WO2022000135A1 - 处理器和变量的预测方法 - Google Patents

处理器和变量的预测方法 Download PDF

Info

Publication number
WO2022000135A1
WO2022000135A1 PCT/CN2020/098564 CN2020098564W WO2022000135A1 WO 2022000135 A1 WO2022000135 A1 WO 2022000135A1 CN 2020098564 W CN2020098564 W CN 2020098564W WO 2022000135 A1 WO2022000135 A1 WO 2022000135A1
Authority
WO
WIPO (PCT)
Prior art keywords
processor core
cycle
value
predicted
power consumption
Prior art date
Application number
PCT/CN2020/098564
Other languages
English (en)
French (fr)
Inventor
胡荻
刘臻
李刚
王哲
李继明
Original Assignee
华为技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Priority to CN202080102431.2A priority Critical patent/CN115769172A/zh
Priority to PCT/CN2020/098564 priority patent/WO2022000135A1/zh
Publication of WO2022000135A1 publication Critical patent/WO2022000135A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F1/00—Details not covered by groups G06F3/00 - G06F13/00 and G06F21/00
    • G06F1/26—Power supply means, e.g. regulation thereof
    • G06F1/32—Means for saving power
    • G06F1/3203—Power management, i.e. event-based initiation of a power-saving mode
    • G06F1/3234—Power saving characterised by the action undertaken

