A method and system for deep sleep control of an end-side AI audio chip

Abstract

The application relates to the technical field of chip control, and provides a deep sleep control method and system for an end-side AI audio chip, which comprises the following steps: after current energy consumption data and task execution information of all audio processing units in the end-side AI audio chip are acquired, an energy consumption optimization graph is constructed for analysis, the audio processing units in a low load state are identified, tasks are mapped to high-efficiency audio processing units, dynamic sleep calculation is performed on the audio processing units in the low load state after a task reconstruction sequence is obtained, a deep sleep target unit and sleep parameters thereof are obtained, the sleep parameters are used for configuring a power consumption control module of the end-side AI audio chip, and the deep sleep target unit is enabled to enter a low-power mode. Through analysis and processing of the current energy consumption data and the task execution information, flexible and efficient energy consumption management is realized, and the problem that energy consumption optimization is not flexible when task load frequently changes and energy consumption cannot be flexibly controlled is solved.

Description

Technical Field

[0001] This application relates to the technical field of chip control, and in particular to a deep sleep control method and system for an edge AI audio chip. Background Technology

[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, and smart wearable devices. These devices achieve more intelligent audio processing and interaction functions through edge AI audio chips, meeting users' needs for high efficiency and low power consumption.

[0003] Among the relevant technical means, the audio chip sleep control technology mainly reduces 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] While the above technical solutions can reduce the power consumption of the audio processing unit through fixed-cycle sleep and wake-up strategies, they cannot flexibly control energy consumption when the task load changes frequently. Furthermore, the power consumption optimization effect is limited by the sleep cycle setting, resulting in insufficient flexibility in energy consumption optimization. Summary of the Invention

[0005] To address the issue of insufficient flexibility in energy consumption control when task loads change frequently, this application provides a deep sleep control method and system for edge AI audio chips.

[0006] This invention provides a deep sleep control method for an edge AI audio chip, comprising: acquiring current energy consumption data and task execution information of all audio processing units within the edge AI audio chip; constructing an energy consumption optimization map based on the current energy consumption data and task execution information; performing topology analysis on the energy consumption optimization map to identify audio processing units in a low-load state; mapping the tasks of the low-load audio processing units to high-efficiency audio processing units according to a preset task migration mapping strategy to obtain a task reconstruction sequence; performing dynamic sleep calculation on the low-load audio processing units based on the task reconstruction sequence to obtain a deep sleep target unit and its sleep parameters; configuring the power consumption control module of the edge 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 mode and recording the 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 edge AI audio chip to obtain a wake-up timing and a corresponding energy consumption threshold; and 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 embodiment, the steps of acquiring all current energy consumption data and task execution information within the edge AI audio chip, constructing an energy consumption optimization map based on the current energy consumption data and task execution information, performing topology analysis on the energy consumption optimization map, and identifying audio processing units in a low-load state include: acquiring current energy consumption data and task execution information of each audio processing unit in the edge AI audio chip through a power monitoring module and a timestamp recording module; preprocessing the current energy consumption data and task execution information using Z-score normalization to obtain energy consumption preprocessed 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 preprocessed data to obtain energy consumption load assessment data; constructing task dependencies between audio processing units based on the task execution feature data to obtain task topology data; optimizing the energy consumption load assessment data using the task topology data, and constructing an energy consumption optimization map based on the optimized energy consumption load assessment data; and performing topology analysis on the energy consumption optimization map to identify audio processing units in a low-load state.

[0008] As a preferred embodiment, after performing topological analysis on the energy consumption optimization map to identify audio processing units in low-load states, the method further includes: parsing the audio processing units in low-load states to obtain low-load state identification data and low-load state task data; calculating the energy consumption change trend of the audio processing units in low-load states under different task migration conditions to obtain energy consumption change trend data; and identifying a set of migrateable tasks and corresponding task migration priorities based on the energy consumption change trend data to obtain task migration priority data.

[0009] As a preferred embodiment, the step of mapping the low-load audio processing unit tasks to high-performance audio processing units according to a preset task migration mapping strategy to obtain a task reconstruction sequence, and performing dynamic hibernation calculations on the low-load audio processing units based on the task reconstruction sequence to obtain deep hibernation target units and their hibernation parameters, includes: determining candidate high-performance audio processing units based on the low-load state identification data, low-load state task data, and task migration priority data combined with the preset task migration mapping strategy; generating a candidate set of high-performance audio processing units based on the candidate high-performance audio processing units; and calculating the task carrying capacity and task matching degree of the candidate set of high-performance audio processing units. The process involves obtaining task capacity data and task execution matching data; selecting the optimal high-performance audio processing unit based on the task execution matching data, and mapping the low-load state task data to the optimal high-performance audio processing unit to obtain task reconstruction sequence data; calculating the impact of the task reconstruction sequence data on the low-load state audio processing unit to obtain the energy consumption data of the low-load state audio processing unit after task migration; calculating the optimal sleep mode of the low-load state audio processing unit using the energy consumption data after task migration to obtain sleep mode data and sleep time data; and determining the deep sleep target unit based on the sleep mode data and sleep time data, and calculating the sleep parameters corresponding to the deep sleep target unit.

