Methods for predicting the power consumption of integrated circuits

By analyzing the clock gating cell hierarchy and memory operation state of integrated circuits, and using a weighted lookup table to calculate the power consumption of integrated circuits, the problems of long prediction time and low accuracy in existing technologies are solved, and efficient and accurate power consumption prediction is achieved.

CN115617151BActive Publication Date: 2026-11-17BAUM DESIGN SYST CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202111253384.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Priority Date
2021-07-12
Filing Date
2021-10-27
Publication Date
2026-11-17
Estimated Expiration
2041-10-27

AI Technical Summary

Technical Problem

Existing integrated circuit power consumption prediction methods suffer from problems such as excessively long prediction time or low accuracy, are difficult to automate, and fail to effectively consider the impact of hierarchical structure and memory operation state on power consumption.

Method used

By determining the hierarchical structure of multiple clock gating units in the integrated circuit, calculating the power consumption of each clock gating domain, considering the logic level of the clock gating enable signal and the operating state of the memory, determining the weights using a weighted lookup table, and calculating the total power consumption of the integrated circuit.

Benefits of technology

It achieves accurate prediction of integrated circuit power consumption, taking into account hierarchical structure and memory operation state, thus improving the accuracy and automation of prediction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115617151B_ABST
    Figure CN115617151B_ABST
Patent Text Reader

Abstract

The technical concept of this disclosure relates to a method for predicting the power consumption of an integrated circuit and a system for performing such prediction. One embodiment of the method for predicting the power consumption of an integrated circuit may include: determining a hierarchical structure of a plurality of clock gating units included in the integrated circuit; determining, based on the hierarchical structure, a first clock gating domain corresponding to a first clock gating unit and a second clock gating domain corresponding to a second clock gating unit located below the first clock gating unit; calculating the power consumption of the second clock gating domain based on a first logic level of a first clock gating enable signal applied to the first clock gating unit; and predicting the power consumption of the integrated circuit based on the power consumption of the second clock gating domain.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to a method for designing integrated circuits, and more specifically, to a method for predicting the power consumption of integrated circuits and a power consumption prediction system for performing such prediction. Background Technology

[0002] An integrated circuit (IC) is an electronic device that integrates various functions such as computing and storage onto a single semiconductor chip. With the increasing societal sophistication, various forms of mobile devices are being developed. Mobile devices strive for miniaturization and weight reduction by incorporating IC devices. A drawback of mobile devices is their limited power supply, such as a battery. To extend the lifespan of mobile devices, research is underway to increase battery capacity and reduce the power consumption of IC devices.

[0003] Predicting the power consumption of integrated circuit devices is a prerequisite for achieving low-power integrated circuit devices. Existing power consumption prediction methods suffer from problems such as excessively long prediction times or low accuracy.

[0004] To address this problem, various solutions have been proposed. In these methods, a person who understands the operation of the target integrated circuit device extracts power-consuming states as power states and generates a power model based on them. However, this approach suffers from large deviations due to the knowledge and experience of the person executing the process and is difficult to automate.

[0005] Existing technical documents

[0006] Patent documents

[0007] Korean Patent Publication No. 10-1139603 (Granted on April 17, 2012) Summary of the Invention

[0008] The problem the invention aims to solve

[0009] One object of the present invention is to provide an integrated circuit power consumption prediction method and a power consumption prediction system that performs the method, which takes into account the hierarchical structure to calculate the power consumption of each domain.

[0010] Another object of the present invention is to provide an integrated circuit power consumption prediction method and a power consumption prediction system that performs the method, taking into account whether or not a memory is available, to calculate the power consumption of each domain.

[0011] Solution for solving the problem

[0012] The method for predicting the power consumption of an integrated circuit according to the present invention may include: determining a hierarchical structure of a plurality of clock gating units included in the integrated circuit; determining, based on the hierarchical structure, a first clock gating domain corresponding to a first clock gating unit and a second clock gating domain corresponding to a second clock gating unit located below the first clock gating unit; calculating the power consumption of the second clock gating domain based on a first logic level of a first clock gating enable signal applied to the first clock gating unit; and predicting the power consumption of the integrated circuit based on the power consumption of the second clock gating domain.

[0013] In one embodiment, the step of calculating the power consumption of the second clock gate domain may include: determining the first logic level of the first clock gate enable signal and the second logic level of the second clock gate enable signal applied to the second clock gate unit; and calculating the power consumption of the second clock gate domain by multiplying the first logic level, the second logic level and the weight corresponding to the second clock gate domain.

[0014] In one embodiment, the power consumption prediction method may further include: calculating the power consumption of the first clock gate domain by multiplying the first logic level and the weight corresponding to the first clock gate domain; and predicting the power consumption of the integrated circuit based on the power consumption of the first clock gate domain and the power consumption of the second clock gate domain.

[0015] In one embodiment, the power consumption prediction method may further include: determining whether a memory is included in the third clock gating domain corresponding to the third clock gating unit; when the memory is included in the third clock gating domain, determining a weight corresponding to the third clock gating domain based on the operating state of the memory; and calculating the power consumption of the third clock gating domain based on the determined weight corresponding to the third clock gating domain and the logic level of the third clock gating enable signal applied to the third clock gating unit.

[0016] In one embodiment, the step of determining the weight corresponding to the third clock gating domain may include: obtaining information related to the signal applied to the memory; and determining the weight corresponding to the signal based on a weight lookup table.

[0017] In one embodiment, the present invention is characterized in that the signals applied to the memory include a chip enable signal and an optical enable signal, and the weight lookup table includes information relating to the weights of each combination of the chip enable signal and the optical enable signal.

