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?
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?
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.
[flagged]