mclaughlin6464/pearce

Pickle computed kernels?

Opened this issue · 0 comments

Building the kernel is annoyingly slow. It'd be nice if I could somehow pickle ones that have been computed. However, I don't know if that'll help much, since loading a several GB array may be just as slow as computing it. I also am not sure what unique identifying information could be used to pin a unique kernel. Additionally, should I pickle the kernel, or the whole Emu object? Or Define a scheme to load a pickled emu from file? This is something similar to what I'd want to do for software release, so that may be where I want to go.