C. D. Polychronopoulos, M. B. Girkar, M. R. Haghighat, C. L. Lee, B. P. Leung, and D. A. Schouten. The structure of Parafrase-2: An advanced parallelizing compiler for C and Fortran. In LCPC Workshop, pages 423--453, 1989.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....into two classes, subject to their memory organization, and compare several code transformation issues of parallelizing compilers between these classes of machines. The first class of machines we consider is shared memory multiprocessors with Uniform Memory Access(UMA) architecture. Much work [13, 43, 58, 73] has been done in recent years on automatic parallelization of conventional sequential codes for these machines. This work has led to significant progress and has opened the door to several new strategies currently under study. Parallelization techniques for UMA machines are better understood ....

C. Polychronopoulos, M. Girkar, M. Haghighat, C. Lee, B Leung, and D. Schouten. The structure of parafrase-2: An advanced parallelizing compiler for c and fortran. In Lecture Notes in Computer Science. Springer Verlag, New York, New York, August 1990.

....references, which improves readability and allows the re use of existing sequential code for data parallel applications. The implementation of a ZPL compiler began in April, 1993 and is expected to be completed in the summer of 1993. We are modifying the Parafrase 2 source to source translator [13, 14] to compile ZPL code down to C for a variety of parallel computers. We believe that this system will provide an ideal testbed to experiment with novel compiler optimizations for parallel computers. Acknowledgments. We thank Alex Klaiber, Chuck Koelbel, and Guy Blelloch for providing C , HPF, and ....

C. D. Polychronopoulos, M. B. Girkar, M. R. Haghighat, C. L. Lee, B. P. Leung, and D. A. Schouten. The structure of parafrase-2: an advanced parallelizing compiler for c and fortran. In Workshop on Languages and Compilers for Parallel Computing, pages 423--453.

....done at Illinois as part of the Cedar project, by a group of people of which the author was a member. Similar studies have been done at Stanford [21] Researchers at Stanford [22] and Minnesota [20] plus the PIPS group at Ecole des mines de Paris [13, 14] and the Parafrase 2 group at Illinois [31], have implemented compilers including the same basic transformations found to be important in the Cedar Project. The results have been similar enough to form a general consensus as to which analysis techniques are important to provide in a parallelizing compiler. The Illinois results will be ....

C. Polychronopoulos, M. Girkar, M. Haghighat, C. Lee, B Leung, and D. Schouten. The Structure of Parafrase-2: An Advanced Parallelizing Compiler for C and Fortran. In Lecture Notes in Computer Science. Springer Verlag, New York, New York, August 1990. 137

....machine with a global memory space and noncoherent caches. 1 Introduction Of the three main classes of today s parallel computers, namely, message passing multiprocessors, cache coherent multiprocessors, and noncoherent cache multiprocessors with a global address space, parallelizing compilers [2, 9, 11, 12] have been extensively studied for only the first two. In this paper, we present a preliminary study on the automatic parallelization of Fortran programs for the third class machine. Our translation algorithms were implemented in the Polaris restructurer [2] which was developed by the authors and ....

C. Polychronopoulos, et al., The Structure of Parafrase-2: An Advanced Parallelizing Compiler for C and Fortran, Languages and Compilers for Parallel Computing, MIT Press, 1990

....Also, more recently, some researchers in the Fx project made efforts to generate Put Get for NCC machines. However, our work differs from theirs since they take explicitly parallel code written in Split C or an HPF dialect as the source instead of sequential code. SUIF [16] and Parafrase 2 [44] are typical examples of conventional parallelizing compilers that mainly target bus based shared memory architectures. Although several researchers [1] extended SUIF to retarget distributed memory machines, our work also differs from theirs since their techniques were implemented only for ....

C. Polychronopoulos, M. Girkar, MR. Haghighat, C. Lee, B. Leung, and D. Schouten. The Structure of Parafrase2: An Advanced Parallelizing Compiler for C and Fortran. Workshop on Language and Compilers for Parallel Computing, August 1990.

