| Message ID | 1500279271-15249-6-git-send-email-wxt@rock-chips.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Hi Caesar, Am Montag, 17. Juli 2017, 16:14:31 CEST schrieb Caesar Wang: > As RK3399 had used the Power allocator thermal governor by default, > enabled this to manage thermals by dynamically allocating and limiting > power to devices. > > Also, this patch supported the dynamic-power-coefficient/sustainable_power > and GPU's power model for needed parameters with thermal IPA. > > The Thermal power allocator governor works optimatly with two passive trip > points, for the better performance we will use the trip-point0 with 70 > degree above which the governor control starts operating and trip-point1 > with 85 degree is the target temperature by controlling. > > Signed-off-by: Caesar Wang <wxt@rock-chips.com> > > --- > > Changes in v2: > - foo@ will produce warnings when used without reg property. > - update the commit to explain the two passive trip points changed. > > arch/arm64/boot/dts/rockchip/rk3399.dtsi | 62 +++++++++++++++----------------- > 1 file changed, 29 insertions(+), 33 deletions(-) > > diff --git a/arch/arm64/boot/dts/rockchip/rk3399.dtsi b/arch/arm64/boot/dts/rockchip/rk3399.dtsi > index 77d67cb..6d8a5eb 100644 > --- a/arch/arm64/boot/dts/rockchip/rk3399.dtsi > +++ b/arch/arm64/boot/dts/rockchip/rk3399.dtsi > @@ -147,7 +147,7 @@ > enable-method = "psci"; > #cooling-cells = <2>; /* min followed by max */ > clocks = <&cru ARMCLKB>; > - dynamic-power-coefficient = <100>; > + dynamic-power-coefficient = <436>; > }; > > cpu_b1: cpu@101 { > @@ -156,7 +156,7 @@ > reg = <0x0 0x101>; > enable-method = "psci"; > clocks = <&cru ARMCLKB>; > - dynamic-power-coefficient = <100>; > + dynamic-power-coefficient = <436>; Adjusting the coefficients should be a separate patch and the commit message should explain how they were calculated and why they are the exacter ones over the old values. > }; > }; > > @@ -690,24 +690,25 @@ > }; > > thermal_zones: thermal-zones { > - cpu_thermal: cpu { > + soc_thermal: soc-thermal { > polling-delay-passive = <100>; > polling-delay = <1000>; > + sustainable-power = <1000>; > > thermal-sensors = <&tsadc 0>; > > trips { > - cpu_alert0: cpu_alert0 { > + threshold: trip-point0 { > temperature = <70000>; > hysteresis = <2000>; > type = "passive"; > }; > - cpu_alert1: cpu_alert1 { > - temperature = <75000>; > + target: trip-point1 { > + temperature = <85000>; > hysteresis = <2000>; > type = "passive"; > }; > - cpu_crit: cpu_crit { > + soc_crit: soc-crit { > temperature = <95000>; > hysteresis = <2000>; > type = "critical"; > @@ -716,45 +717,31 @@ > > cooling-maps { > map0 { > - trip = <&cpu_alert0>; > + trip = <&target>; still both maps use &target as trip point. Is that intentional and if so, why is the &threshold trip point never referenced? > cooling-device = > - <&cpu_b0 THERMAL_NO_LIMIT THERMAL_NO_LIMIT>; > + <&cpu_l0 THERMAL_NO_LIMIT THERMAL_NO_LIMIT>; > + contribution = <4096>; > }; > map1 { > - trip = <&cpu_alert1>; > + trip = <&target>; > cooling-device = > - <&cpu_l0 THERMAL_NO_LIMIT THERMAL_NO_LIMIT>, > <&cpu_b0 THERMAL_NO_LIMIT THERMAL_NO_LIMIT>; > + contribution = <1024>; > + }; > + map2 { > + trip = <&target>; > + cooling-device = > + <&gpu THERMAL_NO_LIMIT THERMAL_NO_LIMIT>; > + contribution = <4096>; > }; > }; > }; > > - gpu_thermal: gpu { > + gpu_thermal: gpu-thermal { > polling-delay-passive = <100>; > polling-delay = <1000>; > > thermal-sensors = <&tsadc 1>; > - > - trips { > - gpu_alert0: gpu_alert0 { > - temperature = <75000>; > - hysteresis = <2000>; > - type = "passive"; > - }; > - gpu_crit: gpu_crit { > - temperature = <95000>; > - hysteresis = <2000>; > - type = "critical"; > - }; > - }; > - > - cooling-maps { > - map0 { > - trip = <&gpu_alert0>; > - cooling-device = > - <&cpu_b0 THERMAL_NO_LIMIT THERMAL_NO_LIMIT>; > - }; > - }; > }; > }; > > @@ -1451,8 +1438,17 @@ > <GIC_SPI 21 IRQ_TYPE_LEVEL_HIGH 0>; > interrupt-names = "gpu", "job", "mmu"; > clocks = <&cru ACLK_GPU>; > + #cooling-cells = <2>; > power-domains = <&power RK3399_PD_GPU>; > status = "disabled"; > + > + gpu_power_model: power_model { > + compatible = "arm,mali-simple-power-model"; > + static-coefficient = <1079403>; > + dynamic-coefficient = <977>; > + ts = <32000 4700 (-80) 2>; > + thermal-zone = "gpu-thermal"; > + }; You might want to have the gpu thermal work without the power-model-thingy for now, so most likely just drop that gpu-related change for now. Heiko
