WO2006068817A2

WO2006068817A2 - Method and apparatus for on-demand power management

Info

Publication number: WO2006068817A2
Application number: PCT/US2005/044053
Authority: WO
Inventors: Joel A. Jorgenson; Divyata Kakumanu; Brian M. Morlock
Original assignee: Packet Digital
Priority date: 2004-12-21
Filing date: 2005-12-05
Publication date: 2006-06-29
Also published as: US20080263382A1; US7398407B2; US20080263377A1; US20120089856A1; US8020015B2; EP1836545B1; WO2006068817A3; US8095818B2; KR20070100314A; JP5159316B2; US7228446B2; EP1836545A2; JP2008524745A; EP1836545A4; US20060136763A1; US20070198867A1; KR101282126B1

Abstract

An apparatus for on-demand power management including a system controller, a clock domain manager coupled to the system controller and a power distribution manager coupled to the system controller. The system controller monitors a processing demand in a processing system. The clock domain manager provides one or more clock frequencies and, in response to the processing demand, switches between a first set of clock frequencies and a second set of clock frequencies without halting the processing system. The power distribution manager provides one or more operating voltages and, in response to the processing demand, switches between a first set of voltages and a second set of voltages without halting the processing system.

Description

METHOD AND APPARATUS FOR ON-DEMAND POWER MANAGEMENT

TECHNICAL FIELD

[0001] The present invention relates generally to power management and in particular to managing voltages and frequencies in response to application processing demands.

BACKGROUND

[0002] As digital electronic processing systems trend toward higher operating frequencies and smaller device geometries, power management has become increasingly important to prevent thermal overload while maintaining system performance and prolonging battery life in portable systems. [0003] The two principal sources of power dissipation in digital logic circuits are static power dissipation and dynamic power dissipation. Static power dissipation is dependent on temperature, device technology and processing variables, and is composed primarily of leakage currents. Dynamic power dissipation is the predominant loss factor in digital circuitry and is proportional to the operating clock frequency, the square of the operating voltage and the capacitive load. Capacitive load is highly dependent on device technology and processing variables, so most approaches to dynamic power management focus on frequency and voltage control.

[0004] One conventional approach to power management halts the processing system to adjust core clock frequencies and voltages, during which time the processor does not execute operating system code or application code, and then restarts the system after the new frequencies and voltages have stabilized. Such an approach is described in U.S. Patent No. 6,754,837, as illustrated in Figure 1. Figure 1 illustrates a processor or processing system 1 contains a programmable voltage ID (VID) register 3, a clock frequency control register 4 and a count register 5. When the processor determines that a change in the voltage and/or frequency is desired, the desired voltage and frequency control information is loaded into the VID register and the clock frequency control register, respectively. Access to those registers triggers a stop request 9 to the CPU core logic 11. In response to the stop request, the CPU completes the current instruction and issues a stop grant signal 13 to indicate to a power controller 7 that processing has stopped. The stop grant state is maintained, for a time determined by a value in the count register, while the voltage and/or frequency are changed and stabilized. In addition to the processing time lost during the stop grant state, this approach may also result in large transient power surges when the processor restarts. [0005] Another conventional approach to power management, described in U.S. Patent No. 6,788,156, changes the clock frequency of a processor while the processor is operating, but requires the frequency changes to be made in small increments to avoid processing errors that large frequency steps would cause. As a result, this approach may require a significant time period to achieve a desired operating frequency.

[0006] Yet another conventional approach to power management, described in U.S. Patent No. 6,778,418, employs a fixed relationship between voltage and frequency, either through a lookup table or by use of a frequency to voltage converter. In this approach, a frequency increase is always preceded by a voltage increase and a frequency decrease always precedes a voltage decrease. In addition, a frequency increase is delayed while the voltage is ramped up to a corresponding voltage. The new frequency and voltage are not scaled independently, and the new operating point may not be optimum with respect to an application's processing demand.

BRIEF DESCRIPTION OF THE DRAWINGS

[0007] The present invention is illustrated by way of example, and not of limitation, in the figures of the accompanying drawings in which: [0008] Figure 1 illustrates a conventional power management system; [0009] Figure 2 A illustrates one embodiment of on-demand power management in a processing system;

[0010] Figure 2B illustrates one embodiment of on-demand power management in a distributed processing system;

[0011] Figure 2C illustrates one embodiment of an on-demand power manager;

[0012] Figure 3 illustrates a compensation engine in one embodiment of on- demand power management;

[0013] Figure 4 illustrates a power distribution manager in one embodiment of on-demand power management;

[0014] Figure 5 illustrates a clock domain manager in one embodiment of on- demand power management;

[0015] Figure 6 illustrates one embodiment of phase-matching in on-demand power management; and

[0016] Figure 7 is a state diagram illustrating one embodiment of on-demand power management;

[0017] Figure 8 illustrates voltage and frequency control in one embodiment of on-demand power management;

[0018] Figure 9A illustrates a method in one embodiment of on-demand power management;

[0019] Figure 9B illustrates one embodiment of the method illustrated by Figure 9A;

[0020] Figure 9C illustrates a further embodiment of the method illustrated by Figure 9A; and [0021] Figure 9D illustrates another further embodiment of the method illustrated by Figure 9A.

