Implementation of Sequential Loops in Data

The implementation of sequential loops in dataaow computation had traditionally not received very much attention as it was assumed that most loops would be executed in parallel. This assumption was valid for earlier dataaow machines such as the MIT Tagged Token Dataaow Architecture (TTDA))2], Sigma-119] but not for the newest generation of dataaow machines including Monsoonn6], EM-4411] and Epsilon-227]. On the latter machines, sequential loops use less memory, and can execute in fewer instructions, albeit with lower parallelism then the parallel versions. This char-acterisation of sequential and parallel loops suggests that programs should have parallel outer loops and sequential inner loops. The run time of sequential loops therefore become signiicant in the overall run time. We also found that previous implementations of sequential loops can incur fairly high overheads. In this paper, we present two new ways of implementing sequential loops that have lower overhead then previous methods. We studied this problem in the context of compiling Idd14, 15] for Monsoon. 1 Introduction Loops executing on dataaow machines can be classiied into three categories according to the amount of inter-iteration parallelism allowed. At one extreme is an unbounded loop, where there is no artiicial bound on the number of concurrent iterations. Such a loop is compiled into a recursive procedure which together with non-strict procedure call convention lead to uncontrolled unraveling of the loop. Next, we have what is often referred to as k-bounded loopss5], where the number of iterations executing concurrently is bounded