Electronic component and operation control method of electronic component

Abstract

The present application discloses an electronic component and an operation control method of the electronic component, which relate to the field of computer hardware technology. The electronic component includes a control component and multiple computing units, and the multiple computing units are divided into a computing domain cluster; in the computing domain cluster, different computing domains include different computing units; the control component is used to monitor a set of state parameters of the activated computing domains when there are K activated computing domains in the computing domain cluster; when there is a first computing domain among the K activated computing domains, in which the parameter value of a specified state parameter in a set of state parameters meets the corresponding threshold condition, the operation parameters of the first computing domain are adjusted according to the adjustment method corresponding to the specified state parameter, so that the operation parameters of the computing units can be adjusted by region, which solves the problem of poor timeliness of operation control of electronic components in related technologies.

Description

Technical Field

[0001] The present application relates to the field of computer hardware technology, and in particular to an electronic component and an operation control method of the electronic component. Background Art

[0002] To improve the efficiency of computationally intensive tasks, such as those in the field of artificial intelligence (AI), electronic components can be used to accelerate them. However, high-performance computing often comes with high power consumption and heat dissipation, especially in mobile devices and edge computing scenarios. Therefore, the operating parameters of electronic components need to be adjusted.

[0003] Related technologies for adjusting the operating parameters of electronic components primarily rely on dynamic voltage and frequency scaling (DVFS) and clock gating. However, these electronic components typically rely on static strategies or simple feedback control, resulting in poor timeliness in operational control. Summary of the Invention

[0004] The present application provides an electronic component and an operation control method of the electronic component, so as to at least solve the problem of poor timeliness of operation control of electronic components in the related art.

[0005] The present application provides an electronic component, including a control component and multiple computing units, wherein the multiple computing units are divided into a computing domain cluster, and the computing domains in the computing domain cluster are a set of computing units that independently adjust operating parameters; in the computing domain cluster, one computing domain includes some computing units in the multiple computing units, and different computing domains include different computing units; wherein the control component is used to monitor a set of state parameters of the activated computing domains when there are K activated computing domains in the computing domain cluster, wherein K is a positive integer greater than or equal to 1; when there is a first computing domain among the K activated computing domains, in which the parameter value of a specified state parameter in the set of state parameters meets the corresponding threshold condition, the operating parameters of the first computing domain are adjusted according to the adjustment method corresponding to the specified state parameter.

[0006] The present application also provides an operation control method for an electronic component, wherein the electronic component includes multiple computing units, and the multiple computing units are divided into a computing domain cluster, and the computing domains in the computing domain cluster are a set of computing units that independently adjust operating parameters; in the computing domain cluster, one computing domain includes some computing units in the multiple computing units, and different computing domains include different computing units; the method includes: when there are K activated computing domains in the computing domain cluster, monitoring a set of state parameters of the activated computing domains, wherein K is a positive integer greater than or equal to 1; when there is a first computing domain among the K activated computing domains, in which the parameter value of a specified state parameter in the set of state parameters meets the corresponding threshold condition, adjusting the operating parameters of the first computing domain according to the adjustment method corresponding to the specified state parameter.

[0007] Through the present application, since the electronic component includes multiple computing units, the multiple computing units are divided into a computing domain cluster, the computing domains in the computing domain cluster include some computing units in the multiple computing units, and the computing domains in the computing domain cluster are a set of computing units that independently adjust operating parameters. In the computing domain cluster, one computing domain includes some computing units in multiple computing units, and different computing domains include different computing units, so that the computing domains in the computing domain cluster can be individually monitored and adjusted in terms of status. The control component is used to monitor a set of state parameters of the activated computing domains when there are K activated computing domains in the computing domain cluster, where K is a positive integer greater than or equal to 1; when there is a first computing domain in the K activated computing domains whose parameter value of a specified state parameter in a set of state parameters meets the corresponding threshold condition, the operating parameters of the first computing domain are adjusted according to the adjustment method corresponding to the specified state parameter. Therefore, the problem of poor timeliness of operation control of electronic components in related technologies can be solved, and the technical effect of improving the timeliness and flexibility of operation control of electronic components can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0009] Figure 1 A schematic structural diagram of an optional electronic component provided in an embodiment of the present application.

[0010] Figure 2 A schematic diagram of an optional electronic component provided in an embodiment of the present application.

[0011] Figure 3 A schematic diagram of another optional electronic component provided in an embodiment of the present application.

[0012] Figure 4 A flowchart of an optional operation control method of an electronic component provided in an embodiment of the present application. DETAILED DESCRIPTION

[0013] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0014] It should be noted that, in the description of this application, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. The terms "first," "second," etc., in this application are used to distinguish similar objects, and are not used to describe a particular order or sequence.

[0015] In order to enable those skilled in the art to better understand the present application, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0016] According to one aspect of the embodiments of the present application, an electronic component is provided. Figure 1 As shown, the above-mentioned electronic component 101 includes a control component 1011 and multiple computing units 1012, and the multiple computing units 1012 are divided into a computing domain cluster. The computing domain 1013 in the computing domain cluster is a set of computing units 1012 that independently adjust operating parameters; in the computing domain cluster, one computing domain 1013 includes some computing units 1012 in the multiple computing units 1012, and different computing domains 1013 include different computing units 1012.

[0017] The electronic component in this embodiment can be applied to the field of computer hardware technology, and can be applied to scenarios where the electronic component is required to process computing tasks and perform operation control on the electronic component.

[0018] With the rapid development of artificial intelligence (AI), electronic components are playing an increasingly important role in compute-intensive tasks. They are typically used to provide powerful parallel computing performance to accelerate AI model training and inference while maintaining low latency. These components can be designed as hardware components specifically for inference and training of deep learning models. Their architecture typically includes multiple dedicated computing units, high-bandwidth memory, high-speed interconnects, and sophisticated power and clock management systems, making them suitable for a variety of AI tasks, including but not limited to image classification, natural language processing, speech recognition, and recommendation systems. However, high-performance computing often comes with high power consumption and heat dissipation, posing significant challenges in mobile and edge device scenarios. Mobile devices, due to their size constraints, have limited heat dissipation capabilities, while edge computing devices may face even more extreme operating environments, such as high humidity and temperature. Consequently, operational control of electronic components is essential. This operational control not only extends device battery life but also directly impacts performance stability and user experience.

[0019] In these scenarios, operational control must find an optimal balance between performance, power consumption, temperature, and cost, but the relevant electronic components have limitations in this regard. Operational control of these electronic components often relies on dynamic voltage and frequency scaling and clock gating technologies, which lack timely operational control.

[0020] Dynamic voltage and frequency scaling (DVFS) is a widely used operational control technology that adjusts the operating voltage and frequency in real time to adapt to the current workload. When the load is low, the voltage and frequency are reduced to reduce dynamic power consumption; when the load is high, the voltage and frequency are increased to ensure sufficient computing performance. However, DVFS regulation is usually global, that is, the voltage and frequency are adjusted simultaneously throughout the entire electronic component. This leads to inefficiency in different internal functional modules when the load is uneven. For example, compute-intensive modules may be performance-limited due to overall frequency reduction, while non-compute-intensive modules may run at high voltage, wasting energy. In addition, DVFS regulation relies on feedback control and requires time to respond to load changes. This can lead to performance loss and unnecessary functional waste when the load changes rapidly.

[0021] Clock gating is another common operation control technology used here. It reduces the power consumption of the limiting circuit by pausing the clock signal when calculation is not required. For example, when the Central Processing Unit (CPU) is idle, some clock signals can be turned off until a new task arrives and then activated. However, clock gating usually decides to turn the clock on or off based on preset static rules rather than real-time load conditions, which lacks timeliness. In addition, frequent clock gating switching can cause performance delays, especially when switching tasks, which may lead to performance degradation.

[0022] Furthermore, the operational control of electronic components in related technologies is also based on heterogeneous computing resource scheduling and near-threshold computing (NTC) models. Here, heterogeneous computing refers to the use of multiple types of processors working together in a system to achieve higher computing efficiency and energy efficiency. Heterogeneous computing architectures are very common in electronic components because different computing tasks may be better suited to the processing methods of different types of processors. However, current heterogeneous resource scheduling algorithms are mostly based on preset rules or simple load balancing strategies, such as assigning high-computational-intensive tasks to the graphics processing unit (GPU) and low-load tasks to the CPU. While this technology improves resource utilization efficiency to a certain extent, given the diversity and uncertainty of AI tasks, static rules may not be able to make timely and optimal resource allocations. Moreover, such scheduling algorithms often ignore the correlation between key factors such as real-time performance, temperature, and power consumption. For example, when the temperature of a GPU area is too high, continuing to assign high-load tasks to it may exacerbate overheating and affect system stability and lifespan.

[0023] Here, near-threshold computing is another operational control technique. Since power consumption is proportional to the square of the voltage, power consumption can be significantly reduced by reducing the operating voltage of electronic components to an operating point close to the transistor threshold voltage. However, when transistors operate near the threshold voltage, timing errors are prone to occur due to factors such as process deviations, temperature changes, and environmental noise, resulting in inaccurate data processing. To address this issue, a large voltage margin is often reserved within a safe range, and a more conservative strategy is adopted, such as setting a higher voltage lower limit or frequent voltage recovery. This can improve system reliability, but also reduces the effect of energy efficiency optimization.

[0024] It can be seen that the operation control of electronic components in related technologies is usually based on static strategies or simple feedback control, with high adjustment delays and difficulty in adapting to sudden load changes or fast dynamic scenarios. Traditional DVFS and scheduling technologies cannot perform fine-grained adjustments to different modules inside electronic components, and usually only focus on a single dimension (such as performance or temperature). There is a lack of coordinated optimization of performance, power consumption, temperature and reliability, which can easily lead to local bottlenecks or resource waste. There is also a lack of predictive ability for load changes, and it is impossible to adjust voltage and frequency in advance, resulting in adjustment lag and performance loss. Even if the NTC mode is applied, due to the lack of an efficient error detection and recovery mechanism, a large voltage margin needs to be reserved, which limits the room for energy efficiency improvement. That is, the electronic components in related technologies have the problem of poor timeliness of operation control.

[0025] In order to at least partially solve the above technical problems, in this embodiment, an electronic component with a hierarchical adaptive energy efficiency management architecture is proposed. The computing units on the electronic component can be divided into multiple computing domain clusters. The computing domains in the computing domain cluster can include some computing units from multiple computing units. The computing domains in the computing domain cluster can be a set of computing units that independently perform operation control; in the computing domain cluster, a computing domain includes some computing units from multiple computing units, and different computing domains include different computing units, that is, the computing units in the electronic component can be divided into different areas, each of which can be an independent dynamic voltage and frequency adjustment domain, which means that they can independently adjust their operating voltage and frequency to adapt to different load requirements, thereby improving the accuracy and flexibility of the adjustment, and ensuring that different computing modules (such as CPU, GPU) inside the electronic component can achieve the optimal energy efficiency ratio according to their specific computing tasks.

[0026] Optionally, the division of computing units may be pre-set.

[0027] Optionally, each computing domain can also include a distributed sensor network. These sensors are used to monitor the status of the computing domain in real time. Each domain can have 2-4 sensors, or even more sensors. The data sampling rate of these sensors is as high as 1-10kHz, which can capture changes at the microsecond level and provide accurate information for dynamic energy efficiency management.

[0028] Optionally, each computing domain may include one or more computing units. For example, multiple computing units may be included as processing units (PUs), including PU0, PU1, PU2, ... PUm. The above computing units may be CPUs, GPUs, or other types of computing units. Each computing unit may adjust its working mode according to the configuration of the computing domain in which it is located to achieve optimal energy efficiency.

[0029] Here, different computing domains may include different computing units, that is, a computing unit does not belong to two computing domains at the same time. Operation control of a computing domain may be operation control of computing units in this area.

[0030] Optionally, the electronic component may include a control component, which may include a central arbitrator, a central scheduler, and other components. During the power-on startup phase, the electronic component may be initialized, including initializing the electronic component power supply, activating the central arbitrator, and loading firmware. The central arbitrator may be a module within the electronic component that coordinates and manages the component's operational control strategy. It may be responsible for key functions such as task scheduling, performance monitoring, and resource allocation. Upon confirming that the power supply is stable based on the Power_ON signal, the central arbitrator may immediately activate and begin the initialization sequence, preparing for the next steps.

[0031] Optionally, during the system initialization phase, the sensor network may be calibrated. For example, the central arbitrator may be configured to send an analog-to-digital converter (ADC) calibration instruction to the sensor network distributed throughout the electronic components. In response to the instruction, the ADC may perform zero-input reference measurement and gain coefficient adjustment.

[0032] Optionally, during the system initialization phase, key registers (such as clock dividers, interrupt controllers, etc.) can also be initialized. Here, the clock divider is used to adjust the clock frequency of each part within the electronic component to ensure that all components can operate at the expected frequency. The interrupt controller is used to control various events and interrupt requests inside and outside the electronic component. By initializing key registers, it can be ensured that the system can respond to important events in a timely manner and efficiently coordinate the execution of various tasks.

[0033] Optionally, after the sensor completes self-calibration, the calibration results can be stored in local registers. This is because the sensor network may be spread across the entire chip. To reduce data transmission latency and power consumption, local storage of calibration results ensures fast access when needed and immediate application in data processing.

[0034] Optionally, the electronic component may read and load a firmware program from a read-only memory (ROM). The firmware may include codes and data necessary for starting and initializing the electronic component.

[0035] In this embodiment, the machine learning engine can also be initialized to accurately predict task loads and participate in dynamic regulation in subsequent processes. This process may include reading model weights, accelerator configuration, and lightweight self-testing. Model weights can be read by a central arbitrator from a non-volatile memory (Electrically Erasable Programmable Read-Only Memory, or EEPROM). Accelerator configuration can refer to loading model weights into a static random-access memory (SRAM) and performing address mapping so that the accelerator can efficiently access this data. In addition, hyperparameters such as the time step of the long short-term memory (LSTM) network can also be set to prepare for the operation of the prediction model. Lightweight self-testing verifies whether key hardware components of the machine learning engine (such as the matrix multiplication and accumulation unit) are functioning properly.

