Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Endoscopy tool tracking not working with MPS after PR 550 note #590

Open
wants to merge 3 commits into
base: main
Choose a base branch
from
Open

Endoscopy tool tracking not working with MPS after PR 550 note #590

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
462e5fb
Endoscopy tool tracking not working with MPS after PR 550 note
sohamm17 Nov 14, 2024
2f5ca5b
Update tutorials/cuda_mps/README.md
sohamm17 Nov 14, 2024
053806a
Merge branch 'main' into ss/mps_clarification
sohamm17 Nov 14, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 5 additions & 0 deletions tutorials/cuda_mps/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -73,6 +73,11 @@ MPS, it can also improve the overall GPU utilization.

## Performance Benefits on x86 System

> **Note:** Endoscopy Tool Tracking does not work with CUDA MPS after [holohub/Holoscan-SDK-v2.6.0](https://github.com/nvidia-holoscan/holohub/tree/holoscan-sdk-2.6.0) due to the
> unavailability of CUDA dynamic parallelism implemented in this
> [PR](https://github.com/nvidia-holoscan/holohub/pull/550) in CUDA MPS. In case endoscopy tool tracking needs
> to be tested with CUDA MPS, please use the [`holoscan-sdk-v2.6.0`](https://github.com/nvidia-holoscan/holohub/tree/holoscan-sdk-2.6.0) tag or earlier.

Suppose, we want to run the endoscopy tool tracking and ultrasound segmentation applications
concurrently on an x86 workstation with RTX A6000 GPU. The below table shows the maximum end-to-end latency performance
without and with CUDA MPS, where the active thread percentage is set to 40\% for each application.
Expand Down