How do you find out where those gaps are coming from:
If you refresh timings, the gaps move to different places or vary in length. The Top SOLs are all very low.
In the actions view at the top you sometimes get a long line and sometimes it’s just a gap. But even the ones with a long line don’t always seem to be real (for example, the Hidden/Internal Stencil on the second picture is not showing up on the first picture).
What is causing this and how do you distinguish between actual long draw calls and false positives? I’ve also tried NSight Systems but the frame looked fine there (no gaps). Also, if you save it as C++ file and profile that, most of the gaps disappear - but there are still some.
Thank you for using Nsight Graphics and the Range Profiler feature. I’m sorry you ran into this issue. I will follow up with the engineering team and get back to you. We will let you know if we need any further details.
Thanks for the response, dwoods. Any news from the engineering team?
The nature of the D3D11 replay can introduce these types of gaps. This is one of the reasons we encourage people to look at smaller ranges or sections of the frame when doing profiling. Not only does it allow you to look at workloads that have common setup like shaders used, etc., but it reduces the cross pollution that can occur with the measurements where one part is, say, shader bound and the other is maybe blending bound, and the overall larger range looks like a mix of both.
Thanks Jeff, appreciate the response.