SLDB: An End-To-End Heterogeneous System-on-Chip Benchmark Suite for LLM-Aided Design

Abstract: Over the last few years, LLMs have emerged as a valuable tool for Electronic Design Automation (EDA). State-of-the-art research in LLM-aided design has demonstrated the ability of LLMs to generate syntactically correct RTL code, showcasing encouraging prospects for integrating AI into the hardware design process. A key enabler of these advancements is the availability of high-quality benchmarks to evaluate new approaches. However, existing datasets and benchmarks fall short of system-level design, as they focus primarily on component-level information and low-complexity designs. To address this gap, we introduce the System-Level Design Benchmark (SLDB), a dataset tailored for evaluating LLMs in system-level integration and configuration tasks. SLDB includes a curated benchmark suite of 10 baseline SoC designs, whose components can be combined into an exponential number of distinct tile-based SoCs through a synthetic library. The dataset provides full SoC configurations, accelerator integration code, communication parameters, and accelerator-aware system configurations, along with testing-application code, compatible with the ESP platform[1].