Skip to main content
OSTI.GOVU.S. Department of Energy Office of Scientific and Technical Information
U.S. Department of Energy
Office of Scientific and Technical Information
 

TEMPI: An Interposed MPI Library with Canonical Representation of MPI Datatypes [Poster]

Conference ·
DOI:https://doi.org/10.2172/1873267· OSTI ID:1873267
 [1];  [2];  [3];  [3];  [4]
  1. Sandia National Lab. (SNL-CA), Livermore, CA (United States)
  2. Univ. of Illinois at Urbana-Champaign, IL (United States)
  3. IBM T. J. Watson Research, Ossining, NY (United States)
  4. NVIDIA Corporation, Santa Clara, CA (United States)
+ Show Author Affiliations
TEMPI provides a transparent non-contiguous data-handling layer compatible with various MPIs. MPI Datatypes are a powerful abstraction for allowing an MPI implementation to operate on non-contiguous data. CUDA-aware MPI implementations must also manage transfer of such data between the host system and GPU. The non-unique and recursive nature of MPI datatypes mean that providing fast GPU handling is a challenge. The same noncontiguous pattern may be described in a variety of ways, all of which should be treated equivalently by an implementation. This work introduces a novel technique to do this for strided datatypes. Methods for transferring non-contiguous data between the CPU and GPU depends on the properties of the data layout. This work shows that a simple performance model can accurately select the fastest method. Unfortunately, the combination of MPI software and system hardware available may not provide sufficient performance. The contributions of this work are deployed on OLCF Summit through an interposer library which does not require privileged access to the system to use
Research Organization:
Sandia National Laboratories (SNL-NM), Albuquerque, NM (United States); Oak Ridge National Laboratory (ORNL), Oak Ridge, TN (United States). Oak Ridge Leadership Computing Facility (OLCF)
Sponsoring Organization:
USDOE National Nuclear Security Administration (NNSA); IBM AI Horizon Network, IBM-ILLINOIS Center for Cognitive Computing Systems Research (C3SR); USDOE Office of Science (SC); National Science Foundation (NSF)
DOE Contract Number:
NA0003525; AC05-00OR22725
OSTI ID:
1873267
Report Number(s):
SAND2021-7110C; 696876
Country of Publication:
United States
Language:
English

Similar Records

TEMPI: An Interposed MPI Library with Canonical Representation of MPI Datatypes [Slides]
Conference · Thu Jun 24 00:00:00 EDT 2021 · OSTI ID:1872311
Runtime checking of datatype signatures in MPI
Conference · Mon Jun 19 00:00:00 EDT 2000 · OSTI ID:757553
High-order finite-element seismic wave propagation modeling with MPI on a large GPU cluster
Journal Article · Fri Oct 01 00:00:00 EDT 2010 · Journal of Computational Physics · OSTI ID:21418106

Related Subjects

97 MATHEMATICS AND COMPUTING