C. D. Polychronopoulos, M. Girkar, M. R. Haghighat, C. L. Lee, B. Leung, and D. Schouten. Parafrase-2: An Environment for Parallelizing, Partitioning, Synchronizing and Scheduling Programs on Multiprocessors. In Proc. of the 18th Int'l Conf. on Parallel Processing, pages II:39--48, St. Charles, IL, Aug. 1989.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....means to parallelize programs, written in a serial programming model, for efficient execution on distributed memory multicomputers. The compiler accepts either a sequential FORTRAN 77 or HPF program and produces a parallel program optimized for a target machine. PARADIGM uses Parafrase 2 [51] as a preprocessing platform to parse the sequential program into an intermediate representation, to analyze the code and generate flow, dependence, and call graphs, and to perform transformations such as constant propagation and induction variable elimination. In addition to the traditional ....

....means to parallelize programs, written in a serial programming model, for efficient execution on distributed memory multicomputers. The compiler accepts either a sequential FORTRAN 77 or HPF program and produces a parallel program optimized for a target machine. PARADIGM uses Parafrase 2 [51] as a preprocessing platform to parse the sequential program into an intermediate representation, to analyze the code and generate flow, dependence, and call graphs, and to perform transformations such as constant propagation and induction variable elimination. In addition to the traditional ....

[Article contains additional citation context not shown here]

C. D. Polychronopoulos, M. Girkar, M. R. Haghighat, C. L. Lee, B. Leung, and D. Schouten. Parafrase-2: An Environment for Parallelizing, Partitioning, Synchronizing and Scheduling Programs on Multiprocessors. In Proceedings of the 18th International Conference on Parallel Processing, pages II:39--48, St. Charles, IL, August 1989. 91

....[29] and the compiler should be able to figure out the remaining details which the programmer does not want to worry about. With these goals in mind, we have developed an automatic data partitioning system for multicomputers [23, 25] Our system has been built on top of the Parafrase 2 compiler [57], which is used to provide vital information about the program, such as the parse tree structure, the data dependences, and the control flow information. Our system, in conjunction with a small machine dependent module that provides the cost model for the target machine, can be used on almost any ....

....an overview of the design of Paradigm, and then explain how this problem is resolved in the system. 2.2.3 Overview of the System The structure of the overall system for automatic data distribution is shown in Figure 2.3. Paradigm has been developed as an extension to the Parafrase 2 system [57]. Like the base system, it is organized in the form of passes through the program. Parafrase 2 builds an internal representation for the program, providing information such as the parse tree structure, the data dependences, and the control flow information. It also performs constant propagation ....

C. Polychronopoulos, M. Girkar, M. Haghighat, C. Lee, B. Leung, and D. Schouten. Parafrase-2: An environment for parallelizing, partitioning, synchronizing and scheduling 131 programs on multiprocessors. In Proc. 1989.

....implicit reuse opportunities. This motivates us to design and implement a High Performance Fortran assistant Tool (HPFT) that assists vector compilers in vectorization, reuse exploitation, and synchronization reduction. Related translators developed in last decade can be found such as Parafrase [18], PFC [2, 3] and SUPERB [23] Parafrase is a source to source translator applied to a Fortran 66 or Fortran 77 program. The system can be retargeted to produce code for different classes of concurrent architectures, including register to register and memory to memory vector machines, array ....

....is located in latter section. For a sequential program written in Fortran 77, the HPFT parses the program and then constructs the abstract syntax tree. HPFT then analyzes the data dependence relation and constructs the EDG of the original sequential program. Most of the previous designs [18, 2, 23] use dependence vectors [19] or direction vectors [21] to present the data dependence relation existed in program. In HPFT design, we use EDG to present the data dependence restriction of a loop program. We combine the dependence vector and statement order into an extended dependence vector such ....

C. D. Polychronopoulos, M. Girkar, M. R. Haghighat, C. L. Lee, B. Leng, and D. Schouten, Parafrase-2: an environment for parallelizing, partitioning, synchronizing and scheduling programs on multiprocessors, in Proceedings of International Conference on Parallel Processing, vol. II, Aug. 1989, pp. 39--48.

....can be conflicts at each step of our strategy because of different constraints implied by various loops not being consistent with each other. Such conflicts get resolved on the basis of quality measures. 5 Results We are currently in the process of implementing our approach using Parafrase 2 [9] (a source to source restructurer being developed at the University of Illinois) as our basic tool for identifying the dependencies in a program. Earlier, to determine the applicability of our proposed ideas to real programs, we performed a study using five Fortran programs taken from the Linpack ....

