Deep sleep control method and system for end-side AI audio chip

Abstract

The invention relates to the technical field of chip control, and provides a deep dormancy control method and system for an end-side AI audio chip, and the method comprises the steps: obtaining the current energy consumption data and task execution information of all audio processing units in the end-side AI audio chip, constructing an energy consumption optimization map for analysis, recognizing the audio processing units in a low-load state, and obtaining the current energy consumption data and task execution information of all audio processing units in the end-side AI audio chip; mapping the task to a high-efficiency audio processing unit, and after a task reconstruction sequence is obtained, performing dynamic dormancy calculation on the audio processing unit in a low-load state to obtain a deep dormancy target unit and dormancy parameters of the deep dormancy target unit; and configuring a power consumption control module of the end-side AI audio chip by using the sleep parameter, so that the deep sleep target unit enters a low-power-consumption mode. By analyzing and processing the current energy consumption data and the task execution information, flexible and efficient energy consumption management is realized, and the problems that the energy consumption cannot be flexibly controlled and the energy consumption optimization is not flexible enough when the task load is frequently changed are solved.

Description

Technical Field

[0001] The present application relates to the technical field of chip control, and in particular, to a deep sleep control method and system for an end-side AI audio chip. Background Art

[0002] In recent years, with the rapid development of artificial intelligence technology, edge AI audio chips have been widely used in smart homes, smart speakers, smart wearable devices and other fields. These devices use edge AI audio chips to achieve more intelligent audio processing and interactive functions, meeting users' needs for high efficiency and low power consumption.

[0003] Among the related technical means, the audio chip sleep control technology is mainly used to reduce power consumption through fixed-cycle sleep and wake-up strategies, thereby reducing energy consumption and extending the battery life of the device to some extent. However, its power consumption optimization effect is limited by the setting of the sleep cycle.

[0004] Regarding the above technical solution, although the power consumption of the audio processing unit can be reduced through a fixed-cycle sleep and wake-up strategy, it is impossible to flexibly control the energy consumption when the task load changes frequently, and the power consumption optimization effect is limited by the setting of the sleep cycle, there is a problem that the energy consumption optimization is not flexible enough. Summary of the invention

[0005] In order to improve the problem that energy consumption cannot be flexibly controlled when the task load changes frequently and the energy consumption optimization is not flexible enough, the present application provides a deep sleep control method and system for an end-side AI audio chip.

[0006] The present invention provides a deep sleep control method for an end-side AI audio chip, comprising: obtaining current energy consumption data and task execution information of all audio processing units in the end-side AI audio chip, constructing an energy consumption optimization map based on the current energy consumption data and task execution information, performing topological analysis on the energy consumption optimization map, and identifying an audio processing unit in a low-load state; mapping the audio processing unit task in the low-load state to a high-efficiency audio processing unit according to a preset task migration mapping strategy to obtain a task reconstruction sequence, performing dynamic sleep calculation on the audio processing unit in the low-load state based on the task reconstruction sequence to obtain a deep sleep target unit and its sleep parameters; configuring a power consumption control module of the end-side AI audio chip using the sleep parameters of the deep sleep target unit to enable the deep sleep target unit to enter a low power consumption mode, and recording a sleep state sequence; performing adaptive wake-up calculation on the deep sleep target unit according to the sleep state sequence and the task scheduling information of the end-side AI audio chip to obtain a wake-up timing and a corresponding energy consumption threshold; performing recovery control on the deep sleep target unit according to the wake-up timing and the corresponding energy consumption threshold to enable the deep sleep target unit to re-enter the task execution state.

[0007] As a preferred scheme, the steps of obtaining all current energy consumption data and task execution information in the end-side AI audio chip, constructing an energy consumption optimization map based on the current energy consumption data and task execution information, performing topological analysis on the energy consumption optimization map, and identifying the audio processing unit in a low-load state include: obtaining the current energy consumption data and task execution information of each audio processing unit in the end-side AI audio chip through a power monitoring module and a timestamp recording module, preprocessing the current energy consumption data and the task execution information using Z-score standardization to obtain energy consumption preprocessing data and task execution feature data; calculating the energy consumption load level and task execution intensity of each audio processing unit based on the energy consumption preprocessing data to obtain energy consumption load evaluation data; constructing task dependency relationships between audio processing units according to the task execution feature data to obtain task topological structure data; optimizing the energy consumption load evaluation data through the task topological structure data, and constructing an energy consumption optimization map based on the optimized energy consumption load evaluation data; performing topological analysis on the energy consumption optimization map to identify the audio processing unit in a low-load state.

[0008] As a preferred scheme, after the step of performing topological analysis on the energy consumption optimization map and identifying the audio processing unit in a low-load state, it includes: parsing the audio processing unit in the low-load state to obtain low-load state identification data and low-load state task data; calculating the energy consumption change trend of the audio processing unit in the low-load state under different task migration conditions to obtain energy consumption change trend data; identifying the set of migratable tasks and the corresponding task migration priority according to the energy consumption change trend data to obtain task migration priority data.

[0009] As a preferred scheme, the step of mapping the audio processing unit task in the low-load state to the high-efficiency audio processing unit according to a preset task migration mapping strategy to obtain a task reconstruction sequence, and performing dynamic sleep calculation on the audio processing unit in the low-load state based on the task reconstruction sequence to obtain a deep sleep target unit and its sleep parameters includes: judging candidate high-efficiency audio processing units according to the low-load state identification data, low-load state task data and task migration priority data in combination with a preset task migration mapping strategy, and generating high-efficiency audio processing unit candidate set data based on the candidate high-efficiency audio processing units; calculating the task carrying capacity and task matching degree of the high-efficiency audio processing unit candidate set data to obtain to task carrying capacity data and task execution matching data; according to the task execution matching data, select the optimal high-efficiency audio processing unit, and map the low-load state task data to the optimal high-efficiency audio processing unit to obtain task reconstruction sequence data; calculate the impact of the task reconstruction sequence data on the audio processing unit in the low-load state, and obtain the energy consumption data of the audio processing unit in the low-load state after task migration; calculate the optimal sleep mode of the audio processing unit in the low-load state through the energy consumption data after task migration, and obtain the sleep mode data and sleep time data; based on the sleep mode data and the sleep time data, determine the deep sleep target unit, and calculate the sleep parameters corresponding to the deep sleep target unit.

