Ray Bittner, Erik Ruf, and Alessandro Forin
17 May 2013
We describe a mechanism for connecting GPU and FPGA devices directly via the PCI Express bus, enabling the transfer of data between these heterogeneous computing units without the intermediate use of system memory. We evaluate the performance benefits of this approach over a range of transfer sizes, and demonstrate its utility in a computer vision application. We find that bypassing system memory yields improvements as high as 2.2x in data transfer speed, and 1.9x in application performance.
|Published in||Cluster Computing|