Static time sequence analysis method and device before wiring
Through graph-based modeling and network delay prediction model of distributed gradient enhancement library, the low generalization and low parallelism problems in multi-design modes in static timing analysis before cabling are solved, and high-precision and fast timing analysis are achieved As a result, the accuracy and speed of the prior art are significantly improved.
Patent Information
- Application Number
- CN202411121715.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-15
- Publication Date
- 2024-11-26
AI Technical Summary
In the prior art, static timing analysis before wiring has low generalization across multiple design patterns and low parallelism of local delay prediction, resulting in insufficient accuracy of timing analysis and slow slack prediction speed in multi-mode design.
Using graph-based modeling method, the integrated circuit is divided into networks, the network delay prediction feature set is extracted, the network delay prediction model is constructed using a distributed gradient enhancement library, and the arrival time and relaxation are calculated through a linear complexity algorithm to distinguish different working conditions. network latency prediction.
High-precision static timing analysis in multi-design mode is realized, which significantly accelerates the slack prediction speed, improves the generalization and parallelism of network delay prediction, and has a lower computing complexity than the existing technology.
Smart Images

Figure CN119026543A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of chip design methods, and in particular to a static timing analysis method and device before wiring. Background Art
[0002] Pre-wiring static timing analysis is an important step in the contemporary chip design process, and is used to guide pre-wiring optimization to determine the location of fixed components, including macros and cells, in the subsequent process. Therefore, the accuracy of pre-wiring static timing analysis is crucial, providing valuable guidance for timing optimization during circuit layout; in addition, the speed of pre-wiring static timing analysis can reduce the iteration time of circuit layout and optimization, thereby simplifying the design process of the entire chip.
[0003] At present, a method for pre-wiring timing analysis by extracting features from the network has been proposed in the prior art. However, there are two main problems and disadvantages in this method:
[0004] (1) Low generalization across multiple design modes limits the accuracy of analyzing the timing of multi-mode design implementations of circuits;
[0005] (2) The low parallelism of local delay prediction slows down the relaxation prediction of timing paths involving multiple networks.
[0006] For the above-mentioned technical problems, it is urgent to reconstruct a feature extraction method to solve the above two defects by extracting features of the machine learning regression model for local delay prediction, so as to achieve generalization of pre-wiring static timing analysis under multiple design modes and speed up the relaxation prediction process. Summary of the invention
[0007] In order to solve the problems of low generalization across multiple design modes and low parallelism of local delay prediction in the prior art, a fast and accurate pre-routing static timing analysis method and system are implemented.
[0008] In a first aspect, an embodiment of the present application provides a static timing analysis method before wiring, the method comprising:
[0009] Modeling step: modeling the integrated circuit before wiring based on a graph, dividing the circuit into a network, and outputting a network modeling result of the circuit, wherein the network includes: a driving unit, interconnection lines, and receiving pins;
[0010] Feature extraction step: extracting a network delay prediction feature set from pre-routing design data, wherein the timing features of the prediction feature set include: reference delay and input slew;
[0011] Network delay prediction step: based on the network delay prediction feature set, a distributed gradient enhancement library is used to build a network delay prediction model, and the network delay prediction results under different working conditions are output;
[0012] Timing result calculation step: Based on the network modeling results and the network delay prediction results under various working conditions, a linear complexity algorithm is used to calculate the arrival time and relaxation, and the final pre-routing static timing analysis results are output.
[0013] As an optional implementation, in the first aspect of the present invention, the modeling step further includes:
[0014] Network division step: divide the circuit into a directed acyclic graph for network division, wherein the driving unit includes an input pin, the node of the directed acyclic graph is each of the networks, and the edge of the directed acyclic graph is each input pin of the driving unit.
[0015] As an optional implementation, in the first aspect of the present invention, in the above-mentioned feature extraction step, the network delay prediction feature set includes: physical features and timing features, wherein the timing features include reference delay, reference slew and input slew.
[0016] As an optional implementation manner, in the first aspect of the present invention, the above-mentioned network delay prediction step further includes:
[0017] Under different working conditions, network delay prediction is performed based on a distributed gradient enhancement library, wherein a data set of the enhancement library is the total number of working conditions of all the networks, and a single data of the data set is one of the working conditions of each network, wherein the data is two-dimensional data, and the two-dimensional data includes: a feature set of the data and a true value of the network delay of the data;
[0018] The working conditions include: different input pins, receiving pins, logical responses of network signals to input signals, and types of input signals.
