Hi Biju,
CC clk
On Mon, Oct 14, 2019 at 9:46 AM Biju Das <biju.das@xxxxxxxxxxxxxx> wrote:
> I have further investigated this issue, please find my findings below
>
> The "cpg_z_clk_round_rate" function is called 2 times from "dev_pm_opp_set_rate" (1 is direct call and other through "clk_set_rate") function.
>
> For 500000000 frequency, after doing math operation(mult = div_u64(rate * 32ULL, prate);), it gets a factor 10--> 10.66 truncated to 10 and the frequency returned by the system is 468748125
>
> On the second " cpg_z_clk_round_rate ", the below function make the value to worse, it gets a value of 9.9999, since it is integer division it is truncated to 9.
> mult = div_u64(rate * 32ULL, prate);
>
> Now the frequency 468748125, after doing math operation becomes 421874 KHz. This is the problem.
>
> So I think the div_64 is wrong here, Instead we could go with DIV64_U64_ROUND_CLOSEST.
>
> With "DIV64_U64_ROUND_CLOSEST" the factor is 10.66 which is rounded to 11.
>
> So for 500000, after math operation, it becomes 515624
>
> Please share your opinion this issue.
I don't know what's the best way to solve this.
Rounding instead of truncating may lead to programming a too high clock
frequency, which might damage the hardware.
Please note that the current code no longer uses div_u64(), but div64_ul()
instead, which should have no impact on your findings.
> > -----Original Message-----
> > From: Biju Das
> > Sent: Thursday, September 19, 2019 11:49 AM
> > To: Geert Uytterhoeven <geert+renesas@xxxxxxxxx>; Simon Horman
> > <horms@xxxxxxxxxxxx>; linux-renesas-soc@xxxxxxxxxxxxxxx; Dien Pham
> > <dien.pham.ry@xxxxxxxxxxxxxxx>; TAKESHI KIHARA
> > <takeshi.kihara.df@xxxxxxxxxxx>; Niklas Söderlund
> > <niklas.soderlund+renesas@xxxxxxxxxxxx>
> > Subject: Regarding CPU frequency reported by Salvator-X board
> >
> > Hi All,
> >
> > I started testing frequency mentioned in OPP table with M3-W1.1 Salvator-X
> > board (NOT salvator-XS) using user space governor
> >
> > The actual frequency reported for 0.5GHz(500000 KHz) is too much deviated
> > from the target frequency [INFO] Target frequency: 500000 KHz [INFO]
> > Actual frequency: 421874 KHz
> >
> > But if I change the extal value, as per the board schematic (16.6666MHz), the
> > value is some what closer to target frequency.
> > &extal_clk {
> > - clock-frequency = <16666666>;
> > + clock-frequency = <16666600>;
> > };
> > [INFO] Target frequency: 500000 KHz
> > [INFO] Actual frequency: 468748 KHz
> >
> > Q1) Have any one seen this issue? Please share your thoughts on this issue.
> >
> > Note:-
> > I am not seeing this issue on Salvator-XS board, where the extal_clk value=
> > 16640000(16.64MHz).
> >
> > Please see the logs
> > with clock-frequency = <16666666>;
> > -----------------------------------------------
> > root@salvator-x:/cip-test-scripts# ./opp_t_001.sh [INFO] Testing cpufreq for
> > policy:
> > [INFO] /sys/devices/system/cpu/cpufreq/policy0
> > [INFO]
> > [INFO] CPUs affected:
> > [INFO] 0 1
> > [INFO]
> > [INFO] Available frequencies:
> > [INFO] 500000 1000000 1500000
> > [INFO]
> > [INFO] Target frequency: 500000 KHz
> > [INFO] Actual frequency: 421874 KHz
> > [INFO]
> > [INFO] Target frequency: 1000000 KHz
> > [INFO] Actual frequency: 937499 KHz
> > [INFO]
> > [INFO] Target frequency: 1500000 KHz
> > [INFO] Actual frequency: 1499999 KHz
> > [INFO]
> > [INFO] Testing cpufreq for policy:
> > [INFO] /sys/devices/system/cpu/cpufreq/policy2
> > [INFO]
> > [INFO] CPUs affected:
> > [INFO] 2 3 4 5
> > [INFO]
> > [INFO] Available frequencies:
> > [INFO] 800000 1000000 1200000
> > [INFO]
> > [INFO] Target frequency: 800000 KHz
> > [INFO] Actual frequency: 749999 KHz
> > [INFO]
> > [INFO] Target frequency: 1000000 KHz
> > [INFO] Actual frequency: 974999 KHz
> > [INFO]
> > [INFO] Target frequency: 1200000 KHz
> > [INFO] Actual frequency: 1199999 KHz
> >
> > With clock-frequency = <16666600>
> > ----------------------------------------------
> > root@salvator-x:/cip-test-scripts# ./opp_t_001.sh [INFO] Testing cpufreq for
> > policy:
> > [INFO] /sys/devices/system/cpu/cpufreq/policy0
> > [INFO]
> > [INFO] CPUs affected:
> > [INFO] 0 1
> > [INFO]
> > [INFO] Available frequencies:
> > [INFO] 500000 1000000 1500000
> > [INFO]
> > [INFO] Target frequency: 500000 KHz
> > [INFO] Actual frequency: 468748 KHz
> > [INFO]
> > [INFO] Target frequency: 1000000 KHz
> > [INFO] Actual frequency: 937496 KHz
> > [INFO]
> > [INFO] Target frequency: 1500000 KHz
> > [INFO] Actual frequency: 1499994 KHz
> > [INFO]
> > [INFO] Testing cpufreq for policy:
> > [INFO] /sys/devices/system/cpu/cpufreq/policy2
> > [INFO]
> > [INFO] CPUs affected:
> > [INFO] 2 3 4 5
> > [INFO]
> > [INFO] Available frequencies:
> > [INFO] 800000 1000000 1200000
> > [INFO]
> > [INFO] Target frequency: 800000 KHz
> > [INFO] Actual frequency: 787496 KHz
> > [INFO]
> > [INFO] Target frequency: 1000000 KHz
> > [INFO] Actual frequency: 974996 KHz
> > [INFO]
> > [INFO] Target frequency: 1200000 KHz
> > [INFO] Actual frequency: 1199995 KHz
> >
> > Regards,
> > Biju
> >
>
--
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