[0036] Optionally, the compute domain, as an independent voltage and frequency control area, can be configured to a default low-power safe state at startup to prepare for upcoming computing tasks.

[0037] Optionally, the central arbitrator can also be used to send configuration packages to all computing domains through the power management bus to set the initial low-power state, for example, the voltage is set to 0.6V and the frequency is set to 800MHz, which can not only ensure the rapid startup of the acceleration device, but also reduce the power consumption during startup.

[0038] Optionally, once all modules have completed initialization, the system can enter a ready state, ready to receive and process computing tasks.

[0039] In this embodiment, the control component can be used to monitor a set of status parameters of the activated computing domains when there are K activated computing domains in the computing domain cluster, where K is a positive integer greater than or equal to 1. Here, the number of K can be variable, that is, the number of activated and working computing domains in the computing domain cluster can be variable. A computing domain in a dormant state can be activated, and an activated computing domain can be frozen and put into a dormant state.

[0040] Optionally, each computing domain in the computing domain cluster can have a set of status parameters, i.e., parameters that reflect the current operating status of the electronic component or its computing unit, which can be used to monitor and diagnose the status. These parameters may include but are not limited to temperature, power consumption, computing load (such as the number of instructions per cycle), cache hit rate, etc. The sensor network distributed on each computing domain in the electronic component can be used to monitor the above parameters. For example, the temperature sensor can be used to measure the local temperature of the computing domain, and the hardware performance counter can be used to measure the computing power load and other performance indicators of the computing domain.

[0041] Optionally, a star-shaped power supply network can be used in the computing domain architecture, which means that each computing domain can have its own power supply path, forming a radial network from the central power supply point to the voltage regulation unit of each domain. This design reduces the complexity of the power supply network and supports faster voltage switching. At the same time, the voltage regulation of each domain can be performed independently, which improves the flexibility of regulation.

[0042] Optionally, a decoupling capacitor can be integrated in each computing domain to suppress voltage transient noise and ensure power supply quality.

[0043] Optionally, if there is a first computing domain among the K activated computing domains for which a parameter value of a specified state parameter in a set of state parameters satisfies a corresponding threshold condition, the control component may be configured to adjust the operating parameters of the first computing domain according to the adjustment method corresponding to the specified state parameter. Each state parameter may have its own specific threshold condition, which may be pre-set based on factors such as the physical properties of the electronic component, the current workload requirements, and the system's operational control strategy. For example, a temperature threshold may be set to 85°C to prevent overheating of the corresponding component; while an instructions per cycle threshold may be dynamically adjusted based on changes in process, voltage, and temperature.

[0044] Optionally, operating parameters refer to parameters that can directly control the working state of electronic components or computing units, and their adjustment can directly affect state parameters such as power consumption, performance and temperature. For example, operating parameters may include voltage and frequency.

[0045] Optionally, when a specified parameter in a set of state parameters of a first computing domain among the K activated computing domains meets a pre-set threshold condition, it indicates that the operating state of the computing domain requires intervention. For example, if the temperature of a domain exceeds a safety threshold or its power consumption reaches a predefined upper limit, the corresponding control strategy will be triggered.

[0046] Alternatively, operational control is typically based on specified state parameters that meet threshold conditions. For example, if the temperature is too high, the heat can be distributed by reducing the frequency and voltage of the computational domain or migrating some of the load to other computational domains with lower temperatures. If power consumption is too high, it may be necessary to reduce computing tasks through scheduling or modify the computing mode of the computing unit.

[0047] Once the first compute domain is determined to require adjustment, the control component can intervene according to the corresponding operating parameter adjustment method. This process can be guided by a machine learning (ML) engine or a central arbitrator. For example, if the ML engine predicts that the load of a particular compute domain will increase, it can preemptively adjust the operating parameters by adjusting the voltage and frequency in advance, avoiding performance bottlenecks and sudden increases in thermal stress.

[0048] This monitoring and control mechanism enables the compute domain cluster to maintain optimal operating conditions under varying workloads. On the one hand, it ensures the stability and lifespan of electronic components by preventing overheating and excessive power consumption. On the other hand, precise resource scheduling and dynamic adjustment maximize overall system performance while ensuring the response speed and accuracy requirements of computing tasks. Operational control under this mechanism is not limited to a single compute domain; it enables thermal load balancing and dynamic resource allocation across compute domains, forming a collaboratively optimized whole.

[0049] Through the embodiments provided in the present application, the control component is used to monitor a set of state parameters of the activated computing domains when there are K activated computing domains in the computing domain cluster, where K is a positive integer greater than or equal to 1; when there is a first computing domain among the K activated computing domains in which the parameter value of a specified state parameter in a set of state parameters meets the corresponding threshold condition, the operating parameters of the first computing domain are adjusted according to the adjustment method corresponding to the specified state parameter, which can solve the problem of poor timeliness of operation control of electronic components in related technologies and achieve the technical effect of improving the timeliness and flexibility of operation control of electronic components.

[0050] In an exemplary embodiment, the electronic component further comprises at least one of the following:

[0051] A voltage regulating unit is configured for each computing domain in the computing domain cluster, and is used to regulate the voltage of the corresponding computing domain, wherein the operating parameters of the computing domains in the computing domain cluster include the voltages of the computing domains in the computing domain cluster, and the voltage regulating unit includes at least one of the following: a low-dropout regulator, a switched capacitor converter;

[0052] A local clock tree synthesis element, an asynchronous bridge, and a digital phase-locked loop configured for each computational domain in the computational domain cluster. The local clock tree synthesis element is used to distribute the global clock signal of the electronic component to different computational domains in the computational domain cluster to form clock domains corresponding to the different computational domains in the computational domain cluster. The asynchronous bridge establishes a data transmission path between different clock domains. The digital phase-locked loop is used to adjust the frequency of the corresponding computational domain. The operating parameters of the computational domain in the computational domain cluster include the frequency of the computational domain in the computational domain cluster.

[0053] Optionally, each domain in a compute domain cluster can be independently voltage- and frequency-controlled. This means each domain can have its own power management and clock frequency to adapt to the energy efficiency requirements of different computing tasks or load conditions. In a compute domain cluster, each domain's operating parameters include, but are not limited to, its operating voltage and frequency. Dynamic adjustment of these parameters allows for adjustments in computing efficiency and energy efficiency. For example, under light load conditions, power consumption can be reduced by lowering the voltage; under heavy load conditions, computing performance can be improved by increasing the voltage and frequency.

[0054] Optionally, the electronic component may include a voltage regulation unit configured separately for the computing domains in the computing domain cluster. Here, the voltage regulation unit is a hardware component responsible for adjusting the voltage of the computing domain. The voltage regulation unit may include a low dropout regulator (LDO), which can be used to provide a stable operating voltage for the computing domain and stably convert the input voltage (for example, the global 1.8V) into a voltage level (0.4-1.2V) suitable for the load of the computing domain to meet the power supply requirements of different computing tasks while reducing power consumption and heat.

[0055] Optionally, the voltage regulating unit may also include a switched capacitor converter. Here, the switched capacitor converter is a circuit that uses a combination of capacitors and switches to achieve voltage conversion. It does not require an external power supply, but instead uses the charge and discharge cycle of the internal power supply of the electronic component to adjust the voltage.

[0056] Optionally, the electronic component can also include a local clock tree element for distributing the electronic component's global clock signal to different computational domains within the computational domain cluster, forming clock domains corresponding to the different computational domains within the computational domain cluster. This can be dynamically adjusted based on task load requirements without affecting the operation of other computational domains. Each computational domain can be integrated with a digital phase-locked loop (DPLL) to generate and adjust the operating frequency, allowing the computational units within the domain to operate at different frequencies. The frequency can be dynamically adjusted based on the load within the domain. For example, the frequency adjustment range can be stepped from 10 MHz to 100 MHz, providing flexible clock management capabilities for the computational domain.

[0057] Optionally, in order to ensure the synchronization of cross-domain data, the electronic components may also include an asynchronous bridge, for example, a dual-clock domain first-in-first-out buffer (FIFO). This bridge can transmit data without distortion between different clock frequencies. The depth of the FIFO is at least 8, thereby ensuring the integrity and timing correctness of the data, and maintaining data synchronization and correctness even when data is exchanged between domains with different frequencies.

[0058] Through this embodiment, through the local clock tree synthesis element in the electronic component, the asynchronous bridge, and the digital phase-locked loop and voltage regulation unit respectively configured for the computing domains in the computing domain cluster, the voltage and frequency of each computing domain can be adjusted independently, thereby improving the flexibility and reliability of the operation control of the electronic component.

[0059] In an exemplary embodiment, the control component and the computing units in the computing domains of the computing domain cluster are connected via an on-chip interconnection network.

[0060] A Network-on-Chip (NoC) is a high-speed, low-latency communication network that can be used within large-scale integrated circuits (LSIs). It can replace traditional bus architectures and provide a more flexible and reliable data transmission mechanism. A NoC can include multiple routing nodes and links, enabling parallel communication between multiple computing units.

[0061] Optionally, the NoC topology can be a regular grid, ring, tree, or more complex hybrid topology. The choice of topology may depend on the number and location of computing domains and the data transmission requirements. For example, a grid topology can be used for large-scale evenly distributed computing units, and a tree topology can be used for a hierarchical cluster of computing domains.

[0062] Optionally, through the NoC, the control component can send control signals to the computing domain and each computing unit within the computing domain to achieve dynamic task allocation, voltage / frequency adjustment and other operational controls. For example, in the event of a sudden high load, the control component can send frequency reduction and voltage reduction commands to the computing units in the computing domain cluster through the NoC to reduce power consumption and temperature.

[0063] Through this embodiment, the control component and the computing units in the computing domains of the computing domain cluster are connected via the on-chip interconnection network, which can improve the efficiency and flexibility of data transmission.

[0064] In an exemplary embodiment, the set of state parameters includes a number of instructions per cycle, and the threshold condition corresponding to the number of instructions per cycle is that the value of the number of instructions per cycle is less than or equal to a first instruction number threshold;

[0065] The control component is further configured to adjust the operating parameters of the first computing domain according to the adjustment method corresponding to the specified state parameter, including:

[0066] When the specified state parameter includes the number of instructions per cycle, a first adjustment operation is performed on the first computing domain to adjust the operating parameters of the first computing domain, wherein the first adjustment operation includes: reducing the frequency of the first computing domain and reducing the voltage of the first computing domain.

[0067] Here, instructions per cycle (IPC) refers to the number of instructions that can be executed within one clock cycle. It can be the number of instructions per cycle of the computing domain or the number of instructions per cycle of the computing unit. In this embodiment, a set of state parameters can include the number of instructions per cycle. A high IPC value means that the computing domain can efficiently process more instructions, thereby achieving higher computing performance. Generally, higher frequencies and voltages can support the execution of more instructions, but also result in higher power consumption and heat generation.

[0068] Optionally, a set of status parameters can include multiple indicators, including IPC, to comprehensively monitor and evaluate the operating status of electronic components. Within this set of parameters, the IPC threshold condition can be set to a first instruction count threshold. When the IPC value remains below this threshold, it indicates that the computational load of the electronic component is low, and the current frequency and voltage settings may exceed actual requirements, resulting in unnecessary power consumption.

[0069] Optionally, when the specified state parameter includes the number of instructions per cycle, the control component can be used to perform a first adjustment operation on the first computing domain to adjust the operating parameters of the first computing domain, wherein the first adjustment operation may include: lowering the frequency of the first computing domain and lowering the voltage of the first computing domain. Here, by lowering the clock frequency of the computing domain, the number of instructions executed in each clock cycle can be reduced, thereby reducing power consumption. In addition, lower voltage also means lower power consumption.

[0070] Optionally, in the operation control strategy, the performance of the computing domain can be measured in the form of TOPS / W (computing power per watt), which can reflect the computing power of electronic components per unit power consumption. The computing domain scoring formula can be performance R / (power consumption × temperature), which can balance performance and power consumption and take into account the impact of temperature. In this computing domain scoring, although a high IPC value represents high performance, if power consumption and temperature are also high, the advantage of high IPC will be weakened. Therefore, the goal of operation control can be to optimize the performance of the entire system by adjusting frequency and voltage to minimize power consumption and temperature while meeting performance requirements.

[0071] In this embodiment, by dynamically monitoring IPC and comparing it with a preset threshold, low-load status can be intelligently identified, and the frequency and voltage can be adjusted down appropriately, thereby avoiding energy waste when computing demand is low and improving the overall energy efficiency of the system.

[0072] Optionally, both the IPC threshold setting and the frequency / voltage adjustment strategy can be dynamic, allowing for flexible adjustments based on different application environments and computing loads. This operational control strategy is suitable not only for bursty computing tasks but also for long-term, high-load scenarios, ensuring the system maintains optimal operating conditions under various operating conditions.

[0073] Through this embodiment, by monitoring the number of instructions per cycle and adjusting corresponding operating parameters when the number of instructions per cycle is lower than a corresponding threshold, the timeliness and reliability of the operation control of the electronic component can be improved.

[0074] In an exemplary embodiment, the electronic component further includes: a hardware performance counter for monitoring the number of instructions per cycle of the K activated computing domains in real time.

[0075] Here, the hardware performance counter (HPC) can be used to collect and count the performance indicator data of each unit in the electronic component in real time. The unit can be a computing unit or a computing domain.

[0076] Optionally, hardware performance counters may be used to monitor, including but not limited to, number of instructions per cycle, cache hit rate, branch hit rate, instruction type mix, data flow pattern, and the like.

[0077] In this embodiment, hardware performance counters can be used to continuously monitor the IPC values ​​of K activated compute domains, where K refers to the number of currently active compute domains, i.e., the number of activated compute domains. By monitoring the IPC of these domains in real time, it is possible to quickly identify which compute domains are executing instructions efficiently and which are relatively lightly loaded. Furthermore, when the IPC of a compute domain consistently falls below a preset threshold, the operating frequency and voltage of that compute domain can be reduced.