[0010] As a preferred solution, the task carrying capacity and task matching degree of the candidate set data of the high-efficiency audio processing unit are calculated to obtain the calculation formula of the task carrying capacity data and the task execution matching degree data as follows: in, The task-carrying capacity of the high-performance audio processing unit, is the number of candidate high-performance audio processing units, For the The processing capabilities of the candidate audio processing units, For the The load factor of the candidate audio processing units; in, For task execution matching, is the task intensity coefficient, The processing power required for the current task.

[0011] As a preferred solution, the calculation formula for calculating the influence of the task reconstruction sequence data on the audio processing unit in the low-load state to obtain the energy consumption data of the audio processing unit in the low-load state after task migration is as follows: in, Energy consumption data after task migration. The number of audio processing units in the task reconstruction sequence, For the The power consumption of the migrated audio processing unit, For the The time required for the migration task of each audio processing unit.

[0012] As a preferred scheme, the step of using the sleep parameters to configure the power consumption control module of the end-side AI audio chip so that the deep sleep target unit enters a low power consumption mode and records a sleep state sequence includes: transmitting the sleep parameters to the power consumption control module of the end-side AI audio chip, and configuring the power consumption control module to control the start of the low power consumption mode so that the deep sleep target unit enters a sleep state; in the low power consumption mode, recording the sleep state change sequence of each audio processing unit to obtain a sleep state sequence.

[0013] As a preferred scheme, the step of performing adaptive wake-up calculation on the deep sleep target unit according to the sleep state sequence and the task scheduling information of the end-side AI audio chip to obtain the wake-up timing and the corresponding energy consumption threshold includes: obtaining the task scheduling information of the end-side AI audio chip, establishing a mapping relationship between the task scheduling information and the sleep state sequence, and analyzing the mapping relationship by using a decision tree algorithm to determine the correlation between task scheduling and sleep timing, and obtaining a correlation result; predicting the task load information in the next cycle through the task scheduling information, and determining the wake-up requirement of the deep sleep target unit according to the task load information and the correlation result; calculating the wake-up timing of each task based on the wake-up requirement, and calculating the energy consumption threshold of the deep sleep target unit according to the wake-up timing.

[0014] As a preferred scheme, the step of performing recovery control on the deep sleep target unit according to the wake-up timing and the corresponding energy consumption threshold so that the deep sleep target unit re-enters the task execution state includes: generating a wake-up control instruction based on the wake-up timing and the corresponding energy consumption threshold, sending the wake-up control instruction to the deep sleep target unit through the power consumption control module, and starting the recovery process; judging whether the current power consumption meets the preset energy consumption threshold, and if not, delaying the wake-up timing; if it meets the preset energy consumption threshold, starting the wake-up program of the deep sleep target unit to restore the task execution capability of the deep sleep target unit.

[0015] The present application also provides a deep sleep control system for an end-side AI audio chip, including: an acquisition unit, used to acquire current energy consumption data and task execution information of all audio processing units in the end-side AI audio chip, and based on the current energy consumption data and task execution information, construct an energy consumption optimization map, perform topological analysis on the energy consumption optimization map, and identify the audio processing unit in a low-load state; a mapping unit, used to map the audio processing unit task in the low-load state to a high-efficiency audio processing unit according to a preset task migration mapping strategy, obtain a task reconstruction sequence, and perform dynamic sleep calculation on the audio processing unit in the low-load state based on the task reconstruction sequence. A deep sleep target unit and its sleep parameters are obtained; a configuration unit is used to configure the power consumption control module of the end-side AI audio chip using the sleep parameters of the deep sleep target unit, so that the deep sleep target unit enters a low power consumption mode and records a sleep state sequence; a calculation unit is used to perform adaptive wake-up calculation on the deep sleep target unit according to the sleep state sequence and the task scheduling information of the end-side AI audio chip, and obtain the wake-up timing and the corresponding energy consumption threshold; a control unit is used to perform recovery control on the deep sleep target unit according to the wake-up timing and the corresponding energy consumption threshold, so that the deep sleep target unit re-enters the task execution state.

[0016] Compared with the prior art, the present application has the following beneficial effects: long battery life and flexible energy consumption optimization. By acquiring the current energy consumption data and task execution information of all audio processing units in the end-side AI audio chip, an energy consumption optimization map is constructed, topological analysis is performed to identify audio processing units in low-load states, and tasks are reconstructed and dynamically dormant for these low-load state units according to the preset task migration mapping strategy, thereby obtaining deep sleep target units and their sleep parameters, and using these sleep parameters to configure the power consumption control module, so that the deep sleep target unit enters a low-power mode and records its sleep state sequence; based on the sleep state sequence and task scheduling information, an adaptive wake-up calculation is performed to determine the wake-up timing and energy consumption threshold, and the deep sleep target unit is restored and controlled to re-enter the task execution state. By dynamically adapting to changes in task loads, more flexible and efficient energy consumption management is achieved, effectively extending the battery life of the device, and improving the problem that energy consumption cannot be flexibly controlled when the task load changes frequently, and there is an inflexible problem in energy consumption optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0018] The structures, proportions, sizes, etc. illustrated in the drawings of this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with this technology. They are not used to limit the conditions under which the present invention can be implemented, and therefore have no substantive technical significance. Any structural modification, change in proportion or adjustment of size, without affecting the effects and purposes that can be achieved by the present invention, should still fall within the scope of the technical contents disclosed by the present invention.

[0019] Figure 1 It is a flowchart of a deep sleep control method of a terminal-side AI audio chip provided by an embodiment of the present invention; Figure 2 It is a schematic block diagram of the structure of a deep sleep control system of an end-side AI audio chip provided in an embodiment of the present invention.

