Efficient Multithreading Implementation of H.264 Encoder on Intel Hyper-Threading Architectures

Exploiting thread-level parallelism is a promising way to improve the performance of multimedia applications running on multithreading general-purpose processors. This paper describes our work in developing the first multithreading implementation of the H.264 encoder. We parallelize the encoder using the OpenMP programming model, which allows us to leverage the advanced compiler technology in the Intel C++ compiler for Intel Hyper-Threading architectures. We present our design considerations in the parallelization process. We describe an efficient multi-level data partitioning scheme that increases performance of a multithreaded H.264 encoder. Our experiments show parallel speedups ranging from 4.31x to 4.69x on a 4-CPU Intel XeonTM system with Hyper-Threading Technology.