#include "ShaderLibrary/Common.hlsl"
#include "ShaderLibrary/BC6H.hlsl"
#include "ShaderLibrary/Sampling.hlsl"

TextureCube<float4> _Source;
RWTexture2DArray<uint4> _Target;
int _MipIndex;

SamplerState s_point_clamp;

static const uint2 kOffsets[] =
{
    uint2(0, 0), uint2(1, 0), uint2(2, 0), uint2(3, 0),
    uint2(0, 1), uint2(1, 1), uint2(2, 1), uint2(3, 1),
    uint2(0, 2), uint2(1, 2), uint2(2, 2), uint2(3, 2),
    uint2(0, 3), uint2(1, 3), uint2(2, 3), uint2(3, 3),
};

#pragma kernel KEncodeFastCubemapMip
[numthreads(1, 1, 1)]
void KEncodeFastCubemapMip(uint3 groupId : SV_GroupID, uint3 groupThreadId : SV_GroupThreadID, uint3 dispatchThreadId : SV_DispatchThreadID)
{
    uint width = 0;
    uint height = 0;
    _Source.GetDimensions(width, height);
    int size = width >> _MipIndex;

    // Load 4x4 pixel block
    float3 texels[16];

    uint2 topLeftSourceID = dispatchThreadId.xy << 2;
    uint faceId = dispatchThreadId.z;

    float2 nvc;
    float3 dir;
    for (uint i = 0u; i < 16; ++i)
    {
        nvc = CubemapTexelToNVC(topLeftSourceID + kOffsets[i], size);
        dir = CubemapTexelToDirection(nvc, faceId);
        texels[i] = _Source.SampleLevel(s_point_clamp, dir, _MipIndex).rgb;
    }

    uint4 block = uint4(0, 0, 0, 0);
    float blockMSLE = 0;

    EncodeMode11(block, blockMSLE, texels);

    _Target[dispatchThreadId] = block;
}