while surfing I found this fantastic little physic game
very cool
This is so much fun.
note that this is also a game, once you have set the player and some fighter you move the player with keyboard arrows, this is really an amazing and fun applet
Really cool, I wish it was open source. How did he squeeze so much optimization out of Java2D, I wonder?
Small window size & direct pixel rendering, bypassing the accelerated pipeline entirely.
It’s all about memory bandwidth.
Fun physics demo!
I don’t get it. What am i supposed to do?
I’d classify it as a toy rather than a game.
Very interessting never-the-less. Thanks for sharing 
How is “bypassing the accelerated pipeline” going to improve performance?
I was under the impression that direct pixel rendering was incredible slow…
I may have misunderstud something though…
- Scarzzurs
Sorry, I didn’t mean to imply doing this improves performance.
Going back to DemonPants original question:
The Applet achieves what it does in spite of Java2D, not because of it.
The perceived performance is high because the speed limitations have been well understood & factored into the overall design.
You’d be surprised what you can achieve when your rendering is just manipulating values in a ~120000 element array. (400x300)
How do you do that? All I know of is drawing to the raster in a BufferedImage, which obviously isn’t what you’re talking about.
Essencially yes, the fastest way to do pixel rendering in 1.2+ is to write to the array exposed through the Rasters underlying DataBuffer.
Obviously by doing this you have to tailor your per-pixel rasterizing code to a particular pixel format, so be sure you pick one that won’t require conversion when the Image’s raster is blitted to the screen.
I wrote a benchmark that draws 100 000 fractal generated pixels (varying positions) at 60 fps (400 000 with q6600 using all cores).
There is a 800x600 image drawn pixel per pixel per frame.
You can easily loop through an 800x600 int[] array several times at 60 fps and alter all pixels.