J. J. Dongarra, R. A. van de Geijn, and R. C. Whaley. A users' guide to the BLACS. Manuscript. Department of Computer Science, University of Tennessee, Knoxville, TN 37996., 1993.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....memory, vector registers) the level 3 BLAS involve less data movement per floating point operation, leading to faster execution. For a more detailed explanation see, for example, 52] 65] or [69, Ch. 1] Work is ongoing to extend the BLAS standards to support parallelism and sparsity; see [48]. It is important to realise that the BLAS comprise subprogram specifications only; there is freedom in the method used to match the specifications. This freedom is most relevant in the case of the level 3 BLAS, where fast matrix multiplication techniques can be applied. There has been interest ....

Jack Dongarra, Sven Hammarling, and Susan Ostrouchov. BLAS technical workshop. Technical Report CS-95-317, Department of Computer Science, University of Tennessee, Knoxville, November 1995. 21 pp. LAPACK Working Note 109.

....lowest layer s interface cannot be adequately matched to the underlying architecture, then the performance of the entire library will suffer. Historically, highly application specific approaches such as the BLAS and BLACS linear algebra libraries have been the most successful [Dongarra et al. 90, Dongarra et al. 93] These libraries also support simple forms of interoperability. Such a feature is necessary in these libraries because of their highly specialized nature. For instance, because KeLP sticks to managing distributed data and interprocessor communication, it provides no special support for serial ....

J. J. Dongarra, R. A. van de Geijn, and R. C. Whaley. A users' guide to the BLACS. Manuscript. Department of Computer Science, University of Tennessee, Knoxville, TN 37996., 1993.

....2. calculate matrices A and B 3. LU decomposition to factorize A 4. multiply b = B Omega 5. forward backward substitution to solve for flux Aq = b In the following the parallel code only is described. ScaLAPACK requires that matrices are stored in a format known as block cyclic distribution [2], 4] This determines which subset of matrix elements is stored locally on each process. Block cyclic distribution is intended to ensure good load balance in the LU decomposition. ScaLAPACK did not include any subroutines for the creation of matrices. Using knowledge of the blockcyclic ....

J. Chou, J. J. Dongarra, L. S. Ostrouchov, A. P. Petitet, D. W. Walker, and R. C. Whaley, The design and implementation of the scaLAPACK LU, QR, and cholesky factorization routines, Tech. Rep. LAPACK Working Note 80, Department of Computer Science, University of Tennessee, 1994.

....sending Matlab commands to them or by executing routines that, in turn, use them. The key component in this architecture is the MultiMatlab interface module. The interface module, shown in Figure 2, is responsible for initializing an underlying communication layer, such as MPI [2] PVM [3] BLACS [4], or any other package available on the platform, and exposing it to the rest of the system. MEX routines, also shown in the Figure 2, are executables originally written in C or Fortran that may be run directly from Matlab. All parallel functionality in the MultiMatlab system is provided through ....

J. Dongarra and R. C. Whaley. A user's guide to the BLACS. Technical Report CS-95281, Department of Computer Science, University of Tennessee, Knoxville, TN, 1995. Also LAPACK Working Note No.94.

....parallelism according to the data flow paradigm. Communication among WebFlow modules is specified graphically by appropriate communication links. At the time a module is instantiated, all necessary communication links are set up and standard CORBA protocols are used for communication. NetSolve [47, 49] is a client server application designed to solve computational science problems over a network. NetSolve allows the use of network resources for numerical computing by providing convenient access to problem solving techniques (software) in addition to managing computational resources. With the ....

H. Casanova and J. Dongarra. NetSolve Version 1.2: Design and Implementation. Technical Report UT-CS-98-406, Department of Computer Science, University of Tennessee, November 1998.

....etc. Among those libraries we can notice MinPack (Optimization) The subroutines of this library require the user to supply a piece of code that implements the function to be minimized for instance. The current version of the NetSolve software handles user supplied functions in a way described in [13]. The ScaLAPACK library is also a particular case since it is a parallel library and hence necessited the design of a customized NetSolve server. This server is available in NetSolve 1.2 and uses MPI [14] Integrating ScaLAPACK in NetSolve allows a user on a workstation to access MPP systems to ....