[0018] In one embodiment, the third clock-gated domain includes a first memory and a second memory. The step of calculating the power consumption of the third clock-gated domain may include: obtaining a first sub-weight from the weight lookup table based on a combination of control signals applied to the first memory during a first clock cycle; obtaining a second sub-weight from the weight lookup table based on a combination of control signals applied to the second memory during the first clock cycle; adding the first sub-weight and the second sub-weight to calculate a first weight corresponding to the third clock-gated domain of the first clock cycle; calculating the power consumption corresponding to the third clock-gated domain of the first clock cycle based on the first weight; obtaining a third sub-weight from the weight lookup table based on a combination of control signals applied to the first memory during a second clock cycle; obtaining a fourth sub-weight from the weight lookup table based on a combination of control signals applied to the second memory during the second clock cycle; adding the third sub-weight and the fourth sub-weight to calculate a second weight corresponding to the third clock-gated domain of the second clock cycle; and calculating the power consumption corresponding to the third clock-gated domain of the second clock cycle based on the second weight.

[0019] In one embodiment, the power consumption prediction method may further include: distinguishing independent elements not included in the clock gating domain of the plurality of clock gating units; measuring the power consumption of combinational logic in the independent elements; determining a first constant corresponding to the power consumption of the combinational logic; determining a second constant corresponding to the power consumption of clock logic in the independent elements; and predicting the power consumption of the integrated circuit based on the first and second constants.

[0020] The power consumption prediction method of this disclosure may include: a step of determining whether a memory is included in a clock gating domain corresponding to a clock gating unit; a step of determining a weight corresponding to the clock gating domain based on the operating state of the memory when the memory is included in the clock gating domain; and a step of calculating the power consumption of the clock gating domain based on the determined weight corresponding to the clock gating domain and the logic level of the clock gating enable signal applied to the clock gating unit.

[0021] In one embodiment, the step of determining the weights corresponding to the clock gating domain may include: obtaining information related to the signal combination applied to the memory; and determining the weights corresponding to the signal combination based on a weight lookup table.

[0022] In one embodiment, the present invention is characterized in that the signals applied to the memory include a chip enable signal and an optical enable signal, and the weight lookup table includes information relating to the weights of each combination of the chip enable signal and the optical enable signal.

[0023] In one embodiment, the clock gating domain includes a first memory and a second memory. The step of calculating the power consumption of the clock gating domain may include: obtaining a first sub-weight from the weight lookup table based on a combination of control signals applied to the first memory during a first clock cycle; obtaining a second sub-weight from the weight lookup table based on a combination of control signals applied to the second memory during the first clock cycle; adding the first sub-weight and the second sub-weight to calculate a first weight corresponding to the third clock gating domain of the first clock cycle; calculating the power consumption corresponding to the clock gating domain of the first clock cycle based on the first weight; obtaining a third sub-weight from the weight lookup table based on a combination of control signals applied to the first memory during a second clock cycle; obtaining a fourth sub-weight from the weight lookup table based on a combination of control signals applied to the second memory during the second clock cycle; adding the third sub-weight and the fourth sub-weight to calculate a second weight corresponding to the third clock gating domain of the second clock cycle; and calculating the power consumption corresponding to the clock gating domain of the second clock cycle based on the second weight.

[0024] Invention Effects

[0025] According to one embodiment of the invention, the power consumption of each domain is calculated considering the hierarchical structure, thereby predicting power consumption similar to the actual power consumption. Furthermore, according to another embodiment of the invention, the power consumption of each domain is calculated based on whether or not a memory is present, thereby calculating the power consumption according to the operating state of the memory, and thus predicting power consumption similar to the actual power consumption. Attached Figure Description

[0026] Figure 1 A flowchart illustrating a power consumption prediction method for an integrated circuit according to an exemplary embodiment of this disclosure.

[0027] Figure 2 A flowchart illustrating a power consumption prediction method according to an exemplary embodiment of this disclosure is provided.

[0028] Figure 3 A circuit diagram illustrating an exemplary embodiment of the integrated circuit of this disclosure is provided.

[0029] Figure 4 A flowchart illustrating a power consumption prediction method according to an exemplary embodiment of this disclosure is provided.

[0030] Figure 5 A circuit diagram illustrating an exemplary embodiment of the integrated circuit of this disclosure is provided.

[0031] Figure 6 This disclosure illustrates a method for determining weights based on the operating state of a memory, according to an exemplary embodiment of the present disclosure.

[0032] Figure 7 A flowchart illustrating a power consumption prediction method according to an exemplary embodiment of this disclosure is provided.

[0033] Figure 8 and Figure 9 A diagram illustrating the power consumption prediction method of an exemplary embodiment of this disclosure.

[0034] Figure 10 A flowchart illustrating a power consumption prediction method according to an exemplary embodiment of this disclosure is provided.

[0035] Figure 11 A circuit diagram illustrating an exemplary embodiment of the integrated circuit of this disclosure is provided.

[0036] Figure 12 A block diagram illustrating a power consumption prediction system according to an embodiment of the present disclosure is provided. Detailed Implementation

[0037] The preferred embodiments of this disclosure are described in detail below with reference to the accompanying drawings. The advantages, features, and methods for achieving these advantages and features of this disclosure will become clear from the detailed embodiments and drawings described below. However, the technical concept of this disclosure is not limited to the following embodiments and can be embodied in various different implementations. The following embodiments are provided only to make the technical concept of this disclosure more complete and to fully inform those skilled in the art of the scope of this disclosure. The technical concept of this disclosure is defined only by the scope of the invention claims.

[0038] It should be noted that when assigning reference numerals to structural elements in the various figures, the same reference numerals should be assigned to the same structural elements as much as possible, even if they appear in different figures. Furthermore, in describing this disclosure, detailed descriptions related to relevant well-known structures or functions will be omitted if it is determined that such detailed descriptions would unnecessarily obscure the spirit of this disclosure.

