#include "ShaderLibrary/Common.hlsl"

Texture2D<float> _Source;
RWTexture2D<float> _Result;

SamplerState sampler_PointClamp;    //TODO: could we use min-sampler instead of using ALU?

CBUFFER_START(cb)
float4 _SrcSize;
CBUFFER_END

#pragma kernel KMain
[numthreads(8, 8, 1)]
void KMain(uint2 groupId : SV_GroupID, uint2 groupThreadId : SV_GroupThreadID, uint2 dispatchThreadId : SV_DispatchThreadID)
{
    // Upper-left pixel coordinate of quad that this thread will read
    int2 threadUL = dispatchThreadId;

    // Downsample the block
    float2 offset = float2(threadUL) * 2.0f + 1.0f;
    float4 depths = _Source.GatherRed(sampler_PointClamp, offset * _SrcSize.zw, 0.0);

    float minDepth = min(min(depths.x, depths.y), min(depths.z, depths.w));

    // Write to the final target
    _Result[dispatchThreadId] = minDepth;
}