Any entity involved in power management is referred to as a node. The following sections describe how the power management framework (PMF) works with slave nodes allocated to the APU and the RPU.
Generally, the APU or the RPU inform the power management controller about their usage of a slave node, by requesting for it. They then inform the power management controller about the capability requirement needed from the slave node. At this point, the power management controller powers up the slave node so that it can be initialized by the APU or the RPU.
Requesting and Releasing Slave Nodes
When a PU requires a slave node, either peripheral or memory, it must request that slave node using the power management API. After the slave node has performed its function and is no longer required, it may be released, allowing the slave node to be powered off.
The following figure shows the call flow for a use-case in which the APU and the RPU are sharing an OCM memory bank, ocm0.
XStatus
XPm_ReleaseNode, because the RPU has also requested the same slave
node. It is after the RPU also releases the ocm0 node that the PMU powers off the
ocm0 memory.Processor Unit Suspend and Resume
To allow a processor unit (PU) to be powered off, as opposed to just entering an idle state, an external entity is required to take care of the power-down and power-up transitions. For the Zynq UltraScale+ MPSoC, the platform management unit (PMU) is the responsible entity for performing all power state changes.
The processor unit (PU) notifies the PMU that a power state transition is being requested. The following figure illustrates the process.
The Self-Suspending a CPU/PU section in Implementing Power Management on a Processor Unit provides more details on the suspend or resume procedure. Each PU usually depends on a number of slave nodes to be able to operate.