This talk will go over the ongoing effort to provide generic GPU offloading infrastructure in LLVM. We will cover the new API into the offloading runtime as well as the new GPU device-side runtime and language support in LLVM. The talk will demonstrate how to take a generic C/C++ or offloading program and execute it on your AMD or NVIDIA GPU and how SPIR-V will be supported in the future with the `gpuintrin.h` header. We also cover the growing list of GPU runtimes (compiler-rt, libc, libc++, libc++abi, flang-rt, openmp) and how the offloading runtime supports them. The end goal of this talk is to exhibit the full pipeline of creating and executing GPU programs in LLVM as well as the useful features we have.