[0010] As a preferred embodiment, the calculation formulas for the task carrying capacity and task matching degree of the candidate set data of the high-performance audio processing unit are as follows:

[0011]

[0012] in, To enhance the task-bearing capacity of the high-performance audio processing unit, The number of candidate high-performance audio processing units. For the first The processing power of each candidate audio processing unit For the first Load factor of each candidate audio processing unit;

[0013]

[0014] in, For task execution matching degree, This is the task intensity coefficient. The processing power required for the current task.

[0015] As a preferred embodiment, the calculation formula for the energy consumption data of the low-load audio processing unit after task migration is as follows: The impact of the task reconstruction sequence data on the low-load audio processing unit is calculated based on the following:

[0016]

[0017] in, Energy consumption data after task migration. The number of audio processing units in the task reconstruction sequence. For the first The power consumption of each migrated audio processing unit. For the first The time required for each audio processing unit to migrate tasks.

[0018] As a preferred embodiment, the step of configuring the power control module of the edge AI audio chip using the sleep parameters to enable the deep sleep target unit to enter a low-power mode and recording the sleep state sequence includes: transmitting the sleep parameters to the power control module of the edge AI audio chip and configuring the power control module to control the start of the low-power mode, so that the deep sleep target unit enters a sleep state; and recording the sleep state change sequence of each audio processing unit in the low-power mode to obtain a sleep state sequence.

[0019] As a preferred embodiment, the step of performing adaptive wake-up calculation on the deep sleep target unit based on the sleep state sequence and the task scheduling information of the edge AI audio chip to obtain the wake-up timing and the corresponding energy consumption threshold includes: acquiring the task scheduling information of the edge AI audio chip, establishing a mapping relationship between the task scheduling information and the sleep state sequence, and analyzing the mapping relationship 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 based on 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 based on the wake-up timing.

[0020] As a preferred embodiment, the step of restoring 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 command based on the wake-up timing and the corresponding energy consumption threshold; sending the wake-up control command to the deep sleep target unit through the power consumption control module to start the recovery process; determining whether the current power consumption meets the preset energy consumption threshold; if not, delaying the wake-up timing; if the preset energy consumption threshold is met, starting the wake-up program of the deep sleep target unit to restore the task execution capability of the deep sleep target unit.

[0021] This application also provides a deep sleep control system for an edge AI audio chip, comprising: an acquisition unit, configured to acquire current energy consumption data and task execution information of all audio processing units within the edge AI audio chip, construct an energy consumption optimization map based on the current energy consumption data and task execution information, perform topology analysis on the energy consumption optimization map, and identify audio processing units in a low-load state; and a mapping unit, configured to map the tasks of the low-load audio processing units to high-efficiency audio processing units according to a preset task migration mapping strategy, obtain a task reconstruction sequence, and perform dynamic sleep calculations on the low-load audio processing units based on the task reconstruction sequence. The system obtains a deep sleep target unit and its sleep parameters; a configuration unit is used to configure the power control module of the edge AI audio chip using the sleep parameters of the deep sleep target unit, so that the deep sleep target unit enters a low power mode and records the 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 edge AI audio chip, so as to obtain the wake-up timing and the corresponding energy consumption threshold; and 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.

[0022] Compared with existing technologies, this application has the following advantages: longer battery life and more flexible energy consumption optimization. By acquiring the current energy consumption data and task execution information of all audio processing units in the edge AI audio chip, an energy consumption optimization map is constructed. Topology analysis is performed to identify audio processing units in low-load states. Based on a preset task migration mapping strategy, these low-load state units are reconfigured for task reconstruction and dynamic sleep calculations, thereby obtaining the target units for deep sleep and their sleep parameters. These sleep parameters are used to configure the power consumption control module, enabling the target units for deep sleep to enter a low-power mode and recording their sleep state sequence. Based on the sleep state sequence and task scheduling information, adaptive wake-up calculations are performed to determine the wake-up timing and energy consumption threshold. The target units for deep sleep are then restored to the task execution state. By dynamically adapting to changes in task load, more flexible and efficient energy consumption management is achieved, effectively extending the device's battery life and improving the problem of insufficient flexibility in energy consumption optimization when task load changes frequently. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] The structures, proportions, sizes, etc., shown in the accompanying drawings of this specification are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed in the specification, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0025] Figure 1 This is a flowchart illustrating a deep sleep control method for an edge AI audio chip provided in an embodiment of the present invention;

[0026] Figure 2 This is a schematic block diagram of the structure of a deep sleep control system for an edge AI audio chip provided in an embodiment of the present invention.

[0027] Explanation of reference numerals in the attached figures:

[0028] 10. Deep sleep control system for edge AI audio chip; 11. Acquisition unit; 12. Mapping unit; 13. Configuration unit; 14. Calculation unit; 15. Control unit. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0031] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0032] It should also be further understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0033] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0034] Example 1:

[0035] like Figure 1 As shown, this application provides a deep sleep control method for an edge AI audio chip, including steps S100 to S500.

[0036] Step S100: Obtain the current energy consumption data and task execution information of all audio processing units in the edge AI audio chip. Based on the current energy consumption data and task execution information, construct an energy consumption optimization map, perform topology analysis on the energy consumption optimization map, and identify audio processing units in a low-load state.

