I believe this symptom was analyzed in an earlier thread (was it for erikd's emulator?).
Ssshht! don't mention that to everyone
There's some things that I learned during that excercise:
- Don't assume anything or make premature conclusions
- Don't write particulary for one VM by practicing bad programming habits (like manual inlining).
*Maybe* the 3d demo's suffer from the same problem, but since we haven't seen the source, we don't know so we can't tell.
The end of my excercise had the emulator running fine on the Sun 1.4 VM with better code, but still generally slower than on the MS VM (ranging from significantly slower till a little bit faster, depending on the emulation driver being used).
For these kind of things, MS VM seems generally faster than Sun still. I can't explain why, but most applets simply prove that. This does not mean the Sun VM is generally slower than the MS VM, but for these kind of applet it seems true.
Read the discussion here, but don't tell anyone I started that topic: http://www.java-gaming.org/discus/messages/27/1531.html?1029399057http://www.java-gaming.org/discus/messages/27/1546.html?1029399057
As a side note, I'm still waiting for a 1.4 VM for windows coming from IBM ... (for games it seems faster than Sun VM and faster than MS VM. My emulator runs best on IBM 1.3 JRE).
I tested some applets from both links on Sun 1.4.1_01 and MS VM and I could hardly see the difference to be honest, although the MS VM seems a *little bit* more smooth. I also did some profiling on the Sun JVM which shows that it doesn't suffer from the problem I had in the past (The Sun JVM not compiling because of manual inlining).