| Message ID | cover.1544782279.git.viresh.kumar@linaro.org |
|---|---|
| Headers | show |
| Series | PM / Domains: Allow performance state propagation | expand |
On Fri, Dec 14, 2018 at 11:15 AM Viresh Kumar <viresh.kumar@linaro.org> wrote: > > Hi, > > This series adds performance state propagation support in genpd core. > The propagation happens from the sub-domains to their masters. More > details can be found in the individual commit logs. > > This is tested on hikey960 by faking power domains in such a way that > the CPU devices have two power domains and both of them have the same > master domain. The CPU device, as well as its power domains have > "required-opps" property set and the performance requirement from the > CPU eventually configures all the domains (2 sub-domains and 1 master). > > Based on opp/linux-next branch (which is 4.20-rc1 + > multiple-power-domain-support-in-opp-core + some OPP fixes). > > Rajendra has already tested the previous version of this series and so I > have included his Tested-by for all patches. I'm assuming that this set will go in via the OPP tree. Cheers, Rafael
On 14-12-18, 11:18, Rafael J. Wysocki wrote: > On Fri, Dec 14, 2018 at 11:15 AM Viresh Kumar <viresh.kumar@linaro.org> wrote: > > > > Hi, > > > > This series adds performance state propagation support in genpd core. > > The propagation happens from the sub-domains to their masters. More > > details can be found in the individual commit logs. > > > > This is tested on hikey960 by faking power domains in such a way that > > the CPU devices have two power domains and both of them have the same > > master domain. The CPU device, as well as its power domains have > > "required-opps" property set and the performance requirement from the > > CPU eventually configures all the domains (2 sub-domains and 1 master). > > > > Based on opp/linux-next branch (which is 4.20-rc1 + > > multiple-power-domain-support-in-opp-core + some OPP fixes). > > > > Rajendra has already tested the previous version of this series and so I > > have included his Tested-by for all patches. > > I'm assuming that this set will go in via the OPP tree. Yes. -- viresh
Quoting Viresh Kumar (2018-12-14 02:15:26) > Hi, > > This series adds performance state propagation support in genpd core. > The propagation happens from the sub-domains to their masters. More > details can be found in the individual commit logs. > > This is tested on hikey960 by faking power domains in such a way that > the CPU devices have two power domains and both of them have the same > master domain. The CPU device, as well as its power domains have > "required-opps" property set and the performance requirement from the > CPU eventually configures all the domains (2 sub-domains and 1 master). > > Based on opp/linux-next branch (which is 4.20-rc1 + > multiple-power-domain-support-in-opp-core + some OPP fixes). > For the series Reviewed-by: Stephen Boyd <sboyd@kernel.org>
Hi, This series adds performance state propagation support in genpd core. The propagation happens from the sub-domains to their masters. More details can be found in the individual commit logs. This is tested on hikey960 by faking power domains in such a way that the CPU devices have two power domains and both of them have the same master domain. The CPU device, as well as its power domains have "required-opps" property set and the performance requirement from the CPU eventually configures all the domains (2 sub-domains and 1 master). Based on opp/linux-next branch (which is 4.20-rc1 + multiple-power-domain-support-in-opp-core + some OPP fixes). Rajendra has already tested the previous version of this series and so I have included his Tested-by for all patches. V3->V4: - dev_pm_opp_xlate_performance_state() returns negative values on error now. - of_get_required_opp_performance_state() is also updated to do the same. - _genpd_set_performance_state() is not called anymore from _genpd_reeval_performance_state() and we get rid of the extra declaration. - Improved commit log in the last patch. - dropped an unlikely. V2->V3: - Include Ulf's patch (sent separately earlier) with this series. - The performance state update code doesn't rely anymore on the power on/off state of the genpd, it sets and propagates rate in all cases. - That simplified a lot of code from V2 in _genpd_power_on(). - commit logs improved for few commits. - s/mstate/master_state/ - and few more minor changes. v1->V2: - First patch (1/5) is new and an improvement to earlier stuff. - Move genpd_status_on() check to _genpd_reeval_performance_state() from _genpd_set_performance_state(). - Improve dev_pm_opp_xlate_performance_state() to handle 1:1 pstate mapping between genpd and its master and also to fix a problem while finding the dst_table. - Handle pstate=0 case properly. -- viresh Ulf Hansson (1): PM / Domains: Make genpd performance states orthogonal to the idlestates Viresh Kumar (6): OPP: Improve _find_table_of_opp_np() OPP: Add dev_pm_opp_xlate_performance_state() helper OPP: Don't return 0 on error from of_get_required_opp_performance_state() PM / Domains: Save OPP table pointer in genpd PM / Domains: Factorize dev_pm_genpd_set_performance_state() PM / Domains: Propagate performance state updates drivers/base/power/domain.c | 202 ++++++++++++++++++++++++++---------- drivers/opp/core.c | 63 +++++++++++ drivers/opp/of.c | 24 +++-- include/linux/pm_domain.h | 6 ++ include/linux/pm_opp.h | 13 ++- 5 files changed, 242 insertions(+), 66 deletions(-) -- 2.19.1.568.g152ad8e3369a