https://bugs.kde.org/show_bug.cgi?id=521727

--- Comment #8 from sunghwan <[email protected]> ---
I've found a simpler fix. The approach focuses on fixing the CPU-GPU race
condition triggered during the MultiGpuSwapchain::deleteResources call.
MultiGpuSwapchain::deleteResources destroys Vulkan swapchain memory allocations
on the CPU without waiting for the GPU to finish, causing the NVIDIA driver to
unmap the page tables while the Copy Engine is still actively writing to those
addresses, triggering write page faults (Xid 31).

This patch (https://pastebin.com/4Yv9F0fj) adds a crucial Vulkan
synchronization call (waitIdle()) inside MultiGpuSwapchain::deleteResources()

>From my tests, I haven't encountered any system freezes or performance issues.
I hope this helps in resolving the root cause.

-- 
You are receiving this mail because:
You are watching all bug changes.

Reply via email to