On 09-01-20, 10:35, Saravana Kannan wrote: > Agreed for the example you are giving where PM domains/voltages are > dropped automatically when dropping the device freq to suspend freq. > I'm just wondering about a different scenario where if some power > domain needed to be at say 0.5v when it's suspended (no consumer using > it) The domain should be powered off in this case I think. > to not lose state, or to come back up without brownouts, etc then > suspend OPP for PM domains might be useful. But I don't know enough > about that to speak with authority, so I'll leave it at this. > > I see this suspend-opp as a way to mark to what level the bandwidth > needs to be dropped to/brought back up from during suspend/resume by > the driver making interconnect bandwidth requests. For example, what > if the CPU -> DDR needed to be at some level to avoid suspend/resume > issues (say CPU bug with respect to timing/latencies)? In this > example, the CPU driver would be the one making bandwidth requests for > CPU -> DDR bandwidth during normal operation and during > suspend/resume. So it's basically exactly the same way it would treat > CPU freq OPP. I understand your concerns but to me it all looks hypothetical right now. I am not saying we won't support suspend-opp for interconnect or domains, but that we should do it only if it is required. > Btw, I don't have a strong opinion on this. But, even if we do only a > rate comparison, what does it even mean to compare rates for genpd or > BW opp tables? We will never do the comparison because those tables will never have the suspend OPP in the respective tables. -- viresh
