How convert this shader code [passing parameter on "compile" line], so can build with MGPT?

Content
1

This is a fragment from an XNA 3.1 “v3” .fx file (this won’t build on its own, but this is part of a large file; hopefully this is enough to understand the problem):

float4 GaussianBlur_PS1_V1(VS_Simple_Terrain_Out input, uniform float2 vDir) : COLOR0
{
	float4 vColor = 0.0f;

	for (int i = 0; i < iGB_KERNEL_SIZE; i++) {
		float4 vTexCoords = float4(input.vTexCoords + vDiffuseTexelSize.xy * fGB_TexelScale.xx * vDir.xy * fGB_PixelKernel[i].xx, 0, iGB_LOD);

		vColor += tex2Dlod(sDiffuseTexture, vTexCoords) * fGB_BlurWeights[i] * fGB_Strength;
	}
	
	return vColor;
}

technique GaussianBlurH1_V1 {
	pass p0 {
		VertexShader = compile vs_4_0 VS_Simple_Terrain();
		PixelShader = compile ps_4_0 GaussianBlur_PS1_V1(float2(1.0f, 0.0f));
	}
}

technique GaussianBlurV1_V1 {
	pass p0 {
		VertexShader = compile vs_4_0 VS_Simple_Terrain();
		PixelShader = compile ps_4_0 GaussianBlur_PS1_V1(float2(0.0f, 1.0f));
	}
}


technique GaussianBlurV1 {
	pass horizontal {
		VertexShader = compile vs_4_0 VS_Simple_Terrain();
		PixelShader = compile ps_4_0 GaussianBlur_PS1_V1(float2(1.0f, 0.0f));
	}
	pass vertical {
		VertexShader = compile vs_4_0 VS_Simple_Terrain();
		PixelShader = compile ps_4_0 GaussianBlur_PS1_V1(float2(0.0f, 1.0f));
	}
}

Above, I’ve replaced “vs_3_0”/“ps_3_0” with “vs_4_0”/“ps_4_0”; otherwise MGPT won’t compile it. Most of the file compiles (and I have successfully compiled other files as well, by making this change).

But on all the “compile ps_4_0” lines above I get error:

Unexpected token ‘f’ found. Expected CloseParenthesis.
Unexpected toekn ‘f’ found. Expected CloseBracket.

I can get this to build by removing the parameter being passed:

		PixelShader = compile ps_4_0 GaussianBlur_PS1_V1(float2(1.0f, 0.0f));

=>

		PixelShader = compile ps_4_0 GaussianBlur_PS1_V1();

BUT that won’t produce the desired effect - that parameter is being used to make different variants of the PixelShader.

Q: How should I make these different variants, in ps_4_0 / MGPT?
That is, how can I pass in this parameter during compilation, or what is an alternative technique for producing these variants?

(Using “4_0” under the theory that I should first make the minimum version upgrade necessary to build with MGPT. If it helps, I can use any shader version supported by D3D 11 - though I would need to know if that will require additional changes.)

2

I remind having read somewhere here shaders could no longer have parameter (dx 10 or 11)

You can define this variable as you do for the world matrix for ex (at the top of the file) , with setparameter in the c#
Or as this is for a simple gaussuan blur, create 2 functions dedicated to each direction (this saves a setparameter per update) one for each technique
Many other solutions are possible

1 Like
3

Yeah, shader parameters like that are not supported in MG. You could just set the blur direction as a uniform in cpu code instead.