Results 1 - 10 of 17858
Results 1 - 10 of 17858. Search took: 0.039 seconds
|Sort by: date | relevance|
[en] The Double-Layer Master-Slave Model (DMSM) is a suitable hybrid model for executing a workload that consists of multiple independent tasks of varying length on a cluster consisting of multicore nodes. In this model, groups of individual tasks are first deployed to the cluster nodes through an MPI based Master-Slave model. Then, each group is processed by multiple threads on the node through an OpenMP based All-Slave approach. The lack of thread safety of most MPI libraries has to be addressed by a judicious use of OpenMP critical regions and locks. The HPCVL DMSM Library implements this model in Fortran and C. It requires a minimum of user input to set up the framework for the model and to define the individual tasks. Optionally, it supports the dynamic distribution of task-related data and the collection of results at runtime. This library is freely available as source code. Here, we outline the working principles of the library and on a few examples demonstrate its capability to efficiently distribute a workload on a distributed-memory cluster with shared-memory nodes.
[en] HPF (High Performance Fortran) is a data parallel language designed for programming on distributed memory parallel systems. The first draft of HPF1.0 was defined in 1993 as a de facto standard language. Recently, relatively reliable HPF compilers have become available on several distributed memory parallel systems. Many projects to parallelize real world programs have started mainly in the U.S. and Europe, and the weak and strong points in the current HPF have been made clear. In this paper, major data transfer patterns required to parallelize numerical simulations, such as SHIFT, matrix transposition, reduction, GATHER/SCATTER and irregular communication, and the programming methods to implement them with HPF are described. The problems in the current HPF interface for developing efficient parallel programs and recent activities to deal with them is presented as well. (author)
[en] In preparation for building a 250 VUP second generation Fermilab ACP System at Michigan for the CERN L3 experiment, we have developed a semi-automatic pre-compiler to convert event oriented programs into parallel code for the ACP. (orig.)
[en] The kpx is a program analyzer, developed as a common technological basis for promoting parallel processing. The kpx consists of three tools. The first is ktool, that shows how much execution time is spent in program segments. The second is ptool, that shows parallelization overhead on the Paragon system. The last is xtool, that shows parallelization overhead on the VPP system. The kpx, designed to work for any FORTRAN cord on any UNIX computer, is confirmed to work well after testing on Paragon, SP2, SR2201, VPP500, VPP300, Monte-4, SX-4 and T90. (author)