Scalable Multidimensional DFT using Processing In Memory

This paper presents a fully parallel and scalable implementation of the multidimensional discrete Fourier transform (M-D DFT). The underlying idea is to utilize the processing-in-memory (PIM) approach, leveraging memristors to perform M-D DFT computations in a fully parallel manner, thereby enabling ultra-high throughput and low latency for M-D DFT operations. Moreover, the proposed scheme supports arbitrary transform sizes, not limited to powers of 2, and it is independent of the DFT radix. One significant advantage of the proposed scheme is its substantial reduction in computational complexity compared to traditional M-D DFT implementations.