ignite-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Vladimir Ozerov (JIRA)" <j...@apache.org>
Subject [jira] [Comment Edited] (IGNITE-2762) Optimized StripedCompositeReadWriteLock to avoid TLS lookups.
Date Sun, 06 Mar 2016 16:58:40 GMT

    [ https://issues.apache.org/jira/browse/IGNITE-2762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15182234#comment-15182234
] 

Vladimir Ozerov edited comment on IGNITE-2762 at 3/6/16 4:58 PM:
-----------------------------------------------------------------

I implemented this thing after I saw a little hotspot in this place during sampling of PUT
in ATOMIC cache. However, it seems that this sample was not very precise because it was collection
for 1-2 minutes or so. 

Anyways, it is done. No observable speedup, but it could potentially give us protection from
collisions on thread-local entries (we see them from time to time during porfiling/sampling
of other benchmarks). 

But in future we should better analyze performance data before jumping on things like this.



was (Author: vozerov):
I implemented this thing after I saw a little hotspot in this place during sampling of PUT
in ATOMIC cache. However, it seems that this sample was not very precise because it was collection
for 1-2 minutes or so. 

Anyways, it is done. No observable speedup, but it could potentially give us protection from
collisions on thread-local entries (we see them from time to time during porfiling/sampling
of other benchmarks. 

But in future we should better analyze performance data before jumping on things like this.


> Optimized StripedCompositeReadWriteLock to avoid TLS lookups.
> -------------------------------------------------------------
>
>                 Key: IGNITE-2762
>                 URL: https://issues.apache.org/jira/browse/IGNITE-2762
>             Project: Ignite
>          Issue Type: Sub-task
>          Components: general
>    Affects Versions: 1.5.0.final
>            Reporter: Vladimir Ozerov
>            Assignee: Vladimir Ozerov
>             Fix For: 1.6
>
>
> Currently {{StripedCompositeReadWriteLock}} relies on thread-local index to find-out
thread index. 
> As these locks are usually used inside IgniteThread, we can assign special index to each
thread instead.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message