Method, system, terminal and medium for estimating the effect of a process on the ir drop distribution

Abstract

The application provides a process influence prediction method, system, terminal and medium for IR voltage drop distribution, and specifically relates to the chip design technical field, and the scheme comprises the following steps: obtaining an estimated IR voltage drop distribution map produced under a target process; using a trained target deep learning model to predict the estimated IR voltage drop distribution map, obtaining an optimized IR voltage drop distribution map, and using a real IR voltage drop distribution map corresponding to the target process obtained after pre-taping and a preset circuit characteristic index threshold to train the target deep learning model. The scheme uses the real IR voltage drop distribution map corresponding to the target process to train the target deep learning model, which can improve the training efficiency, and make the trained target deep learning model accurately and efficiently predict the IR voltage drop distribution of the integrated circuit, optimize the IR voltage drop distribution map, guide the circuit design, optimization or fault diagnosis, and improve the yield of the integrated circuit, which has important production application value.

Description

Technical Field

[0001] The present invention relates to the field of chip design technology, and in particular to a method, system, terminal and medium for estimating the impact of a manufacturing process on IR voltage drop distribution. Background Art

[0002] Interconnect voltage drop (IR drop), also known as IR voltage drop, refers to the voltage drop and rise on the power and ground networks within an integrated circuit (IC). With the continuous advancement of semiconductor processes, metal interconnects are becoming narrower, their resistance is increasing, and the supply voltage is decreasing, making the IR drop effect increasingly pronounced. Therefore, in the chip industry, IR drop analysis is a crucial step in chip signoff.

[0003] Currently, the placement of individual sensors within the integrated circuit (IC) is typically determined by engineers based on their own experience and SPICE simulation results. This lack of effective consideration of uncontrollable factors in the manufacturing process can easily lead to differences in circuit characteristics between individual chips, resulting in suboptimal IR drop measurement results and, consequently, low chip yield. This demonstrates that the existing technology suffers from a flaw in chip design due to inaccurate IR drop analysis, which in turn leads to low chip yield. Summary of the Invention

[0004] In view of the above-mentioned deficiencies in the prior art, the purpose of the present invention is to provide a method, system, terminal and medium for estimating the impact of the process on the IR drop distribution, aiming to solve the problem in the prior art of low chip design yield due to inaccurate analysis of IR drop.

[0005] To achieve the above objectives, the present invention provides, in a first aspect, a method for estimating the impact of a process on IR drop distribution, comprising:

[0006] Obtain the estimated IR drop distribution diagram produced under the target process;

[0007] The IR drop distribution map to be estimated is estimated using the trained target deep learning model to obtain an optimized IR drop distribution map. The trained target deep learning model is obtained by training the target deep learning model using the actual IR drop distribution map corresponding to the target process obtained after pre-tapeout and the preset circuit characteristic index threshold corresponding to the target process.

[0008] Optionally, the training process of the target deep learning model includes:

[0009] Obtaining actual IR drop distribution graphs corresponding to a plurality of target processes;

[0010] Training the target deep learning model using the real IR pressure drop distribution map to obtain an updated target deep learning model;

[0011] generating an updated IR drop distribution graph using the updated target deep learning model, and measuring a circuit characteristic index value of the updated IR drop distribution graph;

[0012] The circuit characteristic index value is compared with a preset circuit characteristic index threshold, and the updated target deep learning model is iteratively optimized according to the comparison result to obtain a trained target deep learning model.

[0013] Optionally, the target deep learning model is a GAN model, the GAN model includes at least an initial generator and an initial discriminator, and the using the real IR drop distribution map to train the target deep learning model to obtain an updated target deep learning model includes:

[0014] adding noise to the real IR pressure drop distribution graph to generate a plurality of pseudo IR pressure drop distribution graphs;

[0015] constructing a training set using the real IR pressure drop distribution map and the pseudo IR pressure drop distribution map;

[0016] The initial generator and the initial discriminator in the target deep learning model are alternately and iteratively trained using the training set to obtain an updated target deep learning model.

