DYNAMIC CORE SWITCHING
A system including a first core to execute instructions associated with an application at a first speed based on a first instruction set and a second core to execute the instructions associated with the application at a second speed based on a second instruction set. The first speed is greater than the first speed. The second instruction set is a subset of the first instruction set. A first memory stores an operating system. The operating system includes a kernel that provides services to the application. A core switching module loads into a second memory after the operating system is booted, where the second memory is separate from the first memory, switches execution of the instructions associated with the application between the first core and the second core, and switches the execution of the instructions associated with the application between the first core and the second core transparently to the operating system.
This application is a continuation of U.S. application Ser. No. 12/215,760 (now U.S. Pat. No. 9,158,355), filed Jun. 30, 2008, which is a continuation of U.S. application Ser. No. 12/145,660, filed Jun. 25, 2008, which claims the benefit of U.S. Provisional Application No. 60/968,143, filed Aug. 27, 2007; U.S. Provisional Application No. 60/978,936, filed Oct. 10, 2007; U.S. Provisional Application No. 60/981,606, filed Oct. 22, 2007; U.S. Provisional Application No. 61/022,431 filed Jan. 21, 2008; U.S. Provisional Application No. 61/029,476, filed Feb. 18, 2008; U.S. Provisional Application No. 61/049,641, filed May 1, 2008; and U.S. Provisional Application No. 61/058,050, filed Jun. 2, 2008 and is a continuation-in-part of U.S. application Ser. No. 11/523,996, filed Sep. 20, 2006, which claims the benefit of U.S. Provisional Application No. 60/825,368, filed Sep. 12, 2006; U.S. Provisional Application No. 60/823,453, filed Aug. 24, 2006; and U.S. Provisional Application No. 60/822,015, filed Aug. 10, 2006. The disclosures of the above applications are incorporated herein by reference in their entirety.
FIELDThe present disclosure relates to mobile computing devices, and more particularly to dynamically switching cores of multi-core processing systems of mobile computing devices.
BACKGROUNDThe background description provided herein is for the purpose of generally presenting the context of the disclosure. Work of the presently named inventors, to the extent it is described in this background section, as well as aspects of the description that may not otherwise qualify as prior art at the time of filing, are neither expressly nor impliedly admitted as prior art against the present disclosure.
Mobile computing devices such as cellular phones, MP3 players, global positioning system (GPS) devices, etc. are typically powered using both line power and battery power. The mobile computing devices typically include a processor, memory and a display, which consume power during operation. The processor generally executes both simple applications that are less processing-intensive and complex applications that are more processing-intensive. Therefore, the capabilities of the processor such as processing speed are typically selected to match the performance requirements of the most processing-intensive applications that will be executed.
One significant limitation of mobile computing devices relates to the amount of time for which the devices can be operated using batteries without recharging. Using a high-speed processor that meets the performance requirements of the processing-intensive applications generally increases power consumption, which corresponds to a relatively short battery life.
Referring now to
In
In
A system comprises a first asymmetric core, a second asymmetric core, and a core switching module. The first asymmetric core executes an application when the system operates in a first mode and is inactive when the system operates in a second mode. The second asymmetric core executes the application when the system operates in the second mode. The core switching module switches operation of the system between the first mode and the second mode. The core switching module selectively stops processing of the application by the first asymmetric core after receiving a first control signal. The core switching module transfers a first state of the first asymmetric core to the second asymmetric core. The second asymmetric core resumes executing the application in the second mode.
In another feature, the first control signal indicates that interrupts are disabled.
In another feature, the second asymmetric core executes instructions without instruction translation when the second asymmetric core resumes executing the application during the second mode.
In other features, the system further comprises an operating system (OS) that provides services to the application. The core switching module switches execution of the application between the first asymmetric core and the second asymmetric core transparently to the OS.
In other features, a first maximum speed of the first asymmetric core is greater than a second maximum speed of the second asymmetric core. The first asymmetric core operates at frequencies greater than a predetermined frequency. The second asymmetric core operates at frequencies less than the predetermined frequency. A first maximum operating power level of the first asymmetric core is greater than a second maximum operating power level of the second asymmetric core.
In other features, the first asymmetric core uses a first instruction set architecture (ISA). The second asymmetric core uses a second ISA. The first ISA is compatible with the second ISA. A first set of instructions of the first ISA is a superset of a second set of instructions of the second ISA. The first set includes more instructions than the second set.
In other features, the OS comprises a kernel. The core switching module executes above a level of the kernel.
In other features, the system further comprises a hypervisor module. The core switching module is integrated with the hypervisor module.
In other features, the core switching module saves the first state when the core switching module selectively stops processing of the application by the first asymmetric core. The core switching module powers up the second asymmetric core and initializes the second asymmetric core using the first state. The interrupts are enabled after the second asymmetric core resumes executing the application. The core switching module shuts down the first asymmetric core when the second asymmetric core powers up. One of no power and standby power is supplied to the first asymmetric core after the first asymmetric core is shut down.
In other features, the system further comprises a level-2 (L2) cache that communicates with the first asymmetric core. One of no power and standby power is supplied to the L2 cache after the first asymmetric core is shut down.
In other features, the core switching module initializes the first asymmetric core using the first state when the second asymmetric core fails to power up. The first asymmetric core resumes executing the application in the first mode and interrupts are enabled.
In other features, the core switching module switches operation of the system between the second mode and the first mode. The core switching module selectively stops processing of the application by the second asymmetric core after receiving the first control signal. The core switching module transfers a second state of the second asymmetric core to the first asymmetric core. The first asymmetric core resumes executing the application in the first mode. The first control signal indicates that interrupts are disabled.
In another feature, instructions are executed without instruction translation when the first asymmetric core resumes executing the application during the first mode.
In other features, the system further comprises an operating system (OS) that provides services to the application. The core switching module switches execution of the application between the second asymmetric core and the first asymmetric core transparently to the OS.
In other features, the core switching module saves the second state when the core switching module selectively stops processing of the application by the second asymmetric core. The core switching module powers up the first asymmetric core and initializes the first asymmetric core using the second state. The interrupts are enabled after the first asymmetric core resumes executing the application.
In other features, the system further comprises a level-2 (L2) cache that communicates with the first asymmetric core. Power is supplied to the L2 cache after the first asymmetric core powers up.
In other features, the core switching module shuts down the second asymmetric core when the first asymmetric core powers up. One of no power and standby power is supplied to the second asymmetric core after the second asymmetric core is shut down.
In other features, the core switching module initializes the second asymmetric core using the second state when the first asymmetric core fails to power up. The second asymmetric core resumes executing the application in the second mode, and the interrupts are enabled.
In another feature, the system further comprises a core profile module that generates a second control signal based on at least one of core utilization, resource utilization, and performance of the application.
In other features, the system further comprises a core change sequence (CCS) module that initiates a CCS based on the second control signal. The core switching module switches execution of the application between one of the first and second asymmetric cores and another of the first and second asymmetric cores based on the CCS.
In other features, the CCS module initiates the CCS when the core utilization of the second asymmetric core by the application is greater than or equal to a first predetermined threshold. The CCS module initiates the CCS when the core utilization of the second asymmetric core by the application is greater than or equal to a second predetermined threshold for a first predetermined time period. The CCS module initiates the CCS when an anticipated core utilization of the second asymmetric core is greater than or equal to a third predetermined threshold. The anticipated core utilization is determined based on at least one of a type of the application and a history of execution of the application. The CCS module initiates the CCS when the core utilization of the first asymmetric core by the application is less than or equal to a fourth predetermined threshold. The CCS module initiates the CCS when the core utilization of the first asymmetric core by the application is less than or equal to a fifth predetermined threshold for a second predetermined time period.
In another feature, the CCS module initiates the CCS based on at least one of a type of the application, a number of applications, and a type of instruction executed by one of the first and second asymmetric cores.
In another feature, the core profile module generates the second control signal based on a number of times execution of applications is switched between one of the first and second asymmetric cores and another of the first and second asymmetric cores.
In other features, the system further comprises a power control module that controls power consumption of the system and that generates a third control signal based on the second control signal and the power consumption. The system further comprises a core change sequence (CCS) module that initiates a CCS based on the third control signal. The core switching module switches execution of the application between one of the first and second asymmetric cores and another of the first and second asymmetric cores based on the CCS.
In another feature, the system further comprises a frequency change sequence (FCS) module that initiates a FCS based on the second control signal and that selects an operating frequency of at least one of the first and second asymmetric cores based on the FCS.
In another feature, the system further comprises a voltage change sequence (VCS) module that initiates a VCS based on the second control signal and that selects a supply voltage of at least one of the first and second asymmetric cores based on the VCS.
In another feature, the system further comprises a power management module that disables the interrupts when the system switches operation between the first mode and the second mode and that disables the interrupts when the system switches operation between the second mode and the first mode.
In other features, the system further comprises a plurality of the first asymmetric core. The core switching module selectively activates and deactivates more than one of the plurality of the first asymmetric core based on the CCS when the system operates in the first mode.
In another feature, the system further comprises a glue logic module that selectively communicates with the core switching module, that receives interrupts, that receives first signals from the application, and that routes the interrupts and the first signals to the one of the first and second asymmetric cores activated by the core switching module.
In another feature, a system-on-chip (SOC) comprises the system.
In another feature, a system-in-package (SIP) comprises the system.
In still other features, a method comprises executing an application using a first asymmetric core when operating in a first mode. The first asymmetric core is inactive when operating in a second mode. The method further comprises switching operation between the first mode and the second mode using a core switching module. The method further comprises selectively stopping processing of the application by the first asymmetric core using the core switching module after receiving a first control signal. The method further comprises transferring a first state of the first asymmetric core to the second asymmetric core using the core switching module and resuming execution of the application in the second mode using the second asymmetric core.
In another feature, the method further comprises disabling interrupts and indicating via the first control signal that the interrupts are disabled.
In another feature, the method further comprises executing instructions without instruction translation when the second asymmetric core resumes executing the application during the second mode.
In another feature, the method further comprises providing services to the application via an operating system (OS) and switching execution of the application between the first asymmetric core and the second asymmetric core transparently to the OS using the core switching module.
In other features, the method further comprises selectively operating the first asymmetric core at a first maximum speed and selectively operating the second asymmetric core at a second maximum speed. The first maximum speed is greater than the second maximum speed. The method further comprises selectively operating the first asymmetric core at a first maximum operating power level and selectively operating the second asymmetric core at a second maximum operating power level. The first maximum operating power level is greater than the second maximum operating power level. The method further comprises selectively operating the first asymmetric core at frequencies greater than a predetermined frequency and selectively operating the second asymmetric core at frequencies less than the predetermined frequency.
In other features, the method further comprises operating the first asymmetric core using a first instruction set architecture (ISA) and operating the second asymmetric core using a second ISA. The first ISA is compatible with the second ISA. The method further comprises operating the first asymmetric core using a first set of instructions of the first ISA and operating the second asymmetric core using a second set of instructions of the second ISA. The first set is a superset of the second set. The first set includes more instructions than the second set.
In another feature, the method further comprises providing a kernel of the OS and executing the core switching module above a level of the kernel.
In another feature, the method further comprises providing a hypervisor module and integrating the core switching module with a hypervisor module.
In other features, the method further comprises saving the first state when the core switching module selective stops processing of the application by the first asymmetric core. The method further comprises powering up the second asymmetric core and initializing the second asymmetric core using the first state. The method further comprises enabling the interrupts after the second asymmetric core resumes executing the application. The method further comprises shutting down the first asymmetric core when the second asymmetric core powers up and supplying one of no power and standby power to the first asymmetric core after the first asymmetric core is shut down.