[0019] As an optional implementation manner, in the first aspect of the present invention, the above-mentioned static timing parameter calculation step further includes:
[0020] Arrival time calculation steps: according to the network delay, the arrival time is calculated using a graph traversal algorithm, and the arrival time of the network receiving pin is calculated simultaneously under different working conditions; after determining the rising arrival time and falling arrival time of the end point of the timing path, the maximum value operation is used to calculate the final arrival time of the end point.
[0021] As an optional implementation manner, in the first aspect of the present invention, the above-mentioned static timing parameter calculation step further includes:
[0022] Slack time calculation step: Based on the final arrival time, slack is calculated using a specified clock cycle and timing information associated with a historical database.
[0023] As an optional implementation manner, in the first aspect of the present invention, the above-mentioned arrival time calculation step further includes:
[0024] Rise arrival time calculation step: for the calculation of the rise arrival time, select the rise network delay under the positive response from the input pin, add the selected network delay to the rise arrival time at the input pin, and calculate the rise arrival time at the receiving pin under a set of working conditions; calculate the rise network delay under the negative response of the input pin, and set the rise arrival time at the receiving pin to the maximum value of all calculated rise arrival times;
[0025] Fall arrival time calculation steps: for the calculation of the fall arrival time, select the fall network delay under the positive response from the input pin, add the selected network delay to the fall arrival time at the input pin, and calculate the fall arrival time at the receiving pin under a set of working conditions; calculate the fall network delay under the negative response of the input pin, and set the fall arrival time at the receiving pin to the maximum value of all calculated fall arrival times.
[0026] In a second aspect, an embodiment of the present application provides a pre-wiring static timing analysis device. According to the pre-wiring static timing analysis method, the device includes:
[0027] Modeling unit: used for modeling the integrated circuit before wiring based on graph, dividing the circuit into networks, and outputting the network modeling result of the circuit, wherein the network includes: a driving unit, interconnection lines and receiving pins;
[0028] A feature extraction unit: used to extract a network delay prediction feature set from pre-routing design data, wherein the timing features of the prediction feature set include: reference delay and input slew;
[0029] A network delay prediction unit is used to construct a network delay prediction model based on the network delay prediction feature set and adopt a distributed gradient enhancement library to output network delay prediction results under different working conditions;
[0030] Timing result calculation unit: based on the network modeling results and the network delay prediction results under various working conditions, a linear complexity algorithm is used to calculate the arrival time and relaxation, and the final static timing analysis result before wiring is output.
[0031] In a third aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned pre-wiring static timing analysis method.
[0032] In a fourth aspect, an embodiment of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the pre-wiring static timing analysis method as described above when executing the program.
[0033] Compared with the related prior art, it has the following outstanding beneficial effects:
[0034] 1. The present invention realizes a fast and accurate static timing analysis method before wiring; the method of the present invention can provide consistent and higher accuracy than the existing technical solutions in various design modes, and the speed is much faster than the existing technical solutions;
[0035] 2. The present invention realizes the customization of a concise network delay prediction feature set, in which the timing features include reference delay and input slew; the method of the present invention enhances the generalization and parallelism of network delay prediction respectively;
[0036] 3. The present invention realizes the use of a distributed gradient enhancement library model for solving static timing analysis before wiring; the method of the present invention has a higher prediction accuracy of network delay than the existing technical solutions and other integrated learning methods under the same conditions;
[0037] 4. The present invention implements a linear complexity algorithm to accelerate the calculation of arrival time; the computational complexity and time complexity of the method of the present invention are much lower than those of the prior art methods. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:
[0039] Figure 1 A schematic diagram of a static timing analysis method before wiring according to an embodiment of the present invention;
[0040] Figure 2 A schematic diagram of a graph-based modeling method according to an embodiment of the present invention;
[0041] Figure 3 It is a schematic diagram of an unreasonable approximate embodiment in the prior art of the present invention;
[0042] Figure 4 It is a schematic diagram of timing characteristics of a specific embodiment of the present invention;
[0043] Figure 5It is a schematic diagram of a method for calculating arrival time according to a specific embodiment of the present invention;
[0044] Figure 6 It is a schematic diagram of a static timing analysis device before wiring according to an embodiment of the present invention;
[0045] Figure 7 It is a schematic diagram of computer hardware of the present invention;
[0046] Figure 8 It is a schematic diagram of a timing analysis method before wiring in the prior art of the present invention.
