Performance Analysis with NVIDIA Nsight Systems Timeline | CUDA Developer Tools

NVIDIA Developer October 27, 2023
Video Thumbnail
NVIDIA Developer Logo

NVIDIA Developer

View Channel

About

Welcome to the NVIDIA Developer YouTube Channel Subscribe to this channel for easy-to-follow “how-to” videos to learn about the latest technologies for developers from NVIDIA. Whether you’re a student, professional developer, or tech enthusiast, discover: 🧑‍💻 CUDA Programming: Parallel computing, debugging, and performance tips ✨ Agentic & Generative AI: Build intelligent agents and generative apps with AgentIQ, NeMo, and open-source tools 🤖 Robotics: Unlock smart automation and robotics solutions 📊 Data Science & Analytics: Accelerate data workflows with GPU-powered libraries like RAPIDS and popular tools 🛠️ And More: Deep learning, computer vision, simulation, high-performance computing, SDK tutorials, and expert guides Join a vibrant developer community, stay ahead with emerging tech, get real-world examples, and tips from NVIDIA engineers. Subscribe and start creating, optimizing, and deploying innovations with NVIDIA. 🙌

Video Description

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. Check out the blog post the code being analyzed in this tutorial is based on: https://developer.nvidia.com/blog/how-overlap-data-transfers-cuda-cc/ 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. CUDA Developer Tools | NVIDIA Nsight Tools Ecosystem: https://youtu.be/xdFQZSV5IrU CUDA Developer Tools | Intro to NVIDIA Nsight Systems: https://youtu.be/dUDGO66IadU CUDA Developer Tools | Intro to NVIDIA Nsight Compute: https://youtu.be/Iuy_RAvguBM Thanks for watching, and stay tuned for more episodes. Learn more about CUDA Developer Tools: https://developer.nvidia.com/tools-overview Get started with NVIDIA Nsight Systems: https://developer.nvidia.com/nsight-systems Get started with NVIDIA Nsight Compute: https://developer.nvidia.com/nsight-compute Join the NVIDIA Developer Program: https://nvda.ws/3OhiXfl Read and subscribe to the NVIDIA Technical Blog: https://nvda.ws/3XHae9F #CUDA #Nsight #developertools #NVIDIA #HPC #LLM #CUDAtutorials

You May Also Like