The Cudafy.Translator library must be rewritten, hence the question arises whether it makes sense to continue the work of the Hybrid DSP, or to rewrite everything from scratch with the MIT license.
If anyone is willing to cooperate, please contact me. Currently, everything is basically rewritten to .NET 6. The Cudafy.Translator library fails to compile due to problems with Decompiler, NRefactory and ILSpy.
If you want, you can download the last successful compilation of the library here.
Source code of the library, before .NET6 refactor, can be downloaded here
The original repository is no longer available for download. CodePlex has been shutted down
CUDA 11 & Visual Studio 2022
CUDAfy.NET access to work with Visual Studio 2022 and the latest NVIDIA Toolkit CUDA 11 library.
This repository uses the source code created by Hybrid DSP to create a derivative work under the LGPL license.
CUDAfy .NET allows easy development of high performance GPGPU applications completely from the .NET. It's developed in C#.
Modern graphics cards provide the potential of massive speed increase over CPUs for non-graphics related intensive numeric operations. Many large data set operations such as matrices can see a 100x or more speed up. CUDAfy allows .NET developers to easily create complex applications that split processing cleanly between host and GPU. There are no separate CUDA cu files or complex set-up procedures to launch GPU device functions. It follows the CUDA programming model and any knowledge gained from tutorials or books on CUDA can be easily transferred to CUDAfy, only in a clean .NET fashion.
- x64 Windows or Linux
- Visual Studio 2022
- MSVC v142 x64 / 86 build tools (v.14.24) or higher
- .NET Framework 4.8 | .NET 6
- Download the latest repository
- Open the Cudafy.sln project
- Choose the "Debug" solution
- Choose the "x64" solution platforms
- Choose the "Cudafy.Demo" startup project
- Rebuild whole project
- Start "Cudafy.Demo"
- Automatic support for versions 10.2, 10.1 and 11.
- A new way to detect Visual Studio locations with the MSVC package.
- Corrections in the code of some test programs.
- Almost complete removal of x86 support.
- Removal of unnecessary Solutions.
- Some new errors to display.
You can download the latest CUDA version from the official NVIDIA website NVIDIA CUDA Toolkit
Here is an example program you can run using compiled libraries.
class Program
{
public const int N = 10;
public static void Main()
{
Console.WriteLine("CUDAfy Example\nCollecting necessary resources...");
CudafyModes.Target = eGPUType.Cuda; // To use OpenCL, change this enum
CudafyModes.DeviceId = 0;
CudafyTranslator.Language = CudafyModes.Target == eGPUType.OpenCL ? eLanguage.OpenCL : eLanguage.Cuda;
//Check for available devices
if (CudafyHost.GetDeviceCount(CudafyModes.Target) == 0)
throw new System.ArgumentException("No suitable devices found.", "original");
//Init device
GPGPU gpu = CudafyHost.GetDevice(CudafyModes.Target, CudafyModes.DeviceId);
Console.WriteLine("Running example using {0}", gpu.GetDeviceProperties(false).Name);
//Load module for GPU
CudafyModule km = CudafyTranslator.Cudafy();
gpu.LoadModule(km);
//Define local arrays
int[] a = new int[N];
int[] b = new int[N];
int[] c = new int[N];
// allocate the memory on the GPU
int[] dev_c = gpu.Allocate<int>(c);
// fill the arrays 'a' and 'b' on the CPU
for (int i = 0; i < N; i++)
{
a[i] = i;
b[i] = i * i;
}
// copy the arrays 'a' and 'b' to the GPU
int[] dev_a = gpu.CopyToDevice(a);
int[] dev_b = gpu.CopyToDevice(b);
gpu.Launch(1, N).add(dev_a, dev_b, dev_c);
// copy the array 'c' back from the GPU to the CPU
gpu.CopyFromDevice(dev_c, c);
// display the results
for (int i = 0; i < N; i++)
Console.WriteLine("{0} + {1} = {2}", a[i], b[i], c[i]);
// free the memory allocated on the GPU
gpu.FreeAll();
Console.WriteLine("Done!");
Console.ReadKey();
}
[Cudafy]
public static void add(GThread thread, int[] a, int[] b, int[] c)
{
int tid = thread.threadIdx.x;
if (tid < N)
c[tid] = a[tid] + b[tid];
}
}Cudafy.NET was created by Hybrid DSP under LGPL v2.1 License.
The LGPL v2.1 License applies to CUDAfy .NET. If you wish to modify the code then changes should be re-submitted to Hybrid DSP. If you wish to incorporate Cudafy.NET into your own application instead of redistributing the dll's then please consider a commerical license. Visit http://www.hybriddsp.com. This will also provide you with priority support and contribute to on-going development.
The following libraries are made use of:
The MIT license applies to ILSpy, NRefactory and ICSharpCode.Decompiler (Copyright (c) 2011 AlphaSierraPapa for the SharpDevelop team).
Mono.Cecil also uses the MIT license (Copyright JB Evain).
CUDA.NET is a free for use license (Copyright Company for Advanced Supercomputing Solutions Ltd)