Simple Physics Engine Platform question

Posted By: indiGLOW

Simple Physics Engine Platform question - 05/18/04 03:52

Can someone give me the minimum specifications for running a game that uses the GS-PE. I have a very simple destruction derby car game that uses the PE to bounce the cars off one and another, and the players car. This runs fine on my dev platform, P4 3.06, 512 DDR etc. However on the target platform, a Celeron PC (Not sure of Processor, trying to find this out) with 256 RAM, when a car is added that uses the PE the game hangs...

Could someone give me a minimum spec for the target platform PC to run GS-PE.

Note: I am trying to get the info on the target platform PC and will post it here asap. I will also test this on another PC to see if there is some other issue...
Posted By: Aaron_H

Re: Simple Physics Engine Platform question - 05/18/04 05:06

Wouldn't it depend on the ammount of physics active at any given moment...?
If so, couldn't that change the specs quite a bit.

Please post details about the ammount of physics objects that you are using and maybe somebody could help (although I do not know much about physics myself, So I cannot help ).
Posted By: ventilator

Re: Simple Physics Engine Platform question - 05/18/04 05:10

sometimes i read the ODE mailing list out of curiosity. i read that their current solver is a memory hog with lots of constrained entities. maybe 256mb isn't enough? just a thought...
Posted By: indiGLOW

Re: Simple Physics Engine Platform question - 05/18/04 05:57

Yes maybe your right Ventilator... considering we are talking about extreamly simply PE use, i.e 10 cars using simple box collision PE, gravity and friction...

I will research a little more and post here when I know more...
Posted By: indiGLOW

Re: Simple Physics Engine Platform question - 05/18/04 20:06

I have removed all references to the PE and now the game runs on the target platform so I presume this is as you say the fact that the PE is system resource hungry...

I am not getting a warning or error message, just the game either drops out to windows or hangs and requires a reboot of the machine....

Could someone from conitec confirm/deny the 256Mb PE spec suggested? Or am I missing something? Thank you.
Posted By: ello

Re: Simple Physics Engine Platform question - 05/18/04 20:12

i didnt have problems with the PE and 128MB Ram
Posted By: ventilator

Re: Simple Physics Engine Platform question - 05/18/04 20:16

with 128mb you will have problems with windows alone! ...i think the problem just would occur with lots of contrained entities (like 10 cars? )...

ode will get a new solver soon. -> http://q12.org/pipermail/ode/2004-May/012806.html i don't know when/if this solver will be reasonably stable and when marco will implement it into a6.
Posted By: indiGLOW

Re: Simple Physics Engine Platform question - 05/18/04 20:57

The target machines im using have 256MB RAM and the crash happens when I add the first car! (I am adding the car in-game with a button press)

Although I am trying to get at least 9 of these very simple cars (Just Box Physics)
Posted By: indiGLOW

Re: Simple Physics Engine Platform question - 05/19/04 17:22

Out of curiosity is there a way to simply initialise the PE? i.e turn it on without doing any actualy calculations to see if the solver is too large for the target platform memory/spec.

This project is due for release at the end of this month, with a final beta due the end of this week so a quick response would be most beneficial. For now I will have to work without the PE as that seems to be the only way to run the game.

Thanks in advance guyz what would I do with out you?
Posted By: indiGLOW

Re: Simple Physics Engine Platform question - 05/21/04 02:00

Can someone point me at a minimum spec for just starting the PE, i.e adding one cube/box that is bound by gravity etc. As I suspect that there is another reason for this system crash... This is the first time i have used the PE so it could easily be user error!

I will email conitec directly anyway.
Posted By: indiGLOW

Re: Simple Physics Engine Platform question - 05/22/04 22:24

Ok I have done some further investigation into the problems I am having. I have set some -d handles, that enable/disable the physics within my game. Also running the game on the target platform in a window, I find that it is not just locking up the computer, but is actually generating an invalid instruction.

I don't know if this means anything to anyone, but here is that error anyway.

Code:

3DGS_6_23 executed an invalid instruction in
module 3DGS_6_23.EXE at 0167:004eacee.
Registers:
EAX=44e1d271 CS=0167 EIP=004eacee EFLGS=00010203
EBX=028a1cf8 SS=016f ESP=01bef494 EBP=028a1c98
ECX=028a1be0 DS=016f ESI=028a1cec FS=2fd7
EDX=44c6d275 ES=016f EDI=028a1d04 GS=0000
Bytes at CS:EIP:
db c1 db f2 db c2 d9 5c 24 14 de d9 d9 44 24 14 
Stack dump:
035aa594 028a1be0 00000008 08471af0 006b4e20 03bb9d6c 44e1d271 44c6d275 44c619e3 0000000c 0000b38c 0044c1cd 00000000 3d0a0c4a 01bef634 03b69c44


My real question is this due to poor code on my behalf, a bug in the physics engine, the solver 256MB that Ventilator mentions or something else?

