GPUmat: GPU toolbox for MATLAB
The GP-you Group develops high-level software tools to easily access the parallel computing capabilities of the GPUs from different platforms. Thanks to the annual experience in software engineering and development (real time, embedded systems, numerical simulations), the GP-you Group proposes solutions that allow the end user to take full profit of the GPU capabilities without facing the costs related to the GPU code optimization and development: our solutions implement complex low level methods needed to access the GPU computing power on the back-end while keeping a user-friendly and easy-to-use front-end. The first product the GP-you Group proposes is a Freeware library named GPUmat which brings the GPU parallel computing power directly in MATLABĀ©. The GP-you Group offers support on developing GPU-based software on demand, as well as to customize our already existing products in function of the user requests. |
|
GPUmat
GPUmat allows standard MATLAB code to run on GPUs. The execution is transparent to the user as shown in the following example:
|
|
Read more...
|
GPUmat News - GPUmat 0.252 for CUDA 3.1 has been released.
GPUmat version 0.252 is now available for CUDA 3.1. |
GPUmat News - GPUmat 0.25 has been released.
GPUmat 0.25 has been released. More functions are now available with source code on the GPUmat User Modules project (http://sourceforge.net/projects/gpumatmodules/). Please check the Wiki documents at http://sourceforge.net/apps/mediawiki/gpumatmodules/ for more details about implemented functions. GPUmat 0.25 will be soon updated also to CUDA 3.0.
New features:
- GPUmat compiler (BETA). The compiler allows the user to record GPU operations into a new compiled function. The compiled function is faster than the native GPUmat implementation. Moreover, for-loops can be optimized using the GPUfor statement. Please check the documentation for more information and examples.
|
August 10.10 - GPUmat 0.252 for CUDA 3.1 has been released. |
|
|
|
July 18.10 - GPUmat 0.251 for CUDA 3.1 has been released. |
|
|
|
|
|
|