Hi,
We've been seeing a sporadic freeze in our Vulkan-based app (Quest 2, Quest Pro). The freeze appears as the view becoming mostly unresponsive to user input, and the image only changing every few seconds, with black areas towards the side of the screen when moving the view.
From the ADB logs, it looks like this may be due to a GPU hang. Here's an extract of what we're seeing from the logs:
MISC: GPU hang detectedgpu fault ctx 31 ctx_type VK ts 61149 status 00C01005 rb 1bb5/1bb5 ib1 00000040137B1000/0067 ib2 0000004013980000/0000gpu fault rb 2 rb sw r/w 1bb5/1c62GMU snapshot started at 0x16ad33d86 ticksset FENCE to ALLOW mode:0kgsl-3d0: GPU snapshot froze 144Kb of GPU buffersGPU snapshot created at pa 0x0000000351000000++0x14529cplatform 3d6a000.qcom,gmu: Suspended GMU
We've tried to check if the crash is caused by a particular state in our app, but there's nothing obvious. For example, we've tracked the number of polygons we render and it is about the same when the freeze occurs compared to when the app is working.
Does anyone know if any of the numbers / information in those logs are something we can interpret to help debug this?
Thanks!
Could you send a full logcat of when this is happening? I wonder if there are driver/vulkan impl issues popping up there. Also, have you tried enabling the Vulkan validation layer to see if it is able to spot something weird?