C. Polychronopoulos, M. Girkar, M. Haghighat, C. Lee, B. Leung, and D. Schouten. Parafrase2: An environment for parallelizing, partitioning, synchronizing and scheduling programs on multiprocessors. In Proc. 1989 International Conference on Parallel Processing, August 1989.

....convex Application Compiler was the first on the market [58] it includes several complex analyses, such as interprocedural pointer tracking or array section analysis. The acknowledged origins for this compiler have to be found in several research projects: ptran [72] ParaScope [20] Parafrase 2 [68], and pat [74] Other commercial products, such as the product line developed by Applied Parallel Research, the xlf compiler from ibm (option qipa) or the sgi, foresys, kap, and vast compilers, have since included various interprocedural analyses. But most interprocedural techniques are still ....

....that their potential has not yet been measured. 2. 5 PIPS and Other Research Tools Many research Fortran optimizing tools include some kind of interprocedural analyses, but the closest ones to pips certainly are fiat suif [46, 47] ParaScope [26] and the D system [44] Polaris [11] Parafrase 2 [68], and Panorama [65, 37, 38] FIAT SUIF suif is an intraprocedural compiler for parallel machines developed at Stanford. To enable the parallelization of loops containing procedure calls, fiat [46] an interprocedural engine, has been added on top of it. fiat is a framework which allows ....

C. Polychronopoulos, M. Girkar, M. Haghighat, C. Lee, B. Leung, and D. Schouten. Parafrase-2: An environment for parallelizing, partitionning, synchronizing and scheduling programs on multiprocessors. In International Conference on Parallel Processing, August 1989.

....software for providing service to a diverse user community [15] The large parallel machines have been mainly used by scientists at national laboratories with in house expertise on programming such machines. Developing new parallel software for a given parallel architecture can be quite expensive [3, 19]. In order for parallel processing to be as common as using a workstation, there is a critical need for software tools to support users. In the scientific computing community, most researchers have written sequential programs for their problems of interest. These researchers would benefit ....

....machine, to yield significant speedups. 2. Background Automatic transformation of sequential programs to execute on a parallel machines has been the dream of investigators for several decades ( the dusty deck problem [13] The work was pioneered by Kuck s group at the University of Illinois [5, 13, 19] and Ken Kennedy s group at Rice University [3, 8, 9] Research in this area continues at 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 ....

Polychronopoulos, C., et al. Parafrase-2: An Environment for Parallelizing, Partitioning, Synchronizing and Scheduling Programs on Multiprocessors. in International Conference on Parallel Processing. 1989.

....Research, Inc. Maple is a registered trademark of Waterloo Maple Software, Inc. The PARADIGM project at the University of Illinois addresses this problem by developing an automated means to parallelize and optimize sequential programs for efficient execution on multicomputers. It uses Parafrase 2 [21] as a preprocessing platform to parse the sequential program into an intermediate representation, to analyze the code and generate flow, dependence, and call graphs, and to perform transformations such as constant propagation and induction variable substitution. Some of the other major research ....

Polychronopoulos, C. D., Girkar, M., Haghighat, M. R., Lee, C. L., Leung, B., and Schouten, D. Parafrase-2: An Environment for Parallelizing, Partitioning, Synchronizing and Scheduling Programs on Multiprocessors. In Proceedings of the 18th International Conference on Parallel Processing (St. Charles, IL, Aug. 1989), pp. II:39--48.

....problem of parallelizing programs would be a compiler that takes a sequential program as input and generates an optimal parallel program while preserving the sequential semantics of the original code. There have been some attempts to develop just such a compiler, such as SUIF [18] Parafrase 2 [32], Paradigm [2] javar [6] and javab [7] 6 Reflection in object oriented languages is a mechanism that allows the internals of an object to be examined, and usually includes a facility for invoking a method based on a name and signature. The principle weakness with parallelizing compilers is ....

C. D. Polychronopoulos, M. B. Girkar, M. Resa Haghighat, C. L. Lee, B. P. Leung, and D. A. Schouten. Parafrase--2: An environment for parallelizing, partitioning, synchronizing and scheduling programs on multiprocessors. In Proceedings of the 1989 International Conference on Parallel Processing, volume II, pages 39--48, August 1989.

....than a certain threshold. For an array with fewer dimensions than the processor mesh, the distribution functions of the missing array dimensions are restricted to constant values. The overall structure of Paradigm is shown in Figure 1. Paradigm has been built on top of the Parafrase2 compiler [15], which is used to provide information about the program, such as the parse tree structure, the data dependences, and the control flow information. The internal program representation used by Parafrase2 has been extended to incorporate some additional information, such as a detailed ....

C. Polychronopoulos, M. Girkar, M. Haghighat, C. Lee, B. Leung, and D. Schouten. Parafrase2: An environment for parallelizing, partitioning, synchronizing and scheduling programs on multiprocessors. In Proc. 1989 International Conference on Parallel Processing, August 1989.

