A Chip Thermal Sensor Distribution Method, System, Device and Medium

Abstract

The present invention discloses a chip thermal sensor distribution method, system, device, and medium. The method comprises: constructing a data set by performing thermal simulation on a chip; calculating mutual information on the data set to determine a mutual information matrix; filtering the mutual information matrix to determine a sensor distribution set; and reconstructing the sensor distribution set to determine a target sensor distribution map. Embodiments of the present invention can effectively restore global chip thermal distribution information and are widely applicable to the field of chip thermal management technology.

Description

Technical Field

[0001] The present invention relates to the field of chip thermal management technology, and in particular to a chip thermal sensor distribution method, system, equipment and medium. Background Art

[0002] With technological advancements, the complexity of multi-core systems continues to increase. Furthermore, the high power density resulting from the high integration density of multi-core systems based on advanced processes is exacerbating thermal issues. To effectively monitor and manage the thermal distribution information of multi-core systems, circuit systems integrate a certain number of thermal sensors to perform temperature monitoring tasks at selected locations. Secondly, temperature data acquired from thermal sensors at designated locations is used to predict the overall temperature distribution of the system, thereby providing effective data support for dynamic thermal management. To control the number of thermal sensors deployed on a chip and reduce the chip area occupied by the thermal sensors, thereby lowering costs, it is necessary to utilize a limited number of thermal sensors and achieve high-precision global thermal distribution information recovery based on the temperature data fed back by the thermal sensors. However, no specific method for distributing thermal sensors on a chip can effectively restore the global thermal distribution information of a chip. Summary of the Invention

[0003] The present invention aims to address at least one of the technical problems existing in the prior art. To this end, the present invention proposes a chip thermal sensor distribution method, system, device, and medium that can plan the layout of thermal sensors and generate a thermal sensor distribution map, thereby assisting in dynamic thermal management at the system level.

[0004] In one aspect, the present invention provides a chip thermal sensor distribution method, comprising:

[0005] By performing thermal simulation on the chip, a data set is constructed;

[0006] performing mutual information calculation on the data set to determine a mutual information matrix;

[0007] Filtering the mutual information matrix to determine a sensor distribution set;

[0008] The sensor distribution set is reconstructed to determine a target sensor distribution map.

[0009] Optionally, the step of constructing a data set by performing thermal simulation on the chip includes:

[0010] Perform thermal simulation on the chip to determine thermal simulation data;

[0011] The thermal simulation data is exported and preprocessed to construct a data set.

[0012] Optionally, calculating the mutual information of the data set to determine the mutual information matrix includes:

[0013] The correlation of each feature in the data set is calculated according to the mutual information calculation formula to construct the mutual information matrix.

[0014] Optionally, filtering the mutual information matrix to determine the sensor distribution set includes:

[0015] Performing variance filtering on the mutual information matrix to determine a selected feature set;

[0016] Performing mutual information calculation on the selected feature set to determine a correlation matrix set;

[0017] Perform correlation filtering on the correlation matrix set to determine a sensor distribution set.

[0018] Optionally, reconstructing the sensor distribution set to determine a target sensor distribution map includes:

[0019] Polynomial processing is performed on the data in the sensor distribution set, and the processed data is substituted into a linear regression with a penalty term to calculate the regressor algorithm error, thereby determining a target sensor distribution map.

[0020] Optionally, performing variance filtering on the mutual information matrix to determine a selected feature set includes:

[0021] Perform cyclic variance filtering on the mutual information matrix according to the first proportional step size to determine the selected feature set.

[0022] Optionally, performing correlation filtering on the correlation matrix set to determine the sensor distribution set includes:

[0023] Calculating node degrees of correlation matrices in the set of correlation matrices to determine a first node, where the first node is a node with the largest degree in the correlation matrix;

[0024] Sort the correlation between the node with the largest degree and other nodes in the correlation matrix, filter the nodes, and determine the correlation sequence;

[0025] Calculating the degree of an undirected graph of the correlation sequence to determine a second node, where the second node is a node with the largest degree of the undirected graph in the correlation sequence;

[0026] A sensor distribution set is determined according to the first node and the second node.