[0047] In the above picture:
[0048] 101 modeling unit 102 feature extraction unit
[0049] 103 Network delay prediction unit 104 Timing result calculation unit. DETAILED DESCRIPTION
[0050] It should be noted that the processor described in the present invention is the control center of the electronic device, which may be a processor or a general term for multiple processing elements. For example, it may be one or more central processing units (CPUs), or application specific integrated circuits (ASICs), or one or more integrated circuits configured to implement the embodiments of the present invention, such as one or more digital signal processors (DSPs), or one or more field programmable gate arrays (FPGAs).
[0051] Optionally, the processor can perform various functions of the electronic device by running or executing a software program stored in the memory, and calling data stored in the memory.
[0052] In a specific implementation, as an embodiment, the processor may include one or more CPUs. Each of these processors may be a single core processor or a multi-core processor. The processor here may refer to one or more devices, circuits, and / or processing cores for processing data (e.g., computer program instructions). Electronic devices may include: servers, desktop computers, laptops, smart phones, tablet computers, embedded computers, etc., wherein the embedded computers include vehicles and robots, etc.
[0053] The memory is used to store the software program for executing the solution of the present invention, and the execution is controlled by the processor. The specific implementation method can refer to the above method embodiment, which will not be repeated here.
[0054] It should be noted that the structure of the electronic device shown in the drawings of the present invention does not constitute a limitation thereto, and the actual knowledge structure recognition device may include more or fewer components than shown in the drawings, or a combination of certain components, or a different arrangement of components.
[0055] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.
[0056] It should also be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.
[0057] In the present invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can be represented by: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or multiple.
[0058] It should also be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order 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.
[0059] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
[0060] 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, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0061] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
[0062] 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 can be essentially or partly embodied in the form of a software product that contributes to the prior art. 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 perform all or part of the steps of the methods described in various embodiments 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.
[0063] In order to make the above features and effects of the present invention more clearly and understandably described, embodiments are given below and described in detail with reference to the accompanying drawings. This specification discloses one or more embodiments that include the features of the present invention. The disclosed embodiments are only for illustration. The scope of protection of the present invention is not limited to the disclosed embodiments, and the present invention is defined by the attached claims.
[0064] The following is a system embodiment corresponding to the above method embodiment. This embodiment can be implemented in conjunction with the above embodiment. The relevant technical details mentioned in the above embodiment are still valid in this embodiment. In order to reduce repetition, they are not repeated here. Accordingly, the relevant technical details mentioned in this embodiment can also be applied in the above embodiment.
[0065] The present invention aims to propose a method for performing static timing analysis before generalized wiring in multiple design modes and to accelerate the process of relaxation prediction.
[0066] Embodiment 1
[0067] like Figure 1 As shown, an embodiment of the present application provides a static timing analysis method before wiring, the method comprising:
[0068] Modeling step S101: The present application performs graph-based modeling on the integrated circuit before wiring, and outputs the network modeling result of the circuit; the integrated circuit before wiring is divided into a network (net). In some embodiments, Figure 2 As shown, the network consists of a driver cell U1 (including input pin(s) and driver pin), interconnection wires and sink pin(s);
[0069] Accordingly, the above modeling step S101 further includes:
[0070] Network partitioning step: divide the circuit into a directed acyclic graph, where each network is a node and each input pin of the driving unit is an edge;
[0071] The beneficial effect of the modeling method based on directed acyclic graphs in the present application is that the modeling method based on directed acyclic graphs proposed in the present application simplifies the problem by converting the complex relaxation prediction problem into a network delay prediction task. In each network, the network delay is the time it takes for a signal to propagate from the input pin of a driver unit to the receiving pin within the same network.
[0072] However, in the worst conditions, simply predicting the network delay and using it to calculate the arrival time may lead to inaccurate results due to unreasonable approximations.
[0073] Figure 3 An example of an unreasonable approximation in the prior art is given, in which the different input pins of the driving unit are not distinguished. In the network driven by unit U4, the network delays of input pins C and D of U4 are t CF and t DF The existing methods include unreasonable approximation methods that choose a larger value of t DF The delay of the network driven by U4. Due to the path delay t from FF4 to FF6 A +t AF Than the path delay t from FF5 (D Flip-Flop 5) to FF6 B +t BF Therefore, the critical path corresponding to the timing path endpoint F is the path from FF4 to FF6. Therefore, t is accumulated in the calculation of the arrival time corresponding to F. DF -t CF The detrimental effects of inappropriate approximations introduced by other operating conditions are similar to those in this example.