[0039] Unless otherwise defined, all terms (including technical and scientific terms) used in this specification are to be understood in the same sense as those skilled in the art to which this disclosure pertains. Furthermore, unless explicitly defined, terms as defined in commonly used dictionaries should not be interpreted ideally or excessively. The terms used in this specification are for illustrative purposes and do not limit the scope of this disclosure. In this specification, singular forms include plural forms unless specifically mentioned herein.

[0040] Furthermore, when describing the structural elements of this disclosure, terms such as first, second, A, B, (a), (b), etc., may be used. These terms are used only to distinguish one structural element from another, and the nature, order, or sequence of the corresponding structural elements are not limited by these terms. It should be understood that when a structural element is described as being "connected," "combined," or "linked" with another structural element, it can be understood that one structural element can be directly connected or linked to another structural element, and it can also be understood that other structural elements are "connected," "combined," or "linked" between the structural elements.

[0041] As used in this disclosure, “comprises” and / or “comprising” means that the mention of a structural element, step, action and / or component does not exclude the presence or addition of more than one other structural element, step, action and / or component.

[0042] Structural elements having the same function as those included in any embodiment may be described using the same names in other embodiments. Unless otherwise stated, the descriptions provided in any embodiment are also applicable to other embodiments, and specific descriptions may be omitted to the extent that they are repeated or clearly understood by one of ordinary skill in the art.

[0043] The present invention will now be described in detail with reference to preferred embodiments and accompanying drawings.

[0044] Figure 1 A flowchart illustrating a power consumption prediction method for an integrated circuit according to an exemplary embodiment of this disclosure.

[0045] Reference Figure 1 The integrated circuit power consumption prediction method may include a domain extraction step S11, a hierarchy detection step S12, a power consumption calculation step S13 for each domain, and a power consumption prediction step S14 for the integrated circuit. In one example, the power consumption prediction method may be performed at the gate level, register transfer level, or system level.

[0046] Power consumption prediction methods can be implemented through a power consumption prediction system. This system can include various structures used in designing integrated circuits (ICs). Power consumption prediction systems can be implemented in various terminal devices such as cellular phones, smartphones, laptops, personal computers (PCs), navigators, personal communication systems (PCS), Global System for Mobile communications (GSM), personal digital cellular (PDC), personal handyphone systems (PHS), personal digital assistants (PDAs), International Mobile Telecommunications (IMT)-2000 systems, Code Division Multiple Access (CDMA)-2000, Wideband Code Division Multiple Access (W-CDMA), Wireless Broadband Internet (Wibro) terminals, smartpads, and tablet PCs.

[0047] Integrated circuits can operate based on clock gating. An integrated circuit is divided into a clock-gated input cone and a clock-gated domain, centered around the clock-gated cell. The clock-gated input cone responds to input signals to control the clock-gated enable signal. For example, the clock-gated input cone can control the clock-gated enable signal to be either logic low or logic high.

[0048] The clock gating unit can respond to a clock gating enable signal to terminate and resume providing clock to the clock gating domain. For example, the clock gating unit receives a clock and responds to a clock gating enable signal to output the received clock to the clock gating domain.

[0049] The clock gating domain operates in response to the gated clock as the output of the clock gating unit. For example, the clock gating domain may include a clock network connected to the gated clock.

[0050] If the clock gating unit terminates the output of the gated clock, the operation of the clock gating domain can be terminated. For example, the clock gating domain can correspond to the termination of the gated clock, which is the output of the clock gating unit, thereby terminating the operation while maintaining the data. Therefore, power consumption of the clock gating domain can be prevented.

[0051] In this specification, a clock-gated field refers to the fan-out logic cone of a clock-gated unit that supplies a clock to a clock gated unit.

[0052] In the domain extraction step S11, the power consumption prediction system can extract multiple clock-gated domains corresponding to multiple clock-gated units included in the integrated circuit. As described above, the power consumption prediction system can extract multiple elements applying clocks as clock-gated domains based on a clock-gated enable signal of each clock-gated unit.

[0053] In the hierarchical structure detection step S12, the power consumption prediction system can determine the hierarchical structure of multiple clock gating units. In this specification, the hierarchical structure of multiple clock gating units refers to a signal gating level based on the flow of clock signals. In one example, a lower-level clock gating unit in the hierarchy can further gating a clock signal that has been gated by an upper-level clock gating unit. The power consumption prediction system can determine the lower-level clock gating units and the upper-level clock gating units by detecting the hierarchical structure of the multiple clock gating units.

[0054] In the power consumption calculation step S13 for each domain, the power consumption prediction system may consider the hierarchical structure to calculate the power consumption of each clock-gated domain. In one embodiment, when calculating the power consumption of a clock-gated domain, the power consumption prediction system may calculate the power consumption based on the clock gating enable signal of the clock gating unit and the clock gating enable signal applied to the upper-level clock gating unit. In one example, the power consumption may be the clock network power consumed by the clocked element included in the clock-gated domain. In another example, in addition to predicting the clock network power, the power consumption prediction system may also predict the power consumed by combinational logic.

[0055] In the integrated circuit power consumption prediction step S14, the power consumption prediction system can predict the integrated circuit power consumption using the power consumption of each clock gate domain. In one embodiment, the power consumption prediction system can add the power consumption of multiple clock gate domains to predict the integrated circuit power consumption.

[0056] In the power consumption prediction method of this disclosure, when calculating the power consumption of each clock gate domain, the clock gate enable signal corresponding to the clock gate domain and the upper-level clock gate enable signal are considered to calculate the power consumption, thereby calculating the accurate power consumption.

