1. Purpose[edit source]
This article explains how to change the CPU operating point (also known as OPP). An operating point corresponds to the frequency of the processor and the voltage that needs to be supplied to sustain it.
It also shows how it is possible to define multiple operating points, allowing the system to jump among them at run time: this is the dynamic voltage and frequency scaling (DVFS). Some cautions are given at the end of the article, to help the user in its DVFS deployment.
2. Hardware side[edit source]
On STM32 Arm® Cortex® MPUs 
products, the Cortex-A core:
- is clocked by PLL1 from the RCC internal peripheral in OpenSTLinux:
the PLL1 output frequency can be directly propagated to the core, or it can go through an intermediate MPUDIV divider - is supplied with
- VDDCPU voltage on STM32MP13x lines and STM32MP25x lines
- VDDCORE voltage on STM32MP15x lines
- VDDCPU voltage on STM32MP13x lines
The part number tells the device maximum supported frequency with associated usage conditions
Otherwise, the frequency must be kept below the nominal frequency:
- 650 MHz for STM32MP1 series
- 1.2 GHz for STM32MP2 series.
3. Software side[edit source]
3.1. Boot time[edit source]
In OpenSTLinux distribution, TF-A BL2 sets the CPU nominal frequency, that is the maximum frequency sustainable with the nominal voltage.
Notice that the processor must receive the nominal voltage during TF-A BL2 execution, whether configuring the STPMIC1 from TF-A BL2 itself or getting it from a discrete power supply, depending on the hardware board configuration.
3.1.1. On STM32MP13x lines [edit source]
3.1.1.1. VDDCPU voltage configuration[edit source]
The example below sets STPMIC1 BUCK1 minimal voltage to 1.25 V, allowing to provide the expected nominal voltage on VDDCPU for the CPU:
&cpu0 {
cpu-supply = <&vddcpu>;
};
...
pmic: stpmic@33 {
vddcpu: buck1 {
regulator-name = "vddcpu";
regulator-min-microvolt = <1250000>;
...
};
3.1.1.2. CPU frequency configuration[edit source]
TF-A is using an algorithm to calculate the PLL1 settings in drivers/st/clk/clk-stm32mp13.c .
3.1.2. On STM32MP15x lines [edit source]
3.1.2.1. VDDCORE voltage configuration[edit source]
The example below sets STPMIC1 BUCK1 minimal voltage to 1.2 V, allowing to provide the expected nominal voltage on VDDCORE for the CPU:
&cpu0 {
cpu-supply = <&vddcore>;
};
...
pmic: stpmic@33 {
vddcore: buck1 {
regulator-name = "vddcore";
regulator-min-microvolt = <1200000>;
...
};
3.1.2.2. CPU frequency configuration[edit source]
TF-A is using an algorithm to calculate the PLL1 settings in drivers/st/clk/stm32mp1_clk.c .
3.1.3. On STM32MP25x lines [edit source]
3.1.3.1. VDDCPU voltage configuration[edit source]
The example below sets STPMIC25 BUCK1 minimal voltage to 0.8 V, allowing to provide the expected nominal voltage on VDDCPU for the CPU:
&cpu0 {
cpu-supply = <&vddcpu>;
};
...
pmic2: stpmic2@33 {
vddcpu: buck1 {
regulator-name = "vddcpu";
regulator-min-microvolt = <800000>;
...
};
3.1.3.2. CPU frequency configuration[edit source]
TF-A is using an algorithm to calculate the PLL1 settings in drivers/st/clk/clk-stm32mp2.c .
3.2. Runtime[edit source]
3.2.1. On STM32MP13x lines [edit source]
3.2.1.1. Overview[edit source]
By default, in OpenSTLinux distribution, the OP-TEE "soc" device tree file defines an OPP table that contains one or several frequency / voltage pair(s).
3.2.1.2. OPP table[edit source]
At runtime, Linux will automatically request to OP-TEE SCMI interface to switch between the available operating points according to the CPU load and thanks to Linux cpufreq framework, configured with "ondemand" governor policy (see admin-guide/pm/cpufreq.html). This feature is called dynamic voltage and frequency scaling (DVFS).
