The purpose of these benchmarks is to illustrate that even though opqit may help to reduce compile times and ultimately reduce coupling in your software, it does not come without cost.
Wrapping native iterators in
opaque_iterators will have a detrimental effect on performance. So it may be wise to avoid using them in performance-critical code.
I only ran tests on Windows, but I don't see how the general trend would be any different on another O.S. The benchmarks application is in the opqit repository, though, if you would like to try for yourself.
MinGW g++ 4.2.1
Microsoft Visual C++ 2009