[0057] Figure 2 A flowchart illustrating a power consumption prediction method according to an exemplary embodiment of this disclosure is provided. In detail, Figure 2 This paper illustrates a method for calculating the power consumption of a second clock-gated domain by considering the hierarchical structure.

[0058] Reference Figure 2 The power consumption prediction system can determine a first clock gating domain located at the upper layer and a second clock gating domain located at the lower layer (S110). In one example, the first clock gating domain can receive a clock gated by a first clock gating unit controlled by a first clock gating enable signal, and the second clock gating domain can receive clocks gated by both the first clock gating unit and a second clock gating unit controlled by a second clock gating enable signal. Thus, it can be expressed that the second clock gating domain is located below the first clock gating domain.

[0059] The power consumption prediction system can acquire information related to a first logic level of the first clock gating enable signal and a second logic level of the second clock gating enable signal (S120). In one example, the first logic level and the second logic level can each have a value that indicates either "1 or logic high (H)" meaning the clock is output to the clock gating domain or "0 or logic low (L)" meaning the clock is not output to the clock gating domain. In one example, the power consumption prediction system can determine the first logic level and the second logic level based on the scenario of operating the integrated circuit.

[0060] The power consumption prediction system can obtain weights corresponding to the second clock gate domain (S130). In this specification, the weights can indicate the weights for the power consumption on a clock network that vary depending on the type and number of components included in the clock gate domain.

[0061] The power consumption prediction system calculates the power consumption of the second clock-gated domain by multiplying the first logic level, the second logic level, and a determined weight (S140). That is, the power consumption of the second clock-gated domain can only be calculated when the second logic level and the first logic level are both "logic high". When only the second logic level is considered in calculating the power consumption of the second clock-gated domain, even if the clock is not applied to the second clock-gated domain through the first clock-gated unit, it can be calculated as if the clock is applied to the second clock-gated domain, thus resulting in a value different from the actual power consumption. Conversely, according to an embodiment of the present invention, a hierarchical structure can be considered to calculate the dynamic power consumption when the clock is actually applied to the clock-gated domain. Therefore, when the clock is actually applied to the second clock-gated domain and dynamic power is consumed, the actual dynamic power consumption is reflected, thereby accurately calculating the power consumption.

[0062] Figure 3 A circuit diagram of an integrated circuit illustrating an exemplary embodiment of this disclosure is provided. In detail, Figure 3 This is a diagram illustrating a method for calculating the power consumption of a clock-gated domain based on a hierarchical structure.

[0063] Reference Figure 3 The integrated circuit IC1 may include: a first clock gating unit CG1, controlled by a first clock gating enable signal CGE1; a first clock gating domain CGD1, which receives clock Clk through the first clock gating unit CG1; a second clock gating unit CG2, controlled by a second clock gating enable signal CGE2; and a second clock gating domain CGD2, which receives clock Clk through the second clock gating unit CG2.

[0064] The second clock gating unit CG2 can output the clock Clk, which is gated by the first clock gating unit CG1, to the second clock gating domain CGD2. Therefore, it can be said that the second clock gating unit CG2 is located below the first clock gating unit CG1.

[0065] The power consumption CP1 of the first clock-gated domain CGD1 can be calculated as follows: Since there is no upper-level clock-gated unit above the first clock-gated unit CG1, for the first weight Wt1 corresponding to the first clock-gated domain CGD1, such as " ", can be calculated by multiplying the first weight Wt1 and the first clock gating enable signal CGE1. In one example, the power consumed CP1 refers to the clock network power consumed by the clock-applied element in the first clock gating domain CGD1.

[0066] Conversely, the power consumption CP2 of the second clock-gated domain CGD2 can be calculated as follows: Since there exists a first clock-gated unit CG1 above the second clock-gated unit CG2, the first clock-gated enable signal CGE1 and the second clock-gated enable signal CGE2 can be considered for calculation. For the second weight Wt2 corresponding to the second clock-gated domain CGD2, it can be calculated from… "calculate.

[0067] Therefore, the power consumption CP of integrated circuit IC1 can be calculated using the following mathematical formula 1 (C1 is a constant value caused by leakage power, etc.).

[0068] Mathematical formula 1:

[0069] exist Figure 3 In the example, when considering the upper layer, the first clock gating enable signal CGE1 is logic low ("0") and the second clock gating enable signal CGE2 is logic high ("1"). Therefore, the power consumption CP of integrated circuit IC1 can be calculated by reflecting the actual power consumed.

[0070] Without considering the hierarchical structure, even if the clock Clk is not passed to the second clock-gated domain CGD2 via the first clock-gated enable signal CGE1, the second clock-gated enable signal CGE2 is still logic high ("1"). Therefore, the calculated power consumption may be greater than the actual power consumption. According to an embodiment of this disclosure, considering the hierarchical structure, the power consumption of the clock-gated domain can be calculated based on whether the clock Clk is actually passed, taking into account the clock enable signals of the upper layers, thereby accurately predicting the power consumption.

[0071] exist Figures 1 to 3 The example shown is of two-layer clock gating units CG1 and CG2 included in an integrated circuit. This is just an example. In cases where more than two layers of clock gating units are included in an integrated circuit, the technical concept of this disclosure can also be applied in embodiments where the power consumption of the lower-layer clock gating domain is calculated by considering the clock gating enable signal of the upper layer. This is to be expected.

[0072] Figure 4 A flowchart illustrating a power consumption prediction method according to an exemplary embodiment of this disclosure is provided. In detail, Figure 4 This demonstrates a method for predicting power consumption based on the operating state of the memory.

