On Thu, Sep 12, 2019 at 8:58 AM Adam Ford <aford173@xxxxxxxxx> wrote: > > On Wed, Sep 11, 2019 at 12:49 PM Adam Ford <aford173@xxxxxxxxx> wrote: > > > > On Wed, Sep 11, 2019 at 12:43 PM H. Nikolaus Schaller <hns@xxxxxxxxxxxxx> wrote: > > > > > > Hi Adam, > > > > > > > Am 11.09.2019 um 18:01 schrieb H. Nikolaus Schaller <hns@xxxxxxxxxxxxx>: > > > > > > > >> > > > >> Am 11.09.2019 um 17:56 schrieb Adam Ford <aford173@xxxxxxxxx>: > > > >> > > > >>> There are some subtleties for testing. > > > >>> > > > >>> * I have added turbo-mode; to OPP6 / OPP1G > > > >>> * which means they are available but not used by the ondemand govenor > > > >>> * to enable them one has to echo 1 >/sys/devices/system/cpu/cpufreq/boost > > > >> > > > >> Will that be documented somewhere? If not, can we put a comment in the > > > >> device tree so people know how to enable it? > > > > > > > > It seems to be pretty standard on i86 systems if you google for "turbo mode". > > > > I have added it to the commit message which adds the vbb regulator. > > > > > > And, I am not sure if DT maintainers will accept comments about the > > > Linux /sys implementation in device tree files or bindings. Those > > > should be independent of Linux. > > > > OK. > > > > > > Basically the turbo-mode property is a hint to the OPP system (which > > > may or may not use of it). > > > > > > So I think it is indeed better to have it in the commit message and > > > not the code. > > > > That makes sense. > > > > > > > > One more thought: as long as we do not have junction temperature monitoring > > > we should keep it off by default... We may even remove the turbo-mode > > > designator if we have the 90°C limit and smart reflex working. > > I found some info on the thermal framework [1]. It seems to show that > we can put placeholders in the device tree to help facilitate this. > > An excerpt from [1] shows: > Cooling devices provide control on power dissipation. > There are essentially two ways to provide control on power dissipation. > First is by means of regulating device performance, which is > known as passive cooling (DVFS). > ... [snip stuff about cooling fans] > > It also shows there are ways to link the cpu node to the cooling info. > Since it shows DVFS can be used to regulate temperature, it seems like > the hooks might already be in place > > [1] - https://blog.linuxplumbersconf.org/2015/ocw//system/presentations/2613/original/thermal-framework-status-no-transitioning.pdf > > The downside is that the OMAP thermal sensor is unreliable. I submitted an RFC for thermal throttling. I am not sure if I did it right, and I don't have a good way to test it. adam > > > > We're almost there! > > > > adam > > > > > > BR, > > > Nikolaus > > >