[0074] It can be seen that the methods of the prior art cannot distinguish the network delays under different working conditions, which include: different input pins, receiving pins, logical responses of network signals to input signals (including: positive responses or negative responses) and types of network signals (rising signals or falling signals). Among them, the logical response of the network signal to the input signal specifically refers to the way the network signal changes with the input signal. Each working condition needs to select one from each item; that is, the predicted network delay under a certain working condition refers to the predicted network delay under a certain input pin, a certain output pin, a certain network signal’s logical response to the input signal and a certain type of network signal.
[0075] In order to maintain accuracy and avoid unreasonable approximations, the present invention replaces the "network level" analysis in the prior art method with "pin level" analysis. Specifically, the present invention distinguishes between input pins, receiving pins, logical responses of network signals to input signals, types of input signals, and other conditions. The present invention predicts network delays under different conditions and uses the prediction results to calculate arrival times.
[0076] In summary, based on the modeling method proposed in the embodiment of the present application, the present invention formalizes the network delay prediction task as a machine learning regression problem. This problem can be further solved by machine learning regression algorithms such as distributed gradient boosting libraries. After completing the network delay prediction, the present invention uses a graph traversal algorithm to calculate the arrival time and relaxation of the circuit.
[0077] Furthermore, based on the completion of the modeling, the specific embodiment of the present invention continues to perform a customized feature extraction step.
[0078] Feature extraction step S102: extracting a network delay prediction feature set from the pre-routing design data, wherein the network delay prediction feature set includes: physical features and timing features, wherein the timing features include reference delay, reference slew and input slew.
[0079] The present invention designs a customized and concise feature set to enhance generalizability and parallelism. These customized features are extracted from the design and design report generated by the pre-layout design tool.
[0080] In a specific embodiment of the present invention, experiments were conducted using designs from the ITC'99 and OpenROAD benchmarks. Each circuit is constrained by 40 different clock cycles, representing different design modes. In each design, the lower limit of the clock cycle is approximately the minimum achievable clock cycle, while the upper limit is approximately the maximum achievable value. In order to distinguish the configurations of the verification set and the test set, the verification set allows a relaxation of up to 1 nanosecond (ns), and the test set allows a relaxation of up to 1 picosecond (ps). Other clock cycles are evenly divided into the range encompassed by the maximum and minimum values. Each clock cycle is rounded to two decimal places.
[0081] In a specific embodiment of the present invention, the circuit is designed and implemented using the Synaptics Design Compiler and the Synaptics Integrated Circuit Compiler and the TSMC N45GS standard cell library. The Synaptics Design Compiler performs logic synthesis and provides clock cycle constraints, and the Synaptics Integrated Circuit Compiler performs layout and routing. Based on this process, features and circuit connection relationships can be extracted from the pre-routing design and design report, and real values can be extracted from the post-routing design and design report.
[0082] The machine learning model in the present invention is implemented, trained and tested using Python 3.9 on a 16-core CPU, but the present invention is not limited thereto and other programming languages may also be used.
[0083] In the network delay prediction model, extracting a concise and suitable feature set is crucial to the effectiveness of the model. The adopted features and their sizes are summarized in Table 1, which can be extracted from the pre-routing design and design report generated by Synopsys IC Compiler.
[0084] Unlike machine learning regression problems that usually include hundreds of features, in the specific embodiment of the present invention, the feature set used in the present invention consists of only 11-dimensional features, making it a concise feature set. These features can be divided into physical features and timing features. Physical features include the capacitance of the driving pin, the total capacitance of the receiving pin, the Manhattan distance between the driving pin and the receiving pin, and the relative position of the related receiving pins (median and standard deviation); Figure 4 As shown in Figure 1, the difference between the input pin waves and the sink pin waves is the reference delay or net delay, and the timing characteristics include reference delay, reference slew, and input slew. The physical characteristics are easy to understand, and only the meaning and extraction process of the timing characteristics are described here.
[0085] Table 1 Feature list
[0086]
[0087] Reference delay refers to the reference value of network delay, which combines cell delay and line delay. Among them, cell delay includes cell information and reflects cell differences. Since the prior art method lacks cell-related features, its generalizability to multiple design patterns is limited. In order to solve this problem, the present invention extracts reference delay from the design report before wiring to improve the accuracy of network delay prediction under various cell types.
