A massively parallel RNS architecture

Parallelism on the algorithmic, architectural, and arithmetic levels is exploited in the design of a residue number system (RNS) based architecture. The architecture is based on modulo processors. Each modulo processor is implemented by a two-dimensional systolic array composed of very simple cells. The decoding state is implemented using a two-dimensional array. The decoding bottleneck is eliminated. The whole architecture is pipelined, which leads to a high throughput rate. High speed algorithms for modulo addition, modulo multiplication, and RNS decoding are presented