[8] RFR: Save vector registers before LRB slowpath call

[Roman Kennke]

Have you checked aarch64?

Am 24. Oktober 2019 18:17:49 MESZ schrieb Aleksey Shipilev <shade at redhat.com>:
>jcstress caught a failure in sh/jdk8:
>
>$
>build/linux-x86_64-normal-server-fastdebug/images/j2sdk-image/bin/java
>-jar
>jcstress-tests-all-20190811.jar --jvmArgs
>"-XX:+UnlockExperimentalVMOptions -XX:+UseShenandoahGC
>-XX:TieredStopAtLevel=1 -XX:ShenandoahGCHeuristics=aggressive" -t
>tearing.buffers.FloatBufferInterleaveTest
>
>  1 matching test results.
>  [FAILED] o.o.j.t.tearing.buffers.FloatBufferInterleaveTest
>(JVM args: [-XX:+UnlockExperimentalVMOptions, -XX:+UseShenandoahGC,
>-XX:TieredStopAtLevel=1,
>-XX:ShenandoahGCHeuristics=aggressive])
>  Observed state   Occurrences   Expectation  Interpretation
>
>    0, 128, 128       883,871    ACCEPTABLE  Seeing all updates intact.
>
>0, 129, 127            42     FORBIDDEN  No default case provided,
>assume FORBIDDEN
>
>1, 127, 128           141     FORBIDDEN  No default case provided,
>assume FORBIDDEN
>
>1, 128, 127           144     FORBIDDEN  No default case provided,
>assume FORBIDDEN
>
>2, 127, 127            93     FORBIDDEN  No default case provided,
>assume FORBIDDEN
>
>
>Roman suggested we are not caring enough about xmm registers around LRB
>slowpath call, and indeed we
>are not. This is the fix:
>https://cr.openjdk.java.net/~shade/shenandoah/8u-lrb-xmm-save/webrev.01/
>
>Testing: hotspot_gc_shenandoah; affected jcstress tests
>
>-- 
>Thanks,
>-Aleksey

-- 
Diese Nachricht wurde von meinem Android-Gerät mit K-9 Mail gesendet.