A parallel algorithm for multiplication of quaternary signed digit (QS
D) numbers is developed in this paper. The number of minterms is signi
ficant ly minimized due to the symmetric encoding involved in the algo
rithm. With a shared content-addressable memory (CAM), the implementat
ion requires a fixed number of minterms for operands of arbitrary size
. An incoherent correlator based optical CAM processor can be used as
the basic unit for QSD multiplication, and preliminary experiment is d
emonstrated. Large scale matrix-vector multiplication call be performe
d in a pipelined iterative architecture using a limited number of mult
ipliers.