### Table 1. UIUC results as the number of training images M is var- ied: (a) fractal dimension; (b) fractal length; (c) and (d) the meth- ods of [20] and [37] respectively. Means and standard deviations have been computed over 1000 random splits of the training and test set.

2007

"... In PAGE 6: ... Means and standard deviations have been computed over 1000 random splits of the training and test set. Table1 compares the performance of our local fractal dimension and length descriptors with the methods of [20] and [37]. As can be seen, the performance using either D or L is better than that achieved by the state-of-the-art fractal based method of [37].... ..."

Cited by 1

### Table 2. Safe prediction durations and stability for the three duration prediction schemes.

"... In PAGE 6: ... A second aspect of a predictor is the typical duration it is able to successfully predict. We show this with the mean safe prediction duration in Table2 . This mean does not include predictions that over- shot.... ..."

### Table 2. Safe prediction durations and stability for the three duration prediction schemes.

"... In PAGE 6: ... A second aspect of a predictor is the typical duration it is able to successfully predict. We show this with the mean safe prediction duration in Table2 . This mean does not include predictions that over- shot.... ..."

### Table 1: Summary of current popular prediction schemes

1996

"... In PAGE 2: ... 2.2 Overview of current branch prediction schemes Using the conceptual model just introduced, we can sum- marize current popular branch prediction schemes in Table1 . This table describes the basic components used in each prediction scheme.... ..."

Cited by 72

### Table 1: Summary of current popular prediction schemes

"... In PAGE 2: ... 2.2 Overview of current branch prediction schemes Using the conceptual model just introduced, we can summa- rize current popular branch prediction schemes in Table1 . This table describes the basic components used in each prediction scheme.... ..."

### Table 1: Summary of current popular prediction schemes

"... In PAGE 2: ... 2.2 Overview of current branch prediction schemes Using the conceptual model just introduced, we can summa- rize current popular branch prediction schemes in Table1 . This table describes the basic components used in each prediction scheme.... ..."

### Table 7. Sensitivity of results to the branch prediction scheme.

"... In PAGE 8: ... To test the sensitivity of the results to the branch prediction scheme, a perfect branch predictor is tested on pipeline depths of 7, 15, 25, and 50. Table7 shows that a perfect branch predictor improves instruction through- put. However, the trend of shorter pipelines outperforming longer pipelines remains unchanged.... ..."

Cited by 1

### Table 3. Overshoot for the three duration prediction schemes.

"... In PAGE 6: ... In these cases, stable cover- age of FXby8 drops as we discard the overshot predictions at the end of each phase. The last figure of merit in designing dura- tion predictors is the degree of overshoot they exhibit; we show this metric in Table3 . FXX displays poor performance, with very long 44 LOW POWER IEEE MICRO Table 1.... ..."

### Table 3. Overshoot for the three duration prediction schemes.

"... In PAGE 6: ... In these cases, stable cover- age of FXby8 drops as we discard the overshot predictions at the end of each phase. The last figure of merit in designing dura- tion predictors is the degree of overshoot they exhibit; we show this metric in Table3 . FXX displays poor performance, with very long 44 LOW POWER IEEE MICRO Table 1.... ..."

### Table 4: Different Prediction Schemes Evaluated

1994

Cited by 3