In other features, the method further comprises providing a level-2 (L2) cache. The first asymmetric core communicates with the L2 cache when the first asymmetric core is active. The method further comprises supplying one of no power and standby power to the L2 cache after the first asymmetric core is shut down.
In another feature, the method further comprises initializing the first asymmetric core using the first state when the second asymmetric core fails to power up, resuming execution of the application in the first mode using the first asymmetric core, and enabling interrupts.
In another feature, the method further comprises switching operation between the second mode and the first mode using a core switching module, selectively stopping processing of the application by the second asymmetric core using the core switching module after receiving the first control signal indicating that interrupts are disabled, transferring a second state of the second asymmetric core to the first asymmetric core using the core switching module, and resuming execution of the application in the first mode using the first asymmetric core.
In another feature, the method further comprises executing instructions without instruction translation when the first asymmetric core resumes executing the application during the first mode.
In another feature, the method further comprises providing services to the application via an operating system (OS) and switching execution of the application between the second asymmetric core and the first asymmetric core transparently to the OS using the core switching module.
In other features, the method further comprises saving the second state when the core switching module selectively stops processing of the application by the second asymmetric core. The method further comprises powering up the first asymmetric core and initializing the first asymmetric core using the second state. The method further comprises enabling the interrupts after the first asymmetric core resumes executing the application.
In other features, the method further comprises providing a level-2 (L2) cache. The first asymmetric core communicates with the L2 cache when the first asymmetric core is active. The method further comprises supplying power to the L2 cache after the first asymmetric core powers up.
In other features, the method further comprises shutting down the second asymmetric core when the first asymmetric core powers up and supplying one of no power and standby power to the second asymmetric core after the second asymmetric core is shut down.
In another feature, the method further comprises initializing the second asymmetric core using the second state when the first asymmetric core fails to power up, resuming execution of the application in the second mode using the second asymmetric core, and enabling the interrupts.
In another feature, the method further comprises generating a second control signal based on at least one of core utilization, resource utilization, and performance of the application.
In another feature, the method further comprises initiating a core change sequence (CCS) based on the second control signal and switching execution of the application between one of the first and second asymmetric cores and another of the first and second asymmetric cores based on the CCS using the core switching module.
In other features, the method further comprises initiating the CCS when the core utilization of the second asymmetric core by the application is greater than or equal to a first predetermined threshold. The method further comprises initiating the CCS when the core utilization of the second asymmetric core by the application is greater than or equal to a second predetermined threshold for a first predetermined time period. The method further comprises initiating the CCS when the core utilization of the first asymmetric core by the application is less than or equal to a third predetermined threshold. The method further comprises initiating the CCS when the core utilization of the first asymmetric core by the application is less than or equal to a fourth predetermined threshold for a first predetermined time period.
In another feature, the method further comprises determining an anticipated core utilization based on at least one of a type of the application and a history of execution of the application and initiating the CCS when the anticipated core utilization of the second asymmetric core is greater than or equal to a predetermined threshold.
In another feature, the method further comprises initiating the CCS based on at least one of a number of applications executed by one of the first and second asymmetric cores, a type of the application executed by one of the first and second asymmetric cores, and a type of instruction executed by one of the first and second asymmetric cores.
In another feature, the method further comprises generating the second control signal based on a number of times execution of applications is switched between one of the first and second asymmetric cores and another of the first and second asymmetric cores.
In another feature, the method further comprises controlling power consumption, generating a third control signal based on the second control signal and the power consumption, initiating a core change sequence (CCS) based on the third control signal, and switching execution of the application between one of the first and second asymmetric cores and another of the first and second asymmetric cores based on the CCS using the core switching module.
In another feature, the method further comprises initiating a frequency change sequence (FCS) based on the second control signal and selecting an operating frequency of at least one of the first and second asymmetric cores based on the FCS.
In another feature, the method further comprises initiating a voltage change sequence (VCS) based on the second control signal and selecting a supply voltage of at least one of the first and second asymmetric cores based on the VCS.
In another feature, the method further comprises providing a plurality of the first asymmetric core, and selectively activating and deactivating more than one of the plurality of the first asymmetric core based on the CCS when operating in the first mode using the core switching module.
In another feature, the method further comprises providing a glue logic module, selectively communicating with the core switching module using the glue logic module, receiving interrupts via the glue logic module, receiving first signals from the application via the glue logic module, and routing the interrupts and the first signals to the one of the first and second asymmetric cores activated by the core switching module using the glue logic module.
In another feature, the method further comprises providing the first and second asymmetric cores and the core switching module in one of a system-on-chip (SOC) and a system-in-package (SIP).
In still other features, a system comprises first core means for executing an application when the system operates in a first mode. The first asymmetric core means is inactive when the system operates in a second mode. The system further comprises second asymmetric core means for executing the application when the system operates in the second mode. The system further comprises core switching means for switching operation between the first mode and the second mode. The core switching means selectively stops processing of the application by the first asymmetric core means after receiving a first control signal. The core switching means transfers a first state of the first asymmetric core means to the second asymmetric core means. The second asymmetric core means resumes executing the application in the second mode.
In another feature, the first control signal indicates that interrupts are disabled.
In another feature, the second asymmetric core means executes instructions without instruction translation when the second asymmetric core means resumes executing the application during the second mode.
In other features, the system further comprises an operating system (OS) that provides services to the application. The core switching module switches execution of the application between the first asymmetric core means and the second asymmetric core means transparently to the OS.
In other features, a first maximum speed of the first asymmetric core means is greater than a second maximum speed of the second asymmetric core means. The first asymmetric core means operates at frequencies greater than a predetermined frequency. The second asymmetric core means operates at frequencies less than the predetermined frequency. A first maximum operating power level of the first asymmetric core means is greater than a second maximum operating power level of the second asymmetric core means.
In other features, the first asymmetric core means uses a first instruction set architecture (ISA). The second asymmetric core means uses a second ISA. The first ISA is compatible with the second ISA. A first set of instructions of the first ISA is a superset of a second set of instructions of the second ISA. The first set includes more instructions than the second set.
In other features, the OS comprises a kernel. The core switching means executes above a level of the kernel.
In other features, the system further comprises hypervisor means for switching cores. The core switching means is integrated with the hypervisor means.
In other features, the core switching means saves the first state when the core switching means selectively stops processing of the application by the first asymmetric core means. The core switching means powers up the second asymmetric core means and initializes the second asymmetric core means using the first state. The interrupts are enabled after the second asymmetric core means resumes executing the application. The core switching means shuts down the first asymmetric core means when the second asymmetric core means powers up. One of no power and standby power is supplied to the first asymmetric core means after the first asymmetric core means is shut down.
In other features, the system further comprises level-2 (L2) cache means for caching instructions and data of the first asymmetric core means. One of no power and standby power is supplied to the L2 cache means after the first asymmetric core means is shut down.
In other features, the core switching means initializes the first asymmetric core means using the first state when the second asymmetric core means fails to power up. The first asymmetric core means resumes executing the application in the first mode and interrupts are enabled.
In other features, the core switching means switches operation of the system between the second mode and the first mode. The core switching means selectively stops processing of the application by the second asymmetric core means after receiving the first control signal indicating that interrupts are disabled. The core switching means transfers a second state of the second asymmetric core means to the first asymmetric core means. The first asymmetric core means resumes executing the application in the first mode.
In another feature, instructions are executed without instruction translation when the first asymmetric core means resumes executing the application during the first mode.
In other features, the system further comprises an operating system (OS) that provides services to the application. The core switching means switches execution of the application between the second asymmetric core means and the first asymmetric core means transparently to the OS.
In other features, the core switching means saves the second state when the core switching means selectively stops processing of the application by the second asymmetric core means. The core switching means powers up the first asymmetric core means and initializes the first asymmetric core means using the second state. The interrupts are enabled after the first asymmetric core means resumes executing the application.
In other features, the system further comprises level-2 (L2) cache means for caching instructions and data of the first asymmetric core means. Power is supplied to the L2 cache means after the first asymmetric core means powers up.
In other features, the core switching means shuts down the second asymmetric core means when the first asymmetric core means powers up. One of no power and standby power is supplied to the second asymmetric core means after the second asymmetric core means is shut down.
In other features, the core switching means initializes the second asymmetric core means using the second state when the first asymmetric core means fails to power up. The second asymmetric core means resumes executing the application in the second mode, and the interrupts are enabled.
In another feature, the system further comprises core profile means for generating a second control signal based on at least one of core utilization, resource utilization, and performance of the application.
In other features, the system further comprises core change sequence (CCS) means for initiating a CCS based on the second control signal. The core switching means switches execution of the application between one of the first and second asymmetric core means and another of the first and second asymmetric core means based on the CCS.
In other features, the CCS means initiates the CCS when the core utilization of the second asymmetric core means by the application is greater than or equal to a first predetermined threshold. The CCS means initiates the CCS when the core utilization of the second asymmetric core means by the application is greater than or equal to a second predetermined threshold for a first predetermined time period. The CCS means initiates the CCS when an anticipated core utilization of the second asymmetric core means is greater than or equal to a third predetermined threshold. The anticipated core utilization is determined based on at least one of a type of the application and a history of execution of the application. The CCS means initiates the CCS when the core utilization of the first asymmetric core means by the application is less than or equal to a fourth predetermined threshold. The CCS means initiates the CCS when the core utilization of the first asymmetric core means by the application is less than or equal to a fifth predetermined threshold for a second predetermined time period.
In another feature, the CCS means initiates the CCS based on at least one of a type of the application, a number of applications, and a type of instruction that is executed by one of the first and second asymmetric core means.
In another feature, the core profile means generates the second control signal based on a number of times execution of applications is switched between one of the first and second asymmetric core means and another of the first and second asymmetric core means.
In others feature, the system further comprises power control means for controlling power consumption of the system and for generating a third control signal based on the second control signal and the power consumption. The system further comprises core change sequence (CCS) means that initiates a CCS based on the third control signal. The core switching means switches execution of the application between one of the first and second asymmetric core means and another of the first and second asymmetric core means based on the CCS.
In another feature, the system further comprises frequency change sequence (FCS) means for initiating a FCS based on the second control signal and for selecting an operating frequency of at least one of the first and second asymmetric core means based on the FCS.
In another feature, the system further comprises voltage change sequence (VCS) means for initiating a VCS based on the second control signal and for selecting a supply voltage of at least one of the first and second asymmetric core means based on the VCS.
In another feature, the system further comprises power management means for disabling the interrupts when the system switches the operation between the first mode and the second mode and for disabling the interrupts when the system switches the operation between the second mode and the first mode.
In other features, the system further comprises a plurality of the first asymmetric core means. The core switching module selectively activates and deactivates more than one of the plurality of the first asymmetric core means based on the CCS when the system operates in the first mode.
In other features, the system further comprises glue logic means for selectively communicating with the core switching module. The glue logic means receives interrupts, receives first signals from the application, and routes the interrupts and the first signals to the one of the first and second asymmetric core means activated by the core switching module.
In another feature, a system-on-chip (SOC) comprises the system.
In another feature, a system-in-package (SIP) comprises the system.
In still other features, a computer program stored on a computer readable medium and executed by a processor comprises executing an application using a first asymmetric core when operating in a first mode. The first asymmetric core is inactive when operating in a second mode. The computer program further comprises switching operation between the first mode and the second mode using a core switching module. The computer program further comprises selectively stopping processing of the application by the first asymmetric core using the core switching module after receiving a first control signal. The computer program further comprises transferring a first state of the first asymmetric core to the second asymmetric core using the core switching module and resuming execution of the application in the second mode using the second asymmetric core.
