Composing broadly useful programs to exploit NVIDIA GPUs has long implied utilizing NVIDIA's CUDA platform. While CUDA supports a few diverse programming langiages, composing superior code commonly required the utilization of C or C++. 

Author: Mohamed Musthaque


Another compiler tool from Altimesh, Hybridizer, is planned to address this issue by furnishing C# engineers with an approach to target GPUs by delivering a CUDA paired from their C# source. Hybridizer is part into two versions, with each focusing on various needs and spending plans. Hybridizer Fundamentals is allowed to all and is given as a Visual Studio extension. It produces binaries for the CUDA stage. Hybridizer Programming Suite (HSE) is authorized programming that permits the focusing of CUDA, and additionally different stages that incorporate AVX, AVX2, and AX512. The product suite produces binaries, however will likewise alternatively deliver CUDA source code to enable clients to review what is being aggregated.

Either alternative combined with NVIDIA's Nsight Visual Studio Edition gives engineers the way to compose and debug C# code in Visual Studio while the subsequent code executes on the NVIDIA GPU. Since the HSE in a general sense works on MSIL (Microsoft Intermediate Language), it can coordinate with existing activities regardless of whether the source code isn't accessible. This additionally gives indirect support to individual .NET platform languages F# and VB.NET.

Since one of the objectives of composing C/C++ code for the CUDA platform is most extreme execution, it merits looking at the execution C# code assembled by Hybridizer. As per Altimesh, the binaries created from C# accomplished 83% of the execution from the transcribed C++ code focusing on CUDA. With a few hands-on contribution of its own, the C# code was additionally enhanced to coordinate the execution of the C++ comparable.

The Hybridizer programming gives C# engineers inquisitive about CUDA and GPU programming with an approach to investigate these advances without abandoning their favored innovation. Sample code is accessible on GitHub and the Hybridizer Essentials extension is accessible on the Visual Studio Marketplace.

Source: InfoQ

 



Tags: .NET C# CUDA NVIDIA GPU
Views: 776
Register for more exciting articles

Comments

Please login or register to post a comment.


There are currently no comments.