{% macro impl_mat4_inverse() %}
    // Based on https://github.com/g-truc/glm `glm_mat4_inverse`
    let fac0 = {
        let swp0a = i32x4_shuffle::<3, 3, 7, 7>(self.w_axis.0, self.z_axis.0);
        let swp0b = i32x4_shuffle::<2, 2, 6, 6>(self.w_axis.0, self.z_axis.0);

        let swp00 = i32x4_shuffle::<2, 2, 6, 6>(self.z_axis.0, self.y_axis.0);
        let swp01 = i32x4_shuffle::<0, 0, 4, 6>(swp0a, swp0a);
        let swp02 = i32x4_shuffle::<0, 0, 4, 6>(swp0b, swp0b);
        let swp03 = i32x4_shuffle::<3, 3, 7, 7>(self.z_axis.0, self.y_axis.0);

        let mul00 = f32x4_mul(swp00, swp01);
        let mul01 = f32x4_mul(swp02, swp03);
        f32x4_sub(mul00, mul01)
    };
    let fac1 = {
        let swp0a = i32x4_shuffle::<3, 3, 7, 7>(self.w_axis.0, self.z_axis.0);
        let swp0b = i32x4_shuffle::<1, 1, 5, 5>(self.w_axis.0, self.z_axis.0);

        let swp00 = i32x4_shuffle::<1, 1, 5, 5>(self.z_axis.0, self.y_axis.0);
        let swp01 = i32x4_shuffle::<0, 0, 4, 6>(swp0a, swp0a);
        let swp02 = i32x4_shuffle::<0, 0, 4, 6>(swp0b, swp0b);
        let swp03 = i32x4_shuffle::<3, 3, 7, 7>(self.z_axis.0, self.y_axis.0);

        let mul00 = f32x4_mul(swp00, swp01);
        let mul01 = f32x4_mul(swp02, swp03);
        f32x4_sub(mul00, mul01)
    };
    let fac2 = {
        let swp0a = i32x4_shuffle::<2, 2, 6, 6>(self.w_axis.0, self.z_axis.0);
        let swp0b = i32x4_shuffle::<1, 1, 5, 5>(self.w_axis.0, self.z_axis.0);

        let swp00 = i32x4_shuffle::<1, 1, 5, 5>(self.z_axis.0, self.y_axis.0);
        let swp01 = i32x4_shuffle::<0, 0, 4, 6>(swp0a, swp0a);
        let swp02 = i32x4_shuffle::<0, 0, 4, 6>(swp0b, swp0b);
        let swp03 = i32x4_shuffle::<2, 2, 6, 6>(self.z_axis.0, self.y_axis.0);

        let mul00 = f32x4_mul(swp00, swp01);
        let mul01 = f32x4_mul(swp02, swp03);
        f32x4_sub(mul00, mul01)
    };
    let fac3 = {
        let swp0a = i32x4_shuffle::<3, 3, 7, 7>(self.w_axis.0, self.z_axis.0);
        let swp0b = i32x4_shuffle::<0, 0, 4, 4>(self.w_axis.0, self.z_axis.0);

        let swp00 = i32x4_shuffle::<0, 0, 4, 4>(self.z_axis.0, self.y_axis.0);
        let swp01 = i32x4_shuffle::<0, 0, 4, 6>(swp0a, swp0a);
        let swp02 = i32x4_shuffle::<0, 0, 4, 6>(swp0b, swp0b);
        let swp03 = i32x4_shuffle::<3, 3, 7, 7>(self.z_axis.0, self.y_axis.0);

        let mul00 = f32x4_mul(swp00, swp01);
        let mul01 = f32x4_mul(swp02, swp03);
        f32x4_sub(mul00, mul01)
    };
    let fac4 = {
        let swp0a = i32x4_shuffle::<2, 2, 6, 6>(self.w_axis.0, self.z_axis.0);
        let swp0b = i32x4_shuffle::<0, 0, 4, 4>(self.w_axis.0, self.z_axis.0);

        let swp00 = i32x4_shuffle::<0, 0, 4, 4>(self.z_axis.0, self.y_axis.0);
        let swp01 = i32x4_shuffle::<0, 0, 4, 6>(swp0a, swp0a);
        let swp02 = i32x4_shuffle::<0, 0, 4, 6>(swp0b, swp0b);
        let swp03 = i32x4_shuffle::<2, 2, 6, 6>(self.z_axis.0, self.y_axis.0);

