SNU College of Engineering Professor Chung-Kil Hur’s Team Adds New Command to LLVM Compiler

Even Google pays attention to the freeze commands that fix errors without compromising performance
 

• From left: Professor Chung-Kil Hur (SNU), Juneyoung Lee (Ph.D. student of SNU), LLVM Compiler

 
A team of Korean researchers developed a command that addresses LLVM errors, one of the most commonly used compilers in the world.
 
SNU College of Engineering (Dean Kookheon Char) announced on the 23rd that the Freeze command proposed by Professor Chung-Kil Hur (Software Foundations Laboratory) in the Department of Computer Science and Engineering was officially added to the LLVM in November to solve the problem of the LLVM compiler.
 
LLVM is currently used in several of the world' leading companies like Apple, Google and Facebook as well as numerous open source projects. Its representative projects include C/C++ Compiler Clang, Apple's Swift language, Google's Tensorflow Project and lastly, the programming language Rust.
 
Researchers found problems with the concept of 'undefined behavior' in the LLVM compiler intermediate representation as well as the compiler errors that can arise from it. In this regard, experiments have demonstrated that a new command called freeze can be proposed as a solution and the problem can be solved without any performance degradation.
 
The research was first recognized in academia and in 2017, was published in Programming Language Design and Implementation (PLDI), one of the top academic societies in the programming language field. Since then, the industry has also become aware of the severity of the problem and active discussion has led to the official addition of the freeze command to the LLVM in November, two years since the initial discussion.
"This research is gaining more spotlight overseas," said Professor Chung-Kil Hur of SNU. "After the publication of the paper, the problem caused LLVM compilers to incorrectly compile Google's software, which has led to such attention to the research," he explained.                
 
Under the lead of Professor Chung-Kil Hur, this research is the result of a joint collaboration with Google, Microsoft, Azul Systems and The University of Utah in the U.S. More details can be found in the following link:
 
• Page for Technical Explanation: http://sf.snu.ac.kr/freeze/ 
• Official Explanation of LLVM's method: https:// https://llvm.org/docs/LangRef.html#freeze-instruction