June 22–26, 2014
Leipzig, Germany

Presentation Details

Name: (22a) Efficient Execution of Multiple Applications Using Process Migration
Time: Thursday, June 26, 2014
10:30 am - 11:00 am
Room:   Hall 4
CCL - Congress Center Leipzig
Breaks:10:30 am - 11:00 am Coffee Break
07:30 am - 10:30 am Welcome Coffee
Presenter:   Taichirou Suzuki, Tokyo Institute of Technology
Abstract:   NVIDIA's Kepler generation GPUs and successors support a hardware function called Hyper-Q that enables multiple CUDA applications to run efficiently on single device. Using Hyper-Q can reduce the total execution time of applications when applications have differing characteristic. However, normally, a CUDA application assumes that it can use the whole device memory on GPU exclusively. So we propose Mobile CUDA which enables CUDA applications to migrate in order to avoid running out of memory.

Taichirou Suzuki, Akira Nukada & Satoshi Matsuoka, Tokyo Institute of Technology