G. Ammons and J. R. Larus. Improving data- ow analysis with path pro les. In Proceedings of the ACM SIGPLAN '98 Conference on Programming Language Design and Implementation (PLDI), pages 72-84, Montreal, Canada, 17-19 June 1998.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....they do not keep track of the contents of memory when registers are saved restored. Every one of these drawbacks have been addressed in this work. There is a considerably body of work on interprocedural data ow analyses design to analyze only part, but not all, of a program (see, for example, [3, 6, 16, 29]) although only some of them use pro le information to guide their decisions. This pro le information is, however, widely used when performing optimizations [26, 9, 12, 20] On the other hand, while speculation has been commonly used in the compiler world for optimizing programs [21, 17, 22, ....

G. Ammons and J. R. Larus. Improving data- ow analysis with path pro les. In Proceedings of the ACM SIGPLAN '98 Conference on Programming Language Design and Implementation, pages 72-84, Montreal, Canada, June 17-19, 1998. 17

....to interpreter. Figure 10: An example of redirecting the rare path. On the left, the dotted block is rare, so we redirect it to a block that calls the interpreter. is typically used to allow more precise data ow information after the merge point and thereby increase optimization opportunities [13, 40, 1]. By removing edges from rare blocks to non rare blocks, we are simply splitting the rare path from the common path, which obviously does not change the program semantics on the common path. The second observation is the code that we replace the rare path with the instruction to transfer to the ....

....frequentlyexecuted paths and form superblocks from those paths, using tail duplication to avoid joins. Our technique achieves a similar e ect. Ammons and Larus describe a technique of identifying and duplicating hot paths in order to improve the precision of data ow analysis along those paths [1]. The hot paths are constructed using an acyclic path pro le [7] Their technique uses quali ed data ow analysis [22] which couples a conventional data ow problem with a deterministic nite automaton, to recognize hot paths. To avoid unnecessary code growth, the duplications that ended up to ....

G. Ammons and J. R. Larus. Improving data- ow analysis with path pro les. In Proceedings of the ACM SIGPLAN'98 Conference on Programming Language Design and Implementation (PLDI), pages 72-84, Montreal, Canada, June 17-19, 1998.

....di erence is that superblock scheduling performs tail duplication along a single hot trace, whereas our hot path splitting allows duplication of multiple hot paths through a method. Previous work has used data ow information combined with o ine pro ling to guide code duplication. Ammons and Larus [2] used path pro les to guide code duplication for improving data ow analysis. Bodik et al. 9] uses edge or debugged for this benchmark, it will be re incorporated. 9 path pro les to guide code restructuring for partial redundancy elimination and code motion. Our feedback directed splitting ....

G. Ammons and J. R. Larus. Improving data- ow analysis with path proles. In Proceedings of the ACM SIGPLAN'98 Conference on Programming Language Design and Implementation (PLDI), pages 72-84, 1998.

....16] Neither of the latter two prior works were designed with predicates as one of the data ow analysis frameworks, and none of the three techniques derives run time tests. Recently, additional approaches that, in some way, exploit control ow information in data ow analysis have been proposed [1, 5, 20]. Ammons and Larus s approach improves the precision of data ow analysis along frequently taken control ow paths, called hot paths, by using pro le information. Bod k et al. describe a demand driven interprocedural correlation analysis that eliminates some branches by path specialization. Both ....

Ammons, G., and Larus, J. R. Improving data- ow analysis with path proles. In Proceedings of the ACM SIGPLAN '98 Conference on Programming Language Design and Implementation (Montreal, Canada, June 1998), pp. 72-84.

....23] Neither of the latter two prior works were designed with predicates as one of the data ow analysis frameworks, and none of the three techniques derives run time tests. Recently, additional approaches that, in some way, exploit control ow information in data ow analysis have been proposed [2, 6, 27]. Ammons and Larus s approach improves the precision of data ow analysis along frequently taken control ow paths, called hot paths, by using pro le information. Bod k et al. describe a demand driven interprocedural correlation analysis that eliminates some branches by path specialization. Both ....

Glenn Ammons and James R. Larus. Improving data- ow analysis with path proles. In Proceedings of the ACM SIGPLAN '98 Conference on Programming Language Design and Implementation, pages 72-84, Montreal, Canada, June 1998.

No context found.

G. Ammons and J. R. Larus. Improving data- ow analysis with path pro les. In Proceedings of the ACM SIGPLAN '98 Conference on Programming Language Design and Implementation (PLDI), pages 72-84, Montreal, Canada, 17-19 June 1998.

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