.... communication primitives have been implemented as part of a system that, given a sequential program and parameters for data partitioning, statically estimates the communication costs of the program when parallelized on a multicomputer [7] Our system has been built on top of the Parafrase 2 system [15], which is used to set up data structures describing the syntax of the given A B 1 n 1 n n Figure 7: Parallel OneToManyMulticasts in Example 3 program and the dependences in it. In order to generate code implementing the required communication, the compiler has to further specify the actual ....

C. Polychronopoulos, M. Girkar, M. Haghighat, C. Lee, B. Leung, and D. Schouten. Parafrase2: An environment for parallelizing, partitioning, synchronizing and scheduling programs on multiprocessors. In Proc. 1989 International Conference on Parallel Processing, August 1989.

....codes that have been optimized for the modeled architecture. These parallel Cedar Fortran [11] codes have prefetching and forwarding applied via algorithms implemented with the EPG source code instrumentation tools [12] a set of passes implemented within the Parafrase 2 parallelizing compiler [13]. The remainder of this paper is organized as follows. Section 2 presents a brief overview of existing data prefetching and data forwarding schemes and provides additional motivation for the schemes presented and evaluated in the following sections. Section 3 presents the multiprocessor software ....

Polychronopoulos, C. D., et al. Parafrase-2: an environment for parallelizing, partitioning, synchronizing and scheduling programs on multiprocessors. Proc. 1989 International Conference on Parallel Processing. 1989, pp. II-39-48.

....a communication graph is constructed to examine the cost of communication in relation to the flow of data within the program. Phase Phase 1 Phase 2 Fig. 3. Phase decomposition We define a communication graph as the flow information from the dependence graph (generated by Parafrase 2 [16]) weighted by the cost of communication. The nodes of the communication graph correspond to individual statements while the edges correspond to flow dependencies that exist between the statements. As a heuristic, the cost of communication performed for a given reference in a statement is assigned ....

C. D. Polychronopoulos, M. Girkar, M. R. Haghighat, C. L. Lee, B. Leung, and D. Schouten. Parafrase-2: An Environment for Parallelizing, Partitioning, Synchronizing and Scheduling Programs on Multiprocessors. In Proc. of the 18th Int'l Conf. on Parallel Processing, pages II:39--48, St. Charles, IL, Aug. 1989.

.... for Distributed memory General purpose Multicomputers Optimized Parallel Program Multithreading Transformation and Run time Support Analysis Irregular Pattern PILAR Irregular Run time Support Sequential FORTRAN 77 Program HPF Figure 1: PARADIGM Compiler Overview Program Analysis Parafrase 2 [18] is used as a preprocessing platform to parse the sequential program into an intermediate representation and to analyze the code to generate flow, dependence, and call graphs. Various code transformations, such as constant propagation and induction variable substitution, are also performed at this ....

C. D. Polychronopoulos, M. Girkar, M. R. Haghighat, C. L. Lee, B. Leung, and D. Schouten. Parafrase-2: An Environment for Parallelizing, Partitioning, Synchronizing and Scheduling Programs on Multiprocessors. In Proc. of the 18th Int'l Conf. on Parallel Processing, pages II:39--48, St. Charles, IL, Aug. 1989.

....of Perfect Benchmarks [10] codes that have been optimized for the modeled architecture. These parallel Cedar Fortran [11] codes have prefetching and forwarding applied by using the EPG source code instrumentation tools [12] a set of passes implemented within the Parafrase 2 parallelizing compiler [13]. The remainder of this paper is organized as follows. Section 2 presents a brief overview of existing data prefetching and data forwarding schemes and provides additional motivation for the schemes presented and evaluated in the following sections. Section 3 presents the multiprocessor software ....

Polychronopoulos, C. D., et al. Parafrase-2: an environment for parallelizing, partitioning, synchronizing and scheduling programs on multiprocessors. Proc. 1989 International Conference on Parallel Processing. 1989, pp. II-39-48.

....Such conflicts get resolved on the basis of quality measures. 6 Study of Numeric Programs Our approach to automatic data partitioning presupposes the compiler s ability to identify various dependences in a program. We are currently in the process of implementing our approach using Parafrase 2 [17], a source to source restructurer being developed at the University of Illinois, as the underlying tool for analyzing programs. Prior to that, we performed an extensive study using some well known scientific application programs to determine the applicability of our proposed ideas to real ....

C. Polychronopoulos, M. Girkar, M. Haghighat, C. Lee, B. Leung, and D. Schouten. Parafrase-2: An environment for parallelizing, partitioning, synchronizing and scheduling programs on multiprocessors. In Proc. 1989 International Conference on Parallel Processing, August 1989.

