F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the PTRAN analysis system for multiprocessing. In Proceedings of the 1987.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....The initial algorithms typically require reaching definitions and loop invariant expressions. The algorithms are conservative and find simple linear induction variables. The PTRAN compiler uses an optimistic approach and assumes variables in loops are induction variables until proven otherwise [2]. Gerlek, Stoltz, and Wolfe present a demand driven SSA approach for detecting general induction variables by identifying strongly connected components in the SSA graph [11] Gerlek et al. present a lattice for classifying different types of induction variables. Haghighat and Polychronopoulos also ....

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the PTRAN analysis system for multiprocessing. Journal of Parallel and Distributed Computing, 5(5):617--640, Oct. 1988.

....calls, or restricted ourselves to individual procedures while analyzing real application codes with Paradigm. Clearly, there is a need to develop techniques for interprocedural analysis. A number of researchers have worked on this problem for improving the effectiveness of parallelizing compilers [9, 10, 76, 49, 1]. Those ideas need to be extended to allow determination of constraints and their quality measures across procedure boundaries, and to summarize such information for data accessed in various procedures. ffl Redistribution of data : Currently, Paradigm assigns a fixed distribution to each array, ....

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the PTRAN analysis system for multiprocessing. Journal of Parallel and Distributed Computing, 5:617-- 640, 1988.

....These issues become important when creating a complete compiler implementation, in which the described techniques are integrated into a user friendly tool. Several compiler research infrastructures have played pioneering roles in this regard. Among them are the Parafrase [1, 2] PFC [3] PTRAN [4], ParaScope [5] Polaris [6] and SUIF [7] compilers. The following paragraphs describe a number of issues that have to be addressed by such infrastructures. An adequate compiler internal representation of the program must be chosen, the large number of transformation passes need to be put in the ....

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante", "An overview of the PTRAN analysis system for multiprocessing," Proc.oftheInt'l Conf. on Supercomputing, 1987, pages 194--211.

....lower level code using a code template. Each construct can be lowered independently. The expander makes no attempt to generate tight, fast code. Instead, later optimizations tune and customize the result to fit the surrounding code. For example, when expanding an array instruction representing a[3][4], the expander does not fold any constants; it is done later by a constant folding and propagation phase. Because high SUIF includes low SUIF as a subset, we have built many compiler passes that accept high SUIF by extending passes that accepts low SUIF. Most of our scalar optimizations fit this ....

....data flow relationships, data dependences, and control dependences must be implicitly represented as part of the flow values. While this focus on CFG suits many optimizations, a uniform representation for all dependencies can benefit many optimizations. For example, several powerful techniques [4, 8, 12, 26, 29, 73, 89] are formulated on graphs in which control dependence and data flow are uniformly represented. Although Sharlit can support these techniques through flow values abstractions, it would be interesting to see how Sharlit can support these techniques more directly as Sharlit can do with sparse ....

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the PTRAN analysis system for multiprocessing. Tech. Rept. Technical Report RC 13115, IBM, Sept, 1987.

....to obtain parallel code semantically equivalent to the original sequential one. DO loops offer a great amount of potential parallelism in numerical programs. Several parallelizing compilers such as Parafrase I and II [KKLW80, PGHL89] from Illinois, PFC from Rice University [AlKe87] IBM PTRAN [ABCC88], KAP from Kuck and Associates, VAST from Pacific Sierra, to name a few, have been developed over the years. Such parallelizing compilers rely on data dependence analysis. In loop partitioning it is important to analyze data dependences between statements involving array references within the ....

F. Allen, M. Burke, P. Charles, R. Cytron and J. Ferrante, "An Overview of the PTRAN Analysis System for Multiprocessing", Journal of Parallel and Distributed Computing, October 1988.

....be partitioned into relatively independent pieces, and therefore the operations performed on these pieces can proceed in parallel. It is this property of scientific programs that has led to impressive results in the development of vectorizing and parallelizing compilers [Allen and Kennedy 1987; Allen et al. 1988; Wolfe 1989] More recently, this property has been exploited by researchers investigating methods for automatically generating parallel programs for SPMD (Single Program, Multiple Data) execution on distributed memory machines. In this article, we address the problem of supporting programs that ....

