High Energy Physics code has been known for making limited use of high performance computing architectures. Efforts in optimising HEP code on vector and RISC architectures have yield limited results and recent studies have shown that, on modern architectures, it achieves a performance between 10% and 50% of the theoretical one. Although several successful attempts have been made to port selected codes on GPUs, no major HEP code suite has a "High Performance” implementation for accelerators. With LHC undergoing a major upgrade and a number of challenging experiments on the drawing board, HEP has to try making the best usage of the new hardware. This activity is one of the foci of the SFT group at CERN, which hosts, among others, the Root and Geant4 projects. One important initiative is centred on the development of a high-performance prototype for particle transport, the GeantV project. A good concurrency level has been achieved by Geant4 with the parallelisation at event level. However, apart the sharing of data structures, this does not increase the number of events per second produced on a loaded system, which requires an efficient use of the low level instruction parallelism and pipelining of modern processors. In GeantV project we have implemented a framework that allows scheduling vectors of particles to an arbitrary number of computing resources in a fine grain parallel approach. Via template code specialisation we are able to recast simple computational kernels do different architectures, minimising code duplication. This approach is already providing very promising results in the treatment of the detector geometry and we plan to extend it soon to the optimisation of the physics code. The talk will review the current satis and the perspectives of the development of a simulation framework able to profit best from the recent technology evolution in computing.
