Abstract
This chapter deals with the microarchitecture design of MAC unit. MAC as hardware module stands for multiplication and accumulation unit. A MAC unit is probably the most important hardware module in any datapath of DSP processors. A MAC unit is a hardware module consisting of at least a multiplier and an accumulator. The multiplier in a MAC unit is normally a signed two complement general multiplier. The accumulator in a MAC unit is a full adder supporting two's complement arithmetic computing with double-precision. The accumulator register keeps the result from the accumulator supporting iterative computing. By specifying pre- and postoperations around the multiplier and the accumulator, a MAC unit can support general arithmetic and logic functions with high precision. The design of the MAC module is essential in DSP hardware design. It supports algorithms including filtering, auto/cross correlations, transforms, and double-precision arithmetic operations. The acronym MAC has two meanings: an instruction or a hardware module. MAC instruction stands for multiplication and accumulation operation. When executing a convolution algorithm in a general-purpose processor, only one MAC unit is usually available, and one step of the convolution can be executed in each clock cycle assuming that the memory subsystem can supply the right data and the right coefficients each clock cycle.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call