Jacket v0.4 Release: The GPU Engine for MATLAB


We are happy to announce the release of Jacket beta version 0.4 - adding Windows 64-bit to our list off supported operating systems! You can download your free version of Jacket by logging in and downloading the new release:


In this newer version of Jacket, you will find:

  • Increased performance: Jacket speed improvements for many applications are very close to speeds achieved with raw CUDA code.
  • Increased operating system support: Jacket now supports Windows 32-bit, Windows 64-bit, Linux 32-bit, Linux 64-bit, and MacOSX 32-bit.
  • Improved benchmarking tools: gforce() now properly blocks until all computations are finished.
  • New installers: Jacket now comes with an automated installer to guide you through the installation process.
  • New features:
    ** find() no longer uses CUDPP and now performs well under memory pressure
    ** gsingles can now be used in subscripted reference and assignment
    ** gselect() allows you to select your computational GPU in multi-GPU setups

Also, along with Jacket v0.4, we are releasing new documentation and examples (both included in the download and available online):

We look forward to hearing your feedback and guidance on our forums (http://www.accelereyes.com/forums) and look forward to continually upgrading Jacket to meet your needs. If you are going to NVISION 2008, please stop by our booth - we’d love to meet with you!

Best regards,

The AccelerEyes Team
Atlanta, GA, USA
T: 770.315.1099
E: sales@accelereyes.com

You guys seem to be pretty fast! Good Luck with your work!

how do you handle the ~5sec kernel limit? will matlab crash, or will you catch the aborted kernel and notify the user?

We break the program down into many smaller kernels, so none run anywhere close to 5s.