[0017] Optionally, the alternately iteratively training the initial generator and the initial discriminator in the target deep learning model using the training set to obtain an updated target deep learning model includes:

[0018] Fixing the initial discriminator, training the initial generator using the training set, generating a plurality of updated IR drop distribution graphs and obtaining an updated generator;

[0019] Updating the training set using the updated IR pressure drop distribution graph to obtain an updated training set;

[0020] The updated generator is fixed, and the initial discriminator is trained using the updated training set to obtain an updated target deep learning model.

[0021] Optionally, obtaining a plurality of actual IR drop distribution graphs corresponding to the target processes includes:

[0022] Obtain all bare dies on all wafers of the same batch produced under the target process;

[0023] The IR voltage drop distribution diagrams of all the bare chips are measured to obtain actual IR voltage drop distribution diagrams corresponding to several target processes.

[0024] Optionally, after obtaining the optimized IR pressure drop distribution diagram, the method further includes:

[0025] According to the optimized IR voltage drop distribution diagram, placement points of each sensor corresponding to the IR voltage drop distribution diagram to be estimated in the integrated circuit are determined.

[0026] A second aspect of the present invention provides a system for estimating the impact of a process on IR drop distribution, the system comprising:

[0027] The module for acquiring data to be estimated is used to obtain the IR drop distribution diagram to be estimated produced under the target process;

[0028] The IR drop distribution estimation module is used to use the trained target deep learning model to estimate the IR drop distribution map to be estimated to obtain an optimized IR drop distribution map. The trained target deep learning model is obtained by training the target deep learning model using the actual IR drop distribution map corresponding to the target process obtained after pre-tapeout and the preset circuit characteristic indicator threshold corresponding to the target process.

[0029] Optionally, the system further includes a target deep learning model training module, which includes a training data acquisition module, a model update module, a circuit characteristic index value measurement module and a trained model output module, wherein:

[0030] The training data acquisition module is used to obtain a plurality of real IR drop distribution diagrams corresponding to the target processes;

[0031] The model updating module is configured to train the target deep learning model using the real IR pressure drop distribution map to obtain an updated target deep learning model;

[0032] The circuit characteristic index value measurement module is used to generate an updated IR voltage drop distribution graph using the updated target deep learning model, and measure the circuit characteristic index value of the updated IR voltage drop distribution graph;

[0033] The trained model output module is used to compare the circuit characteristic index value with a preset circuit characteristic index threshold, and iteratively optimize the updated target deep learning model according to the comparison result to obtain a trained target deep learning model.

[0034] A third aspect of the present invention provides an intelligent terminal, which includes a memory, a processor, and a process impact estimation program stored in the memory and runnable on the processor. When the process impact estimation program is executed by the processor, it implements any step of the above-mentioned process impact estimation method.

[0035] A fourth aspect of the present invention provides a computer-readable storage medium, which stores a program for estimating the impact of a process on IR pressure drop distribution. When the program for estimating the impact of a process on IR pressure drop distribution is executed by a processor, the program implements any step of the above-mentioned method for estimating the impact of a process on IR pressure drop distribution.

[0036] Compared with the existing technology, the beneficial effects of this solution are as follows:

[0037] The present invention obtains an IR drop distribution map to be estimated produced under a target process, and uses a trained target deep learning model to estimate the IR drop distribution map to be estimated, thereby obtaining an optimized IR drop distribution map. Since different processes correspond to different integrated circuit manufacturing processes and design rules, the present invention uses the actual IR drop distribution map corresponding to the target process as a training sample to train the target deep learning model, which can improve the effectiveness and efficiency of training the target deep learning model. The trained model is detected using the circuit characteristic indicator threshold corresponding to the target process, which can effectively improve the accuracy of using the trained model to detect the IR drop distribution map produced under the corresponding process, thereby facilitating the use of the trained target deep learning model to accurately and efficiently predict the IR drop distribution of the integrated circuit, and has important production application value for optimizing the IR drop distribution map, guiding circuit design, optimization or fault diagnosis, and improving the yield of the integrated circuit. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0039] Figure 1 A flow chart of an embodiment of a method for estimating the effect of a manufacturing process on IR drop distribution according to the present invention;