Allen, F., Burke, M., Charles, P., Cytron, R., and Ferrante, J. 1988. An overview of the PTRAN analysis system for multiprocessing. J. Parallel Distrib.

....which satisfy the CTL e formula. The compiler can then safely apply to these nodes the optimization or parallelization transformation described by the formula. 6. 1 Graph representations of programs A compiler can apply many well understood optimizing and parallelizing program transformations [2, 1, 20, 21, 34, 32, 43, 57, 59, 80, 82] that can significantly reduce the execution time of a program. The effectiveness of these transformations depends on both the target machine architecture and the semantics of the programming language constructs which are affected by these transformations. Although parallelizing compilers have ....

....) Initialization items . APTList[0] MI.PREFIXimage = MACROptr ) malloc ( sizeof ( MACROstruct ) dstrinit ( APTList[0] MI.PREFIXimage rep ) APTList[0] MI.PREFIXimage label [ 0 ] 1 ; APTList[0] MI.PREFIXimage label [ 1 ] 1 ; APTList[0] MI.PREFIXimage label [ 2 ] = 1 ; rep( 3) return(0) MacroBody = repkeyword StmtList ; Rep rep: rep( 2) MacroBody = repkeyword LexParameter ; Rep rep: rep( 2) StmtList = StmtList StmtLine ; Rep rep: rep( 1) rep( 2) StmtList = StmtLine ; Rep 128 rep: ....

F.E. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the ptran analysis system for multiprocessing. Journal of Parallel and Distributed Computing, 5(October):617--640, 1988.

....like arrays and lists, require new processing techniques, different from those for sets. There has been a good deal of work on data parallel techniques and languages that are all or in part data parallel: there exist parallel extensions of FORTRAN, like High Performance Fortran [43] and PTRAN [3, 24], parallel extensions of C, like C [69] C [45] of Lisp, like CM List [39] and Paralation Lisp [60, 51] and applicative parallel programming languages, like NESL [8, 10, 11] Sisal [31, 62, 30] Crystal [21] Proteus [54, 34, 59] and Data parallel ML [32, 37, 38] None of these have been ....

.... : fk i t t1 Theta : Theta t k where f i : t t i , i = 1; n if f then g1 else g2 t1 t2 where f : t1 [N] g i : t1 t2 , i = 1; 2 [ unit [s] empty sequence singleton unit [unit ] returns [hi] s] Theta [s] s] concatenation length [s] N] e.g. length( a; b; c] [3] select [N Theta s] s] see text zip [s] Theta [s 0 ] s Theta s 0 ] enumerate [s] N] bmRoute [s 0 ] Theta [N] Theta [s] s] see text sbmRoute [s 0 ] Theta [N] Theta [s] Theta [N] s] see text while(p; f) t t 0 where p : t [N] f : t t Figure 9: The ....

Frances Allen, Michael Burke, Philippe Charles, Ron Cytron, and Jeanne Ferrante. An overview of the PTRAN analysis system for multiprocessing. Journal of Parallel and Distributed Computing, 5:617--640, 1988.

....present their own set of challenges to the compiler. A large and active research community is attacking the problem of compiling code that makes effective use of these machines. Over the past decade, progress has been made in detecting parallelism and in improving performance on parallel programs [2, 3, 21]. The key tool that compilers use to improve performance on multiprocessor systems is static program analysis. Restructuring compilers perform data flow analysis, data dependence analysis, and control dependence analysis to understand the sequence of memory references and computation that can ....

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the PTRAN analysis system for multiprocessing. In Proceedings of the First International Conference on Supercomputing. Springer-Verlag, Athens, Greece, June 1987.

....data decomposition, the compiler performs process decomposition by analyzing the program and specializing it, for each processor, to the data that resides on that processor. Thus, our approach to process decomposition is data driven rather than program driven as are more traditional approaches[1, 19]. 1.1 System Overview Our compiler takes a sequential program and a domain decomposition and generates C code for the Intel iPSC 2. The system generates code for each process based on the data residing in that process. In this section, we present an overview of our system which includes two code ....

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the ptran analysis system for multiprocessing. J. of Parallel and Distributed Computing, 5:617--640, 1988.

