T. Brandes. Adaptor: A compilation system for data parallel Fortran programs. Technical report, German National Center for Computer Science (GMD), St. Augustin, Germany, 1994. ftp://ftp.gmd.de/GMD/adaptor/docs/ adaptor.ps.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....and transformation phases of KaHPF are shown in fig. 3. KaHPF Communication Generation Communication Analysis Partitioning Analysis F90 E register (SPMD) Forall Statements Data Distributions HPF Dependence Analysis Fig. 3. Architectural Overview of KaHPF KaHPF is based on the ADAPTOR [3] front end developed by Brandes at the German National Center for Computer Science (GMD) All subsequent analysis and transformation phases operate on the abstract syntax tree built by the frontend. The dependence and partitioning analysis phases use common techniques to perform their tasks. 3.1 ....

T. Brandes. Adaptor: A compilation system for data parallel Fortran programs. Technical report, German National Center for Computer Science (GMD), St. Augustin, Germany, 1994. ftp://ftp.gmd.de/GMD/adaptor/docs/adaptor.ps.

....Performance Fortran Forum was set up to select the most useful functionalities and to standardize the syntax. The initial definition of the new language, Hpf, was frozen in May 1993, and corrections were added in November 1994 [36] Prototype compilers incorporating some Hpf features are available [18, 19, 26, 81, 88, 14]. Commercial compilers from APR [64, 65] DEC [71, 17] IBM [42] and PGI [34, 68] are also being developed or are already available. These compilers implement part or all of the Hpf Subset, which only allows static distribution of data and prohibits dynamic redistributions. This paper deals with ....

....the dividend is positive, we can use the Fortran division. 6 Related work Techniques to generate distributed code from sequential or parallel code using a uniform memory space have been extensively studied since 1988 [22, 70, 89] Techniques and prototypes have been developed based on Fortran [38, 39, 47, 18, 69, 88, 19, 20], C [8, 63, 6, 60, 7, 61] or others languages [74, 75, 58, 66, 57] The most obvious, most general and safest technique is called run time resolution [22, 70, 74] Each instruction is guarded by a condition which is only true for processors that must execute it. Each memory address is checked ....

Thomas Brandes. Adaptor: A compilation system for data parallel fortran programs. Technical report, High Performance Computing Center, German National Research Institute for Computer Science, August 1993.

....parallel processors. Furthermore, new performance measuring and visualization tools have spurred [5] and tools used in sequential machines have been redesigned for parallel machines, like the ARRAYTRACER [6] based on PVM. New standard parallel languages, such as HPF [7] along with the ADAPTOR [8] tool, are used to rewrite sequential application code into parallel code. Thus, workload performance also depends on the efficient implementation of such underlined software systems onto the parallel machine. In spite of the complexity of parallel systems software, the measures metrics by which ....

Brandes, T., Adaptor: A Compilation System for Data Parallel Fortran Programs, Internal Report, GMD, June 1993 ???

....uses only Cray T3E s E register operations for communication, see Figure 3. KarHPFn s program transformations concentrate on the forall statement. node program (SPMD) data parallel) HPF KarHPFn F90 E register F90 Figure 3: Overview of KarHPFn KarHPFn is based on the ADAPTOR front end [Bra94] All subsequent analysis and transformation phases use the Cocktail Toolbox [GE90] to operate on the abstract syntax tree built by that front end. Dependence and partitioning analysis phases use common techniques to perform their tasks. 3.2 Transformation within KarHPFn KarHPFn transforms a ....

Thomas Brandes. Adaptor: A compilation system for data parallel fortran programs. Technical report, German National Center for Computer Science (GMD), St. Augustin, Germany, 1994. ftp://ftp.gmd.de/GMD/adaptor/docs/adaptor.ps.

....products had shown that this goal could be achieved and the High Performance Fortran Forum was set up to select the most useful functionalities and to standardize the syntax. The definition of the new language, hpf, was frozen in May 1993 [13] and the first compilers are already available [7, 8, 11, 36, 24]. This quick delivery was made possible by defining a subset of the language which only allows static distribution of data and prohibits dynamic redistribution. This paper deals with this hpf static subset and shows how simple changes of basis and affine constraints can be used to relate the ....

