Abstract: Selective refresh techniques for memory devices are disclosed. In one aspect, a memory device that is used with an application that has frequent repeated read or write commands to certain memory segments may be able to set a flag or similar indication that exempts these certain memory segments from being actively refreshed. By exempting these memory segments from being actively refreshed, these memory segments are continuously available, thereby improving performance. Likewise, because these memory segments are so frequently the subject of a read or write command, these memory segments are indirectly refreshed through the execution of the read or write command.

Abstract: The energy consumed by data transfer in a computing device may be reduced by transferring data that has been encoded in a manner that reduces the number of one “1” data values, the number of signal level transitions, or both. A data destination component of the computing device may receive data encoded in such a manner from a data source component of the computing device over a data communication interconnect, such as an off-chip interconnect. The data may be encoded using minimum Hamming weight encoding, which reduces the number of one “1” data values. The received data may be decoded using minimum Hamming weight decoding. For other computing devices, the data may be encoded using maximum Hamming weight encoding, which increases the number of one “1” data values while reducing the number of zero “0” values, if reducing the number of zero values reduces energy consumption.

Abstract: The energy consumed by data transfer in a computing device may be reduced by transferring data that has been encoded in a manner that reduces the number of one “1” data values, the number of signal level transitions, or both. A data destination component of the computing device may receive data encoded in such a manner from a data source component of the computing device over a data communication interconnect, such as an off-chip interconnect. The data may be encoded using minimum Hamming weight encoding, which reduces the number of one “1” data values. The received data may be decoded using minimum Hamming weight decoding. For other computing devices, the data may be encoded using maximum Hamming weight encoding, which increases the number of one “1” data values while reducing the number of zero “0” values, if reducing the number of zero values reduces energy consumption.

Abstract: In some aspects, the present disclosure provides a method for power management. The method includes receiving, by a power management unit (PMU), signaling indicative of a first plurality of latency durations from a first plurality of clients, each of the first plurality of latency durations corresponding to one of the first plurality of clients, wherein each of the first plurality of clients is configured to utilize a first shared resource for communication of data. In certain aspects, the method also includes selecting, by the PMU, a first latency duration from the first plurality of latency durations based on a determination that the first latency duration is the shortest latency duration of the first plurality of latency durations, and transitioning, by the PMU, the first shared resource from an active state to the first idle state.

Abstract: In some aspects, the present disclosure provides a method for power management. The method includes receiving, by a power management unit (PMU), signaling indicative of a first plurality of latency durations from a first plurality of clients, each of the first plurality of latency durations corresponding to one of the first plurality of clients, wherein each of the first plurality of clients is configured to utilize a first shared resource for communication of data. In certain aspects, the method also includes selecting, by the PMU, a first latency duration from the first plurality of latency durations based on a determination that the first latency duration is the shortest latency duration of the first plurality of latency durations, and transitioning, by the PMU, the first shared resource from an active state to the first idle state.

Abstract: Systems and methods are disclosed method for controlling instantaneous current changes in parallel processors with arrays of parallel computing elements, such as neural processors. An exemplary method comprises monitoring the array of computing elements and determining a transition from a first activity level of the array to a second activity level of the array, such as an idle-to-active or active-to-idle transition. Once a transition is determined, the array is selectively controlled to minimize the instantaneous current change from the transition from the first activity level to the second activity level.

Abstract: In certain aspects, an apparatus includes a first power chain, a second power chain, and an enable circuit having an output coupled to an input of the first power chain. The apparatus also includes a multiplexer having a first input coupled to an output of the first power chain, a second input coupled to the output of the enable circuit, and an output coupled to an input of the second power chain, wherein the multiplexer is configured to receive a select signal, and couple the first input or the second input to the output of the multiplexer based on the select signal.

Abstract: Various aspects include methods for managing memory subsystems on a computing device. Various aspect methods may include determining a period of time to force a memory subsystem on the computing device into a low power mode, inhibiting memory access requests to the memory subsystem during the determined period of time, forcing the memory subsystem into the low power mode for the determined period of time, and executing the memory access requests to the memory subsystem inhibited during the determined period of time in response to expiration of the determined period of time.

Abstract: In certain aspects, an apparatus includes a first power chain, a second power chain, and an enable circuit having an output coupled to an input of the first power chain. The apparatus also includes a multiplexer having a first input coupled to an output of the first power chain, a second input coupled to the output of the enable circuit, and an output coupled to an input of the second power chain, wherein the multiplexer is configured to receive a select signal, and couple the first input or the second input to the output of the multiplexer based on the select signal.

