Performance Study of Kotlin and Java Program Considering Bytecode Instructions and JVM JIT Compiler

Abstract

Java has been widely used in the information technology field such as smartphones and server-side software. Java language and its runtime environments, which is called the Java ecosystem, have been developed and improved for some decades. This has become a huge and sophisticated ecosystem. Therefore, optimizing software codes, such as bytecodes program, should be well matched for this eco-system. Recently, a new programing language, which is called Kotlin, based on JVM (Java virtual machine) was developed. Kotlin is a well-designed program language, but is newly proposed and may not be exhaustively optimized for this Java ecosystem. In this paper, we discuss the performances of programs that are written in Java and Kotlin languages based on this ecosystem implementation. We first evaluate the performances of the loop function, which is a basic function of programming languages, of these two languages, then show that the performance of Kotlin is less than that of Java in the case of subjectively typical and common description way. Also, we show that even the performance of Java is less in some cases. We seconds discuss the reason for these performance declines comparing the bytecodes that are output from compilers of these languages. We reveal a small difference in bytecode sequences causes a large performance change due to the optimizer of JIT (Just-in-time) compiler implementation. We third discuss a method for improving the performance by solving this issue. One of the easiest ways is translating a JVM bytecode file. Finally, we evaluate our performance improving method and show that the method can improve the performance by 33% in the largest case.