[0078] Similarly, if a compute domain has a high IPC value, it indicates that the compute domain is operating at high voltage, which may require additional cooling or voltage adjustments to prevent overheating and excessive power consumption.

[0079] Through this embodiment, the number of instructions per cycle of the activated computing domain is monitored in real time through the hardware performance counter, and the immediate working status of each computing domain inside the electronic component can be obtained in real time, so that changes can be responded to quickly and corresponding operation control measures can be implemented, thereby improving the timeliness of the operation control of the electronic component.

[0080] In an exemplary embodiment, a set of state parameters includes a calculation domain temperature, and a threshold condition corresponding to the calculation domain temperature is: a temperature value of the calculation domain temperature is greater than or equal to a first temperature threshold;

[0081] The control component is also used to perform a second adjustment operation on the first computing domain to adjust the operating parameters of the first computing domain when the specified state parameters include the computing domain temperature, wherein the second adjustment operation includes at least one of the following: lowering the frequency of the first computing domain and performing voltage compensation on the second computing domain, wherein the second computing domain is an activated computing domain adjacent to the first computing domain and does not meet the threshold condition corresponding to the computing domain temperature; migrating the task executed by the first computing domain to the activated computing domain that does not meet the threshold condition corresponding to the computing domain temperature.

[0082] In this embodiment, the control component can be used to adjust the computing domain based on the monitoring results of the computing domain temperature. Here, the computing domain temperature can be used to evaluate the thermal status of each computing domain (i.e., different computing areas of the electronic component). Once the temperature of a computing domain reaches or exceeds a predefined temperature threshold, it means that the computing domain has entered an overheating state, which not only affects its computing performance, but may also affect the stability and life of the hardware.

[0083] Optionally, the first temperature threshold may be preset, or may be modified as needed.

[0084] Optionally, when it is monitored that the temperature of the first computing domain has reached or exceeded a preset temperature threshold, a corresponding second adjustment operation can be triggered to optimize overall energy efficiency and heat distribution to avoid performance degradation or hardware damage caused by excessive temperature.

[0085] Optionally, the second adjustment operation may include: lowering the frequency of the first computing domain and performing voltage compensation on the second computing domain. Here, lowering the frequency can directly reduce power consumption and heat generation. Through this adjustment operation, the power consumption of the first computing domain can be quickly reduced, its temperature can be lowered, and overheating can be avoided. After lowering the frequency and voltage of the first computing domain, in order to maintain the performance stability of the overall system, voltage compensation can be performed on the second computing domain that is adjacent to the first computing domain and whose temperature has not yet reached the overheating threshold. This is because due to the electrical connection between the computing domains, the frequency reduction operation may indirectly affect its neighboring computing domains. Specifically, the frequency reduction of the first computing domain causes its power consumption and current demand to decrease, which may cause the current distribution of the overall power supply network to change, thereby increasing the voltage drop (voltage drop due to IR resistance, referred to as IR drop) effect in the area where the adjacent computing domain is located. Here, IR drop refers to the problem of a decrease in power supply voltage at various points inside the electronic component due to changes in resistance and current inside the electronic component. To counteract this effect, the voltage of the second computing domain can be moderately increased while the frequency of the first computing domain is reduced. By increasing the supply voltage of the second computing domain, the local voltage drop in the power supply network caused by the frequency reduction of the first computing domain can be offset, thereby maintaining stable operation and high-performance output of the second computing domain.

[0086] Optionally, the second computing domain may include only activated computing domains.

[0087] Optionally, multiple high-temperature calculation domains whose temperatures exceed a preset temperature threshold, i.e., multiple hot spots, can be identified. For each identified hot spot, the temperature conditions around it can be further analyzed, and adjacent points within a certain range of distance from the hot spot and with temperatures higher than the threshold can be merged into the same cluster to form an adjustment area.

[0088] Optionally, the second adjustment operation may further include: migrating part of the computing tasks on the first computing domain to other computing domains with lower temperatures, and further controlling the heat source by redistributing the tasks to achieve thermal balance.

[0089] Through this embodiment, by taking corresponding adjustment operations when overheating of the computing domain is detected, the reliability and stability of the operation control of the electronic component can be improved.

[0090] In an exemplary embodiment, the control component is also used to lower the frequency of the first calculation domain by a specified frequency value, wherein the specified frequency value is the product of the first coefficient and the specified temperature difference value, and the specified temperature difference value is the difference between the temperature value of the calculation domain temperature of the first calculation domain and the first temperature value; and to increase the voltage of the second calculation domain by a specified voltage value, wherein the specified voltage value is the product of the second coefficient, the specified difference value and the specified voltage compensation value, and the specified difference value is the difference obtained by subtracting the ratio of the temperature value of the calculation domain temperature of the second calculation domain to the temperature value of the calculation domain temperature of the first calculation domain from 1.

[0091] In this embodiment, a high-temperature computing domain (i.e., a computing domain whose temperature exceeds a first temperature threshold) may be subjected to a gradient frequency reduction, as shown in formula (1):

[0092] Δf = -k×(T_now - T_max) (1)

[0093] Wherein, Δf refers to the specified frequency value to be lowered, k refers to the first coefficient, T_now refers to the temperature value of the calculation domain temperature of the first calculation domain, and T_max refers to the first temperature value, that is, the first temperature threshold.

[0094] Optionally, the specified frequency value refers to a dynamic value calculated based on the first coefficient and the specified temperature difference. The first coefficient can be a value predetermined by experiment, which is used to reflect the strength of the relationship between temperature and frequency adjustment. The larger the first coefficient, the more obvious the impact of temperature change on frequency adjustment, and vice versa. The specified temperature difference refers to the difference between the actual temperature of the first calculation domain and the first temperature threshold. That is, the more the calculation domain temperature exceeds the threshold, the greater the frequency reduction amplitude, and the amplitude of the frequency reduction is proportional to the degree of temperature exceeding the limit, thereby ensuring that the response speed of thermal control matches the degree of temperature increase, avoiding the risk of overheating.

[0095] Optionally, while lowering the frequency, the supply voltage of the computational domain can be adjusted according to a dynamic voltage frequency scaling (DVFS) table to ensure stable circuit operation at different frequencies while further reducing power consumption. The DVFS table can include recommended voltages for different frequencies. For example, when the frequency drops from 2.0 GHz to 1.5 GHz, the supply voltage can also be reduced from 1.0 V to 0.8 V. This voltage-following strategy ensures that the voltage adjustment matches the frequency change during frequency reduction, maintaining efficient and reliable circuit operation.

[0096] Optionally, when performing cold zone compensation, that is, performing voltage compensation on the second calculation domain, voltage compensation may be performed based on formula (2):

[0097] ΔV_cold =α× (1 - (T_cold / T_hot)) × ΔV_max (2)

[0098] Where ΔV_cold is the specified voltage value, α is the second coefficient, T_cold is the temperature value of the second calculation domain, T_hot is the temperature value of the first calculation domain, (1 - (T_cold / T_hot)) is the specified difference, and ΔV_max is the specified voltage compensation value.

[0099] Here, the second coefficient can be a value predetermined through experiments, which can be used to reflect the correlation between temperature changes and voltage compensation effects, that is, the degree of influence of temperature differences on the voltage compensation amount. The specified difference is defined as 1 minus the difference between the temperature of the second calculation domain and the temperature of the first calculation domain. It can be used to represent the difference between the hot and cold temperatures of the second calculation domain relative to the first calculation domain. The larger the temperature difference, the greater the need for compensation. The specified voltage compensation value can be an upper limit value used to limit the compensation amount to avoid new problems caused by excessive voltage.

[0100] Through the above-mentioned anti-gradient regulation algorithm, it can be achieved that when the temperature difference between the cold zone and the high temperature domain is large, the cold zone voltage compensation will be more significant to offset the voltage loss caused by the increase in hot spot current in the power supply network and ensure the normal operation of the cold zone circuit.

[0101] Through this embodiment, by performing gradient frequency reduction on the high-temperature calculation domain and compensating the voltage of the adjacent cold zone, the local heat source can be finely adjusted. Through dynamic adjustment and precise control, efficient control is achieved under complex and changing load and environmental conditions, ensuring performance optimization while reducing energy waste.

[0102] In an exemplary embodiment, the electronic component also includes a distributed sensor network; wherein, the sensor network is used to detect the temperature of the activated computing domain separately when a set of state parameters includes the computing domain temperature, and obtain sensor data corresponding to the activated computing domain; the control component is also used to parse the sensor data corresponding to the activated computing domain, and obtain the actual temperature value of the activated computing domain at the current moment; based on the actual temperature value of the activated computing domain at the current moment, the current power value of the activated computing domain and the ambient temperature value at the current moment, the temperature value of the activated computing domain at the next moment of the current moment is predicted, wherein the current power value of the activated computing domain is the power value of the activated computing domain at the current moment, and the temperature value of the computing domain temperature of the activated computing domain is the predicted temperature value of the activated computing domain at the next moment.

[0103] Electronic components in related technologies can integrate temperature, voltage, and current sensors for real-time monitoring of component status. However, these sensor data are usually only used for simple threshold alarms or global adjustments, and fail to be deeply integrated with dynamic operation adjustment strategies.

[0104] In this embodiment, the electronic components also include a distributed sensor network, wherein sensors can be distributed in each computing domain for real-time monitoring of temperature changes in the computing domain. Correspondingly, the sensor network can detect the temperature in each activated computing domain separately to obtain sensor data corresponding to the activated computing domain.

[0105] The control component can be used to parse the sensor data corresponding to the activated computing domain to obtain the actual temperature value of the activated computing domain at the current moment, wherein the sensor data corresponding to the activated computing domain is the sensor data obtained by the sensor network by detecting the activated computing domain separately, that is, by parsing the data sent by the sensor network, it can be parsed and decoded to obtain the precise temperature value of each activated computing domain at the current time point.

[0106] Optionally, in addition to detecting the temperature within the computing domain, the current ambient temperature can also be taken into account. This is because the thermal condition of the modules in the electronic components is not only affected by the internal computing activities, but the external ambient temperature is also an important factor. The inclusion of the ambient temperature helps to establish a more accurate temperature model, especially for edge computing devices, whose working environment may be more variable.

[0107] Optionally, based on real-time temperature, power, and environmental data, the control component can predict the temperature value of the activated computing domain at the next time point and use it as the computing domain temperature of the activated computing domain, so as to take control actions before the temperature actually exceeds the threshold, such as reducing the frequency in advance, transferring the load, etc., thereby avoiding performance degradation and hardware damage caused by overheating.

[0108] Through this embodiment, the actual temperature and predicted temperature of the calculation domain can be obtained through sensor data obtained based on a distributed sensor network, which can improve the foresight and flexibility of the operation control of the electronic components.

[0109] In an exemplary embodiment, the control component is also used to determine the predicted temperature value of the activated computing domain at the next moment by summing the product of the actual temperature value of the activated computing domain at the current moment and the thermal inertia coefficient, the product of the current power value of the activated computing domain and the power-temperature conversion coefficient, and the product of the ambient temperature value at the current moment and the specified heat exchange rate, wherein the specified heat exchange rate is the heat exchange rate between the electronic component and the environment in which the electronic component is located.

[0110] In this embodiment, the control component can be used to determine the predicted temperature value of the activated computing domain at the next moment by summing the product of the actual temperature value of the activated computing domain at the current moment and the thermal inertia coefficient, the product of the current power value of the activated computing domain and the power-temperature conversion coefficient, and the product of the ambient temperature value at the current moment and the specified heat exchange rate, as shown in formula (3):

[0111] T_new = A×T_prev + B×Power_map + C×Ambient (3)

[0112] Here, T_new represents the new temperature state, T_prev is the previous temperature, Power_map indicates the power distribution within the electronic component, and Ambient represents the ambient temperature. The coefficient matrices A, B, and C respectively reflect the thermal inertia coefficient of the electronic component, the power-to-temperature conversion coefficient, and the heat exchange rate between the ambient environment and the electronic component (i.e., the ambient coupling coefficient).

[0113] Among them, the thermal inertia coefficient A can be used to indicate the speed of change of electronic components when facing temperature changes. The larger the coefficient A, the slower the electronic components react to temperature changes; the power-temperature conversion coefficient B can be used to reflect the degree of temperature rise caused by unit power consumption per unit time; the environmental coupling coefficient C can be used to measure the ability of electronic components to exchange heat with the surrounding environment, that is, the intensity of the impact of ambient temperature on the internal temperature of electronic components.

[0114] Optionally, the temperature, power, and ambient temperature of the computational domain can be monitored in real time. These three parameters are then multiplied by the corresponding coefficient matrices A, B, and C, respectively. The results are summed to obtain the predicted temperature at the next moment. This process can be updated every 100 μs, ensuring the real-time and accuracy of the temperature prediction.

[0115] Alternatively, in the later stages of electronic component design (i.e., post-silicon testing), the temperature distribution can be recorded using an infrared thermal imager through a heating resistor under different power consumption conditions, and the three coefficient matrices A, B, and C can be finely calibrated to ensure the accuracy and effectiveness of the prediction model.

[0116] Through this embodiment, temperature prediction is performed by combining the actual temperature, power value, ambient temperature and thermal coefficient of electronic components with the thermal prediction model, which can improve the real-time performance and accuracy of temperature prediction, thereby improving the accuracy of operation control.

[0117] In an exemplary embodiment, the sensor network includes a set of sensor nodes in a grid layout, and a computing domain in the computing domain cluster corresponds to some of the sensor nodes in the set of sensor nodes;

[0118] The control component is further used to obtain sensor data detected by the sensor nodes corresponding to the K activated computing domains by adopting an address polling method, and obtain sensor data corresponding to the K activated computing domains.

[0119] Optionally, the sensor network can adopt a uniform grid layout, integrating one sensor node per square millimeter, a computational domain in the computational domain cluster can correspond to some of the sensor nodes in the sensor node set, and the sensors can be high-precision sensors (for example, thermal gradient detection accuracy of ±0.1°C / mm).

