Efficient maximal convex custom instruction enumeration for extensible processors

Abstract

In recent years, the use of extensible processors has been increased. Extensible processors extend the base instruction set of a general-purpose processor with a set of custom instructions. Custom instructions that can be implemented in special hardware units make it possible to improve performance and decrease power consumption in extensible processors. The key issue involved is to generate and select automatically the custom instructions from a high-level application code. However, enumerating all possible custom instructions of a given dataflow graph is a computationally difficult problem. In this paper, we propose an efficient algorithm for the exact enumeration of maximal convex custom instructions. The state of the art algorithms use either a bottom-up manner or a top-down manner to solve the problem. The proposed algorithm enumerates all maximal convex custom instructions by using a sandwich manner that combines the advantage of the bottom-up manner and the top-down manner. Compared to the latest algorithm, our algorithm can achieve orders of magnitude speedup.