[0040] Figure 2 A flow chart of another embodiment of a method for estimating the effect of a manufacturing process on IR drop distribution according to the present invention;

[0041] Figure 3A flow chart of another embodiment of a method for estimating the effect of a manufacturing process on IR drop distribution according to the present invention;

[0042] Figure 4 Schematic diagram of the GAN model training process for estimating IR drop distribution according to the present invention;

[0043] Figure 5 A flow chart of another embodiment of a method for estimating the effect of a manufacturing process on IR drop distribution according to the present invention;

[0044] Figure 6 A flow chart of another embodiment of a method for estimating the effect of a manufacturing process on IR drop distribution according to the present invention;

[0045] Figure 7 is a contour map of the actual IR voltage drop distribution corresponding to the die inside the wafer of the present invention;

[0046] Figure 8 Schematic diagram of the system module for estimating the impact of the process on IR drop distribution of the present invention;

[0047] Figure 9 It is a schematic diagram of the structure of the intelligent terminal of the present invention. DETAILED DESCRIPTION

[0048] In the following description, specific details such as particular system structures and techniques are provided for purposes of illustration, not limitation, to facilitate a thorough understanding of the embodiments of the present invention. However, it will be apparent to those skilled in the art that the present invention may be practiced in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present invention with unnecessary detail.

[0049] It will be understood that when used in this specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0050] It should also be understood that the terms used in the present specification are only for the purpose of describing particular embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0051] It should be further understood that the term "and / or" used in the present description and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0052] The following is a clear and complete description of the technical solutions in the embodiments of the present invention, in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0053] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0054] In response to the problem in the prior art of low chip design yield due to inaccurate IR drop analysis, the present invention proposes a method for predicting the impact of the process on the IR drop distribution. The method mainly uses the real IR drop distribution map to train the target deep learning model to obtain the trained target deep learning model, and then uses the trained target deep learning model to predict the IR drop distribution of the target integrated circuit. This method has important production application value for optimizing the IR drop distribution map, guiding circuit design, optimization or fault diagnosis, and improving the yield of integrated circuits.

[0055] The embodiment of the present invention provides a method for estimating the impact of the process on the IR voltage drop distribution, which is deployed on electronic devices such as computers and servers. The application scenario is to detect the yield rate of integrated circuit designs. It is aimed at detecting whether the integrated circuit design is qualified by analyzing the IR voltage drop distribution of the integrated circuit. The type of the above-mentioned integrated circuit is not limited. It can be taxis, private cars and other vehicles that can provide online car-hailing services. Specifically, Figure 1 As shown, the steps of the method in this embodiment include:

[0056] Step S1000: obtaining an IR drop distribution diagram to be estimated produced under a target process;

[0057] Specifically, an IR drop distribution map to be estimated produced under a target process is obtained, wherein the IR drop distribution map to be estimated may be an IR drop distribution map of bare dies on some or all wafers of the same batch generated during the initial tape-out, may be a partial or overall IR drop distribution map of an integrated circuit that has been produced, or may be a module or overall IR drop distribution map of other integrated circuits to be tested, and the present invention does not impose any specific restrictions thereto.

[0058] Step S2000: Use the trained target deep learning model to estimate the IR voltage drop distribution map to obtain an optimized IR voltage drop distribution map. The trained target deep learning model is obtained by training the target deep learning model using the actual IR voltage drop distribution map corresponding to the target process obtained after pre-tapeout and the preset circuit characteristic index threshold corresponding to the target process.

[0059] Specifically, in order to more accurately estimate the IR drop distribution map to be estimated, this embodiment uses the actual IR drop distribution map corresponding to the target process obtained after pre-tapeout as a training sample to train the target deep learning model, and uses the circuit characteristic index threshold set according to the integrated circuit process design standard corresponding to the target process to detect whether the trained model meets the detection accuracy requirements. Only when the trained target deep learning model meets the detection accuracy requirements is it considered to be a trained target deep learning model, and then the trained target deep learning model is deployed in practical applications, such as circuit design software, monitoring systems, detection systems, etc.

