RFR(M) 8201447: C1 does backedge profiling incorrectly

Vladimir Kozlov vladimir.kozlov at oracle.com
Wed May 16 01:32:53 UTC 2018


Vladimir K

> On May 15, 2018, at 4:53 PM, Tom Rodriguez <tom.rodriguez at oracle.com> wrote:
> Looks good.
> tom
> Igor Veresov wrote:
>> C1 incorrectly incremented a back branch counter for both arms of a conditional branch if one of the arms were a back branch. This fix makes the increment conditional and applicable only to the appropriate arm.
>> Webrev: http://cr.openjdk.java.net/~iveresov/8201447/webrev.01/
>> Internal performance testing showed no adverse effects of the change.
>> Thanks,
>> igor

More information about the hotspot-compiler-dev mailing list