2 pointsby fbcouto9 hours ago2 comments
  • panny9 hours ago
    I can see your comment, but I'm afraid I can't do anything to vouch for you since it's [flagged] and [dead]. Maybe build some karma with submissions and try again later. Anyway, regarding

    >By implementing a parallel Multimerge algorithm in Rust

    Are you comparing your parallel sort vs Java's Arrays.parallelSort() or are you saying your multithreaded parallel sort is faster than single threaded Arrays.sort()?

    >Benchmarks show up to 38x speedups on high-entropy (random) datasets at the 10M element scale compared to standard Timsort/Arrays.sort.

    Also, did you do a proper JVM warmup for your benchmarks if it is a comparison of parallel sorts?

    • exabrial7 hours ago
      I was going to ask a similar question about threading. The Timsort in Java is known to be quite optimal and is an easy C2 compile target.
  • fbcouto9 hours ago
    [flagged]