[0060] The trained target deep learning model is used to perform real-time IR Drop estimation on the IR drop distribution map to be estimated, and an optimized IR drop distribution map is obtained to guide circuit design, optimization, or fault diagnosis.

[0061] In this embodiment, the target deep learning model is trained using the actual IR drop distribution map corresponding to the target process obtained by pre-tapeout (such as the first tape-out or the Nth small amount of tape-out), which can improve the effectiveness and efficiency of the target deep learning model training. By detecting the trained model according to the circuit characteristic index threshold set according to the integrated circuit process design standard corresponding to the target process, the detection accuracy of the trained model can be effectively improved, which is conducive to using the trained target deep learning model to accurately and efficiently predict the IR drop distribution of integrated circuits produced according to the integrated circuit manufacturing process and design rules corresponding to the target process. It has important production application value for optimizing the IR drop distribution map, guiding circuit design, optimization or fault diagnosis, and improving the yield of integrated circuits.

[0062] like Figure 2 As shown, in one embodiment, the training process of the target deep learning model in step S2000 includes:

[0063] Step S2100: obtaining a plurality of actual IR drop distribution graphs corresponding to the target processes;

[0064] Specifically, the real IR drop distribution diagrams of all bare crystals in silicon wafers used for batch semiconductor integrated circuit production under the integrated circuit manufacturing process and design rules corresponding to the target process obtained by pre-tapeout (such as the first tape-out or the Nth small-scale tape-out) are obtained. By using the complete IR drop distribution diagrams of multiple wafers in the integrated circuits produced in the same batch as the training set, the trained target deep learning model can make a complete and comprehensive analysis of the IR drop distribution of the integrated circuit as a whole.

[0065] Step S2200: training the target deep learning model using the real IR drop distribution graph to obtain an updated target deep learning model;

[0066] Specifically, based on the characteristics of the real IR drop profiles in the training set, a suitable deep learning model for image classification is selected as the target deep learning model. Initialization parameters are set for the target deep learning model to construct an initialized target deep learning model. The target deep learning model is then trained using the complete IR drop profiles of multiple wafers from the same batch of integrated circuits, which improves the effectiveness and accuracy of the trained model.

[0067] Step S2300: generating an updated IR drop distribution graph using the updated target deep learning model, and measuring a circuit characteristic index value of the updated IR drop distribution graph;

[0068] Specifically, during the training of the target deep learning model, after each training to obtain an updated target deep learning model, the target deep learning model after the last update is used to generate an updated IR voltage drop distribution map to measure the circuit characteristic index value of the updated IR voltage drop distribution map, wherein the circuit characteristic index value includes performance indicators related to the IR voltage drop distribution in the integrated circuit, including at least changes in voltage on the power supply and ground networks in the integrated circuit, and / or changes in the current and resistance in the integrated circuit, etc.

[0069] Step S2400: Compare the circuit characteristic index value with a preset circuit characteristic index threshold, and iteratively optimize the updated target deep learning model according to the comparison result to obtain a trained target deep learning model.

[0070] Specifically, by using the circuit characteristic indicator threshold set according to the integrated circuit process design standard to detect the circuit characteristic indicator value, it is possible to simply, directly and effectively detect whether the prediction accuracy of the IR drop distribution diagram of the updated target deep learning model obtained after this training meets the integrated circuit process design standard.

[0071] In this embodiment, based on the characteristics of the real IR voltage drop distribution map in the training set, a suitable target deep learning model is selected as the detection model, and the target deep learning model is trained using the real IR voltage drop distribution map and a preset strict circuit characteristic index threshold. This can improve the effectiveness of model training and at the same time improve the accuracy of the trained target deep learning model.

[0072] like Figure 3 and Figure 4 As shown, in one embodiment, the target deep learning model in step S2200 is a GAN model, and the GAN model includes at least an initial generator (Generator) and an initial discriminator (Discriminator). The target deep learning model is trained using the real IR drop distribution map to obtain an updated target deep learning model, including:

[0073] Step S2210: adding noise to the real IR pressure drop distribution graph to generate a plurality of pseudo IR pressure drop distribution graphs;

[0074] Specifically, in actual operation, the real IR drop distribution maps that can be used as training samples are often limited, but generally the more training samples there are, the higher the accuracy of the trained model. Therefore, in order to enhance the robustness of the target deep learning model, random noise (Z) can be added to the real IR drop distribution map (C) to construct multiple pseudo IR drop distribution maps to expand the number of training samples.

