On 25-05-22, 09:04, Stephen Boyd wrote: > I'm saying that each OPP table would be for a single clk, but they would > be connected through required-opps for the device's OPP table. Ahh, okay. > It would > mean that dev_pm_opp_set_clkname() would need extension to let a driver > indicate which clk is associated with an OPP table. Hmm, just that it complicates simple cases. Lets see. > From your other > reply on v3 it seems that you're leaning towards having an array of > frequency values in the OPP table instead of doing table linking? I am not against that to be honest, we have done that for voltages and current already. I am just not fine with having any one of them as the primary clock. I liked your idea of reusing "level" for that. I have started some rewriting of the core, to simplify things and reduce the number of ever increasing APIs (which you suggested earlier once). Lets see where we land eventually. -- viresh
