Message Passing Interface Forum. Draft Document for a Standard Message-Passing Interface. Available from Oak Ridge National Laboratory, November 1993.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....as defined in [4] SCL also supports nested parallelism by allowing distributed structures as elements of a distributed structure and by permitting a parallel operation to be applied to each of elements in parallel. An element of a nested array corresponds to the concept of group in MPI [11]. The leaves of a nested array contain any valid sequential data structure of the base computing language. Skeletons supporting nested parallelism are specified in [4] Elementary Skeletons: Parallel Arrays Operators In SCL, we use a set of second order functions as elementary skeletons to ....

....language in the SPP scheme with Fortran. Fortran S can be implemented by transforming Fortran S programs into conventional parallel Fortran programs, that is sequential Fortran augmented with message passing libraries. Currently, we are building a prototype system based on Fortran 77 plus MPI [11] targeted at a Fujitsu AP1000 machine [8] The matrix multiplication example has been translated to Fortran77 plus MPI on an AP1000. Due to the richness of information provided by the Fortran S code, the performance data is very encouraging, as shown in Figure 3 for an array size of 400 Theta ....

Message Passing Interface Forum. Draft Document for a Standard Message-Passing Interface. Available from Oak Ridge National Laboratory, November 1993.

...., DAn = f 1 DA 1 , fn DAn SCL supports nested parallelism by allowing ParArrays as elements of a ParArray and by permitting a parallel operation to be applied to each of elements (ParArrays) in parallel. An element of a nested array corresponds to the concept of group in MPI [7]. The leaves of a nested array contain any valid sequential data structure of the base computing language. The following skeleton gather collects together a distributed array: gather : ParArray index (SeqArray index ff) SeqArray index ff Another pair of configuaration skeletons are split and ....

.... level transformation can be applied to optimise the parallel behaviour of the program, including granularity adjustment, nested parallelism flattening, optimised data distribution and interprocessor communication [2] Currently, we are building a prototype system based on Fortran 77 plus MPI [7] targeted at a Fujitsu AP1000 machine [6] The matrix multiplication example has been translated to Fortran77 plus MPI on an AP1000. Due to the richness of information provided by the Fortran S code, the performance data is very encouraging, as shown in Fig. 4 for an array size of 400 Theta 400. ....

Message Passing Interface Forum. Draft Document for a Standard Message-Passing Interface. Available from Oak Ridge National Laboratory, November 1993.

....machines. However, the problem is exacerbated in a heterogeneous system where the number of possible options is greatly increased and the differences in potential performance more marked. 1 Traditional approaches to parallel programming, such as the use of lower level communication libraries [10] embedded into existing sequential language, are often unsatisfactory and too restrictive for this task. The main drawback of these approaches lies with the low level and unstructured nature of the mechanisms used for co ordinating parallel computation. In these approaches it is difficult to both ....

....entire argument vector on each of the processors. The result can then be computed locally with the result distributed in the same manner as the original matrix. Currently we are building a prototype compiler translator for SPF and SPC which translates SPP(X) programs into Fortran or C plus MPI [10] targeted at the AP1000. In this paper, SPC is used to program the different approaches to the Conjugate Gradient algorithm. 5 3 Performance Guided Resource Organisation Organising the computational resources in a parallel machine is crucial for achieving high performance. However, this is ....

Message Passing Interface Forum. Draft Document for a Standard Message-Passing Interface. Available from Oak Ridge National Laboratory, November 1993.

.... communications, namely broadcasts and multicasts (that is, broadcasts to a specified set of processes) CALL pvmfbcast (group, msgtag, info) CALL pvmfmcast (ntask, tids, msgtag, info) MPI: collective communications In contrast, the forthcoming informal standard for message passing, MPI [6], defines a number of collective communication routines, some of which also include computation to provide global computation operations such as global summation. They are summarised in Table 6. Because of project priorities and the fact that MPI is not yet widely used, we shall initially ....

Message Passing Interface Forum. `Draft Document for a Standard Message-Passing Interface', University of Tennessee, Knoxville, 1993. Can be obtained by anonymous ftp from netlib2.cs.utk.edu, in files mpi/draft1102.ps or mpi/draft1102-tar.Z.

....distributed memory supercomputers support the explicit communication programming model. In such programs, programmers explicitly use communication primitives to perform the communication required in a program. The communication primitives can be high level library routines, such as PVM [60] or MPI [54], or low level communication primitives such as the shared memory operations in the CRAY T3D [59] and the CRAY T3E. Communication patterns in a program with explicit communication primitives can be obtained from the analysis of the communication primitives in the program. Implicit communication: ....

"The Message Passing Interface Forum". Draft Document for a Standard Message Passing Interface, November 1993.

