Internal error attempting to replay event

Hi!

I am trying to capture a frame in a program that uses Vulkan raytracing. However, the process crashes at some point during the capturing process and the Output Messages tab in Nsight Graphics shows

1,,Host,NVIDIA Nsight Graphics,Project 'simple' updated from version 1.2 to version 1.3.
2,,Target,NVIDIA Nsight Graphics,Graphics interception initialized
3,,Target,NVIDIA Nsight Graphics,"Adding VK_BUFFER_CREATE_DEVICE_ADDRESS_CAPTURE_REPLAY_BIT to buffer with device address usage. This can be disabled via ""Vulkan > Device Address C++ Support"""
4,,Target,NVIDIA Nsight Graphics,"Adding VK_MEMORY_ALLOCATE_DEVICE_ADDRESS_CAPTURE_REPLAY_BIT to memory allocation with device address support. This can be disabled via ""Vulkan > Device Address C++ Support"""
5,,Target,Nsight Capture Timings,Capture progress checkpoint: Preparing for capture @ 0 s
6,,Target,Nsight Capture Timings,Capture progress checkpoint: Saving initial state @ 0.033 s
7,,Target,Nsight Capture Timings,Capture progress checkpoint: Capturing function stream @ 0.043 s
8,,Target,Nsight Capture Timings,Capture progress checkpoint: Building event database @ 0.525 s
9,,Target,Nsight Capture Timings,Capture progress checkpoint: Processing object mutations @ 0.526 s
10,,Target,Nsight Capture Timings,Capture progress checkpoint: Starting replay @ 0.528 s
11,,Target,NVIDIA Nsight Graphics,Vulkan device lost: Internal error attempting to replay event 117
12,,Target,NVIDIA Nsight Graphics,Failed to replay event 117 due to an internal error (subsequent data will be incorrect)

I am assuming that my program is faulty in a way that makes the frame capture fail. However, the program does not trigger any Vulkan validation layer messages and doesn’t crash when run outside Nsight Graphics. I have yet to spot anything obviously wrong with my code. What can I do to get the crash resolved / work around it?

Gregor

Hello,
Thank you for using Nsight Graphics and I’m sorry you ran into this issue with your application crashing while using Nsight Graphics. Have you seen this section of our Nsight Graphics User Guide that may help to troubleshoot? If you’re still experiencing an issue with your application crashing please capture a Crash Dump by following the steps here and send to myself at dwoods@nvidia.com.

Thanks for the informative reply. I’m sorry for not coming back to you earlier. I’ve upgraded to Nsight Graphics 2022.7, but the crash still persists. Disabling multithreaded capture changes the behavior from a CPU null pointer access inside the capture DLL to a GPU crash (“Failed to translate virtual address” at address 0). The other troubleshooting options don’t seem to have any effect.

I still believe that the root cause is something in my application that I need to fix. But the application is behaving as expected outside Nsight Graphics so far.

I have captured crash dumps with Visual Studio 2022 and Aftermath. I’ll send the files by email.

Ok thanks for the heads up. I’ll keep an eye open for the files you plan to send.