Alexey Verkhovsky wrote:
On Jan 30, 2008 4:05 PM, Chris Nelson
<[EMAIL PROTECTED]> wrote:
Here are the slides for a presentation I gave at the Columbus Ruby Brigade
this month. Some of them are recycled from previous talks by Charles and
Tom. BTW, for the performance demo I gave I used the quicksort benchmark.
JRuby 1.1RC1 totally smokes, I highly recommend it for demos: MRI: 5 secs,
JRuby: 0.99 secs on my dual core laptop.
The problem with using microbenchmarks for performance demos is that
they are misleading. You can easily make some gullible person believe
that JRuby totally kicks MRI's ass performance-wise. Reality is not so
cool though: applications are typically a tad slower on JRuby than on
MRI.
I think it more has to do with the way a given app is designed.
PetStore, for example, has confounded most attempts to optimize. There's
something about it that causes JRuby trouble that we haven't identified
yet. But Nick's project at Sun has been benchmarked multiple times and
everyone doing the testing has seen much faster performance with JRuby
than with MRI, even comparing mongrel to mongrel.
So like Chris says, I think it's in flux. Overall, however...it should
be faster. PetStore not being faster right now is a bug as far as I'm
concerned.
- Charlie
---------------------------------------------------------------------
To unsubscribe from this list please visit:
http://xircles.codehaus.org/manage_email
