        let r: i16x8 = transmute(vmvnq_s16(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmvn_s32() {
        let a = i32x2::new(0, 1);
        let e = i32x2::new(-1, -2);
        let r: i32x2 = transmute(vmvn_s32(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmvnq_s32() {
        let a = i32x4::new(0, 1, 2, 3);
        let e = i32x4::new(-1, -2, -3, -4);
        let r: i32x4 = transmute(vmvnq_s32(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmvn_u8() {
        let a = u8x8::new(0, 1, 2, 3, 4, 5, 6, 7);
        let e = u8x8::new(255, 254, 253, 252, 251, 250, 249, 248);
        let r: u8x8 = transmute(vmvn_u8(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmvnq_u8() {
        let a = u8x16::new(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15);
        let e = u8x16::new(
            255, 254, 253, 252, 251, 250, 249, 248, 247, 246, 245, 244, 243, 242, 241, 240,
        );
        let r: u8x16 = transmute(vmvnq_u8(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmvn_u16() {
        let a = u16x4::new(0, 1, 2, 3);
        let e = u16x4::new(65_535, 65_534, 65_533, 65_532);
        let r: u16x4 = transmute(vmvn_u16(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmvnq_u16() {
        let a = u16x8::new(0, 1, 2, 3, 4, 5, 6, 7);
        let e = u16x8::new(
            65_535, 65_534, 65_533, 65_532, 65_531, 65_530, 65_529, 65_528,
        );
        let r: u16x8 = transmute(vmvnq_u16(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmvn_u32() {
        let a = u32x2::new(0, 1);
        let e = u32x2::new(4_294_967_295, 4_294_967_294);
        let r: u32x2 = transmute(vmvn_u32(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmvnq_u32() {
        let a = u32x4::new(0, 1, 2, 3);
        let e = u32x4::new(4_294_967_295, 4_294_967_294, 4_294_967_293, 4_294_967_292);
        let r: u32x4 = transmute(vmvnq_u32(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmvn_p8() {
        let a = u8x8::new(0, 1, 2, 3, 4, 5, 6, 7);
        let e = u8x8::new(255, 254, 253, 252, 251, 250, 249, 248);
        let r: u8x8 = transmute(vmvn_p8(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmvnq_p8() {
        let a = u8x16::new(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15);
        let e = u8x16::new(
            255, 254, 253, 252, 251, 250, 249, 248, 247, 246, 245, 244, 243, 242, 241, 240,
        );
        let r: u8x16 = transmute(vmvnq_p8(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmovn_s16() {
        let a = i16x8::new(1, 2, 3, 4, 5, 6, 7, 8);
        let e = i8x8::new(1, 2, 3, 4, 5, 6, 7, 8);
        let r: i8x8 = transmute(vmovn_s16(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmovn_s32() {
        let a = i32x4::new(1, 2, 3, 4);
        let e = i16x4::new(1, 2, 3, 4);
        let r: i16x4 = transmute(vmovn_s32(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmovn_s64() {
        let a = i64x2::new(1, 2);
        let e = i32x2::new(1, 2);
        let r: i32x2 = transmute(vmovn_s64(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmovn_u16() {
        let a = u16x8::new(1, 2, 3, 4, 5, 6, 7, 8);
        let e = u8x8::new(1, 2, 3, 4, 5, 6, 7, 8);
        let r: u8x8 = transmute(vmovn_u16(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmovn_u32() {
        let a = u32x4::new(1, 2, 3, 4);
        let e = u16x4::new(1, 2, 3, 4);
        let r: u16x4 = transmute(vmovn_u32(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmovn_u64() {
        let a = u64x2::new(1, 2);
        let e = u32x2::new(1, 2);
        let r: u32x2 = transmute(vmovn_u64(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmovl_s8() {
        let e = i16x8::new(1, 2, 3, 4, 5, 6, 7, 8);
        let a = i8x8::new(1, 2, 3, 4, 5, 6, 7, 8);
        let r: i16x8 = transmute(vmovl_s8(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmovl_s16() {
        let e = i32x4::new(1, 2, 3, 4);
        let a = i16x4::new(1, 2, 3, 4);
        let r: i32x4 = transmute(vmovl_s16(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmovl_s32() {
        let e = i64x2::new(1, 2);
        let a = i32x2::new(1, 2);
        let r: i64x2 = transmute(vmovl_s32(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmovl_u8() {