DETAILED DESCRIPTION

[0022] In the following description, numerous specific details are set forth such as examples of specific components, devices, methods, etc., in order to provide a thorough understanding of embodiments of the present invention. It will be apparent, however, to one skilled in the art that these specific details need not be employed to practice embodiments of the present invention. In other instances, well-known materials or methods have not been described in detail in order to avoid unnecessarily obscuring embodiments of the present invention. It should be noted that the "line" or "lines" discussed herein, that connect elements, may be single lines or multiple lines. The term "coupled" as used herein, may mean directly coupled or indirectly coupled through one or more intervening components. It will also be understood by one having ordinary skill in the art that lines and/or other coupling elements may be identified by the nature of the signals they carry (e.g., a "clock line " may implicitly carry a "clock signal") and that input and output ports may be identified by the nature of the signals they receive or transmit (e.g., "clock input" may implicitly receive a "clock signal"). [0023] A method and apparatus for on-demand power management is described. In one embodiment, the method includes monitoring a processing demand in a processing system operating at a first one or more voltages and a first one or more clock frequencies phase-locked to a reference frequency. The method also includes generating a second one or more clock frequencies in response to the processing demand, wherein the second one or more clock frequencies is phase-locked to the reference frequency and phase-matched to the first one or more clock frequencies. The method also includes switching from the first one or more clock frequencies to the second one or more clock frequencies without halting the processing system. In one embodiment, the method further includes generating a second one or more voltages in response to the processing demand, and switching from the first one or more voltages to the second one or more voltages without halting the processing system. [0024] In one embodiment, the apparatus includes a system controller to monitor an application processing demand on a processing system and to determine one or more clock frequencies and one or more voltages at which the processing system operates. The apparatus also includes a power distribution manger, coupled with the system controller, to provide one or more operating voltages to the processing system and to switch between a first one or more voltages and a second one or more voltages without halting the processing system. The apparatus also includes a clock domain manager, coupled with the system controller, to provide one or more clock signals to the processing system and to switch between a first one or more clock frequencies and a second one or more clock frequencies without halting the processing system. The first one or more clock frequencies and the second one or more clock frequencies are phase- locked to a common reference frequency and the second one or more clock frequencies are phase-matched to the first one or more clock frequencies. In one embodiment, the apparatus also includes a compensation engine coupled with the system controller, the power distribution manager and the clock domain manager, to receive voltage and frequency commands from the system controller and to compensate the voltage and frequency commands for temperature and processing variables.

[0025] Figure 2 A illustrates one embodiment of on-demand power management in a processing system 100. Processing system 100 may include a system processor 101, which may be a general-purpose processing device such as a microprocessor or central processing unit, or the like. Alternatively, system processor 101 may also be a special-purpose processing device such as an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a digital signal processor (DSP) or the like. System processor 101 may also be any combination of a general-purpose processing device and a special- purpose processing device. System processor 101 may be coupled to a system bus 102 which may carry system data and commands to and from system processor 101. System bus 102 may be coupled to memory 103 which may store programs and data. Memory 103 may be any type of memory, including, but not limited to, random access memory (RAM) and read only memory (ROM). System bus 102 may also be coupled with peripherals 104-1 through 104-k to carry system commands and data to and from peripherals 104-1 through 104-k. [0026] Processing system 100 may also include power manager 105, which maybe coupled to system bus 102, frequency source 108 and voltage source 109. Power manager 105 may also be coupled to system processor 101 and peripherals 104-1 through 104-k via a clock bus 106 and voltage bus 107. In one embodiment, as illustrated in Figure 2a, power manager 105 may be coupled to an external frequency source 108. Power manager 105 may be capable of converting a reference frequency f₀ from frequency source 108 into one or more clock frequencies fj through f_m, phase-locked to reference frequency f₀, to provide clock signals to system processor 101 and peripherals 104-1 through 104- k. In other embodiments, frequency source 108 may be integrated with power manager 105 and reside with power manager 105 on a common carrier substrate such as, for example, an integrated circuit (IC) die substrate, a multi-chip module substrate, or the like. Power manager 105 may also be capable of converting a voltage V₀ from voltage soμrce 109 into one or more operating voltages Vi through V_n to provide voltages to system processor 101 and peripherals 104-1 through 104-k. In the embodiment illustrated in Figure 2a, each of system processor 101 and peripherals 104-1 through 104-k are shown to have one voltage input and one clock input such that m = n = k+1. It will be appreciated that in other embodiments, any of system processor 101 and peripherals 104-1 through 104-k may require more than one operating voltage and/or more than one clock signal. In one embodiment, two or more of system processor 101, memory 103, power manager 105, frequency source 108 and peripherals 104-1 through 104-k may reside on a common carrier substrate, for example, a printed circuit board (PCB) such as motherboard 110 illustrated in Figure 2B, a daughter board 111 in Figure 2B, or a line card. Alternatively, the common carrier substrate on which the two or more of system processor 101, memory 103, power manager 105, frequency source 108 and peripherals 104-1 through 104-k may reside can be an integrated circuit (IC) die substrate.