In another feature, the computer program further comprises disabling interrupts and indicating via the first control signal that the interrupts are disabled.
In another feature, the computer program further comprises executing instructions without instruction translation when the second asymmetric core resumes executing the application during the second mode.
In another feature, the computer program further comprises providing services to the application via an operating system (OS) and switching execution of the application between the first asymmetric core and the second asymmetric core transparently to the OS using the core switching module.
In other features, the computer program further comprises selectively operating the first asymmetric core at a first maximum speed and selectively operating the second asymmetric core at a second maximum speed. The first maximum speed is greater than the second maximum speed. The computer program further comprises selectively operating the first asymmetric core at a first maximum operating power level and selectively operating the second asymmetric core at a second maximum operating power level. The first maximum operating power level is greater than the second maximum operating power level. The computer program further comprises selectively operating the first asymmetric core at frequencies greater than a predetermined frequency and selectively operating the second asymmetric core at frequencies less than the predetermined frequency.
In other features, the computer program further comprises operating the first asymmetric core using a first instruction set architecture (ISA) and operating the second asymmetric core using a second ISA. The first ISA is compatible with the second ISA. The computer program further comprises operating the first asymmetric core using a first set of instructions of the first ISA and operating the second asymmetric core using a second set of instructions of the second ISA. The first set is a superset of the second set. The first set includes more instructions than the second set.
In another feature, the computer program further comprises providing a kernel of the OS and executing the core switching module above a level of the kernel.
In another feature, the computer program further comprises providing a hypervisor module and integrating the core switching module with a hypervisor module.
In other features, the computer program further comprises saving the first state when the core switching module selective stops processing of the application by the first asymmetric core. The computer program further comprises powering up the second asymmetric core and initializing the second asymmetric core using the first state. The computer program further comprises enabling the interrupts after the second asymmetric core resumes executing the application. The computer program further comprises shutting down the first asymmetric core when the second asymmetric core powers up and supplying one of no power and standby power to the first asymmetric core after the first asymmetric core is shut down.
In other features, the computer program further comprises providing a level-2 (L2) cache. The first asymmetric core communicates with the L2 cache when the first asymmetric core is active. The computer program further comprises supplying one of no power and standby power to the L2 cache after the first asymmetric core is shut down.
In another feature, the computer program further comprises initializing the first asymmetric core using the first state when the second asymmetric core fails to power up, resuming execution of the application in the first mode using the first asymmetric core, and enabling interrupts.
In another feature, the computer program further comprises switching operation between the second mode and the first mode using a core switching module, selectively stopping processing of the application by the second asymmetric core using the core switching module after receiving the first control signal indicating that interrupts are disabled, transferring a second state of the second asymmetric core to the first asymmetric core using the core switching module, and resuming execution of the application in the first mode using the first asymmetric core.
In another feature, the computer program further comprises executing instructions without instruction translation when the first asymmetric core resumes executing the application during the first mode.
In another feature, the computer program further comprises providing services to the application via an operating system (OS) and switching execution of the application between the second asymmetric core and the first asymmetric core transparently to the OS using the core switching module.
In other features, the computer program further comprises saving the second state when the core switching module selectively stops processing of the application by the second asymmetric core. The computer program further comprises powering up the first asymmetric core and initializing the first asymmetric core using the second state. The computer program further comprises enabling the interrupts after the first asymmetric core resumes executing the application.
In other features, the computer program further comprises providing a level-2 (L2) cache. The first asymmetric core communicates with the L2 cache when the first asymmetric core is active. The computer program further comprises supplying power to the L2 cache after the first asymmetric core powers up.
In other features, the computer program further comprises shutting down the second asymmetric core when the first asymmetric core powers up and supplying one of no power and standby power to the second asymmetric core after the second asymmetric core is shut down.
In another feature, the computer program further comprises initializing the second asymmetric core using the second state when the first asymmetric core fails to power up, resuming execution of the application in the second mode using the second asymmetric core, and enabling the interrupts.
In another feature, the computer program further comprises generating a second control signal based on at least one of core utilization, resource utilization, and performance of the application.
In another feature, the computer program further comprises initiating a core change sequence (CCS) based on the second control signal and switching execution of the application between one of the first and second asymmetric cores and another of the first and second asymmetric cores based on the CCS using the core switching module.
In other features, the computer program further comprises initiating the CCS when the core utilization of the second asymmetric core by the application is greater than or equal to a first predetermined threshold. The computer program further comprises initiating the CCS when the core utilization of the second asymmetric core by the application is greater than or equal to a second predetermined threshold for a first predetermined time period. The computer program further comprises initiating the CCS when the core utilization of the first asymmetric core by the application is less than or equal to a third predetermined threshold. The computer program further comprises initiating the CCS when the core utilization of the first asymmetric core by the application is less than or equal to a fourth predetermined threshold for a first predetermined time period.
In another feature, the computer program further comprises determining an anticipated core utilization based on at least one of a type of the application and a history of execution of the application and initiating the CCS when the anticipated core utilization of the second asymmetric core is greater than or equal to a predetermined threshold.
In another feature, the computer program further comprises initiating the CCS based on at least one of a number of applications executed by one of the first and second asymmetric cores, a type of the application executed by one of the first and second asymmetric cores, and a type of instruction executed by one of the first and second asymmetric cores.
In another feature, the computer program further comprises generating the second control signal based on a number of times execution of applications is switched between one of the first and second asymmetric cores and another of the first and second asymmetric cores.
In another feature, the computer program further comprises controlling power consumption, generating a third control signal based on the second control signal and the power consumption, initiating a core change sequence (CCS) based on the third control signal, and switching execution of the application between one of the first and second asymmetric cores and another of the first and second asymmetric cores based on the CCS using the core switching module.
In another feature, the computer program further comprises initiating a frequency change sequence (FCS) based on the second control signal and selecting an operating frequency of at least one of the first and second asymmetric cores based on the FCS.
In another feature, the computer program further comprises initiating a voltage change sequence (VCS) based on the second control signal and selecting a supply voltage of at least one of the first and second asymmetric cores based on the VCS.
In another feature, the computer program further comprises providing a plurality of the first asymmetric core, and selectively activating and deactivating more than one of the plurality of the first asymmetric core based on the CCS when operating in the first mode using the core switching module.
In another feature, the computer program further comprises providing a glue logic module, selectively communicating with the core switching module using the glue logic module, receiving interrupts via the glue logic module, receiving first signals from the application via the glue logic module, and routing the interrupts and the first signals to the one of the first and second asymmetric cores activated by the core switching module using the glue logic module.
In still other features, a core switching system comprises a mode switching module, a core activation module, and a state transfer module. The mode switching module receives a switch signal to switch operation between a first mode and a second mode. During the first mode, instructions associated with applications are executed by a first asymmetric core, and a second asymmetric core is inactive. During the second mode, the instructions are executed by the second asymmetric core, and the first asymmetric core is inactive. The core activation module stops processing of the applications by the first asymmetric core after interrupts are disabled. The state transfer module transfers a state of the first asymmetric core to the second asymmetric core. The core activation module allows the second asymmetric core to resume execution of the instructions and enables the interrupts.
In another feature, the second asymmetric core executes the instructions without instruction translation when the second asymmetric core resumes executing the instructions during the second mode.
In another feature, a first maximum speed and a first maximum operating power level of the first asymmetric core are greater than a second maximum speed and a second maximum operating power level of the second asymmetric core, respectively.
In another feature, the core activation module supplies one of no power and standby power to the first asymmetric core when the second asymmetric core resumes execution of the instructions.
In other features, the core switching system further comprises a core profile module that generates an anticipated instruction execution rate for executing the applications based on at least one of a data cache miss rate, an instruction cache miss rate, and instructions per cycle executed by the first asymmetric core. The core profile module generates the switch signal based on the anticipated instruction execution rate.
In other features, the core switching system further comprises a core change sequence (CCS) module that generates a CCS based on the switch signal. The core activation module powers up one of the first and second asymmetric cores and shuts down another of the first and second asymmetric cores based on the CCS.
In still other features, a device comprises memory and a core switching module. The memory stores an operating system (OS) including a kernel that provides services to applications. The core switching module switches operation between a first mode and a second mode. During the first mode, instructions associated with the applications are executed by a first asymmetric core, and a second asymmetric core is inactive. During the second mode, the instructions are executed by the second asymmetric core, and the first asymmetric core is inactive. The core switching module operates at a level above the kernel.
In another feature, the core switching module switches between the first mode and the second mode transparently to the OS.
In other features, the core switching module comprises a mode switching module, a core activation module, and a state transfer module. The mode switching module receives a switch signal to switch operation between the first mode and the second mode. The core activation module stops processing of the applications by the first asymmetric core after interrupts are disabled. The state transfer module transfers a state of the first asymmetric core to the second asymmetric core. The core activation module allows the second asymmetric core to resume execution of the instructions and enables the interrupts.
In another feature, instructions are executed without instruction translation when the second asymmetric core resumes executing the instructions during the second mode.
In another feature, a first maximum speed and a first maximum operating power level of the first asymmetric core are greater than a second maximum speed and a second maximum operating power level of the second asymmetric core, respectively.
In another feature, the core switching module supplies one of no power and standby power to the first asymmetric core when the second asymmetric core resumes execution of the instructions.
In other features, the core switching system further comprises a core profile module that generates an anticipated instruction execution rate for executing the applications based on at least one of a data cache miss rate, an instruction cache miss rate, and instructions per cycle executed by the first asymmetric core. The core switching module switches operation between one of the first and second modes and another of the first and second modes based on the anticipated instruction execution rate.
In other features, the core switching system further comprises a core change sequence (CCS) module that generates a CCS based on at least one of a data cache miss rate, an instruction cache miss rate, and instructions per cycle executed by one of the first and second asymmetric cores. The core switching module powers up one of the first and second asymmetric cores and shuts down another of the first and second asymmetric cores based on the CCS.
In still other features, a method comprises executing instructions associated with applications using a first asymmetric core when a second asymmetric core is inactive during a first mode and executing the instructions using the second asymmetric core when the first asymmetric core is inactive during a second mode. The method further comprises receiving a switch signal to switch operation between the first mode and the second mode. The method further comprises disabling interrupts during the first mode based on the switch signal. The method further comprises stopping processing of the applications by the first asymmetric core after interrupts are disabled. The method further comprises transferring a state of the first asymmetric core to the second asymmetric core, allowing the second asymmetric core to resume execution of the instructions and enabling the interrupts.
In another feature, the method further comprises executing the instructions without instruction translation when the second asymmetric core resumes executing the instructions.
In other features, the method further comprises providing the first asymmetric core having a first maximum speed and a first maximum operating power level and providing the second asymmetric core having a second maximum speed and a second maximum operating power level. The first maximum speed and the first maximum operating power level are greater than the second maximum speed and the second maximum operating power level, respectively.
In another feature, the method further comprises supplying one of no power and standby power to the first asymmetric core when the second asymmetric core resumes execution of the instructions.
In other features, the method further comprises generating an anticipated instruction execution rate for executing the applications based on at least one of a data cache miss rate, an instruction cache miss rate, and instructions per cycle executed by the first asymmetric core. The method further comprises generating the switch signal based on the anticipated instruction execution rate.
In another feature, the method further comprises generating a core change sequence (CCS) based on the switch signal, powering up one of the first and second asymmetric cores based on the CCS, and shutting down another of the first and second asymmetric cores based on the CCS.
