ATOM: a system for building customized program analysis tools
PLDI '94 Proceedings of the ACM SIGPLAN 1994 conference on Programming language design and implementation
The predictability of data values
MICRO 30 Proceedings of the 30th annual ACM/IEEE international symposium on Microarchitecture
Predictive techniques for aggressive load speculation
MICRO 31 Proceedings of the 31st annual ACM/IEEE international symposium on Microarchitecture
Delphi: Predition-based Page Prefetching to Improve the Performance of Shared Virtual Memory Systems
PDPTA '02 Proceedings of the International Conference on Parallel and Distributed Processing Techniques and Applications - Volume 1
Efficacy and Performance Impact of Value Prediction
PACT '98 Proceedings of the 1998 International Conference on Parallel Architectures and Compilation Techniques
The Alpha 21264 Microprocessor Architecture
ICCD '98 Proceedings of the International Conference on Computer Design
Differential FCM: Increasing Value Prediction Accuracy by Improving Table Usage Efficiency
HPCA '01 Proceedings of the 7th International Symposium on High-Performance Computer Architecture
SlicK: slice-based locality exploitation for efficient redundant multithreading
Proceedings of the 12th international conference on Architectural support for programming languages and operating systems
Hi-index | 0.00 |
The most promising value predictors to date are the finite context method predictor and a recent improvement thereof, the differential finite context method predictor. Both predictors comprise two levels and the index into the second level is a function of the content of the first level. This index function is crucial for good performance. However, our research shows that the currently used select-fold-shift-xor function performs poorly on range-limited sequences of values. For example, it does not predict the results of byte loads well. The problem with the current function is that it often cannot reach the predictor's entire second-level table. We propose an improved index function that does not suffer from this shortcoming. On the 15 SPECcpu2000 C programs, our new index function improves the average load-value predictability by about 1% to 5% without increase in predictor size. On byte loads, the improvement is over 6% for 4096-entry predictors.