On Friday 01 Sep 2023 at 15:03:08 (+0200), Vincent Guittot wrote:
> This is the 1st part of consolidating how the max compute capacity is
> used in the scheduler and how we calculate the frequency for a level of
> utilization.
>
> Fix some unconsistancy when computing frequency for an utilization. There
> can be a mismatch between energy model and schedutil.
There are a few more pieces of functionality that would be worth
consolidating in this set as well, if you'd like to consider them:
- arch_set_freq_scale() still uses policy->cpuinfo.max_freq. It might be
good to use the boot time stored max_freq here as well. Given that
arch_scale_cpu_capacity() would be based on that stored value, if
arch_scale_freq_capacity() ends up using a different value, it could
have interesting effects on the utilization signals in case of
boosting.
- As Pierre mentioned in a previous comment, there is already a
cpufreq_get_hw_max_freq() weak function that returns
policy->cpuinfo.max_freq and it's only used at boot time by
the setup code for AMU use for frequency invariance. I'm tempted to
suggest to use this to initialize what is now "freq_factor" as my
intention when I created that function was to provide platform
providers with the possibility to implement their own and decide on
the frequency they choose as their maximum. This could have been an
arch_ function as well, but as you mentioned before, mobile and server
platforms might want to choose different maximum values even if they
are using the same architecture.
Thanks,
Ionela.
>
> Next step will be to make a difference between the original
> max compute capacity of a CPU and what is currently available when
> there is a capping applying forever (i.e. seconds or more).
>
> Vincent Guittot (4):
> sched: consolidate and cleanup access to CPU's max compute capacity
> topology: add a new arch_scale_freq_reference
> cpufreq/schedutil: use a fixed reference frequency
> energy_model: use a fixed reference frequency
>
> arch/arm/include/asm/topology.h | 1 +
> arch/arm64/include/asm/topology.h | 1 +
> arch/riscv/include/asm/topology.h | 1 +
> drivers/base/arch_topology.c | 9 +++------
> include/linux/arch_topology.h | 7 +++++++
> include/linux/energy_model.h | 20 +++++++++++++++++---
> kernel/sched/core.c | 2 +-
> kernel/sched/cpudeadline.c | 2 +-
> kernel/sched/cpufreq_schedutil.c | 29 +++++++++++++++++++++++++++--
> kernel/sched/deadline.c | 4 ++--
> kernel/sched/fair.c | 18 ++++++++----------
> kernel/sched/rt.c | 2 +-
> kernel/sched/sched.h | 6 ------
> kernel/sched/topology.c | 7 +++++--
> 14 files changed, 75 insertions(+), 34 deletions(-)
>
> --
> 2.34.1
>
>
On Thu, 21 Sept 2023 at 12:12, Ionela Voinescu <[email protected]> wrote:
>
> On Friday 01 Sep 2023 at 15:03:08 (+0200), Vincent Guittot wrote:
> > This is the 1st part of consolidating how the max compute capacity is
> > used in the scheduler and how we calculate the frequency for a level of
> > utilization.
> >
> > Fix some unconsistancy when computing frequency for an utilization. There
> > can be a mismatch between energy model and schedutil.
>
> There are a few more pieces of functionality that would be worth
> consolidating in this set as well, if you'd like to consider them:
>
> - arch_set_freq_scale() still uses policy->cpuinfo.max_freq. It might be
> good to use the boot time stored max_freq here as well. Given that
> arch_scale_cpu_capacity() would be based on that stored value, if
> arch_scale_freq_capacity() ends up using a different value, it could
> have interesting effects on the utilization signals in case of
> boosting.
That's a good point, I will have a look at this part too. From a quick
look, this seems to be only used by architecture that are using
arch_topology.c too
>
> - As Pierre mentioned in a previous comment, there is already a
> cpufreq_get_hw_max_freq() weak function that returns
> policy->cpuinfo.max_freq and it's only used at boot time by
> the setup code for AMU use for frequency invariance. I'm tempted to
> suggest to use this to initialize what is now "freq_factor" as my
> intention when I created that function was to provide platform
> providers with the possibility to implement their own and decide on
> the frequency they choose as their maximum. This could have been an
> arch_ function as well, but as you mentioned before, mobile and server
> platforms might want to choose different maximum values even if they
> are using the same architecture.
>
> Thanks,
> Ionela.
>
> >
> > Next step will be to make a difference between the original
> > max compute capacity of a CPU and what is currently available when
> > there is a capping applying forever (i.e. seconds or more).
> >
> > Vincent Guittot (4):
> > sched: consolidate and cleanup access to CPU's max compute capacity
> > topology: add a new arch_scale_freq_reference
> > cpufreq/schedutil: use a fixed reference frequency
> > energy_model: use a fixed reference frequency
> >
> > arch/arm/include/asm/topology.h | 1 +
> > arch/arm64/include/asm/topology.h | 1 +
> > arch/riscv/include/asm/topology.h | 1 +
> > drivers/base/arch_topology.c | 9 +++------
> > include/linux/arch_topology.h | 7 +++++++
> > include/linux/energy_model.h | 20 +++++++++++++++++---
> > kernel/sched/core.c | 2 +-
> > kernel/sched/cpudeadline.c | 2 +-
> > kernel/sched/cpufreq_schedutil.c | 29 +++++++++++++++++++++++++++--
> > kernel/sched/deadline.c | 4 ++--
> > kernel/sched/fair.c | 18 ++++++++----------
> > kernel/sched/rt.c | 2 +-
> > kernel/sched/sched.h | 6 ------
> > kernel/sched/topology.c | 7 +++++--
> > 14 files changed, 75 insertions(+), 34 deletions(-)
> >
> > --
> > 2.34.1
> >
> >