Near-Pruned single assignment transformation of programs

Abstract

This paper introduces Near-Pruned SSA, a novel variant of the SSA form that attains precision close to the Pruned version while prioritizing its efficient generation without the need for costly data flow analysis. This is realized by leveraging variables’ usage information within the program’s augmented CFG. Furthermore, we propose a direct method for generating DSA form of programs that bypasses the traditional process of $\varphi$-node destruction into its immediate predecessor-blocks, thereby streamlining the process. Experimental evaluation on a range of Solidity programs, including real-world smart contracts deployed on the Ethereum mainnet, demonstrates that our method outperforms existing SSA variants, except for the Pruned version, by minimizing the number of introduced $\varphi$-statements compared to state-of-the-art techniques. In particular, the proposed Near-Pruned variant demonstrates a computational cost that is approximately one-third of that of the Pruned variant while achieving a nearly 92% reduction in the introduction of additional statements compared to the Semi-Pruned variant.