Bridging the Gap between Hardware Fuzzing and Industrial Verification Conference: Great Lakes Symposium on VLSI 2025; June 30-July 2, 2025; New Orleans, LA, USA Great Lakes Symposium on VLSI 2025 (GLSVLSI ’25), June 30-July 2, 2025, New Orleans, LA, USA DOI: 10.1145/3716368.3735190 ISBN: 979-8-4007-

Ruiyang Ma ∗ , Tianhao Wei ∗ , Jiaxi Zhang, Chun Yang, Jiangfang Yi, Guojie Luo † · 2025

Read Bridging the Gap between Hardware Fuzzing and Industrial Verification Conference: Great Lakes Symposium on VLSI 2025; June 30-July 2, 2025; New Orleans, LA, USA Great Lakes Symposium on VLSI 2025 (GLSVLSI ’25), June 30-July 2, 2025, New Orleans, LA, USA DOI: 10.1145/3716368.3735190 ISBN: 979-8-4007- by Ruiyang Ma ∗ , Tianhao Wei ∗ , Jiaxi Zhang, Chun Yang, Jiangfang Yi, Guojie Luo † (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-00461.txt.

Introduction

Ruiyang Ma and co-authors examine why hardware fuzzing, a promising way to test chip designs, is not yet ready for industry use. Fuzzing feeds random inputs to a design to hunt for bugs, but industrial verification has its own tools and habits. The authors review recent fuzzing methods from an industry viewpoint and set criteria for whether a method fits industrial practice.

They also check whether current verification tools can support fuzzing well. Several bottlenecks trace back to weak support in the industrial environment. To close the gap, the team built a prototype environment called HwFuzzEnv that supplies the missing support. With it, an earlier fuzzing method ran several hundred times faster in industrial settings. The authors offer the work as a reference for tool companies, urging them to build better fuzzing support into verification products.

Catalog record

CREATORRuiyang Ma ∗ , Tianhao Wei ∗ , Jiaxi Zhang, Chun Yang, Jiangfang Yi, Guojie Luo †
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE37 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2506-00461