[0037] In this step, current energy consumption data and task execution information of all audio processing units within the edge AI audio chip are collected, and an energy consumption optimization map is constructed using data analysis tools. Specifically, graph theory algorithms are used to perform topological analysis on the energy consumption optimization map to identify audio processing units currently in a low-load state and mark them as potential deep sleep target units.

[0038] For example, at a specific moment, the power consumption data of audio processing units A, B, and C are 0.1W, 0.05W, and 0.02W, respectively, while the power consumption data of audio processing units D and E are 0.4W and 0.35W, respectively. Then, after graph analysis, A, B, and C are identified as low-load state units.

[0039] Step S200: Map the low-load audio processing unit tasks to the high-performance audio processing unit according to the preset task migration mapping strategy to obtain the task reconstruction sequence. Perform dynamic sleep calculation on the low-load audio processing unit based on the task reconstruction sequence to obtain the deep sleep target unit and its sleep parameters.

[0040] In this step, a pre-defined task migration mapping strategy is used to reallocate tasks from low-load audio processing units to high-performance audio processing units. Specifically, a task reconstruction sequence is generated using a task migration mapping algorithm, and dynamic sleep calculations are performed on low-load audio processing units based on this reconstruction sequence to determine the target unit for deep sleep and its corresponding sleep parameters.

[0041] For example, suppose that after task T1 of audio processing unit A is mapped to audio processing unit D, the power consumption of D rises to 0.45W, and A becomes the target unit for deep sleep, with its sleep parameter set to a sleep cycle of 30 minutes.

[0042] Step S300: Configure the power control module of the edge AI audio chip using the sleep parameters of the deep sleep target unit, so that the deep sleep target unit enters a low power mode and records the sleep state sequence.

[0043] In this step, the sleep parameters of the deep sleep target units are used to configure the power control module of the edge AI audio chip. Specifically, the sleep parameters of each deep sleep target unit are input into the power control module to put it into a low-power mode, and its sleep state sequence is recorded.

[0044] For example, after inputting the sleep parameters of audio processing unit A into the power consumption control module, A enters a low-power mode, its power consumption drops to 0.01W, and the sleep state sequence of A is recorded as "30-minute sleep".

[0045] Step S400: Based on the sleep state sequence and the task scheduling information of the edge AI audio chip, perform adaptive wake-up calculation on the deep sleep target unit to obtain the wake-up timing and the corresponding energy consumption threshold.

[0046] 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 edge AI audio chip. Specifically, by analyzing the task execution priority in the task scheduling information and the current power consumption state of the deep sleep target unit, an appropriate wake-up time and corresponding power consumption threshold are determined.

[0047] For example, if audio processing unit A has its task execution priority increased 5 minutes before the end of its sleep cycle, and its current power consumption is close to the preset threshold of 0.01W, then the wake-up time is set to the end of the sleep cycle, and the power consumption threshold is 0.02W.

[0048] Step S500: Perform recovery control on the deep sleep target unit according to the wake-up time and the corresponding energy consumption threshold, so that the deep sleep target unit re-enters the task execution state.

[0049] In this step, recovery control is performed on the target unit in deep sleep based on its wake-up timing and power consumption threshold. Specifically, when the preset wake-up timing and power consumption threshold are reached, the low-power mode of the target unit in deep sleep is deactivated, allowing it to re-enter the task execution state.

[0050] For example, when the sleep cycle of audio processing unit A ends and the current power consumption reaches 0.02W, A is restored to resume task T1.

[0051] In this embodiment, current energy consumption data and task execution information of all audio processing units within the edge AI audio chip are acquired, and an energy consumption optimization map is constructed based on this data. Then, topology analysis is performed on the energy consumption optimization map to identify audio processing units in a low-load state. According to a preset task migration mapping strategy, the tasks of low-load audio processing units are mapped to high-efficiency audio processing units, resulting in a task reconstruction sequence. Based on the task reconstruction sequence, dynamic sleep calculations are performed on the low-load audio processing units to obtain deep sleep target units and their sleep parameters. Finally, the sleep parameters of the deep sleep target units are used to configure the power consumption control module of the edge AI audio chip, enabling the deep sleep target units to enter a low-power mode and recording the sleep state sequence. Simultaneously, based on the sleep state sequence and the task scheduling information of the edge AI audio chip, adaptive wake-up calculations are performed on the deep sleep target units to obtain the wake-up timing and corresponding energy consumption threshold, thereby achieving recovery control of the deep sleep target units. This achieves accurate identification and dynamic sleep of low-load audio processing units, enabling more flexible and efficient energy consumption management. Compared to fixed-cycle sleep and wake-up strategies, this solution provides better power consumption control when task loads change frequently, thereby extending device battery life, reducing overall energy consumption, improving the working efficiency of edge AI audio chips, and addressing the problem of insufficient flexibility in energy consumption control when task loads change frequently.

[0052] Example 2:

[0053] In step S100, the current energy consumption data and task execution information of each audio processing unit in the edge AI audio chip are obtained through the power monitoring module and the timestamp recording module. The current energy consumption data and task execution information are preprocessed using Z-score standardization to obtain energy consumption preprocessed data and task execution feature data.

