DewarpCompute.compute

#pragma kernel WCtoBM_Resize
#pragma kernel bmOutputToUVTexture
RWBuffer<float> result_buffer;
Buffer<float> input_buffer;
RWTexture2D<float4> result_texure;


// Specialized 256x256 to 128x128
// Dispatch(WCtoBM_ResizeKernelID, 4, 4, 3);
[numthreads(32, 32, 1)]
void WCtoBM_Resize(uint3 id : SV_DispatchThreadID)
{
	result_buffer[((id.y) * 128 + id.x) + id.z * 128*128] = saturate((
		input_buffer[(id.y * 2 + 0) * 256 + (id.x * 2 + 0) + id.z * 256 * 256] +
		input_buffer[(id.y * 2 + 1) * 256 + (id.x * 2 + 0) + id.z * 256 * 256] +
		input_buffer[(id.y * 2 + 0) * 256 + (id.x * 2 + 1) + id.z * 256 * 256] +
		input_buffer[(id.y * 2 + 1) * 256 + (id.x * 2 + 1) + id.z * 256 * 256]) * 0.25);
}

//Dispatch(bmOutputToUVTextureKernelID, 4, 4, 1);
[numthreads(32, 32, 1)]
void bmOutputToUVTexture(uint2 id : SV_DispatchThreadID)
{
	float u = 0;
	float v = 0;
	const int r = 1;
	const int2 xRange = int2(max((int)id.x - r, 0), min(id.x + r, 127));
	const int2 yRange = int2(max((int)id.y - r, 0), min(id.y + r, 127));

	for (int j = yRange.x; j <= yRange.y; j++)
	{
		for (int i = xRange.x; i <= xRange.y; i++)
		{
			u += input_buffer[j * 128 + i];
			v += input_buffer[j * 128 + i + 128 * 128];
		}
	}
	const int kernelSize = (xRange.y - xRange.x + 1) * (yRange.y - yRange.x + 1);
	result_texure[id] = saturate(float4((u + kernelSize) * (1.0 / (2 * kernelSize)), (v + kernelSize) * (1.0 / (2 * kernelSize)), 0, 1));
}