Hi,

The threads are created only 1 time*.

Have I missed something? I checked sample_sort and I saw several sets of calls to std::async which is likely to create and destroy threads several times.

*parallel_sort, sample_sort* : this algorithm is commonly accepted as the fastest for the parallel stable sort. The implementation is *notoriously faster than the provided by the GCC compiler* ( based on Open MP ) using less memory. TBB don?t provide parallel stable sort, and only have an experimental code in their web pages

Indeed, I found it highly interesting. Cheers, Christophe