Presentation Details |
|
| |
| Name: |
 |
(06a) Efficient Utilization of Memory Hierarchy on GPU Clusters: Optimization Methods & Performance Models |
|
| |
| 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: |
|
Guanghao Jin, Tokyo Institute of Technology |
|
| |
| Abstract: |
|
The domain of the stencil computation is limited by the memory capacity of GPUs on a GPU cluster. As the domain grows to cope with higher accuracy requirements, more GPUs need to be employed to extend the memory capacity. In this poster, we introduce new optimization methods which apply temporal blocking method to device memory and registers of a set of GPUs to allow computations on the domain that is bigger than the memory capacity of GPUs while maintaining high performance on TSUBAME2.5 GPU cluster. We also build the performance model to estimate the performance and decide the parameters for our optimization methods.
Authors
Guanghao Jin, Toshio Endo & Satoshi Matsuoka, Tokyo Institute of Technology |
|
| |
|
;){kind=link}