pojeda / juliacon21-gpu_workshop

Material for the 2021 GPU workshop at JuliaCon

JuliaCon 2021 - GPU Workshop material

This repository contains the notebooks and other material for the GPU workshop at JuliaCon 2021.

• deep_dive: a notebook that explains the different GPU programming models, array programming and kernel programming, and demonstrates what they can and cannot be used for using a series of examples. Depending on the exact back-end, different tools are shown to facilitate GPU development.
• case_studies: more hands-on demonstrations of the GPU programming functionality to implement specific applications and algorithms.
• kernelabstractions: an in-depth demonstration of KernelAbstractions.jl, an alternative way to program GPUs in Julia using a vendor-agnostic kernel programming abstraction
• enzyme: a sneak peek at the new LLVM-based autodifferentiation support, demonstrated on a parallel GPU kernel.