Abstract: Systems and methods are disclosed method for controlling instantaneous current changes in parallel processors with arrays of parallel computing elements, such as neural processors. An exemplary method comprises monitoring the array of computing elements and determining a transition from a first activity level of the array to a second activity level of the array, such as an idle-to-active or active-to-idle transition. Once a transition is determined, the array is selectively controlled to minimize the instantaneous current change from the transition from the first activity level to the second activity level.

Abstract: Various aspects include methods for managing memory subsystems on a computing device. Various aspect methods may include determining a period of time to force a memory subsystem on the computing device into a low power mode, inhibiting memory access requests to the memory subsystem during the determined period of time, forcing the memory subsystem into the low power mode for the determined period of time, and executing the memory access requests to the memory subsystem inhibited during the determined period of time in response to expiration of the determined period of time.

Abstract: Various aspects include methods for managing memory subsystems on a computing device. Various aspect methods may include determining a period of time to force a memory subsystem on the computing device into a low power mode, inhibiting memory access requests to the memory subsystem during the determined period of time, forcing the memory subsystem into the low power mode for the determined period of time, and executing the memory access requests to the memory subsystem inhibited during the determined period of time in response to expiration of the determined period of time.

Abstract: An apparatus includes a first circuit and a second circuit sharing an instruction stream. A voltage controller circuit is configured to provide an operation voltage and at least one low-power voltage to the second circuit independent of a supply voltage of the first circuit in response to a sequence of the instruction stream. In another aspect, a method of operating a power management function is presented. The method includes providing an instruction stream for a first circuit and a second circuit and providing selectively an operation voltage and at least one low-power voltage to the second circuit independent of a supply voltage of the first circuit in response to a sequence of the instruction stream.

Abstract: Data retention circuitry, such as at least one integrated circuit (IC), is disclosed herein for power multiplexing with flip-flops having a retention feature. In an example aspect, an IC includes a first power rail and a second power rail. The IC further includes a flip-flop and power multiplexing circuitry. The flip flop includes a master portion and a slave portion. The master portion is coupled to the first power rail for a regular operational mode and for a retention operational mode. The power multiplexing circuitry is configured to couple the slave portion to the first power rail for the regular operational mode and to the second power rail for the retention operational mode.

Abstract: A method for operating an electronic apparatus is provided. The method includes receiving a token, activating a power switch for powering up a core in response to the receiving the token, and outputting the token based on a state of powering up the core. The outputting of the received token is delayed until the state of powering up the core is reached. In one aspect, an electronic apparatus includes a power switch configured to power up to a core is provided. A power-switch control circuit is configured to receive a token, activate the power switch for powering up the core in response to receiving the token, output the received token based on a state of powering up the core. The outputting of the received token is delayed until the state of powering up the core is reached. A plurality of the power-switch control circuits is configured as a ring.

Abstract: Data retention circuitry, such as at least one integrated circuit (IC), is disclosed herein for power multiplexing with flip-flops having a retention feature. In an example aspect, an IC includes a first power rail and a second power rail. The IC further includes a flip-flop and power multiplexing circuitry. The flip flop includes a master portion and a slave portion. The master portion is coupled to the first power rail for a regular operational mode and for a retention operational mode. The power multiplexing circuitry is configured to couple the slave portion to the first power rail for the regular operational mode and to the second power rail for the retention operational mode.

Abstract: An apparatus includes a first circuit and a second circuit sharing an instruction stream. A voltage controller circuit is configured to provide an operation voltage and at least one low-power voltage to the second circuit independent of a supply voltage of the first circuit in response to a sequence of the instruction stream. In another aspect, a method of operating a power management function is presented. The method includes providing an instruction stream for a first circuit and a second circuit and providing selectively an operation voltage and at least one low-power voltage to the second circuit independent of a supply voltage of the first circuit in response to a sequence of the instruction stream.

Abstract: A method for operating an electronic apparatus is provided. The method includes receiving a token, activating a power switch for powering up a core in response to the receiving the token, and outputting the token based on a state of powering up the core. The outputting of the received token is delayed until the state of powering up the core is reached. In one aspect, an electronic apparatus includes a power switch configured to power up to a core is provided. A power-switch control circuit is configured to receive a token, activate the power switch for powering up the core in response to receiving the token, output the received token based on a state of powering up the core. The outputting of the received token is delayed until the state of powering up the core is reached. A plurality of the power-switch control circuits is configured as a ring.