Improving R Interpreter’s Performance Through Specialization

R, a dynamic scripting language designed for statistical computing, has grown in popularity in recent years. However, the low performance of R, due to inefficiencies in the interpretation, limits its usability. Our previous study classifies R programs into three types, including Type I (looping over data), Type II (vector programming), and Type III (glue codes). The most serious performance problems of R are mostly manifested on Type I R codes. We have proposed and implemented two approaches based on specialization to improve R’s performance for Type I R codes. Firstly, ORBIT VM, an extension of the GNU R VM, to perform aggressive object allocation removal and instruction path length reduction in the GNU R VM via profile-driven specialization techniques. Secondly, the VALOR compiler, transforming one specific R Type I code into Type II to reduce the interpretation overhead. These two approaches improved the running speedup from 3x to 7x in different contexts.