Tracking at level 2 for the ATLAS high level trigger

A set of algorithms for fast pattern recognition and track reconstruction using 3D space points for the ATLAS Level 2 (LVL2) Trigger at the LHC is described. The presented algorithms work by first identifying the z position along the beamline of any interesting interaction. This allows selection of groups of spacepoints pointing back to this position using a histogramming technique that greatly reduces the combinatorial overhead and execution time before moving on to fitting the tracks and vertices. Results on the timing and performance of the tracking at LVL2 are presented using both simulated data and real test beam data and cosmic data.