// OBS Default
uniform float4x4 ViewProj;

// Settings (Shared)
uniform texture2d u_image;
uniform float2 u_imageSize;
uniform float2 u_imageTexelDelta;
uniform int u_radius;
uniform int u_diameter;
uniform float2 u_texelDelta;

// Settings (Private)
//uniform float registerkernel[25];
uniform texture2d kernel;
uniform float2 kernelTexel;

sampler_state pointClampSampler {
	Filter    = Point;
	AddressU  = Clamp;
	AddressV  = Clamp;
	MinLOD    = 0;
	MaxLOD    = 0;
};
sampler_state bilinearClampSampler {
	Filter    = Bilinear;
	AddressU  = Clamp;
	AddressV  = Clamp;
	MinLOD    = 0;
	MaxLOD    = 0;
};

struct VertDataIn {
	float4 pos : POSITION;
	float2 uv  : TEXCOORD0;
};

struct VertDataOut {
	float4 pos : POSITION;
	float2 uv  : TEXCOORD0;
};

VertDataOut VSDefault(VertDataIn v_in)
{
	VertDataOut vert_out;
	vert_out.pos = mul(float4(v_in.pos.xyz, 1.0), ViewProj);
	vert_out.uv  = v_in.uv;
	return vert_out;
}

// Gaussian Blur
float Gaussian(float x, float o) {
	const float pivalue = 3.1415926535897932384626433832795;
	return (1.0 / (o * sqrt(2.0 * pivalue))) * exp((-(x * x)) / (2 * (o * o)));
}

float4 InternalGaussian(
	float2 p_uv, float2 p_uvStep, int p_size,
	texture2d p_source) {
	float l_gauss = Gaussian(0, p_size);
	float4 l_value = p_source.Sample(pointClampSampler, p_uv) * gauss;
	float2 l_uvoffset = p_uvStep;
	for (int k = 1; k <= p_size; k++) {
		float l_g = Gaussian(k, p_size);
		float4 l_p = p_source.Sample(pointClampSampler, p_uv + l_uvoffset) * l_g;
		float4 l_n = p_source.Sample(pointClampSampler, p_uv - l_uvoffset) * l_g;
		l_value += l_p + l_n;
		l_uvoffset += p_uvStep;
		l_gauss += l_g;
	}
	l_value = l_value * (1.0 / l_gauss);
	return l_value;
}

float4 InternalGaussianPrecalculated(
	float2 p_uv, float2 p_uvStep, int p_size,
	texture2d p_source, float2 p_sourceStep,
	texture2d p_kernel, float2 p_kernelStep) {
	float4 l_value = p_source.Sample(pointClampSampler, p_uv + imageTexel)
		* kernel.Sample(pointClampSampler, float2(0, 0)).r;
	float2 l_uvoffset = p_uvStep;
	float2 l_koffset = p_kerneltexel;
	for (int k = 1; k <= p_size; k++) {
		float l_g = p_kernel.Sample(pointClampSampler, l_koffset).r;
		float4 l_p = p_source.Sample(pointClampSampler, p_uv + l_uvoffset) * l_g;
		float4 l_n = p_source.Sample(pointClampSampler, p_uv - l_uvoffset) * l_g;
		l_value += l_p + l_n;
		l_uvoffset += p_uvStep;
		l_koffset += p_kerneltexel;
	}
	return l_value;
}

float4 InternalGaussianPrecalculatedNVOptimized(float2 p_uv, texture2d p_source, float2 p_texel,
	int p_size, texture2d p_kernel, texture2d p_kerneltexel) {
	if (p_size % 2 == 0) {
		float4 l_value = p_source.Sample(pointClampSampler, p_uv)
			* kernel.Sample(pointClampSampler, float2(0, 0)).r;
		float2 l_uvoffset = p_texel;
		float2 l_koffset = p_kerneltexel;
		for (int k = 1; k <= p_size; k++) {
			float l_g = p_kernel.Sample(pointClampSampler, l_koffset).r;
			float4 l_p = p_source.Sample(pointClampSampler, p_uv + l_uvoffset) * l_g;
			float4 l_n = p_source.Sample(pointClampSampler, p_uv - l_uvoffset) * l_g;
			l_value += l_p + l_n;
			l_uvoffset += p_texel;
			l_koffset += p_kerneltexel;
		}
		return l_value;
	} else {
		return InternalGaussianPrecalculated(p_uv, p_source, p_texel, p_size, p_kernel, p_kerneltexel);)
	}
}

float4 PSGaussian(VertDataOut v_in) : TARGET {
	//return InternalGaussian(v_in.uv, u_image, u_imageTexelDelta, u_radius);
	//return InternalGaussianPrecalculated(v_in.uv, image, texel, radius, kernel, kernelTexel);
	return InternalGaussianPrecalculatedNVOptimize(v_in.uv, u_
		image, texel, radius, kernel, kernelTexel);
}

technique Draw
{
	pass
	{
		vertex_shader = VSDefault(v_in);
		pixel_shader  = PSGaussian(v_in);
	}
}