Boost histogram performance slows down if we add more histo of same kind(which are not used)

Dear experts,

We plan to use boost histogram in QC. While checking performance of boost histogram vs ROOT histograms I found that in the "Fill process " boost histogram slows down if you define another boost histogram (the second histogram is outside the clock, which I technically don’t even use anywhere).
However the performance of ROOT histograms do not change much if I add more THn histograms which I don’t use in the macro. A small macro which I used to check is attached here. TestSpeed.cpp (2.3 KB)

A detailed checks and findings are shown in the slides attached (from slide 8) QCTUM_16092021.pdf (3.2 MB)

The behavior of boost libraries under optimization is puzzling to me. Moreover to see if this behaviors is associated with particular clock, I have tried steady clock instead of high resolution clock, the results are same nevertheless.

I am wondering if someone is already see such behavior.
Any suggestion/inputs will be much appreciated.

Thanks a lot!
Bhawani Singh