Math.min(II) polluted compilation

[Galder Zamarreno]

Turns out you can pass in -XX:CompileThresholdScaling to delay the
compilation. I just tried with a value of 2.0 and that was enough to push
Math::min compilation into the benchmark and it has the desired effect.
Thanks Tobias H for the suggestion

On Thu, Feb 27, 2025 at 6:54 PM Andrew Haley <aph at redhat.com> wrote:

> On 2/27/25 16:28, Galder Zamarreno wrote:
> > Is there a way to avoid this issue? Can JMH somehow instruct HotSpot to
> deopt Math.min(II) before the warmup phase of the benchmark runs to avoid
> pollution?
>
> I don't think that would help, because even after deopt the MDO
> remains. Surely you're asking for something completely unrealistic,
> though? In the real world of real apps, Math.min() is bound to be
> polluted by all of the different ways it's been called, from many
> users.
>
> You may be able to use WhiteBox in your JMH code. There is a way to
> deoptimize a specific method or the whole call stack, which you can
> call from Java. There are lots of examples of WhiteBox use in the
> JDK test suite. Using WhiteBox is rather fiddly, though.
>
> --
> Andrew Haley  (he/him)
> Java Platform Lead Engineer
> Red Hat UK Ltd. <https://www.redhat.com>
> https://keybase.io/andrewhaley
> EAC8 43EB D3EF DB98 CC77 2FAD A5CD 6035 332F A671
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/jmh-dev/attachments/20250320/ac664eb3/attachment-0001.htm>