Repository navigation
Conversation
github-actions
Bot
added
testing
Contributor
|
I didn't have time to dig into the code today. I ran test-backend-ops and the perf is improved. codex suggested using V=32, and I get another almost 2x speedup from that. |
This branch has not been deployed
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Overview
This is a Vulkan port of the CUDA chunked kernel in #29353 to NV_coopmat2 and KHR_coopmat_maintenance1. The perf delta is much smaller than on CUDA, because I am missing some tuning options that CUDA has and because the Vulkan scan kernel is significantly better than CUDA's. This may not be worth the complexity, but I want to open it as draft to get some feedback from @jeffbolznv. It is a significant op-level speedup, but not on the model-level yet.
On the AMD side it requires radv built from upstream for maintenance1 support, since no released version supports it yet.
Requirements