Learn More
Matching data distribution to workload distribution is important to improve the performance of distributed-memory multiprocessors. While data and workload distribution can be tailored to fit a particular problem to a particular distributed-memory architecture, it is often difficult to do so for various reasons including complexity of address computation,(More)
17 According to the experimental results, we have made the following key observations: First, the minimum communication time occurred for both machines when the message size is approximately 512 to 2K integers (4K bytes). We have compared our experimental results with the analytical results using the LogP and LogGP models and found that they do not seem to(More)
Processing logic programs in parallel is much more complex than for conven tional applications. Indeed, the inference mechanism, particularly the environment stacking and the runtime traversal of a search tree, cannot be easily implemented on distributed memory machines. The USC Research Team has therefore designed a parallel execution model for logic(More)
  • 1