A fine-grained mapping method, system and medium for storage-computing integrated chip
By analyzing the chip thread status, filtering and processing kernel status information and generating mapping rules, the problems of poor mapping accuracy and poor robustness in the existing technology are solved, and higher mapping accuracy and system robustness are achieved.
Patent Information
- Application Number
- CN202410875421.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-02
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2044-07-02
AI Technical Summary
Existing chip fine-grained mapping methods cannot dynamically filter mapping rules based on the operating status of the processing core, resulting in poor mapping accuracy and robustness.
By obtaining chip parameters, analyzing thread status, filtering processing kernel status information, generating mapping rules, and adjusting fine-grained filtering standards according to the difference threshold to achieve dynamic mapping of processing kernel data.
It improves the quantization accuracy of the mapping and the robustness of the system, and enhances the mapping accuracy of processing core running data.
Smart Images

Figure CN118445595B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of chip fine-grained analysis, and more specifically, to a fine-grained mapping method, system and medium for a storage-computing integrated chip. Background Art
[0002] In recent years, RRAM (Resistive Random Access Memory)-based integrated storage and computing chips have great potential in neural network acceleration. The weights of the neural network can be directly mapped to the crossbar array based on RRAM for in-situ calculation, effectively avoiding the overhead of data transfer in the von Neumann architecture and obtaining a higher accuracy. Before mapping the weights of the neural network to the crossbar array based on memristors, it is usually necessary to quantize the weights and deploy the quantized weights of each layer to the memristor array; the existing chip fine-grained mapping method cannot perform fine-grained screening of the processing core state according to the chip thread, so it is difficult to establish different mapping rules according to different processing core state information. At present, most algorithms for RRAM-based integrated storage and computing chips adopt a hierarchical quantization strategy, with some layers of the network using high quantization accuracy and some layers using low quantization accuracy, which reduces hardware overhead while ensuring accuracy. However, the quantization accuracy of a certain layer is consistent, that is, the number of columns occupied by all convolution kernels in the layer is the same, and it is not flexible enough, which will result in a relatively larger mapping accuracy loss. For the above problems, effective technical solutions are urgently needed. Summary of the invention
[0003] The purpose of the embodiments of the present application is to provide a fine-grained mapping method, system and medium for a storage and computing integrated chip, so as to solve the problem in the prior art that it is difficult to establish a dynamic mapping relationship according to the operating status of the processing core, and the mapping accuracy and robustness are poor.
[0004] The embodiment of the present application also provides a fine-grained mapping method for a storage-computing integrated chip, including:
[0005] Obtain chip parameters, analyze thread status based on chip parameters, and process core status information based on thread status analysis;
[0006] Perform granular analysis based on the processing core status information, filter out fine-grained processing core status information, sort the processing core status information, and generate mapping rules;
[0007] Based on the difference between the processing core state information analysis and the preset state, determining whether the difference is greater than or equal to a set difference threshold;
[0008] If it is greater than or equal to, then generate correction information, and adjust the fine-grained screening criteria based on the correction information;
[0009] If it is less than, the processing core operation data is mapped according to the mapping rule to obtain a mapping result, and the mapping result is transmitted to the terminal in real time.
[0010] Optionally, in the fine-grained mapping method of the storage-computing integrated chip described in the embodiment of the present application, obtaining chip parameters, analyzing thread states based on the chip parameters, and analyzing and processing core state information according to the thread states specifically include:
[0011] Obtain chip model, chip power and chip memory resource information, and divide the chip into multiple processing cores based on the chip model and chip power;
[0012] Analyze multiple processing core threads and coordination information based on dynamic changes in chip memory resource information;
[0013] Based on the collaborative information, different processing core threads are operated collaboratively to obtain collaborative data;
[0014] Compensation information is generated according to the collaborative data, and the operation data of a single processing core is adjusted based on the compensation information to obtain the processing core state information.
