Method and equipment for selecting mapping strategy

By using neural network models to analyze performance parameters and formulas in the decoding unit, selecting the mapping strategy with the best performance, solving the problem of matching complex tasks and hardware configurations in the existing technology, and achieving high-performance and low-power mapping strategy selection .

CN118916152APending Publication Date: 2024-11-08SHANGHAI CAMBRICON INFORMATION TECH CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202310508404.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-05-06
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The existing technology is difficult to effectively deal with the matching of complex computer tasks and hardware configurations when selecting mapping strategies, which makes it difficult to achieve the goal of high performance and low power consumption, mainly because developers have difficulty covering all possibilities to make ideal mapping solutions. .

Method used

Using a decoding unit including predictors and classifiers, the performance parameters and formulas of the task are imported into the neural network model for analysis. Based on the analysis results of the neural network model, the mapping strategy with the best performance is selected, and the performance formulas and parameters in the training set are used for Training to achieve efficient mapping strategy selection.

Benefits of technology

Through computer computing performance formulas, better mapping solutions are comprehensively and efficiently analyzed or speculated, and the technical effect of high performance and low power consumption is achieved, solving the problem of errors in mapping strategy selection in the existing technology.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118916152A_ABST
    Figure CN118916152A_ABST
Patent Text Reader

Abstract

The present disclosure relates to a method and apparatus for selecting a mapping policy, where a computing device of the present disclosure is included in an integrated circuit device that includes a universal interconnection interface and other processing devices. And the computing device interacts with other processing devices to jointly complete the computing operation specified by the user. The integrated circuit device can further comprise a storage device, and the storage device is connected with the computing device and the other processing devices and used for data storage of the computing device and the other processing devices.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure generally relates to the field of computers. More specifically, the present disclosure relates to a decoding unit, an integrated circuit device, a board, a method, a computer-readable storage medium, a computer program product, and a computer device for selecting a mapping strategy. Background Art

[0002] When using a computer to perform a task, the task needs to be split based on hardware conditions in order to make good use of the hardware configuration and efficiently perform the task. This process is called mapping. In the process of mapping the task to the chip computing and memory access devices, many hardware splitting and optimization strategies are involved. The performance and power consumption effects of each mapping strategy are different. If the selection is inappropriate, hardware resources will be wasted. Therefore, it can be understood that the goal of the final selected mapping strategy should be to tend to high performance and / or low power consumption solutions.

[0003] The existing selection method is to list the performance of a limited number of typical task scenarios, and developers manually calculate the performance results, compare and classify them to obtain a suitable mapping strategy. As computer tasks and hardware configurations become increasingly complex, it is difficult to obtain an ideal mapping solution based solely on intuitive analysis or speculation by developers, which is likely to lead to errors in mapping strategy selection and failure to achieve the goals of high performance and low power consumption. Summary of the invention

[0004] In order to at least solve one or more of the technical problems mentioned above, the present disclosure proposes a decoding unit, an integrated circuit device, a board, a method, a computer-readable storage medium, a computer program product, and a computer device for selecting a mapping strategy in multiple aspects.

[0005] In the first aspect, the present disclosure provides a decoding unit for selecting a mapping strategy for a task, wherein the mapping strategy represents an implementation scheme for mapping the task to the computing device and memory access device of the processor. The decoding unit includes a predictor and a classifier. The predictor is used to import a performance parameter and b performance formula of the task into a neural network model for analysis, wherein the b performance formulas are performance expressions corresponding to the b mapping strategies one by one; the performance parameter is a parameter related to the performance index in the performance formula, and the performance index is used to quantify the performance of the mapping strategy in different scenarios; a and b are integers greater than one respectively. The predictor is also used to obtain the performance values ​​corresponding to the b mapping strategies respectively based on the analysis results of the neural network model. The classifier is used to select the mapping strategy corresponding to the optimal performance value.

[0006] In a second aspect, the present disclosure provides an integrated circuit device, comprising the above-mentioned decoding unit.

[0007] In a third aspect, the present disclosure provides a board comprising the above-mentioned integrated circuit device.

[0008] In a fourth aspect, the present disclosure provides an integrated circuit device for selecting a mapping strategy, including a processing device and a computing device, the processing device being used to: obtain a training set of the mapping strategy, the training set including a performance formula corresponding to the computing device and its corresponding performance parameters, the performance formula being a performance expression corresponding one-to-one to the mapping strategy, the performance parameter being a parameter in the performance formula related to the performance indicator, the performance indicator being used to quantify the performance of the mapping strategy in different scenarios; and importing the training set into a neural network model for training.

[0009] In a fifth aspect, the present disclosure provides a method for selecting a mapping strategy for a task, wherein the mapping strategy represents an implementation scheme for mapping the task to the computing device and memory access device of the processor. The method comprises: importing a performance parameter and b performance formula of the task into a neural network model for analysis, wherein the b performance formulas are performance expressions corresponding to the b mapping strategies one by one, the performance parameters are parameters related to the performance index in the performance formula, the performance index is used to quantify the performance of the mapping strategy in different scenarios, and a and b are integers greater than one respectively; based on the analysis results of the neural network model, obtaining the performance values ​​corresponding to the b mapping strategies respectively; and selecting the mapping strategy corresponding to the optimal performance value.

[0010] In a sixth aspect, the present disclosure provides a method for selecting a mapping strategy, the method comprising: obtaining a training set of the mapping strategy; the training set comprises a performance formula and its corresponding performance parameters, the performance formula being a performance expression corresponding one-to-one to the mapping strategy; the performance parameters are parameters in the performance formula related to the performance indicator, the performance indicator being used to quantify the performance of the mapping strategy in different scenarios; and importing the training set into a neural network model for training.

[0011] In a seventh aspect, the present disclosure provides a computer-readable storage medium having stored thereon a computer program code for selecting a mapping strategy, and when the computer program code is executed by a processing device, any of the above methods is executed.

[0012] In an eighth aspect, the present disclosure provides a computer program product, comprising a computer program for selecting a mapping strategy, wherein the computer program implements the steps of any of the above methods when executed by a processor.

[0013] In a ninth aspect, the present disclosure provides a computer device, comprising a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of any one of the above methods.

