Hi all, I thought you might be interested in this. I've run a couple of experiments with OpenLane &a...

Amr Gouhar

03/03/2021, 7:31 PM

Hi all, I thought you might be interested in this. I've run a couple of experiments with OpenLane & Sky130: 1. ADDER_TYPE comparison. 2. Resizer Vs OpenPhySyn comparison. 3. Synthesis Strategies comparison. I'm sharing the results here: https://docs.google.com/spreadsheets/d/1hmsmxpW9iaq0NfHA3fme3cl6m87pZ5Xs_pFltLCJQUE/edit?usp=sharing I'll be conducting more experiments over time and adding them to the same spreadsheets.

🎉 1

💡 1

👍 7

❤️ 1

Anton Blanchard

03/05/2021, 8:03 PM

Nice work @Amr Gouhar! Lots of interesting stuff to look into. I'm guessing this wasn't using the

openroad_update

branch, and even if it was, it doesn't have the recent resizer fix https://github.com/The-OpenROAD-Project/OpenROAD/commit/df1041a256fd9603f45fa58c2fc61c1de447b623

Anton Blanchard

03/05/2021, 8:03 PM

It could be interesting to look at OpenPhySyn vs Resizer with that (might be another issue or two we need to find in the Resizer)

Anton Blanchard

03/05/2021, 8:09 PM

Most of the FA/FCA/yosys adder tests matched what I saw: • RCA has less area but much worse timing than RCA and yosys (expected) • FA and yosys are similar in terms of timing (expected) • FA has a little less area than yosys (expected)

Anton Blanchard

03/05/2021, 8:11 PM

There are a few unexpected things in there, eg

sha512

where RCA has much better wns than FA or yosys

Open in Slack

Previous Next

What is Linen?

Linen is a search-engine friendly community platform. We offer integrations with existing Slack/Discord communities and make those conversations Google-searchable.