Task allocation method and device for multiple computing power units and storage medium

Abstract

The invention provides a task allocation method and device for multiple computing power units and a storage medium. The method comprises the following steps: in response to a received computing power task, obtaining a mapping relationship between a used computing power resource corresponding to each computing power unit and temperature; on the basis of the first computing power resource currently used by each computing power unit, the second computing power resource added for completing the computing power task and the mapping relation corresponding to each computing power unit, predicting the temperature of each computing power unit after the second computing power resource is added; and allocating the computing power task to at least one computing power unit based on the predicted temperature of each computing power unit. According to the method, the computing power task can be distributed based on the predicted temperature, it can be ensured that the computing power unit operates within a reasonable temperature range, the influence of the too high temperature of the computing power unit on the overall computing efficiency can be avoided, unnecessary energy consumption is reduced, the hardware service life of the computing power unit is prolonged, and more stable performance is provided.

Description

Technical Field

[0001] The present disclosure relates to the field of computing technology, and in particular to a method, device and storage medium for allocating tasks to multiple computing units. Background Art

[0002] With the rapid development of artificial intelligence technology, the computing power requirements for computing power terminal devices are increasing. In order to meet the computing power requirements, computing power terminal devices will use multiple computing power units to work together. However, in the process of multiple computing power units working together, the resource usage of different computing power units is not the same, resulting in different energy consumption and heat generation of different computing power units. In addition, different computing power units are arranged at different positions of the computing power terminal equipment, and the heat dissipation conditions at each position are not the same, which further aggravates the problem of uneven temperature distribution of multiple computing power units. Computing power units with poor heat dissipation conditions may work at high loads, while computing power units with good heat dissipation conditions may work at low loads, which in turn causes the temperature of individual computing power units to be too high.

[0003] If the temperature of the computing unit is too high, the internal resistance of the computing unit will increase, reducing the transmission efficiency and computing efficiency of the data inside the computing unit. In addition, it is possible to reduce the temperature of the computing unit by reducing the main frequency of the computing unit, resulting in a significant slowdown in the computing speed of the computing unit. It may even be necessary to reduce the computing tasks of the computing unit and reallocate them to other computing units, resulting in a decrease in the overall computing speed of the computing terminal device. Summary of the invention

[0004] In view of this, the embodiments of the present disclosure propose a task allocation method, device and storage medium for multiple computing units to solve the deficiencies in the related art.

[0005] According to a first aspect of an embodiment of the present disclosure, a method for allocating tasks to multiple computing units is provided, the method comprising:

[0006] In response to receiving the computing task, obtaining a mapping relationship between the computing resources used and the temperature corresponding to each computing unit;

[0007] Based on the first computing power resources currently used by each computing power unit, the second computing power resources added after completing the computing power task, and the mapping relationship corresponding to each computing power unit, predict the temperature of each computing power unit after adding the second computing power resources;

[0008] Based on the predicted temperatures of the respective computing units, the computing task is allocated to at least one computing unit.

[0009] According to a second aspect of an embodiment of the present disclosure, an electronic device is proposed, comprising a control unit and multiple computing units, wherein the control unit is used to execute the steps in the method for allocating tasks to multiple computing units described in the first aspect above.

[0010] According to a third aspect of an embodiment of the present disclosure, a non-volatile computer-readable storage medium is proposed, on which a computer program is stored. When the program is executed by a processor, the steps in the task allocation method for multiple computing units described in the first aspect are implemented.

[0011] According to an embodiment of the present disclosure, after receiving a computing power task, the temperature of each computing power unit after adding the second computing power resource is predicted based on the first computing power resource currently used by each computing power unit, the second computing power resource added to complete the computing power task, and the mapping relationship between the computing power resources used by each computing power unit and the temperature. Computing power tasks are allocated based on the predicted temperature, which can ensure that the computing power unit operates within a reasonable temperature range, avoid the impact of excessive temperature of the computing power unit on the overall computing efficiency, reduce unnecessary energy consumption, extend the hardware life of the computing power unit, and provide more stable performance.

[0012] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0014] Figure 1 It is a schematic diagram of an implementation environment according to an embodiment of the present disclosure.

[0015] Figure 2 It is a flowchart of a method for allocating tasks to multiple computing units according to an exemplary embodiment of the present disclosure.

[0016] Figure 3 The flowchart of a method for generating a mapping relationship shown in an exemplary embodiment of the present disclosure.

[0017] Figure 4 The figure is a flow chart of updating a mapping relationship shown in an exemplary embodiment of the present disclosure.

[0018] Figure 5 It is a flowchart of a method for allocating tasks of multiple computing units under different temperature ranges shown in an exemplary embodiment of the present disclosure.

[0019] Figure 6 It is a schematic diagram of an RK3588 terminal computing power board shown in an exemplary embodiment of the present disclosure.

[0020] Figure 7 It is a schematic diagram of a mapping relationship shown in an exemplary embodiment of the present disclosure.

[0021] Figure 8 It is a flowchart of a method for allocating tasks to multiple computing units according to an exemplary embodiment of the present disclosure.

[0022] Fig. 9 It is a schematic block diagram of an electronic device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of the present disclosure to clearly and completely describe the technical solutions in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present disclosure.

[0024] The task allocation method of multiple computing units provided in the present disclosure can be applied to electronic devices. In one embodiment, the electronic device is an AI (artificial intelligence) computing terminal device, which refers to a high-performance device specifically used to perform artificial intelligence-related computing tasks. For example, a server, an AI supercomputer, a computing box, etc. In another embodiment, the electronic device is a consumer electronic product, such as a mobile phone, a tablet computer, a laptop computer, a desktop computer, a smart home, etc. In another embodiment, the electronic device can also be a smart car, etc. The present disclosure does not limit the type of electronic device.

[0025] In one embodiment, if Figure 1 As shown, the electronic device includes a control unit and multiple computing units. In response to receiving a computing task, the control unit allocates the computing task to at least one of the multiple computing units. The embodiment of the present disclosure does not limit the number of the multiple computing units, and the number of the multiple computing units can be 2, 5, 10, or any integer greater than 1.

[0026] In one embodiment, the computing unit may be a chip, and multiple computing units are multiple chips. In another embodiment, the computing unit may be an NPU (Neural Processing Unit), and multiple computing units are multiple NPUs. Of course, the computing unit may also be a GPU (Graphics Processing Unit), a TPU (Tensor Processing Unit), etc., and the present disclosure does not limit the type of computing unit.

