Open zhiayang opened 3 years ago
Due to the current architecture of the library, it is not possible to arrange the locks in an appropriate way to achieve this.
Maybe this will be looked at again when we need to eke more performance out of nabs in the future.
When building in parallel, it is often the case that multiple threads will simultaneously invoke
find_library()
. While this isn't detrimental (since the global state is correctly synchronised), it is wasted work.