Hi Heiko, Thanks your comments. 在 2017年07月23日 05:48, Heiko Stuebner 写道: > Hi Caesar, > > Am Montag, 17. Juli 2017, 16:14:31 CEST schrieb Caesar Wang: >> As RK3399 had used the Power allocator thermal governor by default, >> enabled this to manage thermals by dynamically allocating and limiting >> power to devices. >> >> Also, this patch supported the dynamic-power-coefficient/sustainable_power >> and GPU's power model for needed parameters with thermal IPA. >> >> The Thermal power allocator governor works optimatly with two passive trip >> points, for the better performance we will use the trip-point0 with 70 >> degree above which the governor control starts operating and trip-point1 >> with 85 degree is the target temperature by controlling. >> >> Signed-off-by: Caesar Wang <wxt@rock-chips.com> >> >> --- >> >> Changes in v2: >> - foo@ will produce warnings when used without reg property. >> - update the commit to explain the two passive trip points changed. >> >> arch/arm64/boot/dts/rockchip/rk3399.dtsi | 62 +++++++++++++++----------------- >> 1 file changed, 29 insertions(+), 33 deletions(-) >> >> diff --git a/arch/arm64/boot/dts/rockchip/rk3399.dtsi b/arch/arm64/boot/dts/rockchip/rk3399.dtsi >> index 77d67cb..6d8a5eb 100644 >> --- a/arch/arm64/boot/dts/rockchip/rk3399.dtsi >> +++ b/arch/arm64/boot/dts/rockchip/rk3399.dtsi >> @@ -147,7 +147,7 @@ >> enable-method = "psci"; >> #cooling-cells = <2>; /* min followed by max */ >> clocks = <&cru ARMCLKB>; >> - dynamic-power-coefficient = <100>; >> + dynamic-power-coefficient = <436>; >> }; >> >> cpu_b1: cpu@101 { >> @@ -156,7 +156,7 @@ >> reg = <0x0 0x101>; >> enable-method = "psci"; >> clocks = <&cru ARMCLKB>; >> - dynamic-power-coefficient = <100>; >> + dynamic-power-coefficient = <436>; > Adjusting the coefficients should be a separate patch and the > commit message should explain how they were calculated and > why they are the exacter ones over the old values. Okay, i don't know why the dynamic-power-coefficient is 100 for b-cluster before.:-) > > >> }; >> }; >> >> @@ -690,24 +690,25 @@ >> }; >> >> thermal_zones: thermal-zones { >> - cpu_thermal: cpu { >> + soc_thermal: soc-thermal { >> polling-delay-passive = <100>; >> polling-delay = <1000>; >> + sustainable-power = <1000>; >> >> thermal-sensors = <&tsadc 0>; >> >> trips { >> - cpu_alert0: cpu_alert0 { >> + threshold: trip-point0 { >> temperature = <70000>; >> hysteresis = <2000>; >> type = "passive"; >> }; >> - cpu_alert1: cpu_alert1 { >> - temperature = <75000>; >> + target: trip-point1 { >> + temperature = <85000>; >> hysteresis = <2000>; >> type = "passive"; >> }; >> - cpu_crit: cpu_crit { >> + soc_crit: soc-crit { >> temperature = <95000>; >> hysteresis = <2000>; >> type = "critical"; >> @@ -716,45 +717,31 @@ >> >> cooling-maps { >> map0 { >> - trip = <&cpu_alert0>; >> + trip = <&target>; > still both maps use &target as trip point. Is that intentional > and if so, why is the &threshold trip point never referenced? For the power allocator governor, the &threshold trip point just control starts operating, not need for map. For other governor (e.g: step_wise) will need the first trip point. Looks like we have to think about how to support them. > }; > }; > > @@ -1451,8 +1438,17 @@ > <GIC_SPI 21 IRQ_TYPE_LEVEL_HIGH 0>; > interrupt-names = "gpu", "job", "mmu"; > clocks = <&cru ACLK_GPU>; > + #cooling-cells = <2>; > power-domains = <&power RK3399_PD_GPU>; > status = "disabled"; > + > + gpu_power_model: power_model { > + compatible = "arm,mali-simple-power-model"; > + static-coefficient = <1079403>; > + dynamic-coefficient = <977>; > + ts = <32000 4700 (-80) 2>; > + thermal-zone = "gpu-thermal"; > + }; > You might want to have the gpu thermal work without the > power-model-thingy for now, so most likely just drop that > gpu-related change for now. Well, I guess the mediatek/hisilicon/rockchip ARM SoCs should support the gpu's power mode with private kernel, not in the upstream kernel. -Caesar > > > Heiko > > _______________________________________________ > Linux-rockchip mailing list > Linux-rockchip@lists.infradead.org > http://lists.infradead.org/mailman/listinfo/linux-rockchip