In still other features, a method comprises storing an operating system (OS) including a kernel that provides services to applications and executing instructions associated with the applications using a first asymmetric core when a second asymmetric core is inactive during a first mode. The method further comprises executing the instructions using the second asymmetric core when the first asymmetric core is inactive during a second mode. The method further comprises switching operation between the first mode and the second mode using a core switching module and executing the core switching module at a level above the kernel.
In another feature, the method further comprises switching between the first mode and the second mode transparently to the OS.
In another feature, the method further comprises receiving a switch signal to switch operation between the first mode and the second mode, stopping processing of the applications by the first asymmetric core after interrupts are disabled, transferring a state of the first asymmetric core to the second asymmetric core, allowing the second asymmetric core to resume execution of the instructions and enabling the interrupts.
In another feature, the method further comprises executing the instructions without instruction translation when the second asymmetric core resumes executing the instructions during the second mode.
In other features, the method further comprises providing the first asymmetric core having a first maximum speed and a first maximum operating power level and providing the second asymmetric core having a second maximum speed and a second maximum operating power level. The first maximum speed and the first maximum operating power level are greater than the second maximum speed and the second maximum operating power level, respectively.
In another feature, the method further comprises supplying one of no power and standby power to the first asymmetric core when the second asymmetric core resumes execution of the instructions.
In other features, the method further comprises generating an anticipated instruction execution rate for executing the applications based on at least one of a data cache miss rate, an instruction cache miss rate, and instructions per cycle executed by the first asymmetric core. The method further comprises switching operation between one of the first and second modes and another of the first and second modes based on the anticipated instruction execution rate.
In another feature, the method further comprises generating a core change sequence (CCS) based on at least one of a data cache miss rate, an instruction cache miss rate, and instructions per cycle executed by the first asymmetric core, powering up one of the first and second asymmetric cores based on the CCS, and shutting down another of the first and second asymmetric cores based on the CCS.
In still other features, a core switching system comprises mode switching means for switching operating modes. The mode switching means receives a switch signal to switch between a first mode and a second mode. During the first mode, instructions associated with applications are executed by first asymmetric core means for executing the instructions, and second asymmetric core means for executing the instructions is inactive. During the second mode, the instructions are executed by the second asymmetric core means, and the first asymmetric core means is inactive. The system further comprises core activation means for activating the first and second asymmetric core means. The core activation means stops processing of the applications by the first asymmetric core means after interrupts are disabled. The system further comprises state transfer means for transferring a state of the first asymmetric core means to the second asymmetric core means. The core activation means allows the second asymmetric core means to resume execution of the instructions and enables the interrupts.
In another feature, the instructions are executed without instruction translation when the second asymmetric core means resumes executing the instructions during the second mode.
In another feature, a first maximum speed and a first maximum operating power level of the first asymmetric core means are greater than a second maximum speed and a second maximum operating power level of the second asymmetric core means, respectively.
In another feature, the core activation means supplies one of no power and standby power to the first asymmetric core means when the second asymmetric core means resumes execution of the instructions.
In another feature, the core switching system further comprises core profile means for generating an anticipated instruction execution rate for executing the applications based on at least one of a data cache miss rate, an instruction cache miss rate, and instructions per cycle executed by the first asymmetric core means, and for generating the switch signal based on the anticipated instruction execution rate.
In other features, the core switching system further comprises core change sequence (CCS) means for generating a CCS based on the switch signal. The core activation means powers up one of the first and second asymmetric core means and shuts down another of the first and second asymmetric cores based on the CCS.
In still other features, a device comprises memory that stores an operating system (OS) including a kernel that provides services to applications. The device further comprises core switching means for switching operation between a first mode and a second mode. During the first mode, instructions associated with the applications are executed by first asymmetric core means for executing the applications, and second asymmetric core means for executing the applications is inactive. During the second mode, the instructions are executed by the second asymmetric core means, and the first asymmetric core means is inactive. The core switching means operates at a level above the kernel.
In another feature, the core switching means switches between the first mode and the second mode transparently to the OS.
In other features, the core switching means comprises mode switching means for receiving a switch signal to switch operation between the first mode and the second mode. The core switching means further comprises core activation means for stopping processing of the applications by the first asymmetric core means after interrupts are disabled. The core switching means further comprises state transfer means for transferring a state of the first asymmetric core means to the second asymmetric core means. The core activation means allows the second asymmetric core means to resume execution of the instructions and enables the interrupts.
In another feature, the instructions are executed without instruction translation when the second asymmetric core means resumes executing the instructions during the second mode.
In another feature, a first maximum speed and a first maximum operating power level of the first asymmetric core means are greater than a second maximum speed and a second maximum operating power level of the second asymmetric core means, respectively.
In another feature, the core switching means supplies one of no power and standby power to the first asymmetric core means when the second asymmetric core means resumes execution of the instructions.
In other features, the device further comprises core profile means for generating an anticipated instruction execution rate for executing the applications based on at least one of a data cache miss rate, an instruction cache miss rate, and instructions per cycle executed by the first asymmetric core means. The core switching means switches between one of the first and second modes and another of the first and second modes based on the anticipated instruction execution rate.
In other features, the device further comprises core change sequence (CCS) means for generating a CCS based on at least one of a data cache miss rate, an instruction cache miss rate, and instructions per cycle executed by the first asymmetric core means. The core switching means powers up one of the first and second asymmetric core means and shuts down another of the first and second asymmetric core means based on the CCS.
In still other features, a computer program stored on a computer readable medium and executed by a processor comprises executing instructions associated with applications using a first asymmetric core when a second asymmetric core is inactive during a first mode and executing the instructions using the second asymmetric core when the first asymmetric core is inactive during a second mode. The method further comprises receiving a switch signal to switch operation between the first mode and the second mode. The method further comprises disabling interrupts during the first mode based on the switch signal. The method further comprises stopping processing of the applications by the first asymmetric core after interrupts are disabled. The method further comprises transferring a state of the first asymmetric core to the second asymmetric core, allowing the second asymmetric core to resume execution of the instructions and enabling the interrupts.
In another feature, the computer program further comprises executing the instructions using without instruction translation when the second asymmetric core resumes executing the instructions during the second mode.
In other features, the computer program further comprises providing the first asymmetric core having a first maximum speed and a first maximum operating power level and providing the second asymmetric core having a second maximum speed and a second maximum operating power level. The first maximum speed and the first maximum operating power level are greater than the second maximum speed and the second maximum operating power level, respectively.
In another feature, the computer program further comprises supplying one of no power and standby power to the first asymmetric core when the second asymmetric core resumes execution of the instructions.
In other features, the computer program further comprises generating an anticipated instruction execution rate for executing the applications based on at least one of a data cache miss rate, an instruction cache miss rate, and instructions per cycle executed by the first asymmetric core. The computer program further comprises generating the switch signal based on the anticipated instruction execution rate.
In another feature, the computer program further comprises generating a core change sequence (CCS) based on the switch signal, powering up one of the first and second asymmetric cores based on the CCS, and shutting down another of the first and second asymmetric cores based on the CCS.
In still other features, a computer program stored on a computer readable medium and executed by a processor comprises storing an operating system (OS) including a kernel that provides services to applications and executing instructions associated with the applications using a first asymmetric core when a second asymmetric core is inactive during a first mode. The method further comprises executing the instructions using the second asymmetric core when the first asymmetric core is inactive during a second mode. The method further comprises switching operation between the first mode and the second mode using a core switching module and executing the core switching module at a level above the kernel.
In another feature, the computer program further comprises switching between the first mode and the second mode transparently to the OS.
In another feature, the computer program further comprises receiving a switch signal to switch operation between the first mode and the second mode, stopping processing of the applications by the first asymmetric core after interrupts are disabled, transferring a state of the first asymmetric core to the second asymmetric core, allowing the second asymmetric core to resume execution of the instructions and enabling the interrupts.
In another feature, the computer program further comprises executing the instructions without instruction translation when the second asymmetric core resumes executing the instructions during the second mode.
In other features, the computer program further comprises providing the first asymmetric core having a first maximum speed and a first maximum operating power level and providing the second asymmetric core having a second maximum speed and a second maximum operating power level. The first maximum speed and the first maximum operating power level are greater than the second maximum speed and the second maximum operating power level, respectively.
In another feature, the computer program further comprises supplying one of no power and standby power to the first asymmetric core when the second asymmetric core resumes execution of the instructions.
In other features, the computer program further comprises generating an anticipated instruction execution rate for executing the applications based on at least one of a data cache miss rate, an instruction cache miss rate, and instructions per cycle executed by the first asymmetric core. The computer program further comprises switching operation between one of the first and second modes and another of the first and second modes based on the anticipated instruction execution rate.
In another feature, the computer program further comprises generating a core change sequence (CCS) based on at least one of a data cache miss rate, an instruction cache miss rate, and instructions per cycle executed by the first asymmetric core, powering up one of the first and second asymmetric cores based on the CCS, and shutting down another of the first and second asymmetric cores based on the CCS.
Further areas of applicability of the present disclosure will become apparent from the detailed description provided hereinafter. It should be understood that the detailed description and specific examples are intended for purposes of illustration only and are not intended to limit the scope of the disclosure.
The present disclosure will become more fully understood from the detailed description and the accompanying drawings, wherein:
The following description is merely exemplary in nature and is in no way intended to limit the disclosure, its application, or uses. For purposes of clarity, the same reference numbers will be used in the drawings to identify similar elements. As used herein, the phrase at least one of A, B, and C should be construed to mean a logical (A or B or C), using a non-exclusive logical or. It should be understood that steps within a method may be executed in different order without altering the principles of the present disclosure.
As used herein, the term module refers to an Application Specific Integrated Circuit (ASIC), an electronic circuit, a processor or core (shared, dedicated, or group) and memory that execute one or more software or firmware programs, a combinational logic circuit, and/or other suitable components that provide the described functionality.
Most mobile computing devices execute applications that have different levels of complexity and/or processing requirements. Some applications can be efficiently executed using a single-core, low-speed processor while other applications can be efficiently executed using a single-core, high-speed processor. For example only, the low-speed processor may operate at speeds that are less than or equal to 500 MegaHertz (MHz) while the high-speed processor may operate at speeds that are greater than or equal to 1 GigaHertz (GHz). Other speed thresholds may be used.
Most mobile computing devices currently use only one single-core processor. Using the single-core processor to execute all types of applications requires power-performance tradeoffs. For example, using the single-core, low-speed processor may increase power savings and extend battery life at the expense of performance. Some processing-intensive applications may run slowly or not at all. Conversely, using the single-core, high-speed processor may increase performance while reducing battery life since the high-speed processor consumes more power even when executing applications requiring lower processing speed.
According to the present disclosure, both increased power savings and increased performance may be achieved in a mobile device by using a multi-core processing system with a low-speed, low-power (LP) core and a high-speed, high-power (HP) core. The LP core and the HP core may be implemented on the same IC or SOC or as separate ICs that are combined into a system-in-package (SIP). As will be described in further detail below, the multi-core processing system dynamically switches between the LP and HP cores depending on the type of applications being executed, the processing load, and/or other factors described herein.
Referring now to
According to the present disclosure, the asymmetric MP system 10 is operated in a pseudo-symmetric manner by achieving compatibility between the LP core 12 and the HP core 16. Compatibility between the LP and the HP cores 12 and 16 may be achieved in many ways. For example, the LP and the HP cores 12 and 16 may use the same or similar key core registers/interfaces; the LP core 12 may use register mapping; and/or the LP core 12 may use an instruction set that is the same as or a subset of the instruction set of the HP core 16.
