#include "CoreRP/ShaderLibrary/Common.hlsl"

#pragma kernel KMain_8                KERNEL_SIZE=8     KERNEL_NAME=KMain_8
#pragma kernel KMain_1                KERNEL_SIZE=1     KERNEL_NAME=KMain_1

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


[numthreads(KERNEL_SIZE, KERNEL_SIZE, 1)]
void KERNEL_NAME(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;
#if defined(PLATFORM_SUPPORT_GATHER)
    float4 depths = GATHER_RED_TEXTURE2D(_Source, sampler_PointClamp, offset * _SrcSize.zw, 0.0);
#else
    // Downsample the block
    float p00 = SAMPLE_TEXTURE2D_LOD(_Source, sampler_PointClamp, (offset) * _SrcSize.zw, 0.0).x;
    float p10 = SAMPLE_TEXTURE2D_LOD(_Source, sampler_PointClamp, (offset + float2(1.0, 0.0)) * _SrcSize.zw, 0.0).x;
    float p01 = SAMPLE_TEXTURE2D_LOD(_Source, sampler_PointClamp, (offset + float2(0.0, 1.0)) * _SrcSize.zw, 0.0).x;
    float p11 = SAMPLE_TEXTURE2D_LOD(_Source, sampler_PointClamp, (offset + float2(1.0, 1.0)) * _SrcSize.zw, 0.0).x;
    float4 depths = float4(p00, p10, p01, p11);
#endif

    // Select the nearest sample
#if UNITY_REVERSED_Z
    float minDepth = max(min(depths.x, depths.y), max(depths.z, depths.w));
#else
    float minDepth = min(min(depths.x, depths.y), min(depths.z, depths.w));
#endif

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