Static Performance Analysis of Faust Programs Using Dataflow Modelling

The processing capabilities of modern System-on-Chips allow artists to push the boundaries of digital art by enabling the implementation of complex real-time signal processing applications. The Faust framework has simplified the process of implementing signal processing applications for these systems. With the Faust language, it is possible to generate implementations for a wide variety of target contexts, allowing users to take a "write once, implement everywhere" approach. Nonetheless, this variety of hardware also comes with the issue of determining what kind of processing capability is required for a given project.

Evaluation tools of Faust currently rely on runtime analysis to measure the performance of implementations on target hardware; this method can be time consuming and requires the user to already own the hardware they wish to evaluate.

Instead, we attempt to address this problem by considering static performance analysis. We propose to use a dataflow model of computation to represent Faust programs, and to predict performance for specific target hardware. This approach could provide useful heuristics in the future for users to decide which hardware is needed to reach their required performance.