Accordingly, an operating system (OS) and applications can run on both the LP and the HP cores 12 and 16. The asymmetric MP system 10 dynamically switches between the LP and the HP cores 12 and 16 depending on the processing load and the types of applications being executed to increase power savings and improve performance as needed. The asymmetric MP system 10 can be used for general purpose processing (GPP).
The present disclosure uses core morphing to optimize power consumption and improve performance. Core morphing includes adaptive and dynamic switching of cores of MP systems. Specifically, low and high power cores are switched securely and transparently to the OS and/or the applications. In other words, the OS and applications may be unaware of which core is currently executing the processing load.
Core morphing can optimize battery life and performance. For example, battery life can be extended using the LP core to run applications when the processing load is low. On the other hand, high performance can be provided using the HP core to run applications when the processing load is high. Both power consumption and performance can be optimized by dynamically switching between the LP and the HP cores as needed.
In core morphing, cores are dynamically enabled (i.e., activated) or disabled (i.e., deactivated) based on the system load. The system load may be partly determined based on resource utilization and performance demanded by the applications. In some implementations, only one core may be active (i.e., executing applications) at a time. When one core is active, other cores may be disabled (i.e., deactivated) to save power. The other cores may be disabled in many ways. For example, the other cores may be put in a standby mode wherein the clock frequency and/or the supply voltage of the other cores may be decreased to values that are lower than when the cores are active. Alternatively, the other cores may be completely shut down by disconnecting the power supply to the cores.
When execution is switched from one core to another, the active core may transparently run the OS and applications. Specifically, when execution is switched from one core to another, the other core executes instructions without translation. Accordingly, the switching of cores is transparent or invisible to the OS and applications, which continue to run as if the system comprises a single core instead of multiple cores. Core morphing may be used in multi-core systems comprising one LP core and one HP core. When applications demand still higher performance than that provided by one HP core, multiple HP cores may be used.
Although dynamic core switching is discussed throughout the disclosure using mobile computing devices as an example, the dynamic core switching can be used in other computing systems including desktop computers, servers, etc. to reduce power consumption.
Before a detailed discussion, a brief overview of the figures is presented.
Referring now to
In the HP mode, the HP core 24 is in the active state and processes threads. The LP core 28 may also operate during the HP mode. In other words, the LP core 28 may be in the active state during all or part of the HP mode. To reduce power, however, the LP core 28 may be operated in the inactive state while the HP core 24 is in the active state.
In the LP mode, the LP core 28 operates in the active state and processes threads and the HP core 24 is in the inactive state. The inactive core may be completely turned off (no power supplied). Alternatively, standby power, which includes lower power levels than active power levels, may be supplied to the core in the inactive state to allow quicker transitions to the active state.
The HP and LP cores 24 and 28 may use the same or a similar operating system. The LP core 28 may use an ISA that is the same as or a subset of the ISA of the HP core 24. The HP and LP cores 24 and 28 may be capable of running the same code and/or applications, albeit at different speeds. The HP and LP cores 24 and 28 may have the same or a similar architecture. The HP core 24 may have more processing capabilities (e.g., pipelines, stages, etc.) than the LP core 28. Both HP and LP cores 24 and 28 may temporarily operate in the active state at the same time when transitioning from the LP mode to the HP mode and from the HP mode to the LP mode.
Memory 40 such as volatile and/or non-volatile memory stores first applications 42-1, 42-2, . . . , and 42-P (collectively first applications 42), second applications 44-1, 44-2, . . . , and 44-Q (collectively second applications 44), and/or third applications 46-1, 46-2, . . . , and 46-R (collectively third applications 46), where P, Q and R are integers greater than zero. The first applications 42 may include applications that require relatively low processing speed during operation and may be executed during LP or HP modes. In other words, the first applications 42 may comprise applications that are generally not very processor intensive. However, if a sufficient number of the first applications 42 are being executed concurrently, the processing system 20 may need to transition to and operate in the HP mode.
The second applications 44 may comprise applications that require medium processing speed (e.g., typically higher than the first applications 42 and lower than the third applications 46 on average). Depending upon the circumstances, the second applications 44 may require low, medium, or high processing speed. The third applications 46 may be applications that tend to require high processing speed and may be best operated in the HP mode. Both HP and LP cores 24 and 28 may be capable of running all three types of applications. Alternately, some applications may be limited to running using only the HP core 24.
A control module 34 may be provided to selectively control transitions between the HP and LP modes. The control module 34 may receive a mode change request signal from another module or device. The control module 34 may monitor the transfer of instructions, states, threads, and/or information relating to the transfer such as registers, states, checkpoints, and/or program counters. Once the transfer is complete, the control module 34 may transition one of the HP and LP cores 24 and 28 to the inactive state.
The processing system 20 may be packaged in a variety of ways. For example, both the HP and LP cores 24 and 28 of the processing system 20 may optionally be implemented as a system on chip (SOC). Alternately, the processing system 20 may be packaged as multiple chips arranged as a system-in-package (SIP) or arranged on a printed circuit board (PCB). Still other packaging methods are contemplated.
Referring now to
The HP core 24 may have a more complex architecture than the LP core 28. For example, the LP core 28 may have a smaller width and/or depth than the HP core 24. The width may be defined by the number of parallel pipelines. The HP core 24 may include PHP parallel pipelines 142 and the LP core 28 may include PLP parallel pipelines 146. In some implementations, PLP may be less than PHP. PLP may be an integer greater than or equal to zero. When PLP=0, the LP core 28 does not include any parallel pipelines. The depth may be defined by the number of stages. The HP core 24 may include SHP stages 144 and the LP core 28 may include SLP stages 148. In some implementations, SLP may be less than SHp. SLP may be an integer greater than or equal to one.
The register file 112 may be shared between the HP core 24 and the LP core 28. Alternatively, each of the HP and the LP cores 24 and 28 may have a register file. The contents of the register file 112 may be transferred from an active core that is soon to be deactivated to the inactive core that is soon to be activated. The register file 112 may use predetermined address locations for registers, checkpoints, and/or program counters. For example, registers, checkpoints, and/or program counters that are used by the HP and/or LP cores 24 and/or 28, respectively, may be stored in the same locations in register file 112. Therefore, the HP core 24 and the LP core 28 can locate a particular register, checkpoint, and/or program counter when transitioning from one core to the other core. The register file 112 may be in addition to register files (not shown) in each of the HP and LP cores 24 and 28, respectively. Threading may include single threading and/or multi-threading.
The control module 114 may selectively control transitions between the HP and LP modes. The control module 114 may receive a mode change request signal from another module or device. The control module 114 may monitor the transfer of instructions, states, threads, and/or information relating to the transfer such as registers, states, checkpoints, and/or program counters. Once the transfer is complete, the control module 114 may transition one of the HP and LP cores 24 and 28 to the inactive state. The HP core 24, the LP core 28, the register file 112, and/or the control module 114 may be implemented as a system on chip (SOC) 130.
The HP core 24 may include cache 26, and the LP core 28 may include cache 30. The cache 26 may be larger in size than the cache 30. The cache 26 may consume more power than the cache 30. For example only, the cache 26 may include four-way cache while the cache 30 may include direct map cache. For example only, the HP core 24 may operate at speeds greater than or equal to 1 GHz while the LP core 28 may operate at speeds less than or equal to 500 MHz.
Referring now to
The performance profile module 116 may monitor the performance of the applications being executed by the HP and the LP cores 24 and 28 and may generate profiles of system loading. Specifically, the performance profile module 116 may dynamically detect the types of workloads (e.g., whether the system load is CPU-bound, memory-bound, or both). Additionally, the performance profile module 116 may detect performance bottlenecks when the clock frequency and/or the supply voltage of the HP and LP cores 24 and 28 are varied.
The idle profile module 118 may monitor CPU utilization and determine a lowest possible clock frequency and/or supply voltage for operating the HP and LP cores 24 and 28. The idle profile module 118 may minimize impact on performance based on workload while saving power for workloads having a high idle duty cycle.
Additionally, the idle profile module 118 may be used to maintain a lower power level in the HP and LP cores 24 and 28 when the HP and LP cores 24 and 28 are inactive, respectively. For example, the idle profile module 118 may operate a power supply to supply a low power level that is lower than an active operating power level (and greater than no power) and that is sufficient to allow the core that is inactive to quickly transition to an active state.
The cache monitoring module 124 may monitor cache miss rates when the HP and LP cores 24 and 28 execute applications. For example, the cache monitoring module 124 may monitor instruction cache miss rate and data cache miss rate of the HP and LP cores 24 and 28.
The core profile module 120 may analyze information generated by the performance profile module 116, the idle profile module 118, and/or the cache monitoring module 124. Based on the information, the core profile module 120 may monitor core utilization of the HP and the LP cores 24 and 28 when the applications are being executed. The core profile module 120 may indicate when the core utilization is at or below certain predetermined thresholds. The core profile module 120 may generate a mode change signal when a switch from the LP mode to the HP mode or from the HP mode to the LP is needed to optimize power consumption and performance. A power management system may utilize the mode change signal in determining whether to change mode as requested.
More specifically, the core profile module 120 may monitor processing speed of the HP and LP cores 24 and 28, respectively. When the HP core 24 begins operating at a first processing speed below a first predetermined threshold, the core profile module 120 may indicate a need to switch operation to the LP core 28. When the LP core 28 begins operating at a second processing speed above a second predetermined threshold, the core profile module 120 may indicate a need to switch operation to the HP core 24.
The first and second predetermined thresholds may be equal or unequal (to provide hysteresis). In other words, the first predetermined threshold may be lower than the second predetermined threshold. The first and second processing speeds may be averaged over first and second periods to prevent switching during momentary changes in the processing speed. The first and second periods can be set to different periods to facilitate or retard switching. The first period, the second period, the first predetermined threshold, and the second predetermined threshold may be preset by programmers or may be used-defined. By doing so, the user may control the power dissipation of the device including the processing system 20. The thread transfer module 122 facilitates transfer of threads between the HP core 24 and the LP core 28 when the cores are switched.
Referring now to
The core profile module 120 may use a single threshold. In other words, the transition to HP core 24 may occur when the CPU usage of the LP core 28 is above predetermined percentage during a predetermined period, the page file usage is greater than a predetermined number of page files, or the thread count is greater than a predetermined number of threads. The core profile module 120 may also combine criteria into more complex conditions. For example only, the core profile module 120 may require the processing speed of the LP core 28 to be above a predetermined percentage of maximum speed during a predetermined period and the input/output loading to be greater than a predetermined value. Hysteresis may be used as well. Still other combinations are contemplated.
Referring now to
If step 306 is equal to LP, control continues with step 328 and determines whether the mode signal is set to the HP mode. If step 328 is true, control continues with step 330 and increases power to the HP core 24 and the cache 26. In step 332, control determines whether the HP core 24 and the cache 26 are ready. In step 338, control transfers instructions, states, and/or threads running on the LP core 28 to the HP core 24.
Referring now to
The LP and HP cores 200 and 204 may also have some or all of the attributes described herein. The LP core 200 may have a smaller core than the HP core 204. In addition to the L1 cache 202 and 206, level 2 (L2) cache 208 may also be provided. The L1 cache 202 and 206 may be integrated with the respective LP and HP cores 200 and 204. The L2 cache 208 may be utilized by one or both of the cores 200 and 204. Since the L2 cache 208 generally tends to have high leakage current and/or power consumption, use of the L2 cache may be restricted to the HP core 204 if desired. In some implementations, the HP core 204 may comprise multiple identical cores.