[0020] Description of reference numerals: 10. Deep sleep control system of the end-side AI audio chip; 11. Acquisition unit; 12. Mapping unit; 13. Configuration unit; 14. Calculation unit; 15. Control unit. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.

[0023] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0024] It should be further understood that the term “and / or” used in the specification and appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0025] The technical solution of the present invention is further described below with reference to the accompanying drawings and through specific implementation methods.

[0026] Embodiment 1: like Figure 1 As shown, the present application provides a deep sleep control method for a terminal-side AI audio chip, comprising steps S100 to S500; Step S100: Obtain the current energy consumption data and task execution information of all audio processing units in the end-side AI audio chip, build an energy consumption optimization map based on the current energy consumption data and task execution information, perform topological analysis on the energy consumption optimization map, and identify the audio processing unit in a low-load state.

[0027] In this step, the current energy consumption data and task execution information of all audio processing units in the end-side AI audio chip are collected, and an energy consumption optimization map is constructed using data analysis tools. Specifically, a graph theory algorithm is used to perform topological analysis on the energy consumption optimization map, identify the audio processing units that are currently in a low-load state, and mark them as potential deep sleep target units.

[0028] For example, at a specific moment, the energy consumption data of audio processing units A, B and C are 0.1W, 0.05W and 0.02W respectively, while the energy consumption data of audio processing units D and E are 0.4W and 0.35W respectively. After spectrum analysis, A, B and C are identified as low-load state units.

[0029] Step S200: Map the audio processing unit task in a low-load state to a high-performance audio processing unit according to a preset task migration mapping strategy to obtain a task reconstruction sequence, and perform dynamic sleep calculation on the audio processing unit in a low-load state based on the task reconstruction sequence to obtain a deep sleep target unit and its sleep parameters.

[0030] In this step, the pre-set task migration mapping strategy is used to reallocate the tasks of the audio processing units in the low-load state to the high-performance audio processing units. Specifically, a task reconstruction sequence is generated through the task migration mapping algorithm, and a dynamic sleep calculation is performed on the audio processing units in the low-load state based on the reconstruction sequence to obtain the deep sleep target units and their corresponding sleep parameters.

[0031] For example, assuming that task T1 of audio processing unit A is mapped to audio processing unit D, the energy consumption data of D rises to 0.45W, and A is used as a deep sleep target unit, and its sleep parameter is set to a sleep period of 30 minutes.

[0032] Step S300: Use the sleep parameters of the deep sleep target unit to configure the power consumption control module of the end-side AI audio chip, so that the deep sleep target unit enters a low power consumption mode, and records the sleep state sequence.

[0033] In this step, the sleep parameters of the deep sleep target unit are used to configure the power consumption control module of the end-side AI audio chip. Specifically, the sleep parameters of each deep sleep target unit are input into the power consumption control module to put it into low power consumption mode and record its sleep state sequence.

[0034] For example, after the sleep parameters of the audio processing unit A are input into the power consumption control module, A enters the low power consumption mode, its power consumption data drops to 0.01W, and the sleep state sequence of A is recorded as "30 minutes sleep".

[0035] Step S400: According to the sleep state sequence and the task scheduling information of the end-side AI audio chip, adaptive wake-up calculation is performed on the deep sleep target unit to obtain the wake-up timing and the corresponding energy consumption threshold.

[0036] In this step, adaptive wake-up calculation is performed based on the sleep state sequence of the deep sleep target unit and the task scheduling information of the end-side AI audio chip. Specifically, by analyzing the task execution priority in the task scheduling information and the current energy consumption state of the deep sleep target unit, the appropriate wake-up time and the corresponding energy consumption threshold are determined.

[0037] For example, 5 minutes before the end of the sleep cycle of audio processing unit A, its task execution priority is increased, and the current energy consumption state is close to the preset threshold value 0.01W, then the wake-up time is set to the end of the sleep cycle, and the energy consumption threshold is 0.02W.

[0038] Step S500: performing recovery control on the deep sleep target unit according to the wake-up timing and the corresponding energy consumption threshold, so that the deep sleep target unit re-enters the task execution state.

[0039] In this step, the deep sleep target unit is restored and controlled according to its wake-up timing and energy consumption threshold. Specifically, when the preset wake-up timing and energy consumption threshold are reached, the low power consumption mode of the deep sleep target unit is released to make it re-enter the task execution state.

[0040] For example, when the sleep cycle of the audio processing unit A ends and the current energy consumption reaches 0.02 W, A is restored and controlled to execute task T1 again.

[0041] In this embodiment, by obtaining the current energy consumption data and task execution information of all audio processing units in the end-side AI audio chip, an energy consumption optimization map is constructed based on the current energy consumption data and task execution information. Then, the energy consumption optimization map is topologically analyzed to identify the audio processing unit in a low-load state. According to the preset task migration mapping strategy, the audio processing unit task in a low-load state is mapped to a high-efficiency audio processing unit to obtain a task reconstruction sequence. Based on the task reconstruction sequence, the audio processing unit in a low-load state is dynamically dormant to obtain a deep sleep target unit and its sleep parameters. Finally, the power consumption control module of the end-side AI audio chip is configured using the sleep parameters of the deep sleep target unit to enable the deep sleep target unit to enter a low-power mode, and record the sleep state sequence. At the same time, according to the sleep state sequence and the task scheduling information of the end-side AI audio chip, the deep sleep target unit is adaptively awakened and calculated to obtain the wake-up time and the corresponding energy consumption threshold, thereby realizing the recovery control of the deep sleep target unit. Accurately identify and dynamically sleep audio processing units in a low-load state, and realize more flexible and efficient energy consumption management. Compared with the fixed-cycle sleep and wake-up strategy, this solution provides better power consumption control when the task load changes frequently, thereby extending the battery life of the device, while reducing overall energy consumption, improving the working efficiency of the end-side AI audio chip, and improving the problem of inflexible energy consumption control when the task load changes frequently, and the lack of flexibility in energy consumption optimization.

