Binaryzation neural network accelerator based on RISC-V

Abstract

The invention relates to the field of binary neural networks, in particular to a binary neural network accelerator based on RISC-V. A target network architecture requirement and a current task requirement thereof are meticulously considered, and an adaptive RISC-V instruction set executed relative to a current task of a target network in a specified accelerator is screened; in the driving process of the adaptive RISC-V instruction set aiming at the current task execution of the target network, the communication guarantee capability of the specified accelerator aiming at the current task execution of the target network is effectively evaluated, and after the current task execution of the target network is completed, the quality guarantee capability of the specified accelerator aiming at the current task execution of the target network is effectively evaluated. According to the method, the communication guarantee capability and the quality guarantee capability are integrated, the driving effect of the specified accelerator for the current task execution of the target network is evaluated and fed back, the task driving performance and the processing capability of the RISC-V-based binary neural network accelerator are helped to be deeply understood, and then the use safety and the decision support capability of the accelerator are enhanced.

Description

Technical Field

[0001] The present invention relates to the field of binarized neural networks, and in particular to a RISC-V-based binarized neural network accelerator. Background Art

[0002] RISC-V, as an open source, modular instruction set architecture, has set off a wave of innovation in the field of computer architecture in recent years. Its open source nature has not only attracted a large number of developers, but also promoted innovation in chip design. Especially in the field of neural network accelerators, the introduction of RISC-V provides a new opportunity for the efficient execution of binary neural networks.

[0003] The existing technology has combined the flexibility of the RISC-V architecture and the computational efficiency of the binary neural network to design a series of RISC-V-based binary neural network accelerators. However, the existing accelerators still have limitations in the analysis of the driving performance of the binary neural network task execution process. Specifically: 1. The existing accelerators lack a personalized matching mechanism for the RISC-V instruction set for the binary neural network execution task. A low configuration of the RISC-V instruction set may result in limited performance of the accelerator when processing the binary neural network task and fail to meet the real-time or high efficiency requirements. A high configuration of the RISC-V instruction set may increase the risk of resource waste.

[0004] 2. The communication performance evaluation of existing accelerators for binary neural network tasks is limited to the overall communication parameter performance level, ignoring the detailed consideration of the adaptive communication configuration parameter control performance of existing accelerators, resulting in the lack of reliability and comprehensiveness of the communication performance evaluation results.

[0005] 3. The quality performance evaluation of existing accelerators for binary neural network tasks is limited to the accuracy of output results and the actual degree of acceleration, and ignores the resource allocation efficiency of existing accelerators for binary neural network tasks, resulting in the lack of completeness of the quality performance evaluation results. Summary of the invention

[0006] In view of this, in order to solve the problems raised in the above background technology, a binary neural network accelerator based on RISC-V is proposed.

[0007] The technical solution adopted by the present invention to solve its technical problem is: the present invention provides a RISC-V-based binary neural network accelerator, including: an adaptation instruction set screening module, a communication assurance capability evaluation module, a quality assurance capability evaluation module, a task-driven effect feedback module and a cloud database.

[0008] The adaptation instruction set screening module is connected to the communication assurance capability evaluation module, the communication assurance capability evaluation module is connected to the quality assurance capability evaluation module, the quality assurance capability evaluation module is connected to the task-driven effect feedback module, and the cloud database is respectively connected to the communication assurance capability evaluation module, the quality assurance capability evaluation module, and the task-driven effect feedback module.

[0009] The adaptation instruction set screening module is used to import the target binary neural network configuration data and its current task requirement data into the specified accelerator, record the target binary neural network as the target network, and screen the adaptation RISC-V instruction set for the current task execution of the target network from various RISC-V instruction sets configured by the specified accelerator.

[0010] The communication assurance capability evaluation module is used to monitor in real time the communication link status of the target network during the execution of the current task driven by the RISC-V instruction set, and evaluate the communication assurance capability of the specified accelerator for the current task of the target network.

[0011] The quality assurance capability evaluation module is used to analyze the recognition accuracy, acceleration degree and resource efficiency of the designated accelerator for the execution of the current task of the target network after the current task of the target network is completed, and evaluate the quality assurance capability of the designated accelerator for the execution of the current task of the target network.