In core/arch/arm/dts/stm32mp131.dtsi , two operating points ([1 GHz ; 1.35 V]; [650 MHz ; 1.25 V]) are defined and the supported OPP are defined with the bitfield opp-supported-hw
- bit 0: 650 MHz capable part number = STM32MP13xA / STM32MP13xC
- bit 1: 1 GHz capable part number = STM32MP13xD / STM32MP13xF
cpu0_opp_table: cpu0-opp-table {
compatible = "operating-points-v2";
opp-650000000 {
opp-hz = /bits/ 64 <650000000>;
opp-microvolt = <1250000>;
opp-supported-hw = <0x3>;
st,opp-default;
};
opp-1000000000 {
opp-hz = /bits/ 64 <1000000000>;
opp-microvolt = <1350000>;
opp-supported-hw = <0x2>;
st,opp-default;
};
};
The PLL1 configurations needed to reach the above frequencies must be described via the st,clk_opp, st,pll@0 and st,pll_vco property in 'rcc' device tree node on OP-TEE device tree.
For example in core/arch/arm/dts/stm32mp135f-dk.dts :
&rcc {
st,pll_vco {
pll1_vco_2000Mhz: pll1-vco-2000Mhz {
src = <CLK_PLL12_HSE>;
divmn = <1 82>;
frac = <0xAAA>;
};
pll1_vco_1300Mhz: pll1-vco-1300Mhz {
src = <CLK_PLL12_HSE>;
divmn = <2 80>;
frac = <0x800>;
};
...
};
/* VCO = 1300.0 MHz => P = 650 (CPU) */
pll1: st,pll@0 {
compatible = "st,stm32mp1-pll";
reg = <0>;
st,pll = <&pll1_cfg1>;
pll1_cfg1: pll1_cfg1 {
st,pll_vco = <&pll1_vco_1300Mhz>;
st,pll_div_pqr = <0 1 1>;
};
pll1_cfg2: pll1_cfg2 {
st,pll_vco = <&pll1_vco_2000Mhz>;
st,pll_div_pqr = <0 1 1>;
};
};
....
st,clk_opp {
st,ck_mpu {
cfg_1 {
hz = < 1000000000 >;
st,clksrc = < CLK_MPU_PLL1P >;
st,pll = < &pll1_cfg2 >;
};
cfg_2 {
hz = < 650000000 >;
st,clksrc = < CLK_MPU_PLL1P >;
st,pll = < &pll1_cfg1 >;
};
};
};
};
Notes:
- OP-TEE selects the highest st,opp-default supported OPP during the cold boot
- This description is valid for cold boot, but also when coming back from Standby low power mode.
3.2.2. On STM32MP15x lines [edit source]
3.2.2.1. Overview[edit source]
By default, in OpenSTLinux distribution, the OP-TEE OS and Linux kernel "soc" device tree file defines an OPP table that contains one or several frequency / voltage pair(s).
3.2.2.2. OPP table[edit source]
At runtime, the Linux kernel will automatically switch between the available operating points according to the CPU load and thanks to Linux cpufreq framework, configured with "ondemand" governor policy (see admin-guide/pm/cpufreq.html). This feature is called dynamic voltage and frequency scaling (DVFS).
In {Linux device tree arch/arm/boot/dts/stm32mp151.dtsi and in OP-TEE device tree core/arch/arm/dts/stm32mp151.dtsi , two operating points ([800 MHz ; 1.35 V] and [400 MHz ; 1.2 V]) are defined for a 800 MHz capable part number (opp-supported-hw = <0x2>), one operating points ([400 MHz ; 1.2 V]) are defined for a 650 MHz capable part number (opp-supported-hw = <0x1>):
cpu0_opp_table: cpu0-opp-table {
compatible = "operating-points-v2";
opp-shared;
opp-400000000 {
opp-hz = /bits/ 64 <400000000>;
opp-microvolt = <1200000>;
opp-supported-hw = <0x2>;
opp-suspend;
};
opp-650000000 {
opp-hz = /bits/ 64 <650000000>;
opp-microvolt = <1200000>;
opp-supported-hw = <0x1>;
};
opp-800000000 {
opp-hz = /bits/ 64 <800000000>;
opp-microvolt = <1350000>;
opp-supported-hw = <0x2>;
};
};
OP-TEE is using an algorithm to calculate the PLL1 settings in core/drivers/clk/clk-stm32mp15.c , pll1_config_from_opp_khz().
Notes:
- This same OPP table must be present in both the BL32 (OP-TEE) and Linux device tree
- OP-TEE configure the OPP selected by supported OPP with st,opp-default and with the higher frequency during the cold boot
- This description is valid for cold boot, but also when coming back from Standby low power mode.
- During cold boot, BL32 (OP-TEE computes and saves the PLL1 settings for all operating points available in the device tree in compliance with the hardware capabilities. These saved parameters are used later to increase the performance of the system-state transitions.