[0014] The present disclosure uses computer calculation performance formulas to cope with complex computer tasks and hardware configurations, and comprehensively and efficiently analyze or infer better mapping solutions to achieve high performance and low power consumption technical effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] By reading the detailed description below with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present disclosure will become readily understood. In the accompanying drawings, several embodiments of the present disclosure are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0016] Figure 1 is a structural diagram of a board card showing an embodiment of the present disclosure;

[0017] Figure 2 is a structural diagram showing an integrated circuit device according to an embodiment of the present disclosure;

[0018] Figure 3 is a schematic diagram showing the internal structure of a computing device according to an embodiment of the present disclosure;

[0019] Figure 4 is a schematic diagram showing the internal structure of a processor core according to an embodiment of the present disclosure;

[0020] Figure 5 is a schematic diagram showing when a processor core wants to write data to a processor core of another cluster;

[0021] Figure 6 is a schematic diagram showing that a decoding unit of an embodiment of the present disclosure selects a mapping strategy for a task;

[0022] Figure 7 is a structural diagram showing a decoding unit of an embodiment of the present disclosure;

[0023] Figure 8 is a flow chart showing the selection of a mapping strategy for a convolution task by a decoding unit according to an embodiment of the disclosure;

[0024] Fig. 9 is a simplified schematic diagram showing a decoding unit selection mapping strategy according to an embodiment of the present disclosure;

[0025] Fig.10 FIG. 4 is a flowchart showing a method of selecting a mapping strategy according to another embodiment of the present disclosure. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of the present disclosure to clearly and completely describe the technical solutions in the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present disclosure.

[0027] It should be understood that the terms "include" and "comprising" used in the specification and claims of the present disclosure indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0028] It should also be understood that the terms used in this disclosure are only for the purpose of describing specific embodiments and are not intended to limit the disclosure. As used in this disclosure and claims, the singular forms of "a", "an", and "the" are intended to include the plural forms unless the context clearly indicates otherwise. It should also be further understood that the term "and / or" used in this disclosure and claims refers to any combination of one or more of the associated listed items and all possible combinations, including these combinations.

[0029] As used in this specification and claims, the term "if" may be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context.

[0030] The specific implementation of the present disclosure is described in detail below with reference to the accompanying drawings.

[0031] Figure 1 FIG. 1 is a schematic diagram showing the structure of a board 10 according to an embodiment of the present disclosure. Figure 1 As shown, the board 10 includes a chip 101, which is a system-on-chip (SoC), or system on chip, integrated with one or more combined processing devices, which are artificial intelligence computing units that support various deep learning and machine learning algorithms to meet the intelligent processing needs in complex scenarios in the fields of computer vision, speech, natural language processing, data mining, etc. In particular, deep learning technology is widely used in the field of cloud intelligence. A notable feature of cloud intelligence applications is the large amount of input data, which places high demands on the storage capacity and computing power of the platform. The board 10 of this embodiment is suitable for cloud intelligence applications and has huge off-chip storage, on-chip storage, and a large amount of computing power.

[0032] The chip 101 is connected to an external device 103 via an external interface device 102. The external device 103 is, for example, a server, a computer, a camera, a display, a mouse, a keyboard, a network card, or a wifi interface. The data to be processed can be transmitted from the external device 103 to the chip 101 via the external interface device 102. The calculation result of the chip 101 can be transmitted back to the external device 103 via the external interface device 102. According to different application scenarios, the external interface device 102 can have different interface forms, such as a PCIe interface.

[0033] The board 10 also includes a storage device 104 for storing data, which includes one or more storage units 105. The storage device 104 is connected to the control device 106 and the chip 101 through a bus and transmits data. The control device 106 in the board 10 is configured to control the state of the chip 101. To this end, in an application scenario, the control device 106 may include a microcontroller (MCU).

[0034] Figure 2 1 is a block diagram showing the combined processing device in the chip 101 of this embodiment. Figure 2 As shown in , the combined processing device 20 includes a computing device 201 , an interface device 202 , a processing device 203 and a storage device 204 .

[0035] The computing device 201 is configured to execute user-specified operations, and is mainly implemented as a single-core intelligent processor or a multi-core intelligent processor to perform deep learning or machine learning calculations. It can interact with the processing device 203 through the interface device 202 to jointly complete the user-specified operations.

[0036] The interface device 202 is used to transmit data and control instructions between the computing device 201 and the processing device 203. For example, the computing device 201 can obtain input data from the processing device 203 via the interface device 202 and write it into the storage device on the computing device 201 chip. Further, the computing device 201 can obtain control instructions from the processing device 203 via the interface device 202 and write them into the control cache on the computing device 201 chip. Alternatively or optionally, the interface device 202 can also read data in the storage device of the computing device 201 and transmit it to the processing device 203.

[0037] The processing device 203, as a general processing device, performs basic controls including but not limited to data handling, starting and / or stopping the computing device 201. Depending on the implementation, the processing device 203 can be a central processing unit (CPU), a graphics processing unit (GPU) or one or more types of processors in other general and / or special processors, these processors include but are not limited to digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., and their number can be determined according to actual needs. As mentioned above, only with respect to the computing device 201 disclosed in the present invention, it can be regarded as having a single-core structure or a homogeneous multi-core structure. However, when the computing device 201 and the processing device 203 are integrated and considered together, the two are regarded as forming a heterogeneous multi-core structure.

[0038] The storage device 204 is used to store data to be processed, which may be DRAM or DDR memory, and is usually 16G or larger in size, and is used to store data of the computing device 201 and / or the processing device 203 .

[0039] Figure 3 The figure shows a schematic diagram of the internal structure of the computing device 201. The computing device 201 is used to process input data such as computer vision, speech, natural language, and data mining. The computing device 201 in the figure adopts a multi-core hierarchical structure design. The computing device 201 is a system on a chip, which includes multiple computing clusters, and each computing cluster includes multiple processor cores. In other words, the computing device 201 is composed of a hierarchy of system on a chip-computing cluster-processor core.

[0040] At the system-on-chip level, Figure 3 As shown, the computing device 201 includes an external storage controller 301 , a peripheral communication module 302 , an on-chip interconnect module 303 , a synchronization module 304 , and multiple computing clusters 305 .

