Patch to improve primitives Array.sort()

Paul Sandoz paul.sandoz at oracle.com
Tue May 26 07:49:51 UTC 2015


On May 22, 2015, at 9:56 PM, "O'Leary, Kristen" <Kristen.O'Leary at gs.com> wrote:
> Hi Paul,
> 
> We've created an additional test  based on your suggestion: an array of size 10,000,000, 32 pair flips, a run of zeroes in the middle, and 32 pair flips at the end. Here are the results for int:
> Benchmark                                                   (listType)                   Mode  Cnt   Score      Error   Units
> SortingIntTestJMH.sortCurrentWay  pairFlipZeroPairFlip  thrpt   10    4.886   ± 0.031   ops/s
> SortingIntTestJMH.sortNewWay        pairFlipZeroPairFlip  thrpt   10    14.793 ± 0.217   ops/s
> 
> We also created a similar test which is 10, 5 repeated 32 times, a run of 100 in the middle, and 10, 5 repeated 32 times at the end. Here are the results again for int:
> Benchmark                                                   (listType)                                       Mode  Cnt   Score      Error   Units
> SortingIntTestJMH.sortCurrentWay   pairFlipOneHundredPairFlip   thrpt   10      4.936 ± 0.040  ops/s
> SortingIntTestJMH.sortNewWay         pairFlipOneHundredPairFlip    thrpt   10    18.472 ± 0.217  ops/s
> 
> As Moh mentioned on a different thread, we will work with Sunny on getting the tests to you.
> 

Thanks those number look good.

Paul.



More information about the core-libs-dev mailing list