Anthony Skjellum, Steven G. Smith, Nathan E. Doss, Charles H. Still, Alvin P. Leung, and Manfred Morari. Zipcode: A Portable Communication Layer for High Performance Multicomputing. Technical Report UCRL-JC-106725 (revised 9/92, 12/93, 4/94), Lawrence Livermore National Laboratory, March 1991. To appear in Concurrency:  Home/Search   Document Not in Database   Summary   Related Articles   Check

....Algorithms like reduce and allreduce have strong enough source selectivity properties so that they are inherently okay (no backmasking) provided that MPI provides basic guarantees. So are multiple calls to a typical tree broadcast algorithm with the same root or different roots (see [6]) Here we rely on two guarantees of MPI: pairwise ordering of messages between processes in the same context, and source selectivity deleting either feature removes the guarantee that backmasking cannot be required. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 ....

Anthony Skjellum, Steven G. Smith, Nathan E. Doss, Charles H. Still, Alvin P. Leung, and Manfred Morari. Zipcode: A Portable Communication Layer for High Performance Multicomputing. Technical Report UCRL-JC-106725 (revised 9/92, 12/93, 4/94), Lawrence Livermore National Laboratory, March 1991. To appear in Concurrency:

.... the language does not limit the global access primitives to simple read and write, it does not rely on (or attempt to justify) novel architectural features, nor does it does it assume communication is fabulously expensive, as with global address models built on top of conventional message passing[15, 8]. These differences arise because of differences in the implementation assumptions. Split C is targeted toward the current and emerging distributed memory multiprocessors with fast, flexible network hardware, including the Thinking Machines CM 5, Meiko CS 1, Cray T3D and other. Within the global ....

A. Skjellum. Zipcode: A portable communication layer for high performance multicomputing -- practice and experience. Unpublished draft, March 1991.

....masking. Algorithms like reduce and allreduce have strong enough source selectivity properties so that they are inherently okay (no backmasking) provided that MPI provides basic guarantees. So are multiple calls to a typical tree broadcast algorithm with the same root or different roots (see [28]) Here we rely on two guarantees of MPI: pairwise ordering of messages between processes in the same context, and source selectivity deleting either feature removes the guarantee that backmasking cannot be required. Algorithms that try to do non deterministic broadcasts or other calls that ....

Anthony Skjellum, Steven G. Smith, Nathan E. Doss, Charles H. Still, Alvin P. Leung, and Manfred Morari. Zipcode: A Portable Communication Layer for High Performance Multicomputing. Technical Report UCRL-JC-106725 (revised 9/92, 12/93, 4/94), Lawrence Livermore National Laboratory, March 1991. To appear in Concurrency: Practice & Experience.

....differs from previous shared memory languages by providing a rich set of memory operations, not simply read and write. It does not rely on novel architectural features, nor does it assume communication has enormous overhead, thereby making bulk operations the only reasonable form of communication [16, 9]. These differences arise because of differences in the implementation assumptions. Split C 8 16 32 64 128 256 512 1024 400 350 300 250 200 150 100 50 0 Mflops Matrix dimension Blocked Cannon Unopt Unblocked Figure 9: Performance of multiple versions of matrix multiply on 64 Sparc processor ....

A. Skjellum. Zipcode: A Portable Communication Layer for High Performance Multicomputing -- Practice and Experience. Unpublished draft, March 1991.

....To do this, we chose to support a generalized receipt selectivity scheme, based on a mailer class, and include a variable length matching field ( PO Box ) with each letter sent by the system. Each MPI implementation specifies the number of bits in its tag; an early MPI implementation has chosen 31 bits for its tags [12] At the outset in 1988, and even more so after new discussions during the MPI Forum, we were convinced that source selectivity is a clear requirement for making collective message passing operations (based on send receive) work reliably. Source selectivity was consequently ....

....groups define a rank naming for processes in point to point communication relative to the group. In addition, groups define the scope of collective operations. This scope allows us to make strong statements about the non interference of sequential collective operations in the same context (see [31] In Zipcode, groups are static objects, not shared dynamic objects as they are in PVM [5] ffl Contexts provide the ability to have separate safe universes of message passing in Zipcode. A context is conceptually implemented via a secondary or hyper tag, that differentiates messages from ....

[Article contains additional citation context not shown here]

Anthony Skjellum, Steven G. Smith, Nathan E. Doss, Charles H. Still, Alvin P. Leung, and Manfred Morari. Zipcode: A Portable Communication Layer for High Performance Multicomputing. Technical Report UCRL-JC-106725 (revised 9/92, 11/93), Lawrence Livermore National Laboratory, March 1991. To appear in Concurrency: Practice & Experience.

