cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Blake Eggleston (JIRA)" <j...@apache.org>
Subject [jira] [Created] (CASSANDRA-7621) Use jvm server machine code compiler
Date Fri, 25 Jul 2014 04:49:39 GMT
Blake Eggleston created CASSANDRA-7621:
------------------------------------------

             Summary: Use jvm server machine code compiler
                 Key: CASSANDRA-7621
                 URL: https://issues.apache.org/jira/browse/CASSANDRA-7621
             Project: Cassandra
          Issue Type: Improvement
          Components: Config
            Reporter: Blake Eggleston
            Priority: Minor
             Fix For: 2.0.10
         Attachments: 0001-adding-server-compiler-switch.patch

As far as I can tell, Cassandra is using the jvm's default client compiler. The attached patch
adds a line to cassandra-env.sh to tell it to use the jvm's server compiler. The server compiler
is meant for long running processes and is more aggressive in it's optimizations to the machine
code it generates, at the expense of machine code size and initial compilation speed.

I've run {{cassandra-stress  --num-keys=2000000 -K 50}} and {{cassandra-stress  --num-keys=2000000
-K 50 --operation=READ}} with both settings, and the runs with the server flag are 2-5% faster
on average on my Macbook Pro, with a 2.3GHz i7, 16GB of memory and an SSD.

http://www.oracle.com/technetwork/java/whitepaper-135217.html#2



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Mime
View raw message