ACT Ecosystem

Research Lead: Devansh Jain (PhD Student, UIUC)
Principal Investigator: Prof. Charith Mendis (UIUC)
Meet the research team →

Recent years have seen a proliferation of specialized AI accelerators – proposed in both academia (e.g., Gemmini, FEATHER, EVA) and industry (e.g., Google TPU, Intel AMX, AWS Trainium) – that depart significantly from traditional CPU/GPU architectures. However, research on compiler and systems support for these accelerators remains sparse, largely due to the lack of mature open-source compiler infrastructures capable of targeting them from popular ML frameworks like PyTorch, and JAX. Building such support involves considerable manual effort, slowing innovation and creating a gap between hardware and software research communities.

To bridge this gap, we present ACT (Accelerator Compiler Toolkit) Ecosystem, an ecosystem that automatically generates complete compiler backends and essential software tooling from high-level ISA specifications of AI accelerators.

The ACT ecosystem consists of:

TAIDL (Tensor Accelerator ISA Definition Language): A Python-based DSL for specifying AI accelerator ISAs. TAIDL leverages tensor IRs like XLA-HLO to compactly and precisely model execution semantics of AI accelerator ISAs.
TAIDL-TO (Test Oracle) Generator: Automatically generates scalable functional simulators just from TAIDL specifications, enabling correctness testing of the software stack. TAIDL-TOs are orders of magnitude faster than existing simulators.
ACT Backend Generator: Automatically generates sound and complete compiler backends just from TAIDL specification. ACT backends match or outperform state-of-the-art expert-written libraries, while maintaining low compile times (<1 sec).
XLA Integration: Enables end-to-end compilation from popular ML frameworks like JAX, TensorFlow, and PyTorch

Getting Started

We recommend starting with our MICRO 2025 tutorial. The tutorial provides a step-by-step walkthrough of the ACT ecosystem, requiring no prior experience with AI accelerators or ML compilers, and provides hands-on exercises to get familiar with the TAIDL Python DSL and automated tool generation. We encourage you to follow along and experiment with the code.

A revised tutorial will be presented at ASPLOS 2026 (Pittsburgh, Mar 22) – updated materials coming soon.

Collaboration Opportunities

We are open to collaborations to extend the ACT ecosystem in various directions, including but not limited to:

AI accelerator designers – hardware teams that have built or are designing custom AI accelerators and want to leverage the ACT ecosystem to automatically get software tooling, without the overhead of building a separate software stack
ISA-driven automation – researchers working on code generation, formal verification, or simulation from high-level hardware specifications, who want to leverage TAIDL as a common language for modeling and automation
ML compiler teams – teams building or maintaining compilers for existing commercial accelerators (e.g., AWS Trainium, custom ASICs) who want to explore ISA-driven automation or benchmark against ACT-generated backends
MLSys research – researchers working on compiler optimizations, autotuning, or hardware-software co-design for AI accelerators, who want to use ACT as a platform for rapid prototyping and evaluation

If any of these resonate – or if you have related ideas along these themes – reach out to Devansh Jain or Prof. Charith Mendis.

Selected Publications

arXiv

ACT: Automatically Generating Compiler Backends from Tensor Accelerator ISA Descriptions

Devansh Jain, Akash Pardeshi, Marco Frigo, Krut Patel, Kaustubh Khulbe, Jai Arora, and Charith Mendis

Oct 2025

DOI Bib PDF

@misc{act-arxiv,
  title = {ACT: Automatically Generating Compiler Backends from Tensor Accelerator ISA Descriptions},
  author = {Jain, Devansh and Pardeshi, Akash and Frigo, Marco and Patel, Krut and Khulbe, Kaustubh and Arora, Jai and Mendis, Charith},
  year = {2025},
  archiveprefix = {arXiv},
  primaryclass = {cs.PL},
  url = {https://arxiv.org/abs/2510.09932},
  doi = {10.48550/arXiv.2510.09932},
  month = oct
}

MICRO

TAIDL: Tensor Accelerator ISA Definition Language with Auto-generation of Scalable Test Oracles

Devansh Jain, Marco Frigo, Jai Arora, Akash Pardeshi, Zhihao Wang, Krut Patel, and Charith Mendis

In 58th IEEE/ACM International Symposium on Microarchitecture (MICRO 2025)

Oct 2025

DOI Bib PDF

@inproceedings{taidl-micro2025,
  author = {Jain, Devansh and Frigo, Marco and Arora, Jai and Pardeshi, Akash and Wang, Zhihao and Patel, Krut and Mendis, Charith},
  title = {TAIDL: Tensor Accelerator ISA Definition Language with Auto-generation of Scalable Test Oracles},
  year = {2025},
  isbn = {9798400715730},
  publisher = {Association for Computing Machinery},
  address = {New York, NY, USA},
  url = {https://doi.org/10.1145/3725843.3756075},
  doi = {10.1145/3725843.3756075},
  booktitle = {Proceedings of the 2025 58th IEEE/ACM International Symposium on Microarchitecture},
  pages = {1316–1333},
  numpages = {18},
  series = {MICRO '25},
  month = oct
}

POPL

TensorRight: Automated Verification of Tensor Graph Rewrites

Jai Arora, Sirui Lu, Devansh Jain, Tianfan Xu, Farzin Houshmand, Phitchaya Mangpo Phothilimthana, Mohsen Lesani, Praveen Narayanan, Karthik Srinivasa Murthy, Rastislav Bodik, Amit Sabne, and Charith Mendis

In 52nd ACM SIGPLAN Symposium on Principles of Programming Languages

Jan 2025

Distinguished Paper Award

DOI Bib PDF

Distinguished Paper Award

@article{tensorright-popl2025,
  author = {Arora, Jai and Lu, Sirui and Jain, Devansh and Xu, Tianfan and Houshmand, Farzin and Phothilimthana, Phitchaya Mangpo and Lesani, Mohsen and Narayanan, Praveen and Murthy, Karthik Srinivasa and Bodik, Rastislav and Sabne, Amit and Mendis, Charith},
  title = {TensorRight: Automated Verification of Tensor Graph Rewrites},
  year = {2025},
  issue_date = {January 2025},
  publisher = {Association for Computing Machinery},
  address = {New York, NY, USA},
  volume = {9},
  number = {POPL},
  url = {https://doi.org/10.1145/3704865},
  doi = {10.1145/3704865},
  journal = {Proc. ACM Program. Lang.},
  month = jan,
  articleno = {29},
  numpages = {32},
  keywords = {Denotational Semantics, Tensor Compilers, Unbounded Verification}
}

Collaborating Universities

ACT is developed at the University of Illinois Urbana-Champaign (UIUC), with collaborations across multiple universities through the ACE Center for Evolvable Computing, an SRC JUMP 2.0 Center.

Getting Started

Collaboration Opportunities

Selected Publications

Collaborating Universities

Sponsors