....tomography 1 Introduction Parallel and distributed processing has been applied to almost every computeintensive task to speed up its execution. To that end, libraries (DSM [27,28] PVM [4] MPI [2,3] etc. and programming environments [9] FPT [13] SUIF [15] Polaris [10] Parafrase 2 [25], etc. have been developed. The longterm goal of all this work is eventually to come to the fully automatic parallelization of code and to make optimal use of the available computing resources. With the advent of sophisticated superscalar processor architectures, multiprocessor parallelism is ....

C. Polychronopoulos, M. Girkar, M. Haghighat, C. Lee, B. Leung, and D. Schouten. The structure of Parafrase-2: An advanced parallelizing compiler for C and Fortran. In Languages and Compilers for Parallel Computing, pages 423--453, Cambridge, MA, USA, 1990. MIT Press.

....all program parallelism, from coarse to fine grain parallelism. In this paper we describe the implementation design of the HTG Visualization Tool 2 (HTGviz) a graphic tool for user interaction with the compilation process. HTGviz is built on top of Parafrase 2 3 parallelizing compiler [16, 17] and uses its internal program representation (HTG) as interaction medium. Parafrase 2 is a research tool for experimenting with program transformations and compiler techniques for parallel computing on shared memory computers. HTGviz is designed as an interface for program parallelization tuning ....

Polychronopoulos, C.D., Gyrkar, M.B., Haghighat, M.R., Lee, C.L., Leung, B.P., Schouten, D.A.: The Structure of Parafrase-2: An Advanced Parallelizing Compiler for C and Fortran. Languages and Compilers for Par. Comp., MIT Press (1990)

....information in the next round of the tuning cycle, in a feed back process that should converge to the optimal solution. All the GPE components are described in more details in section 4. Figure 2 describes the GPE architecture, its components and how they interact. The GPE core is the Parafrase 2 [21, 22] parallelizing compiler developed at the Center for Supercomputing Research and Development at the University of Illinois (UrbanaChampaign, Illinois, USA) Parafrase 2 is a source to source multilanguage parallelizing compiler that provides a reliable portable and efficient research tool for ....

....and showed to the user in an eye pleasant manner. On the other side, user actions in the graphic environment modify the compiler data structure and or execute compiler routines or passes by means of a call back mechanism. 4 GPE Components 4. 1 The parallelizing compiler: Parafrase 2 Parafrase 2 [21, 22] is aimed a source to source multilanguage restructuring compiler. It provides a reliable portable and efficient research tool for experimenting with program transformations and other compiler techniques for parallel shared memory supercomputers. Figure 2 shows the different components of the ....

Polychronopoulos C.D., Gyrkar M.B., Haghighat M.R., Lee C.L., Leung B.P., and Schouten D.A. The Structure of Parafrase-2: An Advanced Parallelizing Compiler for C and Fortran. Languages and Compilers for Parallel Computing, MIT Press (1990).

....transformation techniques. We applied these techniques by hand and automatically to several benchmark programs, and present the recent experiment results in Section 5.2. Finally, we conclude our discussion with future research goals in Section 6. 2 Polaris Restructurer Parallelizing compilers[2, 15, 10, 16] have been extensively studied during the past twenty years or so. Polaris was developed to overcome limitations in the analysis and transformation techniques implemented in other systems and to be robust enough to allow serious experimental studies. As shown in Figure 1, Polaris contains two ....

C. Polychronopoulos, et al., The Structure of Parafrase-2: An Advanced Parallelizing Compiler for C and Fortran, Languages and Compilers for Parallel Computing, MIT Press, 1990

....versus shared memory architectures, are implementation issues, not user issues. One criteria is whether the parallel system is an evolutionary advance on an existing sequential imperative system, or a revolutionary break with the imperative tradition. Vectorizing compilers such as Parafrase 2 [19], which accept sequential code as is and find fine grained parallelism automatically, are at the evolutionary end of this scale. Systems such as PCN [11] in which serial Fortran fragments are nested in a parallel harness constructed in a system based on dataflow and logic programming, represent ....