[0012] The task-driven effect feedback module is used to integrate communication assurance capabilities and quality assurance capabilities, evaluate the driving effect of the specified accelerator on the current task of the target network and provide feedback.

[0013] The cloud database is used to store the reference communication assurance capability evaluation indicators and reference quality assurance capability evaluation indicators planned for the task-driven effect evaluation standards, store the reasonable communication bandwidth utilization preset by the communication link during the binary neural network task execution process, and store the real label content of the current task.

[0014] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: (1) The present invention carefully considers the requirements of the target network architecture and its current task requirements, and selects the adaptive RISC-V instruction set for the current task execution of the target network from various RISC-V instruction sets configured by the specified accelerator, which not only helps to achieve the optimal energy efficiency ratio of the accelerator RISC-V instruction set, but also ensures the efficiency and security of task execution.

[0015] (2) The present invention carefully considers the overall communication quality coefficient and the overall communication maintenance coefficient of the designated accelerator for the current task execution driven by the target network, and accurately evaluates the communication guarantee capability of the designated accelerator for the current task execution of the target network, which is beneficial to the continuous optimization of the communication quality of the accelerator task driving process, thereby enhancing the safety of accelerator use and decision support capabilities.

[0016] (3) The present invention analyzes the recognition accuracy, acceleration degree and resource efficiency of the designated accelerator for the current task execution of the target network, accurately evaluates the quality assurance capability of the designated accelerator for the current task execution of the target network, directly measures the accelerator's task driving performance and processing capability for the binary neural network, and provides a basis for system optimization and decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The present invention is further described using the accompanying drawings, but the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative work.

[0018] Figure 1 It is a schematic diagram of module connection of the present invention. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] Reference Figure 1 As shown, the present invention provides a RISC-V-based binary neural network accelerator, including: an adaptation instruction set screening module, a communication assurance capability evaluation module, a quality assurance capability evaluation module, a task-driven effect feedback module and a cloud database.

[0021] The adaptation instruction set screening module is connected to the communication assurance capability evaluation module, the communication assurance capability evaluation module is connected to the quality assurance capability evaluation module, the quality assurance capability evaluation module is connected to the task-driven effect feedback module, and the cloud database is respectively connected to the communication assurance capability evaluation module, the quality assurance capability evaluation module, and the task-driven effect feedback module.

[0022] The adaptation instruction set screening module is used to import the target binary neural network configuration data and its current task requirement data into the designated accelerator, record the target binary neural network as the target network, and screen the adaptation RISC-V instruction set for the current task execution of the target network from various RISC-V instruction sets configured by the designated accelerator.

[0023] Specifically, the specific analysis process of the adaptation instruction set screening module includes: extracting each configuration level and its corresponding various logical operation requirements and various data processing method requirements from the target network configuration data, searching for various RISC-V instruction sets configured by the specified accelerator to determine whether there are support operation instructions for various logical operation requirements and support operation instructions for various data processing method requirements for each configuration level of the target network, so as to analyze the instruction coverage IC of various RISC-V instruction sets configured by the specified accelerator for the target network architecture. j , j is the number of various RISC-V instruction sets configured for the specified accelerator, j = 1, 2, ..., b.

[0024] Retrieve the execution log content of each task in the historical preset period of various RISC-V instruction sets of the specified accelerator, including the task difficulty level, precision timing change curve, speed timing change curve and power consumption timing change curve, and filter the minimum precision value, minimum speed value and maximum power consumption value per unit time, so as to calculate the average precision limit y of the execution of tasks of each difficulty level in the historical preset period of various RISC-V instruction sets of the specified accelerator jh , average speed limit v jh and the average power consumption limit p jh , h is the number of the task at each difficulty level, h = 1, 2, ..., l, extract the current task requirement data of the target network, including power consumption limit, accuracy limit and speed limit, and compare and analyze the matching degree RM of various RISC-V instruction sets configured by the specified accelerator for the current task of the target network j .