[0041] There may be multiple external storage controllers 301, two of which are shown in the figure as an example, and are used to respond to access requests issued by the processor core and access external storage devices, such as Figure 2The storage device 204 in the chip is used to read data from outside the chip or write data. The peripheral communication module 302 is used to receive the control signal from the processing device 203 through the interface device 202 to start the computing device 201 to perform tasks. The on-chip interconnect module 303 connects the external storage controller 301, the peripheral communication module 302 and multiple computing clusters 305 to transmit data and control signals between each module. The synchronization module 304 is a global synchronization barrier controller (GBC) used to coordinate the work progress of each computing cluster and ensure information synchronization. Multiple computing clusters 305 are the computing cores of the computing device 201. Four are shown as examples in the figure. With the development of hardware, the computing device 201 disclosed in the present invention can also include 8, 16, 64, or even more computing clusters 305. The computing clusters 305 are used to efficiently execute deep learning algorithms.

[0042] From the perspective of computing clusters, Figure 3 As shown, each computing cluster 305 includes multiple processor cores (IPU cores) 306 and a memory core (MEM core) 307 .

[0043] The figure shows four processor cores 306 as an example, but the present disclosure does not limit the number of processor cores 306. Figure 4 Each processor core 306 includes three modules: a control module 41 , a computing module 42 and a storage module 43 .

[0044] The control module 41 is used to coordinate and control the operation of the computing module 42 and the storage module 43 to complete the deep learning task, and includes an instruction fetch unit (IFU) 411 and an instruction decode unit (IDU) 412. The instruction fetch unit 411 is used to obtain instructions from the processing device 203, and the instruction decode unit 412 decodes the obtained instructions and sends the decoding results to the computing module 42 and the storage module 43 as control information.

[0045] The operation module 42 includes a vector operation unit 421 and a matrix operation unit 422. The vector operation unit 421 is used to perform vector operations and can support complex operations such as vector multiplication, addition, and nonlinear transformation; the matrix operation unit 422 is responsible for the core calculation of the deep learning algorithm, namely matrix multiplication and convolution.

[0046] The storage module 43 is used to store or move related data, including a neuron RAM (NRAM) 431, a weight RAM (WRAM) 432, an input / output direct memory access module (IODMA) 433, and a move direct memory access module (MVDMA) 434. NRAM 431 is used to store input and output data and intermediate results for calculation by the processor core 306; WRAM 432 is used to store the weights of the deep learning network; IODMA 433 controls the memory access between NRAM 431 / WRAM 432 and the storage device 204 through the broadcast bus 309; MVDMA 434 is used to control the memory access between NRAM 431 / WRAM 432 and SRAM 308. It should be noted that the NRAM and WRAM here can be two storage areas formed by dividing the same memory in the logical storage space, or can be two independent memories, which is not specifically limited here.

[0047] Back to Figure 3 The storage core 307 is mainly used for storage and communication, that is, to store shared data or intermediate results between the processor cores 306, and to perform communication between the computing cluster 305 and the storage device 204, between the computing clusters 305, and between the processor cores 306. In other embodiments, the storage core 307 has the ability of scalar operation and is used to perform scalar operation.

[0048] The storage core 307 includes a shared memory unit (SRAM) 308, a broadcast bus 309, a cluster direct memory access module (CDMA) 310, and a global direct memory access module (GDMA) 311. The SRAM 308 plays the role of a high-performance data transfer station. The data reused between different processor cores 306 in the same computing cluster 305 does not need to be obtained from the storage device 204 by each processor core 306, but is transferred between the processor cores 306 via the SRAM 308. The storage core 307 only needs to quickly distribute the reused data from the SRAM 308 to multiple processor cores 306, so as to improve the efficiency of inter-core communication and greatly reduce on-chip and off-chip input / output access.

[0049] The broadcast bus 309, CDMA 310 and GDMA 311 are respectively used to perform communication between processor cores 306, communication between computing clusters 305 and data transmission between computing clusters 305 and storage device 204. They will be described below.

[0050] The broadcast bus 309 is used to complete high-speed communication between the processor cores 306 in the computing cluster 305. The broadcast bus 309 of this embodiment supports inter-core communication modes including unicast, multicast and broadcast. Unicast refers to point-to-point (i.e., single processor core to single processor core) data transmission, multicast is a communication mode of transmitting a copy of data from SRAM 308 to specific processor cores 306, and broadcast is a communication mode of transmitting a copy of data from SRAM 308 to all processor cores 306, which is a special case of multicast.

[0051] CDMA 310 is used to control memory access to SRAM 308 between different computing clusters 305 within the same computing device 201 . Figure 5 A schematic diagram is shown when a processor core wants to write data to a processor core of another computing cluster, to illustrate the working principle of CDMA310. In this application scenario, the same computing device includes multiple computing clusters. For the convenience of explanation, only computing cluster 0 and computing cluster 1 are shown in the figure. Computing cluster 0 and computing cluster 1 each include multiple processor cores. Also for the convenience of explanation, computing cluster 0 in the figure only shows processor core 0, and computing cluster 1 only shows processor core 1. Processor core 0 wants to write data to processor core 1.

[0052] First, processor core 0 sends a unicast write request to write data into the local SRAM 0. CDMA 0 acts as the master and CDMA1 acts as the slave. The master pushes the write request to the slave, that is, the master sends the write address AW and write data W to transfer the data to SRAM 1 of computing cluster 1. Then the slave sends a write response B in response. Finally, processor core 1 of computing cluster 1 sends a unicast read request to read the data from SRAM 1.

[0053] Back to Figure 3, GDMA 311 cooperates with external memory controller 301 to control the access of SRAM 308 of computing cluster 305 to storage device 204, or reads data from storage device 204 to SRAM 308. As can be seen from the above, the communication between storage device 204 and NRAM 431 or WRAM 432 can be realized through two channels. The first channel is to directly connect storage device 204 and NRAM 431 or WRAM 432 through IODAM 433; the second channel is to first transmit data between storage device 204 and SRAM 308 through GDMA 311, and then transmit data between SRAM 308 and NRAM 431 or WRAM 432 through MVDMA 434. Although it seems that the second channel requires more components to participate and the data flow is longer, in fact, in some embodiments, the bandwidth of the second channel is much greater than that of the first channel, so the communication between storage device 204 and NRAM 431 or WRAM 432 may be more efficient through the second channel. The embodiments of the present disclosure can select a data transmission channel according to the hardware conditions thereof.