[0027] With reference to Figure 2B, peripherals 104-1 through 104-k may be any type of device, component, circuit, subsystem or system capable of communicating with system processor 101 via system bus 102. For example, any of peripheral devices 104-1 through 104-k may be a single chip device 112 such as a system on a chip, an ASIC, an FPGA, a memory chip or like device. Any of peripherals 104-1 through 104-k may also be a multi-chip module 113 including any combination of single chip devices on a common integrated circuit substrate. Alternatively, peripherals 104-1 through 104-k may reside on one or more printed circuit boards such as, for example, a mother board 110, a daughter board 114 or other type of circuit card.

[0028] Figure 2C illustrates a power manager 105 in one embodiment of on- demand power management. Power manager 105 may include a system controller 201 to monitor the application processing demand in processing system 100 and to select an operating point for processing system 100. Power manager 105 may also include a power distribution manager 202, coupled with the system controller 201, to provide the one or more operating voltages Vi-V_n to processing system 100 and to switch between a first one or more voltages Vi '-V_n' and a second one or more voltages V₁ "-V_n" without halting processing system 100 as described below. Power manager 105 may also include a clock domain manager 203, coupled with system controller 201, to provide one or more clock signals f_\- f_m to processing system 100 and to switch between a first one or more clock signals fi'-f_m' and a second one or more clock signals fi"-f_m" without halting processing system 100 as described below. In one embodiment, power manager 105 may also include a compensation engine 204 coupled with system controller 201, power distribution manager 202 and clock domain manager 203. Compensation engine 204 may be configured to compensate the operating point selected by system controller 201 for temperature and process variables as described in detail below.

[0029] In one embodiment, power manager 105 may be configured to monitor processing activity on system bus 102 while supplying the first one or more clock frequencies fi'-f_m' and the first one or more voltages Vj '-V_n ¹ to system processor 101 and peripherals 104-1 through 104-k. Power manager 105 may also be configured to determine a processing demand based on the monitored processing activity and to generate the second one or more clock frequencies fi"- f_m" and the second one or more voltages Vi "-V_n" in response to the processing demand. Power manager 105 may also be configured to switch from the first one or more voltages to the second one or more voltages without halting the processing system 100, and to switch from the first one or more clock frequencies to the second one or more clock frequencies without halting the processing system 100.

[0030] System controller 201 may include a bus interface unit 205 to monitor processing activity on system bus 102 and to select a new operating point for the processing system 100. System controller 201 may also include a programmable memory 206 coupled with the bus interface unit 205. Programmable memory 206 may include programmed information to enable the bus interface unit 205 to correlate activity on the system bus 102 with the application processing demand in processing system 100.

[0031] In one embodiment, bus interface unit 205 may be configured to detect a plurality of commands on the system bus 102 and to recognize a command pattern, programmed in programmable memory 206, associated with a change in the application processing demand. The command pattern may be a generic processing command pattern, or a command pattern and bus transaction cycles associated with a specific system processor 101 or a processor family of which system processor 101 may be a member. In response to recognizing the command pattern, bus interface unit 205 may select the new operating point for the processing system 100. The new operating point may include a new set of operating voltages V₁ "-V_n" which are different from a current set of operating voltages Vi '-V_n", and/or a new set of clock frequencies fi"-f_m" which are different from a current set of operating clock frequencies fi'-f_m'. In one embodiment, the current sets of operating voltages and clock frequencies and the new sets of operating voltages and clock frequencies may be written to hardware registers (not shown) within system controller 201 or software defined registers (e.g., memory locations in programmable memory 206). [0032] Alternatively, bus interface unit 205 may be configured to detect an average number of processing events per unit time on system bus 102 and to compare the average number of processing events with one or more current clock frequencies 112. Based on the comparison, bus interface unit 205 may select a new operating point as described above.

