Ralph M. Butler and Ewing L. Lusk. Monitors, messages, and clusters : the p4 parallel programming system. Technical report, University of North Florida and Argonne National Laboratory, 1993. http://wwwfp. mcs.anl.gov/~lusk/p4/p4-paper/paper.html.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....C as the implementation language. There are several freely distributed parallel enhancements of C. These enhancements fall roughly into two categories: message based C libraries and enhancements to the C syntax and semantics. Among the message based C libraries, there are several notable examples [2, 7, 12]. Our choice of MPI [9, 12] was in uenced by its proposal as a portable standard and its appropriately layered feature set. Several developers of other message based libraries are participating in the development of the MPI standard, and there are several good implementations of MPI on top of ....

....below, myfactor(number) would then be called on master and slaves at the same time. A standard sieve algorithm is used. The master assigns 1000 consecutive numbers to a slave to test for the presence of factors. It returns a list of factors and their exponents. For example, myfactor(12) returns [[2,2], 3,1] 2 3 = 12) function myfactor (num) sqrt : ceiling(sqrt(num) currNum : 1; answer : MasterSlave( getTask: function( ifcurrNum sqrt then currNum : currNum 1; return currNum; doTask: 4 function(tstNum) for j from tstNum to tstNum 1000 do ifnum mod j = 0 ....

[Article contains additional citation context not shown here]

R.M. Butler and E.L. Lusk, \Monitors, messages, and clusters: The p4 parallel programming system", Parallel Computing 20, (1994) pp. 547-564.

....an acceptable price for the bene ts of using a supercomputer. This attitude was not sustainable as one parallel architecture gave way to another, and the cost of porting software became exorbitant. For several years now, portability across platforms had been a central concern in parallel computing [4, 5, 14, 13, 23]. More fundamentally, the assumption that high performance computing will be done primarily on specialized supercomputers is questioned increasingly. Rapid strides in performance and connectivity of ordinary workstations and PCs make it look equally possible that the future of parallel computing ....

Ralph Butler and Ewing Lusk. Monitors, messages, and clusters: The p4 parallel programming system. Parallel Computing, 20:547-564, April 1994.

....and eciently. In this section of the paper we will describe the WMPI implementation of mpiJava for NT. 7 3.2 WMPI WMPI from the Instituto Supererior de Engenharia de Coimbra, Portugal is a full implementation of MPI for Microsoft Win32 platforms. WMPI is based on MPICH and includes a P4 [2] device standard. P4 provides provides the communication internals and a startup mechanism. The WMPI package is a set of libraries (for Borland C , Microsoft Visual C and Microsoft Visual FORTRAN) The release of WMPI provides libraries, header les, examples and daemons for remote start up. ....

Ralph Butler and Ewing Lusk. Monitors, messages, and clusters: The p4 parallel programming system. Parallel Computing, 20:547-564, April 1994.

....constituting a compromise between LINDA, CORBA and the more explicit process models mentioned above. 2. 2 Data Models Global data Many programming models allow the discrimination between private (local) and shared (global) data (among others, LINDA, Split C[Cul96] CC [CK94] the P4 package[BL94] With the exception of SPLIT C all these models support both shared and distributed memory. However, in order to get the expected performance results on specific platforms, one has to use different paradigms within these languages. Remote communication and synchronization are typically treated ....

R. Butler and E. Lusk. Monitors, messages, and clusters: The p4 parallel programming system. Parallel Computing 20, 1994.

....J. Dongarra, Steve W. Otto, Marc Snir, and David Walker Standard MPP and Work st at ions for COMMUNICATIONS OF THE ACM July 1996 Vol. 39, No. 7 85 most useful features of several systems, rather than choosing a single system as the standard. MPI has roots in PVM [4, 5] Express [9] P4 [1], Zipcode [10] and PARMACS [2] and in systems sold by IBM, Intel, Meiko Scientific, Cray Research, and nCube. MPI is used to specify the communication among a set of processes forming a concurrent program. The message passing paradigm is attractive because of its wide portability and ....