[0042] Embodiment 2: In step S100, the current energy consumption data and task execution information of each audio processing unit in the end-side AI audio chip are obtained through the power monitoring module and the timestamp recording module, and the current energy consumption data and task execution information are preprocessed using Z-score standardization to obtain energy consumption preprocessing data and task execution feature data.

[0043] The power monitoring module monitors the power consumption of the audio processing unit in real time, and collects the current energy consumption data and task execution information of all audio processing units in the end-side AI audio chip in combination with the time information of the timestamp recording module. Specifically, these data are Z-score standardized to eliminate the energy consumption differences between different processing units and the dimensional impact of task execution feature data, ensuring data consistency and comparability.

[0044] For example, the current energy consumption data of the audio processing unit A is 0.1 W, and the task execution information is 20 times / second. After standardization, the energy consumption pre-processing data is -0.5, and the task execution characteristic data is 1.5.

[0045] The energy consumption load level and task execution intensity of each audio processing unit are calculated based on the energy consumption preprocessing data to obtain energy consumption load evaluation data.

[0046] By analyzing the energy consumption preprocessing data, the energy consumption load level of each audio processing unit is calculated, and the task execution intensity is calculated in combination with the task execution characteristic data. Specifically, the energy consumption preprocessing data is processed using a weighted average algorithm to obtain the energy consumption load level of each audio processing unit, and the task execution intensity is calculated using a linear regression model in combination with the task execution characteristic data.

[0047] For example, the energy consumption load level of the audio processing unit A is 0.2, the task execution intensity is 0.8, and the obtained energy consumption load evaluation data is (0.2, 0.8).

[0048] The task dependency relationship between the audio processing units is constructed according to the task execution feature data to obtain the task topology structure data.

[0049] By analyzing the task execution feature data, the task dependency relationships between the audio processing units are established. Specifically, the task execution feature data are processed using a correlation analysis method to identify task dependencies with high correlation, and task topology data are constructed based on these dependencies.

[0050] For example, the task execution correlation between audio processing units A and B is 0.85, and a task dependency relationship between A and B is established.

[0051] The energy consumption load evaluation data is optimized through the task topology data, and an energy consumption optimization map is constructed based on the optimized energy consumption load evaluation data.

[0052] The energy consumption load evaluation data is optimized by combining the task topology data with the energy consumption load evaluation data. Specifically, the task topology data is optimized by using a graph optimization algorithm to obtain optimized energy consumption load evaluation data, and an energy consumption optimization graph is constructed based on these data.

[0053] For example, the optimized energy consumption load evaluation data of audio processing units A and B are 0.15 and 0.75 respectively. When constructing the energy consumption optimization graph, A and B are connected to represent the task dependency.

[0054] Perform topological analysis on the energy optimization map to identify audio processing units in low-load state.

[0055] By performing topological analysis on the constructed energy consumption optimization graph, the audio processing units in low load state are identified. Specifically, the energy consumption optimization graph is analyzed node by node using a graph traversal algorithm to identify audio processing units with low energy consumption load level and task execution intensity.

[0056] For example, the energy consumption load level and task execution intensity of the audio processing unit A are 0.1 and 0.2 respectively, and the audio processing unit A is identified as an audio processing unit in a low load state.

[0057] Among them, after performing topological analysis on the energy consumption optimization map and identifying the audio processing unit in the low-load state, it includes: parsing the audio processing unit in the low-load state to obtain low-load state identification data and low-load state task data.

[0058] By performing a detailed analysis on the audio processing unit in the low-load state, its low-load state identification data and low-load state task data are extracted. Specifically, by analyzing the audio processing units in the low-load state one by one, recording their current task execution status and energy consumption data, forming low-load state identification data, and recording their corresponding task information as low-load state task data.

[0059] For example, the low load state identification data of the audio processing unit A is “low load”, and the task data is “task T1”.

[0060] The energy consumption variation trend of the audio processing unit in a low-load state under different task migration conditions is calculated to obtain energy consumption variation trend data.

[0061] By simulating and analyzing the low-load audio processing unit under different task migration conditions, its energy consumption change trend is calculated. Specifically, an energy consumption simulation tool is used to simulate the energy consumption of the audio processing unit under different task migration conditions to obtain energy consumption change trend data.

[0062] For example, when task T1 is migrated from audio processing unit A to audio processing unit B, the energy consumption of A decreases from 0.1W to 0.02W.

[0063] According to the energy consumption change trend data, the set of migratable tasks and the corresponding task migration priority are identified to obtain the task migration priority data.

[0064] By analyzing the energy consumption change trend data, the set of migratable tasks and the corresponding task migration priority are identified. Specifically, the task migration priority is determined by sorting the energy consumption change trends under different task migration conditions.

[0065] For example, tasks T1 and T2 can both be migrated from the audio processing unit A, and the migration priority of T1 is higher than that of T2, so the task migration priority data is obtained as {T1, T2}.

[0066] In step S200, candidate high-efficiency audio processing units are determined based on low-load state identification data, low-load state task data and task migration priority data in combination with a preset task migration mapping strategy, and high-efficiency audio processing unit candidate set data is generated based on the candidate high-efficiency audio processing units.

[0067] By analyzing the low-load state identification data, the low-load state task data and the task migration priority data, combined with the preset task migration mapping strategy, the most suitable candidate high-efficiency audio processing unit is determined. Specifically, according to the task migration mapping strategy, the low-load state task data is matched to the candidate high-efficiency audio processing unit to generate the high-efficiency audio processing unit candidate set data.

[0068] For example, task T1 is suitable for migration from audio processing unit A to audio processing units B and C, forming a candidate set {B, C}.

[0069] The task carrying capacity and task matching degree of the candidate set data of the high-efficiency audio processing unit are calculated to obtain the task carrying capacity data and the task execution matching degree data.

[0070] By analyzing the candidate set data of the high-efficiency audio processing unit, its task carrying capacity and task matching degree are calculated. Specifically, by evaluating the processing capacity and load of the candidate high-efficiency audio processing unit, its task carrying capacity is calculated, and by comparing the task execution feature data, the task matching degree is calculated.