The LP and HP cores 200 and 204 may execute the same operating system (OS) with different ISAs and can execute the same code. A kernel module 220, which is a component of the OS, manages system resources. For example only, the OS may include Linux® or other suitable operating system. The kernel module 220 provides a lowest level abstraction layer for resources (e.g., the high-speed memory 210 or main memory 228, LP and HP cores 200 and 204, and input/output devices 234) that applications 224 (e.g., applications 224-1, 224-2, . . . and 224-P) control when performing their functions. The kernel module 220 typically provides these resources to processes of the applications 224 through inter-process communication mechanisms and system calls.
During boot up, a boot loader typically starts executing the kernel module 220. The kernel module 220 initializes itself and starts a first process. Subsequently, the kernel module 220 generally executes in response to external events such as system calls used by applications 224 to request services from the kernel module 220 or via interrupts used by hardware components to notify the kernel module 220 of events. The kernel module 220 may also execute an idle process loop when no processes are available to run.
A power management system (PMS) module 242 may be provided to manage power consumption and to initiate or control changes in system operation based on the monitored system loading. A core profile module 240 may monitor system loading such as core loading, input/output loading, application loading, and/or loading based on other system operating parameters.
For example, the PMS module 242 may control switching between different operating frequencies of the LP core 200 based on demand. In other words, the PMS module 242 may handle transitions between one of T different operating speeds of the LP core 200, where T is an integer greater than one, based on system loading. For example, under low load conditions, the LP core 200 may initially operate at speeds below 100 MHz. As system loading increases, the PMS module 242 may need to increase operating speed to 200 MHz. As can be appreciated, operation at higher speeds may also be associated with higher power consumption. The increase in speeds may be accomplished by raising supply voltage Vdd to the LP core 200 and/or to related clocking circuitry. Additional speed steps may be employed to increase the operating speed of the LP core 200 up to a maximum processing speed thereof.
When the LP core 200 reaches its highest operating speed and additional system loading occurs, the core profile module 240, the PMS module 242, and/or the kernel module 220 disable interrupts and generate a system call to a hypervisor module 213 or a core switching module 212 to trigger a transition. The core switching module 212 may be standalone or may be integrated with the hypervisor module 213. The core switching module 212 may transition (switch) the operation from the LP core 200 to the HP core 204 in response to the system call and return control to the kernel module 220 after the transition (switching) is complete.
The term hypervisor generally refers to a virtualization platform that allows multiple operating systems to run at the same time on a host computer. The hypervisor may be used to safely suspend operation of one OS while the other OS is operating. The hypervisor typically operates above the kernel. Alternatively, the hypervisor may operate below the kernel.
In the present disclosure, the term hypervisor refers to a virtualization platform that allows a host device to use multiple asymmetric processors using the same OS. The hypervisor is operated in a trusted zone in contrast to other user applications that are not in a trusted zone. The trusted zone is a safe environment providing code isolation where services can be executed securely without external interrupts. The hypervisor manages transfers between the processors and suspends operation of one asymmetric processor while the other asymmetric processor operates. Alternatively, the hypervisor may be used to run multiple asymmetric processors using multiple OSs.
Depending on whether the core switching module 212 is integrated in the hypervisor module 213, the hypervisor module 213 or the core switching module 212 is loaded into the high-speed memory 210 from RAM or other storage after boot up. The core switching module 212 switches cores in a trusted zone. The PMS module 242 disables interrupts to one of the LP and HP cores 200 and 204 when operation of one core is suspended and handed over to another core. The hypervisor module 213 and/or the core switching module 212 provide hardware abstraction by isolating the cores from the OS and applications and make switching cores transparent to the OS and applications.
The core switching module 212 executes at a level above the kernel module 220. Since the core switching module 212 switches cores transparently to the OS and applications, the OS and applications can run unaltered when the core switching module 212 switches cores.
After the interrupts are disabled by the PMS module 242, the core switching module 212 copies the state of the LP core 200, activates the HP core 204, and initializes the HP core 204 with the state of the LP core 200. After the HP core 204 is initialized, the core switching module 212 deactivates the LP core 200. The HP core 204 resumes executing the application, and the PMS module 242 enables the interrupts.
After executing instructions using the HP core 204, system loading may eventually decrease. When the system loading decreases to a predetermined point, the core switching module 212 may transition operation from the HP core 204 back to the LP core 200. As can be appreciated, systems employing core switching may benefit from the low power consumption of the LP core 200 when relatively low processing speed is required. When high speed processing is desired, the HP core 204 can be used and then turned off when the processing load decreases sufficiently.
Referring now to
If step 256 is true, a control message is sent to the core switching module 212. The core switching module 212 reads the state (pointers, registers, opcodes, operands, program counters (PC), etc.) of the LP core 200 and copies the state or information representative of the state of the LP core 200 into the high-speed memory 210 in step 260. In step 264, the core switching module 212 transitions the HP core 204 to the active state. The core switching module 212 loads the state of the LP core 200 from the high-speed memory 210 into the HP core 204 in step 268. In step 272, the core switching module 212 transitions the LP core 200 to the inactive state. In step 276, the HP core 204 resumes execution at the state loaded into the HP core 204. In other words, operation resumes at the same program counter value where the LP core 200 entered the inactive state.
In step 280, after completing the instructions loaded from the LP core 200, the HP core 204 may begin executing at the application level. In step 288, the core profile module 240 monitors system loading of the HP core 204. In step 292, the core profile module 240 determines whether there is a need to switch to the LP core 200.
If step 292 is false, control returns to step 288. Otherwise if step 292 is true, the core switching module 212 reads the state of the HP core 204 and copies the state or information representative of the state of the HP core 204 into the high-speed memory 210 in step 296. Alternatively, the state may be copied directly to the core to be activated. In step 300, the core switching module 212 transitions the LP core 200 to the active state. In step 306, the core switching module 212 loads the state of the HP core 204 from the high-speed memory 210 into the LP core 200.
In step 310, the HP core 204 transitions to the inactive state. In step 320, the LP core 200 resumes execution at the loaded state. In step 324, after completing the loaded instructions, the LP core 200 executes at the application level, and control returns to step 252.
Referring now to
When an application is designated as the HP type application, an immediate transition to the HP core 204 may occur when the application is launched. HP type applications may be programmed and/or user designated. Alternately, HP applications may be selected based on a battery charge state. In other words, the application may default to HP core operation except when the batteries are low. When the batteries are low, the application may be forced to operate on the LP core with reduced operating speed or not at all. Still other applications may execute exclusively on the HP core 204.
If step 342 is false, control determines in step 348 whether the system loading is greater than a first threshold. If step 348 is false, control returns to step 330. If step 348 is true, control continues with step 352 and transitions to the HP core 204 using the core switching module 212. In step 356, the core profile module 240 monitors the system loading of the HP core 204. In step 360, control determines whether the system loading is less than a second threshold. If step 360 is false, control returns to step 356. If step 360 is true, control transitions to the LP core 200 in step 376, and control returns to step 330.
If step 342 is true, control continues with step 364 and transitions to the HP core 204 using the core switching module 212. In step 368, control determines whether the HP type application is closed or otherwise terminated. If step 368 is false, the HP core 204 continues executing the HP type application. If step 368 is true, control continues with step 372 and determines whether the system loading after closing the HP type application is less than a second speed threshold. If step 372 is false, control returns to step 372. If step 372 is true, control transitions to the LP core 200 in step 376, and control returns to step 330.
Core switching can be transparent to the OS and the applications when the LP core can resemble or look like the HP core to the rest of the hardware and software of the mobile computing device. When the LP core resembles the HP core, threads scheduled for the LP core may be scheduled for the HP core and vice versa without any modification or thread-coloring. In other words, applications may be executed by either core without regard to characteristics of threads of either core. Accordingly, the core switching becomes transparent to the OS and applications.
The LP core can resemble the HP core when the LP and the HP cores have identical or compatible key core registers and interfaces. Examples of key core registers and interfaces include a CPUID register, a debug register, a cache organization & control register, a CP register, a trace macro, a memory management unit (MMU), and a performance management unit (PMU). Additionally, or alternatively, the LP core may implement a register mapping interface and/or a microsequencer to achieve dynamic compatibility with the HP core. Alternatively, or additionally, the instruction set of the LP core may be the same as or a subset of the instruction set of the HP core.
Referring now to
Specifically, in
The mux control module 386 may receive the wakeup exchange signals and may generate a mux control signal based on the wakeup exchange signals. The mux control signal may indicate which one of the LP and HP cores 200 and 204 is active at a time after core switching is complete. Alternatively, the core switching module may control the interrupt control and mux device 382 and the trace macro and debug mux device 384 based on the wakeup exchange signals.
Based on the mux control signal, the interrupt control and mux device 382 may route the masked interrupts and the WFI signals to one of the LP and the HP cores 200 and 204 that is active after cores are switched. The trace macro and debug mux device 384 may route the trace macro and/or debug signals to one of the LP and the HP cores 200 and 204 that is active after cores are switched. Thus, only one of the LP and the HP cores 200 and 204 is active and executes applications at a time.
The LP and HP cores 200 and 204 have symmetric communication capabilities since the glue logic module 380 symmetrically communicates with the LP and HP cores 200 and 204. Specifically, the signals generated by the interrupt control and mux device 382 and the trace macro and debug mux device 384 are symmetrically communicated to the LP and HP cores 200 and 204. Thus, regardless of which of the LP and HP cores 200 and 204 is active at a time, the hardware interfacing and signaling capabilities of the LP and HP cores 200 and 204 are similar. This enables other hardware of the mobile device to communicate with either core without regard or without identifying which core is active.
Additionally, due to the symmetric communication, the programmability of the LP and HP cores 200 and 204 is similar. Specifically, the applications can run transparently on the LP or the HP core 200, 204 that is active without regard or without identifying which core is active. The hardware and software transparency or symmetry provided by the glue logic module 380 enables hardware and software vendors to use, modify, develop, and/or add products normally as if only one core is present.
When the HP core 204 comprises multiple identical cores, the interrupt control and mux device 382 may generate masked interrupts and WFI signals for the respective cores. Additionally, the trace macro and debug mux device 384 may route the trace macro and debug signals to the respective cores.
Referring now to
The power control module 404 may receive inputs related to power/performance from a user via user profiles. For example, the user may specify performance levels for applications regardless of power consumption entailed. Additionally, the power control module 404 may receive inputs from the performance profile module 406, the idle profile module 408, the OS task profile module 410, and the core profile module 412.
The performance profile module 406 may generate outputs indicating current performance and demanded or desired performance for applications. For example, the performance profile module 406 may indicate to the power control module 404 and the core profile module 412 parameters including current core utilization by the applications, cache miss rates, and whether the system load is memory bound, core bound, or I/O bound. The performance profile module 406 may generate the outputs partly based on inputs received from an OS services module 416, a device driver module 418, and cores 420. The cores 420 may comprise LP and one or more HP cores.
The OS services module 416 may comprise a kernel input/output control (IOCTL) module 422 and a scheduling module 424. The kernel IOCTL module 422 may report current I/O usage to the power control module 404. The scheduling module 424 may schedule OS tasks. Examples of OS tasks include number of tasks, number of processes, and thread quantum. The schedule of OS tasks is monitored by the OS task profile module 410 and reported to the power control module 404. The idle profile module 408 may receive inputs from the OS services module 416 and may report to the power control module 404 when the cores 420 are idle.
The device driver module 418 may comprise device drivers that control the hardware devices of the mobile computing device. For example, the device driver module 418 may comprise a battery driver module 426 that controls the battery of the mobile computing device based on signals received from the PMS module 402. The battery driver module 426 may report status of the battery to the power control module 404.