[0054] In other embodiments, the functions of GDMA 311 and IODMA 433 can be integrated into the same component. For the convenience of description, this disclosure regards GDMA 311 and IODMA 433 as different components. For those skilled in the art, as long as the functions they implement and the technical effects they achieve are similar to those of this disclosure, they fall within the protection scope of this disclosure. Furthermore, the functions of GDMA 311, IODMA 433, CDMA 310, and MVDMA 434 can also be implemented by the same component. Similarly, as long as the functions they implement and the technical effects they achieve are similar to those of this disclosure, they fall within the protection scope of this disclosure.

[0055] The hardware structure of the computing device 201 of this embodiment is complex. When selecting a mapping strategy for a specific task, it is necessary to quantify the performance of each scenario. These scenarios (parameters) include, for example, input data size, bandwidth of GDMA 311, bandwidth of CDMA 310, bandwidth of IODMA433, bandwidth of MVDMA434, storage space of SRAM, storage space of NRAM 431, storage space of WRAM 432, computing power of vector operation unit 421, computing power of matrix operation unit 422, execution of vector operation, execution of matrix operation, etc. In this embodiment, the mapping strategy specifically represents the implementation scheme of mapping tasks to various computing devices and memory access devices of the computing device 201. There are quite a lot of parameters that determine performance, and the change of any parameter will lead to completely different performance results. These parameters form a huge set of scenarios. The task of the developer is to find the mapping strategy that achieves the best performance in this set. It is difficult to fully cover all possibilities by simply analyzing and calculating with human power, so that the mapping strategy cannot be effectively selected.

[0056] This embodiment provides a solution for selecting a mapping strategy for a task by a computer, specifically, using the decoding unit of the processing device 203 to implement the selection of the mapping strategy for the computing device 201. For the convenience of description, unless otherwise specified, the following description will take the task of convolution calculation as an example.

[0057] Figure 6 The schematic diagram of the decoding unit of this embodiment selecting a mapping strategy for a matrix multiplication operation is shown. The decoding unit receives the performance parameters (i.e., hardware scenario parameters) of the computing device 201. In this embodiment, it is assumed that there are a performance parameters C1, C2, C1a, C2b, C1c, and C1c, and b mapping strategies, each of which includes M performance indicator values ​​trained with the model, for example, the performance indicator values ​​of the first mapping strategy are M11, M12, C1c, and C1d, and the performance indicator values ​​of the second mapping strategy are M21, M22, C1c, and C1d. Figure 6 Only two mapping strategies are shown as examples. The performance parameters are bitwise multiplied with the performance index values ​​to obtain a performance value matrix. b mapping strategies can obtain b performance value matrices, where the first performance value matrix is ​​R11, R12,,, R1a, and the second performance value matrix is ​​R21, R22,,,, R2a. Figure 6 Only two performance value matrices are shown as an example. These performance value matrices are respectively element-accumulated to obtain the corresponding performance values, where the first performance value is R1 and the second performance value is R2. Figure 6 Only two performance values ​​are shown by way of example. The decoding unit finally selects the mapping strategy with the maximum performance value as the task execution scheme of the computing device 201 .

[0058] Figure 7The structure diagram of the decoding unit of this embodiment is shown, which includes a quantizer 701, a selector 702, an operator 703, a filter 704 and a decision maker 705. The decision maker 705 includes a predictor 706 and a classifier 707. Figure 8 FIG. 4 is a flowchart showing the embodiment of selecting a mapping strategy for a convolution task by a decoding unit. Figure 7 and Figure 8 Elaborate in more detail on how the selection mapping strategy works.

[0059] In step 801, N mapping strategies and their corresponding performance formulas are obtained. The performance formula is a performance expression corresponding to each mapping strategy. The performance formula is used to indicate the degree to which the corresponding mapping strategy achieves the expected performance and is set by the developer.

[0060] The quantizer 701 quantizes the performance results of the N mapping strategies. For example, the performance formula is the utilization rate of the multiplier-accumulator in the matrix operation unit 422, that is, the performance of the convolution instruction is the efficiency of the multiplier-accumulator. The performance formula can be set as the theoretical operation time divided by the actual operation time, where the theoretical operation time is the number of operations divided by the computing power of the matrix operation unit 422. The higher the efficiency of the multiplier-accumulator, the better the performance. The quantizer 701 generates N quantization results after quantization.

[0061] The selector 702 then selects data that can cover the performance points from the N quantization results as the full performance set. The best case is to cover all performance points, but considering that there will be a lot of performance points, the actual situation may not be able to cover all of them. Another way is to select data of performance points of performance bottlenecks as the full performance set.

[0062] In step 802, the performance indicators corresponding to each mapping strategy are calculated according to the performance formulas corresponding to each mapping strategy. One mapping strategy calculates M performance indicator values, where M is equal to the number of types of performance indicators. The performance indicators are used to quantify the performance of the mapping strategy in different scenarios, for example: calculating the memory access ratio (the ratio of the number of calculated operations to the number of memory operations), the utilization rate of the operator (operation module 42, vector operation unit 421 or matrix operation unit 422), and the amount of data interacting with the dynamic random access memory (storage device 204). The operator 703 calculates the performance indicators corresponding to each mapping strategy according to the performance formula obtained in step 801.

[0063] In step 803, b mapping strategies that meet the performance indicator condition are selected from the N mapping strategies, where the performance indicator condition refers to that the M performance indicator values ​​of the mapping strategy all meet the performance thresholds corresponding to the M performance indicators. Not every mapping strategy can necessarily meet the mapping indicator condition, and the filter 704 selects b mapping strategies that meet the performance indicator condition for subsequent calculation.

[0064] In step 804, a performance parameters and b performance formulas of the task are imported into the neural network model for analysis, where a and b are integers greater than one, and N is greater than or equal to b.

