Multi-Valued Network Simplification using Redundancy Removal

We introduce a scheme to simplify a multi-valued network using redundancy removal techniques. Recent methods [1], [2] for binary redundancy removal avoid the use of state traversal. Additionally, [2] finds multiple compatible redundancies simultaneously. We extend these powerful advances in the field of binary redundancy removal to perform redundancy removal for multi-valued networks. First we perform a one-hot encoding of all the multi-valued variables of the design. Multi-valued tables are written out as binary tables, using this one-hot encoding. At the end of this step, we have a binary network which is equivalent to the multi-valued network modulo encoding. Next, binary redundancy removal is invoked on the resulting network. In case a binary signal si is determined to be stuck-at-0 redundant, this means that the multi-valued signal s can never take on a value i. Further, if the binary signal si is determined to be stuck-at-1 redundant, this means that the multi-valued signal s takes on a constant value i. All redundant binary signals are recorded in a file. The original multi-valued network is modified based on the binary redundancies thus computed. Initial experiments using this technique show a 10-20% reduction in the size of the multi-valued description.