AI compilers logopeterindia.net
Knowledge Hub

AI Compilers: MLIR and the Compilers Built on It

A directory of AI compilers and compiler infrastructure, starting with MLIR, the LLVM project framework for multi-level intermediate representations, and the compilers, dialects and hardware back ends that build on it.

20 resources•3 categories•Updated November 16, 2026

MLIR and LLVM Foundations

6 entries

MLIR (Multi-Level Intermediate Representation) is compiler infrastructure from the LLVM project. Instead of one fixed intermediate representation, it lets a compiler define dialects at different levels of abstraction, from tensor operations down to hardware instructions. Descriptions follow the projects' own pages.

Compilers and Dialects Built on MLIR

10 entries

Frontends, compilers and hardware back ends that use MLIR or are described as MLIR-based.

Papers and Background

4 entries

Where the ideas come from.

Related Pages

5 entries

Related directories on this site.

Frequently Asked Questions

What is an AI compiler?

An AI compiler translates a model written as high-level operations into efficient code for a specific processor. It optimises the computation graph, for example by fusing operators, then lowers it step by step to GPU kernels or CPU and accelerator instructions.

What is MLIR?

MLIR, the Multi-Level Intermediate Representation, is compiler infrastructure from the LLVM project. It lets a compiler define its own dialects at different abstraction levels and convert between them, which suits ML programs that move from tensor operations to hardware code.

Which AI compilers are built on MLIR?

Projects such as IREE, StableHLO, torch-mlir, ONNX-MLIR, Triton, Mojo and tt-mlir use or are described as building on MLIR. This page links to each project's own site, and adoption details change, so check the project documentation.

How is this page different from the ML Compilers page?

The ML Compilers page covers NVIDIA TensorRT, Apache TVM and XLA and related runtimes. This page focuses on MLIR and the compilers, dialects and hardware back ends built on it.

Is this page a ranking?

No. It is a curated directory. Each entry links to the project's own site, and the notes follow the projects' descriptions.