[0015] Optionally, in the fine-grained mapping method of the storage-computing integrated chip described in the embodiment of the present application, granular analysis is performed based on the processing core state information, fine-grained processing core state information is screened out, the processing core state information is sorted, and mapping rules are generated, specifically including:
[0016] Acquire the processing core state information, classify the processing core state information according to multiple granularity standards, and obtain different categories of processing core state information;
[0017] Compare different categories of processing core status information with fine-grained set indicators;
[0018] Filter the processing core status information that is less than or equal to the fine-grained set index to obtain the processing core status information that meets the mapping requirements;
[0019] Sorting different categories of processing core state information according to fine-grained sizes, and generating mapping rules for different categories of processing core state information;
[0020] If the processing core state information is greater than the fine-grained setting index, it is determined to be coarse-grained processing core state information, and the coarse-grained processing core state information is subdivided into secondary granularity.
[0021] Optionally, in the fine-grained mapping method of the storage-computing integrated chip described in the embodiment of the present application, based on the difference between the processing core state information analysis and the preset state, judging whether the difference is greater than or equal to a set difference threshold specifically includes:
[0022] Obtaining the status information of the processing cores at different time nodes, comparing the status information of the processing cores at different nodes, and obtaining status difference information;
[0023] Compare the state difference information with the preset state to obtain the degree of difference;
[0024] The difference is compared with a set difference threshold to determine whether the difference exceeds the set difference threshold.
[0025] Optionally, in the fine-grained mapping method of the storage-computing integrated chip described in the embodiment of the present application, if it is greater than or equal to, then generating correction information, and adjusting the fine-grained screening criteria based on the correction information, specifically including:
[0026] Obtaining a difference and a set difference threshold, where the set difference threshold includes a first difference threshold and a second difference threshold, and the first difference threshold is less than the second difference threshold;
[0027] If the difference is greater than the first difference threshold and less than the second difference threshold, generating first correction information, setting a first adjustment rule based on the first correction information, and adjusting the fine-grained screening criteria according to the first adjustment rule;
[0028] If the difference is greater than or equal to the second difference threshold, second correction information is generated, a second adjustment rule is set based on the second correction information, and the fine-grained screening standard is adjusted according to the second adjustment rule.
[0029] Optionally, in the fine-grained mapping method of the storage-computing integrated chip described in the embodiment of the present application, if less than, the processing core operation data is mapped according to the mapping rule to obtain a mapping result, and the mapping result is transmitted to the terminal in real time, specifically including:
[0030] Acquire mapping rules and processing core state information, divide the processing core state information into fine-grained categories, and obtain different categories of processing core state information;
[0031] Matching the mapping rules with different categories of processing core state information to obtain a matching degree;
[0032] sorting the mapping matching degrees of different categories of processing core state information to generate a mapping order;
[0033] Mapping different types of processing core state information in sequence according to a mapping order to obtain multiple mapping results;
[0034] The multiple mapping results are transmitted to the terminal for storage respectively.
[0035] In a second aspect, an embodiment of the present application provides a fine-grained mapping system for a storage-computing integrated chip, the system comprising: a memory and a processor, the memory comprising a program of a fine-grained mapping method for the storage-computing integrated chip, and the program of the fine-grained mapping method for the storage-computing integrated chip implements the following steps when executed by the processor:
[0036] Obtain chip parameters, analyze thread status based on chip parameters, and process core status information based on thread status analysis;
[0037] Perform granular analysis based on the processing core status information, filter out fine-grained processing core status information, sort the processing core status information, and generate mapping rules;
[0038] Based on the difference between the processing core state information analysis and the preset state, determining whether the difference is greater than or equal to a set difference threshold;
[0039] If it is greater than or equal to, then generate correction information, and adjust the fine-grained screening criteria based on the correction information;
[0040] If it is less than, the processing core operation data is mapped according to the mapping rule to obtain a mapping result, and the mapping result is transmitted to the terminal in real time.
