Merging two cargo bench logs by picking the faster result for each bench


Does there already exist a tool that takes the outputs from two cargo bench runs and merges them by choosing the faster result for each bench?

Use case: RPi3 seems to throttle the CPU by a factor of 2 for a moment during each cargo bench run (for the duration of one to four individual benches in my case), so I’d like to filter out that noise before I compare the performance of different changesets/branches with cargo benchcmp. Obviously, since the hardware can’t become magically impossibly fast, it makes sense to pick the faster results, since the slower ones have to be the ones with noise from thermal throttling.


I patched cargo-benchcmp a bit (in a very unpolished not-suited-for-PR way) to accomplish this.