Exploiting multilevel parallelism using processor groups is becoming increasingly important for programming on high-end systems. This paper describes a group-aware run-time support for shared-/global- address space programming models. The current effort has been undertaken in the context of the Aggregate Remote Memory Copy Interface (ARMCI) [5], a portable runtime system used as a communication layer for Global Arrays [6], Co-Array Fortran (CAF) [9], GPSHMEM [10], Co-Array Python [11], and also end-user applications. The paper describes the management of shared memory, integration of shared memory communication and RDMA on clusters with SMP nodes, and registration. These are all required for efficient multi- method and multi-protocol communication on modern systems. Focus is placed on techniques for supporting process groups while maximizing communication performance and efficiently managing global memory system-wide.
Revised: March 31, 2018 |
Published: December 7, 2004
Citation
Krishnan M., V. Tipparaju, B.J. Palmer, and J. Nieplocha. 2004.Processor-Group Aware Runtime Support for Shared-and Global-Address Space Models. In 33rd International Conference on Parallel Processing (ICPP-2004). Workshop on Compile and Runtime Techniques for Parallel Computing, 15–18 August 2004, Montreal, Quebec, Canada, 506-513. Piscataway, New Jersey:Institute of Electrical and Electronics Engineers. IEEE Computer Society.PNNL-SA-41767.