// Normalize the light and view vectors and transform it to the tangent space:
float3x3 mWorldToTangent = float3x3( vTangentWS, vBinormalWS, vNormalWS );
// Propagate the view and the light vectors (in tangent space):
Out.vViewTS = mul( mWorldToTangent, vViewWS );
Out.vLightTS = mul( vLightWS, mWorldToTangent );
vLightWS, vViewWS is the lgith vector and view vector in world space.
My question is that the last two line codes are doing the same operation but with inverse order to multiply matrix. I think the last is wrong, but the program runs well. Why?
I think the variable "mWorldToTangent" may be not right if it will be used to multiply a vector at the right side of mul function.
Because mul() function give different result according to the position of params.
We can see [color="#000000"]mul[color="#666600"]([color="#000000"] mWorldToTangent[color="#666600"],[color="#000000"] vViewWS ) != m( vViewWS, mWorldToTangent ).
[font=courier new,courier,monospace]Out.vLightTS = mul( vLightWS, mWorldToTangent );[/font]
is the same as
[font=courier new,courier,monospace]Out.vLightTS = mul( transpose(mWorldToTangent), vLightWS );[/font]
and in the case of pure 3x3 rotation matrices
[font=courier new,courier,monospace]transpose(rotation) == inverse(rotation)[/font]
So what that code is doing is multiplying by the inverse of that rotation matrix.
If [font=courier new,courier,monospace]vLightWS[/font] and [font=courier new,courier,monospace]vViewWS[/font] are both in world-space, then yes, the code looks fishy, as one is being transformed by a TBN rotation and one by the inverse...
So the last codes ---[color="#660066"]Out[color="#666600"].[color="#000000"]vLightTS [color="#666600"]=[color="#000000"] mul[color="#666600"]([color="#000000"] vLightWS[color="#666600"],[color="#000000"] mWorldToTangent [color="#666600"]), is wrong? Do you think so?
But the program runs well with the last code. Why? Who can tell me?