[0025] The demand matching degree and instruction coverage are accumulated, and the RISC-V instruction set corresponding to the maximum accumulated value is selected as the adapted RISC-V instruction set for the current task execution of the target network.

[0026] Specifically, the IC j The specific analysis process includes: integrating the total number of logic operation requirement types and the total number of data processing method requirement types of the target network configuration level, calculating the support instruction coverage of various RISC-V instruction sets of the specified accelerator configuration for the target network configuration level logic operation requirements and the support instruction coverage of the data processing method requirements, and accumulating the two to obtain the instruction coverage of various RISC-V instruction sets of the specified accelerator configuration for the target network architecture.

[0027] It should be noted that the support instruction coverage rate of the various RISC-V instruction sets of the above-mentioned specified accelerator configuration for the target network configuration level logic operation requirements is obtained by performing a ratio analysis on the number of support instructions for the target network configuration level logic operation requirements of the various RISC-V instruction sets of the specified accelerator configuration and the total number of logic operation requirement types of the target network configuration level. Similarly, the support instruction coverage rate of the various RISC-V instruction sets of the specified accelerator configuration for the target network configuration level data processing method requirements is obtained by performing a ratio analysis on the number of support instructions for the target network configuration level data processing method requirements of the various RISC-V instruction sets of the specified accelerator configuration and the total number of data processing method requirement types of the target network configuration level.

[0028] Specifically, the RM j The specific analysis process includes: Analyze the quality scores of tasks of various difficulty levels within the historical preset cycles of various RISC-V instruction sets of the specified accelerator, where y0, v0, and p0 are the preset reference accuracy values, reference speed values, and reference power consumption values, respectively. Similarly, calculate the required quality scores of the current task execution of the target network.

[0029] According to the difficulty level of the current task of the target network, the quality score Q of the execution of tasks with relatively high difficulty levels within the historical preset period of various RISC-V instruction sets of the specified accelerator is screened. jh′ , the quality score Q of each relatively low difficulty level task execution jh″ The quality score of the task execution of the same difficulty level Where h′ is the number of each relatively high difficulty level task, h′=1, 2, ..., l′, and h″ is the number of each relatively low difficulty level task, h″=1, 2, ..., l″, in order to analyze the reference quality score of various RISC-V instruction sets of the specified accelerator for the current task of the target network.

[0030] It should be noted that the specific calculation formula for the reference quality score of the various RISC-V instruction sets of the above-mentioned specified accelerator for the current task execution of the target network is: Among them, l′ and l″ are the number of tasks with relatively high difficulty level and the number of tasks with relatively low difficulty level respectively, and δ1, δ2, and δ3 are the preset reference value weights corresponding to the relatively high difficulty level, the relatively low difficulty level and the same difficulty level respectively.

[0031] Exemplarily, the above δ1, δ2, and δ3 may specifically be 0.3, 0.3, and 0.4.

[0032] By computational model Get the matching degree of various RISC-V instruction sets configured by the specified accelerator to the requirements of the current task of the target network.

[0033] The embodiment of the present invention carefully considers the target network architecture requirements and its current task requirements, and screens the adaptive RISC-V instruction set for the current task execution of the target network from various RISC-V instruction sets configured for the specified accelerator, which not only helps to achieve the optimal energy efficiency ratio of the accelerator RISC-V instruction set, but also ensures the efficiency and security of task execution.

[0034] The communication assurance capability evaluation module is used to monitor in real time the communication link status of the target network during the execution and driving process of the current task of the target network adapted by the RISC-V instruction set, and evaluate the communication assurance capability of the designated accelerator for the current task of the target network.

[0035] Specifically, the specific analysis process of the communication assurance capability evaluation module includes: respectively recording the communication bandwidth utilization change curve, communication delay change curve, packet loss rate change curve and bit error rate change curve of the communication link during the drive process of the current task execution of the target network adapted by the RISC-V instruction set, and analyzing the overall communication quality coefficient CQ of the specified accelerator for the current task execution drive of the target network.

[0036] Retrieve the adapted RISC-V instruction set recorded in the communication management log of the specified accelerator to adjust the timestamps of the communication link configuration parameters during the current task execution drive of the target network and detect the corresponding communication improvement effect, so as to analyze the overall communication maintenance coefficient CM of the specified accelerator for the current task execution drive of the target network.