Definitions

  • the present application relates to the field of computer technology, and in particular, to a processor and a method for predicting variables.
  • a tuning strategy based on various variables (eg power consumption, temperature). By adjusting the strategy, it can be ensured that during the running process of the processor, abnormal states such as overcurrent or overtemperature will not occur, thereby ensuring that the processor runs within a safe range.
  • the adjustment process of the power consumption-based adjustment strategy includes: determining the relationship between the measured value of the power consumption of each processor core in the processor in the current cycle and the preset waterline, and determining the relationship between the power consumption of each processor core
  • the relationship between the measured value in the current cycle and the preset waterline determines the adjustment strategy of each processor core, and adjusts the operating state of the corresponding processor core in the next cycle according to the adjustment strategy of each processor core, thereby ensuring that the processor runs within safe limits.
  • the processor may behave unexpectedly, for example, for a given 32-core processor, if all the processor cores of the current cycle processor are in a low power state, then When the system requests all the processor cores of the processor to enter the Turbo state in the next cycle, since all the processor cores of the processor in the current cycle are in a low power consumption state, the determined adjustment strategy determines that the power consumption exists A margin is allowed, and all processor cores are allowed to enter the Turbo state. Obviously, all the processor cores currently enter the Turbo state, and the processor is at risk of overcurrent.
  • the adjustment strategy determined in the above manner may not be able to cope with the running state of the processor in the next cycle, resulting in The processor has the risk of overcurrent, which affects the normal operation of the processor. Similarly, there are similar problems for other variables.
  • the present application provides a processor and a method for predicting a variable, which are used to solve the problem of determining the variable based on the measured value of the variable in the current cycle because the measured value of the variable in the current cycle does not consider the running state of the processor core in the next cycle.
  • the adjustment strategy cannot deal with the problem of the running state of the processor in the next cycle.
  • a processor including at least one processor core, control logic, and at least one sensing device; the at least one sensing device is used to obtain the at least one processor core's variable to be predicted in the current cycle. a measured value; the control logic is configured to determine the predicted value of the to-be-predicted variable of the at least one processor core in the next cycle according to the measured value of the to-be-predicted variable of the at least one processor core in the current cycle.
  • a method for determining the predicted value of the variable to be predicted in the processor core in the next cycle is provided, so that the adjustment strategy determined based on the predicted value of the variable to be predicted in the next cycle can cope with the running state of the processor in the next cycle, thereby ensuring The processor operates within safe limits.
  • the variable to be predicted is the power consumption of the processor core; the at least one process is determined according to the measured value of the variable to be predicted of the at least one processor core in the current cycle
  • the predicted value of the variable to be predicted of the processor core in the next cycle includes: obtaining the measured value of the frequency and voltage of the at least one processor core in the current cycle; determining the frequency and voltage of the at least one processor core in the next cycle.
  • the values of the frequency and voltage at the next cycle determine a predicted value of the power consumption of the at least one processor core at the next cycle.
  • the predicted value of the power consumption of the processor core in the next cycle provides a method for determining the predicted value of the power consumption of the processor core in the next cycle, and the determination method is simple and easy to execute, which improves the determination of the processor core. Efficiency of the predicted value of power consumption in the next cycle.
  • the voltage and power values of the at least one processor core in the next cycle are considered, that is, the power consumption of the at least one processor core in the next cycle is considered.
  • the working state therefore, improves the accuracy of determining the predicted value of the power consumption of the processor core over the next week.
  • the predicted value of the power consumption of the processor core in the next cycle is determined according to the following formula:
  • P(t+1) is the predicted value of the power consumption of the processor core in the next cycle
  • P(t) is the measured value of the power consumption of the processor core in the current cycle
  • f (t+1) is the value of the frequency of the processor core in the next cycle
  • f(t) is the measured value of the frequency of the processor core in the current cycle
  • V(t+1) is The value of the voltage of the processor core in the next cycle
  • V(t) is the measured value of the voltage of the processor core in the current cycle.
  • control logic is further configured to: acquire the measurement value of the temperature of the at least one processor core in the current cycle; acquire the measurement value of the temperature of the at least one processor core in the previous cycle obtain the measured value of the power consumption of the at least one processor core in the previous cycle; according to the measured value of the power consumption and temperature of the at least one processor core in the previous cycle and the current cycle, respectively,
  • the predicted value of the power consumption of the at least one processor core in the next cycle is determined as the predicted value of the temperature of the at least one processor core in the next cycle.
  • the predicted value of the temperature in the next cycle provides a method for determining the predicted value of the temperature of the processor core in the next cycle, and the determination method is simple and easy to execute, which improves the determination of the temperature of the processor core in the next cycle. Efficiency of the predicted value.
  • the value of the power consumption of the at least one processor core in the next cycle is considered, that is, the working state of the at least one processor core in the next cycle is considered , thus improving the accuracy of determining the predicted value of the temperature of the processor core for the next week.
  • the predicted value of the temperature of the processor core in the next cycle is determined according to the following formula:
  • T(t+1) is the predicted value of the temperature of the processor core in the next cycle
  • T(t) is the measured value of the temperature of the processor core in the current cycle
  • T(t -1) is the measured value of the temperature of the processor core in the previous cycle
  • P(t+1) is the predicted value of the power consumption of the processor core in the next cycle
  • P(t) is the measured value of the power consumption of the processor core in the current cycle
  • P(t-1) is the measured value of the power consumption of the processor core in the previous cycle.
  • the determining, according to the measured value of the variable to be predicted of the at least one processor core in the current cycle, the predicted value of the variable to be predicted of the at least one processor core in the next cycle includes: according to The measured value of the variable to be predicted of the at least one processor core in the current cycle is combined with a calibration table to determine the predicted value of the variable to be predicted of the at least one processor core in the next cycle.
  • the variable to be predicted is the power consumption of the processor core
  • the correction table includes a first correction table and a second correction table
  • Determining the predicted value of the variable to be predicted in the next cycle by combining the measured value of the variable to be predicted in the current cycle with a correction table in the next cycle includes: according to the power consumption of the at least one processor core in the current cycle The measured value of the at least one processor core determines the first value of the power consumption of the at least one processor core in the next cycle; according to the first correction table, it is determined that the load of the at least one processor core is in the next cycle.
  • the correction value of the load of the processor core corrects the first value of the power consumption of the at least one processor core to obtain a prediction of the power consumption of the at least one processor core in the next cycle value.
  • control logic is further configured to: acquire a measurement value of the power consumption of the at least one processor core in the previous cycle; acquire the temperature of the at least one processor core in the The measured values of the current cycle and the previous cycle; according to the predicted value of the power consumption of the at least one processor core in the next cycle, the power consumption of the at least one processor core in the current cycle and the previous cycle.
  • the measured value of the cycle, the measured value of the temperature of the at least one processor core in the current cycle and the previous cycle determine the predicted value of the temperature of the at least one processor core in the next cycle.
  • the predicted value of the power consumption of at least one processor core in the next cycle the measured value of the power consumption of at least one processor core in the current cycle and the previous cycle, the temperature of at least one processor core in the current cycle and the previous cycle
  • the measurement value determines the predicted value of the temperature of at least one processor core in the next cycle, and provides a method for determining the predicted value of the temperature of the processor core in the next cycle.
  • the efficiency of the predicted value of the core temperature in the next cycle is improved.
  • the accuracy of the predicted value of processor core temperature for the next week is considered, and further improved.
  • the accuracy of the predicted value of the power consumption of the processor core in the next cycle is further improved.
  • the first correction table comprises a first index number T N T N of first information and trends, wherein: the index information of first T N T N-th and the One change trend corresponds one to one; the first index information is used to indicate a change trend of the load of the processor core in N consecutive cycles, and T is the change of the load of the processor core in one cycle The number of types of trends; a first change trend is used to indicate the change trend of the value of the load of the processor core in the first cycle compared to the value of the load of the processor core in the second cycle, wherein the The first period is the next period of N consecutive periods indicated by the first index information corresponding to the first change trend, and the second period is N indicated by the first index information corresponding to the first change trend The last cycle of consecutive cycles; the second correction table includes T N second index information and T N correction values, wherein: the T N second index information and the T N correction values are equal to one One correspondence; one of the second index information is used to indicate that the change trend of the load of the processor core in N consecutive
  • a correction value is used to indicate the percentage change of the value of the load of the processor core in the third cycle compared to the value of the load of the processor core in the fourth cycle, wherein the The third period is the next period of N consecutive periods indicated by the second index information corresponding to the correction value, and the fourth period is the N consecutive periods indicated by the second index information corresponding to the correction value. in the last cycle.
  • control logic is further configured to: determine the adjustment strategy of the at least one processor core according to the predicted value of the variable to be predicted of the at least one processor core in the next cycle, wherein the The adjustment strategy is used to indicate the adjustment mode of the voltage and/or frequency of the processor core.
  • the determining the adjustment strategy of the at least one processor core according to the predicted value of the variable to be predicted of the at least one processor core in the next cycle includes: according to the at least one processor core The adjustment strategy generates a frequency adjustment instruction and/or a voltage adjustment instruction of the at least one processor core; the processor further includes at least one frequency and voltage adjustment circuit; the control logic is also used to adjust the at least one processor The frequency adjustment instruction and/or the voltage adjustment instruction of the core is sent to the at least one frequency and voltage adjustment circuit; the at least one frequency and voltage adjustment circuit is used for adjusting the frequency and/or the voltage adjustment instruction according to the at least one processor core The instructions adjust the frequency and/or voltage of the at least one processor core.
  • a method for predicting a variable comprising: acquiring a measurement value of the variable to be predicted of the at least one processor core in the current cycle; according to the measurement of the variable to be predicted of the at least one processor core in the current cycle The value determines the predicted value of the variable to be predicted for the at least one processor core in the next cycle.
  • the variable to be predicted is the power consumption of the processor core; the at least one process is determined according to the measured value of the variable to be predicted of the at least one processor core in the current cycle
  • the predicted value of the variable to be predicted of the processor core in the next cycle includes: obtaining the measured value of the frequency and voltage of the at least one processor core in the current cycle; determining the frequency and voltage of the at least one processor core in the next cycle.
  • the values of the frequency and voltage at the next cycle determine a predicted value of the power consumption of the at least one processor core at the next cycle.
  • the predicted value of the power consumption of the processor core in the next cycle is determined according to the following formula:
  • P(t+1) is the predicted value of the power consumption of the processor core in the next cycle
  • P(t) is the measured value of the power consumption of the processor core in the current cycle
  • f (t+1) is the value of the frequency of the processor core in the next cycle
  • f(t) is the measured value of the frequency of the processor core in the current cycle
  • V(t+1) is The value of the voltage of the processor core in the next cycle
  • V(t) is the measured value of the voltage of the processor core in the current cycle.
  • the method further includes: acquiring a measured value of the temperature of the at least one processor core in the current cycle; acquiring a measured value of the temperature of the at least one processor core in a previous cycle; Obtain the measured value of the power consumption of the at least one processor core in the previous cycle; according to the measured value of the power consumption and temperature of the at least one processor core in the previous cycle and the current cycle, the The predicted value of the power consumption of at least one processor core in the next cycle is determined, and the predicted value of the temperature of the at least one processor core in the next cycle is determined.
  • the predicted value of the temperature of the processor core in the next cycle is determined according to the following formula:
  • T(t+1) is the predicted value of the temperature of the processor core in the next cycle
  • T(t) is the measured value of the temperature of the processor core in the current cycle
  • T(t -1) is the measured value of the temperature of the processor core in the previous cycle
  • P(t+1) is the predicted value of the power consumption of the processor core in the next cycle
  • P(t) is the measured value of the power consumption of the processor core in the current cycle
  • P(t-1) is the measured value of the power consumption of the processor core in the previous cycle.
  • the determining, according to the measured value of the variable to be predicted of the at least one processor core in the current cycle, the predicted value of the variable to be predicted of the at least one processor core in the next cycle includes: according to The measured value of the variable to be predicted of the at least one processor core in the current cycle is combined with a calibration table to determine the predicted value of the variable to be predicted of the at least one processor core in the next cycle.
  • the variable to be predicted is the power consumption of the processor core
  • the correction table includes a first correction table and a second correction table
  • Determining the predicted value of the variable to be predicted in the next cycle by combining the measured value of the variable to be predicted in the current cycle with a correction table in the next cycle includes: according to the power consumption of the at least one processor core in the current cycle The measured value of the at least one processor core determines the first value of the power consumption of the at least one processor core in the next cycle; according to the first correction table, it is determined that the load of the at least one processor core is in the next cycle.
  • the correction value of the load of the processor core corrects the first value of the power consumption of the at least one processor core to obtain a prediction of the power consumption of the at least one processor core in the next cycle value.
  • the method further includes: acquiring a measured value of the power consumption of the at least one processor core in the previous cycle; acquiring the temperature of the at least one processor core in the current cycle respectively and the measured value of the previous cycle; according to the predicted value of the power consumption of the at least one processor core in the next cycle, the power consumption of the at least one processor core is between the current cycle and the previous cycle.
  • the measured value, the measured value of the temperature of the at least one processor core in the current cycle and the previous cycle determines the predicted value of the temperature of the at least one processor core in the next cycle.
  • the first correction table comprises a first index number T N T N of first information and trends, wherein: the index information of first T N T N-th and the One change trend corresponds one to one; the first index information is used to indicate a change trend of the load of the processor core in N consecutive cycles, and T is the change of the load of the processor core in one cycle The number of types of trends; a first change trend is used to indicate the change trend of the value of the load of the processor core in the first cycle compared to the value of the load of the processor core in the second cycle, wherein the The first period is the next period of N consecutive periods indicated by the first index information corresponding to the first change trend, and the second period is N indicated by the first index information corresponding to the first change trend The last cycle of consecutive cycles; the second correction table includes T N second index information and T N correction values, wherein: the T N second index information and the T N correction values are equal to one One correspondence; one of the second index information is used to indicate that the change trend of the load of the processor core in N consecutive
  • a correction value is used to indicate the percentage change of the value of the load of the processor core in the third cycle compared to the value of the load of the processor core in the fourth cycle, wherein the The third period is the next period of N consecutive periods indicated by the second index information corresponding to the correction value, and the fourth period is the N consecutive periods indicated by the second index information corresponding to the correction value. in the last cycle.
  • the method further includes: determining an adjustment strategy of the at least one processing core according to the predicted value of the variable to be predicted of the at least one processor core in the next cycle, wherein the adjustment strategy Used to indicate how the voltage and/or frequency of the processor core is adjusted.
  • the determining the adjustment strategy of the at least one processing core according to the predicted value of the variable to be predicted of the at least one processor core in the next cycle includes: according to the at least one processor core's predicted value
  • the adjustment strategy generates a frequency adjustment instruction and/or a voltage adjustment instruction of the at least one processor core; adjusts the frequency and/or voltage adjustment instruction of the at least one processor core according to the frequency adjustment instruction and/or the voltage adjustment instruction of the at least one processor core. / or voltage.
  • a computer-readable storage medium comprising a computer program, which, when executed on a computer, causes the computer to perform the method of any one of the second aspect.
  • a computer program for performing the method according to any one of the second aspects when the computer program is executed by a computer.
  • FIG. 1 is a schematic structural diagram of a processor provided by an embodiment of the present application.
  • FIG. 2 is a schematic flowchart of determining the predicted value of the power consumption of processor core 0 in the next cycle provided by an embodiment of the present application:
  • FIG. 3 is a schematic flowchart of determining the predicted value of the temperature of the processor core 0 in the next cycle according to an embodiment of the present application;
  • FIG. 4 is a schematic flowchart of a method for predicting a variable according to an embodiment of the present application.
  • At least one (item) refers to one or more, and "a plurality” refers to two or more.
  • “And/or” is used to describe the relationship between related objects, indicating that there can be three kinds of relationships, for example, “A and/or B” can mean: only A, only B, and both A and B exist , where A and B can be singular or plural.
  • the character “/” generally indicates that the associated objects are an “or” relationship.
  • At least one item(s) below” or similar expressions thereof refer to any combination of these items, including any combination of single item(s) or plural items(s).
  • At least one (a) of a, b or c can mean: a, b, c, "a and b", “a and c", “b and c", or "a and b and c" ", where a, b, c can be single or multiple.
  • FIG. 1 is a schematic structural diagram of the processor provided by an embodiment of the present application.
  • the processor includes 4 processor cores, control logic, 4 sensing devices, 4 frequency voltage regulation circuits, external cache, internal memory, general purpose unit, accelerator, input/output control unit & interface unit.
  • 4 processor cores are in one-to-one correspondence with 4 sensing devices
  • 4 processor cores are in one-to-one correspondence with 4 frequency voltage adjustment circuits.
  • the four processor cores are respectively processor core 0, processor core 1, processor core 2, and processor core 3.
  • Each sensing device is used to acquire the measured value of the variable to be predicted of its corresponding processor core in the current cycle, and transmit the acquired measured value of the variable to be predicted of the corresponding processor core in the current cycle to the control logic.
  • the control logic is used to determine the predicted value of the variable to be predicted for each processor core in the next cycle. Specifically, it can include but is not limited to the following two ways:
  • control logic determines the predicted value of each processor core's variable to be predicted in the next cycle according to the measured value of the to-be-predicted variable of each processor core in the current cycle.
  • the control logic determines the predicted value of the to-be-predicted variable of each processor core in the next cycle according to the measured value of the to-be-predicted variable of each processor core in the current cycle and combined with a correction table.
  • the correction table is used to indicate the change trend and correction value of the variables to be predicted in the next cycle.
  • the variables related to the variable to be predicted refer to the variables that affect the predicted value of the variable to be predicted in the next cycle. For example, if the variable to be predicted is the power consumption or temperature of the processor core, the temperature and power consumption of the processor core will be affected by changes in the load of the processor core. Therefore, the related variables of the power consumption and temperature of the processor core are: The load on the processor core.
  • variable to be predicted may be, for example, the power consumption of the processor core, the temperature of the processor core, and the like, which are not particularly limited in this embodiment of the present application.
  • FIG. 2 is a schematic flowchart of determining a predicted value of power consumption of processor core 0 in the next cycle according to an embodiment of the present application. As shown in Figure 2, including:
  • Step 201 the sensing device in the processor core 0 obtains the measured value of the power consumption of the processor core 0 in the current cycle. Specifically, it includes the following two methods:
  • the sensing device acquires the value of the power consumption of the processor core 0 at any moment in the current cycle or a specified moment, and determines the acquired value of the power consumption as the power consumption of the processor core 0 in the current cycle measured value.
  • a specified time may be, for example, an intermediate time or a starting time in the current cycle, which is not specifically limited in this application.
  • the sensing device obtains the values of the power consumption of the processor core 0 at multiple times in the current cycle, and calculates the average value of the power consumption at multiple times, and determines the average value as the processor core 0
  • the measured value of the power consumption in the current cycle, wherein, multiple moments in the current cycle can be set according to experience, which is not specially limited in this application.
  • the sensing device for acquiring power consumption may be, for example, a power sensor (power sensor), etc., which is not particularly limited in this embodiment of the present application.
  • Step 202 the sensing device in the processor core 0 sends the measured value of the power consumption of the processor core 0 in the current cycle to the control logic.
  • Step 203 The control logic determines the predicted value of the power consumption of the processor core 0 in the next cycle. Specifically, it can be determined in the following two ways.
