On Tue, Nov 20, 2018 at 12:02 PM Andrzej Hajda <a.hajda@xxxxxxxxxxx> wrote: > > + /* > > + * The datasheet (table 39) specifies "limited clock channel speed" > > + * for 4 lanes as 550 Mbps for RGB888. As this is 4 bits at the time, > > + * the maximum HS frequency is 550/4 = 137.5 MHz. > > + */ > > Are you sure it is OK? > > 550Mbps is per lane, so since DSI is DoubleDataRate clock should be the > half of the speed - 275MHz. That is a bit confusing indeed. > Anyway more I think about it more doubts I have. hs_rate can be helpful > for command mode panels - panel refresh rate (provided by timings) > imposes only lower limit on the speed, max hs rate will impose upper limit. Yeah. I added it for the command mode-only Samsung s6d16d0. > In case of video mode, timings determines exactly value of the clock, so > max hs rate is unnecessary, am I right? As far as I can tell, yes, provided that the video mode timings are <= HS rate. (Here it would mainly be a safeguard/sanity check.) I wonder what happens with a video mode panel in LP mode really. It would be good to get feedback from some experts here. I assume that in practice in LP mode the frequency is set down and with the rest of the frame properties remaining the same, the frame rate will go down (as expected). So the relationship between mode clock, HS and LP in practice is a bit of mystery to me. I haven't seen anything in the upstream kernel making it possible for panels to enter LP mode, so I assume that is only done by vendor out-of-tree code as of today. It would be great if some people who worked with this in practice could provide some input. My naive idea is that the panel should provide two "resolutions", one for HS and one for LP, and they only really differ in frame rate. So then userspace choose the resolution with the lower frame rate to save power. Yours, Linus Walleij _______________________________________________ dri-devel mailing list dri-devel@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/dri-devel
