Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DESIL: Detecting Silent Bugs in MLIR Compiler Infrastructure

Chenyao Suo, Jianrong Wang, Yongjia Wang, Jiajun Jiang, QingChao Shen, Junjie Chen

TL;DR

DESIL tackles silent bugs in the MLIR compiler by (i) eliminating MLIR-specific undefined behaviors to produce UB-free programs, and (ii) optimizing the lowering path to generate executable forms, enabling reliable differential testing across diverse optimization sequences. By integrating these components with a checksum-based oracle, DESIL detects inconsistencies that reveal silent bugs beyond the reach of prior fuzzing approaches. Evaluated on recent MLIR revisions, DESIL uncovers dozens of silent and crash bugs, with several confirmed or fixed by developers, and demonstrates clear advantages over lifted baselines. The approach is modular, extensible to additional dialects, and has practical impact for improving reliability of MLIR-based tooling and downstream compilers.

Abstract

MLIR (Multi-Level Intermediate Representation) compiler infrastructure provides an efficient framework for introducing a new abstraction level for programming languages and domain-specific languages. It has attracted widespread attention in recent years and has been applied in various domains, such as deep learning compiler construction. Recently, several MLIR compiler fuzzing techniques, such as MLIRSmith and MLIRod, have been proposed. However, none of them can detect silent bugs, i.e., bugs that incorrectly optimize code silently. The difficulty in detecting silent bugs arises from two main aspects: (1) UB-Free Program Generation: Ensures the generated programs are free from undefined behaviors to suit the non-UB assumptions required by compiler optimizations. (2) Lowering Support: Converts the given MLIR program into an executable form, enabling execution result comparisons, and selects a suitable lowering path for the program to reduce redundant lowering pass and improve the efficiency of fuzzing. To address the above issues, we propose DESIL. DESIL enables silent bug detection by defining a set of UB-elimination rules based on the MLIR documentation and applying them to input programs to produce UB-free MLIR programs. To convert dialects in MLIR program into the executable form, DESIL designs a lowering path optimization strategy to convert the dialects in given MLIR program into executable form. Furthermore, DESIL incorporates the differential testing for silent bug detection. To achieve this, it introduces an operation-aware optimization recommendation strategy into the compilation process to generate diverse executable files. We applied DESIL to the latest revisions of the MLIR compiler infrastructure. It detected 23 silent bugs and 19 crash bugs, of which 12/14 have been confirmed or fixed

DESIL: Detecting Silent Bugs in MLIR Compiler Infrastructure

TL;DR

DESIL tackles silent bugs in the MLIR compiler by (i) eliminating MLIR-specific undefined behaviors to produce UB-free programs, and (ii) optimizing the lowering path to generate executable forms, enabling reliable differential testing across diverse optimization sequences. By integrating these components with a checksum-based oracle, DESIL detects inconsistencies that reveal silent bugs beyond the reach of prior fuzzing approaches. Evaluated on recent MLIR revisions, DESIL uncovers dozens of silent and crash bugs, with several confirmed or fixed by developers, and demonstrates clear advantages over lifted baselines. The approach is modular, extensible to additional dialects, and has practical impact for improving reliability of MLIR-based tooling and downstream compilers.

Abstract

MLIR (Multi-Level Intermediate Representation) compiler infrastructure provides an efficient framework for introducing a new abstraction level for programming languages and domain-specific languages. It has attracted widespread attention in recent years and has been applied in various domains, such as deep learning compiler construction. Recently, several MLIR compiler fuzzing techniques, such as MLIRSmith and MLIRod, have been proposed. However, none of them can detect silent bugs, i.e., bugs that incorrectly optimize code silently. The difficulty in detecting silent bugs arises from two main aspects: (1) UB-Free Program Generation: Ensures the generated programs are free from undefined behaviors to suit the non-UB assumptions required by compiler optimizations. (2) Lowering Support: Converts the given MLIR program into an executable form, enabling execution result comparisons, and selects a suitable lowering path for the program to reduce redundant lowering pass and improve the efficiency of fuzzing. To address the above issues, we propose DESIL. DESIL enables silent bug detection by defining a set of UB-elimination rules based on the MLIR documentation and applying them to input programs to produce UB-free MLIR programs. To convert dialects in MLIR program into the executable form, DESIL designs a lowering path optimization strategy to convert the dialects in given MLIR program into executable form. Furthermore, DESIL incorporates the differential testing for silent bug detection. To achieve this, it introduces an operation-aware optimization recommendation strategy into the compilation process to generate diverse executable files. We applied DESIL to the latest revisions of the MLIR compiler infrastructure. It detected 23 silent bugs and 19 crash bugs, of which 12/14 have been confirmed or fixed

Paper Structure

This paper contains 26 sections, 7 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Background and Motivation
  3. Terminology
  4. A Motivating Example
  5. Approach
  6. Undefined Behavior Elimination
  7. Undefined Behaviors from Shape Inconsistency
  8. Undefined Behaviors from Index Out-of-Bounds
  9. Undefined Behaviors from Invalid Memory References.
  10. Undefined Behaviors from Scalar Calculations.
  11. UB-Irrelevant Fix for Normal Compilation.
  12. Lowering Path Optimization
  13. Differential Testing
  14. Bug Detection by DESIL
  15. Evaluation
  16. ...and 11 more sections

Figures (7)

  • Figure 1: Motivating example.
  • Figure 2: Overview of DESIL
  • Figure 3: The lowering process for the dialect operation affine.for by applying passes $\{p_1,p_2,p_3\}$.
  • Figure 4: previously unknown bug examples detected by DESIL.
  • Figure 5: The number of unique bugs detected by DESIL and lifted existing techniques in bug detection.
  • ...and 2 more figures