NxOgre and Bump cg material effect

Hi all

I'm french and new to ogre, so I'm sorry for all dumb things that I could say !

So,

I have a problem with the cg script "Example_BumpMapping.cg" when I use it with NxOgre.

When i let this code like this :
NxWorld = new NxOgre::World("FrameListener: Yes, log: none");
I fall down to 6 FPS when the mesh with the material using "Example_BumpMapping.cg" is rendered. ( I try various World's options, nothing help. )

But when I comment it :
//NxWorld = new NxOgre::World("FrameListener: Yes, log: none");
Everything goes back to normal. ( 500 fps when rendering the mesh )

This is the only line of NxOgre code in my program.

( I try to find answers on Google and on this forum, but nothing. )

If it is a NxOgre issue, how can I fix this ? Or it is a know bug from PHYSX ?

PhysX/NxOgre have nothing to do with CG/Shaders, and I haven't seen an issue or a problem like yours before.

Hmm strange. And there is no fps slowdown if you create the NxOgre world but do not use the shader?
Do you run PhysX in software mode? Or do you have a newer NVidia graphics card?

@betajaen: Yes that's also my idea beacause cuda is, like the shader, computed on the gpu. But I also never heard about a problem like it before.

Or do you have a newer NVidia graphics card?

Perhaps it could be CUDA. But again; I've never heard of a problem like it before.

When I put another texture on my mesh with NxOgre ( without bump mapping, just flat texture ), no fps slowdown.

I run PHYSX in software mode, i have a geforce 8800 GTS 320 Mb.

It's definitly a "PhysX/NxOgre problem" because if I remove NxOgre header file or if I change the texture material, all work fine again.

I hope make a mistake because if it isn't, I will have no solutions :/

Most likely it's a PhysX problem. Probably something to do with CUDA. NxOgre doesn't touch shaders and it barely touches Ogre.

Have you you tried upgrading the Nvidia drivers and PhysX software?

I just make the upgrade. My drivers was the 178.24 and I upgrade to the 178.28.

Same problem. But i find what was the problem! My normal map was too huge ! I resize it and everything is OK.

I think the PHYSX use too much CPU time for this big normal map when enabled.

I have no CUDA processor, i hope is can help someone else

Thanks BetaJaen for help and NxOgre, and thx Caphalor too

As far as I know the Geforce 8 series is able to run Cuda/GPU PhysX.
Good to hear that you solved your problem.

I think the PHYSX use too much CPU time for this big normal map when enabled. PhysX doesn't use a normal map. Neither does the CPU the computing of a normal map.

As far as I know the Geforce 8 series is able to run Cuda/GPU PhysX. Wink
Good to hear that you solved your problem.

I must believe that is wrong, i download the latest driver, et the CUDE toolkit + SDK, and when I run an example of CUDA 1.0 ( and + ), the app show this error message :

VideoCodec: MPEG 2
Frame rate: 30000/1001fps ~ 29.97fps
Sequence format: Progressive
Coded frame size: [1280, 720]
Display area: [0, 0, 1280, 720]
Chroma format: 4:2:0
Bitrate: 14116kBit/s
Aspect ratio: 16:9

cudaVideoDecode:
<GeForce 8800 GTS> CUDA 1.0 device 0 detected
This device not CUDA 1.1 capable, exiting

Test PASSED

Press ENTER to exit...

Maybe I don't understand the message, I am not really good in english But i think i have "0 device detected" for CUDA.
I have the first generation of GeForce8's series, maybe that's the only one reason.

--------

Right for the fact that bump mapping is not a CPU task, it's CG programming :/ I'am dumb

So I don't understand why there is a conflict with CG scripts and PHYSX with a huge sized normal map.

