“Don’t use this feature unless you can demonstrate that this is where the biggest bottleneck is” is such a buzzkill. Like, why would you not want a multipurpose multitree multithread-friendly data structure with a custom allocator??

Comments