Efficient implementations from libraries: Analyzing the influence of configuration parameters on key performance properties

Library based waveform (WF) development approaches have the potential to address one of the key challenges in software defined radio technology, developing portable and at the same time implementation-efficient WFs. The term WF, in this paper, represents a complete wireless standard with several modes. Upon standardizing the library and the interfaces of its components, different vendors can provide Flavors, which are efficient implementations, for some/all components of a WF as a board-support-package. Flavors might provide different trade-offs with respect to key performance related properties like bit error rate, throughput, latency, area, etc. This paper analyzes such a scenario by using FFT as an example. We use 10 Flavors of FFT from 2 vendors for 3 processing elements to analyze the influence of configuration parameters like input data-width, scaling, etc. on the performance properties. Analysis has been performed by doing two tests: using a simple OFDM transceiver system, calculating root mean square error. Based on our analysis, we have identified the main parameters that influence the properties. Our investigations stress the need for modeling the differences in performance on key properties in terms of the influencing parameters. Such a model can assist in selecting a set of appropriate Flavors for implementing a WF and enable tool assisted WF-development. Furthermore, it is evident from our investigations that Flavors exhibit trade-offs in properties and pose tough constraints in several aspects.