[0041] Optionally, in the fine-grained mapping system of the storage-computing integrated chip described in the embodiment of the present application, obtaining chip parameters, analyzing thread states based on the chip parameters, and analyzing and processing core state information according to the thread states specifically include:
[0042] Obtain chip model, chip power and chip memory resource information, and divide the chip into multiple processing cores based on the chip model and chip power;
[0043] Analyze multiple processing core threads and coordination information based on dynamic changes in chip memory resource information;
[0044] Based on the collaborative information, different processing core threads are operated collaboratively to obtain collaborative data;
[0045] Compensation information is generated according to the collaborative data, and the operation data of a single processing core is adjusted based on the compensation information to obtain the processing core state information.
[0046] Optionally, in the fine-grained mapping system of the storage-computing integrated chip described in the embodiment of the present application, granular analysis is performed based on the processing core state information, fine-grained processing core state information is screened out, the processing core state information is sorted, and mapping rules are generated, specifically including:
[0047] Acquire the processing core state information, classify the processing core state information according to multiple granularity standards, and obtain different categories of processing core state information;
[0048] Compare different categories of processing core status information with fine-grained set indicators;
[0049] Filter the processing core status information that is less than or equal to the fine-grained set index to obtain the processing core status information that meets the mapping requirements;
[0050] Sorting different categories of processing core state information according to fine-grained sizes, and generating mapping rules for different categories of processing core state information;
[0051] If the processing core state information is greater than the fine-grained setting index, it is determined to be coarse-grained processing core state information, and the coarse-grained processing core state information is subdivided into secondary granularity.
[0052] In the third aspect, an embodiment of the present application also provides a computer-readable storage medium, which includes a fine-grained mapping method program for a storage and computing integrated chip. When the fine-grained mapping method program for the storage and computing integrated chip is executed by a processor, the steps of the fine-grained mapping method for the storage and computing integrated chip as described in any one of the above items are implemented.
[0053] From the above, it can be seen that the fine-grained mapping method, system and medium of a storage and computing integrated chip provided in the embodiments of the present application have the following technical effects: fine-grained screening of the processing core status is performed by analyzing the chip threads, so as to realize dynamic mapping of the processing core running data according to the established mapping rules, and during the analysis process, granular analysis is performed based on the processing core status information to screen out fine-grained processing core status information, sort the processing core status information, generate the best mapping rules, improve the quantitative accuracy of the mapping, and increase the robustness of the system. BRIEF DESCRIPTION OF THE DRAWINGS
[0054] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.
[0055] Figure 1 A flowchart of a fine-grained mapping method for a storage-computing integrated chip provided in an embodiment of the present application;
[0056] Figure 2 A flowchart of processing core state information analysis of a fine-grained mapping method for a storage-computing integrated chip provided in an embodiment of the present application;
[0057] Figure 3 A flow chart of a mapping rule generation method for a fine-grained mapping method for a storage-computing integrated chip provided in an embodiment of the present application. DETAILED DESCRIPTION
[0058] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.
[0059] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.
[0060] Please refer to Figure 1 , Figure 1 The flowchart of a fine-grained mapping method of a storage-computing integrated chip in some embodiments of the present application. The fine-grained mapping method of the storage-computing integrated chip is used in a terminal device, and the fine-grained mapping method of the storage-computing integrated chip includes the following steps:
[0061] S101, obtaining chip parameters, analyzing thread status based on the chip parameters, and processing core status information according to the thread status analysis;
[0062] S102, performing granular analysis based on the processing core state information, screening out fine-grained processing core state information, sorting the processing core state information, and generating a mapping rule;
[0063] S103, based on the difference between the processing core state information analysis and the preset state, determining whether the difference is greater than or equal to a set difference threshold;
[0064] S104, if it is greater than or equal to, generating correction information, and adjusting the fine-grained screening criteria based on the correction information;
[0065] S105: If it is less than, mapping the processing core operation data according to the mapping rule to obtain a mapping result, and transmitting the mapping result to the terminal in real time.