[0027] The task allocation scheme of the multi-computing units disclosed in the present invention is described in detail below in conjunction with the accompanying drawings and corresponding embodiments.

[0028] Figure 2 FIG. 1 is a flowchart of a method for allocating tasks to multiple computing units, shown in an exemplary embodiment of the present disclosure. Figure 2 As shown, the method is applied to an electronic device and includes the following steps 202-206:

[0029] Step 202, in response to receiving a computing task, obtain a mapping relationship between the computing resources used and the temperature corresponding to each computing unit.

[0030] In the embodiments of the present disclosure, a computing task is a task that requires a computing unit to complete. The embodiments of the present disclosure do not limit the type of computing tasks. For example, the computing task may be a computing task related to a neural network model, such as an image processing task, a feature extraction task, a prediction task, etc. A computing unit may be a chip, NPU, GPU, TPU, etc. in an electronic device. The embodiments of the present disclosure do not limit the type of computing units.

[0031] The computing resources used by the computing unit during operation are constantly changing with the computing operations it performs. The computing resources used by the computing unit in step 202 above refer to the computing resources used by the computing unit at a certain moment during operation. The more computing resources a computing unit is using, the more heat the computing unit generates, and the higher the temperature of the computing unit. The mapping relationship in step 202 above is used to reflect the relationship between the computing resources used by the computing unit and the temperature of the computing unit.

[0032] In one embodiment, the mapping relationship may be a curve of the relationship between the computing power resources used and the temperature. In another embodiment, the mapping relationship may be a function of the relationship between the computing power resources used and the temperature. In another embodiment, the mapping relationship may also be a mapping relationship table, which includes a plurality of computing power resources used and the temperatures corresponding to the plurality of computing power resources used. It should be noted that the embodiments of the present disclosure are only exemplary illustrations of the mapping relationship and do not limit the form of expression of the mapping relationship.

[0033] Step 204, based on the first computing resources currently used by each computing unit, the second computing resources added to complete the computing task, and the corresponding mapping relationship between each computing unit, predict the temperature of each computing unit after the second computing resources are added.

[0034] The first computing power resource refers to the computing power resource currently used by the computing power unit, and the second computing power resource refers to the computing power resource that will be increased if the computing power unit is used to complete the computing task. The temperature of each computing unit after adding the second computing power resource refers to the increase of the computing power resource used by each computing unit by the second computing power resource, that is, the computing power resource used by each computing unit is changed from the first computing power resource to the sum of the first computing power resource and the second computing power resource.

[0035] After a computing task is assigned to a computing unit, the computing resources used by the computing unit will inevitably increase, and thus the temperature of the computing unit will also increase. The above step 204 is to predict in advance what the future temperature of the computing unit will be if the computing task is assigned to the computing unit before assigning the computing task to the computing unit. Subsequently, it can be determined whether to assign the computing task to the computing unit based on whether the future temperature of the computing unit is within a reasonable range.

[0036] In one embodiment, based on the first computing power resources currently used by each computing power unit, the second computing power resources added to complete the computing power task, and the mapping relationship corresponding to each computing power unit, the temperature of each computing power unit after the second computing power resources are added is predicted, including: for each computing power unit, the first computing power resources currently used by the computing power unit are increased by the second computing power resources to obtain a fourth computing power resource, and the temperature corresponding to the fourth computing power resource in the mapping relationship is determined.

[0037] Step 206: Allocate the computing task to at least one computing unit based on the predicted temperatures of each computing unit.

[0038] According to the record in the above step 204, the predicted temperature of the computing unit indicates the temperature that the computing unit will reach in the future after the computing task is assigned to the computing unit. The electronic device can assign the computing task to the appropriate computing unit based on whether the future temperature of each computing unit is within a reasonable range to ensure that each computing unit operates within a reasonable temperature range.

[0039] According to an embodiment of the present disclosure, after receiving a computing power task, the temperature of each computing power unit after adding the second computing power resource is predicted based on the first computing power resource currently used by each computing power unit, the second computing power resource added to complete the computing power task, and the mapping relationship between the computing power resources used by each computing power unit and the temperature. Computing power tasks are allocated based on the predicted temperature, which can ensure that the computing power unit operates within a reasonable temperature range, avoid the impact of excessive temperature of the computing power unit on the overall computing efficiency, reduce unnecessary energy consumption, extend the hardware life of the computing power unit, and provide more stable performance.

[0040] Figure 3 is a flowchart of a method for generating a mapping relationship shown in an exemplary embodiment of this specification, such as Figure 3 As shown, the method is applied to electronic equipment. Figure 3 The generation process of the mapping relationship corresponding to the method described in the embodiment of the present disclosure is described in detail, and the process may include the following steps 301-308.

[0041] Step 301: During the operation of the computing unit, the computing resources used and the measured temperature of the computing unit at the current moment are collected to obtain a first data group.

[0042] During the operation of the computing unit, the computing resources used by the computing unit are constantly changing, and accordingly, the temperature of the computing unit also changes with the continuous change of the computing resources used by the computing unit. Therefore, during the operation of the computing unit, the computing resources used by the computing unit and the measured temperature at the current moment can be collected to obtain a first data group, and based on the first data groups collected at multiple moments, the mapping relationship between the computing resources used by the computing unit and the temperature can be analyzed.

[0043] The first data group includes the used computing resources and the measured temperature. Exemplarily, the first data group is represented as (xi, Ti), where xi represents the used computing resources of the computing unit at the i-th moment, and Ti represents the measured temperature of the computing unit at the i-th moment.

[0044] The embodiments of the present disclosure do not limit the collection method of the computing resources used by the computing unit and the measured temperature at the current moment, and are only exemplified by the following embodiments:

[0045] In one embodiment, the electronic device includes a control unit, which can monitor the operating status of the computing unit (such as the computing resources used, energy consumption, etc.), and the electronic device collects the computing resources used by the computing unit at the current moment through the control unit. Among them, the control unit can be a CPU main control unit. In another embodiment, the computing unit can actively report the computing resources used at the current moment.