[0075] Step S2220: constructing a training set using the real IR pressure drop distribution map and the pseudo IR pressure drop distribution map;

[0076] Specifically, before the training samples in the training set are input into the target deep learning model, the training samples are first preprocessed, such as normalization, denoising, feature extraction, etc., so that they can be input into the GAN model.

[0077] Step S2230: Use the training set to alternately iteratively train the initial generator and the initial discriminator in the target deep learning model to obtain an updated target deep learning model.

[0078] Specifically, since the GAN model can train any generator network, there is no need to use Markov chains for repeated sampling. It is simple and efficient, and can better model the data distribution, making the generated images sharper and clearer. Therefore, this embodiment selects the GAN model as the target deep learning model, and the GAN model includes at least one initial generator and one initial discriminator, and then uses the training set to alternately iteratively train the initial generator and initial discriminator in the GAN model. During the training process, the parameters of the generator and the discriminator are updated alternately, mainly using the generator to generate pseudo image samples, and at the same time, all the labels of the pseudo images are defined as real labels; the discriminator is trained using a training set constructed from real IR drop distribution maps and pseudo IR drop distribution maps to improve the discriminator's ability to identify real images and pseudo images until the entire GAN model converges.

[0079] It should be stated that in order to control the computing resources and training time of the training target deep learning model, in actual applications, the appropriate network structure, optimization algorithm and training strategy can be flexibly selected according to the specific application requirements and the characteristics of the data set.

[0080] like Figure 5 As shown, in one embodiment, the step S2230 of alternately iteratively training the initial generator and the initial discriminator in the target deep learning model using the training set to obtain an updated target deep learning model includes:

[0081] Step S2231: fixing the initial discriminator, using the training set to train the initial generator, generating a plurality of updated IR drop distribution graphs and obtaining an updated generator;

[0082] Step S2232: updating the training set using the updated IR drop distribution graph to obtain an updated training set;

[0083] Step S2233: Fix the updated generator, use the updated training set to train the initial discriminator, and obtain the updated target deep learning model.

[0084] Specifically, during the training process, the initial discriminator is first fixed and trained, and then the updated generator is fixed and trained. The parameters of the generator and discriminator are updated alternately over multiple cycles. As the number of alternating training increases, the pseudo IR drop distribution map generated by the generator becomes increasingly close to the true IR Drop distribution. Therefore, to improve training efficiency, the training set is updated using the updated IR drop distribution map generated by the generator after the latest training. The updated training set is obtained, so the training samples in the training set are the true IR Drop distribution map and the updated IR Drop distribution map generated by the generator after the latest training. After multiple cycles of alternating training, the generator can generate fake data similar to the true IR Drop distribution, and the discriminator can accurately distinguish whether the input data is real or fake data generated by the generator. Finally, the true IR Drop distribution map is used as a validation set to evaluate the performance of the GAN model to detect whether the generated IR Drop distribution map can accurately reflect the distribution of the true IR Drop. Based on the evaluation results, the model structure, parameters, or training strategy are adjusted to optimize the model performance.

[0085] Use the trained GAN model to generate an estimated IR drop distribution map. Analyze the generated estimated IR drop distribution map to extract useful information, such as the average, standard deviation, and maximum value of the voltage drop, to evaluate the performance of the integrated circuit or system.

[0086] It's important to note that when training a GAN model, the training of the discriminator and generator are interdependent. The discriminator's training relies on the fake data generated by the generator, while the generator's training relies on the error feedback from the discriminator. Furthermore, to prevent the model from falling into local optimal solutions, regularization terms such as weight decay and dropout are also required during training.

[0087] like Figure 6 As shown, in one embodiment, obtaining a plurality of actual IR drop distribution graphs corresponding to the target processes in step S2100 includes:

[0088] Step S2110: Acquire all bare dies on all wafers of the same batch corresponding to the target process;

[0089] Step S2120: measuring the IR voltage drop distribution graphs of all the bare chips to obtain actual IR voltage drop distribution graphs corresponding to several target processes.

