GIMPACT v 0.2 Released

[projectileman]

Not really.
GIMPACT 0.1 doesn't have margins.

About my machine is a Pentium4, 2.4ghz, 32bits, 1 Core.

And I've tested your GimpactDemo with GIMPACT 0.1, and you're right. It runs faster. The new GIMPACT 0.2 takes 2 times longer in my machine, and I don't know why ..
GIMPACT 0.2 uses the same algorithms than 0.1, but GIMPACT 0.2 is more dependant of Bullet.

I'm start to thinking that it could be the contact generation process. GIMPACT 0.1 uses its own contact filter process for giving the minimal amount of points. GIMPACT 0.2 uses the btManifoldResult class for filtering points.

However, I'm not sure if that is causing the slowdown. Let me doing some tests ok?

[Erwin Coumans]

@Erwin: Can you please add this Demo to the next Bullet release!?


DevO

Hi Dev0,

The demo has been added in Bullet 2.62. This will hopefully help Francisco improve GIMPACT.

Thanks,
Erwin

[projectileman]

Hi again.

I was profiling DevO's demo with both GIMPACTs, and I've discovered that GIMPACT 0.2 fails in bounding tree construction. So this version test more nodes due of its unbalanced trees;it calculates around 6xtimes compared with 0.1.

My apologies.

I'm solving this problem as soon as I can, for the upcomming GIMPACT version 0.3.

Stay tuned.