  • the first type according to the measured value of the power consumption of the processor core 0 in the current cycle, the predicted value of the power consumption of the processor core 0 in the next cycle is determined.
  • the specific process includes:
  • the measured value of the frequency of processor core 0 in the current cycle can be the value of the frequency of processor core 0 at any moment in the current cycle or a specified moment, and the measured value of the frequency of processor core 0 in the current cycle can also be is the average value of the frequency of the processor core 0 at multiple times in the current cycle, etc., which is not specifically limited in this application.
  • the measured value of the voltage of processor core 0 in the current cycle can be the value of the voltage of processor core 0 at any moment in the current cycle or a specified moment, and the measured value of the voltage of processor core 0 in the current cycle The value may also be the average value of the voltage of the processor core 0 at multiple times in the current cycle, etc., which is not specifically limited in this application.
  • the value of the frequency and voltage of the processor core 0 in the next cycle can be determined by a control algorithm or an operating system, respectively.
  • the predicted value of power consumption in the next cycle can be determined by the following formula:
  • P(t+1) is the predicted value of the power consumption of the processor core 0 in the next cycle
  • P(t) is the measured value of the power consumption of the processor core 0 in the current cycle
  • f(t+1) is the processing value is the value of the frequency of processor core 0 in the next cycle
  • f(t) is the measured value of the frequency of processor core 0 in the current cycle
  • V(t+1) is the value of the voltage of processor core 0 in the next cycle
  • V( t) is the measured value of the voltage of processor core 0 in the current cycle.
  • the measured values of the frequency, voltage and power consumption of the processor core 0 in the current cycle are determined, the frequency and voltage of the processor core 0 in the next cycle are determined, and the measured values determined according to the above and the determined values are determined.
  • Each value can determine the predicted value of the power consumption of the processor core 0 in the next cycle, providing a method for determining the predicted value of the power consumption of the processor core 0 in the next cycle, and the determination method is simple, easy to execute, and improves the To determine the efficiency of the predicted value of the power consumption of processor core 0 in the next cycle.
  • the voltage and power of the processor core 0 in the next cycle are considered, that is, the working state of the processor core 0 in the next cycle is considered, Therefore, the accuracy of determining the predicted value of the power consumption of processor core 0 for the next week is improved.
  • the second according to the measured value of the power consumption of the processor core 0 in the current cycle and combined with a calibration table to determine the predicted value of the power consumption of the processor core 0 in the next cycle, wherein the calibration table includes the first calibration table and the second calibration table.
  • a correction table the first correction table is used to predict the change trend of the load of the processor core in the next cycle
  • the second correction table is used to predict the correction value of the load of the processor core in the next cycle.
  • the first value of the power consumption of the processor core 0 in the next cycle is determined according to the measured value of the power consumption of the processor core 0 in the current cycle. Because the principle of this step is the same as the principle of determining the predicted value of the power consumption of the processor core 0 in the next cycle in the first method, that is, the first value here is the power consumption of the processor core 0 in the first method. The predicted value consumed in the next cycle, so it will not be repeated here.
  • the change trend of the load of the processor core 0 in the next cycle is determined according to the first correction table.
  • the first correction table includes TN pieces of first index information and TN pieces of first change trends, wherein the pieces of TN pieces of first index information are in one-to-one correspondence with the pieces of TN pieces of first change trends.
  • a first index information is used to indicate a variation trend of the load of the processor core in N consecutive cycles, and a variation trend of the load of the processor core in the N consecutive cycles refers to N of the loads of the processor core.
  • a combination of change trends, N change trends correspond to N consecutive cycles one-to-one, one change trend is used to indicate the change trend of the processor core load in the corresponding cycle, and the change trend of the processor core load in the corresponding cycle is: The change trend of the value of the load of the processor core in the corresponding cycle compared to the value of the load of the processor core in the previous cycle of the corresponding cycle.
  • T is the number of types of changes in the load of the processor core in one cycle.
  • the value of T is 3.
  • the change trend of the load of the processor core in one cycle includes two situations of increasing and decreasing, that is, including two types, the value of T is 2.
  • the variation trends of the load of the processor core in a cycle include T types, that is, there are T types of variation trends, there are T N types of combinations of the N variation trends, then, the number of the first index information is T N Piece.
  • a first change trend is used to indicate the change trend of the value of the load of the processor core in the first cycle compared with the value of the load of the processor core in the second cycle, wherein the first cycle is the first cycle corresponding to the first change trend.
  • the next cycle of the N consecutive cycles indicated by the index information, and the second cycle is the last cycle among the N consecutive cycles indicated by the first index information corresponding to the first change trend.
  • a first change trend is used to indicate a change trend of the load of the processor core in the next cycle of N consecutive cycles indicated by the corresponding first index information. For example, if the value of N is 8, and if the 8 consecutive cycles are the 5th cycle to the 12th cycle, the first cycle is the 13th cycle, and the second cycle is the 12th cycle.
  • the first change trend includes, for example, three cases of becoming larger, smaller, and unchanged, and the first change trend also includes, for example, two cases of becoming larger or equal, and smaller, which is not specifically limited in this application.
  • the principle of determining the change trend of the load of the processor core 0 in the next cycle is to obtain the change trend of the load of the processor core 0 in N consecutive cycles, wherein the last cycle in the N consecutive cycles is the current cycle; take the change trend of the load of the processor core 0 in N consecutive cycles as the first index information to be compared, and match the first index information to be compared with the first index information in the first correction table, and The first change trend corresponding to the first index information matching the first index information to be compared is determined as the change trend of the load of the processor core 0 in the next cycle.
  • the process of obtaining the change trend of the load of the processor core 0 in N consecutive cycles includes: obtaining the values of the power consumption, frequency, and voltage of the processor core 0 in each cycle in the N consecutive cycles, and calculating the value of the processor core 0
  • the product of the value of the frequency in one cycle and the square of the value of the voltage of the processor core 0 in this cycle is characterized by the ratio of the value of the power consumption of the processor core 0 in this cycle and the product of the load of the processor core 0 in The value of this cycle, in other words, the ratio can be regarded as the value of the load of processor core 0 in this cycle.
  • the second correction table includes TN pieces of second index information and TN pieces of correction values, wherein: the pieces of TN pieces of second index information are in one-to-one correspondence with the pieces of TN correction values.
  • a second index information is used to indicate a result obtained after XOR operation is performed between the change trend of the load of the processor core in N consecutive cycles and the value of the load of the processor core in the last cycle of the N consecutive cycles. Since the variation trend of the load of the processor core in N consecutive cycles has been described above, it will not be repeated here. It should be noted that the representation of the value of the load of the processor core in the last cycle in N consecutive cycles needs to be the same as the representation of the change trend of the load of the processor core in N consecutive cycles, so that the XOR operation.
  • the change trend of the load of the processor core in N consecutive cycles is represented by a series of numbers
  • the series of numbers includes 8 digits and each digit is 0 or 1
  • the load of the processor core is
  • the value of the last cycle of N consecutive cycles is also converted into a series of 8-bit numbers, each of which has a value of 1 or 0, for XOR operation.
  • the conversion method may be, for example, a hash algorithm, etc., which is not specifically limited in this application.
  • the quantity of the second index information obtained after the XOR operation is also T N .
  • N and T have been described above, so they will not be repeated here.
  • a correction value is used to indicate the percentage change of the value of the load of the processor core in the third cycle compared to the value of the load of the processor core in the fourth cycle, wherein the third cycle is the value of the second index information corresponding to the correction value.
  • the next cycle of the indicated N consecutive cycles, and the fourth cycle is the last cycle among the N consecutive cycles indicated by the second index information corresponding to the correction value.
  • the percentage change of the value of the load of the processor core in the third cycle compared to the value of the load of the processor core in the fourth cycle is the value of the load of the processor core in the third cycle and the load of the processor core in the fourth cycle The ratio of the difference of the values to the value of the load on the processor core in the fourth cycle.
  • the principle of determining the correction value of the load of the processor core 0 in the next cycle is to obtain the change trend of the load of the processor core 0 in N consecutive cycles, wherein the last cycle in the N consecutive cycles is the current cycle; obtain the value of the load of processor core 0 in the current cycle; perform the XOR operation on the change trend of the load of processor core 0 in N consecutive cycles and the value of the load of processor core 0 in the current cycle to obtain the first Second, the index information to be compared is matched with the second index information in the second correction table, and the correction value corresponding to the second index information that matches the second index information to be compared is determined as The correction value of the load of processor core 0 in the next cycle.
  • the predicted value of the power consumption of core 0 in the next cycle is determined by the following formula:
  • P(t+1) is the predicted value of the power consumption of processor core 0 in the next cycle
  • P(t+1) / is the first value of the power consumption of processor core 0 in the next cycle
  • ⁇ A(t +1) is the correction value of the load of processor core 0 in the next cycle. It should be noted that if the change trend of the load of processor core 0 in the next cycle is smaller or unchanged, then before ⁇ A(t+1) Take a minus sign. If the load of processor core 0 has a trend of increasing in the next cycle, take a plus sign before ⁇ A(t+1).
  • the first value of the power consumption of the processor core 0 in the next cycle is determined, and the change trend and correction value of the load of the processor core 0 in the next cycle are determined according to the first calibration table and the second calibration table, and Correct the first value of the power consumption of processor core 0 in the next cycle according to the change trend of the load of processor core 0 in the next cycle and the correction value, so as to obtain the predicted value of the power consumption of processor core 0 in the next cycle, and provide A method for determining the predicted value of the power consumption of the processor core 0 in the next cycle is provided, and the determination method is simple and easy to execute, which improves the efficiency of determining the predicted value of the power consumption of the processor core 0 in the next cycle;
  • the voltage and power of the processor core 0 in the next cycle are considered, that is, the working state of the processor core 0 in the next cycle is considered, so , which improves the accuracy of determining
  • the change trend of the load of the processor core in one cycle includes two types: larger, smaller or unchanged. That is, the value of T is 2, and the load of the processor core changes in one cycle. The trend is represented by 0 or 1. If the change trend of the load of the processor core in a cycle is 0, the value of the load of the processor core in this cycle is less than or equal to the value of the load of the processor core in the previous cycle of the cycle.
  • first index information of 2 8 wherein the index information from the first 8 digits, each digit of the value is 1 or 0, index information for indicating a first processor core load 8 A variation trend in consecutive cycles, and the 8-bit numbers in the first index information correspond to 8 consecutive cycles one-to-one, and each number represents a variation trend of the load of the processor core in its corresponding cycle.
  • the first change trend also includes two situations of becoming larger, smaller or unchanged.
  • the first change trend is represented by 0 or 1, wherein if the first change trend is 0, the value of the load of the processor core in the first cycle is less than or equal to the value of the load of the processor core in the second cycle (that is, the first change If the first change trend is 1, the value of the load of the processor core in the first cycle is greater than the value of the load of the processor core in the second cycle (that is, the first change trend is to increase ). Since the number of first index information of 2 8, and therefore, the number of the first trend is also 2. 8.
  • the first change trend consists of one digit, and the value of this digit is 1 or 0.
  • Table 1 is the first calibration table provided in the embodiment of the present application.
  • first index information The first trend 00000000 0 ... ... 00110010 1 ... ... 01100101 0 ... ... 10011001 1 ... ... 11111110 1 11111111 0
  • the first calibration table of the initial state is generated. Specifically, since the first index information consists of 8 digits, and the value of each digit is 1 or 0, the 2 8 first index information are respectively: 00000000, 00000001, 00000010, 00000011, ..., 11111110, 11111111.2 8 pieces of first index information constitute the first column of the first correction table.
  • the initial value of the first change trend corresponding to each first index information is set to obtain the second column of the first correction table.
  • the initial value of the first change trend corresponding to the first index information may be set to 0 or 1, etc., which is not specifically limited in this application. It should be noted that, in other embodiments of the present application, the initial value of the first change trend corresponding to the first index information may not be set.
  • the first correction table in the initial state is trained to obtain the first correction table shown in Table 1 above.
  • the specific process includes: acquiring a large amount of training data, and training the first correction table in the initial state through the large amount of training data to obtain the first correction table shown in Table 1 above.
  • the current cycle is the sth cycle in the running process of the processor core
  • the values of the power consumption, voltage and frequency of the processor core in each cycle from the sth cycle to the s-8th cycle are obtained.
  • the value of the load of the processor core in each cycle is determined. It should be noted that, since the principle of determining the value of the load of the processor core in one cycle has been described above, it will not be repeated here.
  • the eight changing trends of the load of the processor core are combined according to the sequence of the number of the corresponding period from small to large to obtain the first index information to be compared.
  • the first index information to be compared is matched with the first index information in Table 1, and the first change trend corresponding to the first index information matched with the first index information to be compared is determined as the load of the processor core below
  • the change trend of a cycle ie, the s+1th cycle).
  • the value of the load of the processor core in the next cycle is less than or equal to the value of the load of the processor core in the current cycle. If the change trend of the load in the next cycle is 1, the value of the load of the processor core in the next cycle is greater than the value of the load of the processor core in the current cycle.
  • the first index information to be compared is 01100101. It can be seen from Table 1 that the change trend corresponding to the first index information that matches the first index information to be compared is 0, that is, the load of the processor core is 0 in the next cycle. The change trend is 0, that is, the value of the load of the processor core in the next cycle is less than or equal to the value of the load of the processor core in the current cycle.
  • the value of the load of the processor core in each cycle is calculated in each cycle during the operation of the processor core, and the processing value is determined according to the value of the load of the processor core in each cycle
  • the change trend of the load of the processor core in each cycle, and the change trend of the load of the processor core in the last 8 consecutive cycles is determined according to the change trend of the load of the processor core in each cycle, and the processor core is always saved.
  • the change trend in the last 8 consecutive cycles so, when determining the change trend of the load of the processor core in the next cycle based on Table 1, because the load of the processor core in the last 8 consecutive cycles is always saved change trend, so the change trend of the load of the processor core in the last 8 consecutive cycles can be directly used as the first index information to be compared, so as to determine the load of the processor core in the next cycle according to the first index information to be compared.
  • the change trend of improves the efficiency of determining the change trend of the load of the processor core in the next cycle.
  • the change trend of the load of the processor core from the 7th cycle to the 14th cycle will be saved.
  • the change trend of the load of the core from the 7th cycle to the 14th cycle is used as the first index information to be compared, so as to determine the change trend of the load of the processor core in the next cycle (ie, the 15th cycle).
  • the next cycle (the 15th cycle) comes, measure the power consumption, frequency and voltage of the processor core in the next cycle, and determine the load of the processor core according to the power consumption, frequency and voltage of the processor core in the next cycle
  • the value in the next cycle compare the value of the load of the processor core in the next cycle with the value of the load of the processor core in the current cycle to determine the change trend of the load of the processor core in the next cycle, and delete the first pending Compare the first digit of the index information, and supplement the change trend of the load of the processor core in the next cycle after deleting the first index information to be compared with the first digit, so as to obtain a new first to be compared. index information, and determining a change trend of the load of the processor core in the next cycle according to the new first index information to be compared.
  • the values of the power consumption, frequency and voltage of the processor core in the next cycle are obtained, and the values of the power consumption, frequency and voltage of the processor core in the next cycle are obtained.
  • Determine the value of the load of the processor core in the next cycle and determine the change trend of the load of the processor core in the next cycle according to the relationship between the value of the processor core load in the next cycle and the value of the processor core load in the current cycle , and determine whether the change trend of the load of the processor core in the next cycle is the same as the first change trend corresponding to the first index information matching the first index information to be compared in Table 1.
  • the first change trend corresponding to the first index information matched by the index information is corrected.
  • the first index matching the first index information to be compared is corrected according to the change trend of the load of the processor core in the next cycle.
  • the first change trend corresponding to the information is corrected.
  • the corresponding first change trend in the first correction table is corrected by the measured value, which further improves the accuracy of the first correction table.
  • the first index information to be compared is 11111110.
  • the first change trend corresponding to the first index information that matches the first index information to be compared is 1. If the next cycle comes , the value of the load of the processor core in the next cycle is less than the value of the load of the processor core in the current cycle, that is, the actual change trend of the load of the processor core in the next cycle is 0. The first change trend corresponding to the first index information matched by the comparison index information is corrected from 1 to 0.
  • the revised Table 1 is as follows.
  • the bolded items in the revised Table 1 are the first index information matching the first index information to be compared and the corresponding first change trend.
  • the change trend of the load of the processor core in the corresponding cycle includes two types: larger, smaller or unchanged. That is, if the value of T is 2, the load of the processor core changes in one cycle. The trend is represented by 0 or 1. If the change trend of the load of the processor core in a cycle is 0, the value of the load of the processor core in this cycle is less than or equal to the value of the load of the processor core in the previous cycle of the cycle.
  • the number of variation trends of the load of the processor core in 8 consecutive cycles is 2 8 , wherein each variation trend of the load of the processor core in the 8 consecutive cycles consists of 8 digits, and each bit The value of the number is 1 or 0, and the 8-digit number corresponds to 8 consecutive cycles one-to-one, and each number represents the change trend of the load of the processor core in its corresponding cycle.
  • the value of the load of the processor core in the last cycle of 8 consecutive cycles is also converted into an 8-bit number by a hash algorithm, and each digit takes the value of 1 or 0.
  • the second index information obtained after the XOR operation is performed on the variation trend of the load of the processor core in 8 consecutive cycles and the value of the load of the processor core in the last cycle of the 8 consecutive cycles is obtained.
  • the number is also 2 8 , wherein the second index information is composed of 8 digits, and each digit takes a value of 1 or 0.
  • Table 2 is a second correction table provided in this embodiment of the present application.
  • a second correction table of the initial state is generated. Specifically, since the second index information consists of 8 digits, and the value of each digit is 1 or 0, the 2 8 second index information are respectively: 00000000, 00000001, 00000010, 00000011, ..., 11111110, 11111111.2 8 pieces of second index information constitute the first column of the second correction table.
  • the initial value of the correction value corresponding to each second index information is set to obtain the second column of the second correction table.
  • the initial value of the correction value corresponding to the second index information may be set to any value, which is not specifically limited in this application. It should be noted that, in other embodiments of the present application, the initial value of the correction value corresponding to the second index information may not be set.
  • the second correction table in the initial state is trained to obtain the second correction table shown in Table 2 above.
  • the specific process includes: acquiring a large amount of training data, and training the second correction table in the initial state through the large amount of training data to obtain the second correction table shown in Table 2 above.
  • the value of the power consumption, frequency, and voltage of the processor core in each cycle in 10 consecutive cycles (ie, one training data), and calculate the value of the processor core's frequency in one cycle and the processor core's voltage in that cycle.
  • the product of the square of the value of , and the ratio of the power consumption of the processor core in the cycle to the product represents the value of the load of the processor core in the cycle, that is, it can be determined that the ratio is the load of the processor core in the cycle. value of .
  • the first sequence includes 8 digits, and each digit takes a value of 0 or 1.
  • the third sequence is obtained by XORing the first sequence and the second sequence.
  • the ratio is determined as the percentage of the value of the load on the processor core at the 10th cycle compared to the value of the load on the processor core at the 9th cycle.
  • the current cycle is the sth cycle in the running process of the processor core
  • the values of the power consumption, voltage and frequency of the processor core in each cycle from the sth cycle to the s-8th cycle are obtained.
  • the value of the load of the processor core in each cycle is determined. It should be noted that, since the principle of determining the value of the load of the processor core in one cycle has been described above, it will not be repeated here.
  • the change trend of the load of the processor core in 8 consecutive cycles is 00110010
  • the hash value of the value of the processor core's load in the sth cycle is 00000110.
  • the second index information to be compared obtained after the XOR operation of the above two values is performed is 00110100. It can be seen from Table 2 that the correction value corresponding to the second index information matching the second index information to be compared is 8%, that is, the correction value of the load of the processor core in the next cycle is 8%, that is, the load of the processor core is 8%.
  • the value at the next cycle has a percentage change of 8% compared to the value of the load on the processor core at the current cycle.
