[
https://issues.apache.org/jira/browse/RNG-161?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17398524#comment-17398524
]
Alex Herbert commented on RNG-161:
----------------------------------
If recreating a RNG is too slow then I assume you are doing this very
frequently, and as such you are not using many numbers from the RNG. In this
case you should choose a RNG with a very small state such as SPLIT_MIX_64.
There is a performance test in the JMH examples module that compares creating a
RNG with the constructor verses using RandomSource.create. If you already have
a native seed then the difference is very small.
The fastest option through the RandomSource API and its equivalent would be:
{code:java}
Long seed = 123; // auto-boxing
UniformRandomProvider rng = RandomSource.create(RandomSource.SPLIT_MIX_64,
seed);
// Equal to
UniformRandomProvider rng = new SplitMix64(seed);
{code}
Can you let me know your use case for reseeding where the speed is critical?
> 是否可以动态设置seed？
> -------------
>
> Key: RNG-161
> URL: https://issues.apache.org/jira/browse/RNG-161
> Project: Commons RNG
> Issue Type: Improvement
> Components: client-api
> Affects Versions: 1.3
> Reporter: sun
> Priority: Major
>
> Does Commons RNG can set the seed by method,isn't use
> RandomSource.create，like random.setSeed() to set the seed ?
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