[0037] The accumulated value of the overall communication quality coefficient and the overall communication maintenance coefficient is used as an evaluation indicator of the communication assurance capability of the designated accelerator for the current task of the target network.

[0038] Specifically, the specific analysis process of the CQ includes: importing the communication bandwidth utilization change curve of the communication link during the current task execution driving process of the target network adapted to the RISC-V instruction set into the matlab software to obtain its corresponding best fitting function, recorded as F(t), collecting the start time point t1 and the end time point t2 of the current task execution of the target network, and using the formula The communication bandwidth compliance coefficient of the communication link in the process of driving the current task execution of the target network is analyzed and adapted to the RISC-V instruction set, where η0 is the preset reasonable communication bandwidth utilization rate of the communication link in the binary neural network task execution process stored in the cloud database.

[0039] Analyze and adapt the RISC-V instruction set to the communication delay compliance coefficient χ2, packet loss rate compliance coefficient χ3 and bit error rate compliance coefficient χ4 of the communication link during the current task execution driving process of the target network.

[0040] Similarly, Among them, G(t), R(t), and Z(t) are the best fitting functions corresponding to the communication delay change curve, packet loss rate change curve, and bit error rate change curve of the communication link in the process of driving the current task execution of the target network by adapting the RISC-V instruction set, respectively. σ0, ψ0, The cloud database stores the preset permitted communication delay threshold, permitted packet loss rate threshold, and permitted bit error rate threshold of the communication link during the execution process of the binarized neural network task respectively.

[0041] According to the formula CQ = χ l +χ2+χ3+χ4 to obtain the overall communication quality coefficient of the specified accelerator driven by the current task execution of the target network.

[0042] Specifically, the specific analysis process of the CM includes: comparing the communication bandwidth utilization, communication delay, packet loss rate and bit error rate of each control timestamp of the communication link configuration parameters during the current task execution driving process of the target network according to the adapted RISC-V instruction set with the preset reasonable communication bandwidth utilization, permitted communication delay threshold, permitted packet loss rate threshold and permitted bit error rate threshold of the communication link in the binary neural network task execution process, calculating the deviation ratio corresponding to the communication bandwidth utilization, communication delay, packet loss rate and bit error rate of each control timestamp of the communication link configuration parameters during the current task execution driving process of the target network according to the adapted RISC-V instruction set and accumulating them, taking the accumulated value as the communication anomaly risk coefficient of each control timestamp of the communication link configuration parameters during the current task execution driving process of the target network according to the adapted RISC-V instruction set, and further performing ratio analysis with the communication anomaly risk coefficient warning threshold corresponding to the preset communication link configuration parameter control trigger to retrieve the control trigger compliance of each control timestamp.

[0043] It should be noted that the communication bandwidth utilization deviation ratio of each adjustment timestamp of the communication link configuration parameters in the above-mentioned adapted RISC-V instruction set for the target network current task execution driving process is obtained by calculating the absolute difference between the communication bandwidth utilization of each adjustment timestamp of the communication link configuration parameters in the adapted RISC-V instruction set for the target network current task execution driving process and the reasonable communication bandwidth utilization preset for the communication link in the binary neural network task execution process, and further performing ratio analysis on the calculated absolute difference with the reasonable communication bandwidth utilization preset for the communication link in the binary neural network task execution process.

[0044] The communication delay deviation ratio of each adjustment timestamp of the communication link configuration parameters during the current task execution driving process of the target network adapted to the RISC-V instruction set is calculated by performing a difference calculation between the communication delay of each adjustment timestamp of the communication link configuration parameters during the current task execution driving process of the target network adapted to the RISC-V instruction set and the permitted communication delay threshold preset for the communication link in the binary neural network task execution process. If the calculated difference is a non-positive number, the communication delay deviation ratio is set to 0. If the calculated difference is a positive number, the calculated difference is further obtained by performing a ratio analysis with the permitted communication delay threshold preset for the communication link in the binary neural network task execution process.

