R. Gupta, D. Berson, and J.Z. Fang, "Resource-Sensitive Profile-Directed Data Flow Analysis for Code Optimization," IEEE/ACM International Symposium on Microarchitecture (MICRO), pages 558-568, December 1997  Home/Search   Document Details and Download   Summary   Related Articles   Check

....partial dead stores. The technique uses path profiling information to target only statements in frequently executed paths. A cost benefit data flow analysis technique determines the profitability of sinking, taking into account the frequencies of each path considered. The same approach is used in [GBF97b] to speculatively hoist computations in PRE. Decisions to speculate are made locally at individual merge or split points based on the affected paths. Acyclic and cyclic code are treated by different versions of the algorithm. 3 Overview of Approach In our PRE based approach to register ....

R. Gupta, D. Berson, and J. Fang. Resourcesensitive profile-directed data flow analysis for code optimization. In Proceedings of the 30th Annual International Symposium on Microarchitecture, pages 358--368, December 1997.

....through demand driven analysis as in branch elimination [BGS97a] c) all profitable opportunities are exploited with economical transformations (code motion and speculation) and d) infrequent program regions will be avoided using the profile navigated demand driven analysis. Gupta et al. [GBF97a, GBF98, GBF97b] also use path profiles to carry out path sensitive optimization. Similar to the method of Ammons and Larus, path profiles are used to maintain path specific context along each hot path. The results of the analysis are then used to integrate code motion with control speculation. In ....

R. Gupta, D. Berson, and J.Z. Fang. Resource-sensitive profile-directed data flow analysis for code optimization. In 30th Annual IEEE/ACM International Symposium on Microarchitecture, pages 358--368, December 1997.

....decrease performance. GURRR can be used to assess more precisely if the code produced by transformations has the required resources available in order to improve performance, and perform transformations only when the resources are sufficient to expect an improvement in code quality. Gupta et al. GBF97] use the same idea to do program optimizations guided both by profiles and by resource sensitivity. Their algorithms do code sinking or hoisting only when sufficient resources are available at the locations to which instruction would get moved. In particular, they use this resource sensitive ....

Rajiv Gupta, David A. Berson, and Jesse Z. Fang. Resource-sensitive profiledirected data flow analysis for code optimization. In Proceedings of MICRO30: 30th Annual IEEE/ACM International Symposium on Microarchitecture, page ??, 1997.

....the lifetime of temporaries in Lazy Code Motion [10] They also proposed the algorithm for partial dead code elimination [11] These are machine independent optimizations that do not consider machine resources or instruction latencies during code motion. More recently, Gupta, Berson, and Fang [6] provided a theoretical description of partial redundancy elimination and partial dead code elimination that considers resource constraints and pro 21 file information. Though useful as an optimization, this work is not an algorithm for global instruction scheduling. In another paper, Gupta ....

R. Gupta, D. A. Berson, and J. Z. Fang, "Resource-Sensitive Profile-Directed Data Flow Analysis for Code Optimization," Proc. 30th Annual IEEE/ACM Intl. Symp. on Microarchitecture, Dec. 1997, pp. 358-368.

No context found.

R. Gupta, D. Berson, and J.Z. Fang, "Resource-Sensitive Profile-Directed Data Flow Analysis for Code Optimization," IEEE/ACM International Symposium on Microarchitecture (MICRO), pages 558-568, December 1997

....through demand driven analysis as in branch elimination [BGS97a] 15 c) all profitable opportunities are exploited with economical transformations (code motion and speculation) and d) infrequent program regions will be avoided using the profile navigated demand driven analysis. Gupta et al. GBF97b,GBF98,GBF97c] also use path profiles to carry out path sensitive optimization. Similar to the method of Ammons and Larus, path profiles are used to maintain path specific context along each hot path. The results of the analysis are then used to integrate code motion with control speculation. In ....

....the value flow patterns are abstracted and summarized. Our estimators offer an alternative to frequency dataflow analysis; they are not as general (not all our estimators can compute the per node estimate) but are cheaper and fit the need of PRE transformations. Path profiles Gupta et al. [GBF97a,GBF97b,GBF98] present various versions of profile directed PRE, for exploiting various hardware features. Common to all these algorithms, all the algorithms are navigated using Ball Larus path profiles [BL96a, BMS98] While the algorithms do not explicitly compute an estimate, the estimate is computed ....

[Article contains additional citation context not shown here]

R. Gupta, D. Berson, and J.Z. Fang. Resource-sensitive profile-directed data flow analysis for code optimization. In 30th Annual IEEE/ACM International Symposium on Microarchitecture, pages 358--368, December 1997.

....cost of allowing speculation. Thus, in this situation it is beneficial to use the placement shown in Figure 11b. In [9] we describe in detail the combination of code speculation with the PRE optimization [16, 15] Speculation is enabled at conditional nodes based upon cost benefit analysis. In [10] we present predication based PRE and speculation based PRE algorithms that are resource sensitive. These algorithms compute code placements where the functional unit required for the execution of the placed algorithm is available. Only the optimization opportunities for which such placements ....

R. Gupta, D. Berson, and J.Z. Fang, "ResourceSensitive Profile-Directed Data Flow Analysis for Code Optimization," The 30th Annual IEEE/ACM International Symposium on Microarchitecture, Research Triangle Park, North Carolina, December 1997.

