[jdk21u-dev] RFR: 8328085: C2: Use after free in PhaseChaitin::Register_Allocate()
Richard Reingruber
rrich at openjdk.org
Tue Dec 17 08:12:44 UTC 2024
On Tue, 10 Dec 2024 08:47:30 GMT, Richard Reingruber <rrich at openjdk.org> wrote:
> Hi all,
>
> This pull request contains a backport of commit https://github.com/openjdk/jdk/commit/93aa7e2fcf87c4dc62de4ea71be543ee677b11be.
>
> The commit being backported was authored by Richard Reingruber on 21 Nov 2024 and was reviewed by Tobias Hartmann and Martin Doerr.
>
> The original commit did not apply because the `TracePhase` constructor used above the removed line was changed.
> EDIT: the bot considers the backport to be clean.
> // Actually I had to resolve the minimal conflict I mentioned above.
>
> I'd consider the risk medium. There could be scenarios with higher memory usage in c2 register allocation.
> DaCapo tests didn't show higher memory usage.
> (the measuring code was part of the [original pull request](https://github.com/openjdk/jdk/pull/22200/commits))
>
>
> Max. ResourceArea size in KB after C2 PhaseChaitin::merge_multidefs
>
> DaCapo Benchmark Basline Pull Request
>
> avrora 2167 2167
> batik 2600 2509
> biojava 3125 3125
> cassandra 482 514
> eclipse 3408 3400
> fop 2863 2895
> graphchi 1331 1331
> h2 3204 3290
> h2o 638 514
> jme 2160 2160
> jython 9617 9649
> kafka 2577 2586
> luindex 2738 2902
> lusearch 2645 2645
> pmd 2931 2723
> spring 2958 3383
> sunflow 1437 1437
> tomcat 2668 2643
> tradebeans 2975 2663
> tradesoap 3124 2663
> xalan 3025 2826
> zxing 2473 2505
>
>
> The fix passed our CI testing: JTReg tests: tier1-4 of hotspot and jdk. All of Langtools and jaxp. JCK, SPECjvm2008, SPECjbb2015, Renaissance Suite, and SAP specific tests.
> Testing was done with fastdebug builds on the main platforms and also on Linux/PPC64le and AIX.
>
> Thanks, Richard.
Retrying... /integrate
-------------
PR Comment: https://git.openjdk.org/jdk21u-dev/pull/1217#issuecomment-2547755968
More information about the jdk-updates-dev
mailing list