[0045] The calculation process of the bit error rate deviation ratio and packet loss rate deviation ratio of each adjustment time stamp of the communication link configuration parameters during the current task execution driving process of the target network adapted to the RISC-V instruction set is consistent with the calculation process of the communication delay deviation ratio.

[0046] The communication bandwidth utilization change curve, communication delay change curve, packet loss rate change curve and bit error rate change curve of the communication link in the process of executing the current task of the target network by adapting the RISC-V instruction set are respectively marked with the timestamps of each adjustment and control of the configuration parameters, and each inspection time period combination after each adjustment and control timestamp is planned. The relative improvement degree of the communication bandwidth utilization, communication delay, packet loss rate and bit error rate of the communication link in each inspection time period combination after each adjustment and control timestamp is detected. According to the preset influence weights corresponding to each inspection time period combination, the relative communication improvement degree of each adjustment and control timestamp of the communication link configuration parameters in the process of executing the current task of the target network by adapting the RISC-V instruction set is analyzed.

[0047] It should be noted that the detection process of the relative improvement of the communication bandwidth utilization, communication delay, packet loss rate and bit error rate of the communication link for each inspection period combination after each of the above-mentioned control timestamps is as follows: the average communication bandwidth utilization, average communication delay, average packet loss rate and average bit error rate of the communication link corresponding to each inspection period combination after a certain control timestamp are obtained, and compared with the communication bandwidth utilization, communication delay, packet loss rate and bit error rate of the communication link of the control timestamp, and the communication delay reduction rate, packet loss rate reduction rate and average bit error rate reduction rate of the communication link corresponding to each inspection period combination after the control timestamp are respectively used as the relative improvement corresponding to the communication delay, packet loss rate and bit error rate, and the deviation ratio reduction rate of the communication bandwidth utilization of the communication link corresponding to each inspection period combination after the control timestamp and the reasonable communication bandwidth utilization preset by the communication link in the binary neural network task execution process is used as the relative improvement of the communication bandwidth utilization, thereby obtaining the relative improvement of the communication bandwidth utilization, communication delay, packet loss rate and bit error rate of the communication link for each inspection period combination after each control timestamp.

[0048] It should also be noted that the relative communication improvement of each adjustment timestamp of the communication link configuration parameters during the current task execution driving process of the target network adapted by the RISC-V instruction set is achieved by accumulating the relative improvement of the communication bandwidth utilization, communication delay, packet loss rate and bit error rate of the communication link for each inspection period combination after each adjustment timestamp to obtain the overall relative improvement of the communication link for each inspection period combination after each adjustment timestamp, and accumulating the overall relative improvement of the communication link for each inspection period combination and the product of its corresponding preset influence weight to obtain the relative communication improvement of each adjustment timestamp of the communication link configuration parameters during the current task execution driving process of the target network adapted by the RISC-V instruction set.

[0049] The product of the control trigger compliance and the relative communication improvement is used as the communication improvement effect detection index. The communication improvement effect detection index of each control timestamp of the communication link configuration parameters in the process of driving the current task execution of the target network adapted to the RISC-V instruction set is collected and the average is calculated to obtain the overall communication maintenance coefficient of the specified accelerator for the current task execution drive of the target network.

[0050] The embodiment of the present invention carefully considers the overall communication quality coefficient and the overall communication maintenance coefficient of the designated accelerator driven by the current task execution of the target network, and accurately evaluates the communication guarantee capability of the designated accelerator for the current task execution of the target network, which is beneficial to the continuous optimization of the communication quality of the accelerator task driving process, thereby enhancing the safety of accelerator use and decision support capabilities.

[0051] The quality assurance capability evaluation module is used to analyze the recognition accuracy, acceleration degree and resource efficiency of the designated accelerator for the execution of the current task of the target network after the current task of the target network is completed, and evaluate the quality assurance capability of the designated accelerator for the execution of the current task of the target network.

[0052] Specifically, the specific analysis process of the quality assurance capability evaluation module includes: comparing the output label content of the designated accelerator for the current task execution drive of the target network with the real label content of the current task stored in the cloud database, analyzing the precision ζ and recall φ of the designated accelerator for the current task execution drive of the target network, and calculating the recognition accuracy of the designated accelerator for the current task execution drive of the target network.

