{"ddc":["000"],"publication_status":"accepted","page":"2","file":[{"file_name":"witschen20_axc.pdf","file_id":"20749","success":1,"file_size":250870,"creator":"witschen","date_created":"2020-12-15T15:11:06Z","date_updated":"2020-12-15T15:11:06Z","relation":"main_file","access_level":"closed","content_type":"application/pdf"}],"has_accepted_license":"1","project":[{"_id":"12","name":"SFB 901 - Subproject B4"},{"_id":"3","name":"SFB 901 - Project Area B"},{"name":"SFB 901","_id":"1"}],"abstract":[{"lang":"eng","text":"On the circuit level, the design paradigm Approximate Computing seeks to trade off computational accuracy against a target metric, e.g., energy consumption. This trade-off is possible for many applications due to their inherent resiliency against inaccuracies.\r\nIn the past, several automated approximation frameworks have been presented, which either utilize designated approximation techniques or libraries to replace approximable circuit parts with inaccurate versions. The frameworks invoke a search algorithm to iteratively explore the search space of performance degraded circuits, and validate their quality individually. \r\nIn this paper, we propose to reverse this procedure. Rather than exploring the search space, we delineate the approximate parts of the search space which are guaranteed to lead to valid approximate circuits. Our methodology is supported by formal verification and independent of approximation techniques. Eventually, the user is provided with quality bounds of the individual approximable circuit parts. Consequently, our approach guarantees that any approximate circuit which implements these parts within the determined quality constraints satisfies the global quality constraints, superseding a subsequent quality verification.\r\nIn our experimental results, we present the runtimes of our approach."}],"file_date_updated":"2020-12-15T15:11:06Z","date_created":"2020-12-15T15:13:49Z","language":[{"iso":"eng"}],"type":"preprint","date_updated":"2022-01-06T06:54:35Z","publication":"Fifth Workshop on Approximate Computing (AxC 2020)","year":"2020","status":"public","author":[{"first_name":"Linus Matthias","last_name":"Witschen","id":"49051","full_name":"Witschen, Linus Matthias"},{"full_name":"Wiersema, Tobias","last_name":"Wiersema","id":"3118","first_name":"Tobias"},{"last_name":"Platzner","id":"398","full_name":"Platzner, Marco","first_name":"Marco"}],"title":"Search Space Characterization for AxC Synthesis","_id":"20748","department":[{"_id":"78"}],"user_id":"3118","citation":{"apa":"Witschen, L. M., Wiersema, T., &#38; Platzner, M. (n.d.). Search Space Characterization for AxC Synthesis. <i>Fifth Workshop on Approximate Computing (AxC 2020)</i>.","mla":"Witschen, Linus Matthias, et al. “Search Space Characterization for AxC Synthesis.” <i>Fifth Workshop on Approximate Computing (AxC 2020)</i>.","ama":"Witschen LM, Wiersema T, Platzner M. Search Space Characterization for AxC Synthesis. <i>Fifth Workshop on Approximate Computing (AxC 2020)</i>.","ieee":"L. M. Witschen, T. Wiersema, and M. Platzner, “Search Space Characterization for AxC Synthesis,” <i>Fifth Workshop on Approximate Computing (AxC 2020)</i>. .","short":"L.M. Witschen, T. Wiersema, M. Platzner, Fifth Workshop on Approximate Computing (AxC 2020) (n.d.).","chicago":"Witschen, Linus Matthias, Tobias Wiersema, and Marco Platzner. “Search Space Characterization for AxC Synthesis.” <i>Fifth Workshop on Approximate Computing (AxC 2020)</i>, n.d.","bibtex":"@article{Witschen_Wiersema_Platzner, title={Search Space Characterization for AxC Synthesis}, journal={Fifth Workshop on Approximate Computing (AxC 2020)}, author={Witschen, Linus Matthias and Wiersema, Tobias and Platzner, Marco} }"}}