MPI: A Message-Passing Interface Standard, MPI Forum, Mar. 1994.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....behavior of irregular parallel computations. Similarly, high level parallel languages and programming environments often do not contain support for specifying the I O activities of the program, although the recent incorporation of I O constructs into the MPI message passing standard is promising [MPI97] 2.4 Run time libraries Perhaps one of the most active areas of software development in parallel I O research has been in run time libraries. Run time libraries are the quickest way to provide I O support to a wide range of applications on a wide range of platforms, avoiding the need to modify ....

.... support a compiler, such as ViC [CC94, CH97] or PASSION [TCB 96] Others are oriented toward scientific applications in general [TG96, SW96] Still others are designed for specific application domains, such as computational chemistry [NFK98] Although few attempt to be standards, except MPI 2 [MPI97] these libraries allow the application programmer to take advantage of carefully tuned algorithms and proven techniques. 2.5 Operating systems The operating system of a parallel or distributed computer has to strike a delicate balance when it comes to I O. On one hand, it must provide the ....

[Article contains additional citation context not shown here]

MPI-2: Extensions to the message-passing interface. The MPI Forum, July 1997.

....for a parallel programming environment is a trade off between portability and performance. Higher level portable libraries are generally slower than platform dependent libraries. The next version of the MPI library (MPI 2) provides one sided communications along with synchronization routines [22]. MPI 2 routines such as MPI Put, MPI Get, MPI Win Lock and MPI Win Unlock support a shared memory programming paradigm. Thus, both our RCPQ and BCPQ (cf. Section 3) concurrent data structure implementations can become portable by using the MPI 2 library. At present, there is no MPI 2 ....

MPI-2: Extensions to the message-passing interface. MPI Forum, July 1997.

....physically distributed but logically single data structure (e.g. one could use the same scheme shown in Fig. 3 for arrays in MP programs) an MP process would need help from another applicationlevel process to access remote data. With MP system that provides one sided communication (e.g. MPI 2 [52]) this help may be delayed but is still needed, otherwise the principle of pivot computes will be violated sooner or later. Table I gives a taxonomy of variables in MESSENGERS. In addition to the primitive node and agent variables, shared variable programming in distributed memory requires the ....

"MPI-2: Extensions to the message-passing interface," The MPI Forum, July 1997.

....and then provided to a virtual classroom through the use of a lightweight Java application. 1 2 Main Architecture Description Before starting the project, we had to define the operating context as far as parallel computation concerns. At the moment the best of brand was the model defined by MPI [3] specifications that allow the user to write concurrent programs based on message passing paradigm and following a well defined and flexible communication standard. The animation architecture we propose is distributed into three distinct levels, these are: ffl Parallel animated program A pool of ....

"MPI standard 2.0 ". MPI Forum, 18 July 1997.

....within row and columns of processors configured in a 2 dimensional geometry. The example from the right employs halo updates in an irregular multiblock stencil based computation. Multicast A Multicast B Local DGemm GHOST CELLS INTERIOR 2 On a multicomputer, communication interfaces like MPI [6] support many of the above communication algorithms. For example, MPI provides a data type mechanism to express regular sections with strides, communication domains to configure processor geometries, and a variety of global communication primitives. But MPI has limitations. Support for irregular ....

, 1995, MPI Forum. http://wwwunix. mcs.anl.gov/mpi/index.html

....Such tasks include decision trees, neural networks, image processing, and dimension reduction. Each class of algorithms is designed using objectoriented principles and implemented as a C class library. Parallelism is supported through the use of MPI for distributedmemory parallel processing [9]. We use domain speci c software for tasks such as reading, writing, and the display of data. To support many di erent input data formats, such as FITS, View, and netCDF, we rst convert each format into Sapphire s internal data format, prior to any processing. We are using RDB, a public domain ....

MPI Forum. http://www.mpi-forum.org.

....typically provide routines to initialise and configure the messaging environments as well as sending the receiving packets of data. The two most popular high level message passing systems for scientific and engineering applications are MPI (Message Passing Interface) 28] defined by the MPI Forum [29] and the PVM (Parallel Virtual Machine) 27] from Oak Ridge National Laboratory and the University of Tennessee at Knoxville, USA. MPI I is the de facto standard for parallel programming, both on clusters, and on traditional parallel supercomputers, such as the Cray T3E and IBM SP2. MPI consists ....

MPI Forum, http://www.mpi-forum.org/

....independently of one another, as in stencil algorithms [11] A great deal of research has already addressed this I O bottleneck. The mismatch between the traditional Unix like I O interface and the access patterns of parallel applications have lead to a re evaluation of file system interfaces [10, 15, 16]. These interfaces present a higher level of abstraction than the traditional interface, thus facilitating the recognition of access patterns and allowing effective prefetching. Unfortunately, the above approaches require code be rewritten to utilize the specialized interfaces. Such a requirement ....

MPI-2: Extensions to the message-passing interface. The MPI Forum, July 1997.

....specialized, typically expensive, parallel computing platforms. An important factor that has made the usage of clusters a practical proposition is the standardization of many of the tools and utilities used by parallel applications. Examples of these standards are the message passing library MPI [8] and data parallel language HPF [3] In this context, standardization enables 1 Clusters, Network of Workstations (NOW) Cluster of Workstations (COW) and Workstation Clusters are synonymous. 8 Cluster Computing at a Glance Chapter 1 applications to be developed, tested, and even run on NOW, ....

....as well as sending and receiving packets of data. Currently, the two most popular high level message passing systems for scientific and engineering application are the PVM (Parallel Virtual Machine) 22] from Oak Ridge National Laboratory, and MPI (Message Passing Interface) defined by MPI Forum [8]. PVM is both an environment and a message passing library, which can be used to run parallel applications on systems ranging from high end supercomputers through to clusters of workstations. Whereas MPI is a message passing specification, designed to be standard for distributed memory parallel ....

MPI Forum. http://www.mpi-forum.org/docs/docs.html

....(5) each subdomain mesh is then assigned to a processor within a portable distributed object oriented framework for clusters of workstations and distributed memory supercomputers such as the Cray T3E. Communication among processors is handled using the standard message passing interface MPI [25] protocols. e.g. All data are sent to processors using a non blocking protocol called MPI Isend. This allows a processor to proceed to the next level of computation without waiting for the receiving processor to read the data. To prevent bottlenecks, all processors receive data using a blocking ....

MPI: A Message-Passing Interface Standard, MPI Forum, April 21, 1994.

....[9] which provides a mechanism for connecting to, searching and modifying internet directories, and remote database access APIs such as OLEDB over DCOM. Parallel I O is about high performance I O rather than adaptive I O: the goal of our work. However we were influenced by MPI IO, part of MPI 2. 0 [3], in that we wanted to use a mechanism analogous to communication for I O. 7 Discussion The basic ideas presented in the previous sections have been implemented. This has provided a simple and effective means to perform I O in Gardens. So far we have not implemented sophisticated caching ....

MPI Forum. MPI-2.0. http://www.mpi-forum.org/.

....of multiple processes so that (1) the user can periodically monitor a long running application and (2) multiple scientists can collaborate while visualizing their data. Several communication packages or specifications meet these design requirements, including Nexus [17] the MPI 2 specification [18] (although dynamic process management has not yet been implemented on most MPP architectures) PVM [19] and various tools built with these communication infrastructures. For our initial tests we use the CAVEcomm library [12] developed at Argonne National Laboratory, which uses Nexus. CAVEcomm ....

MPI-2: Extensions to the message passing interface. http://www.mpi-forum.org/docs/docs.html, 1997.

....an SGI Origin 2000 show that our MPI prototype called TMPI using the proposed techniques is competitive with SGI s native MPI implementation in a dedicated environment, and it has significant performance advantages in a multiprogrammed environment. 1 Introduction MPI is a message passing standard [3, 34] widely used for developing high performance parallel applications. MPI standard 1.1 was initially designed for distributed memory machines and workstation PC clusters. Because SMMs become popular due to their commercial success, it is important to address performance portability of MPI code on ....

....previous MPI research is focused on distributed memory machines or workstation clusters, e.g. 10] The MPI SIM project [8] has used multi threading to simulate MPI execution on distributed memory machines for performance prediction as we will discuss in Section 3.1. Thread safety is addressed in [3, 30, 33]. However, their concern is how multiple threads can be invoked in each MPI node, but not how to execute each MPI node as a thread. These studies are useful for us to relax our assumptions in the future. Previous work has also illustrated the importance of lock free management for reducing ....

[Article contains additional citation context not shown here]

MPI Forum. http://www.mpi-forum.org.

....systems were very rarely ported among different platforms. This resulted from the fact that implementations were tightly coupled to the underlying O.S. communication layers, either UNIX BSD sockets or other specific operating system message passing interfaces. The Message passing standard (MPI) [14] changed this scenario by providing a common platform for the development of message passing applications. Furthermore, it includes some additional features specifically to support the development of parallel libraries [20] Process groups, Virtual Topologies, Contexts, and Communicators. In ....

"MPI: A Message Passing Interface Standard", The MPI Forum, March 31, 1994.

....which limits its use. This paper presents performance results showing that the efficiency of Linda implementations can be significantly increased if the Linda system accepts hints from the programmer concerning data (tuple) usage. This was accomplished in a Linda library developed on top of MPI, named Eilean, which provides means for the transference of knowledge from the programmer to the run time system without compromising Linda s high level programming concept. Performance results for several Linda benchmarks were collected in a CRAY T3D, with and without the inclusion of programmer ....

