Overview

Instructor(s): Jian Tao

Time: Tuesday, April 22, 2025 — 1:30PM-4:00PM CT

Location: Online using Zoom

Prerequisite(s): Current ACCESS ID, Linux/Unix skills, experience with C or C++

This short course covers basic topics in CUDA programming on NVIDIA GPUs. Topics include CUDA architecture; basic language usage of CUDA C/C++; and writing and executing CUDA code.

Course Materials

• Introduction to CUDA slides (Spring 2025): PDF
• Exercises: tarball

Previous Course Materials

• Introduction to CUDA slides (Fall 2024): PDF
• Introduction to CUDA slides (Spring 2024): PDF
• Introduction to CUDA slides (Fall 2023): PDF
• Introduction to CUDA slides (Spring 2023): PDF
• Introduction to CUDA exercises (Spring 2023): TGZ

Learning Objectives and Agenda

Learning objectives:

• Understand the Basics of CUDA: Learn the fundamentals of CUDA, including its architecture, and how it exposes GPU parallelism for general-purpose computing.
• Familiarize with CUDA Programming: Understand how to write CUDA programs, including how to manage devices, memory, and execute kernels on the GPU
• Learn Parallel Programming Concepts: Grasp the concepts of parallel programming in CUDA, including how to launch kernels, manage threads, and optimize performance.

Agenda:

• GPU as an Accelerator
• Running CUDA Code on ACES
• CUDA C/C++ Basics
• Unified Memory Programming
• Device Management

Note: During the class sessions many aspects of the material will be demonstrated live on the ACES cluster. Attendees will also log in to ACES to complete exercises themselves.