Claire Xen  🏳️‍⚧️ 🏳️‍🌈 🧙🏻‍♀️ BLM  🏴 🚩

Asynchronous resets are bad design practice. Reset your flops synchronously or not at all.https://twitter.com/iowahawkblog/status/932309842123022336 …

In most standard cell libraries that I used DFF with async reset smaller then DFF with sync reset. 90% of projects I worked over last 20 years, async reset was BKM

Yes, it is incredibly wide spread and most people who use it are completely unaware of all the things that could go wrong with it. That's why it's my "most trivial hill on which I am willing to die". :)

Re size: That's why I added the "or not at all" part. Most flops in a design actually don't need resetting. If area is a concern: Only reset the flops that actually need resetting. No reset needs even less area than async reset.

At least thinking about size of sync vs async rest _is_ thinking about it. AFAICT, many designers either don't consider sync vs async reset at all, or have a _very_ mistaken idea that async reset is somehow better.

Well, needing less area is a way it's better, isn't it? I assume that what can go wrong is that that asynchronous reset signal is basically another clock domain with all the worries about metastability that go along with that?

Also: sync reset tree (being part of the same clock domain) becomes a part of critical path during the timing closure of high speed designs. Ability to exclude async reset tree from timing closure is big plus. IMHO