[0046] In one embodiment, a temperature sensor is provided at a position corresponding to the computing power unit, and the electronic device obtains the measured temperature of the computing power unit through the temperature sensor. Exemplarily, a thermistor is arranged near the computing power unit, and the voltage on the thermistor is collected by utilizing the characteristic that the resistance of the thermistor changes with temperature. The temperature corresponding to the voltage is determined by the mapping relationship between the voltage and temperature of the thermistor, and the temperature is the measured temperature of the computing power unit. In another embodiment, the energy consumption of the computing power unit can also be determined, and the temperature of the computing power unit is calculated based on the energy consumption of the computing power unit, and the calculated temperature is used as the measured temperature of the computing power unit.

[0047] In an illustrated embodiment, the electronic device may perform the step of collecting the computing power resources used and the measured temperature of the computing unit at the current moment every first time period. The first time period may be 1 second, 2 seconds, 5 seconds, 10 seconds, 1 minute, etc., and the first time period is not limited in the embodiment of the present disclosure.

[0048] In an illustrated embodiment, considering that when the temperature of the computing unit is relatively low, allocating computing tasks to the computing unit will hardly affect the computing efficiency, energy consumption, etc. of the computing unit. In order to reduce the amount of calculation, when the temperature of the computing unit is relatively low, it is not necessary to perform the above steps 202 to 206 to allocate computing tasks. Since there is no need to perform the above steps 202 to 206 to allocate computing tasks, there is no need to generate a mapping relationship corresponding to the computing unit. Only when the temperature of the computing unit is high, the continued increase in temperature will affect the computing efficiency, energy consumption, etc. of the computing unit, so it is necessary to perform the above steps 202 to 206 to allocate computing tasks, and therefore, it is also necessary to generate a mapping relationship corresponding to the computing unit. In one embodiment, during the operation of the computing unit, in response to the measured temperature of the computing unit reaching the first temperature, the computing resources used and the measured temperature of the computing unit at the current moment are collected to obtain a first data group. Subsequently, the first data groups collected at multiple moments are fitted to obtain a mapping relationship corresponding to the computing unit.

[0049] Among them, the first temperature can be regarded as the temperature dividing line of whether it affects the computing efficiency of the computing power unit. When the temperature of the computing power unit is lower than the first temperature, the temperature change of the computing power unit will hardly affect the computing efficiency, energy consumption, etc. of the computing power unit. When the temperature of the computing power unit is higher than the first temperature, the temperature change of the computing power unit will have a certain impact on the computing efficiency, energy consumption, etc. of the computing power unit. In one embodiment, the first temperature can be determined based on the mapping relationship corresponding to the computing power unit under normal heat dissipation conditions. Exemplarily, the mapping relationship between the used computing power resources and the temperature in different intervals is different. According to the mapping relationship corresponding to the computing power unit, it can be determined from which used computing power resources the mapping relationship between the used computing power resources and the temperature begins to change, and the temperature corresponding to this change point can be used to determine the first temperature.

[0050] Of course, the first temperature may also be a value set by a technician, an empirical value, etc., and the first temperature is not limited in the embodiment of the present disclosure. For example, the first temperature may be 60°C.

[0051] Step 302: Perform fitting processing on the first data groups collected at multiple moments to obtain a first mapping relationship corresponding to the computing power unit.

[0052] In the embodiment of the present disclosure, any fitting algorithm can be used to fit the first data group collected at multiple moments, and the embodiment of the present disclosure does not limit the fitting algorithm. Exemplarily, the least squares method can be used to fit the first data group collected at multiple moments. Exemplarily, a neural network model can be used to fit the first data group collected at multiple moments.

[0053] The execution timing of the above step 302 can be set according to actual needs. The embodiments of the present disclosure are not limited to this, and are only exemplified by the following two embodiments. In one embodiment, when the number of collected first data groups reaches a first threshold, the collected multiple first data groups are fitted to obtain a first mapping relationship corresponding to the computing power unit. The first threshold can be any value, for example, 10, 20, 50, 100, etc.

[0054] In another embodiment, in order to avoid the use of computing resources in the collected multiple first data groups being concentrated in a small used computing resource interval, the mapping relationship fitted based on the collected first data group can only accurately represent the mapping relationship between the computing resources used in the small interval and the temperature, but cannot accurately represent the mapping relationship between the computing resources used in other intervals and the temperature. Before the fitting process is performed, the richness of the computing resources used in the multiple first data groups will be verified. Exemplarily, when the difference between the maximum used computing resources and the minimum used computing resources in the collected multiple first data groups reaches the second threshold, the collected multiple first data groups are fitted to obtain the first mapping relationship corresponding to the computing unit. Among them, the second threshold can be any value, for example, 5Tops (TeraOperations Per Second, an indicator to measure the performance of a computing device, referring to one trillion operations that can be completed per second), 10Tops, 15Tops, etc. The second threshold can be determined based on the total computing resources of the computing unit, or it can be an empirical value, or it can be a value set by a technician, and the embodiment of the present disclosure does not limit this.

[0055] Step 303, after obtaining the first mapping relationship corresponding to the computing power unit, collect the computing power resources used and the measured temperature of the computing power unit at the current moment to obtain a second data group.

[0056] Considering that the environment in which the electronic device is located may change, the heat dissipation conditions of the computing unit may also change. After the heat dissipation conditions of the computing unit change, it is obvious that the first mapping relationship that has been obtained is no longer accurate. In addition, the mapping relationship between the used computing resources and temperature in different intervals is different. If the first mapping relationship is obtained based on the used computing resources in a certain interval, then the first mapping relationship may not be very accurate for the used computing resources in other intervals.

[0057] In order to ensure that the temperature predicted based on the mapping relationship is accurate, after obtaining the first mapping relationship corresponding to the computing power unit, the embodiment of the present disclosure will continue to collect data groups, and verify whether the first mapping relationship is accurate based on the newly collected data groups. If the first mapping relationship is inaccurate, the mapping relationship can be re-fitted based on the newly collected data groups.

[0058] It should be noted that the embodiment of the present disclosure only distinguishes the data group collected before the mapping relationship is generated and the data group collected after the mapping relationship is generated by using the first data group and the second data group, which has no special meaning, and the collection method and data content of the first data group and the second data group are the same. In addition, the embodiment of the present disclosure also distinguishes the mapping relationships generated twice by using the first mapping relationship and the second mapping relationship, which has no special meaning, and the fitting algorithm, expression form, etc. of the first mapping relationship and the second mapping relationship are the same.

