What does this shader compiler error mean?

[PeterV 116]

Hello, I'm trying to compile a vertex shader but I get this error about problems when writing to output registers, but I can't figure out what the problem is nor can I find any documentation that explains the meaning of this error code. The standalone compiler (fxc) gives this error: error X5467: Dest register type for LRP must be temp (r#). And the line that produces this error is: Out.Depth.x = Depth*FarZMult; Here is the entire shader source: float4x4 ViewProjMatrix : register(c0); float4 EyePos : register(c4); float4 EyeDir : register(c5); float4 NearPlane : register(c6); float FarZMult : register(c7); float4 ChunkPos : register(c16); float4 ChunkRange : register(c17); float4 PosLow[216] : register(c18); // ----------------------------------------------------------------------------------------- // vertex shader input channels. // ----------------------------------------------------------------------------------------- struct VSInput { float4 Pos1 : COLOR0; float4 NormalAndDitherFactor1 : COLOR1; float4 LightComponentsAndMaterial1 : COLOR2; ///////////////////////////////////////////////////////////////////////////////////////// float4 Pos2 : COLOR0; float4 NormalAndDitherFactor2 : COLOR1; float4 LightComponentsAndMaterial2 : COLOR2; }; // ----------------------------------------------------------------------------------------- // vertex shader output channels. // ----------------------------------------------------------------------------------------- struct VSOutput { float4 Pos : POSITION; float Depth : TEXCOORD0; float4 NormalAndDitherFactor : COLOR0; float4 LightComponentsAndMaterial : COLOR1; }; // ----------------------------------------------------------------------------------------- // vertex shader function (input channels). // ----------------------------------------------------------------------------------------- VSOutput main(VSInput In) { VSOutput Out = (VSOutput)0; ///////////////////////////////////////////////////////////////////////////////////////// float3 Pos1, Pos2; float4 Pos; float3 Normal1, Normal2; /////////////////////// /////// Unpack Position 1 ////////////////////////////////////// int PosLowIndex = D3DCOLORtoUBYTE4(In.Pos1.w); Pos1 = ChunkPos + ChunkPos.w*(In.Pos1 + PosLow[PosLowIndex]); /////////////////////// /////// Unpack Position 2 ////////////////////////////////////// PosLowIndex = D3DCOLORtoUBYTE4(In.Pos2.w); Pos2 = ChunkPos + ChunkPos.w*(In.Pos2 + PosLow[PosLowIndex]); ///////////////////////////////////////////////////////////////////////////////////////// float Dist = distance(EyePos.xyz, Pos1.xyz); float Ratio = (Dist - ChunkRange.z)/ChunkRange.w; // change to multiply float Ratio2 = saturate(Ratio); float Ratio1 = 1.0f - Ratio2; ////////////////////////////// transform Position /////////////////////////////////////// Pos.xyz = Pos1*Ratio1 + Pos2*Ratio2; Pos.w = 1.0f; Out.Pos = mul(ViewProjMatrix, Pos); ////////////////////////// Compute distance to near plane /////////////////////////////// float Depth = Pos.x*NearPlane.x + Pos.y*NearPlane.y + Pos.z*NearPlane.z + NearPlane.w; Out.Depth.x = Depth*FarZMult; //////////////////////// Pass on Normal Vector And Dither Factor //////////////////////// Out.NormalAndDitherFactor = Ratio1*In.NormalAndDitherFactor1 + Ratio2*In.NormalAndDitherFactor2; /////////////////////// Pass on Light Components and Material /////////////////////////// Out.LightComponentsAndMaterial.xyz = In.LightComponentsAndMaterial1.xyz*Ratio1 + In.LightComponentsAndMaterial2.xyz*Ratio2; Out.LightComponentsAndMaterial.w = In.LightComponentsAndMaterial1.w; ///////////////////////////////////////////////////////////////////////////////////////// return Out; }

[MJP 19921]

Looks like it's talking about the lrp instruction. I don't know why the shader compiler would mess this up though...

On an unrelated (or maybe related) note...why do you have all of your inputs mapped to COLOR's?

[PeterV 116]

I've found the problem. As you can see in the code above, I had used the wrong input register numbers in the VSInput struct, COLOR0, COLOR1, COLOR2 and then again COLOR0, COLOR1, COLOR2. Should have been 0, 1, 2, 3, 4, 5.

The reason I'm using all colors for input is because all values are compressed to bytes. I'm rendering scenes with millions of vertices, streamed from the hard-disk. I need to save a lot of both video and system mem and increase data transfer speed from hard-disk.