Fork me on GitHub
#data-science
<
2016-03-13
>
blueberry15:03:52

New Matrix Multiplication benchmarks - Neanderthal up to 60 times faster than core.matrix with Vectorz

mikera23:03:45

It would be great if Neanderthal simply implemented the core.matrix protocols, then people could use it as a core.matrix implementation for situations where it makes sense. I really think it is an architectural dead-end for Neanderthal to develop a separate API. You'll simply get less users for Neanderthal and fragment the Clojure library ecosystem which doesn't help anyone. In the absence of that, we'll just need to develop separate BLAS implementations for core.matrix.

mikera23:03:07

I'm happy to take PRs if needed to make core.matrix work well with Neanderthal, I appreciate the amount of engineering work that has gone in to it and I'd really like to see it become a proper part of the ecosystem rather than a standalone incompatible library.