[0033] As described in greater detail below, system controller 201 may also include a state machine 207, coupled with the bus interface unit 206 and a command bus 208, to control the provision of voltages Vi-V_n in the power distribution manager 202 and the provision of clock frequencies fi-f_m in the clock domain manager 203.

[0034] It will be appreciated by one having ordinary skill in the art that system controller 201 may be configured to automatically monitor the processing activity on system bus 102 and to and autonomously command the one or more voltages Vi-V_n and the one or more clock frequencies fi-f_m to select a new operating point as the application processing demand in processing system 100 changes. However, system controller 201 may also include a command interrupt line 209, coupled with state machine 207, to override the automatic control of the one or more voltages Vi-V_n and the one or more clock frequencies fi-f_m (e.g., in response to a critical power demand from the system processor 101 or one or more of peripherals 104-1 through 104-n). Command interrupt line 209 may be used to set processing system 100 to a predetermined operating point wherein the system controller 201 commands the power distribution manager 202 to provide one or more predetermined voltages to the processing system 100 and wherein the system controller commands the clock domain manager to provide one or more predetermined clock frequencies to the processing system 100. [0035] Figure 3 illustrates a compensation engine 204 in one embodiment of on-demand power management. Compensation engine 204 may include a receiver 301 to receive one or more voltage commands and one or more frequency commands from system controller 201 which are selected by system controller 201 to change the operating point of system 100 in response to the application processing demand. The voltage and frequency commands received by receiver 301 may be digital commands. Compensation engine 204 may also include a temperature sensor 302 to measure and report a temperature which may be, for example, a device temperature, a system temperature, an ambient temperature or any temperature which may have an effect on the operating point of processing system 100. Compensation engine 204 may also include a nonvolatile memory 303, coupled with temperature sensor 302, to store calibration data for processing system 100. The calibration data stored in non- volatile memory 303 may contain temperature dependent voltage and frequency correction factors for a device or system processing technology (e.g., CMOS processes) or one or more individual devices such as system processor 101 and peripherals 104-1 through 104-k. Compensation engine 204 may also include a compensation module 304 which may be coupled with receiver 301, temperature sensor 302 and non-volatile memory 303. Compensation module 304 may be configured to compensate voltage and frequency commands from receiver 301 for temperature, and temperature dependent processing and device variables. Compensation module 304 may be coupled with a scaling circuit 305 to provide one or more scaled voltage commands to the power distribution manager 202 and one or more scaled frequency commands to the clock domain manager 203 via command bus 306.

[0036] Figure 4 illustrates a power distribution manager 202 in one embodiment of on-demand power management. Power distribution manager 202 may include one or more voltage control channels 401-1 through 401-n corresponding to one or more operating voltages Vi-V_n. Each voltage control channel 401-1 through 401-n may include a dual voltage regulator 403 coupled between a ping-pong controller 402 and a multiplexer 404. The ping-pong controller may receive commands from the state machine 207 in system controller 201, through compensation engine 204, via command bus 306. The ping-pong controller 402 may set a first voltage regulator 403 a to a first voltage, a second voltage regulator 403b to a second voltage, and select between the first voltage and the second voltage in response to voltage commands from state machine 207. For example, in voltage control channel 401-1, voltage regulator 403a may be set to a first voltage Vi' and voltage regulator 403b may be set to a second voltage Vi". Power distribution manager 202 may also include a sequence controller 405, controlled by the system controller 201, to sequence the transitions between the first one or more voltages Vi '-V_n' and a second one or more voltages Vi "-V_n" in order to manage transient power demands. It will be appreciated that because the voltage changes described above may be made independently of any frequency changes, the voltages may be switched without halting the processing system 100.

[0037] Figure 5 illustrates a clock domain manager 203 in one embodiment of on-demand power management. Clock domain manager 203 may include one or more frequency control channels 501-1 through 501-m corresponding to one or more clock signals fl-fm. Each frequency control channel 501-1 through 501-m may include a dual phase-locked loop (PLL) 503 coupled between a ping-pong controller 502 and a multiplexer 504. The ping-pong controller may receive commands from the state machine 207 in system controller 201, through compensation engine 204, via command bus 306. The ping-pong controller 502 may set a first PLL 503a to a first clock frequency, a second PLL 503b to a second clock frequency, and select between the first clock frequency and the second clock frequency in response to frequency commands from state machine 207. For example, in frequency control channel 501-1, PLL 503a may be set to a first clock frequency rV and PLL 503b may be set to a second clock frequency fi". Each PLL 503a and 503b may be phase-locked to the reference frequency 110 from frequency source 108 (not shown), such that the clock frequencies provide by PLL's 503a and 503b are all multiples or sub-multiples of the reference frequency 110. Frequency multiplying PLL's and frequency dividing PLL's are known in the art and will not be discussed in detail here. Clock domain manger 203 may also include a jitter and phase controller 505, controlled by the system controller 201, to adjust for differential propagation delays among clock frequencies fi-f_m and to control the combined spectral content of the clock frequencies fi-f_m. .

[0038] It will be appreciated by one of ordinary skill in the art that all clock frequencies fl-frn will be harmonically related because all are phase-locked to the common reference frequency 110. In particular, any two clock frequencies in a single frequency control channel (e.g., clock frequencies fi' and fi" in frequency control channel 501-1) will be harmonically related. Figure 6 illustrates how this harmonic relationship may be used to switch between a first clock frequency and a second clock frequency without halting the processing system 100. Figure 6 depicts reference frequency 110 having frequency f₀ and period T₀ a= 1/ fo, clock frequency fi' = Af₀ and period Ti = To/A, and frequency fi"= Bf₀ and period T₂ = T₀ /B. As shown in Figure 6, the phase of clock frequency fi ' will periodically align with the phase of clock frequency fl " (e.g., at times ti, t₂, t₃, etc.) at time intervals corresponding to the lowest common multiples of Ti and T₂. This time interval may be calculated, for example, by system controller 201. Therefore, when a new operating point is commanded by the system controller in response to an application processing demand, the switch from the first clock frequency (e.g., fj') to the second clock frequency (e.g., fi") may be timed to occur when the phases of the first clock frequency and the second clock frequency are aligned. If the phases of the first clock frequency and the second clock frequency are aligned when the frequencies are switched (e.g., by a multiplexer 504), there is no phase discontinuity in the processing system 100 and the frequencies may be switched without halting the processing system 100. The ratio of the second clock frequency to the first clock frequency may be very large, approximately up to six orders of magnitude depending on the stability of the reference frequency 109. [0039] As noted above, the ping-pong controllers 402 in the power distribution manager 202 may receive commands from state machine 207 in system controller 201 to control the dual voltage regulators 403, and the ping- pong controllers 502 in the clock domain manger 203 may receive commands from the state machine 207 in the system controller 201 to control the dual PLL's 503. Figure 7 illustrates a state diagram for state machine 207 in one embodiment of on-demand power management for the exemplary voltage control channel 401-1 (where the dual voltage regulators 403 a and 403b are designated as VRi and VR₂, respectively) and the exemplary frequency control channel 501-1 (where the dual PLL's 503a and 503b are designated as PLLi and PLL₂, respectively), as shown in Figure 8. It will be appreciated that a state diagram, such as the state diagram in Figure 7 may be applied to each voltage control channel in power distribution manager 202 and each frequency control channel in clock domain manager 203.

[0040] In one embodiment, when a new operating voltage and/or a new clock frequency is commanded by the system controller, state machine 207 may operate in a ping-pong mode or a steady-state mode. Ping-pong mode is a symmetric mode where a new steady-state operating voltage is provided alternately by VRi and VR₂ with each change, and where the new steady-state clock frequency is provided alternately by PLLi and PLL₂ with each change. Steady-state mode is an asymmetrical mode where a new steady-state voltage is always provided by one voltage regulator (e.g., VRi) after a transient change is provided by the other voltage regulator (e.g., VR₂) and where a new steady-state clock frequency is always provided by one PLL (e.g., PLLi) after a transient change is provided by the other PLL (e.g., PLL₂). Table 1 defines the state variables used in Figure 7 and in the following description.

