Hi, Robert Schuster has sent me some very interesting performance results with JamVM. I don't know if he wants them made public, but on the fosdemo, JamVM 1.2.5 performs almost as well as Sun's Hotspot or IBM's VM (within ~0.5%). Admittedly I think it's spending most of it's time calling native code via JNI, but it's interesting none the less.
If anybody else sees any dramatic speed improvements with 1.2.5 I'd be very grateful if you could provide feedback (equally, if you _don't_ see any improvements). It's not just an ego-trip -- feedback will help me tune the VM and suggest future optimisations. I only have a limited number of architectures and some defaults are pure guess-work. It's also nice to get positive feedback once in a while, instead of the usual bug reports :) Thanks, Rob. On Wed, 2 Mar 2005 10:30:24 +0000, Robert Lougher <[EMAIL PROTECTED]> wrote: > Hi, > > I'm pleased to announce the release of JamVM 1.2.5 > (http://jamvm.sourceforge.net). This release has an improved > interpreter, which shows typical speed-ups of between 60% and 100% > over JamVM 1.2.4. On some micro-benchmarks on a Athlon it's over 300% > faster! The full list of changes are here: > > http://sourceforge.net/project/shownotes.php?release_id=309491 > > Thanks, > > Rob. > _______________________________________________ Classpath mailing list [email protected] http://lists.gnu.org/mailman/listinfo/classpath