....that the arrays can be partitioned into relatively independent pieces and therefore the operations performed on these pieces can proceed in parallel. It is this property of scientific programs that has led to impressive results in the development of vectorizing and parallelizing compilers [Allen et al. 1988; Allen and Kennedy 1987; Wolfe 1989] More recently, this property has been exploited by researchers investigating methods for automatically generating parallel programs for SPMD (Single Program, Multiple Data) execution on distributed memory machines. In this paper, we address the problem of ....

Allen, F., Burke, M., Charles, P., Cytron, R., and Ferrante, J. 1988. An overview of the PTRAN analysis system for multiprocessing. J. of Parallel and Distributed Computing 5, 5 (Oct), 617--640.

....The high cost of machine specific software development and the lack of highly portable parallel programming languages is a major obstacle in their rewriting. A common approach has been to employ parallelizing compilers which automatically extract the parallelism present in sequential applications [21,3,2]. Most of the concurrency present in these programs is expressed in the form of loops and considerable efforts have been devoted to loop parallelization ( 22,17,8,16,20] to name a few) The problem studied in this paper is optimum and nearly optimum cyclic scheduling. More specifically we propose ....

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante, An overview of the PTRAN analysis system for multiprocessing, Journal of Parallel and Distributed Computing, 5 (1988), pp. 617--640.

....this optimization, and therefore the most important induction variable candidates, are array address expressions in inner loops. Techniques for detecting and classifying linear induction variables have a long history and are well known [Aho, Sethi, and Ullman 1986; Allen, Cocke, and Kennedy 1981; Allen, Burke, Charles, Cytron, and Ferrante 1988]. This research was supported by ARPA under grant F3062 92 C 135. A preliminary paper, Beyond Induction Variables , appeared in the Proceedings of the ACM SIGPLAN Conference on Programming Language Design and Implementation (June 1992) Authors address: Oregon Graduate Institute of Science ....

....however, since the other variable on the right hand side is not known to be in any family of induction variables. In the PTRAN compiler, such cases are solved by an optimistic dataflow technique which initially assumes all variables are linear induction variables until a contradiction exists [Allen, Burke, Charles, Cytron, and Ferrante 1988]. Allen et al. present a comprehensive treatment of strength reduction by recognizing more general linear cases [Allen, Cocke, and Kennedy 1981] Tu and Padua have used our SSA technique for linear induction (and monotonic) variable detection in order to perform symbolic analysis of array bounds ....

Allen, F., Burke, M., Charles, P., Cytron, R., and Ferrante, J. 1988. An overview of the PTRAN analysis system for multiprocessing. Journal of Parallel and Distributed Computing 5, 5 (Oct.), 617--640.

....technology. The idea of providing early performance feedback in the upper, compile time layer of the performance prediction hierarchy has been applied in many parallel programming environments, in which parallelization is performed either explicitly [12, 35, 26] or through restructuring [2, 11, 20, 36], and data partitioning [3, 22, 27, 32, 34, 37, 39, 40, 47] Due to the procedure oriented (e.g. fork2 join) source level paradigm, the computational task graph has a series parallel structure, which permits an efficient prediction, typically implemented through a recursive reduction scheme, ....

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante, "An overview of the PTRAN analysis system for multiprocessing," in Proc. 1st ACM Int. Conf. on Supercomputing, 1987, pp. 194--211.

....with strength reduction; the most common candidates for this optimization, and therefore the most important induction variable candidates, are array address expressions in inner loops. Techniques for detecting and classifying linear induction variables have a long history and are well known [1, 2, 3]. Many modern compilers now include advanced loop transformations such as loop distribution and loop interchanging [17] that have proven useful on a wide variety of systems, ranging from uniprocessor workstations to vector multiprocessors and massively parallel processors. These transformations ....

....however, since the other variable on the right hand side is not known to be in any family of induction variables. In the PTRAN compiler, such cases are solved by an optimistic data flow technique which initially assumes all variables are linear induction variables until a contradiction exists [2]. Allen et al. present a comprehensive treatment of strength reduction by recognizing 52 more general linear cases [3] Tu and Padua have used our SSA technique for linear induction (and monotonic) variable detection in order to perform symbolic analysis of array bounds and subscript expressions ....