        let e = u16x8::new(1, 2, 3, 4, 5, 6, 7, 8);
        let a = u8x8::new(1, 2, 3, 4, 5, 6, 7, 8);
        let r: u16x8 = transmute(vmovl_u8(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmovl_u16() {
        let e = u32x4::new(1, 2, 3, 4);
        let a = u16x4::new(1, 2, 3, 4);
        let r: u32x4 = transmute(vmovl_u16(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vmovl_u32() {
        let e = u64x2::new(1, 2);
        let a = u32x2::new(1, 2);
        let r: u64x2 = transmute(vmovl_u32(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vrsqrt_f32() {
        let a = f32x2::new(1.0, 2.0);
        let e = f32x2::new(0.9980469, 0.7050781);
        let r: f32x2 = transmute(vrsqrte_f32(transmute(a)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vpmin_s8() {
        let a = i8x8::new(1, -2, 3, -4, 5, 6, 7, 8);
        let b = i8x8::new(0, 3, 2, 5, 4, 7, 6, 9);
        let e = i8x8::new(-2, -4, 5, 7, 0, 2, 4, 6);
        let r: i8x8 = transmute(vpmin_s8(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vpmin_s16() {
        let a = i16x4::new(1, 2, 3, -4);
        let b = i16x4::new(0, 3, 2, 5);
        let e = i16x4::new(1, -4, 0, 2);
        let r: i16x4 = transmute(vpmin_s16(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vpmin_s32() {
        let a = i32x2::new(1, -2);
        let b = i32x2::new(0, 3);
        let e = i32x2::new(-2, 0);
        let r: i32x2 = transmute(vpmin_s32(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vpmin_u8() {
        let a = u8x8::new(1, 2, 3, 4, 5, 6, 7, 8);
        let b = u8x8::new(0, 3, 2, 5, 4, 7, 6, 9);
        let e = u8x8::new(1, 3, 5, 7, 0, 2, 4, 6);
        let r: u8x8 = transmute(vpmin_u8(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vpmin_u16() {
        let a = u16x4::new(1, 2, 3, 4);
        let b = u16x4::new(0, 3, 2, 5);
        let e = u16x4::new(1, 3, 0, 2);
        let r: u16x4 = transmute(vpmin_u16(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vpmin_u32() {
        let a = u32x2::new(1, 2);
        let b = u32x2::new(0, 3);
        let e = u32x2::new(1, 0);
        let r: u32x2 = transmute(vpmin_u32(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vpmin_f32() {
        let a = f32x2::new(1., -2.);
        let b = f32x2::new(0., 3.);
        let e = f32x2::new(-2., 0.);
        let r: f32x2 = transmute(vpmin_f32(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vpmax_s8() {
        let a = i8x8::new(1, -2, 3, -4, 5, 6, 7, 8);
        let b = i8x8::new(0, 3, 2, 5, 4, 7, 6, 9);
        let e = i8x8::new(1, 3, 6, 8, 3, 5, 7, 9);
        let r: i8x8 = transmute(vpmax_s8(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vpmax_s16() {
        let a = i16x4::new(1, 2, 3, -4);
        let b = i16x4::new(0, 3, 2, 5);
        let e = i16x4::new(2, 3, 3, 5);
        let r: i16x4 = transmute(vpmax_s16(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vpmax_s32() {
        let a = i32x2::new(1, -2);
        let b = i32x2::new(0, 3);
        let e = i32x2::new(1, 3);
        let r: i32x2 = transmute(vpmax_s32(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vpmax_u8() {
        let a = u8x8::new(1, 2, 3, 4, 5, 6, 7, 8);
        let b = u8x8::new(0, 3, 2, 5, 4, 7, 6, 9);
        let e = u8x8::new(2, 4, 6, 8, 3, 5, 7, 9);
        let r: u8x8 = transmute(vpmax_u8(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vpmax_u16() {
        let a = u16x4::new(1, 2, 3, 4);
        let b = u16x4::new(0, 3, 2, 5);
        let e = u16x4::new(2, 4, 3, 5);
        let r: u16x4 = transmute(vpmax_u16(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vpmax_u32() {
        let a = u32x2::new(1, 2);
        let b = u32x2::new(0, 3);
        let e = u32x2::new(2, 3);
        let r: u32x2 = transmute(vpmax_u32(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }

    #[simd_test(enable = "neon")]
    unsafe fn test_vpmax_f32() {
        let a = f32x2::new(1., -2.);
        let b = f32x2::new(0., 3.);
        let e = f32x2::new(1., 3.);
        let r: f32x2 = transmute(vpmax_f32(transmute(a), transmute(b)));
        assert_eq!(r, e);
    }
}

#[cfg(test)]
#[cfg(target_endian = "little")]
#[path = "table_lookup_tests.rs"]
mod table_lookup_tests;