[0120] Optionally, the sensing node may include: a PN junction diode temperature sensor for temperature detection, with a nominal accuracy of ±0.5°C; a micro Hall sensor for current detection, with a range of 0-100mA and a resolution of 1mA, thereby accurately measuring the current flowing through each computational domain; and an on-chip 12-bit Successive Approximation Register (SAR) ADC for converting analog temperature and current signals into digital signals. The ADC sampling rate may be 1kHz (i.e., 1000 samples per second) to ensure real-time data, and an effective number of bits (ENOB) ≥ 10 bits indicates that the actual effective number of bits of the ADC is sufficiently high under the influence of noise, and the data is reliable.

[0121] Optionally, the control component may interact with the sensor network by way of address polling. In the address polling, the control component may send data requests to each sensor node in the sensor network in turn according to a predefined order, requesting them to return current monitoring data.

[0122] Optionally, the communication between the control component and the sensor network generally follows a standardized protocol, such as the Inter-Integrated Circuit (I2C).

[0123] Optionally, the control component may send a request to each sensor node at a fixed time interval (eg, 100 μs) and wait for a response. This time interval ensures that the request can cover all sensor nodes in the activated computing domain without affecting the overall system performance.

[0124] Optionally, after receiving the sensor data, the control component may perform preliminary analysis and preprocessing on the data for use in formulating dynamic adjustment strategies.

[0125] Optionally, the control component may send data requests only to the sensor nodes corresponding to the activated computing domain. After receiving the request, the sensor nodes corresponding to the activated computing domain may return the current temperature, voltage, current and other data to the control component through the I2C bus.

[0126] Through this embodiment, the control component periodically polls the sensor nodes corresponding to the activated computing domain, thereby improving the real-time performance and reliability of data collection.

[0127] In an exemplary embodiment, the electronic component further includes a temperature sensor; wherein the temperature sensor is configured to periodically detect the ambient temperature;

[0128] The temperature sensor or control component is also used to calculate the temperature difference between the detected ambient temperature value and the recorded reference ambient temperature value; when the temperature difference between the detected ambient temperature value and the recorded reference ambient temperature value is greater than or equal to the temperature difference threshold, the product of the detected ambient temperature value and the specified coefficient is added to the second temperature value to determine it as the updated first temperature threshold, and the recorded reference ambient temperature value is updated to the detected ambient temperature value.

[0129] In this embodiment, the temperature sensor may be configured to detect the ambient temperature periodically (eg, every minute or when a significant system state changes), thereby obtaining real-time temperature changes in the operating environment of the electronic component.

[0130] Optionally, after receiving the latest ambient temperature value from the temperature sensor, it can be compared with a pre-recorded reference ambient temperature value to calculate the difference between the two temperature values. The reference ambient temperature value here can be the ambient temperature at the time of the last detection or the initial ambient temperature recorded when the system is initialized. This calculation process can be performed by the control component or autonomously by the temperature sensor.

[0131] Similar to the above embodiment, the temperature threshold can be dynamically updated and adjusted. Here, a temperature difference threshold can be set to determine whether the current ambient temperature change is sufficient to affect the operation control strategy of the electronic component. If the difference between the detected ambient temperature value and the reference value is greater than or equal to this threshold (for example, ΔT ≥ 5°C), it can be considered that the environmental conditions have changed significantly and the temperature threshold in the electronic component needs to be adjusted. Specifically, the new first temperature threshold (the upper temperature limit for triggering the adjustment operation) can be defined as the product of the detected ambient temperature value and a specified coefficient plus a fixed second temperature value. The specified coefficient here reflects the sensitivity of the ambient temperature to the internal temperature of the electronic component, and the second temperature value can represent the temperature offset of the electronic component under ideal cooling conditions.

[0132] T_max = base value + β × ambient temperature (4)

[0133] Wherein, T_max is the first temperature threshold, and β is the specified coefficient.

[0134] Optionally, after the temperature threshold is adjusted, the newly detected ambient temperature value may replace the original reference ambient temperature value and serve as a comparison benchmark for the next period of ambient temperature detection.

[0135] By adjusting the temperature threshold, it is possible to maximize its performance output under different environmental conditions while staying within a safe energy consumption range. For example, in a low-temperature environment, electronic components may be able to withstand higher internal temperatures.

[0136] Through this embodiment, by dynamically updating the first temperature threshold according to the ambient temperature, the accuracy and timeliness of temperature judgment can be improved, thereby improving the accuracy and timeliness of operation control of the electronic component.

[0137] In an exemplary embodiment, the control component is also used to, when there is a first computing task to be assigned, select a third computing domain from the K activated computing domains to execute the first computing task based on the task tag of the first computing task and the operating characteristic data of the K activated computing domains, wherein the task tag of the first computing task is used to indicate the task attributes of the first computing task and the execution constraints of the first computing task; and assign the first computing task to the third computing domain so that the first computing task is executed by the first computing unit of the third computing domain.

[0138] In this embodiment, when there is a first computing task to be assigned, that is, a new computing task enters the central task scheduler queue in the control component and waits for resource allocation, the most suitable computing domain can be selected to execute the task based on the task requirements and the operating status of the currently activated computing domain to maximize efficiency and optimize performance.

[0139] Optionally, task requirements can be determined based on the compute task's task tag, which is part of the Task Descriptor (TD) and can include task attributes and execution constraints. For example, the TD for an AI inference request might include max_power_allowed (maximum allowed power consumption), required precision (e.g., single-precision floating point, half-precision floating point, or 8-bit integer), and task type (e.g., image classification, natural language processing, etc.).

[0140] Optionally, the central scheduler may parse the task tag of the first computing task to understand the task type, accuracy requirements, and execution constraints.

[0141] Similar to the aforementioned embodiment, the operating status parameters of all activated computing domains on the electronic components can be monitored in real time, including their current temperature and load status. Furthermore, the central scheduler can calculate the adaptability score of each computing domain based on the operating characteristic data of the computing domain. For example, the computing domain score (performance² / power consumption × temperature), load balancing status, and whether the current power consumption of the computing domain meets the execution constraints of the task can be calculated.

[0142] Optionally, the central scheduler can continuously monitor the real-time process-voltage-temperature (PVT) data of the electronic components. This data reflects the current operating environment and status of the electronic components and can be used to evaluate the computing resources available on the electronic components.

[0143] Optionally, after evaluating all K activated computational domains, the computational domain with the highest score that best meets the requirements of the first computational task can be selected as the third computational domain. If multiple domains have the same score, they can be selected based on other priorities (such as temperature, power consumption, or availability of computing resources).

[0144] Optionally, once the third computing domain is determined, the central scheduler can assign the first computing task to the first computing unit of the computing domain. Subsequently, the first computing unit executes the task under the execution constraints specified in the task label, and can continuously monitor its operating status to ensure consistency in energy efficiency and performance.

[0145] When a task arrives, the central scheduler adds the new computing task to the task queue and begins to allocate resources based on the task tag and real-time PVT data. This step is the starting point of the dynamic operation control strategy, in which the task tag and precision requirements (single-precision floating point / half-precision floating point / 8-bit integer) play a guiding role, ensuring that the task is assigned to the computing unit that best suits its needs. For example, a high-load, low-precision image detection task may be directed to a Neural Processing Unit (NPU) in 8-bit integer mode because the NPU can provide high energy efficiency in 8-bit integer mode while meeting the precision requirements of the task.

[0146] Through this embodiment, by allocating tasks based on the task tags and the operating characteristic data of the activated computing domains when the tasks arrive, it is possible to achieve reasonable allocation of computing resources, avoid resource waste and excessive power consumption, and improve the flexibility of the system.

[0147] In an exemplary embodiment, a task level prediction engine is run on the control component, and the task level prediction engine is used to predict the task level of the computing task corresponding to the input task label based on the input task label and the input historical performance data; wherein,

[0148] The control component is also used to input the task label of the first computing task and the historical performance data of the activated computing domain into the task level prediction engine to obtain the task level of the first computing task predicted by the task level prediction engine, wherein the task level of the first computing task is used to characterize the computing power requirement for executing the first computing task; from the K activated computing domains, an activated computing domain that matches the task level of the first computing task is selected to obtain a third computing domain.

[0149] Here, the task level prediction engine refers to a machine learning component integrated in the central scheduler, which can be used to predict the task level of the computing task corresponding to the input task label based on the input task label and the input historical performance data. For example, the prediction engine can analyze the execution status of past tasks with similar attributes to the first computing task (such as accuracy requirements, power consumption constraints, etc.), including their load patterns, power consumption characteristics, execution time and temperature changes, etc. Through these historical data, the prediction engine can build a model to predict the behavior patterns and computing power requirements of new tasks on different computing domains.

[0150] Optionally, the control component can parse the task descriptor of the first computing task to understand its attributes and execution constraints, and pass the task's TD information and the historical performance data of the corresponding activated computing domains as input to the task level prediction engine. Based on the input data, the task level prediction engine can use a pre-trained model to predict the computing power requirement level of the first computing task. The prediction result can also reflect the expected performance of the task on different computing domains and can be used to guide resource allocation.

[0151] Optionally, after completing the task level prediction, the control component can select a computing domain that best matches the predicted task level from the K activated computing domains and use it as the third computing domain. The selection criteria may include the current status of the computing domain, historical performance and computing power requirements predicted by the task level. Finally, the scheduler assigns the first computing task to the third computing domain and adjusts the operating parameters (such as voltage and frequency) of the computing domain based on the predicted performance requirements and execution constraints of the task to optimize the efficiency and energy efficiency of task execution.

[0152] Through this embodiment, the computing power requirements of tasks are predicted by the task level prediction engine, which can improve the accuracy of task allocation and reduce resource waste.

[0153] In an exemplary embodiment, one of the plurality of computing units allows for executing a computing task using at least one precision mode, wherein a higher computing precision corresponding to the precision mode results in a greater number of floating-point bits used to execute the computing task; the task attribute of the first computing task includes a precision requirement of the first computing task, wherein the precision requirement of the first computing task is used to indicate a minimum precision mode required to execute the first computing task;

[0154] The control component is further configured to, when there are at least two candidate computing domains among the K activated computing domains whose allowed precision modes match the precision requirement of the first computing task, calculate a computing domain score of the candidate computing domain based on the performance of the candidate computing domain, the power consumption of the candidate computing domain, and the temperature of the candidate computing domain, wherein the computing domain score of the candidate computing domain is positively correlated with the performance of the candidate computing domain and negatively correlated with the power consumption of the candidate computing domain and the temperature of the candidate computing domain, and the performance of the candidate computing domain is the number of instructions per cycle of the candidate computing domain;

[0155] A candidate computing domain with the highest computing domain score among the at least two candidate computing domains is determined as the third computing domain.

[0156] In modern heterogeneous computing systems, such as AI electronic components, computing units that can support multiple precision modes are widely used to meet the needs of different computing tasks. The choice of precision mode has a direct impact on power consumption, performance, and operating status. Although high-precision calculations (such as single-precision floating point) provide higher accuracy, they also consume relatively more resources (including power consumption and computing time). Lower-precision calculations (such as 8-bit integers) achieve higher performance ratios while sacrificing a certain degree of accuracy. They are especially suitable for scenarios such as inference that have high real-time requirements but a slightly higher tolerance for accuracy.

[0157] In this embodiment, each computing unit can support one or more precision modes, such as single-precision floating-point mode, half-precision floating-point mode, and 8-bit integer mode. The higher the precision, the more floating-point bits are used when executing the computing task, and the more accurate the calculation process is, but the corresponding computing resource consumption and power consumption are also greater.

[0158] Specifically, the following precision modes are possible:

[0159] Single-precision floating-point mode (2'b00): uses 32-bit floating-point calculations and is suitable for scenarios with extremely high precision requirements, such as scientific computing tasks such as model training.

[0160] Half-precision floating-point mode (2'b01): uses 16-bit floating-point calculations, which can achieve a good balance between accuracy and energy efficiency and is often used for mixed-precision training.

[0161] 8-bit integer mode (2'b11): uses 8-bit fixed-point calculations and has the highest computing performance. It is mainly used in inference scenarios such as image classification and object detection.

[0162] Taking the reconfigurable multiplier as an example, different calculation methods can be selected according to the value of the precision_mode function:

[0163] When precision_mode is 2'b00, the fp32_mult function is called to perform 32-bit floating-point multiplication.

[0164] When precision_mode is 2'b01, 16-bit floating-point multiplication fp16_mult is first performed on the lower 16 bits of the input, and then the result is converted to 32-bit floating point.

[0165] When precision_mode is 2'b11, 8-bit fixed-point multiplication int8_mult is performed on the lower 8 bits of the input and zero-filled in the upper bits to extend to 32 bits.

[0166] To further optimize the area of ​​electronic components, the reconfigurable multiplier can reuse some hardware resources, such as the mantissa processing unit, between the three modes. This means that when switching between different precision modes, the multiplier can process data of different bit widths through the same circuit block, and can adapt to different computing needs simply by adjusting the internal configuration. In addition, when switching modes, the system delay can be less than 10 clock cycles. This short delay is mainly due to the update of the hardware configuration and ensuring that there is no residual data from the previous mode in the computing pipeline. Therefore, the system needs to clear the pipeline before and after switching to avoid data confusion.

[0167] For example, different precision modes of the reconfigurable multiplier correspond to different power consumption levels and applicable scenarios:

[0168] Single-precision floating-point mode: Provides the highest precision and is suitable for scientific computing.

[0169] Half-precision floating-point mode: provides medium precision and is suitable for mixed-precision training scenarios.

[0170] 8-bit integer mode: Provides the lowest accuracy but the highest energy efficiency, suitable for inference tasks such as image classification and object detection.

[0171] Optionally, as part of the task descriptor, the precision requirement of the first computing task may indicate the minimum precision mode required to execute the task. For example, an image recognition task may only require 8-bit integer precision, while a scientific computing task may require single-precision floating-point precision.

[0172] Optionally, the control component is also used to determine the computational domain as the third computational domain when there is only one candidate computational domain among the K activated computational domains whose allowed accuracy mode matches the accuracy requirement of the first computational task; and to calculate the computational domain score of the candidate computational domain based on the performance of the candidate computational domain, the power consumption of the candidate computational domain, and the temperature of the candidate computational domain when there are at least two candidate computational domains among the K activated computational domains whose allowed accuracy mode matches the accuracy requirement of the first computational task.

