>[EMAIL PROTECTED] wrote:
>>So for me 1.5 out of the 7 seconds is due to the verification (if I
>>understand the working of the option "-Xverify:none" right). The
>>remaining 5.77 seconds I supppose is JRuby:s own initialization.
>
>For the fastest startup (and slowest runtime performance, incidentally), use
>the following options:
>
>-J-Djruby.reflection=true
> disable loading the generated invoker stubs
>-C
> disable compilation
>-J-Xverify:none
>
>It will still be slower than I'd like it. It also makes me sad to see how fast
>"jruby -help" is, which doesn't initialize the runtime at all. That's more the
>true startup speed of "just the JVM" and the rest is loading all our classes
>and initializing the runtime.
Here's timimg data on all those options in one place.
Testing with r5169
$ svn up
$ ant clean
$ ant
$ ant jruby-nailgun
$ time jruby -e 'puts :hello'
real 0m6.306s
user 0m4.006s
sys 0m0.304s
$ time jruby -J-Xverify:none -e 'puts :hello'
real 0m4.327s
user 0m3.412s
sys 0m0.249s
$ time jruby -J-Xverify:none -J-Djruby.reflection=true -e 'puts :hello'
real 0m2.914s
user 0m2.690s
sys 0m0.180s
$ time jruby -J-Xverify:none -J-Djruby.reflection=true -C -e 'puts :hello'
real 0m2.613s
user 0m2.430s
sys 0m0.170s
There's some problem with nailgun -- it's times are much slower than I expected
-- the second request took 5s.
$ jruby-ng-server
NGServer started on all interfaces, port 2113.
$ time jruby-ng -e "puts 'hello'"
real 0m12.770s
user 0m0.003s
sys 0m0.009s
$ time jruby-ng -e "puts 'hello'"
real 0m4.984s
user 0m0.003s
sys 0m0.009s
Testing just Java startup:
$ time jruby -h
real 0m1.111s
user 0m0.164s
sys 0m0.050s
---------------------------------------------------------------------
To unsubscribe from this list please visit:
http://xircles.codehaus.org/manage_email
