Re: big.LITTLE status for rk3399/rockpro64?

On Thu, 30 Jul 2020 11:08:11 -0700,
Greg V wrote:
> 
> 
> 
> On Thu, Jul 30, 2020 at 10:05, Josh Howard <[email protected]> wrote:
> > On Thu, 30 Jul 2020 09:02:56 -0700,
> > William Carson via freebsd-arm wrote:
> >> 
> >> 
> >>  > On Jul 28, 2020, at 10:43 AM, Vincent Milum Jr
> >> <[email protected]> wrote:
> >>  >
> >>  > The downside is that the two big cores are not fully clocking up
> >> to their max
> >>  > speed, because FreeBSD doesn't have support yet to clock cores
> >> or groups of
> >>  > cores independently.
> >> 
> >>  Is this true? I'm running r363091 on a RockPro64 with
> >> sysutils/powerdxx and it
> >>  supports the different frequencies just fine:
> >> 
> >>  # sysctl dev.cpu.{0,1,2,3,4,5}.freq
> >>  dev.cpu.0.freq: 1416
> >>  dev.cpu.1.freq: 1416
> >>  dev.cpu.2.freq: 1416
> >>  dev.cpu.3.freq: 1416
> >>  dev.cpu.4.freq: 1800
> >>  dev.cpu.5.freq: 1800
> > 
> > powerdxx definitely does help. I think the other issue is that if
> > you set ncpu=4,
> > you get get the max freq_levels of 1512 for the 4 CPUs, but when all
> > 6 are
> > enabled, it finds the highest common freq, which is
> > 1416. Furthermore, we only
> > set the big cores at 1800, but they should be 2016.
> 
> Wait, "it" == stock powerd? oh sure it might not support that. I was
> only considering manual setting of these sysctls.
> 
> re: what the maximum should be, that depends on your device tree.
> I have overclocked mine to 2200.
> 

I'm probably conflating too many things, between what I was observing on
Linux, which seemed to default to max clock of 1512 on little cores and
what I see on FreeBSD. I see in rk3399-opp.dtsi that those freq_levels
match what's in the freq_levels in sysctl, so not sure the discrepency.

What device tree are you using to get up to 2200?
_______________________________________________
[email protected] mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-arm
To unsubscribe, send any mail to "[email protected]"