Abstract:Despite the great success of high-level synthesis (HLS) tools, we observe several unresolved challenges: 1) the high-level abstraction of HLS programming styles sometimes conceals optimization opportunities; 2) the actual quality of resulting RTL designs is hard to predict; and 3) existing HLS tools do not provide flexible tradeoff (Pareto) solutions among different objectives and constraints. To this end, we propose an end-to-end framework, namely, IronMan-Pro. The primary goal is to enable a flexible and automated design space exploration (DSE), to provide either optimized solutions under user-specified constraints or Pareto tradeoffs among different objectives (such as resource types, area, and latency). IronMan-Pro consists of three components: 1) GPP, a highly accurate graph-neural-network-based performance and resource predictor; 2) RLMD, a reinforcement-learning-based multiobjective design exploration engine for optimal resource allocation strategies, aiming to provide Pareto solutions among different objectives; and 3) CT, a code transformer to assist RLMD and GPP, which extracts the data flow graphs from original HLS C/C++ and automatically generates synthesizable code with optimized HLS directives. Experimental results show that, 1) GPP achieves high prediction accuracy, reducing the prediction errors of HLS tools by in resource utilization and in critical path (CP) timing; 2) compared with meta-heuristic-based techniques, IronMan-Pro generates superior solutions improving resource utilization by and CP timing by 7.6%–16.5%; and - ) under user-specified constraints, IronMan-Pro can find satisfying solutions over 96% of the cases, more than twice as many as that of meta-heuristic-based techniques and with a speedup of up to . This work demonstrates the great potential of applying machine learning algorithms in the electronic design automation domain, especially for the hard-to-solve problems, such as timing estimation and optimization. IronMan-Pro is available at https://github.com/lydiawunan/IronMan.

LOSTIN: (L)under-barogic (O)under-barptimization via (S)under-barpatio-(T)under-baremporal (In)under-barformation with Hybrid Graph Models

MTLSO: A Multi-Task Learning Approach for Logic Synthesis Optimization

Logic Synthesis Optimization Sequence tuning using RL-based LSTM and Graph Isomorphism Network

ML-based AIG Timing Prediction to Enhance Logic Optimization

BoolGebra: Attributed Graph-learning for Boolean Algebraic Manipulation

The prediction of the quality of results in Logic Synthesis using Transformer and Graph Neural Networks

Deep Neural Network-Based Surrogate-Assisted Inverse Optimization for High-Speed Interconnects

OpenLS-DGF: An Adaptive Open-Source Dataset Generation Framework for Machine Learning Tasks in Logic Synthesis

Logic Synthesis Optimization with Predictive Self-Supervision via Causal Transformers

Deep-PowerX: A Deep Learning-Based Framework for Low-Power Approximate Logic Synthesis

LaMAGIC: Language-Model-based Topology Generation for Analog Integrated Circuits

DE-HNN: An effective neural model for Circuit Netlist representation

LLM-Enhanced Bayesian Optimization for Efficient Analog Layout Constraint Generation

IronMan-Pro: Multiobjective Design Space Exploration in HLS via Reinforcement Learning and Graph Neural Network-Based Modeling

MasterRTL: A Pre-Synthesis PPA Estimation Framework for Any RTL Design

Graph Neural Networks for High-Level Synthesis Design Space Exploration

E-Syn: E-Graph Rewriting with Technology-Aware Cost Functions for Logic Synthesis

ECO-GNN: Signoff Power Prediction using Graph Neural Networks with Subgraph Approximation

INVICTUS: Optimizing Boolean Logic Circuit Synthesis via Synergistic Learning and Search

Intelligent Circuit Design and Implementation with Machine Learning

Accuracy Versus Simplification in an Approximate Logic Neural Model