apfp
apfp copied to clipboard
spcl
FPGA acceleration of arbitrary precision floating point computations.
Results
3
apfp issues
Sort by
recently updated
recently updated
newest added
Isolate Karatsuba multiplication, then partition into streaming subcomponents, to aid routing.
Pull into a free-running kernel that communicates with the matrix multiplier using streams.
We will want some integration tests for this, but it's probably more important to not have a huge diff piling up