....It necessary because Fortran remainder is not positive for negative numbers. It was added to insure the semantics of the output code. 6 Related work Techniques to generate distributed code from sequential or parallel code using a uniform memory space have been extensively studied since 1988 [9, 28, 40, 31, 15, 22, 16, 6, 25, 2, 21, 27, 32, 23, 5, 19, 4, 8, 11, 39]. The most obvious, most general and safest technique is called run time resolution [9, 28, 31] Each instruction is guarded by a condition that is true for processors that must execute it. Each memory address is checked before it is referenced to decide whether the address is local and the ....

Thomas Brandes. Adaptor: A Compilation System for Data Parallel Fortran Programs. Technical report, High Performance Computing Center, German National Research Institute for Computer Science, August 1993.

....sequential or shared memory parallel programs that are annotated with directives specifying data decomposition. The compilers for these languages are responsible for partitioning the computation, and generating the communication necessary to fetch values of non local data referenced by a processor [15, 30, 4, 3, 5, 12]. Accessing remote data is usually orders of magnitude slower than accessing local data. This gap is growing Computer Science Division, U.C. Berkeley, CA 94720. Partly supported by ARPA DOD (DABT63 92 C 0026) DOE (DE FG0394ER25206) and NSF (CCR 9210260, CDA 8722788 and CDA9401156) Part of ....

T. Brandes. ADAPTOR: A compilation system for data-parallel Fortran programs. In C. W. Kessler, editor, Automatic parallelization -- new approaches to code generation, data distribution, and performance prediction. Vieweg Advanced Studies in Computer Science, Vieweg, Wiesbaden, Jan. 1994.

....the compiler. The symbolic analysis ability of the compiler maintains the same level of performance when the number of processors is not known at compile time. 5 Related Work There are several groups which have looked at the problem of compiling HPF like languages on distributed memory machines [11, 23, 19, 3, 4, 15, 18]. Our work has also benefited from other early projects like Kali [14] Id Nouveau [17] and Crystal [16] The Fortran D compiler [11] performs several optimizations like message vectorization, using collective communication, and exploiting pipeline parallelism. It also performs analysis to ....

....I O and handling out of core programs [20] The PARADIGM compiler [19, 2] is targeted to Fortran 77 programs and provides an option for automatic data partitioning for regular computations. It also supports exploitation of functional parallelism in addition to data parallelism. The ADAPTOR system [4] supports the HPF subset and performs optimizations both for handling Fortran 90 array constructs and for improving cache locality, in addition to those for reducing communication costs. The ADAPTOR determines the communication schedules at runtime. The SUIF compiler [1] performs loop ....

T. Brandes. ADAPTOR: A compilation system for data-parallel Fortran programs. In C. W. Kessler, editor, Automatic parallelization -- new approaches to code generation, data distribution, and performance prediction. Vieweg Advanced Studies in Computer Science, Vieweg, Wiesbaden, January 1994.

....sequential or shared memory parallel programs that are annotated with directives specifying data decomposition. The compilers for these languages are responsible for partitioning the computation and generating the communication necessary to fetch values of non local data referenced by a processor [72, 123, 20, 5, 21, 68]. Accessing remote data is usually orders of magnitude slower than accessing local data, for the following reasons. It is getting increasingly cost effective to build multiprocessors from commodity hardware components and system software. Most current generation CPU s are well beyond the 100 ....

....references and their relation with this thesis are given in specific chapters. 6. 1 Communication scheduling The results in Chapter 2 follow (and generalize) a large number of preliminary results in communication optimization, such as message vectorizing, coalescing, and redundancy elimination [72, 123, 20, 5, 21, 68]. Message vectorization is the technique of hoisting communication of array elements or sections out of loops to produce a single large message. Typically this works locally on single loop nests [123, 72, 67, 83, 68] As compilers were applied to more complex code, eliminating redundant ....

T. Brandes. ADAPTOR: A compilation system for data-parallel Fortran programs. In C. W. Kessler, editor, Automatic parallelization -- new approaches to code generation, data distribution, and performance prediction. Vieweg Advanced Studies in Computer Science, Vieweg, Wiesbaden, Jan. 1994.

