Training method and related device for DERR pin of high bandwidth memory

By setting circuit delay on the DERR pin of HBM, correcting the offset between the DERR signal and the sampling clock signal, the problem of sampling error in HBM is solved, and the accuracy and reliability of data transmission are improved.

CN115116489BActive Publication Date: 2024-12-06TENCENT TECHNOLOGY (SHENZHEN) CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202210762703.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-06-30
Publication Date
2024-12-06
Estimated Expiration
2042-06-30

AI Technical Summary

Technical Problem

In high bandwidth memory (HBM), the DERR signal and the sampling clock signal are easily affected by environmental changes and crosstalk between signals, resulting in sampling offsets, and errors occur, affecting the accuracy of data transmission.

Method used

By calculating the appropriate circuit delay on the transmission path represented by the DERR pin, the offset between the DERR signal and the sample clock signal is corrected to achieve sampling alignment and reduce the possibility of sampling errors.

Benefits of technology

Ensure that the HBM host can detect verification errors in a timely manner and process them, reduce errors in storing or reading data, and improve the accuracy and reliability of data transmission.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115116489B_ABST
    Figure CN115116489B_ABST
Patent Text Reader

Abstract

The present application discloses a training method and related device for the DERR pin of a high-bandwidth memory, in which a data word pattern with a check error is sent to a dynamic random access memory. When the check error is determined, a DERR signal can be fed back to the host through the DERR pin. Each bit signal included in the DERR signal is used as a target bit signal, and a first circuit delay of the target bit signal on the transmission path is determined so that the target bit signal satisfies the sampling alignment with the sampling clock signal. Under the first circuit delay, the clock pulses of the sampling alignment of multiple bit signals and the sampling clock signal are checked. According to the check result, a second circuit delay of each bit signal on the transmission path is determined so that the same clock pulse of the multiple bit signals and the sampling clock signal satisfies the sampling alignment, and the training is completed. The trained DERR pin can be used to adjust the offset back, re-achieve sampling alignment, and reduce the possibility of sampling errors.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a training method and related device for a DERR pin of a high-bandwidth memory. Background Art

[0002] As data centers increasingly utilize artificial intelligence and machine learning, large amounts of data are constantly being generated and consumed, which poses a huge challenge to data centers in storing, moving, and analyzing data quickly and efficiently. Data center servers are in urgent need of greater memory bandwidth and capacity, and for this reason, High-Bandwidth Memory (HBM) has emerged to better support the continued development of artificial intelligence and machine learning.

[0003] In the process of storing or reading a data word (Data Word, DWORD), the transmission of DWORD from the HBM host to the HBM dynamic random access memory (Dynamic Random Access Memory, DRAM) is easily affected by environmental changes and crosstalk between signals, which causes the DWORD data signal and sampling clock signal to be offset, resulting in DWORD sampling errors. For this reason, a data word error (DWORD Error, DERR) pin is set on the HBM. When the HBMDRAM determines that a DWORD sampling error occurs, it will promptly feedback a DERR signal that lasts for 1 clock cycle to the HBM Host through the DERR pin, so that the HBM Host can promptly determine that there is a DWORD transmission error, thereby implementing measures such as retransmission or line diagnosis to ensure the correctness of DWORD transmission.

[0004] However, the transmission of the DERR signal through the DERR pin may also be affected by environmental changes and crosstalk between signals, resulting in an offset between the DERR signal and the sampling clock signal, leading to errors in the sampling of the DERR signal. When a DERR signal sampling error occurs, it is very likely to cause problems with mis-sampling and missing DERR signals. As a result, the DERR signal fed back from the HBM DRAM to the HBMHost will not be responded to in a timely manner, which may cause errors in storing or reading data in the HBM DRAM. Summary of the invention

[0005] In order to solve the above technical problems, the present application provides a training method and related devices for the DERR pin of a high-bandwidth memory, which can calculate what size of circuit delay should be set on the transmission path represented by the DERR pin to correct the sampling offset of the DERR signal, so as to set the delay circuit on the transmission path based on the circuit delay (for example, the second circuit delay) obtained in this training, thereby ensuring that when the DERR pin is actually used, the offset between the DERR signal and the sampling clock signal is adjusted back through the delay circuit, and the sampling alignment is re-achieved, reducing the possibility of DERR signal sampling errors, thereby ensuring that the HBM host promptly detects the verification error, promptly performs retransmission or other processing, and reduces the possibility of errors in HBMDRAM storage or reading data.