[0073] Reference Figure 4The power consumption prediction system can determine whether the memory is included in the clock gate domain (S210). In this specification, memory refers to a device used for storing data, including non-volatile memory such as NAND, NOR, flash memory, hard disk, etc., and volatile memory such as dynamic random access memory (DRAM), static random access memory (SRAM), etc.

[0074] In the case of a memory, the power consumption prediction system may determine the weights for calculating the power consumption of the clock-gated domain including the memory based on the operating state of the memory (S220). In one embodiment, the power consumption prediction system may determine the weights based on various combinations of signals applied to the memory.

[0075] The power consumption prediction system can determine the power consumption of the clock-gated domain based on the weights and the logic level of the clock gating enable signal (S230).

[0076] According to one embodiment of the present disclosure, when a memory is included in the clock-gated domain, the power consumption prediction system determines the weights as weights of power consumption based on the operating state of the memory. When predicting the power consumption of the integrated circuit, it can reflect the power consumption of the memory, which varies according to each operating state, thereby accurately predicting the power consumption.

[0077] Figure 5 A circuit diagram of an integrated circuit illustrating an exemplary embodiment of this disclosure is provided. In detail, Figure 5 A diagram illustrating a method for calculating the power consumption of a clock-gated domain based on the operating state of a memory.

[0078] Reference Figure 5 The integrated circuit IC2 may include: a first clock gating unit CG1, controlled by a first clock gating enable signal CGE1; a first clock gating domain CGD1, receiving clock Clk through the first clock gating unit CG1; a second clock gating unit CG2, controlled by a second clock gating enable signal CGE2; a second clock gating domain CGD2, receiving clock Clk through the second clock gating unit CG2; a third clock gating unit CG3, controlled by a third clock gating enable signal CGE3; a third clock gating domain CGD3, receiving clock Clk through the third clock gating unit CG3; a fourth clock gating unit CG4, controlled by a fourth clock gating enable signal CGE4; and a fourth clock gating domain CGD4, receiving clock Clk through the fourth clock gating unit CG4.

[0079] The first clock-gated domains CGD1 to the third clock-gated domains CGD3 do not include memory, thus allowing access via... Figures 1 to 3The method described herein calculates the power consumption CP1, CP2, and CP3 consumed by the first clock gate domain CGD1 to the third clock gate domain CGD3, therefore, the description of it will be omitted.

[0080] The fourth clock-gated domain CGD4 includes the memory Mem, and therefore the weights can be determined based on the memory's operating state. In one embodiment, the fourth weight Wt4 corresponding to the fourth clock-gated domain CGD4 can be determined based on various combinations of signals applied to the memory Mem; in one example, it can be determined based on the chip enable signal Cen and the optical enable signal Wen. In this specification, "Wt4(Cen, Wen)" can represent the fourth weight Wt4 determined by the chip enable signal Cen and the optical enable signal Wen.

[0081] In a power consumption prediction method according to an embodiment of the present disclosure, for a clock-gated domain including memory Mem, the power consumption can be calculated based on the operating state of memory Mem, thereby predicting the power consumption of integrated circuit IC2 by reflecting the power consumption differences that arise according to the operating state of each memory Mem.

[0082] As in Figures 1 to 3 As mentioned above, both the hierarchical structure and the operational state of the memory can be considered simultaneously to determine the " Calculate the power consumption CP4 of the fourth clock-gated domain.

[0083] Therefore, for the third weight Wt3 of the third clock gate domain CGD3, the power consumption CP of integrated circuit IC2 can be calculated by the following mathematical formula 2 (C2 is a constant value caused by leakage power, etc.).

[0084] Mathematical formula 2:

[0085] Figure 6 This disclosure illustrates an exemplary embodiment of a method for determining weights based on the operating state of a memory. In detail, Figure 6 The method for generating a weight lookup table WL for determining weights and using it to determine weights based on the operating state of the memory is shown.

[0086] Reference Figure 5 and Figure 6 The power consumption prediction system can determine the weights based on a weight lookup table WL, which is composed of combinations of input signals from memory. In one example, the weight lookup table WL may contain information related to the weight Wt_m of each combination of control signals Cen and Wen.

[0087] In one embodiment, the weight Wt_m of each control signal combination can be determined through simulation results. In one example, the clock power consumption and access counts (#Counts) can be obtained for each address and control signal combination through simulation. In an example where the chip enable signal Cen is logic low ("0") and the optical enable signal Wen is logic low ("0"), when the address is "00", a first power p1 is consumed, and the corresponding address can be accessed with a first number n1. In this way, in the example of the corresponding control signal combination, the clock power consumption and access counts (#Counts) for all addresses can be obtained, and the weight (Wt_m(0,0)) of each control signal combination can be obtained through the following mathematical formula 3.

[0088] Mathematical formula 3:

[0089] In a similar manner, for all cases of the chip enable signal Cen and the optical enable signal Wen, the weights of each control signal (Wt_m(0,1), Wt_m(1,0), Wt_m(1,1)) can be obtained, thereby generating a weight lookup table WL.

[0090] According to one embodiment of this disclosure, the weight of each control signal combination is obtained from a weight lookup table WL generated by simulation results, and the power consumption of the memory is calculated using the obtained weights, thereby enabling accurate prediction of power consumption based on the memory state.

[0091] exist Figure 6 The illustration shows an example of calculating power consumption based on the chip enable signal Cen and the optical enable signal Wen as control signals. The technical concept of this disclosure is not limited to this, and can also be applied to examples of calculating power consumption based on various signals of memory (e.g., address signals, data signals, etc.), which is to be expected.

[0092] Figure 7 A flowchart illustrating a power consumption prediction method according to an exemplary embodiment of this disclosure is provided. In detail, Figure 7 A method for calculating the power consumption of a clock-gated domain comprising multiple memories is shown.

