This is the mail archive of the
kawa@sourceware.org
mailing list for the Kawa project.
Re: Performance
- From: Per Bothner <per at bothner dot com>
- To: Tony <tonys-mailbox at hotmail dot com>
- Cc: kawa at sourceware dot org
- Date: Sat, 21 Jun 2008 11:43:51 -0700
- Subject: Re: Performance
- References: <BAY115-W22D4D18E1168CCA4F506AFE1AA0@phx.gbl>
Tony wrote:
Can anyone indicate the relative performance of Kawa as opposed to Java?
In particular, if static types are used in the Kawa code, how does performance compare with Java?
It will of course very a lot from application to application.
I'm slowly trying to port some of the benchmarks from
http://shootout.alioth.debian.org
So far I've ported the binarytrees and nsieve benchmarks.
Some very preliminary results:
binarytrees.java takes about 1.6s, while binarytrees.scm takes
about 1.2s. These are unofficial numbers, of course.
nsieve seems to be dominated by JVM startup time, which is
worse for Kawa. Running the test 10 times makes both take about 5s,
but when running the test just once Kawa takes about twice as long.
Java 5 and Java 6 give different results, but I don't have a
picture of that yet. (I'd like to implement "StackMap" support
for Kawa, which is likely to improve at least starup-time.
However, I don't know when that will be.)
--
--Per Bothner
per@bothner.com http://per.bothner.com/