D. Loveman (Ed.), `Draft high performance Fortran language specification, version 1.0', Technical Report CRPC-TR92225, Center for Research on Parallel Computation, Rice University, (January 1993).  Home/Search   Document Not in Database   Summary   Related Articles   Check

....to avoid the access to remote data. When non local references are necessary, and in order to amortize their remote access, block transfer of data and data reuse are additional aspects to be considered and optimized to improve the efficiency. The programming model offered by Fortran D [3] and HPF [4] gives the programmer control over how to align and distribute data structures across processors. The compiler has to be able to assign work to processors (the ownership rule is the simplest way to do this [5] and restructure loop nests with the aim of avoiding non local accesses as much as ....

David Loveman (ed.), Draft High Performance Fortran Language Specification Version 1.0, Technical Report TR92-225, CRPC, Rice University, January 1993.

....to avoid the access to remote data. When non local references are necessary, and in order to amortize their remote access, block transfer of data and data reuse are additional aspects to be considered and optimized to improve the efficiency. The programming model offered by Fortran D [3] and HPF [4] gives the programmer control over how to align and distribute data structures across processors. The compiler has to be able to assign work to processors (the ownership rule is the simplest way to do this [5] and restructure loop nests with the aim of avoiding non local accesses as much as ....

David Loveman (ed.), Draft High Performance Fortran Language Specification Version 1.0, Technical Report TR92-225, CRPC, Rice University, January 1993.

....and compiler support for distributed memory machines. On these machines, large data arrays are typically partitioned among local memories of individual processors. The languages supporting such distributed arrays include Fortran D [9, 17] Vienna Fortran [3, 27] and High Performance Fortran (HPF) [8]. Current prototype compilers for these HPF like languages produce Single Program Multiple Data (SPMD) code, combining sequential code for operating on each processor s data with calls to message passing or runtime communication primitives for sharing data with other processors. Reducing ....

D. Loveman (Ed.). Draft High Performance Fortran language specification, version 1.0. Technical Report CRPC-TR92225, Center for Research on Parallel Computation, Rice University, January 1993.

....into equal sized contiguous subarrays and assignment of each subarray to a different processor. Another example would be to assign consecutively indexed array elements to processors in a round robin fashion. These two data distribution schemes are often called BLOCK and CYCLIC data distributions [13], respectively. Researchers have developed a variety of heuristic methods to obtain data mappings that are designed to optimize irregular problem communication requirements [39, 44, 27, 25, 3, 17] The distribution produced by these methods typically results in a table that lists a processor ....

....language extensions could be used for a wide range of languages and compilers such as Vienna Fortran, pC , and HPF. Vienna Fortran, Fortran D and HPF provide a rich set of data decomposition specifications. A definition of such language extensions may be found in Fox et al. [14] Loveman et al. [13], i and Chapman et al. [7, 8] Fortran D and HPF require that users explicitly define how data is to be distributed. Vienna Fortran allows users to write procedures to generate user defined distributions. The techniques described in this paper are being adapted to implement user defined ....

D. Loveman (Ed.). Draft High Performance Fortran language specification, version 1.0. Technical Report CRPCTR92225, Center for Research on Parallel Computation, Rice University, January 1993.

....are not shown in the example. Even if the loop bounds are known at compile time, runtime resolution may be required if complete information about distribution of arrays A, B and C is not available to the compiler. This can happen primarily because of two reasons, use of dynamic data distributions [5] or the compiler not having sufficient support for inter procedural information propagation [8] Again, in this case the compiler can handle this by inserting calls to our runtime primitives, as shown in Figure 2. 4 Regular Section Analysis In this section we discuss the regular section analysis ....

D. Loveman (Ed.). Draft High Performance Fortran language specification, version 1.0. Technical Report CRPC-TR92225, Center for Research on Parallel Computation, Rice University, January 1993.

