On Mon, Sep 18, 2023 at 08:05:48PM +0200, Benjamin Bara wrote: > Hi Frank! > > On Mon, 18 Sept 2023 at 19:24, Frank Oltmanns <frank@xxxxxxxxxxxx> wrote: > > On 2023-09-18 at 00:39:56 +0200, Benjamin Bara <bbara93@xxxxxxxxx> wrote: > > Thank you very much for including me in the discussion. If I understood > > Maxime correctly, your proposal is close to what he was suggesting in > > the discussion you referenced. Unfortunately, it doesn't cover the > > rounding aspect (which you also mentioned in your cover letter and the > > description for clk_detect_unintended_rate_changes in patch 7. I've been > > pondering the last three weeks how to find a good solution to this > > problem, but so far haven't found any. > > I think if we stick to the idea of always enforcing the exact "typical > rate", we cannot avoid physically impossible cases. IMHO, it might make > sense to add a set_rate() function with a "timing_entry" (e.g. used by > display_timing.h[1]) to the clock API, which gives a suggestion but also > defines the "real" boundaries. This would provide a shared parent PLL > more freedom to provide a satisfying rate for all its children. It's definitely something we should do, and I've wanted to do that for a while. The clock rate is not the only thing we can change though. The usual trick is to modify the blanking areas to come up with a rate that matches what the hardware can provide without modifying the framerate. It belongs more in a KMS helper Maxime
Attachment:
signature.asc
Description: PGP signature
