New Software Techniques in Particle Physics and Improved Track Reconstruction for the CMS Experiment

This thesis presents improvements of the processing speed of HEP applications by making better use of the vector units in modern CPUs. To cope with the increased amount of LHC data, the parallel computing power of GPUs are one option. GPU-optimized reconstruction algorithms which provide a high level of parallelism are presented. Furthermore, the decay of neutral kaons was used to show the benefits of the Geant4-based fitting method when applied to the mass reconstruction of particles.