Adaptive Mesh Refinement for Parallel Processors