GPU V3 makes developers more productive by radically simplifying GPU programming.Developers can now harness the power of GPUs without any GPU programming experience.
Offloading compute intensive calculations to a GPU can significantly speed up multi-threaded .NET C# applications. Impressive performance gains of a factor 50 to 100 or more have been observed in real production code.
Alea GPU comes with a large collection of fully documented and ready to use samples. They guide the programmer from simple yet powerful applications of parallel-for to advanced GPU programming. Alea GPU runs on Windows, Linux or Mac OS X, is conveniently installed from NuGet packages and easily deployed without recompilation on the target platform.
Alea GPU provides highly efficient GPU parallel-for and parallel aggregate methods to map data parallel computations to the GPU.
Automatic Memory Management
Alea GPU automatically manages GPU memory and economically transfers data between the main memory and the GPU. The programmer can concentrate on the core logic and the algorithm.
Commonly used parallel performance primitives and NVIDIA libraries such cuRand, cuBlas or cuDNN are seamlessly integrated with Alea GPU.
Alea TK is an open source machine learning library based on Alea GPU. It uses tensors and automatic differentiation to build and train deep networks on GPUs efficiently. Join the community and help us to extend the library!