[0090] Specifically, this embodiment selects all the bare dies on all wafers of the same batch corresponding to the target process obtained in the initial tape-out as the carriers for obtaining training samples, and obtains the real IR drop distribution map corresponding to the target process as the training sample by measuring the IR drop distribution map of all the bare dies. In this way, it can be ensured that the training sample can include the integrity and comprehensiveness of the IR drop distribution map of the integrated circuit to be tested, thereby avoiding the situation where a high-precision target deep learning model cannot be trained using only part of the IR drop distribution map of the integrated circuit. Figure 7 The figure shows the actual IR voltage drop distribution contour map corresponding to the die inside the wafer. Specifically, it shows the die array schematic diagram inside the wafers of the same batch, and the actual IR voltage drop distribution contour map corresponding to the die inside each wafer, namely Die_1 to Die_n.

[0091] In one embodiment, after obtaining the optimized IR drop distribution diagram in step S2000, step S3000 is further included. Step S3000 specifically includes:

[0092] According to the optimized IR voltage drop distribution diagram, placement points of each sensor corresponding to the IR voltage drop distribution diagram to be estimated in the integrated circuit are determined.

[0093] Specifically, after using the trained target deep learning model to predict the IR voltage drop distribution map and obtaining the optimized IR voltage drop distribution map, by comparing and analyzing the differences and similarities between the optimized IR voltage drop distribution map and the IR voltage drop distribution map to be estimated, it is possible to clearly determine the direction and distance in which the placement points of each sensor corresponding to the IR voltage drop distribution map to be estimated within the integrated circuit need to be adjusted, thereby accurately adjusting the placement points of each sensor within the integrated circuit to take into account the impact on the process design of the integrated circuit during the manufacturing process, thereby effectively improving the yield rate of the produced integrated circuits.

[0094] like Figure 8 As shown, corresponding to the above-mentioned method for estimating the impact of a process on IR pressure drop distribution, an embodiment of the present invention further provides a system for estimating the impact of a process on IR pressure drop distribution. The system for estimating the impact of a process on IR pressure drop distribution includes:

[0095] The module 810 for acquiring data to be estimated is used to acquire an IR drop distribution diagram to be estimated produced under a target process;

[0096] The IR drop distribution estimation module 820 is used to use the trained target deep learning model to estimate the IR drop distribution map to be estimated to obtain an optimized IR drop distribution map. The trained target deep learning model is obtained by training the target deep learning model using the actual IR drop distribution map corresponding to the target process obtained after pre-tapeout and the preset circuit characteristic indicator threshold corresponding to the target process.

[0097] Furthermore, the system also includes a target deep learning model training module, which includes a training data acquisition module, a model update module, a circuit characteristic index value measurement module and a trained model output module, wherein,

[0098] The training data acquisition module is used to obtain a plurality of real IR drop distribution diagrams corresponding to the target processes;

[0099] The model updating module is configured to train the target deep learning model using the real IR pressure drop distribution map to obtain an updated target deep learning model;

[0100] The circuit characteristic index value measurement module is used to generate an updated IR voltage drop distribution graph using the updated target deep learning model, and measure the circuit characteristic index value of the updated IR voltage drop distribution graph;

[0101] The trained model output module is used to compare the circuit characteristic index value with a preset circuit characteristic index threshold, and iteratively optimize the updated target deep learning model according to the comparison result to obtain a trained target deep learning model.

[0102] Specifically, in this embodiment, the specific functions of the above-mentioned process impact on IR pressure drop distribution prediction system can also refer to the corresponding description in the above-mentioned process impact on IR pressure drop distribution prediction method, which will not be repeated here.