C. Polychronopoulos, M. Girkar, M.Haghighat, C. Lee, B. Leung and D. Schouten. The Structure of Parafrase-2: an Advanced Parallelizing Compiler for C and Fortran. Languages and Compilers for Parallel Computing, D. Gelernter (ed), A. Nicolau and D. Padua, Pitman, London, 1990.

....into two classes, subject to their memory organization, and compare several code transformation issues of parallelizing compilers between these classes of machines. The first class of machines we consider is shared memory multiprocessors with Uniform Memory Access(UMA) architecture. Much work [13, 43, 58, 73] has been done in recent years on automatic parallelization of conventional sequential codes for these machines. This work has led to significant progress and has opened the door to several new strategies currently under study. Parallelization techniques for UMA machines are better understood ....

C. Polychronopoulos, M. Girkar, M. Haghighat, C. Lee, B Leung, and D. Schouten. The structure of parafrase-2: An advanced parallelizing compiler for c and fortran. In Lecture Notes in Computer Science. Springer Verlag, New York, New York, August 1990.

....Parafrase, program transformations are structured in phases and are always applied where applicable. Batch analysis is performed after each transformation phase to update the dependence information for the entire program. Parafrase2 adds scheduling and improved program analysis and transformations [44]. More advanced interprocedural and symbolic analysis is planned [27] Parafrase 2 uses Faust as a front end to provide interactive parallelization and graphical displays [26] Ptran is also an automatic parallelizer with extensive program analysis. It computes the SSA and program dependence ....

C. Polychronopoulos, M. Girkar, M. Haghighat, C. Lee, B. Leung, and D. Schouten. The structure of Parafrase-2: An advanced parallelizing compiler for C and Fortran. In D. Gelernter, A. Nicolau, and D. Padua, editors, Languages and Compilers for Parallel Computing. The MIT Press, 1990.

....and scheduling results; a code generator that inserts synchronization primitives and performs code optimization for nCUBE I, nCUBE II and INTEL iPSC 860 hypercube machines. A more detailed description of PYRROS is given in [YG92] There are several other systems related to PYRROS. PARAFRASE 2 [Pol90] by Polychronopoulos et al. is a parallelizing compiler system that performs dependence analysis, partitioning and dynamic scheduling on shared memory machines. SCHEDULER by Dongarra and Sorensen [DS87] uses centralized dynamic scheduling for a shared memory machine. KALI by Koelbel and Mehrota ....

C. Polychronopoulos, M. Girkar, M. Haghighat, C. Lee, B. Leung, and D. Schouten, The Structure of Parafrase-2: an Advanced Parallelizing Compiler for C and Fortran, in Languages and Compilers for Parallel Computing, D. Gelernter, A. Nicolau and D. Padua (Eds.), 1990.

....to generate parallel code. These compilers perform dependence analysis to determine a partial order on loop nest iterations, but the analysis techniques work only when array subscripts are linear functions of loop index variables, allowing standard tests like the GCD and Banerjee tests to be used [18, 91, 126]. When complicated subscripts such as those involving indirection arrays are used, dependence analysis fails. With conventional compiler technology, the matrix vector product code shown above is doomed to run sequentially. The Bernoulli project at Cornell is addressing this problem through the use ....

C. D. Polychronopoulos, M. B. Gikar, M. R. Haghighat, C. L. Lee, B. P. Leung, and D. A. Schouten. The structure of Parafrase-2: An advanced parallelizing compiler for C and FORTRAN. In D. Gelernter, A. Nicolau, and D. Padua, editors, Languages and Compilers for Parallel Computing, pages 423--453. Pitman Publishing, 1990.

....memory and runtime overhead. The scheme proposed in this paper provides special syntax for a special class of user defined data distributions, as proposed in Vienna Fortran and HPF [12] On the other hand, in the area of the automatic parallelization, the most outstanding tools we know (Parafrase [20], Polaris [6] are not intended to be a framework for the parallelization of sparse algorithms such as those addressed in our present work. The methods proposed by Saltz et al. for handling irregular problems consists in endowing the compiler with a runtime library [23] to facilitate the search ....

