RFR: 8326227: Fix a rare rounding error affecting RandomSupport::computeNextGaussian

[Jaikiran Pai]

On Mon, 5 Feb 2024 04:25:16 GMT, Chris Hennick <duke at openjdk.org> wrote:

> This provides a slightly more accurate bounding limit for `computeNextExponentialSoftCapped` when the computed bound is greater than `(1.0p53 - 1.0) * DoubleZigguratTables.exponentialX0`. This could cause the `while (computeNextExponentialSoftCapped(rng, limit) < limit)` check in `computeNextGaussian` on line 1402 to always be true, making the `nextGaussian` runtime unbounded in the worst case.

Hello Chris @Pr0methean, I don't have knowledge of this area to know if this is indeed a bug that is being addressed. I think it will be better to open a bug for this through https://bugs.java.com/bugdatabase/ and once it gets triaged and assigned JDK bug id then you can link this PR to that issue so that it goes through the usual review process.

Hello Chris, the issue has now made it to the JDK project https://bugs.openjdk.org/browse/JDK-8326227. Please update the title of this PR to `8326227: Rounding error that may distort computeNextGaussian results` to officially trigger the review.

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

PR Comment: https://git.openjdk.org/jdk/pull/17703#issuecomment-1951881768
PR Comment: https://git.openjdk.org/jdk/pull/17703#issuecomment-1953587306