[0093] Reference Figure 7 The power consumption prediction system can obtain a first sub-weight based on a combination of control signals applied to the first memory during the first clock cycle (S310). The power consumption prediction system can obtain a second sub-weight based on a combination of control signals applied to the second memory during the first clock cycle (S320). The first sub-weight and the second sub-weight can be weights corresponding to the first memory or the second memory, respectively, such as in... Figure 6As mentioned above, it can be obtained through a weighted lookup table.

[0094] The power consumption prediction system can add the acquired first sub-weight and second sub-weight to determine the weight corresponding to the first clock cycle (S330). The power consumption prediction system can calculate the power consumption of the clock-gated domain, including the first memory and the second memory, within the first clock cycle based on the determined weight (S340).

[0095] Figure 8 and Figure 9 This is a diagram illustrating the power consumption prediction method of an exemplary embodiment of this disclosure. In detail, Figure 8 and Figure 9 A method is shown for calculating the power consumption of a clock-gated domain (CGD) comprising multiple memories Mem1 and Mem2.

[0096] Reference Figure 8 The integrated circuit IC3 may include: a clock gating unit CG, controlled by a clock gating enable signal CGE; a first memory Mem1, controlled by a first chip enable signal Cen1 and a first optical enable signal Wen1; and a second memory Mem2, controlled by a second chip enable signal Cen2 and a second optical enable signal Wen2.

[0097] The power consumption of integrated circuit IC3 can be determined based on the first sub-weight Wt_m1 corresponding to the first memory Mem1 and the second sub-weight Wt_m2 corresponding to the second memory Mem2. In one embodiment, as in Figures 1 to 8 As described in the text, it can be derived from " (C3 is a constant value caused by leakage power, etc.) The power consumption CP of integrated circuit IC3 can be determined based on the weight lookup tables WL1 and WL2 to obtain each sub-weight. The first weight lookup table WL1 corresponds to the first memory Mem1, and the second weight lookup table WL2 corresponds to the second memory Mem2.

[0098] Reference Figure 9 The diagram shows the first chip enable signal Cen1, the first optical enable signal Wen1, the second chip enable signal Cen2, and the second optical enable signal Wen2 for each of the first clock cycles t1 to the fourth clock cycle t4. In the first clock cycle t1, the first chip enable signal Cen1 may have a logic high (H), the first optical enable signal Wen1 may have a logic high (H), the second chip enable signal Cen2 may have a logic high (H), and the second optical enable signal Wen2 may have a logic low (L).

[0099] Reference Figure 8In the first weight lookup table WL1, when the first chip enable signal Cen1 is logic high (H) and the first optical enable signal Wen1 is logic high (H), the first sub-weight Wt_m1 can have a value of "3". Referring to the second weight lookup table WL2, when the third chip enable signal Cen3 is logic high (H) and the second optical enable signal Wen2 is logic low (L), the second sub-weight Wt_m2 can have a value of "12". Therefore, within the first clock cycle, the weight Wt_m corresponding to the clock gate domain CGD can be determined as "15", which is the sum of the first sub-weight Wt_m1 and the second sub-weight Wt_m2. Similarly, the weight Wt_m for the second clock cycle can be determined as "5", the weight Wt_m for the third clock cycle can be determined as "3", and the weight Wt_m for the fourth clock cycle can be determined as "18".

[0100] In one example, when the clock gating enable signal CGE is logic high H, the power consumed during the clock cycles t1, t2, t3, and t4 of integrated circuit IC3 can be determined as "41" by adding the weights Wt_m of the first clock cycle t1 to the fourth clock cycle t4.

[0101] According to one embodiment of this disclosure, when multiple memories are in different operating states, the sub-weights of each memory can be added together to determine the weight, which can be used to predict power consumption, thereby accurately predicting power consumption.

[0102] exist Figure 8 and Figure 9 The illustration shows an example of two memories included in a clock-gated domain (CGD). This is just one example; the technical concept of this disclosure can also be applied to examples of three or more memories included in a clock-gated domain (CGD), which is self-evident.

[0103] Furthermore, embodiments that include other elements besides the memory can also be implemented by, as in... Figures 1 to 3 The weights with fixed values ​​as a basis, as described in the embodiments, and as in Figure 6 The method described above, which calculates power consumption by adding the weights of the changes in the memory-based input signal, is logical.

[0104] Figure 10 A flowchart illustrating a power consumption prediction method according to an exemplary embodiment of this disclosure is provided.

[0105] The power consumption prediction system can distinguish multiple independent components not included in the clock gating domain (S410). An independent component is a component that does not receive a clock through the clock gating unit or is not clocked in the entire circuit. Independent components may include: clock logic that does not receive a clock through the clock gating unit; combinational logic that does not receive a clock in the entire circuit; and multiple memories that do not receive a clock through the clock gating unit.

[0106] The power consumption prediction system measures the average power consumption of the combinational logic and can determine a first constant (S420) of the measured average power consumption. The first constant corresponds to the power consumption of the combinational logic.

[0107] The power consumption prediction system can determine a second constant (S430) for the power consumption of a clock-receiving element (e.g., a memory that does not receive a clock through clock logic and / or clock gating units). The second constant can be a value corresponding to the power consumed by applying a clock to the clock-receiving element. The power consumption prediction system can determine constants for multiple independent elements based on the first and second constants (S440).

[0108] Furthermore, although not illustrated, the power consumption prediction system can predict power consumption by determining a constant for the combination of input signals for each of multiple memories included in independent elements. In one embodiment, the power consumption prediction system can predict power consumption by... Figure 6 The method for determining weights described herein is similar to the method for determining constants included in memory in independent components.

