Asynchronous resets are bad design practice. Reset your flops synchronously or not at all.https://twitter.com/iowahawkblog/status/932309842123022336 …
-
Show this thread
-
Replying to @oe1cxw
Sync reset tree need to be synthesized at Fmax of main clock. Tree has high fan-out and spans big distances. At high Fmax that means over-buffering + over-pipelining. Means big area / power overhead.
1 reply 0 retweets 1 like -
Replying to @wavedrom
All that is also true for async clock tree because you have to release synchronously. If you only reset with slow clock (as you described in other tweet) then your sync reset tree does not need to be synthesized for Fmax. This is a strawman. Has nothing to do with sync vs async.
1 reply 0 retweets 0 likes -
Replying to @oe1cxw
Async reset tree = separate clock domain, and yes it is synthesized for another (slow) clock. Sync reset tree is part of main clock domain, so has to be synthesized together, for the fastest clock that circuit targeting.
2 replies 0 retweets 0 likes -
You could probably exclude sync reset tree as the false / multi path, but I don't know how to do it safely.
1 reply 0 retweets 0 likes -
Replying to @wavedrom
There are a few ways to do that. Would need to look up the exact sdc syntax, but it is possible. If you are worried about getting it right in combined STA, you can always re-run STA with 2nd set of rules with everything in slow clock domain.
2 replies 0 retweets 0 likes
But if you manage to cleanly switch between different clock speeds without a risk of clock glitches, and manage to verify statically that you did that correctly, then setting up STA for slower constraints on reset edges is easy in comparison.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.