On Dec 6, 2007 7:00 PM, Stephen Bannasch <[EMAIL PROTECTED]> wrote:
>
> What version of Java are you running? I'm running 1.5.0_07 on MacOS 10.4.11
> -- but that shouldn't be the issue because I was getting times like yours
> about 3 weeks ago.
>
[...]
>
> My times are absurdly long
>
> $ time jruby-ng -e "puts 'hello'"
> hello
>
> real 0m37.544s
> user 0m0.004s
> sys 0m0.017s
>
[...]
> Does anybody else see the kind of behavior I'm reporting?
>
I just tried nailgun for the first time, and below are my
measurements. I use Java 1.5.0_13 on MacOS 10.5.1 (i.e. Leopard). I
used the same SVN revision as you did (-r5169). First a normal run of
jruby:
$ time ./bin/jruby -e 'print :hello'
hello 7.05 real 5.82 user 0.60 sys
And then the nailgun runs:
$ ./bin/jruby-ng-server &
$ while true ; do time ./bin/jruby-ng -e 'print :hello' ; sleep 10 ; done
hello 6.65 real 0.01 user 0.04 sys
hello 1.05 real 0.01 user 0.04 sys
hello 0.82 real 0.01 user 0.04 sys
hello 0.84 real 0.01 user 0.04 sys
hello 0.61 real 0.01 user 0.03 sys
hello 0.82 real 0.01 user 0.04 sys
hello 0.58 real 0.01 user 0.04 sys
hello 0.82 real 0.01 user 0.04 sys
hello 0.55 real 0.01 user 0.04 sys
hello 0.85 real 0.01 user 0.04 sys
hello 0.67 real 0.01 user 0.04 sys
hello 0.79 real 0.01 user 0.04 sys
hello 0.53 real 0.01 user 0.04 sys
So my MacOS setup seem to work more normally than yours.
I noticed a somewhat strange pattern in these numbers: every second
number is close to 0.80 and the other half are close to 0.55. This has
been very consistent over several runs. I have absolutely no idea why.
(I added the "sleep 10" in the loop to "stabilize" the system between
each iteration in the loop, but it didn't affect the results.)
/Johan Holmberg
---------------------------------------------------------------------
To unsubscribe from this list please visit:
http://xircles.codehaus.org/manage_email
