RFR: 8296746: NativePRNG SecureRandom doesn't scale with threads

Sun Nov 13 18:44:30 UTC 2022

On Wed, 9 Nov 2022 20:15:57 GMT, Xubo Zhang <duke at openjdk.org> wrote:

> NativePRNG SecureRandom doesn’t scale with number of threads. The performance starts dropping as we increase the number of threads. Even going from 1 thread to 2 threads shows significant drop. The bottleneck is the singleton RandomIO instance. Making the RandomIO ThreadLocal helps in removing this bottleneck.
> 
> Here are the jmh  thrpt ops/s data for SecureRandomBench.nextBytes, algorithm=NativePRNGNonBlocking, datasize=64, notshared:
> 
> <html xmlns:v="urn:schemas-microsoft-com:vml"
> xmlns:o="urn:schemas-microsoft-com:office:office"
> xmlns:x="urn:schemas-microsoft-com:office:excel"
> xmlns="http://www.w3.org/TR/REC-html40">
> 
> <head>
> 
> <meta name=ProgId content=Excel.Sheet>
> <meta name=Generator content="Microsoft Excel 15">
> <link id=Main-File rel=Main-File
> href="file:///C:/Users/xbzhang/AppData/Local/Temp/msohtmlclip1/01/clip.htm">
> <link rel=File-List
> href="file:///C:/Users/xbzhang/AppData/Local/Temp/msohtmlclip1/01/clip_filelist.xml">
> 
> 
> 
> </head>
> 
> <body link="#0563C1" vlink="#954F72">
> 
> 
> 
> #threads | singleton RandomIO | ThreadLocal RandomIO
> -- | -- | --
> 1 | 1269133 | 1279088
> 2 | 862923 | 1362066
> 3 | 819734 | 1630522
> 4 | 809128 | 1614500
> 5 | 821514 | 1594965
> 6 | 808795 | 1545045
> 7 | 783050 | 1499388
> 8 | 787236 | 1470004
> 
> 
> 
> </body>
> 
> </html>

By using thread-local, is it possible to get the same random number in two threads?

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

PR: https://git.openjdk.org/jdk/pull/11069