> The reset duration could vary for each reset ID, since they refer to > different hardware. There's probably no "optimum" value here. Figures 7.3. and 7.4. say "5us delay is recommended" when changing XXX.RSTN bits, at least for the NoC parts. Yes, we can agree to trust the driver authors to get it right, but we could also decide to make it hard to get wrong. Bigger delays can still be open-coded if needed.