High computation performance through efficient access of data structures

The authors propose the use of an access coprocessor (AP) to reduce the adverse effects of the access overhead on the performance of a VLSI processor. The AP offers the possibility of achieving substantial speedup in program execution time by means of concurrency in the access and execute processes, and by the use of special operations to perform efficiently address calculations. System operation is illustrated by means of a matrix multiplication algorithm. A preliminary analysis of the performance of a VLSI processor, with and without the AP, is presented.<
>