use crate::{sse2::*, BVec4A, Vec2, Vec3, Vec3A};
#[cfg(not(target_arch = "spirv"))]
use core::fmt;
use core::iter::{Product, Sum};
use core::{f32, ops::*};
#[cfg(target_arch = "x86")]
use core::arch::x86::*;
#[cfg(target_arch = "x86_64")]
use core::arch::x86_64::*;
#[cfg(feature = "libm")]
#[allow(unused_imports)]
use num_traits::Float;
union UnionCast {
a: [f32; 4],
v: Vec4,
}
#[inline(always)]
pub const fn vec4(x: f32, y: f32, z: f32, w: f32) -> Vec4 {
Vec4::new(x, y, z, w)
}
#[derive(Clone, Copy)]
#[repr(transparent)]
pub struct Vec4(pub(crate) __m128);
impl Vec4 {
pub const ZERO: Self = Self::splat(0.0);
pub const ONE: Self = Self::splat(1.0);
pub const NEG_ONE: Self = Self::splat(-1.0);
pub const NAN: Self = Self::splat(f32::NAN);
pub const X: Self = Self::new(1.0, 0.0, 0.0, 0.0);
pub const Y: Self = Self::new(0.0, 1.0, 0.0, 0.0);
pub const Z: Self = Self::new(0.0, 0.0, 1.0, 0.0);
pub const W: Self = Self::new(0.0, 0.0, 0.0, 1.0);
pub const NEG_X: Self = Self::new(-1.0, 0.0, 0.0, 0.0);
pub const NEG_Y: Self = Self::new(0.0, -1.0, 0.0, 0.0);
pub const NEG_Z: Self = Self::new(0.0, 0.0, -1.0, 0.0);
pub const NEG_W: Self = Self::new(0.0, 0.0, 0.0, -1.0);
pub const AXES: [Self; 4] = [Self::X, Self::Y, Self::Z, Self::W];
#[inline(always)]
pub const fn new(x: f32, y: f32, z: f32, w: f32) -> Self {
unsafe { UnionCast { a: [x, y, z, w] }.v }
}
#[inline]
pub const fn splat(v: f32) -> Self {
unsafe { UnionCast { a: [v; 4] }.v }
}
#[inline]
pub fn select(mask: BVec4A, if_true: Self, if_false: Self) -> Self {
Self(unsafe {
_mm_or_ps(
_mm_andnot_ps(mask.0, if_false.0),
_mm_and_ps(if_true.0, mask.0),
)
})
}
#[inline]
pub const fn from_array(a: [f32; 4]) -> Self {
Self::new(a[0], a[1], a[2], a[3])
}
#[inline]
pub const fn to_array(&self) -> [f32; 4] {
unsafe { *(self as *const Vec4 as *const [f32; 4]) }
}
#[inline]
pub const fn from_slice(slice: &[f32]) -> Self {
Self::new(slice[0], slice[1], slice[2], slice[3])
}
#[inline]
pub fn write_to_slice(self, slice: &mut [f32]) {
unsafe {
assert!(slice.len() >= 4);
_mm_storeu_ps(slice.as_mut_ptr(), self.0);
}
}
#[inline]
pub fn truncate(self) -> Vec3 {
use crate::swizzles::Vec4Swizzles;
self.xyz()
}
#[inline]
pub fn dot(self, rhs: Self) -> f32 {
unsafe { dot4(self.0, rhs.0) }
}
#[inline]
pub fn dot_into_vec(self, rhs: Self) -> Self {
Self(unsafe { dot4_into_m128(self.0, rhs.0) })
}
#[inline]
pub fn min(self, rhs: Self) -> Self {
Self(unsafe { _mm_min_ps(self.0, rhs.0) })
}
#[inline]
pub fn max(self, rhs: Self) -> Self {
Self(unsafe { _mm_max_ps(self.0, rhs.0) })
}
#[inline]
pub fn clamp(self, min: Self, max: Self) -> Self {
glam_assert!(min.cmple(max).all(), "clamp: expected min <= max");