[0059] Step 304: predict the temperature corresponding to the computing resources used in the second data group based on the first mapping relationship.

[0060] In one embodiment, the first mapping relationship is a relationship curve between the computing power resources used and the temperature. Based on the first mapping relationship, predicting the temperature corresponding to the computing power resources used in the second data group includes: obtaining the temperature corresponding to the computing power resources used in the relationship curve. In another embodiment, the first mapping relationship is a relationship function between the computing power resources used and the temperature. Based on the first mapping relationship, predicting the temperature corresponding to the computing power resources used in the second data group includes: using the relationship function to calculate the temperature corresponding to the computing power resources used in the second data group. In another embodiment, the first mapping relationship is a mapping table of the computing power resources used and the temperature. Based on the first mapping relationship, predicting the temperature corresponding to the computing power resources used in the second data group includes: determining the temperature corresponding to the computing power resources used in the mapping table as the predicted temperature.

[0061] Step 305: If the difference between the predicted temperature and the measured temperature in the second data group meets the mapping relationship update condition, fitting processing is performed based on the latest collected multiple second data groups to obtain the second mapping relationship corresponding to the computing power unit.

[0062] If the predicted temperature differs greatly from the measured temperature in the second data group, it means that the first mapping relationship is inaccurate and the mapping relationship needs to be updated. Therefore, the mapping relationship update condition can be set based on the difference between the predicted temperature and the measured temperature in the second data group. Exemplarily, the mapping relationship update condition is that the error rate between the predicted temperature and the measured temperature in the second data group is higher than the target error rate. When the error rate is not higher than the target error rate, it means that the first mapping relationship is still accurate and has a strong predictive ability, and the first mapping relationship can continue to be used for temperature prediction. When the error rate is higher than the target error rate, it means that the first mapping relationship is not accurate enough, the predictive ability is reduced, and the mapping relationship needs to be refitted. Among them, the target error rate can be a pre-set error rate or a fitting error rate of the first mapping relationship. The embodiment of the present disclosure does not limit the target error rate.

[0063] In one embodiment, when refitting the mapping relationship, only the latest collected multiple second data groups may be fitted, or the latest collected multiple second data groups and the multiple first data groups used to fit the first mapping relationship may be fitted. That is, when refitting the mapping relationship, a new data group may be added for fitting. For example, Figure 4 As shown, when the error rate between the predicted temperature and the measured temperature in the second data group is not higher than the target error rate, the first mapping relationship can continue to be used to predict the temperature without performing the above step 305. When the error rate between the predicted temperature and the measured temperature in the second data group is higher than the target error rate, and the difference between the error rate and the target error rate is not greater than the third threshold value, the latest collected multiple second data groups can be added for fitting processing to obtain the second mapping relationship. In other words, the latest collected multiple second data groups and the multiple first data groups used to fit the first mapping relationship are fitted to obtain the second mapping relationship. When the error rate between the predicted temperature and the measured temperature in the second data group is higher than the target error rate, and the difference between the error rate and the target error rate is greater than the third threshold value, it means that the first mapping relationship can no longer meet the requirements of temperature prediction and needs to be refitted based on the new data group. In other words, only the latest collected multiple second data groups are fitted to obtain the second mapping relationship.

[0064] The third threshold value may be any value, and the embodiment of the present disclosure does not limit the third threshold value. Exemplarily, the third threshold value is 10%. The number of the multiple second data groups collected most recently may be any number, for example, 10, 20, etc., and may be set according to actual needs.

[0065] Exemplarily, the newly collected second data set is represented as (xj, Tj), and the temperature corresponding to xj predicted using the first mapping relationship is Tjy. The error rate between the predicted temperature Tjy and the measured temperature Tj in the second data set is expressed as: δj = |(Tjy-Tj) / Tj|, where |*| represents the absolute value of *. δj is the error rate between the predicted temperature and the measured temperature in the second data set.

[0066] It should be noted that the fitting methods of the first mapping relationship and the second mapping relationship are the same, and the embodiments of the present disclosure will not be described one by one here.

[0067] In an illustrated embodiment, the mapping relationship corresponding to the computing unit when the heat dissipation condition is normal is different from the mapping relationship corresponding to the computing unit when the heat dissipation condition is abnormal. In the above step 305, the reason why the mapping relationship update condition is met may be due to the temperature change characteristics of the computing unit itself, or it may be due to abnormal heat dissipation of the computing unit. Therefore, when the difference between the predicted temperature and the measured temperature in the second data group meets the mapping relationship update condition, the second data group can be recorded, and the characteristics of the second data group can be analyzed later to determine whether the computing unit has abnormal heat dissipation.

[0068] The method also includes: if the difference between the predicted temperature and the measured temperature in the second data group meets the mapping relationship update condition, the second data group is recorded as the mapping relationship transition point of the computing power unit; based on the recorded mapping relationship transition point of the computing power unit, determining whether the heat dissipation of the computing power unit is abnormal.

[0069] According to the above records, the mapping relationship update condition may include two conditions, one condition is: the error rate between the predicted temperature and the actual measured temperature in the second data group is higher than the target error rate, and the difference between the error rate and the target error rate is not greater than a third threshold; the other condition is: the error rate between the predicted temperature and the actual measured temperature in the second data group is higher than the target error rate, and the difference between the error rate and the target error rate is greater than the third threshold. In one embodiment, considering that in the case of "the error rate between the predicted temperature and the actually measured temperature in the second data group is higher than the target error rate, and the difference between the error rate and the target error rate is greater than the third threshold", the first mapping relationship and the second mapping relationship are quite different, and the corresponding second data group has more analysis value. Therefore, only in the case of "the error rate between the predicted temperature and the actually measured temperature in the second data group is higher than the target error rate, and the difference between the error rate and the target error rate is greater than the third threshold", the second data group is recorded as the mapping relationship transition point of the computing power unit, and in the case of "the error rate between the predicted temperature and the actually measured temperature in the second data group is higher than the target error rate, and the difference between the error rate and the target error rate is not greater than the third threshold", the second data group will not be recorded as the mapping relationship transition point.

