#pragma kernel CSVFXCopyBuffer
#pragma kernel CSVFXCopyStructBuffer
#pragma kernel CSVFXInitDeadListBuffer
#pragma kernel CSVFXZeroInitBuffer
#pragma kernel CSVFXBatchCopyCountUint
#pragma kernel CSVFXBatchCopyCountKvp
#pragma kernel CSVFXInitBoundsBuffer
#pragma kernel CSVFXCopyStructBufferSelf

#pragma only_renderers d3d11 playstation xboxone xboxseries vulkan metal switch glcore gles3 webgpu

#include "HLSLSupport.cginc"

CBUFFER_START(Uniform)
    uint dstOffset;
    uint srcOffset;
    uint srcStride;
    uint size;
    uint dispatchWidth;
    uint instanceIndex;
CBUFFER_END

struct Kvp
{
    float sortKey;
    uint index;
};

StructuredBuffer<uint> srcStructBufferUint;
StructuredBuffer<Kvp> srcStructBufferKvp;
ByteAddressBuffer srcBuffer;
RWByteAddressBuffer dstBuffer;
RWStructuredBuffer<uint> dstStructBuffer;


#define NB_THREADS_PER_GROUP 64

[numthreads(NB_THREADS_PER_GROUP,1,1)]
void CSVFXCopyBuffer(uint3 id : SV_DispatchThreadID)
{
    if (id.x < size) /* One dword is copied by thread */
    {
        uint srcAddress = (id.x + srcOffset) << 2;
        uint dstAddress = (id.x + dstOffset) << 2;
        uint load = srcBuffer.Load(srcAddress);
        dstBuffer.Store(dstAddress, load);
    }
}

[numthreads(NB_THREADS_PER_GROUP,1,1)]
void CSVFXCopyStructBuffer(uint3 id : SV_DispatchThreadID)
{
    if (id.x < size) /* One dword is copied by thread */
    {
        uint srcIndex = id.x + srcOffset ;
        uint dstIndex = id.x + dstOffset ;
        dstStructBuffer[dstIndex] = srcStructBufferUint[srcIndex];
    }
}

[numthreads(NB_THREADS_PER_GROUP, 1, 1)]
void CSVFXInitDeadListBuffer(uint3 groupId : SV_GroupID, uint3 threadId : SV_DispatchThreadID)
{
    uint index = groupId.y * dispatchWidth * NB_THREADS_PER_GROUP + threadId.x;
    if (index < size)
        dstStructBuffer[dstOffset + index] = size - index - 1u;
    if(index == 0)
        dstStructBuffer[instanceIndex] = size;
}

[numthreads(NB_THREADS_PER_GROUP, 1, 1)]
void CSVFXZeroInitBuffer(uint3 groupId : SV_GroupID, uint3 threadId : SV_DispatchThreadID)
{
    uint index = groupId.y * dispatchWidth * NB_THREADS_PER_GROUP + threadId.x;
    if (index < size)
        dstBuffer.Store((dstOffset + index) << 2, 0u);
}

[numthreads(NB_THREADS_PER_GROUP,1,1)]
void CSVFXBatchCopyCountUint(uint3 id : SV_DispatchThreadID)
{
    if (id.x < size) /* One dword is copied by thread */
    {
        uint load = srcStructBufferUint[id.x];
        dstBuffer.Store((id.x + dstOffset) << 2, load);
    }
}
[numthreads(NB_THREADS_PER_GROUP,1,1)]
void CSVFXBatchCopyCountKvp(uint3 id : SV_DispatchThreadID)
{
    if (id.x < size) /* One dword is copied by thread */
    {
        uint load = srcStructBufferKvp[id.x].index;
        dstBuffer.Store((id.x + dstOffset) << 2, load);
    }
}

[numthreads(NB_THREADS_PER_GROUP,1,1)]
void CSVFXCopyStructBufferSelf(uint3 id : SV_DispatchThreadID)
{
    if (id.x < size) /* One dword is copied by thread */
    {
        uint srcIndex = id.x + srcOffset ;
        uint dstIndex = id.x + dstOffset ;
        dstStructBuffer[dstIndex] = dstStructBuffer[srcIndex];
    }
}

[numthreads(NB_THREADS_PER_GROUP, 1, 1)]
void CSVFXInitBoundsBuffer(uint3 groupId : SV_GroupID, uint3 threadId : SV_DispatchThreadID)
{
    uint index = groupId.y * dispatchWidth * NB_THREADS_PER_GROUP + threadId.x;
    if (index < size)
        dstStructBuffer[index] = index % 6 < 3 ? 0xffffffff : 0u;
}