[0027] On the other hand, an embodiment of the present invention further discloses an electronic device, including a processor and a memory;

[0028] The memory is used to store programs;

[0029] The processor executes the program to implement the method described above.

[0030] On the other hand, an embodiment of the present invention further discloses a computer-readable storage medium, wherein the storage medium stores a program, and the program is executed by a processor to implement the method described above.

[0031] In another aspect, embodiments of the present invention further disclose a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform the aforementioned method.

[0032] Compared with the prior art, the present invention adopts the above technical solution and has the following technical effects: the embodiment of the present invention calculates the mutual information of the data set to determine the mutual information matrix; it can introduce more relationship information between nodes through mutual information, improve the rationality of sensor layout, and thus improve the accuracy of the reconstructed chip thermal distribution map; furthermore, the embodiment of the present invention filters the mutual information matrix to determine the sensor distribution set; it can make the layout position of the sensors more flexible and changeable; in addition, the embodiment of the present invention reconstructs the sensor distribution set to determine the target sensor distribution map, which can further improve the restoration accuracy of the thermal sensor distribution map. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0034] Figure 1 The figure is a flow chart of a method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solutions and advantages of this application more clearly understood, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0036] Reference Figure 1 , an embodiment of the present invention provides a chip thermal sensor distribution method, comprising:

[0037] S101, constructing a data set by performing thermal simulation on the chip;

[0038] S102, performing mutual information calculation on the data set to determine a mutual information matrix;

[0039] S103, filtering the mutual information matrix to determine a sensor distribution set;

[0040] S104: Reconstruct the sensor distribution set to determine a target sensor distribution map.

[0041] As a further preferred embodiment, in the above step S101, the step of constructing a data set by performing thermal simulation on the chip includes:

[0042] Perform thermal simulation on the chip to determine thermal simulation data;

[0043] The thermal simulation data is exported and preprocessed to construct a data set.

[0044] Specifically, the embodiment of the present invention uses COMSOL software to perform thermal simulation on the chip. The specific operation process is as follows: construct a structural diagram of the chip in COMSOL, set the material (such as Si, Ge, etc.) and thermal conductivity parameters of each area of the chip, select several relatively prominent heat sources in the chip, set the heat transfer power of the heat source, perform simulation, and obtain thermal simulation data. Utilize the embedded function in COMSOL and the corresponding Java program to automatically export the thermal simulation data as a TXT file, and then use Python to perform an automated preprocessing on the data, and finally obtain thermal distribution map data similar to 10x10, that is, the data set in this solution. At this point, the automatic construction of the data set is completed, providing a data basis for the testing of subsequent algorithms. It is conceivable that the embodiment of the present invention can also use other simulation software to obtain simulation data, and the materials and thermal conductivity parameters in the thermal simulation process can also be adaptively set according to actual conditions. At the same time, applications in other computer languages can also be used to preprocess the simulation data.

[0045] As a further preferred embodiment, in the above step S102, the mutual information calculation of the data set to determine the mutual information matrix includes:

[0046] The correlation of each feature in the data set is calculated according to the mutual information calculation formula to construct the mutual information matrix.

[0047] Specifically, the embodiment of the present invention can also use the correlation graph algorithm to calculate the correlation matrix, but the embodiment of the present invention uses the mutual information algorithm to calculate the mutual information matrix. For the correlation matrix of linear relationships, the Pearson correlation matrix of the traditional correlation graph algorithm can only describe the linear relationship between nodes (features), and the embodiment of the present invention introduces mutual information to find more correlations between nodes (features) and uses the mutual information method to calculate the relationship between continuous variables. The embodiment of the present invention uses the mutual information method to calculate the correlation of each feature of the data set and constructs a mutual information matrix. The mutual information calculation formula is:

[0048] I(X, Y)=ψ(k)-<ψ(n x +1)+ψ(n y +1)>+ψ(N);

[0049] In the above mutual information calculation formula, X represents the first continuous random variable, Y represents the second continuous random variable, I(X, Y) represents the average mutual information between continuous variables, ψ() represents the gamma function, k represents the number of nearest points set by the k-nearest neighbor algorithm, and n x Indicates the number of elements in vector X whose distance to the i-th element is less than the threshold, excluding itself, n y Represents the number of elements in vector Y (excluding itself) whose distance to the i-th element is less than the threshold, i is a positive integer, and N represents the total number of elements in variables X and Y.

