### Table 2. Numerical results for reliabilityPath Decomposition Path Lower Bound Upper Bound Basic Algorithm [6] Enhanced Algorithm

### Table 2.11: Reduction in schedule length of transformed hard basic blocks compared to the untransformed basic blocks using critical-path list scheduling.

2007

### Table 2. Comparison of targeted path profiling (TPP) with Ball-Larus path profiling (PP) for a subset of SPEC95 benchmarks. Over- head is the increase in execution time due to profiling. Attribution of Definite Flow is the fraction of dynamic paths (weighted by path length in basic blocks) that can be derived from the instrumentation.

2004

"... In PAGE 8: ...1. Results Our experiments demonstrate that targeted path profiling (TPP) reduces the overhead of the original Ball-Larus algo- rithm by about half (SPEC95, Table2 ) to almost two-thirds 5 Comparison of profiles collected by PP and TPP is complicated by the fact that the two tools may truncate paths along different edges. To en- able a quantitative comparison of the two profiles, we split all recorded paths at the union of both truncated edge sets.... ..."

Cited by 11

### Table 3. Overlapping paths in the CFG.

2004

"... In PAGE 3: ... Note that this path is nothing but BL path BJ extended to C8BE, that is, extended by 1 predicate block. The basic block sequences for few OL paths of different overlaps are shown in Table3 . Also, the number of OL paths of each degree are shown.... ..."

Cited by 6

### Table 3. Overlapping paths in the CFG.

"... In PAGE 3: ... Note that this path is nothing but BL path BJ extended to C8 BE , that is, extended by 1 predicate block. The basic block sequences for few OL paths of different overlaps are shown in Table3 . Also, the number of OL paths of each degree are shown.... ..."

### Table 2. Comparison of targeted path pro ling (TPP) with Ball-Larus path pro ling (PP) for a subset of SPEC95 benchmarks. Over- head is the increase in execution time due to pro ling. Attribution of De nite Flow is the fraction of dynamic paths (weighted by path length in basic blocks) that can be derived from the instrumentation.

2004

"... In PAGE 8: ...1. Results Our experiments demonstrate that targeted path pro ling (TPP) reduces the overhead of the original Ball-Larus algo- rithm by about half (SPEC95, Table2 ) to almost two-thirds 5 Comparison of pro les collected by PP and TPP is complicated by the fact that the two tools may truncate paths along different edges. To en- able a quantitative comparison of the two pro les, we split all recorded paths at the union of both truncated edge sets.... ..."

Cited by 11

### Table 5: Performance guarantees in terms of worst-case factors from optimality for the critical-path heuristic (hcp) and the decision tree heuristic (hdt), for ranges of basic block sizes and various issue widths.

"... In PAGE 10: ...ic, there were between 2.9 and 7.8 basic blocks where the decision tree heuristic found a better schedule than the critical-path heuristic. Table5 shows performance guarantees in terms of the worst-case factor from optimality, a measure of the robustness of a heuristic. For each basic block, we calculated the ratio of the length of the schedule found by the heuristic over the length of the optimal schedule.... ..."

### Table 3.1: Average transformation time for the hard basic blocks using critical-path distance (CP) and range-based distance for the superior subgraph transformation with the semi-superior node transformation. Average Schedule

2007

### Table 4. Frequencies of interesting paths.

2004

"... In PAGE 4: ... C7BYCX AX CZB4C8CYB5 BP CGBYCX AX D0 (2) where CX AX D0 are all the interesting paths that contain the basic block sequence of the OL path (CX AX CZB4C8CYB5). For instance, for the loop in Table4 , C7BYBD AX BFB4C8BEB5 BP BYBD AX BE B7 BYBD AX BF, and this can be easily proved. We derive the estimates for the frequencies of the over- lapping paths by getting the upper and lower bounds of the frequencies of the paths in the following sections.... In PAGE 4: ...2.3 An Example Consider the control flow graph shown in Table4 . It is the same loop as used before except that numbers mark the flow values of different loop paths.... ..."

Cited by 6

### Table 4. Frequencies of interesting paths.

"... In PAGE 4: ... C7BY CX AX CZB4C8 CY B5 BP CG BY CX AX D0 (2) where CX AX D0 are all the interesting paths that contain the basic block sequence of the OL path (CX AX CZB4C8 CY B5). For instance, for the loop in Table4 , C7BY BDAXBFB4C8 BE B5 BP BY BDAXBE B7 BY BDAXBF , and this can be easily proved. We derive the estimates for the frequencies of the over- lapping paths by getting the upper and lower bounds of the frequencies of the paths in the following sections.... In PAGE 4: ...2.3 An Example Consider the control flow graph shown in Table4 . It is the same loop as used before except that numbers mark the flow values of different loop paths.... ..."