[0071] For example, the task carrying capacity of audio processing units B and C are 0.8 and 0.7 respectively, and the task matching degrees are 0.9 and 0.85 respectively. The obtained task carrying capacity data and task matching degree data are {(B, 0.8, 0.9), (C, 0.7,0.85)}.

[0072] According to the task execution matching data, the optimal high-efficiency audio processing unit is selected, and the low-load state task data is mapped to the optimal high-efficiency audio processing unit to obtain the task reconstruction sequence data.

[0073] By analyzing the task execution matching data, the optimal high-performance audio processing unit is selected, and the low-load state task data is mapped for task reconstruction. Specifically, according to the task matching data, the high-performance audio processing unit with the highest matching degree is selected, and the low-load state task data is mapped to the unit to generate task reconstruction sequence data.

[0074] For example, the optimal high-performance audio processing unit for task T1 is B, and task T1 is mapped to audio processing unit B to form task reconstruction sequence data {(T1, B)}.

[0075] The influence of the task reconstruction sequence data on the audio processing unit in the low-load state is calculated, and the energy consumption data of the audio processing unit in the low-load state after task migration is obtained.

[0076] By simulating and analyzing the task reconstruction sequence data, the impact on the energy consumption of the audio processing unit in the low-load state is calculated. Specifically, the task reconstruction sequence data is simulated using an energy consumption simulation tool to calculate the energy consumption data of the audio processing unit in the low-load state after task migration.

[0077] For example, after task T1 is migrated to B, the energy consumption of audio processing unit A is reduced to 0.02 W, and the energy consumption data of A after task migration is obtained to be 0.02 W.

[0078] The optimal sleep mode of the audio processing unit in a low-load state is calculated through the energy consumption data after task migration, and the sleep mode data and sleep time data are obtained.

[0079] By analyzing the energy consumption data of the audio processing unit after task migration in a low-load state, its optimal sleep mode is calculated. Specifically, the energy consumption simulation tool is used to simulate different sleep modes, select the sleep mode with the lowest energy consumption, and set the sleep time according to the task execution characteristic data to generate sleep mode data and sleep time data.

[0080] For example, for audio processing unit A, after simulation analysis, it is determined that its optimal sleep mode is "deep sleep" and the sleep time is "45 minutes".

[0081] Based on the sleep mode data and the sleep time data, a deep sleep target unit is determined, and a sleep parameter corresponding to the deep sleep target unit is calculated.

[0082] By comprehensively analyzing the sleep mode data and sleep time data, the deep sleep target unit is determined and its sleep parameters are calculated. Specifically, the sleep parameters are set according to the optimal sleep mode and sleep time of the audio processing unit in the low-load state to ensure the lowest energy consumption while ensuring the quality of task execution.

[0083] For example, for the audio processing unit A, the sleep parameters include "deep sleep mode" and "45 minutes sleep time".

[0084] Among them, the task carrying capacity and task matching degree of the candidate set data of the high-efficiency audio processing unit are calculated, and the calculation formula for obtaining the task carrying capacity data and the task execution matching degree data is as follows: in, The task-carrying capacity of the high-performance audio processing unit, is the number of candidate high-performance audio processing units, For the The processing capabilities of the candidate audio processing units, For the The load factor of the candidate audio processing units.

[0085] By applying the above formula to the data of the candidate high-performance audio processing units, the task carrying capacity of each unit is calculated. Specifically, the processing capacity of each candidate unit is multiplied by the load factor, and the total task carrying capacity is accumulated.

[0086] For example, the processing capability of candidate high-performance audio processing unit B is 0.8, and the load factor is 0.6, so its task carrying capacity is 0.48.

[0087] in, For task execution matching, is the task intensity coefficient, The processing power required for the current task.

[0088] The task execution matching degree is calculated by the above formula. Specifically, the task carrying capacity is multiplied by the task intensity coefficient, and then divided by the processing capacity required by the current task to obtain the task matching degree.

[0089] For example, if the task intensity coefficient of task T1 is 1.2 and the processing capacity required for the current task is 0.5, the matching degree is 1.15.

[0090] The influence of the task reconstruction sequence data on the audio processing unit in the low-load state is calculated, and the calculation formula for obtaining the energy consumption data of the audio processing unit in the low-load state after task migration is as follows: in, Energy consumption data after task migration. The number of audio processing units in the task reconstruction sequence, For the The power consumption of the migrated audio processing unit, For the The time required for the migration task of each audio processing unit.

[0091] The energy consumption data of the audio processing unit after task migration in the low-load state is calculated by the above formula. Specifically, the power consumption of each audio processing unit after migration is multiplied by the time required for the migration task, and the total energy consumption data after migration is accumulated.

[0092] For example, the power consumption of audio processing unit A is 0.1 W, and the time required for task migration is 30 minutes, so its energy consumption after task migration is 0.05 W.

[0093] In step S300, the sleep parameters are transmitted to the power consumption control module of the end-side AI audio chip, and the power consumption control module is configured to control the start of the low power consumption mode so that the deep sleep target unit enters the sleep state.

[0094] The sleep parameters are transmitted to the power consumption control module of the end-side AI audio chip, and the power consumption control module is configured to start the low power consumption mode. Specifically, according to the setting of the sleep parameters, the configuration of the power consumption control module is adjusted to put the deep sleep target unit into a sleep state.

[0095] For example, the sleep parameters of the audio processing unit A are transmitted to the power consumption control module and set to "deep sleep mode". After startup, A enters the sleep state.

[0096] In the low power consumption mode, the sleep state change sequence of each audio processing unit is recorded to obtain a sleep state sequence.

[0097] By real-time monitoring of the low power consumption mode of the audio processing unit, the sleep state change sequence is recorded. Specifically, the sleep state of each audio processing unit is recorded one by one using a state recording module to generate a sleep state sequence.

[0098] For example, the audio processing unit A is in sleep state, and its sleep state sequence is {"00:00:00", "deep sleep", "00:30:00", "low power consumption"}.

