Introduction to GPU Computing
Introduction to GPU Computing
What You'll Learn
- Understand GPU architecture and parallel computing principles
- Master CUDA programming for NVIDIA GPUs
- Write OpenCL code for cross-platform GPU computing
- Optimize memory access patterns and thread divergence
- Implement parallel algorithms: reduction, scan, and matrix operations
- Use GPU libraries: cuBLAS, cuFFT, Thrust, and cuDNN
- Profile and debug GPU applications with NSight
- Complete a GPU-accelerated scientific application capstone
Course Content
Course Preview Video
Course Requirements
Strong C/C++ programming skills required. Basic understanding of computer architecture and parallel computing concepts is helpful. GPU access is provided via cloud labs.
- Level: Intermediate
- Delivery: Online / Self-paced
- Prerequisites: Strong C/C++ programming skills
- Certificate: Certificate of Completion upon successful completion
Frequently Asked Questions
Do I need a GPU?
No personal GPU is required. We provide free cloud GPU labs through Google Colab and other providers for all hands-on exercises and the capstone project.
Do I need C/C++ experience?
Yes, strong C/C++ skills are required. CUDA is a C/C++ extension, so you should be comfortable with pointers, memory management, and compilation.
What will I learn about GPU libraries?
The course covers cuBLAS, cuFFT, Thrust, cuDNN, and OpenCL, giving you a complete toolkit for GPU-accelerated scientific and AI applications.
Will I optimize real applications?
Yes, you'll learn profiling with NSight, identify bottlenecks, and apply optimization techniques to achieve significant speedups in real-world scenarios.
How long is the course access?
You receive lifetime access to all course materials, including future updates on new GPU architectures and CUDA versions.
Will I receive a certificate?
Yes, you will receive a Certificate of Completion after successfully completing the capstone project and all assessments.
Course Tutor
Dr. Kevin Liu
Dr. Kevin Liu is a Principal GPU Computing Engineer and CUDA Architect with over 19 years of experience developing GPU-accelerated applications. He has led GPU computing teams at NVIDIA, AMD, and Microsoft, contributing to the development of CUDA and DirectX technologies. Dr. Liu holds a PhD in Computer Science from Stanford University where his research focused on parallel algorithms and GPU architecture. He has published over 70 papers on GPU computing and is a co-author of the official CUDA programming guide. His work has accelerated scientific simulations, deep learning training, and real-time graphics across multiple industries. Dr. Liu is a frequent speaker at GTC, SC, and SIGGRAPH conferences.
Throughout this course, Kevin will guide you through hands-on CUDA programming, performance optimization, and the exact techniques used by professional GPU developers to achieve maximum computational performance.
Course Reviews
Dr. Liu's NVIDIA experience is invaluable. The CUDA programming modules are exceptional. My molecular dynamics simulations now run 100x faster on GPU. Best GPU course available!
The memory optimization module transformed how I think about GPU performance. Dr. Liu explains coalescing and bank conflicts brilliantly. The capstone was challenging and rewarding.
As a ML engineer, this course helped me understand GPU hardware deeply. The cuDNN and profiling sections are gold. I now optimize deep learning models like a pro.
The parallel algorithms module is outstanding. I learned reduction, scan, and matrix multiplication from scratch. The OpenCL section is also great for cross-platform work.
PhD-level content made practical. The NSight profiling module saved me weeks of optimization work. Now I write GPU code that achieves near-peak performance. Exceptional course!
The capstone GPU image processing project was challenging and educational. Dr. Liu's insights into warp-level programming and memory access are invaluable. Highly recommended!