[0065] Predictor 706 is used to extract the features of input data. It is a model trained based on a machine learning method. The model includes but is not limited to convolutional neural networks (CNN), fully connected networks (FCNN), graph neural networks (GNN) and reinforcement learning methods (RL). Common convolutional neural networks include AlexNet, VGG, Resnet, MobileNet, etc., common graph neural networks include GCN, GraphSAGE, GAT, etc., and common reinforcement learning methods include Q-learning, DDPG, A2C, PPO, etc. The input data is a number of data obtained by sampling, and the data contains several dimensional features, which can be divided into numerical features and type features. In this embodiment, the classification labels can be pre-calibrated, or there can be no labels, so that the model can learn the connection between the features within the data by itself. The training method includes supervised training and unsupervised training methods. During the training process, the model can adjust its own weight according to the pre-calibrated classification labels, or it can adjust its own weight by learning the internal association of data features, so that the classification results are more accurate.

[0066] Specifically, the predictor 706 will first determine the performance parameters. Performance parameters are parameters related to performance indicators in the performance formula, which can be divided into numerical parameters (corresponding to numerical features) and type parameters (corresponding to type features). Numerical parameters are parameters related to numerical values, and type parameters are parameters related to types. For convolution tasks, parameters related to the convolution scale are numerical parameters, and parameters related to the convolution operation data type are type parameters. For example, the input data size, the bandwidth of GDMA 311, the bandwidth of CDMA310, the bandwidth of IODMA 433, the bandwidth of MVDMA 434, the storage space of SRAM, the storage space of NRAM 431, the storage space of WRAM 432, the computing power of the vector operation unit 421, the computing power of the matrix operation unit 422, etc. are all numerical parameters, while the vector operation and matrix operation in the convolution belong to type parameters.

[0067] In this step, the predictor 706 imports a performance parameters and b performance formulas of the convolution task into the neural network model for training. The training methods include supervised training and unsupervised training methods. During the training process, the model can adjust its own weights according to pre-calibrated classification labels, and can also learn the internal correlation of data features to adjust its own weights, so that the classification results are more accurate.

[0068] In step 805, based on the analysis results of the neural network model, the performance values ​​corresponding to the b mapping strategies are obtained. In the early stage of processor design, the predictor 706 obtains the a*b first performance value matrix output by the neural network model. The first performance value matrix is ​​trained by algorithms such as neural networks using data from the performance script. In this embodiment, the b mapping strategies are the weights of each performance parameter for the final performance result, because the b results are obtained by multiplying and accumulating the performance parameters and the weights. Then the predictor 706 multiplies the 1*a performance parameter matrix with the a*b first performance value matrix based on the 1*a performance parameter matrix composed of a performance parameters to obtain a 1*b second performance value matrix. The second performance value matrix is ​​used to represent the performance values ​​corresponding to the b mapping strategies.

[0069] In step 806, a mapping strategy corresponding to the best performance value is selected. In this embodiment, the classifier 707 selects a mapping strategy corresponding to the maximum value from the second performance value matrix of 1*b, where the maximum value may be a multiplication result of the 1*a performance parameter matrix and the first performance value matrix of a*b, and / or selects the maximum value from the second performance value matrix of 1*b using hardware.

[0070] Predictor 706 performs one-hot encoding on all type-related parameters. One-hot encoding is also called one-bit effective encoding. It uses a multi-bit state register to encode multiple states one-to-one. Each state has its own independent register bit, and only one bit is valid at any time. Taking the encoding of six states as an example, the natural sequence code is: 000, 001, 010, 011, 100, 101, and the one-hot encoding is: 000001, 000010, 000100, 001000, 010000, 100000. After predictor 706 performs one-hot encoding on the type parameter, the parameter will be regarded as a new independent parameter. In this embodiment, the a-bit state register is divided into a1 state register corresponding to the type parameter and a2 state registers corresponding to the value parameter, that is, a=a1+a2.

[0071] The predictor 706 obtains a1 parameters related to the type according to the result of the one-hot encoding, and temporarily stores them in a1 state registers corresponding to the type parameters, and obtains a2 parameters related to the value, and temporarily stores them in a2 state registers corresponding to the value parameters. Finally, a1 and a2 are added together to obtain the total number of a performance parameters.

[0072] The classifier 707 selects a strategy based on the data features extracted by the predictor 706. The classifier 707 may be a binary classifier or a multi-class classifier, and selects the best strategy based on the input data and the task. When there are multiple performance strategies with the same maximum value, the classifier 707 preferentially selects the performance strategy with lower power consumption.

[0073] Fig. 9 A simplified schematic diagram of selecting a mapping strategy by the decoding unit of this embodiment is shown. The figure shows two mapping strategies: mapping strategy 1 and mapping strategy 2. Assuming that each matrix operation module 422 includes four multiplication-accumulation matrices 901, mapping strategy 1 is to operate the four multiplication-accumulation matrices 901 of the matrix operation module 422 independently, while mapping strategy 2 is to plan the four multiplication-accumulation matrices 901 of the matrix operation module 422 into two matrix groups for operation. The decoding unit adopts Figure 8 The method is used to determine which mapping strategy has a larger value of the second performance value matrix. If the value of the second performance value matrix of mapping strategy 1 is larger, the four multiplication-accumulation matrices 901 are selected to run independently. If the value of the second performance value matrix of mapping strategy 2 is larger, the four multiplication-accumulation matrices 901 are selected to be planned into two matrix groups for operation.

[0074] Another embodiment of the present disclosure is to utilize Figures 1 to 4 The structure of the mapping strategy is used to select the method. Fig.10 A flowchart showing the process of selecting a mapping strategy according to this embodiment is shown.

[0075] In step 1001, the processing device 203 obtains a training set of mapping strategies. The training set includes the performance formula corresponding to the computing device 201 and its corresponding performance parameters as described in the previous embodiment, the performance formula is a performance expression corresponding to the mapping strategy one by one, the performance parameter is a parameter in the performance formula related to the performance index, and the performance index is used to quantify the performance of the mapping strategy in different scenarios.

