From Pragmas to Partners: A Symbiotic Evolution of Agentic High-Level Synthesis

TL;DR

This paper argues that high-level synthesis (HLS) remains essential in the agentic hardware design era, serving as a fast, portable, and permutable abstraction that enables agent-driven optimization. It analyzes three contributions: (i) positioning HLS as a practical golden reference for agentic design, (ii) identifying core HLS limitations—performance feedback, rigid interfaces, and end-to-end verification—and (iii) proposing an autonomy-inspired taxonomy that traces the evolution from human copilots to autonomous design partners. The method centers on evaluating HLS’s role in rapid design space exploration, portability, and verification, while outlining how agentic systems can address current tool gaps through mixed-fidelity modeling, interface adaptation, and end-to-end validation strategies. The authors argue for a mixed workflow where agents and humans alternate between HLS sources and RTL, enabling scalable, platform-agnostic optimization and closer collaboration between HLS and AI communities. The significance lies in a concrete pathway toward symbiotic HLS-AI design that preserves the benefits of HLS while leveraging agentic capabilities to improve feedback, interoperability, and verification across the stack.

Abstract

The rise of large language models has sparked interest in AI-driven hardware design, raising the question: does high-level synthesis (HLS) still matter in the agentic era? We argue that HLS remains essential. While we expect mature agentic hardware systems to leverage both HLS and RTL, this paper focuses on HLS and its role in enabling agentic optimization. HLS offers faster iteration cycles, portability, and design permutability that make it a natural layer for agentic optimization.This position paper makes three contributions. First, we explain why HLS serves as a practical abstraction layer and a golden reference for agentic hardware design. Second, we identify key limitations of current HLS tools, namely inadequate performance feedback, rigid interfaces, and limited debuggability that agents are uniquely positioned to address. Third, we propose a taxonomy for the symbiotic evolution of agentic HLS, clarifying how responsibility shifts from human designers to AI agents as systems advance from copilots to autonomous design partners.