Understanding systematic errors

[jowens]

Performance variance study of min-/max-/avg-process-time, and looking at the impact of GPU clock/thermals/power, caching, etc.

[neoblizz]

• We can do a mini case study later on on the impact of GPU clocks, etc...
  --

• Performance variance study of min-/max-/avg-process-time
  @jowens Can we try this for delaunay graphs or any of the synthetic graphs where number of nodes and edges scale up?

[jowens]

Sure, but do we have enough runs to do that? I mean, each run we have has 10 data points, right?, but that's all. It would be hard to get anything meaningful out of that. I think we'd need a lot more data points. For instance, if the OS decided to do something compute-heavy right when a run was starting, that might completely skew the results. But you tell me!