13 ms·
I'm surprised they didn't try std::atomic_long or a compare-and-swap.
by titzer 4y ago
I'm surprised they didn't try std::atomic_long or a compare-and-swap.
- waynesonfire 4y agowhy? it doesn't matter. that'll will only get you so far, as you scale, you'll still run into synchronization contention. Maybe going from 16 to 64 threads you'll get some nice wins, but so what? What are you going to do when you need to double the number of threads again?