[0088] Input slew refers to the slew of the input pin of the driving unit, which significantly affects the unit delay. Since its value is approximately equal to the slew of the previous network, the existing technology method predicts the network slew sequentially along the timing path to obtain the input slew. However, this sequential prediction method seriously hinders parallelism and generalizability.
[0089] In order to solve this problem, the present invention extracts input slew from the pre-wiring design report, thereby effectively parallelizing the prediction of network delay and saving the time of predicting network slew. Reference slew refers to the reference value of network slew extracted from the pre-wiring design report in this work. However, the present invention is not limited to this, and other timing features can also be extracted.
[0090] Furthermore, in the specific embodiments of the present application, network delay prediction is based on the distributed gradient boosting library. The distributed gradient boosting library is a popular and efficient machine learning algorithm based on the gradient boosting framework, which is usually used for tasks such as regression, classification and ranking, and is well-known for its speed, efficiency and accuracy.
[0091] In order to improve the accuracy, the present invention implements a network delay prediction model based on a distributed gradient boosting library, and uses this model to predict network delay.
[0092] Network delay prediction step S103: Based on the network delay prediction feature set, a distributed gradient enhancement library is used to build a network delay prediction model, and the network delay prediction results under different working conditions are output;
[0093] As an optional implementation, under different working conditions, network delay prediction is performed based on a distributed gradient enhancement library, the data set of the enhancement library is the total number of working conditions of all networks, and a single data of the data set is one of the working conditions of each network, wherein the data is two-dimensional data, and the two-dimensional data includes: a feature set of the data and the true value of the network delay of the data; wherein the working conditions include: different input pins, receiving pins, logical responses of network signals to input signals, and types of network signals.
[0094] In the specific embodiment of the present application, in the network delay prediction based on the distributed gradient boosting library, the data set is D={(x i ,y i )}. Where |D| represents the total number of working conditions of all networks in the data set, and one condition of each network corresponds to one piece of data i; x i Represents the feature set of the i-th data, such as
[0095] As shown in Table 3; i It refers to the true value of the network delay of the i-th data.
[0096]
[0097] Where F is the space of regression trees in the distributed gradient boosting library. The goal of the training phase of the distributed gradient boosting library is to obtain:
[0098]
[0099] in is a differentiable convex function used to measure the error between the true value and the predicted network delay. The present invention uses the mean square error. θ represents a set of hyperparameter values of the distributed gradient enhancement library used, such as learning rate, number of trees (ie, K), maximum depth, sampling rate, minimum child node weight, etc.
[0100] Use the training set to train the distributed gradient boosting library model, and use the validation set to determine the values of the hyperparameters.
[0101] The beneficial effects of using the distributed gradient boosting library in the present invention are: the present invention adopts the distributed gradient boosting library, so that the present invention can predict network delay with high accuracy and high speed; the net delay prediction model based on the distributed gradient boosting library is superior to the random forest-based model in the prior art method in terms of accuracy, because the random forest emphasizes robustness and the tendency to exclude outliers.
[0102] Further, after determining the net delay, the specific embodiment of the present invention uses a graph traversal algorithm to calculate the arrival time and relaxation time. Since the present invention distinguishes multiple working conditions, the complexity of calculating the arrival time and relaxation may increase exponentially with the number of networks, so the present invention implements Figure 5 The algorithm calculates the worst result of the rising and falling arrival times (i.e., the arrival time when the signal type of the pin is rising or falling respectively) of each endpoint and is a recursive algorithm.
[0103] The specific embodiment of the present invention adopts a linear complexity algorithm to accelerate the calculation of arrival time and relaxation. Using this algorithm, the present invention can accurately calculate the arrival time and relaxation. Figure 5 As shown:
[0104] Timing result calculation step S104: Based on the network modeling results and the network delay prediction results under various working conditions, a linear complexity algorithm is used to calculate the arrival time and calculation relaxation, and the final pre-routing static timing analysis results are output.