[0066] It should be noted that by analyzing the thread status of the chip and performing fine-grained sorting, different mapping rules are established to improve the mapping accuracy of the processing core running data.
[0067] Please refer to Figure 2 , Figure 2 This is a flowchart of processing core status information analysis of a fine-grained mapping method for a storage-computing integrated chip in some embodiments of the present application. According to an embodiment of the present invention, obtaining chip parameters, analyzing thread status based on chip parameters, and analyzing processing core status information according to thread status specifically include:
[0068] S201, obtaining chip model, chip power and chip memory resource information, and dividing the chip into multiple processing cores based on the chip model and chip power;
[0069] S202, analyzing multiple processing core threads and coordination information based on dynamic change data of chip memory resource information;
[0070] S203, coordinating the different processing core threads based on the coordination information to obtain coordination data;
[0071] S204, generating compensation information according to the collaborative data, and adjusting the operation data of the single processing core based on the compensation information to obtain the processing core state information.
[0072] It should be noted that different chip models correspond to different chip memories. By analyzing the changes in chip memory resource information, analyzing the collaborative information between processing core threads and different threads, dynamic compensation is performed on a single processing core to improve the accuracy of obtaining processing core status information.
[0073] Please refer to Figure 3 , Figure 3 This is a flowchart of a mapping rule generation method of a fine-grained mapping method for a storage-computing integrated chip in some embodiments of the present application. According to an embodiment of the present invention, granular analysis is performed based on the processing core state information, fine-grained processing core state information is screened out, the processing core state information is sorted, and mapping rules are generated, specifically including:
[0074] S301, acquiring processing core state information, and classifying the processing core state information according to a plurality of granularity standards to obtain processing core state information of different categories;
[0075] S302, comparing different categories of processing core status information with fine-grained setting indicators;
[0076] S303, filtering the processing core status information that is less than or equal to the fine-grained setting index to obtain the processing core status information that meets the mapping requirements;
[0077] S304, sorting different types of processing core state information according to fine-grained sizes, and generating mapping rules for different types of processing core state information;
[0078] S305: If the processing core state information is greater than the fine-grained setting index, it is determined to be coarse-grained processing core state information, and the coarse-grained processing core state information is subdivided into secondary granularity.
[0079] It should be noted that by analyzing and classifying the core status information, the core status information is screened and sorted according to fine-grained setting indicators, and mapping rules of different categories are established.
[0080] According to an embodiment of the present invention, based on the difference between the processing core state information analysis and the preset state, determining whether the difference is greater than or equal to a set difference threshold specifically includes:
[0081] Obtaining the status information of the processing cores at different time nodes, comparing the status information of the processing cores at different nodes, and obtaining status difference information;
[0082] Compare the state difference information with the preset state to obtain the degree of difference;
[0083] The difference is compared with a set difference threshold to determine whether the difference exceeds the set difference threshold.
[0084] It should be noted that by analyzing the processing core state information at different time nodes, the difference in the processing core state information between different time nodes is obtained, so as to determine whether the difference exceeds the set difference threshold, thereby improving the accuracy of difference analysis.
[0085] According to an embodiment of the present invention, if it is greater than or equal to, then correction information is generated, and the fine-grained screening criteria are adjusted based on the correction information, specifically including:
[0086] Obtaining a difference and a set difference threshold, where the set difference threshold includes a first difference threshold and a second difference threshold, and the first difference threshold is less than the second difference threshold;
[0087] If the difference is greater than the first difference threshold and less than the second difference threshold, generating first correction information, setting a first adjustment rule based on the first correction information, and adjusting the fine-grained screening criteria according to the first adjustment rule;
[0088] If the difference is greater than or equal to the second difference threshold, second correction information is generated, a second adjustment rule is set based on the second correction information, and the fine-grained screening standard is adjusted according to the second adjustment rule.
