473 lines
13 KiB
Rust
473 lines
13 KiB
Rust
use super::*;
|
|
|
|
pick! {
|
|
if #[cfg(target_feature="sse2")] {
|
|
#[derive(Default, Clone, Copy, PartialEq, Eq)]
|
|
#[repr(C, align(16))]
|
|
pub struct i16x8 { sse: m128i }
|
|
} else if #[cfg(target_feature="simd128")] {
|
|
use core::arch::wasm32::*;
|
|
|
|
#[derive(Clone, Copy)]
|
|
#[repr(transparent)]
|
|
pub struct i16x8 { simd: v128 }
|
|
|
|
impl Default for i16x8 {
|
|
fn default() -> Self {
|
|
Self::splat(0)
|
|
}
|
|
}
|
|
|
|
impl PartialEq for i16x8 {
|
|
fn eq(&self, other: &Self) -> bool {
|
|
u16x8_all_true(i16x8_eq(self.simd, other.simd))
|
|
}
|
|
}
|
|
|
|
impl Eq for i16x8 { }
|
|
} else {
|
|
#[derive(Default, Clone, Copy, PartialEq, Eq)]
|
|
#[repr(C, align(16))]
|
|
pub struct i16x8 { arr: [i16;8] }
|
|
}
|
|
}
|
|
|
|
int_uint_consts!(i16, 8, i16x8, i16x8, i16a8, const_i16_as_i16x8, 128);
|
|
|
|
unsafe impl Zeroable for i16x8 {}
|
|
unsafe impl Pod for i16x8 {}
|
|
|
|
impl Add for i16x8 {
|
|
type Output = Self;
|
|
#[inline]
|
|
#[must_use]
|
|
fn add(self, rhs: Self) -> Self::Output {
|
|
pick! {
|
|
if #[cfg(target_feature="sse2")] {
|
|
Self { sse: add_i16_m128i(self.sse, rhs.sse) }
|
|
} else if #[cfg(target_feature="simd128")] {
|
|
Self { simd: i16x8_add(self.simd, rhs.simd) }
|
|
} else {
|
|
Self { arr: [
|
|
self.arr[0].wrapping_add(rhs.arr[0]),
|
|
self.arr[1].wrapping_add(rhs.arr[1]),
|
|
self.arr[2].wrapping_add(rhs.arr[2]),
|
|
self.arr[3].wrapping_add(rhs.arr[3]),
|
|
self.arr[4].wrapping_add(rhs.arr[4]),
|
|
self.arr[5].wrapping_add(rhs.arr[5]),
|
|
self.arr[6].wrapping_add(rhs.arr[6]),
|
|
self.arr[7].wrapping_add(rhs.arr[7]),
|
|
]}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
impl Sub for i16x8 {
|
|
type Output = Self;
|
|
#[inline]
|
|
#[must_use]
|
|
fn sub(self, rhs: Self) -> Self::Output {
|
|
pick! {
|
|
if #[cfg(target_feature="sse2")] {
|
|
Self { sse: sub_i16_m128i(self.sse, rhs.sse) }
|
|
} else if #[cfg(target_feature="simd128")] {
|
|
Self { simd: i16x8_sub(self.simd, rhs.simd) }
|
|
} else {
|
|
Self { arr: [
|
|
self.arr[0].wrapping_sub(rhs.arr[0]),
|
|
self.arr[1].wrapping_sub(rhs.arr[1]),
|
|
self.arr[2].wrapping_sub(rhs.arr[2]),
|
|
self.arr[3].wrapping_sub(rhs.arr[3]),
|
|
self.arr[4].wrapping_sub(rhs.arr[4]),
|
|
self.arr[5].wrapping_sub(rhs.arr[5]),
|
|
self.arr[6].wrapping_sub(rhs.arr[6]),
|
|
self.arr[7].wrapping_sub(rhs.arr[7]),
|
|
]}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
impl Mul for i16x8 {
