We are getting zero timings for vector randomization
I thought our clock was progressive. Seems like it is not. Repeated queries in quick succession result in calls with 0 duration time. See this for example:
timings[000] [mapping] t= 13.80000 us
timings[000] [randvec] t= 0.00000 s
timings[000] [getpart] t= 100.00002 ns
Could it be that this is a limited precision issue? The smallest I have seen other times go is 99.99997 ns:
timings[000] [mapping] t= 12.70000 us
timings[000] [randvec] t= 0.00000 s
timings[000] [getpart] t= 99.99997 ns
We should investigate if it makes sense to keep the timings for randvec
and getpart
.