CUDA: Scalable parallel programming for high-performance scientific computing

Abstract

Graphics processing units (GPUs) originally designed for computer video cards have emerged as the most powerful chip in a high-performance workstation. Unlike multicore CPU architectures, which currently ship with two or four cores, GPU architectures are "manycore" with hundreds of cores capable of running thousands of threads in parallel. NVIDIA's CUDA is a co-evolved hardware-software architecture that enables high-performance computing developers to harness the tremendous computational power and memory bandwidth of the GPU in a familiar programming environment - the C programming language. We describe the CUDA programming model and motivate its use in the biomedical imaging community.