Is there a way to define a frame of offscreen work for capture and profile with Nsight?
Context: I’m trying to debug, profile, and optimize our product which uses a number of cameras to do markerless motion capture. We use a combination of DirectCompute and rendering to various render targets to compute our data which we then pull back to the CPU and stream to networked clients. There’s no on-screen rendering.
Is seems like Nsight is using the call to Present to define the frame. Is there a way to insert other calls that bracket the frame somehow and tell Nsight what to capture?