Roofline Analysis on Jetson system

Hello, recently I’ve been researching heterogeneous computing on Jetson AGX Orin, and I can use NCU to profile and compute the GPU of the Roofline.

However, I am wondering if there is an analysis method that can perform a Roofline-like analysis for both the CPU and GPU resources in Jetson, or if there is another method to show the utilization of both CPU and GPU.

(For instance, if my application sequentially switches between using the CPU and GPU or only uses the CPU, what metrics can be used to quantify this underutilization of computational resources?)

Hi,
Please try Nsight:

Jetson Software Architecture — NVIDIA Jetson Linux Developer Guide 1 documentation

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.