MINIMIZATION OF MEMORY ACCESS OVERHEAD FOR MULTIDIMENSIONAL DSP APPLICATIONS VIA MULTILEVEL PARTITIONING AND SCHEDULING

Massive uniform nested loops are broadly used in multidimensional digi tal signal processing (DSP) applications, Due Co the large amount of d ata handled by such applications, the optimization of data accesses by fully utilizing the local memory and minimizing communication overhea d is important in order to improve the overall system performance, Mos t of the traditional partition strategies do not consider the effect o f data access on the computational performance, In this paper, a multi level partitioning method, based on a static data scheduling technique known as carrot-hole data scheduling, is proposed to control the data traffic between different levels of memory, Based on this data schedu le, optimal partition vector, scheduling vector and the partition size are chosen in such a way to minimize communication overhead, Nonhomog eneous size partitions are the final result of the partition scheme wh ich produces a significant performance improvement. Experiments show t hat by using this technique, local memory misses are significantly red uced as compared to results obtained from traditional methods, This me thod can be used in application specific DSP system design and compile r for DSP processors.