Search papers, labs, and topics across Lattice.
1
0
3
10
Achieve up to 3.48x faster sparse matrix-vector multiplication on GPUs by compressing matrices with a new entropy coding technique, dtANS, that's faster to decode in parallel than existing methods.