[0103] Based on the above embodiment, the present invention also provides an intelligent terminal, whose principle block diagram can be shown as follows: Figure 9As shown. The above-mentioned intelligent terminal includes a processor, a memory, a network interface and a display screen connected through a system bus. Among them, the processor of the intelligent terminal is used to provide computing and control capabilities. The memory of the intelligent terminal includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a process impact estimation program on IR pressure drop distribution. The internal memory provides an environment for the operation of the operating system in the non-volatile storage medium and the process impact estimation program on IR pressure drop distribution. The network interface of the intelligent terminal is used to communicate with an external terminal through a network connection. When the process impact estimation program on IR pressure drop distribution is executed by the processor, the steps of any of the above-mentioned process impact estimation methods on IR pressure drop distribution are implemented. The display screen of the intelligent terminal can be a liquid crystal display or an electronic ink display.

[0104] Those skilled in the art will understand that Figure 9 The principle block diagram shown in the figure is only a block diagram of a partial structure related to the solution of the present invention and does not constitute a limitation on the smart terminal to which the solution of the present invention is applied. The specific smart terminal may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0105] In one embodiment, a smart terminal is provided, comprising a memory, a processor, and a program for estimating the impact of a process on IR drop distribution, the program being stored in the memory and executable on the processor. When the program for estimating the impact of a process on IR drop distribution is executed by the processor, the steps of any one of the methods for estimating the impact of a process on IR drop distribution provided in an embodiment of the present invention are implemented.

[0106] An embodiment of the present invention also provides a computer-readable storage medium, which stores a program for estimating the impact of a process on IR drop distribution. When the program for estimating the impact of a process on IR drop distribution is executed by a processor, the steps of any method for estimating the impact of a process on IR drop distribution provided in an embodiment of the present invention are implemented.

[0107] It should be understood that the sequence numbers of the steps in the above embodiments do not imply a specific order of execution; the order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0108] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the above-mentioned device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0109] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0110] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.

[0111] In the embodiments provided by the present invention, it should be understood that the disclosed apparatus / terminal device and method can be implemented in other ways. For example, the apparatus / terminal device embodiments described above are merely illustrative. For example, the division of the modules or units described above is merely a logical functional division. In actual implementation, other division methods may be used. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not implemented.

[0112] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, it should be understood by those skilled in the art that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A method for estimating the effect of a process on IR drop distribution, characterized in that: The following steps are involved: Obtaining an IR drop profile to be estimated produced under a target process, wherein the IR drop profile to be estimated is an IR drop profile of bare dies on some or all wafers of the same batch generated during an initial tape-out, or a partial or overall IR drop profile of an integrated circuit that has been produced, or an IR drop profile of a module or the entirety of another integrated circuit to be tested; Using a trained target deep learning model to estimate the IR drop distribution map to be estimated, to obtain an optimized IR drop distribution map, the trained target deep learning model is obtained by training the target deep learning model using a real IR drop distribution map corresponding to the target process obtained after pre-tapeout and a preset circuit characteristic index threshold corresponding to the target process, the circuit characteristic index threshold being set according to an integrated circuit process design standard corresponding to the target process; The training process of the target deep learning model includes: Obtaining true IR voltage drop distribution maps of all bare wafers in silicon wafers used in batch semiconductor integrated circuit fabrication, produced under the integrated circuit manufacturing process and design rules corresponding to the target process obtained during pre-tapeout, and using the true IR voltage drop distribution maps of all wafers in the integrated circuits produced in the same batch as a training set to train the target deep learning model, thereby obtaining an updated target deep learning model; generating an updated IR drop distribution graph using the updated target deep learning model, and measuring a circuit characteristic index value of the updated IR drop distribution graph; Comparing the circuit characteristic index value with a preset circuit characteristic index threshold, and iteratively optimizing the updated target deep learning model according to the comparison result to obtain a trained target deep learning model; The circuit characteristic index values ​​include performance indicators related to IR drop distribution in the integrated circuit, including voltage changes on the power supply and ground networks in the integrated circuit, and / or changes in the magnitude of current and resistance in the integrated circuit; After obtaining the optimized IR pressure drop distribution diagram, the method further includes: According to the optimized IR voltage drop distribution diagram, placement points of each sensor corresponding to the IR voltage drop distribution diagram to be estimated in the integrated circuit are determined.

