Thirimadura Charith Mendis

Graduate Student

Charith Mendis is a PhD candidate advised by Prof. Saman Amarasinghe. His research interests include Compilers, Program Analysis and Machine Learning. He was awarded the William A. Martin Thesis Prize for best Masters thesis in Computer Science at MIT, best student paper award at IEEE Big Data conference and a best paper award at ML for Systems workshop at ISCA. He has published work at both top programming language venues such as PLDI and OOPSLA as well as at top machine learning venues such as ICML and NeurIPS.



Program Performance Prediction with Deep Learning

Predicting the number of clock cycles a processor takes to execute a block of assembly instructions in steady-state (the throughput) is important for both compiler designers and performance engineers.

However, building an analytical model to do so is especially complicated in modern x86-64 Complex Instruction Set Computer (CISC) machines with sophisticated processor microarchitectures in that it is tedious, error-prone, and must be performed from scratch for each processor generation.

Ithemal is the first tool that learns to predict the throughput of a set of instructions. It does so more accurately than state-of-the-art hand-written tools currently used in compiler backends and static machine code analyzers. In particular, Ithemal has less than half the error of state-of-the-art analytical models (LLVM's llvm-mca and Intel's IACA).