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.

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

DOCS/FAQ: add MNNVL section #10285

Open
wants to merge 1 commit into
base: master
Choose a base branch
from

Conversation

Akshay-Venkatesh
Copy link
Contributor

What?

Add FAQ section for MNNVL

Copy link
Contributor

@tvegas1 tvegas1 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

cosmetic review


There are some memory type combinations (particularly involving
cudaMallocManaged memory, and host memory source buffers) where end-to-end
bandwidth is ub-optimal. These stem from existing driver issues or owing to
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

typo sub-optimal?

There are some memory type combinations (particularly involving
cudaMallocManaged memory, and host memory source buffers) where end-to-end
bandwidth is ub-optimal. These stem from existing driver issues or owing to
deficiencies in ipeline protocols. These will be addressed in future releases.
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

typo pipeline

#### Does UCX support use of NVLINK across nodes?

Yes. This feature is not enabled by default and users need to set
UCX_CUDA_IPC_ENABLE_MNNVL=y. In addition, the following criteria have to be met:
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

backquotes around UCX env variable

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants