WIP

bitshifter · Feb 6, 2024 · 1332b6a · 1332b6a
1 parent bcbc026
commit 1332b6a
Show file tree

Hide file tree

Showing 36 changed files with 214 additions and 142 deletions.
diff --git a/codegen/templates/mat.rs.tera b/codegen/templates/mat.rs.tera
@@ -451,9 +451,9 @@ impl {{ self_t }} {
     pub fn from_mat4(m: {{ mat4_t }}) -> Self {
         {% if self_t == "Mat3A" %}
             Self::from_cols(
-                m.x_axis.into(),
-                m.y_axis.into(),
-                m.z_axis.into(),
+                m.x_axis.xyz().into(),
+                m.y_axis.xyz().into(),
+                m.z_axis.xyz().into(),
             )
         {% else %}
             Self::from_cols(
@@ -2278,6 +2278,32 @@ impl From<Mat3> for Mat3A {
 }
 {% endif %}
 
+{% if self_t == "Mat4" %}
+impl From<Mat4A> for Mat4 {
+    #[inline]
+    fn from(m: Mat4A) -> Self {
+        Self {
+            x_axis: m.x_axis.into(),
+            y_axis: m.y_axis.into(),
+            z_axis: m.z_axis.into(),
+            w_axis: m.w_axis.into(),
+        }
+    }
+}
+{% elif self_t == "Mat4A" %}
+impl From<Mat4> for Mat4A {
+    #[inline]
+    fn from(m: Mat4) -> Self {
+        Self {
+            x_axis: m.x_axis.into(),
+            y_axis: m.y_axis.into(),
+            z_axis: m.z_axis.into(),
+            w_axis: m.w_axis.into(),
+        }
+    }
+}
+{% endif %}
+
 impl Sum<Self> for {{ self_t }} {
     fn sum<I>(iter: I) -> Self
     where

diff --git a/codegen/templates/vec.rs.tera b/codegen/templates/vec.rs.tera
@@ -30,7 +30,6 @@
         {% endif %}
         {% set vec2_t = "Vec2" %}
         {% set vec3_t = "Vec3" %}
-        {% set vec3a_t = "Vec3A" %}
         {% set vec4_t = "Vec4" %}
     {% elif scalar_t == "f64" %}
         {% set self_t = "DVec" ~ dim %}
@@ -144,8 +143,11 @@
     {% if self_t != vec3_t %}
         {{ vec3_t }},
     {% endif %}
-    {% if dim == 4 and scalar_t == "f32" %}
-        {{ vec3a_t }},
+    {% if self_t == "Vec4A" or self_t == "Vec4" %}
+        Vec3A,
+    {% endif %}
+    {% if self_t == "Vec4" %}
+        Vec4A,
     {% endif %}
     {% if dim > 2 and self_t != vec4_t %}
         {{ vec4_t }},
@@ -478,11 +480,7 @@ impl {{ self_t }} {
     #[inline]
     #[must_use]
     pub(crate) fn from_vec4(v: {{ vec4_t }}) -> Self {
-        {% if is_scalar %}
-            Self { x: v.x, y: v.y, z: v.z }
-        {% else %}
-            Self(v.0)
-        {% endif %}
+        Self::new(v.x, v.y, v.z)
     }
 
     /// Creates a 4D vector from `self` and the given `w` value.
@@ -512,8 +510,7 @@ impl {{ self_t }} {
     #[inline]
     #[must_use]
     pub fn truncate(self) -> {{ vec3_t }} {
-        use crate::swizzles::Vec4Swizzles;
-        self.xyz()
+        {{ vec3_t }}::new(self.x, self.y, self.z)
     }
 {% endif %}
 
@@ -3066,12 +3063,13 @@ impl From<Vec3> for Vec3A {
     }
 }
 
-impl From<Vec4> for Vec3A {
+{#
+impl From<Vec4A> for Vec3A {
     /// Creates a [`Vec3A`] from the `x`, `y` and `z` elements of `self` discarding `w`.
     ///
     /// On architectures where SIMD is supported such as SSE2 on `x86_64` this conversion is a noop.
     #[inline]
-    fn from(v: Vec4) -> Self {
+    fn from(v: Vec4A) -> Self {
         {% if is_scalar %}
             Self {
                 x: v.x,
@@ -3083,6 +3081,7 @@ impl From<Vec4> for Vec3A {
         {% endif %}
     }
 }
+#}
 
 impl From<Vec3A> for Vec3 {
     #[inline]
@@ -3116,7 +3115,13 @@ impl From<Vec3A> for Vec3 {
 impl From<(Vec3A, f32)> for {{ self_t }} {
     #[inline]
     fn from((v, w): (Vec3A, f32)) -> Self {
-        v.extend(w)
+        {% if is_simd %}
+            let v = Self(v.0);
+            v.w = w;
+            v
+        {% else %}
+            Self::new(v.x, v.y, v.z, w)
+        {% endif %}
     }
 }
 
