Invited: Toward Sustainable and Transparent Benchmarking for Academic Physical Design Research
This paper introduces RosettaStone 2.0, an open-source framework built on OpenROAD-Research that enables transparent, reproducible, and rigorous benchmarking of both conventional 2D and hybrid-bonded 3D physical designs through standardized RTL-to-GDS flows, continuous integration testing, and a community-governed leaderboard.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine the world of chip design (making the brains of computers) as a giant, high-stakes cooking competition. For years, different chefs (researchers) have been trying to prove their recipes are the best. But there's a major problem: they aren't cooking in the same kitchen.
Some chefs use a 20-year-old stove, others use a brand-new induction cooktop. Some have a full pantry with every spice imaginable, while others are missing half the ingredients. When they say, "My soup is faster to make!" or "My cake tastes better!", it's impossible to know if they are actually better cooks, or if they just had better equipment and ingredients.
This paper introduces RosettaStone 2.0, a new project designed to fix this chaos. Think of it as building a universal, fair-play kitchen where every chef must use the exact same stove, the exact same ingredients, and the exact same measuring cups.
Here is a breakdown of what they built, using simple analogies:
1. The Problem: "Apples to Oranges" Comparisons
In the past, researchers tried to compare their new chip designs. But because everyone used different tools and missing data, it was like comparing a car built in a garage to a car built in a factory. You couldn't tell if the garage car was faster because the driver was better, or just because the factory car had a flat tire. This made it hard to know what was actually working in the field of chip design.
2. The Solution: A "RosettaStone" for Chips
Just as the real Rosetta Stone helped us translate ancient languages into modern ones, this RosettaStone 2.0 translates old, messy research data into a clean, standard format.
- The Kitchen (The Framework): They built a "kitchen" called OpenROAD-Research. It's an open-source platform where anyone can cook.
- The Recipe Book (The Flow): They created a complete, step-by-step recipe (called a "flow") that takes a raw idea (RTL) and turns it into a finished, physical blueprint (GDS) that a factory can actually build.
- The Scorecard (Metrics): They introduced a strict rulebook called METRICS2.1. This ensures that when someone says "My chip is faster," they are measuring speed in the exact same way as everyone else. No more guessing.
3. The New Challenge: 3D Stacking (The "Layer Cake")
The paper focuses heavily on a new way of building chips called Pin-3D or Face-to-Face (F2F) Hybrid Bonding.
- The Old Way (2D): Imagine a single-layer pancake. You spread the batter flat.
- The New Way (3D): Imagine a multi-layer cake. You bake two separate cakes (chips) and then stick them face-to-face with a super-strong, super-thin glue (Hybrid Bonding). This makes the chip much more powerful and compact.
The paper claims to be the first to provide a complete, open-source "recipe" for baking these 3D layer cakes. Before this, people could only bake the flat pancakes (2D) or had to use secret, expensive recipes (commercial tools) to make the 3D cakes. Now, anyone can try to bake a 3D cake in this new kitchen.
4. How They Tested It
To prove their kitchen works, they baked several "cakes" (chip designs) and compared them:
- The Test: They baked the same designs using their new open-source kitchen and compared the results to a "commercial kitchen" (a top-tier, expensive industry tool).
- The Findings:
- They showed that their open-source kitchen can successfully bake complex 3D cakes.
- They found that while the open-source kitchen is great at planning and placing the ingredients, it sometimes takes longer to "route" the wires (like untangling a mess of spaghetti) compared to the expensive commercial tools.
- They discovered that the size of the "glue" (Hybrid Bonding Terminals) matters a lot. If the glue spots are too big, they block the roads on the cake, causing traffic jams (errors). If they are smaller, the traffic flows much better.
5. The "Leaderboard"
Finally, they set up a public Leaderboard. Think of this like a sports scoreboard. Because everyone is using the same kitchen and the same rules, researchers can now submit their results, and the community can see exactly who is winning. This is governed by strict rules to ensure no one cheats or fakes their data.
Summary
In short, this paper says: "We built a fair, open, and transparent kitchen for chip designers. We included a new way to make 3D layer-cake chips. We provided a strict rulebook so everyone measures success the same way. And we set up a scoreboard so we can finally see who is truly making the best chips, without any confusion."
This allows researchers to stop arguing about whose tools are better and start focusing on actually inventing better chip designs.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.