[0054] The power monitoring module monitors the power consumption of the audio processing unit in real time, and combines this with the time information from the timestamp recording module to collect current energy consumption data and task execution information for all audio processing units in the edge AI audio chip. Specifically, this data is Z-score standardized to eliminate energy consumption differences between different processing units and the influence of the unit of measurement on task execution characteristic data, ensuring data consistency and comparability.

[0055] For example, the current energy consumption data of audio processing unit A is 0.1W, the task execution information is 20 times / second, and after standardization, the energy consumption preprocessing data is -0.5, and the task execution feature data is 1.5.

[0056] Based on the energy consumption preprocessing data, the energy consumption load level and task execution intensity of each audio processing unit are calculated to obtain energy consumption load assessment data.

[0057] By analyzing energy consumption preprocessing data, the energy consumption load level of each audio processing unit is calculated, and the task execution intensity is calculated in conjunction with task execution characteristic data. Specifically, a weighted average algorithm is used to process the energy consumption preprocessing data 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 conjunction with task execution characteristic data.

[0058] For example, the power consumption load level of audio processing unit A is 0.2, the task execution intensity is 0.8, and the resulting power consumption load assessment data is (0.2, 0.8).

[0059] Based on the task execution characteristic data, the task dependency relationship between audio processing units is constructed to obtain the task topology data.

[0060] By analyzing task execution feature data, task dependencies between audio processing units are established. Specifically, correlation analysis is used to process the task execution feature data, identify highly correlated task dependencies, and construct task topology data based on these dependencies.

[0061] For example, the task execution correlation between audio processing units A and B is 0.85, thus establishing a task dependency relationship between A and B.

[0062] The energy consumption load assessment data is optimized using task topology data, and an energy consumption optimization map is constructed based on the optimized energy consumption load assessment data.

[0063] The energy load assessment data is optimized by combining task topology data with energy load assessment data. Specifically, a graph optimization algorithm is used to optimize the task topology data to obtain optimized energy load assessment data, and an energy optimization graph is constructed based on this data.

[0064] For example, the optimized energy consumption load assessment data for audio processing units A and B are 0.15 and 0.75, respectively. When constructing the energy consumption optimization map, A and B are connected to represent the task dependency relationship.

[0065] Topological analysis of the energy consumption optimization map identifies audio processing units operating under low load conditions.

[0066] By performing topological analysis on the constructed energy consumption optimization graph, audio processing units in a low-load state were identified. Specifically, a graph traversal algorithm was used to analyze the energy consumption optimization graph node by node to identify audio processing units with both low energy consumption load levels and low task execution intensity.

[0067] For example, audio processing unit A has a power consumption load level and task execution intensity of 0.1 and 0.2, respectively, and is identified as an audio processing unit in a low-load state.

[0068] The process includes, after performing topological analysis on the energy consumption optimization map to identify audio processing units in low-load states, parsing the audio processing units in low-load states to obtain low-load state identification data and low-load state task data.

[0069] By performing a detailed analysis of the audio processing units in a low-load state, their low-load state identification data and low-load state task data are extracted. Specifically, by analyzing each audio processing unit in a low-load state individually, their current task execution status and energy consumption data are recorded to form low-load state identification data, and their corresponding task information is recorded as low-load state task data.

[0070] For example, the low load status identifier data of audio processing unit A is "low load", and the task data is "task T1".

[0071] The energy consumption trend of the audio processing unit under low load conditions is calculated under different task migration conditions to obtain energy consumption trend data.

[0072] The energy consumption trend of the audio processing unit under low load conditions was calculated by simulation analysis under different task migration conditions. Specifically, energy consumption simulation tools were used to simulate the energy consumption of the audio processing unit under different task migration conditions to obtain energy consumption trend data.

[0073] For example, when task T1 is moved from audio processing unit A to audio processing unit B, the power consumption of A decreases from 0.1W to 0.02W.

[0074] Based on energy consumption change trend data, a set of transferable tasks and corresponding task migration priorities are identified, and task migration priority data is obtained.

[0075] By analyzing energy consumption trend data, a set of transferable tasks and their corresponding migration priorities are identified. Specifically, the migration priorities are determined by ranking the energy consumption trends under different task migration conditions.

[0076] For example, both tasks T1 and T2 can be migrated from audio processing unit A, but T1 has a higher migration priority than T2, resulting in task migration priority data of {T1, T2}.

[0077] In step S200, candidate high-performance audio processing units are identified based on low-load status identification data, low-load status task data, and task migration priority data, combined with a preset task migration mapping strategy. Based on the candidate high-performance audio processing units, a candidate set of high-performance audio processing units is generated.

[0078] By analyzing low-load state identification data, low-load state task data, and task migration priority data, and combining this with a preset task migration mapping strategy, the most suitable candidate high-performance audio processing units are determined. Specifically, according to the task migration mapping strategy, low-load state task data is matched to candidate high-performance audio processing units, generating a candidate set of high-performance audio processing units.

[0079] 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}.

[0080] Calculate the task carrying capacity and task matching degree of the candidate set data of high-performance audio processing units to obtain task carrying capacity data and task execution matching degree data.

[0081] By analyzing the candidate set of high-performance audio processing units, their task capacity and task matching degree are calculated. Specifically, the task capacity is calculated by evaluating the processing power and load of the candidate high-performance audio processing units, and the task matching degree is calculated by comparing task execution characteristic data.

