ArrayFire is an open-source mathematical library for CPU and GPU acceleration of mathematical functions. ArrayFire gives you the option to run mathematical functions on the CPU and as well as the GPU. If you execute code on your GPU/video card, it will free up the CPU to do other tasks. The better the video card you have is, the more noticeable the difference in execution time will be. This is clearly visible when you have larger matrices, FFT, convolution and more. ArrayFire supports all CUDA –capable NVIDIA video cards and OpenCL devices such as the AMD video cards. In addition, ArrayFire also works on ordinary CPUs, meaning code does not have to be executed on the GPU for it to be accelerated. For more information on ArrayFire, see arrayfire.com/why-arrayfire.
MatDeck allows you to set which GPU modes ArrayFire will work in. ArrayFire functions can be mixed with other MatDeck functions, GUIs and features. For instance, a single MatDeck document can run multiple video cards from different manufacturers, this speeds up the function execution time drastically and is crucial for areas such as AI development.
Please note, to use ArrayFire in MatDeck, Microsoft Visual Studio Run Time will have to be installed before installing ArrayFire. Your video card must also have the latest driver.
Examples
- Cholesky decomposition using OpenCL
- FFT using CUDA
- FFT using OpenCL
- Hardware acceleration environment setup
- LU decomposition using CUDA
- LU decomposition using OpenCL
- QR decomposition using CUDA
- QR decomposition using OpenCL
- SVD decomposition using CUDA
- SVD decomposition using OpenCL
References
www.arrayfire.org ArrayFire is a high performance software library for parallel computing
www.arrayfire.com The Fastest Library for GPUs