Volume 7, Issue 1 -
Spring/Summer 1999

Volume 6, Issue 3
Fall 1998

Volume 6, Issue 2
Spring/Summer 1998

Volume 6, Issue 1
Winter 1998

Volume 5, Issue 4
Fall 1997

Volume 5, Issue 3
Summer 1997

Volume 5, Issue 2
Spring 1997

Volume 5, Issue 1
Winter 1997

Volume 4, Issue 4
Fall 1996

Volume 4, Issue 3
Summer 1996

Volume 4, Issue 2
Spring 1996

Volume 4, Issue 1
Winter 1996

Volume 3, Issue 4
Fall 1995

Volume 3, Issue 3
Summer 1995

Volume 3, Issue 2
Spring 1995

Volume 3, Issue 1
January 1995

Volume 2, Issue 4
October 1994

Volume 2, Issue 3
July 1994

Volume 2, Issue 2
April 1994

Volume 2, Issue 1
January 1994

Volume 1, Issue 4
October 1993

Volume 1, Issue 3
July 1993

Volume 1, Issue 2
April 1993

Volume 1, Issue 1
January 1993


A host of programming languages for message-passing machines were compared to determine the best programming language an applications programmer should use when writing a portable code for a message- passing machine (e.g., the Intel Paragon or a workstation cluster). The languages studied included the p4 and PVM languages developed in part by researchers at Argonne National Laboratory and the University of Tennessee/Oak Ridge National Laboratory. PVM and p4 are explicit message-passing systems (i.e., one processor explicitly sends data to another processor).

Computer scientists Craig Douglas and Martin Schultz of Yale University and Timothy Mattson of Intel performed benchmark studies using various message-passing systems (PVM, p4, TCGMSG, C-Linda, and POSYBL) on a wide variety of platforms. Ease of use and run-time effects were among the factors that the group considered. The group performed several types of benchmarks on these languages: the first benchmark measured the communication rates for a ping-pong test; a second benchmark on a ring of four SPARCstation 1 workstations created an array at each workstation and then passed the array to its neighbor; the third benchmark answered the question "How do these various languages impact a code running over a collection of processors?" using a commercial application from the pharmaceutical industry with heavy communication requirements running on a cluster of four IBM RS/6000 workstations.

The study drew several conclusions. Although message-passing languages like PVM, p4, and TCGMSG were measured using the same network protocol, the communication rates differed due to buffer-management costs on either end of communication. Also, as communication patterns grew in complexity, the difference in performance decreased. The study also noted that run-time efficiency told only part of the story, since the system's ease-of-use and debugging tools can greatly aid in the programming effort. The complete report of the study, TR975.ps, is available via ftp at: casper.cs.yale.edu in the directory pub.

PVM, an ongoing research project at University of Tennessee/Oak Ridge National Laboratory, can be obtained by anonymous ftp from www.netlib.org in the directory pvm3. The p4 language, developed at Argonne National Laboratory, is available by anonymous ftp from info.mcs.anl.gov in the directory pub/p4.

Source: Carl Scarbnick, San Diego Supercomputer Center, Gather/Scatter January 1994

Table of Contents

News | From the Director | Parallel Profile | Research Focus | Work in Progress | Education / Outreach | Resources | Calendar | CRPC Home