self.max(min).min(max)
}
#[inline]
pub fn min_element(self) -> f32 {
unsafe {
let v = self.0;
let v = _mm_min_ps(v, _mm_shuffle_ps(v, v, 0b00_00_11_10));
let v = _mm_min_ps(v, _mm_shuffle_ps(v, v, 0b00_00_00_01));
_mm_cvtss_f32(v)
}
}
#[inline]
pub fn max_element(self) -> f32 {
unsafe {
let v = self.0;
let v = _mm_max_ps(v, _mm_shuffle_ps(v, v, 0b00_00_11_10));
let v = _mm_max_ps(v, _mm_shuffle_ps(v, v, 0b00_00_00_01));
_mm_cvtss_f32(v)
}
}
#[inline]
pub fn cmpeq(self, rhs: Self) -> BVec4A {
BVec4A(unsafe { _mm_cmpeq_ps(self.0, rhs.0) })
}
#[inline]
pub fn cmpne(self, rhs: Self) -> BVec4A {
BVec4A(unsafe { _mm_cmpneq_ps(self.0, rhs.0) })
}
#[inline]
pub fn cmpge(self, rhs: Self) -> BVec4A {
BVec4A(unsafe { _mm_cmpge_ps(self.0, rhs.0) })
}
#[inline]
pub fn cmpgt(self, rhs: Self) -> BVec4A {
BVec4A(unsafe { _mm_cmpgt_ps(self.0, rhs.0) })
}
#[inline]
pub fn cmple(self, rhs: Self) -> BVec4A {
BVec4A(unsafe { _mm_cmple_ps(self.0, rhs.0) })
}
#[inline]
pub fn cmplt(self, rhs: Self) -> BVec4A {
BVec4A(unsafe { _mm_cmplt_ps(self.0, rhs.0) })
}
#[inline]
pub fn abs(self) -> Self {
Self(unsafe { crate::sse2::m128_abs(self.0) })
}
#[inline]
pub fn signum(self) -> Self {
unsafe {
let result = Self(_mm_or_ps(_mm_and_ps(self.0, Self::NEG_ONE.0), Self::ONE.0));
let mask = self.is_nan_mask();
Self::select(mask, self, result)
}
}
#[inline]
pub fn is_negative_bitmask(self) -> u32 {
unsafe { _mm_movemask_ps(self.0) as u32 }
}
#[inline]
pub fn is_finite(self) -> bool {
self.x.is_finite() && self.y.is_finite() && self.z.is_finite() && self.w.is_finite()
}
#[inline]
pub fn is_nan(self) -> bool {
self.is_nan_mask().any()
}
#[inline]
pub fn is_nan_mask(self) -> BVec4A {
BVec4A(unsafe { _mm_cmpunord_ps(self.0, self.0) })
}
#[doc(alias = "magnitude")]
#[inline]
pub fn length(self) -> f32 {
unsafe {
let dot = dot4_in_x(self.0, self.0);
_mm_cvtss_f32(_mm_sqrt_ps(dot))
}
}
#[doc(alias = "magnitude2")]
#[inline]
pub fn length_squared(self) -> f32 {
self.dot(self)
}
#[inline]
pub fn length_recip(self) -> f32 {
unsafe {
let dot = dot4_in_x(self.0, self.0);
_mm_cvtss_f32(_mm_div_ps(Self::ONE.0, _mm_sqrt_ps(dot)))
}
}
#[inline]
pub fn distance(self, rhs: Self) -> f32 {
(self - rhs).length()
}
#[inline]
pub fn distance_squared(self, rhs: Self) -> f32 {
(self - rhs).length_squared()
}
#[must_use]
#[inline]
pub fn normalize(self) -> Self {
unsafe {
let length = _mm_sqrt_ps(dot4_into_m128(self.0, self.0));
#[allow(clippy::let_and_return)]
let normalized = Self(_mm_div_ps(self.0, length));
glam_assert!(normalized.is_finite());
normalized
}
}
#[must_use]
#[inline]
pub fn try_normalize(self) -> Option<Self> {
