Closed NeverGivinUp closed 4 years ago
The simd methods provided will likely cover about as much of f32/f64 cover in terms of trig and utilities. Additional content would probably be better suited for an external crate built on top of core::simd
.
I think it would be a good idea to keep BLAS explicitly out of scope.
(To be clear, I think you should be able to build competitive BLAS using these components, but this should probably be focused on being smaller)
In my ideal world, there would be a BLAS package that can run in Web Assembly in browsers, so that interacting with special data types is encapsulated and as a programmer I can just use BLAS level 2 and 3 and know the that the optimal performance of the machine will happen.
Is that in scope of this project group? If not, what would be required for that and which parts are in scope of this project group and which are outside of its scope?