HPCwire is reporting that NVIDIA has demonstrated that its CUDA GPU programming API can be coupled with high-level programming language MATLAB.
NVIDIA’s MATLAB plug-in example allows MATLAB programs to utilize standard GPU libraries for application speed ups. The plug-in is also outfitted to allow users to write their own libraries enabling them to take the performance critical piece of their code and harness the capabilities of the GPU through the NVIDIA CUDA software environment.
…”A typical MATLAB simulation of 2D isotropic turbulence at a resolution suitable for scientific publications (1024×1024) until recently would take a couple of days,” said Andy Keane, general manager of the GPU computing business unit at NVIDIA. “With the CUDA plug-ins we can perform the same simulation in 4 hours, a 12X increase, and with more optimizations, we can get this even faster.”
The MATLAB plug-in can be downloaded here.