Data parallel programming with parallel virtual machine, Computer Networking

Data Parallel Programming

In the data parallel programming model, main focus is on performing simultaneous operations on a data set. The data set is typically prepared into a common structure, such as an hypercube or array.  Processors work collectively on the similar data structure. Though, every task works on a different partition of the similar data structure. It is more restrictive because not all algorithms can be shows in the data-parallel terms. For the reasons, data parallelism, although main, is not a universal parallel programming paradigm.

Programming with the data parallel model is generally accomplished by writing a program with the data parallel constructs. The constructs can be called to a compiler directives or data parallel subroutine library. Data parallel languages give facilities to states the data decomposition and mapping to the processors. The languages have data distribution statements, which let the programmer to control which data goes on what processor to minimize the quantity of communication between the processors. Directives show how arrays are to be aligned and distributed over the processors and therefore specify agglomeration and mapping. Communication operations are not particular explicitly by the programmer, but are instead inferred by the compiler from the program. Data parallel languages are more suitable for SIMD architecture however some languages for MIMD structure have also been executed. Data parallel approach is more effective for highly regular troubles, but are not very effective for irregular troubles.

The major languages used for this are Fortran 90, High Performance Fortran (HPF) and HPC++. We shall talk about HPF in detail in the next unit. Now, we shall give a short overview of some of the early data parallel languages:

  • Computational Fluid Dynamics: CFD was a FORTRAN-like language developed in the early 70s at the Computational Fluid Dynamics Branch of Ames Research Center for ILLIAC IV machines, a SIMD computer for array processing. The language design was very pragmatic. No attempt was made to cover the hardware peculiarities from the user; in fact, each attempt was made to give the programmers the access and control of all of the hardware to help constructing efficient programs. This language made the architectural types of the ILLIAC IV very apparent to the programmer, but it also contained the seeds of a few practical programming language abstractions for data-parallel programming. In spite of its ad hoc and simplicity machine-dependencies, CFD permits the researchers at Ames to develop a range of application programs that efficiently used the ILLIAC IV.
  • Connection Machine Fortran: Connection Machine Fortran was a later SIMD language developed by Thinking Machines Corporation. Connection Machine Fortran included all of FORTRAN 77, together with the latest array syntax of Fortran 90. It added a variety of machine shows features, but unlike CFD or DAP FORTRAN these appeared as compiler directives rather than usual syntax in Fortran declarations or implementable statements. A main improvement over the earlier languages was that, distributed array dimensions were no longer constrained to precisely fit in the size of the processing element array; the compiler could transparently map dimensions of arbitrary extent across the existing processor grid dimensions. At last the language added an explicitly parallel looping construct called FORALL. While CM Fortran looked syntactically like standard Fortran, the programmer had to be aware of numerous nuances. Like the ILLIAC IV, the Connection.

Machine permitted the Fortran arrays to either be distributed across the processing nodes (called distributed arrays, or CM arrays), or allocated in the memory of the front-end computer (called sequential arrays or front-end arrays). Unlike the control unit of the ILLIAC, the Connection Machine front-end was a ordinary, general-purpose computer--typically a Sun or VAX. But there were still significant restrictions on how arrays could be manipulate, reflecting the two possible homes.

Glypnir ,IVTRAN and *LISP are a little of the other early data parallel languages.

Let us conclude this unit with the introduction of a typical data parallel programming

style known as SPMD.

Posted Date: 3/2/2013 7:31:47 AM | Location : United States

Related Discussions:- Data parallel programming with parallel virtual machine, Assignment Help, Ask Question on Data parallel programming with parallel virtual machine, Get Answer, Expert's Help, Data parallel programming with parallel virtual machine Discussions

Write discussion on Data parallel programming with parallel virtual machine
Your posts are moderated
Related Questions
What are the responsibilities of Application Layer? The Application Layer enables the user, whether human or software, to access the network. It gives user interfaces and suppo

File servers are useful for sharing files through a network. With a file server, the client passes requests for file records over network to file server

To relieve some programming burden, project 5 is a group project that two students can form a group to do and submit one report. In your submitted report, please specify who is you

Which layer is responsible for coordinating communication between systems Ans) The Session layer performs the following: Responsible for establishing and maintaining connections

What is difference between Baseband and Broadband Transmission in ccna? Ans) In a baseband transmission, the entire bandwidth of the cable is consumed by a one signal. In bro

Normal 0 false false false EN-IN X-NONE X-NONE MicrosoftInternetExplorer4

What is a Transaction server? With a transaction server, the client includes remote procedures that reside on the server with an SQL database engine. These remote processes on

- The model was developed by the ISO (International Organisation for Standardisation) in 1984. It is currently considered the primary Architectural model for inter-computer communi