In this workshop we will discuss past and ongoing efforts utilizing GPUs via LLVM. We will discuss the LLVM/libc effort to bring standard library functions onto the GPU; the "new" offload driver in Clang (and Flang) that allows for interoperability between programming languages (CUDA/HIP/OpenMP offload) and features such as device-side LTO; as well as other infrastructure and research projects that touch LLVM and GPUs. Anyone, from newcomers to the topic up to core developers are equally welcome.