Submitted by Rachel Gardner on Mon, 25/05/2026 - 12:04
Throughout May, code-hosting platform GitHub is celebrating maintainer month, "a month for open source maintainers to gather, share, and be celebrated." As Computer Science research at Cambridge has a long history in open source, we're very glad to see our project xDSL being celebrated as part of GitHub's maintainer month.
xDSL is an open-source framework for developing compilers (i.e. software that translates computer code written in one programming language into another). Compilers are frequently used to translate code from a high-level language into executable code that will be used in a specific computing domain.
The use of xDSL is currently growing fast in the compiler community, where xDSL received over 500 GitHub stars, and now has over 100 community members who contributed code to its development.
xDSL is written in Python, and industry partners who value developer productivity and Python ecosystem integration have repeatedly chosen xDSL as the foundation of their industry-grade software products.
A Formula 1 car for family car drivers
"Imagine how difficult it would be for someone who's only ever driven a family car to drive a Formula 1 race car (a C++-based compiler like MLIR)," explains project leader Dr Tobias Grosser, Associate Professor here in Compiler Design. "xDSL offers the convenience of a family car, while exposing the technological foundations of a race car.
"This is important in today's world," he adds, "as AI is speeding up the pace of innovation and research and industrial teams need to be able to keep up with it when building the compiler stack that fuels AI."
Indeed, xDSL was developed in part to address the issue that – while central for the success of innovative hardware companies – development speed in compiler design had become an innovation bottleneck in academia.
xDSL was developed by a collaborative team, including Dr Grosser, and is being maintained by him and two of his PhD students, Sasha Lopoukhine and Mathieu Fehr (pictured right). It's one of the projects being celebrated in May by GitHub.
"Open source maintainers are behind the software we use everyday," the GitHub website explains. "But they don't always have the community or support they need. That's why we're celebrating open source maintainers during the month of May."
The need for tools like xDSL
The need for tools like xDSL is growing due to the combination of increasing diversity in hardware capabilities, pace of innovation in neural network architecture, and pressure on companies and research teams to stay competitive by iterating quickly to reach a high performance when either the hardware or software changes.
To do that, developers often turn to a framework that provides some functionality for part of the compilation.
A popular solution for this is LLVM, a toolkit for the construction of highly optimized compilers, optimizers, and run-time environments. But it was developed initially to compile C and C-like programming languages, so a sub-project called MLIR (Multi-Level Intermediate Representation) has been created to extend its use in different domains.
xDSL is compatible with MLIR, so its users can tap into LLVM's rich ecosystem of production compilers and backends.
Helping less experienced users try out their ideas
"LLVM and MLIR use programming patterns optimised for expert engineers compiling very large programs such as operating systems or browsers for industrial-level workloads," says Sasha Lopoukhine. "So it's difficult for students and researchers to experiment with what they are doing.
"The value of xDSL is that it provides a lighter-weight framework that is designed to be more approachable and make it easier for less experienced users to try out different ideas."
