Discussion:
[RNG] Towards release 1.2
Gilles
2018-11-15 16:44:46 UTC
Permalink
Hi.

Only 1 issue left that is targeted for the next release.[1]
Any work in progress that is ready for inclusion?

Gilles

P.S. Review(s) welcome (as usual).

[1]
https://issues.apache.org/jira/projects/RNG?selectedItem=com.atlassian.jira.jira-projects-plugin%3Arelease-page&status=unreleased


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-***@commons.apache.org
For additional commands, e-mail: dev-***@commons.apache.org
Gary Gregory
2018-11-15 16:59:25 UTC
Permalink
Go for it :-)

Gary
Post by Gilles
Hi.
Only 1 issue left that is targeted for the next release.[1]
Any work in progress that is ready for inclusion?
Gilles
P.S. Review(s) welcome (as usual).
[1]
https://issues.apache.org/jira/projects/RNG?selectedItem=com.atlassian.jira.jira-projects-plugin%3Arelease-page&status=unreleased
---------------------------------------------------------------------
Alex Herbert
2018-11-15 23:09:03 UTC
Permalink
Hi,

The outstanding issue (RNG-57) is a performance enhancement in the core library [1].

The speed improvement has been demonstrated. However the effect on the randomness has not been tested using the 3rd party DieHarder and BigCrush test suites. I would be reluctant to change the core library without verifying the impact of the change.

A second issue is that the change causes the current test suite to fail. This has been investigated and is due to the fact that the test suite is performing many tests of randomness with statistical tests for acceptance. The modification alters the sequences generated by the test suite from their fixed seeds and the results are no longer accepted. Basically after enough tests for randomness something will almost certainly fail. That is what is happening after the enhancement is put into the core module.

Ideally the suite would work with random seeds and use a more selective set of tests (to reduce the chance for failures) but I have not had time to look at reengineering the test suite.

It may be better to delay this for a later release.

Regards,

Alex


[1] https://issues.apache.org/jira/browse/RNG-57 <https://issues.apache.org/jira/browse/RNG-57>
Post by Gary Gregory
Go for it :-)
Gary
Post by Gilles
Hi.
Only 1 issue left that is targeted for the next release.[1]
Any work in progress that is ready for inclusion?
Gilles
P.S. Review(s) welcome (as usual).
[1]
https://issues.apache.org/jira/projects/RNG?selectedItem=com.atlassian.jira.jira-projects-plugin%3Arelease-page&status=unreleased
---------------------------------------------------------------------
Gilles
2018-11-16 01:29:11 UTC
Permalink
Hello.
Post by Alex Herbert
Hi,
The outstanding issue (RNG-57) is a performance enhancement in the core library [1].
The speed improvement has been demonstrated. However the effect on
the randomness has not been tested using the 3rd party DieHarder and
BigCrush test suites.
I could not do it before getting your latest PR update with
this enhancement... :-)
Post by Alex Herbert
I would be reluctant to change the core library
without verifying the impact of the change.
+1
Post by Alex Herbert
A second issue is that the change causes the current test suite to
fail. This has been investigated and is due to the fact that the test
suite is performing many tests of randomness with statistical tests
for acceptance. The modification alters the sequences generated by
the
test suite from their fixed seeds and the results are no longer
accepted. Basically after enough tests for randomness something will
almost certainly fail. That is what is happening after the
enhancement
is put into the core module.
Ideally the suite would work with random seeds
+1

I'll make that change.
Post by Alex Herbert
and use a more
selective set of tests (to reduce the chance for failures)
In the meantime, we can configure junit to rerun "flaky" tests.
Post by Alex Herbert
but I have
not had time to look at reengineering the test suite.
It is perhaps not necessary if it is expected that the same
test won't fail several times in a row.
The testing does not take very long (~10 s for "core");
hence there isn't much to be gained by having the minimal
set of unit tests.
Post by Alex Herbert
It may be better to delay this for a later release.
Sure.

Best,
Gilles
Post by Alex Herbert
Regards,
Alex
[1] https://issues.apache.org/jira/browse/RNG-57
<https://issues.apache.org/jira/browse/RNG-57>
Post by Gary Gregory
Go for it :-)
Gary
On Thu, Nov 15, 2018 at 9:44 AM Gilles
Post by Gilles
Hi.
Only 1 issue left that is targeted for the next release.[1]
Any work in progress that is ready for inclusion?
Gilles
P.S. Review(s) welcome (as usual).
[1]
https://issues.apache.org/jira/projects/RNG?selectedItem=com.atlassian.jira.jira-projects-plugin%3Arelease-page&status=unreleased
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-***@commons.apache.org
For additional commands, e-mail: dev-***@commons.apache.org

Loading...