[0173] Optionally, for each candidate computing domain, its computing domain score can be calculated. This score can comprehensively consider the performance of the computing domain (such as the number of instructions per cycle), power consumption and temperature. The computing domain temperature here can be the actual temperature of the computing domain or the predicted temperature of the computing domain at the next moment. This is not limited in this embodiment.

[0174] Alternatively, the scoring formula can reflect the positive correlation between performance and the computational domain score, and the negative correlation between power consumption and temperature and the computational domain score, as shown in formula (5):

[0175] Computational domain score = (performance^2) / (power consumption*temperature) (5)

[0176] Among them, performance is measured by instructions per cycle (IPC), which can reflect the computing efficiency of the computing domain, while power consumption and temperature represent the energy consumption and thermal state of the computing domain.

[0177] Optionally, among at least two candidate computing domains, the candidate computing domain with the highest computing domain score can be determined as the third computing domain, so as to ensure that the task can be allocated to the computing resources with the best energy efficiency ratio while meeting the accuracy requirements, thereby achieving a balance between performance and energy efficiency.

[0178] Through this embodiment, the computational domain of the task is selected based on the accuracy requirement and the computational domain score, which can achieve a balance between performance and energy efficiency, improve resource utilization, and optimize system performance.

[0179] In an exemplary embodiment, one of the plurality of computing units allows for executing a computing task using at least one precision mode, wherein a higher computing precision corresponding to the precision mode results in a greater number of floating-point bits used to execute the computing task; the task attributes of the first computing task include a precision requirement of the first computing task, wherein the precision requirement of the first computing task is used to indicate a minimum precision mode required to execute the first computing task; and the third computing domain includes at least two types of computing units;

[0180] The third computing domain is used to screen computing units from the computing units of the third computing domain according to preset screening conditions, and when at least two candidate computing units are screened out, calculate the computing unit score of the candidate computing unit according to the performance of the candidate computing unit, the power consumption of the candidate computing unit and the temperature of the candidate computing unit; when the accuracy mode allowed by the candidate computing unit with the highest computing unit score among the at least two candidate computing units matches the accuracy requirement of the first computing task, the candidate computing unit with the highest computing unit score is determined as the first computing unit; when the accuracy mode allowed by the candidate computing unit with the highest computing unit score among the at least two candidate computing units does not match the accuracy requirement of the first computing task, the computing unit specified in the third computing domain is determined as the first computing unit; wherein the preset screening conditions include one of the following: when the temperature of the electronic component is greater than or equal to the second temperature threshold, screening computing units with a temperature less than the third temperature threshold; screening computing units that meet the execution constraint conditions; screening computing units whose allowed accuracy mode matches the accuracy requirement of the first computing task;

[0181] The computing unit score of the candidate computing unit is positively correlated with the performance of the candidate computing unit, and negatively correlated with the power consumption of the candidate computing unit and the temperature of the candidate computing unit. The performance of the candidate computing unit is the number of instructions per cycle of the candidate computing unit.

[0182] Similar to the aforementioned embodiment, one of the multiple computing units allows the use of at least one precision mode to execute computing tasks. The higher the computing precision corresponding to the precision mode, the more floating-point bits are used to execute the computing task. The task attributes of the first computing task include the precision requirement of the first computing task, and the precision requirement of the first computing task is used to indicate the minimum precision mode required to execute the first computing task. The third computing domain includes at least two types of computing units, that is, each computing unit supports at least one precision mode to execute computing tasks. The high-precision mode (such as single-precision floating point) uses more floating-point bits, provides higher mathematical precision, and is suitable for application scenarios with extremely high accuracy requirements. The low-precision mode (such as 8-bit integer) uses fewer floating-point bits. Although the precision is lower, it can significantly improve energy efficiency and reduce power consumption in scenarios such as inference.

[0183] Optionally, when selecting a computing unit for performing the first computing task, the computing unit may be selected from the third computing domain according to a preset screening condition. The preset screening condition may include:

[0184] 1. When the temperature of an electronic component is greater than or equal to a second temperature threshold, the system filters out computing units with a temperature less than a third temperature threshold. The second and third temperature thresholds are preset and are used to automatically exclude high-power computing cores when the overall temperature of the electronic component is too high to ensure the stability of the electronic component.

[0185] 2. Select computing units that meet the execution constraints, that is, computing units that perform well in terms of power consumption and temperature, to optimize the overall system energy efficiency.

[0186] 3. Select the computing units whose allowed precision modes match the precision requirements of the first computing task to ensure that the computing units can meet the precision requirements of the task.

[0187] After screening out at least two candidate computing units, the computing unit score of each candidate computing unit can be calculated. The calculation formula can be shown as formula (5), where the computing unit score of the candidate computing unit is positively correlated with the performance of the candidate computing unit and negatively correlated with the power consumption of the candidate computing unit and the temperature of the candidate computing unit. The performance of the candidate computing unit is the number of instructions per cycle of the candidate computing unit.

[0188] Optionally, each computing unit can provide real-time power consumption data through a built-in current sensor, while using distributed thermal diodes to collect temperature data at a sampling rate of 1kHz to provide a basis for scheduling decisions.

[0189] Optionally, under the condition that the accuracy requirement is met, the candidate computing unit with the highest computing unit score will be determined as the first computing unit.

[0190] Optionally, if the computing unit with the highest computing unit score does not match the accuracy requirement of the first computing task, a computing unit preset in the third computing domain that matches the accuracy requirement can be selected as the first computing unit to execute the first computing task.

[0191] Through this embodiment, the dynamic screening and selection strategy of computing units based on accuracy requirements and execution constraints can dynamically find a balance between meeting task accuracy requirements and optimizing overall energy efficiency, ensuring that when processing diverse computing tasks, it can continuously provide high-performance and low-power operating states and optimize system performance.

[0192] In an exemplary embodiment, one of the plurality of computing units allows for performing a computing task using at least one precision mode, wherein a higher precision mode corresponds to a higher computing accuracy, and a greater number of floating-point bits are used to perform the computing task;

[0193] A fourth computing domain among the K activated computing domains is used to extract a second computing task to be executed from a computing task queue of the fourth computing domain, wherein the second computing task has at least two computing stages, and the precision requirements of the second computing task are different in different computing stages of the at least two computing stages, and the precision requirements of the second computing task are used to indicate the minimum precision mode required to execute the second computing task; in different computing stages of the second computing task, the second computing task is executed by adopting different precision modes through at least one computing unit in the fourth computing domain, so that the precision mode adopted by the computing unit of the fourth computing domain matches the precision requirement of the computing stage of the second computing task.

[0194] In the field of high-performance computing and AI acceleration, computing tasks often include multiple stages, and the computing accuracy requirements of each stage may be different. This diversity of requirements requires the ability to flexibly adjust the accuracy mode of its computing units to adapt to the needs of tasks at different stages while achieving the best energy efficiency ratio.

[0195] In this embodiment, each computing unit, such as a CPU, a GPU, etc., can execute computing tasks in different precision modes.

[0196] Optionally, a fourth computing domain among the K activated computing domains can be used to extract a second computing task to be executed from the computing task queue of the fourth computing domain, wherein the second computing task has at least two computing stages, and the precision requirements of the second computing task are different in different computing stages of the at least two computing stages. The precision requirements of the second computing task are used to indicate the minimum precision mode required to execute the second computing task. For example, in an image processing task, the preprocessing stage may only require lower precision calculations (such as 8-bit integers) for image scaling and cropping; while in the feature extraction or classification stage, higher precision calculations (such as single-precision floating points) may be required to ensure the accuracy and robustness of the model.

[0197] Optionally, the precision requirement of each computation stage can be used to indicate the minimum precision mode required to execute that stage.

[0198] Optionally, in each different computing stage of the second computing task, the second computing task can be executed using different precision modes through at least one computing unit in the fourth computing domain. The task can be executed using a precision mode that matches the precision requirements of that stage, so that the precision mode can be dynamically adjusted in different computing stages, thereby ensuring the accuracy of the calculation results in each stage while reducing unnecessary power consumption and computing resource consumption.

[0199] For example, taking continuous face detection on a mobile AI camera as an example, the computing mode can be dynamically switched according to the needs of different stages: in the initial stage, since high precision is required at this stage, the system uses the GPU to run a half-precision floating-point model to ensure the accuracy of face positioning. At this time, the computing unit score is 15TOPS / W; in the stable stage, that is, when the detection results of 10 consecutive frames are consistent, it means that low-precision calculations can also meet the needs in the current scenario, and it can automatically switch to the NPU to run an 8-bit integer quantization model, and the computing unit score is increased to 42TOPS / W; when exception processing is triggered, such as if the detection confidence drops by more than 5%, it can immediately fall back to half-precision floating-point mode, and trigger the recalculation of the precision-energy efficiency weight to ensure the accuracy of detection.

[0200] Through this embodiment, by matching the most suitable precision mode for different calculation stages, the calculation efficiency and accuracy of each stage can be ensured, performance bottlenecks caused by insufficient precision or waste of resources can be avoided, and efficient resource utilization and maximum energy efficiency can be achieved.

[0201] In an exemplary embodiment, an activated computing domain is configured with multiple load levels, and a load level prediction engine is run on a control component. The load level prediction engine is used to predict the load level of the computing domain corresponding to the input prediction reference data in M ​​future cycles based on input prediction reference data. The prediction reference data includes the following data of the computing domains in the computing domain cluster in N historical cycles: instruction mixing, cache miss rate, and task queue depth of the computing task queue, where M is a positive integer greater than or equal to 1, and N is a positive integer greater than or equal to 1; wherein,

[0202] The control component is also used to obtain the predicted reference data of the activated computing domain, input the predicted reference data of the activated computing domain into the load level prediction engine, obtain the load level of the activated computing domain in M ​​future cycles predicted by the load level prediction engine, and pre-adjust the voltage of the activated computing domain and the frequency of the activated computing domain based on the predicted load level of the activated computing domain in M ​​future cycles.

[0203] In this embodiment, the control component can obtain the operating indicators of the activated computing domain in the last N cycles as prediction reference data. These indicators may include: instruction mixing, which is used to reflect the proportion and combination of different types of instructions (such as integer, floating point, and memory access) to evaluate the computing requirements and load characteristics of the computing domain; cache miss rate, which is used to indicate the efficiency of cache access. A higher cache miss rate means frequent data exchange, and a higher frequency may be required to compensate for the resulting delay; the task queue depth of the computing task queue, which is used to indicate the current state of the task queue. The larger the queue depth, the more tasks to be processed, and the higher the load of the computing domain may be.

[0204] Optionally, the selection of the N value may depend on the characteristics of the task and the system's response requirements to the historical cycle. Typically, N is a positive integer greater than or equal to 1, meaning that the reference data covers at least one complete cycle to capture the dynamic changes of recent operations.

[0205] Optionally, the control component can be used to input the collected prediction reference data into a load level prediction engine. This engine can be based on a machine learning model, such as a recurrent neural network, a long short-term memory network, or other time series analysis model. This is not limited in this embodiment. It can be used to predict the load level of the computing domain in the next M cycles, where M is also a positive integer greater than or equal to 1, representing the predicted time window.

[0206] Optionally, the load level prediction engine can be a quantized 2-layer LSTM network running using a dedicated tensor accelerator. The quantized LSTM network can convert data types such as weights and activation functions from floating point numbers to fixed point numbers, thereby reducing computational complexity and memory usage while maintaining the network's predictive capability.

[0207] Optionally, the number of future cycles that the load level prediction engine can predict can be predetermined based on experiments.

[0208] Optionally, based on the prediction results output by the load level prediction engine, the voltage and frequency of the activated computing domain can be pre-adjusted. For example, the voltage or frequency can be pre-adjusted 50-100 clock cycles in advance. Through pre-adjustment, the voltage and frequency can be increased before the load actually increases to prevent performance degradation. Correspondingly, when it is predicted that the load will be reduced, the voltage and frequency can also be lowered in advance to save power consumption.

[0209] Similar to the aforementioned embodiment, traditional DVFS responds to load changes with a certain delay, but prediction-based pre-adjustment can significantly reduce this delay and improve the overall response speed and stability of the system.

[0210] Optionally, based on the prediction results, the voltage can be adjusted by the Power Management Unit (PMU), which can follow a predefined load level and voltage-frequency correspondence table (as shown in Table 1) to ensure that computing needs are met while minimizing power consumption.

[0211] Table 1

[0212]

[0213] In addition to voltage regulation, the frequency can also be fine-tuned through a digital phase-locked loop to adapt to the predicted load level, ensuring that the computing unit operates in the most appropriate state.

[0214] Through this embodiment, by pre-adjusting the calculation domain voltage and frequency based on the load level prediction engine, the hysteresis and resource allocation problems in dynamic voltage and frequency adjustment can be solved, and the timeliness and flexibility of the operation control of electronic components can be improved.

[0215] In an exemplary embodiment, the control component is also used to control the activated computing domain to enter a near-threshold computing mode when the number of instructions per cycle of the activated computing domain is less than or equal to a second instruction number threshold and the temperature of the activated computing domain is less than or equal to a fourth temperature threshold; in the near-threshold computing mode, the voltage offset of the activated computing domain is adjusted based on the operating status data of the activated computing domain, wherein the operating status data of the activated computing domain is used to indicate the temperature of the activated computing domain, the voltage of the activated computing domain and the process deviation coefficient of the activated computing domain, and the voltage offset of the activated computing domain is obtained by summing the difference between the temperature of the activated computing domain and the specified temperature and the product of the third coefficient, the difference between the voltage of the activated computing domain and the specified voltage and the product of the fourth coefficient, and the product of the process deviation coefficient and the fifth coefficient.

[0216] Similar to the aforementioned embodiment, in the hierarchical adaptive energy efficiency management architecture, a near-threshold computing mode (i.e., a near-threshold voltage mode) can be introduced. In this mode, the operating voltage of the computing unit is very close to the threshold voltage of the transistor. Therefore, the voltage needs to be carefully controlled to avoid timing errors caused by factors such as process deviations and temperature fluctuations.

