Compiled execution of the reduce-or process model on multiprocessors
Balkrishna Ramkumar, Laxmikant V. Kalé · 1989
In this paper, the authors describe the abstract machine developed for the reduce - or process model (ROPM) and its implementation on a variety of multiprocessors. In keeping with the objective behind the ROPM, the abstract machine is suitable for execution on both shared and nonshared memory machines. It uses structure sharing unlike most of the abstract machines based on the WAM. This is due to significant benefits in a nonshared memory context. It has currently been implemented on the Encore Multimax, the Sequent Symmetry, the Alliant, and the InteliPSC/2. The authors provide preliminary performance data of our implementations on these machines in this paper. The benchmarks chosen illustrate the range of programs which ROPM can parallelize - and, or, as well as and/or parallel programs are effectively parallelized and speeded up.