"cuDNN: Efficient Primitives for Deep Learning."

Sharan Chetlur et al. (2014)
a service of Schloss Dagstuhl - Leibniz Center for Informatics