General Processing Units (GPUs) are proving to be very useful in the area of scientific programming for speeding up the execution of a large number of scientific codes. A new field of programming called GPU programming now exists to take full advantage of the capabilities of GPUs. For a certain type of algorithm GPUs can offer a speedup of a factor of 40 or more compared to traditional single core CPUs.
A lot of algorithms currently implemented for CPUs could profit from being ported to GPUs. This work package proposes to continue the work in this promising field by porting typical synchrotron data analysis programs to GPUs and explore the possibility to enable these programs to run efficiently on many core CPUs, the next step in parallel computing, using OpenCL.
You may access further information on WP12 ongoing activities at http://eniac.cyi.ac.cy/