Sequentialization / Granularity Control [UPM]

• State of the art prior to project:
  • Basic, cost function based granularity control approach proposed.
• Achievements:
  • Proposed conditions for controlling parallelism based on both upper and lower bounds on task cost.
  • Proposed transformations for dynamic term size computation and reduction of run-time granularity control overhead.
  • Extended techniques to OR-Parallelism.
  • Developed lower bound cost analysis (including non-failure analysis) and a specialized analysis for divide-and-conquer programs.
  • Achieved first (prototype) completely automatic granularity control system by integrating CASLOG (upper bound analysis) with automatic inference of mode information, size measure, and types.
  • Tested and assessed the approach and techniques (with ECLPS, MUSE and Ciao-Prolog).