Evolution of SVALint: Open-Source Guardrails for AI-Generated Verification

At AsFigo, we have always believed that the best verification tools grow organically from real engineering friction. Over the last couple of years, as AI coding assistants have flooded into hardware workflows, that friction has fundamentally shifted. The core bottleneck is no longer just writing code—it is safely and predictably integrating AI-generated code into complex, existing environments.

As a small but dedicated contributor to the open-source EDA community, we want to share the journey of how SVALint evolved from a localized parser experiment into an essential guardrail for modern AI-assisted verification.

From Experiment to Ecosystem: The SVALint Journey

Our work on SVALint began in early 2023, we began experimenting with the early stages of slang and PySlang. That work led to PySlint, and eventually, we built out a custom linting infrastructure specifically focused on SystemVerilog testbench (including SVA).

We took those early iterations to the community, gathering invaluable feedback at:

The consensus from the open-source community was clear: a linter is only as useful as its ability to adapt to diverse environments. This explains why traditional EDA vendors are unable to sell such a product for decades!

By 2025, as we began deploying SVALint at customer sites, we ported the backend to support Google’s Verible. Later that same year, responding directly to customer requests, we re-architected the code to support commercial tools such as Verdi, Verisium, and others. In early 2026, we reached a major accessibility milestone, releasing the tool via PyPI so that integrating it is now as simple as running

  pip install svalint.

The AI Reality Check: Generation Without Validation is a Liability

Recently, we noticed a fascinating trend among our users: SVALint was increasingly being deployed not just to check human code, but to act as a frontline defense against Large Language Models (LLMs).

The new wave of AI-for-EDA startups is making incredible leaps in rapid code generation, but they often severely underestimate the necessity of structural and stylistic guardrails. LLMs can write textbook-perfect SVA in a vacuum, but hardware verification doesn't happen in a vacuum. Customers were finding that AI models frequently generated code that simply couldn't merge with their existing codebases due to systemic integration failures:

  • Proprietary Context: LLMs lack context for a team’s internal macro wrappers, reusable properties, and specific error-messaging conventions.
  • Target Tool Limitations: Different tools support different subsets of SVA. AI-generated assertions often fail when fed into different simulators. This is where SVALint shines—it acts as a filter to ensure AI output perfectly conforms to the specific SVA subsets supported by simulators including open-source Verilator, allowing teams to leverage LLMs without breaking their toolchains.

Without a localized guardrail, unverified AI code creates massive integration headaches. We love what the AI generation tools are doing, but SVALint provides the necessary reality check before that code is ever committed.

Build Your Own Linter (BYOL)

This integration gap is exactly why we champion the Build Your Own Linter (BYOL) philosophy.

SVALint provides a flexible, open-source framework where teams can swap out the underlying parser (be it slang, Verible, or commercial tools) and inject their own custom rulesets. Instead of rigid, top-down enforcement, SVALint acts as a translation layer. It forces AI-generated SVA to respect local macros, package structures, and specific simulator constraints at the desktop level.

By establishing this open, adaptable infrastructure, the goal is to give verification teams the tools they need to embrace automation securely—ensuring that whatever the AI generates, it still speaks the exact language of your local codebase.

Comments

Post a Comment

Popular posts from this blog

Cracking the UVM-Verilator Code: 50+ IPs, AI Guardrails, and the Open-Source North Star

UVM on Verilator Isn’t Impossible. It’s Just Hard. We Did It Anyway. The “standard” line in the industry: UVM and Verilator don’t mix. Enterprise-grade open-source verification is a fantasy. So we put that claim under load. ✔ 50+ production UVCs ported ✔ 35 + non-obvious failure modes uncovered ✔ Converted into UVMLint and SVALint rules ✔ AI guardrails to prevent regressions This post is the why and what; the how and numbers are for the room at Hacker Dojo.  This is production-grade code—the kind that tapes out chips—running on open-source infrastructure. The implication is bigger than a port: Reduced dependency on closed ecosystems. Negotiating leverage with EDA vendors. Strategic control over your verification stack. If you’re a semiconductor company exploring Verilator for serious IP, don’t start from scratch. Start from the team that already mapped the minefield. 📅 Join us at the Hacker Dojo Next Thursday, the AsFigo team...
Read more

Call for Collaboration: Seeking Public UVM Environments for Verilator 5.0+ Porting

The transition of SystemVerilog UVM environments to open-source simulators presents specific technical challenges. While Verilator 5.0+ has expanded support for various SystemVerilog constructs, the industry still lacks a documented and repeatable path for certain UVM-specific behaviors. In parallel, we aim to curate a public vault of UVM repositories that have been successfully ported to Verilator. AsFigo is initiating a collaborative effort to document these gaps, and we are looking for engineers, researchers, and students to contribute original UVM codebases to this effort.  Link to Google Form: Submit Your code base The Technical Focus This effort intentionally moves beyond minimal or illustrative examples. We are seeking unique, original UVM implementations that reflect realistic UVM environments and non-trivial SystemVerilog usage found in real-world verification flows. Infrastructure Components: Memo...
Read more

Breaking the License Barrier: The World’s First UVM + Verilator Hands-On Bootcamp - Mar-1, Sunday 3-5 PM PST

Remote link:  AsFigo UVM Verilator Bootcamp | Meeting-Join | Microsoft Teams For decades, the power of UVM (Universal Verification Methodology) has been locked behind the high-cost gates of commercial simulators. At AsFigo , we believe the future of silicon verification belongs to the community. We are breaking those gates down to empower a new generation of verification engineers. A Community Milestone The dream of open-source UVM is the result of a massive, multi-year industry effort. Foundations laid by the CHIPS Alliance, DARPA, Google, and Antmicro have reached a critical tipping point. This movement was fueled by the relentless technical contributions of pioneers such as Wilson Snyder, Geza Lore, Krzysztof Bieganski, et al. , whose work over the last several years made the "impossible" possible. The public breakthrough for UVM support in Verilator was announced in late 2023. On that very day, we at AsFigo successfully ported a complete APB UVC and...
Read more