Table 1

[0041] In an initial state (701), VRj is set to a first voltage, which is selected by multiplexer 404 and provided to processing system 100. In the initial state PLLi is set to a first clock frequency, which is selected by multiplexer 504 and provided to processing system 100. Bus interface unit 205 periodically checks the system bus 102 for processing activity. If bus interface unit 205 does not detect a change in processing activity, the change frequency flag is cleared (cmd_fc = 0) and the change voltage flag is cleared (cmd_vc = 0). If bus interface unit 205 detects a change in processing activity on system bus 102 that warrants a change in the operating point of processing system 100, bus interface unit 205 will select the new operating point from programmable memory 206, which may require a new voltage and/or new clock frequency. [0042] If a new voltage is required (cmd vc = 1), VR₂ is commanded to the new voltage (702). After the new voltage is stabilized (chk_st = 1), the output of VR₂ is selected (703). In ping-pong mode (mode_pp = 1), VR₂ continues to be selected while the voltage requirement does not change (cmd_vc = 0). If the voltage requirement changes (cmd vc = 1), VR₁ is commanded to the new voltage (704a). After the new voltage is stabilized (chk st = 1), the output of VR₁ is selected (705) and the system returns to the initial state with the new voltage. In steady-state mode (mode ss = 1) at 703, the output OfVR₁ is commanded to equal the output of VR₂ (704b) and the output OfVR₁ is selected (705) when VRi is stabilized (chk st = 1) and the system returns to the initial state with the new voltage.

[0043] If a new clock frequency is required (cmd fc = 1), PLL₂ is commanded to the new frequency (706). After the new frequency is stabilized (chk_st = 1), the output of PLL₂ is selected (707). hi ping-pong mode (mode_pp = 1), PLL₂ continues to be selected while the frequency requirement does not change (cmd_fc = 0). If the frequency requirement changes (cmd_fc = 1), PLLi is commanded to the new frequency (708a). After the new frequency is stabilized (chk_st = 1), the output of PLLl is selected (709) and the system returns to the initial state (701) with the new frequency. In steady-state mode (mode_ss = 1) at 707, the output of PLLi is commanded to equal the output of PLL₂ (708b) and the output of PLLi is selected (709) when PLLi is stabilized (chk_st = 1) and the system returns to the initial state (701) with the new frequency. [0044] Figure 9 A illustrates one embodiment of a method 900 for on-demand power management. With reference to Figures 1 through 4, the method may include: monitoring a processing demand in processing system 100 operating at a first one or more voltages 113 and a first one or more clock frequencies 112 phase-locked to a reference frequency 110 (step 910); generating a second one or more clock frequencies 112 in response to the processing demand, the second one or more clock frequencies 112 phase-locked to the reference frequency 109 and phase-matched to the first one or more clock frequencies 112 (step 920); generating a second one or more voltages 113 in response to the processing demand (step 930); switching from the first one or more voltages 110 to the second one or more voltages 113 without halting the processing system 100 (step 940); and switching from the first one or more clock frequencies 112 to the second one or more clock frequencies 112 without halting the processing system 100 (step 950).

[0045] In one embodiment, as illustrated in Figure 9B, monitoring the processing demand (step 910) may include: detecting a plurality of processing events on a system bus 102 with a bus interface unit 205 (step 911); and correlating a clock frequency requirement with the plurality of processing events (step 912).

[0046] In one embodiment, as illustrated in Figure 9C, generating the second one or more clock frequencies 111 in response to the processing demand (step 920) may include: determining values for the second one or more clock frequencies 112 from the processing demand (step 921); scaling the values of the second one or more clock frequencies in a compensation engine 204 to compensate for a system temperature and a processing variable (step 922); synthesizing the scaled values of the second one or more clock frequencies 112 in one or more dual phase-locked loops 503 (step 923); and stabilizing the scaled values of the second one or more clock frequencies 112 before switching from the first one or more clock frequencies 112 to the second one or more clock frequencies 112 with one or more multiplexers 504.

[0047] In one embodiment, as illustrated in Figure 9D, generating the second one or more voltages 113 in response to the processing demand (step 930) may include: determining values for the second one or more voltages 113 from the processing demand (step 931); scaling the values of the second one or more voltages 113 in a compensation engine 204 to compensate for a system temperature and a processing variable (step 932); setting the scaled values of the second one or more voltages 113 in one or more dual voltage regulators 403 (step 933); and stabilizing the scaled values of the second one or more voltages 113 before switching from the first one or more voltages 113 to the second one or more voltages 113 with one or more multiplexers 404.

[0048] Thus, a method and apparatus for on-demand power management has been described. It will be apparent from the foregoing description that aspects of the present invention may be embodied, at least in part, in software. That is, the techniques may be carried out in a computer system or other data processing system in response to its processor, such as system controller 201, executing sequences of instructions contained in a memory, such as programmable memory 206. In various embodiments, hardwired circuitry may be used in combination with software instructions to implement the present invention. Thus, the techniques are not limited to any specific combination of hardware circuitry and software or to any particular source for the instructions executed by the data processing system. In addition, throughout this description, various functions and operations may be described as being performed by or caused by software code to simplify description. However, those skilled in the art will recognize what is meant by such expressions is that the functions result from execution of the code by a processor or controller, such as system controller 201. [0049] A machine-readable medium can be used to store software and data which when executed by a data processing system causes the system to perform various methods of the present invention. This executable software and data may be stored in various places including, for example, memory 103 and programmable memory 206 or any other device that is capable of storing software programs and/or data.