  • the value of the load of the processor core in each cycle is calculated in each cycle during the operation of the processor core, and the processing value is determined according to the value of the load of the processor core in each cycle
  • the change trend of the load of the processor core in each cycle, and the change trend of the load of the processor core in the last 8 consecutive cycles is determined according to the change trend of the load of the processor core in each cycle, and the processor core is always saved.
  • the change trend in the last 8 consecutive cycles in this way, when determining the correction value of the load of the processor core in the next cycle based on Table 2, because the load of the processor core in the last 8 consecutive cycles is always saved change trend, so the change trend of the load of the processor core in the last 8 consecutive cycles can be directly used as the fourth sequence, and the value of the load of the processor core in the current cycle can be hashed to obtain the fifth sequence, XOR operation is performed on the fourth sequence and the fifth sequence to obtain the second index information to be compared, so as to determine the correction value of the load of the processor core in the next cycle according to the second index information to be compared, because the processor is stored in advance.
  • the variation trend of the core load in the last 8 consecutive cycles does not need to be recalculated, thus improving the efficiency of determining the correction value of the processor core load in the next cycle.
  • the processor core can be calculated The hash value of the core load value in the 16th cycle, and the XOR operation is performed on the hash value and the change trend of the processor core load in the 9th cycle to the 16th cycle, and the second index information to be compared is obtained. , so as to determine the correction value of the load of the processor core in the next cycle (ie, the 17th cycle) according to the second index information to be compared.
  • the trend determines the second index information to be compared, and the correction value of the load of the processor core in the next cycle is determined according to the second index information to be compared.
  • the values of the power consumption, frequency and voltage of the processor core in the next cycle are obtained, and the values of the power consumption, frequency and voltage of the processor core in the next cycle are obtained.
  • Determine the value of the load of the processor core in the next cycle calculate the value of the load of the processor core in the next cycle according to the value of the load of the processor core in the next cycle and the value of the load of the processor core in the current cycle.
  • the percentage of the value of the core load in the current cycle, and whether the calculated percentage is the same as the correction value corresponding to the second index information that matches the second index information to be compared. If they are the same, do not match the second index to be compared.
  • the correction value corresponding to the second index information whose information is matched is corrected. If they are not the same, the correction value corresponding to the second index information matching the second index information to be compared is corrected by the calculated percentage.
  • the second index information to be compared is 11011010.
  • the correction value corresponding to the second index information that matches the second index information to be compared is 5%. If the next cycle comes, The percentage of the value of the load of the processor core in the next cycle compared to the value of the load of the processor core in the current cycle is 8%, then in Table 2, the second index information that matches the second index information to be compared The corresponding correction value was revised from 5% to 8%.
  • the revised Table 2 is as follows, and the bolded items in the revised Table 2 are the correction values corresponding to the second index information matching the second index information to be compared.
  • the second correction table may also be a model obtained by training a neural network, or the like.
  • FIG. 3 is a schematic flowchart of determining a predicted value of the temperature of the processor core 0 in the next cycle according to an embodiment of the present application. As shown in Figure 3, including:
  • Step 301 the sensing device in the processor core 0 obtains the measured value of the temperature of the processor core 0 in the current cycle. Since the principle of acquiring the measured value of the temperature of the processor core 0 in the current cycle is the same as the principle of acquiring the measured value of the power consumption of the processor core 0 in the current cycle, it is not repeated here. It should be noted that the sensing device may be, for example, a temperature sensor (T-sensor), which is not particularly limited in the embodiment of the present application.
  • T-sensor temperature sensor
  • Step 302 the sensing device in the processor core 0 sends the measured value of the temperature of the processor core 0 in the current cycle to the control logic.
  • Step 303 The control logic determines the predicted value of the temperature of the processor core 0 in the next cycle. Specifically, it can be determined in the following two ways.
  • the first type the predicted value of the temperature of processor core 0 in the next cycle is determined according to the measured value of the temperature of processor core 0 in the current cycle.
  • the specific process includes:
  • the measured values of the power consumption of the processor core 0 in the current cycle and the previous cycle, and the predicted value of the power consumption of the processor core 0 in the next cycle are obtained. Since the principle of obtaining the measured value of the power consumption of the processor core 0 in the current cycle has been described above, it will not be repeated here. Since the principle of obtaining the measured value of the power consumption of the processor core 0 in the previous cycle is the same as the above-mentioned principle of obtaining the measured value of the power consumption of the processor core 0 in the current cycle, it will not be repeated here. Since the principle of obtaining the predicted value of the power consumption of the processor core 0 in the next cycle has been described above, it will not be repeated here.
  • the predicted value of the power consumption of the processor core 0 in the next cycle determines the prediction of the temperature of the processor core 0 in the next cycle. value. Since the increase in power consumption in a short period of time is positively correlated with the increase in temperature, the predicted value of the temperature of processor core 0 in the next cycle can be determined by the following formula:
  • T(t+1) is the predicted value of the temperature of the processor core 0 in the next cycle
  • T(t) is the measured value of the temperature of the processor core 0 in the current cycle
  • T(t-1) is the processor core
  • P(t+1) is the predicted value of the power consumption of processor core 0 in the next cycle
  • P(t) is the measured value of the power consumption of processor core 0 in the current cycle
  • P(t-1) is the measured value of the power consumption of processor core 0 in the previous cycle.
  • the measured value of the power consumption of processor core 0 in the current cycle and the previous cycle, the predicted value of the power consumption of processor core 0 in the next cycle, and the temperature of processor core 0 in the current cycle and the previous cycle are obtained.
  • the predicted value of the temperature of the processor core 0 in the next cycle can be determined according to the measured values and predicted values obtained above, and a method for determining the predicted value of the temperature of the processor core 0 in the next cycle is provided.
  • the determination method is simple and easy to implement, which improves the efficiency of determining the predicted value of the temperature of the processor core 0 in the next cycle.
  • the predicted value of the power consumption of the processor core 0 in the next cycle is considered, that is, the working state of the processor core 0 in the next cycle is considered, so , which improves the accuracy of determining the predicted value of the temperature of processor core 0 for the next week.
  • the second according to the measured value of the temperature of the processor core 0 in the current cycle and combined with a calibration table to determine the predicted value of the temperature of the processor core 0 in the next cycle, wherein the calibration table includes a first calibration table and a second calibration table , the first calibration table is used to predict the change trend of the load of the processor core in the next cycle, and the second calibration table is used to predict the calibration value of the load of the processor core in the next cycle.
  • the specific process includes:
  • the first value of the power consumption of the processor core 0 in the next cycle is determined according to the measured value of the power consumption of the processor core 0 in the current cycle. Since the principle of this step has been described above, it will not be repeated here.
  • the change trend of the load of the processor core 0 in the next cycle is determined according to the first calibration table, and the calibration value of the load of the processor core 0 in the next cycle is determined according to the second calibration table.
  • the predicted value of the power consumption of the processor core 0 in the next cycle the measured value of the power consumption of the processor core 0 in the current cycle and the previous cycle, and the temperature of the processor core 0 in the current cycle and the previous cycle. value to determine the predicted value of the temperature of processor core 0 in the next cycle. Specifically, the predicted value of the temperature of processor core 0 in the next cycle is determined according to the following formula:
  • T(t+1) is the predicted value of the temperature of the processor core 0 in the next cycle
  • T(t) is the measured value of the temperature of the processor core 0 in the current cycle
  • T(t-1) is the processor core
  • P(t+1) is the predicted value of the power consumption of processor core 0 in the next cycle
  • P(t) is the measured value of the power consumption of processor core 0 in the current cycle
  • P(t-1) is the measured value of the power consumption of processor core 0 in the previous cycle.
  • a method for determining the predicted value of the temperature of the processor core 0 in the next cycle is provided, and The determination method is simple and easy to execute, which improves the efficiency of determining the predicted value of the temperature of the processor core 0 in the next cycle.
  • the value of the power consumption of the processor core 0 in the next cycle is considered, that is, the working state of the processor core 0 in the next cycle is considered, therefore, Improved the accuracy of determining the predicted value of processor core 0 temperature for the next week.
  • the influence of the change of the load of the processor core 0 in the next cycle on the predicted value of the power consumption of the processor core 0 in the next cycle is considered.
  • the control logic can also determine the adjustment strategy of each processor core according to the predicted value of the variable to be predicted of each processor core in the next cycle,
  • the scaling policy is used to indicate how the voltage and/or frequency of the processor core is scaled.
  • the control logic respectively generates a frequency adjustment instruction and/or a voltage adjustment instruction of the corresponding processor core according to the adjustment policy of each processor core.
  • the voltage and/or frequency of the corresponding processor core is adjusted through the adjustment policy of each processor core in the next cycle to ensure that the processor operates within a safe range.
  • control logic is further configured to send the frequency adjustment instruction and/or the voltage adjustment instruction of each processor core to the frequency and voltage adjustment circuit of the corresponding processor core;
  • Each frequency and voltage adjustment circuit adjusts the frequency and/or voltage of the corresponding processor core according to the frequency adjustment instruction and/or the voltage adjustment instruction of the corresponding processor core.
  • the control logic calls the adjustment strategy to reduce the power consumption of the processor core, and in the next cycle
  • a voltage regulation command is generated according to the regulation strategy to reduce the power consumption of the processor core, and the voltage regulation command is sent to the frequency voltage regulation circuit of the processor core, and the frequency voltage regulation circuit adjusts the processor core according to the voltage regulation command.
  • voltage so as to reduce the temperature of the processor core by reducing the voltage of the processor core, so as to ensure that the temperature of the processor core is below a preset level, thereby ensuring that the processor operates at a normal temperature.
  • control logic may separately determine the adjustment strategy of each processor core according to the predicted value of the variable to be predicted of each processor core in the next cycle.
  • the control logic may also comprehensively consider the predicted value of the variable to be predicted of each processor core in the next cycle to determine the adjustment strategy of each processor core.
  • the adjustment strategy is determined by the predicted value of the variable to be predicted of the processor core in the next cycle, and the predicted value of the variable to be observed of the processor core in the next cycle takes into account the running state of the processor core in the next cycle, so The adjustment strategy also considers the impact of the operating state of the processor core in the next cycle, so that when the frequency and/or voltage of the processor core in the next cycle is adjusted according to the adjustment strategy, the adjustment strategy can cope with the processor's next cycle. One-cycle working state, so that the processor can run normally and avoid the risk of overcurrent.
  • the processor may also include more or fewer processor cores, ie the processor includes at least one processor core.
  • sensing devices may also be included in the processor.
  • the number of sensing devices is the same as the number of processor cores in the processor, and the sensing devices are in one-to-one correspondence with the processor cores, and the predicted value of the variable to be predicted of the corresponding processor core in the next cycle is obtained through each sensing device .
  • the number of sensing devices is one, the processor cores in the processor share the sensing device, and the sensing device can obtain the measured values of the variables to be predicted in the current cycle of all the processor cores of the processor, that is, the processor can be All the processor cores in the system are regarded as a whole to obtain the measured value of the overall variable to be predicted in the current cycle, so as to predict the predicted value of the entire variable to be predicted in the next cycle based on the above principles; or the one sensing device can also be separately Obtain the measured value of the variable to be predicted in the current cycle of each processor core in the processor, so as to determine the predicted value of the variable to be predicted in the next cycle based on the above-mentioned principle.
  • the measured value of the processor core's variable to be predicted in the current cycle may share one sensing device, and another part of the processor cores may share another sensing device, which is not particularly limited in this embodiment of the present application.
  • More or fewer frequency voltage regulation circuits may also be included in the processor.
  • the number of frequency voltage adjustment circuits is the same as the number of processor cores in the processor, and the frequency voltage adjustment circuits are in one-to-one correspondence with the processor cores, so that the frequency voltage adjustment circuits adjust the frequency and/or frequency of the corresponding processor cores Voltage.
  • the number of the frequency voltage adjustment circuit is one, the processor cores in the processor share the frequency voltage adjustment circuit, and the frequency voltage can adjust the frequency and/or voltage of each processor core in time division.
  • a part of the processor cores in the processor share one frequency voltage adjustment circuit, and another part of the processor cores in the processor share another frequency voltage adjustment circuit.
  • the processor may further include more or less components, which are not particularly limited in the embodiments of the present application.
  • the predicted value of the variable to be predicted in the next cycle of each component in the processor other than the processor core may also be determined based on the above-mentioned principle.
  • Each component in the processor other than the processor core may share an induction device and a frequency voltage adjustment circuit, or may be provided with an induction device and a frequency voltage adjustment circuit for each component, etc., which are not made in this embodiment of the present application. Special restrictions.
  • FIG. 4 is a schematic flowchart of a method for predicting a variable provided by an embodiment of the present application.
  • the execution body of the method may be, for example, a sensing device and a control logic in a processor, which are not particularly limited in the embodiment of the present application. As shown in Figure 4, the method includes the following steps:
  • Step 401 obtaining the measured value of the variable to be predicted of the at least one processor core in the current cycle
  • Step 402 Determine the predicted value of the variable to be predicted of the at least one processor core in the next cycle according to the measured value of the variable to be predicted of the at least one processor core in the current cycle.
  • the variable to be predicted is the power consumption of the processor core; the at least one process is determined according to the measured value of the variable to be predicted of the at least one processor core in the current cycle
  • the predicted value of the variable to be predicted of the processor core in the next cycle includes: obtaining the measured value of the frequency and voltage of the at least one processor core in the current cycle; determining the frequency and voltage of the at least one processor core in the next cycle.
  • the values of the frequency and voltage at the next cycle determine a predicted value of the power consumption of the at least one processor core at the next cycle.
  • the predicted value of the power consumption of the processor core in the next cycle is determined according to the following formula:
  • P(t+1) is the predicted value of the power consumption of the processor core in the next cycle
  • P(t) is the measured value of the power consumption of the processor core in the current cycle
  • f (t+1) is the value of the frequency of the processor core in the next cycle
  • f(t) is the measured value of the frequency of the processor core in the current cycle
  • V(t+1) is The value of the voltage of the processor core in the next cycle
  • V(t) is the measured value of the voltage of the processor core in the current cycle.
  • the method further includes: acquiring a measured value of the temperature of the at least one processor core in the current cycle; acquiring a measured value of the temperature of the at least one processor core in a previous cycle; Obtain the measured value of the power consumption of the at least one processor core in the previous cycle; according to the measured value of the power consumption and temperature of the at least one processor core in the previous cycle and the current cycle, the The predicted value of the power consumption of at least one processor core in the next cycle is determined, and the predicted value of the temperature of the at least one processor core in the next cycle is determined.
  • the predicted value of the temperature of the processor core in the next cycle is determined according to the following formula:
  • T(t+1) is the predicted value of the temperature of the processor core in the next cycle
  • T(t) is the measured value of the temperature of the processor core in the current cycle
  • T(t -1) is the measured value of the temperature of the processor core in the previous cycle
  • P(t+1) is the predicted value of the power consumption of the processor core in the next cycle
  • P(t) is the measured value of the power consumption of the processor core in the current cycle
  • P(t-1) is the measured value of the power consumption of the processor core in the previous cycle.
  • the determining, according to the measured value of the variable to be predicted of the at least one processor core in the current cycle, the predicted value of the variable to be predicted of the at least one processor core in the next cycle includes: according to The measured value of the variable to be predicted of the at least one processor core in the current cycle is combined with a calibration table to determine the predicted value of the variable to be predicted of the at least one processor core in the next cycle.
  • the variable to be predicted is the power consumption of the processor core
  • the correction table includes a first correction table and a second correction table
  • Determining the predicted value of the variable to be predicted in the next cycle by combining the measured value of the variable to be predicted in the current cycle with a correction table in the next cycle includes: according to the power consumption of the at least one processor core in the current cycle The measured value of the at least one processor core determines the first value of the power consumption of the at least one processor core in the next cycle; according to the first correction table, it is determined that the load of the at least one processor core is in the next cycle.
  • the correction value of the load of the processor core corrects the first value of the power consumption of the at least one processor core to obtain a prediction of the power consumption of the at least one processor core in the next cycle value.
  • the method further includes: acquiring a measured value of the power consumption of the at least one processor core in the previous cycle; acquiring the temperature of the at least one processor core in the current cycle respectively and the measured value of the previous cycle; according to the predicted value of the power consumption of the at least one processor core in the next cycle, the power consumption of the at least one processor core is between the current cycle and the previous cycle.
  • the measured value, the measured value of the temperature of the at least one processor core in the current cycle and the previous cycle determines the predicted value of the temperature of the at least one processor core in the next cycle.
  • the first correction table comprises a first index number T N T N of first information and trends, wherein: the index information of first T N T N-th and the One change trend corresponds one to one; the first index information is used to indicate a change trend of the load of the processor core in N consecutive cycles, and T is the change of the load of the processor core in one cycle
  • the number of types of trends; a first change trend is used to indicate the change trend of the value of the load of the processor core in the first cycle compared to the value of the load of the processor core in the second cycle, wherein the The first period is the next period of N consecutive periods indicated by the first index information corresponding to the first change trend, and the second period is N indicated by the first index information corresponding to the first change trend
  • the second correction table includes T N second index information and T N correction values, wherein: the T N second index information and the T N correction values are equal to one One correspondence; one of the second index information is used to indicate that the change trend of the load of the processor core in N consecutive
  • a correction value is used to indicate the percentage change of the value of the load of the processor core in the third cycle compared to the value of the load of the processor core in the fourth cycle, wherein the The third period is the next period of N consecutive periods indicated by the second index information corresponding to the correction value, and the fourth period is the N consecutive periods indicated by the second index information corresponding to the correction value. in the last cycle.
  • the method further includes: determining an adjustment strategy of the at least one processing core according to the predicted value of the variable to be predicted of the at least one processor core in the next cycle, wherein the adjustment strategy Used to indicate how the voltage and/or frequency of the processor core is adjusted.
  • the determining the adjustment strategy of the at least one processing core according to the predicted value of the variable to be predicted of the at least one processor core in the next cycle includes: according to the prediction value of the at least one processor core
  • the adjustment strategy generates a frequency adjustment instruction and/or a voltage adjustment instruction of the at least one processor core; adjusts the frequency and/or voltage adjustment instruction of the at least one processor core according to the frequency adjustment instruction and/or the voltage adjustment instruction of the at least one processor core. / or voltage.
  • the present application further provides a computer-readable storage medium, including a computer program, which, when executed on a computer, causes the computer to execute the technical solutions of any one of the above method embodiments.
  • the present application also provides a computer program, which, when the computer program is executed by a computer, is used to execute the technical solution of any one of the above method embodiments.
  • the disclosed system, apparatus and method may be implemented in other manners.
  • the apparatus embodiments described above are only illustrative.
  • the division of the units is only a logical function division. In actual implementation, there may be other division methods.
  • multiple units or components may be combined or Can be integrated into another system, or some features can be ignored, or not implemented.
  • the shown or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, indirect coupling or communication connection of devices or units, and may be in electrical, mechanical or other forms.
  • the units described as separate components may or may not be physically separated, and components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution in this embodiment.
  • each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit.
  • the functions, if implemented in the form of software functional units and sold or used as independent products, may be stored in a computer-readable storage medium.
  • the technical solution of the present application can be embodied in the form of a software product in essence, or the part that contributes to the prior art or the part of the technical solution.
  • the computer software product is stored in a storage medium, including Several instructions are used to cause a computer device (personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application.
  • the aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk and other media that can store program codes .

