This repository is a curated list of research papers, tools, and datasets focused on applying machine learning to compiler design and systems optimization. It serves as a comprehensive resource for researchers and engineers interested in leveraging ML for tasks like auto-tuning, phase ordering, instruction-level optimization, and program representation.

How It Works

The repository organizes a vast collection of academic literature and practical tools, categorizing them by specific compiler optimization areas. This structured approach allows users to quickly find relevant research and software for their specific needs, covering topics from iterative compilation and instruction-level optimization to memory modeling and domain-specific optimizations.

Quick Start & Requirements

• Installation: No direct installation is required as this is a curated list. Users will need to follow individual tool instructions.
• Prerequisites: Access to academic papers (often via institutional subscriptions or arXiv), and potentially specific software libraries (e.g., Python, MLIR, LLVM) for the listed tools.
• Resources: Primarily requires internet access for research and tool downloads.

Highlighted Details

• Extensive coverage of papers from major conferences (PLDI, CGO, PACT, ASPLOS, etc.) and journals (TACO).
• Links to numerous open-source tools and frameworks like TVM, CompilerGym, OpenTuner, and MLIR-based projects.
• Includes datasets and benchmarks such as TenSet, SPEC CPU® 2017, Project CodeNet, and CodeXGLUE.
• Categorization spans a wide range of ML applications in compilers, including program representation, cost modeling, and domain-specific optimizations.

Maintenance & Community

• The list appears to be actively curated, with recent entries from 2024 and 2025.
• Contribution guidelines are provided, encouraging community input via pull requests.

Licensing & Compatibility

• Licensing varies by the individual tools and datasets linked. The repository itself does not impose a specific license.
• Compatibility for commercial use depends on the licenses of the referenced projects.

Limitations & Caveats

This is a curated list, not a software project itself. Users must evaluate and integrate the individual tools and papers independently. The sheer volume of information may require significant effort to navigate and synthesize.