Choose timezone
Your profile timezone:
MPP Colloquium
Event reconstruction on modern and future computer architectures
by
→
Europe/Berlin
Auditorium
Auditorium
Description
Event reconstruction is one of the most challenging problems of the data analysis in modern high energy physics experiments. This is particularly important for heavy-ion experiments like ALICE (CERN, Switzerland) and CBM (FAIR/GSI, Germany) with thousands of tracks per central collision.
The track finding is the most time consuming part of the event reconstruction. Every track finder has to solve a very complicated combinatorial optimization problem in order to find the particle trajectories. The advantage of the cellular automaton (CA) method consists in building up short track segments already before the beginning of the combinatorial search. The CA method is intrinsically local and parallel. Because of the high particle track densities, the CA track finder includes the Kalman filter (KF) algorithm for the track parameters estimation.
Future many-core CPU and GPU architectures require a fundamental redesign of the traditional approaches to data processing. A massive hardware parallelization at the level of cores, threads and vectors should be adequately reflected in a mathematical and programming optimization of the algorithms used for reconstruction and analysis. According to this principle the reconstruction algorithms were mathematically and numerically optimized, and vectorized by inline operator overloading. This approach allows using the same code on different many-core computer platforms, such as IBM Cell, AMD Opteron, Intel Nehalem, Intel Knights Ferry and Nvidia GPU. In this way the speed of the algorithms becomes increased up to several orders of magnitude. In addition, the algorithms show strong many-core scalability.
New parallel programming languages, like Intel Ct and OpenCL, being less architecture dependent, provide considerable support of the algorithms with respect to the future computer systems.