|
|
type Output = Self;
|
|
#[inline]
|
|
#[must_use]
|
|
fn mul(self, rhs: Self) -> Self::Output {
|
|
pick! {
|
|
if #[cfg(target_feature="sse2")] {
|
|
Self { sse: mul_i16_keep_low_m128i(self.sse, rhs.sse) }
|
|
} else if #[cfg(target_feature="simd128")] {
|
|
Self { simd: i16x8_mul(self.simd, rhs.simd) }
|
|
} else {
|
|
Self { arr: [
|
|
self.arr[0].wrapping_mul(rhs.arr[0]),
|
|
self.arr[1].wrapping_mul(rhs.arr[1]),
|
|
self.arr[2].wrapping_mul(rhs.arr[2]),
|
|
self.arr[3].wrapping_mul(rhs.arr[3]),
|
|
self.arr[4].wrapping_mul(rhs.arr[4]),
|
|
self.arr[5].wrapping_mul(rhs.arr[5]),
|
|
self.arr[6].wrapping_mul(rhs.arr[6]),
|
|
self.arr[7].wrapping_mul(rhs.arr[7]),
|
|
]}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
impl Add<i16> for i16x8 {
|
|
type Output = Self;
|
|
#[inline]
|
|
#[must_use]
|
|
fn add(self, rhs: i16) -> Self::Output {
|
|
self.add(Self::splat(rhs))
|
|
}
|
|
}
|
|
|
|
impl Sub<i16> for i16x8 {
|
|
type Output = Self;
|
|
#[inline]
|
|
#[must_use]
|
|
fn sub(self, rhs: i16) -> Self::Output {
|
|
self.sub(Self::splat(rhs))
|
|
}
|
|
}
|
|
|
|
impl Mul<i16> for i16x8 {
|
|
type Output = Self;
|
|
#[inline]
|
|
#[must_use]
|
|
fn mul(self, rhs: i16) -> Self::Output {
|
|
self.mul(Self::splat(rhs))
|
|
}
|
|
}
|
|
|
|
impl Add<i16x8> for i16 {
|
|
type Output = i16x8;
|
|
#[inline]
|
|
#[must_use]
|
|
fn add(self, rhs: i16x8) -> Self::Output {
|
|
i16x8::splat(self).add(rhs)
|
|
}
|
|
}
|
|
|
|
impl Sub<i16x8> for i16 {
|
|
type Output = i16x8;
|
|
#[inline]
|
|
#[must_use]
|
|
fn sub(self, rhs: i16x8) -> Self::Output {
|
|
i16x8::splat(self).sub(rhs)
|
|
}
|
|
}
|
|
|
|
impl Mul<i16x8> for i16 {
|
|
type Output = i16x8;
|
|
#[inline]
|
|
#[must_use]
|
|
fn mul(self, rhs: i16x8) -> Self::Output {
|
|
i16x8::splat(self).mul(rhs)
|
|
}
|
|
}
|
|
|
|
impl BitAnd for i16x8 {
|
|
type Output = Self;
|
|
#[inline]
|
|
#[must_use]
|
|
fn bitand(self, rhs: Self) -> Self::Output {
|
|
pick! {
|
|
if #[cfg(target_feature="sse2")] {
|
|
Self { sse: bitand_m128i(self.sse, rhs.sse) }
|
|
} else if #[cfg(target_feature="simd128")] {
|
|
Self { simd: v128_and(self.simd, rhs.simd) }
|
|
} else {
|
|
Self { arr: [
|
|
self.arr[0].bitand(rhs.arr[0]),
|
|
self.arr[1].bitand(rhs.arr[1]),
|
|
self.arr[2].bitand(rhs.arr[2]),
|
|
self.arr[3].bitand(rhs.arr[3]),
|
|
self.arr[4].bitand(rhs.arr[4]),
|
|
self.arr[5].bitand(rhs.arr[5]),
|
|
self.arr[6].bitand(rhs.arr[6]),
|
|
self.arr[7].bitand(rhs.arr[7]),
|
|
