A parallel divider for architecture numbers of arbitrary word length

<jats:title>Abstract</jats:title><jats:p>The high‐speed and highly accurate arithmetic unit is becoming more important. Especially, one factor that determines the system performance is the execution of the multiplication/division, which arises frequently and takes a long computation time. The most appropriate idea in improving the speed is the parallel hardware architecture.</jats:p><jats:p>Often in the system design, the capacity of the initially designed arithmetic unit becomes insufficient due to the change of specification and the upgrading of the system function after the design making the circuit redesign necessary. Especially, in the case of the divider, such a system redesign is not easy since it cannot be used if the number exceeds the word length. If the number exceeds the operational capacity even by a bit in the divider, it is difficult to cope with the requirement, even if a memory or an external circuit is added. Hence, there is less advantage in fabricating a dedicated integrated circuit (IC) and there are only a few examples of the one‐chip divider.</jats:p><jats:p>The high‐speed dedicated arithmetic unit includes digital signal processing (DSP) circuitry and a DSP coprocessor, but the division is executed by the software since it is not advantageous to include the divider. This requires a long computation time.</jats:p><jats:p>From such a viewpoint, this paper proposes an architecture for the parallel divider which can cope with any requirement for the accuracy only by employing several similar dividers. Detailed problems in the expansion of the operational accuracy are discussed, and the proposed divider architecture is described. The one‐chip implementation also is discussed.</jats:p><jats:p>By this structure, an advantage is produced to prepare the dedicated divider, and the cheap, efficient and high‐speed division can be realized by a relatively small‐scale divider. The proposed divider will be applied widely, not only to the general scientific computations but also to various control systems, signal processing, and computer communication.</jats:p>