aboutsummaryrefslogtreecommitdiff
path: root/src/f32/sse2/vec4.rs
diff options
context:
space:
mode:
Diffstat (limited to 'src/f32/sse2/vec4.rs')
-rw-r--r--src/f32/sse2/vec4.rs1078
1 files changed, 1078 insertions, 0 deletions
diff --git a/src/f32/sse2/vec4.rs b/src/f32/sse2/vec4.rs
new file mode 100644
index 0000000..0e0882c
--- /dev/null
+++ b/src/f32/sse2/vec4.rs
@@ -0,0 +1,1078 @@
+// Generated from vec.rs.tera template. Edit the template, not the generated file.
+
+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,
+}
+
+/// Creates a 4-dimensional vector.
+#[inline(always)]
+pub const fn vec4(x: f32, y: f32, z: f32, w: f32) -> Vec4 {
+ Vec4::new(x, y, z, w)
+}
+
+/// A 4-dimensional vector with SIMD support.
+///
+/// This type uses 16 byte aligned SIMD vector type for storage.
+#[derive(Clone, Copy)]
+#[repr(transparent)]
+pub struct Vec4(pub(crate) __m128);
+
+impl Vec4 {
+ /// All zeroes.
+ pub const ZERO: Self = Self::splat(0.0);
+
+ /// All ones.
+ pub const ONE: Self = Self::splat(1.0);
+
+ /// All negative ones.
+ pub const NEG_ONE: Self = Self::splat(-1.0);
+
+ /// All NAN.
+ pub const NAN: Self = Self::splat(f32::NAN);
+
+ /// A unit-length vector pointing along the positive X axis.
+ pub const X: Self = Self::new(1.0, 0.0, 0.0, 0.0);
+
+ /// A unit-length vector pointing along the positive Y axis.
+ pub const Y: Self = Self::new(0.0, 1.0, 0.0, 0.0);
+
+ /// A unit-length vector pointing along the positive Z axis.
+ pub const Z: Self = Self::new(0.0, 0.0, 1.0, 0.0);
+
+ /// A unit-length vector pointing along the positive W axis.
+ pub const W: Self = Self::new(0.0, 0.0, 0.0, 1.0);
+
+ /// A unit-length vector pointing along the negative X axis.
+ pub const NEG_X: Self = Self::new(-1.0, 0.0, 0.0, 0.0);
+
+ /// A unit-length vector pointing along the negative Y axis.
+ pub const NEG_Y: Self = Self::new(0.0, -1.0, 0.0, 0.0);
+
+ /// A unit-length vector pointing along the negative Z axis.
+ pub const NEG_Z: Self = Self::new(0.0, 0.0, -1.0, 0.0);
+
+ /// A unit-length vector pointing along the negative W axis.
+ pub const NEG_W: Self = Self::new(0.0, 0.0, 0.0, -1.0);
+
+ /// The unit axes.
+ pub const AXES: [Self; 4] = [Self::X, Self::Y, Self::Z, Self::W];
+
+ /// Creates a new vector.
+ #[inline(always)]
+ pub const fn new(x: f32, y: f32, z: f32, w: f32) -> Self {
+ unsafe { UnionCast { a: [x, y, z, w] }.v }
+ }
+
+ /// Creates a vector with all elements set to `v`.
+ #[inline]
+ pub const fn splat(v: f32) -> Self {
+ unsafe { UnionCast { a: [v; 4] }.v }
+ }
+
+ /// Creates a vector from the elements in `if_true` and `if_false`, selecting which to use
+ /// for each element of `self`.
+ ///
+ /// A true element in the mask uses the corresponding element from `if_true`, and false
+ /// uses the element from `if_false`.
+ #[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),
+ )
+ })
+ }
+
+ /// Creates a new vector from an array.
+ #[inline]
+ pub const fn from_array(a: [f32; 4]) -> Self {
+ Self::new(a[0], a[1], a[2], a[3])
+ }
+
+ /// `[x, y, z, w]`
+ #[inline]
+ pub const fn to_array(&self) -> [f32; 4] {
+ unsafe { *(self as *const Vec4 as *const [f32; 4]) }
+ }
+
+ /// Creates a vector from the first 4 values in `slice`.