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Power Sources (AREA)

Abstract

一种处理器和一种变量的预测方法，其中，处理器包括：至少一个处理器核心、控制逻辑和至少一个感应装置；至少一个感应装置，用于获取至少一个处理器核心的待预测变量在当前周期的测量值；控制逻辑，用于根据至少一个处理器核心的待预测变量在当前周期的测量值确定至少一个处理器核心的待预测变量在下一周期的预测值。所述处理器及预测方法提供了一种确定待预测变量在下一周期的预测值的方式，以使根据待预测变量在下一周期的预测值确定的调节策略可以应对处理器在下一周期的运行状态。

Description

处理器和变量的预测方法 技术领域
本申请涉及计算机技术领域，尤其涉及一种处理器和一种变量的预测方法。
背景技术
在具备Turbo功能的处理器中，通常存在基于各项变量(例如功耗、温度)的调节策略。通过调节策略可以保证处理器在运行的过程中，不会发生过流或过温等异常状态，从而保证处理器在安全范围内运行。
例如，基于功耗的调节策略的调节过程包括：确定处理器中的每个处理器核心的功耗在当前周期的测量值与预设水线的关系，以及根据每个处理器核心的功耗在当前周期的测量值与预设水线的关系确定每个处理器核心的调节策略，以及根据每个处理器核心的调节策略调节对应处理器核心在下一周期的运行状态，进而确保处理器运行在安全范围内。然而，由于在下一周期内，处理器可能存在超出预期的行为，例如，对一个给定32核的处理器，若当前周期处理器的全部处理器核心均处在功耗较低的状态，那么在系统请求处理器的全部处理器核心在下一周期均进入Turbo状态时，由于当前周期的处理器的全部处理器核心均处在功耗较低的状态，因此，确定的调节策略判断功耗存在裕量，并允许全部处理器核心进入Turbo状态，显然，当前全部处理器核心进入Turbo状态，处理器存在过流的风险。
显然，由于处理器核心的功耗在当前周期的测量值未考虑处理器核心在下一周期的运行状态，因此，通过上述方式确定的调节策略可能无法应对处理器在下一周期的运行状态，从而导致处理器存在过流的风险，影响处理器的正常运行，同理，针对其他变量也存在类似的问题。
发明内容
本申请提供一种处理器和一种变量的预测方法，用于解决由于变量在当前周期的测量值未考虑处理器核心在下一周期的运行状态，而导致的根据变量在当前周期的测量值确定的调节策略无法应对处理器在下一周期的运行状态的问题。
第一方面，提供一种处理器，包括至少一个处理器核心、控制逻辑和至少一个感应装置；所述至少一个感应装置，用于获取所述至少一个处理器核心的待预测变量在当前周期的测量值；所述控制逻辑，用于根据所述至少一个处理器核心的待预测变量在当前周期的测量值确定所述至少一个处理器核心的待预测变量在下一周期的预测值。
提供了一种确定处理器核心的待预测变量在下一周期的预测值的方法，从而使得基于待预测变量在下一周期的预测值确定的调节策略能够应对处理器在下一周期的运行状态，进而确保处理器在安全范围内运行。
在一种可能的实现方式中，所述待预测变量为所述处理器核心的功耗；所述根据所述 至少一个处理器核心的待预测变量在当前周期的测量值确定所述至少一个处理器核心的待预测变量在下一周期的预测值包括：获取所述至少一个处理器核心的频率、电压在当前周期的测量值；确定所述至少一个处理器核心的频率、电压在所述下一周期的值；根据所述至少一个处理器核心的功耗在所述当前周期的测量值、所述至少一个处理器核心的频率和电压在当前周期的测量值以及所述至少一个处理器核心的频率和电压在所述下一周期的值确定所述至少一个处理器核心的功耗在所述下一周期的预测值。
根据至少一个处理器核心的功耗在当前周期的测量值、至少一个处理器核心的频率和电压在当前周期的测量值以及至少一个处理器核心的频率和电压在下一周期的值确定至少一个处理器核心的功耗在所述下一周期的预测值，提供了一种确定处理器核心的功耗在下一周期的预测值的方式，且确定方式简单，易于执行，提高了确定处理器核心的功耗在下一周期的预测值的效率。此外，由于在确定至少一个处理器核心的功耗在下一周期的预测值时，考虑了至少一个处理器核心的电压、功率在下一周期的值，即考虑了至少一个处理器核心在下一周期的工作状态，因此，提高了确定处理器核心的功耗在下一周的预测值的准确性。
在一种可能的实现方式中，根据下述公式确定所述处理器核心的功耗在所述下一周期的预测值：
Figure PCTCN2020098564-appb-000001
其中：P(t+1)为所述处理器核心的功耗在所述下一周期的预测值，P(t)为所述处理器核心的功耗在所述当前周期的测量值，f(t+1)为所述处理器核心的频率在所述下一周期的值，f(t)为所述处理器核心的频率在所述当前周期的测量值，V(t+1)为所述处理器核心的电压在所述下一周期的值，V(t)为所述处理器核心的电压在所述当前周期的测量值。
在一种可能的实现方式中，所述控制逻辑还用于：获取所述至少一个处理器核心的温度在当前周期的测量值；获取所述至少一个处理器核心的温度在前一周期的测量值；获取所述至少一个处理器核心的功耗在前一周期的测量值；根据所述至少一个处理器核心的功耗和温度分别在所述前一周期和所述当前周期的测量值、所述至少一个处理器核心的功耗在所述下一周期的预测值，确定所述至少一个处理器核心的温度在所述下一周期的预测值。
根据至少一个处理器核心的功耗和温度分别在前一周期和所述当前周期的测量值、至少一个处理器核心的功耗在所述下一周期的预测值，确定至少一个处理器核心的温度在所述下一周期的预测值，提供了一种确定处理器核心的温度在下一周期的预测值的方式，且确定方式简单，易于执行，提高了确定处理器核心的温度在下一周期的预测值的效率。此外，由于在确定至少一个处理器核心的温度在下一周期的预测值时，考虑了至少一个处理器核心的功耗在下一周期的值，即考虑了至少一个处理器核心在下一周期的工作状态，因此，提高了确定处理器核心的温度在下一周的预测值的准确性。
在一种可能的实现方式中，根据下述公式确定所述处理器核心的温度在所述下一周期的预测值：
Figure PCTCN2020098564-appb-000002
其中，T(t+1)为所述处理器核心的温度在所述下一周期的预测值，T(t)为所述处理器核心的温度在所述当前周期的测量值，T(t-1)为所述处理器核心的温度在所述前一周期的测量值，P(t+1)为所述处理器核心的功耗在所述下一周期的预测值，P(t)为所述处理器核心的功耗在所述当前周期的测量值，P(t-1)为所述处理器核心的功耗在所述前一周期的测量值。
在一种可能的实现方式中，所述根据所述至少一个处理器核心的待预测变量在当前周期的测量值确定所述至少一个处理器核心的待预测变量在下一周期的预测值包括：根据所述至少一个处理器核心的待预测变量在当前周期的测量值并结合一校正表确定所述至少一个处理器核心的待预测变量在下一周期的预测值。
在一种可能的实现方式中，所述待预测变量为所述处理器核心的功耗，所述校正表包括第一校正表和第二校正表；所述根据所述至少一个处理器核心的待预测变量在当前周期的测量值并结合一校正表确定所述至少一个处理器核心的待预测变量在下一周期的预测值包括：根据所述至少一个处理器核心的功耗在所述当前周期的测量值确定所述至少一个处理器核心的功耗在所述下一周期的第一取值；根据所述第一校正表确定所述至少一个处理器核心的负载在所述下一周期的变化趋势；根据所述第二校正表确定所述至少一个处理器核心的负载在所述下一周期的校正值；根据所述至少一个处理器核心的负载的所述变化趋势和所述至少一个处理器核心的负载的所述校正值对所述至少一个处理器核心的功耗的所述第一取值进行校正，得到所述至少一个处理器核心的功耗在所述下一周期的预测值。
通过确定至少一个处理器核心的功耗在下一周期的第一取值，并根据第一校正表和第二校正表确定至少一个处理器核心的负载在下一周期的变化趋势和校正值，以及根据至少一个处理器核心的负载在下一周期的变化趋势和校正值校正处理器核心的功耗在下一周期的第一取值，以得到至少一个处理器核心的功耗在下一周期的预测值，提供了一种确定处理器核心的功耗在下一周期的预测值的方式，且确定方式简单，易于执行，提高了确定处理器核心的功耗在下一周期的预测值的效率；另外，由于根据至少一个处理器核心的负载在下一周期的变化趋势和校正值校正至少一个处理器核心的功耗在下一周期的第一取值，得到至少一个处理器核心的功耗在下一周期的预测值，即在确定至少一个处理器核心的功耗在下一周期的预测值的过程中，考虑了至少一个处理器核心的负载在下一周期的变化对至少一个处理器核心的功耗在下一周期的预测值的影响，提高了确定至少一个处理器核心的功耗在下一周期的预测值的准确性。
在一种可能的实现方式中，所述控制逻辑还用于：获取所述至少一个处理器核心的功耗在前一周期的测量值；获取所述至少一个处理器核心的温度分别在所述当前周期和所述前一周期的测量值；根据所述至少一个处理器核心的功耗在下一周期的预测值，所述至少一个处理器核心的功耗在所述当前周期和所述前一周期的测量值，所述至少一个处理器核心的温度在所述当前周期和所述前一周期的测量值，确定所述至少一个处理器核心的温度在下一周期的预测值。
根据至少一个处理器核心的功耗在下一周期的预测值，至少一个处理器核心的功耗在当前周期和前一周期的测量值，至少一个处理器核心的温度在当前周期和前一周期的测量值，确定至少一个处理器核心的温度在下一周期的预测值，提供了一种确定处理器核心的温度在下一周期的预测值的方式，且确定方式简单，易于执行，提高了确定处理器核心的 温度在下一周期的预测值的效率。另外，由于在确定处理器核心的温度在下一周期的预测值时，考虑了处理器核心的功耗在下一周期的值，即考虑了处理器核心在下一周期的工作状态，因此，提高了确定处理器核心的温度在下一周的预测值的准确性。此外，在确定处理器核心的功耗在下一周期的预测值的过程中，考虑了处理器核心的负载在下一周期的变化对处理器核心的功耗在下一周期的预测值的影响，进一步提高了确定处理器核心的功耗在下一周期的预测值的准确性，从而进一步提高了由处理器核心的功耗在下一周期的预测值确定的处理器核心的温度在下一周期的预测值的准确性。
在一种可能的实现方式中，所述第一校正表包括T N个第一索引信息和T N个第一变化趋势，其中：所述T N个第一索引信息与所述T N个第一变化趋势一一对应；一个所述第一索引信息用于指示所述处理器核心的负载在N个连续周期中的一种变化趋势，T为所述处理器核心的负载在一个周期的变化趋势的种类数量；一个第一变化趋势用于指示所述处理器核心的负载在第一周期的值相比于所述处理器核心的负载在第二周期的值的变化趋势，其中，所述第一周期为所述第一变化趋势对应的第一索引信息所指示的N个连续周期的下一个周期，所述第二周期为所述第一变化趋势对应的第一索引信息所指示的N个连续周期中的最后一个周期；所述第二校正表包括T N个第二索引信息和T N个校正值，其中：所述T N个第二索引信息与所述T N个校正值一一对应；一个所述第二索引信息用于指示所述处理器核心的负载在N个连续周期中的变化趋势与所述处理器核心的负载在N个连续周期中的最后一个周期的值进行异或运算后得到的结果；一个校正值用于指示所述处理器核心的负载在第三周期的值相比于所述处理器核心的负载在第四周期的值的变化百分比，其中，所述第三周期为所述校正值对应的第二索引信息所指示的N个连续周期的下一周期，所述第四周期为所述校正值对应的第二索引信息所指示的N个连续周期中的最后一个周期。
在一种可能的实现方式中，所述控制逻辑还用于：根据所述至少一个处理器核心的待预测变量在下一周期的预测值确定所述至少一个处理器核心的调节策略，其中，所述调节策略用于指示所述处理器核心的电压和/或频率的调节方式。
在一种可能的实现方式中，所述根据所述至少一个处理器核心的待预测变量在下一周期的预测值确定所述至少一个处理器核心的调节策略包括：根据所述至少一个处理器核心的调节策略生成所述至少一个处理器核心的频率调节指令和/或电压调节指令；所述处理器还包括至少一个频率电压调节电路；所述控制逻辑，还用于将所述至少一个处理器核心的频率调节指令和/或电压调节指令发送至所述至少一个频率电压调节电路；所述至少一个频率电压调节电路，用于根据所述至少一个处理器核心的频率调节指令和/或电压调节指令调节所述至少一个处理器核心的频率和/或电压。
第二方面，提供一种变量的预测方法，包括：获取所述至少一个处理器核心的待预测变量在当前周期的测量值；根据所述至少一个处理器核心的待预测变量在当前周期的测量值确定所述至少一个处理器核心的待预测变量在下一周期的预测值。
在一种可能的实现方式中，所述待预测变量为所述处理器核心的功耗；所述根据所述至少一个处理器核心的待预测变量在当前周期的测量值确定所述至少一个处理器核心的待预测变量在下一周期的预测值包括：获取所述至少一个处理器核心的频率、电压在当前周期的测量值；确定所述至少一个处理器核心的频率、电压在所述下一周期的值；根据所述至少一个处理器核心的功耗在所述当前周期的测量值、所述至少一个处理器核心的频率 和电压在当前周期的测量值以及所述至少一个处理器核心的频率和电压在所述下一周期的值确定所述至少一个处理器核心的功耗在所述下一周期的预测值。
在一种可能的实现方式中，根据下述公式确定所述处理器核心的功耗在所述下一周期的预测值：
Figure PCTCN2020098564-appb-000003
其中：P(t+1)为所述处理器核心的功耗在所述下一周期的预测值，P(t)为所述处理器核心的功耗在所述当前周期的测量值，f(t+1)为所述处理器核心的频率在所述下一周期的值，f(t)为所述处理器核心的频率在所述当前周期的测量值，V(t+1)为所述处理器核心的电压在所述下一周期的值，V(t)为所述处理器核心的电压在所述当前周期的测量值。
在一种可能的实现方式中，所述方法还包括：获取所述至少一个处理器核心的温度在当前周期的测量值；获取所述至少一个处理器核心的温度在前一周期的测量值；获取所述至少一个处理器核心的功耗在前一周期的测量值；根据所述至少一个处理器核心的功耗和温度分别在所述前一周期和所述当前周期的测量值、所述至少一个处理器核心的功耗在所述下一周期的预测值，确定所述至少一个处理器核心的温度在所述下一周期的预测值。
在一种可能的实现方式中，根据下述公式确定所述处理器核心的温度在所述下一周期的预测值：
Figure PCTCN2020098564-appb-000004
其中，T(t+1)为所述处理器核心的温度在所述下一周期的预测值，T(t)为所述处理器核心的温度在所述当前周期的测量值，T(t-1)为所述处理器核心的温度在所述前一周期的测量值，P(t+1)为所述处理器核心的功耗在所述下一周期的预测值，P(t)为所述处理器核心的功耗在所述当前周期的测量值，P(t-1)为所述处理器核心的功耗在所述前一周期的测量值。
在一种可能的实现方式中，所述根据所述至少一个处理器核心的待预测变量在当前周期的测量值确定所述至少一个处理器核心的待预测变量在下一周期的预测值包括：根据所述至少一个处理器核心的待预测变量在当前周期的测量值并结合一校正表确定所述至少一个处理器核心的待预测变量在下一周期的预测值。
在一种可能的实现方式中，所述待预测变量为所述处理器核心的功耗，所述校正表包括第一校正表和第二校正表；所述根据所述至少一个处理器核心的待预测变量在当前周期的测量值并结合一校正表确定所述至少一个处理器核心的待预测变量在下一周期的预测值包括：根据所述至少一个处理器核心的功耗在所述当前周期的测量值确定所述至少一个处理器核心的功耗在所述下一周期的第一取值；根据所述第一校正表确定所述至少一个处理器核心的负载在所述下一周期的变化趋势；根据所述第二校正表确定所述至少一个处理器核心的负载在所述下一周期的校正值；根据所述至少一个处理器核心的负载的所述变化趋势和所述至少一个处理器核心的负载的所述校正值对所述至少一个处理器核心的功耗的所述第一取值进行校正，得到所述至少一个处理器核心的功耗在所述下一周期的预测值。
在一种可能的实现方式中，所述方法还包括：获取所述至少一个处理器核心的功耗在前一周期的测量值；获取所述至少一个处理器核心的温度分别在所述当前周期和所述前一周期的测量值；根据所述至少一个处理器核心的功耗在下一周期的预测值，所述至少一个处理器核心的功耗在所述当前周期和所述前一周期的测量值，所述至少一个处理器核心的温度在所述当前周期和所述前一周期的测量值，确定所述至少一个处理器核心的温度在下一周期的预测值。
在一种可能的实现方式中，所述第一校正表包括T N个第一索引信息和T N个第一变化趋势，其中：所述T N个第一索引信息与所述T N个第一变化趋势一一对应；一个所述第一索引信息用于指示所述处理器核心的负载在N个连续周期中的一种变化趋势，T为所述处理器核心的负载在一个周期的变化趋势的种类数量；一个第一变化趋势用于指示所述处理器核心的负载在第一周期的值相比于所述处理器核心的负载在第二周期的值的变化趋势，其中，所述第一周期为所述第一变化趋势对应的第一索引信息所指示的N个连续周期的下一个周期，所述第二周期为所述第一变化趋势对应的第一索引信息所指示的N个连续周期中的最后一个周期；所述第二校正表包括T N个第二索引信息和T N个校正值，其中：所述T N个第二索引信息与所述T N个校正值一一对应；一个所述第二索引信息用于指示所述处理器核心的负载在N个连续周期中的变化趋势与所述处理器核心的负载在N个连续周期中的最后一个周期的值进行异或运算后得到的结果；一个校正值用于指示所述处理器核心的负载在第三周期的值相比于所述处理器核心的负载在第四周期的值的变化百分比，其中，所述第三周期为所述校正值对应的第二索引信息所指示的N个连续周期的下一周期，所述第四周期为所述校正值对应的第二索引信息所指示的N个连续周期中的最后一个周期。
在一种可能的实现方式中，所述方法还包括：根据所述至少一个处理器核心的待预测变量在下一周期的预测值确定所述至少一个处理核心的调节策略，其中，所述调节策略用于指示所述处理器核心的电压和/或频率的调节方式。
在一种可能的实现方式中，所述根据所述至少一个处理器核心的待预测变量在下一周期的预测值确定所述至少一个处理核心的调节策略包括：根据所述至少一个处理器核心的调节策略生成所述至少一个处理器核心的频率调节指令和/或电压调节指令；根据所述至少一个处理器核心的频率调节指令和/或电压调节指令调节所述至少一个处理器核心的频率和/或电压。
第三方面，提供一种计算机可读存储介质，包括计算机程序，所述计算机程序在计算机上被执行时，使得所述计算机执行第二方面中任一项所述的方法。
第四方面，提供一种计算机程序，当所述计算机程序被计算机执行时，用于执行第二方面中任一项所述的方法。
附图说明
图1为本申请实施例提供的处理器的结构示意图；
图2为本申请实施例提供的确定处理器核心0的功耗在下一周期的预测值的流程示意图：
图3为本申请实施例提供的确定处理器核心0的温度在下一周期的预测值的流程示意图；
图4为本申请实施例提供的一种变量的预测方法的流程示意图。
具体实施方式
下面将结合附图，对本申请中的技术方案进行描述。
为使本申请的目的、技术方案和优点更加清楚，下面将结合本申请中的附图，对本申请中的技术方案进行清楚、完整地描述，显然，所描述的实施例是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
本申请的说明书实施例和权利要求书及附图中的术语“第一”、“第二”等仅用于区分描述的目的，而不能理解为指示或暗示相对重要性，也不能理解为指示或暗示顺序。此外，术语“包括”和“具有”以及他们的任何变形，意图在于覆盖不排他的包含，例如，包含了一系列步骤或单元。方法、系统、产品或设备不必限于清楚地列出的那些步骤或单元，而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它步骤或单元。
应当理解，在本申请中，“至少一个(项)”是指一个或者多个，“多个”是指两个或两个以上。“和/或”，用于描述关联对象的关联关系，表示可以存在三种关系，例如，“A和/或B”可以表示：只存在A，只存在B以及同时存在A和B三种情况，其中A，B可以是单数或者复数。字符“/”一般表示前后关联对象是一种“或”的关系。“以下至少一项(个)”或其类似表达，是指这些项中的任意组合，包括单项(个)或复数项(个)的任意组合。例如，a，b或c中的至少一项(个)，可以表示：a，b，c，“a和b”，“a和c”，“b和c”，或“a和b和c”，其中a，b，c可以是单个，也可以是多个。
为了解决由于变量在当前周期的测量值未考虑处理器核心在下一周期的运行状态，而导致的根据变量在当前周期的测量值确定的调节策略无法应对处理器在下一周期的运行状态的问题，本申请实施例提供了一种处理器，图1为本申请实施例提供的处理器的结构示意图。如图1所示，处理器包括4个处理器核心、控制逻辑、4个感应装置、4个频率电压调节电路、外部缓存、内部存储器、通用单元、加速器、输入/输出控制单元&接口单元。其中，4个处理器核心与4个感应装置一一对应，4个处理器核心与4个频率电压调节电路一一对应。4个处理器核心分别为处理器核心0、处理器核心1、处理器核心2、处理器核心3。
每个感应装置均用于获取其对应的处理器核心的待预测变量在当前周期的测量值，以及将其获取的对应的处理器核心的待预测变量在当前周期的测量值传输至控制逻辑。
控制逻辑用于确定每个处理器核心的待预测变量在下一周期的预测值。具体可以包括但不限于以下两种方式：
第一种，控制逻辑根据每个处理器核心的待预测变量在当前周期的测量值分别确定每个处理器核心的待预测变量在下一周期的预测值。
第二种，控制逻辑根据每个处理器核心的待预测变量在当前周期的测量值并结合一校正表确定每个处理器核心的待预测变量在下一周期的预测值。校正表用于指示待预测变量的相关变量在下一周期的变化趋势和校正值。待预测变量的相关变量指影响待预测变量在下一周期的预测值的变量。例如，若待预测变量为处理器核心的功耗或者温度，则由于处 理器核心的负载的变化会影响处理器核心的温度和功耗，因此，处理器核心的功耗和温度的相关变量为处理器核心的负载。
待预测变量例如可以为处理器核心的功耗、处理器核心的温度等，本申请实施例对此不作特殊限定。
下面，以待预测变量为处理器核心的功耗为例，对确定处理器核心0的功耗在下一周期的预测值的过程进行说明。图2为本申请实施例提供的确定处理器核心0的功耗在下一周期的预测值的流程示意图。如图2所示，包括：
步骤201，处理器核心0中的感应装置获取处理器核心0的功耗在当前周期的测量值。具体的，包括以下两种方式：
第一种，感应装置获取处理器核心0的功耗在当前周期内的任一时刻或者指定的一个时刻的值，并将获取的功耗的值确定为处理器核心0的功耗在当前周期的测量值。其中，指定的一个时刻例如可以为当前周期中的中间时刻或者起始时刻等，本申请对此不作特殊限定。
第二种，感应装置获取处理器核心0的功耗在当前周期中的多个时刻的值，以及计算功耗在多个时刻的值的平均值，以及将该平均值确定为处理器核心0的功耗在当前周期的测量值，其中，当前周期中的多个时刻可以根据经验设置，本申请对此不作特殊限定。
获取功耗的感应装置例如可以为功耗传感器(power sensor)等，本申请实施例对此不作特殊限定。
需要说明的是，上述获取处理器0的功耗在当前周期的测量值的方式仅为示例性的，并不用于限定本申请。
步骤202、处理器核心0中的感应装置将处理器核心0的功耗在当前周期的测量值发送至控制逻辑。
步骤203、控制逻辑确定处理器核心0的功耗在下一周期的预测值。具体的，可以采用以下两种方式确定。
第一种：根据处理器核心0的功耗在当前周期的测量值确定处理器核心0的功耗在下一周期的预测值。具体过程包括：
首先，获取处理器核心0的频率、电压在当前周期的测量值。
处理器核心0的频率在当前周期的测量值可以是处理器核心0的频率在当前周期中的任意一个时刻或者指定的一个时刻的值，处理器核心0的频率在当前周期的测量值还可以是处理器核心0的频率在当前周期的多个时刻的值的平均值等，本申请对此不作特殊限定。
同理，处理器核心0的电压在当前周期的测量值可以是处理器核心0的电压在当前周期中的任一个时刻或者指定的一个时刻的值，处理器核心0的电压在当前周期的测量值还可以是处理器核心0的电压在当前周期中的多个时刻的值的平均值等，本申请对此不作特殊限定。
然后，确定处理器核心0的频率、电压在下一周期的值。例如，可以通过控制算法或者操作系统分别确定处理器核心0的频率和电压在下一周期的值。
最后，根据处理器核心0的功耗在当前周期的测量值，处理器核心0频率和电压在当前周期的测量值以及处理器核心0的频率、电压在下一周期的值确定处理器核心0的功耗在下一周期的预测值。例如，可以通过下述公式确定处理器核心0的功耗在下一周期的预 测值：