[0217] Optionally, the number of instructions per cycle and the temperature of the activated computing domain can be continuously monitored. When the IPC is less than or equal to a third instruction count threshold (that is, the computing load of the current computing domain is low, and there is enough room to reduce the operating voltage without affecting performance) and the temperature is less than or equal to a third temperature threshold (ensuring that the computing domain runs in a cool state to reduce the risk of timing errors caused by temperature increase), the computing domain can be controlled to enter a near-threshold computing mode to further reduce power consumption.

[0218] Optionally, in the near-threshold calculation mode, the operating voltage can be fine-tuned according to the real-time operating status data of the calculation domain to achieve the best energy efficiency ratio while ensuring the reliability of the calculation. Here, the operating status data can mainly include: temperature, voltage, and process deviation coefficient.

[0219] Optionally, a voltage offset (delta_V) can be calculated based on the above parameters to guide voltage fine-tuning. The calculation formula of the voltage offset is shown in formula (6):

[0220] (6)

[0221] Among them, k1, k2, and k3 are the third coefficient, the fourth coefficient, and the fifth coefficient, respectively, which represent the degree of influence of temperature, voltage, and process deviation on the voltage offset.

[0222] Optionally, the voltage of the computational domain is dynamically adjusted according to the calculated voltage offset, so that a suitable operating point can be found on the boundary between safety and energy efficiency.

[0223] Optionally, after adjustment, the timing status of the computing domain can be re-checked to verify the stability of the system. If no timing violations or errors are detected within the next N cycles (N is usually set to 10 or more), the system state is considered stable and the task can continue to be executed under the current voltage and computing mode. If errors are detected continuously, further voltage adjustment can be triggered or fall back to a safer operating mode (such as an operating mode under standard voltage) to avoid data corruption or computing failure.

[0224] Through this embodiment, by introducing a near-threshold computing mode and performing voltage fine-tuning in this mode, it is possible to achieve extremely low power consumption operation and reduce resource loss while ensuring the correct execution of computing tasks.

[0225] In an exemplary embodiment, the electronic component further comprises: a main latch and a shadow latch, wherein the main latch and the shadow latch have different trigger edges; wherein,

[0226] The main latch and the shadow latch are used to sample data from the activated computing domain when the activated computing domain enters a near-threshold computing mode;

[0227] The control component is also used to sample data from the activated computing domain through the main latch and the shadow latch, and perform timing error detection on the activated computing domain, wherein the trigger edges of the main latch and the shadow latch are different; when a timing error is detected in the activated computing domain, the voltage of the activated computing domain is continuously increased until the error instruction is successfully executed, wherein the error instruction is an instruction executed by the activated computing domain and corresponds to the timing error in the activated computing domain; or, the activated computing domain is controlled to exit the near-threshold computing mode.

[0228] In this embodiment, in the near-threshold calculation mode, the data sampling executed in the calculation domain can be monitored in real time through the comparison mechanism of the main latch and the shadow latch. Here, the trigger edges of the main latch and the shadow latch are different. For example, the main latch (positive edge triggered) and the shadow latch (negative edge triggered) can be triggered by the rising edge and falling edge of the clock pulse respectively in the same clock cycle to sample the data signal. Under normal circumstances, since they sample the same signal at the same time, the output results should be consistent. However, if after the sampling is completed, the output results of the main latch and the shadow latch are compared, and the difference in the results exceeds a preset tolerance (for example, ±3%), it can be considered that a timing violation has occurred, that is, Error=1, indicating that the timing consistency of the data sampling has been destroyed.

[0229] Optionally, the Razor trigger can monitor and determine the occurrence of timing errors in real time by detecting the inconsistency between the outputs of the main latch and the shadow latch.

[0230] Optionally, once a timing error is detected, action can be taken to prevent the error from further propagating. For example, a pipeline stall can be immediately triggered to freeze the execution of the current and subsequent instructions, preventing the result of the erroneous instruction from being passed to the next stage of the computing process, thereby preventing the error from spreading.

[0231] Optionally, during a pipeline stall, subsequent instruction execution will be suspended until it is confirmed that the error has been corrected.

[0232] Optionally, it is possible to trace back to the most recent correct state, that is, use the correct instruction copy pre-stored in the buffer and re-execute to ensure the continuity and correctness of the calculation.

[0233] In this embodiment, the required voltage compensation amount (delta_V) can be calculated based on the current PVT parameters to restore the timing consistency of the transistor. This adjustment process is shown in formula (7):

[0234] (7)

[0235] Among them, k1, k2, and k3 are compensation coefficients related to temperature, voltage, and process deviation coefficients, respectively.

[0236] Through the above dynamic voltage regulation process, the voltage can be gradually adjusted to a stable operating point to prevent the recurrence of timing errors. For example, the voltage can be gradually increased in steps of 10mV / cycle until the erroneous instruction can be successfully executed.

[0237] Optionally, if successive voltage regulation attempts fail to stabilize, or if timing errors are frequently detected over a period of time, it can be determined that the risk of operating in NTC mode is too high. In this case, the voltage of the activated computing domain can be controlled to rise to a safer level, and the near-threshold computing mode can be exited, returning to normal operating mode to ensure the normal operation of the computing platform under all conditions.

[0238] In this embodiment, a hardware-accelerated load prediction engine combined with Razor trigger protection enables sub-millisecond response. This architecture achieves high-precision energy efficiency regulation through hardware-accelerated key control loops (prediction engine and error recovery mechanism) and a hierarchical power supply design. Predictive regulation reduces latency penalties compared to traditional DVFS, and ensures reliability in NTC mode through detection circuits, balancing recovery speed and energy efficiency with voltage regulation step size.

[0239] Through this embodiment, by detecting timing errors in real time in the near-threshold calculation mode and performing error recovery through a dynamic voltage regulation strategy, when faced with an unrecoverable situation, the near-threshold calculation mode is exited in time and switched to a safer operating state, effectively avoiding the risk of data corruption and calculation failure.

[0240] As an optional embodiment, different computing domains in the computing domain cluster are configured with different activation levels. The higher the activation level, the higher the computing power of the computing domain. The highest activation level among the activation levels of the computing domains in the computing domain cluster is the first activation level.

[0241] The control component is also used to continuously monitor the following information of the K activated computing domains: the number of instructions per cycle, the task queue depth of the computing task queue, wherein the highest activation level among the activation levels of the K activated computing domains is the second activation level; when the second activation level is not the first activation level, when the value of the number of instructions per cycle of the activated computing domain of the second activation level is greater than the third instruction number threshold and the task queue depth of the computing task queue of the activated computing domain of the second activation level is greater than the specified depth threshold, activate the computing domain of the third activation level in the computing domain cluster, wherein the third activation level is the next activation level of the second activation level.

[0242] In this embodiment, different computing domains in the computing domain cluster can be configured with different activation levels. The higher the activation level, the higher the computing power of the computing domain. The highest activation level among the activation levels of the computing domains in the computing domain cluster is the first activation level.

[0243] Exemplarily, the computing domain corresponding to the lowest activation level can be set as the zeroth activation level domain, that is, the L0 domain. The L0 domain can be in a normally active state, always powered, and responsible for basic control, sensor data acquisition and simple tasks (such as background synchronization); the computing domains corresponding to the secondary activation level and higher activation levels can be set as L1, L2, L3...Ln domains, respectively, where the L1-Ln domains can be turned off and can be sorted according to the system topology.

[0244] For example, Figure 2 As shown, the computing units on the electronic components can be divided into computing domain clusters, and the computing domains in the computing domain clusters can be configured as different activation levels, for example, the zeroth activation level domain, the second activation level domain, the third activation level domain, etc. The above computing domains can all be connected to the control components (including input / output interfaces, double data rate memory controllers, general processors, and central arbiters) through the on-chip interconnection network.

[0245] Optionally, the activation level of the computing domain refers to the level of computing power that can be provided when the computing domain is activated. Generally, the higher the activation level, the stronger the computing power of the computing domain, but the corresponding power consumption and heat may also increase accordingly.

[0246] Optionally, in the initial state, only the L0 domain and necessary input / output (I / O) domains may be active, and other domains are power-gated (ie, the supply voltage is 0 and the clock is turned off).

[0247] Optionally, if the L0 domain cannot meet the load and computing power requirements of the task, primary activation can be performed to activate the L1 (corresponding to the second activation level) domain. At this time, the central arbitrator can send an activation signal to the power controller of the L1 domain. The power controller can release the power gating (for example, turn on the LDO, and the voltage rises from 0V to 0.7V, which takes about 200ns) and release the clock gating (for example, the DPLL is locked to 1.0GHz, which takes 150ns). When the L1 domain has been activated, the control component can migrate the basic tasks to the L1 domain.

[0248] For example, the activation order of computed fields can be as follows Figure 3 As shown, when the trigger condition is met, the first activation level domain (i.e., L1 domain) is activated, and the load is continuously monitored to determine whether it meets the demand. If it is still insufficient, the next level of computing domain is activated, and this process is repeated until the load is balanced.

[0249] Optionally, performance indicators of the K activated computing domains may be monitored in real time, including the number of instructions per cycle and the task queue depth of the computing task queue. These indicators reflect the computing efficiency and task load of the computing domains.

[0250] Optionally, when the second activation level is not the first activation level (i.e., there are computational domains that have not yet activated the highest level), it can be determined based on the performance indicators of the K activated computational domains monitored in real time whether the activated computational domains of the second activation level meet the following two conditions:

[0251] 1. The number of instructions per cycle is greater than the third instruction count threshold, indicating that the computing resources in the compute domain are close to saturation and may require more computing power.

[0252] 2. The task queue depth of the computing task queue is greater than the specified depth threshold, which means that there are a large number of tasks waiting to be processed in the task queue and the current computing power is insufficient to respond quickly.

[0253] Optionally, when the above conditions are met, a computing domain of a third activation level in the computing domain cluster may be further activated. The third activation level here is the next activation level after the second activation level, that is, a computing domain with stronger computing power.

[0254] Optionally, if the load of a computing domain drops below a preset threshold, a degradation process can be initiated to reallocate computing tasks to a computing domain with lower power consumption to save energy.

[0255] Similarly, when a higher-level computing domain (such as L2 or L3 domain) needs to be activated, the PMU can also quickly adjust the voltage and frequency to the predetermined values, such as setting the L2 domain to 0.9V / 2.0GHz.

[0256] Optionally, voltage and frequency settings corresponding to different load levels may be predefined by a table lookup method to minimize power consumption during regulation. For example, the table defined here may be the same as Table 1 in the aforementioned embodiment.

[0257] Optionally, when a computing task is migrated, in order to ensure the continuity of the task, the computing domain can load the complete computing context before receiving the task, including but not limited to the current state of the task, processed data, and cache information, so as to continue execution from the interruption point of the previous processing unit.

[0258] Optionally, when the IPC of a computing domain is continuously lower than a preset threshold (for example, 0.3 indicates light load or idle) for more than N cycles (for example, 1000 cycles), the sleep process of the computing domain can be triggered. At this time, the control component can migrate the unfinished tasks on the computing domain to a computing domain with a low load level, and then freeze the clock signal, save the status of key registers on the computing domain to the retention register for rapid restoration of the status when subsequently activated, and apply power gating to reduce the voltage of the computing domain to 0V, entering a deep sleep state, thereby reducing the power consumption of the system.

[0259] Through this embodiment, the computing domain can be dynamically activated through the computing domain dynamic expansion strategy based on the activation level, which not only meets the computing needs but also realizes the effective utilization of resources and improves the flexibility and adaptability of the system.

[0260] As an optional embodiment, the computing task executed by the electronic component is configured with a corresponding task priority, and the task priority of the computing task executed by the electronic component corresponds to the activation level of the computing domain in the computing domain cluster;

[0261] The control component is also used to activate the fifth computing domain in the computing domain cluster whose activation level corresponds to the task priority of the third computing task when there is a third computing task to be assigned and the activation levels of the K activated computing domains are all lower than the activation level corresponding to the task priority of the third computing task, and assign the third computing task to the fifth computing domain.

[0262] In this embodiment, tasks can be allocated based on the correspondence between task priorities and activation levels of computing domains. Here, each computing task can be configured with a specific task priority for the urgency and importance of the task. For example, real-time response, critical business processing, or user interaction may belong to high-priority scenarios, while background updates, data preprocessing, etc. may belong to low-priority scenarios.

[0263] Similar to the aforementioned embodiment, the activation level of the computing domain is proportional to the computing power it can provide. A higher activation level means higher voltage and frequency settings of the computing domain, thereby providing stronger computing power, but also consuming more energy.

[0264] Optionally, the correspondence between task priority and computing domain activation level can be predefined. Then, when a high-priority task arrives, it can be compared with the activation level of the currently activated computing domain to ensure that the task is executed on a computing domain with sufficient computing power level.

[0265] For example, there is a third computing task to be assigned, and its task priority requires a computing domain that provides a higher level of computing power than the currently K activated computing domains (the activation levels of which are all lower than the activation levels corresponding to the task priority of the third computing task). In this case, the computing domain with the fifth activation level in the computing domain cluster can be automatically activated. This level just matches the task priority of the third computing task. In other words, the fifth computing domain can provide sufficient computing resources to meet the requirements of the third computing task. Subsequently, the third computing task will be directly assigned to the fifth computing domain for execution, skipping the steps that require layer-by-layer activation level matching to ensure that the task can respond quickly and be completed efficiently.

[0266] In this embodiment, for high-priority tasks, the conventional activation level sequence can be skipped and the highest-level computing domain can be directly activated, which can ensure that critical tasks can receive immediate responses and avoid delays caused by untimely preparation of computing resources.

[0267] Optionally, when multiple tasks simultaneously request to use computing resources of the same computing domain, the priority of the tasks may be determined based on the ratio of the performance requirements of the tasks to the power consumption of the current domain.