[0050] As a further preferred embodiment, in the above step S103, filtering the mutual information matrix to determine the sensor distribution set includes:

[0051] Performing variance filtering on the mutual information matrix to determine a selected feature set;

[0052] Performing mutual information calculation on the selected feature set to determine a correlation matrix set;

[0053] Perform correlation filtering on the correlation matrix set to determine a sensor distribution set.

[0054] Specifically, embodiments of the present invention first perform variance filtering on the mutual information matrix. Variance filtering is based on a simple concept: variance, to a certain extent, represents the amount of information, and features with small variance carry less information. Therefore, embodiments of the present invention can initially use variance filtering to pre-screen nodes with small variance in the dataset, filtering a certain percentage of nodes to obtain a selected feature set. Mutual information is then calculated on the selected feature set to obtain a correlation matrix set. Finally, correlation filtering is performed on the correlation matrix set to obtain a sensor distribution set.

[0055] As a further preferred embodiment, in the above step S104, the reconstructing the sensor distribution set to determine the target sensor distribution map includes:

[0056] Polynomial processing is performed on the data in the sensor distribution set, and the processed data is substituted into a linear regression with a penalty term to calculate the regressor algorithm error, thereby determining a target sensor distribution map.

[0057] Specifically, the present embodiment improves upon the linear regressor algorithm to obtain a reconstruction algorithm. First, polynomial processing is performed on the data features to obtain high-order features. The processed data is then used as features and the original labels for linear regression. This approach achieves polynomial regression without incurring excessive computational costs. Furthermore, because polynomial regression can fit nonlinear relationships, it significantly improves the accuracy of heat maps. However, the introduction of polynomials can lead to overfitting, which can cause high-order polynomial models to fit the training set well but perform poorly on the test set. To address this issue, the present embodiment introduces other linear regression models: Ridge regression, Lasso (least absolute shrinkage and selection operator), and ElasticNet. Compared to conventional linear regression models, Ridge introduces the L2 norm as a penalty term, while Lasso introduces the L1 norm as a penalty term. ElasticNet combines the Ridge and Lasso algorithms. Experimental comparisons revealed that combining these penalized linear regressions with polynomials effectively reduces overfitting and improves the accuracy of thermal distribution compared to the original polynomial model. Therefore, the present invention applies polynomial processing to the data in the sensor distribution set and substitutes the processed data into a penalized linear regression to calculate the regressor algorithm error, thereby obtaining the target sensor distribution map.

[0058] As a further preferred embodiment, performing variance filtering on the mutual information matrix to determine the selected feature set includes:

[0059] Perform cyclic variance filtering on the mutual information matrix according to the first proportional step size to determine the selected feature set.

[0060] Specifically, the embodiment of the present invention performs cyclic variance filtering on the mutual information matrix using a first scale step size, and then cycles the variance filtering scale from 0 to 1 with a step size of 0.2. In each cycle, a set of corresponding sensor layout positions is found within the selected variance filtering scale to obtain a selected feature set.

[0061] As a further preferred embodiment, performing correlation filtering on the correlation matrix set to determine the sensor distribution set includes:

[0062] Calculating node degrees of correlation matrices in the set of correlation matrices to determine a first node, where the first node is a node with the largest degree in the correlation matrix;

[0063] Sort the correlation between the node with the largest degree and other nodes in the correlation matrix, filter the nodes, and determine the correlation sequence;

[0064] Calculating the degree of an undirected graph of the correlation sequence to determine a second node, where the second node is a node with the largest degree of the undirected graph in the correlation sequence;

[0065] A sensor distribution set is determined according to the first node and the second node.

[0066] Specifically, the embodiment of the present invention calculates the node degree of each correlation matrix in the correlation matrix set to obtain the first node; then sorts the correlation between the node with the largest degree and other nodes in the correlation matrix, and filters the nodes to obtain a correlation sequence, wherein the calculation formula for the number of filtered nodes is as follows:

[0067]

[0068] where N filter is the number of nodes to be filtered, N total is the total number of nodes, K is the number of layout sensors, r filter is the filtering ratio. In the experiment, it was found that r filter =1 has better effect.

