+Red King Subscriber² Posted January 24, 2015 Subscriber² Share Posted January 24, 2015 I might have asked this in the past, but, What is the current best solution to run a kernel on GPU without explicitly coding in C++? That is, is there a more accessible / easy (and FREE) way to do computing on the GPU these days? Link to comment Share on other sites More sharing options...
0 Max Norris Posted January 24, 2015 Share Posted January 24, 2015 Language should be .NET compatible or, as end product, produce DLL which can be called from .NET Yea, probably start with CUDAfy then, that's purely dotNET and actively developed. Note you'll probably need the various SDK's from the GPU manufacturers as well depending on what platform you're targeting. https://cudafy.codeplex.com/ https://www.nuget.org/packages/CUDAfy.NET/ Link to comment Share on other sites More sharing options...
0 Max Norris Posted January 24, 2015 Share Posted January 24, 2015 What platform are we talking about? Off the top of my head I recall CUDAfy.NET, supports CUDA (nVidia specific) or OpenCL (nVidia, AMD and Intel). Open source on CodePlex or just pull the package from the NuGet repository. Python has a couple packages, PyCUDA, PyOpenCL, etc. There's JOCL for Java. Sure there's plenty of others. Link to comment Share on other sites More sharing options...
0 +Red King Subscriber² Posted January 24, 2015 Author Subscriber² Share Posted January 24, 2015 What platform are we talking about? Off the top of my head I recall CUDAfy.NET, supports CUDA (nVidia specific) or OpenCL (nVidia, AMD and Intel). Open source on CodePlex or just pull the package from the NuGet repository. Python has a couple packages, PyCUDA, PyOpenCL, etc. Windows 8.1 x64 using NV 970 Language should be .NET compatible or, as end product, produce DLL which can be called from .NET Link to comment Share on other sites More sharing options...
0 +Red King Subscriber² Posted January 24, 2015 Author Subscriber² Share Posted January 24, 2015 Yea, probably start with CUDAfy then, that's purely dotNET. Note you'll probably need the various SDK's from the GPU manufacturers as well depending on what platform you're targeting. https://cudafy.codeplex.com/ https://www.nuget.org/packages/CUDAfy.NET/ Still the best one out there, eh? Link to comment Share on other sites More sharing options...
0 Max Norris Posted January 24, 2015 Share Posted January 24, 2015 Still the best one out there, eh? *shrug* I've only dabbled, won't say it's the best but it's actively developed and reasonably popular on NuGet, didn't seem terribly hard but again, dabbling, a lot of it is either over my head or something I didn't really need, a curiosity thing for me. There's others, take a gander. https://www.nuget.org/packages?q=opencl https://www.nuget.org/packages?q=cuda Link to comment Share on other sites More sharing options...
0 SirFaizdat Posted March 3, 2015 Share Posted March 3, 2015 OpenCL allows for utilizing the GPU to make calculations. It is supported on almost every platform we see today, including Intel, ARM, Nvidia, and AMD, which are very popular. It is very well documented as well. C and C++ are its main languages, however there are many other bindings that support it. If you're willing to use Java, a great one would be LWJGL (http://lwjgl.org). It provides bindings for OpenCL, which is what I believe you are looking for, and also provides bindings for OpenGL and OpenAL. It's stable and mature and has a good documentation and community. Link to comment Share on other sites More sharing options...
Question
+Red King Subscriber²
I might have asked this in the past, but,
What is the current best solution to run a kernel on GPU without explicitly coding in C++?
That is, is there a more accessible / easy (and FREE) way to do computing on the GPU these days?
Link to comment
Share on other sites
6 answers to this question
Recommended Posts