Figure PCTCN2020098564-appb-000005
其中：P(t+1)为处理器核心0的功耗在下一周期的预测值，P(t)为处理器核心0的功耗在当前周期的测量值，f(t+1)为处理器核心0的频率在下一周期的值，f(t)为处理器核心0的频率在当前周期的测量值，V(t+1)为处理器核心0的电压在下一周期的值，V(t)为处理器核心0的电压在当前周期的测量值。
由上可知，确定处理器核心0的频率、电压和功耗在当前周期的测量值，确定处理器核心0的频率和电压在下一周期的值，以及根据上述确定的各项测量值和确定的各项值即可确定处理器核心0的功耗在下一周期的预测值，提供了一种确定处理器核心0的功耗在下一周期的预测值的方式，且确定方式简单，易于执行，提高了确定处理器核心0的功耗在下一周期的预测值的效率。此外，由于在确定处理器核心0的功耗在下一周期的预测值时，考虑了处理器核心0的电压、功率在下一周期的值，即考虑了处理器核心0在下一周期的工作状态，因此，提高了确定处理器核心0的功耗在下一周的预测值的准确性。
第二种：根据处理器核心0的功耗在当前周期的测量值并结合一校正表确定处理器核心0的功耗在下一周期的预测值，其中，校正表包括第一校正表和第二校正表，第一校正表用于预测处理器核心的负载在下一周期的变化趋势，第二校正表用于预测处理器核心的负载在下一周期的校正值。具体过程包括：
首先，根据处理器核心0的功耗在当前周期的测量值确定处理器核心0的功耗在下一周期的第一取值。由于该步骤的原理与第一种方式中确定处理器核心0的功耗在下一周期的预测值的原理相同，即此处的第一取值为第一种方式中的处理器核心0的功耗在下一周期的预测值，因此此处不再赘述。
然后，根据第一校正表确定处理器核心0的负载在下一周期的变化趋势。
第一校正表包括T N个第一索引信息和T N个第一变化趋势，其中，T N个第一索引信息与T N个第一变化趋势一一对应。
一个第一索引信息用于指示处理器核心的负载在N个连续周期中的一种变化趋势，处理器核心的负载在N个连续周期中的一种变化趋势指处理器核心的负载的N个变化趋势的一种组合，N个变化趋势与N个连续周期一一对应，一个变化趋势用于指示处理器核心的负载在对应周期的变化趋势，处理器核心的负载在对应周期的变化趋势为处理器核心的负载在对应周期的值相比于处理器核心的负载在对应周期的前一周期的值的变化趋势。T为处理器核心的负载在一个周期的变化趋势的种类数量。例如，若处理器核心的负载在一个周期的变化趋势包括变大、变小、不变三种情况，即包括三种种类，则T的取值为3。再例如，若处理器核心的负载在一个周期的变化趋势包括变大、变小两种情况，即包括两种种类，则T的取值为2。需要说明的是，N个变化趋势是按照对应的N个连续周期的先后顺序组合的。N的取值可以根据设计需求进行设置，本申请对此不作特殊限定。
由于处理器核心的负载在一个周期的变化趋势包括T种，即一个变化趋势有T种类型，因此，N个变化趋势的组合方式有T N种，那么，第一索引信息的数量为T N个。
一个第一变化趋势用于指示处理器核心的负载在第一周期的值相比于处理器核心的 负载在第二周期的值的变化趋势，其中，第一周期为第一变化趋势对应的第一索引信息所指示的N个连续周期的下一个周期，第二周期为第一变化趋势对应的第一索引信息所指示的N个连续周期中的最后一个周期。换言之，一个第一变化趋势用于指示处理器核心的负载在对应的第一索引信息所指示的N个连续周期的下一周期的变化趋势。例如，若N的取值为8，且若8个连续周期为第5周期至第12周期，则第一周期为第13周期，第二周期为第12周期。第一变化趋势例如包括变大、变小、不变三种情况，第一变化趋势还例如包括变大或等于、变小两种情况，本申请对此不作特殊限定。
需要说明的是，关于第一校正表的构建原理将在下文中进行说明，因此此处不再赘述。
基于此，确定处理器核心0的负载在下一周期的变化趋势的原理为：获取处理器核心0的负载在N个连续周期中的变化趋势，其中，N个连续周期中的最后一个周期为当前周期；将处理器核心0的负载在N个连续周期中的变化趋势作为第一待比对索引信息，将第一待比对索引信息与第一校正表中的第一索引信息进行匹配，以及将与第一待比对索引信息匹配的第一索引信息对应的第一变化趋势确定为处理器核心0的负载在下一周期的变化趋势。
获取处理器核心0的负载在N个连续周期中的变化趋势的过程包括：获取处理器核心0的功耗、频率、电压在N个连续周期中的每个周期的值，计算处理器核心0的频率在一个周期的值与处理器核心0的电压在该周期的值的平方的乘积，通过处理器核心0的功耗在该周期的值与该乘积的比值表征处理器核心0的负载在该周期的值，换言之，可以认定该比值为处理器核心0的负载在该周期的值。通过上述原理，可以得到处理器核心0的负载在每个周期的值。将处理器核心0的负载在每个周期的值与前一周期的值进行比较，得到处理器核心0的负载在每个周期的变化趋势；将处理器核心0的负载在每个周期的变化趋势按照对应周期由小到大的顺序进行组合，即可得到第一待比对索引信息。
再然后，根据第二校正表确定处理器核心0的负载在下一周期的校正值。第二校正表包括T N个第二索引信息和T N个校正值，其中：T N个第二索引信息与T N个校正值一一对应。
一个第二索引信息用于指示处理器核心的负载在N个连续周期中的变化趋势与处理器核心的负载在N个连续周期中的最后一个周期的值进行异或运算后得到的结果。由于处理器核心的负载在N个连续周期中的变化趋势已经在上文中进行了说明，因此此处不再赘述。需要说明的是，处理器核心的负载在N个连续周期中的最后一个周期的值的表示方式与处理器核心的负载在N个连续周期中的变化趋势的表示方式需要相同，这样才可进行异或运算。例如，若处理器核心的负载在N个连续周期中的变化趋势用一组数列表示，该组数列包括8位数字且每位数字的取值为0或1，那么，处理器核心的负载在N个连续周期中的最后一个周期的值也要转换为由8位数字组成且每位数字的取值为1或0的一组数列，以便于做异或运算。需要说明的是，转换的方式例如可以采用哈希算法等，本申请对此不作特殊限定。
由于处理器核心的负载在N个连续周期中的变化趋势有T N种，因此，异或运算之后得到的第二索引信息的数量也为T N。N和T已经在上文中进行了说明，因此此处不再赘述。
一个校正值用于指示处理器核心的负载在第三周期的值相比于处理器核心的负载在第四周期的值的变化百分比，其中，第三周期为校正值对应的第二索引信息所指示的N个 连续周期的下一周期，第四周期为校正值对应的第二索引信息所指示的N个连续周期中的最后一个周期。处理器核心的负载在第三周期的值相比于处理器核心的负载在第四周期的值的变化百分比为处理器核心的负载在第三周期的值与处理器核心的负载在第四周期的值的差值与处理器核心的负载在第四周期的值的比值。
需要说明的是，关于第二校正表的构建原理将在下文中进行说明，因此此处不再赘述。
基于此，确定处理器核心0的负载在下一周期的校正值的原理为：获取处理器核心0的负载在N个连续周期中的变化趋势，其中，N个连续周期中的最后一个周期为当前周期；获取处理器核心0的负载在当前周期的值；对处理器核心0的负载在N个连续周期中的变化趋势和处理器核心0的负载在当前周期的值进行异或运算，得到第二待比对索引信息，将第二待比对索引信息与第二校正表中的第二索引信息进行匹配，将与第二待比对索引信息匹配的第二索引信息对应的校正值确定为处理器核心0的负载在下一周期的校正值。
由于获取处理器核心0的负载在N个连续周期中的变化趋势的原理以及确定负载的原理已经在上文中进行了说明，因此此处不再赘述。
最后，根据处理器核心0的负载在下一周期的变化趋势、处理器核心0的负载在下一周期的校正值对处理器核心0的功耗在下一周期的第一取值进行校正，得到处理器核心0的功耗在下一周期的预测值。具体的，通过下述公式确定处理器核心0的功耗在下一周期的预测值：
P(t+1)＝P(t+1) /*(1±ΔA(t+1))
其中，P(t+1)为处理器核心0的功耗在下一周期的预测值，P(t+1) /为处理器核心0的功耗在下一周期的第一取值，ΔA(t+1)为处理器核心0的负载在下一周期的校正值，需要说明的是，若处理器核心0的负载在下一周期的变化趋势为变小或不变，则ΔA(t+1)之前取减号，若处理器核心0的负载在下一周期的变化趋势为变大，则ΔA(t+1)之前取加号。
由上可知，确定处理器核心0的功耗在下一周期的第一取值，并根据第一校正表和第二校正表确定处理器核心0的负载在下一周期的变化趋势和校正值，以及根据处理器核心0的负载在下一周期的变化趋势和校正值校正处理器核心0的功耗在下一周期的第一取值，以得到处理器核心0的功耗在下一周期的预测值，提供了一种确定处理器核心0的功耗在下一周期的预测值的方式，且确定方式简单，易于执行，提高了确定处理器核心0的功耗在下一周期的预测值的效率；此外，由于在确定处理器核心0的功耗在下一周期的第一取值时，考虑了处理器核心0的电压、功率在下一周期的值，即考虑了处理器核心0在下一周期的工作状态，因此，提高了确定第一取值的准确率，进而提高了确定处理器核心0的功耗在下一周的预测值的准确性；另外，由于根据处理器核心0的负载在下一周期的变化趋势和校正值校正处理器核心0的功耗在下一周期的第一取值，得到处理器核心0的功耗在下一周期的预测值，即在确定处理器核心0的功耗在下一周期的预测值的过程中，考虑了处理器核心0的负载在下一周期的变化对处理器核心0的功耗在下一周期的预测值的影响，进一步提高了确定处理器核心0的功耗在下一周期的预测值的准确性。
需要说明的是，上述确定处理器核心0的功耗在下一周期的预测值的方式仅为示例性的，并不用于限定本申请。
需要说明的是，计算其他各处理器核心的功耗在下一周期的预测值的原理同上，此处 不再赘述。
下面，举例对第一校正表进行说明。
若N的取值为8，处理器核心的负载在一个周期的变化趋势包括变大、变小或不变两种种类，即T的取值为2，处理器核心的负载在一个周期的变化趋势用0或1表示，若处理器核心的负载在一个周期的变化趋势为0，则处理器核心的负载在该周期的值小于或者等于处理器核心的负载在该周期的前一周期的值(即处理器核心的负载在该周期的变化趋势为变小或不变)，若处理器核心的负载在一个周期的变化趋势为1，则处理器核心的负载在该周期的值大于处理器核心的负载在该周期的前一周期的值(即处理器核心的负载在该周期的变化趋势为变大)。
则，第一索引信息的数量为2 8个，其中，第一索引信息由8位数字组成，每位数字的取值为1或0，第一索引信息用于指示处理器核心的负载在8个连续周期中的一种变化趋势，且第一索引信息中的8位数字与8个连续周期一一对应，每个数字表示处理器核心的负载在其对应周期的变化趋势。
第一变化趋势同样也包括变大、变小或不变两种情况。第一变化趋势用0或1表示，其中若第一变化趋势为0，则处理器核心的负载在第一周期的值小于或者等于处理器核心的负载在第二周期的值(即第一变化趋势为变小或不变)，若第一变化趋势为1，则处理器核心的负载在第一周期的值大于处理器核心的负载在第二周期的值(即第一变化趋势为变大)。由于第一索引信息的数量为2 8个，因此，第一变化趋势的数量也为2 8个。第一变化趋势由1位数字组成，该位数字的取值为1或0。
表1为本申请实施例提供的第一校正表。
第一索引信息 第一变化趋势
00000000 0
…… ……
00110010 1
…… ……
01100101 0
…… ……
10011001 1
…… ……
11111110 1
11111111 0
表1
下面，对生成表1所示的第一校正表的过程进行说明。
生初始状态的第一校正表。具体的，由于第一索引信息由8位数字组成，每位数字的 取值为1或0，因此，2 8个第一索引信息分别为：00000000、00000001、00000010、00000011、……、11111110、11111111。2 8个第一索引信息构成了第一校正表的第一列。设置每个第一索引信息对应的第一变化趋势的初始值，以得到第一校正表的第二列。第一索引信息对应的第一变化趋势的初始值可以设置为0或者1等，本申请对此不作特殊限定。需要说明的是，在本申请的其他实施例中，还可以不设置第一索引信息对应的第一变化趋势的初始值。
对初始状态的第一校正表进行训练，以得到上述表1中示出的第一校正表。具体过程包括：获取大量的训练数据，并通过大量的训练数据对初始状态的第一校正表进行训练，得到上述表1中示出的第一校正表。
下面，对获取一个训练数据，并通过该训练数据对初始状态的第一校正表进行训练的过程进行说明。
获取处理器核心的功耗、频率、电压在10个连续周期中的每个周期的值(即一个训练数据)，计算处理器核心的频率在一个周期的值与处理器核心的电压在该周期的值的平方的乘积，通过处理器核心的功耗在该周期的值与该乘积的比值表征处理器核心的负载在该周期的值，换言之，可以认定该比值为处理器核心的负载在该周期的值。通过上述原理，可以得到处理器核心的负载在每个周期的值。
将处理器核心的负载在第i个周期的值与处理器核心的负载在第i-1个周期的值进行比较，得到处理器核心的负载的一个变化趋势，该变化趋势与第i个周期对应，若处理器核心的负载在第i个周期的值大于处理器核心的负载在第i-1个周期的值，则处理器核心的负载在第i个周期的变化趋势用1表示，若处理器核心的负载在第i个周期的值小于或者等于处理器核心的负载在第i-1个周期的值，则处理器核心的负载在第i个周期的变化趋势用0表示，2≤i≤10。基于此，可得到处理器核心的负载的9个变化趋势，处理器核心的负载的9个变化趋势与10个连续周期中的第2个周期至第10个周期中的9个周期一一对应。
按照对应周期的编号由小到大的顺序对处理器核心的负载的9个变化趋势进行组合，得到一个数列。取该数列的前8位数字，并将该前8位数字与初始状态的第一校正表中的第一索引信息进行匹配，并将与该前8位数字匹配的第一索引信息对应的第一变化趋势更新为该数列的第9位数字。这样，完成了该训练数据对初始状态的第一校正表的训练。
需要说明的是，重复上述过程，并通过其他训练数据进行训练，得到表1中示出的第一校正表。
下面，对基于表1中所示的第一校正表确定处理器核心的负载在下一周期的变化趋势的原理进行说明。
若当前周期为处理器核心运行过程中的第s周期，则获取处理器核心的功耗、电压、频率在第s周期至第s-8周期中的每个周期的值。然后，根据处理器核心的功耗、电压、频率在每个周期的值，确定处理器核心的负载在每个周期的值。需要说明的是，由于确定处理器核心的负载在一个周期的值的原理已经在上文中进行了说明，因此此处不再进行赘 述。
将处理器核心的负载在第j周期的值与处理器核心的负载在第j-1周期的值进行比较，得到处理器核心的负载的一个变化趋势，该变化趋势与第j周期对应，若处理器核心的负载在第j周期的值大于处理器核心的负载在第j-1周期的值，则处理器核心的负载在第j周期的变化趋势用1表示，若处理器核心的负载在第j周期的值小于或者等于处理器核心的负载在第j-1周期的值，则处理器核心的负载在第j周期的变化趋势用0表示，s-7≤j≤s。基于此，得到处理器核心的负载的8个变化趋势，处理器核心的负载的8个变化趋势与第s周期至第s-7周期中的8个周期一一对应。
按照对应周期的编号由小到大的顺序对处理器核心的负载的8个变化趋势进行组合，得到第一待比对索引信息。将第一待比对索引信息与表1中的第一索引信息进行匹配，以及将与第一待比对索引信息匹配的第一索引信息对应的第一变化趋势确定为处理器核心的负载在下一周期(即第s+1周期)的变化趋势。
需要说明的是，若处理器核心的负载在下一周期的变化趋势为0，则处理器核心的负载在下一周期的值小于或者等于处理器核心的负载在当前周期的值，若处理器核心的负载在下一周期的变化趋势为1，则处理器核心的负载在下一周期的值大于处理器核心的负载在当前周期的值。
例如，第一待比对索引信息为01100101，从表1中可知，与该第一待比对索引信息匹配的第一索引信息对应的变化趋势为0，即处理器核心的负载在下一周期的变化趋势为0，即处理器核心的负载在下一周期的值小于或等于处理器核心的负载在当前周期的值。
在本申请的另一实施例中，在处理器核心运行过程中的每个周期均计算处理器核心的负载在每个周期的值，以及根据处理器核心的负载在每个周期的值确定处理器核心的负载在每个周期的变化趋势，以及根据处理器核心的负载在每个周期的变化趋势确定处理器核心的负载在最近的8个连续周期中的变化趋势，并始终保存处理器核心在最近的8个连续周期中的变化趋势，这样，在基于表1确定处理器核心的负载在下一周期的变化趋势时，由于始终保存了处理器核心的负载在最近的8个连续周期中的变化趋势，因此可以直接将处理器核心的负载在最近的8个连续周期中的变化趋势作为第一待比对索引信息，以根据第一待比对索引信息确定处理器核心的负载在下一周期的变化趋势，提高了确定处理器核心的负载在下一周期的变化趋势的效率。
例如，若当前周期为处理器核心的运行过程中的第14周期，那么在当前周期中，会保存处理器核心的负载在第7周期至第14周期中的变化趋势，因此，可以将处理器核心的负载在第7周期至第14周期中的变化趋势作为第一待比对索引信息，以确定处理器核心的负载在下一周期(即第15周期)的变化趋势。在下一周期(第15周期)到来后，测量处理器核心的功耗、频率、电压在下一周期的值，根据处理器核心的功耗、频率、电压在下一周期的值确定处理器核心的负载在下一周期的值，将处理器核心的负载在下一周期的值与处理器核心的负载在当前周期的值进行比较，以确定处理器核心的负载在下一周期的变化趋势，以及删除第一待比对索引信息的第一位，并将处理器核心的负载在下一周期的变化趋势补充在删除了第一位数字的第一待比对索引信息之后，以得到一个新的第一待比对索引信息，以及根据新的第一待比对索引信息确定处理器核心的负载在下下一个周期的变化趋势。
为了进一步提高第一校正表的准确性，在下一周期到来后，获取处理器核心的功耗、频率、电压在下一周期的值，根据处理器核心的功耗、频率、电压在下一周期的值确定处理器核心的负载在下一周期的值，根据处理器核心的负载在下一周期的值和处理器核心的负载在当前周期的值的大小关系，确定处理器核心的负载在下一周期的变化趋势，以及判断处理器核心的负载在下一周期的变化趋势是否与表1中与第一待比对索引信息匹配的第一索引信息对应的第一变化趋势相同，若相同，则不对与第一待比对索引信息匹配的第一索引信息对应的第一变化趋势进行修正，若不相同，则通过处理器核心的负载在下一周期的变化趋势对与第一待比对索引信息匹配的第一索引信息对应的第一变化趋势进行修正。综上，在下一周期到来后，通过实测值对第一校正表中对应的第一变化趋势进行修正，进一步提高了第一校正表的准确性。
例如，在当前周期，第一待比对索引信息为11111110，在表1中，与第一待比对索引信息匹配的第一索引信息对应的第一变化趋势为1，若在下一周期到来时，处理器核心的负载在下一周期的值小于处理器核心的负载在当前周期的值，即处理器核心的负载在下一周期的实际变化趋势为0，则在表1中，将与第一待比对索引信息匹配的第一索引信息对应的第一变化趋势从1修正为0。
修正后的表1如下所示，修正后的表1中的加粗项为与第一待比对索引信息匹配的第一索引信息和对应的第一变化趋势。
第一索引信息 变化趋势
00000000 0
…… ……
00110010 1
…… ……
01100101 0
…… ……
10011001 1
…… ……
11111110 0
11111111 0
下面，举例对第二校正表进行说明。
若N的取值为8，处理器核心的负载在对应周期的变化趋势包括变大、变小或不变两种种类，即T的取值为2，处理器核心的负载在一个周期的变化趋势用0或1表示，若处理器核心的负载在一个周期的变化趋势为0，则处理器核心的负载在该周期的值小于或者等于处理器核心的负载在该周期的前一周期的值(即处理器核心的负载在该周期的变化趋势为变小或不变)，若处理器核心的负载在一个周期的变化趋势为1，则处理器核心的负载在该周期的值大于处理器核心的负载在该周期的前一周期的值(即处理器核心的负载在该周期的变化趋势为变大)。
则，处理器核心的负载在8个连续周期中的变化趋势的数量为2 8个，其中，处理器核心的负载在8个连续周期中的每个变化趋势均由8位数字组成，每位数字的取值为1或0，8位数字与8个连续周期一一对应，每个数字表示处理器核心的负载在其对应的周期的变化趋势。处理器核心的负载在8个连续周期中的最后一个周期的值也通过哈希算法转换为8位数字，且每位数字的取值为1或0。在此基础上，对处理器核心的负载在8个连续周期中的变化趋势和处理器核心的负载在8个连续周期中的最后一个周期的值执行异或运算之后得到的第二索引信息的数量也为2 8，其中，第二索引信息由8位数字组成，每位数字的取值为1或0。
表2为本申请实施例提供的第二校正表。
第二索引信息 校正值
00000000 0
…… ……
00110100 8％
…… ……
01100101 1％
…… ……
11011010 5％
…… ……
11111110 15％
11111111 2％
下面，对生成表2的过程进行说明。
生成初始状态的第二校正表。具体的，由于第二索引信息由8位数字组成，每位数字的取值为1或0，因此，2 8个第二索引信息分别为：00000000、00000001、00000010、00000011、……、11111110、11111111。2 8个第二索引信息构成了第二校正表的第一列。设置每个第二索引信息对应的校正值的初始值，以得到第二校正表的第二列。第二索引信息对应的校正值的初始值可以设置为任意一个值，本申请对此不作特殊限定。需要说明的是，在本申请的其他实施例中，还可以不设置第二索引信息对应的校正值的初始值。
对初始状态的第二校正表进行训练，得到上述表2中示出的第二校正表。具体过程包括：获取大量的训练数据，并通过大量的训练数据对初始状态的第二校正表进行训练，得到上述表2中示出的第二校正表。
下面，对获取一个训练数据，并通过该训练数据对初始状态的第二校正表进行训练的过程进行说明。
获取处理器核心的功耗、频率、电压在10个连续周期中的每个周期的值(即一个训练数据)，计算处理器核心的频率在一个周期的值与处理器核心的电压在该周期的值的平方的乘积，通过处理器核心的功耗在该周期的值与该乘积的比值表征处理器核心的负载在该周期的值，即可以认定该比值为处理器核心的负载在该周期的值。通过上述原理，可以得到处理器核心的负载在每个周期的值。
将处理器核心的负载在第i个周期的值与处理器核心的负载在第i-1个周期的值进行比较，得到处理器核心的负载的一个变化趋势，该变化趋势与第i个周期对应，若处理器核心的负载在第i个周期的值大于处理器核心的负载在第i-1个周期的值，则处理器核心的负载在第i个周期的变化趋势用1表示，若处理器核心的负载在第i个周期的值小于或者等于处理器核心的负载在第i-1个周期的值，则处理器核心的负载在第i个周期的变化趋势用0表示，2≤i≤9。基于此，得到处理器核心的负载的8个变化趋势，处理器核心的负载的8个变化趋势与连续的10个周期中的第2个周期至第9个周期中的8个周期一一对应。
按照对应周期的编号由小到大的顺序对处理器核心的负载的8个变化趋势进行组合，得到一个第一数列，即得到处理器核心的负载在8个连续周期中的一种变化趋势，该第一数列包括8位数字，每位数字的取值为0或1。将处理器核心的负载在连续的10个周期中的第9个周期的值通过哈希算数转换为第二数列，第二数列包括8位数字，每位数字的取值为1或0。对第一数列和第二数列进行异或运算即可得到第三数列。将处理器核心的负载在连续的10个周期中的第10个周期的值与处理器核心的负载在第9个周期的值的差值与处理器核心的负载在第9个周期的值的比值确定为处理器核心的负载在第10个周期的值相比于处理器核心的负载在第9个周期的值的百分比。
将第三数列与初始状态的第二校正表中的每个第二索引信息进行匹配，并将与第三数列匹配的第二索引信息对应的校正值更新为处理器核心的负载在第10个周期的值相比于处理器核心的负载在第9个周期的值的百分比。
需要说明的是，重复上述过程，并通过其他各训练数据进行训练，得到表2中示出的第二校正表。
下面，对基于表2中所示的第二校正表确定处理器核心的负载在下一周期的校正值的原理进行说明。
若当前周期为处理器核心运行过程中的第s周期，则获取处理器核心的功耗、电压、频率在第s周期至第s-8周期中的每个周期的值。然后，根据处理器核心的功耗、电压、频率在每个周期的值，确定处理器核心的负载在每个周期的值。需要说明的是，由于确定处理器核心的负载在一个周期的值的原理已经在上文中进行了说明，因此此处不再进行赘述。
