Platform Improvement [UPM]

• Achievements:
  • Devised improved memory management methods:
    [UPM, in collaboration with Manchester]
    • Detected high memory consumption when using non-restricted scheduling in and-parallel programs
    • Proposed implementation techniques to reduce it
    • Assessment showed greatly reduced memory usage

  • Studied implementing issues for nondeterministic or+and parallel systems: [UPM, in collaboration with NMSU]
    • Proposed a practical execution scheme (ACE)
    • Studied overhead imposed by parallel execution in terms of execution time and memory consumption
    • Devised techniques to ameliorate this overhead by compile time and run time identification of special cases

  • Evaluated application of data-parallel techniques in and-parallelism: [UPM]
    • Better memory usage through task collapsing
    • Faster task startup by splitting goal set-up among agents
    • Achieved improved speedups, comparable with native data-parallel only systems