Additionally, the device driver module 418 may comprise a power management interrupt control (PMIC) module 428 that controls power consumed by the hardware devices based on inputs received from the PMS module 402. The device driver module 418 may comprise a performance monitoring unit (PMU) module 430 that monitors performance or resource utilization parameters. The resource utilization parameters may include memory usage, cache usage, bus usage, I/O usage, and CPU usage by the OS and the applications being executed by the cores 420. The PMU module 430 may report resource utilization parameters to the performance profile module 406. Specifically, the PMU module 430 may report when the various resource utilization parameters cross (i.e., exceed or fall below) respective predetermined thresholds when applications are executed by one of the cores 420. The thresholds may be programmable.
The device driver module 418 may further comprise a device timeout and state monitoring module 432 that monitors states of the hardware devices. The device timeout and state monitoring module 432 reports to the power control module 404 when any of the hardware devices are idle or have timed out due to lack of adequate resources available to perform requested tasks. The power control module 404 may shut down idle hardware devices to save power. Alternatively, the power control module 404 may activate (i.e., initiate power up) additional resources so that the hardware devices may render performance by completing the requested tasks instead of timing out.
The core profile module 412, the core sequence module 413, and the core switching module 414 enable core morphing. The core profile module 412 may anticipate or predict power/performance demand based on a history or log of execution of applications. For example, the core profile module 412 may predict a number of million instructions per second (MIPS) requirement of an application based on data cache miss rate, instruction cache miss rate, and/or instructions per cycle being executed by one of the cores 420 that is active. The core profile module 412 may receive the history from at least one of the performance profile module 406, the idle profile module 408, and the OS task profile module 410. Based on the history, the core profile module 412 can anticipate or predict the power/performance that the applications may demand.
The core profile module 412 may generate an event when the power/performance requirement changes to a value greater or less than a threshold value. For example, the core profile module 412 may generate an event when the core utilization or cache miss rate exceeds respective predetermined threshold values. The event is a request for switching cores according to the change in the power/performance demand. Additionally, the core profile module 412 may monitor a frequency of core switching and generate the event to minimize thrashing (i.e., unnecessary switching of cores 420). For example, the core profile module 412 may monitor a number of times the cores 420 are switched in a predetermined time period. The core profile module 412 may generate the event when the number exceeds a dynamically programmable threshold. When the event is received, the power control module 404 may decide to switch cores based solely on the event. Alternatively, the power control module 404 may analyze all the information received including resource utilization (past, current, and projected) and determine whether to switch cores.
Occasionally, the power control module 404 may override the core profile module 412 and decide not to switch cores notwithstanding the event. For example, when the mobile device is powered by line power and the HP core is the active core, the power control module 404 may not switch cores although the event may indicate that the performance requirement is low. Alternatively, when the HP core is the active core and the application is designated as a high-performance application, the power control module 404 may not switch cores although the mobile device is not powered by line power and the event requests a switch to the LP core due to low battery.
In
Based on the frequency change sequence received, the FCS module 434 may generate signals that change the clock frequency of the active core. Based on the voltage change sequence received, the VCS module 436 may generate signals that change the supply voltage of the active core. Based on the core change sequence received, the CCS module 438 may generate signals that switch the cores 420. The CCS module 438 may initiate the core change sequence in conjunction with the frequency and/or the voltage change sequence. The core change sequences decrease system overhead, system latency, and core transition time. The CCS module 438 is now described in more detail.
When the power control module 404 decides to switch the cores, the PMS module 402 disables interrupts and generates a system call to the hypervisor module 415 and/or the core switching module 414 to switch cores 420. Based on the system call, the CCS module 438 generates a call and outputs the call to the hypervisor module 415 and/or the core switching module 414. The hypervisor module 415 and/or the core switching module 414 switches the cores 420. After the cores 420 are switched, the hypervisor module 415 and/or the core switching module 414 returns control to the kernel IOCTL module 422 and/or the OS services module 416. Subsequently, the newly activated core resumes operation, and the PMS module 402 enables interrupts.
Following are some examples when the CCS module 438 may initiate the core change sequence. The CCS module 438 may initiate the core change sequence to switch from the LP core to the HP core when the core utilization of the LP core is greater than or equal to a high predetermined threshold. For example only, the high predetermined threshold may be a percentage such as 80% or 90%. The CCS module 438 may initiate the core change sequence when the core utilization of the LP core is greater than or equal to the high predetermined threshold for a predetermined time period. The predetermined time period may be short or long. For example only, the predetermined time period may be two or ten seconds. The high predetermined threshold and the predetermined time period can be programmed to optimize performance, power, and minimize core switching that may be unnecessary.
The CCS module 438 may initiate the core change sequence when an anticipated core utilization of the LP core is greater than or equal to the high predetermined threshold. The anticipated core utilization may be determined based on the type of the application being executed and/or the history of execution of the application. For example only, applications tagged by the user as requiring high performance may initiate a core switch from the LP core to the HP core.
The CCS module 438 may initiate the core change sequence to switch from the HP core to the LP core when the core utilization of the HP core is less than or equal to a low predetermined threshold. For example only, the low predetermined threshold may be a percentage such as 5% or 10%. The CCS module 438 may initiate the core change sequence when the core utilization of the HP core is less than or equal to the high predetermined threshold for a predetermined time period. The predetermined time period may be short or long. For example only, the predetermined time period may be two or ten seconds. The low predetermined threshold and the predetermined time period can be programmed to optimize performance, power, and minimize core switching that may be unnecessary.
The CCS module 438 may initiate the core change sequence based on the number of applications being executed by one of LP and HP cores. For example only, the cores may be switched from the LP core to the HP core when the number of applications being executed is 10 or more and from the HP to the LP core when the number of applications being executed is 10 or less. The CCS module 438 may initiate the core change sequence based on the type of applications. For example only, the cores may be switched from the LP core to the HP core when the application being executed is designated as requiring high performance. Finally, the CCS module 438 may initiate the core change sequence based on the type of instruction that is scheduled for execution by one of the cores. For example only, cores may be switched from the LP core to the HP core when the LP core is executing an application and an instruction scheduled for execution can be executed by the HP core instead of the LP core.
More specifically, the CCS module 438 may generate signals that switch the cores 420 as follows. When the CCS module 438 receives the core change sequence, the CCS module 438 outputs the call to the hypervisor module 415 and/or the core switching module 414. The core switching module 414 switches the cores 420 according to the core change sequence. Specifically, the core switching module 414 transfers the state of the active core to the inactive core, performs handshaking between the active and the inactive cores 420, and performs core switching as follows.
The core switching module 414 begins core switching when the PMS module 402 disables interrupts to the cores 420. The core switching module 414 suspends execution of the OS and the applications and isolates the cores 420 from the OS and the applications. The core switching module 414 may save the state of the active core that include core registers, CP registers, stack pointer, program counter, PMU events, and MMU configuration. The core switching module 414 activates the inactive core by turning on or increasing power supply to the clock, L2 cache (e.g., if the inactive core is the HP core), and the inactive core.
Based on the wakeup exchange signals received from the inactive core, the core switching module 414 determines if the inactive core wakes up (i.e., powers up) within a predetermined time after wakeup signals are output to the inactive core. If the inactive core does not acknowledge (i.e., does not wake up or power up) within the predetermined time, the core switching module 414 reloads the state of the active core. The core switching module 414 returns control to the kernel IOCTL module 422 and/or the OS services module 416. The active core resumes executing the application according to the reloaded state, and the PMS module 402 enables interrupts to the active core.
Instead, if the core switching module 414 receives an acknowledgement from the inactive core within the predetermined time, the core switching module 414 transfers the state of the previously active core to the newly activated core. The core switching module 414 initializes the newly activated core with the state of the previously active core. After the newly activated core is initialized, the core switching module 414 shuts down the previously active core to the inactive state by turning off or reducing the power supply to the clock, L2 cache (e.g., if the previously active core is the HP core), and the previously active core. The core switching module 414 returns control to the kernel IOCTL module 422 and/or the OS services module 416. The newly activated core resumes executing the application according to the loaded state. The PMS module 402 enables interrupts to the active core. Thus, the core switching module 414 switches cores securely and in isolation from the OS and the applications and makes the core switching transparent to the OS and the applications.
When the core switching module 414 switches operation from the LP core to the HP core, the OS and applications continue to run as if they would continue to execute on a single core with a stepped-up frequency or supply voltage. On the other hand, when the core switching module 414 switches operation from the HP core to the LP core, the OS and applications continue to run as if they would continue to execute on a single core with a stepped-down frequency or supply voltage.
The core switching module 414 and/or the hypervisor module 415 may communicate to the glue logic module 380 the handshaking information exchanged by the cores 420 during core switching. The handshaking information may include signals that indicate which one of the cores 420 is active, when interrupts are disabled/enabled, etc. Based on the handshaking signals, the glue logic module 380 routes interrupts and other signals from the application to one of the cores 420 that is active.
In use, Core1 may be the LP core, and Core2 may be the HP core. The mobile computing device may normally execute most applications using Core1. Core1 may operate at a frequency and voltage level specified by the frequency change sequence and the voltage change sequence, respectively. The frequency change sequence may include control signals that increase or decrease the clock frequency of the active core sequentially as needed. The voltage change sequence may include control signals that increase or decrease the supply voltage of the active core sequentially as needed.
When high performance is required (e.g., when utilization of Core1 is maximum at current clock frequency and supply voltage), the FCS module 434 may vary the clock frequency of Core1 within a frequency range from low to high based on the frequency change sequence generated for Core1 until the highest and/or desired operating frequency of Core1 is reached. Alternatively, or additionally, the VCS module 436 may change the supply voltage of Core1 within a voltage range from low to high based on the voltage change sequence generated for Core1 until the highest and/or desired operating supply voltage of Core1 is reached.
When still higher performance is demanded by the application (i.e., when utilization of Core1 is maximum at the highest frequency and voltage), the core switching module 414 may begin core switching. The core switching module 414 may activate Core2, handover the state of Core1 to Core2, and shut down Core1 or otherwise render Core1 inactive. Specifically, the core switching module 414 may deactivate Core1 and copy the state of Core1. Then the core switching module 414 may power up Core2. On powering up, when Core2 initializes, the core switching module 414 may handover the state of Core1 to Core2. That is, the core switching module 414 may reset Core2 with the state of Core1. Thereafter, the core switching module 414 may shut down Core1. Core2 may resume executing the application where Core1 entered the inactive state using the state of Core1. The PMS module 402 may enable interrupts. The scheduling module 424 may schedule threads that were to be executed by Core1 for execution by Core2.
Initially, Core2 may operate at a frequency and voltage level specified by the frequency change sequence and the voltage change sequence, respectively. Subsequently, the FCS module 434 may change the clock frequency of Core2 within a frequency range from low to high and then from high to low according to the performance level demanded by the application. Alternatively, or additionally, the VCS module 436 may vary the supply voltage of Core2 within a voltage range from low to high and from high to low according to the performance level demanded by the application. If the HP core comprises a plurality of Core2, each additional Core2 may be activated and deactivated as the performance level demanded by the application continues to vary.
When the performance level demanded by the application decreases to a level at which the performance rendered by Core2 is no longer optimal, efficient, desirable, and/or possible, the core switching module 414 may switch cores. The core switching module 414 may activate Core1, handover the state of Core2 to Core1, and shut down Core2 or otherwise render Core2 inactive. Specifically, the core switching module 414 may deactivate Core2 and copy the state of Core2. Then the core switching module 414 may power up Core1. On powering up, when Core1 initializes, the core switching module 414 may handover the state of Core2 to Core1. That is, the core switching module 414 may reset Core1 with the state of Core2. Core1 may resume executing the application where Core2 entered the inactive state using the state of Core2. The PMS module 402 may enable interrupts. Thereafter, the core switching module 414 may shut down Core2. The scheduling module 424 may schedule threads that were to be executed by Core2 for execution by Core1.