@@ -3126,6 +3131,21 @@ impl From<(f32, Vec3A)> for {{ self_t }} {
         Self::new(x, v.x, v.y, v.z)
     }
 }
+{% if is_align %}
+impl From<Vec4> for Vec4A {
+    #[inline]
+    fn from(v: Vec4) -> Self {
+        Self::new(v.x, v.y, v.z, v.w)
+    }
+}
+{% else %}
+impl From<Vec4A> for Vec4 {
+    #[inline]
+    fn from(v: Vec4A) -> Self {
+        Self::new(v.x, v.y, v.z, v.w)
+    }
+}
+{% endif %}
 {% endif %}
 
 {% if dim == 3 %}

diff --git a/src/f32.rs b/src/f32.rs
@@ -66,6 +66,7 @@ pub use mat2::{mat2, Mat2};
 pub use mat3::{mat3, Mat3};
 pub use mat3a::{mat3a, Mat3A};
 pub use mat4::{mat4, Mat4};
+pub use mat4a::{mat4a, Mat4A};
 pub use quat::{quat, Quat};
 pub use vec2::{vec2, Vec2};
 pub use vec3::{vec3, Vec3};

diff --git a/src/f32/coresimd/mat3a.rs b/src/f32/coresimd/mat3a.rs
@@ -152,7 +152,11 @@ impl Mat3A {
     #[inline]
     #[must_use]
     pub fn from_mat4(m: Mat4) -> Self {
-        Self::from_cols(m.x_axis.into(), m.y_axis.into(), m.z_axis.into())
+        Self::from_cols(
+            m.x_axis.xyz().into(),
+            m.y_axis.xyz().into(),
+            m.z_axis.xyz().into(),
+        )
     }
 
     /// Creates a 3D rotation matrix from the given quaternion.

diff --git a/src/f32/coresimd/mat4a.rs b/src/f32/coresimd/mat4a.rs
@@ -1334,6 +1334,18 @@ impl MulAssign<f32> for Mat4A {
     }
 }
 
+impl From<Mat4> for Mat4A {
+    #[inline]
+    fn from(m: Mat4) -> Self {
+        Self {
+            x_axis: m.x_axis.into(),
+            y_axis: m.y_axis.into(),
+            z_axis: m.z_axis.into(),
+            w_axis: m.w_axis.into(),
+        }
+    }
+}
+
 impl Sum<Self> for Mat4A {
     fn sum<I>(iter: I) -> Self
     where

diff --git a/src/f32/coresimd/vec3a.rs b/src/f32/coresimd/vec3a.rs
@@ -143,7 +143,7 @@ impl Vec3A {
     #[inline]
     #[must_use]
     pub(crate) fn from_vec4(v: Vec4) -> Self {
-        Self(v.0)
+        Self::new(v.x, v.y, v.z)
     }
 
     /// Creates a 4D vector from `self` and the given `w` value.
@@ -1209,16 +1209,6 @@ impl From<Vec3> for Vec3A {
     }
 }
 
-impl From<Vec4> for Vec3A {
-    /// Creates a [`Vec3A`] from the `x`, `y` and `z` elements of `self` discarding `w`.
-    ///
-    /// On architectures where SIMD is supported such as SSE2 on `x86_64` this conversion is a noop.
-    #[inline]
-    fn from(v: Vec4) -> Self {
-        Self(v.0)
-    }
-}
-
 impl From<Vec3A> for Vec3 {
     #[inline]
     fn from(v: Vec3A) -> Self {

diff --git a/src/f32/coresimd/vec4a.rs b/src/f32/coresimd/vec4a.rs
@@ -149,8 +149,7 @@ impl Vec4A {
     #[inline]
     #[must_use]
     pub fn truncate(self) -> Vec3 {
-        use crate::swizzles::Vec4Swizzles;
-        self.xyz()
+        Vec3::new(self.x, self.y, self.z)
     }
 
     /// Computes the dot product of `self` and `rhs`.
@@ -1117,7 +1116,9 @@ impl From<Vec4A> for (f32, f32, f32, f32) {
 impl From<(Vec3A, f32)> for Vec4A {
     #[inline]
     fn from((v, w): (Vec3A, f32)) -> Self {
-        v.extend(w)
+        let v = Self(v.0);
+        v.w = w;
+        v
     }
 }
 
@@ -1128,6 +1129,13 @@ impl From<(f32, Vec3A)> for Vec4A {
     }
 }
 
+impl From<Vec4> for Vec4A {
+    #[inline]
+    fn from(v: Vec4) -> Self {
+        Self::new(v.x, v.y, v.z, v.w)
+    }
+}
+
 impl From<(Vec3, f32)> for Vec4A {
     #[inline]
     fn from((v, w): (Vec3, f32)) -> Self {

diff --git a/src/f32/mat4.rs b/src/f32/mat4.rs
@@ -1240,6 +1240,18 @@ impl MulAssign<f32> for Mat4 {
     }
 }
 
