Parallel Inference Engine

This text describes the machine model designed to support parallel interface, the design of the Kleng language, the design and implementation of the parallel interface engine, the programming tools, the runtime system, and some evaluation results. The architecture of the PIE 64 is tuned specially to support parallel inference. The compiler and runtime systems proposed here are designed to reduce the overhead that inevitably incurrs when using fine granularity processing. A higher degree of parallelism can be expected by using parallel logic programming languages, especially since it is difficult to gain extra parallelism from description in standard languages such as C or Fortran. The contributors to this book believe that the results of their research will be useful for designing a new high-performance processor/computer architecture.