....DAn = f 1 DA 1 , fn DAn SCL supports nested parallelism by allowing ParArrays as elements of a ParArray and by permitting a parallel operation to be applied to each of the elements of the ParArrays in parallel. An element of a nested array corresponds to the concept of a group in MPI [13]. The leaves of a nested array may contain any valid sequential data structure of the base programming language. The following skeleton gather collects together a distributed array: gather : ParArray index (SeqArray index ff) SeqArray index ff The functions split and combine provide another ....

Message Passing Interface Forum. Draft Document for a Standard Message-Passing Interface. Available from Oak Ridge National Laboratory, November 1993.

....programs are usually strictly tied to the platform for which they have been developed. In order to make parallel programming more useful and more efficient, it is necessary to consolidate some of the low level functionality. One such effort is the standardization of a message passing interface [1]. On such a platform, it is easier to develop programming models that can be supported on a variety of architectures, rather than on just one type of machine. The GRIDS [2] parallel programming system for unstructured grid computations, described in this paper is one such effort. Its goal is to ....

Message Passing Interface Forum, DRAFT Document for a Standard Message-Passing Interface, June 30, 1993

.... functions all of which operate on sequences of atomic values (scalars are implemented as sequences of length 1) A Vcode interpreter has been implemented for running Vcode on the Cray C90 and J90, the Connection Machine CM 5, or any machine serial machine with a C compiler [8] We also have an MPI [20] version of VCODE [25] which will run on machines that support MPI, such as the IBM SP 2, the Intel Paragon, or clusters of workstations. The sequence functions in this interpreter have been highly optimized [7, 17] and, for large sequences, the interpretive overhead becomes relatively small ....

Message Passing Interface Forum. Draft document for a standard message-passing interface. Technical Report CS-93-214, University of Tennessee, November 1993.

....e cube router with the torus routing capability. To compare a specialized AAPC architecture with a general message passing system, we are most concerned with the issues of routing and data transfer. For this reason, we did not use a standard message passing interface such as PVM[25] or MPI[18]. These standard interfaces have software overheads resulting from error and protection checking in the operating system, buffer management, and standardized synchronization semantics, which result in copying data. These aspects of message passing libraries are an important issue of research, but ....

THE MESSAGE PASSING INTERFACE FORUM. Draft Document for a Standard Message Passing Interface, November 1993.

....to investigate how to incorporate such features into our model, and especially how lookahead calculations need to be changed for the resulting (more complex) models. In addition, we plan to investigate how to port LAPSE to run under and model the newly emerging MPI (Message Passing Interface [17]) standard on other parallel platforms besides the Paragon. Acknowledgments We are grateful to Jeff Earickson of Intel Supercomputers for his generous assistance in helping us to better understand the Intel Paragon system. ....

Message Passing Interface Forum, The University of Tennessee, Knoxville, Tennessee. Draft: Document for a Standard Message-Passing Interface, November 2, 1993.

....message passing; PVM (Parallel Virtual Machine) which supports message passing and was designed specifically for heterogeneous networks of workstations [33, 34] ffl 3.5 Types of System 3.5.1 Dedicated, Sequential System 3.5. 2 Non dedicated, Sequential System MPI (Message Passing Interface) [35, 36, 37, 38], which is supposed to bring together the main ideas from all of the other message passing systems, to create a single standard set of libraries for message passing in workstations clusters. With backing from many computer manufacturers, this system appears likely to become the standard method for ....

Message Passing Interface Forum. Draft document for a standard Message Passing Interface. Technical report, Netlib, November 1993.

....moves the data using the skeleton for two dimensional rotation rotate2d. Thus each processor passes its portion of A to its left neighbour and passes its portion of B to its north neighbour. 3. 2 Prototype SPF Compiler A prototype compiler for SPF has been built which generates Fortran 77 and MPI [4] code for the Fujitsu AP1000. The resulting code can then be compiled using a Fortran 77 compiler to generate executable code. Since the parallel behaviour of the skeletons in SCL is known a priori, the co ordination primitives can be implemented as pre defined libraries or code templates in the ....

Message Passing Interface Forum. Draft Document for a Standard Message-Passing Interface. Available from Oak Ridge National Laboratory, November 1993.

....e cube router with the torus routing capability. To compare a specialized AAPC architecture with a general message passing system, we are most concerned with the issues of routing and data transfer. For this reason, we did not use a standard message passing interface such as PVM[Sun90] or MPI[MPI93] These standard interfaces have software overheads resulting from error and protection checking in the operating system, buffer management, and standardized synchronization semantics, which result in copying data. These aspects of message passing libraries are an important issue of research, ....

The Message Passing Interface Forum. Draft Document for a Standard Message Passing Interface, November 1993.

No context found.

Message Passing Interface Forum, DRAFT Document for a Standard Message-Passing Interface, June 30, 1993

No context found.

Message Passing Interface (MPI) Forum. "DRAFT Document for a Standard Message-Passing Interface," (University of Tennessee, 1993).

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