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Bridging the Gap between Hardware Fuzzing and Industrial Verification Conference: Great Lakes Symposium on VLSI 2025; June 30-July 2, 2025; New Orleans, LA, USA Great Lakes Symposium on VLSI 2025 (GLSVLSI ’25), June 30-July 2, 2025, New Orleans, LA, USA DOI: 10.1145/3716368.3735190 ISBN: 979-8-4007-1496-2/2025/06 CCS: Hardware Functional verification CCS: Hardware Simulation and emulation CCS: Hardware Software tools for EDA Ruiyang Ma ∗ , Tianhao Wei ∗ , Jiaxi Zhang, Chun Yang, Jiangfang Yi, Guojie Luo † Abstract . As hardware design complexity increases, hardware fuzzing emerges as a promising tool for automating the verification process. However, a significant gap still exists before it can be applied in industry. This paper aims to summarize the current progress of hardware fuzzing from an industry-use perspective and propose solutions to bridge the gap between hardware fuzzing and industrial verification. First, we review recent hardware fuzzing methods and analyze their compatibilities with industrial verification. We establish criteria to assess whether a hardware fuzzing approach is compatible. Second, we examine whether current verification tools can efficiently support hardware fuzzing. We identify the bottlenecks in hardware fuzzing performance caused by insufficient support from the industrial environment. To overcome the bottlenecks, we propose a prototype, HwFuzzEnv, providing the necessary support for hardware fuzzing. With this prototype, the previous hardware fuzzing method can achieve a several hundred times speedup in industrial settings. Our work could serve as a reference for EDA companies, encouraging them to enhance their tools to support hardware fuzzing efficiently in industrial verification. † † footnotetext: *Both authors contributed equally to this research. † Corresponding author. 1. Introduction The growing complexity of hardware designs becomes a significant challenge for verification engineers. It is increasingly difficult for them to identify potential vulnerabilities during the hardware development process ( Laeufer et al. 2018 ; Canakci et al. 2021 ) . Extensive effort is dedicated to creating test cases of the design under test (DUT), which is crucial for achieving DUT coverage closure and discovering bugs in corner cases. Hardware fuzzing has recently emerged as a promising technique for reducing the substantial human effort required in hardware verification ( Laeufer et al. 2018 ; Trippel et al. 2022 ; Canakci et al. 2021 ; Hur et al. 2021 ; Canakci et al. 2023 ; Kande et al. 2022 ; Ma et al. 2024b ; Fajardo et al. 2021 ; Lin et al. 2023 ; Hossain et al. 2023 ) . This innovative approach employs a gray-box test generation strategy, using coverage as feedback to heuristically mutate previous high-quality inputs and generate new stimuli. It has demonstrated a remarkable ability to automatically complete test plans, achieving both code and functional coverage closures ( Laeufer et al. 2018 ; Canakci et al. 2021 ; Trippel et al. 2022 ) . Furthermore, fuzzing methods have enabled researchers to uncover previously hidden vulnerabilities ( Kande et al. 2022 ; Hur et al. 2021 ; Canakci et al. 2023 ) , some even within industrial hardware designs ( Thomas et al. 2024 ) . There is significant potential to apply hardware fuzzing in real-world industrial hardware development. However, a key question remains: What is the current gap between hardware fuzzing and its practical application in industrial verification? In this paper, we aim to address this topic by summarizing and exploring two sub-questions, as illustrated in Figure 1 . • What types of hardware fuzzing methods are compatible with industrial verification? • What additional support is needed from the industrial verification environment to facilitate hardware fuzzing? For the first sub-question, we review recent influential research on hardware fuzzing and find that many of these works are actually not compatible with industrial verification. We analyze their compatibilities from three dimensions: simulation platform, coverage metric, and mutation strategy, which will be detailed in Section 3 . In each dimension, we identify the most appropriate approach for industry. Firstly, the framework should utilize industrial simulators rather than acceleration platforms. Secondly, it should focus on achieving traditional coverage metrics that align with real-world test plans. Finally, it should eliminate the effort of creating design-specific mutation strategies for each individual design. Figure 1 . Our sub-questions about the current gap between hardware fuzzing and industrial verification. After summarizing the first sub-question, we turn to the second: Can current industrial hardware verification tools efficiently support hardware fuzzing? We select a hardware fuzzing work that aligns with our proposed standards and conduct a quantitative analysis, which will be detailed in Section 4 . Our findings indicate that existing industrial verification tools are not adequately equipped to support hardware fuzzing, resulting in significant inefficiencies. In fact, over 90% of the performance loss is attributed to the lack of necessary support from the industrial verification environment. We identify three key limitations in current industrial tools that hinder the efficiency of hardware fuzzing in practical applications. Firstly, existing industrial simulators lack support for efficient coverage collection. In fuzzing iterations, coverage is collected frequently as feedback and an inefficient coverage interface wastes a large portion of time. Secondly, industrial tools do not facilitate efficient mutation operations. Current hardware fuzzing works mainly rely on software fuzzers, which are time-consuming due to complex mutation processes and high communication overheads. Thirdly, industrial tools do not incorporate efficient parallel mechanisms for hardware fuzzing, leaving significant room for optimization. To overcome the limitations listed above, we develop a prototype hardware fuzzing environment called HwFuzzEnv , with three simple but effective enhancements. HwFuzzEnv incorporates essential support for hardware fuzzing within the hardware verification environment and associated tools. The experiments demonstrate that our enhancements provide a significant speedup for previous hardware fuzzing practices in industrial applications. In summary, our main contributions are as follows: • We assess the compatibilities of current hardware fuzzing methods for industrial verification and summarize our criteria that determine their suitabilities for industry use. • We identify the limitations in industrial tools that do not support hardware fuzzing efficiently. To the best of our knowledge, we are the first to address the lack of support for hardware fuzzing in industrial verification environments. • We develop a prototype industrial hardware fuzzing environment named HwFuzzEnv 1 , which incorporates simple but effective enhancements for each stage of hardware fuzzing, including coverage collection, mutation, and simulation. • We evaluate the efficiency and utility of HwFuzzEnv . With the support of HwFuzzEnv , RtlFuzzLab ( Fajardo et al. 2021 ) achieves a speedup of up to 21.5\times (on average 12.7\times ) with a single thread, and up to 621\times (on average 361\times ) with 64 threads. † † footnotetext: 1 The project is available at https://github.com/magicYang1573/fast-hw-fuzz . 2. Background 2.1. Industrial Hardware Verification In the industrial hardware development lifecycle ( Molina and Cadenas 2007 ; Cieplucha 2019 ) , engineers use hardware description languages (HDLs) at the register-transfer level (RTL) to describe microarchitectural modules. Electronic design automation (EDA) tools synthesize these RTL modules into gate-level designs, which are then mapped to the transistor level and eventually to physical layout for manufacturing. During hardware development, writing HDL code at the RTL is error-prone. As a result, an overwhelming 60% of a project’s total effort is dedicated to verification ( Foster 2022 ) . Simulation-based verification is a key technique in industrial hardware verification,

Read the complete text