]}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
impl BitOr for i16x8 {
|
|
type Output = Self;
|
|
#[inline]
|
|
#[must_use]
|
|
fn bitor(self, rhs: Self) -> Self::Output {
|
|
pick! {
|
|
if #[cfg(target_feature="sse2")] {
|
|
Self { sse: bitor_m128i(self.sse, rhs.sse) }
|
|
} else if #[cfg(target_feature="simd128")] {
|
|
Self { simd: v128_or(self.simd, rhs.simd) }
|
|
} else {
|
|
Self { arr: [
|
|
self.arr[0].bitor(rhs.arr[0]),
|
|
self.arr[1].bitor(rhs.arr[1]),
|
|
self.arr[2].bitor(rhs.arr[2]),
|
|
self.arr[3].bitor(rhs.arr[3]),
|
|
self.arr[4].bitor(rhs.arr[4]),
|
|
self.arr[5].bitor(rhs.arr[5]),
|
|
self.arr[6].bitor(rhs.arr[6]),
|
|
self.arr[7].bitor(rhs.arr[7]),
|
|
]}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
impl BitXor for i16x8 {
|
|
type Output = Self;
|
|
#[inline]
|
|
#[must_use]
|
|
fn bitxor(self, rhs: Self) -> Self::Output {
|
|
pick! {
|
|
if #[cfg(target_feature="sse2")] {
|
|
Self { sse: bitxor_m128i(self.sse, rhs.sse) }
|
|
} else if #[cfg(target_feature="simd128")] {
|
|
Self { simd: v128_xor(self.simd, rhs.simd) }
|
|
} else {
|
|
Self { arr: [
|
|
self.arr[0].bitxor(rhs.arr[0]),
|
|
self.arr[1].bitxor(rhs.arr[1]),
|
|
self.arr[2].bitxor(rhs.arr[2]),
|
|
self.arr[3].bitxor(rhs.arr[3]),
|
|
self.arr[4].bitxor(rhs.arr[4]),
|
|
self.arr[5].bitxor(rhs.arr[5]),
|
|
self.arr[6].bitxor(rhs.arr[6]),
|
|
self.arr[7].bitxor(rhs.arr[7]),
|
|
]}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
macro_rules! impl_shl_t_for_i16x8 {
|
|
($($shift_type:ty),+ $(,)?) => {
|
|
$(impl Shl<$shift_type> for i16x8 {
|
|
type Output = Self;
|
|
/// Shifts all lanes by the value given.
|
|
#[inline]
|
|
#[must_use]
|
|
fn shl(self, rhs: $shift_type) -> Self::Output {
|
|
pick! {
|
|
if #[cfg(target_feature="sse2")] {
|
|
let shift = cast([rhs as u64, 0]);
|
|
Self { sse: shl_all_u16_m128i(self.sse, shift) }
|
|
} else if #[cfg(target_feature="simd128")] {
|
|
Self { simd: i16x8_shl(self.simd, rhs as u32) }
|
|
} else {
|
|
let u = rhs as u64;
|
|
Self { arr: [
|
|
self.arr[0] << u,
|
|
self.arr[1] << u,
|
|
self.arr[2] << u,
|
|
self.arr[3] << u,
|
|
self.arr[4] << u,
|
|
self.arr[5] << u,
|
|
self.arr[6] << u,
|
|
self.arr[7] << u,
|
|
]}
|
|
}
|
|
}
|
|
}
|
|
})+
|
|
};
|
|
}
|
|
impl_shl_t_for_i16x8!(i8, u8, i16, u16, i32, u32, i64, u64, i128, u128);
|
|
|
|
macro_rules! impl_shr_t_for_i16x8 {
|
|
($($shift_type:ty),+ $(,)?) => {
|
|
$(impl Shr<$shift_type> for i16x8 {
|
|
type Output = Self;
|
|
/// Shifts all lanes by the value given.
