Timing rtTrace via NVAPI

Gaming and Visualization Technologies Visualization OptiX
1

Hi,
I just came across this blog post:
https://developer.nvidia.com/blog/profiling-dxr-shaders-with-timer-instrumentation/
It is showing timing capabilities per pixel in DXR using Nvidias NVAPI.

My question: is this available in general CUDA code too and in Optix especially? This would be a nice debugging/profiling tool. If so, how can it be done? As far as I understood the NVAPI documentation the functions described in the post exist in DXR extensions only so not readily available in CUDA/Optix.

Thanks and kind regards

2

Yes, you can simply use the CUDA clock() instruction for that and scale it accordingly.

Please find more information and links to example code in this post where I mentioned that blog post as well:
https://forums.developer.nvidia.com/t/optix-6-5-wireframe-rendering/119619/10

Actually the OptiX SDKs before 7.0.0 also had that feature inside the pinhole_camera.cu example. Search for TIME_VIEW.

3

Thank you! I assumed there is something similarly easy but could not find it directly.