FOGA: Flag Optimization with Genetic Algorithm

TL;DR

This paper presents FOGA, a genetic algorithm-based autotuning method for optimizing compiler flags in embedded systems, achieving significant speedups over existing frameworks like OpenTuner.

Contribution

FOGA introduces a novel GA-based autotuning approach with hyperparameter tuning and a maximum iteration stop criterion for efficient compiler flag optimization.

Findings

Achieved notable execution time speedups in C++ programs.

Outperformed the state-of-the-art OpenTuner framework.

Demonstrated effectiveness in resource-constrained embedded systems.

Abstract

Recently, program autotuning has become very popular especially in embedded systems, when we have limited resources such as computing power and memory where these systems run generally time-critical applications. Compiler optimization space gradually expands with the renewed compiler options and inclusion of new architectures. These advancements bring autotuning even more important position. In this paper, we introduced Flag Optimization with Genetic Algorithm (FOGA) as an autotuning solution for GCC flag optimization. FOGA has two main advantages over the other autotuning approaches: the first one is the hyperparameter tuning of the genetic algorithm (GA), the second one is the maximum iteration parameter to stop when no further improvement occurs. We demonstrated remarkable speedup in the execution time of C++ source codes with the help of optimization flags provided by FOGA when compared to the state of the art framework OpenTuner.