Page 1 of 2

From Blender Game Engine to Ogre,Girona,Crystalblend, Nebula

Posted: Wed Jul 06, 2005 12:41 pm
by erwin

I hope projects like Girona, Ogre, Crystalblend, Nebula Blender Gameplayers will take off. I offer some assistance in transition from Ketsji. I posted my plans on bf-committers:

"Ton asked me to outline plans of my new game engine contributions.

I haven't spoken to Kester yet, so plans might still change, but roughly I would like to contribute:

- Improving Physics Engine in Ketsji (current game engine)
I don't want to spoil all details yet. Physics abstraction has silently been completed in 3.37.

- Supporting CrystalBlend with Jorrit
Games won't be backwards compatible but it fits with Blender as good as possible

- Transition: Assist in interfacing CrystalBlend with Blender
I think it's best to focus on getting CrystalBlend to the point that it is a reasonable blenderplayer. After that people can start thinking about more tight integration.

I hope I can make one more contribution to Ketsji, before it will be replaced. Hmmm, this sounds a bit too dramatic ;-) "


Posted: Wed Jul 06, 2005 2:20 pm
by malCanDo
Hi Erwin,

I'm looking forward to testing those new features!!! ;)


alpha test

Posted: Fri Jul 08, 2005 6:26 pm
by erwin
Hi Malachy,

Today I stayed home (in London), so I had some quality blender codingtime.
Hence, I have a special blender version for you today. It has box, cylinder and sphere support and runs your first scenes very nice. I will add triangle meshes and convex support in a few days, just a matter of hooking up things.

UPDATE: There is a windows testbuild for bullet physics in Blender 2.37a (gameblender, updated 19 july 2005) ... der/ErwinC

Please visit bullet game blender forum at ... /index.php


Re: alpha test

Posted: Fri Jul 08, 2005 7:03 pm
by kidb
erwin wrote: These links are only for serious alpha testers, else my bandwidth goes out of control :)
If you don't mind, I can put it on, for the less serious. I even could provide you with a login. mailto kidbatkidbdotde.


Posted: Mon Jul 11, 2005 4:18 pm
by erwin
Hi Kidb,

that comes in handy, thanks ! I will make a bit more stable and full featured release soon, which is more suitable for testing.

Could you get in touch with me,
blender at ?


Posted: Wed Jul 13, 2005 2:41 pm
by malCanDo

I tried out the Blender build, it's great to see the sample scenes working, with stacked boxes now remaining pretty solid!

Keep up the great work!

Posted: Fri Jul 15, 2005 3:45 pm
by pildanovak
the link to the build is not working...

new location of testbuild

Posted: Fri Jul 15, 2005 5:13 pm
by erwin
There is a windows testbuild for bullet physics in Blender 2.37a (gameblender, updated 15 july 2005) ... der/ErwinC

Please visit bullet game blender forum at ... /index.php

Posted: Fri Jul 15, 2005 7:27 pm
Incredible ! .... But true !

My English is horrible, and I haven't post since 2 years on this forum. But it's so great to see that someone is working on the GE again ! I am going to open a bottle of champain !

Great great work Erwin. I hope your bullet engine will be a full part of a future Blender release.

I applauded well extremely! And I will follow your work step by step :wink:


Posted: Fri Jul 15, 2005 7:46 pm
by Metallica
hey erwin, i've been paying attention to your code work :D

i was testing and playing around with the build... its very very stable. only thing on the concave demo, the cone mesh goes through the concave mesh that it falls into just a little bit, but does not fall through like old physics ;) also in that demo the cone seems to just stop moving at a certain point. (maybe because the concave mesh doesn't have polyheader bounds on?)

in the convex demo the top object seems to have enough friction to not slide down. is that how its supposed to be? i would think its too topheavy

the other demos seem to work very well. i like that its not jittery at all and nothing falls through the meshes like the old phsyics/collision. =)

one more thing, i tryed a radiosity floor (very stupid thing to do) and had a cube fall on it, seems that it crashes blender. :/ but its not good to do radiosity meshes anyway.

Posted: Fri Jul 15, 2005 7:47 pm
by Saluk
Poor ketsji, it will be sad to see you go :( But it's nice to have the original guy back! Thanks for all you're hard work in the past and begining again.

Posted: Fri Jul 15, 2005 11:38 pm
by ElBarto
Kester did a lot and he did a lot well. Besides the lot of game engines there also a lot of realtime 3d solution that are very interesting because most of them:

- offer a very easy interface for programming and interaction like logic bricks of Blender
- have very nice fetures like multitexuring and shadows
- with many of them you can built games and interactive applications too


Virtools ... atures.asp


Director ... /features/


VisX3D (based on the X3D technology)

Posted: Mon Jul 18, 2005 10:14 am
by pildanovak
crashes blender here - Win 2000 proffesional, intel proc, don't know what card. I sucesfully crashed these 2.25 demos:
ballercoaster demo
conveyor head demo

edit> stupid me, i tested gabios build and without the demos provided with your build...

This physics is great!!! all the domos run nicely. The framerate stays almost the same also with huge heaps of stacked or moving objects, also tryed the p2p constrains, makes good chains and again - I had to make cca 30 chains with 4 or 5 constraints each to drop the framerate from cca 22 to the half of it(never got under 10)


Posted: Tue Jul 19, 2005 7:02 am
by pildanovak
I've tested at home yesterday(Win 2000 pro, athlon 1.8,512mb, radeon 9000):
100 frames per sec with 127 cubes falling in a stack, gets down to 60 with them on the floor.

more than 127 objects crash blender
polyheder collision works a little strange and slows down a lot. I tested on a falling monkey(has many faces, but i also tried to make an envelope of a few faces). --- First the simulation runs smooth, monkey falls and lies down, but in the moment she shoud rest already she starts jumping, and in that moment gets framerate dramatically low.

testbuild 2005july19: fix for random collision problem ?

Posted: Tue Jul 19, 2005 7:25 am
by erwin
"First the simulation runs smooth, monkey falls and lies down, but in the moment she shoud rest already she starts jumping, and in that moment gets framerate dramatically low."
Isn't the monkey a he, instead of a she ? ;-)

There is a hardcoded limit of 128 in the number of dynamic objects, I will remove that. Also, triangle meshes are not optimized, some kind of compressed kdop-tree needs to be added.

Please test if this new build fixes the occasional 'random' collision and rotational problems: ...

Thanks for the feedback,