[0082] For example, the task capacity of audio processing units B and C are 0.8 and 0.7 respectively, and the task matching degree is 0.9 and 0.85 respectively. The obtained task capacity data and task matching degree data are {(B, 0.8, 0.9), (C, 0.7, 0.85)}.

[0083] Based on the task execution matching data, the optimal high-performance audio processing unit is selected, and the low-load state task data is mapped to the optimal high-performance audio processing unit to obtain the task reconstruction sequence data.

[0084] By analyzing task execution matching data, the optimal high-performance audio processing unit is selected, and low-load task data is reconstructed and mapped. Specifically, based on the task matching data, the high-performance audio processing unit with the highest matching degree is selected, and the low-load task data is mapped to that unit to generate task reconstruction sequence data.

[0085] For example, if the optimal high-performance audio processing unit for task T1 is B, then task T1 is mapped to audio processing unit B to form task reconstruction sequence data {(T1, B)}.

[0086] The impact of the task reconstruction sequence data on the audio processing unit under low load is calculated, and the energy consumption data of the audio processing unit after task migration under low load is obtained.

[0087] The impact of task reconstruction sequence data on the energy consumption of the audio processing unit under low load was calculated through simulation analysis. Specifically, energy consumption simulation tools were used to simulate the task reconstruction sequence data and calculate the energy consumption data of the audio processing unit under low load after task migration.

[0088] For example, after task T1 is migrated to B, the power consumption of audio processing unit A drops to 0.02W, and the power consumption data of A after task migration is 0.02W.

[0089] The optimal sleep mode for the audio processing unit under low load is calculated by using energy consumption data after task migration, and sleep mode data and sleep time data are obtained.

[0090] By analyzing the energy consumption data of the audio processing unit after task migration under low load conditions, its optimal sleep mode is calculated. Specifically, different sleep modes are simulated using energy consumption simulation tools, the sleep mode with the lowest energy consumption is selected, and the sleep time is set according to the task execution characteristic data to generate sleep mode data and sleep time data.

[0091] For example, after simulation analysis, the optimal sleep mode for audio processing unit A is determined to be "deep sleep" with a sleep time of "45 minutes".

[0092] Based on hibernation mode data and hibernation time data, the target unit for deep hibernation is determined, and the corresponding hibernation parameters for the target unit are calculated.

[0093] By comprehensively analyzing sleep mode data and sleep time data, the target unit for deep sleep was identified, and its sleep parameters were calculated. Specifically, based on the optimal sleep mode and sleep time of the audio processing unit under low load, sleep parameters were set to ensure the lowest power consumption while maintaining task execution quality.

[0094] For example, for audio processing unit A, the sleep parameters include "deep sleep mode" and "45-minute sleep time".

[0095] The calculation formulas for the task carrying capacity and task matching degree of the candidate set data of high-performance audio processing units are as follows:

[0096]

[0097] in, To enhance the task-bearing capacity of the high-performance audio processing unit, The number of candidate high-performance audio processing units. For the first The processing power of each candidate audio processing unit For the first Load factor of each candidate audio processing unit.

[0098] By applying the above formula to the data of the candidate high-performance audio processing units, the task capacity of each unit is calculated. Specifically, the processing capacity of each candidate unit is multiplied by a load factor, and the results are summed to obtain the total task capacity.

[0099] For example, if the processing power of candidate high-performance audio processing unit B is 0.8 and the load factor is 0.6, then its task carrying capacity is 0.48.

[0100]

[0101] in, For task execution matching degree, This is the task intensity coefficient. The processing power required for the current task.

[0102] The task execution matching degree is calculated using the formula described above. 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.

[0103] For example, if the task intensity coefficient of task T1 is 1.2 and the processing power required for the current task is 0.5, then the matching degree is 1.15.

[0104] The impact of computational task reconstruction sequence data on the audio processing unit under low load is calculated, and the formula for calculating the energy consumption data of the audio processing unit after task migration under low load is as follows:

[0105]

[0106] in, Energy consumption data after task migration. The number of audio processing units in the task reconstruction sequence. For the first The power consumption of each migrated audio processing unit. For the first The time required for each audio processing unit to migrate tasks.

[0107] The energy consumption data of the audio processing unit after task migration under low load is calculated using 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 results are summed to obtain the total energy consumption data after migration.

[0108] For example, if the power consumption of audio processing unit A is 0.1W and the time required for task migration is 30 minutes, then its energy consumption after task migration is 0.05W.

[0109] In step S300, the sleep parameters are transmitted to the power control module of the edge AI audio chip, and the power control module is configured to control the start of the low power mode, so that the deep sleep target unit enters the sleep state.

[0110] By transmitting sleep parameters to the power control module of the edge AI audio chip, the power control module is configured to initiate a low-power mode. Specifically, based on the sleep parameter settings, the configuration of the power control module is adjusted to put the target unit into a deep sleep state.

[0111] For example, the sleep parameters of audio processing unit A are transmitted to the power consumption control module and set to "deep sleep mode". After startup, A enters sleep state.

[0112] In low-power mode, the sleep state change sequence of each audio processing unit is recorded to obtain the sleep state sequence.

