Should libcxx+libcxxabi+libunwind speed up concurrent stack unwinding?

Hi,

We are seeing terrible performance when throwing exceptions in multiple threads on Linux and FreeBSD but on macOS the performance is fine. It appears that there is a mutex in the unwinding code (if I read things correctly its in Unwind_Find_FDE). I tried linking with libcxx+libcxxabi+libunwind in the hopes that it would speed things up on Linux but it did not.

Should linking with libcxx+libcxxabi+libunwind speed up concurrent stack unwinding?

I wrote a little C++ demo that illustrates the problem…
https://github.com/clasp-developers/ctak