将处理器核心的负载在第j周期的值与处理器核心的负载在第j-1周期的值进行比较，得到处理器核心的负载的一个变化趋势，该变化趋势与第j周期对应，若处理器核心的负载在第j周期的值大于处理器核心的负载在第j-1周期的值，则处理器核心的负载的变化趋势用1表示，若处理器核心的负载在第j周期的值小于或者等于处理器核心的负载在第j-1周期的值，则处理器核心的负载的变化趋势用0表示，s-7≤j≤s。基于此，得到处理器核心的负载的8个变化趋势，处理器核心的负载的8个变化趋势与第s周期至第s-7周 期中的8个周期一一对应。
按照对应周期的编号由小到大的顺序对处理器核心的负载的8个变化趋势进行组合，得到第四数列，通过哈希算法将处理器核心的负载在第s周期的值转换为第五数列。对第四数列和第五数列进行异或运算，得到第二待比对索引信息。将第二待比对索引信息与表2中的第二索引信息进行匹配，以及将与第二待比对索引信息匹配的第二索引信息对应的校正值确定为处理器核心的负载在下一周期(即第s+1周期)的校正值。
例如，处理器核心的负载在8个连续周期(即第s-7周期至第s周期)中的变化趋势为00110010，处理器核心的负载在第s周期的值的哈希值为00000110，对上述两个值进行异或运算之后得到的第二待比对索引信息为00110100。从表2中可知，与第二待比对索引信息匹配的第二索引信息对应的校正值为8％，即处理器核心的负载在下一周期的校正值为8％，即处理器核心的负载在下一周期的值相比于处理器核心的负载在当前周期的值的变化百分比为8％。
在本申请的另一实施例中，在处理器核心运行过程中的每个周期均计算处理器核心的负载在每个周期的值，以及根据处理器核心的负载在每个周期的值确定处理器核心的负载在每个周期的变化趋势，以及根据处理器核心的负载在每个周期的变化趋势确定处理器核心的负载在最近的8个连续周期中的变化趋势，并始终保存处理器核心在最近的8个连续周期中的变化趋势，这样，在基于表2确定处理器核心的负载在下一周期的校正值时，由于始终保存了处理器核心的负载在最近的8个连续周期中的变化趋势，因此可以直接将处理器核心的负载在最近的8个连续周期中的变化趋势作为第四数列，以及对处理器核心的负载在当前周期的值进行哈希运算，得到第五数列，对第四数列和第五数列进行异或运算，得到第二待比对索引信息，以根据第二待比对索引信息确定处理器核心的负载在下一周期的校正值，由于预先保存了处理器核心的负载在最近的8个连续周期的变化趋势，无需重新计算，因此提高了确定处理器核心的负载在下一周期的校正值的效率。
例如，若当前周期为处理器核心的运行过程中的第16周期，那么在当前周期中，会保存处理器核心的负载在第9周期至第16周期中的变化趋势，因此，可以计算处理器核心的负载在第16周期的值的哈希值，并对哈希值和处理器核心的负载在第9周期至第16周期中的变化趋势进行异或运算，得到第二待比对索引信息，以根据第二待比对索引信息确定处理器核心的负载在下一周期(即第17周期)的校正值。在下一周期到来后，测量处理器核心的功耗、频率、电压在下一周期的值，根据处理器核心的功耗、频率、电压在下一周期的值确定处理器核心的负载在下一周期的值，将处理器核心的负载在下一周期的值与处理器核心的负载在当前周期的值进行比较，以确定处理器核心的负载在下一周期(即第17周期)的变化趋势，以及删除处理器核心的负载在第9周期至第16周期中的变化趋势的第一位数字，并将处理器核心的负载在下一周期的变化趋势补充在删除了第一位数字的处理器核心的负载在第9周期至第16周期中的变化趋势之后，以得到处理器核心的负载在第10周期至第17周期中的变化趋势，以及根据处理器核心的负载在第10周期至第17周期中的变化趋势确定第二待比对索引信息，以及根据第二待比对索引信息确定处理器核心的负载在下下一个周期的校正值。
为了进一步提高第二校正表的准确性，在下一周期到来后，获取处理器核心的功耗、频率、电压在下一周期的值，根据处理器核心的功耗、频率、电压在下一周期的值确定处 理器核心的负载在下一周期的值，根据处理器核心的负载在下一周期的值和处理器核心的负载在当前周期的值计算处理器核心的负载在下一周期的值相比于处理器核心的负载在当前周期的值的百分比，以及判断计算得到的百分比与第二待比对索引信息匹配的第二索引信息对应的校正值是否相同，若相同，则不对与第二待比对索引信息匹配的第二索引信息对应的校正值进行修正，若不相同，则通过计算得到的百分比修正与第二待比对索引信息匹配的第二索引信息对应的校正值。
例如，在当前周期，第二待比对索引信息为11011010，在表2中，与第二待比对索引信息匹配的第二索引信息对应的校正值为5％，若在下一周期到来时，处理器核心的负载在下一周期的值相比于处理器核心的负载在当前周期的值的百分比为8％，则在表2中，将与第二待比对索引信息匹配的第二索引信息对应的校正值从5％修正为8％。
修修正的表2如下所示，修正后的表2中的加粗项为与第二待比对索引信息匹配的第二索引信息对应的校正值。
第二索引信息 校正值
00000000 0
…… ……
00110100 8％
…… ……
01100101 1％
…… ……
11011010 8％
…… ……
11111110 15％
11111111 2％
需要说明的是，第一校正表和第二校表的生成方式和第一校正表和第二校正表的表示方式仅为示例性的，并不用于限定本申请，例如第一校正表和第二校正表还可以是一由神经网络训练得到的模型等。
下面，以待预测变量为处理器核心的温度为例，对确定处理器核心0的温度在下一周期的预测值的过程进行说明。图3为本申请实施例提供的确定处理器核心0的温度在下一周期的预测值的流程示意图。如图3所示，包括：
步骤301，处理器核心0中的感应装置获取处理器核心0的温度在当前周期的测量值。由于获取处理器核心0的温度在当前周期的测量值的原理与获取处理器核心0的功耗在当前周期的测量值的原理相同，因此此处不再赘述。需要说明的是，感应装置例如可以为温度传感器(T-sensor)等，本申请实施例对此不作特殊限定。
步骤302、处理器核心0中的感应装置将处理器核心0的温度在当前周期的测量值发送至控制逻辑。
步骤303、控制逻辑确定处理器核心0的温度在下一周期的预测值。具体的，可以采用以下两种方式确定。
第一种：根据处理器核心0的温度在当前周期的测量值确定处理器核心0的温度在下一周期的预测值。具体过程包括：
首先，获取处理器核心0的功耗在当前周期和前一周期的测量值、处理器核心0的功耗在下一周期的预测值。由于获取处理器核心0的功耗在当前周期的测量值的原理已经在上文中进行了说明，因此此处不再赘述。由于获取处理器核心0的功耗在前一周期的测量值的原理与上文中获取处理器核心0的功耗在当前周期的测量值的原理相同，因此此处不再赘述。由于获取处理器核心0的功耗在下一周期的预测值的原理已经在上文中进行了说明，因此此处不再赘述。
然后，获取处理器核心0的温度在前一周期的测量值。由于该步骤的原理与获取处理器核心0的温度在当前周期的测量值的原理相同，因此此处不再赘述。
最后，根据处理器核心0的功耗和温度分别在前一周期和当前周期的测量值、处理器核心0的功耗在下一周期的预测值，确定处理器核心0的温度在下一周期的预测值。由于在短时间内功耗的增量和温度的增量成正相关关系，因此可以通过下述公式确定处理器核心0的温度在下一周期的预测值：
Figure PCTCN2020098564-appb-000006
其中，T(t+1)为处理器核心0的温度在下一周期的预测值，T(t)为处理器核心0的温度在当前周期的测量值，T(t-1)为处理器核心0的温度在前一周期的测量值，P(t+1)为处理器核心0的功耗在下一周期的预测值，P(t)为处理器核心0的功耗在当前周期的测量值，P(t-1)为处理器核心0的功耗在前一周期的测量值。
需要说明的是，上述确定处理器核心0的温度在下一周期的预测值的过程仅为示例性的，并不用于限定本申请。
由上可知，获取处理器核心0的功耗在当前周期和前一周期的测量值、处理器核心0的功耗在下一周期的预测值、处理器核心0的温度在当前周期和前一周期的测量值，以及根据上述获取的各项测量值和预测值即可确定处理器核心0的温度在下一周期的预测值，提供了一种确定处理器核心0的温度在下一周期的预测值的方式，且确定方式简单，易于执行，提高了确定处理器核心0的温度在下一周期的预测值的效率。此外，由于在确定处理器核心0的温度在下一周期的预测值时，考虑了处理器核心0的功耗在下一周期的预测值，即考虑了处理器核心0在下一周期的工作状态，因此，提高了确定处理器核心0的温度在下一周的预测值的准确性。
第二种：根据处理器核心0的温度在当前周期的测量值并结合一校正表确定处理器核心0的温度在下一周期的预测值，其中，校正表包括第一校正表和第二校正表，第一校正表用于预测处理器核心的负载在下一周期的变化趋势，第二校正表用于预测处理器核心的负载在下一周期的校正值。具体过程包括：
首先，获取处理器核心0的功耗在当前周期和前一周期的测量值。由于该步骤的原理已经在上文中进行了说明，因此此处不再进行赘述。
然后，获取处理器核心0的温度在前一周期的测量值。由于获取处理器核心0的温度在前一周期的测量值的原理与获取处理器核心0的温度在当前周期的测量值的原理相同，因此此处不再赘述。
再然后，根据处理器核心0的功耗在当前周期的测量值确定处理器核心0的功耗在下一周期的第一取值。由于该步骤的原理已经在上文中进行了说明，因此此处不再赘述。
接下来，根据第一校正表确定处理器核心0的负载在下一周期的变化趋势，根据第二校正表确定处理器核心0的负载在下一周期的校正值。
由于第一校正表和第二校正表、确定处理器核心0的负载在下一周期的变化趋势和校正值的原理已经在上文中进行了说明，因此此处不再赘述。
再接下来，根据处理器核心0的负载在下一周期的变化趋势和校正值对处理器核心0的功耗在下一周期的第一取值进行校正，得到处理器核心0的功耗在下一周期的预测值。该步骤的原理已经在上文中进行了说明，因此此处不再赘述。
最后，根据处理器核心0的功耗在下一周期的预测值、处理器核心0的功耗在当前周期和前一周期的测量值、处理器核心0的温度在当前周期和前一周期的测量值，确定处理器核心0的温度在下一周期的预测值。具体的，根据下述公式确定处理器核心0的温度在下一周期的预测值：
Figure PCTCN2020098564-appb-000007
其中，T(t+1)为处理器核心0的温度在下一周期的预测值，T(t)为处理器核心0的温度在当前周期的测量值，T(t-1)为处理器核心0的温度在前一周期的测量值，P(t+1)为处理器核心0的功耗在下一周期的预测值，P(t)为处理器核心0的功耗在当前周期的测量值，P(t-1)为处理器核心0的功耗在前一周期的测量值。
由上可知，通过确定处理器核心0的功耗在下一周期的第一取值，并根据第一校正表和第二校正表确定处理器核心0的负载在下一周期的变化趋势和校正值，以及根据处理器核心0的负载在下一周期的变化趋势和校正值校正处理器核心0的功耗在下一周期的第一取值，以得到处理器核心0的功耗在下一周期的预测值，以及根据处理器核心0的功耗在下一周期的预测值确定处理器核心0的温度在下一周期的预测值，提供了一种确定处理器核心0的温度在下一周期的预测值的方式，且确定方式简单，易于执行，提高了确定处理器核心0的温度在下一周期的预测值的效率。另外，由于在确定处理器核心0的温度在下一周期的预测值时，考虑了处理器核心0的功耗在下一周期的值，即考虑了处理器核心0在下一周期的工作状态，因此，提高了确定处理器核心0的温度在下一周的预测值的准确性。此外，在确定处理器核心0的功耗在下一周期的预测值的过程中，考虑了处理器核心0的负载在下一周期的变化对处理器核心0的功耗在下一周期的预测值的影响，进一步提高了确定处理器核心0的功耗在下一周期的预测值的准确性，从而进一步提高了由处理器核心0的功耗在下一周期的预测值确定的处理器核心0的温度在下一周期的预测值的准确性。
需要说明的是，上述计算处理器核心0的温度在下一周期的预测值的方式仅为示例性的，并不用于限定本申请。
需要说明的是，上述其他各处理器核心的温度在下一周期的预测值的确定原理如上所述，此处不再赘述。
需要说明的是，针对其他待预测变量，确定处理器核心的待预测变量在下一周期的预测值的原理与上述原理类似，因此此处不再赘述。
进一步的，在确定处理器核心的待预测变量在下一周期的预测值之后，控制逻辑还可以根据每个处理器核心的待预测变量在下一周期的预测值确定每个处理器核心的调节策 略，调节策略用于指示处理器核心的电压和/或频率的调节方式。具体的，控制逻辑根据每个处理器核心的调节策略分别生成对应的处理器核心的频率调节指令和/或电压调节指令。通过生成每个处理器核心的调节策略，以在下一周期通过每个处理器核心的调节策略对对应的处理器核心的电压和/或频率进行调节，以确保处理器在安全范围内运行。
在此基础上，控制逻辑还用于将每个处理器核心的频率调节指令和/或电压调节指令发送至对应的处理器核心的频率电压调节电路；
每个频率电压调节电路根据对应处理器核心的频率调节指令和/或电压调节指令对对应的处理器核心的频率和/或电压进行调节。
例如，若待预测变量为处理器核心的温度，且处理器核心的温度在下一周期的预测值大于预设的水准线，则控制逻辑调用降低处理器核心的功耗的调节策略，以及在下一周期到来时，根据降低处理器核心的功耗的调节策略生成电压调节指令，并将电压调节指令发送至该处理器核心的频率电压调节电路，频率电压调节电路根据电压调节指令调整处理器核心的电压，以通过降低处理器核心的电压降低处理器核心的温度，从而确保处理器核心的温度在预设的水准线之下，进而确保处理器工作在正常温度下。
需要说明的是，控制逻辑可以根据每个处理器核心的待预测变量在下一周期的预测值分别确定每个处理器核心的调节策略。控制逻辑还可以综合考虑每个处理器核心的待预测变量在下一周期的预测值确定每个处理器核心的调节策略。
由上可知，由于调节策略由处理器核心的待预测变量在下一周期的预测值确定，而处理器核心的待观测变量在下一周期的预测值考虑了处理器核心在下一周期的运行状态，因此调节策略中也考虑了处理器核心在下一周期的运行状态所带来的影响，使得根据调节策略对处理器核心的下一周期的频率和/或电压进行调节时，调节策略可以应对处理器在下一周期的工作状态，进而使得处理器能够正常运行，避免出现过流的风险。
需要说明的是，上述处理器的结构仅为示例性的，并不用于限定本申请。
例如，处理器还可以包括更多或者更少的处理器核心，即处理器包括至少一个处理器核心。
处理器中还可以包括更多或者更少的感应装置。例如，感应装置的数量与处理器中的处理器核心的数量相同，且感应装置与处理器核心一一对应，通过每个感应装置获取对应的处理器核心的待预测变量在下一个周期的预测值。再例如，感应装置的数量为一个，处理器中的处理器核心共用该感应装置，该感应装置可以获取处理器的所有处理器核心的待预测变量在当前周期的测量值，即可以将处理器中的所有处理器核心看做一个整体获取这个整体的待预测变量在当前周期的测量值，以基于上述原理预测这个整体的待预测变量在下一周期的预测值；或者该一个感应装置还可以分别获取处理器中的每个处理器核心的待预测变量在当前周期的测量值，以基于上述原理分别确定待预测变量在下一周期的预测值，由于仅存在一个感应装置，因此可以分时获取不同处理器核心的待预测变量在当前周期的测量值。再例如，处理器中的一部分处理器核心可以共用一个感应装置，另一部分处理器核心可以共用另一个感应装置，本申请实施例对此不作特殊限定。
处理器中还可以包括更多或这更少的频率电压调节电路。例如，频率电压调节电路的数量与处理器中的处理器核心的数量相同，且频率电压调节电路与处理器核心一一对应，以使频率电压调节电路调节对应的处理器核心的频率和/或电压。再例如，频率电压调节 电路的数量为一个，处理器中的处理器核心共用该频率电压调节电路，该频率电压可以分时调节每个处理器核心的频率和/或电压。再例如，处理器中的一部分处理器核心共用一个频率电压调节电路，处理器中的另一部分处理器核心共用另一个频率电压调节电路。
在本申请的其他实施例中，处理器还可以包括更多或者更少的部件，本申请实施例对此不作特殊限定。
在本申请的其他实施例中，还可以基于上述原理确定处理器中除过处理器核心之外的各个部件的待预测变量在下一周期的预测值。处理器中除过处理器核心之外的各个部件可以共用一个感应装置和一个频率电压调节电路，也可以分别为每个部件设置一个感应装置和频率电压调节电路等，本申请实施例对此不作特殊限定。
图4为本申请实施例提供的一种变量的预测方法的流程示意图，该方法的执行主体例如可以为处理器中的感应装置和控制逻辑等，本申请实施例对此不作特殊限定。如图4所示，该方法包括以下步骤：
步骤401，获取所述至少一个处理器核心的待预测变量在当前周期的测量值；
步骤402，根据所述至少一个处理器核心的待预测变量在当前周期的测量值确定所述至少一个处理器核心的待预测变量在下一周期的预测值。
在一种可能的实施方式中，所述待预测变量为所述处理器核心的功耗；所述根据所述至少一个处理器核心的待预测变量在当前周期的测量值确定所述至少一个处理器核心的待预测变量在下一周期的预测值包括：获取所述至少一个处理器核心的频率、电压在当前周期的测量值；确定所述至少一个处理器核心的频率、电压在所述下一周期的值；根据所述至少一个处理器核心的功耗在所述当前周期的测量值、所述至少一个处理器核心的频率和电压在当前周期的测量值以及所述至少一个处理器核心的频率和电压在所述下一周期的值确定所述至少一个处理器核心的功耗在所述下一周期的预测值。
在一种可能的实施方式中，根据下述公式确定所述处理器核心的功耗在所述下一周期的预测值：
Figure PCTCN2020098564-appb-000008
其中：P(t+1)为所述处理器核心的功耗在所述下一周期的预测值，P(t)为所述处理器核心的功耗在所述当前周期的测量值，f(t+1)为所述处理器核心的频率在所述下一周期的值，f(t)为所述处理器核心的频率在所述当前周期的测量值，V(t+1)为所述处理器核心的电压在所述下一周期的值，V(t)为所述处理器核心的电压在所述当前周期的测量值。
在一种可能的实施方式中，所述方法还包括：获取所述至少一个处理器核心的温度在当前周期的测量值；获取所述至少一个处理器核心的温度在前一周期的测量值；获取所述至少一个处理器核心的功耗在前一周期的测量值；根据所述至少一个处理器核心的功耗和温度分别在所述前一周期和所述当前周期的测量值、所述至少一个处理器核心的功耗在所述下一周期的预测值，确定所述至少一个处理器核心的温度在所述下一周期的预测值。
在一种可能的实施方式中，根据下述公式确定所述处理器核心的温度在所述下一周期的预测值：
Figure PCTCN2020098564-appb-000009
其中，T(t+1)为所述处理器核心的温度在所述下一周期的预测值，T(t)为所述处理器核心的温度在所述当前周期的测量值，T(t-1)为所述处理器核心的温度在所述前一周期的测量值，P(t+1)为所述处理器核心的功耗在所述下一周期的预测值，P(t)为所述处理器核心的功耗在所述当前周期的测量值，P(t-1)为所述处理器核心的功耗在所述前一周期的测量值。
在一种可能的实施方式中，所述根据所述至少一个处理器核心的待预测变量在当前周期的测量值确定所述至少一个处理器核心的待预测变量在下一周期的预测值包括：根据所述至少一个处理器核心的待预测变量在当前周期的测量值并结合一校正表确定所述至少一个处理器核心的待预测变量在下一周期的预测值。
在一种可能的实施方式中，所述待预测变量为所述处理器核心的功耗，所述校正表包括第一校正表和第二校正表；所述根据所述至少一个处理器核心的待预测变量在当前周期的测量值并结合一校正表确定所述至少一个处理器核心的待预测变量在下一周期的预测值包括：根据所述至少一个处理器核心的功耗在所述当前周期的测量值确定所述至少一个处理器核心的功耗在所述下一周期的第一取值；根据所述第一校正表确定所述至少一个处理器核心的负载在所述下一周期的变化趋势；根据所述第二校正表确定所述至少一个处理器核心的负载在所述下一周期的校正值；根据所述至少一个处理器核心的负载的所述变化趋势和所述至少一个处理器核心的负载的所述校正值对所述至少一个处理器核心的功耗的所述第一取值进行校正，得到所述至少一个处理器核心的功耗在所述下一周期的预测值。
在一种可能的实施方式中，所述方法还包括：获取所述至少一个处理器核心的功耗在前一周期的测量值；获取所述至少一个处理器核心的温度分别在所述当前周期和所述前一周期的测量值；根据所述至少一个处理器核心的功耗在下一周期的预测值，所述至少一个处理器核心的功耗在所述当前周期和所述前一周期的测量值，所述至少一个处理器核心的温度在所述当前周期和所述前一周期的测量值，确定所述至少一个处理器核心的温度在下一周期的预测值。
在一种可能的实施方式中，所述第一校正表包括T N个第一索引信息和T N个第一变化趋势，其中：所述T N个第一索引信息与所述T N个第一变化趋势一一对应；一个所述第一索引信息用于指示所述处理器核心的负载在N个连续周期中的一种变化趋势，T为所述处理器核心的负载在一个周期的变化趋势的种类数量；一个第一变化趋势用于指示所述处理器核心的负载在第一周期的值相比于所述处理器核心的负载在第二周期的值的变化趋势，其中，所述第一周期为所述第一变化趋势对应的第一索引信息所指示的N个连续周期的下一个周期，所述第二周期为所述第一变化趋势对应的第一索引信息所指示的N个连续周期中的最后一个周期；所述第二校正表包括T N个第二索引信息和T N个校正值，其中：所述T N个第二索引信息与所述T N个校正值一一对应；一个所述第二索引信息用于指示所述处理器核心的负载在N个连续周期中的变化趋势与所述处理器核心的负载在N个连续周期中的最后一个周期的值进行异或运算后得到的结果；一个校正值用于指示所述处理器核心的负载在第三周期的值相比于所述处理器核心的负载在第四周期的值的变化百分比，其中，所述第三周期为所述校正值对应的第二索引信息所指示的N个连续周期的下一周期，所述第 四周期为所述校正值对应的第二索引信息所指示的N个连续周期中的最后一个周期。
在一种可能的实施方式中，所述方法还包括：根据所述至少一个处理器核心的待预测变量在下一周期的预测值确定所述至少一个处理核心的调节策略，其中，所述调节策略用于指示所述处理器核心的电压和/或频率的调节方式。
在一种可能的实施方式中，所述根据所述至少一个处理器核心的待预测变量在下一周期的预测值确定所述至少一个处理核心的调节策略包括：根据所述至少一个处理器核心的调节策略生成所述至少一个处理器核心的频率调节指令和/或电压调节指令；根据所述至少一个处理器核心的频率调节指令和/或电压调节指令调节所述至少一个处理器核心的频率和/或电压。
本申请的上述方法的实现原理和技术效果与上述处理器中感应装置、控制逻辑、频率电压调节电路中所执行的步骤的原理和技术效果类似，此处不再赘述。
本申请还提供一种计算机可读存储介质，包括计算机程序，所述计算机程序在计算机上被执行时，使得所述计算机执行上述中的任一种方法实施例的技术方案。
本申请还提供一种计算机程序，当所述计算机程序被计算机执行时，用于执行上述中的任一种方法实施例的技术方案。
本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
所属领域的技术人员可以清楚地了解到，为描述的方便和简洁，上述描述的系统、装置和单元的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。
在本申请所提供的几个实施例中，应该理解到，所揭露的系统、装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，所述单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性，机械或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。
所述功能如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台计算机设备(个人计算机，服务器，或者网络设备等)执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器(read-only memory，ROM)、随机存取存 储器(random access memory，RAM)、磁碟或者光盘等各种可以存储程序代码的介质。以上所述，仅为本申请的具体实施方式，但本申请的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本申请揭露的技术范围内，可轻易想到变化或替换，都应涵盖在本申请的保护范围之内。因此，本申请的保护范围应以所述权利要求的保护范围为准。