....Performance Fortran Forum was set up to select the most useful functionalities and to standardize the syntax. The initial definition of the new language, Hpf, was frozen in May 1993, and corrections were added in November 1994 [36] Prototype compilers incorporating some Hpf features are available [18, 19, 26, 81, 88, 14]. Commercial compilers from APR [64, 65] DEC [71, 17] IBM [42] and PGI [34, 68] are also being developed or are already available. These compilers implement part or all of the Hpf Subset, which only allows static distribution of data and prohibits dynamic redistributions. This paper deals with ....

....division. 34 Ancourt et al. A Linear Algebra Framework: 6 Related work Techniques to generate distributed code from sequential or parallel code using a uniform memory space have been extensively studied since 1988 [22, 70, 89] Techniques and prototypes have been developed based on Fortran [38, 39, 47, 18, 69, 88, 19, 20], C [8, 63, 6, 60, 7, 61] or others languages [74, 75, 58, 66, 57] The most obvious, most general and safest technique is called run time resolution [22, 70, 74] Each instruction is guarded by a condition which is only true for processors that must execute it. Each memory address is checked ....

Thomas Brandes. Adaptor: A compilation system for data parallel fortran programs. Technical report, High Performance Computing Center, German National Research Institute for Computer Science, August 1993.

....erlebacher which is from ICASE. For our multiprocessor fusion experiments, we used lumm, shallow, hydflo, and helmholtz which are explicitly parallel programs written in a subset HPF (High Performance Fortran) These four programs are part of the ADAPTOR source to source translator from SCAI GMD [11, 12]. Table 3 summarizes important features of both uniprocessor and multiprocessor programs. In this table, we list the number of lines, loops, the number of fusion graphs for each program (one per candidate procedure) the number of nodes in each fusion graph, and the number of loop nests that our ....

Brandes, T. (1994) ADAPTOR: A compilation system for data-parallel Fortran programs. In Automatic Parallelization - New Approaches to Code Generation, Data Distribution, and Performance Prediction (ed. Kessler, C. W.). Vieweg, Wiesbaden.

....and prototype compilers for DMMs was published during the last few years. The Vienna Fortran Compiling System [13] and Fortran D [12] are early developments of data parallel languages on which hpf is based. The Fortran D prototype allows only one dimension to be distributed, and no I O. Adaptor [2] is a pragmatic and efficient public domain software which implements some features of hpf and which was initially based on CM Fortran. Many vendors have announced commercial compilers: Applied Parallel Research, Digital Equipment, Pacific Sierra Research, the Portland group: These products ....

T. Brandes. Adaptor: A compilation system for data parallel fortran programs. Technical report, High Performance Computing Center, German National Research Institute for Computer Science, Aug. 1993.

....products had shown that this goal could be achieved and the High Performance Fortran Forum was set up to select the most useful functionalities and to standardize the syntax. The definition of the new language, Hpf, was frozen in May 1993 [29] and the first compilers are already available [15, 16, 22, 63, 50]. This quick delivery was made possible by defining a subset of the language which only allows static distribution of data and prohibits dynamic redistribution. This paper deals with this Hpf static subset and shows how simple changes of basis and affine constraints can be used to relate the ....

.... A Linear Algebra: Submitted to Scientific Programming 47 6 Related work Techniques to generate distributed code from sequential or parallel code using a uniform memory space have been extensively studied since 1988 [18, 54, 71] Techniques and prototypes have been developed based on Fortran [31, 32, 37, 15, 53, 70, 16, 17], C [8, 49, 6, 47, 7] or others languages [57, 58, 45, 51, 44] The most obvious, most general and safest technique is called run time resolution [18, 54, 57] Each instruction is guarded by a condition that is true for processors that must execute it. Each memory address is checked before it is ....

Thomas Brandes. Adaptor: A compilation system for data parallel fortran programs. Technical report, High Performance Computing Center, German National Research Institute for Computer Science, August 1993. Ancourt et al., A Linear Algebra: : : , Submitted to Scientific Programming 53

....of the system is that the base distributed memory parallelization tool (DMP) does not attempt code restructuring transformations to enhance the detection or generation of parallel code. Adaptor Adaptor is an HPF like compilation system that is structured very much like the CM Fortran compiler [33, 34, 35]. It only takes advantage of parallelism present in the explicitly parallel HPF constructs. It has no feature for automatic parallelization. In addition to batch compilation support, Adaptor also has a graphical user environment which allows the user to assist in directing the translation process. ....