[0268] Optionally, in complex computing systems, the dynamic allocation and activation of resources may encounter delay issues, such as activation signal transmission, power management unit (PMU) response time, clock tree stability, etc., which may increase the activation delay. Long activation delays may cause resource allocation deadlock, that is, multiple tasks or modules wait for each other to release resources, and thus cannot execute normally. In order to prevent the occurrence of deadlock, a maximum activation delay constraint (such as 100μs) can be set in advance. If the timeout exceeds, all domains will be forced to activate and an alarm will be issued.

[0269] Through this embodiment, by activating the computing domain based on the correspondence mechanism between task priority and computing domain activation level, timely scheduling and response of computing resources can be achieved, thereby improving the stability and reliability of the system when processing sudden computing needs.

[0270] The embodiments of the present application also provide an operation control method for an electronic component, which can be implemented by the hardware of the above embodiments and preferred embodiments, and will not be repeated here. The electronic component includes multiple computing units, which are divided into computing domain clusters. The computing domains in the computing domain cluster are sets of computing units that independently adjust operating parameters. In the computing domain cluster, a computing domain includes some computing units from the multiple computing units, and different computing domains include different computing units. Figure 4 is a flow chart of an optional method for controlling the operation of an electronic component according to an embodiment of the present application, such as Figure 4 As shown, the method includes:

[0271] Step S402: When there are K activated computing domains in the computing domain cluster, a set of status parameters of the activated computing domains is monitored, where K is a positive integer greater than or equal to 1.

[0272] Step S404: If there is a first computing domain among the K activated computing domains whose parameter value of a specified state parameter in a set of state parameters meets the corresponding threshold condition, the operating parameters of the first computing domain are adjusted according to the adjustment method corresponding to the specified state parameter.

[0273] It can be understood that the above steps S402 and S404 can be executed by the control component 1011 in the aforementioned embodiment. The specific execution process has been described in the aforementioned embodiment and will not be repeated here.

[0274] Through the above method, when there are K activated computing domains in the computing domain cluster, a set of state parameters of the activated computing domains are monitored, where K is a positive integer greater than or equal to 1; when there is a first computing domain in the K activated computing domains whose parameter value of a specified state parameter in a set of state parameters meets the corresponding threshold condition, the operating parameters of the first computing domain are adjusted according to the adjustment method corresponding to the specified state parameter, which can solve the problem of poor timeliness of operation control of electronic components in related technologies and achieve the technical effect of improving the timeliness and flexibility of operation control of electronic components.

[0275] In an exemplary embodiment, a set of state parameters includes the number of instructions per cycle, and the threshold condition corresponding to the number of instructions per cycle is that the value of the number of instructions per cycle is less than or equal to a first instruction number threshold; according to the adjustment method corresponding to the specified state parameters, the operating parameters of the first computing domain are adjusted, including: when the specified state parameters include the number of instructions per cycle, performing a first adjustment operation on the first computing domain to adjust the operating parameters of the first computing domain, wherein the first adjustment operation includes: lowering the frequency of the first computing domain and lowering the voltage of the first computing domain.

[0276] In an exemplary embodiment, the electronic component further includes a hardware performance counter; the above method further includes: monitoring the number of instructions per cycle of the K activated computing domains in real time through the hardware performance counter.

[0277] In an exemplary embodiment, a set of state parameters includes a computing domain temperature, and a threshold condition corresponding to the computing domain temperature is: a temperature value of the computing domain temperature is greater than or equal to a first temperature threshold; operating parameters of the first computing domain are adjusted according to an adjustment method corresponding to the specified state parameters, including: when the specified state parameters include the computing domain temperature, performing a second adjustment operation on the first computing domain to adjust the operating parameters of the first computing domain, wherein the second adjustment operation includes at least one of the following: lowering the frequency of the first computing domain and performing voltage compensation on the second computing domain, wherein the second computing domain is an activated computing domain adjacent to the first computing domain and does not meet the threshold condition corresponding to the computing domain temperature; migrating the task executed by the first computing domain to the activated computing domain that does not meet the threshold condition corresponding to the computing domain temperature.

[0278] In an exemplary embodiment, a second adjustment operation is performed on the first calculation domain, including: lowering the frequency of the first calculation domain by a specified frequency value, wherein the specified frequency value is the product of a first coefficient and a specified temperature difference value, and the specified temperature difference value is the difference between the temperature value of the calculation domain temperature of the first calculation domain and the first temperature value; and increasing the voltage of the second calculation domain by a specified voltage value, wherein the specified voltage value is the product of the second coefficient, the specified difference value and a specified voltage compensation value, and the specified difference value is 1 minus the difference obtained by subtracting the ratio of the temperature value of the calculation domain temperature of the second calculation domain to the temperature value of the calculation domain temperature of the first calculation domain.

[0279] In an exemplary embodiment, the electronic component also includes a distributed sensor network; monitoring a set of state parameters of the activated computing domain, including: when a set of state parameters includes the computing domain temperature, parsing the sensor data corresponding to the activated computing domain to obtain the actual temperature value of the activated computing domain at the current moment, wherein the sensor data corresponding to the activated computing domain is the sensor data obtained by the sensor network by detecting the activated computing domain separately; based on the actual temperature value of the activated computing domain at the current moment, the current power value of the activated computing domain and the ambient temperature value at the current moment, predicting the temperature value of the activated computing domain at the next moment after the current moment, wherein the current power value of the activated computing domain is the power value of the activated computing domain at the current moment, and the temperature value of the computing domain temperature of the activated computing domain is the predicted temperature value of the activated computing domain at the next moment.

[0280] In an exemplary embodiment, the temperature value of the activated computing domain at the next moment after the current moment is predicted based on the actual temperature value of the activated computing domain at the current moment, the current power value of the activated computing domain and the ambient temperature value at the current moment, including: summing the product of the actual temperature value of the activated computing domain at the current moment and the thermal inertia coefficient, the product of the current power value of the activated computing domain and the power-temperature conversion coefficient, and the product of the ambient temperature value at the current moment and the specified heat exchange rate, to determine the predicted temperature value of the activated computing domain at the next moment, wherein the specified heat exchange rate is the heat exchange rate between the electronic component and the environment in which the electronic component is located.

[0281] In an exemplary embodiment, the electronic component further includes a control component, the sensor network includes a set of sensor nodes in a grid layout, and a computing domain in the computing domain cluster corresponds to some of the sensor nodes in the sensor node set; the above method further includes: obtaining sensor data detected by the sensor nodes corresponding to K activated computing domains by the control component using address polling to obtain sensor data corresponding to the K activated computing domains.

[0282] In an exemplary embodiment, the electronic component also includes a temperature sensor; the above method also includes: periodically performing ambient temperature detection through the temperature sensor, and calculating the temperature difference between the detected ambient temperature value and the recorded reference ambient temperature value; when the temperature difference between the detected ambient temperature value and the recorded reference ambient temperature value is greater than or equal to the temperature difference threshold, the product of the detected ambient temperature value and the specified coefficient is added to the second temperature value, and determined as the updated first temperature threshold, and the recorded reference ambient temperature value is updated to the detected ambient temperature value.

[0283] In an exemplary embodiment, the above method also includes: in the case where there is a first computing task to be assigned, selecting a third computing domain from the K activated computing domains to execute the first computing task based on the task label of the first computing task and the operating characteristic data of the K activated computing domains, wherein the task label of the first computing task is used to indicate the task attributes of the first computing task and the execution constraints of the first computing task; assigning the first computing task to the third computing domain so that the first computing task is executed by the first computing unit of the third computing domain.

[0284] In an exemplary embodiment, the operating status data of the activated computing domain includes the historical performance data of the activated computing domain; based on the task label of the first computing task and the operating characteristic data of K activated computing domains, a third computing domain that performs the first computing task is selected from the K activated computing domains, including: inputting the task label of the first computing task and the historical performance data of the activated computing domain into the task level prediction engine to obtain the task level of the first computing task predicted by the task level prediction engine, wherein the task level of the first computing task is used to characterize the computing power requirement for performing the first computing task; and selecting an activated computing domain that matches the task level of the first computing task from the K activated computing domains to obtain the third computing domain.

[0285] In an exemplary embodiment, a computing unit among multiple computing units allows at least one precision mode to be used to execute a computing task, and the higher the computing precision corresponding to the precision mode, the more floating-point bits used to execute the computing task; the task attributes of the first computing task include the precision requirement of the first computing task, and the precision requirement of the first computing task is used to indicate the minimum precision mode required to execute the first computing task; from K activated computing domains, an activated computing domain that matches the task level of the first computing task is selected to obtain a third computing domain, including: when there are at least two candidate computing domains in the K activated computing domains whose allowed precision modes match the precision requirement of the first computing task, the computing domain score of the candidate computing domain is calculated according to the performance of the candidate computing domain, the power consumption of the candidate computing domain, and the temperature of the candidate computing domain, wherein the computing domain score of the candidate computing domain is positively correlated with the performance of the candidate computing domain and negatively correlated with the power consumption of the candidate computing domain and the temperature of the candidate computing domain, and the performance of the candidate computing domain is the number of instructions per cycle of the candidate computing domain; and the candidate computing domain with the highest computing domain score among the at least two candidate computing domains is determined as the third computing domain.

[0286] In an exemplary embodiment, one of the plurality of computing units allows at least one precision mode to be used to execute a computing task, and the higher the computing precision corresponding to the precision mode, the more floating-point bits used to execute the computing task; the task attributes of the first computing task include the precision requirement of the first computing task, and the precision requirement of the first computing task is used to indicate the minimum precision mode required to execute the first computing task; the third computing domain includes at least two types of computing units; the above method also includes: according to preset screening conditions, filtering computing units from the computing units of the third computing domain, wherein the preset screening conditions include one of the following: when the temperature of the electronic component is greater than or equal to the second temperature threshold, filtering computing units whose temperature is less than the third temperature threshold; filtering computing units that meet the execution constraints; filtering computing units whose allowed precision modes match the precision requirements of the first computing task; after filtering out at least two candidates When selecting a computing unit, a computing unit score of the candidate computing unit is calculated based on the performance of the candidate computing unit, the power consumption of the candidate computing unit, and the temperature of the candidate computing unit, wherein the computing unit score of the candidate computing unit is positively correlated with the performance of the candidate computing unit and negatively correlated with the power consumption of the candidate computing unit and the temperature of the candidate computing unit, and the performance of the candidate computing unit is the number of instructions per cycle of the candidate computing unit; if, among at least two candidate computing units, the precision mode allowed to be adopted by the candidate computing unit with the highest computing unit score matches the precision requirement of the first computing task, the candidate computing unit with the highest computing unit score is determined as the first computing unit; if, among at least two candidate computing units, the precision mode allowed to be adopted by the candidate computing unit with the highest computing unit score does not match the precision requirement of the first computing task, the computing unit specified in the third computing domain is determined as the first computing unit.

[0287] In an exemplary embodiment, one of the multiple computing units allows the use of at least one precision mode to execute a computing task, and the higher the computing precision corresponding to the precision mode, the more floating-point bits used to execute the computing task; the above method also includes: extracting a second computing task to be executed from the computing task queue of the fourth computing domain among the K activated computing domains, wherein the second computing task has at least two computing stages, and the precision requirements of the second computing task are different in different computing stages of the at least two computing stages, and the precision requirements of the second computing task are used to indicate the minimum precision mode required to execute the second computing task; in different computing stages of the second computing task, different precision modes are used by at least one computing unit in the fourth computing domain to execute the second computing task, so that the precision mode adopted by the computing unit of the fourth computing domain matches the precision requirement of the computing stage of the second computing task.

[0288] In an exemplary embodiment, the activated computing domain is configured with multiple load levels; the above method also includes: obtaining predicted reference data of the activated computing domain, wherein the predicted reference data includes the instruction mix, cache miss rate and task queue depth of the computing task queue of the activated computing domain in N historical cycles, where N is a positive integer greater than or equal to 1; inputting the predicted reference data into the load level prediction engine to obtain the load level of the activated computing domain in M ​​future cycles predicted by the load level prediction engine, wherein M is a positive integer greater than or equal to 1; based on the predicted load level of the activated computing domain in M ​​future cycles, pre-adjusting the voltage of the activated computing domain and the frequency of the activated computing domain.

[0289] In an exemplary embodiment, the above method also includes: when the number of instructions per cycle of the activated computing domain is less than or equal to the second instruction number threshold and the temperature of the activated computing domain is less than or equal to the fourth temperature threshold, controlling the activated computing domain to enter a near-threshold computing mode; in the near-threshold computing mode, adjusting the voltage offset of the activated computing domain based on the operating status data of the activated computing domain, wherein the operating status data of the activated computing domain is used to indicate the temperature of the activated computing domain, the voltage of the activated computing domain and the process deviation coefficient of the activated computing domain, and the voltage offset of the activated computing domain is obtained by summing the difference between the temperature of the activated computing domain and the specified temperature and the product of the third coefficient, the difference between the voltage of the activated computing domain and the specified voltage and the product of the fourth coefficient, and the product of the process deviation coefficient and the fifth coefficient.

[0290] In an exemplary embodiment, after controlling the activated computing domain to enter the near-threshold computing mode, the above method also includes: in the near-threshold computing mode, sampling data on the activated computing domain through the main latch and the shadow latch, and performing timing error detection on the activated computing domain, wherein the trigger edges of the main latch and the shadow latch are different; when a timing error is detected in the activated computing domain, continuously increasing the voltage of the activated computing domain until the error instruction is successfully executed, wherein the error instruction is an instruction executed by the activated computing domain and corresponds to the timing error in the activated computing domain; or, controlling the activated computing domain to exit the near-threshold computing mode.

[0291] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus the necessary general hardware platform, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method.

[0292] The above is a detailed introduction to an electronic component and an operation control method of an electronic component provided by the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core ideas of the present application. It should be pointed out that for ordinary technicians in this technical field, without departing from the principles of the present application, several improvements and modifications can be made to the present application, and these improvements and modifications also fall within the scope of protection of the claims of the present application.

