Mapping Multi-Loop Nest Algorithms on to Reconfigurable Architecture

Recently, Field Programmable Gate Array technologies have made significant advances in both speed and capacity. Specifically, the design methodology to map the loop dependencies in a do loop algorithm onto a linear array of processors after the extraction of parallelism is a challenging task. The mapping of Full Search Block Motion Estimation (FSBM) and edge detection algorithms are taken up here as they represent multi loop nested algorithms that are intensive in computations. Also, a method of prefixing the elements in the mapping vectors has been used, which reduces the search space for both the algorithms. The formulation of dependence vectors for the FSBM is explained, the architecture for the computationally intensive FSBM and Edge detection algorithm is developed, and the simulation and synthesis results are presented.