[
https://issues.apache.org/jira/browse/HBASE-13307?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
stack updated HBASE-13307:
--------------------------
Attachment: traces.9.svg
[~larsh] Yeah, will be interested to know if you see a difference. There are
some pretty critical methods that could be broken up if there is speed to be
had here. Below is partial list:
{code}
@ 249 org.apache.hadoop.hbase.regionserver.ScanQueryMatcher::match (823
bytes) hot method too big
@ 29
org.apache.hadoop.hbase.regionserver.HRegion$RegionScannerImpl::nextInternal
(611 bytes) hot method too big
! @ 20
org.apache.hadoop.hbase.regionserver.StoreScanner::next (774 bytes) hot
method too big
! @ 20
org.apache.hadoop.hbase.regionserver.StoreScanner::next (774 bytes) hot
method too big
! @ 52
org.apache.hadoop.hbase.io.hfile.HFileReaderV3$ScannerV3::blockSeek (734 bytes)
hot method too big
{code}
I doubled the MaxInlineSize making it 72 instead of default 34. This seemed to
do about same as default but better than the 512. It made my flame graph have a
different shape though (the HBB._get is showing again which is odd since it is
inlined... previous it seemed to indicate that the hosted method was not
inlined... need to dig more).
> Making methods under ScannerV2#next smaller allows them to be inlined gaining
> us 7% more throughput
> ---------------------------------------------------------------------------------------------------
>
> Key: HBASE-13307
> URL: https://issues.apache.org/jira/browse/HBASE-13307
> Project: HBase
> Issue Type: Sub-task
> Components: Scanners
> Reporter: stack
> Assignee: stack
> Fix For: 2.0.0, 1.1.0
>
> Attachments: 13307.txt, traces.8.svg, traces.9.svg
>
>
> See parent issue for patch and evidence.
> I was looking at graphs of our scan and found that methods were 'too big' to
> be inlined (looking at jvm compilation and inlining output flags -- see
> parent for list). Changing method size helped some. Let me commit the
> resultant patch.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
