Bilkent University
Department of Computer Engineering


Feature Extraction for Code Optimization


Melih Peker
MS Student
(Supervisor: Asst. Prof. Eray Tüzün)
Computer Engineering Department
Bilkent University

Abstract: In order to increase the quality and efficiency of a code, optimization methods can be used during compilation. GCC compiler has more than 250 optimization methods for generating more efficient code. Those optimizations are divided into levels by GCC and while compiling the code, the degree of this optimization can be specified with the -O flag. The goal of our research is to develop a machine learning model for choosing right optimization for any code, rather than specifying levels. The model will take the code as an input and will output suggested optimization flags. In order to train the model, we need to extract features from the code. For that, we will use a machine code analyzer LLVM-MCA. Using this performance analysis tool, we can extract features and have information about both input code and optimized code as well. In this presentation, I will discuss the feature extraction process.


DATE: 22 November 2021, Monday @ 15:30 Zoom