[jira] [Commented] (CASSANDRA-3286) Performance issue in ByteBufferUtil

["Jonathan Ellis (Commented) (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/CASSANDRA-3286?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13118204#comment-13118204 ] 

Jonathan Ellis commented on CASSANDRA-3286:
-------------------------------------------

it looks like the 5th and 6th parameters to compareUnsigned are lengths, not offsets, so we should just pass remaining() instead of p + remaining
                
> Performance issue in ByteBufferUtil
> -----------------------------------
>
>                 Key: CASSANDRA-3286
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-3286
>             Project: Cassandra
>          Issue Type: Improvement
>          Components: Core
>            Reporter: T Jake Luciani
>            Assignee: T Jake Luciani
>            Priority: Minor
>             Fix For: 1.0.1
>
>         Attachments: v1-0001-CASSANDRA-3286-perf-improvement.txt
>
>
> Profiling 1.0 we can see ByteBufferUtil.compareUnsigned is slow.
> {code}
> Excl. Incl. Incl. Incl. Name 
> User CPU User CPU Sync Wait Sync Wait 
> sec. sec. sec. Count 
> 318.491 318.491 1.400 113786 <Total> 
> 40.561 40.561 0. 0 <Unknown> 
> 18.972 19.093 0. 0 <static>@0xd949 (<libpthread-2.5.so>) 
> 17.718 18.730 0. 0 sun.security.provider.MD5.implCompress(byte[], int) 
> 14.396 14.396 0. 0 __pthread_cond_signal 
> 8.908 8.908 0. 0 org.apache.cassandra.utils.ByteBufferUtil.compareUnsigned(java.nio.ByteBuffer, java.nio.ByteBuffer) 
> 7.435 7.688 0. 0 __pthread_cond_timedwait 
> 7.127 7.182 0. 0 <static>@0xd8c9 (<libpthread-2.5.so>) 
> 7.072 7.072 0. 0 jbyte_disjoint_arraycopy 
> 6.764 39.065 0. 0 org.apache.cassandra.utils.ReducingIterator.computeNext() 
> 6.533 17.575 0. 0 java.util.concurrent.ConcurrentSkipListMap.doPut(java.lang.Object, java.lang.Object, boolean) 
> 6.346 6.346 0. 0 com.sun.crypto.provider.SunJCE_c.a(byte[], int, byte[], int) 
> 5.378 5.433 0. 0 send 
> 4.861 6.643 0.000 1 org.apache.cassandra.utils.ByteBufferUtil.read(java.io.DataInput, int) 
> 4.410 9.260 0. 0 org.apache.commons.collections.iterators.CollatingIterator.least() 
> 4.355 4.355 0. 0 java.io.ByteArrayOutputStream.write(int) 
> 4.300 6.632 0. 0 java.io.ByteArrayOutputStream.write(byte[], int, int) 
> 3.827 30.190 0. 0 org.apache.cassandra.dht.RandomPartitioner.decorateKey(java.nio.ByteBuffer) 
> 3.783 23.954 0. 0 org.apache.cassandra.utils.FBUtilities.hash(java.nio.ByteBuffer[]) 
> 3.783 3.860 0. 0 <static>@0xd486c (<libc-2.5.so>) 
> 3.739 3.739 0. 0 clock_gettime
> {code}
> We can avoid the problem when the ByteBuffer has a backing array

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira