Algorithmic Profiling for Real-World Complexity Problems
Tue 10 May 2022 22:00 - 22:05 at ICSE room 5-even hours - Software Testing 6 Chair(s): Leonardo Sousa
Complexity problems are a common type of performance issues, caused by algorithmic inefficiency. Algorithmic profiling aims to automatically attribute execution complexity to an executed code construct. It can identify code constructs in superlinear complexity to facilitate performance optimizations and debugging. However, existing algorithmic profiling techniques suffer from several severe limitations, missing the opportunity to be deployed in production environment and failing to effectively pinpoint root causes for performance failures caused by complexity problems. In this paper, we design a tool, ComAir, which can effectively conduct algorithmic profiling in production environment. We propose several novel instrumentation methods to significantly lower runtime overhead and enable the production-run usage. We also design an effective ranking mechanism to help developers identify root causes of performance failures due to complexity problems. Our experimental results show that ComAir can effectively identify root causes and generate accurate profiling results in production environment, while incurring a negligible runtime overhead.
Tue 10 MayDisplayed time zone: Eastern Time (US & Canada) change
22:00 - 23:00 | Software Testing 6SEIP - Software Engineering in Practice / Technical Track / Journal-First Papers at ICSE room 5-even hours Chair(s): Leonardo Sousa | ||
22:00 5mTalk | Algorithmic Profiling for Real-World Complexity Problems Journal-First Papers Boqin Qin China Telecom Cloud Computing Corporation, Tengfei Tu Beijing University of Posts and Telecommunications, Ziheng Liu University of California, San Diego, Tingting Yu University of Cincinnati, Linhai Song Pennsylvania State University, USA DOI Pre-print Media Attached | ||
22:05 5mTalk | To What Extent Do DNN-based Image Classification Models Make Unreliable Inferences? Journal-First Papers Yongqiang TIAN The Hong Kong University of Science and Technology; University of Waterloo, Shiqing Ma Rutgers University, Ming Wen Huazhong University of Science and Technology, Yepang Liu Southern University of Science and Technology, Shing-Chi Cheung Hong Kong University of Science and Technology, Xiangyu Zhang Purdue University DOI Pre-print Media Attached | ||
22:10 5mTalk | Testing Machine Learning Systems in Industry: An Empirical Study SEIP - Software Engineering in Practice Shuyue Li Xi'an Jiaotong University, Jiaqi Guo Xi'an Jiaotong University, Jian-Guang Lou Microsoft Research, Ming Fan Xi'an Jiaotong University, Ting Liu Xi'an Jiaotong University, Dongmei Zhang Microsoft Research DOI Pre-print Media Attached | ||
22:15 5mTalk | R2Z2: Detecting Rendering Regressions in Web Browsers through Differential Fuzz Testing Technical Track Suhwan Song Seoul National University, South Korea, Jaewon Hur Seoul National University, Sunwoo Kim Samsung Research, Samsung Electronics, Philip Rogers Google, Byoungyoung Lee Seoul National University, South Korea Pre-print Media Attached | ||
22:20 5mTalk | Fuzzing Class Specifications Technical Track Facundo Molina University of Rio Cuarto and CONICET, Argentina, Marcelo d'Amorim Federal University of Pernambuco, Nazareno Aguirre University of Rio Cuarto and CONICET, Argentina Pre-print Media Attached | ||
22:25 5mTalk | GIFdroid: Automated Replay of Visual Bug Reports for Android Apps Technical Track DOI Pre-print Media Attached | ||