// No atomic support on CPU
//DISABLE_TEST(compute):COMPARE_COMPUTE_EX:-cpu -compute -shaderobj
// No support for int64_t on DX11
//DISABLE_TEST(compute):COMPARE_COMPUTE_EX:-slang -compute -shaderobj
// No support for int64_t on fxc - we need SM6.0 and dxil
// https://docs.microsoft.com/en-us/windows/win32/direct3dhlsl/hlsl-shader-model-6-0-features-for-direct3d-12
//DISABLE_TEST(compute):COMPARE_COMPUTE_EX:-slang -compute -dx12 -nvapi-slot u0 -shaderobj
//TEST(compute):COMPARE_COMPUTE_EX:-slang -compute -dx12 -profile cs_6_0 -render-features atomic-int64 -compile-arg -O2 -shaderobj
//TEST(compute, vulkan):COMPARE_COMPUTE_EX:-vk -compute -render-features atomic-int64 -shaderobj
//TEST(compute):COMPARE_COMPUTE_EX:-cuda -compute -shaderobj

// The test doesn't directly use this, but having this defined makes the 0 slot available if NVAPI is going to be used
// Only strictly necessary on the D3D12 path
//TEST_INPUT:ubuffer(data=[0 0 0 0 ], stride=4):name=nvapiBuffer
RWStructuredBuffer<int> nvapiBuffer;

//TEST_INPUT:ubuffer(data=[0 1 2 3 4 5 6 7]):out,name=outputBuffer
RWByteAddressBuffer outputBuffer;

[numthreads(16, 1, 1)]
void computeMain(int3 dispatchThreadID : SV_DispatchThreadID)
{    
    int tid = dispatchThreadID.x;
    int idx = (tid & 3) ^ (tid >> 2); 

    // Try directly reading
    uint2 currentValue2 = outputBuffer.Load2(idx << 3);
    uint64_t currentValue = uint64_t(currentValue2.y) | currentValue2.x;    
    
    while (true)
    {
        // This is probably not a great way to do this - InterlockedAddI64 would be better
        // but we are doing this to test CAS.
        
        uint64_t readValue;
        outputBuffer.InterlockedCompareExchangeU64(idx << 3, currentValue, currentValue + 1, readValue);
        
        if (readValue == currentValue)
        {
            break;
        }
        
        currentValue = readValue;
    }
}