Claims

1. An electronic component, characterized in that The system comprises a control component and a plurality of computing units, wherein the plurality of computing units are divided into a computing domain cluster, wherein a computing domain in the computing domain cluster is a set of computing units that independently adjust operating parameters; In the computing domain cluster, one computing domain includes some computing units among the multiple computing units, and different computing domains include different computing units; wherein, The control component is configured to, when there are K activated computing domains in the computing domain cluster, monitor a set of state parameters of the activated computing domains, where K is a positive integer greater than or equal to 1; and, when there is a first computing domain among the K activated computing domains whose parameter value of a specified state parameter in the set of state parameters meets a corresponding threshold condition, adjust the operating parameters of the first computing domain according to the adjustment method corresponding to the specified state parameter; Wherein, the set of status parameters includes the number of instructions per cycle, and the threshold condition corresponding to the number of instructions per cycle is that the value of the number of instructions per cycle is less than or equal to the first instruction number threshold; the control component is also used to perform a first adjustment operation on the first computing domain when the specified status parameter includes the number of instructions per cycle, so as to adjust the operating parameters of the first computing domain, wherein the first adjustment operation includes: lowering the frequency of the first computing domain and lowering the voltage of the first computing domain; the electronic component also includes: a hardware performance counter, which is used to monitor the number of instructions per cycle of the K activated computing domains in real time.

2. The electronic component according to claim 1, wherein The electronic component further includes at least one of the following: A voltage regulating unit is configured for each computational domain in the computational domain cluster, and is used to regulate the voltage of the corresponding computational domain, wherein the operating parameters of the computational domains in the computational domain cluster include the voltages of the computational domains in the computational domain cluster, and the voltage regulating unit includes at least one of the following: a low-dropout regulator, a switched capacitor converter; A local clock tree synthesis element, an asynchronous bridge, and a digital phase-locked loop configured respectively for the computational domains in the computational domain cluster, wherein the local clock tree synthesis element is used to distribute the global clock signal of the electronic component to different computational domains in the computational domain cluster to form clock domains corresponding to different computational domains in the computational domain cluster, the asynchronous bridge establishes a data transmission path between different clock domains, the digital phase-locked loop is used to adjust the frequency of the corresponding computational domain, and the operating parameters of the computational domains in the computational domain cluster include the frequency of the computational domains in the computational domain cluster.

3. The electronic component according to claim 1, wherein The control component and the computing units in the computing domains of the computing domain cluster are connected via an on-chip interconnection network.

4. The electronic component according to claim 1, wherein The set of state parameters includes a calculation domain temperature, and a threshold condition corresponding to the calculation domain temperature is: a temperature value of the calculation domain temperature is greater than or equal to a first temperature threshold; The control component is also used to perform a second adjustment operation on the first computing domain to adjust the operating parameters of the first computing domain when the specified state parameters include the computing domain temperature, wherein the second adjustment operation includes at least one of the following: lowering the frequency of the first computing domain and performing voltage compensation on the second computing domain, wherein the second computing domain is the activated computing domain that is adjacent to the first computing domain and does not meet the threshold condition corresponding to the computing domain temperature; migrating the task executed by the first computing domain to the activated computing domain that does not meet the threshold condition corresponding to the computing domain temperature.

5. The electronic component according to claim 4, wherein The control component is further used to lower the frequency of the first calculation domain by a specified frequency value, wherein the specified frequency value is the product of a first coefficient and a specified temperature difference, and the specified temperature difference is the difference between the temperature value of the calculation domain temperature of the first calculation domain and the first temperature value; and to increase the voltage of the second calculation domain by a specified voltage value, wherein the specified voltage value is the product of a second coefficient, a specified difference and a specified voltage compensation value, and the specified difference is 1 minus the difference obtained by subtracting the ratio of the temperature value of the calculation domain temperature of the second calculation domain to the temperature value of the calculation domain temperature of the first calculation domain.

6. The electronic component according to claim 4, wherein The electronic components also include a distributed sensor network; wherein, The sensor network is configured to detect the temperatures of the activated computing domains respectively when the set of state parameters includes the computing domain temperature, and obtain sensor data corresponding to the activated computing domains; The control component is also used to parse the sensor data corresponding to the activated computing domain to obtain the actual temperature value of the activated computing domain at the current moment; based on the actual temperature value of the activated computing domain at the current moment, the current power value of the activated computing domain and the ambient temperature value at the current moment, predict the temperature value of the activated computing domain at the next moment after the current moment, wherein the current power value of the activated computing domain is the power value of the activated computing domain at the current moment, and the temperature value of the computing domain temperature of the activated computing domain is the predicted temperature value of the activated computing domain at the next moment.

7. The electronic component according to claim 6, wherein: The control component is further used to determine the predicted temperature value of the activated computing domain at the next moment by summing the product of the actual temperature value of the activated computing domain at the current moment and the thermal inertia coefficient, the product of the current power value of the activated computing domain and the power-temperature conversion coefficient, and the product of the ambient temperature value at the current moment and the specified heat exchange rate, wherein the specified heat exchange rate is the heat exchange rate between the electronic component and the environment in which the electronic component is located.

8. The electronic component according to claim 6, wherein The sensor network includes a set of sensor nodes in a grid layout, and a computing domain in the computing domain cluster corresponds to some of the sensor nodes in the set of sensor nodes; The control component is further configured to acquire sensor data detected by the sensor nodes corresponding to the K activated computing domains by adopting an address polling method, thereby obtaining sensor data corresponding to the K activated computing domains.

9. The electronic component according to claim 6, wherein The electronic component further includes a temperature sensor; wherein, The temperature sensor is used to periodically detect the ambient temperature; The temperature sensor or the control component is also used to calculate the temperature difference between the detected ambient temperature value and the recorded reference ambient temperature value; when the temperature difference between the detected ambient temperature value and the recorded reference ambient temperature value is greater than or equal to a temperature difference threshold, the product of the detected ambient temperature value and the specified coefficient is added to a second temperature value to determine the updated first temperature threshold, and the recorded reference ambient temperature value is updated to the detected ambient temperature value.

10. The electronic component according to claim 1, wherein The control component is also used to, when there is a first computing task to be assigned, select a third computing domain from the K activated computing domains to execute the first computing task based on the task label of the first computing task and the operating characteristic data of the K activated computing domains, wherein the task label of the first computing task is used to indicate the task attributes of the first computing task and the execution constraints of the first computing task; and assign the first computing task to the third computing domain so that the first computing task is executed by the first computing unit of the third computing domain.

11. The electronic component according to claim 10, wherein The control component runs a task level prediction engine, which is used to predict the task level of the computing task corresponding to the input task label based on the input task label and the input historical performance data; wherein, The control component is also used to input the task label of the first computing task and the historical performance data of the activated computing domain into the task level prediction engine to obtain the task level of the first computing task predicted by the task level prediction engine, wherein the task level of the first computing task is used to characterize the computing power requirement for executing the first computing task; and select one of the K activated computing domains that matches the task level of the first computing task to obtain the third computing domain.

12. The electronic component according to claim 11, wherein One of the plurality of computing units is allowed to use at least one precision mode to perform a computing task, wherein the higher the computing precision corresponding to the precision mode, the more floating-point bits used to perform the computing task; the task attribute of the first computing task includes a precision requirement of the first computing task, and the precision requirement of the first computing task is used to indicate a minimum precision mode required to perform the first computing task; The control component is further configured to, when there are at least two candidate computing domains among the K activated computing domains whose allowed precision modes match the precision requirement of the first computing task, calculate a computing domain score of the candidate computing domain based on the performance of the candidate computing domain, the power consumption of the candidate computing domain, and the temperature of the candidate computing domain, wherein the computing domain score of the candidate computing domain is positively correlated with the performance of the candidate computing domain and negatively correlated with the power consumption of the candidate computing domain and the temperature of the candidate computing domain, and the performance of the candidate computing domain is the number of instructions per cycle of the candidate computing domain; The candidate computing domain with the highest computing domain score among the at least two candidate computing domains is determined as the third computing domain.

13. The electronic component according to claim 11, wherein One of the plurality of computing units is allowed to use at least one precision mode to perform a computing task, and the higher the computing precision corresponding to the precision mode, the more floating-point bits are used to perform the computing task; the task attribute of the first computing task includes the precision requirement of the first computing task, and the precision requirement of the first computing task is used to indicate the minimum precision mode required to perform the first computing task; the third computing domain includes at least two types of computing units; The third computing domain is configured to filter computing units from the computing units in the third computing domain according to a preset filtering condition; and when at least two candidate computing units are filtered out, calculate computing unit scores for the candidate computing units based on the performance of the candidate computing units, the power consumption of the candidate computing units, and the temperature of the candidate computing units; If, among the at least two candidate computing units, the accuracy mode allowed to be adopted by the candidate computing unit with the highest computing unit score matches the accuracy requirement of the first computing task, the candidate computing unit with the highest computing unit score is determined as the first computing unit; if, among the at least two candidate computing units, the accuracy mode allowed to be adopted by the candidate computing unit with the highest computing unit score does not match the accuracy requirement of the first computing task, the computing unit specified in the third computing domain is determined as the first computing unit; The preset screening condition includes one of the following: when the temperature of the electronic component is greater than or equal to the second temperature threshold, screening the computing unit whose temperature is less than the third temperature threshold; screening the computing unit that meets the execution constraint condition; screening the computing unit whose allowed precision mode matches the precision requirement of the first computing task; The computing unit score of the candidate computing unit is positively correlated with the performance of the candidate computing unit, and negatively correlated with the power consumption of the candidate computing unit and the temperature of the candidate computing unit. The performance of the candidate computing unit is the number of instructions per cycle of the candidate computing unit.

14. The electronic component according to claim 1, wherein One of the plurality of computing units is allowed to adopt at least one precision mode to perform a computing task, wherein the higher the computing precision corresponding to the precision mode, the more floating-point bits are used to perform the computing task; The fourth computing domains in the K activated computing domains are used to extract a second computing task to be executed from the computing task queue of the fourth computing domain, wherein the second computing task has at least two computing stages, and the precision requirements of the second computing task are different in different computing stages of the at least two computing stages, and the precision requirements of the second computing task are used to indicate the minimum precision mode required to execute the second computing task; in different computing stages of the second computing task, the second computing task is executed by adopting different precision modes through at least one computing unit in the fourth computing domain, so that the precision mode adopted by the computing unit of the fourth computing domain matches the precision requirement of the computing stage of the second computing task.

15. The electronic component according to claim 1, wherein The activated computing domain is configured with multiple load levels, and a load level prediction engine runs on the control component. The load level prediction engine is used to predict the load level of the computing domain corresponding to the input prediction reference data in M ​​future cycles based on input prediction reference data, and the prediction reference data includes the following data of the computing domains in the computing domain cluster in N historical cycles: instruction mixing, cache miss rate, and task queue depth of the computing task queue, M is a positive integer greater than or equal to 1, and N is a positive integer greater than or equal to 1; wherein, The control component is also used to obtain the predicted reference data of the activated computing domain; input the predicted reference data of the activated computing domain into the load level prediction engine to obtain the load level of the activated computing domain in the M future cycles predicted by the load level prediction engine; and pre-adjust the operating parameters of the activated computing domain based on the predicted load level of the activated computing domain in the M future cycles.

16. The electronic component according to claim 1, wherein The control component is further configured to control the activated computing domain to enter a near-threshold computing mode when the number of instructions per cycle of the activated computing domain is less than or equal to a second instruction number threshold and the temperature of the activated computing domain is less than or equal to a fourth temperature threshold; In the near-threshold calculation mode, the voltage offset of the activated calculation domain is adjusted based on the operating status data of the activated calculation domain, wherein the operating status data of the activated calculation domain is used to indicate the temperature of the activated calculation domain, the voltage of the activated calculation domain and the process deviation coefficient of the activated calculation domain, and the voltage offset of the activated calculation domain is obtained by summing the product of the difference between the temperature of the activated calculation domain and the specified temperature and the third coefficient, the product of the difference between the voltage of the activated calculation domain and the specified voltage and the fourth coefficient, and the product of the process deviation coefficient and the fifth coefficient.

17. The electronic component according to claim 16, wherein: The electronic component further comprises: a main latch and a shadow latch, wherein the main latch and the shadow latch have different trigger edges; wherein, The main latch and the shadow latch are used to perform data sampling on the activated computing domain when the activated computing domain enters the near-threshold computing mode; The control component is also used to sample data from the activated computing domain through the main latch and the shadow latch, and perform timing error detection on the activated computing domain; when a timing error is detected in the activated computing domain, continuously increase the voltage of the activated computing domain until an error instruction is successfully executed, wherein the error instruction is an instruction executed by the activated computing domain and corresponds to the timing error in the activated computing domain; or, control the activated computing domain to exit the near-threshold computing mode.

18. A method for controlling the operation of an electronic component, characterized in that: The electronic component includes a plurality of computing units, the plurality of computing units are divided into a computing domain cluster, and a computing domain in the computing domain cluster is a set of computing units that independently adjust operating parameters; In the computing domain cluster, one computing domain includes some computing units among the plurality of computing units, and different computing domains include different computing units; the method includes: When there are K activated computing domains in the computing domain cluster, a set of state parameters of the activated computing domains is monitored, where K is a positive integer greater than or equal to 1; If there is a first computing domain among the K activated computing domains whose parameter value of a specified state parameter in the set of state parameters meets a corresponding threshold condition, operating parameters of the first computing domain are adjusted according to an adjustment method corresponding to the specified state parameter; Wherein, the set of status parameters includes the number of instructions per cycle, and the threshold condition corresponding to the number of instructions per cycle is that the value of the number of instructions per cycle is less than or equal to the first instruction number threshold; the operation parameters of the first computing domain are adjusted according to the adjustment method corresponding to the specified status parameters, including: when the specified status parameters include the number of instructions per cycle, performing a first adjustment operation on the first computing domain to adjust the operation parameters of the first computing domain, wherein the first adjustment operation includes: lowering the frequency of the first computing domain and lowering the voltage of the first computing domain; the electronic component also includes a hardware performance counter, and the method also includes: real-time monitoring of the number of instructions per cycle of the K activated computing domains through the hardware performance counter.

Images