- Selecting OPP frequencies that can be reached with RCC MPUDIV divider and without configuring again the PLL1 is recommended in order to get the fastest switch time between the OPP.
3.2.3. On STM32MP25x lines [edit source]
3.2.3.1. Overview[edit source]
By default, in OpenSTLinux distribution, the OP-TEE "soc" device tree file defines an OPP table that contains one or several frequency / voltage pair(s).
3.2.3.2. OPP table[edit source]
At runtime, Linux will automatically request to OP-TEE SCMI interface to switch between the available operating points according to the CPU load and thanks to Linux cpufreq framework, configured with "schedutil" governor policy (default in drivers/cpufreq/Kconfig for ARM64 see admin-guide/pm/cpufreq.html). This feature is called dynamic voltage and frequency scaling (DVFS).
In core/arch/arm/dts/stm32mp251.dtsi , the two operating points [1.5GHz ; 0.9.1V] and [1.2GHz ; 0.8V] are defined with the associated bit-field opp-supported-hw
- bit 0: supported by 1.2 GHz part number = STM32MP25xA / STM32MP25xC
- bit 1: supported by 1.5 GHz part number = STM32MP25xD / STM32MP25xF
cpu0_opp_table: cpu0-opp-table {
compatible = "operating-points-v2";
opp-1200000000 {
opp-hz = /bits/ 64 <1200000000>;
opp-microvolt = <800000>;
opp-supported-hw = <0x3>;
st,opp-default;
};
opp-1500000000 {
opp-hz = /bits/ 64 <1500000000>;
opp-microvolt = <910000>;
opp-supported-hw = <0x2>;
};
};
The PLL1 configurations needed to reach the above frequencies must be described via the st,clk_opp, st,pll-1 and st,pll_vco property in 'rcc' device tree node on OP-TEE device tree.
For example in core/arch/arm/dts/stm32mp257f-dk-ca35tdcid-rcc.dtsi , generated by STM32MP25 clock tree:
&rcc {
pll1: st,pll-1 {
st,pll = <&pll1_cfg_1200Mhz>;
pll1_cfg_1200Mhz: pll1-cfg-1200Mhz {
cfg = <30 1 1 1>;
src = <MUX_CFG(MUX_MUXSEL5, MUXSEL_HSE)>;
};
pll1_cfg_1500Mhz: pll1-cfg-1500Mhz {
cfg = <75 2 1 1>;
src = <MUX_CFG(MUX_MUXSEL5, MUXSEL_HSE)>;
};
};
And in core/arch/arm/dts/stm32mp257f-dk.dts :
&rcc {
st,c1msrd = <2>;
st,clk_opp {
st,ck_cpu1 {
cfg_1 {
hz = <1500000000>;
st,clksrc = <0>;
st,pll = <&pll1_cfg_1500Mhz>;
};
cfg_2 {
hz = <1200000000>;
st,clksrc = <0>;
st,pll = <&pll1_cfg_1200Mhz>;
};
};
};
};
Notes:
- OP-TEE selects the highest st,opp-default supported OPP during the cold boot
- This description is valid for cold boot, but also when coming back from Standby low power mode.
4. Dynamic voltage and frequency scaling (DVFS) caution[edit source]
As stated above, as soon as at least two operating points are defined in the OPP table, Linux kernel will automatically switch between them at runtime thanks to Linux cpufreq framework, configured with "ondemand" (for ARM, STM32MP1 series) or "schedutil" (for ARM64, STM32MP2 series) governor policy (see admin-guide/pm/cpufreq.html). It is important to notice that cpufreq framework is only monitoring the CPU load to select the OPP because this can lead to some limitation during use cases where the CPU is not loaded a lot but high reactivity is needed to respect some real time constraints, like interrupt management. If you face some system issue where the CPU reactivity may be the root cause whereas DVFS is enabled, consider doing a trial with "performance" governor (see admin-guide/pm/cpufreq.html).
echo performance > /sys/devices/system/cpu/cpufreq/policy0/scaling_governor
cat /sys/devices/system/cpu/cpufreq/policy0/cpuinfo_cur_freq
5. How to go further[edit source]
- admin-guide/pm/cpufreq.html
- cpu-freq/cpufreq-stats.html
- Guidelines for Lifetime Usage Estimates: AN5438 and AN5729
- Clock device tree configuration and "how to build a clock tree".
Arm® is a registered trademark of Arm Limited (or its subsidiaries) in the US and/or elsewhere. 