Samenvatting
Current multi-functional embedded systems such as smartphones and tablets support multiple 2G/3G/4G radio standards including Long Term Evolution (LTE) and LTE-Advanced. LTE-Advanced is the latest industry standard that improves upon LTE by introducing several feature-rich and complex improvements. Moreover, both LTE and LTE-Advanced have real-time requirements. LTE and LTE-Advanced receivers are typically scheduled on a heterogeneous multi-core processor to satisfy real-time and low-power requirements.
Manual simulation-based real-time analysis of such applications is infeasible. Dataflow can be used for real-time analysis. Static dataflow allows a rich set of analysis techniques to support real-time scheduling, but is too restrictive to accurately and conveniently model the dynamic data-dependent behavior for many practical applications, including LTE-Advanced. Dynamic dataflow allows modeling of such applications, but in general does not support rigorous real-time analysis.
Mode-controlled Dataflow (MCDF) is a restricted form of dynamic dataflow that captures realistic features in such applications and allows rigorous timing analysis. We stepwise refine and develop complete and fine-grained MCDF models of an LTE-Advanced receiver that include two key features: 1) carrier aggregation and 2) enhanced physical downlink control channel processing. We schedule the MCDF models on an industrial platform to benchmark them against (static) Single-rate Dataflow (SRDF) using existing buffer allocation techniques to demonstrate that these models are analyzable and practically applicable. Moreover, we also develop latency analysis techniques for single-rate and mode-controlled dataflow. For our LTE-Advanced receiver, relative to SRDF models, MCDF models offer 1) up to 15% smaller memory consumption, and 2) up to 1.6% smaller LTE-Advanced sub-frame processing latency.
Manual simulation-based real-time analysis of such applications is infeasible. Dataflow can be used for real-time analysis. Static dataflow allows a rich set of analysis techniques to support real-time scheduling, but is too restrictive to accurately and conveniently model the dynamic data-dependent behavior for many practical applications, including LTE-Advanced. Dynamic dataflow allows modeling of such applications, but in general does not support rigorous real-time analysis.
Mode-controlled Dataflow (MCDF) is a restricted form of dynamic dataflow that captures realistic features in such applications and allows rigorous timing analysis. We stepwise refine and develop complete and fine-grained MCDF models of an LTE-Advanced receiver that include two key features: 1) carrier aggregation and 2) enhanced physical downlink control channel processing. We schedule the MCDF models on an industrial platform to benchmark them against (static) Single-rate Dataflow (SRDF) using existing buffer allocation techniques to demonstrate that these models are analyzable and practically applicable. Moreover, we also develop latency analysis techniques for single-rate and mode-controlled dataflow. For our LTE-Advanced receiver, relative to SRDF models, MCDF models offer 1) up to 15% smaller memory consumption, and 2) up to 1.6% smaller LTE-Advanced sub-frame processing latency.
Originele taal-2 | Engels |
---|---|
Pagina's (van-tot) | 216-230 |
Aantal pagina's | 15 |
Tijdschrift | Microprocessors and Microsystems |
Volume | 47 |
Nummer van het tijdschrift | Part A |
DOI's | |
Status | Gepubliceerd - nov. 2016 |