LLVM Adds MLIR-Vulkan-Runner To Run MLIR On Vulkan-Enabled GPUs

Added to the LLVM source tree today is mlir-vulkan-runner as a new utility for testing with some interesting possibilities.

For those out of the loop, MLIR is a new intermediate representation (IR) in the LLVM ecosystem that has grown immensely in popularity since Google developers announced it last year. MLIR was designed as a machine learning IR for the likes of TensorFlow and has seen significant adoption by the LLVM ecosystem in working out well for heterogeneous hardware among other advantages over the traditional LLVM IR.

The mlir-vulkan-runner added to the LLVM source tree today is an execution driver for executing MLIR files on Vulkan by translating MLIR modules into SPIR-V for execution on GPUs while the host portion is converted to LLVM IR and JIT'ed on the system. This is similar to the MLIR CUDA runner that has already existed for NVIDIA platforms.

More details within this commit.

The mlir-vulkan-runner is a basic implementation for taking MLIR to GPUs/SPIR-V while Google engineers have already been working on the likes of IREE for taking MLIR to Vulkan-enabled GPUs for machine learning tasks.