OpenCL is an open, royalty-free standard being created by the Khronos Group for programming heterogeneous parallel computing across GPUs and CPUs. OpenCL is being driven by industry-leading companies including AMD, Apple, ARM, Codeplay, Ericsson, Freescale, Imagination Technologies, IBM, Intel, Nokia, NVIDIA, Motorola, RapidMind and Texas Instruments. This informal gathering will provide one of the first opportunities for the HPC community to gain an insight into the architecture and direction of this exciting development. Please register early as seating is limited - we look forward to seeing you in Austin!
Recently Apple came to the Khronos Group with a proposal to establish the Heterogeneous Computing Working Group, based on OpenCL which Apple had been working on with companies like Intel, ATI, and NVIDIA. While OpenCL is a heterogeneous solution targeting CPUs and GPUs alike, the article concentrates on the GPU side. In a nut-shell, if you have an ATI or NVIDIA graphics card installed, chances are your current computer has 100 or more processor cores in it, and currently, this Graphics Processor Unit (GPU) only handles graphics. Similarly, this also applies to multi-core SIMD CPUs, which OpenCL can harness to yield many times the performance of a sequential CPU programming model. OpenCL is intended to be an API that will let developers use the GPU and CPU for the compute-intensive portions of their application, across many platforms, from desktop to mobile. “It was an idea whose time had come,” says Neil Trevett, president of Khronos and chairman of the OpenCL TSG.