Fast Implementation of 2D DCT Based on Media Instructions
Wei Li · Jisuanji gongcheng · 2007
2D DCT transform is frequently used in video compression protocol H.264 and MPEG family standards. Conventional operations in general-purpose processors can not meet the demand of the real-time CODEC. A fast implementation of 2D DCT transform is presented and the data structure is designed for the demand. It can make full use of the operations of media instructions, and avoid the data hazard. The performance results show that it can meet the real-time requirement and can be served as a reference for compiler development on media instructions.