### Table 6: Large benchmark speedups and breakdowns.

1996

Cited by 4

### Table 6: Large benchmark speedups and breakdowns.

1996

Cited by 4

### Table 6: Large benchmark speedups and breakdowns.

1996

Cited by 4

### Table 4.2: Large benchmark speedups and breakdowns.

### Table VI: Speedup Comparison on Large CM-5

1995

Cited by 18

### Table 5.6: Speedups on the KSR1 for the large data set.

### Table 1. The large matrices that we use to measure speedups.

"... In PAGE 28: ... The selection criterion for these matrices was a factorization time of 20 seconds or more (by our code). These matrices are listed in Table1 . We do not claim that our code scales well on matrices that can be factored in several seconds on a uniprocessor.... ..."

### Table 3: Speedup summaries for large problem sizes; indicates an estimate on the speedup as we were unable to run the largest problem size on a single processor.

1999

"... In PAGE 19: ...97 124416 7 Table 5: Algorithmic and model summaries for large problem sizes on 16-processor SGI system. Table3 shows speedup results for large input sizes, for each application and system. To obtain meaningful values for speedup, we limit the problem sizes so no swapping to disk is necessary.... In PAGE 24: ... This suggests that we could perform MST computations on more highly connected graphs without much degradation in performance. Finally, as discussed earlier, the good speedup results for the minimum spanning tree application on large input sizes shown in Table3 should be quali ed, since the total work for sixteen processors (3.9 seconds) is signi cantly less than the total work for a single processor (6.... ..."

Cited by 9

### Table 3: Speedup summaries for large problem sizes; indicates an estimate on the speedup as we were unable to run the largest problem size on a single processor.

1999

"... In PAGE 19: ...97 124416 7 Table 5: Algorithmic and model summaries for large problem sizes on 16-processor SGI system. Table3 shows speedup results for large input sizes, for each application and system. To obtain meaningful values for speedup, we limit the problem sizes so no swapping to disk is necessary.... In PAGE 24: ... This suggests that we could perform MST computations on more highly connected graphs without much degradation in performance. Finally, as discussed earlier, the good speedup results for the minimum spanning tree application on large input sizes shown in Table3 should be quali ed, since the total work for sixteen processors (3.9 seconds) is signi cantly less than the total work for a single processor (6.... ..."

Cited by 9