Gießen 2007 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Downloads | Hilfe

HK: Fachverband Physik der Hadronen und Kerne

HK 34: Instrumentation und Anwendungen

HK 34.7: Vortrag

Mittwoch, 14. März 2007, 15:45–16:00, B

PCI Express DMA Engine for Active Buffer Project in CBM Experiment — •Wenxue Gao, Andreas Kugel, Reinhard Männer, Holger Singpiel, and Andreas Wurz for the CBM collaboration — Lehrstuhl Informatik V, Universität Mannheim, Deutschland

Due to the specific character of the planned CBM experiment at GSI a trigger design with self-triggered detector channels has been proposed. The data of all sub-detectors is used for the first level trigger decision. Event data of about 10 Gbps per link has to be handled by a sophisticated network and buffer management, namely Active Buffer. This presentation talks about an efficient DMA engine of the Active Buffer Board unit that runs on the transaction layer of the PCI Express protocol. Along with a Xilinx PCI Express logic core (1-lane or 4-lane, 32-bit or 64-bit) working on physical layer and data link layer, the DMA engine provides general PCI Express services between the high-bandwidth network and the host computer. It uses virtual channels between receive and transmit ports and has minimum latency. Downstream (host to endpoint) and upstream (endpoint to host) DMA transactions are running independently to get a higher bidirectional throughput. Balanced scheduling policy is also practiced for fairness and against starvation. The design is implemented into Xilinx Virtex4FX FPGA and both logic verification and system test are done. Supported by EU-FP6 HADRONPHYSICS