[0089] It should be noted that by analyzing the difference between the difference and the set difference threshold, different correction information is generated for different situations, and the fine-grained screening criteria are adjusted in different ways according to different correction information to improve the fine-grained screening effect.
[0090] According to an embodiment of the present invention, if , is less than , mapping the processing core operation data according to the mapping rule to obtain a mapping result, and transmitting the mapping result to the terminal in real time, specifically including:
[0091] Acquire mapping rules and processing core state information, divide the processing core state information into fine-grained categories, and obtain different categories of processing core state information;
[0092] Matching the mapping rules with different categories of processing core state information to obtain a matching degree;
[0093] sorting the mapping matching degrees of different categories of processing core state information to generate a mapping order;
[0094] Mapping different types of processing core state information in sequence according to a mapping order to obtain multiple mapping results;
[0095] The multiple mapping results are transmitted to the terminal for storage respectively.
[0096] It should be noted that by analyzing the matching degree between the mapping rule and the processing core status information, different categories of processing core status information are mapped in sequence, and different mapping results are transmitted to the terminal, thereby realizing dynamic transmission of the mapping results.
[0097] According to an embodiment of the present invention, the method further includes: obtaining chip memory resource occupancy information at different time nodes, and calculating the chip resource occupancy rate based on the chip resource occupancy information;
[0098] Determine whether the chip resource occupancy rate is greater than or equal to a set occupancy rate threshold;
[0099] If it is greater than or equal to, adjust the coordination parameters between multiple threads;
[0100] If it is less than, the memory resource fluctuation information of a single thread is obtained.
[0101] It should be noted that the chip resource occupancy rate at different time nodes can be used to analyze whether the threads meet the requirements, thereby dynamically adjusting the coordination parameters between threads and improving the coordination effect between different threads.
[0102] In a second aspect, an embodiment of the present application provides a fine-grained mapping system for a storage-computing integrated chip, the system comprising: a memory and a processor, the memory comprising a program of a fine-grained mapping method for the storage-computing integrated chip, and when the program of the fine-grained mapping method for the storage-computing integrated chip is executed by the processor, the following steps are implemented:
[0103] Obtain chip parameters, analyze thread status based on chip parameters, and process core status information based on thread status analysis;
[0104] Perform granular analysis based on the processing core status information, filter out fine-grained processing core status information, sort the processing core status information, and generate mapping rules;
[0105] Based on the difference between the processing core state information analysis and the preset state, determining whether the difference is greater than or equal to a set difference threshold;
[0106] If it is greater than or equal to, then generate correction information, and adjust the fine-grained screening criteria based on the correction information;
[0107] If it is less than, the processing core operation data is mapped according to the mapping rule to obtain a mapping result, and the mapping result is transmitted to the terminal in real time.
[0108] It should be noted that by analyzing the thread status of the chip and performing fine-grained sorting, different mapping rules are established to improve the mapping accuracy of the processing core running data.
[0109] According to an embodiment of the present invention, obtaining chip parameters, analyzing thread status based on the chip parameters, and analyzing and processing core status information according to the thread status specifically include:
[0110] Obtain chip model, chip power and chip memory resource information, and divide the chip into multiple processing cores based on the chip model and chip power;
[0111] Analyze multiple processing core threads and coordination information based on dynamic changes in chip memory resource information;
[0112] Based on the collaborative information, different processing core threads are operated collaboratively to obtain collaborative data;
[0113] Compensation information is generated according to the collaborative data, and the operation data of a single processing core is adjusted based on the compensation information to obtain the processing core state information.
[0114] It should be noted that different chip models correspond to different chip memories. By analyzing the changes in chip memory resource information, analyzing the collaborative information between processing core threads and different threads, dynamic compensation is performed on a single processing core to improve the accuracy of obtaining processing core status information.