[0053] The target network is simulated to execute the current task without the help of the specified accelerator and the task completion time is recorded. The ratio analysis is performed with the completion time of the specified accelerator for the target network's current task execution drive to obtain the acceleration degree of the specified accelerator for the target network's current task execution drive.

[0054] The computing resource utilization ε1, storage resource occupancy ε2 and hardware cumulative power consumption μ of the specified accelerator driven by the current task execution of the target network are sorted out by the formula Analyze the resource efficiency of the specified accelerator driven by the current task execution of the target network, where ε′1, ε′2, and μ0 are the reasonable reference computing resource utilization, reasonable reference storage resource occupancy, and reasonable reference hardware cumulative power consumption preset by the specified accelerator binary neural network task driver, respectively.

[0055] The recognition accuracy, acceleration degree, and resource efficiency of the designated accelerator driven by the current task execution of the target network are accumulated to obtain the quality assurance capability evaluation index of the designated accelerator for the current task execution of the target network.

[0056] The embodiment of the present invention analyzes the recognition accuracy, acceleration degree and resource efficiency of the designated accelerator for the current task execution of the target network, accurately evaluates the quality assurance capability of the designated accelerator for the current task execution of the target network, directly measures the accelerator's task driving performance and processing capability for the binary neural network, and provides a basis for system optimization and decision-making.

[0057] Specifically, the calculation of the recognition accuracy of the designated accelerator for the current task execution drive of the target network is expressed as:

[0058] The task driving effect feedback module is used to comprehensively evaluate the driving effect of the specified accelerator on the current task of the target network and provide feedback.

[0059] Specifically, the specific analysis process of the task-driven effect feedback module includes: extracting the communication assurance capability evaluation index TX and the quality assurance capability evaluation index ZL of the specified accelerator for the current task of the target network, and using the formula Evaluate the driving effect of the specified accelerator on the current task of the target network, where TX0 and ZL0 are the reference communication assurance capability evaluation index and reference quality assurance capability evaluation index of the cloud database storage task driving effect evaluation standard planning respectively.

[0060] The cloud database is used to store reference communication assurance capability evaluation indicators and reference quality assurance capability evaluation indicators planned for task-driven effect evaluation standards, store reasonable communication bandwidth utilization, permitted communication delay threshold, permitted packet loss rate threshold and permitted bit error rate threshold preset for the communication link in the binary neural network task execution process, and store the real label content of the current task.

[0061] The data sources in the cloud database of this embodiment are shown in Table 1 below.

[0062] Table 1 Detailed description of data sources in cloud database

[0063]

[0064]

[0065] The above contents are merely examples and explanations of the structure of the present invention. The technicians in this technical field may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the structure of the invention or exceed the scope defined by the present invention, they should all fall within the protection scope of the present invention.

Claims

1. A RISC-V-based binary neural network accelerator, characterized in that: include: An adaptation instruction set screening module is used to import the target binary neural network configuration data and its current task requirement data into the specified accelerator, record the target binary neural network as the target network, and screen the adaptation RISC-V instruction set for the current task execution of the target network from various RISC-V instruction sets configured by the specified accelerator; The communication assurance capability evaluation module is used to monitor the communication link status of the target network during the execution of the current task driven by the RISC-V instruction set in real time, and evaluate the communication assurance capability of the specified accelerator for the current task of the target network; The quality assurance capability evaluation module is used to analyze the recognition accuracy, acceleration degree and resource efficiency of the designated accelerator for the execution of the current task of the target network after the current task of the target network is completed, and evaluate the quality assurance capability of the designated accelerator for the execution of the current task of the target network; The task driving effect feedback module is used to integrate communication assurance capabilities and quality assurance capabilities, evaluate the driving effect of the specified accelerator on the current task of the target network and provide feedback; The cloud database is used to store the reference communication assurance capability evaluation indicators and reference quality assurance capability evaluation indicators planned for the task-driven effect evaluation standards, store the reasonable communication bandwidth utilization preset by the communication link during the binary neural network task execution process, and store the real label content of the current task.

