bracevac/sai

Sai is a research project that makes a specific kind of software testing tool much faster. When developers write code, they use tools called "static analyzers" to automatically scan the program and find bugs or security issues before the code ever runs. The problem is that these scanners can be painfully slow on large codebases. This project speeds up that scanning process dramatically. It achieves this speed using a technique called "multi-stage programming." Instead of running the analysis step-by-step on a program, the tool essentially writes a customized, highly optimized piece of code specifically designed to analyze that exact program. Think of it like tailoring a suit rather than buying one off the rack. By generating a specialized tool for each piece of software it examines, the analysis avoids a lot of generic overhead and runs much faster. The primary audience for this project is programming language researchers and academics. The repository contains the code and artifacts for several published research papers, including submissions to conferences like OOPSLA and PEPM. A programming language researcher might use this code to study advanced compilation techniques, build upon the published findings, or explore new ways to optimize how software is tested and verified. It is not an off-the-shelf application meant for everyday developers to download and run. What is notable about the project is its heavy reliance on meta-programming, essentially writing code that writes other code. The README does not go into deep detail about the exact setup or how to configure the current development version, but it lists dependencies on specialized compiler libraries. This highlights that the project is a proof-of-concept exploring a cutting-edge tradeoff: investing extra effort upfront to generate a custom analyzer in order to save massive amounts of time during the actual scanning process.