Butler, R., and Lusk, E. Monitors, messages, and clusters: The P4 parallel programming system. Parallel Comput. 20 (April 1994), 547--564.

.... and services; the technique used to achieve flow and error control is fixed for a given communication protocol (e.g. window based scheme is used in TCP IP protocol suite) There has been intensive research to de velop new communication systems to support network centric computing environment [1, 2, 3, 4, 5, 6, 7, 8, 9, 11]. There has been some success in improving communication performance and in reducing the communication overhead. However, they were not able to provide dynamic and programmable communication services that can meet the requirements of a wide range of HPDC applications. As a result, the application ....

R. Butler and E. Lusk. Monitors, Message, and Clusters: The p4 Parallel Programming System. Parallel Computing, Vol. 20, pp. 547--564, April 1994.

....the main purpose of achieving maximum efficiency. Therefore, one among the main goals of PARMA 2 project is to develop also a dedicated version of MPI, evolution of the MPICH implementation. MPIPR library is mainly devoted to greatly simplify the internal protocol used by the MPI ch p4 device [17], in order to dramatically decrease software latencies. Until now only synchronous MPIPR primitives have been implemented, as they mainly affect the performance of distributed applications. The MPICH implementation of MPI adopts a useful layered approach [18] The channel interface is responsible ....

....small user messages sending two packets can determine a substantial latency overhead. Therefore, packets with size below a preset cut off are sent together with the control part. The default cut off value is 1024 bytes. The ch p4 device derives from the preexisting p4 parallel programming system [17], developed at Argonne National Laboratory too. Its protocol imposes that, during a send operation, a suitable 40 byte header is built, with all necessary information (source, destination, message type, length, etc. and sent apart from data message, in a separate socket write( system call. This ....

R. M. Butler and E. L. Lusk: Monitors, Messages, and Clusters: The p4 Parallel Programming System. Parallel Computing Vol. 20 (1994) 547--564

....(host to grid and vice versa) and support for reduction operations and file input output. We have developed for this archetype an implementation consisting of a code skeleton and an archetype specific library of communication routines, 3 with versions based on Fortran M [40] Fortran with p4 [17], and Fortran with NX [60] The implementation is described in detail in [57] it has been used to run applications on the IBM SP, the Intel Delta, the Intel Paragon, and a network of Sun workstations. 7.3 Applications This section discusses the second phase of the archetypes related ....

R. M. Butler and E. L. Lusk. Monitors, messages, and clusters --- the p4 parallel programming system. Parallel Computing, 20(4):547--564, 1994.

....35 new val(i,j) f(old val(i,j) old val(i 1,j) old val(i 1,j) old val(i,j 1) old val(i,j 1) Figure 3.1: Basic mesh computation. and other housekeeping. This archetype code (template and library) has been implemented in Fortran M, Fortran with Intel s NX Library, and Fortran with p4 [BL94] All implementations have essentially the same application programmer interface, so applications developed using one implementation are trivially ported to another. We note that for the two mesh archetype applications given in Sections 3.1 and 3.2, the total process count includes a host ....

R. M. Butler and E. L. Lusk. Monitors, Messages, and Clusters--- The p4 Parallel Programming System. Parallel Computing, 20:547-- 564, 1994.

....routine, which is implemented in a way that is likely to cause hot spots in any network. 4.2 The ch p4 device This is the network of workstations implementation of MPICH. P4 (Portable Programs for Parallel Processors) is an older message passing library that was used to implement the MPICH ADI[9]. The ch in ch p4 stands for channel. The ADI is in fact implemented in terms of a simpler channel interface, and the channel interface is implemented in terms of P4. The layering is not strict. The ch p4 device is characterized by the following. ffl P4 runs on Sun SunOS, Sun Solaris, ....