let rcp = self.length_recip();
if rcp.is_finite() && rcp > 0.0 {
Some(self * rcp)
} else {
None
}
}
#[must_use]
#[inline]
pub fn normalize_or_zero(self) -> Self {
let rcp = self.length_recip();
if rcp.is_finite() && rcp > 0.0 {
self * rcp
} else {
Self::ZERO
}
}
#[inline]
pub fn is_normalized(self) -> bool {
(self.length_squared() - 1.0).abs() <= 1e-4
}
#[must_use]
#[inline]
pub fn project_onto(self, rhs: Self) -> Self {
let other_len_sq_rcp = rhs.dot(rhs).recip();
glam_assert!(other_len_sq_rcp.is_finite());
rhs * self.dot(rhs) * other_len_sq_rcp
}
#[must_use]
#[inline]
pub fn reject_from(self, rhs: Self) -> Self {
self - self.project_onto(rhs)
}
#[must_use]
#[inline]
pub fn project_onto_normalized(self, rhs: Self) -> Self {
glam_assert!(rhs.is_normalized());
rhs * self.dot(rhs)
}
#[must_use]
#[inline]
pub fn reject_from_normalized(self, rhs: Self) -> Self {
self - self.project_onto_normalized(rhs)
}
#[inline]
pub fn round(self) -> Self {
Self(unsafe { m128_round(self.0) })
}
#[inline]
pub fn floor(self) -> Self {
Self(unsafe { m128_floor(self.0) })
}
#[inline]
pub fn ceil(self) -> Self {
Self(unsafe { m128_ceil(self.0) })
}
#[inline]
pub fn fract(self) -> Self {
self - self.floor()
}
#[inline]
pub fn exp(self) -> Self {
Self::new(self.x.exp(), self.y.exp(), self.z.exp(), self.w.exp())
}
#[inline]
pub fn powf(self, n: f32) -> Self {
Self::new(
self.x.powf(n),
self.y.powf(n),
self.z.powf(n),
self.w.powf(n),
)
}
#[inline]
pub fn recip(self) -> Self {
Self(unsafe { _mm_div_ps(Self::ONE.0, self.0) })
}
#[doc(alias = "mix")]
#[inline]
pub fn lerp(self, rhs: Self, s: f32) -> Self {
self + ((rhs - self) * s)
}
#[inline]
pub fn abs_diff_eq(self, rhs: Self, max_abs_diff: f32) -> bool {
self.sub(rhs).abs().cmple(Self::splat(max_abs_diff)).all()
}
#[inline]
pub fn clamp_length(self, min: f32, max: f32) -> Self {
glam_assert!(min <= max);
let length_sq = self.length_squared();
if length_sq < min * min {
self * (length_sq.sqrt().recip() * min)
} else if length_sq > max * max {
self * (length_sq.sqrt().recip() * max)
} else {
self
}
}
pub fn clamp_length_max(self, max: f32) -> Self {
let length_sq = self.length_squared();
if length_sq > max * max {
self * (length_sq.sqrt().recip() * max)
} else {
self
}
}
pub fn clamp_length_min(self, min: f32) -> Self {
let length_sq = self.length_squared();
if length_sq < min * min {
self * (length_sq.sqrt().recip() * min)
} else {
self
}
}
#[inline]
pub fn mul_add(self, a: Self, b: Self) -> Self {
#[cfg(target_feature = "fma")]
unsafe {
Self(_mm_fmadd_ps(self.0, a.0, b.0))
}
#[cfg(not(target_feature = "fma"))]
Self::new(
self.x.mul_add(a.x, b.x),
self.y.mul_add(a.y, b.y),
