Write a Blog >>
ESEC/FSE 2022
Mon 14 - Fri 18 November 2022 Singapore
Wed 16 Nov 2022 14:45 - 15:00 at SRC LT 53 - Program Repair/Synthesis Chair(s): Saikat Chakraborty

Improving software performance is an important yet challenging part of the software development cycle. Today, the majority of performance inefficiencies are identified and patched by performance experts. Recent advancements in deep learning approaches and the wide-spread availability of open-source data creates a great opportunity to automate the identification and patching of performance problems. In this paper, we present DeepDev-PERF, a transformer-based approach to suggest performance improvements for C# applications. We pretrain DeepDev-PERF on English and Source code corpora, followed by finetuning for the task of generating performance improvement patches for C# applications. Our evaluation shows that our model can generate the same performance improvement suggestion as the developer fix in ~53% of the cases, getting $\sim$34% of them verbatim in our expert-verified dataset of performance changes made by C# developers. Additionally, we evaluate DeepDev-PERF on 50 open-source C# repositories on GitHub using both benchmark and unit tests and find that our model is able to suggest valid performance improvements that can improve both CPU usage and Memory allocations. So far we've submitted 19 pull-requests with 28 different performance optimizations and 11 of these PRs have been approved by the project owners.

Wed 16 Nov

Displayed time zone: Beijing, Chongqing, Hong Kong, Urumqi change

14:00 - 15:30
Program Repair/SynthesisResearch Papers / Industry Paper at SRC LT 53
Chair(s): Saikat Chakraborty Microsoft Research
14:00
15m
Talk
PyTER: Effective Program Repair for Python Type Errors
Research Papers
Wonseok Oh Korea University, Hakjoo Oh Korea University
DOI
14:15
15m
Talk
VulRepair: A T5-Based Automated Software Vulnerability Repair
Research Papers
Micheal Fu Monash University, Chakkrit Tantithamthavorn Monash University, Trung Le Monash University, Australia, Van Nguyen Monash University, Australia, Dinh Phung Monash University, Australia
DOI
14:30
15m
Talk
An Empirical Study of Deep Transfer Learning-Based Program Repair for Kotlin Projects
Industry Paper
Misoo Kim Sungkyunkwan University, Youngkyoung Kim Sungkyunkwan University, Hohyeon Jeong Sungkyunkwan University, Jinseok Heo Sungkyunkwan University, Sungoh Kim Samsung Electronics, Hyunhee Chung Samsung Electronics, Eunseok Lee Sungkyunkwan University
DOI
14:45
15m
Talk
DeepDev-PERF: A Deep Learning-Based Approach for Improving Software Performance
Research Papers
Spandan Garg Microsoft, Roshanak Zilouchian Moghaddam Microsoft, Colin Clement Microsoft, Neel Sundaresan Microsoft, Chen Wu Microsoft
DOI
15:00
15m
Talk
Less Training, More Repairing Please: Revisiting Automated Program Repair via Zero-Shot Learning
Research Papers
Chunqiu Steven Xia University of Illinois at Urbana-Champaign, Lingming Zhang University of Illinois at Urbana-Champaign
DOI