We are sincerely grateful to those who have provided comments on an earlier draft of this manuscript, including Leslie Valiant, Torsten Suel, Zhiwei Xu, Ravi Shankar, and Sanjay Ranka.
We would like to thank the CASTLE/Split-C group at The University of California, Berkeley, especially for the help and encouragement from David Culler, Arvind Krishnamurthy, and Lok Tin Liu. Computational support on UC Berkeley's 64-processor TMC CM-5 was provided by NSF Infrastructure Grant number CDA-8722788.
The University of California, Santa Barbara, parallel radix sort code was provided to us by Mihai Ionescu. Also, Klaus Schauser, Oscar Ibarra, Chris Scheiman, and David Probert of UC Santa Barbara, provided help and access to the UCSB 64-node Meiko CS-2. The Meiko CS-2 Computing Facility was acquired through NSF CISE Infrastructure Grant number CDA-9218202, with support from the College of Engineering and the UCSB Office of Research, for research in parallel computing.
Arvind Krishnamurthy provided additional help with his port of Split-C to the Cray Research T3D [3]. The Jet Propulsion Lab/Caltech 256-node Cray T3D Supercomputer used in this investigation was provided by funding from the NASA Offices of Mission to Planet Earth, Aeronautics, and Space Science. We also acknowledge William Carlson and Jesse Draper from the Center for Computing Science (formerly Supercomputing Research Center) for writing the parallel compiler AC (version 2.6) [15] on which the T3D port of Split-C \ has been based.
This work also utilized the CM-5 at National Center for Supercomputing Applications, University of Illinois at Urbana-Champaign, under grant number ASC960008N.
We also thank Jeffrey Hollingsworth from UMCP's Computer Science Department for his suggestions and encouragement.
We would like to acknowledge the use of the UMIACS 16-node IBM SP-2-TN2, which was provided by an IBM Shared University Research award and an NSF Academic Research Infrastructure Grant No. CDA9401151.
Please see http://www.umiacs.umd.edu/research/EXPAR for additional performance information. In addition, all the code used in this paper is freely available for interested parties from our anonymous ftp site, ftp://ftp.umiacs.umd.edu/pub/EXPAR. We encourage other researchers to compare with our results for similar inputs.