feat: Add comprehensive NCCL operations support for CubeCL CUDA runtime #792
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
So took a step back to the grouping abstraction as discussed. Hope thats more the direction :) |
Comment on lines
+310
to
+313
| pub async fn barrier(&self) { | ||
| for (_, device) in &self.map { | ||
| device.client.sync().await; | ||
| } |
There was a problem hiding this comment.
We should await in parallel instead of sequential using join_all or something equivalent.
| let expected_recv_size = send_data.len() * nccl_op.device_count; | ||
| let cu_device = CudaDevice { index: 0 }; | ||
| let client = <CudaRuntime as Runtime>::client(&cu_device); | ||
| let send_handle = create_test_data(&client, send_data.clone()); |
There was a problem hiding this comment.
It's weird to test the send/recv on the same device no?
There was a problem hiding this comment.
Just got one gpu ^^ Getting my Cluster later this week. But had to test somehow.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
Ok, so I made the first NCCL implementation which works now. The design isn't for bigger clusters since it's a single-threaded async approach. For small device counts, this works but not really for HPC - I'll be able to make this more interesting for HPC in a few weeks... But this is a usable approach without grouping, just a simple straightforward approach for now.
Tests work, have a short doc entry with that, and no more memory leaks :)