RFR 8221507: Implement JFR Events for Shenandoah

[Aleksey Shipilev]

On 3/29/19 11:08 PM, Roman Kennke wrote:
>> On Tue, Mar 26, 2019 at 5:16 PM Roman Kennke <rkennke at redhat.com <mailto:rkennke at redhat.com>> wrote:
>>
>>      > Please review this patch that adds support for two new JFR events
>>      > ShenandoahHeapRegionStateChange and ShenandoahHeapRegionInformation.
>>      >
>>      > Bug: https://bugs.openjdk.java.net/browse/JDK-8221507
>>      > Webrev: http://cr.openjdk.java.net/~kdobson/53476/webrev/
>>      >
>>      > The events appears to result in ~2% overhead though there is some
>>      > uncertainty as to whether some of that is noise.
>>      >
>>      > Thanks,
>>      >
>>      > Ken Dobson
>>
>>     Very nice!
>>
>>     The overhead is only present with JFR active, I assume?
>>
>> We've noticed that there is some overhead when JFR is included in the jdk but not running. This
>> may be because the event object has to be instantiated at each transition whether it's enabled or
>> not. You can see the exact numbers in the email I previously sent you.
> 
> Hmm, ok. This only ever happens when regions change state, right? I doubt that this is frequent
> enough to make an actual difference. I suspect that what you see is noise. I'm good. Let's also hear
> Aleksey's opinion though.

It looks fine to me. I ran a few tests of my own, and overheads clearly visible with JFR turned on,
but are not bad. The overheads with JFR turned off are slim to none.

Ken, you might want to clean up trailing whitespaces here:

+void ShenandoahJFRSupport::register_jfr_type_serializers() {
+  JfrSerializer::register_serializer(TYPE_SHENANDOAHHEAPREGIONSTATE,
+                                     false,
+                                     true,
+                                     new ShenandoahHeapRegionStateConstant());
+}
+#endif


-Aleksey