this document.

Presently, different vendors' MPI implementations cannot interoperate directly with each other. As a result, performance of distributed computing across different vendors' machines requires use of a single MPI implementation, such as MPICH. This solution may be sub-optimal since it cannot utilize the vendors' own optimized MPI implementations. PVMPI, a software package currently under development at the University of Tennessee, provides the needed interoperability between different vendors' optimized MPI implementations. As the name suggests PVMPI is a powerful combination of the proven and widely ported Parallel Virtual Machine (PVM) system and MPI. PVMPI is transparent to MPI applications thus allowing intercommunication via all the MPI point-to-point calls. Additionally, PVMPI allows flexible control over MPI applications by providing access to all the process control and resource control functions available in the PVM virtual machine.

jmpi is a 100% Java-based implementation of the Message-Passing Interface (MPI-1) standard. jmpi comes with a consistent MPI object model suitable for Java. Its Application Programming Interface (API) is similar to the standard C bindings of MPI. jmpi is integrated with a performance instrumentation, analysis, and visualization system called JPVS, that is also implemented in Java. Instrumented jmpi routines generate execution trace files in Pablo's SDDF format. These trace files are processed by the JPVS and processor- and communication-oriented static and dynamic performance displays are generated to help jmpi users to observe the behavior of their programs. We give sample displays from the JPVS along with some early performance results of a set of jmpi benchmark codes on a cluster of SUN UltraSparc workstations. Keywords: MPI, Java, Pablo SDDF, JPVM, performance instrumentation and visualization. 1.#INTRODUCTION We have recently witnessed a strong trend towards building high-perfor...

We discuss the use of PVM as a system for controlling the execution of MPI applications, by allowing the user access to both the MPI API and an enhanced set of the PVM API. The intention is to give the user community flexible control over MPI applications using a system that is both portable and familiar---without having to wait for new MPI-2 systems to be developed. Our system, called PVMPI, uses the already proven and widely ported MPI message-passing system within PVM to to enable interoperation with different implementations executing on distributed hardware. PVMPI also takes advantage of contexts under PVM3.4 to provide more security. Additional benefits will be available to those who currently already use resource managers that interface to PVM, in that PVMPI can control MPI applications. 1 Introduction PVM is one of a number of parallel distributed computing environments (DCEs) [16] that were introduced to assist users wishing to create portable parallel applications [19]. The ...

this report are those of the author(s) and should not be construed as an official Department of Defense position, policy, or decision unless so designated by other official documentation. MPI Inter-connection and Control

This paper presents a new kind of portability system, Unify, which modifies the PVM message passing system to provide (currrently a subset of) the Message Passing Interface (MPI) standard notation for message passing. Unify is designed to reduce the effort of learning MPI while providing a sensible means to make use of MPI libraries and MPI calls while applications continue to run in the PVM environment. We are convinced that this strategy will reduce the costs of porting completely to MPI, while providing a gradual environment within which to evolve. Furthermore, it will permit immediate use of MPI-based parallel libraries in applications, even those that use PVM for user code. We describe several paradigms for supporting MPI and PVM message passing notations in a single environment, and note related work on MPI and PVM implementations. We show the design options that existed within our chosen paradigm (which is an MPI interface added to the base PVM system), and why we chose that par...

. In this paper, we present a portable, object-oriented, pure Java implementation of the Message-Passing Interface (MPI), called jmpi. jmpi is a class library of Java-routines for specifying and coordinating parallel codes. Our pure Java implementation is distinguished from earlier implementation efforts that pervasively use native methods and provide a Java wrapper functionality to some specific traditional MPI implementations. While bringing in a consistent MPI object model suitable for Java, we also follow the standard MPI Application Programming Interface (API) definitions as closely as possible to keep the learning curve short for experienced MPI programmers. We tested the performance of jmpi, by running a set of benchmark programs written in Java with calls to jmpi library routines on a cluster of SUN UltraSparc workstations. 1#Introduction The use of a collection of general-purpose heterogeneous computer systems interconnected by existing networks and support services as a sing...