Support #1450
Feature #1393: Redesign of Hybrid API (GPU Manager) to support concurrent kernel execution
Clean up and add CUDA example programs
Start date:
02/28/2017
Due date:
% Done:
0%
Description
CUDA example programs under examples/charm++/cuda need to be refactored with the new design of GPUManager.
Especially, lazy host memory allocation & deallocation is no longer possible due to the removal of polling,
and those calls should be replaced by either hapi_poolMalloc or hapi_poolFree.
Related issues
History
#1 Updated by Jaemin Choi about 1 year ago
- Status changed from New to Resolved
On gerrit for review.
[[https://charm.cs.illinois.edu/gerrit/#/c/2275/]]
#3 Updated by Jaemin Choi 8 months ago
- Status changed from Resolved to In Progress
- Subject changed from Refactor CUDA example programs to fit new GPUManager design to Clean up and add CUDA example programs
Current example programs in examples/cuda should be cleaned up, and some new ones such as stencil2d and AMR should be added.
#4 Updated by Jaemin Choi 7 months ago
- Related to Cleanup #980: Cleanup examples/charm++/cuda/hello added
#5 Updated by Jaemin Choi 7 months ago
- Status changed from In Progress to Feedback