-------

2nd Question :

Why my callback don't work ?

using namespace NxOgre;

class MyCallBack : public NxOgre::GroupCallback::InheritedCallback
{

public:

void onStartTouch( NxOgre::Actor* Actor1 , NxOgre::Actor* Actor2 /*, NxOgre::ContactStream* Stream*/ )
{
LogManager::getSingleton().logMessage( "-= Entered onStartTouch Callback! =-" );
}

void onEndTouch( NxOgre::Actor* Actor1 , NxOgre::Actor* Actor2 /*, NxOgre::ContactStream* Stream*/)
{
LogManager::getSingleton().logMessage( "-= Entered onEndTouch Callback! =-" );
}

void onTouch( NxOgre::Actor* Actor1 , NxOgre::Actor* Actor2 /*,NxOgre::ContactStream* Stream*/ )
{
LogManager::getSingleton().logMessage( "-= Entered onTouch Callback! =-" );
}

};
.....

Class foo{
public:
NxOgre::ActorGroup * MainPlayer;
NxOgre::ActorGroup * MapObjects;

void Init( )
{

MainPlayer = NxScene->createActorGroup( "MainPlayer" );
MapObjects = NxScene->createActorGroup( "MapObjects" );
}
void foofunc( )
{
MyCallBack * Joueur_VS_Map = new MyCallBack();
MainPlayer->setCallback(Joueur_VS_Map);
MainPlayer->setCollisionCallback( NxScene->getActorGroup("MapObjects"), NX_NOTIFY_ALL, true);
MainPlayer->setCallback<MyCallBack>(Joueur_VS_Map, &MyCallBack::onStartTouch, &MyCallBack::onEndTouch, &MyCallBack::onTouch );

}
}

I have this error msg :

nxogre\include\NxOgreGroup.h(173) : error C2661: 'NxOgre::GroupCallback::GroupCallback' : overloaded function does not requires 5 arguments

Humm, guess what ? Empty log. ( Yes, i collide 2 mesh of opposite groups )

Solved, I don't know why but I must make my declaration function like this :


virtual void onStartTouch( NxOgre::Actor* Actor1 , NxOgre::Actor* Actor2 /*, NxOgre::ContactStream* Stream*/ )
{
LogManager::getSingleton().logMessage( "-= Entered onStartTouch Callback! =-" );
}

virtual void onEndTouch( NxOgre::Actor* Actor1 , NxOgre::Actor* Actor2 /*, NxOgre::ContactStream* Stream*/)
{
LogManager::getSingleton().logMessage( "-= Entered onEndTouch Callback! =-" );
}

virtual void onTouch( NxOgre::Actor* Actor1 , NxOgre::Actor* Actor2 /*,NxOgre::ContactStream* Stream*/ )
{
LogManager::getSingleton().logMessage( "-= Entered onTouch Callback! =-" );
}

I see that with the "NxOgreGroup.h" header file. I hope that will help newbies like me.

I had a major slowdown issue from like 200fps to 11fps. Turned out it was texture thrashing...have you compressed your textures as dds? DXT1 compression? (DXT3 or DXT5 if you're using alpha)

Textures are not yet compressed, I think that's was the resolution in DPI, since i'm in 72 DPI with a reasonnable size, everything is ok.

A little detail to care of.

even at 72 DPI textures can eat up memory quickly.

even a 38k png file can end up taking up 4 meg of video memory depending on how big the image actually is.

Talking about numbers

My buggy huge normal map : 4000 x 2000 - 92 DPI - JPG - 8000Ko @ 6 FPS

My actual normal map : 2500 x 1250 - 72 DPI - JPG - 955Ko @ 500 FPS

Talking about numbers

My buggy huge normal map : 4000 x 2000 - 92 DPI - JPG - 8000Ko @ 6 FPS

My actual normal map : 2500 x 1250 - 72 DPI - JPG - 955Ko @ 500 FPS

2500 * 1250 * 24 bits per pixel = over 9 meg video memory

video memory != disk memory

jpg is compressed on disk but is completely uncompressed in video memory

use the directX texture editor and it'll probably take that down to about 2-4 meg and is loaded into video memory as-is.

You're probably texture-thrashing and physx is using your GPU a bit so the swapping takes forever to do.

Compress your textures to dds with DXT1 compression (DXT 3 or 5 if you're using alpha) and I bet that'll fix your FPS issue. I fell into the same problem myself a couple months ago. Now I'm using NxOgre and I've got twice as many textures and I'm still running at a great FPS rate.

Thanks for the tips

I'll try it soon, for now I play with PHYSX :p