Ralph Butler and Ewing Lusk. Monitors, messages, and clusters: The p4 parallel programming system. Parallel Computing, 20:547--564, April 1994.

....will com4 pare the performance of this software with our work in the performance evaluation section (see section 5) The design of the whole system uses similar techniques as BIP SMP. Finally, one of the first message passing interfaces to manage CLUMPs as a platform is the well known device P4 [3] used for MPICH. P4 provides mechanisms to start multiple processes on machines and uses either message passing or shared memory copies to communicate between these processes. However, the programmer must explicitly select the appropriate library call. P4 also provides several useful reduction ....

Ralph M. Butler and Ewing L. Lusk. Monitors, messages, and clusters : the p4 parallel programming system. Technical report, University of North Florida and Argonne National Laboratory, 1993. http://www-fp.mcs.anl.gov/ lusk/p4/p4paper /paper.html.

....AM as polling for external messages is more expensive than for internal messages. However, Multi Protocol AM is the rst message passing interface to eciently manage CLUMPs. Finally, one of the rst message passing interfaces to manage CLUMPs as a platform is the well known device P4 [2] used for MPICH. P4 provides mechanisms to start multiple processes on hosts and uses 6 either message passing or shared memory copies to communicate between these processes. However, the programmer must explicitly select the appropriate library call. We can also cite implementations of MPI ....

Ralph M. Butler and Ewing L. Lusk. Monitors, Messages, and Clusters : the p4 Parallel Programming System. Technical report, University of North Florida and Argonne National Laboratory, 1993. http://wwwfp. mcs.anl.gov/ lusk/p4/p4-paper/paper.html.

....the upper layer exports MPI s application programming interface (API) while the lower layer contains platform specific message passing libraries. For our testbed environment consisting of a network of workstations connected by an Ethernet, MPICH s lower layer is implemented using the p4 library [4]. To integrate Egida with MPICH, we replaced in the MPICH s upper layer all the send and receive calls to p4 with corresponding calls to Egida s API; modules in Egida in turn invoke the message passing operations of p4. To prevent data copying, MPICH messages and the information piggybacked by ....

R. Butler and E. Lusk. Monitors, Message, and Clusters: the p4 Parallel Programming System.

....is the rst message passing interface to eciently manage CLUMPs. Another project, closely related to our solution, comes from MPICH PM CLUMPs [7] and can be used on commodity SMP nodes. Finally, one of the rst message passing interfaces to manage CLUMPs as a platform is the well known device P4 [2] with MPICH. P4 provides mechanisms to start multiple threads on machines and uses either message passing or shared memory copies to communicate between these threads. However, the programmer must explicitly select the appropriate library call. 4 The parallel simulation application 4.1 ....

Ralph M. Butler and Ewing L. Lusk. Monitors, messages, and clusters : the p4 parallel programming system. Technical report, University of North Florida and Argonne National Laboratory, 1993.

....and exposing the heterogeneous nature of networks and applications. Some communication libraries permit different communication methods to coexist. For example, p4 and PVM on the Intel Paragon use the NX communication library for internal communication and TCP for external communication [5, 17]; p4 supports NX and TCP within a single process, while PVM uses a forwarding process for TCP. In both systems, the choice of method is hard coded and cannot be extended or changed without substantial re engineering. 13 The x kernel [24] and the Horus distributed systems toolkit [30] both ....

R. Butler and E. Lusk. Monitors, messages, and clusters: The p4 parallel programming system. Parallel Computing, 20:547--564, 1994.

....different machines. 7 Several research groups solved the lack of portability by defining vendorindependent application programming interfaces (in the following called portability APIs) and by implementing them on various machines. Some of the more widely used examples are Express [28] P4 [1, 2], PARMACS [3, 20, 21] and PVM [34] A program using such an API was portable between computers. The remaining problem, however, was the diversity of portability APIs, making applications and library software incompatible if they used different APIs. In section 4.2 we will describe some ....