....supported, in part, by the NSF Engineering Research Center for Computational Field Simulation, Mississippi State University. 1 Overview of Zipcode Zipcode is a message passing system designed and developed originally by the first author, beginning at Caltech in July, 1988 and continuing to date [16, 18, 20, 12]. Zipcode was created to address issues and provide features absent in then existing message passing systems (most of these issues and features remain largely unaddressed in extant systems) Currently, Zipcode continues to serve as a vehicle to demonstrate high level message passing research ....

.... safe distributed memory, and or distributed computing libraries, communication contexts are needed to restrict the scope of messages. This is done to prevent messages from being selected improperly by processes when they do message passing. We described contexts in several papers on Zipcode [16, 18, 20, 19]. Without this type of scope restriction, it quickly becomes intractable to build up code without globalizing the details of how each portion of a code utilizes the message passing resource. Communication contexts are therefore central to creating reusable library code, and to maintaining ....

[Article contains additional citation context not shown here]

Anthony Skjellum, Steven G. Smith, Charles H. Still, Alvin P. Leung, and Manfred Morari. Zipcode: A Portable Communication Layer for High Performance Multicomputing. Technical Report UCRL-JC-106725 (revised 9/92), Lawrence Livermore National Laboratory, March 1991. To appear in Concurrency: Practice & Experience.

....Furthermore, an advanced user can add new virtual topologies to the system. Some libraries might like to have tree or other graph topology, for instance, to make them most natural to program. A communication context is an abstraction that was introduced by the author in the original (1988) Zipcode system, and which also will appear in the MPI standard [8, 16] In order to write practical, safe distributed memory, and or distributedcomputing libraries, communication contexts are needed to restrict the scope of messages. This is done to prevent messages from being selected ....

Anthony Skjellum and Manfred Morari. Zipcode: A Portable Communication Layer for High Performance Multicomputing. Technical Report UCRL-JC-106725, Lawrence Livermore National Laboratory, March 1991. To appear in Concurrency: Practice & Experience.

....mechanism for much higher message passing performance than first generation multicomputers, as described in [2] Mesh edges provide a rich potential source of input output bandwidth for visualization, or disk farms. methods for communication primitives and collections of communicating processes [14, 18]. A lot more can be said about the communication issues that the Toolbox addresses and supports, and that underly grid based communication described here, as a particular example; this discussion appears elsewhere [14, 18] and we will not consider these issues further at present. We will also ....

.... for communication primitives and collections of communicating processes [14, 18] A lot more can be said about the communication issues that the Toolbox addresses and supports, and that underly grid based communication described here, as a particular example; this discussion appears elsewhere [14, 18], and we will not consider these issues further at present. We will also omit details of the data structures involved in the abstraction from communication layers to concurrent data structures, leaving this for [18] and further planned publications. Skjellum Baldwin The Multicomputer ....

[Article contains additional citation context not shown here]

Anthony Skjellum and Manfred Morari. Zipcode: A Portable Communication Layer for High Performance Multicomputing -- Practice and Experience. Technical Report UCRL-JC-106725, Lawrence Livermore National Laboratory, March 1991. Accepted by Concurrency: Practice & Experience. In minor revision.

....as a correctness metric for optimized versions of the library. Our CBLAS library is based on the Multicomputer Toolbox with its Zipcode message passing system. Zipcode was designed to be a general purpose message passing system that is easy to implement on any high performance multicomputer (see [10, 8, 11, 9]) Current implementations layer on the Reactive Kernel calls. Planned Zipcode ports to programming 2 Falgout, Skjellum, Smith Still The Multicomputer Toolbox : models like CMMD, PVM, PICL, and Express offer the potential for application level message passing in both homogeneous and ....

....Like other message passing systems, Zipcode provides the message passing functions needed to perform communication effectively on several node topologies. Unlike other systems, it also interfaces to a grid oriented description of message passing as well as a Toolbox library of data distributions [8, 11]. By building the CBLAS routines on this platform, we provide data distribution independence, portability, and hardware abstractions. By marrying the CBLAS to general data distributions from the outset, we significantly increase the number of real parallel applications, with varied data ....

Anthony Skjellum and Manfred Morari. Zipcode: A Portable Communication Layer for High Performance Multicomputing -- Practice and Experience. Technical Report UCRL-JC-106725, Lawrence Livermore National Laboratory, March 1991. Accepted by Concurrency: Practice & Experience. In minor revision.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC