Time sequence convergence method and device based on clock tree
By using a clock tree-based timing convergence method, the timing path of a one-to-many structure is optimized in a targeted manner, which solves the problem of increased power consumption caused by Hold timing violations in large-scale integrated circuit design, achieves higher precision clock tree adjustment, and reduces power consumption and design cost.
Patent Information
- Application Number
- CN202610032629.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-12
- Publication Date
- 2026-02-10
- Estimated Expiration
- 2046-01-12
AI Technical Summary
In large-scale integrated circuit design, existing technologies cannot effectively reduce the power consumption increase caused by fixing Hold timing violations.
By using a clock tree-based timing convergence method, timing paths in one-to-many structures are specifically detected, and automatic layout and timing margin optimization are performed. The setting pattern for useful clock skew adjustment is determined, and the clock tree is optimized and adjusted through positive or negative feedback to generate a clock tree that meets the predetermined requirements.
It achieves higher precision clock tree adjustment, reduces Hold timing violations, suppresses the increase in cell density, reduces design size and power consumption, shortens the design cycle, and reduces design costs.
Smart Images

Figure CN121503414A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of circuit design, and particularly relates to a clock tree based timing convergence method and device. BACKGROUND
[0002] With the increasingly mature large-scale integrated circuit design, power consumption gradually becomes the focus. In the design process, in order to repair the Hold timing violation, the automatic layout and routing tool needs to insert buffers and delays, which leads to the increase of the design size, and the larger the size is, the greater the power consumption is. Moreover, it also leads to the higher local cell density, which increases the difficulty of routing and produces detour routing. In order to repair the design rule violation caused by the detour routing, more buffers need to be inserted, which further increases the power consumption. Therefore, how to reduce the power consumption increase caused by repairing the Hold timing violation has become one of the problems that the low-power design needs to face. SUMMARY
[0003] The present application provides a clock tree based timing convergence method and device to solve the problem of power consumption increase caused by repairing the Hold timing violation.
[0004] The present application is implemented by the following technical solutions.
[0005] In a first aspect, a clock tree based timing convergence method is provided, which comprises: automatically laying out a device; determining a timing path that needs to be optimized in a one-to-many structure timing path of the automatically laid out device; performing setup time slack optimization on the timing path that needs to be optimized, and determining a common adjustment amount based on the optimized setup time slack as a setting pattern of useful clock skew adjustment; determining a buffer number based on the setting pattern of useful clock skew adjustment and the optimized setup time slack, and generating a clock tree; determining whether a Hold timing violation of the timing path that needs to be optimized meets a predetermined requirement based on the generated clock tree, and in the case that the predetermined requirement is not met, adjusting the setting pattern of useful clock skew adjustment based on positive feedback optimization or negative feedback optimization, and returning to the step of generating the clock tree until the predetermined requirement is met; and completing automatic routing based on the clock tree that meets the predetermined requirement.
[0006] In some embodiments, determining the timing path to be optimized in the one-to-many structure timing path of the automatically placed device includes: obtaining the setup time timing of all timing paths, and extracting the start trigger, end trigger, and setup time timing margin of each timing path in all timing paths; traversing all start triggers in all timing paths and recording the occurrence frequency of each start trigger to filter out timing paths with an occurrence frequency greater than a first threshold; extracting timing paths containing the same start trigger from the filtered timing paths, and after deduplicating the end triggers, filtering out timing paths with the first threshold number; and using the timing paths with the first threshold number as the timing paths to be optimized.
[0007] In some embodiments, optimizing the setup time timing margin of the timing path to be optimized includes: obtaining setup time timing margin from the timing path to be optimized, and filtering out timing paths to be optimized based on the setup time timing margin; clustering all devices on the timing path to be optimized and deleting redundant buffers; obtaining the setup time timing of the optimized timing path, and replacing the original devices on the optimized timing path with devices less than the threshold if the setup time timing margin of the optimized timing path is less than a second threshold.
[0008] In some embodiments, determining the common adjustment amount based on the optimized setup time timing margin as a setting pattern for the useful clock skew adjustment includes: setting the adjustment amount for the transmit clock and the timing path where the transmit clock is located based on the optimized setup time timing margin; sorting all the adjustment amounts for the timing path where the same transmit clock is located, taking the smallest adjustment amount among all the adjustment amounts as the common adjustment amount, and outputting it as a setting pattern for the useful clock skew adjustment; and inserting a buffer at the end of the timing path where the transmit clock is located based on the setting pattern for the useful clock skew adjustment.
[0009] In some embodiments, based on the setting pattern of the useful clock skew adjustment, inserting a buffer at the end of the timing path where the transmit clock is located includes: obtaining the delay value of the clock line dedicated buffer as the minimum unit value, calculating the number of buffers to be inserted, and adding constraints to the inserted buffers and the connections between them.
[0010] In some embodiments, based on the generated clock tree, it is determined whether the hold-time timing violation of the timing path to be optimized meets a predetermined requirement, and if the expected requirement is not met, the setting pattern of the useful clock skew adjustment is adjusted based on positive feedback optimization or negative feedback optimization. This includes: determining whether the hold-time timing violation in the hold-time timing of the timing path to be optimized in the clock tree meets the expected target; if the hold-time timing violation meets the expected target, based on the hold-time timing violation of each timing path in the one-to-many structure timing path, determining whether to use positive feedback optimization or negative feedback optimization to adjust the delay of the timing path to obtain the adjustment amount of the timing path, wherein the positive feedback optimization is for those with the hold-time timing violation, and the negative feedback optimization is for those without the hold-time timing violation; sorting all the adjustment amounts of the timing path where the same transmit clock is located, taking the smallest adjustment amount among all the adjustment amounts as the useful clock skew adjustment amount, and outputting it as the setting pattern of the useful clock skew adjustment; and returning to the step of generating the clock tree based on the setting pattern of the useful clock skew adjustment.
