I created this crate to solve a fairly narrow issue on a personal project where exp was taking up enough time that some rudimentary SIMD improvement was a very big performance win, but I wanted to have less boilerplate than using the wide crate and handling chunking my vectors and dealing with the remainder manually. I don't know if it will compete with intrinsics for all or most cases but it may be useful for some people with similar circumstances, and it compiles on stable Rust today (again thanks to the amazing wide crate for implementing many SIMD math operations in pure Rust, it makes things so much easier).
I also think that the concept of using macros to dispatch SIMD operations uniformly across scalars and SIMD types without boilerplate or needing to abandon native Vecs for special containers is something worth exploring further.