CS Compiler Seminar: Zirui Zhou & Denzel Brian Budii

- Sponsor
- Siebel School of Computing and Data Science
- Speaker
- Zirui Zhou & Denzel Brian Budii
- Contact
- Dushyant Bharadwaj & Avaljot Singh
- db50@illinois.edu | avaljot2@illinois.edu
- Originating Calendar
- Siebel School Speakers Calendar
Speaker: Zirui Zhou
Title: Automatic Rewrite Rule Discovery for GPU Tensor Compilers with Agents
Abstract: Discovering profitable tensor graph rewrites for GPU compilers remains challenging due to the large search space and the complexity of downstream GPU compilation. We present AutoRCA, a workload-driven, agent-based framework that uses compiler feedback to discover reusable graph rewrites for XLA. AutoRCA explores alternative computation representations to improve GPU execution through fusion, kernel boundary changes, and more efficient hardware implementations, and try to lift the concrete transformations into reusable rules. Using three LLMs and six training workloads on A100, the strongest configuration discovers 37 rewrite rules, including 22 absent from XLA’s existing algebraic simplifier. We integrate these rules into XLA and evaluate them on 13 workloads, including seven held-out workloads, across A100 and H100 GPUs. AutoRCA achieves up to 2.48× speedup and 1.50× geometric-mean speedup, compared with 1.11× for XLA’s algebraic simplifier, demonstrating generalization across workloads and GPU architectures.
Speaker Bio: Zirui Zhou is a second year Ph.D. student in the Siebel School of Computing and Data Science at the University of Illinois Urbana-Champaign (UIUC) advised by Prof. Charith Mendis. Zirui received a Bachelor’s degree from the School of Electronics Engineering and Computer Science at Peking University in 2025, advised by Prof. Xin Zhang. His research interests lie at the intersection of programming languages, formal methods, and AI, particularly reliable agentic compilers and formally verified optimizations for efficient machine learning on modern hardware accelerators.
Speaker: Denzel Brian Budii
Title: xDSL: Sidekick Compilation for SSA-Based Compilers
Abstract: Traditionally, compiler researchers either conduct experiments within an existing production compiler or develop their own prototype compiler; both options come with trade-offs. On one hand, prototyping in a production compiler can be cumbersome, as they are often optimized for program compilation speed at the expense of software simplicity and development speed. On the other hand, the transition from a prototype compiler to production requires significant engineering work. To bridge this gap, we introduce the concept of sidekick compiler frameworks, an approach that uses multiple frameworks that interoperate with each other by leveraging textual interchange formats and declarative descriptions of abstractions. Each such compiler framework is specialized for specific use cases, such as performance or prototyping. Abstractions are by design shared across frameworks, simplifying the transition from prototyping to production. We demonstrate this idea with xDSL, a sidekick for MLIR focused on prototyping and teaching. xDSL interoperates with MLIR through a shared textual IR and the exchange of IRs through an IR Definition Language. The benefits of sidekick compiler frameworks are evaluated by showing on three use cases how xDSL impacts their development: teaching, DSL compilation, and rewrite system prototyping. We also investigate the trade-offs that xDSL offers, and demonstrate how we simplify the transition between frameworks using the IRDL dialect. With sidekick compilation, we envision a future in which engineers minimize the cost of development by choosing a framework built for their immediate needs, and later transitioning to production with minimal overhead.
Speaker Bio: Denzel is from Romania, where he received a bachelor's degree in computer science from West University of Timisoara. As part of his undergraduate thesis, he worked on implementing backward and forward analysis for an interactive query language called MLIR-Query.
He has previously worked as a software engineer at Intel, where he worked on an NPU compiler for machine learning workloads.