[0069] After filtering, the remaining nodes are obtained, and a correlation matrix is constructed for these nodes. The above process is repeated to ultimately determine the sensor layout locations. A specific example of an embodiment of the present invention is: one embodiment requires selecting two sensors from six nodes: A, B, C, D, E, and F. According to a traditional correlation graph algorithm, the undirected graph degrees of all nodes are sorted, and the top two nodes, A and B, are selected as sensor layout locations. However, if node A is further analyzed, it is found that the correlation between B and A is as high as 0.9, which means that node A is a good predictor of node B. Therefore, selecting node B as the sensor placement node would result in a waste of sensor resources. Based on this idea, the correlation filtering method of an embodiment of the present invention is to sort the correlations between node A and other nodes after node A is selected, filtering out nodes A, B, and D with high correlations with node A. Among the remaining nodes E, F, and C, the undirected graph degrees are then sorted, and node C with the highest undirected graph degree is selected. This completes the sensor layout location selection.

[0070] As can be imagined, the present invention also provides an implementation: the number of thermal sensors is input, and then the correlation filtering loop is looped through the number of thermal sensors, K, to sequentially find all sensor layout locations. The larger loop loops through the variance filtering ratios to find the optimal variance filtering ratio for the K-sensor layout. Through this process, the present invention addresses three shortcomings of traditional correlation graph algorithms: a linear correlation matrix, sensor clustering issues, and relatively fixed sensor locations. The following is a more detailed description of this overall process. First, the present invention requires input of a training set and the number of sensors, K, in the layout. Then, using a mutual information calculation method, the mutual information matrix for all features in the training set is obtained. The variance filtering ratio is then looped from 0 to 1, with a step size of 0.2. In each loop, the present invention finds a set of corresponding sensor layout locations within the selected variance filtering ratio. This embodiment then applies these layout locations to a reconstruction algorithm—the regressor algorithm (i.e., polynomial regression and its variants)—to select the set of sensor layout locations with the lowest error. In this way, the optimal variance filtering ratio corresponding to the number of sensors being K can be determined, thereby selecting the optimal layout position under the algorithm.

[0071] and Figure 1 Corresponding to the method, an embodiment of the present invention further provides an electronic device, including a processor and a memory; the memory is used to store a program; the processor executes the program to implement the method as described above.

[0072] and Figure 1 Corresponding to the method, an embodiment of the present invention further provides a computer-readable storage medium, wherein the storage medium stores a program, and the program is executed by a processor to implement the method as described above.

[0073] The embodiment of the present invention also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs Figure 1 The method shown.

[0074] In summary, the embodiments of the present invention have the following advantages:

[0075] (1) The embodiment of the present invention uses mutual information as a measure of correlation and combines correlation filtering and variance filtering to propose a thermal sensor layout algorithm, which can achieve higher-precision thermal distribution map restoration compared to other algorithms;

[0076] (2) The embodiment of the present invention adds polynomial processing to the reconstruction algorithm of the traditional linear regressor, and proposes a new heat distribution map reconstruction algorithm by combining it with the ridge regression algorithm, the least absolute shrinkage and selection operator and the elastic net algorithm, which can effectively reduce overfitting and improve the accuracy of heat distribution.

[0077] (3) The embodiments of the present invention can achieve high-precision restoration of thermal distribution maps using fewer sensors.

[0078] In some optional embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the present invention is provided in an exemplary manner for the purpose of providing a more comprehensive understanding of the technology. The disclosed method is not limited to the operation and logic flow presented herein. Optional embodiments are contemplated in which the order of the various operations is changed and the sub-operations described as a part of a larger operation are performed independently.