[0070] Optionally, based on the recorded mapping relationship transition points of the computing power unit, determining whether the heat dissipation of the computing power unit is abnormal includes at least one of the following items: if the number of recorded mapping relationship transition points of the computing power unit exceeds the target number, determining that the heat dissipation of the computing power unit is abnormal; if the computing power resources used at the recorded mapping relationship transition points of the computing power unit are less than the third computing power resources, determining that the heat dissipation of the computing power unit is abnormal.

[0071] Under normal heat dissipation conditions, a computing unit has only one mapping relationship transition point. When the number of recorded mapping relationship transition points of a computing unit exceeds the target number, it means that the computing unit has too many mapping relationship transition points and the heat dissipation of the computing unit is abnormal.

[0072] When the computing power used by the computing unit reaches a certain value, if the computing power used continues to increase, the temperature will rise at a faster rate; however, when the computing power used by the computing unit has not reached a certain value, the temperature has already risen at a faster rate when the computing power used continues to increase. Obviously, the computing unit is also abnormal. The third computing power resource can be determined based on the mapping relationship obtained by the computing unit under normal heat dissipation conditions. The third computing power resource is determined based on the computing power used corresponding to the transition point in the mapping relationship obtained by the computing unit under normal heat dissipation conditions.

[0073] The target number can be any number, and the embodiment of the present disclosure does not limit the target number. In one embodiment, the target number is an empirical value or a value set by a technician.

[0074] It should be noted that after determining that the heat dissipation of the computing unit is abnormal, a corresponding reminder can also be made. For example, a pop-up window, notification message, text message, etc. are used to remind the user to check whether the cooling fan, cooling duct, and radiator are installed in place.

[0075] Step 306: Update the mapping relationship corresponding to the computing power unit from the first mapping relationship to the second mapping relationship.

[0076] Step 307, obtaining the computing resources used and the measured temperature of the computing unit at multiple moments within a preset time period, and obtaining third data groups corresponding to the multiple moments respectively.

[0077] The preset duration may be any duration, for example, 2 days, 3 days, etc. The embodiment of the present disclosure does not impose any limitation on the preset duration.

[0078] Some electronic devices are not moved after deployment, and the surrounding environment is relatively stable, and the heat dissipation conditions in the short term are also relatively stable. At this time, a fixed mapping relationship can be fitted according to the computing power resources used by the computing power unit at multiple moments within a preset time and the measured temperature. The fixed mapping relationship can be a short-term fixed mapping relationship or a long-term fixed mapping relationship, and the embodiments of the present disclosure do not limit this. If the fixed mapping relationship is a short-term fixed mapping relationship, the short-term fixed mapping relationship can be recalibrated at regular intervals.

[0079] In one embodiment, when the solution provided by the embodiment of the present disclosure is first used, the electronic device executes the above steps 301 to 306. After a period of use (for example, after 2 days), it can determine whether the data group within this period is stable and whether a fixed mapping relationship needs to be fitted based on the data group recorded during this period.

[0080] It should be noted that the third data group acquired in the above step 307 may include the first data group and the second data group collected in the process of executing the above steps 301 to 306, which is not limited in this embodiment of the present disclosure.

[0081] Step 308: If the multiple third data groups obtained meet the fitting condition, fitting processing is performed on the multiple third data groups obtained to obtain a fixed mapping relationship corresponding to the computing power unit.

[0082] If the multiple third data groups obtained are unstable, the multiple third data groups will be relatively scattered, and the fitting effect will be relatively poor, so it can be considered that the fitting condition is not met. If the multiple third data groups obtained are stable, the multiple third data groups will be relatively concentrated, and the fitting effect will be relatively good, so it can be considered that the fitting condition is met. Therefore, when the multiple third data groups obtained meet the fitting condition, the multiple third data groups obtained are fitted to obtain a fixed mapping relationship of the computing power unit.

[0083] In the disclosed embodiment, during the operation of the computing unit, the corresponding data is collected to generate the mapping relationship corresponding to the computing unit, thereby ensuring the accuracy of the mapping relationship. In addition, new data is continuously collected to verify whether the generated mapping relationship is accurate. When the generated mapping relationship is not accurate enough, the generated mapping relationship is updated in time, thereby further improving the accuracy of the mapping relationship.

[0084] In addition, whether the computing power unit has abnormal heat dissipation will be judged based on the transition point of the mapping relationship. When the computing power unit has abnormal heat dissipation, it can be discovered in time, and the heat dissipation abnormality of the computing power unit can be solved in time, thereby extending the service life of the computing power unit.

[0085] Figure 5 is a flowchart of a method for allocating tasks of multiple computing units in different temperature ranges, as shown in an exemplary embodiment of this specification. Figure 5 As shown, the method is applied to electronic equipment. Figure 5 The computing task allocation process of the computing unit in different temperature ranges corresponding to the method described in the embodiment of the present disclosure is described in detail, and the process may include the following steps 501-504.

[0086] Step 501, when the currently measured temperature of at least one computing power unit is lower than a first temperature, in response to receiving a computing power task, allocate the computing power task to the computing power unit whose currently measured temperature is lower than the first temperature.

[0087] according to Figure 3 It can be seen from the records in the illustrated embodiment that when the temperature of a computing power unit is lower than the first temperature, allocating computing power tasks to the computing power unit will hardly have an impact on the computing power unit. Therefore, when there are computing power units whose current measured temperature is lower than the first temperature, computing power tasks will be preferentially allocated to computing power units whose current measured temperature is lower than the first temperature.

[0088] In one embodiment, when the current measured temperature of at least one computing power unit is lower than the first temperature, in response to receiving the computing power task, the computing power task is allocated to the computing power unit whose current measured temperature is lower than the first temperature, including: when the current measured temperature of each computing power unit is lower than the first temperature, in response to receiving the computing power task, the computing power task is allocated to the target computing power unit based on at least one of the communication rate, main frequency and total computing power resources of each computing power unit. The target computing power unit is the optimal computing power unit determined based on at least one of the communication rate, main frequency and total computing power resources of each computing power unit.

[0089] In a case where the current measured temperature of at least one computing power unit is lower than the first temperature, in response to receiving a computing power task, the computing power task is allocated to the computing power unit whose current measured temperature is lower than the first temperature, and also includes: in a case where the current measured temperature of some computing power units is lower than the first temperature, and the some computing power units are at least two computing power units, in response to receiving the computing power task, based on at least one of the communication rate, main frequency and total computing power resources of the at least two computing power units, the computing power task is allocated to a target computing power unit among the at least two computing power units.