Referring now to
Thus, the core switching module 414 may dynamically switch the cores 420 and decrease the power consumption and increase the performance of the mobile device. Additionally, since the core switching module 414 hands over the state of one core to another securely and in isolation when the cores 420 are switched, switching of the cores 420 is transparent to the OS and the applications running on the cores 420. Accordingly, the cores 420 are plug compatible with the existing processing hardware and can replace the existing processing hardware without any hardware modifications.
Referring now to
Control disables interrupts in step 508. Core1 completes any pending read/write (R/W) commands being executed by Core1 in step 510. Control generates a system management interrupt (SMI) call to the core switching module 414 in step 512 to switch cores 420 (e.g., deactivate the active core, Core1, and activate the inactive core, Core2).
Control saves the state of the active core, Core1, in step 514. Control flushes L1 cache of Core1 in step 516 for coherency. Control sends the event to the inactive core, Core2, in step 518. Control determines in step 520 whether Core2 has acknowledged the event by powering up and by transitioning to the active state. If the result of step 520 is true, control turns off or otherwise reduces power to Core1 in step 521. Control loads the saved state of Core1 into Core2 in step 522. Core2 resumes execution of applications in step 524 where Core1 left off. Control enables interrupts in step 526. Control returns to step 504.
If, however, the result of step 520 is false (i.e., if Core2 does not acknowledge the event and does not power up), control reloads the state of Core1 into Core1 in step 528. Core1 resumes execution of applications in step 530. Control enables interrupts in step 532. Control returns to step 504.
The cell phone control module 860 may implement the multi-core control module according to the teachings of the present disclosure. The cell phone control module 860 may receive input signals from the cellular network interface 867, the network interface 868, the microphone 870, and/or the user input device 876. The cell phone control module 860 may process signals, including encoding, decoding, filtering, and/or formatting, and generate output signals. The output signals may be communicated to one or more of memory 864, the storage device 866, the cellular network interface 867, the network interface 868, and the audio output 872.
Memory 864 may include random access memory (RAM) and/or nonvolatile memory. Nonvolatile memory may include any suitable type of semiconductor or solid-state memory, such as flash memory (including NAND and NOR flash memory), phase change memory, magnetic RAM, and multi-state memory, in which each memory cell has more than two states. The storage device 866 may include an optical storage drive, such as a DVD drive, and/or a hard disk drive (HDD). The power supply 862 provides power to the components of the cellular phone 858.
In
The handset control module 960 may implement the multi-core control module according to the teachings of the present disclosure. The handset control module 960 may receive input signals from the base station interface 967, the network interface 968, the microphone 970, and/or the user input device 976. The handset control module 960 may process signals, including encoding, decoding, filtering, and/or formatting, and generate output signals. The output signals may be communicated to one or more of memory 964, the storage device 966, the base station interface 967 and the audio output 972.
Memory 964 may include random access memory (RAM) and/or nonvolatile memory. Nonvolatile memory may include any suitable type of semiconductor or solid-state memory, such as flash memory (including NAND and NOR flash memory), phase change memory, magnetic RAM, and multi-state memory, in which each memory cell has more than two states. The storage device 966 may include an optical storage drive, such as a DVD drive, and/or a hard disk drive (HDD). The power supply 962 provides power to the components of the wireless handset 958.
In
The mobile device control module 990 may implement the multi-core control module according to the teachings of the present disclosure. The mobile device control module 990 may receive input signals from the network interface 994 and/or the external interface 999. The external interface 999 may include USB, infrared, and/or Ethernet. The input signals may include compressed audio and/or video, and may be compliant with the MP3 format. Additionally, the mobile device control module 990 may receive input from a user input 996 such as a keypad, touchpad, or individual buttons. The mobile device control module 990 may process input signals, including encoding, decoding, filtering, and/or formatting, and generate output signals.
The mobile device control module 990 may output audio signals to an audio output 997 and video signals to a display 998. The audio output 997 may include a speaker and/or an output jack. The display 998 may present a graphical user interface, which may include menus, icons, etc. The power supply 991 provides power to the components of the mobile device 989. Memory 992 may include random access memory (RAM) and/or nonvolatile memory.
Nonvolatile memory may include any suitable type of semiconductor or solid-state memory, such as flash memory (including NAND and NOR flash memory), phase change memory, magnetic RAM, and multi-state memory, in which each memory cell has more than two states. The storage device 993 may include an optical storage drive, such as a DVD drive, and/or a hard disk drive (HDD). The mobile device may include a personal digital assistant, a media player, a laptop computer, a gaming console, or other mobile computing device.
In
The GPS control module 1090 may implement the multi-core control module according to the teachings of the present disclosure. The GPS control module 1090 may receive input signals from the external interface 1099. The external interface 1099 may include wireless signals, USB, infrared, and/or Ethernet. The input signals may include compressed audio and/or video, and may be compliant with the MP3 format. Additionally, the GPS control module 1090 may receive input from a user input 1096 such as a keypad, touchpad, or individual buttons. The GPS control module 1090 may process input signals, including encoding, decoding, filtering, and/or formatting, and generate output signals.
The GPS control module 1090 may output audio signals to an audio output 1097 and video signals to a display 1098. The audio output 1097 may include a speaker and/or an output jack. The display 1098 may present a graphical user interface, which may include menus, icons, etc. The power supply 1091 provides power to the components of the mobile device 1089. Memory 1092 may include random access memory (RAM) and/or nonvolatile memory.
Nonvolatile memory may include any suitable type of semiconductor or solid-state memory, such as flash memory (including NAND and NOR flash memory), phase change memory, magnetic RAM, and multi-state memory, in which each memory cell has more than two states. The storage device 1093 may include an optical storage drive, such as a DVD drive, and/or a hard disk drive (HDD). The mobile device may include a personal digital assistant, a media player, a laptop computer, a gaming console, or other mobile computing device.
In
The multi-core processing system 1102, the I/O core 1104, and memory 1106 may communicate via a system bus 1110. The I/O devices 1108 may include hard disk drives (HDDs), compact disc (CD) drives, digital versatile disc (DVD) drives, display, keyboard, and other I/O devices. The I/O core 1104 may comprise I/O controllers that control the I/O devices 1108. The multi-core processing system 1102 may execute an operating system (OS) and applications stored on the HDDs, CD drives, and/or the DVD drives. The multi-core control module 1103 may provide increased power savings and increased performance when the multi-core processing system 1102 executes applications.
Those skilled in the art can now appreciate from the foregoing description that the broad teachings of the disclosure can be implemented in a variety of forms. Therefore, while this disclosure includes particular examples, the true scope of the disclosure should not be so limited since other modifications will become apparent to the skilled practitioner upon a study of the drawings, the specification, and the following claims.
Claims
1. A system comprising:
- a first core to execute instructions associated with an application at a first speed based on a first instruction set;
- a second core to execute the instructions associated with the application at a second speed based on a second instruction set, wherein the first speed is greater than the first speed, and wherein the second instruction set is a subset of the first instruction set;
- a first memory to store an operating system, wherein the operating system includes a kernel that provides services to the application; and
- a core switching module to
- load into a second memory after the operating system is booted, wherein the second memory is separate from the first memory;
- operate at a level above the kernel of the operating system; and
- switch execution of the instructions associated with the application between the first core and the second core, transparently to the operating system and the application.
2. The system of claim 1, wherein the first core and the second core execute the instructions associated with the application without translating the instructions in response to the core switching module switching the execution of the instructions between the first core and the second core.
3. The system of claim 1, wherein the core switching module switches the execution of the instructions associated with the application between the first core and the second core based on one or more of processing load, type of the application, and resource utilization of the application.
4. The system of claim 1, further comprising a glue logic module to activate only one of the first core and the second core at a time in response to the core switching module switching the execution of the instructions associated with the application between the first core and the second core.
5. The system of claim 1, further comprising a core profile module to:
- generate an anticipated instruction execution rate for executing the instructions associated with the application based on at least one of a data cache miss rate, an instruction cache miss rate, and instructions per cycle executed by at least one of the first core and the second core;
- wherein the core switching module is configured to switch the execution of the instructions associated with the application between the first core and the second core based on the anticipated instruction execution rate.
6. The system of claim 5, further comprising:
- a core change sequence module to generate a core change sequence based on the anticipated instruction execution rate,
- wherein the core switching module powers up one of the first core and the second core and shut down another of the first core and the second core based on the core change sequence.
7. The system of claim 1, further comprising:
- a core activation module to stop processing of the application by the first core in response to disabling interrupts to the first core; and
- a state transfer module to transfer a state of the first core to the second core in response to the stopping of the processing of the application by the first core.
8. The system of claim 7, wherein the core activation module:
- allows the second core to resume execution of the instructions based on the state of the first core at the time of the stopping of the processing of the application by the first core; and
- enables interrupts to the second core.
9. The system of claim 7, wherein the core switching module supplies no power or standby power to the first core in response to the second core resuming execution of the instructions.
10. A method comprising:
- storing an operating system in a first memory, wherein the operating system includes a kernel that provides services to an application;
- loading a core switching module into a second memory after the operating system is booted, wherein the second memory is separate from the first memory;
- operating the core switching module at a level above the kernel of the operating system; and
- switching execution of instructions associated with the application between a first core and a second core transparently to the operating system and the application;
- wherein the first core is configured to execute the instructions associated with the application at a first speed based on a first instruction set;
- wherein the second core is configured to execute the instructions associated with the application at a second speed based on a second instruction set;
- wherein the first speed is greater than the first speed; and
- wherein the second instruction set is a subset of the first instruction set.
11. The method of claim 10, further comprising executing the instructions associated with the application on the first core and the second core without translating the instructions in response to the switching of the execution of the instructions between the first core and the second core.
12. The method of claim 10, further comprising switching the execution of the instructions associated with the application between the first core and the second core based on one or more of processing load, type of the application, and resource utilization of the application.
13. The method of claim 10, further comprising activating only one of the first core and the second core at a time in response to switching the execution of the instructions associated with the application between the first core and the second core.
14. The method of claim 10, further comprising:
- generating an anticipated instruction execution rate for executing the instructions associated with the application based on at least one of a data cache miss rate, an instruction cache miss rate, and instructions per cycle executed by at least one of the first core and the second core; and
- switching the execution of the instructions associated with the application between the first core and the second core based on the anticipated instruction execution rate.
15. The method of claim 14, further comprising:
- generating a core change sequence based on the anticipated instruction execution rate; and
- powering up one of the first core and the second core and shutting down another of the first core and the second core based on the core change sequence.
16. The method of claim 10, further comprising:
- stopping processing of the application by the first core in response to disabling interrupts to the first core; and
- transferring a state of the first core to the second core in response to the stopping of the processing of the application by the first core.
17. The method of claim 16, further comprising:
- allowing the second core to resume execution of the instructions based on the state of the first core at the time of the stopping of the processing of the application by the first core; and
- enabling interrupts to the second core.
18. The method of claim 16, further comprising supplying no power or standby power to the first core in response to the second core resuming execution of the instructions.
Type: Application
Filed: Oct 12, 2015
Publication Date: Feb 4, 2016
Inventors: Sehat Sutardja (Los Altos Hills, CA), Hong-Yi Chen (Fremont, CA), Premanand Sakarda (Acton, MA), Mark N. Fullerton (Austin, TX), Jay Heeb (Gilbert, AZ)
Application Number: 14/880,670
