A multiplier with redundant operands

A drawback of using redundant number representation in arithmetic operations for general purpose processors is the need to convert the result to conventional notation before retiring the instruction. A possible solution is to allow forwarding of redundant-form results. To improve the efficiency of forwarding, functional units which can accept operands in redundant form are needed. To this end, a multiplier which accepts a redundant multiplicand and multiplier has been designed, simulated and compared with a conventional multiplier. The redundant multiplicand is converted to radix-2/sup k/ carry/save form before the multiples are formed. Simulations of a 32/spl times/32 bit multiplier show that for instructions which can use redundant form operands there is as much as 20% speedup compared to using conventional form.