[0115] According to an embodiment of the present invention, granular analysis is performed based on the processing core state information, fine-grained processing core state information is screened out, the processing core state information is sorted, and a mapping rule is generated, which specifically includes:
[0116] Acquire the processing core state information, classify the processing core state information according to multiple granularity standards, and obtain different categories of processing core state information;
[0117] Compare different categories of processing core status information with fine-grained set indicators;
[0118] Filter the processing core status information that is less than or equal to the fine-grained set index to obtain the processing core status information that meets the mapping requirements;
[0119] Sorting different categories of processing core state information according to fine-grained sizes, and generating mapping rules for different categories of processing core state information;
[0120] If the processing core state information is greater than the fine-grained setting index, it is determined to be coarse-grained processing core state information, and the coarse-grained processing core state information is subdivided into secondary granularity.
[0121] It should be noted that by analyzing and classifying the core status information, the core status information is screened and sorted according to fine-grained setting indicators, and mapping rules of different categories are established.
[0122] According to an embodiment of the present invention, based on the difference between the processing core state information analysis and the preset state, determining whether the difference is greater than or equal to a set difference threshold specifically includes:
[0123] Obtaining the status information of the processing cores at different time nodes, comparing the status information of the processing cores at different nodes, and obtaining status difference information;
[0124] Compare the state difference information with the preset state to obtain the degree of difference;
[0125] The difference is compared with a set difference threshold to determine whether the difference exceeds the set difference threshold.
[0126] It should be noted that by analyzing the processing core state information at different time nodes, the difference in the processing core state information between different time nodes is obtained, so as to determine whether the difference exceeds the set difference threshold, thereby improving the accuracy of difference analysis.
[0127] According to an embodiment of the present invention, if it is greater than or equal to, then correction information is generated, and the fine-grained screening criteria are adjusted based on the correction information, specifically including:
[0128] Obtaining a difference and a set difference threshold, where the set difference threshold includes a first difference threshold and a second difference threshold, and the first difference threshold is less than the second difference threshold;
[0129] If the difference is greater than the first difference threshold and less than the second difference threshold, generating first correction information, setting a first adjustment rule based on the first correction information, and adjusting the fine-grained screening criteria according to the first adjustment rule;
[0130] If the difference is greater than or equal to the second difference threshold, second correction information is generated, a second adjustment rule is set based on the second correction information, and the fine-grained screening standard is adjusted according to the second adjustment rule.
[0131] It should be noted that by analyzing the difference between the difference and the set difference threshold, different correction information is generated for different situations, and the fine-grained screening criteria are adjusted in different ways according to different correction information to improve the fine-grained screening effect.
[0132] According to an embodiment of the present invention, if , is less than , mapping the processing core operation data according to the mapping rule to obtain a mapping result, and transmitting the mapping result to the terminal in real time, specifically including:
[0133] Acquire mapping rules and processing core state information, divide the processing core state information into fine-grained categories, and obtain different categories of processing core state information;
[0134] Matching the mapping rules with different categories of processing core state information to obtain a matching degree;
[0135] sorting the mapping matching degrees of different categories of processing core state information to generate a mapping order;
[0136] Mapping different types of processing core state information in sequence according to a mapping order to obtain multiple mapping results;
[0137] The multiple mapping results are transmitted to the terminal for storage respectively.
[0138] It should be noted that by analyzing the matching degree between the mapping rule and the processing core status information, different categories of processing core status information are mapped in sequence, and different mapping results are transmitted to the terminal, thereby realizing dynamic transmission of the mapping results.
[0139] According to an embodiment of the present invention, the method further includes: obtaining chip memory resource occupancy information at different time nodes, and calculating the chip resource occupancy rate based on the chip resource occupancy information;
[0140] Determine whether the chip resource occupancy rate is greater than or equal to a set occupancy rate threshold;
[0141] If it is greater than or equal to, adjust the coordination parameters between multiple threads;
[0142] If it is less than, the memory resource fluctuation information of a single thread is obtained.