[0099] In step S400, the task scheduling information of the end-side AI audio chip is obtained, a mapping relationship between the task scheduling information and the sleep state sequence is established, and the mapping relationship is analyzed using a decision tree algorithm to determine the correlation between the task scheduling and the sleep timing, and obtain a correlation result.

[0100] A mapping relationship is established by analyzing the task scheduling information of the end-side AI audio chip and the sleep state sequence of the audio processing unit. Specifically, the task scheduling information is obtained through the task scheduling module and associated with the sleep state sequence. The mapping relationship is analyzed using the decision tree algorithm to determine the correlation between task scheduling and sleep timing.

[0101] For example, the correlation between the scheduling information of task T1 and the sleep state "deep sleep" of the audio processing unit A is 0.95, and the obtained correlation result is {T1, A, "deep sleep", 0.95}.

[0102] The task load information in the next cycle is predicted through the task scheduling information, and the wake-up requirement of the deep sleep target unit is determined according to the task load information and the correlation result.

[0103] By analyzing and predicting the task scheduling information, the task load information in the next cycle is estimated, and the awakening requirement of the deep sleep target unit is determined in combination with the correlation result. Specifically, the task scheduling information is modeled using a prediction model to predict the future task load, and the deep sleep target unit is determined in combination with the correlation result.

[0104] For example, it is predicted that the load of task T1 will increase in the next cycle, and the audio processing unit A is currently in a "deep sleep" state and needs to be awakened, and the awakening demand is "high".

[0105] The wake-up timing of each task is calculated based on the wake-up requirement, and the energy consumption threshold of the deep sleep target unit is calculated according to the wake-up timing.

[0106] By quantitatively analyzing the wake-up requirements, the optimal wake-up time for each task is calculated, and the energy consumption threshold of the deep sleep target unit is further calculated. Specifically, according to the task load and wake-up requirements, the wake-up time of each task is calculated in combination with the energy consumption model, and the corresponding energy consumption threshold is determined.

[0107] For example, the best wake-up time for task T1 is “01:00:00”, and the corresponding energy consumption threshold is 0.05W.

[0108] In step S500, a wake-up control instruction is generated based on the wake-up timing and the corresponding energy consumption threshold, and the wake-up control instruction is sent to the deep sleep target unit through the power consumption control module to start the recovery process.

[0109] By analyzing the calculated wake-up timing and the corresponding energy consumption threshold, a corresponding wake-up control instruction is generated and sent to the deep sleep target unit through the power consumption control module to start its recovery process. Specifically, the wake-up control instruction is transmitted to the power consumption control module, and the wake-up recovery operation of the deep sleep target unit is started according to the instruction.

[0110] For example, at "01:00:00", a wake-up control instruction is sent to the audio processing unit A to restore it from the "deep sleep" state.

[0111] Determine whether the current power consumption meets the preset energy consumption threshold. If not, delay the wake-up time.

[0112] By monitoring the current power consumption of the deep sleep target unit, it is determined whether it meets the preset power consumption threshold. Specifically, the power consumption data of the audio processing unit is monitored in real time through the power consumption monitoring module. If the current power consumption does not reach the preset power consumption threshold, the wake-up time is adjusted to delay the wake-up operation.

[0113] For example, if the current power consumption of the audio processing unit A is 0.03 W and does not reach the preset threshold value of 0.05 W, the wake-up timing is delayed to "01:10:00".

[0114] If the preset energy consumption threshold is met, the wake-up procedure of the deep sleep target unit is started to restore the task execution capability of the deep sleep target unit.

[0115] After the energy consumption monitoring module confirms that the current power consumption of the deep sleep target unit reaches or exceeds the preset energy consumption threshold, the wake-up program is started. Specifically, the power consumption control module sends an instruction to start the wake-up program, so that the deep sleep target unit can resume task execution capability.

[0116] For example, at "01:10:00", the current power consumption of the audio processing unit A reaches 0.06W, which meets the preset threshold of 0.05W, and starts its wake-up program to restore the execution capability of task T1.

[0117] In this embodiment, the current energy consumption data and task execution information of each audio processing unit in the end-side AI audio chip are obtained through the power monitoring module and the timestamp recording module, and preprocessed using Z-score standardization to obtain energy consumption preprocessing data and task execution feature data. Based on the energy consumption preprocessing data, the energy consumption load level and task execution intensity of each audio processing unit are calculated, and the task dependency relationship between the audio processing units is constructed to form task topology structure data. Then, the energy consumption load evaluation data is optimized through the task topology structure data, an energy consumption optimization map is constructed, and a topological analysis is performed on it to identify the audio processing unit in a low-load state. The audio processing unit in a low-load state is parsed, and its energy consumption change trend under different task migration conditions is calculated, and the set of migratable tasks and the task migration priority are identified. According to the low-load state identification data, the low-load state task data and the task migration priority data, combined with the preset task migration mapping strategy, the candidate high-efficiency audio processing unit is judged, and the candidate set data of the high-efficiency audio processing unit is generated. By calculating the task carrying capacity and task matching degree of the candidate set of high-efficiency audio processing units, the optimal high-efficiency audio processing unit is selected, and the low-load state task data is mapped to the optimal unit to generate task reconstruction sequence data. The impact of the task reconstruction sequence data on the low-load state audio processing unit is calculated to obtain the energy consumption data after task migration, and the optimal sleep mode is calculated based on this, and the deep sleep target unit and its sleep parameters are determined. The sleep parameters are transmitted to the power consumption control module to make the deep sleep target unit enter the low power consumption mode and record the sleep state sequence. Based on the mapping relationship between task scheduling information and sleep state sequence, the decision tree algorithm is used to determine the correlation between task scheduling and sleep timing, predict the task load information in the next cycle, calculate the wake-up timing and the corresponding energy consumption threshold, generate the wake-up control instruction, and perform recovery control through the power consumption control module. During the wake-up process, if the current power consumption does not reach the preset energy consumption threshold, the wake-up timing is delayed; if the preset energy consumption threshold is met, the wake-up program is started to restore the task execution capability of the deep sleep target unit. In this way, accurate energy consumption management of each audio processing unit in the end-side AI audio chip is achieved, and the energy efficiency and work efficiency of the system are improved.