|
|
#[inline]
|
|
#[must_use]
|
|
fn shr(self, rhs: $shift_type) -> Self::Output {
|
|
pick! {
|
|
if #[cfg(target_feature="sse2")] {
|
|
let shift = cast([rhs as u64, 0]);
|
|
Self { sse: shr_all_i16_m128i(self.sse, shift) }
|
|
} else if #[cfg(target_feature="simd128")] {
|
|
Self { simd: i16x8_shr(self.simd, rhs as u32) }
|
|
} else {
|
|
let u = rhs as u64;
|
|
Self { arr: [
|
|
self.arr[0] >> u,
|
|
self.arr[1] >> u,
|
|
self.arr[2] >> u,
|
|
self.arr[3] >> u,
|
|
self.arr[4] >> u,
|
|
self.arr[5] >> u,
|
|
self.arr[6] >> u,
|
|
self.arr[7] >> u,
|
|
]}
|
|
}
|
|
}
|
|
}
|
|
})+
|
|
};
|
|
}
|
|
impl_shr_t_for_i16x8!(i8, u8, i16, u16, i32, u32, i64, u64, i128, u128);
|
|
|
|
impl CmpEq for i16x8 {
|
|
type Output = Self;
|
|
#[inline]
|
|
#[must_use]
|
|
fn cmp_eq(self, rhs: Self) -> Self::Output {
|
|
pick! {
|
|
if #[cfg(target_feature="sse2")] {
|
|
Self { sse: cmp_eq_mask_i16_m128i(self.sse, rhs.sse) }
|
|
} else if #[cfg(target_feature="simd128")] {
|
|
Self { simd: i16x8_eq(self.simd, rhs.simd) }
|
|
} else {
|
|
Self { arr: [
|
|
if self.arr[0] == rhs.arr[0] { -1 } else { 0 },
|
|
if self.arr[1] == rhs.arr[1] { -1 } else { 0 },
|
|
if self.arr[2] == rhs.arr[2] { -1 } else { 0 },
|
|
if self.arr[3] == rhs.arr[3] { -1 } else { 0 },
|
|
if self.arr[4] == rhs.arr[4] { -1 } else { 0 },
|
|
if self.arr[5] == rhs.arr[5] { -1 } else { 0 },
|
|
if self.arr[6] == rhs.arr[6] { -1 } else { 0 },
|
|
if self.arr[7] == rhs.arr[7] { -1 } else { 0 },
|
|
]}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
impl CmpGt for i16x8 {
|
|
type Output = Self;
|
|
#[inline]
|
|
#[must_use]
|
|
fn cmp_gt(self, rhs: Self) -> Self::Output {
|
|
pick! {
|
|
if #[cfg(target_feature="sse2")] {
|
|
Self { sse: cmp_gt_mask_i16_m128i(self.sse, rhs.sse) }
|
|
} else if #[cfg(target_feature="simd128")] {
|
|
Self { simd: i16x8_gt(self.simd, rhs.simd) }
|
|
} else {
|
|
Self { arr: [
|
|
if self.arr[0] > rhs.arr[0] { -1 } else { 0 },
|
|
if self.arr[1] > rhs.arr[1] { -1 } else { 0 },
|
|
if self.arr[2] > rhs.arr[2] { -1 } else { 0 },
|
|
if self.arr[3] > rhs.arr[3] { -1 } else { 0 },
|
|
if self.arr[4] > rhs.arr[4] { -1 } else { 0 },
|
|
if self.arr[5] > rhs.arr[5] { -1 } else { 0 },
|
|
if self.arr[6] > rhs.arr[6] { -1 } else { 0 },
|
|
if self.arr[7] > rhs.arr[7] { -1 } else { 0 },
|
|
]}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
impl CmpLt for i16x8 {
|
|
type Output = Self;
|
|
#[inline]
|
|
#[must_use]
|
|
fn cmp_lt(self, rhs: Self) -> Self::Output {
|
|
pick! {
|
|
if #[cfg(target_feature="sse2")] {
