SUPERSONIC: Learning to Generate Source Code Optimizations in C/C++
This program is tentative and subject to change.
Software optimization refines programs for resource efficiency while preserving functionality. Traditionally, it is a process done by developers and compilers. This paper introduces a third option, automated optimization at the source code level. We present Supersonic , a neural approach targeting minor source code modifications for optimization. Using a seq2seq model, Supersonic is trained on C/C++ program pairs ( xt , xt+1 ), where xt+1 is an optimized version of xt , and outputs a diff. Supersonic ’s performance is benchmarked against OpenAI’s GPT-3.5-Turbo and GPT-4 on competitive programming tasks. The experiments show that Supersonic not only outperforms both models on the code optimization task but also minimizes the extent of the change with a model more than 600x smaller than GPT-3.5-Turbo and 3700x smaller than GPT-4.
This program is tentative and subject to change.
Thu 1 MayDisplayed time zone: Eastern Time (US & Canada) change
14:00 - 15:30 | |||
14:00 15mTalk | Boosting Path-Sensitive Value Flow Analysis via Removal of Redundant Summaries Research Track Yongchao WANG Hong Kong University of Science and Technology, Yuandao Cai Hong Kong University of Science and Technology, Charles Zhang Hong Kong University of Science and Technology | ||
14:15 15mTalk | Dockerfile Flakiness: Characterization and Repair Research Track Taha Shabani University of British Columbia, Noor Nashid University of British Columbia, Parsa Alian University of British Columbia, Ali Mesbah University of British Columbia | ||
14:30 15mTalk | Evaluating Garbage Collection Performance Across Managed Language Runtimes Research Track Yicheng Wang Institute of Software Chinese Academy of Sciences, Wensheng Dou Institute of Software Chinese Academy of Sciences, Yu Liang Institute of Software Chinese Academy of Sciences, Yi Wang Institute of Software Chinese Academy of Sciences, Wei Wang Institute of Software at Chinese Academy of Sciences, Jun Wei Institute of Software at Chinese Academy of Sciences; University of Chinese Academy of Sciences, Tao Huang Institute of Software Chinese Academy of Sciences | ||
14:45 15mTalk | Module-Aware Context Sensitive Pointer Analysis Research Track Haofeng Li Institute of Computing Technology at Chinese Academy of Sciences, Chenghang Shi SKLP, Institute of Computing Technology, CAS, Jie Lu SKLP, Institute of Computing Technology, CAS, Lian Li Institute of Computing Technology at Chinese Academy of Sciences; University of Chinese Academy of Sciences, Zixuan Zhao Huawei Technologies Co. Ltd | ||
15:00 15mTalk | SUPERSONIC: Learning to Generate Source Code Optimizations in C/C++ Journal-first Papers Zimin Chen KTH Royal Institute of Technology, Sen Fang North Carolina State University, Martin Monperrus KTH Royal Institute of Technology | ||
15:15 15mTalk | T-Rec: Fine-Grained Language-Agnostic Program Reduction Guided by Lexical Syntax Journal-first Papers Zhenyang Xu University of Waterloo, Yongqiang Tian Hong Kong University of Science and Technology, Mengxiao Zhang , Jiarui Zhang University of Waterloo, Puzhuo Liu Beijing Key Laboratory of IOT Information Security Technology, Institute of Information Engineering, CAS, Beijing, China； School of Cyber Security, University of Chinese Academy of Sciences, Beijing, China；, Yu Jiang Tsinghua University, Chengnian Sun University of Waterloo | ||