[0143] It should be noted that the chip resource occupancy rate at different time nodes can be used to analyze whether the threads meet the requirements, thereby dynamically adjusting the coordination parameters between threads and improving the coordination effect between different threads.
[0144] The third aspect of the present invention provides a computer-readable storage medium, which includes a fine-grained mapping method program for a storage-computing integrated chip. When the fine-grained mapping method program for a storage-computing integrated chip is executed by a processor, the steps of the fine-grained mapping method for a storage-computing integrated chip as described in any of the above items are implemented.
[0145] The present invention discloses a fine-grained mapping method, system and medium for a storage-computing integrated chip, which obtains chip parameters, analyzes thread states based on the chip parameters, and analyzes core state information according to the thread states; performs granular analysis based on the core state information, screens out fine-grained core state information, sorts the core state information, and generates mapping rules; determines whether the difference between the core state information analysis and a preset state is greater than or equal to a set difference threshold; if greater than or equal to, generates correction information, and adjusts the fine-grained screening standard based on the correction information; if less than, maps the core operation data according to the mapping rule to obtain a mapping result, and transmits the mapping result to a terminal in real time; performs fine-grained screening on the core state by analyzing the chip thread, thereby dynamically mapping the core operation data according to the established mapping rule.
[0146] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical or other forms.
[0147] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.
[0148] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.
[0149] Those skilled in the art can understand that: all or part of the steps of implementing the above method embodiments can be completed by hardware related to program instructions, the aforementioned program can be stored in a readable storage medium, and when the program is executed, it executes the steps of the above method embodiments; and the aforementioned storage medium includes: mobile storage devices, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), disks or optical disks, and other media that can store program codes.
[0150] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.
Claims
1. A fine-grained mapping method for a storage-computing integrated chip, characterized in that: include: Obtain chip parameters, analyze thread status based on chip parameters, and analyze core status information based on thread status; Perform granular analysis based on the processing core status information, filter out fine-grained processing core status information, sort the processing core status information, and generate mapping rules; Based on the difference between the processing core state information analysis and the preset state, determining whether the difference is greater than or equal to a set difference threshold; If it is greater than or equal to, then generate correction information, and adjust the fine-grained screening criteria based on the correction information; If it is less than, the processing core operation data is mapped according to the mapping rule to obtain a mapping result, and the mapping result is transmitted to the terminal in real time; Obtain chip parameters, analyze thread status based on chip parameters, and analyze and process core status information based on thread status, including: Obtain chip model, chip power and chip memory resource information, and divide the chip into multiple processing cores based on the chip model and chip power; Analyze multiple processing core threads and coordination information based on dynamic changes in chip memory resource information; Based on the collaborative information, different processing core threads are operated collaboratively to obtain collaborative data; Generate compensation information according to the collaborative data, adjust the operation data of a single processing core based on the compensation information, and obtain the processing core state information; Granularity analysis is performed based on the processing core status information to filter out fine-grained processing core status information, sort the processing core status information, and generate mapping rules, including: Acquire the processing core state information, classify the processing core state information according to multiple granularity standards, and obtain different categories of processing core state information; Compare different categories of processing core status information with fine-grained set indicators; Filter the processing core status information that is less than or equal to the fine-grained set index to obtain the processing core status information that meets the mapping requirements; Sorting different categories of processing core state information according to fine-grained sizes, and generating mapping rules for different categories of processing core state information; If the processing core state information is greater than the fine-grained setting index, it is determined to be coarse-grained processing core state information, and the coarse-grained processing core state information is subdivided into secondary granularity.
2. The fine-grained mapping method for a storage-computing integrated chip according to claim 1, characterized in that: Based on the difference between the processing core state information analysis and the preset state, determining whether the difference is greater than or equal to a set difference threshold specifically includes: Obtaining the status information of the processing cores at different time nodes, comparing the status information of the processing cores at different nodes, and obtaining status difference information; Compare the state difference information with the preset state to obtain the degree of difference; The difference is compared with a set difference threshold to determine whether the difference exceeds the set difference threshold.
