labath added a comment. In http://reviews.llvm.org/D13727#268747, @tberghammer wrote:
> Based on Pavel's example and some additional experimenting we done I am not > sure if std::async will give us any benefit even on Windows as it don't limit > the number of threads to the number of cores (because it can't do if it want > to implement the standard). I have no problem with having 2 implementation > (considering that one of them is very simple) but if it don't give us any > benefit, then I would prefer to ignore it. To back these with some numbers, let me explain what I have done: I have modified the f() function in my example to do some actual work (count to a very large number). On my Windows VM with hardware_concurrency()==16, after one second, I had about 70 threads running (more than 4 times the number of cpus). Since our tasks are going to be cpu-intensive, there is no point in having more than 16 of them running, and using std::async will cause a lot of contention, which will slow everything down. I believe the async implementation is not suited for our purpose here and we should not use it. http://reviews.llvm.org/D13727 _______________________________________________ lldb-commits mailing list lldb-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
