RFR: JDK-8266431: Dual-Pivot Quicksort improvements (Radix sort) [v11]

[Vladimir Yaroslavskiy]

On Fri, 26 Jan 2024 17:19:25 GMT, Srinivas Vamsi Parasa <duke at openjdk.org> wrote:

>> Hello Vamsi (@vamsi-parasa),
>> 
>> Could you please run the benchmarking of new DQPS in your environment with AVX?
>> 
>> Take all classes below and put them in the package org.openjdk.bench.java.util.
>> ArraysSort class contains all tests for the new versions and ready to use.
>> (it will run all tests in one execution).
>> 
>> https://github.com/iaroslavski/sorting/blob/master/radixsort/ArraysSort.java
>> https://github.com/iaroslavski/sorting/blob/master/radixsort/DualPivotQuicksort_a15.java
>> https://github.com/iaroslavski/sorting/blob/master/radixsort/DualPivotQuicksort_r20s.java
>> https://github.com/iaroslavski/sorting/blob/master/radixsort/DualPivotQuicksort_r20p.java
>> 
>> Many thanks,
>> Vladimir
>
> Hi Vladimir (@iaroslavski),
> 
> Please see the JMH data below.
> 
> Thanks,
> Vamsi
> 
> Benchmark              (builder)   (size)  Mode  Cnt       Score      Error  Units
> ArraysSort.Int.a15        RANDOM      600  avgt    4       7.096 ±    0.081  us/op
> ArraysSort.Int.a15        RANDOM     2000  avgt    4      44.014 ±    1.717  us/op
> ArraysSort.Int.a15        RANDOM    90000  avgt    4    4451.444 ±   71.168  us/op
> ArraysSort.Int.a15        RANDOM   400000  avgt    4   22751.966 ±  683.597  us/op
> ArraysSort.Int.a15        RANDOM  3000000  avgt    4  190326.306 ± 8008.512  us/op
> ArraysSort.Int.a15      REPEATED      600  avgt    4       1.044 ±    0.016  us/op
> ArraysSort.Int.a15      REPEATED     2000  avgt    4       2.272 ±    0.287  us/op
> ArraysSort.Int.a15      REPEATED    90000  avgt    4     412.331 ±   11.656  us/op
> ArraysSort.Int.a15      REPEATED   400000  avgt    4    1908.978 ±   30.241  us/op
> ArraysSort.Int.a15      REPEATED  3000000  avgt    4   15163.443 ±  100.425  us/op
> ArraysSort.Int.a15       STAGGER      600  avgt    4       1.055 ±    0.057  us/op
> ArraysSort.Int.a15       STAGGER     2000  avgt    4       3.408 ±    0.096  us/op
> ArraysSort.Int.a15       STAGGER    90000  avgt    4     149.220 ±    4.022  us/op
> ArraysSort.Int.a15       STAGGER   400000  avgt    4     663.096 ±   30.252  us/op
> ArraysSort.Int.a15       STAGGER  3000000  avgt    4    5206.890 ±  234.857  us/op
> ArraysSort.Int.a15       SHUFFLE      600  avgt    4       4.611 ±    0.118  us/op
> ArraysSort.Int.a15       SHUFFLE     2000  avgt    4      17.955 ±    0.356  us/op
> ArraysSort.Int.a15       SHUFFLE    90000  avgt    4    1410.357 ±   41.128  us/op
> ArraysSort.Int.a15       SHUFFLE   400000  avgt    4    5739.311 ±  128.270  us/op
> ArraysSort.Int.a15       SHUFFLE  3000000  avgt    4   41501.980 ±  829.443  us/op
> ArraysSort.Int.jdk        RANDOM      600  avgt    4       1.612 ±    0.088  us/op
> ArraysSort.Int.jdk        RANDOM     2000  avgt    4       6.893 ±    0.375  us/op
> ArraysSort.Int.jdk        RANDOM    90000  avgt    4     522.749 ±   19.386  us/op
> ArraysSort.Int.jdk        RANDOM   400000  avgt    4    2424.204 ±   63.844  us/op
> ArraysSort.Int.jdk        RANDOM  3000000  avgt    4   21000.434 ±  801.315  us/op
> ArraysSort.Int.jdk      REPEATED      600  avgt    4       0.496 ±    0.030  us/op
> ArraysSort.Int.jdk      REPEATED     2000  avgt    4       1.037 ±    0.083  us/op
> ArraysSort.Int.jdk      REPEATED    90000  avgt    4      57.763 ±    1.955  us/op
> ArraysSort.Int.jd...

Thank you, Vamsi (@vamsi-parasa)!

I will convert the data to more readable format.

-------------

PR Comment: https://git.openjdk.org/jdk/pull/13568#issuecomment-1912541236