[0011] In some embodiments, both the positive feedback optimization and the negative feedback optimization include optimization under worst-case operating conditions and optimization under best-case operating conditions.
[0012] In some embodiments, the optimization under the worst-case operating conditions includes: for the positive feedback optimization, increasing the setting pattern of the useful clock skew adjustment based on whether the setup time timing is violated, the setup time timing margin, the amount of hold time timing violation, and a predetermined coefficient; and for the negative feedback optimization, decreasing the setting pattern of the useful clock skew adjustment based on whether the setup time timing is violated, the hold time timing margin, and a predetermined coefficient.
[0013] In some embodiments, the method, optimized under the best operating conditions, includes: for the positive feedback optimization, increasing the setting pattern of the useful clock skew adjustment based on whether the setup time timing is violated, the setup time timing margin, the amount of hold time timing violation, and a predetermined coefficient; and for the negative feedback optimization, decreasing the setting pattern of the useful clock skew adjustment based on whether the setup time timing is violated and the hold time timing margin.
[0014] Secondly, a clock tree-based timing convergence device is provided, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method described in any of the above.
[0015] Thirdly, a clock tree-based timing convergence device is provided, comprising: an automatic placement unit, a detection unit, a processing unit, a clock tree synthesis unit, a feedback unit, and an automatic routing unit. The automatic placement unit performs automatic placement of components. The detection unit determines the timing paths requiring optimization in the one-to-many structure timing paths of the automatically placed circuit elements. The processing unit optimizes the setup time timing margin of the timing paths requiring optimization and determines a common adjustment amount based on the optimized setup time timing margin, serving as a setting pattern for useful clock skew adjustment. The clock tree synthesis unit determines the number of buffers and generates a clock tree based on the setting pattern for useful clock skew adjustment and the setup time timing margin. The feedback unit determines, based on the generated clock tree, whether the hold time timing violation of the timing paths requiring optimization meets predetermined requirements. If the predetermined requirements are not met, it adjusts the setting pattern for useful clock skew adjustment based on positive or negative feedback optimization and returns to the clock tree generation step until the predetermined requirements are met. An automatic routing unit is used to complete automatic routing based on a clock tree that meets predetermined requirements.
[0016] Fourthly, a computer-readable storage medium is provided that stores computer program instructions, which, when executed on a computer, cause the computer to perform any of the methods described above.
[0017] Fifthly, a computer program is provided that, when run on a computer, causes the computer to perform any of the methods described above.
[0018] In a sixth aspect, a computer program product containing instructions is provided, which, when run on a computer, causes the computer to perform any of the methods described above.
[0019] Compared with the prior art, the present invention has the following advantages and beneficial effects.
[0020] 1) By employing automatic detection, automatic processing, and automatic feedback, and intelligent optimization, it efficiently identifies one-to-many data transmission structures in the design. It uses a multi-path analysis method to achieve higher precision in adjusting the clock tree, significantly reducing hold timing violations, thereby suppressing cell increment rate, reducing design size, and lowering power consumption.
[0021] 2) The feedback adjustment function can reduce the number of design iterations required for the design data to reach the timing convergence state, shorten the design cycle, avoid excessive consumption of manpower and material resources, and reduce design costs. Attached Figure Description
[0022] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as limiting the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.
[0023] Figure 1 This is a schematic diagram of the basic structure of data transmission.
[0024] Figure 2 This is a schematic diagram of a one-to-many data transmission structure.
[0025] Figure 3 A schematic diagram illustrating the effect of timing margin on useful clock skew.
[0026] Figure 4 This is a diagram illustrating the strategy adjustment for path-by-path analysis.
[0027] Figure 5 This is a diagram illustrating the strategy adjustment for multi-path analysis.
[0028] Figure 6 This is a flowchart of a timing convergence method based on a clock tree according to an embodiment of the present invention.
[0029] Figure 7 This is a flowchart illustrating how to select timing paths that need optimization according to an embodiment of the present invention.
[0030] Figure 8 This is a flowchart for optimizing setup time timing margin according to an embodiment of the present invention.
[0031] Figure 9 A flowchart illustrating the setting pattern for determining useful clock skew adjustment according to an embodiment of the present invention.
[0032] Figure 10 A flowchart illustrating the setting pattern for adjusting the useful clock skew adjustment according to an embodiment of the present invention.
[0033] Figure 11 This is a structural block diagram of a timing convergence device based on a clock tree according to an embodiment of the present invention.
[0034] Figure 12 This is a schematic diagram of a timing convergence device based on a clock tree according to an embodiment of the present invention. Detailed Implementation
[0035] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments and accompanying drawings. The illustrative embodiments and descriptions of this invention are for illustrative purposes only and are not intended to limit the invention.
[0036] In digital circuits, data signal transmission relies on flip-flop structures. When designing digital circuits, setup / hold timing constraints must be met. The following combines... Figure 1 This section explains some of the terms involved in Setup / Hold timing constraints.
[0037] LaunchCLK (Transmit Clock): The clock signal that originates from the clock root node and reaches the CKPin (clock pin) of the transmit flip-flop (FF). Once this clock signal is valid, the transmit flip-flop FF1 begins transmitting data.
[0038] CaptureCLK (Receive Clock): The clock that starts from the clock root node and reaches the CK pin of the receive flip-flop. Once this clock is valid, the receive flip-flop FF2 will start receiving data.
[0039] Setup (Setup Time) Timing Constraint: Data must remain stable for a certain period of time (T) before the clock signal of the receive flip-flop FF2 becomes valid. Setup This is necessary for the receive trigger to read the correct data. Setup timing slack is the difference between the required delay time of the timing path and the actual delay time; its formula is Setup Slack = T 周期 +T CaptureCLK -(T) LaunchCLK +T Data +T Setup When Setup Slack is negative, it is called a Setup timing violation. To reduce the computational load during timing analysis, worst-case operating conditions are generally used (data path delay T under the worst operating conditions such as process deviations, voltage, and temperature). Data (Maximum) to verify whether the design meets the Setup timing constraints.
[0040] Hold (hold time) timing constraint: After the clock signal of the receive flip-flop FF2 becomes valid, the data must remain stable for a certain period of time (T). Hold This is necessary for the receive trigger to read the correct data. Hold timing margin (Slack) refers to the difference between the actual delay time of the timing path and the required delay time. The formula for calculating Hold timing constraint Slack is Hold Slack = T LaunchCLK +T Data -(T) CaptureCLK +THold When Hold Slack is negative, it is called a Hold timing violation. To reduce the computational load during timing analysis, optimal operating conditions are generally used (under optimal process deviations, voltage, temperature, etc., with the data path delay T...). Data The minimum value is used to verify whether the design meets the Hold timing constraints.
[0041] Skew (clock skew): The delay difference between the transmit clock and the receive clock starting from the same clock root node is called clock skew. Skew has always been an important parameter for evaluating clock tree performance; ideally, there is no clock delay, i.e., the skew is zero. The purpose of traditional clock tree synthesis is to reduce skew.
[0042] Useful Skew: When used appropriately, increasing the skew can actually reduce the difficulty of timing convergence; this is called Useful Skew. From Hold Slack = T LaunchCLK +T Data -(T) CaptureCLK +T Hold As can be seen, this can be achieved by reducing T. CaptureCLK Or increase T LaunchCLK The method increases Hold Slack, which is beneficial for Hold timing convergence. However, the calculation formula Setup Slack=T 周期 +T CaptureCLK -(T) LaunchCLK +T Data +T Setup It can be seen that decreasing T CaptureCLK Or increase T LaunchCLK This will reduce Setup Slack. Therefore, adjusting UsefulSkew to favor Hold timing convergence is detrimental to Setup timing convergence. Similarly, adjusting UsefulSkew to favor Setup timing convergence is detrimental to Hold timing convergence.
[0043] The calculation formula for Hold Slack shows that Hold timing violations are mainly caused by Skew, T CaptureCLK >T LaunchCLKFurthermore, the larger the Skew difference, the greater the Hold violation, requiring more buffers to be inserted to fix the violation, resulting in higher power consumption. In traditional design flows, during automatic placement, the clock is an ideal clock (without clock delay), and Hold timing analysis is not performed. Clock tree synthesis also cannot optimize the clock tree based on the Hold timing results, leading to numerous Hold timing violations after the clock tree is generated. Because the tool fixes violations path-by-path, traversing each path and fixing the violation results in excessive buffers and delays being inserted into the data signals. Related techniques suggest adjusting the clock line delay using the UsefulSkew method after clock tree synthesis to reduce clock skew (Tskew). CaptureCLK <T LaunchCLK This is beneficial for Hold timing convergence, thereby reducing Hold violations.
[0044] As integration levels increase and circuit structures become more complex, such as Figure 2 As shown, one-to-many data transmission structures are widely used, characterized by transmitting data from a single transmitting front-end (FF) to a large number of different receiving front-ends (FFs). This data transmission path is more prone to hold timing violations. Due to the large number of data paths, the proportion of violations in these paths is also relatively large. Furthermore, the clock line of the transmitting FF is the common clock path for many data paths; the delay of this clock line can be adjusted using UsefulSkew, thereby significantly reducing hold violations in subsequent data paths and resulting in a significant improvement in power consumption. Therefore, this type of structure requires special attention from designers. However, due to limitations in related technologies, such as the inability to specifically address one-to-many structures, limited adjustment range, and low precision of adjustment, these technologies cannot effectively improve the timing violations of this type of structure.
[0045] Therefore, to address the above problems, this invention proposes a timing convergence method and apparatus based on clock trees, which can specifically detect one-to-many structures and improve hold violations, suppress cell increment rate, reduce design size, and achieve the effect of reducing power consumption by automatically adjusting the delay on the common clock path.
[0046] First, the timing convergence principle of this invention will be explained.
[0047] In related technologies, when the setup margin is small, the adjustment range of the clock line is also very small. For example... Figure 3As shown, when using UsefulSkew for adjustments, it is essential to consider the available setup timing margin for adjustments. In this invention, when there is no setup margin, the setup margin can be increased by optimizing the layout of data transmission paths within the system, removing redundant components, and finely controlling device thresholds, thereby reducing Tdata and ensuring that the clock line has more adjustment possibilities.
[0048] Traditional backend design processes and related technologies are based on path-by-path timing analysis and optimization strategies, lacking a specific one-to-many concept. Therefore, their optimization effects on timing paths in one-to-many structures are not significant. For example... Figure 4 As shown, before clock tree adjustments, the setup margins for paths A / B / C are 0.8ns / 0.4ns / 0.2ns. When adjusting path A, with a margin of 0.8ns, clock line adjustments allow for an increase of 4 buffers in latency, resulting in setup margins of 0ns / -0.4ns / -0.6ns for paths A / B / C. When adjusting path B, its setup margin is -0.4ns. To prevent timing violations, clock line adjustments reduce the latency by 2 buffers, resulting in setup margins of 0.4ns / 0ns / -0.2ns for paths A / B / C. When adjusting path C, its setup margin is -0.2ns. To prevent timing violations, clock line adjustments further reduce the latency by 1 buffer, resulting in setup margins of 0.6ns / 0.2ns / 0ns for paths A / B / C. After paths A, B, and C are adjusted sequentially, the final adjustment on the clock line is only a delay of one buffer. Because T launchCLK The increase is very small, so the improvement on Hold violations on paths A and B is also minimal.
[0049] In this invention, timing paths in one-to-many structures can be detected in advance, and the timing paths that need to be processed can be selected. By processing multiple paths within the system, a joint analysis and adjustment strategy can be implemented, and targeted overall adjustments can be made on common clock paths to significantly reduce hold violations. Figure 5 As shown, paths A, B, and C are analyzed simultaneously, and the path with the maximum margin of 0.8ns is taken as the target adjustment amount. Clock line adjustment allows for an increase in delay of 4 buffers. After adjustment, the setup margins of paths A / B / C become 0ns / -0.4ns / -0.6ns, respectively. Since T launchCLKThe increase is significant, so it greatly improves the hold violation on paths A / B / C. Although paths B / C are currently in a setup timing violation state, the violation can be corrected through setup margin optimization.
[0050] Furthermore, related technologies adjust UsefulSkew based on an existing clock tree, which has limited adjustment range. Excessive adjustment can negatively impact the quality of the existing clock tree, leading to poor overall timing results and even non-convergence. This invention addresses this by implementing UsefulSkew before clock tree synthesis for one-to-many clock paths, maximizing its effectiveness in facilitating Hold timing convergence.
[0051] Meanwhile, the corresponding technology is a one-time adjustment, which may result in insufficient or excessive adjustment. In this invention, the UsefulSkew adjustment amount is iteratively optimized through an internal feedback system, which can achieve higher precision adjustment and optimization of the clock tree. While ensuring timing convergence, it can minimize the amount of Hold violations, thereby reducing the increase in power consumption caused by repairing Hold timing violations.
[0052] like Figure 6 The diagram shown is a flowchart of a timing convergence method based on a clock tree according to an embodiment of the present invention. The following references... Figure 6 The timing convergence method based on clock tree of the present invention will be described in detail.
[0053] refer to Figure 6 The timing convergence method based on clock trees includes the following steps S10 to S60. Steps S10 to S60 are described in detail below with reference to the accompanying drawings.
[0054] In S10, the devices are automatically placed. For example, based on the designer's initial layout plan, the guide tool automatically places the devices (standard cells).
[0055] In S20, the timing path that needs to be optimized in the one-to-many structure timing path of the automatically placed device is determined.
[0056] In S30, the setup time timing margin is optimized for the timing path that needs to be optimized, and a common adjustment amount is determined based on the optimized setup time timing margin as the setting pattern for useful clock skew adjustment.
[0057] In S40, the number of buffers is determined and a clock tree is generated based on the setting pattern of useful clock skew adjustment and optimized setup time timing margin.
[0058] In S50, based on the generated clock tree, it is determined whether the hold time timing violation of the timing path to be optimized meets the predetermined requirements. If the predetermined requirements are not met, the setting pattern of useful clock skew adjustment is adjusted based on positive feedback optimization or negative feedback optimization, and the clock tree generation step is returned until the predetermined requirements are met.
[0059] In S60, automatic routing is completed based on a clock tree that meets predetermined requirements.
[0060] like Figure 7 The diagram shown is a flowchart for selecting timing paths that need optimization according to an embodiment of the present invention. (Reference) Figure 7 In S20, the timing paths that need to be optimized are selected, including S21 to S24.
[0061] In S21, the setup time timings of all timing paths are obtained, and the start trigger, end trigger, and setup time timing margin of each timing path are extracted. For example, the Setup timing results of all paths are obtained, and the start FF, end FF, and Slack results of each path are extracted and output as an array list.
[0062] In S22, all starting triggers in all timing paths are traversed and the frequency of occurrence of each starting trigger is recorded to filter out timing paths whose frequency of occurrence is greater than a first threshold. For example, all starting triggers FF are traversed, the frequency of occurrence of each starting trigger FF is recorded, and timing paths whose frequency of occurrence is greater than the threshold N (N is a dynamic threshold, which is calculated as the highest frequency × 30%, and the percentage can be adjusted according to the computational load) are filtered out.
[0063] In S23, timing paths containing the same start trigger are extracted from the selected timing paths. After deduplicating the end triggers, timing paths with a first threshold number of occurrences are selected. For example, paths containing the same start trigger (FF) are extracted from the selected timing paths and output to different array lists. Each list deduplicates the end triggers and counts the number of occurrences, selecting timing paths with a number greater than the threshold N (the first threshold).
[0064] In S24, the time-series path of the number of the first threshold is taken as the time-series path to be optimized. For example, the time-series path of the number of the first threshold finally selected above is taken as the optimization object of the one to many structure and output.
[0065] like Figure 8 The diagram shown is a flowchart of setup time timing margin optimization according to an embodiment of the present invention. (Reference) Figure 8 In S30, the establishment time timing margin is optimized for the timing path that needs to be optimized, including S31 to S33.
[0066] In step S31, setup time timing margins are obtained from the selected timing paths requiring optimization, and timing paths requiring layout optimization are further selected based on these margins. For example, in a one-to-many structure, timing paths requiring layout optimization are selected based on the size of the setup timing margin. Timing paths with setup timing margins below 500ps (a value recommended to be based on the latency of 5 buffers, ensuring that the LaunchCLK adjustable amount is as high as possible above 5 buffers; this can be adjusted depending on the process technology; 500ps is recommended for 40nm process technology, and 300ps is recommended for 22nm process technology) are considered as targets for layout optimization.
[0067] In S32, all devices on the timing path requiring layout optimization are clustered together, and redundant buffers are removed. For example, for the selected path, all devices on the path are clustered together first (centralized within a radius of 100µm centered on the path midpoint), and then redundant buffer devices are removed.
[0068] In step S33, the setup time timing of the optimized timing path is obtained. If the setup time margin of the optimized timing path is less than a second threshold, the original components on the optimized timing path are replaced with components with a margin less than the threshold. For example, the setup timing results of the above path are obtained again. If the setup timing margin is still less than 500ps, the components on the timing path are replaced with components with a lower threshold.
[0069] like Figure 9 The diagram shows a flowchart of determining the setting pattern for useful clock skew adjustment according to an embodiment of the present invention. In S30, a common adjustment amount is determined based on the optimized setup time timing margin as the setting pattern for useful clock skew adjustment, including S34 to S36.
[0070] In S34, based on the optimized setup time timing margin, the transmit clock and the adjustment amount of the timing path containing the transmit clock are set. For example, the setup timing result of the timing path is obtained, and the adjustment amount (delay) of the LaunchCLK required by the clock line processing system and the timing path containing the LaunchCLK is set according to the setup timing margin. If the setup timing margin is ≥ 1ns, the delay is set to 1ns; if 0ns < setup timing margin < 1ns, the delay is set to the setup timing margin; if the setup margin is ≤ 0ns, the delay is set to 0ns, in which case the clock tree cannot be adjusted.
[0071] In S35, all adjustment values on the timing path containing the same transmit clock are sorted, and the smallest adjustment value among all adjustment values is taken as the common adjustment value and output as the setting pattern for useful clock skew adjustment. For example, all adjustment values for the same LaunchCLK are sorted, the minimum value is taken as the common adjustment value, and the result is output as a list of setting patterns for UsefulSkew adjustment.
[0072] In S36, based on the useful clock skew adjustment configuration pattern, a buffer is inserted at the end of the timing path where the transmit clock is located. For example, the UsefulSkew adjustment configuration pattern list is read in, and a buffer of the specified delay amount is inserted at the end of the LaunchCLK that needs adjustment. If the delay amount is 0ns, no processing is performed.
[0073] In some embodiments, based on the useful clock skew setting pattern, inserting buffers at the end of the timing path where the transmit clock is located includes: obtaining the delay value of the clock line-specific buffer as the minimum unit value, calculating the number of buffers to be inserted, and adding constraints to the inserted buffers. For example, firstly, the delay value of the clock line-specific buffer (taking 1x drive capability, suppressing the number of insertions) needs to be obtained from the library file of the standard cell. This delay is then used as the minimum unit value to calculate the number of buffers to be inserted in the timing path where LaunchCLK is located. This number of clock line-specific buffers is inserted at the end of the LaunchCLK path, and constraints prohibiting touching these buffers are added to prevent them from being deleted during subsequent optimization. Furthermore, constraints prohibiting touching are also added to the connections between these buffers to prevent the insertion of additional buffers with large multiples between them during subsequent optimization, which would violate transition constraints and worsen clock line delay.
[0074] In S40, the data optimized by S30 is synthesized into a clock tree using the traditional process. During clock tree synthesis, the overall clock tree quality is optimized under both worst-case and best-case operating conditions.
[0075] like Figure 10 The diagram shown is a flowchart illustrating the setting pattern for adjusting the useful clock skew adjustment according to an embodiment of the present invention. (Reference) Figure 10 In S50, the setting pattern for useful clock skew adjustment is adjusted, including S51 to S54.
[0076] In S51, it is determined whether the hold time timing violations in the timing path to be optimized in the clock tree have achieved the expected target. For example, the timing results of the adjusted path are obtained to confirm whether the hold timing violations have been reduced to the expected result. If the expected result is achieved, proceed to S40; otherwise, revert to S52. The expected result can be defined as: the total hold violation amount of the adjusted path is below a threshold T (recommended value 10ns; for a design scale of 1 million gates, fixing a 10ns hold violation will result in an increase of approximately 1% in cell size), which can be adjusted in conjunction with the design scale. Furthermore, as the design scale increases, the computation time of each feedback analysis by the feedback unit also increases; the maximum number of feedback iterations C can be set according to the computation time that can be tolerated within the design cycle.
[0077] In S52, when the hold-time timing violation meets the expected target, based on the hold-time timing violation of each timing path in the one-to-many structure, it determines whether to use positive feedback optimization or negative feedback optimization to adjust the delay of the timing path to obtain the adjustment amount. Positive feedback optimization is used for timing paths with hold-time violations, while negative feedback optimization is used for timing paths without hold-time violations. For example, based on the hold-time timing violation of each timing path in the one-to-many structure, it is analyzed whether to use positive feedback optimization or negative feedback optimization. If the timing path has hold-time violations, positive feedback optimization is used; otherwise, negative feedback optimization is used. In addition, the feedback optimization corresponds to the two analysis conditions (worst-case operating condition and best-case operating condition) in clock tree optimization, adapting to the positive and negative feedback optimization schemes under different analysis modes.
[0078] In some embodiments, optimization under worst-case operating conditions includes: for positive feedback optimization, adding a setting pattern for useful clock skew adjustment based on whether setup time timing is violated, setup time timing margin, hold time timing violation amount, and a predetermined coefficient; and for negative feedback optimization, determining a setting pattern for useful clock skew adjustment based on whether setup time timing is violated, hold time timing margin, and a predetermined coefficient.
[0079] For example, the positive feedback optimization scheme is as follows: If there is no violation of the Setup timing, and if the Setup timing margin is greater than or equal to the Hold timing violation amount × coefficient K (where coefficient K is the difference in delay multiples between the slowest Setup analysis and the fastest Hold analysis for the same device), then the delay increase is set to be equal to the Hold violation amount × coefficient K; if the Setup timing margin is less than the Hold timing violation amount × coefficient K, then the delay increase is set to be equal to the Setup margin; if there is a violation of the Setup timing, then the delay amount is not adjusted, and the clock remains in an unadjustable state.
[0080] For example, the negative feedback optimization scheme is as follows: if there is a violation of the Setup timing, the delay reduction amount is set to be equal to the Hold timing margin multiplied by the coefficient K; if there is no violation of the Setup timing, if the delay amount set before this feedback is 0ns, the delay amount is not adjusted and the clock remains in an unadjustable state; if the delay amount set before this feedback is not equal to 0ns, the delay reduction amount is set to be equal to the Hold timing margin multiplied by the coefficient K.
[0081] In some embodiments, optimization under optimal operating conditions includes: for positive feedback optimization, increasing the setting pattern for useful clock skew adjustment based on whether setup time timing is violated, setup time timing margin, hold time timing violation amount, and a predetermined coefficient; and for negative feedback optimization, decreasing the setting pattern for useful clock skew adjustment based on whether setup time timing is violated and hold time timing margin.
[0082] For example, the positive feedback optimization scheme is as follows: if there is no violation of the Setup timing, and if the Setup timing margin is greater than or equal to the Hold timing violation amount × coefficient K, then the increase in delay is set to be equal to the Hold violation amount; if the Setup timing margin is less than the Hold timing violation amount × coefficient K, then the increase in delay is set to be equal to the Setup timing margin ÷ coefficient K; if there is a violation of the Setup timing, then the delay is not adjusted, and the clock remains in an unadjustable state.
[0083] For example, the negative feedback optimization scheme is as follows: if there is no violation of the Setup timing, the amount of delay reduction is set to equal the Hold timing margin; if there is a violation of the Setup timing; if the delay amount set before this feedback is 0ns, the delay amount is not adjusted and the clock remains in an unadjustable state; if the delay amount set before this feedback is ≠ 0ns, the amount of delay reduction is set to equal the Hold margin.
[0084] In S53, all adjustment values on the timing path containing the same transmit clock are sorted, and the smallest adjustment value among all adjustment values is taken as the useful clock skew adjustment value, and output as the setting pattern for useful clock skew adjustment. For example, all adjustment values for the same LaunchCLK are sorted, the minimum value is taken as the common adjustment value, and output as a list of setting patterns for UsefulSkew adjustment.
[0085] In S54, based on the configuration pattern for the useful clock skew adjustment, the step of generating the clock tree is returned. For example, the adjusted configuration pattern list is read in, and the number of buffers already inserted in LaunchCLK is adjusted according to the configuration pattern list. After adjustment, the process returns to S30 for clock tree synthesis. S50 is iteratively executed until the result achieves the expected effect or the worst-case iteration count C is completed, at which point the feedback is exited.
[0086] In S60, the design data that has undergone the above steps and completed clock tree adjustment will be automatically routed.
[0087] Compared to the path-by-path method used in current technology for data optimization, this invention proposes an intelligent optimization approach that combines automatic detection, automatic processing, and automatic feedback to efficiently identify one-to-many data transmission structures in the design. By using a multi-path joint analysis method, the clock tree can be adjusted with higher precision, significantly reducing hold timing violations, thereby suppressing cell increment rate, reducing design size, and contributing 5% to 10% to power consumption reduction.
[0088] In addition, the feedback adjustment function of this invention, besides providing a fast response and the optimal clock tree optimization scheme, suppresses the cell growth rate and reduces power consumption, can also reduce the number of design iterations required for the design data to reach the timing convergence state, shorten the design cycle, avoid excessive consumption of manpower and material resources, and reduce design costs.
[0089] The foregoing primarily describes the solutions of the embodiments of this disclosure from a methodological perspective. It is understood that, to achieve the aforementioned functions, the clock tree-based timing convergence device includes at least one of the hardware structures and software modules corresponding to the execution of each function. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments disclosed herein, the embodiments of this disclosure can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the embodiments of this disclosure.
[0090] This disclosure embodiment can divide the clock tree-based timing convergence device into functional modules according to the above method embodiment. For example, each function can be divided into a separate functional module, or two or more functions can be integrated into one functional module. The integrated module can be implemented in hardware or software. It should be noted that the module division in this disclosure embodiment is illustrative and only represents one logical functional division. In actual implementation, there may be other division methods. The following description uses the example of dividing each function into a separate functional module.
[0091] Figure 11 This is a structural block diagram of a timing convergence device based on a clock tree, provided in an embodiment of this disclosure. Figure 11As shown, the timing convergence device based on clock tree includes: an automatic placement unit, a detection unit, a processing unit, a clock tree synthesis unit, a feedback unit, and an automatic routing unit.
[0092] Automatic placement unit, used for automatic placement of components.
[0093] The detection unit is used to determine the timing path that needs to be optimized in the one-to-many structure timing path of the automatically placed device.
[0094] The processing unit is used to optimize the setup time timing margin of the timing path that needs to be optimized, and to determine the common adjustment amount based on the optimized setup time timing margin, which serves as the setting pattern for the useful clock skew adjustment.
[0095] The clock tree synthesis unit is used to determine the number of buffers and generate a clock tree based on the setting pattern of useful clock skew adjustment and optimized setup time timing margin.
[0096] The feedback unit is used to determine, based on the generated clock tree, whether the hold time timing violation of the timing path to be optimized meets the predetermined requirements. If the predetermined requirements are not met, the setting pattern of the useful clock skew adjustment is adjusted based on positive feedback optimization or negative feedback optimization, and the clock tree generation step is returned until the predetermined requirements are met.
[0097] An automatic routing unit is used to complete automatic routing based on a clock tree that meets predetermined requirements.
[0098] In some embodiments, the detection unit is further configured to: acquire the establishment time timing of all timing paths, and extract the start trigger, end trigger, and establishment time timing margin of each timing path in all timing paths; traverse all start triggers in all timing paths and record the occurrence frequency of each start trigger to filter out timing paths with an occurrence frequency greater than a first threshold; extract timing paths containing the same start trigger from the filtered timing paths, and after deduplicating the end triggers, filter out timing paths with a first threshold number; and select the timing paths with a first threshold number as the timing paths to be optimized.
[0099] In some embodiments, the processing unit is further configured to: obtain setup time timing margins from the selected timing paths that need optimization, and select timing paths that need layout optimization based on the setup time timing margins; cluster all devices on the timing paths that need layout optimization and delete redundant buffers; obtain setup time timings on the optimized timing paths, and if the setup time timing margin of the optimized timing path is less than a second threshold, replace the original devices on the optimized timing path with devices that are less than the threshold.
[0100] In some embodiments, the processing unit is further configured to: set the adjustment amount of the transmit clock and the timing path in which the transmit clock is located based on the setup time timing margin; sort all the adjustment amounts of the transmit clocks in the same timing path, take the smallest adjustment amount among all the adjustment amounts as the common adjustment amount, and output it as the setting pattern of the useful clock skew adjustment; and insert a buffer at the end of the timing path in which the transmit clock is located based on the setting pattern of the useful clock skew adjustment.
[0101] In some embodiments, based on the setting pattern of useful clock skew adjustment, inserting a buffer at the end of the timing path where the transmit clock is located includes: obtaining the delay value of the clock line dedicated buffer as the minimum unit value, calculating the number of buffers to be inserted, and adding constraints to the inserted buffers and the connections between them.
[0102] In some embodiments, the feedback unit is further configured to: determine whether the hold-time timing violation in the hold-time timing of the timing path to be optimized in the clock tree meets the expected target; if the hold-time timing violation meets the expected target, determine whether to use positive feedback optimization or negative feedback optimization to adjust the delay of the timing path based on the hold-time timing violation of each timing path in the one-to-many structure timing path, so as to obtain the adjustment amount of the timing path, wherein positive feedback optimization is for those with hold-time timing violations, and negative feedback optimization is for those without hold-time timing violations; sort all the adjustment amounts of the timing path where the same transmit clock is located, take the smallest delay adjustment amount among all the adjustment amounts as the useful clock skew adjustment amount, and output it as the setting pattern of the useful clock skew adjustment; and return to the step of generating the clock tree based on the setting pattern of the useful clock skew adjustment.
[0103] In some embodiments, both positive feedback optimization and negative feedback optimization include optimization under worst-case operating conditions and optimization under best-case operating conditions.
[0104] In some embodiments, the method optimizes under worst-case operating conditions by: for positive feedback optimization, increasing the setting pattern of useful clock skew adjustment based on whether setup time timing is violated, setup time timing margin, hold time timing violation amount, and predetermined coefficient; and for negative feedback optimization, decreasing the setting pattern of useful clock skew adjustment based on whether setup time timing is violated, hold time timing margin, and predetermined coefficient.
[0105] In some embodiments, the method is optimized under optimal operating conditions, comprising: for positive feedback optimization, increasing the setting pattern of useful clock skew adjustment based on whether setup time timing is violated, setup time timing margin, hold time timing violation amount, and predetermined coefficient; and for negative feedback optimization, decreasing the setting pattern of useful clock skew adjustment based on whether setup time timing is violated and hold time timing margin.
[0106] In implementing the functions of the integrated modules described above in hardware, this disclosure provides a structure for the clock tree-based timing convergence device involved in the above embodiments. For example... Figure 12 As shown, the clock tree-based timing convergence device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the aforementioned clock tree-based timing convergence method.
[0107] A processor can be a set of various exemplary logic blocks, modules, and circuits that implement or execute the embodiments described in connection with this disclosure. The processor can be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in connection with the embodiments of this disclosure. A processor can also be a combination that implements computational functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.
[0108] The memory may be read-only memory (ROM) or other types of static storage devices that can store static information and instructions, random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, electrically erasable programmable read-only memory (EEPROM), disk storage media or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer, but is not limited thereto.
[0109] In one implementation, the memory can exist independently of the processor. The memory can be connected to the processor via a bus and used to store instructions or program code. When the processor calls and executes the instructions or program code stored in the memory, it can implement the clock tree-based timing convergence method provided in this embodiment. In another implementation, the memory can also be integrated with the processor.
[0110] Some embodiments of this disclosure provide a computer-readable storage medium (e.g., a non-transitory computer-readable storage medium) storing computer program instructions that, when executed on a computer, cause the computer to perform a clock tree-based timing convergence method as described in any of the above embodiments.
[0111] Exemplary examples show that the aforementioned computer-readable storage media may include, but are not limited to: magnetic storage devices (e.g., hard disks, floppy disks, or magnetic tapes), optical discs (e.g., compact disks (CDs), digital versatile disks (DVDs), etc.), smart cards, and flash memory devices (e.g., erasable programmable read-only memory (EPROMs), cards, sticks, or key drives, etc.). The various computer-readable storage media described in this disclosure may represent one or more devices and / or other machine-readable storage media for storing information. The term "machine-readable storage media" may include, but is not limited to, wireless channels and various other media capable of storing, containing, and / or carrying instructions and / or data.
[0112] This disclosure provides a computer program that, when run on a computer, causes the computer to execute the clock tree-based timing convergence method of any of the above embodiments.
[0113] This disclosure provides a computer program product containing instructions that, when run on a computer, cause the computer to execute the clock tree-based timing convergence method of any of the above embodiments.
