It's a young library so there are several improvements since our last import.
One that I want in particular is the option to display benchmark results in an inverse sense of their current default "items/s" display.
BM_Print/100 108265082 ns 108161035 ns 6 924.547 items/s
I don't want to know that I can print 924 backtrace frames per second.
Because that number can't be reasoned about as the sum of its component operations.
If I saw the INVERSE, that printing takes 1.08ms per item, I could easily figure out where each of the milliseconds went by glancing at other measurements in the benchmark.
Anyway, it's an important fix for getting usable benchmark results without doing a bunch of annoying python postprocessing.