Software Parser and Analyser for Hardware Performance Estimations
Priit Ruberg, Erki Meinberg, Peeter Ellervee · 2022 International Conference on Electrical, Computer and Energy Technologies (ICECET) · 2022
In this paper, we conclude the work on a novel energy consumption and performance estimation methodology as we complete our toolchain with the development of a parser and an analyser. An embedded software estimation model is created by physical measurements and benchmarking of a hardware platform. To obtain the physical measurements, a specialised semiautomated process has been developed. Although, the model based estimations show promising results as the estimation error is low the estimation process for an arbitrary software has since been a manual labour. Therefore in this work we present a software parser and an analyser for a C-language source code that is also able to preprocess system level as well as custom libraries. Although some C-language parsers are freely available, our estimation methodology requires a more custom solution, as specific data must be obtained by the parser from the abstract syntax tree (AST). In addition to the parser, an analyser is presented in this work that is able to fuse data from both the parser and the software profiler in order to present the number of different atomic operations and the number of repetitions for each atomic operation in the software. Additionally, both the parser and the analyser could be used as standalone products for software assessment. As a result, the construction of the estimation toolchain is complete.