Parallel Computing on GPUs
Another research line performed by the EDNA group is the acceleration of algorithms by their execution on the many-core architecture presented in the Graphics Processing Units (GPUs). Taking advantage of these platforms, it is possible to reduce up to 2 orders of magnitude the execution time.
One direct application of this technique is the simulation of the chemical wet etching process oriented to Micro-Electro-Mechanical-Systems (MEMS) design. Two different methods have been implemented and accelerated with GPUs: the Cellular Automata and the Level Set (LS) method. A commercial software has been licensed with the cellular automata-based algorithm and it is available on http://www.intellisense.com/modules/IntelliEtch.html.
The LS simulator is currently on development progress, but promising results have been already obtained.
Another application in which we are working on, consists in accelerating the image reconstruction process from the PET setup experimental data.