Reduced-ordered binary decision diagrams (BDDs) are a data structure for representation and manipulation of Boolean functions. The variable ordering largely influences the size of the BDD, varying from linear to exponential. In this paper, the authors study the BDD minimization problem based on scatter search optimization. Scatter search offers a reasonable compromise between quality (BDD reduction) and time. On smaller benchmarks it delivers almost optimal BDD size with less time than the exact algorithm. For larger benchmarks it delivers smaller BDD sizes than genetic algorithm or simulated annealing at the expense of longer runtime.

We present an architecture that features dynamic multithreading execution of a single program. Threads are created automatically by hardware at procedure and loop boundaries and executed speculatively on a simultaneous multithreading pipeline. Data prediction is used to alleviate dependency constraints and enable lookahead execution of the threads. A two-level hierarchy significantly enlarges the instruction window. Efficient selective recovery from the second level instruction window takes place after a mispredicted input to a thread is corrected. The second level is slower to access but has the advantage of large storage capacity. We show several advantages of this architecture: (1) it minimizes the impact of ICache misses and branch mispredictions by fetching and dispatching instructions out-of-order, (2) it uses a novel value prediction and recovery mechanism to reduce artificial data dependencies created by the use of a stack to manage run-time storage, and (3) it improves the execution throughput of a superscalar by 15% without increasing the execution resources or cache bandwidth, and by 30% with one additional ICache fetch port. The speedup was measured on the integer SPEC95 benchmarks, without any compiler support, using a detailed performance simulator.

The results presented show that the addition of a secondary network with a wide-tree topology and one or more coordination processors (COPs) to a workstation cluster multicomputer can speed up synchronization, broadcasting, reduction, and several other important global operations by two or three orders of magnitude. The wide tree topology contributes to this by allowing a large number of nodes to simultaneously send commands to a COP and allowing a COP to broadcast a computed value to a large number of nodes simultaneously. Hardware based protection mechanisms in the COP system permit direct user access to the COP network and thus eliminate the overhead of software protocol layers for these global operations. Although the COP system provides a large performance gain for global operations, it increases overall system cost by only 2-3%.