[0113] The low-power mode of the audio processing unit is monitored in real time, and the sequence of changes in its sleep state is recorded. Specifically, a state recording module is used to record the sleep state of each audio processing unit one by one, generating a sleep state sequence.

[0114] For example, when the audio processing unit A is in sleep mode, its sleep state sequence is {"00:00:00", "deep sleep", "00:30:00", "low power"}.

[0115] In step S400, task scheduling information of the edge AI audio chip is obtained, a mapping relationship between task scheduling information and sleep state sequence is established, and the mapping relationship is analyzed using decision tree algorithm to determine the correlation between task scheduling and sleep timing, and obtain the correlation result.

[0116] A mapping relationship is established by analyzing the task scheduling information of the edge AI audio chip and the sleep state sequence of the audio processing unit. Specifically, task scheduling information is obtained through the task scheduling module and associated with the sleep state sequence. The mapping relationship is analyzed using a decision tree algorithm to determine the correlation between task scheduling and sleep timing.

[0117] For example, the correlation between the scheduling information of task T1 and the sleep state "deep sleep" of audio processing unit A is 0.95, and the resulting correlation is {T1, A, "deep sleep", 0.95}.

[0118] The task load information for the next cycle is predicted by the task scheduling information, and the wake-up requirements of the deep sleep target unit are determined based on the task load information and the correlation results.

[0119] By analyzing and predicting task scheduling information, the task load information for the next cycle is estimated, and the wake-up requirement of the deep sleep target unit is determined based on the correlation results. Specifically, a predictive model is used to model the task scheduling information, predict future task load, and determine whether the deep sleep target unit needs to be woken up based on the correlation results.

[0120] For example, it is predicted that the load of task T1 will increase in the next cycle, while audio processing unit A is currently in a "deep sleep" state and needs to be woken up, with a wake-up requirement of "high".

[0121] The wake-up timing for each task is calculated based on the wake-up requirements, and the energy consumption threshold for the deep sleep target unit is calculated based on the wake-up timing.

[0122] By quantitatively analyzing wake-up demands, the optimal wake-up time for each task is calculated, and the energy consumption threshold for the target unit in deep sleep is further calculated. Specifically, based on task load and wake-up demands, the wake-up time for each task is calculated using an energy consumption model, and the corresponding energy consumption threshold is determined.

[0123] For example, the optimal wake-up time for task T1 is "01:00:00", with a corresponding energy consumption threshold of 0.05W.

[0124] In step S500, a wake-up control command is generated based on the wake-up timing and the corresponding energy consumption threshold. The wake-up control command is then sent to the deep sleep target unit through the power consumption control module to start the recovery process.

[0125] By analyzing and calculating the wake-up timing and corresponding energy consumption threshold, a corresponding wake-up control command is generated and sent to the deep sleep target unit through the power consumption control module to initiate its recovery process. Specifically, the wake-up control command is transmitted to the power consumption control module, and the wake-up recovery operation of the deep sleep target unit is initiated according to the command.

[0126] For example, at "01:00:00", a wake-up control command is sent to audio processing unit A to wake it up from "deep sleep" state.

[0127] Determine whether the current power consumption meets the preset power consumption threshold. If not, delay the wake-up time.

[0128] By monitoring the current power consumption of the target unit in deep sleep mode, it is determined whether it meets the preset power consumption threshold. Specifically, the power consumption monitoring module monitors the power consumption data of the audio processing unit in real time. If the current power consumption does not reach the preset power consumption threshold, the wake-up timing is adjusted, and the wake-up operation is delayed.

[0129] For example, if the current power consumption of audio processing unit A is 0.03W, which is less than the preset threshold of 0.05W, then the wake-up time will be delayed to "01:10:00".

[0130] If the preset energy consumption threshold is met, the wake-up procedure of the deep sleep target unit is initiated to restore the task execution capability of the deep sleep target unit.

[0131] Once the power consumption monitoring module confirms that the current power consumption of the target unit in deep sleep has reached or exceeded the preset power consumption threshold, its wake-up procedure is initiated. Specifically, the power consumption control module sends a command to initiate the wake-up procedure, enabling the target unit in deep sleep to resume its task execution capability.

[0132] For example, when the audio processing unit A is at "01:10:00", its current power consumption reaches 0.06W, which meets the preset threshold of 0.05W. It then starts its wake-up program to restore the execution capability of task T1.

