Setup & Installation

This tutorial gets you from zero to a working Hybridizer environment. By the end, you'll have compiled and run your first GPU-accelerated build.

Hybridizer supports Windows and Linux, using .NET 8 (or later).

Prerequisites

Component	Minimum	Recommended
OS	Windows 10 x64 / Ubuntu 22.04	Windows 11 x64 / Ubuntu 24.04
.NET SDK	.NET 8.0	.NET 8.0 (latest patch)
GPU	Any NVIDIA (Compute ≥ 5.0)	RTX 2060 or newer
RAM	8 GB	16 GB
Disk	10 GB free	SSD recommended

Step 1: Install NVIDIA Drivers

Windows
Linux

Go to nvidia.com/drivers or open NVIDIA GeForce Experience
Download and install the latest driver
Verify:

nvidia-smi

Install the recommended driver for your distribution:

# Ubuntu / Debian
sudo apt update
sudo apt install -y nvidia-driver-560
sudo reboot

# After reboot, verify:
nvidia-smi

Alternatively, install via the CUDA toolkit which bundles a compatible driver.

You should see your GPU name, driver version, and CUDA version:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 560.xx    Driver Version: 560.xx    CUDA Version: 12.6          |
|-------------------------------+----------------------+----------------------+
| GPU  Name        | ...       |                      |
| GeForce RTX 4070 | ...       |                      |
+-------------------------------+----------------------+----------------------+

Step 2: Install CUDA Toolkit

Download from developer.nvidia.com/cuda-downloads.

Windows
Linux

Choose Windows → x86_64 → exe (local)
Run the installer — default options are fine
Verify:

nvcc --version

tip

If nvcc is not found, add C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.x\bin to your system PATH.

# Ubuntu 22.04 / 24.04 — network install
wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/x86_64/cuda-keyring_1.1-1_all.deb
sudo dpkg -i cuda-keyring_1.1-1_all.deb
sudo apt update
sudo apt install -y cuda-toolkit-12-6

Add to your shell profile (~/.bashrc or ~/.zshrc):

export PATH=/usr/local/cuda/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH

Verify:

source ~/.bashrc
nvcc --version

Expected output:

nvcc: NVIDIA (R) Cuda compiler driver
Cuda compilation tools, release 12.x, V12.x.xxx

Step 3: Install .NET 8 SDK

Windows
Linux

Download and install from dotnet.microsoft.com/download.

Alternatively, if using Visual Studio 2022 (17.8+), .NET 8 is included with the .NET desktop development workload.

# Ubuntu (via Microsoft packages)
sudo apt install -y dotnet-sdk-8.0

# Or via the install script
curl -sSL https://dot.net/v1/dotnet-install.sh | bash /dev/stdin --channel 8.0

Verify:

dotnet --version    # Should print 8.0.xxx

Step 4: Install a C++ Toolchain

Hybridizer generates C++ code that needs to be compiled by nvcc and a host C++ compiler.

Windows
Linux

Install Visual Studio 2022 with the following workloads:

✅ .NET desktop development
✅ Desktop development with C++ (provides MSVC, needed by nvcc)

Or install the Build Tools for Visual Studio (lighter, no IDE).

# GCC (required by nvcc)
sudo apt install -y build-essential

# Verify
g++ --version

Step 5: Install Hybridizer

Create a new .NET 8 console project and add the Hybridizer packages:

dotnet new console -n MyFirstHybridizer --framework net8.0
cd MyFirstHybridizer
dotnet add package Hybridizer.Runtime.CUDAImports

tip

On Windows with Visual Studio, you can also install the Hybridizer Community Edition extension from the Visual Studio Marketplace for integrated project templates.

Step 6: Verify Your Setup

Replace the content of Program.cs with:

using System;
using Hybridizer.Runtime.CUDAImports;

class Program
{
    [EntryPoint]
    public static void TestKernel(int[] output, int N)
    {
        for (int i = threadIdx.x + blockDim.x * blockIdx.x;
             i < N;
             i += blockDim.x * gridDim.x)
        {
            output[i] = i * 2;
        }
    }

    static void Main()
    {
        // Check GPU
        cuda.GetDeviceProperties(out cudaDeviceProp prop, 0);
        Console.WriteLine($"GPU: {new string(prop.name)}");
        Console.WriteLine($"SMs: {prop.multiProcessorCount}");
        Console.WriteLine($"Memory: {prop.totalGlobalMem / (1024*1024)} MB");

        // Run kernel
        int N = 1024;
        int[] output = new int[N];

        dynamic wrapper = HybRunner.Cuda()
            .SetDistrib(32, 256);
        wrapper.TestKernel(output, N);
        cuda.DeviceSynchronize();

        // Verify
        bool ok = true;
        for (int i = 0; i < N; i++)
        {
            if (output[i] != i * 2) { ok = false; break; }
        }

        Console.WriteLine(ok ? "✅ Hybridizer is working!" : "❌ Something went wrong");
    }
}

Build and run:

dotnet build
dotnet run

Expected output:

GPU: NVIDIA GeForce RTX 4070
SMs: 46
Memory: 12282 MB
✅ Hybridizer is working!

Troubleshooting

Problem	Solution
`nvidia-smi` not found	Install/update NVIDIA drivers
`nvcc` not found	Reinstall CUDA Toolkit, check PATH
`dotnet` not found	Install .NET 8 SDK
Build error "Hybridizer not found"	Verify NuGet package: `dotnet list package`
Runtime "No CUDA device"	Check GPU with `nvidia-smi`, update drivers
`libcudart.so` not found (Linux)	Set `LD_LIBRARY_PATH` to CUDA lib directory
DLL not found at runtime (Windows)	Ensure CUDA bin directory is in PATH

You're ready! Proceed to Your First Kernel →

Prerequisites​

Step 1: Install NVIDIA Drivers​

Step 2: Install CUDA Toolkit​

Step 3: Install .NET 8 SDK​

Step 4: Install a C++ Toolchain​

Step 5: Install Hybridizer​

Step 6: Verify Your Setup​

Troubleshooting​

Next​