[0050] Thus, a machine-readable medium includes any mechanism that provides (i.e., stores and/or transmits) information in a form accessible by a machine (e.g., a computer, network device, personal digital assistant, manufacturing tool, any device with a set of one or more processors, etc.). For example, a machine- readable medium includes recordable/non-recordable media (e.g., read only memory (ROM); random access memory (RAM); magnetic disk storage media; optical storage media; flash memory devices; etc.), as well as electrical, optical, acoustical or other forms of propagated signals (e.g., carrier waves, infrared signals, digital signals, etc.); etc.

[0051] It should be appreciated that references throughout this specification to "one embodiment" or "an embodiment" means that a particular feature, structure or characteristic described in connection with the embodiment is included in at least one embodiment of the present invention. Therefore, it is emphasized and should be appreciated that two or more references to "an embodiment" or "one embodiment" or "an alternative embodiment" in various portions of this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures or characteristics may be combined as suitable in one or more embodiments of the invention. In addition, while the invention has been described in terms of several embodiments, those skilled in the art will recognize that the invention is not limited to the embodiments described. The embodiments of the invention can be practiced with modification and alteration within the scope of the appended claims. The specification and the drawings are thus to be regarded as illustrative instead of limiting on the invention.

Claims

CLAIMSWhat is claimed is:

1. A method, comprising: monitoring a processing demand in a processing system operating at a first one or more voltages and a first one or more clock frequencies phase-locked to a reference frequency; generating a second one or more clock frequencies in response to the processing demand, the second one or more clock frequencies phase-locked to the reference frequency and phase-matched to the first one or more clock frequencies; and switching from the first one or more clock frequencies to the second one or more clock frequencies without halting the processing system.

2. The method of claim 1, wherein monitoring the processing demand comprises: detecting a plurality of processing events; and correlating a clock frequency requirement with the plurality of processing events.

3. The method of claim 2, wherein the plurality of processing events comprises a plurality of commands on a system bus; and correlating the clock frequency requirement with the plurality of processing events comprises recognizing a command pattern.

4. The method of claim 2, wherein the plurality of processing events comprises an average number of processing events per unit time; and correlating the clock frequency requirement with the plurality of processing events comprises comparing the average number of processing events per unit time with a current clock frequency.

5. The method of claim 1, wherein generating the second one or more clock frequencies comprises: determining values for the second one or more clock frequencies from the processing demand; scaling the values of the second one or more clock frequencies to compensate for a system temperature and a processing variable; synthesizing the scaled values of the second one or more clock frequencies; and stabilizing the second one or more clock frequencies before switching from the first one or more clock frequencies.

6. The method of claim 5, wherein the processing variable is a process technology variable.

7. The method of claim 5, wherein the processing variable is a system specific calibration factor.

8. The method of claim 1, wherein switching from the first one or more clock frequencies to the second one or more clock frequencies comprises phase- shifting the second one or more clock frequencies to control a spectral content;

9. The method of claim 1, wherein switching from the first one or more clock frequencies to the second one or more clock frequencies is timed to coincide with periodic phase alignments between the first one or more clock frequencies and the second one or more clock frequencies.

10. The method of claim 1, wherein the ratio of the second one or more clock frequencies to the first one or more clock frequencies is approximately up to six orders of magnitude.

11. The method of claim 1 , further comprising: generating a second one or more voltages in response to the processing demand; and switching from the first one or more voltages to the second one or more voltages without halting the processing system.

12. The method of claim 11, wherein generating the second one or more voltages comprises: determining values for the second one or more voltages from the processing demand; scaling the values of the second one or more voltages to compensate for a system temperature and a processing variable; setting the scaled values of the second one or more voltages; and stabilizing the scaled values of the second one or more voltages before switching from the first one or more voltages.

13. The method of claim 12, wherein the processing variable is a process technology variable.

14. The method of claim 12, wherein the processing variable is a system specific calibration factor.

15. The method of claim 11, wherein switching from the first one or more voltages to the second one or more voltages comprises sequencing the second one or more voltages to reduce a transient power surge.

