#if CUDA_VERSION < 12090
    nppSafeCall(nppGetStreamContext(&nppStreamContext)); 
#else
    // Manual context construction below

int device = 0;
cudaSafeCall(cudaGetDevice(&device));

cudaDeviceProp prop{};
cudaSafeCall(cudaGetDeviceProperties(&prop, device));

nppStreamContext.nCudaDeviceId = device;
nppStreamContext.nMultiProcessorCount = prop.multiProcessorCount;
nppStreamContext.nMaxThreadsPerMultiProcessor = prop.maxThreadsPerMultiProcessor;
nppStreamContext.nMaxThreadsPerBlock = prop.maxThreadsPerBlock;
nppStreamContext.nSharedMemPerBlock = prop.sharedMemPerBlock;
nppStreamContext.nCudaDevAttrComputeCapabilityMajor = prop.major;
nppStreamContext.nCudaDevAttrComputeCapabilityMinor = prop.minor;