....catered for different problem specific domains. These two aspects, portability and application diversity, made the portability APIs an important step on the way towards the message passing standard. This section presents some of the more influential developments. 4.2. 1 The m4 and p4 macros The p4[1, 2] system grew out of a set of Fortran macros that was developed at Argonne National Laboratory (ANL) for use on a HEP shared memory computer. The original macros, called MonMacs, were processed at compile time by the Unix m4 preprocessing utility, and offered the user a set of monitor functions ....

R. Butler and E. Lusk, Monitors, messages, and clusters: the p4 parallel programming system, Parallel Computing 20(4) (1994) 547--564.

....these issues is not immediately clear and is a subject of ongoing investigation. The flexibility gained through our approach opens a wide range of possibilities. We are especially interested in pursuing two areas. 1. Parallel distributed symbolic computation. When done within the context of p4 [6], MPI [17] or PVM, this work is readily portable to tightly coupled, shared memory machines such as the T3D which have optimized implementations of both MPI and PVM. In this scenario, the applications typically know each other quite well, allowing data exchange performance optimizations. 2. ....

R. Butler and E. Lusk. Monitors, messages, and clusters: the p4 parallel programming system. Parallel Computing, 1994.

No context found.

Ralph Butler and Ewing Lusk. Monitors, messages, and clusters: the p4 parallel programming system. Journal of Parallel Computing, 20(4):547--564, April 1994.

....Within a week, major parts of the initial draft were implemented, and MPICH continued to track the development through to completion. What enabled us to implement the initial draft so quickly was considerable experience with the portable parallel message passing systems Chameleon [15] and p4 [3]. As a result of those experiences, we decided to pursue an aggressive program of relying on automated tools both existing ones and those we would write to help us design, write, test, distribute, and maintain MPICH. In our opinion, this program has been quite successful: we have developed ....

Ralph Butler and Ewing Lusk. Monitors, messages, and clusters: The p4 parallel programming system. Parallel Computing, 20:547--564, April 1994.

....2 Related Work All parallel computing environments that support execution of truly parallel programs (those in which any two processes can communicate with one another) have had to address at least some of the issues that we address with MPD. Parallel programming systems, such as PVM [12] p4 [8], and implementations of MPI such as MPICH [16] and LAM [6] all provide some mechanism for starting and running parallel programs, often with a specialized daemon process. MPD differs from these systems in two ways. First, it is independent of any particular programming library, instead ....

Ralph Butler and Ewing Lusk. Monitors, messages, and clusters: The p4 parallel programming system. Parallel Computing, 20:547--564, April 1994.

....2 Related Work All parallel computing environments that support execution of truly parallel programs (those in which any two processes can communicate with one another) have had to address at least some of the issues that we address with MPD. Parallel programming systems, such as PVM [10] P4 [7], and implementations of MPI such as MPICH [13] and LAM [6] all provide some mechanism for starting and running parallel programs, often with a specialized daemon process. Many systems are intended to manage a collection of computing resources for both single process and parallel jobs; see the ....

Ralph Butler and Ewing Lusk. Monitors, messages, and clusters: The p4 parallel programming system. Parallel Computing, 20:547--564, April 1994.

No context found.

Ralph M. Butler and Ewing L. Lusk. Monitors, messages, and clusters : the p4 parallel programming system. Technical report, University of North Florida and Argonne National Laboratory, 1993. http://wwwfp. mcs.anl.gov/~lusk/p4/p4-paper/paper.html.

No context found.

R. Butler and E. Lusk, Monitors, messages, and clusters: The p4 parallel programming system, Parallel Computing, 20 (1994), 547-564.

No context found.

R. Butler and E. Lusk, Monitors, messages, and clusters: The p4 parallel programming system. Parallel Computing, 20:547-564, April 1994.

No context found.

R. M. Butler and E. L. Lusk, Monitors, messages, and clusters: The p4 parallel programming system. Parallel Computing, 20: 547-564, 1994.

First 50 documents  Next 50

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