2. A RISC-V based binary neural network accelerator according to claim 1, characterized in that: The specific analysis process of the adaptation instruction set screening module includes: extracting each configuration level and its corresponding various logic operation requirements and various data processing method requirements from the target network configuration data, searching for various RISC-V instruction sets configured by the specified accelerator to determine whether there are support operation instructions corresponding to various logic operation requirements and support operation instructions for various data processing method requirements for each configuration level of the target network, thereby analyzing the instruction coverage of various RISC-V instruction sets configured by the specified accelerator for the target network architecture. j , j is the number of various RISC-V instruction sets configured for the specified accelerator, j = 1, 2, ..., b; Retrieve the execution log content of each task in the historical preset period of various RISC-V instruction sets of the specified accelerator, including the task difficulty level, precision timing change curve, speed timing change curve and power consumption timing change curve, and filter the minimum precision value, minimum speed value and maximum power consumption value per unit time, so as to calculate the average precision limit y of the execution of tasks of each difficulty level in the historical preset period of various RISC-V instruction sets of the specified accelerator jh , average speed limit v jh and the average power consumption limit p jh , h is the number of the task at each difficulty level, h = 1, 2, ..., l, extract the current task requirement data of the target network, including power consumption limit, accuracy limit and speed limit, and compare and analyze the matching degree RM of various RISC-V instruction sets configured by the specified accelerator for the current task of the target network j ; The demand matching degree and instruction coverage are accumulated, and the RISC-V instruction set corresponding to the maximum accumulated value is selected as the adapted RISC-V instruction set for the current task execution of the target network.

3. A RISC-V based binary neural network accelerator according to claim 2, characterized in that: The IC j The specific analysis process includes: integrating the total number of logic operation requirement types and the total number of data processing method requirement types of the target network configuration level, calculating the support instruction coverage of various RISC-V instruction sets of the specified accelerator configuration for the target network configuration level logic operation requirements and the support instruction coverage of the data processing method requirements, and accumulating the two to obtain the instruction coverage of various RISC-V instruction sets of the specified accelerator configuration for the target network architecture.

4. The RISC-V based binary neural network accelerator according to claim 2, characterized in that: The RM j The specific analysis process includes: Analyze the quality scores of tasks of various difficulty levels within the historical preset cycles of various RISC-V instruction sets of the specified accelerator, where y0, v0, and p0 are the preset reference accuracy values, reference speed values, and reference power consumption values, respectively. Similarly, calculate the required quality scores of the current task execution of the target network. According to the difficulty level corresponding to the current task of the target network, the quality scores of the execution of tasks with relatively high difficulty levels, the quality scores of the execution of tasks with relatively low difficulty levels, and the quality scores of the execution of tasks with the same difficulty levels of various RISC-V instruction sets of the specified accelerator within the historical preset period are screened, so as to analyze the reference quality scores of the execution of various RISC-V instruction sets of the specified accelerator for the current task of the target network. By computational model Get the matching degree of various RISC-V instruction sets configured by the specified accelerator to the requirements of the current task of the target network.

5. The RISC-V-based binary neural network accelerator according to claim 2, characterized in that: The specific analysis process of the communication assurance capability evaluation module includes: recording the communication bandwidth utilization change curve, communication delay change curve, packet loss rate change curve and bit error rate change curve of the communication link in the process of the target network current task execution drive adapted to the RISC-V instruction set, and analyzing the overall communication quality coefficient CQ of the designated accelerator for the target network current task execution drive; Retrieve the timestamps of each adjustment of the communication link configuration parameters during the current task execution drive of the target network recorded in the adapted RISC-V instruction set of the specified accelerator communication management log and detect the corresponding communication improvement effect, so as to analyze the overall communication maintenance coefficient CM of the specified accelerator for the current task execution drive of the target network; The accumulated value of the overall communication quality coefficient and the overall communication maintenance coefficient is used as an evaluation indicator of the communication assurance capability of the designated accelerator for the current task of the target network.

