Page 2 of 2

Re: Eigen MAGMA backend implementation project

Posted: Fri Aug 16, 2013 11:50 am
by bravegag
Doing the following (increasing the DP rate of the Titan card) improved the benchmark results greatly in some cases e.g. now the matrixMulCUBLAS example modified to DGEMM tops at 1.3 TFlop/s:

Image