I'm almost certain this will be slower than OpenMP because it uses a centralized task queue that gets locked. OpenMP uses a decentralized work-stealing task que

I'm almost certain this will be slower than OpenMP because it uses a centralized task queue that gets locked. OpenMP uses a decentralized work-stealing task queue called the Chase-Lev Deque. There's a C implementation in this paper: https://fzn.fr/readings/ppopp13.pdf https://fzn.fr/readings/ppopp13.pdf