[0090] In a case where the current measured temperature of at least one computing power unit is lower than the first temperature, in response to receiving a computing power task, the computing power task is allocated to the computing power unit whose current measured temperature is lower than the first temperature. It also includes: in a case where the current measured temperature of only one computing power unit is lower than the first temperature, in response to receiving the computing power task, the computing power task is allocated to the computing power unit whose current measured temperature is lower than the first temperature.

[0091] Step 502, when the current measured temperature of each computing power unit is not lower than the first temperature, in response to receiving a computing power task, obtain the mapping relationship between the computing power resources used and the temperature corresponding to each computing power unit, based on the first computing power resources currently used by each computing power unit, the second computing power resources added to complete the computing power task, and the mapping relationship corresponding to each computing power unit, predict the temperature of each computing power unit after the second computing power resources are added, and based on the predicted temperature of each computing power unit, allocate the computing power task to at least one of the computing power units.

[0092] That is to say, the steps of acquiring the mapping relationship, predicting the temperature and allocating computing power tasks based on the predicted temperature are performed when the current measured temperature of each computing power unit is not lower than the first temperature.

[0093] In an illustrated embodiment, when allocating computing power tasks, the computing power task can be allocated to one computing power unit for execution, or the computing power task can be split into multiple computing power subtasks, and the multiple computing power subtasks are allocated to multiple computing power units for execution. Based on the predicted temperatures of each computing power unit, the computing power task is allocated to at least one computing power unit, including: allocating the computing power task to the computing power unit with the lowest predicted temperature; or, if the predicted temperatures of each computing power unit are higher than the target temperature, splitting the computing power task into multiple computing power subtasks, and respectively allocating the multiple computing power subtasks to multiple computing power units.

[0094] The target temperature may be a temperature that is about to reach the upper limit of the computing power unit, for example, 80° C. Of course, the target temperature may also be a temperature that has reached the upper limit of the computing power unit, for example, 95° C. Of course, the target temperature may also be other temperatures, which may be set according to actual needs, and the embodiments of the present disclosure are not limited thereto.

[0095] When splitting a computing task into multiple computing subtasks, the number of multiple computing subtasks can be determined based on the predicted temperature of each computing unit. If the predicted temperature of each computing unit is much higher than the target temperature, the computing task can be split into more computing subtasks and assigned to more computing units for execution, ensuring that the computing units to which the computing subtasks are assigned will not have a large temperature rise. If the predicted temperature of each computing unit is only slightly higher than the target temperature, the computing task can be split into fewer computing subtasks and assigned to fewer computing units for execution. Among them, how many computing subtasks the computing task is split into can be determined by a model, and the model is trained using the predicted sample temperatures of each computing unit and the sample number of the split computing subtasks, so that the model can determine the number of computing subtasks based on the predicted temperature of each computing unit. Of course, other methods can also be used to determine the number of split computing subtasks, and the embodiments of the present disclosure are not limited to this. For example, the computing task can also be split into multiple computing subtasks based on the characteristics of the computing task.

[0096] In one embodiment, when multiple computing power subtasks are respectively assigned to multiple computing power units, the multiple computing power units are the computing power units with the lowest predicted temperatures among all computing power units. Of course, the multiple computing power units can also be determined in other ways, and the embodiments of the present disclosure are not limited to this. For example, the multiple computing power units can also be randomly selected.

[0097] Step 503, when the currently measured temperature of each computing power unit is not lower than the second temperature and not higher than the third temperature, reduce the main frequency of each computing power unit until the currently measured temperature of any computing power unit is not higher than the fourth temperature, and restore the main frequency of the computing power unit.

[0098] The second temperature is higher than the first temperature, and the second temperature is used to indicate that the temperature is about to reach the upper limit of the computing unit. If the temperature of the computing unit exceeds the second temperature, it means that the temperature has had a significant impact on the computing efficiency, energy consumption, etc. of the computing unit. The third temperature is used to indicate that the temperature has reached the upper limit of the computing unit. If the temperature of the computing unit exceeds the third temperature, it means that the temperature has had a significant impact on the computing efficiency, energy consumption, etc. of the computing unit.

[0099] The fourth temperature is lower than the second temperature. When the temperature of the computing unit is reduced to the fourth temperature, the influence of the temperature on the computing efficiency, energy consumption, etc. of the computing unit will be significantly reduced. Exemplarily, the fourth temperature is 75°C.

[0100] When the currently measured temperature of each computing power unit is not lower than the second temperature and not higher than the third temperature, the main frequency of each computing power unit is reduced. This can reduce the computing rate and reduce the temperature of the computing power unit. Until the currently measured temperature of any computing power unit is not higher than the fourth temperature, the main frequency of the computing power unit is restored.

[0101] Step 504: When the currently measured temperature of each computing unit is not lower than the third temperature, the main frequency of each computing unit is reduced, and the computing tasks of each computing unit are reduced, until the currently measured temperature of any computing unit is not higher than the fourth temperature, and the main frequency and computing tasks of the computing unit are restored.

[0102] When the currently measured temperature of each computing unit is not lower than the third temperature, it means that the temperature of each computing unit has reached the upper limit. While reducing the main frequency, it is also necessary to reduce the computing tasks and reduce the workload of the computing unit so that the temperature of the computing unit can be reduced as soon as possible. When the temperature drops to the fourth temperature, the main frequency and computing tasks are restored.

[0103] In the embodiment of the present disclosure, temperature intervals are divided, and different operations are performed in different temperature intervals, so as to control the computing unit to operate within a reasonable temperature range as much as possible. It can also avoid the impact of excessive temperature of the computing unit on the overall computing efficiency, reduce unnecessary energy consumption, extend the hardware life of the computing unit, and provide more stable performance.

[0104] Next, the embodiment of the present disclosure takes an electronic device including an RK3588 terminal computing power board, and implements the above solution in the RK3588 terminal computing power board as an example to introduce the solution:

[0105] The RK3588 terminal computing board includes a SoC (System on a Chip) mainboard and an NPU computing sub-board. The NPU computing sub-board is external and inserted into the SoC mainboard through a connector. The SoC mainboard includes a CPU main control unit and a computing unit NPU1. The NPU computing sub-board includes a computing unit NPU2. A thermistor Ta and Tb are respectively set near NPU1 and NPU2 to monitor the temperature of NPU1 and NPU2 in real time. Figure 6 As shown, use the ADC-1 signal line to connect the thermistor Ta to the ADC (Analog-to-Digital Converter) port of the RK3588 master, and use the ADC-2 signal line to connect the thermistor Tb to another ADC port of the RK3588 master.

