Jens Knoop, Bernhard Steen, and Jurgen Vollmer. Parallelism for free: Ecient and optimal bitvector analysis for parallel programs. ACM Transactions on Programming Languages and Systems, 18(3):268-299, May 1996.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....to another node S# , ##### the same variable where S# and S# may run in parallel. The TPDG for the example in Figure 1 is given in Figure 2. The edges of a TPDG can be calculated using standard techniques of data ow analysis for concurrent programs with shared memory and interleaving semantics [8]. 1.1 Previous Work The slicing criterion for a concurrent program is de ned by the triple t;p;V , where t is a thread in the program, p is a program pointint and V is a subset of the program s variables. Given a TPDG, the slicing criterion is clearly identi ed as a node in the TPDG. Cheng ....

J. Knoop and B. Steen. Parallelism for free: ecient and optimal bitvector analyses for parallel programs. #######, 18(3):268-299, May 1996.

....typically suffers from the so called state explosion problem: in general, already the required control structures grow exponentially with the number of parallel components. Bitvector analyses, dominant in most practical compilers, escape this problem in the context of fork join parallelism [11, 9]: a simple pre process is sucient to adapt sequential intra procedural bitvector analyses to directly work on parallel ow graphs which concisely and explicitly represent the program s parallelism. Key for this adaptation was to change from a property analysis (directly associating program points ....

....bit component each of which only requires the consideration of a three point transformer domain. In order to handle also procedures and unbounded parallelism, Esparza and Knoop observed that the described problem pro le also admits an automata 1 Second order analysis in the terminology of [11]. theoretic treatment [5] This observation has been carefully developed by Esparza and Podelski in [6] The resulting algorithm requires involved concepts, like e.g. tree automata, and, from a program analyzer s perspective, the results are rather indirect: the reachability analysis computes ....

[Article contains additional citation context not shown here]

J. Knoop, B. Steen, and J. Vollmer. Parallelism for Free: Ecient and Optimal Bitvector Analyses for Parallel Programs. ACM Transactions on Programming Languages and Systems, 18(3):268-299, 1996.

....analysis typically su ers from the so called state explosion problem: in general, already the required control structures grow exponentially with the number of parallel components. Bitvector analyses, dominant in most practical compilers, escape this problem in the context of fork join parallelism [11, 8]: a simple pre process is sucient to adapt sequential intra procedural bitvector analyses to directly work on parallel ow graphs which concisely and explicitly represent the program s parallelism. Key for this adaptation was to change from a property analysis (directly associating program points ....

....program points with properties) to an e ect analysis 1 associating program points with a property transformer resembling the e ect of the preceding program fragment. The simplicity of the adaption results from the fact that bitvector analyses 1 Second order analysis in the terminology of [11]. Received December 20, 2000. 2 can conceptually be sliced into separate analyses for each individual bitcomponent each of which only requires the consideration of a three point transformer domain. This setting allowed one to elegantly solve the two central problems inherent in the analysis ....

[Article contains additional citation context not shown here]

J. Knoop, B. Steen, and J. Vollmer. Parallelism for Free: Ecient and Optimal Bitvector Analyses for Parallel Programs. ACM Transactions on Programming Languages and Systems, 18(3):268-299, 1996.

No context found.

Jens Knoop, Bernhard Steen, and Jurgen Vollmer. Parallelism for free: Ecient and optimal bitvector analysis for parallel programs. ACM Transactions on Programming Languages and Systems, 18(3):268-299, May 1996.

No context found.

J. Knoop and B. Steen. Parallelism for free: ecient and optimal bitvector analyses for parallel programs. ACM Transactions on Programming Languages and Systems, 18(3):268-299, May 1996.

No context found.

J. Knoop, B. Steen, J. Vollmer, Parallelism for free: Ecient and optimal bitvector analyses for parallel programs, ACM Transactions on Programming Languages and Systems 18 (3) (1996) 268-299.

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