C.D. Polychronopoulos, M.B. Girkar, M.R. Haghighat, C.L. Lee, B.P. Leung and D.A. Schouten, The Structure of Parafrase-2: an Advanced Parallelizing Compiler for C and Fortran, Proceedings 2 nd Workshop on Languages and Compilers for Parallel Computing, pp. 423-453, August 1989.

.... several universities and industrial laboratories [10] Over the past two decades several parallelizing tools have been developed that analyze sequential programs for parallelism (profiling) analyze critical code sections, and help users port existing sequential programs to a parallel platform [5, 8, 10, 20]. Such tools can be extremely valuable in reducing the porting effort and achieving reasonable levels of performance with parallel programs. Current techniques of parallelization, employed by the above tools, are based primarily on analyzing the FORTRAN do loop since this is where one expects to ....

Polychronopoulos, C., et al., The Structure of Parafrase-2: an advanced parallelizing compiler for C and Fortran, in Languages and Compilers for Parallel Computing, D. Gelernter, A. Nicolau, andD. Padua, Editor. 1990,

.... some of these problems through better policies on the reuse of temporary variables and special optimizations, such as loop jamming, which try to recognize related, adjacent loops and merge them into a single outer loop to provide more opportunities for the other optimizations listed above [4, 13]. C makes this very difficult, though, because of the considerable use of pointers and references which hide the effect of many operations and prevent the compiler from making important assumptions allowing code motion and merging. Loop jamming to permit chaining requires C to inline loops ....

C. Polychronopoulos, M. Girkar, M. Haghighat, et al. The Structure of Parafrase-2: An Advanced Parallelizing Compiler for C and Fortran. In D. Gelernter, A. Nicolau, and D. Padua, editors, Languages and Compilers for Parallel Computing, pages 423--453. MIT Press, 1990.

....permit CC and Fortran M programs to be generated from the CODE system. 5. Automatic discovery of task parallelism is the subject of research by a number of investigators, most notably Fran Allen and her group at IBM Research [28,29] and Constantine Polychronopoulos at the University of Illinois [30,31,32,33]. Although this work shows great promise, it is still too early to tell if the compiler technology will be practical and I am aware of no commercial compiler that attempts an aggressive task parallelization. Outside the automatic task parallelization community, not much work on compiler ....

C. D. Polychronopoulos, M. B. Girkar, M. R. Haghighat, C. L. Lee, Bruce P. Leung and D. A. Schouten. The structure of Parafrase-2: an advanced parallelizing compiler for C and Fortran. In D. Gelernter, A. Nicolau and D. Padua, editors, Languages and Compilers for Parallel Computing, pages 114--125. MIT Press, Cambridge Mass., 1990.

....difficulties for distributed memory architectures. The first approach considers the problem of automatic parallelization and scheduling from sequential programs. The emphasis has been in the development of compilers or software tools that will assist in programming parallel architectures [2, 16, 18, 19]. Since message passing architectures require coarse grain parallelism to be efficient, one difficulty is the identification of parallelism especially at the procedural level and another is the scheduling and synchronization. The second approach emphasizes the development of languages with ....

....(1150 seconds) by the actual PYRROS code time. Pyrros(auto) Hand speedup predicted nCUBE p=4 299.0 283.4 288.7 4.0 p=8 155.7 150.5 150.4 7.6 p=16 84.5 81.0 81.4 14.1 p=32 50.0 48.7 47.4 24.3 p=64 33.0 31.8 32.8 35.1 8 Related Work There are several other systems related to PYRROS. PARAFRASE 2 [18] by Polychronopoulos et al. is a parallelizing compiler system that performs dependence analysis, program partitioning and dynamic scheduling on shared memory machines. SCHEDULER by Dongarra and Sorensen [5] uses centralized dynamic scheduling for the shared memory model of computation. The KALI ....

C. D. Polychronopoulos, M.B. Girkar, M. R.Haghighat, C.L. Lee, B.P. Leung, and D.A. Schouten, The Structure of Parafrase-2: an Advanced Parallelizing Compiler for C and Fortran, in D. Gelernter, A. Nicolau and D. Padua (Eds.), Languages and Compilers for Parallel Computing, 1990.

