More about Mathematica »
More User Stories »

CUDA & Heterogeneous Programming with the Wolfram Language

Simplify Access to Performance Computing in Industry and Research

The Wolfram Language combines programming ease-of-use with the computational speed of GPU-equipped hardware to dramatically increase application performance and user productivity.

The Wolfram Language simplifies GPU programming with CUDALink and OpenCLLink, which are high-level interfaces to the GPU built on top of the Wolfram Language's development technologies. They allow you to execute code on your GPU with minimal effort.

By fully integrating and automating the GPU's capabilities using the Wolfram Language, hiding unnecessary complexity associated with GPU programming, and lowering the learning curve for GPU computation, you experience a more productive and efficient development cycle.

Computers that have CUDA or heterogeneous support are supported by CUDALink or OpenCLLink with the Wolfram Language. CUDALink and OpenCLLink are supported on Windows (32 and 64 bit), Mac OS X (64 bit), and Linux (32 and 64 bit).


Wolfram Language GPU programming features include:

  • Intuitive CUDALink and OpenCLLink programming interfaces
  • Automatic memory management, multicore computing, and built-in functions for many applications
  • Elimination of the need to write C/C++, Fortran, or CUDA/heterogeneous wrapper code for GPU computing
  • Support for both single and double precision
  • Integration with Mathematica without the need for add-ons
de es ja pt-br zh