[0118] Embodiment 3: like Figure 2 As shown, the present application also provides a deep sleep control system 10 of an end-side AI audio chip, including an acquisition unit 11, a mapping unit 12, a configuration unit 13, a calculation unit 14 and a control unit 15.

[0119] The acquisition unit 11 is mainly used to obtain the current energy consumption data and task execution information of all audio processing units in the end-side AI audio chip, build an energy consumption optimization map based on the current energy consumption data and task execution information, perform topological analysis on the energy consumption optimization map, and identify the audio processing units in a low-load state.

[0120] The mapping unit 12 is mainly used to map the audio processing unit tasks in a low-load state to a high-performance audio processing unit according to a preset task migration mapping strategy, obtain a task reconstruction sequence, perform dynamic sleep calculation on the audio processing unit in a low-load state based on the task reconstruction sequence, and obtain a deep sleep target unit and its sleep parameters.

[0121] The configuration unit 13 is mainly used to configure the power consumption control module of the end-side AI audio chip using the sleep parameters of the deep sleep target unit, so that the deep sleep target unit enters a low power consumption mode and records the sleep state sequence.

[0122] The calculation unit 14 is mainly used to perform adaptive wake-up calculation on the deep sleep target unit according to the sleep state sequence and the task scheduling information of the terminal AI audio chip to obtain the wake-up timing and the corresponding energy consumption threshold.

[0123] The control unit 15 is mainly used to perform recovery control on the deep sleep target unit according to the wake-up timing and the corresponding energy consumption threshold, so that the deep sleep target unit re-enters the task execution state.

[0124] In this embodiment, the acquisition unit 11 first acquires the current energy consumption data and task execution information of all audio processing units in the end-side AI audio chip, constructs an energy consumption optimization map, and performs a topological analysis on it to identify the audio processing unit in a low-load state. Then, the mapping unit 12 maps the audio processing unit task in a low-load state to a high-efficiency audio processing unit according to the preset task migration mapping strategy, generates a task reconstruction sequence, performs dynamic sleep calculation, and determines the deep sleep target unit and its sleep parameters. Next, the configuration unit 13 configures the power consumption control module of the end-side AI audio chip using the sleep parameters of the deep sleep target unit, so that these units enter a low-power mode and record their sleep state sequence. The calculation unit 14 performs adaptive wake-up calculations based on the sleep state sequence and the task scheduling information of the end-side AI audio chip to obtain the wake-up timing and energy consumption threshold. Finally, the control unit 15 performs recovery control on the deep sleep target unit based on the calculated wake-up timing and energy consumption threshold, so that it re-enters the task execution state. Through the above system, accurate energy consumption management of the audio processing unit can be achieved, more flexible and efficient task scheduling can be achieved, the energy consumption performance of the system can be optimized, and the working efficiency of the end-side AI audio chip and the battery life of the equipment can be improved.

[0125] It should be noted that technical personnel in the relevant technical field can clearly understand that, for the convenience and conciseness of description, the specific working process of the system and each unit described above can refer to the corresponding process in the aforementioned embodiment of the deep sleep control method of the end-side AI audio chip, and will not be repeated here.

[0126] The structures, proportions, sizes, etc. illustrated in the drawings of this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with this technology. They are not used to limit the conditions under which the present invention can be implemented, and therefore have no substantive technical significance. Any structural modification, change in proportion or adjustment of size, without affecting the effects and purposes that can be achieved by the present invention, should still fall within the scope of the technical contents disclosed by the present invention.

[0127] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A deep sleep control method for a terminal-side AI audio chip, characterized in that: include: Obtain current energy consumption data and task execution information of all audio processing units in the end-side AI audio chip, build an energy consumption optimization map based on the current energy consumption data and task execution information, perform topological analysis on the energy consumption optimization map, and identify audio processing units in a low-load state; According to a preset task migration mapping strategy, the audio processing unit task in the low-load state is mapped to a high-performance audio processing unit to obtain a task reconstruction sequence, and based on the task reconstruction sequence, a dynamic sleep calculation is performed on the audio processing unit in the low-load state to obtain a deep sleep target unit and its sleep parameters; The power consumption control module of the end-side AI audio chip is configured using the sleep parameters of the deep sleep target unit to make the deep sleep target unit enter a low power consumption mode and record a sleep state sequence; According to the sleep state sequence and the task scheduling information of the end-side AI audio chip, adaptively wake up the deep sleep target unit to obtain a wake-up timing and a corresponding energy consumption threshold; The deep sleep target unit is restored and controlled according to the wake-up timing and the corresponding energy consumption threshold, so that the deep sleep target unit re-enters the task execution state.

2. The deep sleep control method of the end-side AI audio chip according to claim 1 is characterized in that: The step of obtaining all current energy consumption data and task execution information in the end-side AI audio chip, constructing an energy consumption optimization map based on the current energy consumption data and task execution information, performing topological analysis on the energy consumption optimization map, and identifying the audio processing unit in a low-load state includes: The current energy consumption data and task execution information of each audio processing unit in the end-side AI audio chip are obtained through the power monitoring module and the timestamp recording module, and the current energy consumption data and the task execution information are preprocessed by using Z-score standardization to obtain energy consumption preprocessing data and task execution feature data; Calculating the energy consumption load level and task execution intensity of each audio processing unit based on the energy consumption preprocessing data to obtain energy consumption load evaluation data; Constructing task dependency relationships between audio processing units according to the task execution feature data to obtain task topology structure data; Optimizing the energy consumption load evaluation data through the task topology data, and constructing an energy consumption optimization map based on the optimized energy consumption load evaluation data; A topological analysis is performed on the energy consumption optimization map to identify audio processing units in a low-load state.

3. The deep sleep control method of the end-side AI audio chip according to claim 2 is characterized in that: After the step of topologically analyzing the energy consumption optimization map to identify the audio processing unit in a low-load state, the method further includes: Parsing the audio processing unit in the low-load state to obtain low-load state identification data and low-load state task data; Calculating the energy consumption change trend of the audio processing unit in the low-load state under different task migration conditions to obtain energy consumption change trend data; A set of migratable tasks and corresponding task migration priorities are identified according to the energy consumption change trend data to obtain task migration priority data.

