vulkan: For coopmat2 FA, use fp16 accumulators for the final result#19376
Merged
0cc4m merged 1 commit intoggml-org:masterfrom Feb 6, 2026
Merged
vulkan: For coopmat2 FA, use fp16 accumulators for the final result#193760cc4m merged 1 commit intoggml-org:masterfrom
0cc4m merged 1 commit intoggml-org:masterfrom
Conversation
The cpu and cuda backends use fp16 for the VKQ accumulator type, this change does the same for vulkan. This helps particularly with large head sizes which are very register-limited. I tried this for the coopmat1 path and it slowed down a bit. I didn't try for scalar. I applied the softmax bias that the cuda backend uses to avoid overflow, although I was not able to reproduce the original bug without it.
0cc4m
approved these changes
Feb 6, 2026
liparetejas
pushed a commit
to liparetejas/llama.cpp
that referenced
this pull request
Feb 23, 2026
…gml-org#19376) The cpu and cuda backends use fp16 for the VKQ accumulator type, this change does the same for vulkan. This helps particularly with large head sizes which are very register-limited. I tried this for the coopmat1 path and it slowed down a bit. I didn't try for scalar. I applied the softmax bias that the cuda backend uses to avoid overflow, although I was not able to reproduce the original bug without it.
bartowski1182
pushed a commit
to bartowski1182/llama.cpp
that referenced
this pull request
Mar 2, 2026
…gml-org#19376) The cpu and cuda backends use fp16 for the VKQ accumulator type, this change does the same for vulkan. This helps particularly with large head sizes which are very register-limited. I tried this for the coopmat1 path and it slowed down a bit. I didn't try for scalar. I applied the softmax bias that the cuda backend uses to avoid overflow, although I was not able to reproduce the original bug without it.
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.
The cpu and cuda backends use fp16 for the VKQ accumulator type, this change does the same for vulkan. This helps particularly with large head sizes which are very register-limited.
I tried this for the coopmat1 path and it slowed down a bit. I didn't try for scalar.
I applied the softmax bias that the cuda backend uses to avoid overflow, although I was not able to reproduce the original bug without it.
These perf results were with the change stacked on #19309. Benefits are mostly in GLM-4.7-Flash (head size 576/512) and Qwen3Next (head size 256):