2. The method for estimating the effect of a process on IR drop distribution according to claim 1, wherein: The target deep learning model is a GAN model, which includes at least an initial generator and an initial discriminator. The target deep learning model is trained using the real IR drop distribution map to obtain an updated target deep learning model, including: adding noise to the real IR pressure drop distribution graph to generate a plurality of pseudo IR pressure drop distribution graphs; constructing a training set using the real IR pressure drop distribution map and the pseudo IR pressure drop distribution map; The initial generator and the initial discriminator in the target deep learning model are alternately and iteratively trained using the training set to obtain an updated target deep learning model.

3. The method for estimating the effect of a process on IR drop distribution according to claim 2, wherein: The step of alternately iteratively training the initial generator and the initial discriminator in the target deep learning model using the training set to obtain an updated target deep learning model includes: Fixing the initial discriminator, training the initial generator using the training set, generating a plurality of updated IR drop distribution graphs and obtaining an updated generator; Updating the training set using the updated IR pressure drop distribution graph to obtain an updated training set; The updated generator is fixed, and the initial discriminator is trained using the updated training set to obtain an updated target deep learning model.

4. The system for estimating the impact of the process on the IR drop distribution is characterized by: The system comprises: a module for acquiring data to be estimated, configured to acquire an IR voltage drop distribution map to be estimated produced under a target process, wherein the IR voltage drop distribution map to be estimated is an IR voltage drop distribution map of bare dies on some or all wafers of the same batch generated during an initial tape-out, or a partial or overall IR voltage drop distribution map of an integrated circuit that has been produced, or an IR voltage drop distribution map of a module or the entirety of another integrated circuit to be tested; an IR drop distribution estimation module, configured to estimate the IR drop distribution map to be estimated using a trained target deep learning model to obtain an optimized IR drop distribution map, wherein the trained target deep learning model is obtained by training the target deep learning model using a real IR drop distribution map corresponding to the target process obtained after pre-tapeout and a preset circuit characteristic index threshold corresponding to the target process, wherein the circuit characteristic index threshold is set according to the integrated circuit process design standard corresponding to the target process; The system further comprises: The target deep learning model training module includes a training data acquisition module, a model update module, a circuit characteristic index value measurement module and a trained model output module, wherein: The training data acquisition module is used to obtain a plurality of real IR drop distribution diagrams corresponding to the target processes; The training data acquisition module includes: Obtain the true IR voltage drop distribution maps of all bare silicon wafers used in the production of semiconductor integrated circuits in batches produced under the integrated circuit manufacturing process and design rules corresponding to the target process obtained during pre-tapeout, and use the true IR voltage drop distribution maps of all wafers in the integrated circuits produced in the same batch as the training set; The model updating module is configured to train the target deep learning model using the real IR pressure drop distribution map to obtain an updated target deep learning model; The circuit characteristic index value measurement module is used to generate an updated IR voltage drop distribution graph using the updated target deep learning model, and measure the circuit characteristic index value of the updated IR voltage drop distribution graph; The trained model output module is used to compare the circuit characteristic index value with a preset circuit characteristic index threshold, and iteratively optimize the updated target deep learning model according to the comparison result to obtain a trained target deep learning model; The circuit characteristic index values ​​include performance indicators related to IR drop distribution in the integrated circuit, including voltage changes on the power supply and ground networks in the integrated circuit, and / or changes in the magnitude of current and resistance in the integrated circuit; After obtaining the optimized IR pressure drop distribution diagram, the method further includes: According to the optimized IR voltage drop distribution diagram, placement points of each sensor corresponding to the IR voltage drop distribution diagram to be estimated in the integrated circuit are determined.

5. Intelligent terminal, characterized in that: The intelligent terminal includes a memory, a processor, and a process impact estimation program stored in the memory and runnable on the processor. When the process impact estimation program is executed by the processor, the steps of the process impact estimation method on the IR pressure drop distribution as described in any one of claims 1 to 3 are implemented.

6. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a program for estimating the impact of a process on IR drop distribution. When the program for estimating the impact of a process on IR drop distribution is executed by a processor, the steps of the method for estimating the impact of a process on IR drop distribution as described in any one of claims 1 to 3 are implemented.

Images