+ ///
+ /// # Panics
+ ///
+ /// Panics if `slice` is less than 4 elements long.
+ #[inline]
+ pub const fn from_slice(slice: &[f32]) -> Self {
+ Self::new(slice[0], slice[1], slice[2], slice[3])
+ }
+
+ /// Writes the elements of `self` to the first 4 elements in `slice`.
+ ///
+ /// # Panics
+ ///
+ /// Panics if `slice` is less than 4 elements long.
+ #[inline]
+ pub fn write_to_slice(self, slice: &mut [f32]) {
+ unsafe {
+ assert!(slice.len() >= 4);
+ _mm_storeu_ps(slice.as_mut_ptr(), self.0);
+ }
+ }
+
+ /// Creates a 2D vector from the `x`, `y` and `z` elements of `self`, discarding `w`.
+ ///
+ /// Truncation to `Vec3` may also be performed by using `self.xyz()` or `Vec3::from()`.
+ ///
+ /// To truncate to `Vec3A` use `Vec3A::from()`.
+ #[inline]
+ pub fn truncate(self) -> Vec3 {
+ use crate::swizzles::Vec4Swizzles;
+ self.xyz()
+ }
+
+ /// Computes the dot product of `self` and `rhs`.
+ #[inline]
+ pub fn dot(self, rhs: Self) -> f32 {
+ unsafe { dot4(self.0, rhs.0) }
+ }
+
+ /// Returns a vector where every component is the dot product of `self` and `rhs`.
+ #[inline]
+ pub fn dot_into_vec(self, rhs: Self) -> Self {
+ Self(unsafe { dot4_into_m128(self.0, rhs.0) })
+ }
+
+ /// Returns a vector containing the minimum values for each element of `self` and `rhs`.
+ ///
+ /// In other words this computes `[self.x.min(rhs.x), self.y.min(rhs.y), ..]`.
+ #[inline]
+ pub fn min(self, rhs: Self) -> Self {
+ Self(unsafe { _mm_min_ps(self.0, rhs.0) })
+ }
+
+ /// Returns a vector containing the maximum values for each element of `self` and `rhs`.
+ ///
+ /// In other words this computes `[self.x.max(rhs.x), self.y.max(rhs.y), ..]`.
+ #[inline]
+ pub fn max(self, rhs: Self) -> Self {
+ Self(unsafe { _mm_max_ps(self.0, rhs.0) })
+ }
+
+ /// Component-wise clamping of values, similar to [`f32::clamp`].
+ ///
+ /// Each element in `min` must be less-or-equal to the corresponding element in `max`.
+ ///
+ /// # Panics
+ ///
+ /// Will panic if `min` is greater than `max` when `glam_assert` is enabled.
+ #[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)
+ }
+
+ /// Returns the horizontal minimum of `self`.
+ ///
+ /// In other words this computes `min(x, y, ..)`.
+ #[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)
+ }
+ }
+
+ /// Returns the horizontal maximum of `self`.
+ ///
+ /// In other words this computes `max(x, y, ..)`.
+ #[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)
+ }
+ }
+
+ /// Returns a vector mask containing the result of a `==` comparison for each element of
+ /// `self` and `rhs`.
+ ///
+ /// In other words, this computes `[self.x == rhs.x, self.y == rhs.y, ..]` for all
+ /// elements.
+ #[inline]
+ pub fn cmpeq(self, rhs: Self) -> BVec4A {
+ BVec4A(unsafe { _mm_cmpeq_ps(self.0, rhs.0) })
+ }
+
+ /// Returns a vector mask containing the result of a `!=` comparison for each element of
+ /// `self` and `rhs`.
+ ///
+ /// In other words this computes `[self.x != rhs.x, self.y != rhs.y, ..]` for all
+ /// elements.
