Research And Implemention Of Software Parallelization And Low Power Consumption Optimization For Multimedia Applications

In recent years, with the development of multi-core technology, the performance of processor has been greatly proved. And multi-core systems have been widely applied into mobile devices, embedded devices and PC fields. However, many companies lack of a software technology to take advantage of multi-core computing resources in gengerl. The unbalance between software design and hardware resources results in the degration of system performance and the waste of resource. The research of software parallelization for multi-core systems has great value, and receives extensive attention in both academia and industry.To solve the problem that the resources of multicore processors can not be used fully, this paper makes a research on serial software parallelization. First, this paper design a software parallelization framework based on task dependency of the serial software. Then, we take the JPEG image coding standard as an case, and study the parallizaiton of JPEG encoding algorithm. Third, based on a practical project that H.264 video decoding standard parallelization, we analyze and study the H.264 decoding algorithm parallelization. Specifically, this paper carried out the following work:(1) With the rapid improvement of computing power in multicore era, the system performance has been significantly improved. However, the software technology has not kept pace with hardware technology. In this paper, we present a carefully-defined software parallelization framework by studying the multi-core processor. There are totally four steps to achieve parallelization of serial software, to make full use of the multicore processor resources.(2) Considering that the JPEG encoding algorithm is one of the most popular coding standards in industries, the image encoder based on JPEG coding algorithm is widely used in PC and mobile application. In this paper, we study the JPEG image coding standard algorithm, and design a parallelization scheme of the software based on JPEG. And we realize the goal that software makes full use of multi-core resources.(3) H.264 video decoding standard is currently one of the most popular video codec standard. The multimedia applications based on H.264 are favored because of its high compression ratio and its high-performance. In this paper, we make a carefully study on H.264 video decoding standard, and design a parallelization scheme of H.264 video decoding algorithm. And we optimize the decoding performance and energy consumption of software based on H.264 standard.Finally, the paper test the real cases of two parallelization scheme in experiments. Specifically, we apply OpenMP, the parallel programming model, to parallelized serial code and introduce DVFS technology to further reduce energy consumption. The experimental results show that the above technology can effectively improve the performance of system. We realize the speedup of the system increased by 30%, and reduce energy consumption by at least 25%. The research of this paper make a great contribution in the field of mobile devices, embedded devices, data centers and servers.