Multiprocessor applications are gaining attention as a solution to the problems of increased consumption power due to increasing integration of semiconductors, the slowing of increase in processor operational clock frequency and the increased development time for hardware and software. In information consumer electronics, it is extremely important to increase the speed of multimedia processing such as codec.

By using parallelization on a macroblock level in the MPEG2 encoding process, and by using parallelization on a slice level as well as macroblock parallelization in slice processing, it is possible to decrease power consumption as well as improve speed.

We have proposed a coarse grain task parallel processing method for MPEG2 encoding and decoding, which are used to process video footage, as an example of a multimedia application. We hope to examine its application in the multimedia standards that will develop in the near future.

This parallelization method is effective even on SMP servers with multicore processors, and it will be possible to achieve higher performance and lower power consumption, through parallelization methods for local memory optimization and data transfer optimization in OSCAR multicore processors.

Sponsord research, Collaboration research, Technical consultation

This research seed is the fruit of joint work of Prof. Hironori Kasahara.