|
|
Self { sse: cmp_lt_mask_i16_m128i(self.sse, rhs.sse) }
|
|
} else if #[cfg(target_feature="simd128")] {
|
|
Self { simd: i16x8_lt(self.simd, rhs.simd) }
|
|
} else {
|
|
Self { arr: [
|
|
if self.arr[0] < rhs.arr[0] { -1 } else { 0 },
|
|
if self.arr[1] < rhs.arr[1] { -1 } else { 0 },
|
|
if self.arr[2] < rhs.arr[2] { -1 } else { 0 },
|
|
if self.arr[3] < rhs.arr[3] { -1 } else { 0 },
|
|
if self.arr[4] < rhs.arr[4] { -1 } else { 0 },
|
|
if self.arr[5] < rhs.arr[5] { -1 } else { 0 },
|
|
if self.arr[6] < rhs.arr[6] { -1 } else { 0 },
|
|
if self.arr[7] < rhs.arr[7] { -1 } else { 0 },
|
|
]}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
impl i16x8 {
|
|
#[inline]
|
|
#[must_use]
|
|
pub fn new(array: [i16; 8]) -> Self {
|
|
Self::from(array)
|
|
}
|
|
#[inline]
|
|
#[must_use]
|
|
pub fn blend(self, t: Self, f: Self) -> Self {
|
|
pick! {
|
|
if #[cfg(target_feature="sse4.1")] {
|
|
Self { sse: blend_varying_i8_m128i(f.sse, t.sse, self.sse) }
|
|
} else if #[cfg(target_feature="simd128")] {
|
|
Self { simd: v128_bitselect(t.simd, f.simd, self.simd) }
|
|
} else {
|
|
generic_bit_blend(self, t, f)
|
|
}
|
|
}
|
|
}
|
|
#[inline]
|
|
#[must_use]
|
|
pub fn abs(self) -> Self {
|
|
pick! {
|
|
if #[cfg(target_feature="ssse3")] {
|
|
Self { sse: abs_i16_m128i(self.sse) }
|
|
} else if #[cfg(target_feature="simd128")] {
|
|
Self { simd: i16x8_abs(self.simd) }
|
|
} else {
|
|
let arr: [i16; 8] = cast(self);
|
|
cast([
|
|
arr[0].wrapping_abs(),
|
|
arr[1].wrapping_abs(),
|
|
arr[2].wrapping_abs(),
|
|
arr[3].wrapping_abs(),
|
|
arr[4].wrapping_abs(),
|
|
arr[5].wrapping_abs(),
|
|
arr[6].wrapping_abs(),
|
|
arr[7].wrapping_abs(),
|
|
])
|
|
}
|
|
}
|
|
}
|
|
#[inline]
|
|
#[must_use]
|
|
pub fn max(self, rhs: Self) -> Self {
|
|
pick! {
|
|
if #[cfg(target_feature="sse2")] {
|
|
Self { sse: max_i16_m128i(self.sse, rhs.sse) }
|
|
} else if #[cfg(target_feature="simd128")] {
|
|
Self { simd: i16x8_max(self.simd, rhs.simd) }
|
|
} else {
|
|
self.cmp_lt(rhs).blend(rhs, self)
|
|
}
|
|
}
|
|
}
|
|
#[inline]
|
|
#[must_use]
|
|
pub fn min(self, rhs: Self) -> Self {
|
|
pick! {
|
|
if #[cfg(target_feature="sse4.1")] {
|
|
Self { sse: min_i16_m128i(self.sse, rhs.sse) }
|
|
} else if #[cfg(target_feature="simd128")] {
|
|
Self { simd: i16x8_min(self.simd, rhs.simd) }
|
|
} else {
|
|
self.cmp_lt(rhs).blend(self, rhs)
|
|
}
|
|
}
|
|
}
|
|
|
|
pub fn to_array(self) -> [i16; 8] {
|
|
cast(self)
|
|
}
|
|
|
|
pub fn as_array_ref(&self) -> &[i16; 8] {
|
|
cast_ref(self)
|
|
}
|
|
}
|