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

As part of a joint project with IBM, Argonne National Laboratory took delivery in May of a prototype version of IBM's newest parallel distributed-memory computer, the SP-1. Argonne researchers were eager to evaluate the software environment and to begin porting programming environments and applications to this machine.

They began by porting a variety of programming packages, including:

• BlockSolve - a software library for large, sparse systems of linear equations on massively parallel computers
• Chameleon - a package providing a uniform mechanism for accessing various message-passing systems
• MPI - a message-passing standard currently being developed by CRPC researchers and a broad group of vendors and users
• PCN - a coordination language
• PETSc - a set of portable, extensible tools for scientific computing
• P4 - a portable message-passing and shared- memory library

Each of these programming packages ran with little or no modification. The principal difficulty encountered was problems with the software configuration. Once these packages had been successfully ported, seven applications were ported to the SP-1:

• community climate model
• mesoscale weather model
• program to construct phylogenetic trees
• protein folding program
• model of vortices in high-temperature superconductors
• parallel theorem prover

The use of portability layers offered by Chameleon, PCN, and p4 let researchers port these applications quickly. Having a full, running Unix operating system (IBM AIX) on each node also greatly facilitated the ports.

In addition, the prodigious memory of the SP-1 (128 MB per node, augmented by virtual memory), made it possible to run large problems on small numbers of nodes or even a single node; this feature was very useful for debugging. The large amount of memory also allowed the set up of large buffers that could be used for asynchronous I/O.

Even on this test version of the SP-1, researchers were able to investigate an important scientific question in structural biology. Using heuristic optimization algorithms, they searched for the "best" phylogenetic tree from 47 aligned rRNA sequences. An overnight run produced 72 different outputs based on random orderings of the sequences. The information that was obtained shed new light on the placement of mitochondria within the alpha purple bacteria.

Admittedly, in each of the applications there were some difficulties. Most notably, there are a number of troublesome inconsistencies between IBM's implementation of Fortran and other UNIX implementations. Better documentation would also help. Nevertheless, with the delivery of the fast interconnect switch, CRPC researchers at Argonne can immediately begin to use the SP-1 for large applications as well as continued development of portable parallel programming tools. Since all of the applications are built on top of these tools, they should all run with the fast interconnect without any changes other than relinking.

A technical report (ANL/MCS-TM-177) has been produced documenting these early experiences with the IBM SP-1. For further information or a copy of the report, send email to gropp@mcs.anl.gov .