[0106] Among them, since the CPU main control unit and NPU1 are packaged on the same silicon chip, the signal integrity of the CPU main control unit and NPU1 is high during communication, the pass rate is high, and a higher data throughput rate can be achieved. Generally, NPU1 is used first. The total computing power resources of NPU1 are 6Tops, and the total computing power resources of NPU2 are 26Tops. NPU2 is a computing power unit on an external computing power sub-board, which is connected to the CPU main control unit through the PCIE (Peripheral Component Interconnect Express, a high-speed serial computer expansion bus standard) data bus and the control GPIO (General Purpose Input Output), and the computing power data is bidirectionally transmitted through the PCIE bus. Since the PCIE data bus has a long line on the PCB (Printed Circuit Board) board and passes through the connector, the communication link is affected by the discontinuous characteristic impedance of the line, electromagnetic signal coupling, etc., the signal waveform is easily distorted, the communication rate is limited, and the data throughput rate is lower than that of NPU 1, so NPU 2 is used as a supplementary computing power unit.

[0107] Through experiments, the mapping relationship between the computing resources used by NPU1 and NPU2 and the temperature under normal heat dissipation conditions is obtained, such as Figure 7 As shown in the figure. Since NPU 2 is external and has better heat dissipation conditions, the curvature of the mapping relationship corresponding to NPU2 is lower. Figure 7 It can be seen that the overall trend of the mapping relationship corresponding to NPU1 and NPU2 is the same. After a certain amount of operator resources are exceeded, the temperature rises faster. Therefore, according to the transition point of the mapping relationship between NPU1 and NPU2, four temperature ranges can be divided: less than 60℃, 60℃ to 80℃, 80℃ to 95℃, and greater than 95℃.

[0108] According to these four temperature ranges, different computing task allocation processes can be executed, such as Figure 8 As shown, when the temperatures of NPU1 and NPU2 are both less than 60°C, NPU1 is used first. When the computing power resources of NPU1 are insufficient, they are allocated to NPU2.

[0109] When the temperature of NPU1 and NPU2 is lower than 60℃ or higher than 60℃, the computing task is assigned to the computing unit with a temperature lower than 60℃. During the operation of the computing unit with a temperature higher than 60℃, the computing resources used and the measured temperature (xi, Ti) of the computing unit are collected, and the mapping relationship and error rate are calculated and fitted by the least squares method.

[0110] The mapping relationship is T=k*x+b, where T represents temperature, x represents the computing power resources used, b is a constant, and k is the slope of the curve.

[0111] Error rate: The error rate is the error rate of the fitted mapping relationship, and can also be called the fitting error rate of the mapping relationship. ∑ is a summation function.

[0112] When the temperatures of NPU1 and NPU2 are both between 60°C and 80°C, predict the temperature that NPU1 will reach in the future if the computing power task is assigned to NPU1, and predict the temperature that NPU2 will reach in the future if the computing power task is assigned to NPU2; compare the predicted temperatures, and assign the computing power task to the computing power unit with the lower predicted temperature.

[0113] When the temperatures of NPU1 and NPU2 are both between 80℃-95℃, it means that the chip is close to the upper limit of the operating temperature. It is necessary to reduce the main frequency and the computing rate to restore the normal clock frequency after the temperature drops by 75℃.

[0114] When the measured temperature of NPU1 and NPU2 is greater than 95°C, the chip reaches the upper limit of the operating temperature. While reducing the main frequency, it is necessary to reduce computing resources and reduce the workload. The temperature should be reduced to 75°C before restoring normal clock frequency and computing load.

[0115] An embodiment of the present disclosure further proposes an electronic device, comprising a control unit and multiple computing units, wherein the control unit is configured to implement the task allocation method for multiple computing units described in any of the above embodiments.

[0116] An embodiment of the present disclosure further provides a non-volatile computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the method for allocating tasks to multiple computing units described in any of the above embodiments.

[0117] Regarding the device in the above embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment of the relevant method, and will not be elaborated here.

[0118] Fig. 9 1 is a schematic block diagram of an apparatus 900 for task allocation of multiple computing units according to an embodiment of the present disclosure. For example, the apparatus 900 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.

[0119] Reference Fig. 9, the device 900 may include one or more of the following components: a processing component 902 , a memory 904 , a power component 906 , a multimedia component 908 , an audio component 910 , an input / output (I / O) interface 912 , a sensor component 914 , and a communication component 916 .

[0120] The processing component 902 generally controls the overall operation of the device 900, such as operations associated with display, phone calls, data communications, camera operations, and recording operations. The processing component 902 may include one or more processors 920 to execute instructions to complete all or part of the steps of the above-mentioned method for allocating tasks to multiple computing units. For example, one processor 920 is used to complete all or part of the steps of the above-mentioned method for allocating tasks to multiple computing units to allocate computing tasks to other processors 920. In addition, the processing component 902 may include one or more modules to facilitate the interaction between the processing component 902 and other components. For example, the processing component 902 may include a multimedia module to facilitate the interaction between the multimedia component 908 and the processing component 902.

[0121] The memory 904 is configured to store various types of data to support operations on the device 900. Examples of such data include instructions for any application or method operating on the device 900, contact data, phone book data, messages, pictures, videos, etc. The memory 904 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0122] The power supply component 906 provides power to the various components of the device 900. The power supply component 906 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power for the device 900.

[0123] The multimedia component 908 includes a screen that provides an output interface between the device 900 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundaries of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 908 includes a front camera and / or a rear camera. When the device 900 is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera may receive external multimedia data. Each front camera and rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.

[0124] The audio component 910 is configured to output and / or input audio signals. For example, the audio component 910 includes a microphone (MIC), and when the device 900 is in an operating mode, such as a call mode, a recording mode, and a speech recognition mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in the memory 904 or sent via the communication component 916. In some embodiments, the audio component 910 also includes a speaker for outputting audio signals.

[0125] I / O interface 912 provides an interface between processing component 902 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include but are not limited to: home button, volume button, start button, and lock button.

