Improving GPU Matrix Multiplication by Leveraging Bit Level Granularity and Compression.

Johannes Fett, Christian Schwarz, Urs Kober, Dirk Habich, Wolfgang Lehner

Browse the full BTW paper archive.