diff --git a/arch/arm64/boot/dts/rockchip/rk3399.dtsi b/arch/arm64/boot/dts/rockchip/rk3399.dtsi index 77d67cb..6d8a5eb 100644 --- a/arch/arm64/boot/dts/rockchip/rk3399.dtsi +++ b/arch/arm64/boot/dts/rockchip/rk3399.dtsi @@ -147,7 +147,7 @@ enable-method = "psci"; #cooling-cells = <2>; /* min followed by max */ clocks = <&cru ARMCLKB>; - dynamic-power-coefficient = <100>; + dynamic-power-coefficient = <436>; }; cpu_b1: cpu@101 { @@ -156,7 +156,7 @@ reg = <0x0 0x101>; enable-method = "psci"; clocks = <&cru ARMCLKB>; - dynamic-power-coefficient = <100>; + dynamic-power-coefficient = <436>; }; }; @@ -690,24 +690,25 @@ }; thermal_zones: thermal-zones { - cpu_thermal: cpu { + soc_thermal: soc-thermal { polling-delay-passive = <100>; polling-delay = <1000>; + sustainable-power = <1000>; thermal-sensors = <&tsadc 0>; trips { - cpu_alert0: cpu_alert0 { + threshold: trip-point0 { temperature = <70000>; hysteresis = <2000>; type = "passive"; }; - cpu_alert1: cpu_alert1 { - temperature = <75000>; + target: trip-point1 { + temperature = <85000>; hysteresis = <2000>; type = "passive"; }; - cpu_crit: cpu_crit { + soc_crit: soc-crit { temperature = <95000>; hysteresis = <2000>; type = "critical"; @@ -716,45 +717,31 @@ cooling-maps { map0 { - trip = <&cpu_alert0>; + trip = <&target>; cooling-device = - <&cpu_b0 THERMAL_NO_LIMIT THERMAL_NO_LIMIT>; + <&cpu_l0 THERMAL_NO_LIMIT THERMAL_NO_LIMIT>; + contribution = <4096>; }; map1 { - trip = <&cpu_alert1>; + trip = <&target>; cooling-device = - <&cpu_l0 THERMAL_NO_LIMIT THERMAL_NO_LIMIT>, <&cpu_b0 THERMAL_NO_LIMIT THERMAL_NO_LIMIT>; + contribution = <1024>; + }; + map2 { + trip = <&target>; + cooling-device = + <&gpu THERMAL_NO_LIMIT THERMAL_NO_LIMIT>; + contribution = <4096>; }; }; }; - gpu_thermal: gpu { + gpu_thermal: gpu-thermal { polling-delay-passive = <100>; polling-delay = <1000>; thermal-sensors = <&tsadc 1>; - - trips { - gpu_alert0: gpu_alert0 { - temperature = <75000>; - hysteresis = <2000>; - type = "passive"; - }; - gpu_crit: gpu_crit { - temperature = <95000>; - hysteresis = <2000>; - type = "critical"; - }; - }; - - cooling-maps { - map0 { - trip = <&gpu_alert0>; - cooling-device = - <&cpu_b0 THERMAL_NO_LIMIT THERMAL_NO_LIMIT>; - }; - }; }; }; @@ -1451,8 +1438,17 @@ <GIC_SPI 21 IRQ_TYPE_LEVEL_HIGH 0>; interrupt-names = "gpu", "job", "mmu"; clocks = <&cru ACLK_GPU>; + #cooling-cells = <2>; power-domains = <&power RK3399_PD_GPU>; status = "disabled"; + + gpu_power_model: power_model { + compatible = "arm,mali-simple-power-model"; + static-coefficient = <1079403>; + dynamic-coefficient = <977>; + ts = <32000 4700 (-80) 2>; + thermal-zone = "gpu-thermal"; + }; }; pinctrl: pinctrl {
As RK3399 had used the Power allocator thermal governor by default, enabled this to manage thermals by dynamically allocating and limiting power to devices. Also, this patch supported the dynamic-power-coefficient/sustainable_power and GPU's power model for needed parameters with thermal IPA. The Thermal power allocator governor works optimatly with two passive trip points, for the better performance we will use the trip-point0 with 70 degree above which the governor control starts operating and trip-point1 with 85 degree is the target temperature by controlling. Signed-off-by: Caesar Wang <wxt@rock-chips.com> --- Changes in v2: - foo@ will produce warnings when used without reg property. - update the commit to explain the two passive trip points changed. arch/arm64/boot/dts/rockchip/rk3399.dtsi | 62 +++++++++++++++----------------- 1 file changed, 29 insertions(+), 33 deletions(-)