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the PTRAN analysis system for multiprocessing. Journal of Parallel and Distributed Computing, 5(5):617--640, October 1988.

.... Introduction Compiler techniques for automatic detection of parallelism are often described in the literature and implemented as source to source transformations [Wolf82, Zima91, BENP93] These implementations are usually experimental translators for the parallelization and vectorization of loops [ABCC88, PGHL89, BEKG94]. Low level operations, such as those needed to compute addresses of arrays and scalar parameters, are usually hidden for these source to source translators which only represent internally highlevel constructs. Most compilers for high performance machines include two separate translators: a ....

F. Allen, M. Burke, P. Charles, R. Cytron and J. Ferrante. An overview of the PTRAN analysis system for multiprocessing. Journal of Parallel and Distributed Computing.

....algorithm that can be modeled as the solution of an incremental intraprocedural problem. Their algorithm does not support recursion but does support static call graph cycles, and the propagation of returned constants. No experimental results are given. Allen, Burke, Charles, Cytron, and Ferrante [1] describe a flow insensitive interprocedural algorithm for constant propagation of formal parameters, based on the bindings of formal parameters. It detects immediate argument constants and the transitive effect of passing their corresponding parameters as arguments. They do not handle returned ....

Frances Allen, Michael Burke, Philippe Charles, Ron Cytron, and Jeanne Ferrante. An overview of the ptran analysis system for multiprocessing. Journal of Parallel and Distributed Computing, 5(5):617--640, 1988.

....extensive program analysis. It computes the SSA and program dependence graphs, and performs constant propagation and interprocedural analysis [22] Ptran introduces both task and loop parallelism, but currently the only other program transformations are variable privatization and loop distribution [1]. Sigmacs, a programmable interactive parallelizer in the Faust programming environment, computes and displays call graphs, process graphs, and a statement dependence graph [26, 48] In a process graph each node represents a task or a process, which is a separate entity running in parallel. The ....

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the PTRAN analysis system for multiprocessing. In Proceedings of the First International Conference on Supercomputing. Springer-Verlag, Athens, Greece, June 1987.

.... Introduction Compiler techniques for automatic detection of parallelism are often described in the literature and implemented as source to source transformations [Wolf82, Zima91, BENP93] These implementations are usually experimental translators for the parallelization and vectorization of loops [ABCC88, PGHL89, BEKG94]. Low level operations, such as those needed to compute addresses of arrays and scalar parameters, are usually hidden for these source to source translators which only represent internally highlevel constructs. Most compilers for high performance machines include two separate translators: a ....

F. Allen, M. Burke, P. Charles, R. Cytron and J. Ferrante. An overview of the PTRAN analysis system for multiprocessing. Journal of Parallel and Distributed Computing. Vol.5, 1988.

....however, since the other variable on the right hand side is not known to be in any family of induction variables. In the PTRAN compiler, such cases are solved by an optimistic data flow technique which initially assumes all variables are linear induction variables until a contradiction exists [Allen et al. 1988]. Allen et al. 1981] present a comprehensive treatment of strength reduction by recognizing more general linear cases. Tu and Padua [1993] have used our SSA technique for linear induction (and monotonic) variable detection in order to perform symbolic analysis of array bounds and subscript ....

Allen, F., Burke, M., Charles, P., Cytron, R., and Ferrante, J. 1988. An overview of the PTRAN analysis system for multiprocessing. J. Parallel Distrib. Comput. 5, 5 (Oct.), 617--640.

....an advanced compiler could free the programmer from this concern by automatically converting a sequential program into an equivalent shared memory parallel program. Although a substantial amount of research has been devoted to automatic parallelization, such systems are not consistently successful [1, 9, 6, 30, 33]. The Fourth ACM SIGPLAN Symposium on Principles Practice of Parallel Programming, San Deigo, CA, May, 1993. This research was supported by The Center for Research and Parallel Computation (CRPC) at Rice University, under NFS Cooperative Agreement Number CCR 9120008. y Center for Integrated ....

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the PTRAN analysis system for multiprocessing. In Proceedings of the First International Conference on Supercomputing. Springer-Verlag, Athens, Greece, June 1987.

