[GPU Computing] AMD OpenCL Zone and OpenCL Programming Guide
GPU compute device
AMD has launched a new tech zone about OpenCL mainly dedicated to developers.
AMD’s OpenCL Zone is available HERE.
AMD has also released a cool thing for all OpenCL developers: the OpenCL Programming Guide v1.0.
In this guide you can find a complete glossary of terms about GPU computing:
Thread: One invocation of a kernel corresponding to a single element in the domain of execution. An instance of execution of a shader program on an ALU. Each thread has its own data; multiple threads can share a single program counter.
GPU compute device: A parallel processor capable of executing multiple threads of a kernel in order to process streams of data.
SIMD Engine: A collection of thread processors, each of which executes the same instruction per cycle.
SIMD pipeline: A hardware block consisting of five stream cores, one stream core instruction decoder and issuer, one stream core constant fetcher, and support logic. All parts of a SIMD pipeline receive the same instruction and operate on different data elements. Also known as “slice.”