[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <YfmtxA7fCmbBWK0Z@Ansuel-xps.localdomain>
Date: Tue, 1 Feb 2022 23:01:40 +0100
From: Ansuel Smith <ansuelsmth@...il.com>
To: Stephen Boyd <sboyd@...nel.org>
Cc: Andy Gross <agross@...nel.org>,
Bjorn Andersson <bjorn.andersson@...aro.org>,
Michael Turquette <mturquette@...libre.com>,
Philipp Zabel <p.zabel@...gutronix.de>,
Rob Herring <robh+dt@...nel.org>,
Taniya Das <tdas@...eaurora.org>, devicetree@...r.kernel.org,
linux-arm-msm@...r.kernel.org, linux-clk@...r.kernel.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH v3 10/15] drivers: clk: qcom: gcc-ipq806x: add additional
freq for sdc table
On Tue, Jan 25, 2022 at 02:18:24PM -0800, Stephen Boyd wrote:
> Quoting Ansuel Smith (2022-01-25 13:03:52)
> > On Tue, Jan 25, 2022 at 12:45:53PM -0800, Stephen Boyd wrote:
> > > Quoting Ansuel Smith (2022-01-21 13:03:35)
> > > > Add additional freq supported for the sdc table.
> > > >
> > > > Signed-off-by: Ansuel Smith <ansuelsmth@...il.com>
> > > > ---
> > > > drivers/clk/qcom/gcc-ipq806x.c | 1 +
> > > > 1 file changed, 1 insertion(+)
> > > >
> > > > diff --git a/drivers/clk/qcom/gcc-ipq806x.c b/drivers/clk/qcom/gcc-ipq806x.c
> > > > index 77bc3d94f580..dbd61e4844b0 100644
> > > > --- a/drivers/clk/qcom/gcc-ipq806x.c
> > > > +++ b/drivers/clk/qcom/gcc-ipq806x.c
> > > > @@ -1292,6 +1292,7 @@ static const struct freq_tbl clk_tbl_sdc[] = {
> > > > { 20210000, P_PLL8, 1, 1, 19 },
> > > > { 24000000, P_PLL8, 4, 1, 4 },
> > > > { 48000000, P_PLL8, 4, 1, 2 },
> > > > + { 52000000, P_PLL8, 1, 2, 15 }, /* 51.2 Mhz */
> > >
> > > Why the comment and fake rate? Can it be 51200000 instead and drop the
> > > comment?
> >
> > I will add the related reason in the commit.
> >
> > We cannot achieve exact 52Mhz(jitter free) clock using PLL8.
> > As per the MND calculator the closest possible jitter free clock
> > using PLL8 is 51.2Mhz. This patch adds the values, which will provide
> > jitter free 51.2Mhz when the requested frequency is 52mhz.
>
> Sounds like this clk should use the round down clk_ops instead of the
> round up ones. Then the actual frequency can be in the table.
Some hint on how to do that? This use the rcg generic ops that doesn't
use any round. Should I crate some special ops in the rcg driver to
implement the round ops?
--
Ansuel
Powered by blists - more mailing lists