        let mul00 = f32x4_mul(swp00, swp01);
        let mul01 = f32x4_mul(swp02, swp03);
        f32x4_sub(mul00, mul01)
    };
    let fac5 = {
        let swp0a = i32x4_shuffle::<1, 1, 5, 5>(self.w_axis.0, self.z_axis.0);
        let swp0b = i32x4_shuffle::<0, 0, 4, 4>(self.w_axis.0, self.z_axis.0);

        let swp00 = i32x4_shuffle::<0, 0, 4, 4>(self.z_axis.0, self.y_axis.0);
        let swp01 = i32x4_shuffle::<0, 0, 4, 6>(swp0a, swp0a);
        let swp02 = i32x4_shuffle::<0, 0, 4, 6>(swp0b, swp0b);
        let swp03 = i32x4_shuffle::<1, 1, 5, 5>(self.z_axis.0, self.y_axis.0);

        let mul00 = f32x4_mul(swp00, swp01);
        let mul01 = f32x4_mul(swp02, swp03);
        f32x4_sub(mul00, mul01)
    };
    let sign_a = f32x4(-1.0, 1.0, -1.0, 1.0);
    let sign_b = f32x4(1.0, -1.0, 1.0, -1.0);

    let temp0 = i32x4_shuffle::<0, 0, 4, 4>(self.y_axis.0, self.x_axis.0);
    let vec0 = i32x4_shuffle::<0, 2, 6, 6>(temp0, temp0);

    let temp1 = i32x4_shuffle::<1, 1, 5, 5>(self.y_axis.0, self.x_axis.0);
    let vec1 = i32x4_shuffle::<0, 2, 6, 6>(temp1, temp1);

    let temp2 = i32x4_shuffle::<2, 2, 6, 6>(self.y_axis.0, self.x_axis.0);
    let vec2 = i32x4_shuffle::<0, 2, 6, 6>(temp2, temp2);

    let temp3 = i32x4_shuffle::<3, 3, 7, 7>(self.y_axis.0, self.x_axis.0);
    let vec3 = i32x4_shuffle::<0, 2, 6, 6>(temp3, temp3);

    let mul00 = f32x4_mul(vec1, fac0);
    let mul01 = f32x4_mul(vec2, fac1);
    let mul02 = f32x4_mul(vec3, fac2);
    let sub00 = f32x4_sub(mul00, mul01);
    let add00 = f32x4_add(sub00, mul02);
    let inv0 = f32x4_mul(sign_b, add00);

    let mul03 = f32x4_mul(vec0, fac0);
    let mul04 = f32x4_mul(vec2, fac3);
    let mul05 = f32x4_mul(vec3, fac4);
    let sub01 = f32x4_sub(mul03, mul04);
    let add01 = f32x4_add(sub01, mul05);
    let inv1 = f32x4_mul(sign_a, add01);

    let mul06 = f32x4_mul(vec0, fac1);
    let mul07 = f32x4_mul(vec1, fac3);
    let mul08 = f32x4_mul(vec3, fac5);
    let sub02 = f32x4_sub(mul06, mul07);
    let add02 = f32x4_add(sub02, mul08);
    let inv2 = f32x4_mul(sign_b, add02);

    let mul09 = f32x4_mul(vec0, fac2);
    let mul10 = f32x4_mul(vec1, fac4);
    let mul11 = f32x4_mul(vec2, fac5);
    let sub03 = f32x4_sub(mul09, mul10);
    let add03 = f32x4_add(sub03, mul11);
    let inv3 = f32x4_mul(sign_a, add03);

    let row0 = i32x4_shuffle::<0, 0, 4, 4>(inv0, inv1);
    let row1 = i32x4_shuffle::<0, 0, 4, 4>(inv2, inv3);
    let row2 = i32x4_shuffle::<0, 2, 4, 6>(row0, row1);

    let dot0 = dot4(self.x_axis.0, row2);
    glam_assert!(dot0 != 0.0);

    let rcp0 = f32x4_splat(dot0.recip());

    Self {
        x_axis: Vec4(f32x4_mul(inv0, rcp0)),
        y_axis: Vec4(f32x4_mul(inv1, rcp0)),
        z_axis: Vec4(f32x4_mul(inv2, rcp0)),
        w_axis: Vec4(f32x4_mul(inv3, rcp0)),
    }
{% endmacro impl_mat4_inverse %}