# Automated Data-arithmetic Optimization for Embedded Applications _by Bonnot Justine (Quintech & IETR Vaader) - 2021.05.20 - [Zoom](https://zoom.us/j/91484087470?pwd=andNWU9HNUFaSnJxQitEdHladE0zUT09)_ ###### tags: `VAADER` `Seminar` ![](https://i.imgur.com/BxyjgDE.jpg) ## Abstract **The triptych energy consumption, development time and computing power: the main challenges of embedded electronics.** The global embedded software market is expected to reach more than $20 billion in 2021 with an average annual growth rate of 9.4% over the 2017-2021 period. This strong growth is primarily due to the growing maturity of embedded technologies. Nevertheless, there are still major challenges to optimize resources, either in terms of energy consumption, computing power or development time. **Quintech proposes a solution that goes to the core of these challenges.** The implementation of data processing oriented applications within embedded systems requires the optimization of the arithmetic used to represent the data. Therefore, conversion steps are sometimes necessary to switch from floating-point code to fixed-point or custom floating-point. The goal of this conversion is to obtain the data word-lengths (WL) composed of an integer part (Integer Word-Length, IWL) and a fractional part. However, this conversion process is complex, very time-consuming and error-prone. To overcome these challenges, the automated finite precision conversion solution proposed by Quintech aims at helping the design of embedded systems. This automated and fast solution, integrated upstream of high-level synthesis solutions, optimizes the performance of the embedded system (calculation time, energy consumption and memory footprint) with a reduction between 20 and 40% but above all, reduces the application development phase by up to 50%.