16. An apparatus, comprising: a system controller to monitor an application processing demand on a processing system and to determine one or more clock frequencies and one or more voltages at which the processing system operates; a power distribution manager coupled with the system controller to provide one or more operating voltages to the processing system and to switch between a first one or more voltages and a second one or more voltages without halting the processing system; and a clock domain manager coupled with the system controller to provide one or more clock signals to the processing system and to switch between a first one or more clock frequencies and a second one or more clock frequencies without halting the processing system, wherein the first one or more clock frequencies and the second one or more clock frequencies are phase-locked to a common reference frequency, and wherein the second one or more clock frequencies are phase-matched to the first one or more clock frequencies.

17. The apparatus of claim 16, wherein the system controller comprises: a bus interface unit to monitor activity on a system bus and to select an operating point for the processing system; a programmable memory, coupled with the bus interface unit, including a programmed table to correlate activity on the system bus with the application processing demand; and a state machine coupled with the bus interface unit to control the provision of voltages in the power distribution manager and the provision of clock frequencies in the clock domain manager.

18. The apparatus of claim 17, the bus interface unit to: detect a plurality of commands on a system bus; recognize a command pattern; and select a new operating point from the programmable memory.

19. The apparatus of claim 17, the bus interface unit to: detect an average number of processing events per unit time; compare the average number of processing events per unit time with a current clock frequency; and select a new operating point from the programmable memory.

20. The apparatus of claim 16, wherein the power distribution manager comprises: one or more voltage control channels to provide the first one or more voltages and the second one or more voltages, each voltage control channel comprising: a dual voltage regulator to provide a first voltages a second voltage; a ping-pong controller, coupled with the dual voltage regulator and the state machine, to receive commands from the state machine and to control the dual voltage regulator in response to the application processing demand; a multiplexer coupled with the dual voltage regulator to switch between the first voltage and the second voltage; and a sequence controller coupled to the one or more voltage control channels to sequence the provision of the second one or more voltages.

21. The apparatus of claim 16, wherein the clock domain manager comprises one or more frequency control channels to provide the first one or more clock frequencies and the second one or more clock frequencies, each frequency control channel comprising: a dual phase-locked loop to provide a first clock frequency and a second clock frequency; a ping-pong controller, coupled with the dual phase-locked loop and the state machine, to receive commands from the state machine and to control the dual phase-locked loop in response to the application demand; and a multiplexer coupled with the dual phase-locked loop to switch between the first clock frequency and the second clock frequency.

22. The apparatus of claim 21, further comprising: a jitter and phase controller coupled with the one or more frequency control channels to adjust for differential propagation delays among the one or more clock frequencies and to control a combined spectral content of the one or more clock frequencies.

23. The apparatus of claim 21, wherein: the first one or more clock frequencies and the second one or more clock frequencies are phase-locked to a common reference frequency; and wherein the one or more multiplexers are configured to switch from the first one or more clock frequencies to the second one or more clock frequencies when the phases of the first one or more clock frequencies and the second one or more clock frequencies are aligned.

24. The apparatus of claim 23, wherein a ratio of the second one or more clock frequencies to the first one or more clock frequencies is approximately up to six orders of magnitude.

25. The apparatus of claim 16, further comprising a compensation engine coupled with the system controller, the power distribution manager and the clock domain manager, the compensation engine comprising: a receiver to receive voltage and frequency commands from the system controller; a temperature sensor; a non- volatile memory to store calibration data; a compensation module coupled with the receiver, the temperature sensor and the non-volatile,memory, to compensate the voltage and frequency commands for a system temperature and a processing variable; and a scaling circuit coupled with the compensation module to provide one or more scaled voltage commands to the power distribution manager and one or more scaled frequency commands to the clock domain manager.

26. The apparatus of claim 25, wherein the processing variable is a process technology variable.

27. The apparatus of claim 25, wherein the processing variable is a system specific calibration factor.

28. The apparatus of claim 16, wherein the system controller comprises an interrupt input to receive a commanded mode power setting.

29. The apparatus of claim 16, further comprising: a processing device coupled with the power distribution manger and the clock domain manager; a system bus coupled with the processing device and the system controller; and a memory coupled with the system bus.

30. The apparatus of claim 29, further comprising one or more peripheral devices coupled with the system bus, the power distribution manager and the clock domain manager.

31. The apparatus of claim 30, wherein each peripheral devices comprises one of a single chip device, a multi-chip module and one or more circuit cards.

32. An apparatus, comprising: means for determining a new operating power point in a processing system in response to an application driven processing requirement; and means for changing from a present operating power point to the new operating power point without halting the processing system.

33. The apparatus of claim 32, further comprising means for adjusting the new operating power point to compensate for temperature, technology based processing variables and individual device processing variables.