[0076] In step 1002, the processing device 203 imports the training set into the neural network model for training. The performance parameters in the training set include numerical parameters and type parameters. The definitions of numerical parameters and type parameters are the same as those in the previous embodiment, so they are not repeated. The performance parameters carry classification labels, and the training methods include supervised training and unsupervised training methods. During the training process, the model can adjust its own weight according to the pre-calibrated classification labels, and can also adjust its own weight by learning the internal association of data features to make the classification results more accurate. After the training is completed, this embodiment can obtain the mapping strategy with the best performance value.

[0077] In step 1003 , the computing device 201 executes the task based on the mapping strategy with the best performance value.

[0078] Another embodiment of the present disclosure is a computer-readable storage medium having stored thereon a computer program code for selecting a mapping strategy, which, when executed by a processing device, performs Figure 8 or Fig.10 method.

[0079] Another embodiment of the present disclosure is a computer program product, including a computer program for selecting a mapping strategy, which is executed by a processor to implement Figure 8 or Fig.10 steps.

[0080] Another embodiment of the present disclosure is a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement Figure 8 or Fig.10 steps.

[0081] When the computer program code is run by the processor, the method of each embodiment as described above is executed. In some implementation scenarios, the above-mentioned integrated unit can be implemented in the form of a software program module. If it is implemented in the form of a software program module and sold or used as an independent product, the integrated unit can be stored in a computer readable memory. Based on this, when the scheme of the present disclosure is embodied in the form of a software product (such as a computer-readable storage medium), the software product can be stored in a memory, which may include several instructions to enable a computer device (such as a personal computer, a server or a network device, etc.) to perform some or all of the steps of the method described in the embodiment of the present disclosure. The aforementioned memory may include, but is not limited to, various media that can store program codes, such as a USB flash drive, a flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0082] The present disclosure uses computer calculation performance formulas to cope with complex computer tasks and hardware configurations, and comprehensively and efficiently analyze or infer ideal mapping solutions to achieve high-performance, low-power technical effects.

[0083] According to different application scenarios, the electronic equipment or device disclosed herein may include servers, cloud servers, server clusters, data processing devices, robots, computers, printers, scanners, tablet computers, smart terminals, PC devices, IoT terminals, mobile terminals, mobile phones, driving recorders, navigators, sensors, cameras, cameras, video cameras, projectors, watches, headphones, mobile storage, wearable devices, visual terminals, automatic driving terminals, transportation, household appliances, and / or medical equipment. The transportation includes airplanes, ships and / or vehicles; the household appliances include televisions, air conditioners, microwave ovens, refrigerators, rice cookers, humidifiers, washing machines, electric lights, gas stoves, and range hoods; the medical equipment includes magnetic resonance imaging, ultrasound machines and / or electrocardiographs. The electronic equipment or device disclosed herein may also be applied to the Internet, IoT, data centers, energy, transportation, public administration, manufacturing, education, power grids, telecommunications, finance, retail, construction sites, and medical fields. Further, the electronic equipment or device disclosed herein may also be used in cloud, edge, and terminal applications related to artificial intelligence, big data, and / or cloud computing. In one or more embodiments, electronic devices or devices with high computing power according to the disclosed solution can be applied to cloud devices (such as cloud servers), while electronic devices or devices with low power consumption can be applied to terminal devices and / or edge devices (such as smart phones or cameras). In one or more embodiments, the hardware information of the cloud device and the hardware information of the terminal device and / or edge device are compatible with each other, so that according to the hardware information of the terminal device and / or edge device, appropriate hardware resources can be matched from the hardware resources of the cloud device to simulate the hardware resources of the terminal device and / or edge device, so as to complete the unified management, scheduling and collaborative work of end-to-end or cloud-edge-to-end.

[0084] It should be noted that, for the purpose of simplicity, the present disclosure describes some methods and embodiments thereof as a series of actions and combinations thereof, but those skilled in the art will appreciate that the scheme of the present disclosure is not limited by the order of the actions described. Therefore, based on the disclosure or teaching of the present disclosure, those skilled in the art will appreciate that some of the steps therein may be performed in other orders or simultaneously. Further, those skilled in the art will appreciate that the embodiments described in the present disclosure may be regarded as optional embodiments, i.e., the actions or modules involved therein may not necessarily be necessary for the implementation of one or some of the schemes of the present disclosure. In addition, depending on the different schemes, the present disclosure also has different focuses on the description of some embodiments. In view of this, those skilled in the art may appreciate the parts that are not described in detail in a certain embodiment of the present disclosure, and may also refer to the relevant descriptions of other embodiments.

[0085] In terms of specific implementation, based on the disclosure and teaching of this disclosure, those skilled in the art can understand that several embodiments disclosed in this disclosure can also be implemented by other methods not disclosed herein. For example, with respect to the various units in the electronic device or device embodiments described above, this article splits them on the basis of considering the logical functions, and there may be other splitting methods in actual implementation. For another example, multiple units or components can be combined or integrated into another system, or some features or functions in the units or components can be selectively disabled. In terms of the connection relationship between different units or components, the connection discussed in the above text in conjunction with the accompanying drawings can be a direct or indirect coupling between units or components. In some scenarios, the aforementioned direct or indirect coupling involves a communication connection using an interface, wherein the communication interface can support electrical, optical, acoustic, magnetic or other forms of signal transmission.

[0086] In the present disclosure, the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units. The aforementioned components or units may be located in the same location or distributed on multiple network units. In addition, according to actual needs, some or all of the units may be selected to achieve the purpose of the scheme described in the embodiments of the present disclosure. In addition, in some scenarios, multiple units in the embodiments of the present disclosure may be integrated into one unit or each unit may exist physically separately.

[0087] In some other implementation scenarios, the above-mentioned integrated unit can also be implemented in the form of hardware, that is, a specific hardware circuit, which may include digital circuits and / or analog circuits, etc. The physical implementation of the hardware structure of the circuit may include but is not limited to physical devices, and the physical devices may include but are not limited to devices such as transistors or memristors. In view of this, the various devices described herein (such as computing devices or other processing devices) can be implemented by appropriate hardware processors, such as central processing units, GPUs, FPGAs, DSPs, and ASICs, etc. Further, the aforementioned storage unit or storage device may be any appropriate storage medium (including magnetic storage media or magneto-optical storage media, etc.), which may be, for example, a variable resistive memory (Resistive Random Access Memory, RRAM), a dynamic random access memory (Dynamic Random Access Memory, DRAM), a static random access memory (Static Random Access Memory, SRAM), an enhanced dynamic random access memory (Enhanced Dynamic Random Access Memory, EDRAM), a high bandwidth memory (High Bandwidth Memory, HBM), a hybrid memory cube (Hybrid Memory Cube, HMC), ROM and RAM, etc.

