CUDA Multi Process Service
CUDA Multi Process Service
Tuesday, August 17, 2021
| ORNL | Remote Participation Only | 1:00 PM – 3:00 PM (ET) |
| NERSC | Remote Participation Only | 10:00 AM – 12:00 PM (PT) |
NVIDIA will present “CUDA Multi Process Service” on Tuesday, August 17, 2021. This event is a continuation of the CUDA Training Series and will be presented by Max Katz from NVIDIA.
CUDA allows the user to target a GPU with multiple processes. A common use case in HPC is using multiple MPI ranks to target the same GPU. This session will describe how NVIDIA GPUs work when targeted by multiple processes simultaneously, and present the CUDA multi-process service (MPS) which helps improve the scheduling efficiency of this use case. The multi-instance GPU (MIG) feature on the NVIDIA A100 GPU will also be discussed, and the difference between MPS and MIG (and when you would use one or the other) will be explained. After the presentation, there will be a hands-on session where participants can complete example exercises meant to reinforce the presented concepts and observe their impact on application performance.
Remote Participation
All remote participants can watch the presentations via web broadcast and will have access to the training exercises, but temporary access to the compute systems will be limited as follows:
- Current NERSC users will have Cori-GPU access temporarily added to their accounts.
- Temporary Summit access will not be available for remote participants.
If you have any questions, please contact Subil Abraham ([email protected])
[tw-toggle title=”Registration”] [/tw-toggle] [tw-toggle title=”Remote Connection Details”] [/tw-toggle] [tw-toggle title=”Presentations”] (slides|recording)
[/tw-toggle] [tw-toggle title=”Exercises”] The exercises for this training can be found here: https://github.com/olcf/cuda-training-series/tree/master/exercises/hw11 [/tw-toggle] [tw-toggle title=”Survey”] [/tw-toggle]