....major efforts in developing programming language and compiler support for distributed memory machines. Based on the initial work of projects like Fortran D [9, 13] and Vienna Fortran [5] the High Performance Fortran Forum has recently proposed the first version of High Performance Fortran (HPF) [8]. HPF allows programmer to specify the layout of distributed data and specify parallelism through operations on array sections and through parallel loops. Proposed HPF compilers are being designed to produce Single Program Multiple Data (SPMD) Fortran 77 (F77) code with message passing and or ....

D. Loveman (Ed.). Draft High Performance Fortran language specification, version 1.0. Technical Report CRPC-TR92225, Center for Research on Parallel Computation, Rice University, January 1993.

....into equal sized contiguous subarrays and assignment of each subarray to a different processor. Another example would be to assign consecutively indexed array elements to processors in a roundrobin fashion. These two data distribution schemes are often called BLOCK and CYCLIC data distributions [14], respectively. Researchers have developed a variety of heuristic methods to obtain data mappings that are designed to optimize irregular problem communication requirements [37, 41, 29, 27, 3, 20] The distribution produced by these methods typically results in a table that lists a processor ....

....language extensions could be used for a wide range of languages and compilers such as Vienna Fortran, pC , and HPF. Vienna Fortran, Fortran D and HPF provide a rich set of data decomposition specifications. A definition of such language extensions may be found in Fox et al. [15] Loveman et al. [14], and Chapman et al. [8] 9] Fortran D and HPF require that users explicitly define how data is to be distributed. Vienna Fortran allows users to write procedures to generate user defined distributions. The techniques described in this paper are being adapted to implement user defined ....

D. Loveman (Ed.). Draft High Performance Fortran language specification, version 1.0. Technical Report CRPCTR92225, Center for Research on Parallel Computation, Rice University, January 1993.

....Hampton, VA 23681. The authors assume all responsibility for the contents of the paper. Wing Region (subdomain 1) Control Surface (subdomain 2) subdomain 1 subdomain 2 adjacent cells adjacent cells Figure 1: Block structured grid around a wing, showing an interface between blocks Fortran (HPF) [12]. HPF allows programmer to specify the layout of distributed data and specify parallelism through operations on array sections and through parallel loops. Proposed HPF compilers are being designed to produce Single Program Multiple Data (SPMD) Fortran 77 (F77) code with message passing and or ....

....codes in mind, the runtime primitives can also be used to efficiently handle communication for many other types of applications. Regular section move primitives can be used for handling the communication required when a distribution of an array is changed (using the redistributed statement of HPF [12]. They can also be used to handle the communication required for filling ghost cells when the data distribution is cyclic or block cyclic. The primitives can also be used for handling communication in forall loops and array expressions in many regular applications, especially when strides are ....

D. Loveman (Ed.). Draft High Performance Fortran language specification, version 1.0. Technical Report CRPC-TR92225, Center for Research on Parallel Computation, Rice University, January 1993.

....language extensions could be used for a wide range of languages and compilers such as Vienna Fortran and HPF. Vienna Fortran, Fortran D and HPF (evolved from Fortran D and Fortran 90) provide a rich set of data decomposition specifications; a definition of such language extensions may be found in [10, 8]. These languages, as currently specified, require that users explicitly define how data is to be distributed. Fortran D can be used to explicitly specify an irregular inter processor partition of distributed array elements. In Figure 3, we present an example of such a Fortran D declaration. In ....

D. Loveman (Ed.). Draft High Performance Fortran language specification, version 1.0. Technical Report CRPC-TR92225, Center for Research on Parallel Computation, Rice University, January 1993.

No context found.

D. Loveman (Ed.), `Draft high performance Fortran language specification, version 1.0', Technical Report CRPC-TR92225, Center for Research on Parallel Computation, Rice University, (January 1993).

No context found.

D. Loveman (Ed.), `Draft High Performance Fortran language specification, version 1.0', Technical Report CRPC-TR92225, Center for Research on Parallel Computation, Rice University, (January 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