high performance computing on graphics processing units: hgpu.org

In recent years the need to solve complex problems that require large computing resources in shorter time has especially arisen. Some of these in the scientific field are: weather forecast, seismic simulations, chemical reactions simulation and studies on the human genoma [1]. All of them belong to the "Grand Challenge Problems" set. As can be noted, solving these problems within strict time limits is very interesting for many areas of scientific research. Other relevant problems exist in the computer science and business fields: database management and data-mining [2], search engine development on the web [3], medical diagnostic, advanced graphic [4], virtual reality [5], network broadcasting, highly complex information security [6] and data confidentiality [7]. In the past, the solution of these problems used to take large amounts of time, due to the need of leveraging very large computing capacity to solve them. Now, thanks to the development of parallel computing, larger and cheaper resources are available, that can be used to address such problems. Parallel computing potentially allows: 1) to solve larger problems than before; 2) to improve computing time; 3) to reduce costs (by splitting work among multiple cheaper processors). Despite the hopes placed in parallel computing, some problems, requiring super-polinomial algorithms, are still far from being solved in acceptable time. However parallelization can help reduce computing time in order to speed up results availability.