I'm not an expert, however, while your OCA (see below) is being processed I would recommend to provide more comprehensive stats. Different lengths of an array, different flavours of data: sorted, sorted in the reverse order, random, typical expected case(s), etc.
It seems that this particular functionality ( sort(byte[] ) hasn't changed since the JDK 8. However, you should probably add the current JDK to your comparison. One necessary step towards making this eligible for inclusion in the JDK would be to sign the OCA https://www.oracle.com/technetwork/community/oca-486395.html Keep in mind that it is not by any means a guarantee that your change will be included. Once the OCA has been signed and processed, the code then can be discussed and evaluated by experts. -Pavel > On 14 Jun 2019, at 16:34, Rodion Efremov <codero...@gmail.com> wrote: > > Good evening! > > I managed to improve the JDK 8 java.util.Arrays.sort(byte[]) > performance-wise [1]. The (warmed up) demonstration program produces more > or less optimistic results on arrays of 1e8 bytes: > > seed = 1560526264738 > java.util.Arrays.sort(byte[]) in 87.643701 milliseconds. > java.util.Arrays.parallelSort(byte[]) in 301.329701 milliseconds. > net.coderodde.Arrays.sort(byte[]) in 62.0763 milliseconds. > Algorithms agree: true > > I would like to hear any comments on how to make it eligible for inclusion > in JDK. > > Best regards, > Rodion E. > > References: > [1] https://gist.github.com/coderodde/493407bc1c57352b53c2aa18b5c9a7a8