....propagation and induction variable substitution. In this paper, a framework is proposed for doing a precise and yet efficient program flow analysis for parallelizing compilers. This framework, which is based on symbolic analysis, is implemented as the flow analysis scheme of Parafrase 2 [39]. Symbolic analysis is an abstract interpretation method in which values of program expressions are represented by symbolic expressions. Parafrase 2 uses symbolic analysis for a wide range of program optimizations such as constant propagation, global forward substitution, detection of ....

C. D. Polychronopoulos, M. B. Girkar, M. R. Haghighat, C. L. Lee, B. P. Leung, and D. A. Schouten. Parafrase-2: An environment for parallelizing, partitioning, synchronizing and scheduling programs on multiprocessors. In Proceedings of the 1989 International Conference on Parallel Processing, St. Charles, IL, August 1989. Penn State.

....not completely independent; namely, imprecise information results in inappropriate transformations. The main problem of compile time analysis is the difficulty of collecting sufficient information in an efficient way to utilize the underlying architecture. We have built a framework for Parafrase 2 [21] that supports a very precise compile time analysis of programs. Within this framework, symbolic analysis is used as an abstract interpretation technique to solve many of the program flow analysis problems in a unified way. Through symbolic analysis, the necessary information about program is ....

C. D. Polychronopoulos, M. B. Girkar, M. R. Haghighat, C. L. Lee, B. P. Leung, and D. A. Schouten. Parafrase-2: An environment for parallelizing, partitioning, synchronizing and scheduling programs on multiprocessors. In Proceedings of the 1989 ICPP, St. Charles, IL, August 1989. Penn State.

....experiments are performed on an SGI Power Challenge. This is a 16 processor MIMD system using MIPS 8000; 8 processors are used to execute parallelized codes 6) 4. 2 Automatic Parallelizing Compiler We first apply three parallelizing compilers, Power Fortran Accelerator (PFA) 22] Parafrase2 [17], and Polaris [15] to our MD code and evaluate their resulting codes. We turn on all optimization switches for PFA and employ the recommended passes or switches for Parafrase2 and Polaris respectively. 6) We explicitly specify when the number of processors we have used is other than 8 ....

C. D. Polychronopoulos et al. Parafrase-2: An environment for parallelizing, partitioning, synchronizing and scheduling programs on multiprocessors. In International Conference for Parallel Processing, 1989.

.... compiler with the automatic parallelization vectorization options [1] For the SGI Challenge, we used the SGI Parallel Fortran Accelerator (PFA) 10] Parafrase 2 and Polaris: The original code was compiled by each experimental source to source parallelizing compiler, Parafrase 2 [7] and Polaris [9]. The parallelized source output from the two parallelizers was compiled again by the back end Fortran compiler of each machine. Both the SGI and the Alliant back end compilers parallelize loops only when explicitly indicated. In the following sections, the incremental effects of compilers, system ....

C. D. Polychronopoulos and et al. Parafrase-2: An environment for parallelizing, partitioning, synchronizing and scheduling programs on multiprocessors. In International Conference for Parallel Processing, 1989.

....implements a kernel level scheduling policy that space and time shares the available processors across the active programs, based on the total machine workload and the feedback provided by the programs through a well established kernel interface. The auto scheduling compiler (based on Parafrase2 [15] and extended to support Open MP like directives [3] performs fine grained multilevel parallelization and exploits both functional and loop parallelism from standard applications written in FORTRAN. Applications are decomposed in multiple levels of parallel tasks. The compiler uses the ....

C. Polychronopoulos, M. Girkar, M. Haghighat, C. Lee, B. Leung and D. Schouten, Parafrase-2: An Environment for Parallelizing, Partitioning, Synchronizing and Scheduling Programs on Multiprocessors, Proc. 1989 ICPP St. Charles, Illinois, 1989.

No context found.

C. D. Polychronopoulos, M. Girkar, M. R. Haghighat, C. L. Lee, B. Leung, and D. Schouten. Parafrase-2: An Environment for Parallelizing, Partitioning, Synchronizing and Scheduling Programs on Multiprocessors. In Proc. of the 18th Int'l Conf. on Parallel Processing, pages II:39--48, St. Charles, IL, Aug. 1989.

No context found.

C. D. Polychronopoulos, M. Girkar, M. R. Haghighat, C. L. Lee, B. Leung, and D. Schouten. Parafrase-2: An Environment for Parallelizing, Partitioning, Synchronizing and Scheduling Programs on Multiprocessors. In Proceedings of the 18th International Conference on Parallel Processing, pages II:39--48, St. Charles, IL, Aug. 1989.

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