The Package is installable in GNU Octave and offers OpenCL support for parallelization of numerical computations. It is most effective for computations which are based on large vectors or n-dimensional arrays of numbers, and mostly (but not limited to) identical element-wise operations. The user can select which available OpenCL hardware and drivers are to be used.
The Package offers new numeric data types with many built-in operations (e.g., overloaded operators). It is flexibly extendible by user-written OpenCL C programs. The Package does not, by itself, provide parallelization of higher numerical methods (like BLAS or LAPACK).
The Package is also available from the corresponding Octave Forge webpage https://octave.sourceforge.io/ocl/index.html. More information on GNU Octave can be found at https://www.octave.org.
Features
- parallel computing
- GPU computing
- hardware & driver selection
License
GNU General Public License version 3.0 (GPLv3)Follow octave-ocl
User Reviews
-
Dear Matt, Thank you very much for this Matlab style GPU operation on OpenCL. This is an amazing step for the opensource and scientific communities. Brilliant! Currently, I am reading your readme file and I could feel the ease of research on all GPUs are finally coming. I will put all the stars first so that other could notice, then try on the windows based Octave and share my experience with my colleagues. Fantastic!
-
I'm amazed that it can be so easy to do gpu computing. I was hoping for this since years. Just installed intel-opencl-icd on Ubuntu and this package into octave and some of my matrix operations are 10x faster without much rewrite - wow! I'd like to share one pitfall I ran into. This package is not working if octave is installed as snap or flatpak. It seems that software can't access the icd system library from the package due to restrictions.