Is there a benchmark of TBB concurrent containers versus their standard counterparts ? For example tbb::concurrent_vector compared to std::vector, in purely sequential use ?
In other words, is it possible from now on,to use TBB concurrent containers everywhere, even when not needed ? This would give the possiblility to later add parallelism, without design changes.
If, on the other hand, these concurrent containers are significantly slower than the 'classic' standard ones, and if there is no way to locally disable synchronization, one may have to prioritize one execution mode over the other depending on where the bottlenecks are, this choice becoming maybe obsolete if other changes in his/her software move these performance bottlenecks elsewhere.