+ #[inline]
+ pub fn cmpne(self, rhs: Self) -> BVec4A {
+ BVec4A(unsafe { _mm_cmpneq_ps(self.0, rhs.0) })
+ }
+
+ /// Returns a vector mask containing the result of a `>=` comparison for each element of
+ /// `self` and `rhs`.
+ ///
+ /// In other words this computes `[self.x >= rhs.x, self.y >= rhs.y, ..]` for all
+ /// elements.
+ #[inline]
+ pub fn cmpge(self, rhs: Self) -> BVec4A {
+ BVec4A(unsafe { _mm_cmpge_ps(self.0, rhs.0) })
+ }
+
+ /// Returns a vector mask containing the result of a `>` comparison for each element of
+ /// `self` and `rhs`.
+ ///
+ /// In other words this computes `[self.x > rhs.x, self.y > rhs.y, ..]` for all
+ /// elements.
+ #[inline]
+ pub fn cmpgt(self, rhs: Self) -> BVec4A {
+ BVec4A(unsafe { _mm_cmpgt_ps(self.0, rhs.0) })
+ }
+
+ /// Returns a vector mask containing the result of a `<=` comparison for each element of
+ /// `self` and `rhs`.
+ ///
+ /// In other words this computes `[self.x <= rhs.x, self.y <= rhs.y, ..]` for all
+ /// elements.
+ #[inline]
+ pub fn cmple(self, rhs: Self) -> BVec4A {
+ BVec4A(unsafe { _mm_cmple_ps(self.0, rhs.0) })
+ }
+
+ /// Returns a vector mask containing the result of a `<` comparison for each element of
+ /// `self` and `rhs`.
+ ///
+ /// In other words this computes `[self.x < rhs.x, self.y < rhs.y, ..]` for all
+ /// elements.
+ #[inline]
+ pub fn cmplt(self, rhs: Self) -> BVec4A {
+ BVec4A(unsafe { _mm_cmplt_ps(self.0, rhs.0) })
+ }
+
+ /// Returns a vector containing the absolute value of each element of `self`.
+ #[inline]
+ pub fn abs(self) -> Self {
+ Self(unsafe { crate::sse2::m128_abs(self.0) })
+ }
+
+ /// Returns a vector with elements representing the sign of `self`.
+ ///
+ /// - `1.0` if the number is positive, `+0.0` or `INFINITY`
+ /// - `-1.0` if the number is negative, `-0.0` or `NEG_INFINITY`
+ /// - `NAN` if the number is `NAN`
+ #[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)
+ }
+ }
+
+ /// Returns a bitmask with the lowest 4 bits set to the sign bits from the elements of `self`.
+ ///
+ /// A negative element results in a `1` bit and a positive element in a `0` bit. Element `x` goes
+ /// into the first lowest bit, element `y` into the second, etc.
+ #[inline]
+ pub fn is_negative_bitmask(self) -> u32 {
+ unsafe { _mm_movemask_ps(self.0) as u32 }
+ }
+
+ /// Returns `true` if, and only if, all elements are finite. If any element is either
+ /// `NaN`, positive or negative infinity, this will return `false`.
+ #[inline]
+ pub fn is_finite(self) -> bool {
+ self.x.is_finite() && self.y.is_finite() && self.z.is_finite() && self.w.is_finite()
+ }
+
+ /// Returns `true` if any elements are `NaN`.
+ #[inline]
+ pub fn is_nan(self) -> bool {
+ self.is_nan_mask().any()
+ }
+
+ /// Performs `is_nan` on each element of self, returning a vector mask of the results.
+ ///
+ /// In other words, this computes `[x.is_nan(), y.is_nan(), z.is_nan(), w.is_nan()]`.
+ #[inline]
+ pub fn is_nan_mask(self) -> BVec4A {
+ BVec4A(unsafe { _mm_cmpunord_ps(self.0, self.0) })
+ }
+
+ /// Computes the length of `self`.
+ #[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))
+ }
+ }
+
+ /// Computes the squared length of `self`.
+ ///
+ /// This is faster than `length()` as it avoids a square root operation.
+ #[doc(alias = "magnitude2")]
+ #[inline]
+ pub fn length_squared(self) -> f32 {
+ self.dot(self)
+ }
+
+ /// Computes `1.0 / length()`.
+ ///
+ /// For valid results, `self` must _not_ be of length zero.
+ #[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)))
+ }
+ }
+
+ /// Computes the Euclidean distance between two points in space.
+ #[inline]
+ pub fn distance(self, rhs: Self) -> f32 {
+ (self - rhs).length()
+ }
+
+ /// Compute the squared euclidean distance between two points in space.
+ #[inline]
+ pub fn distance_squared(self, rhs: Self) -> f32 {
+ (self - rhs).length_squared()
+ }
+
+ /// Returns `self` normalized to length 1.0.
+ ///
+ /// For valid results, `self` must _not_ be of length zero, nor very close to zero.
+ ///
+ /// See also [`Self::try_normalize`] and [`Self::normalize_or_zero`].
+ ///
+ /// Panics
+ ///
+ /// Will panic if `self` is zero length when `glam_assert` is enabled.
+ #[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
+ }
+ }
+
+ /// Returns `self` normalized to length 1.0 if possible, else returns `None`.
+ ///
+ /// In particular, if the input is zero (or very close to zero), or non-finite,
+ /// the result of this operation will be `None`.
+ ///
+ /// See also [`Self::normalize_or_zero`].
+ #[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
+ }
+ }
+
+ /// Returns `self` normalized to length 1.0 if possible, else returns zero.
+ ///
+ /// In particular, if the input is zero (or very close to zero), or non-finite,
+ /// the result of this operation will be zero.
+ ///
+ /// See also [`Self::try_normalize`].
+ #[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
+ }
+ }
+
+ /// Returns whether `self` is length `1.0` or not.
+ ///
+ /// Uses a precision threshold of `1e-6`.
+ #[inline]
+ pub fn is_normalized(self) -> bool {
+ // TODO: do something with epsilon
+ (self.length_squared() - 1.0).abs() <= 1e-4
+ }
+
+ /// Returns the vector projection of `self` onto `rhs`.
+ ///
+ /// `rhs` must be of non-zero length.
+ ///
+ /// # Panics
+ ///
+ /// Will panic if `rhs` is zero length when `glam_assert` is enabled.
+ #[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
+ }
+
+ /// Returns the vector rejection of `self` from `rhs`.
+ ///
+ /// The vector rejection is the vector perpendicular to the projection of `self` onto
+ /// `rhs`, in rhs words the result of `self - self.project_onto(rhs)`.
+ ///
+ /// `rhs` must be of non-zero length.
+ ///
+ /// # Panics
+ ///
+ /// Will panic if `rhs` has a length of zero when `glam_assert` is enabled.
+ #[must_use]
+ #[inline]
+ pub fn reject_from(self, rhs: Self) -> Self {
+ self - self.project_onto(rhs)
+ }
+
+ /// Returns the vector projection of `self` onto `rhs`.
+ ///
+ /// `rhs` must be normalized.
+ ///
+ /// # Panics
+ ///
+ /// Will panic if `rhs` is not normalized when `glam_assert` is enabled.
+ #[must_use]
+ #[inline]
+ pub fn project_onto_normalized(self, rhs: Self) -> Self {
+ glam_assert!(rhs.is_normalized());
+ rhs * self.dot(rhs)
+ }
+
+ /// Returns the vector rejection of `self` from `rhs`.
+ ///
+ /// The vector rejection is the vector perpendicular to the projection of `self` onto
+ /// `rhs`, in rhs words the result of `self - self.project_onto(rhs)`.
+ ///
+ /// `rhs` must be normalized.
+ ///
+ /// # Panics
+ ///
+ /// Will panic if `rhs` is not normalized when `glam_assert` is enabled.