[0114] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A timing convergence method based on a clock tree, characterized in that, The method includes: Automatic placement of components; Determine the timing path that needs to be optimized in the one-to-many structure timing path of the device for automatic placement; The setup time timing margin is optimized for the timing path that needs optimization, and a common adjustment amount is determined based on the optimized setup time timing margin as a setting pattern for useful clock skew adjustment. Based on the setting pattern of the useful clock skew adjustment and the optimized setup time timing margin, the number of buffers is determined and a clock tree is generated. Based on the generated clock tree, it is determined whether the hold-time timing violation of the timing path to be optimized meets the predetermined requirements. If the predetermined requirements are not met, the setting pattern of the useful clock skew adjustment is adjusted based on positive feedback optimization or negative feedback optimization, and the process returns to the clock tree generation step until the predetermined requirements are met; and Automatic wiring is completed based on the clock tree that meets the predetermined requirements.
2. The method according to claim 1, characterized in that, Determining the timing path that needs optimization in the one-to-many structure timing path of the automatically placed device includes: Obtain the establishment time sequence of all timing paths, and extract the start trigger, end trigger, and establishment time sequence margin of each timing path in all timing paths; Traverse all starting triggers in all the timing paths and record the frequency of occurrence of each starting trigger to filter out timing paths whose frequency of occurrence is greater than the first threshold. Extract timing paths containing the same starting trigger from the selected timing paths, and after deduplicating the ending triggers, select timing paths with the first threshold number of occurrences; and The time-series path with the number of the first thresholds is taken as the time-series path that needs to be optimized.
3. The method according to claim 1, characterized in that, Optimize the setup time and timing margin of the timing path that needs optimization, including: Obtain the establishment time timing margin from the timing paths that need to be optimized, and filter out the timing paths that need layout optimization based on the establishment time timing margin. All devices on the timing path requiring layout optimization are grouped together, and redundant buffers are removed; and Obtain the setup time timing on the optimized timing path, and if the setup time timing margin of the optimized timing path is less than a second threshold, replace the original device on the optimized timing path with a device whose margin is less than the threshold.
4. The method according to claim 3, characterized in that, The common adjustment amount is determined based on the optimized setup time timing margin, serving as the setting pattern for the useful clock skew adjustment, including: Based on the optimized setup time timing margin, the adjustment amount for the transmit clock and the timing path containing the transmit clock is set; and Sort all adjustment values for the timing path containing the same transmit clock, take the smallest adjustment value among all adjustment values as the common adjustment value, and output it as the setting pattern for the useful clock skew adjustment; and Based on the setting pattern of the useful clock skew adjustment, a buffer is inserted at the end of the timing path where the transmit clock is located.
5. The method according to claim 4, characterized in that, Based on the setting pattern of the useful clock skew adjustment, a buffer is inserted at the end of the timing path where the transmit clock is located, including: Obtain the delay value of the dedicated clock line buffer as the minimum unit value, calculate the number of buffers that need to be inserted, and add constraints to the inserted buffers and the connections between them.
6. The method according to any one of claims 1 to 5, characterized in that, Based on the generated clock tree, it is determined whether the hold-time timing violation of the timing path to be optimized meets the predetermined requirements. If the expected requirements are not met, the setting pattern of the useful clock skew adjustment is adjusted based on positive feedback optimization or negative feedback optimization, including: Determine whether the hold-time timing violation in the hold-time timing path to be optimized in the clock tree achieves the expected goal. If the hold-time timing violation reaches the expected target, based on the hold-time timing violation of each timing path in the one-to-many structure timing path, it is determined whether to use positive feedback optimization or negative feedback optimization to adjust the delay of the timing path to obtain the adjustment amount of the timing path. The positive feedback optimization is for the case with the hold-time timing violation, and the negative feedback optimization is for the case without the hold-time timing violation. Sort all adjustment values for the timing path containing the same transmit clock, select the smallest adjustment value among all adjustment values as the useful clock skew adjustment value, and output it as the setting pattern for the useful clock skew adjustment; and Based on the setting pattern of the useful clock skew adjustment, return to the step of generating the clock tree.
7. The method according to claim 6, characterized in that, Both the positive feedback optimization and the negative feedback optimization include optimization under worst-case operating conditions and optimization under best-case operating conditions.
8. The method according to claim 7, characterized in that, Optimization under the worst-case operating conditions, the method includes: For the positive feedback optimization, based on whether the setup time timing is violated, the setup time timing margin, the hold time timing violation amount, and a predetermined coefficient, the setting pattern for the useful clock skew adjustment is increased; and For the negative feedback optimization, based on whether the setup timing is violated, the maintenance timing margin, and the predetermined coefficient, the setting pattern of the useful clock skew adjustment is reduced.
9. The method according to claim 7, characterized in that, Optimized under the optimal operating conditions, the method includes: For the positive feedback optimization, based on whether the setup time timing is violated, the setup time timing margin, the hold time timing violation amount, and a predetermined coefficient, the setting pattern for the useful clock skew adjustment is increased; and Regarding the negative feedback optimization, based on whether the setup timing is violated and the maintenance timing margin, the setting pattern of the useful clock skew adjustment is reduced.
10. A timing convergence device based on a clock tree, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the timing convergence method based on a clock tree as described in any one of claims 1 to 9.
Citation Information
Patent Citations
A time sequence repairing method based on time headroom
CN109583103A
Clock tree generation and optimization method for quickly realizing time sequence convergence
CN120409367A
Method, tool, computer equipment, medium and program for concurrent clock optimization and data optimization
CN120633585A
Time sequence optimization method of clock tree and computer readable storage medium
CN121257463A
LSI design method
JP2013171503A
Cited By
Timing sequence adjustment method, system and device for expanding chip sign-off conditions and medium
CN122113787A
Method, system, device and medium for adjusting timing of signing condition of extended chip core
CN122113787B
