Getting Started

Create a new .Net Framework 4.7.X (or higher) or a .Net Standard 2.1 compatible project (e.g. .Net Core 3.1) and install the required ILGPU Nuget package.

// Create new  project with .Net 4.7 (or higher) or .Net Standard 2.1 compatible (e.g. .Net Core 3.1)
dotnet new console
nuget install ILGPU

We recommend that you disable the Prefer 32bit option in the application build-settings panel. This typically ensures that the application is executed in native-OS mode (e.g. 64bit on a 64bit OS). You will not able to instantiate an Accelerator instance unless your application runs in native-OS mode since ILGPU requires direct interaction with the graphics-driver API.

The ILGPU compiler has been designed in way that it does not rely on native libraries. Therefore, it is not necessary to worry about such dependencies (except, of course, for the actual GPU drivers) or environment variables. Neither a Cuda SDK nor an OpenCL SDK have to be installed to use ILGPU.

While GPU programming can be done using only the ILGPU package, we recommend using the ILGPU.Algorithms library that realizes useful functions like scan, reduce and sort.

If you want to know about recent changes or new features refer to the upgrade Guide, the milestone descriptions or the change logs.

The ILGPU Context

All ILGPU classes and functions rely on the global ILGPU Context. Instances of classes that require a context reference have to be disposed before disposing of the main context. Note that all operations on a context and its children must be considered as not thread safe.

class ...
{
    static void Main(string[] args)
    {
        using var context = new Context();
        
        // ILGPU-specific functionality goes here

        // Dispose all other classes before disposing the ILGPU context
    }
}

Algorithms Library

The ILGPU.Algorithms library comes in a separate nuget package. In order to use any of the exposed group/warp/math extensions you have to enable the library. This setups all internal ILGPU hooks and custom code-generators to emit code that realizes the extensions in the right places. This is achieved by using the new extension and intrinsic API.

using ILGPU.Algorithms;
class ...
{
    static void ...(...)
    {
        using var context = new Context();

        // Enable all algorithms and extension methods
        context.EnableAlgorithms();

        ...
    }
}

Note that you have to call EnableAlgorithms before instantiating Accelerator objects.