And-Parallel Execution - Granul. Control

• In practice there are overheads associated with parallel execution:
  • Task creation and scheduling.
  • Possible migration of tasks to remote processors.
  • Communication overheads.

• If the granularity of parallel tasks, i.e. the ``work available'' under them, is too small, it may happen that costs are larger than the benefits in their parallel execution.

• Aim of granularity control: change parallel execution to sequential execution or vice-versa based on conditions related to grain size and overheads.

• Benefits greater for systems with greater parallel execution overheads:
  • Small shared memory multiprocessors: beneficial.

  • Distributed memory multiprocessors, workstation ``farms'', etc: essential.

• Useful user support techniques identified:
  • Visualization / trace analysis: e.g., VisAndOr and IDRA.
  • Program transformation to increase granularity.

• Ambitious objective: construction of an automatic granularity control system: granularity analysis + granularity control (program transformation).