self.z.mul_add(a.z, b.z),
self.w.mul_add(a.w, b.w),
)
}
#[inline]
pub fn as_dvec4(&self) -> crate::DVec4 {
crate::DVec4::new(self.x as f64, self.y as f64, self.z as f64, self.w as f64)
}
#[inline]
pub fn as_ivec4(&self) -> crate::IVec4 {
crate::IVec4::new(self.x as i32, self.y as i32, self.z as i32, self.w as i32)
}
#[inline]
pub fn as_uvec4(&self) -> crate::UVec4 {
crate::UVec4::new(self.x as u32, self.y as u32, self.z as u32, self.w as u32)
}
}
impl Default for Vec4 {
#[inline(always)]
fn default() -> Self {
Self::ZERO
}
}
impl PartialEq for Vec4 {
#[inline]
fn eq(&self, rhs: &Self) -> bool {
self.cmpeq(*rhs).all()
}
}
impl Div<Vec4> for Vec4 {
type Output = Self;
#[inline]
fn div(self, rhs: Self) -> Self {
Self(unsafe { _mm_div_ps(self.0, rhs.0) })
}
}
impl DivAssign<Vec4> for Vec4 {
#[inline]
fn div_assign(&mut self, rhs: Self) {
self.0 = unsafe { _mm_div_ps(self.0, rhs.0) };
}
}
impl Div<f32> for Vec4 {
type Output = Self;
#[inline]
fn div(self, rhs: f32) -> Self {
Self(unsafe { _mm_div_ps(self.0, _mm_set1_ps(rhs)) })
}
}
impl DivAssign<f32> for Vec4 {
#[inline]
fn div_assign(&mut self, rhs: f32) {
self.0 = unsafe { _mm_div_ps(self.0, _mm_set1_ps(rhs)) };
}
}
impl Div<Vec4> for f32 {
type Output = Vec4;
#[inline]
fn div(self, rhs: Vec4) -> Vec4 {
Vec4(unsafe { _mm_div_ps(_mm_set1_ps(self), rhs.0) })
}
}
impl Mul<Vec4> for Vec4 {
type Output = Self;
#[inline]
fn mul(self, rhs: Self) -> Self {
Self(unsafe { _mm_mul_ps(self.0, rhs.0) })
}
}
impl MulAssign<Vec4> for Vec4 {
#[inline]
fn mul_assign(&mut self, rhs: Self) {
self.0 = unsafe { _mm_mul_ps(self.0, rhs.0) };
}
}
impl Mul<f32> for Vec4 {
type Output = Self;
#[inline]
fn mul(self, rhs: f32) -> Self {
Self(unsafe { _mm_mul_ps(self.0, _mm_set1_ps(rhs)) })
}
}
impl MulAssign<f32> for Vec4 {
#[inline]
fn mul_assign(&mut self, rhs: f32) {
self.0 = unsafe { _mm_mul_ps(self.0, _mm_set1_ps(rhs)) };
}
}
impl Mul<Vec4> for f32 {
type Output = Vec4;
#[inline]
fn mul(self, rhs: Vec4) -> Vec4 {
Vec4(unsafe { _mm_mul_ps(_mm_set1_ps(self), rhs.0) })
}
}
impl Add<Vec4> for Vec4 {
type Output = Self;
#[inline]
fn add(self, rhs: Self) -> Self {
Self(unsafe { _mm_add_ps(self.0, rhs.0) })
}
}
impl AddAssign<Vec4> for Vec4 {
#[inline]
fn add_assign(&mut self, rhs: Self) {
self.0 = unsafe { _mm_add_ps(self.0, rhs.0) };
}
}
impl Add<f32> for Vec4 {
type Output = Self;
#[inline]
fn add(self, rhs: f32) -> Self {
Self(unsafe { _mm_add_ps(self.0, _mm_set1_ps(rhs)) })
}
}
impl AddAssign<f32> for Vec4 {
#[inline]
fn add_assign(&mut self, rhs: f32) {
self.0 = unsafe { _mm_add_ps(self.0, _mm_set1_ps(rhs)) };
}
}
impl Add<Vec4> for f32 {
type Output = Vec4;
#[inline]
