Algorithm acceleration from GPGPUs for the ATLAS upgrade

Feasibility studies into the use of General Purpose CPUs have been performed on two key algorithms in the ATLAS High Level Trigger. A GPGPU based version of the Z-finder vertex finding algorithm resulted in over 35 times speed-up over serial CPU execution in the best case scenario, whilst a speed-up of over 5 times was observed from a GPGPU-based Kalman filter track finder routine. The approaches taken for converting these algorithms for execution on GPU devices is described.