[0088] The foregoing content can be better understood in accordance with the following terms:

[0089] Item A1, a decoding unit for selecting a mapping strategy for a task, wherein the mapping strategy represents an implementation scheme for mapping the task to the computing device and memory access device of a processor, characterized in that the decoding unit comprises: a predictor, used to: import a performance parameters and b performance formulas of the task into a neural network model for analysis, wherein the b performance formulas are performance expressions corresponding one-to-one to b mapping strategies; the performance parameters are parameters in the performance formulas related to performance indicators, and the performance indicators are used to quantify the performance of the mapping strategy in different scenarios; a and b are respectively integers greater than one; based on the analysis results of the neural network model, performance values ​​corresponding to the b mapping strategies are obtained; and a classifier, used to select the mapping strategy that best corresponds to the performance value.

[0090] Clause A2, according to the decoding unit described in Clause A1, it is characterized in that when the predictor obtains the performance values ​​corresponding to b mapping strategies respectively based on the analysis results of the neural network model, it first obtains the a*b first performance value matrix output by the neural network model, and then multiplies the 1*a performance parameter matrix composed of the a performance parameters by the a*b first performance value matrix to obtain a 1*b second performance value matrix, which represents the performance values ​​corresponding to the b mapping strategies respectively.

[0091] Item A3. A decoding unit according to Item A2, wherein the classifier selects a mapping strategy corresponding to a maximum value from the 1*b second performance value matrix as the optimal corresponding mapping strategy.

[0092] Item A4. A decoding unit according to Item A3, wherein the classifier multiplies the 1*a performance parameter matrix with the a*b first performance value matrix, and / or selects the maximum value from the 1*b second performance value matrix using hardware.

[0093] Clause A5. A decoding unit according to any one of clauses A1-4, characterized in that the performance parameters include: numerical parameters and type parameters; wherein the numerical parameters are parameters related to numerical values, and the type parameters are parameters related to types.

[0094] Clause A6. The decoding unit according to clause A5 is characterized in that the predictor also performs one-hot encoding on all type-related parameters, obtains a1 type-related parameters according to the number of types of the one-hot encoding; obtains a2 value-related parameters; sums a1 and a2 to obtain a total number of a performance parameters.

[0095] Clause A7. A decoding unit according to any one of clauses A1-6, characterized in that the performance formula includes: utilization of the multiplier-accumulator.

[0096] Clause A8. A decoding unit according to any one of clauses A1-7, characterized in that the performance indicators include: computational memory access ratio, utilization of the operator, and the amount of data interacting with the dynamic random access memory; wherein the computational memory access ratio is the ratio of the number of computational operations to the number of memory operations.

[0097] Clause A9. The decoding unit according to Clause A5 is characterized in that the decoding unit also obtains a full set of performance based on N mapping strategies and their corresponding performance formulas, where N is greater than or equal to b, and the decoding unit includes: a quantizer for quantizing the performance results of the N mapping strategies to generate N quantization results; and a selector for selecting data that can cover performance points from the N quantization results as the full set of performance.

[0098] Clause A10, the decoding unit according to clause A9 is characterized in that the decoding unit includes: an operator, used to calculate the performance indicator corresponding to each mapping strategy according to the performance formula corresponding to each mapping strategy; wherein one mapping strategy calculates M performance indicator values, and M is equal to the number of types of performance indicators; a filter, used to select b mapping strategies that meet the performance indicator conditions from N mapping strategies; wherein the performance indicator conditions include: the M performance indicator values ​​of the mapping strategy all meet the performance thresholds corresponding to the M performance indicators.

[0099] Item A11. An integrated circuit device comprising a decoding unit according to any one of Items A1 to 10.

[0100] Item A12. A board comprising the integrated circuit device according to Item A11.

[0101] Item A13, an integrated circuit device for selecting a mapping strategy, comprising a processing device and a computing device, characterized in that the processing device is used to: obtain a training set of mapping strategies; the training set includes a performance formula corresponding to the computing device and its corresponding performance parameters, the performance formula is a performance expression corresponding one-to-one to the mapping strategy; the performance parameter is a parameter in the performance formula related to the performance indicator, and the performance indicator is used to quantify the performance of the mapping strategy in different scenarios; the training set is imported into a neural network model for training.

[0102] Item A14. An integrated circuit device according to Item A11, wherein the performance parameters in the training set include numerical parameters and type parameters, and the performance parameters are provided with classification labels.

[0103] A method for selecting a mapping strategy for a task, wherein the mapping strategy represents an implementation scheme for mapping the task to a computing device and a memory access device of a processor, characterized in that the method comprises: importing a performance parameters and b performance formulas of the task into a neural network model for analysis, wherein the b performance formulas are performance expressions corresponding one-to-one to the b mapping strategies; the performance parameters are parameters in the performance formulas related to performance indicators, and the performance indicators are used to quantify the performance of the mapping strategy in different scenarios; a and b are integers greater than one respectively; based on the analysis results of the neural network model, obtaining performance values ​​corresponding to the b mapping strategies respectively; and selecting the mapping strategy corresponding to the optimal performance value.

[0104] Clause A15, a method for selecting a mapping strategy, characterized in that the method includes: obtaining a training set of mapping strategies; the training set includes a performance formula and its corresponding performance parameters, the performance formula is a performance expression corresponding one-to-one to the mapping strategy; the performance parameter is a parameter in the performance formula related to the performance indicator, and the performance indicator is used to quantify the performance of the mapping strategy in different scenarios; importing the training set into a neural network model for training.

[0105] Item A16. A computer-readable storage medium having stored thereon a computer program code for selecting a mapping strategy, wherein when the computer program code is executed by a processing device, the method described in Item A15 or 16 is executed.

