Hi All, my constant buffer structure for per pass looks like this

struct PerPass{

XMFLOAT4X4 projectionMat;

XMFLOAT4X4 viewMat;

XMFLOAT4X4 textureMat;

Lights maxLights[MAX_LIGHTS]

float3 cameraPosition;

};

I am aware of packing rules so first three variables should be fine since they are all multiple of 4D vectors. My “Light” structure is also a multiple of 4D vector. cameraPosition is last variable so its will be fine too. 

I have tried moving textureMat up and down but it always give me last two rows as “not used” when I do graphics debug in Visual Studio. 

I am initialising it with an identity matrix but it doesn’t come up completely. Setting textureTransform(0,0) = 7; and textureTransform(1,1) = 9 scale the texture along x and y direction. These two changes reflects in first two rows correctly. I wanted to move the coordinates so I am supposed to utilise last row for translation. textureTransform(3,0) and textureTransform(3,1). This row is mentioned “not used”. Maybe I will move textureTransform to different constant buffer but what is going on here? What is my mistake?

A look at the shader would be clarifying. The shader compiler optimizes unused values away, no matter the optimization settings. Chances are, you made a mistake with the texture transformation calculation. Does the compiler issue a warning ?

Hi unbird, Here is my full hlsl content.

 

#include "LightingFunctions.hlsl.txt"
struct VertexIn
{
  float3 Position : POSITION;
  float3 Normal   : NORMAL;
  float4 Color    : COLOR;
  float2 TexCord  : TEXCORD;
  float2 FunData  : FUNDATA;
};

struct VertexOut
{
  float4 OutVertex: SV_POSITION;
  float4 Color : COLOR;
  float3 Position : POSITION;
  float3 Normal : NORMAL;
  float2 TexCord : TEXCORD;
};
cbuffer CbObject : register(b0)
{
  float4x4 gWorld;
};

//Texturing
Texture2D gDiffuseMap:register(t0);

SamplerState gsamLinear  : register(s0);

cbuffer CbPass : register(b1)
{
  float4x4 gView;
  float4x4 gProj;
  float4x4 gTexTransform;
  Light gLights[MaxLights];
  float3 gCameraPosition;

};
cbuffer CbMat : register(b2)
{
  float4 diffuseAlbedo;
  float3 fresnelRo;
  float roughness;
};
VertexOut VS(VertexIn In)
{
  VertexOut out1;
  float4 intermediate = mul(float4(In.Position,1.0f), gWorld);
  out1.OutVertex = mul(intermediate,mul(gView,gProj));
  out1.Position = intermediate.rgb;
  //float4 intermediate = mul(gWorld,float4(In.Position,1.0f));
  //out1.OutVertex = mul(mul(gView,gProj),intermediate);
  out1.Normal = In.Normal;
  out1.Color = In.Color;
  out1.TexCord = In.TexCord;
  return out1;
}
float4 PS(VertexOut in1):SV_Target
{
  in1.Normal = normalize(in1.Normal);
  float4 texT = mul(float4(in1.TexCord,0.0f,0.0f),gTexTransform);
  float3 toEye = normalize(gCameraPosition - in1.Position);
  float4 texDiffuseAlbedo = gDiffuseMap.Sample(gsamLinear,texT.rg);
  float4 tDiffuse = diffuseAlbedo*texDiffuseAlbedo;

  float4 ambient = float4(1.0f,1.0f,1.0f,1.0f)*tDiffuse;
  const float shininess = 1.0f - roughness;
  Material mat = { tDiffuse, fresnelRo, shininess };
  float3 shadowFactor = 1.0f;
  float4 directLight = ComputeLighting(gLights, mat, in1.Position,
      in1.Normal, toEye, shadowFactor);
   float4 litColor = ambient +  directLight;
   // Common convention to take alpha from diffuse material.
  //float4 litColor = tDiffuse;
  litColor.a = diffuseAlbedo.a;
  return litColor;
  //return in1.Color;
}

The bold line in Pixel shader doesn't work as expeced. See the attachment. 

The last two elements of your float4 are both zero. The compiler knows that 0 * anything is 0 and therefore it doesn't need to fetch the last two rows of the matrix or do the work involved in multiplying 0 by gTexTransform[2] and [3].

1 hour ago, ajmiles said:

The last two elements of your float4 are both zero. The compiler knows that 0 * anything is 0 and therefore it doesn't need to fetch the last two rows of the matrix or do the work involved in multiplying 0 by gTexTransform[2] and [3].

Thanks ajmiles! That was he issue.