fn add(self, rhs: Vec4) -> Vec4 {
Vec4(unsafe { _mm_add_ps(_mm_set1_ps(self), rhs.0) })
}
}
impl Sub<Vec4> for Vec4 {
type Output = Self;
#[inline]
fn sub(self, rhs: Self) -> Self {
Self(unsafe { _mm_sub_ps(self.0, rhs.0) })
}
}
impl SubAssign<Vec4> for Vec4 {
#[inline]
fn sub_assign(&mut self, rhs: Vec4) {
self.0 = unsafe { _mm_sub_ps(self.0, rhs.0) };
}
}
impl Sub<f32> for Vec4 {
type Output = Self;
#[inline]
fn sub(self, rhs: f32) -> Self {
Self(unsafe { _mm_sub_ps(self.0, _mm_set1_ps(rhs)) })
}
}
impl SubAssign<f32> for Vec4 {
#[inline]
fn sub_assign(&mut self, rhs: f32) {
self.0 = unsafe { _mm_sub_ps(self.0, _mm_set1_ps(rhs)) };
}
}
impl Sub<Vec4> for f32 {
type Output = Vec4;
#[inline]
fn sub(self, rhs: Vec4) -> Vec4 {
Vec4(unsafe { _mm_sub_ps(_mm_set1_ps(self), rhs.0) })
}
}
impl Rem<Vec4> for Vec4 {
type Output = Self;
#[inline]
fn rem(self, rhs: Self) -> Self {
unsafe {
let n = m128_floor(_mm_div_ps(self.0, rhs.0));
Self(_mm_sub_ps(self.0, _mm_mul_ps(n, rhs.0)))
}
}
}
impl RemAssign<Vec4> for Vec4 {
#[inline]
fn rem_assign(&mut self, rhs: Self) {
*self = self.rem(rhs);
}
}
impl Rem<f32> for Vec4 {
type Output = Self;
#[inline]
fn rem(self, rhs: f32) -> Self {
self.rem(Self::splat(rhs))
}
}
impl RemAssign<f32> for Vec4 {
#[inline]
fn rem_assign(&mut self, rhs: f32) {
*self = self.rem(Self::splat(rhs));
}
}
impl Rem<Vec4> for f32 {
type Output = Vec4;
#[inline]
fn rem(self, rhs: Vec4) -> Vec4 {
Vec4::splat(self).rem(rhs)
}
}
#[cfg(not(target_arch = "spirv"))]
impl AsRef<[f32; 4]> for Vec4 {
#[inline]
fn as_ref(&self) -> &[f32; 4] {
unsafe { &*(self as *const Vec4 as *const [f32; 4]) }
}
}
#[cfg(not(target_arch = "spirv"))]
impl AsMut<[f32; 4]> for Vec4 {
#[inline]
fn as_mut(&mut self) -> &mut [f32; 4] {
unsafe { &mut *(self as *mut Vec4 as *mut [f32; 4]) }
}
}
impl Sum for Vec4 {
#[inline]
fn sum<I>(iter: I) -> Self
where
I: Iterator<Item = Self>,
{
iter.fold(Self::ZERO, Self::add)
}
}
impl<'a> Sum<&'a Self> for Vec4 {
#[inline]
fn sum<I>(iter: I) -> Self
where
I: Iterator<Item = &'a Self>,
{
iter.fold(Self::ZERO, |a, &b| Self::add(a, b))
}
}
impl Product for Vec4 {
#[inline]
fn product<I>(iter: I) -> Self
where
I: Iterator<Item = Self>,
{
iter.fold(Self::ONE, Self::mul)
}
}
impl<'a> Product<&'a Self> for Vec4 {
#[inline]
fn product<I>(iter: I) -> Self
where
I: Iterator<Item = &'a Self>,
{
iter.fold(Self::ONE, |a, &b| Self::mul(a, b))
}
}
impl Neg for Vec4 {
type Output = Self;
#[inline]
fn neg(self) -> Self {
Self(unsafe { _mm_xor_ps(_mm_set1_ps(-0.0), self.0) })
}
}
impl Index<usize> for Vec4 {
type Output = f32;
#[inline]
fn index(&self, index: usize) -> &Self::Output {
match index {
0 => &self.x,
