Skip to content

vulkan: workaround NV queuesubmit driver bug - #28830

Merged
ggerganov merged 1 commit into
ggml-org:masterfrom
jeffbolznv:queuesubmit_bug
Sep 13, 2026
Merged

ggerganov merged 1 commit into
ggml-org:masterfrom
jeffbolznv:queuesubmit_bug

Conversation

@jeffbolznv

Copy link
Copy Markdown
Contributor

Overview

There is a driver bug where two queues on the same VkDevice simultaneously submitting can break some internal synchronization. Until it's fixed, add a mutex around queuesubmit.

This fixes the issue with -j reported at #28740 (comment).

Requirements

There is a driver bug where two queues on the same VkDevice simultaneously
submitting can break some internal synchronization. Until it's fixed, add a
mutex around queuesubmit.
@jeffbolznv
jeffbolznv requested a review from a team as a code owner September 13, 2026 02:22
@github-actions github-actions Bot added Vulkan Issues specific to the Vulkan backend ggml changes relating to the ggml tensor library for machine learning labels Sep 13, 2026
@ggerganov

Copy link
Copy Markdown
Member

I'll fast-track this to try to stabilize the CI.

@ggerganov
ggerganov merged commit f1e44dc into ggml-org:master Sep 13, 2026
22 of 30 checks passed
@BrewTestBot BrewTestBot mentioned this pull request Sep 14, 2026
1 task done
pl752 pushed a commit to pl752/llama.cpp that referenced this pull request Sep 15, 2026
There is a driver bug where two queues on the same VkDevice simultaneously
submitting can break some internal synchronization. Until it's fixed, add a
mutex around queuesubmit.
quimmedes pushed a commit to quimmedes/cafe-llama.cpp that referenced this pull request Sep 16, 2026
There is a driver bug where two queues on the same VkDevice simultaneously
submitting can break some internal synchronization. Until it's fixed, add a
mutex around queuesubmit.
zsogitbe pushed a commit to zsogitbe/llama.cpp that referenced this pull request Sep 17, 2026
There is a driver bug where two queues on the same VkDevice simultaneously
submitting can break some internal synchronization. Until it's fixed, add a
mutex around queuesubmit.
frostyautumnleaf pushed a commit to frostyautumnleaf/llama.cpp that referenced this pull request Oct 5, 2026
There is a driver bug where two queues on the same VkDevice simultaneously
submitting can break some internal synchronization. Until it's fixed, add a
mutex around queuesubmit.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ggml changes relating to the ggml tensor library for machine learning Vulkan Issues specific to the Vulkan backend

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants