FSE 2025
Mon 23 - Fri 27 June 2025 Trondheim, Norway
co-located with ISSTA 2025
Wed 25 Jun 2025 12:10 - 12:30 at Vega - Debugging Chair(s): Chao Peng

In recent years, the MLIR platform has had explosive growth due to the need of building extensible deep learning compilers and hardware accelerator compilers. Such examples include Triton, CIRCT, and ONNX-MLIR. MLIR compilers introduce significant complexities in localizing bugs or inefficiencies because of their layered optimization and transformation process with compilation passes. While existing delta debugging techniques can be used to identify a minimum subset of IR code that reproduces a given bug symptom, their naive application to MLIR is time-consuming, because real-world MLIR compilers usually involve a large number of compilation passes and compiler developers must also identify a minimized set of relevant compilation passes simultaneously, in order to reduce the footprint of MLIR compiler code to be inspected for a bug fix. We propose DuoReduce, a dual-dimensional reduction approach for MLIR bug localization. DuoReduce leverages three key ideas in tandem to design an efficient MLIR debugger. First, DuoReduce reduces the bug-irrelevant compilation passes by identifying ordering dependencies among different compilation passes. Second, DuoReduce uses MLIR-semantics aware transformations to expedite IR code reduction. Finally, DuoReduce leverages cross-dependence between the IR code dimension and the compilation pass dimension by accounting for which IR code segments are related to which compilation passes to reduce the unused passes.

Experiments with three large-scale MLIR compiler projects find that DuoReduce outperforms syntax-aware reducers such as Perses and Vulcan in terms of IR code reduction by 31.6% and 21.5% respectively. If one uses these reducers by enumerating all possible compilation passes (on average 18 passes), it could take up to 145 hours. By identifying ordering dependencies among compilation passes, DuoReduce reduces this time to 9.5 minutes. By identifying which compilation passes are unused for compiling reduced IR code, DuoReduce reduces the number of passes by 14.6%. This translates to not needing to examine 281 lines of MLIR compiler code on average to fix the bugs. DuoReduce has the potential to significantly reduce debugging effort in multi-layer extensible compilers, which serves as an important basis for the current landscape of machine learning and hardware accelerators.

Wed 25 Jun

Displayed time zone: Amsterdam, Berlin, Bern, Rome, Stockholm, Vienna change

11:00 - 12:30
11:00
20m
Talk
ChatDBG: Augmenting Debugging with Large Language Models
Research Papers
Kyla H. Levin University of Massachusetts Amherst, USA, Nicolas van Kempen University of Massachusetts Amherst, USA, Emery D. Berger University of Massachusetts Amherst and Amazon Web Services, Stephen N. Freund Williams College
DOI Pre-print
11:20
10m
Talk
Towards Adaptive Software Agents for Debugging
Ideas, Visions and Reflections
Yacine Majdoub IReSCoMath Research Lab, Faculty of Sciences, University Of Gabes, Tunisia, Eya Ben Charrada IReSCoMath Research Lab, Faculty of Sciences, University Of Gabes, Tunisia, Haifa Touati IReSCoMath Research Lab, Faculty of Sciences, University Of Gabes, Tunisia
Pre-print
11:30
20m
Talk
Empirically Evaluating the Impact of Object-Centric Breakpoints on the Debugging of Object-Oriented Programs
Research Papers
Valentin Bourcier INRIA, Pooja Rani University of Zurich, Maximilian Ignacio Willembrinck Santander Univ. Lille, Inria, CNRS, Centrale Lille, UMR 9189 CRIStAL F-59000 Lille, France, Alberto Bacchelli University of Zurich, Steven Costiou INRIA Lille
DOI
11:50
20m
Talk
An Empirical Study of Bugs in Data Visualization Libraries
Research Papers
Weiqi Lu The Hong Kong University of Science and Technology, Yongqiang Tian , Xiaohan Zhong The Hong Kong University of Science and Technology, Haoyang Ma Hong Kong University of Science and Technology, Zhenyang Xu University of Waterloo, Shing-Chi Cheung Hong Kong University of Science and Technology, Chengnian Sun University of Waterloo
DOI
12:10
20m
Talk
DuoReduce: Bug Isolation for Multi-Layer Extensible Compilation
Research Papers
Jiyuan Wang University of California at Los Angeles, Yuxin Qiu University of California at Riverside, Ben Limpanukorn University of California, Los Angeles, Hong Jin Kang University of Sydney, Qian Zhang University of California at Riverside, Miryung Kim UCLA and Amazon Web Services
DOI Pre-print

Information for Participants
Wed 25 Jun 2025 11:00 - 12:30 at Vega - Debugging Chair(s): Chao Peng
Info for room Vega:

Vega is close to the registration desk.

Facing the registration desk, its entrance is on the left, close to the hotel side entrance.

:
:
:
: