computeLib performance

Open sorear opened this issue 5 years ago • 0 comments

I have a hunch that for the special case we care about - evaluating monomorphic functions which are extractable to ML on ground arguments - we can do significantly better than the general case. I'll either update this with more details or close it after I've had a chance to write and test a proof of concept.

Sep 17 '20 03:09 sorear