4. The deep sleep control method of the end-side AI audio chip according to claim 3 is characterized in that: The step of mapping the audio processing unit task in the low-load state to the high-performance audio processing unit according to the preset task migration mapping strategy to obtain a task reconstruction sequence, and performing dynamic sleep calculation on the audio processing unit in the low-load state based on the task reconstruction sequence to obtain a deep sleep target unit and its sleep parameters includes: According to the low-load state identification data, the low-load state task data and the task migration priority data combined with a preset task migration mapping strategy, a candidate high-efficiency audio processing unit is determined, and based on the candidate high-efficiency audio processing unit, a candidate set of high-efficiency audio processing unit data is generated; Calculating the task carrying capacity and task matching degree of the high-efficiency audio processing unit candidate set data to obtain task carrying capacity data and task execution matching degree data; According to the task execution matching degree data, an optimal high-efficiency audio processing unit is selected, and the low-load state task data is mapped to the optimal high-efficiency audio processing unit to obtain task reconstruction sequence data; Calculating the influence of the task reconstruction sequence data on the audio processing unit in the low-load state, and obtaining the energy consumption data of the audio processing unit in the low-load state after task migration; Calculate the optimal sleep mode of the audio processing unit in the low-load state through the energy consumption data after the task migration, and obtain sleep mode data and sleep time data; Based on the sleep mode data and the sleep time data, a deep sleep target unit is determined, and a sleep parameter corresponding to the deep sleep target unit is calculated.

5. The deep sleep control method of the end-side AI audio chip according to claim 4 is characterized in that: The calculation formula for calculating the task carrying capacity and task matching degree of the high-efficiency audio processing unit candidate set data to obtain the task carrying capacity data and the task execution matching degree data is as follows: in, The task-carrying capacity of the high-performance audio processing unit, For the number of candidate high-performance audio processing units, For the The processing capability of each candidate audio processing unit, For the load factor of each candidate audio processing unit; in, Perform matching degree for tasks, For the task intensity coefficient, Processing capabilities required for the current task.

6. The deep sleep control method of the end-side AI audio chip according to claim 4 is characterized in that: The calculation formula for calculating the influence of the task reconstruction sequence data on the audio processing unit in the low-load state to obtain the energy consumption data of the audio processing unit in the low-load state after task migration is as follows: in, It is energy consumption data after task migration. The number of audio processing units in the task reconstruction sequence, For the power consumption of the audio processing unit after migration, For the The time required for the migration task of each audio processing unit.

7. The deep sleep control method of the end-side AI audio chip according to claim 1, characterized in that: The step of configuring the power consumption control module of the end-side AI audio chip by using the sleep parameters to make the deep sleep target unit enter a low power consumption mode and record a sleep state sequence includes: Transmitting the sleep parameters to the power consumption control module of the AI ​​audio chip on the terminal side, and configuring the power consumption control module to control the start of the low power consumption mode so that the deep sleep target unit enters a sleep state; In the low power consumption mode, the sleep state change sequence of each audio processing unit is recorded to obtain a sleep state sequence.

8. The deep sleep control method of the end-side AI audio chip according to claim 1, characterized in that: The step of performing adaptive wake-up calculation on the deep sleep target unit according to the sleep state sequence and the task scheduling information of the end-side AI audio chip to obtain the wake-up timing and the corresponding energy consumption threshold includes: Acquire task scheduling information of the end-side AI audio chip, establish a mapping relationship between the task scheduling information and the sleep state sequence, and use a decision tree algorithm to analyze the mapping relationship to determine the correlation between task scheduling and sleep timing, and obtain a correlation result; Predicting the task load information in the next cycle through the task scheduling information, and determining the wake-up requirement of the deep sleep target unit according to the task load information and the association result; The wake-up timing of each task is calculated based on the wake-up requirement, and the energy consumption threshold of the deep sleep target unit is calculated according to the wake-up timing.

9. The deep sleep control method of the end-side AI audio chip according to claim 1, characterized in that: The step of performing recovery control on the deep sleep target unit according to the wake-up timing and the corresponding energy consumption threshold so that the deep sleep target unit re-enters the task execution state includes: Generate a wake-up control instruction based on the wake-up timing and the corresponding energy consumption threshold, and send the wake-up control instruction to the deep sleep target unit through the power consumption control module to start the recovery process; Determine whether the current power consumption meets the preset energy consumption threshold. If not, delay the wake-up time. If the preset energy consumption threshold is met, a wake-up procedure of the deep sleep target unit is started to restore the task execution capability of the deep sleep target unit.

10. A deep sleep control system for an end-side AI audio chip, characterized in that: include: An acquisition unit, used to acquire current energy consumption data and task execution information of all audio processing units in the end-side AI audio chip, construct an energy consumption optimization map based on the current energy consumption data and task execution information, perform topological analysis on the energy consumption optimization map, and identify audio processing units in a low-load state; A mapping unit, configured to map the audio processing unit task in the low-load state to a high-performance audio processing unit according to a preset task migration mapping strategy, obtain a task reconstruction sequence, and perform dynamic sleep calculation on the audio processing unit in the low-load state based on the task reconstruction sequence to obtain a deep sleep target unit and its sleep parameters; A configuration unit, configured to configure a power consumption control module of the end-side AI audio chip using the sleep parameters of the deep sleep target unit, so that the deep sleep target unit enters a low power consumption mode, and records a sleep state sequence; A calculation unit, configured to perform adaptive wake-up calculation on the deep sleep target unit according to the sleep state sequence and task scheduling information of the end-side AI audio chip, and obtain a wake-up timing and a corresponding energy consumption threshold; The control unit is used to perform recovery control on the deep sleep target unit according to the wake-up timing and the corresponding energy consumption threshold, so that the deep sleep target unit re-enters the task execution state.

Images