[0126] The sensor assembly 914 includes one or more sensors for providing various aspects of status assessment for the device 900. For example, the sensor assembly 914 can detect the open / closed state of the device 900, the relative positioning of components, such as the display and keypad of the device 900, and the sensor assembly 914 can also detect the position change of the device 900 or a component of the device 900, the presence or absence of user contact with the device 900, the orientation or acceleration / deceleration of the device 900, and the temperature change of the device 900. The sensor assembly 914 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 914 may also include an optical sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 914 may also include an accelerometer, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0127] The communication component 916 is configured to facilitate wired or wireless communication between the device 900 and other devices. The device 900 can access a wireless network based on a communication standard, such as WiFi, 2G or 3G, 4G LTE, 6G NR or a combination thereof. In an exemplary embodiment, the communication component 916 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 916 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0128] In an exemplary embodiment, the device 900 can be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to execute the above-mentioned task allocation method for multiple computing units.

[0129] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 904 including instructions, and the instructions can be executed by the processor 920 of the device 900 to complete the task allocation method of the multi-computing unit. For example, the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0130] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the embodiments disclosed herein. The present disclosure is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and embodiments are to be considered as exemplary only, and the true scope and spirit of the present disclosure are indicated by the following claims.

[0131] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

[0132] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. The terms "include", "comprises" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "comprises a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0133] The method and device provided in the embodiments of the present disclosure are introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present disclosure. The description of the above embodiments is only used to help understand the method of the present disclosure and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present disclosure, there will be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be understood as a limitation on the present disclosure.

Claims

1. A method for allocating tasks to multiple computing units, characterized in that: The method comprises: In response to receiving the computing task, obtaining a mapping relationship between the computing resources used and the temperature corresponding to each computing unit; Based on the first computing power resources currently used by each computing power unit, the second computing power resources added after completing the computing power task, and the mapping relationship corresponding to each computing power unit, predict the temperature of each computing power unit after adding the second computing power resources; Based on the predicted temperatures of the respective computing units, the computing task is allocated to at least one computing unit.

2. The method according to claim 1, characterized in that The generation process of the mapping relationship corresponding to any computing power unit includes: During the operation of the computing unit, the computing resources used and the measured temperature of the computing unit at the current moment are collected to obtain a first data group; Fitting processing is performed on the first data groups collected at multiple moments to obtain a first mapping relationship corresponding to the computing power unit.

3. The method according to claim 2, characterized in that During the operation of the computing unit, the computing resources used and the measured temperature of the computing unit at the current moment are collected to obtain a first data group, including: During the operation of the computing unit, in response to the actually measured temperature of the computing unit reaching a first temperature, the computing resources used and the actually measured temperature of the computing unit at the current moment are collected to obtain the first data group.

4. The method according to claim 2, characterized in that: The method further comprises: After obtaining the first mapping relationship corresponding to the computing power unit, collecting the computing power resources used and the measured temperature of the computing power unit at the current moment to obtain a second data group; Based on the first mapping relationship, predicting the temperature corresponding to the computing resources used in the second data group; If the difference between the predicted temperature and the measured temperature in the second data group meets the mapping relationship update condition, a fitting process is performed based on the latest collected multiple second data groups to obtain a second mapping relationship corresponding to the computing power unit; The mapping relationship corresponding to the computing power unit is updated from the first mapping relationship to the second mapping relationship.

5. The method according to claim 4, characterized in that The method further comprises: If the difference between the predicted temperature and the actually measured temperature in the second data group satisfies the mapping relationship update condition, recording the second data group as a mapping relationship transition point of the computing power unit; Based on the recorded mapping relationship transition point of the computing power unit, determine whether the heat dissipation of the computing power unit is abnormal.

6. The method according to claim 5, characterized in that The determining whether the heat dissipation of the computing unit is abnormal based on the recorded mapping relationship transition point of the computing unit includes at least one of the following: If the number of recorded mapping relationship transition points of the computing power unit exceeds the target number, it is determined that the heat dissipation of the computing power unit is abnormal; If the used computing power resources in the recorded mapping relationship transition point of the computing power unit are less than the third computing power resources, it is determined that the heat dissipation of the computing power unit is abnormal.

7. The method according to claim 1, characterized in that The generation process of the mapping relationship corresponding to the computing power unit includes: Obtaining the computing resources used and the measured temperature of the computing unit at multiple moments within a preset time period, and obtaining third data groups corresponding to the multiple moments respectively; If the multiple third data groups obtained meet the fitting condition, fitting processing is performed on the multiple third data groups obtained to obtain a fixed mapping relationship corresponding to the computing power unit.

8. The method according to claim 1, characterized in that: The steps of acquiring the mapping relationship, predicting the temperature and allocating the computing power task based on the predicted temperature are performed when the currently measured temperature of each computing power unit is not lower than the first temperature.

9. The method according to claim 8, characterized in that The method further comprises at least one of the following: In a case where a currently measured temperature of at least one computing unit is lower than the first temperature, in response to receiving the computing task, allocating the computing task to a computing unit whose currently measured temperature is lower than the first temperature; When the currently measured temperature of each computing unit is not lower than the second temperature and not higher than the third temperature, the main frequency of each computing unit is reduced until the currently measured temperature of any computing unit is not higher than the fourth temperature, and the main frequency of the computing unit is restored; When the currently measured temperature of each computing unit is not lower than the third temperature, the main frequency of each computing unit is reduced, and the computing tasks of each computing unit are reduced, until the currently measured temperature of any computing unit is not higher than the fourth temperature, and the main frequency and computing tasks of the computing unit are restored; The third temperature is higher than the second temperature, the second temperature is higher than the fourth temperature and the first temperature, and the fourth temperature is higher than the first temperature.

10. The method according to claim 1, characterized in that The allocating the computing task to at least one computing unit based on the predicted temperatures of each computing unit includes: Allocate the computing task to the computing unit with the lowest predicted temperature; or, If the predicted temperatures of each computing unit are higher than the target temperature, the computing task is split into multiple computing subtasks, and the multiple computing subtasks are respectively assigned to multiple computing units.

11. An electronic device, characterized in that: It comprises a control unit and a plurality of computing units, wherein the control unit is used to execute the steps described in any one of the methods of claims 1 to 10.

12. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 10 are implemented.

Images