Performance of JRuby StringScanner

Ruby JRuby
1

I’ve been playing around with lexers that rely on the Ruby StringScanner
to recognize lexical tokens. For large texts, the lexers are roughly
twice as slow in JRuby as in MRI. I haven’t probed any further, but am
wondering if there are known performance issues in this area. It looks
like both MRI and JRuby have “native” implementations of StringScanner.

// Gregory

To unsubscribe from this list, please visit:

http://xircles.codehaus.org/manage_email
2

Hi Gregory,

I’m not aware of any performance issues with StringScanner in JRuby.
It should be generally fast.

In my experiments I see that indeed by default JRuby is on par or a
bit faster than MRI, and in -J-server mode it is sometimes TWICE as
fast.

It would be great if you could share the sources of your benchmark or
provide a simplified benchmark that shows the problem.

Thanks,
–Vladimir

On Fri, Jun 13, 2008 at 6:58 AM, Gregory M. [email protected]
wrote:

http://xircles.codehaus.org/manage_email

To unsubscribe from this list, please visit:

http://xircles.codehaus.org/manage_email
3

Yes, you’re right. On closer inspection, the JRuby StringScanner is
out-performing the MRI scanner; it’s the additional string processing
that I have around the scanner that is resulting in a much slower run
with JRuby. I’ll have to narrow this down.

// Gregory

Vladimir S. wrote:

provide a simplified benchmark that shows the problem.

JRuby have “native” implementations of StringScanner.

To unsubscribe from this list, please visit:

http://xircles.codehaus.org/manage_email

To unsubscribe from this list, please visit:

http://xircles.codehaus.org/manage_email
4

What’s the additional string processing that slows JRuby down ?

Gregory M. pisze:

I’m not aware of any performance issues with StringScanner in JRuby.
–Vladimir

there are known performance issues in this area. It looks like both

To unsubscribe from this list, please visit:

http://xircles.codehaus.org/manage_email

To unsubscribe from this list, please visit:

http://xircles.codehaus.org/manage_email
5

On the whole, my application is slower with JRuby 1.2. But when I
micro-benchmark each stage individually, the benchmarks show that JRuby
is faster. I think I need to look more closely at what the JVM is doing
when the application as a whole is running, this may in fact be a
resource issue.

// Gregory

Marcin Mielżyński wrote:

Vladimir S. wrote:

provide a simplified benchmark that shows the problem.

twice as

To unsubscribe from this list, please visit:

http://xircles.codehaus.org/manage_email

To unsubscribe from this list, please visit:

http://xircles.codehaus.org/manage_email