[0006] The embodiments of the present application disclose the following technical solutions:

[0007] On the one hand, an embodiment of the present application provides a method for training a DERR pin of a high bandwidth memory, the method comprising:

[0008] generating a data word pattern with check errors;

[0009] sending the data word pattern to a dynamic random access memory;

[0010] Transmitting a data word error DERR signal fed back by the dynamic random access memory through the DERR pin, wherein the DERR signal includes a plurality of bit signals;

[0011] Taking each bit signal of the plurality of bit signals as a target bit signal, determining a first circuit delay of the target bit signal on a transmission path represented by the DERR pin, wherein the first circuit delay satisfies sampling alignment between the target bit signal and a sampling clock signal;

[0012] Under the first circuit delay, checking the clock pulses of the plurality of bit signals sample-aligned with the sampling clock signal;

[0013] A second circuit delay of each bit signal on the transmission path is determined according to the verification result to complete the training of the DERR pin, and the second circuit delay enables the multiple bit signals to meet sampling alignment with the same clock pulse of the sampling clock signal.

[0014] On the one hand, an embodiment of the present application provides a training device for a DERR pin of a high bandwidth memory, the device comprising a generating unit, a sending unit, a transmission unit, a determining unit and a checking unit:

[0015] The generating unit is used to generate a data word pattern with a check error;

[0016] The sending unit is used to send the data word pattern to the dynamic random access memory;

[0017] The receiving unit is used to transmit a data word error DERR signal fed back by the dynamic random access memory through the DERR pin, wherein the DERR signal includes a plurality of bit signals;

[0018] The determining unit is configured to use each bit signal of the multiple bit signals as a target bit signal, and determine a first circuit delay of the target bit signal on a transmission path represented by the DERR pin, wherein the first circuit delay satisfies sampling alignment between the target bit signal and the sampling clock signal;

[0019] The verification unit is used to verify the clock pulses of the plurality of bit signals sampled and aligned with the sampling clock signal under the delay of the first circuit;

[0020] The determination unit is further used to determine a second circuit delay of each bit signal on the transmission path according to the verification result to complete the training of the DERR pin, and the second circuit delay enables the multiple bit signals to meet sampling alignment with the same clock pulse of the sampling clock signal respectively.

[0021] In one aspect, an embodiment of the present application provides an electronic device, the electronic device comprising a processor and a memory:

[0022] The memory is used to store program code and transmit the program code to the processor;

[0023] The processor is configured to execute the method described in any one of the preceding aspects according to instructions in the program code.

[0024] In one aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium is used to store program code, and the program code is used to execute the method described in any of the aforementioned aspects.

[0025] In one aspect, an embodiment of the present application provides a computer program product, including a computer program, which implements the method described in any of the aforementioned aspects when executed by a processor.

[0026] It can be seen from the above technical solution that the present application provides a training method for the DERR pin of HBM, and the training process may be to generate a data word pattern with a check error, send the data word pattern to the dynamic random access memory, and when the dynamic random access memory checks the data word pattern to determine the check error, the data word error DERR signal can be fed back to the host of the HBM through the DERR pin. Since the DERR signal may have a sampling offset, it is possible to calculate how to set the circuit delay on the transmission path represented by the DERR pin to correct the sampling offset and achieve sampling alignment. At the same time, since the DERR signal includes multiple bit signals, when calculating how to set the circuit delay, each bit signal in the multiple bit signals can be used as a target bit signal, and the first circuit delay of the target bit signal on the transmission path represented by the DERR pin is determined. The first circuit delay satisfies the sampling alignment between the target bit signal and the sampling clock signal. Then, under the first circuit delay, the clock pulses of the sampling alignment of the multiple bit signals and the sampling clock signal are checked, and then the second circuit delay of each bit signal on the transmission path is determined according to the check result. The second circuit delay makes the multiple bit signals respectively meet the sampling alignment with the same clock pulse of the sampling clock signal, and the training of the DERR pin is completed. The training process calculates that the sampling offset of the DERR signal can be corrected and the sampling alignment can be achieved by setting the second circuit delay on the transmission path represented by the DERR pin. The transmission path represented by the DERR pin used to transmit the DERR signal is fixed, so the delay circuit on the transmission path can be set based on the second circuit delay obtained in this training, so as to ensure that when the DERR pin is actually used, the offset between the DERR signal and the sampling clock signal is adjusted back through the delay circuit, and the sampling alignment is re-achieved, reducing the possibility of sampling errors of the DERR signal, thereby ensuring that the HBM host promptly detects the verification error, promptly retransmits or other processing, and reduces the possibility of errors in storing or reading data in the HBM DRAM. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technical members in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0028] Figure 1a , Figure 1b , Figure 1c The following are example diagrams of the situations where 1 DERR signal, 2 DERR signals and 3 consecutive DERR signals appear, respectively, as provided in the related art;