....as Fortran and C) parallel programming systems may be classified using two criteria. The first is whether the parallel system is an evolutionary advance on an existing sequential imperative system, or a revolutionary break with the imperative tradition. Vectorizing compilers such as Parafrase 2 [PGH90], which accept sequential code as is and find fine grained parallelism automatically, are at the evolutionary end of this scale. Systems such as PCN [CT92] in which serial Fortran fragments are nested in a parallel harness constructed in a system based on dataflow and logic programming, ....

C. Polychronopoulos, M. Girkar, M.Haghighat, C. Lee, B. Leung and D. Schouten. The Structure of Parafrase-2: an Advanced Parallelizing Compiler for C and Fortran. Languages and Compilers for Parallel Computing, ed. D. Gelernter, A. Nicolau and D. Padua, Pitman, London, 1990.

....Gradient algorithm, using CRAFT with its data distributions and using two parallelized by hand codes written in C and Fortran with the BRS sparse distribution. 1 Overview In literature, there are several methods for programming parallel machines: automatic parallelizers (Polaris [6] Parafrase [16]) data parallel compilers (VFCS [10] Adaptor [8] and using message passing protocols (PVM [13] MPI [11] PARMACS [9] On the Cray T3D [7] when users program a parallel application, they can choose two main alternatives. In one hand, this parallel machine could be programmed using three ....

C.D. Polychronopoulos, M.B. Girkar, M.R. Haghighat, C.L. Lee, B.P. Leung, D.A. Schouten, The Structure of Parafrase-2: an Advanced Parallelizing Compiler for C and Fortran, Proceedings 2 nd Workshop on Languages and Compilers for Parallel Computing, pp. 423-453, August 1989.

.... The interested reader is referred to any of the following projects which address the basics of distributed memory compilers: the Fortran D compilation system at Rice University [83, 91, 92, 93, 94, 115, 152] the Parafrase 2 and Paradigm compilers at University of Illinois at Urbana Champaign [78, 79, 80, 147, 22, 133], Vienna Fortran and the SUPERB 2 system at the University of Vienna [52, 50, 51, 68] and the SUIF project at Stanford University [16, 145, 157] For the rest of this chapter we concentrate solely on projects whose main purpose is the compilation of Fortran90 constructs. Compass Compilers ....

C. Polychronopoulos, M. Girkar, M. Haghighat, C. Lee, B. Leung, and D. Schouten. The structure of Parafrase-2: An advanced parallelizing compiler for C and Fortran. In D. Gelernter, A. Nicolau, and D. Padua, editors, Languages and Compilers for Parallel Computing. The MIT Press, 1990.

....dynamic scheduling. This approach to scheduling and code generation is appropriate for shared memory architectures with few processors. For message passing architectures centralized scheduling will not perform well because of high control overhead in maintaining a global task queue. PARAFRASE 2 [81] by Polychronopoulos et al. is a parallelizing compiler system that performs dependence analysis, program partitioning and dynamic scheduling on shared memory machines. Recently they have proposed hierarchical task graph (HTG) for modeling parallel computation encapsulating both data and control ....

C. D. Polychronopoulos, M. Girkar, M. Haghighat,C. Lee, B. Leung, and D. Schouten, The Structure of Parafrase-2: An advanced parallelizing compiler for C and Fortran, in Languages and Compilers for Parallel Computing, D. Gelernter, A. Nicolau and D. Padua (Eds.), 1990.

....communication and hoist it from inner loops. HPF and ZPL are similar in the types of parallel operations that they support, though ZPL makes clear to the programmer the execution cost of each operation [22] Considerable research has been devoted to automatically parallelizing Fortran 77 programs [1, 23, 11]. In contrast to the ZPL approach in which the language was designed to facilitate the recognition and exploitation of parallelism, the primary effort for automatically parallelizing compilers is in recognizing, exposing and efficiently exploiting the parallelism hidden in a sequential program. ....

