CopyPositions.compute

#define GROUP_SIZE 128
#define ITEMS_PER_THREAD 48

int _InputPosBufferCount;
int _InputBaseVertex;
int _InputPosBufferStride;
int _InputPosBufferOffset;
int _OutputPosBufferOffset;

ByteAddressBuffer _InputPosBuffer;
RWStructuredBuffer<uint> _OutputPosBuffer;

#pragma kernel CopyVertexBuffer
[numthreads(GROUP_SIZE, 1, 1)]
void CopyVertexBuffer(uint gidx : SV_DispatchThreadID)
{
    for (int i = 0; i < ITEMS_PER_THREAD; ++i)
    {
        int offset = gidx * ITEMS_PER_THREAD + i;
        if (offset >= _InputPosBufferCount)
            return;

        uint inputAddr = _InputPosBufferOffset + (_InputBaseVertex + offset) * _InputPosBufferStride;
        uint3 val = _InputPosBuffer.Load3(inputAddr << 2);

        uint outputAddr = _OutputPosBufferOffset + 3 * offset;
        _OutputPosBuffer[outputAddr] = val.x;
        _OutputPosBuffer[outputAddr + 1] = val.y;
        _OutputPosBuffer[outputAddr + 2] = val.z;
    }
}