[0029] Figure 2 An example diagram of the relationship between a DERR signal and a sampling clock signal provided in an embodiment of the present application;

[0030] Figure 3 An application scenario architecture diagram of a method for training a DERR pin of a high bandwidth memory provided in an embodiment of the present application;

[0031] Figure 4 A flowchart of a method for training a DERR pin of a high bandwidth memory provided in an embodiment of the present application;

[0032] Figure 5 A basic structural diagram of a DERR pin training circuit provided in an embodiment of the present application;

[0033] Figure 6a , Figure 6b The following are example timing diagrams that need to be followed for write data operations (i.e., write data word mode) with BL being 2 and BL being 4 provided in the embodiments of the present application;

[0034] Figure 7 An example diagram of increasing or decreasing a circuit delay based on an initial circuit delay provided in an embodiment of the present application;

[0035] Figure 8 An example diagram of a delay circuit provided in an embodiment of the present application;

[0036] Fig. 9 An exemplary diagram of another delay circuit provided in an embodiment of the present application;

[0037] Fig.10 A structural diagram of a training device for a DERR pin of a high bandwidth memory provided in an embodiment of the present application;

[0038] Fig.11 A structural diagram of a terminal provided in an embodiment of the present application;

[0039] Fig.12 A structural diagram of a server provided in an embodiment of the present application. DETAILED DESCRIPTION

[0040] The embodiments of the present application are described below in conjunction with the accompanying drawings.

[0041] During DWORD storage or reading, if a DWORD sampling error occurs, a DERR signal lasting one clock cycle will be fed back to the HBM Host through the DERR pin in a timely manner. Figure 1a-Figure 1c As shown, Figure 1a-Figure 1cThe following diagrams list the situations where one DERR signal, two DERR signals, and three consecutive DERR signals appear. The DERR signal will return to the DERR signal after the data word parity error output delay (DQ parity error output delay, tPARDQ) after the HBM DRAM samples the parity error. Figure 1a-Figure 1c The tPARDQ in the example will take the clock period of two sampling clock signals (the two sampling clock signals can be expressed as CK_c and CK_t), wherein the method of checking the data word can be a parity check, so the check data used can be parity check data (Parity data, PAR).

[0042] exist Figure 1a In the figure, WDQS_c and WDQS_t represent the clock signals for transmitting data words, DATA represents the transmitted data words, and the transmitted data words here can be represented by Da, Da+1, Db, Db+1, Dc, Dc+1, and the check data can be represented by Pa, Pa+1, Pb, Pb+1, Pc, Pc+1. T0, T1, T2, T3, T4, T5 represent different times. DERR represents the returned DERR signal. Figure 1a It can be seen that the Da data word (including Da, Da+1) and Pa check data (including Pa, Pa+1) sent at time T0 have check errors. After a delay of 2 clock cycles of tPARDQ, a DERR signal of 1 clock cycle is returned at time T2.

[0043] exist Figure 1b In the example, if the Da data and Pa check data sent at time T0 have check errors, then after a delay of 2 clock cycles of tPARDQ, a DERR signal of 1 clock cycle is returned at time T2. If the Dc data (including Dc, Dc+1) and Pc check data (including Pc, Pc+1) sent at time T2 have check errors, then after a delay of 2 clock cycles of tPARDQ, a DERR signal of 1 clock cycle is returned at time T4.