[0109] The power consumption prediction system can predict the power consumption of an integrated circuit based on the constants of multiple independent components (S450).

[0110] In this specification, a constant refers to a constant value used when calculating the power consumption of an integrated circuit, independent of the logic level of the clock gating enable signal. In one example, the power consumption prediction system calculates the power consumption of an integrated circuit by... Figures 1 to 9 The method described herein calculates the power consumption corresponding to multiple clock-gated domains, and the calculated value can be added to a constant determined by S440 to predict the power consumption of the integrated circuit.

[0111] Figure 11 A circuit diagram illustrating an exemplary embodiment of the integrated circuit of this disclosure is provided.

[0112] Reference Figure 11The integrated circuit IC4 may include: a first clock gating unit CG1, controlled by a first clock gating enable signal CGE1; a first clock gating domain CGD1, receiving clock Clk through the first clock gating unit CG1; a second clock gating unit CG2, controlled by a second clock gating enable signal CGE2; a second clock gating domain CGD2, receiving clock Clk through the second clock gating unit CG2; and an independent component ILD. The independent component may include components that do not receive clock signals through additional clock gating units, components that do not receive clock signals throughout the circuit, and memory.

[0113] The power consumption prediction system can measure the average power consumption of combinational logic in an independent element ILD that does not receive a clock. Furthermore, the power consumption prediction system determines a first constant of the combinational logic based on the measured average power consumption, and can determine the constant Cst value of multiple independent element ILDs based on the first constant and a second constant of the element receiving the clock.

[0114] Based on Figure 3 The power consumption CPI of multiple independent components ILD can be predicted by the following mathematical formula 4 using the mathematical formula 1 and basic weights described in the text. (C4 is a constant value caused by leakage power, etc.)

[0115] Mathematical formula 4:

[0116] Figure 12 A block diagram illustrating a power consumption prediction system according to an embodiment of the present disclosure is provided.

[0117] Reference Figure 12 The power consumption prediction system 1000 may include a processor 1100, a memory device 1200, a storage device 1300, a power supply 1500, and an input / output device 1400. Additionally, although not included in... Figure 10 As shown, the power consumption prediction system 1000 may also include multiple ports that can communicate with video cards, sound cards, memory cards, USB devices, etc., or can communicate with other electronic devices.

[0118] As described above, the processor 1100, memory device 1200, storage device 1300, power supply 1500, and input / output device 1400 included in the power consumption prediction system 1000 can perform the operation of the power consumption prediction system according to an embodiment of the technical concept of the present invention. Specifically, in Figures 1 to 11 The operation of the power consumption prediction system 1000 described herein may refer to the operation performed by the processor 1100 based on a computer program containing at least one instruction stored in the memory device 1200 or the storage device 1300.

[0119] Processor 1100 can perform specific calculations or tasks. According to embodiments, processor 1100 may include at least one of a microprocessor, a central processing unit (CPU), a graphics processing unit (GPU), a neural processing unit (NPU), random access memory (RAM), read-only memory (ROM), a system bus, and an application processor. Processor 1100 can communicate with memory device 1200, storage device 1300, and input / output device 1400 via a bus 1600, such as an address bus, control bus, and data bus. According to embodiments, processor 1100 can also be connected to an expansion bus, such as a Peripheral Component Interconnect (PCI) bus.

[0120] Memory device 1200 can store data required for the operation of computing system 1000. For example, memory device 1200 can be embodied as dynamic random access memory (DRAM), mobile dynamic random access memory, static random access memory (SRAM), phase-change random access memory (PRAM), ferroelectric random access memory (FRAM), resistive random access memory (RRAM), and / or magnetic random access memory (MRAM). Storage device 1300 may include solid-state drive, hard disk drive, compact optical disc read-only memory (CD-ROM), etc. Memory device 1200 and storage device 1300 can store and... Figures 1 to 11 The procedures related to the integrated circuit design methods described herein.

[0121] Input / output devices 1400 may include input devices such as a keyboard, keypad, and mouse, and output devices such as a printer and monitor. Power supply 1500 provides the operating voltage required for the operation of the computing system 1000.

[0122] The integrated circuits designed according to the embodiments of the present invention as described above can be embodied in various package forms. For example, at least a portion of an integrated circuit can be constructed using packages on packages (PoP), ball grid arrays (BGAs), chip scale packages (CSPs), plastic leaded chip carriers (PLCCs), plastic dual in-line packages (PDIPs), die in waffle packs, die in wafer form, chip on board (COBs), ceramic dual in-line packages (CERDIPs), plastic metric quad flat packs (PMQFPs), thin quad flat packs (TQFPs), small outline integrated circuits (SOICs), shrink small outline packages (SSOPs), thin small outline packages (TSOPs), and system-in-packages (SIPs). Package assembly including InPackage (SIP), Multi-Chip Package (MCP), Wafer-level Fabricated Package (WFP), and Wafer-Level Processed Stack Package (WSP).

[0123] As described above, exemplary embodiments are disclosed in the accompanying drawings and specification. In this specification, specific terminology is used to describe embodiments, but this is only for illustrating the technical concept of this disclosure and does not limit the technical concept of this disclosure or restrict the scope of this disclosure as set forth in the claims. Therefore, it should be understood that various modifications and equivalent embodiments can be made therein by those skilled in the art. Thus, the true scope of protection of this disclosure must be defined by the technical concept of the claims.

