E-Stitch: Top-Down Library Learning with E-Graphs (POPL 2027 Artifact)

This resource gives you code and instructions to use and test E‑Stitch, a library‑learning system that works from the top down and uses equivalence graphs (e‑graphs) to find reusable subroutines. In program synthesis and compression, library learning looks for repeated patterns in the input programs and builds a small library of helper functions. Older tools such as Stitch and Babble work with concrete syntax trees or with bottom‑up e‑graph representations; these approaches can run into scalability problems or limited search spaces. E‑Stitch adds a top‑down search that works together with e‑graphs, making the discovery of subroutines more efficient. The supplied artifact compares E‑Stitch with current best tools, including Babble and

Stitch, on several benchmark sets: the Babble dataset, molecular‑structure problems, and EPFL logic‑circuit examples. Our documentation helps B.Tech students install the Docker‑based evaluation setup, run the benchmarks, and look at the compression ratios and run times. Since the full evaluation needs a lot of computing power (at least 20 GiB of RAM), we also show how to reduce the benchmark size so you can test it on a typical computer.

Message: Hi FE, I need help with “E-Stitch: Top-Down Library Learning with E-Graphs (POPL 2027 Artifact)” in “Computer Science & Engineering”