[0044] exist Figure 1cIn the example, if the Da data and Pa check data sent at time T0 have check errors, then after a delay of 2 clock cycles of tPARDQ, a DERR signal of 1 clock cycle is returned at time T2. If the Db data (including Db, Db+1) and Pb check data (including Pb, Pb+1) sent at time T1 have check errors, then after a delay of 2 clock cycles of tPARDQ, a DERR signal of 1 clock cycle is returned at time T3. If the Dc data and Pc check data sent at time T2 have check errors, then after a delay of 2 clock cycles of tPARDQ, a DERR signal of 1 clock cycle is returned at time T4.

[0045] When the HBM DRAM returns the DERR signal, the HBM Host will use the rising edge of the CK_t clock to sample the returned DERR signal. To ensure the correct and stable sampling of high-speed signals, in general, the DERR signal needs to meet the sampling alignment with the sampling clock signal. The sampling alignment here, for example, means that the rising edge of CK_t (falling edge of CK_c) needs to be aligned to the center of the DERR signal.

[0046] However, the following situations will cause the DERR signal and the sampling clock signal to deviate:

[0047] (a) The delay offset of the DERR signal transmission path caused by the non-convergence of the back-end timing of HBM;

[0048] (b) HBM production failure causes the shift of DERR signal transmission path delay;

[0049] (c) During HBM operation, the transmission path delay of the DERR signal shifts due to changes in environmental information such as process, voltage and temperature (PVT);

[0050] (d) During HBM operation, the DERR signal and the sampling clock signal are offset due to the influence of crosstalk between signals.

[0051] like Figure 2 As shown, Figure 2 The figure marked by (1) shows that under normal working conditions, the DERR signal and the sampling clock signal meet the sampling alignment, that is, the rising edge of the sampling clock signal (the signal marked by CLK in the figure) and the center position of the DERR signal (the signal marked by Data in the figure) are aligned.

[0052] When the above (a)-(d) situations occur, the DERR signal and the sampling clock signal will appear Figure 2In the figure marked by (2) and (3), the rising edge of the sampling clock signal will drift out of the data window of the DERR signal; in the figure marked by (3), although the rising edge of the sampling clock signal does not drift out of the data window of the DERR signal, the rising edge of the sampling clock signal and the data establishment (rising) or data cancellation (falling) edge of the DERR signal are too close, resulting in a timing violation of the setup timing or hold timing in this sampling. Both of the above situations will lead to sampling errors of the DERR signal, resulting in the loss of data stored in the HBM DRAM.

[0053] To this end, an embodiment of the present application provides a DERR pin training method, which can calculate what size of circuit delay should be set on the transmission path represented by the DERR pin to correct the sampling offset of the DERR signal, so as to set the delay circuit on the transmission path based on the circuit delay (for example, the second circuit delay) obtained in this training, thereby ensuring that when the DERR pin is actually used, the offset between the DERR signal and the sampling clock signal is adjusted back through the delay circuit, and the sampling alignment is re-achieved, thereby reducing the possibility of DERR signal sampling errors, thereby ensuring that the HBM host promptly detects the verification error, promptly performs retransmission or other processing, and reduces the possibility of errors in HBM DRAM storing data or reading data.

[0054] It should be noted that the training method of the DERR pin provided in the embodiment of the present application is mainly used to calculate what size of circuit delay should be set on the transmission path represented by the DERR pin to correct the sampling offset of the DERR signal, so that in the process of using HBM for DWORD storage or reading, the DERR signal can be transmitted through the trained DERR pin, thereby adjusting the offset between the DERR signal and the sampling clock signal, re-aligning the sampling, and reducing the possibility of DERR signal sampling errors. HBM is created to support artificial intelligence, and HBM can be mainly used in the field of artificial intelligence, so the method provided in the embodiment of the present application can involve artificial intelligence.

[0055] Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. Artificial Intelligence is also the study of the design principles and implementation methods of various intelligent machines, so that machines have the functions of perception, reasoning and decision-making.

[0056] Artificial intelligence technology is a comprehensive discipline that covers a wide range of fields, including both hardware-level and software-level technologies. Basic artificial intelligence technologies generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics and other technologies. Artificial intelligence software technology mainly includes computer vision technology, speech processing technology, natural language processing technology, machine learning / deep learning, autonomous driving, smart transportation and other major directions. By training the DERR pins using the method provided in the embodiments of this application, an HBM that can better support artificial intelligence can be obtained.

