[
https://issues.apache.org/jira/browse/CASSANDRA-15215?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17447425#comment-17447425
]
Benedict Elliott Smith commented on CASSANDRA-15215:
----------------------------------------------------
Thanks [~Gerrrr], I took a quick look to see if I could shave any further time
and have the following
[suggestions|https://github.com/belliottsmith/cassandra/tree/15215-trunk]. On
my laptop at least the changes to {{ByteBuffer}} result in ~1ns reduction (or
about 6% improvement). I also took the liberty of modifying the
{{DataOutputPlus}} specification to use the top bytes of the register to permit
the {{ByteBuffer}} and {{DataOutputPlus}} implementations to look approximately
the same, and applied your original optimisation for VIntCoding to the default
implementation of {{writeBytes}}.
We should probably add a unit test or two to cover {{DataOutputPlus}} vint
coding and {{writeBytes}} before we commit, but overall the patch is looking
good.
> VIntCoding should read and write more efficiently
> -------------------------------------------------
>
> Key: CASSANDRA-15215
> URL: https://issues.apache.org/jira/browse/CASSANDRA-15215
> Project: Cassandra
> Issue Type: Bug
> Components: Local/Compaction, Local/SSTable
> Reporter: Benedict Elliott Smith
> Assignee: Aleksandr Sorokoumov
> Priority: Normal
> Fix For: 3.0.x, 3.11.x, 4.x
>
> Attachments: writeUnsignedVInt_megamorphic_BB.png,
> writeUnsignedVInt_megamorphic_DOP.png
>
>
> Most vints occupy significantly fewer than 8 bytes, and most buffers have >=
> 8 bytes spare, in which case we can construct the relevant bytes in a
> register and memcpy them to the correct position. Since we read and write a
> lot of vints, this waste is probably measurable, particularly during
> compaction and flush, and can probably be considered a performance bug.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
