[PATCH v4 7/7] arm64/amu: use capacity_ref_freq to set AMU ratio

[Vincent Guittot]

On Thu, 2 Nov 2023 at 11:31, Beata Michalska <beata.michalska at arm.com> wrote:
>
> Hi Vincent,
>
> On Wed, Nov 01, 2023 at 10:04:27AM +0100, Vincent Guittot wrote:
> > Hi Beata,
> >
> > On Wed, 1 Nov 2023 at 00:59, Beata Michalska <beata.michalska at arm.com> wrote:
> > >
> > > On Fri, Oct 27, 2023 at 10:04:00AM +0200, Vincent Guittot wrote:
> > > > Use the new capacity_ref_freq to set the ratio that is used by AMU for
> > > > computing the arch_scale_freq_capacity().
> > > > This helps to keep everything aligned using the same reference for
> > > > computing CPUs capacity.
> > > >
> >
> > [..]
> >
> > > > @@ -381,6 +385,9 @@ void topology_init_cpu_capacity_cppc(void)
> > > >       }
> > > >
> > > >       for_each_possible_cpu(cpu) {
> > > > +             freq_inv_set_max_ratio(cpu,
> > > > +                                    per_cpu(capacity_ref_freq, cpu) * HZ_PER_KHZ);
> > > > +
> > > The capacity_ref_freq set earlier will still lead to units mismatch,
> > > as at the point of calling topology_init_cpu_capacity_cppc the lowest & nominal
> > > frequencies will be provided in MHz (unless I have missed smth).
> >
> > I don't get your point:
> > the unit of per_cpu(capacity_freq_ref, cpu) is Khz
> >
> > For cppc, we have
> > per_cpu(capacity_freq_ref, cpu) = cppc_perf_to_khz(&perf_caps,
> > raw_capacity[cpu]);
> >
> > freq_inv_set_max_ratio() uses
> > arch_timer_get_rate() which returns a freq in Hz
> > and
> > per_cpu(capacity_freq_ref, cpu) * HZ_PER_KHZ. to get a freq in Hz too.
> >
> Apologies, I should have been more verbose here.
> Before the change made in [1] the cppc_perf_to_khz was indeed operating on
> lowest & nominal frequency values expressed in kHZ, as those were appropriately
> amended:
> cppc_cpufreq_cpu_init
>         |__> cppc_cpufreq_get_cpu_data:
>                 |__>    ...
>                         /* Convert the lowest and nominal freq from MHz to KHz */
>                         cpu_data->perf_caps.lowest_freq *= 1000;
>                         cpu_data->perf_caps.nominal_freq *= 1000;
>
> So far cppc_perf_to_khz (previously cppc_cpufreq_perf_to_khz) was being called
> with the post-processed CPC data (through cppc_cpufreq_get_cpu_data) and thus
> guaranteed to be operating on values in kHz.
> With the new changes the cppc_perf_to_khz will operate on raw CPC data,
> which are expressed in MHz.(as per acpi spec)

Ok, thanks for the explanation. I haven't noticed that cppc_cpufreq
was processing the raw CPC data  (*1000) before using them. I'm going
to fix that.


>
> ---
> Best Regards
> B.
> > > This means that use of both, the capacity_ref_freq and the arch_freq_scale
> > > will generate unexpected results, so I guess this should get amended in the
> > > preceding patch (?) [1]
> > >
> > > ---
> > > BR
> > > B.
> > >
> > > [1] https://lore.kernel.org/linux-arm-kernel/20231027080400.56703-4-vincent.guittot@linaro.org/T/#m42daa167097edc190b1cfc05382c385ed801d909
> > >
> > > >               capacity = raw_capacity[cpu];
> > > >               capacity = div64_u64(capacity << SCHED_CAPACITY_SHIFT,
> > > >                                    capacity_scale);
> > > > @@ -422,8 +429,11 @@ init_cpu_capacity_callback(struct notifier_block *nb,
> > > >
> > > >       cpumask_andnot(cpus_to_visit, cpus_to_visit, policy->related_cpus);
> > > >
> > > > -     for_each_cpu(cpu, policy->related_cpus)
> > > > +     for_each_cpu(cpu, policy->related_cpus) {
> > > >               per_cpu(capacity_ref_freq, cpu) = policy->cpuinfo.max_freq;
> > > > +             freq_inv_set_max_ratio(cpu,
> > > > +                                    per_cpu(capacity_ref_freq, cpu) * HZ_PER_KHZ);
> > > > +     }
> > > >
> > > >       if (cpumask_empty(cpus_to_visit)) {
> > > >               topology_normalize_cpu_scale();
> > > > diff --git a/include/linux/arch_topology.h b/include/linux/arch_topology.h
> > > > index 38ca6c76af56..ffdf0b7c55fa 100644
> > > > --- a/include/linux/arch_topology.h
> > > > +++ b/include/linux/arch_topology.h
> > > > @@ -99,6 +99,7 @@ void update_siblings_masks(unsigned int cpu);
> > > >  void remove_cpu_topology(unsigned int cpuid);
> > > >  void reset_cpu_topology(void);
> > > >  int parse_acpi_topology(void);
> > > > +void freq_inv_set_max_ratio(int cpu, u64 max_rate);
> > > >  #endif
> > > >
> > > >  #endif /* _LINUX_ARCH_TOPOLOGY_H_ */
> > > > --
> > > > 2.34.1
> > > >