B.K. Rosen. High-Level Data Flow Analysis. Communications of the ACM, 20(10):712-724, October 1977.  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... research focusing on identi cation of opportunities for program optimization based on data ow analysis [4] issues related to interprocedural analysis [5, 46] improved eciency [20, 29] the de nition of data ow problems and their related solutions [6] algorithms for use at the source level [45], complications due to variable aliasing [35] more powerful approaches to data ow analysis [28] general improvements to ow analysis algorithms [8] methods to deal with the huge amounts of information that exhaustive data ow analysis generates [9] application to other areas of computer ....

B.K. Rosen. High-Level Data Flow Analysis. Communications of the ACM, 20(10):712-724, October 1977.

....flow using continuations, but in a more direct manner. Work on static analysis in the form of data flow analysis and abstract interpretation is extensive. Performing flow analysis at the source level ( high level data flow analysis ) for specific data flow problems has been considered by Rosen [15] and Babich and Jazayeri [2,3] Generalization of various related flow analysis techniques into uniform frameworks includes the work of the Cousots [5] Kam and Ullman [10] and Kildall [12] Marlowe and Ryder provide an excellent survey of data flow analysis problems and the computational cost of ....

Rosen, B. "High-Level Data Flow Analysis," CACM 20, 10, October 1977, pp. 712-724.

....we will take advantage of the fact that this almost never happens in practice. On the other hand, CFGs nicely describe the arbitrary control flow that GOTOs can induce, and so all other control structures are expressed in terms of GOTO as well. Thus, the GOTO tail wags the control structure dog [Ros77]: high level control structures are sacrificed for the sake of GOTOs. However, as Powell says [Pow84] The value of constructing flow graphs for languages or programs without GOTO statements (as well as the payoff from considering all the perverse possibilities in languages that do have them) is ....

....and Zadeck s, I do not present it here. 10 Related Work Others have noted the merits of using ASTs or other tree like representations for code optimization, but have only discussed their advantages with regard to traditional data flow analysis problems such as computing live variables. Rosen [Ros77] proposes representing each statement in a program by a graph, and an entire program by an AST like hierarchy of graphs. This makes data 30 Compute the PDG; Root CodeMotion(Root) Eliminate any resulting dead code; CodeMotion(thisNode) Replace each child c of thisNode with CodeMotion(c) ....

Barry K. Rosen. High-level data flow analysis. Communications of the ACM, 20(10):712--724, October 1977.

....set of statements that actually affect the value of a variable in a given instance. In addition, slices have been used to represent both executable portions of a program as well as a set of statements that can be effected by a given slice criterion. Using techniques common to data flow analysis [5, 16, 10, 27, 26, 11, 18, 13], program slicing has been shown to be useful to debugging, testing, program integration and maintenance. 4 The SCAS environment The SCAS environment is a collection of tools that can be used to represent source information in a language independent form and for creating reports to aid the user ....

Barry K. Rosen. High-level data flow analysis. Communications of the ACM, 20(10):712--724, Oct. 1977.

....terminate the other paths. Data dependences are determined using traditional static analyses. Register dependences and simple memory dependences (involving named scalar variables) are established by standard dataflow analyses [1] 3] 12] 13] 19] 20] 24] 25] 50] 57] 59] 60] 69] 74] 75] [84] [85] 88] Data dependence information restricted to identifying the producer and the consumer instruction is not sufficient to include the dependence within a task. If the producer and the consumer instructions are not in adjacent basic blocks in the control flow graph, then other basic blocks ....

B. Rosen. High-level data flow analysis. Communications of the ACM, 20:712--724, 1977.

....into the middle of a sub tree, although this almost never happens in practice. On the other hand, CFGs nicely describe the arbitrary control flow that GOTOs can induce, and so all other control structures are expressed in terms of GOTO as well. Thus, the GOTO tail wags the control structure dog [Ros77] high level control structures are sacrificed for the sake of GOTOs. The preceding remarks concerning the hegemony of CFGs are only true in the procedural language community. The Lisp Scheme community uses ASTs in their compilers [Ste78] Kra88] For the most part, however, code optimization ....

....taking the place of DO FOREVER. 1.2 Related Work Others have noted the merits of using ASTs or other tree like representations for code optimization, but have only discussed their advantages with regard to traditional data flow analysis problems, such as computing live variables. Rosen [Ros77] proposes representing each statement in a program by a graph, and an entire program by an AST like hierarchy of graphs. This makes data flow analysis efficient because the graphs for known control structures such as IF THEN can be analyzed when a data flow analysis algorithm is written (at ....

Barry K. Rosen. High-level data flow analysis. Communications of the ACM, 20(10):712--724, October 1977.

....found the contents important. We are currently building an interprocedural analyzer to experiment on aggressive transformations, including array privatization. We adopt a hierarchical approach to dealing with complexity of data flows in large programs. A similar approach is proposed by Rosen in [Ros77] for scalar analysis. This approach takes advantage of structures in high level languages, which makes data flow analysis and dependence analysis efficient. Array privatization fits naturally in this approach, because a condensed node is just one example of using structures. The algorithms ....

B. K. Rosen. High-level data flow analysis. Communication of the ACM, 20(10):712--724, 1977.

No context found.

B.K. Rosen. High-Level Data Flow Analysis. Communications of the ACM, 20(10):712-724, October 1977.

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