Inferring Needless Write Memory Accesses on Ethereum Bytecode

Abstract Efficiency is a fundamental property of any type program, but it even more so in the context programs executing on blockchain (known as smart contracts ). This because optimizing has direct consequences reducing costs deploying and contracts, there are fees to pay related their bytes-size resource consumption (called gas Optimizing memory usage considered challenging problem that, among other things, requires precise inference locations being accessed. also case for Ethereum Virtual Machine (EVM) bytecode generated by most-widely used compiler, , whose rather unconventional low-level challenges automated reasoning. paper presents static analysis, developed at level EVM that infers write accesses needless thus can be safely removed. The application our implementation than 19,000 real detected about 6,200 less 4 hours. Interestingly, many these writes were involved patterns greatly optimized removing entire blocks bytecodes. To best knowledge, existing optimization tools cannot infer such accesses, hence detect inefficiencies affect both deployment execution contracts.