[0133] In this embodiment, the current energy consumption data and task execution information of each audio processing unit in the edge AI audio chip are obtained through a power monitoring module and a timestamp recording module. Z-score normalization is used for preprocessing to obtain energy consumption preprocessed data and task execution feature data. Based on the energy consumption preprocessed data, the energy load level and task execution intensity of each audio processing unit are calculated, and the task dependency relationship between audio processing units is constructed to form task topology data. Then, the energy load assessment data is optimized using the task topology data to construct an energy consumption optimization map, and topology analysis is performed to identify audio processing units in low-load states. The low-load audio processing units are analyzed to calculate their energy consumption change trends under different task migration conditions, identifying the set of migrated tasks and task migration priorities. Based on the low-load state identification data, low-load state task data, and task migration priority data, combined with a preset task migration mapping strategy, candidate high-efficiency audio processing units are determined, and a candidate set of high-efficiency audio processing units is generated. By calculating the task capacity and task matching degree of the candidate set of high-performance audio processing units, the optimal high-performance audio processing unit is selected, and low-load state task data is mapped to the optimal unit to generate task reconstruction sequence data. The impact of task reconstruction sequence data on low-load state audio processing units is calculated to obtain energy consumption data after task migration, and the optimal sleep mode is calculated accordingly to determine the deep sleep target unit and its sleep parameters. The sleep parameters are transmitted to the power control module to enable the deep sleep target unit to enter a low-power mode and record the sleep state sequence. Based on the mapping relationship between task scheduling information and sleep state sequence, the correlation between task scheduling and sleep timing is determined using a decision tree algorithm to predict the task load information in the next cycle, calculate the wake-up timing and corresponding energy consumption threshold, generate wake-up control commands, and perform recovery control through the power 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 enable the deep sleep target unit to restore its task execution capability. In this way, precise energy consumption management of each audio processing unit in the edge AI audio chip is achieved, improving the system's energy efficiency and working efficiency.

[0134] Example 3:

[0135] like Figure 2 As shown, this application also provides a deep sleep control system 10 for an edge 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.

[0136] The acquisition unit 11 is mainly used to acquire the current energy consumption data and task execution information of all audio processing units in the edge AI audio chip. Based on the current energy consumption data and task execution information, it constructs an energy consumption optimization map, performs topology analysis on the energy consumption optimization map, and identifies audio processing units in a low-load state.

[0137] The mapping unit 12 is mainly used to map the low-load audio processing unit tasks to the high-performance audio processing unit according to the preset task migration mapping strategy, obtain the task reconstruction sequence, and perform dynamic sleep calculation on the low-load audio processing unit based on the task reconstruction sequence to obtain the deep sleep target unit and its sleep parameters.

[0138] The configuration unit 13 is mainly used to configure the power control module of the edge AI audio chip using the sleep parameters of the deep sleep target unit, so that the deep sleep target unit enters a low power mode and records the sleep state sequence.

[0139] The computing unit 14 is mainly used to perform adaptive wake-up calculations on the deep sleep target unit based on the sleep state sequence and the task scheduling information of the edge AI audio chip, so as to obtain the wake-up time and the corresponding energy consumption threshold.

[0140] The control unit 15 is mainly used to control the deep sleep target unit to resume operation based on the wake-up time and the corresponding energy consumption threshold, so that the deep sleep target unit can re-enter the task execution state.

[0141] In this embodiment, the acquisition unit 11 first acquires the current energy consumption data and task execution information of all audio processing units within the edge AI audio chip, constructs an energy consumption optimization map, and performs topology analysis to identify audio processing units in a low-load state. Then, the mapping unit 12 maps the tasks of low-load audio processing units to high-efficiency audio processing units according to a preset task migration mapping strategy, generates a task reconstruction sequence, performs dynamic sleep calculations, and determines the deep sleep target unit and its sleep parameters. Next, the configuration unit 13 configures the power consumption control module of the edge AI audio chip using the sleep parameters of the deep sleep target unit, enabling these units to enter a low-power mode and recording 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 edge 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 according to the calculated wake-up timing and energy consumption threshold, causing it to re-enter the task execution state. The above system enables precise energy management of the audio processing unit, more flexible and efficient task scheduling, optimized system energy consumption performance, and improved working efficiency of edge AI audio chips and device battery life.

[0142] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the system and each unit described above can be referred to the corresponding process in the aforementioned embodiment of a deep sleep control method for an edge AI audio chip, and will not be repeated here.

[0143] The structures, proportions, sizes, etc., shown in the accompanying drawings of this specification are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed in the specification, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0144] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A deep sleep control method for an edge-side AI audio chip, characterized in that, include: The power monitoring module and the timestamp recording module acquire the current energy consumption data and task execution information of each audio processing unit in the edge AI audio chip. The current energy consumption data and the task execution information are preprocessed using Z-score standardization to obtain energy consumption preprocessed 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 to obtain energy consumption load assessment data; based on the task execution characteristic data, the task dependency relationship between audio processing units is constructed to obtain task topology data; the energy consumption load assessment data is optimized using the task topology data, and an energy consumption optimization map is constructed based on the optimized energy consumption load assessment data; topology analysis is performed on the energy consumption optimization map to identify audio processing units in a low-load state; According to the preset task migration mapping strategy, the low-load audio processing unit task is mapped to the high-performance audio processing unit to obtain the task reconstruction sequence. Based on the task reconstruction sequence, the low-load audio processing unit is dynamically sleep-inducing to obtain the deep sleep target unit and its sleep parameters. The power consumption control module of the edge 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. Based on the sleep state sequence and the task scheduling information of the edge 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. The deep sleep target unit is restored according to the wake-up time and the corresponding energy consumption threshold, so that the deep sleep target unit can re-enter the task execution state.

2. The deep sleep control method for an edge-side AI audio chip according to claim 1, characterized in that, After the step of performing topological analysis on the energy consumption optimization map to identify audio processing units in low-load states, the following steps are included: The audio processing unit in the low-load state is parsed to obtain low-load state identification data and low-load state task data. Calculate the energy consumption trend of the audio processing unit under low load conditions under different task migration conditions to obtain energy consumption trend data; Based on the energy consumption change trend data, the set of transferable tasks and the corresponding task migration priorities are identified, and task migration priority data is obtained.