....A possible solution is to use higher level programming models that facilitate the task of programming in parallel. Portability of these higher level programming paradigms can be achieved if they are implemented as parallel libraries on top of a standard message passing environments like MPI [MPI94] On the other hand, while there is an inherent performance penalty in using high level libraries instead of MPI, the gains in terms of code reuse and programming effort can be considerable [Clarke94] The Linda parallel programming paradigm [Ahuja86] Carriero87] is a simple and elegant ....

[Article contains additional citation context not shown here]

"MPI: A Message Passing Interface Standard", The MPI Forum, March 31, 1994.

....lower latency than the normal host file system. An example might be a video display program. MPI IO is the ROMIO version of the standard MPI IO interface which has been implemented on top of API0 to support access to BPFS files from parallel programs that use the Message Passing Interface (MPI) [5]. API0 directly interfaces to BPFS, and the other two interfaces are implemented on top of API0, as shown in Figure 1. User Applications ROMIO ADIO API0 BPFS CLI Figure 1: Layering of the API0, CLI, and MPI IO interfaces to BPFS. This report is organized as follows. Section 2 describes the ....

....must exist, to the name given by new, which must not exist. Both functions return zero if they are successful, EOF if they are not. 4 MPI IO over BPFS MPI, which stands for Message Passing Interface , is a well established standard that has recently been extended to a new standard called MPI2 [5]. The most important of the extensions for our work is the addition of I O facilities, which have been given the name MPI IO [3] and are described in Chapter 10 of [5] ROMIO [8] is a high performance, portable implementation of MPI IO that is freely available at ....

[Article contains additional citation context not shown here]

MPI-2: Extensions to the message-passing interface. The MPI Forum, July 1997.

....Multi threading, Communication optimization. 1. INTRODUCTION With the commercial success of SMP architectures, SMP clusters with commodity components have been widely deployed for high performance computing due to the great economic advantage of clustering [1, 2] MPI is a messagepassing standard [14] widely used for high performance parallel applications and has been implemented on a large array of computer systems. This paper studies fast execution of MPI programs on dedicated SMP clusters. In the MPI paradigm, all the MPI nodes execute the same piece of program under separate address ....

MPI-Forum. MPI Forum, 1999. http://www.mpi-forum.org.

....and queues General Terms: Design, Performance, Languages, Algorithms, Experimentation Additional Key Words and Phrases: MPI, program transformation, lock free synchronization, threaded execution, shared memory machines, multiprogrammed environments 1. INTRODUCTION The Message Passing Interface (MPI) MPI Forum 1999; Snir et al. 1996] is the de facto industry standard for developing high performance parallel applications on various platforms. People use MPI on shared memory machines (SMMs) mainly because MPI programs with coarse grain parallelism can perform as well as other parallel programming models for ....

....fit in one process s open file table. This is adjustable by OS reconfiguration. Our lock free data structure design for thread communication makes the following assumption: each MPI node does not spawn multiple threads which call MPI functions simultaneously. As discussed in the MPI 2 standard [MPI Forum 1999], there are four possible levels of thread support: 1) Only one thread runs within each MPI node. 2) Each MPI node may be multithreaded, but only the main thread will make MPI calls. 3) There are multiple threads spawned, however the user program guarantees that MPI functions calls are ....

[Article contains additional citation context not shown here]

MPI-FORUM. 1999. MPI Forum. http://www.mpi-forum.org.

....environments. The improvement ratio can reach as much as 161 or even more depending on the degree of multiprogramming. 1 Introduction Recently shared memory machines (SMMs) have become popular for high end computing because of their success in the commercial market. Even though MPI [2, 12, 26] is designed for distributed memory programming, it is important to address performance portability of MPI code on shared memory machines (SMMs) There are three reasons that people use MPI on SMMs. First of all, MPI has been widely used for many parallel applications on large distributed memory ....

....with a common key. Given the same key value, each thread can store retrieve its own copy of data. The details are in [27] The transformation involves indirect access of TSD based permanent variables. The overhead of such indirection is insignificant in practice. Thread safety is also addressed in [2, 23], however, their concern is how multiple threads can be invoked in each MPI node (process) but not executing each MPI node as a thread. Currently we assume that each MPI node does not spawn multiple threads and we plan to relax this constraint in the future 1 . Message Passing Interface ....

[Article contains additional citation context not shown here]

MPI Forum. http://www.mpi-forum.org/.

No context found.

MPI: A Message-Passing Interface Standard, MPI Forum, Mar. 1994.

No context found.

MPI: A Message Passing Interface Standard, www.mpi-forum.org, June 1995.

No context found.

MPI-2: Extensions to the Message-Passing Interface, The MPI Forum (July 1997).

No context found.

MPI Forum: MPI-2: Extensions to the Message-Passing Interface, http://www.mpi-forum.org/docs/mpi-20.ps.Z (1997).

No context found.

, 1997, MPI Forum. http://wwwunix. mcs.anl.gov/mpi/index.html 13

No context found.

MPI Forum - http://www.mpi-forum.org/docs/docs.html

First 50 documents

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