cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ariel Weisberg (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-8614) Select optimal CRC32 implementation at runtime
Date Wed, 14 Jan 2015 17:36:35 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-8614?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14277276#comment-14277276
] 

Ariel Weisberg commented on CASSANDRA-8614:
-------------------------------------------

I am trying to disable the instruction so I can get the fallback and measure. I read in some
places that Adler32 was also supposed to get accelerated, but I only see stuff for CRC32.
Doesn't mean it doesn't do it on the sly, but we might get the speedup even using Adler32.

The JVM is accepting -XX:-UseCLMUL -XX:-UseCRC32Intrinsic, but it doesn't seem to work. I
keep getting 13 gigabytes/second. Something doesn't add up. The speed up going from 128-bytes
to 1 megabyte is also suspicious.

I also got 13 gigabytes/sec out of JDK 7. ??!!!?? I'm definitely not measuring what I think
I'm measuring.

> Select optimal CRC32 implementation at runtime
> ----------------------------------------------
>
>                 Key: CASSANDRA-8614
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-8614
>             Project: Cassandra
>          Issue Type: Improvement
>          Components: Core
>            Reporter: Ariel Weisberg
>            Assignee: Benedict
>              Labels: performance
>         Attachments: 8614.patch, Sample.java
>
>
> JDK 8 has support for an intrinsic for CRC32 that runs at 12-13 gigabytes/sec per core
in my quick and dirty test. PureJavaCRC32 is < 800 megabytes/sec if I recall and it has
a lookup table that evicts random cache lines every time it runs.
> In order to capture the benefit of that when it is available we can select a CRC32 implementation
at startup in a static block.
> If JDK 8 is not what is running we can fall back to the existing PureJavaCRC32 implementation.



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

Mime
View raw message