+impl From<Mat4A> for Mat4 {
+    #[inline]
+    fn from(m: Mat4A) -> Self {
+        Self {
+            x_axis: m.x_axis.into(),
+            y_axis: m.y_axis.into(),
+            z_axis: m.z_axis.into(),
+            w_axis: m.w_axis.into(),
+        }
+    }
+}
+
 impl Sum<Self> for Mat4 {
     fn sum<I>(iter: I) -> Self
     where

diff --git a/src/f32/scalar/mat3a.rs b/src/f32/scalar/mat3a.rs
@@ -153,7 +153,11 @@ impl Mat3A {
     #[inline]
     #[must_use]
     pub fn from_mat4(m: Mat4) -> Self {
-        Self::from_cols(m.x_axis.into(), m.y_axis.into(), m.z_axis.into())
+        Self::from_cols(
+            m.x_axis.xyz().into(),
+            m.y_axis.xyz().into(),
+            m.z_axis.xyz().into(),
+        )
     }
 
     /// Creates a 3D rotation matrix from the given quaternion.

diff --git a/src/f32/scalar/mat4a.rs b/src/f32/scalar/mat4a.rs
@@ -1247,6 +1247,18 @@ impl MulAssign<f32> for Mat4A {
     }
 }
 
+impl From<Mat4> for Mat4A {
+    #[inline]
+    fn from(m: Mat4) -> Self {
+        Self {
+            x_axis: m.x_axis.into(),
+            y_axis: m.y_axis.into(),
+            z_axis: m.z_axis.into(),
+            w_axis: m.w_axis.into(),
+        }
+    }
+}
+
 impl Sum<Self> for Mat4A {
     fn sum<I>(iter: I) -> Self
     where

diff --git a/src/f32/scalar/vec3a.rs b/src/f32/scalar/vec3a.rs
@@ -150,11 +150,7 @@ impl Vec3A {
     #[inline]
     #[must_use]
     pub(crate) fn from_vec4(v: Vec4) -> Self {
-        Self {
-            x: v.x,
-            y: v.y,
-            z: v.z,
-        }
+        Self::new(v.x, v.y, v.z)
     }
 
     /// Creates a 4D vector from `self` and the given `w` value.
@@ -1327,20 +1323,6 @@ impl From<Vec3> for Vec3A {
     }
 }
 
-impl From<Vec4> for Vec3A {
-    /// Creates a [`Vec3A`] from the `x`, `y` and `z` elements of `self` discarding `w`.
-    ///
-    /// On architectures where SIMD is supported such as SSE2 on `x86_64` this conversion is a noop.
-    #[inline]
-    fn from(v: Vec4) -> Self {
-        Self {
-            x: v.x,
-            y: v.y,
-            z: v.z,
-        }
-    }
-}
-
 impl From<Vec3A> for Vec3 {
     #[inline]
     fn from(v: Vec3A) -> Self {

diff --git a/src/f32/scalar/vec4a.rs b/src/f32/scalar/vec4a.rs
@@ -172,8 +172,7 @@ impl Vec4A {
     #[inline]
     #[must_use]
     pub fn truncate(self) -> Vec3 {
-        use crate::swizzles::Vec4Swizzles;
-        self.xyz()
+        Vec3::new(self.x, self.y, self.z)
     }
 
     /// Computes the dot product of `self` and `rhs`.
@@ -1331,7 +1330,7 @@ impl From<Vec4A> for (f32, f32, f32, f32) {
 impl From<(Vec3A, f32)> for Vec4A {
     #[inline]
     fn from((v, w): (Vec3A, f32)) -> Self {
-        v.extend(w)
+        Self::new(v.x, v.y, v.z, w)
     }
 }
 
@@ -1342,6 +1341,13 @@ impl From<(f32, Vec3A)> for Vec4A {
     }
 }
 
+impl From<Vec4> for Vec4A {
+    #[inline]
+    fn from(v: Vec4) -> Self {
+        Self::new(v.x, v.y, v.z, v.w)
+    }
+}
+
 impl From<(Vec3, f32)> for Vec4A {
     #[inline]
     fn from((v, w): (Vec3, f32)) -> Self {

diff --git a/src/f32/sse2/mat3a.rs b/src/f32/sse2/mat3a.rs
@@ -155,7 +155,11 @@ impl Mat3A {
     #[inline]
     #[must_use]
     pub fn from_mat4(m: Mat4) -> Self {
-        Self::from_cols(m.x_axis.into(), m.y_axis.into(), m.z_axis.into())
+        Self::from_cols(
+            m.x_axis.xyz().into(),
+            m.y_axis.xyz().into(),
+            m.z_axis.xyz().into(),
+        )
     }
 
     /// Creates a 3D rotation matrix from the given quaternion.