NERSC will be hosting a 3-part event series that focuses on using learning introductory GPU Programming concepts with CUDA on Perlmutter at NERSC. In this training series, users will be introduced to the basics of using CUDA on Perlmutter at NERSC. The training will focus on the basics of the Perlmutter architecture and NVIDIA GPUs, programming concepts with CUDA using C/C++. This training series will cover technical training and interactive examples.
Part 3 focuses on understanding CUDA memory types, synchronization techniques, and techniques for tiled algorithms.
This training is meant for all novice users in HPC within the DOE lab environment. No previous experience with CUDA is required to attend this training. Part 1 focused on an introduction to CUDA with C/C++. Part 2 focuses on understanding advanced CUDA kernels. Users are encouraged to watch the recordings from the Part 1 and Part 2 before Part 3.
Users are required to register for this event. Non-NERSC users can register up until October 09, 2025: CUDA C/C++ Part 3 Registration