shared memory

​Join Jai for a hands-on exploration of shared memory in GPU programming. We’ll look at how threads in a CUDA block coordinate, reuse data, and synchronize, then work through how shared memory can improve the performance of parallel computations. We’ll focus on the ideas behind the code, including memory access patterns, synchronization, and the tradeoffs involved in optimizing a GPU kernel. Bring your curiosity and questions, whether you’re following along with the material or looking to deepen your understanding of GPU performance.

​Meet us in SoHo for an informative session and a chance to exchange ideas with fellow attendees.

​Lecture & Exercises: https://ppc.cs.aalto.fi/lecture5/