Claims

1. A method of power consumption prediction for predicting power consumption of an integrated circuit, comprising: include: The steps for determining the hierarchical structure of the multiple clock gating units included in the aforementioned integrated circuit; The steps for determining the first clock gating field corresponding to the first clock gating unit and the second clock gating field corresponding to the second clock gating unit located below the first clock gating unit based on the above hierarchical structure; The step of calculating the power consumption of the second clock-gated domain based on the first logic level of the first clock-gated enable signal applied to the first clock-gated unit; as well as The step of predicting the power consumption of the integrated circuit based on the power consumption of the second clock gate domain. The steps for calculating the power consumption of the second clock-gated domain mentioned above include: The steps of determining the first logic level of the first clock gating enable signal and the second logic level of the second clock gating enable signal applied to the second clock gating unit; as well as The step of calculating the power consumption of the second clock gate domain by multiplying the first logic level, the second logic level, and the weight corresponding to the second clock gate domain.

2. The power consumption prediction method according to claim 1, wherein Also includes: The step of calculating the power consumption of the first clock gate domain by multiplying the first logic level and the weight corresponding to the first clock gate domain. as well as The steps for predicting the power consumption of the integrated circuit based on the power consumption of the first clock gate domain and the power consumption of the second clock gate domain.

3. The power consumption prediction method of claim 1, wherein Also includes: The step of determining whether a memory is included in the third clock gating domain corresponding to the third clock gating unit; When the memory is included in the third clock gate domain, the step of determining the weight corresponding to the third clock gate domain based on the operating state of the memory; as well as The step of calculating the power consumption of the third clock gate domain based on the determined weights corresponding to the third clock gate domain and the logic level of the third clock gate enable signal applied to the third clock gate unit.

4. The power consumption prediction method according to claim 3, wherein The steps for determining the weights corresponding to the third clock-gated domain mentioned above include: The steps of acquiring information related to the signals applied to the aforementioned memory; and The step of determining the weights corresponding to the combinations of signals from the aforementioned memory based on a weight lookup table.

5. The power consumption prediction method according to claim 4, characterized in that, The signals applied to the aforementioned memory include chip enable signals and optical enable signals. The aforementioned weight lookup table includes information related to the weights of each combination of the aforementioned chip enable signal and optical enable signal.

6. The power consumption prediction method according to claim 3, characterized in that, The aforementioned third clock-gated domain includes the first memory and the second memory. The steps for calculating the power consumption of the third clock-gated domain mentioned above include: The step of obtaining the first sub-weight from the weight lookup table based on the combination of control signals applied to the first memory during the first clock cycle; The step of obtaining the second sub-weight from the weight lookup table based on the combination of control signals applied to the second memory during the first clock cycle; The step of adding the first sub-weight and the second sub-weight to calculate the first weight corresponding to the third clock gate domain of the first clock cycle; The step of calculating the power consumption corresponding to the third clock gate domain of the first clock cycle based on the first weight; The step of obtaining the third sub-weight from the weight lookup table based on the combination of control signals applied to the first memory during the second clock cycle; The step of obtaining the fourth sub-weight from the weight lookup table based on the combination of control signals applied to the second memory during the second clock cycle; The step of calculating the second weight corresponding to the third clock gate domain of the second clock cycle by adding the third sub-weight and the fourth sub-weight; and The step of calculating the power consumption corresponding to the third clock gate domain of the second clock cycle based on the second weight.

7. The power consumption prediction method according to claim 1, characterized in that, Also includes: The steps for distinguishing independent components that are not included in the clock gating domain of the aforementioned multiple clock gating units; The steps for determining the power consumption of combinational logic in the aforementioned independent components; The step of determining the first constant corresponding to the power consumption mentioned above; The step of determining the second constant related to the element receiving the clock among the aforementioned independent elements; and The steps for predicting the power consumption of the integrated circuit based on the first and second constants mentioned above.

8. A power consumption prediction method characterized by, include: The step of determining whether a memory is included in the clock gating field corresponding to the clock gating unit; When the memory is included in the clock gate domain, the step of determining the weight corresponding to the clock gate domain based on the operating state of the memory; as well as The step of calculating the power consumption of the clock gated region based on the determined weights corresponding to the clock gated region and the logic level of the clock gate enable signal applied to the clock gated unit. The signals applied to the aforementioned memory include chip enable signals and optical enable signals. The steps for determining the weights corresponding to the aforementioned clock-gated domains include: The step of acquiring information related to the combination of signals applied to the aforementioned memory; as well as The step of determining the weights corresponding to the signal combinations of the above-mentioned memory based on a weight lookup table.

9. The power consumption prediction method according to claim 8, characterized in that, The aforementioned weight lookup table includes information related to the weights of each combination of the aforementioned chip enable signal and optical enable signal.

10. The power consumption prediction method according to claim 8, characterized in that, The aforementioned clock-gated domain includes a first memory and a second memory. The steps for calculating the power consumption of the clock-gated domain mentioned above include: The step of obtaining the first sub-weight from the weight lookup table based on the combination of control signals applied to the first memory during the first clock cycle; The step of obtaining the second sub-weight from the weight lookup table based on the combination of control signals applied to the second memory during the first clock cycle; The step of adding the first sub-weight and the second sub-weight to calculate the first weight corresponding to the third clock gate domain of the first clock cycle; The step of calculating the power consumption corresponding to the clock gate domain of the first clock cycle based on the first weight; The step of obtaining the third sub-weight from the weight lookup table based on the combination of control signals applied to the first memory during the second clock cycle; The step of obtaining the fourth sub-weight from the weight lookup table based on the combination of control signals applied to the second memory during the second clock cycle; The step of calculating the second weight corresponding to the third clock gate domain of the second clock cycle by adding the third sub-weight and the fourth sub-weight; and The step of calculating the power consumption corresponding to the clock gate domain of the second clock cycle based on the second weight.

Citation Information

Patent Citations

  • Low-power design method for wireless sensor network core chip

    CN103324268A

  • Method and apparatus for modelling power consumption of integrated circuit

    CN104541173A