[0057] It is understandable that HBM can run on electronic devices, and the electronic devices used in the embodiments of the present application can be servers or terminals. The server can be a server that provides various artificial intelligence-based services, and the server can be an independent server or a server in a cluster; the terminal includes but is not limited to mobile phones, computers, intelligent voice interaction devices, smart home appliances, vehicle terminals, aircraft, etc.

[0058] like Figure 3 As shown, Figure 3 An application scenario architecture diagram of a training method for the DERR pin of a high-bandwidth memory is shown. The application scenario may include an electronic device, on which a high-bandwidth memory is provided. In the embodiment of the present application, the electronic device is taken as an example of a server 100, on which a high-bandwidth memory 101 is running. The high-bandwidth memory 101 may include a host 1011 and a dynamic random access memory 1012. The training process may be to generate a data word pattern with a check error through the host 1011, and to send the data word pattern to the dynamic random access memory 1012 through the host 1011. When the dynamic random access memory checks the data word pattern and determines a check error, a data word error DERR signal may be fed back to the host 1011 through the DERR pin.

[0059] Since the DERR signal may have a sampling offset, it is possible to calculate how to set the circuit delay on the transmission path represented by the DERR pin to correct the sampling offset and achieve sampling alignment. At the same time, since the DERR signal includes multiple bit signals, when calculating how to set the circuit delay, each of the multiple bit signals can be used as a target bit signal, and the first circuit delay of the target bit signal on the transmission path represented by the DERR pin is determined. The first circuit delay satisfies the sampling alignment between the target bit signal and the sampling clock signal. Then, under the first circuit delay, the clock pulses of the sampling alignment of the multiple bit signals and the sampling clock signal are checked, and then the second circuit delay of each bit signal on the transmission path is determined according to the check result. The second circuit delay satisfies the sampling alignment of the multiple bit signals with the same clock pulse of the sampling clock signal, and the training of the DERR pin is completed. In this way, the transmitted DERR signal can be offset adjusted through the trained DERR pin, so that the host 1011 receives the sampling aligned DERR signal.

[0060] Next, the training method of the DERR pin of the high bandwidth memory provided by the embodiment of the present application will be described in detail with reference to the accompanying drawings. Figure 4 , Figure 4 A flow chart of a method for training a DERR pin of a high bandwidth memory is shown, the method comprising:

[0061] S401. Generate a data word pattern with a check error.

[0062] In order to train the DERR pin and calculate what size of circuit delay should be set on the transmission path represented by the DERR pin to correct the sampling offset of the DERR signal, the HBM Host can generate a data word pattern (DWORD Pattern) with a check error to simulate the transmission of DWORD, so that when the check error is determined, the DERR signal is sent and the circuit delay of the appropriate size is trained to correct the sampling offset of the DERR signal.

[0063] The following is an introduction to the basic knowledge of DWORD data combination and data verification. Usually, DWORD data mainly includes the related interfaces shown in Table 1:

[0064] Table 1

[0065] Function Data bit width Functional Description DQ 128 bits HBM DRAM write data bus DM 16-bit Data Mask, write data mask signal DBI 16-bit Data Bus Inversion PAR 4-digit Parity data

[0066] As can be seen from Table 1, HBM's DWORD contains three groups of signals in the write direction, namely data (DQ), data mask (DM) signal and data bus inversion (DBI) signal. Among them, DQ is a 128-bit write data signal; DM signal is a 16-bit data mask signal, and the DM signal can be configured as an error correction code (ECC) signal by a register (Mode Register). At this time, the function of the DM signal will no longer be supported, and ECC data will be written into the HBM for error correction and detection; DBI signal is a data bus inversion signal. The corresponding relationship between DM signal, DBI signal and data can be shown in Table 2:

[0067] Table 2

[0068]

[0069]

[0070] As can be seen from Table 2, the data bit width of data (DQ) can be divided into multiple segments, each segment includes 8 bits, such as DQ[7:0], DQ[15:8], DQ[23:16], ... DQ[127:120] in Table 2, and each segment of data has a corresponding DBI and DM. In Table 2, Even Byte means an even byte.

[0071] The DWORD check functions supported by HBM DRAM are shown in Table 3. This table only lists the correspondence between DQ 0 to 31 bits, DBI 0 to 3 bits, DM, and ECC 0 to 3 bits. The remaining DQs have similar check relationships with multiple bit signals of DBI / DM / ECC corresponding to each 32 bits.