3. The fine-grained mapping method for a storage-computing integrated chip according to claim 2, characterized in that: If it is greater than or equal to, then generate correction information, and adjust the fine-grained screening criteria based on the correction information, including: Obtaining a difference and a set difference threshold, where the set difference threshold includes a first difference threshold and a second difference threshold, and the first difference threshold is less than the second difference threshold; If the difference is greater than the first difference threshold and less than the second difference threshold, generating first correction information, setting a first adjustment rule based on the first correction information, and adjusting the fine-grained screening criteria according to the first adjustment rule; If the difference is greater than or equal to the second difference threshold, second correction information is generated, a second adjustment rule is set based on the second correction information, and the fine-grained screening standard is adjusted according to the second adjustment rule.
4. The fine-grained mapping method for a storage-computing integrated chip according to claim 3, characterized in that: If it is less than, the processing core operation data is mapped according to the mapping rule to obtain a mapping result, and the mapping result is transmitted to the terminal in real time, specifically including: Acquire mapping rules and processing core state information, divide the processing core state information into fine-grained categories, and obtain different categories of processing core state information; Matching the mapping rules with different categories of processing core state information to obtain a matching degree; sorting the mapping matching degrees of different categories of processing core state information to generate a mapping order; Mapping different types of processing core state information in sequence according to a mapping order to obtain multiple mapping results; The multiple mapping results are transmitted to the terminal for storage respectively.
5. A fine-grained mapping system for a storage-computing integrated chip, characterized in that: The system includes: a memory and a processor, wherein the memory includes a program of a fine-grained mapping method for a storage-computing integrated chip, and when the program of the fine-grained mapping method for the storage-computing integrated chip is executed by the processor, the following steps are implemented: Obtain chip parameters, analyze thread status based on chip parameters, and analyze core status information based on thread status; Perform granular analysis based on the processing core status information, filter out fine-grained processing core status information, sort the processing core status information, and generate mapping rules; Based on the difference between the processing core state information analysis and the preset state, determining whether the difference is greater than or equal to a set difference threshold; If it is greater than or equal to, then generate correction information, and adjust the fine-grained screening criteria based on the correction information; If it is less than, the processing core operation data is mapped according to the mapping rule to obtain a mapping result, and the mapping result is transmitted to the terminal in real time; Obtain chip parameters, analyze thread status based on chip parameters, and analyze and process core status information based on thread status, including: Obtain chip model, chip power and chip memory resource information, and divide the chip into multiple processing cores based on the chip model and chip power; Analyze multiple processing core threads and coordination information based on dynamic changes in chip memory resource information; Based on the collaborative information, different processing core threads are operated collaboratively to obtain collaborative data; Generate compensation information according to the collaborative data, adjust the operation data of a single processing core based on the compensation information, and obtain the processing core state information; Granularity analysis is performed based on the processing core status information to filter out fine-grained processing core status information, sort the processing core status information, and generate mapping rules, including: Acquire the processing core state information, classify the processing core state information according to multiple granularity standards, and obtain different categories of processing core state information; Compare different categories of processing core status information with fine-grained set indicators; Filter the processing core status information that is less than or equal to the fine-grained set index to obtain the processing core status information that meets the mapping requirements; Sorting different categories of processing core state information according to fine-grained sizes, and generating mapping rules for different categories of processing core state information; If the processing core state information is greater than the fine-grained setting index, it is determined to be coarse-grained processing core state information, and the coarse-grained processing core state information is subdivided into secondary granularity.
6. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a fine-grained mapping method program for a storage-computing integrated chip. When the fine-grained mapping method program for the storage-computing integrated chip is executed by a processor, the steps of the fine-grained mapping method for the storage-computing integrated chip as described in any one of claims 1 to 4 are implemented.
