Introduction to CUDA Programming for Physicists - Lecture 2

Delve into the second installment of a four-part course on CUDA programming tailored for physicists. Explore the architecture of GPUs, including computing units, memory structures, and their interaction with host computers. Learn to identify tasks well-suited for GPU acceleration. Master memory operations, including allocation and data transfer between CPU and GPU, as well as the utilization of GPU shared memory. Gain proficiency in writing CUDA kernels and managing streams for task synchronization. Discover techniques for implementing reduction operations on GPUs and familiarize yourself with essential CUDA libraries such as cuFFT, cuBLAS, cuSPARSE, and cuRAND. This comprehensive 1-hour and 50-minute lecture, delivered by François Gelis at IPhT, provides a solid foundation for leveraging GPU computing power in physics-related numerical computations.