Profile time measurements to check differences in graphpass output
greebie opened this issue · comments
Will put the profiling data here in case it is useful for research purposes.
Full output in this gist.
| file | nodes | filesize | Time Graphml | Time GEXF | CPU % | Maximum resident set size (kb) |
|---|---|---|---|---|---|---|
| cpp2.graphml | 218 | 61k | 0.3 s | 0.3s | 95% / 96% | 1488 / 1532 |
| snowden.graphml | 849 | 117k | 4.44 | 4.47 | 97% / 86% | 3960 / 4008 |
| albertahealth.graphml | 8989 | 1.4M | 258.76 | 235.35 | 97% / 98% | 318344 / 318400 |
| anarchist.graphml | 4825 | 2M | 234.42 | 224.51 | 98% / 99% | 96336 / 96768 |
| idlenomore.graphml | 1617 | 2.8M | 17.04 | 16.74 | 98%/99% | 15152 / 15012 |
| 3835-gephi.graphml†l | 13654 | 14.1 MB | 2168.96 | 2219.58 | 99/96% | 613916/614156 |
| 3365-gephi.graphml† | 6135 | 2 MB | 334.80 | 366.47 | 97%/98% | 142552/143668 |
| panama.graphml | 39058 | 6.9 MB | 23446.74 | /83% | / 3075844 | |
| 6671-gephi.graphml | 24197 | 10.5 MB | 11670.89 | 10099.63 | 93%/31% | 2196340/ 2197248 |
| test5kErdos.graphml†† | 5000 | 152 MB | 423.39 | 99% | 581188 | |
| test5kGrowth.graphml†† | 5000 | 312 KB | 179.80 | 99% | 20892 | |
| panama.graphml†† | 39058 | 6.9 MB | 18898.81 | 99% | 4556312 | |
| test50k.graphml†† | 50000 | 3.2 MB | 31621.91 | 99% | 227188 |
† used my faster computer with more memory (16GB) instead of the older one (8GB).
†† used VM with ~31 GB ram.
So far, it looks like the scripts are running at ~ 4 (n)^2, which is about right since we are doing four algorithms that are approximately O(n)^2.
The memory set is proportional to the size of the file.
Resolved for now with PR #23.