How to test Afternath GPU crash dumping?

I integrated the Aftermath SDK into our UE project. How can I setup a testing environment to trigger a GPU crash?
I want to check on my marker resolving and shader debug info generation.
For now, I have no GPU crashes in the project, but I need to prepare for them.

Hi guanning79,

I am not an expert of UE, to trigger a GPU crash in UE, you can try to add some custom HLSL which call some infinite loop, make sure the custom HLSL is called by some certain condition in your main scene. When your UE project runs, toggle the certain condition to hit the infinite loop which makes GPU hang.


1 Like

This solution works. Thanks a lot.
But I have another question. The output marker stacks aren’t always the correct ones.

The case is like this one below:
My compute pass that forged GPU crash is ForgeGPUCrash (CS), and there are also other passes that are active simultaneously on the GPU.
I got many marker-event-resolve callbacks, and I’ll attach a screenshot of the resolved marker stacks in the next reply(I can only embed one texture in one post).
It’s not always like this; sometimes, I can get the correct marker data containing the marker ForgeGPUCrash (CS). But this destabilization is very confusing.

So, when multiple passes are active on the GPU, how can I indicate the marker data that caused the crash?

The screenshot of the resolved marker stacks: