W. Thies, F. Vivien, J. Sheldon, and S. Amarasinghe. A unified framework for schedule and storage optimization. In International Conference on Programming Language Design and Implementation (PLDI'01), pages 232--242. ACM Press, 2001.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....on current architec # The final version will be published in the proceedings of Programming Language Design and Implementation (PLDI) June 2003. tures. Data and loop transformations can fu rther both goals. Until recently the focus has been primarily on compile time transformation frameworks [15, 24, 14, 4, 12, 13, 29, 11, 28] restricted to a#ne loop bounds and a#ne array references. One such framework is that of Kelly and Pugh [12] which represents loop nests as iteration spaces. A compiler can use their framework to transform iteration spaces and implement corresponding data reorderings. The legality of such ....

William Thies, Frederic Vivien, Je#rey Sheldon, and Saman Amarasinghe. A unified framework for schedule and storage optimization. In Cindy Norris and Jr. James B. Fenwick, editors, Conference on Programming Language Design and Implementation (PLDI-01), volume 36.5 of ACM SIGPLAN Notices, pages 232--242, June 2001.

..... 5.00. 1. INTRODUCTION Data locality and parallelism are essential for improving the performance of applications on current architectures. Data and loop transformations can further both goals. Until recently the focus has been primarily on compile time transformation frameworks [19, 27, 18, 3, 16, 17, 32, 14, 31] restricted to a#ne loop bounds and a#ne array references. These frameworks allow for the uniform representation, the composition, the legality determination, and sometimes a benefit model of various compile time transformations. One such framework that of Kelly and Pugh [16] represents ....

W. Thies, F. Vivien, J. Sheldon, and S. Amarasinghe. A unified framework for schedule and storage optimization. In Proceedings of the ACM SIGPLAN '01 Conference on Programming Language Design and Implementation, ACM SIGPLAN Notices, pages 232--242, June 2001.

....at compile time can be reduced to solving a network flow problem. Our compiler provides a richer set of transformations, but probably runs slower. We accept that trade o# because some programs we care about, including multigrid, require the richer set of transformations to run fast. Thies et al. [27] describe a unified mathematical framework for analyzing the tradeo#s between parallelism and storage allocation in a parallelizing compiler. Their work is based on Strout et al. s universal occupancy vectors, and shares some of the same limitations. In particular, their system can at best ....

William Thies, Frederic Vivien, Je#rey Sheldon, and Saman Amarasinghe. A Unified Framework for Schedule and Storage Optimization. In Proceedings of the 2001 SIGPLAN Conference on Programming Language Design and Implementation.

....has to solve (or approximately solve) an NP complete problem. Even worse, their approach is limited to contracting one dimension of an array, because two array indices are in the same equivalence class if and only if they di#er by some multiple of the universal occupancy vector. 15 Thies et al. [35] describe a unified mathematical framework for analyzing the tradeo#s between parallelism and storage allocation in a parallelizing compiler. Their work is based on Strout et al. s universal occupancy vectors, and shares some of the same limitations. In particular, their system can at best ....

William Thies, Frederic Vivien, Je#rey Sheldon, and Saman Amarasinghe. A Unified Framework for Schedule and Storage Optimization. In Proceedings of the 2001 SIGPLAN Conference on Programming Language Design and Implementation.

....contributions of this paper include 1. A synthesis of existing ideas on memoization from declarative and functional languages applied to building a declarative library for an imperative programming language. 2. A discussion of the similarities of three compile time garbage collection methods [13, 18, 27] that were previously unknown to be related. 3. Suggestion for a novel method of garbage collecting memoized facts at runtime. The next section discusses goal directed evaluation with memoization. Section 3 discusses garbage collection of memoized facts. Section 4 discusses a method of reusing ....

....while Section 3.2 discusses a new idea for runtime garbage collection that is more suitable for use in a library. 3. 1 Analysis at Compile Time Three roughly equivalent methods of automatically determining these storage requirements have been independently investigated in the literature [18, 24, 13, 12, 26, 8, 27]. All three methods work at compile time and attempt to recover enough dependence information to conservatively prove that certain memoized facts will never again be accessed. 6 i j Figure 4: Dependencies for the longest common subsequence problem. Shaded circles are results that have already ....

[Article contains additional citation context not shown here]

William Thies, Frederic Vivien, Jeffrey Sheldon, and Saman Amarasinghe. A unified framework for schedule and storage optimization. In Proceedings of the 2001 ACM SIGPLAN Conference on Programming Language Design and Implementation, 2001. (to appear).

No context found.

W. Thies, F. Vivien, J. Sheldon, and S. Amarasinghe. A unified framework for schedule and storage optimization. In Proceedings of the ACM SIGPLAN 2001.

....on a parameterized version of the input program, avoiding the need to expand a graph for varying parameters and problem sizes, and it can often reduce to a linear program for flexible and e#cient optimization. Polyhedral representations have also been utilized for powerful storage optimizations [22, 28, 31, 20]. In this paper, we aim to bridge the gap and employ the polyhedral representations of the scientific community to analyze the synchronous dataflow graphs of the DSP community. Towards this end, we present a translation from a dataflow graph to a System of A#ne Recurrence Equations (SARE) which ....

....optimization problems that are already the focus of the DSP community. The polyhedral model is appealing because it provides a linear algebra framework that is simple, flexible, and e#cient. 7.3. 1 Bu#er Minimization Storage optimization is one area in which both the scientific community [22, 28, 31, 20] and the DSP community [25, 12, 24] have invested a great deal of energy. Both communities have invented schemes for detecting live ranges, collapsing arrays across dead locations, and sharing bu#ers arrays between di#erent 14 statements. It will be an interesting avenue for future work to use ....

W. Thies, F. Vivien, J. Sheldon, and S. Amarasinghe. A unified framework for schedule and storage optimization. In Proceedings of the ACM SIGPLAN 2001.

No context found.

W. Thies, F. Vivien, J. Sheldon, and S. Amarasinghe. A unified framework for schedule and storage optimization. In International Conference on Programming Language Design and Implementation (PLDI'01), pages 232--242. ACM Press, 2001.

No context found.

William Thies, Frederic Vivien, Je#rey Sheldon, and Saman Amarasinghe. A unified framework for schedule and storage optimization. In International Conference on Programming Language Design and Implementation (PLDI'01), pages 232--242. ACM Press, 2001.

No context found.

William Thies, Frederic Vivien, Je#rey Sheldon, and Saman Amarasinghe. A unified framework for schedule and storage optimization. In Proceedings of 2001.

No context found.

W. Thies, F. Vivien, J. Sheldon, and S. Amarasinghe. A Unified Framework for Schedule and Storage Optimization. ACM SIGPLAN Notices, 36(5):232--242, May 2001.

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