jCUDA (Java for CUDA®) offers CUDA® bindings for the Java language, exploiting the full power of GPU computing from Java based applications.

The library is available for both Windows and Linux platforms, running 32 or 64 bit versions.
It can be used to accelerate client or server side applications performing intensive computations or high-end visualization.

Using jCUDA you can create cross-platform CUDA® solutions, that can run on any operating system supported by CUDA® without changing your code.
Either select between Microsoft® Windows® XP, Vista, 7 or 8 by or even Linux/MacOSX/Solaris systems.
Current support is for both 32 and 64 bits of every platform.


jCUDA is provided free of charge, for private, academic, research or commercial use.
The library is provided as is, without warranty or support of any kind.
NOTE: This development is considered legacy and no longer maintained. It is provided only for convenience.
For support or development services: support@cass-hpc.com.


Checkout the archive page.


  • Double precision
  • Object oriented model for CUDA programming
  • CUDA 2.1 Driver API
  • CUDA 2.1 Runtime API
  • CUFFT routines
  • OpenGL interoperability
  • * Support for CUBLAS routines will be added in the future

Operating System Support

  • Microsoft® Windows®
  • Linux
  • * Support for Mac OSX might be added in the future
  • * Support for Sun Solaris (x86) might be added in the future

Requirements: CUDA® version 1.1 and up, Java runtime 1.5 or newer.