Changes to the archive in in April 1995

Newest entries are first. Older changes can be found here.

24th April 1995

/parallel/documents/mpi/anl/sut-1.0.15.tar.Z: Scalable Unix Tools V1.0.15: pps, pls, load, gload, prun, pkill, prm, pdistrib, pfind, fps, pfps etc. by Gropp and Lusk. Includes paper.

21st April 1995

/parallel/transputer/software/compilers/gcc/yaroslavl/changes6: Changes in alpha 6 version
/parallel/transputer/software/compilers/gcc/yaroslavl/gcc-2.6.3-t800.6.dif.gz: Alpha 6 version
/parallel/transputer/software/compilers/gcc/yaroslavl/patch6.gz: Patch from alpha 5 to alpha 6
/parallel/environments/chimp/release/chimp.tar.Z: Updated CHIMP distribution
/parallel/documents/hippi/hippi-atm_1.5x.ps.gz: HIPPI over ATM document Version 1.5
/parallel/documents/hippi/hippi-atm_1.5x_changes.ps.gz: HIPPI over ATM changes
/parallel/documents/hippi/minutes/apr95_hippi_min.ps.gz
/parallel/documents/hippi/minutes/apr95_hippi_min.txt: Minutes for April 1995 HIPPI meeting
/parallel/documents/pario/papers/Kotz/kotz:explore.ps.Z: Exploring the use of I/O Nodes for Computation in a MIMD Multiprocessor by David Kotz and Ting Cai, Department of Computer Science, Dartmouth College, Hanover, NH 03755, USA. fdfk,tcaig@cs.dartmouth.edu ABSTRACT: As parallel systems move into the production scientific-computing world, the emphasis will be on cost-effective solutions that provide high throughput for a mix of applications. Costeffective solutions demand that a system make effective use of all of its resources. Many MIMD multiprocessors today, however, distinguish between compute and I/O nodes, the latter having attached disks and being dedicated to running the file-system server. This static division of responsibilities simplifies system management but does not necessarily lead to the best performance in workloads that need a different balance of computation and I/O. Of course, computational processes sharing a node with a file-system service may receive less CPU time, network bandwidth, and memory bandwidth than they would on a computationonly node. In this paper we begin to examine this issue experimentally. We found that highperformance I/O does not necessarily require substantial CPU time, leaving plenty of time for application computation. There were some complex file-system requests, however, which left little CPU time available to the application. (The impact on network and memory bandwidth still needs to be determined.) For applications (or users) that cannot tolerate an occasional interruption, we recommend that they continue to use only compute nodes. For tolerant applications needing more cycles than those provided by the compute nodes, we recommend that they take full advantage of both compute and I/O nodes for computation, and that operating systems should make this possible.

20th April 1995

/parallel/software/simulators/chaos/docs/minimal.ps.Z: Performance Analysis of a Minimal Adaptive Router by Thu Duc Nguyen and Lawrence Snyder, Dept. of Computer Science and Engineering, University of Washington, Seattle, Washington, USA. In Proceedings of the 1994 Parallel Computer Routing and Communication Workshop, May 1994, pp. 31-44. Copyright 1994, Springer-Verlag. ABSTRACT: Two classes of adaptive routers, minimal and non-minimal, are emerging as possible replacements for the oblivious routers used in current multicomputer networks. In this paper, we compare the simulated performance of three routers, an oblivious, a minimal, and a non-minimal adaptive router, in a twodimensional packet switching torus network. The non-minimal adaptive router is shown to give the best performance and the oblivious router the worst. Significantly, however, for many traffic patterns, the minimal adaptive router's performance degrades sharply as the network saturates. Based on an analysis made using several visualization tools, we argue that this performance drop results from nonuniformities introduced for deadlock prevention. Furthermore, this analysis has led us to believe that network balance is an important performance characteristic that has been largely overlooked by designers of adaptive routing algorithms.
/parallel/software/simulators/chaos/docs/ebn.ps.Z: Updated: "The Express Broadcast Network: A Network for Low-Latency Broadcast of Control Messages" by Kevin Bolding and William Yost, Dept. of Computer Science and Engineering, University of Washington, Seattle, Washington, USA. November 28, 1994. ABSTRACT: We present the Express Broadcast Network (EBN), a network used for quick and reliable broadcast of control messages in multicomputer networks. The EBN can be implemented with a single extra wire per network link and with minimal extra hardware at each routing node. However, it provides very fast broadcast mechanisms that take advantage of all redundancy in the network to deliver messages regardless of faulty network components. We present extensions of the basic network to include multiple-wire, multiple-bit, and bidirectional wire support, as well as describing basic methods of using the EBN for various applications.

