Add nccl parameter to override traffic class for fifo messages #1194
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary: This change adds the support to override the traffic class attribute of the qp on which the cts/fifo messages are sent on. We want to prioritize fifo message to be sent on a network queue different from data traffic. Doing so will remove control path messages from causing slow downs in the case of network congestion when loaded latency is higher. This will only be enabled when NCCL_IB_FIFO_TC variable is set otherwise its disabled by default.