On 08-01-2011 00:12, Hirotsugu Asari wrote:
On Jan 7, 2011, at 8:36 PM, Rodrigo Rosenfeld Rosas wrote:
It seems a lot is needed for running a simple ruby script with 1.9 while much
less is required by 1.8.
Charles Nutter has talked recently about some tips to speed jruby startup time. One of
them is "export JAVA_OPTS='-d32'", but I get this on Linux:
Running a 32-bit JVM is not supported on this platform.
It probably means that you don't have a 32-bit JVM installed.
http://java.com/en/download/manual.jsp You'll want the second one in the Linux
list.
Is there any other tip that I could use for a 64-bit JVM on Linux to speed up
the startup time?
Having said that, if the 1.9 start-up time is driven by disk I/O for reading
lots of files (which I suspect is the case), the changing JVM won't help you
much.
[system]:~ $ time jruby -v -J-Djruby.debug.loadService.timing=true --1.9 -e 'p
1'
jruby 1.6.0.dev (ruby 1.9.2 trunk 136) (2011-01-07 19f7456) (Java HotSpot(TM)
64-Bit Server VM 1.6.0_22) [darwin-x86_64-java]
-> thread.jar
<- thread.jar - 5ms
-> rbconfig
<- rbconfig - 151ms
-> rubygems/defaults/operating_system
<- rubygems/defaults/operating_system - 15ms
-> enumerator
-> generator_internal
-> thread
<- thread - 0ms
<- generator_internal - 83ms
<- enumerator - 112ms
1
real 0m1.660s
user 0m1.502s
sys 0m0.152s
[system]:~ $ JAVA_OPTS=-d32 time jruby -v
-J-Djruby.debug.loadService.timing=true --1.9 -e 'p 1'
jruby 1.6.0.dev (ruby 1.9.2 trunk 136) (2011-01-07 19f7456) (Java HotSpot(TM)
Client VM 1.6.0_22) [darwin-i386-java]
-> thread.jar
<- thread.jar - 7ms
-> rbconfig
<- rbconfig - 204ms
-> rubygems/defaults/operating_system
<- rubygems/defaults/operating_system - 6ms
-> enumerator
-> generator_internal
-> thread
<- thread - 0ms
<- generator_internal - 25ms
<- enumerator - 46ms
1
1.45 real 1.06 user 0.14 sys
Yes, I guess the problem with jars is that the OS can't cache them like
they do with dynamic libraries and they will be loaded from disk each
time so that I would need a super fast HD for getting the program to
start faster :(
Thank you for your time,
Rodrigo
---------------------------------------------------------------------
To unsubscribe from this list, please visit:
http://xircles.codehaus.org/manage_email
