> The workflow is, I set up Vec3x8, and Quaternionx8, which wrap a simd instrinsic for each field (x: f32x8, y: f32x8...) etc. The point is why do that in a hu

> The workflow is, I set up Vec3x8, and Quaternionx8, which wrap a simd instrinsic for each field (x: f32x8, y: f32x8...) etc. The point is why do that in a human, error-prone way every time (in the current software I work on, I would have hundreds of those structs for instance) when you can program your compiler to do it once for you