CUDA Developer Tools | Performance Analysis with NVIDIA Nsight Systems Timeline

In this episode of the CUDA Developer Tools tutorial series, Eyal Soha, senior software engineer at NVIDIA, introduces code performance analysis using the Timeline View in NVIDIA Nsight Systems. Basic knowledge of C++ CUDA programming is recommended.

Highlights include:

◽ Explore the Nsight Systems timeline, a powerful tool for analyzing GPU performance. Learn how the Timeline View helps you understand code processes and uncover optimization opportunities.

◽ Get an overview of the timeline interface and how to navigate the metrics Nsight Systems collects. Read the timeline, customize your view, and understand CPU and GPU utilization.

◽ Learn how NVTX markers in the code add annotations to the timeline, highlighting essential activities such as memory transfers and kernel executions.

◽ Learn how to use the timeline view to make improvements to your code. Understand concepts like "latency hiding" to identify bottlenecks and make informed optimizations. Explore the benefits of parallelism and uncover how asynchronous operations can impact code performance.

00:00 - Introduction
0:58 - Nsight Systems Timeline
3:19 - Correlating CPU and GPU Activity
3:43 - NVTX Markers
4:17 - CUDA Memcpy
6:40 - Optimized Code and Latency Hiding
11:18 - More Optimized Code

This video series will help get you started with NVIDIA Nsight Developer Tools for CUDA. Grow your proficiency with the tools and apply the examples to your own development environment. Or return to specific episodes for a refresher on certain features and functionalities. We walk through analyzing performance reports, offer debugging tips and tricks, and show you the best ways to optimize your CUDA code. The series will focus primarily on Nsight Compute and Nsight Systems.

Thanks for watching, and stay tuned for more episodes.

#CUDA #Nsight #developertools #NVIDIA #HPC #LLM #CUDAtutorials

[Shurgath]

Thank you for this awesome tutorial series! I would personally enjoy an expert analyzing the performance of a pipeline that calls Cuda kernels from a within Python script.

[yuan.pingchen]

what's the reason to buy nvidia's expensive chip, and write cuda-code to help nvidia sell more expensive chips?

[MrNewAmerican]

Enunciate . Clearly. Dont mumble.

You are presenting to an audience. This is a "formal" interaction. You are not snubbing your mother's attempts to feed you vegetables at a high chair as a toddler.

Speak slowly, clearly.