We describe the base-line design and implementation of the Data Flow and High Level Trigger (HLT) part of the ATLAS Trigger and Data Acquisition (TDAQ) system. We then discuss improvements and generalization of the system design to allow the handling of events in parallel data streams and we present the possibility for event duplication, partial Event Building and data stripping. We then present tests on the deployment and integration of the TDAQ infrastructure and algorithms at the TDAQ �pre-series” cluster (~10% of full ATLAS TDAQ). Finally, we tackle two HLT performance issues.