3. The deep sleep control method for an edge-side AI audio chip according to claim 2, characterized in that, The steps of mapping the low-load audio processing unit tasks to high-performance audio processing units according to a preset task migration mapping strategy to obtain a task reconstruction sequence, and performing dynamic sleep calculations on the low-load audio processing units based on the task reconstruction sequence to obtain the deep sleep target unit and its sleep parameters include: Based on the low load status identification data, low load status task data, and task migration priority data, combined with a preset task migration mapping strategy, candidate high-performance audio processing units are identified, and a candidate set of high-performance audio processing units is generated based on the candidate high-performance audio processing units. Calculate the task carrying capacity and task matching degree of the candidate set data of the high-performance audio processing unit to obtain task carrying capacity data and task execution matching degree data; Based on the task execution matching data, the optimal high-performance audio processing unit is selected, and the low-load state task data is mapped to the optimal high-performance 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 after task migration in the low-load state. The optimal sleep mode of the low-load audio processing unit is calculated using the energy consumption data after the task migration, and sleep mode data and sleep time data are obtained. Based on the hibernation mode data and hibernation time data, a deep hibernation target unit is determined, and the hibernation parameters corresponding to the deep hibernation target unit are calculated.

4. The deep sleep control method for an edge-side AI audio chip according to claim 3, characterized in that, The calculation formulas for the task carrying capacity and task matching degree of the candidate set data of the high-performance audio processing unit are as follows: in, To enhance the task-bearing capacity of the high-performance audio processing unit, The number of candidate high-performance audio processing units. For the first The processing power of each candidate audio processing unit For the first Load factor of each candidate audio processing unit; in, For task execution matching degree, This is the task intensity coefficient. The processing power required for the current task.

5. The deep sleep control method for an edge-side AI audio chip according to claim 3, characterized in that, The calculation of the impact of the task reconstruction sequence data on the low-load audio processing unit yields the following formula for calculating the energy consumption data of the low-load audio processing unit after task migration: in, Energy consumption data after task migration. The number of audio processing units in the task reconstruction sequence. For the first The power consumption of each migrated audio processing unit. For the first The time required for each audio processing unit to migrate tasks.

6. The deep sleep control method for an edge-side AI audio chip according to claim 1, characterized in that, The step of configuring the power control module of the edge AI audio chip using the sleep parameters to enable the deep sleep target unit to enter a low-power mode and recording the sleep state sequence includes: The sleep parameters are transmitted to the power control module of the edge AI audio chip, and the power control module is configured to control the start of the low power mode, so that the deep sleep target unit enters the sleep state. In the low-power mode, the sleep state change sequence of each audio processing unit is recorded to obtain the sleep state sequence.

7. The deep sleep control method for an edge-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 based on the sleep state sequence and the task scheduling information of the edge AI audio chip to obtain the wake-up timing and the corresponding energy consumption threshold includes: Obtain task scheduling information from the edge 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 the correlation result. The task load information in the next cycle is predicted by the task scheduling information, and the wake-up requirement of the deep sleep target unit is determined based on the task load information and the correlation result. The wake-up time for each task is calculated based on the wake-up requirements, and the energy consumption threshold of the deep sleep target unit is calculated based on the wake-up time.

8. The deep sleep control method for an edge-side AI audio chip according to claim 1, characterized in that, The step of restoring 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: Based on the wake-up timing and the corresponding energy consumption threshold, a wake-up control command is generated, and the wake-up control command is sent 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, the wake-up procedure of the deep sleep target unit is initiated to restore the task execution capability of the deep sleep target unit.

9. A deep sleep control system for an edge-side AI audio chip, characterized in that, include: The acquisition unit is used to acquire the current energy consumption data and task execution information of each audio processing unit in the edge AI audio chip through the power monitoring module and the timestamp recording module, and to preprocess the current energy consumption data and the task execution information using Z-score standardization to obtain energy consumption preprocessed 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 to obtain energy consumption load assessment data; based on the task execution characteristic data, the task dependency relationship between audio processing units is constructed to obtain task topology data; the energy consumption load assessment data is optimized using the task topology data, and an energy consumption optimization map is constructed based on the optimized energy consumption load assessment data; topology analysis is performed on the energy consumption optimization map to identify audio processing units in a low-load state; The mapping unit is used to map the low-load audio processing unit tasks to the high-performance audio processing unit according to a preset task migration mapping strategy, to obtain a task reconstruction sequence, and to perform dynamic sleep calculation on the low-load audio processing unit based on the task reconstruction sequence to obtain the deep sleep target unit and its sleep parameters. The configuration unit is used to configure the power control module of the edge AI audio chip using the sleep parameters of the deep sleep target unit, so that the deep sleep target unit enters a low power mode and records the sleep state sequence. The computing 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 edge AI audio chip, so as to obtain the wake-up time and the corresponding energy consumption threshold. The control unit is used to perform recovery control on the deep sleep target unit according to the wake-up time and the corresponding energy consumption threshold, so that the deep sleep target unit can re-enter the task execution state.

Images