C. D. Polychronopoulos, M. B. Girkar, M. R. Haghighat, C. L. Lee, B. P. Leung, and D. A. Schouten. The structure of Parafrase-2: an advanced parallelizing compiler for C and Fortran. In Workshop on Languages and Compilers for Parallel Computing, pages 423--453, 1990.

.... Projects Agency and the National Science Foundation (NSF CCR 90 07195) 1 Introduction A number of efforts have been made to build parallelizing or vectorizing compilers which attempt to extract parallelism from code written in traditional sequential programming languages such as FORTRAN [26, 5, 27]. Others have concentrated on compiling specialized languages which are inherently concurrent, such as functional languages [18] and Data Flow [7] Unfortunately, such languages are inadequate to model concurrency in a state based, nondeterministic world [1] A third language paradigm consists of ....

C.D. Polychronopoulos, M.B. Girkar, M.R. Haghighat, C.L. Lee, B.P. Leung, and D.A. Schouten. The Structure of Parafrase-2: an Advanced Parallelizing Compiler for C and Fortran. In D. Gelernter, A. Nicolau, and D. Padua, editors, Languages and Compilers for Parallel Computing, pages 423--453. The MIT press, 1990.

....program transformations are structured in phases and are always applied where applicable. Batch analysis is performed after each transformation phase to update the dependence information for the entire program. Parafrase 2 adds scheduling, improved program analysis, and transformations [38]. More advanced interprocedural and symbolic analysis is planned [22] Parafrase 2 uses Faust as a front end to provide interactive parallelization and graphical displays [21] Ptran is also an automatic parallelizer with extensive program analysis [1] It computes the SSA and program dependence ....

C. Polychronopoulos, M. Girkar, M. Haghighat, C. Lee, B. Leung, and D. Schouten. The structure of Parafrase2: An advanced parallelizing compiler for C and Fortran. In D. Gelernter, A. Nicolau, and D. Padua, editors, Languages and Compilers for Parallel Computing. The MIT Press, 1990.

No context found.

C. D. Polychronopoulos, M. B. Girkar, M. R. Haghighat, C. L. Lee, B. P. Leung, and D. A. Schouten. The structure of Parafrase-2: An advanced parallelizing compiler for C and Fortran. In LCPC Workshop, pages 423--453, 1989.

No context found.

C. Polychronopoulos, M. B. Girkar, et al. The Structure of Parafrase-2: An Advanced Parallelizing Compiler for C and Fortran. In Languages and Compilers for Parallel Computing. MIT Press, 1990.

No context found.

C. Polychronopoulos, M. B. Girkar, et al. The Structure of Parafrase-2: An Advanced Parallelizing Compiler for C and Fortran. In Languages and Compilers for Parallel Computing. MIT Press, 1990.

No context found.

C.D. Polychronopoulos, M.B. Girkar, M.R. Haghighat, C.L. Lee, B.P. Leung, and D.A. Schouten. The Structure of Parafrase--2: An Advanced Parallelizing Compiler for C and Fortran. In Languages and Compilers for Parallel Computing, pages 423--453. MIT Press, 1999.

No context found.

Polychronopoulos, C.D., Gyrkar, M.B., Haghighat, M.R., Lee, C.L., Leung, B.P., Schouten, D.A. The Structure of Parafrase-2: An Advanced Parallelizing Compiler for C and Fortran. Languages and Compilers for Par. Comp., MIT Press, 1990.

No context found.

C. D. Polychronopoulos, M. Girkar, M. Haghighat,C. Lee, B. Leung, and D. Schouten, The Structure of Parafrase-2: An advanced parallelizing compiler for C and Fortran, in Languages and Compilers for Parallel Computing, D. Gelernter, A. Nicolau and D. Padua (Eds.), 1990.

No context found.

C. Polychronopoulos, M. Girkar, M. Haghighat, C. Lee, B. Leung, and D. Schouten. The structure of Parafrase-2: An advanced parallelizing compiler for C and Fortran. In D. Gelernter, A. Nicolau, and D. Padua, editors, Languages and Compilers for Parallel Computing. The MIT Press, 1990.

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