[0105] As an optional implementation, the above-mentioned timing result calculation step S104 further includes:
[0106] Arrival time calculation step S1041: according to the network delay, the arrival time is calculated by using a graph traversal algorithm, and the arrival time of the network receiving pin is calculated simultaneously under different working conditions; after determining the rising arrival time and the falling arrival time of the end point of the timing path, the maximum value operation is used to calculate the final arrival time of the end point; after determining the rising arrival time and the falling arrival time of the end point of the timing path, the maximum value operation is used to calculate the final arrival time of the end point;
[0107] Slack time calculation step S1042: Based on the final arrival time, the slack is calculated using a specified clock cycle and timing information related to the historical database.
[0108] The above-mentioned arrival time calculation step S1041 further includes:
[0109] For the rise arrival time, first select the rise network delay under the positive response from the input pin, add the selected network delay to the rise arrival time at the input pin, and calculate the rise arrival time at the receiving pin under this set of working conditions. Calculate the rise network delay corresponding to the negative response under the input pin. Set the rise arrival time at the receiving pin to the maximum value of all calculated rise arrival times.
[0110] For the falling arrival time, first select the falling network delay under the positive response from the input pin, add the selected network delay to the falling arrival time at the input pin, and calculate the falling arrival time at the receiving pin under this set of working conditions. Calculate the falling network delay corresponding to the negative response under the input pin. Set the falling arrival time at the receiving pin to the maximum value of all calculated falling arrival times.
[0111] Since the calculation of the arrival time of a receiving pin only requires the network delay of the current network and the arrival time of the receiving pin of the previous network, only the rise and fall arrival time results of each network are retained in the specific embodiment of the present application. In addition, the network delay is affected by the network information and the slew of the previous network. In short, in addition to the information of the current network, the calculation of the arrival time of a receiving pin is only affected by the arrival time and slew of the receiving pin of the previous network.
[0112] Therefore, the beneficial effect of the method of the present application is that the method of the present application does not introduce any additional errors. Since the arrival time of the network receiving pin under different working conditions can be calculated simultaneously, the timing complexity of the algorithm is linearly related to the maximum depth of the circuit.
[0113] Embodiment 2
[0114] The embodiment of the present application provides a pre-wiring static timing analysis device, according to the pre-wiring static timing analysis method, such as Figure 6 As shown, the device comprises:
[0115] Modeling unit 101: used to perform graph-based modeling on the integrated circuit before wiring, divide the circuit into networks, and output the network modeling result of the circuit, wherein the network includes: a driving unit, interconnection lines and receiving pins;
[0116] Feature extraction unit 102: used to extract a network delay prediction feature set from pre-routing design data, wherein the prediction feature set includes: reference delay and input slew;
[0117] The network delay prediction unit 103 is used to construct a network delay prediction model according to the network delay prediction feature set and use a distributed gradient enhancement library to output network delay prediction results under different working conditions;
[0118] Timing result calculation unit 104: Based on the network modeling results and the network delay prediction results under various working conditions, a linear complexity algorithm is used to calculate the arrival time and the calculation relaxation, and the final static timing analysis result before wiring is output.
[0119] Embodiment 3
[0120] An embodiment of the present application provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of the above-mentioned pre-wiring static timing analysis method are implemented.
[0121] Embodiment 4
[0122] An embodiment of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of the pre-wiring static timing analysis method as described above are implemented.
[0123] In addition, combined Figure 1 The static timing analysis method before wiring described in the embodiment of the present application can be implemented by an electronic device, such as a computer device. Figure 7 Schematic diagram of the hardware structure of a computer device according to an embodiment of the present application.
[0124] In some of these embodiments, the computer device may further include a communication interface 83 and a bus 80. Figure 7 As shown, the processor 81, the memory 82, and the communication interface 83 are connected via a bus 80 and communicate with each other.
[0125] Specifically, the processor 81 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.
[0126] The memory 82 may be used to store or cache various data files required for processing and / or communication, as well as possible computer program instructions executed by the processor 81 .
[0127] The processor 81 implements any one of the pre-wiring static timing analysis methods in the above embodiments by reading and executing computer program instructions stored in the memory 82 .