Claims (24)

  1. 一种处理器，其特征在于，包括至少一个处理器核心、控制逻辑和至少一个感应装置；
    所述至少一个感应装置，用于获取所述至少一个处理器核心的待预测变量在当前周期的测量值；
    所述控制逻辑，用于根据所述至少一个处理器核心的待预测变量在当前周期的测量值确定所述至少一个处理器核心的待预测变量在下一周期的预测值。
  2. 根据权利要求1所述的处理器，其特征在于，所述待预测变量为所述处理器核心的功耗；
    所述根据所述至少一个处理器核心的待预测变量在当前周期的测量值确定所述至少一个处理器核心的待预测变量在下一周期的预测值包括：
    获取所述至少一个处理器核心的频率、电压在当前周期的测量值；
    确定所述至少一个处理器核心的频率、电压在所述下一周期的值；
    根据所述至少一个处理器核心的功耗在所述当前周期的测量值、所述至少一个处理器核心的频率和电压在当前周期的测量值以及所述至少一个处理器核心的频率和电压在所述下一周期的值确定所述至少一个处理器核心的功耗在所述下一周期的预测值。
  3. 根据权利要求2所述的处理器，其特征在于，根据下述公式确定所述处理器核心的功耗在所述下一周期的预测值：
    Figure PCTCN2020098564-appb-100001
    其中：P(t+1)为所述处理器核心的功耗在所述下一周期的预测值，P(t)为所述处理器核心的功耗在所述当前周期的测量值，f(t+1)为所述处理器核心的频率在所述下一周期的值，f(t)为所述处理器核心的频率在所述当前周期的测量值，V(t+1)为所述处理器核心的电压在所述下一周期的值，V(t)为所述处理器核心的电压在所述当前周期的测量值。
  4. 根据权利要求2或3所述的处理器，其特征在于，所述控制逻辑还用于：
    获取所述至少一个处理器核心的温度在当前周期的测量值；
    获取所述至少一个处理器核心的温度在前一周期的测量值；
    获取所述至少一个处理器核心的功耗在前一周期的测量值；
    根据所述至少一个处理器核心的功耗和温度分别在所述前一周期和所述当前周期的测量值、所述至少一个处理器核心的功耗在所述下一周期的预测值，确定所述至少一个处理器核心的温度在所述下一周期的预测值。
  5. 根据权利要求4所述的处理器，其特征在于，根据下述公式确定所述处理器核心 的温度在所述下一周期的预测值：
    Figure PCTCN2020098564-appb-100002
    其中，T(t+1)为所述处理器核心的温度在所述下一周期的预测值，T(t)为所述处理器核心的温度在所述当前周期的测量值，T(t-1)为所述处理器核心的温度在所述前一周期的测量值，P(t+1)为所述处理器核心的功耗在所述下一周期的预测值，P(t)为所述处理器核心的功耗在所述当前周期的测量值，P(t-1)为所述处理器核心的功耗在所述前一周期的测量值。
  6. 根据权利要求1所述的处理器，其特征在于，所述根据所述至少一个处理器核心的待预测变量在当前周期的测量值确定所述至少一个处理器核心的待预测变量在下一周期的预测值包括：
    根据所述至少一个处理器核心的待预测变量在当前周期的测量值并结合一校正表确定所述至少一个处理器核心的待预测变量在下一周期的预测值。
  7. 根据权利要求6所述的处理器，其特征在于，所述待预测变量为所述处理器核心的功耗，所述校正表包括第一校正表和第二校正表；
    所述根据所述至少一个处理器核心的待预测变量在当前周期的测量值并结合一校正表确定所述至少一个处理器核心的待预测变量在下一周期的预测值包括：
    根据所述至少一个处理器核心的功耗在所述当前周期的测量值确定所述至少一个处理器核心的功耗在所述下一周期的第一取值；
    根据所述第一校正表确定所述至少一个处理器核心的负载在所述下一周期的变化趋势；
    根据所述第二校正表确定所述至少一个处理器核心的负载在所述下一周期的校正值；
    根据所述至少一个处理器核心的负载的所述变化趋势和所述至少一个处理器核心的负载的所述校正值对所述至少一个处理器核心的功耗的所述第一取值进行校正，得到所述至少一个处理器核心的功耗在所述下一周期的预测值。
  8. 根据权利要求7所述的处理器，其特征在于，所述控制逻辑还用于：
    获取所述至少一个处理器核心的功耗在前一周期的测量值；
    获取所述至少一个处理器核心的温度分别在所述当前周期和所述前一周期的测量值；
    根据所述至少一个处理器核心的功耗在下一周期的预测值，所述至少一个处理器核心的功耗在所述当前周期和所述前一周期的测量值，所述至少一个处理器核心的温度在所述当前周期和所述前一周期的测量值，确定所述至少一个处理器核心的温度在下一周期的预测值。
  9. 根据权利要求7或8所述的处理器，其特征在于，所述第一校正表包括T N个第一索引信息和T N个第一变化趋势，其中：
    所述T N个第一索引信息与所述T N个第一变化趋势一一对应；
    一个所述第一索引信息用于指示所述处理器核心的负载在N个连续周期中的一种变化趋势，T为所述处理器核心的负载在一个周期的变化趋势的种类数量；
    一个第一变化趋势用于指示所述处理器核心的负载在第一周期的值相比于所述处理器核心的负载在第二周期的值的变化趋势，其中，所述第一周期为所述第一变化趋势对应的第一索引信息所指示的N个连续周期的下一个周期，所述第二周期为所述第一变化趋势对应的第一索引信息所指示的N个连续周期中的最后一个周期；
    所述第二校正表包括T N个第二索引信息和T N个校正值，其中：
    所述T N个第二索引信息与所述T N个校正值一一对应；
    一个所述第二索引信息用于指示所述处理器核心的负载在N个连续周期中的变化趋势与所述处理器核心的负载在N个连续周期中的最后一个周期的值进行异或运算后得到的结果；
    一个校正值用于指示所述处理器核心的负载在第三周期的值相比于所述处理器核心的负载在第四周期的值的变化百分比，其中，所述第三周期为所述校正值对应的第二索引信息所指示的N个连续周期的下一周期，所述第四周期为所述校正值对应的第二索引信息所指示的N个连续周期中的最后一个周期。
  10. 根据权利要求1～9中任一项所述的处理器，其特征在于，所述控制逻辑还用于：
    根据所述至少一个处理器核心的待预测变量在下一周期的预测值确定所述至少一个处理器核心的调节策略，其中，所述调节策略用于指示所述处理器核心的电压和/或频率的调节方式。
  11. 根据权利要求10所述的处理器，其特征在于，所述根据所述至少一个处理器核心的待预测变量在下一周期的预测值确定所述至少一个处理器核心的调节策略包括：
    根据所述至少一个处理器核心的调节策略生成所述至少一个处理器核心的频率调节指令和/或电压调节指令；
    所述处理器还包括至少一个频率电压调节电路；
    所述控制逻辑，还用于将所述至少一个处理器核心的频率调节指令和/或电压调节指令发送至所述至少一个频率电压调节电路；
    所述至少一个频率电压调节电路，用于根据所述至少一个处理器核心的频率调节指令和/或电压调节指令调节所述至少一个处理器核心的频率和/或电压。
  12. 一种变量的预测方法，其特征在于，包括：
    获取所述至少一个处理器核心的待预测变量在当前周期的测量值；
    根据所述至少一个处理器核心的待预测变量在当前周期的测量值确定所述至少一个处理器核心的待预测变量在下一周期的预测值。
  13. 根据权利要求12所述的方法，其特征在于，所述待预测变量为所述处理器核心的功耗；
    所述根据所述至少一个处理器核心的待预测变量在当前周期的测量值确定所述至少 一个处理器核心的待预测变量在下一周期的预测值包括：
    获取所述至少一个处理器核心的频率、电压在当前周期的测量值；
    确定所述至少一个处理器核心的频率、电压在所述下一周期的值；
    根据所述至少一个处理器核心的功耗在所述当前周期的测量值、所述至少一个处理器核心的频率和电压在当前周期的测量值以及所述至少一个处理器核心的频率和电压在所述下一周期的值确定所述至少一个处理器核心的功耗在所述下一周期的预测值。
  14. 根据权利要求13所述的方法，其特征在于，根据下述公式确定所述处理器核心的功耗在所述下一周期的预测值：
    Figure PCTCN2020098564-appb-100003
    其中：P(t+1)为所述处理器核心的功耗在所述下一周期的预测值，P(t)为所述处理器核心的功耗在所述当前周期的测量值，f(t+1)为所述处理器核心的频率在所述下一周期的值，f(t)为所述处理器核心的频率在所述当前周期的测量值，V(t+1)为所述处理器核心的电压在所述下一周期的值，V(t)为所述处理器核心的电压在所述当前周期的测量值。
  15. 根据权利要求13或14所述的方法，其特征在于，所述方法还包括：
    获取所述至少一个处理器核心的温度在当前周期的测量值；
    获取所述至少一个处理器核心的温度在前一周期的测量值；
    获取所述至少一个处理器核心的功耗在前一周期的测量值；
    根据所述至少一个处理器核心的功耗和温度分别在所述前一周期和所述当前周期的测量值、所述至少一个处理器核心的功耗在所述下一周期的预测值，确定所述至少一个处理器核心的温度在所述下一周期的预测值。
  16. 根据权利要求15所述的方法，其特征在于，根据下述公式确定所述处理器核心的温度在所述下一周期的预测值：
    Figure PCTCN2020098564-appb-100004
    其中，T(t+1)为所述处理器核心的温度在所述下一周期的预测值，T(t)为所述处理器核心的温度在所述当前周期的测量值，T(t-1)为所述处理器核心的温度在所述前一周期的测量值，P(t+1)为所述处理器核心的功耗在所述下一周期的预测值，P(t)为所述处理器核心的功耗在所述当前周期的测量值，P(t-1)为所述处理器核心的功耗在所述前一周期的测量值。
  17. 根据权利要求12所述的方法，其特征在于，所述根据所述至少一个处理器核心的待预测变量在当前周期的测量值确定所述至少一个处理器核心的待预测变量在下一周期的预测值包括：
    根据所述至少一个处理器核心的待预测变量在当前周期的测量值并结合一校正表确定所述至少一个处理器核心的待预测变量在下一周期的预测值。
  18. 根据权利要求17所述的方法，其特征在于，所述待预测变量为所述处理器核心的功耗，所述校正表包括第一校正表和第二校正表；
    所述根据所述至少一个处理器核心的待预测变量在当前周期的测量值并结合一校正表确定所述至少一个处理器核心的待预测变量在下一周期的预测值包括：
    根据所述至少一个处理器核心的功耗在所述当前周期的测量值确定所述至少一个处理器核心的功耗在所述下一周期的第一取值；
    根据所述第一校正表确定所述至少一个处理器核心的负载在所述下一周期的变化趋势；
    根据所述第二校正表确定所述至少一个处理器核心的负载在所述下一周期的校正值；
    根据所述至少一个处理器核心的负载的所述变化趋势和所述至少一个处理器核心的负载的所述校正值对所述至少一个处理器核心的功耗的所述第一取值进行校正，得到所述至少一个处理器核心的功耗在所述下一周期的预测值。
  19. 根据权利要求18所述的方法，其特征在于，所述方法还包括：
    获取所述至少一个处理器核心的功耗在前一周期的测量值；
    获取所述至少一个处理器核心的温度分别在所述当前周期和所述前一周期的测量值；
    根据所述至少一个处理器核心的功耗在下一周期的预测值，所述至少一个处理器核心的功耗在所述当前周期和所述前一周期的测量值，所述至少一个处理器核心的温度在所述当前周期和所述前一周期的测量值，确定所述至少一个处理器核心的温度在下一周期的预测值。
  20. 根据权利要求18或19所述的方法，其特征在于，所述第一校正表包括T N个第一索引信息和T N个第一变化趋势，其中：
    所述T N个第一索引信息与所述T N个第一变化趋势一一对应；
    一个所述第一索引信息用于指示所述处理器核心的负载在N个连续周期中的一种变化趋势，T为所述处理器核心的负载在一个周期的变化趋势的种类数量；
    一个第一变化趋势用于指示所述处理器核心的负载在第一周期的值相比于所述处理器核心的负载在第二周期的值的变化趋势，其中，所述第一周期为所述第一变化趋势对应的第一索引信息所指示的N个连续周期的下一个周期，所述第二周期为所述第一变化趋势对应的第一索引信息所指示的N个连续周期中的最后一个周期；
    所述第二校正表包括T N个第二索引信息和T N个校正值，其中：
    所述T N个第二索引信息与所述T N个校正值一一对应；
    一个所述第二索引信息用于指示所述处理器核心的负载在N个连续周期中的变化趋势与所述处理器核心的负载在N个连续周期中的最后一个周期的值进行异或运算后得到的结果；
    一个校正值用于指示所述处理器核心的负载在第三周期的值相比于所述处理器核心 的负载在第四周期的值的变化百分比，其中，所述第三周期为所述校正值对应的第二索引信息所指示的N个连续周期的下一周期，所述第四周期为所述校正值对应的第二索引信息所指示的N个连续周期中的最后一个周期。
  21. 根据权利要求12～20中任一项所述的方法，其特征在于，所述方法还包括：
    根据所述至少一个处理器核心的待预测变量在下一周期的预测值确定所述至少一个处理核心的调节策略，其中，所述调节策略用于指示所述处理器核心的电压和/或频率的调节方式。
  22. 根据权利要求21所述的方法，其特征在于，所述根据所述至少一个处理器核心的待预测变量在下一周期的预测值确定所述至少一个处理核心的调节策略包括：
    根据所述至少一个处理器核心的调节策略生成所述至少一个处理器核心的频率调节指令和/或电压调节指令；
    根据所述至少一个处理器核心的频率调节指令和/或电压调节指令调节所述至少一个处理器核心的频率和/或电压。
  23. 一种计算机可读存储介质，包括计算机程序，所述计算机程序在计算机上被执行时，使得所述计算机执行权利要求12～22中任一项所述的方法。
  24. 一种计算机程序，当所述计算机程序被计算机执行时，用于执行权利要求12～22中任一项所述的方法。
PCT/CN2020/098564 2020-06-28 2020-06-28 处理器和变量的预测方法 WO2022000135A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN202080102431.2A CN115769172A (zh) 2020-06-28 2020-06-28 处理器和变量的预测方法
PCT/CN2020/098564 WO2022000135A1 (zh) 2020-06-28 2020-06-28 处理器和变量的预测方法

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2020/098564 WO2022000135A1 (zh) 2020-06-28 2020-06-28 处理器和变量的预测方法

Publications (1)

Publication Number Publication Date
WO2022000135A1 true WO2022000135A1 (zh) 2022-01-06

Family

ID=79317805

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/098564 WO2022000135A1 (zh) 2020-06-28 2020-06-28 处理器和变量的预测方法

Country Status (2)

Country Link
CN (1) CN115769172A (zh)
WO (1) WO2022000135A1 (zh)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1991687A (zh) * 2005-12-29 2007-07-04 联想(北京)有限公司 节省处理器功耗的方法
US20080098254A1 (en) * 2006-10-24 2008-04-24 Peter Altevogt Method for Autonomous Dynamic Voltage and Frequency Scaling of Microprocessors
CN110245061A (zh) * 2019-04-30 2019-09-17 上海兆芯集成电路有限公司 计算装置、其功耗预测方法及计算机可读存储介质
CN110580209A (zh) * 2019-07-29 2019-12-17 杭州电子科技大学 一种基于模型预测控制的云服务器弹性伸缩与性能优化方法
CN111164538A (zh) * 2017-12-15 2020-05-15 英特尔公司 处理单元的功率管理

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1991687A (zh) * 2005-12-29 2007-07-04 联想(北京)有限公司 节省处理器功耗的方法
US20080098254A1 (en) * 2006-10-24 2008-04-24 Peter Altevogt Method for Autonomous Dynamic Voltage and Frequency Scaling of Microprocessors
CN111164538A (zh) * 2017-12-15 2020-05-15 英特尔公司 处理单元的功率管理
CN110245061A (zh) * 2019-04-30 2019-09-17 上海兆芯集成电路有限公司 计算装置、其功耗预测方法及计算机可读存储介质
CN110580209A (zh) * 2019-07-29 2019-12-17 杭州电子科技大学 一种基于模型预测控制的云服务器弹性伸缩与性能优化方法

Also Published As

Publication number Publication date
CN115769172A (zh) 2023-03-07

Similar Documents

Publication Publication Date Title
US10983576B2 (en) Method and apparatus for managing global chip power on a multicore system on chip
JP6528893B1 (ja) 学習プログラム、学習方法、情報処理装置
WO2021036905A1 (zh) 数据处理方法、装置、计算机设备和存储介质
CN108512650B (zh) 面向区块链的动态哈希计算方法、装置、节点及存储介质
CN115668097A (zh) 芯片的控制方法和控制装置
JP2019028829A (ja) 並列処理装置、並列処理方法、及びプログラム
WO2022099477A1 (zh) 一种功耗控制方法及装置
EP3798929A1 (en) Information processing apparatus, information processing method, and information processing program
WO2018196750A1 (zh) 处理乘加运算的装置和处理乘加运算的方法
WO2022000135A1 (zh) 处理器和变量的预测方法
CN117117992B (zh) 输出功率调节方法、装置、计算机设备和存储介质
WO2021185262A1 (zh) 计算装置、方法、板卡和计算机可读存储介质
JP2023006529A (ja) データ異常判定装置及び内部状態予測システム
JP2013025467A (ja) 対数関数の近似演算回路
US20200192633A1 (en) Arithmetic processing device and method of controlling arithmetic processing device
JP6646350B2 (ja) 様々な数値フォーマットのデータを用いてデータに基づく関数モデルを計算するためのモデル計算ユニット、および制御装置
CN112713616B (zh) 电力系统发电侧机组的控制方法、装置、设备及介质
CN114243690A (zh) 一种电网有功安全校正方法、装置、电子设备及存储介质
CN104182201A (zh) 尤其在模型计算单元中纯基于硬件计算函数模型的fma单元
KR20170037724A (ko) 복합화력발전기 공급능력 연산 장치, 시스템 및 방법
CN114280939A (zh) 一种模型预测控制方法、装置、设备及存储介质
US20210208849A1 (en) Arithmetic processing device, method for controlling arithmetic processing device, and non-transitory computer-readable storage medium
JP7494801B2 (ja) 情報処理装置、プログラム及び情報処理方法
CN118300195B (zh) 一种微电网智能监控与调度方法
CN112540633B (zh) 一种温度控制方法、装置、设备及介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20942504

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20942504

Country of ref document: EP

Kind code of ref document: A1