HPC++ Run-time Systems

HPC++ can be compiled to run on top of a set of different run-time systems. At HPC++ compilation time, use of proper configuration flags will build HPC++ with the specified run-time system. The run-time system is used by HPC++ for a variety of remote operations such as remote method calls, synchronization between remote contexts, and distributed collective operations. Currently, HPC++ supports three different run-time systems. The first one is based on the Nexus communication library; the second one is based on the Globus toolkit; while the third one is called Tulip which is developed at LANL. All these different run-time are suitable for different purposes. For example, Tulip can be used as a run-time system when running a parallel program on a parallel super-computer with no requirements of security among the communicating contexts (the Nexus based run-time system can also be used when there is no or limited requirement of security). On the other hand, the Globus based run-time system can be used when the computation spans a wide-area network thus requiring authentication and authorization among the communication contexts. Later on, we will also provide either full-blown implementations of the run-time system or bridges to, Corba IIOP and Microsoft DCOM.
  • Run-time System based on Nexus
  • Run-time System based on Globus
  • The Tulip Run-time System

  • hpc++@extreme.indiana.edu

    Last modified: Thu Feb 4 19:32:26 EST 1999