[0128] In summary, compared with the prior art, the method of the present invention has the following beneficial effects:
[0129] (1) There are two main problems and shortcomings in the existing methods of pre-routing timing analysis. First, the low versatility across multiple design modes limits the accuracy of analyzing the timing of multi-mode design implementations of circuits. In order to observe the impact of this problem, the method was experimented using the ITC`99 and OpenROAD benchmarks. The results of network delay prediction are given in Table 2. Figure 8 Results for relaxed prediction are given in . For network delay prediction, the mean square error of this method is one order of magnitude larger than that of the Synopsys IC compiler method. In addition, this method only achieves a coefficient of determination of 0.663 on the ITC`99 benchmark involving multiple design patterns. In contrast, when restricted to a single design pattern, the coefficient of determination is 0.971 on the ITC`99 benchmark. For relaxed prediction, the performance of this method is also one order of magnitude worse than that of the Synopsys IC compiler across all implementations in terms of mean absolute error (MAE), where 20 circuit implementations are shown, involving different design patterns of two circuits (b17 and b22).
[0130] Table 2 Accuracy of the existing technical methods in predicting the offline delay in multi-mode design of circuits
[0131]
[0132] The main inherent reason for this problem is the lack of features related to the unit (specifically the cell in the integrated circuit), which hinders the method's ability to capture unit differences in different design modes, resulting in inaccurate unit delay results. Therefore, even if the training data is comprehensive enough, it cannot provide accurate results for network delay prediction, because unit delay is one of the key components of network delay.
[0133] Another important reason for the loss of accuracy is that when using the predicted network delay to calculate the arrival time, the methods in the prior art ignore the differences between network delays under different working conditions. However, in some implementations, different working conditions may lead to unacceptable errors, which may increase the inaccuracy of the relaxation prediction results.
[0134] (2) Secondly, the low parallelism of local delay prediction in the prior art limits the speed of relaxation prediction of timing paths involving multiple networks. The prior art method claims the negative impact of this limitation, which is 3 times slower than the pre-routing timing prediction of commercial tools. In order to observe the impact of this limitation, the speed of the Synopsys IC compiler and the impact of the prior art method on 15 circuits in the ITC`99 and OpenROAD benchmarks were also evaluated, and a separate implementation was selected for each circuit. The total relaxation estimation time of the prior art method and the Synopsys IC compiler in these 15 implementations is compared, and the results are summarized in Table 3. The table shows that the speed of the prior art method is more than 390 times slower than the relaxation estimation of the Synopsys IC compiler.
[0135] Table 3 Total runtime of methods in the prior art
[0136]
[0137] The main underlying reason for this issue is that the prediction of network latency and slew depends on the prediction results of the previous network. Specifically, input slew is a key feature of the random forest model used, which is predicted as the network slew in the previous network. Therefore, the prediction of network latency cannot be fully decoupled or parallelized, resulting in the network latency along each timing path being predicted sequentially. Another key factor affecting the speed is the time-consuming process of predicting input / network slew. Input / slew prediction takes the same time as network latency prediction, almost doubling the time required for relaxation prediction.
[0138] In summary, the present invention found the above problems in the prior art when conducting static timing analysis research: First, the low versatility across multiple design modes limits the accuracy of analyzing the timing of multi-mode design implementations of circuits. This defect is caused by the lack of unit-related features, and the accuracy is affected by ignoring the differences between network delays under different working conditions; second, the low parallelism of local delay prediction limits the speed of relaxation prediction of timing paths involving multiple networks. This defect is caused by the fact that the prediction of network delay and slew depends on the prediction results of the previous network, and the speed is affected by the time-consuming process of predicting input / network slew.
[0139] After performance evaluation, literature research, algorithm design, experimental verification and other research processes, the present invention found that solving these two defects can be achieved by reconstructing the feature extraction method of the machine learning regression model for local delay prediction.
[0140] Specifically, in order to enhance the generalizability across multiple design modes to achieve accurate static timing analysis, it is crucial to add one or more unit-related features (see Table 3 for details) to the network delay prediction model. In this way, the model can accurately capture unit information to predict network delays. In addition, distinguishing the working conditions of various networks helps to improve accuracy. In order to enhance the parallelism of network delay prediction by static timing analysis before fast wiring, the dependency chain of network delay prediction within the circuit must be systematically broken. The present invention is also based on predicting the network delay of the circuit, which can significantly reduce the time required for relaxation prediction. In addition, applying a simplified network slew generation process will also be helpful.
[0141] In summary, the present invention can consistently obtain static timing analysis results with high accuracy in multiple design modes, which shows that it has robust generalizability. Compared with the Synopsys IC compiler and the prior art method, the present invention achieves more than 12 times and 4800 times speedup respectively. In addition, the mean absolute error of the relaxation prediction of the present invention is more than one order of magnitude lower than that of the prior art method and exceeds the results of the Synopsys IC compiler.
