RFR: 8334816: compiler/c2/irTests/TestIfMinMax.java fails after 8334629

[Jasmine Karthikeyan]

On Mon, 1 Jul 2024 05:20:31 GMT, Tobias Hartmann <thartmann at openjdk.org> wrote:

>> This PR fixes an issue discovered during the review of #19650, where TestIfMinMax produced some IR failures. This issue is similar to the one in #18734, where a rare case could cause the same failure when the branch taken percent was too low. The fix there was to add 50 positive and 50 negative elements to the array, which guarantees that the IfNode will have a branch probability of at least 0.098. However, since the threshold has been increased from 0.01 to 0.18, the amount of sequential elements needs to be increased. I've increased it to 125 so the probability is guaranteed to be at least 0.24, which is larger than the threshold needed to convert to CMove.
>> 
>> Thoughts and reviews would be appreciated!
>
> Looks good to me. I submitted testing and will report back once it passed.

Thanks for the reviews @TobiHartmann and @chhagedorn!

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

PR Comment: https://git.openjdk.org/jdk/pull/19959#issuecomment-2200214160