[0072] Table 3

[0073]

[0074]

[0075]

[0076] It can be seen from Table 3 that if the ECC usage is disabled, when the XOR result of all bits of DQ[31:0], DBI[3:0] and DM[3:0] is an even number (Even), if PAR is low (low can be represented by L), then DERR output is 0; if PAR is high (high can be represented by H), then DERR output is 1.

[0077] If the ECC usage is disabled, the ECC data does not participate in data verification. When the XOR result of all bits of DQ[31:0] and DBI[3:0] is an even number, if PAR is low, the DERR output is 0; if PAR is high, the DERR output is 1.

[0078] In addition, if the use status of DBI is disabled, the DBI signal does not participate in data verification. If the use status of ECC is enabled or the use status of DM is disabled, the DM signal does not participate in data verification.

[0079] From the above introduction, it can be known that DWORD includes data, data mask signal and data bus flip signal, and the data word mode is a generated analog DWORD, so the data word mode can include data, data mask signal and data bus flip signal. In addition, whether a verification error occurs after data verification is related to the usage of data, data mask signal and data bus flip signal, and the usage of data, data mask signal and data bus flip signal can be reflected by data word configuration information. In this case, the implementation method of generating a data word mode with a verification error can be to obtain data word configuration information, and the data word configuration information is used to indicate the usage of data, data mask signal and data bus flip signal, and then generate a data word mode with a verification error according to the usage of data, data mask signal and data bus flip signal. Among them, the data word configuration information can be set by a register.

[0080] Since the purpose of the embodiment of the present application is to generate a data word pattern with a check error, and the occurrence of the check error requires certain usage conditions of data, data mask signals, and data bus flip signals. Since the DQ, DM signals, and DBI signals of HBM are completed in an even check manner, the result of the bit XOR of the remaining data word patterns except the PAR bit and the result generated after the XOR of the PAR bit is 1, a data word pattern with a check error will be generated at this time. Based on the summary of the check relationship shown in Table 3 above, in a possible implementation, the usage conditions based on which the data word pattern with a check error is generated may include any of the following:

[0081] Data, data mask signal and data bus flip signal are all enabled;

[0082] Both the data and data bus flip signals are enabled, and the data mask signal is disabled;

[0083] Data is enabled, and both the data mask signal and the data bus inversion signal are disabled.

[0084] This embodiment takes the case where the check error is a PAR check error as an example to introduce the above three usage scenarios respectively. When the DQ, DM and DBI signals are all enabled and there is a PAR check error:

[0085] In this data word mode, DQ, DM signal, and DBI signal are all enabled and use the sampling clock signal, that is, the pattern of 101010. Then DQ is 0xAAAA_AAAA, DM signal is 0xA, DBI signal is 0xA, and PAR is 1. At this time, the XOR result of DWORD is 0, and PAR is 1. This means that the data word mode has a check error and DERR will be returned as 1.

[0086] When the DQ and DBI signals are enabled, the DM signal is disabled, and the PAR check error occurs:

[0087] In this data word mode, both DQ and DBI signals are enabled and the sampling clock signal is used, that is, the pattern of 101010. Then DQ is 0xAAAA_AAAA, the DBI signal is 0xA, and PAR is 1. At this time, the XOR result of DWORD is 0, and PAR is 1. In this case, the data word mode has a check error, and DERR will be returned as 1.

[0088] When DQ is enabled, DM and DBI signals are disabled, and PAR check error occurs:

[0089] In this data word mode, DQ is enabled and the sampling clock signal is used, that is, the pattern of 101010. Then DQ is 0xAAAA_AAAA. At this time, the XOR result of DWORD is 0, and PAR is 1. Then the data word pattern has a check error, and DERR will be returned as 1.

[0090] It should be noted that in the embodiment of the present application, the above three usage scenarios can generate data word patterns with check errors for training. The specific usage scenario can be configured through the data word configuration information of the register, and then the corresponding data word pattern is generated according to the data word configuration information of the register accessible by the software to complete the training process of the DERR pin. In a possible implementation, the usage scenario configured in the data word configuration information defaults to the first scenario.

