Add benchmarking to continuous integration about lokidb HOT 3 OPEN

Not sure the vm/sandbox which travis will use an provide reliable/consistent benchmarks? If run at 2am maybe but not the same as during peak?

from lokidb.

The idea is to generate benchmarks for the previous and current implementation at one PR. Not the cycles/seconds are interesting but the percentage difference.

I never tried it out but maybe it works. ;)

from lokidb.

Oh, I suppose if you have both old and new paths available at the same time, you might supersample, toggling back and forth 'n' number of times to accomodate fluctuations in cpu activity so that both are being tested along the same peaks and valleys of (shared) cpu activity... better for really fast ops that you can repeat many times (1000+ or so samples, interlacing old and new perf tests)

from lokidb.