Efficient FFTs on iram

is notoriously difficult on conventional general-purpose architectures because FFTs require high memory band-width and strided memory accesses. Since FFTs are im-portant in signal processing, several DSPs have hardware support for performing FFTs; moreover, some DSPs are designed solely for the purpose of computing FFTs and related transforms. In this paper, we show that the per-formance of the general-purpose VIRAM processor in computing floating-point FFTs is competitive with that of existing floating point DSPs and specialized fixed-point FFT chips. Furthermore, VIRAM is a complete “sys-tem on a chip, ” and therefore has power, area, and cost advantages over multi-chip systems based on DSPs. VI-RAM’s advantages result from a combination of hard-ware technology, instruction set design, and algorithmic techniques. In particular our FFT results are achieved by: an integrated processor-in-memory design that provides high memory bandwidth; vector processing, which pro-vides simple and efficient utilization of the high memory bandwidth; a small amount of ISA support for in-register permutations; and an algorithm tuned to the VIRAM in-struction set and implementation.