but, mul (normals, matWorld) doesnt work like i expected it to. only at world coordinates (0,0,0) it looks right. i remember there was a thing about twisted xyz coordinate, but i dont exactly know it anymore.

or what else is wrong ?

hmm..
mul(inNormal,matWorld) ought to work in the general case

Maybe you arent inputing your normals correctly? Or your light vector is not in world space or being derived correctly?

Show your shader code and i can look at it.. maybe can figure it out

i am not even doing some lighting here. i just want the normals in worldspace to be displaced so they dont rotate with the objects.

Code:

---

float4x4 matWorldViewProj;
float4x4 matWorld;

struct vsIn
{
	float4 position : POSITION;
	float4 normal : NORMAL;
};

struct vsOut
{
	float4 position : POSITION;
	float4 normal : TEXCOORD0;
};

vsOut mainVS (vsIn In)
{
	vsOut Out;
	Out.position = mul(In.position,matWorldViewProj);
	Out.normal = mul(In.normal,matWorld);
	return Out;
}

float4 mainPS (vsOut In) : COLOR
{
	return In.normal;
}

technique normals
{
	pass p0
	{
		CullMode = CCW;
		VertexShader = compile vs_2_0 mainVS();
		PixelShader = compile ps_2_0 mainPS();
	}
}	

---

here is an image with untransposed normals, you see the rotated figure which of course has unfitting normals for the overall scene:


and here with transposed normals, i really dont get it:

try using saturate? or multiplying the color by some bias factor like 0.1 or something.. it looks to me like your normal colors are just too overbright (not between 0-1).

Or try renormalizing your normals after you multiply with matWorld (shouldnt need to)

hm, that doesnt give the wanted results, too

normalized:


and saturated:

I dont see why this isnt working.. what shader verison is this compiled to? You wouldn't happen to be using 3.0 are you?

I have had problems with 3.0 with things like this...it changes some things i dont quite understand.

Try doing something else like doing lighting in the world space, see if that gives you correct results.

i am using sm2_0, i went down to sm1_1 i but the results are the same. i dont want to have any lighting. just the normals in correct worldspace. plus i remember that it once worked. i dunno.

it must be something dumb obvious thing i believe

I meant..test the shader by calculating lighting, so you know whether or not the normals are actually being transformed.

mul-ing the normals by the world matrix is going to not only transform the orientation, but also the position, which you don't want- this is causing the overreaction. Maybe it's just me... but I think you want the untransformed normals just raw, maybe normalized...

EDIT: ok.. I see what you mean, reading through again..
You should probably set up your own matrix that doesn't have the translation multiplied in. Or you could mulitply the world transform matrix by an inverse translation for the objects world position...

-Rhuarc

When you want to transform the normal to world-space, the correct way would be to use the inverse transpose of the world-matrix. That's it in HLSL:

Code:

---

float4 worldNormal = mul(matWorldInv, inNormal);

---