....library and hence necessited the design of a customized NetSolve server. This server is available in NetSolve 1.2 and uses MPI [14] Integrating ScaLAPACK in NetSolve allows a user on a workstation to access MPP systems to perform large computation. As outlined in section 1. 3 and detailed in [13], this access is extremely simple and the user is not even be aware that he is using a parallel library. Furthermore, this parallel library is be accessible for C, Fortran, MATLAB, Java programs and even a Java GUI as explained in the following section. 5 1.3 The client interfaces A major ....

[Article contains additional citation context not shown here]

H. Casanova, J. Dongarra, and K. Seymour. NetSolve Users' Guide. Technical report, Department of Computer Science, University of Tennessee, 1998. to appear.

....The NetSolve Users Guide [20] contains general purpose information and examples. Details about the NetSolve agent can be found in [21] Recent developments and applications of NetSolve are described in [22] Lastly, technical details about the current NetSolve implementation are to be found in [23]. 5 Task Farming API 5.1 Basics In this work, we assume that a functional metacomputing environment is available (see Section 2) That environment provides an API that contains two functions: i)submit( to send a request asynchronously for computation and (ii)poll( to poll asynchronously for ....

H. Casanova and J. Dongarra. NetSolve version 1.2: Design and Implementation. Technical Report to appear, Department of Computer Science, University of Tennessee, 1998.

....and communication routines. Thus LAPACK relies on the BLAS #Lawson, Hanson, Kincaid and Krogh, 1979; Dongarra, Du Croz, Hammarlingand Hanson, 1988; Dongarra, Du Croz, Du# and Hammarling, 1990#, particularly the Level 2 and 3 BLAS for computational e#ciency,and ScaLAPACK relies upon the BLACS #Dongarra andWhaley, 1995# for e#ciency of communication andusesasetof parallel BLAS, the PBLAS #Choi, Dongarra, Ostrouchov, Petitet, Walker andWhaley, 1995#, whichthemselves call the BLAS andthe BLACS. LAPACK and ScaLAPACK will runonanymachines for whichthe BLAS andthe BLACS are available. A PVM ##Geist, Beguelin, ....

Dongarra, J. J. andWhaley, R. C. #1995#. A users' guidetothe BLACS v1.0. LAPACK Working Note No.94, Technical Report CS-95-281, Department of Computer Science, UniversityofTennessee, 107 Ayres Hall, Knoxville, TN 37996-1301, USA.

....system has failed or is unavailable. The NetSolve agent is the key to the load balancing strategy, and details about its design can be found in [3] Recent developments and applications of NetSolve are described in [4] Lastly, technical details about NetSolve s implementation are to be found in [5]. Figure 1.2 shows the new software hierarchy of NetSolve version 1.2. 5 Core Functions Client core Mathematica Matlab C Fortran Java NetSolve Client Server daemon Core Functions Numerical Software Server Modules NetSolve Server Network NetSolve Agent Scheduler Data base Core ....

....note the brackets after the problem name. They are necessary because every NetSolve problem is treated as a function defined in Mathematica. The output of that command can be divided to several parts. Let us look at them one by one. The first part is a brief general description of the problem: In[5]: NetSolve[ iqsort[ iqsort: Quicksort Sorts a vector of integers 25 The second part describes the input and output objects, their type and description. The input objects are what the problem requires as input, while the output objects are the result it produces: Input: # 0 : Integer ....

[Article contains additional citation context not shown here]

H. Casanova and Dongarra. NetSolve version 1.2: Design and Implementation. Technical Report to appear, Department of Computer Science, University of Tennessee, 1998.

No context found.

J. J. Dongarra, R. A. van de Geijn, and R. C. Whaley. A users' guide to the BLACS. Manuscript. Department of Computer Science, University of Tennessee, Knoxville, TN 37996., 1993.

No context found.

J. J. Dongarra, R. A. van de Geijn, and R. C. Whaley. Users' guide to the BLACS. Technical report, Department of Computer Science, University of Tennessee, 107 Ayres Hall, Knoxville, TN 37996-1301, USA, 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