7th April 1995

/parallel/environments/lam/distribution/mpi-quick-ref.ps.gz: Updated MPI Quick Reference Card

4th April 1995

/parallel/transputer/software/compilers/gcc/pereslavl/: Alpha release of GCC for T800 by Yury Shevchuk <sizif@botik.yaroslavl.su>
/parallel/transputer/software/compilers/gcc/pereslavl/ANNOUNCE.gcc-t800: Announcement of GCC for T800
/parallel/transputer/software/compilers/gcc/pereslavl/README.T800: Overview of package
/parallel/transputer/software/compilers/gcc/pereslavl/changes5: Changes in alpha 5 version
/parallel/transputer/software/compilers/gcc/pereslavl/gcc-2.6.3-t800.5.dif.gz: Alpha 5 version
/parallel/transputer/software/compilers/gcc/pereslavl/patch5.gz: Patch from alpha 4 to alpha 5
/parallel/transputer/software/compilers/gcc/pereslavl/changes4: Changes in alpha 4
/parallel/transputer/software/compilers/gcc/pereslavl/gcc-2.6.3-t800.4.dif.gz: Alpha 4 version
/parallel/transputer/software/compilers/gcc/pereslavl/patch4.gz: Patch from alpha 3 to alpha 4
/parallel/transputer/software/compilers/gcc/pereslavl/gcc-2.6.3-t800.3.dif.gz: Alpha 3 version
/parallel/software/simulators/chaos/docs/chip.ps.Z: The Chaos Router Chip: Design and Implementation of an Adaptive Router by Kevin Bolding, Sen-Ching Cheung, Sung-Eun Choi, Carl Ebeling, Soha Hassoun, Ton Anh Ngo and Robert Wille, Department of Computer Science and Engineering, University of Washington, Seattle, Washington, USA. In Proceedings of VLSI '93, IFIP, 1993, pp. 311-320. ABSTRACT: Chaotic routers are randomizing, non-minimal adaptive packet routers designed for use in the communication networks of parallel computers. Although adaptive routing, and, specifically, chaotic routing, has been shown to be superior to oblivious routing in most cases, the practical application of adaptive routing to multi-computer networks has been difficult to achieve due to the complex nature of adaptive routers. A prototype two-dimensional (mesh and torus) chaotic router chip has been designed and is being fabricated in a 1:2m CMOS process. The chip exhibits high bandwidth, limited only by the speed of the off-chip drivers, and low input-to-input latency. To achieve this, much attention is given to reducing the critical path complexity of the router. The resulting chip is shown to be as good or better than state-of-the-art oblivious routers in almost all cases.
/parallel/software/simulators/chaos/docs/ftchip.ps.Z: Design of a Router for Fault-Tolerant Networks by Kevin Bolding and William Yost, Department of Computer Science and Engineering, University of Washington, Seattle, WA 98195. In Proceedings of the 1994 Parallel Computer Routing and Communication Workshop, May 1994, pp. 226-240. ABSTRACT: As interconnection networks grow larger and larger, the need for reliable message delivery in the presence of faults grows as well. Unfortunately, most network routing schemes currently in use do not provide graceful tolerance of even the most common faults. Because routing messages around failed components requires non-minimal routing, it makes sense to examine routers which, by design, allow packets to take nonminimal routes. Such routers provide a basic level of fault-tolerance by allowing messages to be routed around faults, without requiring a priori knowledge of their locations. However, the mechanisms can be slow and clumsy at times. We augment Chaotic routing, a non-minimal adaptive routing scheme, with a limited amount of hardware to support fault detection, identification, and reconfiguration so that the network can automatically reconfigure itself when faults occur. We present a high-level design of these mechanisms, driven by the goal of achieving reasonable reliability without exorbitant cost.
/parallel/environments/pvm3/tape-pvm/: The Tape/Pvm event tracing tool developed and maintained at LMC-IMAG.
/parallel/environments/pvm3/tape-pvm/announcement: Announcement of Tape/PVM Author: Eric Maillet <maillet@imag.fr>.
/parallel/environments/pvm3/tape-pvm/ReadMe: ReadMe
/parallel/environments/pvm3/tape-pvm/tape.tgz: Tape/Pvm Source tree contains instructions on setting up, building and installing the distribution. Author: Eric Maillet <maillet@imag.fr>.
/parallel/environments/pvm3/tape-pvm/manual-tape.ps.gz: Tape/Pvm User Manual Author: Eric Maillet <maillet@imag.fr>.