The Data Mover: A Machine-Independent Abstraction for Managing Customized Data Motion

  • Authors:
  • Scott B. Baden;Stephen J. Fink

  • Affiliations:
  • -;-

  • Venue:
  • LCPC '99 Proceedings of the 12th International Workshop on Languages and Compilers for Parallel Computing
  • Year:
  • 1999

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper discusses the Data Mover, an abstraction for expressing machine-independent customized communication algorithms arising in block-structured computations. The Data Mover achieves performance that is competitive with hand-coding in MPI, but enables application-specific optimization to be expressed using intuitive geometric set operations that encapsulate low-level details.