High-Productivity Compiler Development¶

Building a compiler should not require years of expertise. xDSL is a Python-native compiler framework that shares its IR with MLIR, so that compilers can be prototyped in Python and moved to MLIR when they mature. With IRDL, dialects are defined declaratively and registered at runtime; with Tamagoyaki, e-graphs become a first-class compiler abstraction. xDSL is also used to teach compilers to undergraduate students.

xdsl.dev GitHub

Key papers¶

• CGO 2027E-Graphs as a Persistent Compiler AbstractionJules Merckx, Alexandre Lopoukhine, Samuel Coward, Bjorn De Sutter, Jianyi Cheng, Tobias Grosser
• CGO 2025xDSL: Sidekick Compilation for SSA-Based CompilersMathieu Fehr, Michel Weber, Christian Ulmann, Alexandre Lopoukhine, Martin Paul Lücke, Théo Degioanni, Christos Vasiladiotis, Michel Steuwer, Tobias Grosser
• PLDI 2025First-Class Verification Dialects for MLIRMathieu Fehr, Yuyou Fan, Hugo Pompougnac, John Regehr, Tobias Grosser
• CGO 2025A Multi-level Compiler Backend for Accelerated Micro-kernels Targeting RISC-V ISA ExtensionsAlexandre Lopoukhine, Federico Ficarelli, Christos Vasiladiotis, Anton Lydike, Josse Van Delm, Alban Dutilleul, Luca Benini, Marian Verhelst, Tobias Grosser
• PLDI 2022IRDL: An IR Definition Language for SSA CompilersMathieu Fehr, Jeff Niu, River Riddle, Mehdi Amini, Zhendong Su, Tobias Grosser

People¶

Jules Merckx, Mathieu Fehr, Sasha Lopoukhine
Former members: Anton Lydike, Chris Vasiladiotis, Christian Ulmann, Hugo Pompougnac, Michel Weber, Theo Degioanni