1 => &self.y,
2 => &self.z,
3 => &self.w,
_ => panic!("index out of bounds"),
}
}
}
impl IndexMut<usize> for Vec4 {
#[inline]
fn index_mut(&mut self, index: usize) -> &mut Self::Output {
match index {
0 => &mut self.x,
1 => &mut self.y,
2 => &mut self.z,
3 => &mut self.w,
_ => panic!("index out of bounds"),
}
}
}
#[cfg(not(target_arch = "spirv"))]
impl fmt::Display for Vec4 {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
write!(f, "[{}, {}, {}, {}]", self.x, self.y, self.z, self.w)
}
}
#[cfg(not(target_arch = "spirv"))]
impl fmt::Debug for Vec4 {
fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result {
fmt.debug_tuple(stringify!(Vec4))
.field(&self.x)
.field(&self.y)
.field(&self.z)
.field(&self.w)
.finish()
}
}
impl From<Vec4> for __m128 {
#[inline]
fn from(t: Vec4) -> Self {
t.0
}
}
impl From<__m128> for Vec4 {
#[inline]
fn from(t: __m128) -> Self {
Self(t)
}
}
impl From<[f32; 4]> for Vec4 {
#[inline]
fn from(a: [f32; 4]) -> Self {
Self(unsafe { _mm_loadu_ps(a.as_ptr()) })
}
}
impl From<Vec4> for [f32; 4] {
#[inline]
fn from(v: Vec4) -> Self {
use crate::Align16;
use core::mem::MaybeUninit;
let mut out: MaybeUninit<Align16<Self>> = MaybeUninit::uninit();
unsafe {
_mm_store_ps(out.as_mut_ptr().cast(), v.0);
out.assume_init().0
}
}
}
impl From<(f32, f32, f32, f32)> for Vec4 {
#[inline]
fn from(t: (f32, f32, f32, f32)) -> Self {
Self::new(t.0, t.1, t.2, t.3)
}
}
impl From<Vec4> for (f32, f32, f32, f32) {
#[inline]
fn from(v: Vec4) -> Self {
use crate::Align16;
use core::mem::MaybeUninit;
let mut out: MaybeUninit<Align16<Self>> = MaybeUninit::uninit();
unsafe {
_mm_store_ps(out.as_mut_ptr().cast(), v.0);
out.assume_init().0
}
}
}
impl From<(Vec3A, f32)> for Vec4 {
#[inline]
fn from((v, w): (Vec3A, f32)) -> Self {
v.extend(w)
}
}
impl From<(f32, Vec3A)> for Vec4 {
#[inline]
fn from((x, v): (f32, Vec3A)) -> Self {
Self::new(x, v.x, v.y, v.z)
}
}
impl From<(Vec3, f32)> for Vec4 {
#[inline]
fn from((v, w): (Vec3, f32)) -> Self {
Self::new(v.x, v.y, v.z, w)
}
}
impl From<(f32, Vec3)> for Vec4 {
#[inline]
fn from((x, v): (f32, Vec3)) -> Self {
Self::new(x, v.x, v.y, v.z)
}
}
impl From<(Vec2, f32, f32)> for Vec4 {
#[inline]
fn from((v, z, w): (Vec2, f32, f32)) -> Self {
Self::new(v.x, v.y, z, w)
}
}
impl From<(Vec2, Vec2)> for Vec4 {
#[inline]
fn from((v, u): (Vec2, Vec2)) -> Self {
Self::new(v.x, v.y, u.x, u.y)
}
}
impl Deref for Vec4 {
type Target = crate::deref::Vec4<f32>;
#[inline]
fn deref(&self) -> &Self::Target {
unsafe { &*(self as *const Self).cast() }
}
}
impl DerefMut for Vec4 {
#[inline]
fn deref_mut(&mut self) -> &mut Self::Target {
unsafe { &mut *(self as *mut Self).cast() }
}
}