127,165 compiled Rust binaries from 77,004 builds of 10,766 permissively
licensed GitHub repositories — each paired with DWARF-derived function and line
metadata, the source tree it was compiled from, and, for a subset, compiler IR.
Every repository is built across a matrix of codegen backends, build modes and
optimization levels, so the same source appears many times under different
compilation settings. That is the point: it supports studying how compilation… See the full description on the dataset page:
https://huggingface.co/datasets/changliu8541/assemblage-rust.