T. Brandes. Adaptor: A compilation system for data parallel fortran programs. In Christoph W. Kessler, editor, Automatic Parallelization --- New Approaches 143 to Code Generation, Data Distribution, and Performance Prediction. Vieweg, Wiesbaden, 1994.

....goal could be achieved and the High Performance Fortran Forum was set up to select the most useful functionalities and to standardize the syntax. The definition of the new language, hpf, was frozen in May 1993 [13] and the first compilers are already available [7, 8, 11, 35, 24]. This quick delivery was made possible by defining a subset of the language which only allows static distribution of data and prohibits dynamic redistribution. This paper deals with this hpf static subset and shows how simple changes of basis and affine constraints can be used to relate the ....

....It necessary because Fortran remainder is not positive for negative numbers. It was added to insure the semantics of the output code. 6 Related work Techniques to generate distributed code from sequential or parallel code using a uniform memory space have been extensively studied since 1988 [9, 28, 39, 30, 15, 22, 16, 6, 25, 2, 21, 27, 31, 23, 5, 19, 4, 8, 11, 38]. The most obvious, most general and safest technique is called run time resolution[9, 28, 30] Each instruction is guarded by a condition that is true for processors that must execute it. Each memory address is checked before it is referenced to decide whether the address is local and the ....

Thomas Brandes. Adaptor: A Compilation System for Data Parallel Fortran Programs. Technical report, High Performance Computing Center, German National Research Institute for Computer Science, August 1993. ftp ftp.gmd.de, gmd/adaptor/docs.

....are meant as input into NAS s work on HPF (WP 4) 2 Description of prototype HPF like systems 2. 1 Adaptor Adaptor (Automatic Data Parallelism Translator) is an HPF system developed at the GMD for translating data parallel Fortran programs into equivalent Fortran 77 message passing programs [6, 7]. Most features of Connection Machine Fortran (CMF) and many features of HPF are supported. The central idea of the automatic translation is to distribute the large data structures like arrays among the available processors. This should be performed in such a way that most operations could be done ....

T. Brandes. Adaptor: A Compilation System for Data Parallel Fortran Programs. Proc. of AP'93. Saarbrucken, March 1993.

....and recommendations for the HPF compiler designers. 2 Description of prototype HPF like systems 2. 1 Adaptor Adaptor (Automatic Data Parallelism Translator) is an HPF system developed at the GMD for translating data parallel Fortran programs into equivalent Fortran 77 message passing programs [6, 7]. Most features of Connection Machine Fortran (CMF) and many features of HPF are supported. The parallel program can be written in such a way that it could be developed on a serial machine and is also suitable for vector machines or parallel machines with shared memory. Many features supported ....

T. Brandes. Adaptor: A Compilation System for Data Parallel Fortran Programs. Proc. of AP'93. Saarbr ucken, March 1993.

....parallel versions of the application programs have been translated to message passing programs by using the Adaptor tool. Adaptor (Automatic Data Parallelism Translator) is a system developed at the GMD for compiling data parallel Fortran programs to equivalent Fortran 77 message passing programs [Bra93]. Only some small modifications were necessary to get the codes successfully translated by Adaptor with its current restrictions. Some special directives of Adaptor have been used to identify private arrays and parallel loops without communication in the IFS code. The runtime system of the Adaptor ....

T. Brandes. Adaptor: A Compilation System for Data Parallel Fortran Programs. In International Workshop on Automatic Distributed Memory Parallelization, Automatic Data Distribution and Automatic Parallel Performance Prediction, Saarbrucken, March 1993. Vieweg.

No context found.

T. Brandes. Adaptor: A compilation system for data parallel Fortran programs. Technical report, German National Center for Computer Science (GMD), St. Augustin, Germany, 1994. ftp://ftp.gmd.de/GMD/adaptor/docs/ adaptor.ps.

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