Did a big 'ol refactor of the runner to make it much easier to do comparisons of different refs locally. This motivated moving the synced bitcoind into its own docker container for easier reuse over separate bench runner invocations.
While I was at it, I also added a flag called --run-counts (e.g. ibd:3,reindex:2) that allows the user to specify how many times a benchmark should be run. This should help to remove some of the noise of running benchmarks on a multi-use (vs. dedicated) system.
Separate git refs can now be specified for comparison with --commits a,b,c. Passing a single (or no) ref will result in the same behavior as before.
An example invocation with some of the new flags is below, which compares master to the v0.16.0 tag, showing a 24% improvement in IBD time!
Did a big 'ol refactor of the runner to make it much easier to do comparisons of different refs locally. This motivated moving the synced bitcoind into its own docker container for easier reuse over separate bench runner invocations.
While I was at it, I also added a flag called
--run-counts
(e.g.ibd:3,reindex:2
) that allows the user to specify how many times a benchmark should be run. This should help to remove some of the noise of running benchmarks on a multi-use (vs. dedicated) system.Separate git refs can now be specified for comparison with
--commits a,b,c
. Passing a single (or no) ref will result in the same behavior as before.An example invocation with some of the new flags is below, which compares
master
to thev0.16.0
tag, showing a 24% improvement in IBD time!