6. A RISC-V based binary neural network accelerator according to claim 5, characterized in that: The specific analysis process of the CQ includes: importing the communication bandwidth utilization change curve of the communication link during the current task execution driving process of the target network adapted to the RISC-V instruction set into the matlab software to obtain its corresponding best fitting function, recorded as F(t), collecting the start time point t1 and the end time point t2 of the current task execution of the target network, and using the formula Analyze the communication bandwidth compliance coefficient of the communication link in the current task execution driving process of the target network adapted to the RISC-V instruction set, where η0 is the preset reasonable communication bandwidth utilization rate of the communication link in the binary neural network task execution process stored in the cloud database; Analyze and adapt the RISC-V instruction set to the communication delay compliance coefficient χ2, packet loss rate compliance coefficient χ3 and bit error rate compliance coefficient χ4 of the communication link during the current task execution driving process of the target network; The overall communication quality coefficient of the specified accelerator driven by the current task execution of the target network is obtained by the formula CQ=χ1+χ2+χ3+χ4.

7. The RISC-V-based binary neural network accelerator according to claim 5, characterized in that: The specific analysis process of the CM includes: according to the adapted RISC-V instruction set, the communication bandwidth utilization, communication delay, packet loss rate and bit error rate of each control timestamp of the communication link configuration parameters in the current task execution driving process of the target network, and retrieving the control trigger compliance of each control timestamp; The communication bandwidth utilization rate change curve, communication delay change curve, packet loss rate change curve and bit error rate change curve of the communication link in the process of driving the current task of the target network are respectively marked with the timestamps of each adjustment and control of the configuration parameters by the adapted RISC-V instruction set, and each inspection period combination after each adjustment and control timestamp is planned, and the relative improvement of the communication bandwidth utilization rate, communication delay, packet loss rate and bit error rate of the communication link in each inspection period combination after each adjustment and control timestamp is detected, and the relative communication improvement of each adjustment and control timestamp of the communication link configuration parameters in the process of driving the current task of the target network by the adapted RISC-V instruction set is analyzed according to the preset influence weights corresponding to each inspection period combination; The product of the control trigger compliance and the relative communication improvement is used as the communication improvement effect detection index. The communication improvement effect detection index of each control timestamp of the communication link configuration parameters in the process of driving the current task execution of the target network adapted to the RISC-V instruction set is collected and the average is calculated to obtain the overall communication maintenance coefficient of the specified accelerator for the current task execution drive of the target network.

8. The RISC-V based binary neural network accelerator according to claim 5, characterized in that: The specific analysis process of the quality assurance capability evaluation module includes: comparing the output label content of the specified accelerator under the current task execution drive of the target network with the real label content of the current task stored in the cloud database, and analyzing the accuracy of the specified accelerator under the current task execution drive of the target network. and recall φ, to calculate the recognition accuracy of the specified accelerator for the current task execution drive of the target network; The target network is simulated to execute the current task without the help of the specified accelerator and the task completion time is recorded. The ratio analysis is performed with the completion time of the specified accelerator for the target network's current task execution drive to obtain the acceleration degree of the specified accelerator for the target network's current task execution drive; The computing resource utilization, storage resource occupancy, and hardware cumulative power consumption of the specified accelerator driven by the current task execution of the target network are sorted out to analyze the resource efficiency of the specified accelerator driven by the current task execution of the target network; The recognition accuracy, acceleration degree, and resource efficiency of the designated accelerator driven by the current task execution of the target network are accumulated to obtain the quality assurance capability evaluation index of the designated accelerator for the current task execution of the target network.

9. The RISC-V based binary neural network accelerator according to claim 8, characterized in that: The calculation of the recognition accuracy of the designated accelerator for the current task execution drive of the target network is shown as follows:

10. The RISC-V based binary neural network accelerator according to claim 8, characterized in that: The specific analysis process of the task-driven effect feedback module includes: extracting the communication assurance capability evaluation index TX and the quality assurance capability evaluation index ZL of the specified accelerator for the current task of the target network, and using the formula Evaluate the driving effect of the specified accelerator on the current task of the target network, where TX0 and ZL0 are the reference communication assurance capability evaluation index and reference quality assurance capability evaluation index of the cloud database storage task driving effect evaluation standard planning respectively.

Images