[0106] Item A17. A computer program product, comprising a computer program for selecting a mapping strategy, characterized in that when the computer program is executed by a processor, the steps of the method described in Item A15 or 16 are implemented.

[0107] Item A18. A computer device comprising a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described in Item A15 or 16.

[0108] The embodiments of the present disclosure are introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present disclosure. The description of the above embodiments is only used to help understand the method of the present disclosure and its core idea. At the same time, for those skilled in the art, according to the ideas of the present disclosure, there will be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be understood as a limitation on the present disclosure.

Claims

1. A decoding unit for selecting a mapping strategy for a task, wherein the mapping strategy represents an implementation scheme for mapping the task to a computing device and a memory access device of a processor, characterized in that: The decoding unit comprises: Predictors to: Importing a performance parameter and b performance formula of the task into the neural network model for analysis, wherein the b performance formulas are performance expressions corresponding to the b mapping strategies one by one; the performance parameter is a parameter in the performance formula related to the performance index, and the performance index is used to quantify the performance of the mapping strategy in different scenarios; a and b are respectively integers greater than one; Based on the analysis results of the neural network model, obtaining performance values ​​corresponding to the b mapping strategies respectively; and A classifier is used to select a mapping strategy that best corresponds to the performance value.

2. The decoding unit according to claim 1, characterized in that When the predictor obtains the performance values ​​corresponding to b mapping strategies respectively based on the analysis results of the neural network model, it first obtains the a*b first performance value matrix output by the neural network model, and then multiplies the 1*a performance parameter matrix composed of the a performance parameters by the a*b first performance value matrix to obtain a 1*b second performance value matrix, which represents the performance values ​​corresponding to the b mapping strategies respectively.

3. The decoding unit according to claim 2, wherein the classifier selects a mapping strategy corresponding to a maximum value from the 1*b second performance value matrix as the optimal corresponding mapping strategy.

4. The decoding unit according to claim 3, wherein the classifier multiplies the 1*a performance parameter matrix with the a*b first performance value matrix and / or selects the maximum value from the 1*b second performance value matrix using hardware.

5. The decoding unit according to any one of claims 1 to 4, characterized in that: The performance parameters include: numerical parameters and type parameters; wherein the numerical parameters are parameters related to numerical values, and the type parameters are parameters related to types.

6. The decoding unit according to claim 5, characterized in that The predictor also performs one-hot encoding processing on all type-related parameters, obtains a1 type-related parameters according to the number of types of the one-hot encoding; obtains a2 value-related parameters; and sums a1 and a2 to obtain a total number of a performance parameters.

7. The decoding unit according to any one of claims 1 to 6, characterized in that: The performance formula includes: utilization of the multiplier-accumulator.

8. The decoding unit according to any one of claims 1 to 7, characterized in that: The performance indicators include: computing memory access ratio, utilization of the operator and the amount of data interacting with the dynamic random access memory; wherein the computing memory access ratio is the ratio of the number of computing operations to the number of memory operations.

9. The decoding unit according to claim 5, characterized in that: The decoding unit further obtains a full set of performances based on N mapping strategies and their corresponding performance formulas, where N is greater than or equal to b, and the decoding unit includes: A quantizer, used to quantize the performance results of the N mapping strategies to generate N quantized results; A selector is used to select data that can cover the performance points from the N quantization results as the full set of performance.

10. The decoding unit according to claim 9, characterized in that: The decoding unit comprises: An operator, used to calculate the performance index corresponding to each mapping strategy according to the performance formula corresponding to each mapping strategy; wherein one mapping strategy calculates M performance index values, where M is equal to the number of types of performance indicators; A filter is used to select b mapping strategies that meet performance indicator conditions from N mapping strategies; wherein the performance indicator conditions include: M performance indicator values ​​of the mapping strategy all meet performance thresholds corresponding to the M performance indicators.

11. An integrated circuit device comprising a decoding unit according to any one of claims 1 to 10.

12. A card comprising the integrated circuit device according to claim 11.

13. An integrated circuit device for selecting a mapping strategy, comprising a processing device and a computing device, characterized in that: The processing device is used to: Acquire a training set of mapping strategies; the training set includes a performance formula corresponding to the computing device and its corresponding performance parameters, the performance formula being a performance expression corresponding one-to-one to the mapping strategy; The performance parameter is a parameter in the performance formula related to the performance indicator, and the performance indicator is used to quantify the performance of the mapping strategy in different scenarios; The training set is imported into the neural network model for training.

14. The integrated circuit device according to claim 13, wherein the performance parameters in the training set include numerical parameters and type parameters, and the performance parameters are provided with classification labels.

15. A method for selecting a mapping strategy for a task, wherein the mapping strategy represents an implementation scheme for mapping the task to a computing device and a memory access device of a processor, characterized in that: The method comprises: Importing a performance parameter and b performance formula of the task into the neural network model for analysis, wherein the b performance formulas are performance expressions corresponding to the b mapping strategies one by one; the performance parameter is a parameter in the performance formula related to the performance index, and the performance index is used to quantify the performance of the mapping strategy in different scenarios; a and b are respectively integers greater than one; Based on the analysis results of the neural network model, obtaining performance values ​​corresponding to the b mapping strategies respectively; and Select a mapping strategy that best corresponds to the performance value.

16. A method for selecting a mapping strategy, characterized in that: The method comprises: Obtain a training set of mapping strategies; the training set includes a performance formula and its corresponding performance parameters, the performance formula is a performance expression corresponding to the mapping strategy one by one; the performance parameter is a parameter in the performance formula related to the performance indicator, and the performance indicator is used to quantify the performance of the mapping strategy in different scenarios; The training set is imported into the neural network model for training.

17. A computer-readable storage medium having stored thereon a computer program code for selecting a mapping strategy, wherein when the computer program code is executed by a processing device, the method according to claim 15 or 16 is executed.

18. A computer program product comprising a computer program for selecting a mapping strategy, characterized in that When the computer program is executed by a processor, the steps of the method according to claim 15 or 16 are implemented.

19. A computer device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method of claim 15 or 16.

Citation Information

Cited By

  • Memory management method and storage device

    CN121704786A