The Target Platform is a Celeron 900Hmz 256MB Micro-PC (Just means small).
Posted By: indiGLOW

Re: Simple Physics Engine Platform question - 06/02/04 18:22

Ok as no one seems to be able to help me get the PE running on these Celeron900Mhz 256MB Pc's, could someone help me with a work around.

If one car crashes into another, i need to move the car away as IF it had been hit. Only in x,y and not as realistic as the PE. Any ideas ??? Obviously without using the PE now!
Posted By: Phantom88

Re: Simple Physics Engine Platform question - 06/02/04 20:50

Maybe you should post this in bug hunt... Doesn't seem like a bug in your code. ODE(PE) should work with that system, theres something else, it's not the memory, that's for sure

EDIT: Hey, your dev. pc is allmost exactly the same as my new rig i have ordered... (Except i have a amd) (Off-topic, but worth of mentioning)

~Phantom88~
Posted By: indiGLOW

Re: Simple Physics Engine Platform question - 06/03/04 19:41

Interesting, I have contacted Conitec directly about this but as of yet I have had no reply, despite have 12months tech support...duh!

I suspect that the PE should be running on the target machine also, but without tech-sup I am unable to identify the problem. I have tested it on another similiar spec machine and it also crashed out on that one, so I presume it is a under-powered issue. As far as I can tell the PE is not doing anything really complex, which is why im now going to attempt to hard-code the knock-back into the events system.

I am disapointed that the PE could not be used.... I have built in a runtime switch for the PE to future proof it, incase I can resolve the probem....

As for the dev machine. I have found it very capable.... good shopping! although I only ever use intel!
Posted By: Doug

Re: Simple Physics Engine Platform question - 06/04/04 03:51

Quote:

Interesting, I have contacted Conitec directly about this but as of yet I have had no reply, despite have 12months tech support...duh!



You should have an answer in 48 hours (normally 24). If not, double check the address and send again.
Posted By: Marco_Grubert

Re: Simple Physics Engine Platform question - 06/04/04 07:51

A moderately large number of constraints is not a reason for crashes, it would just slow things down. Most likely you are accessing an entity that has not been properly initialized or got removed in the meantime.

Please resend your mail to me: mgrubert@conitec.net
Posted By: indiGLOW

Re: Simple Physics Engine Platform question - 06/04/04 20:48

If that was true, why does it run on my development machine, the other Intel P4 in the office, another AMD P4 equiv in our other office, and not on the 2 target machines that have Celeron2 900 chips?

Please understand I would prefer it to be a error on my side, this is afterall the first time I have seriously set out to use the PE. If you could give me a simple code example that starts the PE and sets an entity to have physics I could test this on the target platform, and if that will define where the problem is. I truly hope the problem is within my code, but I fail to see where it is, considering that it runs fine on all other platforms.

FYI: My first email to conitec was on the 23rd May and then again on the 27th sent to support@conitec.net.
Posted By: Marco_Grubert

Re: Simple Physics Engine Platform question - 06/05/04 06:19

Try using the email address I posted above, I did not get any forwards from support@conitec.net.
Posted By: indiGLOW

Re: Simple Physics Engine Platform question - 06/05/04 20:13

I will do thanks Marc. However it pretty much just points to this thread.

Can you give me a simple PE code example. One that will turn on the PE and apply physics to an object. I will test this on the target platform and then I will know if it will work.

My Code is pretty simple, i use a run-time handle to test if PE has been set on, if it has it runs the following commands. I have just posted my PE commands in case there is something obvious that would prevent them running on Celeron 2 900

Code:
  
// Later in the code............. Entities Action Code
	phent_settype( My, PH_RIGID, 0 ); 
	phent_settype( My, 0, 0 );
	phent_settype( My, PH_RIGID, PH_BOX );	
	phent_setmass( My, 20000, PH_BOX);
	phent_setfriction( My, 45);
	my.enable_friction = on;
  
// Before Move Code............. 
	TruePhysics_On){phent_settype( my, 0, 0 );
 
// After Move Code............. 
	phent_settype( my, PH_RIGID, PH_BOX );


This all works fine, the friction/weight/gravity needs tweaking, but it runs. When you add cars and they hit each other they bounce away, roll and crash. However as soon as you add a car on the target platform the game crashes out to windows without a message.

I have managed to get an error in the Kernal32.dll if I run the game in a window and it crashes.

If this is not enough I can send a copy of my WDL.
© 2024 lite-C Forums