enum class NISGPUArchitecture : uint32_t
{
    NVIDIA_Generic = 0,
    AMD_Generic = 1,
    Intel_Generic = 2
};

struct NISOptimizer
{
    bool isUpscaling;
    NISGPUArchitecture gpuArch;

    NISOptimizer(bool isUpscaling = true,
                 NISGPUArchitecture gpuArch = NISGPUArchitecture::NVIDIA_Generic);
    uint32_t GetOptimalBlockWidth();
    uint32_t GetOptimalBlockHeight();
    uint32_t GetOptimalThreadGroupSize();
};

enum class NISHDRMode : uint32_t
{
    None = 0,
    Linear = 1,
    PQ = 2
};

NISOptimizer opt(true, NISGPUArchitecture::NVIDIA_Generic);
uint32_t blockWidth = opt.GetOptimalBlockWidth();
uint32_t blockHeight = opt.GetOptimalBlockHeight();
uint32_t threadGroupSize = opt.GetOptimalThreadGroupSize();

Defines defines;
defines.add("NIS_SCALER", true);
defines.add("NIS_HDR_MODE", hdrMode);
defines.add("NIS_BLOCK_WIDTH", blockWidth);
defines.add("NIS_BLOCK_HEIGHT", blockHeight);
defines.add("NIS_THREAD_GROUP_SIZE", threadGroupSize);
NVScalerCS = CompileComputeShader(device, "NIS_Main.hlsl”, &defines);

struct NISConfig
{
    float kDetectRatio;
    float kDetectThres;
    float kMinContrastRatio;
    float kRatioNorm;
    ...
};

NISConfig config;
createConstBuffer(&config, &csBuffer);

const int rowPitch = kFilterSize * 4;
const int imageSize = rowPitch * kPhaseCount;

createTexture2D(kFilterSize / 4, kPhaseCount, DXGI_FORMAT_R32G32B32A32_FLOAT, D3D11_USAGE_DEFAULT, coef_scaler, rowPitch, imageSize, &scalerTex);

createTexture2D(kFilterSize / 4, kPhaseCount, DXGI_FORMAT_R32G32B32A32_FLOAT, D3D11_USAGE_DEFAULT, coef_usm, rowPitch, imageSize, &usmTex);

createSRV(scalerTex.Get(), DXGI_FORMAT_R32G32B32A32_FLOAT, &scalerSRV);
createSRV(usmTex.Get(), DXGI_FORMAT_R32G32B32A32_FLOAT, &usmSRV);

createLinearClampSampler(&linearClampSampler);

void NVScalerUpdateConfig(NISConfig& config,
    float sharpness,
    uint32_t inputViewportOriginX, uint32_t inputViewportOriginY,
    uint32_t inputViewportWidth, uint32_t inputViewportHeight,
    uint32_t inputTextureWidth, uint32_t inputTextureHeight,
    uint32_t outputViewportOriginX, uint32_t outputViewportOriginY,
    uint32_t outputViewportWidth, uint32_t outputViewportHeight,
    uint32_t outputTextureWidth, uint32_t outputTextureHeight,
    NISHDRMode hdrMode = NISHDRMode::None
);

NVScalerUpdateConfig(m_config, sharpness,
                0, 0, inputWidth, inputHeight, inputWidth, inputHeight,
                0, 0, outputWidth, outputHeight, outputWidth, outputHeight,
                NISHDRMode::None);

updateConstBuffer(&config, csBuffer.Get());

$> cd samples
$> mkdir build
$> cd build
$> cmake ..