# IronMan: GNN-assisted Design Space Exploration in High-Level Synthesis via Reinforcement Learning

arXiv: https://arxiv.org/abs/2102.08138 · HTML: https://mlsanity.com/paper/2102.08138/

## Key facts

- Cited by 26 papers in the map (data through 14 Oct 2026).
- Builds on 5 of its 22 references that are in the map.
- First posted on arXiv 16 Feb 2021.

## Builds on (most cited first, 5 of 5)

- [Semi-Supervised Classification with Graph Convolutional Networks](https://mlsanity.com/paper/1609.02907/) (2016)
- [Inductive Representation Learning on Large Graphs](https://mlsanity.com/paper/1706.02216/) (2017)
- [Transformations of High-Level Synthesis Codes for High-Performance Computing](https://mlsanity.com/paper/1805.08288/) (2018)
- [Pyramid: Machine Learning Framework to Estimate the Optimal Timing and Resource Usage of a High-Level Synthesis Design](https://mlsanity.com/paper/1907.12952/) (2019)
- [Machine Learning Based Routing Congestion Prediction in FPGA High-Level Synthesis](https://mlsanity.com/paper/1905.03852/) (2019)

## Built on by (most cited first, 26 of 26)

- [Unsupervised Learning for Combinatorial Optimization with Principled Objective Relaxation](https://mlsanity.com/paper/2207.05984/) (2022)
- [High-Level Synthesis Performance Prediction using GNNs: Benchmarking, Modeling, and Advancing](https://mlsanity.com/paper/2201.06848/) (2022)
- [BaCO: A Fast and Portable Bayesian Compiler Optimization Framework](https://mlsanity.com/paper/2212.11142/) (2022)
- [Hierarchical Mixture of Experts: Generalizable Learning for High-Level Synthesis](https://mlsanity.com/paper/2410.19225/) (2024)
- [HLSFactory: A Framework Empowering High-Level Synthesis Datasets for Machine Learning and Beyond](https://mlsanity.com/paper/2405.00820/) (2024)
- [LIFT: LLM-Based Pragma Insertion for HLS via GNN Supervised Fine-Tuning](https://mlsanity.com/paper/2504.21187/) (2025)
- [LOSTIN: Logic Optimization via Spatio-Temporal Information with Hybrid Graph Models](https://mlsanity.com/paper/2201.08455/) (2022)
- [Verilog-to-PyG -- A Framework for Graph Learning and Augmentation on RTL Designs](https://mlsanity.com/paper/2311.05722/) (2023)
- [Theta-Resonance: A Single-Step Reinforcement Learning Method for Design Space Exploration](https://mlsanity.com/paper/2211.02052/) (2022)
- [MetaML-Pro: Cross-Stage Design Flow Automation for Efficient Deep Learning Acceleration](https://mlsanity.com/paper/2502.05850/) (2025)
- [Graph Neural Networks: A Powerful and Versatile Tool for Advancing Design, Reliability, and Security of ICs](https://mlsanity.com/paper/2211.16495/) (2022)
- [Learning to Compare Hardware Designs for High-Level Synthesis](https://mlsanity.com/paper/2409.13138/) (2024)
- [BoolGebra: Attributed Graph-learning for Boolean Algebraic Manipulation](https://mlsanity.com/paper/2401.10753/) (2024)
- [Subequivariant Reinforcement Learning Framework for Coordinated Motion Control](https://mlsanity.com/paper/2403.15100/) (2024)
- [Program-to-Circuit: Exploiting GNNs for Program Representation and Circuit Translation](https://mlsanity.com/paper/2109.06265/) (2021)
- [A Survey and Perspective on Artificial Intelligence for Security-Aware Electronic Design Automation](https://mlsanity.com/paper/2204.09579/) (2022)
- [A Benchmark on Directed Graph Representation Learning in Hardware Designs](https://mlsanity.com/paper/2410.06460/) (2024)
- [ForgeHLS: A Large-Scale, Open-Source Dataset for High-Level Synthesis](https://mlsanity.com/paper/2507.03255/) (2025)
- [MPM-LLM4DSE: Reaching the Pareto Frontier in HLS with Multimodal Learning and LLM-Driven Exploration](https://mlsanity.com/paper/2601.04801/) (2026)
- [Agent Factories for High Level Synthesis: How Far Can General-Purpose Coding Agents Go in Hardware Optimization?](https://mlsanity.com/paper/2603.25719/) (2026)
- [Pyxis: An Open-Source Performance Dataset of Sparse Accelerators](https://mlsanity.com/paper/2110.04280/) (2021)
- [Enabling Automated FPGA Accelerator Optimization Using Graph Neural Networks](https://mlsanity.com/paper/2111.08848/) (2021)
- [TransPlace: Transferable Circuit Global Placement via Graph Neural Network](https://mlsanity.com/paper/2501.05667/) (2025)
- [Ralts: Robust Aggregation for Enhancing Graph Neural Network Resilience on Bit-flip Errors](https://mlsanity.com/paper/2507.18804/) (2025)
- [DAPO: Design Structure-Aware Pass Ordering in High-Level Synthesis with Graph Contrastive and Reinforcement Learning](https://mlsanity.com/paper/2512.11342/) (2025)
- [SyntheticHLS: Building Diverse Synthetic High-Level Synthesis Datasets using LLMs](https://mlsanity.com/paper/2610.00106/) (2026)

## Abstract (arXiv)

Despite the great success of High-Level Synthesis (HLS) tools, we observe several unresolved challenges: 1) the high-level abstraction of programming styles in HLS sometimes conceals optimization opportunities; 2) existing HLS tools do not provide flexible trade-off (Pareto) solutions among different objectives and constraints; 3) the actual quality of the resulting RTL designs is hard to predict. To address these challenges, we propose an end-to-end framework, namelyIronMan. The primary goal is to enable a flexible and automated design space exploration (DSE), to provide either optimal solutions under user-specified constraints, or various trade-offs among different objectives (such as different types of resources, area, and latency). Such DSE either requires tedious manual efforts or is not achievable to attain these goals through existing HLS tools. There are three components in IronMan: 1) GPP, a highly accurate graph-neural-network-based performance and resource predictor; 2) RLMD, a reinforcement-learning-based multi-objective DSE engine that explores the optimal resource allocation strategy, to provide Pareto solutions between different objectives; 3) CT, a code transformer to assist RLMD and GPP, which extracts the data flow graph from original HLS C/C++ and automatically generates synthesizable code with HLS directives. The experimental results show that: 1) GPP achieves high prediction accuracy, reducing prediction errors of HLS tools by 10.9x in resource utilization and 5.7x in timing; 2) RLMD obtains optimal or Pareto solutions that outperform the genetic algorithm and simulated annealing by 12.7% and 12.9%, respectively; 3) IronMan is able to find optimized solutions perfectly matching various DSP constraints, with 2.54x fewer DSPs and up to 6x shorter latency than those of HLS tools while being up to 400x faster than the heuristic algorithms and HLS tools.