Automatic Code generation for Interconnected distributed RAM in the AAA Methodology: H264 Motion Estimation Case Study
Résumé
New video compression standard achieves high compression rates at the cost of a high encoder complexity. High definition context magnifies the difficulty of a real-time embedded implementation. Programmable multicomponent architectures can provide suitable target solutions combining flexibility and computation power. New architectures are designed to use up to 70 processors in a single equipment to address such an application. The result is that a fast and automatic prototyping methodology cannot be ignored for an optimized application development. This paper presents the methodology to automatically generate a distributed implementation of an application over such a multiprocessor platform from a high level application description. The methodology is adapted to handle interconnected distributed RAM media and illustrated in the context of a H264 coding application. The encoder performances are greatly linked to the motion estimation which can reach 60% of the computation load. A multiprocessor implantation of a motion estimation operation is presented.