+ #[must_use]
+ #[inline]
+ pub fn reject_from_normalized(self, rhs: Self) -> Self {
+ self - self.project_onto_normalized(rhs)
+ }
+
+ /// Returns a vector containing the nearest integer to a number for each element of `self`.
+ /// Round half-way cases away from 0.0.
+ #[inline]
+ pub fn round(self) -> Self {
+ Self(unsafe { m128_round(self.0) })
+ }
+
+ /// Returns a vector containing the largest integer less than or equal to a number for each
+ /// element of `self`.
+ #[inline]
+ pub fn floor(self) -> Self {
+ Self(unsafe { m128_floor(self.0) })
+ }
+
+ /// Returns a vector containing the smallest integer greater than or equal to a number for
+ /// each element of `self`.
+ #[inline]
+ pub fn ceil(self) -> Self {
+ Self(unsafe { m128_ceil(self.0) })
+ }
+
+ /// Returns a vector containing the fractional part of the vector, e.g. `self -
+ /// self.floor()`.
+ ///
+ /// Note that this is fast but not precise for large numbers.
+ #[inline]
+ pub fn fract(self) -> Self {
+ self - self.floor()
+ }
+
+ /// Returns a vector containing `e^self` (the exponential function) for each element of
+ /// `self`.
+ #[inline]
+ pub fn exp(self) -> Self {
+ Self::new(self.x.exp(), self.y.exp(), self.z.exp(), self.w.exp())
+ }
+
+ /// Returns a vector containing each element of `self` raised to the power of `n`.
+ #[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),
+ )
+ }
+
+ /// Returns a vector containing the reciprocal `1.0/n` of each element of `self`.
+ #[inline]
+ pub fn recip(self) -> Self {
+ Self(unsafe { _mm_div_ps(Self::ONE.0, self.0) })
+ }
+
+ /// Performs a linear interpolation between `self` and `rhs` based on the value `s`.
+ ///
+ /// When `s` is `0.0`, the result will be equal to `self`. When `s` is `1.0`, the result
+ /// will be equal to `rhs`. When `s` is outside of range `[0, 1]`, the result is linearly
+ /// extrapolated.
+ #[doc(alias = "mix")]
+ #[inline]
+ pub fn lerp(self, rhs: Self, s: f32) -> Self {
+ self + ((rhs - self) * s)
+ }
+
+ /// Returns true if the absolute difference of all elements between `self` and `rhs` is
+ /// less than or equal to `max_abs_diff`.
+ ///
+ /// This can be used to compare if two vectors contain similar elements. It works best when
+ /// comparing with a known value. The `max_abs_diff` that should be used used depends on
+ /// the values being compared against.
+ ///
+ /// For more see
+ /// [comparing floating point numbers](https://randomascii.wordpress.com/2012/02/25/comparing-floating-point-numbers-2012-edition/).
+ #[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()
+ }
+
+ /// Returns a vector with a length no less than `min` and no more than `max`
+ ///
+ /// # Panics
+ ///
+ /// Will panic if `min` is greater than `max` when `glam_assert` is enabled.
+ #[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
+ }
+ }
+
+ /// Returns a vector with a length no more than `max`
+ 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
+ }
+ }
+
+ /// Returns a vector with a length no less than `min`
+ 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
+ }
+ }
+
+ /// Fused multiply-add. Computes `(self * a) + b` element-wise with only one rounding
+ /// error, yielding a more accurate result than an unfused multiply-add.
+ ///
+ /// Using `mul_add` *may* be more performant than an unfused multiply-add if the target
+ /// architecture has a dedicated fma CPU instruction. However, this is not always true,
+ /// and will be heavily dependant on designing algorithms with specific target hardware in
+ /// mind.
+ #[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),
+ )
+ }
+
+ /// Casts all elements of `self` to `f64`.
+ #[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)
+ }
+
+ /// Casts all elements of `self` to `i32`.
+ #[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)
+ }
+
+ /// Casts all elements of `self` to `u32`.
+ #[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() }
+ }
+}
generated by cgit v1.2.3 (git 2.25.1) at 2024-06-13 00:04:33 +0000