Adds an initial version of a plot to show the fold-speedup we get from adding threads to the various programs. This is probably a supplemental figures, as the basic message can be summed up in a sentence. Sgkit and savvy benefit from additional threads, whereas the bcftools based approaches can't effectively used more than 2 threads (although somewhat complicated by the pipelining)
Here's an initial version. We might imagine doing it for 2, 4, 8 and 16 threads to see when sgkit and savvy top out in terms of effective thread usage.
There's problem here though: sgkit is doing mysteriously better than k-fold increase. What gives?
I'm going to merge this and create an issue to track.
Adds an initial version of a plot to show the fold-speedup we get from adding threads to the various programs. This is probably a supplemental figures, as the basic message can be summed up in a sentence. Sgkit and savvy benefit from additional threads, whereas the bcftools based approaches can't effectively used more than 2 threads (although somewhat complicated by the pipelining)
Here's an initial version. We might imagine doing it for 2, 4, 8 and 16 threads to see when sgkit and savvy top out in terms of effective thread usage.
There's problem here though: sgkit is doing mysteriously better than k-fold increase. What gives?
I'm going to merge this and create an issue to track.