Join GitHub today
GitHub is home to over 50 million developers working together to host and review code, manage projects, and build software together.Sign up
This PR changes GC scheduling to use number of blocks rather than number of words for sweeping. I've also removed the 40%/60% split for mark/sweep, which I don't fully understand. With this PR, in most cases, the running time is better, fewer major GCs are performed, and space overhead remains within prescribed limits. See http://ocamllabs.io/multicore/ for updated graphs.