....Index Terms Conditional statements, dependence graph, DO loops, parallelism evaluation, parallelizing compilers, upper and lower bounds of parallelism. 2 1 INTRODUCTION Restructuring compilers, such as Parafrase I and II [9, 14] from Illinois, PFC from Rice University [3] IBM PTRAN [2], KAP from Kuck and Associates, VAST from Pacific Sierra, to name a few, have been developed over the years for high performance parallel computers. Their main goal is to efficiently parallelize or and vectorize sequential programs written in conventional languages to take advantage of its ....

Allen F., Burke M., Charles P., Cytron R. and Ferrante J., "An Overview of the PTRAN Analysis System for Multiprocessing", Journal of Parallel and Distributed Computing, no. 5, pp. 617-640, 1988.

....77 so that they execute efficiently on a parallel machine. In pursuit of this goal, several research projects and prototypes have been built to investigate automatic parallelization. These include Parafrase at the University of Illinois [1] PFC at Rice University [2] PTRAN at IBM Research [3] and SUIF at Stanford University [4] In addition, several commercial systems, such as Vast by Pacific Sierra, KAP by Kuck and Associates, and compilers by Alliant, Convex, and Cray, perform automatic parallelization. While these systems achieve impressive results for some programs, they fail to ....

....Editor Upon completion, the compilation system described in Section 2 will be capable of automatically converting a sequential Fortran program into a parallel version. A substantial amount of work has been conducted on the viability of this approach by researchers at Rice and elsewhere [1, 2, 3, 15, 46]. Ideally, automatically parallelized programs will execute efficiently on the target architecture and users will not need to intervene. Although such systems can effectively parallelize many interesting programs, they have not established an acceptable level of success. Consequently, advanced ....

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante, "An overview of the PTRAN analysis system for multiprocessing," in Proceedings of the First International Conference on Supercomputing, Athens, Greece: Springer-Verlag, June 1987.

....detail. 3 This is often due to the use of explicit SPMD models (e.g. VM EPEX [12] the Force [21] EXPRESS [34] and other parallel dialects [23] or due to implicit fork join parallelism resulting from program restructuring tools (e.g. Faust [17] Parafrase2 [36] ParaScope [9] PTRAN [2]) 7 Conclusion In this paper the Pamela performance modeling methodology is introduced which provides a theoretical framework for modeling and analyzing the performance of parallel systems. Serialization analysis is shown to be an attractive candidate for compile time prediction given its ....

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante, "An overview of the PTRAN analysis system for multiprocessing," in Proc. 1st Int. Conf. Supercomputing, 1987, pp. 194--211.

....ring. Simply put, the problem is to assign each of a set of independent tasks to processors in the system so as to finish the processing of the set of tasks as quickly as possible. Job scheduling arises frequently in parallel computing, for example in algorithms for automatic loop parallelization [2, 10, 11, 18] or in the use of a parallel system to process batches of transactions or independent sequential programs. We restrict our attention to the ring, which is an important network in both theory and practice. From a theoretical perspective, the ring is a basic network structure, and much work has been ....

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the ptran analysis system for multiprocessing. Journal of Parallel and Distributed Computing, 5:617--640, 1988.

....show what parts of the code are being executed using some sort of highlighting. Data visualization systems show pictorial displays of the actual data structures of the program, which can be used for debugging. Numerous parallelizing compilers and parallelization systems, including PTRAN [2], ParaScope Editor [5] Parafrase II [22] Tiny [29] Pat [3] Superb [30] Faust [13] E SP [7] PPSE [16] and MIMDizer [24] have been designed and implemented. Some types of program visualization interfaces do exist for these systems. Most of these systems are limited to a few selected ....

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante, "An Overview of the PTRAN Analysis System for Multiprocessing," Journal of Parallel and Distributed Computing, vol. 5, pp. 617-640, 1988.

....static and dynamic program statistics. One way to compare machines is to do an analysis similar to ours, but at the level of the machine instruction set [Peut77] This approach only permits comparisons between machines which implement the same instruction set. In the context of the PTRAN project [Alle87], execution time prediction has been proposed as a technique to help in the automatic partitioning of parallel programs into tasks. In [Sark89] execution profiles are obtained indirectly by collecting statistics on all the loops of a possible unstructured program, and then combining that with ....

Allen, F., Burke, M., Charles, P., Cytron, R., and Ferrante J., "An Overview of the PTRAN Analysis System for Multiprocessing. ", Proc. of the Supercomputing '87 Conf., 1987.

....and program design issues, which must be faced with the explicit tasking languages, are not present. This approach is, in a real sense, a direct outgrowth of successful research in construction of vectorizing compilers. and is currently being actively explored by several research groups [6, 7, 26, 51, 56, 63, 65]. Since the millions of lines of existing sequential programs cannot be easily replaced, nor are they readily modifiable, there is clear importance to this approach, and it will surely continue. There are, however, a number of difficulties with this approach. The major one is that the semantics of ....

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the PTRAN analysis system for multiprocessing. Research Report RC 13115 (#56866), IBM T. J. Watson Research Center, Yorktown Heights, NY, September 1987.

....users to seek a similarly elegant software solution to the problem of programming parallel computers. A substantial amount of research has been conducted on whether sequential Fortran 77 programs can be automatically converted without user assistance to execute on shared memory parallel machines [2, 3, 7, 8, 42, 49]. The results of this research have been both promising and disappointing. Although such systems can successfully parallelize many interesting programs, they have not established a level of success that will make it possible to avoid explicit parallel programming by the user. Hence, research has ....

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the PTRAN analysis system for multiprocessing. In Proceedings of the First International Conference on Supercomputing. Springer-Verlag, Athens, Greece, June 1987.

....of future work. 4. 3 Explicit Parallel Sections We have designed Nascent to accept explicit parallel section constructs consistent with the Parallel Computing Forum extensions to Fortran [11] An example of parallel section constructs, similar to ideas advocated by Hansen [13] and Allen et al. [2], is given in Figure 7. We needed to design our intermediate representation to translate these programs into SSA form, preserving the unique use def link when applicable, but dealing with multiple (perhaps anomalous) updates among parallel sections. We solved this problem by adding a new operator ....

....the algorithm is intended to be used when generating code to eliminate blocks containing dead OE functions) nor is the set Definers specified to be of factored form necessary to maintain the single reaching definition property of SSA. Other systems have been developed, such as the PTRAN system [2], aimed at automatic restructuring of sequential code for parallel execution. However, the intermediate form is not described in much detail, so it is difficult to assess the methods they use. Many other intermediate representations for compiler analysis and optimizations exist, though most don t ....

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the PTRAN analysis system for multiprocessing. Journal of Parallel and Distributed Computing, pages 617--640, May 1988.

.... as transformations over the PDG [6, 19, 2] Examples include region scheduling [19] and software pipelining [2] In addition, the PDG has been used for various scalar optimizations [15, 29, 24] detecting and improving parallelization for vector machines [32, 5] and multiprocessor machines [34, 3]. Variations of the PDG have also been used for debugging and integrating different versions of a program via program slicing [33, 1, 17, 22, 21, 30] and to enable translation of imperative programs for data flow machines and demand driven graph reducers [4] By using a common program ....

F. E. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the PTRAN analysis system for multiprocessing. Journal of Parallel and Distributed Computing, 5:617--640, 1988.

....Simply put, the problem is to assign each of a set of independent tasks to processors in the system so as to finish the processing of the set of tasks as quickly as possible. Job scheduling arises frequently in parallel computing, for example in algorithms for automatic loop parallelization [3, 18, 19, 27] or in the use of a parallel system to process batches of transactions of independent sequential programs. There is a wealth of literature on the analytical evaluation of parallel machine scheduling (see Lawler et al. 21] for a number of examples) but much of it fails to capture the full ....

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the ptran analysis system for multiprocessing. Journal of Parallel and Distributed Computing, 5:617--640, 1988.

....model. It then falls on the compiler to restructure certain regions into SPMD style for efficiency. 2 Framework We assume as input a fork join program structure, either originally supplied by a programmer and subsequently analyzed, or generated from a program parallelization system such as PTRAN [1, 4]. In such programs, parallelism can be created and terminated by closed, nestable parallel constructs. We call the beginning of a parallel construct the fork point and the end of the parallel construct the join point. We assume that the execution of parallel loops is deterministic: there are no ....