[0079] Furthermore, although the present invention is described in the context of functional modules, it should be understood that, unless otherwise indicated, one or more of the functions and / or features described may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in separate physical devices or software modules. It will also be understood that a detailed discussion of the actual implementation of each module is not necessary for understanding the present invention. More specifically, given the properties, functions, and internal relationships of the various functional modules in the devices disclosed herein, the actual implementation of the module will be understood within the ordinary skill of an engineer. Therefore, a person skilled in the art using ordinary skill will be able to implement the present invention set forth in the claims without undue experimentation. It will also be understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0080] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0081] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0082] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0083] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0084] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0085] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

[0086] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the embodiments. Those skilled in the art can make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of this application.

Claims

1. A chip thermal sensor distribution method, characterized in that: include: By performing thermal simulation on the chip, a data set is constructed; Performing mutual information calculation on the data set to determine a mutual information matrix; Performing filtering on the mutual information matrix to determine a sensor distribution set; Reconstructing the sensor distribution set to determine a target sensor distribution map; The performing mutual information calculation on the data set to determine the mutual information matrix includes: The correlation of each feature in the data set is calculated according to the mutual information calculation formula to construct the mutual information matrix; The filtering process on the mutual information matrix to determine the sensor distribution set includes: Performing variance filtering on the mutual information matrix to determine a selected feature set; Performing mutual information calculation on the selected feature set to determine a correlation matrix set; performing correlation filtering on the correlation matrix set to determine a sensor distribution set; The performing correlation filtering on the correlation matrix set to determine the sensor distribution set includes: Calculating node degrees of correlation matrices in the set of correlation matrices to determine a first node, where the first node is a node with the largest degree in the correlation matrix; Sort the correlation between the node with the largest degree and other nodes in the correlation matrix, filter the nodes, and determine the correlation sequence; Calculating the degree of an undirected graph of the correlation sequence to determine a second node, where the second node is a node with the largest degree of the undirected graph in the correlation sequence; A sensor distribution set is determined according to the first node and the second node.

2. A chip thermal sensor distribution method according to claim 1, characterized in that: The data set is constructed by performing thermal simulation on the chip, including: Perform thermal simulation on the chip to determine thermal simulation data; The thermal simulation data is exported and preprocessed to construct a data set.

3. The chip thermal sensor distribution method according to claim 1, characterized in that: The reconstructing the sensor distribution set to determine a target sensor distribution map includes: Polynomial processing is performed on the data in the sensor distribution set, and the processed data is substituted into a linear regression with a penalty term to calculate the regressor algorithm error, thereby determining a target sensor distribution map.

4. The chip thermal sensor distribution method according to claim 1, characterized in that: The performing variance filtering on the mutual information matrix to determine the selected feature set includes: Perform cyclic variance filtering on the mutual information matrix according to the first proportional step size to determine the selected feature set.

5. A chip thermal sensor distribution system, characterized in that: include: The first module is used to construct a data set by performing thermal simulation on the chip; The second module is used to calculate the mutual information of the data set and determine the mutual information matrix; The third module is used to filter the mutual information matrix to determine the sensor distribution set; A fourth module is configured to reconstruct the sensor distribution set and determine a target sensor distribution map; The second module is used to calculate the mutual information of the data set and determine the mutual information matrix, including: The correlation of each feature in the data set is calculated according to the mutual information calculation formula to construct the mutual information matrix; The third module is configured to filter the mutual information matrix to determine a sensor distribution set, including: Performing variance filtering on the mutual information matrix to determine a selected feature set; Performing mutual information calculation on the selected feature set to determine a correlation matrix set; performing correlation filtering on the correlation matrix set to determine a sensor distribution set; The performing correlation filtering on the correlation matrix set to determine the sensor distribution set includes: Calculating node degrees of correlation matrices in the set of correlation matrices to determine a first node, where the first node is a node with the largest degree in the correlation matrix; Sort the correlation between the node with the largest degree and other nodes in the correlation matrix, filter the nodes, and determine the correlation sequence; Calculating the degree of an undirected graph of the correlation sequence to determine a second node, where the second node is a node with the largest degree of the undirected graph in the correlation sequence; A sensor distribution set is determined according to the first node and the second node.

6. An electronic device, characterized in that: including a processor and a memory; The memory is used to store programs; The processor executes the program to implement the method according to any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that The storage medium stores a program, and the program is executed by a processor to implement the method according to any one of claims 1 to 4.

Images