.... the application of optimizations with scheduling actions [74, 68] Another approach suggests application of optimizations prior to instruction scheduling using new optimization algorithms that are aware of the effects that an optimization may have on register and functional unit demands [69, 70]. Second we must develop new optimization algorithms that take advantage of path execution frequencies and machine characteristics. For example, researchers have developed algorithms for eliminating redundant code [71, 29] and dead code [72, 30] along frequently executed program paths. These ....

R. Gupta, D. Berson, and J.Z. Fang, "Resource-Sensitive Profile-Directed Data Flow Analysis for Code Optimization," Proc. of the 30th Annual IEEE/ACM International Symposium on Microarchitecture, pages 358-368, Research Triangle Park, North Carolina, December 1997.

....also effects register pressure. Our algorithm can be easily extended to perform such code reordering to lower register pressure. However, for such code reordering to be consistent with good instruction scheduling decisions, it is important to consider pressure on functional unit resources. In [12, 10] we presented techniques for code reordering, including for PRE and PDE, that are sensitive to functional unit pressure and consistent with instruction scheduling decisions. Those techniques can be combined with the algorithms presented in this paper. ....

R. Gupta, D. Berson, and J.Z. Fang, "Resource-Sensitive Profile-Directed Data Flow Analysis for Code Optimization," The 30th Annual IEEE/ACM International Symposium on Microarchitecture, pages 558-568, Research Triangle Park, North Carolina, December 1997.

....certain paths. The third algorithm integrates both restructuring and speculation and selects a profitable subgraph of the CMP for each. While profitably balancing the cost and benefit under a given profile is NPhard, the empirically small number of hot program paths promises an efficient algorithm [4, 19]. Finally, to support profile guiding, we show how an estimate of the run time gain thwarted by a CMP region can be obtained using edge profiles, frequency analysis [27] or path profiles [7] 3.1 Selective Restructuring We model the profitability of duplicating a CMP region R with a ....

....PRE. 3.3 Partial Restructuring, Partial Speculation In Section 3.1, edge profiles and frequency analysis were used to estimate the benefit Rem of duplicating a region. An alternative is to use path profiles [3, 7] which are convenient for establishing cost benefit optimization trade offs [4, 19, 20]. To arrive at the value of the region benefit with a path profile, it is sufficient to sum the frequencies of MustMust paths, which are paths that cross any region entry edge that is Must available and any exit edge that is Must anticipated. These are precisely the paths along which value ....

[Article contains additional citation context not shown here]

R. Gupta, D. Berson, and J.Z. Fang. Resource-sensitive profile-directed data flow analysis for code optimization. In 30th Annual IEEE/ACM International Symposium on Microarchitecture, pages 358--368, December 1997.

....of reuse detected by the analysis. In this paper, an estimator is an algorithm that reconstructs the total frequency of reuse paths from a program profile. The estimator returns a profile weighted reuse, which estimates the optimization benefit and thus can guide profile directed optimizations [10, 21]. In this paper, the weighted reuse serves as a measure of PREcompleteness: when the profile used by the estimator is generated by the limit study simulator, the weighted reuse shows what fraction of the simulator detected reuse was found by the analysis, and therefore indicates the precision of ....

R. Gupta, D. Berson, and J.Z. Fang. Resource-sensitive profile-directed data flow analysis for code optimization. In 30th Annual IEEE/ACM International Symposium on Microarchitecture, pages 358--368, December 1997.

.... execution can be used to enable code sinking past merge points resulting in the removal of dead code along frequently executed paths [8] We have also developed analysis techniques in which redundancy and dead code removal decisions are influenced by functional unit resource availability [9]. Other applications of our approach include strength reduction along certain paths and elimination of partially redundant loads and dead stores [2] ....

R. Gupta, D. Berson, and J.Z. Fang, "ResourceSensitive Profile-Directed Data Flow Analysis for Code Optimization," The 30th Annual IEEE/ACM International Symposium on Microarchitecture, pages 558-568, Research Triangle Park, NC, December 1997.

....The anticipability and liveness conditions require that the data flow information be computed with respect to different categories of paths. Conventional analysis techniques compute data flow with respect to all program paths. An approach for path sensitive analysis was introduced in [4, 5] where in addition to computing a data flow fact at a program point, the set of paths along which the fact holds is also computed. The anticipability analysis is performed by starting at node s and propagating data flow information backward along nodes that lie along paths from d to s (set of ....

R. Gupta, D.A. Berson, and J.Z. Fang, "Resource-Sensitive Profile-Directed Data Flow Analysis for Code Optimization," 30th IEEE/ACM Intl. Symp. on Microarchitecture, pages 558-568, Research Triangle Park, NC, Dec. 1997.

No context found.

R. Gupta, D. A. Berson, and J. Z. Fang. 1997. "Resource-Sensitive Profile-Directed Data Flow Analysis for Code Optimization," Proc. 30th Annual IEEE/ACM Intl. Symp. on Microarchitecture, Dec., pp. 358-368.

No context found.

R. Gupta, D. Berson, and J. Fang. Resource-sensitive profile-directed data flow analysis for code optimization. In Proceedings of the 30th Annual Intl. Symp. on Microarchitecture, December 1997. pp. 358-68.

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