Frances Allen, Michael Burke, Philippe Charles, Ron Cytron, and Jeanne Ferrante. An overview of the ptran analysis system for multiprocessing. Proceedings of the 1987.

....identify those conditions affecting statement execution. Informally, a statement is control dependent on a branch if one edge from the branch definitely causes that statement to execute while another edge can cause the statement to be skipped. Such information is vital for detection of parallelism [2], program optimization, and program analysis [28] 1.3 Outline of the Rest of the Paper Section 2 reviews the representation of control flow by a directed graph. Section 3 explains SSA form and sketches how to construct it. This section also considers variants of SSA form as defined here. Our ....

....is thus O(E size(DF ) which amounts to a worst case complexity of O(E N 2 ) However, x8 shows that the size of the mapping DF is usually linear in practice. We have implemented this algorithm and have observed that it is faster than the standard data flow computations in the PTRAN compiler [2]. 4.3 Relating Dominance Frontiers to Joins We start by stating more formally the nonrecursive characterization of where the OE functions should be located. Given a set S of CFG nodes, the set J(S) of join nodes is defined to be the set of all nodes Z such that there are two nonnull CFG paths ....

[Article contains additional citation context not shown here]

F. E. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the PTRAN analysis system for multiprocessing. Journal of Parallel and Distributed Computing, 5:617--640, October 1988.

No context found.

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the PTRAN analysis system for multiprocessing. In Proceedings of the 1987.

No context found.

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the PTRAN analysis system for multiprocessing. Journal of Parallel and Distributed Computing, 5:617-640, 1988.

No context found.

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the PTRAN analysis system for multiprocessing. Journal of Parallel and Distributed Computing, 5(5):617-640, 1988.

No context found.

F. E. Allen, M. Burke, P. Charles, R. Cytron and J. Ferrante, `An overview of the PTRAN analysis system for multiprocessing', Journal of Parallel and Distributed Computing, 5, 617--640 (1988).

No context found.

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An Overview of the PTRAN Analysis System for Multiprocessing. In Proceedings of the 1st International Conference on Supercomputing. June 1987.

No context found.

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the ptran analysis system for multiprocessing. Technical Report RC 13115 (#56866), New-York, September 1987.

No context found.

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the PTRAN analysis system for multiprocessing. In Proceedings of the First International Conference on Supercomputing. Springer-Verlag, Athens, Greece, 1987.

No context found.

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante, "An Overview of the PTRAN Analysis System for Multiprocessing", Proc. International Conference on Supercomputing, 1987.

No context found.

Frances Allen, Michael Burke, Philippe Charles, Ron Cytron, and Jeanne Ferrante. An overview of the PTRAN analysis system for multiprocessing. Journal of Parallel and Distributed Computing, 5:617--640, 1988.

No context found.

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante, An overview of the PTRAN analysis system for multiprocessing, J. of Parallel and Dist Computing, Oct 88.

No context found.

Frances Allen, Michael Burke, Philippe Charles, Ron Cytron, and Jeanne Ferrante. An overview of the PTRAN analysis system for multiprocessing. In Proceedings of the 1987 International Conference on Supercomputing, 1987.

No context found.

F.E. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the ptran analysis system for multiprocessing. Journal of Parallel and Distributed Computing, 5(October):617--640, 1988.

No context found.

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the PTRAN analysis system for multiprocessing. Journal of Parallel and Distributed Computing, 5(5):617--40, October 1988.

No context found.

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the PTRAN analysis system for multiprocessing. J. of Parallel and Distributed Computing, 5:617--640, 1988.

No context found.

F. Allen, M. Burke, P. Charles, R. Cytron, and J. Ferrante. An overview of the PTRAN analysis system for multiprocessing. In Proceedings of the First International Conference on Supercomputing. Springer-Verlag, Athens, Greece, June 1987.

No context found.

Allen87. Allen, F., Burke, M., Charles, P., Cytron, R., and Ferrante, J., "An overview of the PTRAN analysis system for multiprocessing," Res. Rep. RC 13115, IBM T.J. Watson Research Center, Yorktown Heights, NY (September 1987).

First 50 documents  Next 50

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