[0142] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0143] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.
Claims
1. A static timing analysis method before wiring, characterized in that: The method comprises: Modeling step: modeling the integrated circuit before wiring based on a graph, dividing the circuit into a network, and outputting a network-level modeling result of the circuit, wherein the network includes: a driving unit, interconnection lines, and receiving pins; Feature extraction step: extracting a network delay prediction feature set from pre-routing circuit data, wherein the timing features of the prediction feature set include: reference delay and input slew; Network delay prediction step: based on the network delay prediction feature set, a distributed gradient enhancement library is used to build a network delay prediction model, and the network delay prediction results under different working conditions are output; Timing result calculation step: Based on the network modeling results and the network delay prediction results under various working conditions, a linear complexity algorithm is used to calculate the arrival time and calculation relaxation, and the final static timing analysis results before wiring are output.
2. The static timing analysis method before wiring according to claim 1, characterized in that: The modeling step further includes: Network division step: divide the circuit into a directed acyclic graph, wherein the driving unit includes an input pin, the node of the directed acyclic graph is each of the networks, and the edge of the directed acyclic graph is each input pin of the driving unit and the receiving pin of the network.
3. The static timing analysis method before wiring according to claim 1, characterized in that: In the feature extraction step, the network delay prediction feature set includes: physical features and timing features, wherein the timing features include reference delay, reference slew and input slew.
4. The static timing analysis method before wiring according to claim 1, characterized in that: The network delay prediction step further includes: Under different working conditions, network delay prediction is performed based on a distributed gradient enhancement library, the data set size of the enhancement library is the total number of working conditions of all the networks, and a single data of the data set is one of the working conditions of each network, wherein the data is two-dimensional data, and the two-dimensional data includes: a feature set of the data and a true value of the network delay of the data; The working conditions include: different input pins, receiving pins, logical responses of network signals to input signals, and types of network signals.
5. The static timing analysis method before wiring according to claim 1, characterized in that: The timing result calculation step further includes: Arrival time calculation steps: according to the network delay, the arrival time is calculated using a graph traversal algorithm, and the arrival time of the network receiving pin is calculated simultaneously under different working conditions; after determining the rising arrival time and falling arrival time of the end point of the timing path, the maximum value operation is used to calculate the final arrival time of the end point.
6. The static timing analysis method before wiring according to claim 5, characterized in that: The static timing parameter calculation step further includes: Slack time calculation step: Based on the final arrival time, slack is calculated using a specified clock cycle and timing information associated with a historical database.
7. The static timing analysis method before wiring according to claim 5, characterized in that: The arrival time calculation step further includes: Rise arrival time calculation step: for the calculation of the rise arrival time, select the rise network delay under the positive response from the input pin, add the selected network delay to the rise arrival time at the input pin, and calculate the rise arrival time at the receiving pin under a set of working conditions; calculate the rise network delay under the negative response of the input pin, and set the rise arrival time at the receiving pin to the maximum value of all calculated rise arrival times; Fall arrival time calculation steps: for the calculation of the fall arrival time, select the fall network delay under the positive response from the input pin, add the selected network delay to the fall arrival time at the input pin, and calculate the fall arrival time at the receiving pin under a set of working conditions; calculate the fall network delay under the negative response of the input pin, and set the fall arrival time at the receiving pin to the maximum value of all calculated fall arrival times.
8. A pre-routing static timing analysis device, according to the pre-routing static timing analysis method according to any one of claims 1 to 7, characterized in that: The device comprises: Modeling unit: used for modeling the integrated circuit before wiring based on graph, dividing the circuit into networks, and outputting the network-level modeling result of the circuit, wherein the network includes: a driving unit, interconnection lines and receiving pins; A feature extraction unit: used to extract a network delay prediction feature set from pre-routing design data, wherein the timing features of the prediction feature set include: reference delay and input slew; A network delay prediction unit is used to construct a network delay prediction model based on the network delay prediction feature set and adopt a distributed gradient enhancement library to output network delay prediction results under different working conditions; Timing result calculation unit: based on the network modeling results and the network delay prediction results under various working conditions, a linear complexity algorithm is used to calculate the arrival time and calculation relaxation, and the final static timing analysis result before wiring is output.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the pre-wiring static timing analysis method described in any one of claims 1 to 7 are implemented.
10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the pre-wiring static timing analysis method according to any one of claims 1 to 7 are implemented.