[0091] It should be noted that the method provided in the embodiment of the present application can be implemented based on the circuit of DERR pin training. The basic structure diagram of the circuit of DERR pin training can be found in Figure 5 As shown, it may include a host 501 and a dynamic random access memory 502, the host 501 may include a pattern generating module 5011, a data sending module 5012, a delay control circuit 5013 and a DERR receiving unit 5014, and the dynamic random access memory 502 may include a data receiving module 5021, a pattern parsing module 5022, a check error detection module 5023 and a DERR sending unit 5024. Among them, Figure 5This is only an example of a basic circuit structure diagram and does not constitute a limitation of the present application. Figure 5 The circuit for DERR pin training shown implements the present method, and S401 can be generated by the pattern generation module 5011 .

[0092] S402: Send the data word pattern to a dynamic random access memory.

[0093] S403 . Transmitting a data word error DERR signal fed back by the dynamic random access memory through the DERR pin, wherein the DERR signal includes a plurality of bit signals.

[0094] After the HBM Host generates a data word pattern with a check error, the data word pattern can be sent to a dynamic random access memory (HBM DRAM). S402 can be performed by Figure 5 The data transmission module 5012 shown in FIG. 5 is executed. In one possible case, the data transmission module 5012 can automatically detect the burst length (BL) of the data word pattern to be transmitted, so as to transmit the data word pattern in accordance with the timing. Figure 6a and Figure 6b , Figure 6a and Figure 6b The timings that need to be followed for write data operations (ie, write data word modes) when BL is 2 and BL is 4 are described respectively. Figure 6a and Figure 6b It only contains DATA, that is, DQ. If it contains DBI signal and DM signal, the DBI signal and DM signal comply with the same timing requirements as DQ.

[0095] exist Figure 6a In the figure, WDQS_c and WDQS_t represent the clock signals for transmitting data words, DATA represents the transmitted data words (i.e., the data word mode generated in S401), and the transmitted data words here can be represented by Da and Da+1. T0, T1, T2, T3, T4, T5, T6, and T7 represent different times. The command column is used to provide data operation instructions. Figure 6a In the command column, the data operation instruction provided is "WRITE", i.e., the instruction for writing data operation; the address column is used to provide the address instruction, which is used to indicate the address in the HBM targeted by the data operation. When the data operation instruction is "WRITE" and the address instruction is "BAx CAa" and "CAa", it indicates writing data word mode to the space identified by "BAx CAa" and "CAa"; WL indicates write delay, Figure 6a In WL=4. Figure 6a Indicates the write data operation when BL is 2. It can be seen that DATA only lasts for two records, Da and Da+1. Figure 6bIndicates the write data operation when BL is 4. It can be seen that DATA only lasts for four entries: Da, Da+1, Da+2, and Da+3.

[0096] After receiving the data word pattern, the HBM DRAM can parse the data word pattern and perform data verification based on the parsed data word and verification data. If the verification is correct, DERR is returned as 0, and if the verification is wrong, DERR is returned as 1.

[0097] In one possible implementation, Figure 5 The data receiving module 5021 receives a data word pattern with a check error, parses the data word pattern through the pattern parsing module 5022, and uses the check error detection module 5023 to check the data word and the check data to determine the check error.

[0098] After the check error is determined, a DERR signal may be generated, and the DERR signal may include a plurality of bit signals, such as 2 or 4 bit signals.

[0099] Multiple bit signals may be divided according to the data bit width. For example, the DERR signal includes 4 bit signals. The DERR signal may be shown in Table 4:

[0100] Table 4

[0101] Function Data bit width Functional Description DERR 4-digit DWORD Error, data error signal

[0102] In Table 4, the data bit width of the DERR signal is 4 bits, so the DERR signal can include a 4-bit signal. When a check error occurs, the corresponding bit of DERR will output 1. At this time, the corresponding relationship between DERR and data is:

[0103] DQ[31:0], DBI[3:0], DM[3:0] correspond to DERR0;

[0104] DQ[63:32], DBI[7:4], DM[7:4] correspond to DERR1;

[0105] DQ[95:64], DBI[11:8], DM[11:8] correspond to DERR2;

[0106] DQ[127:96], DBI[15:12], DM[15:12] correspond to DERR3.

[0107] After receiving the DERR signal, the DERR signal may be fed back to the HBM Host via the DERR pin so that the HBM Host receives the DERR signal. In a possible implementation, the HBM DRAM may feed back the DERR signal to the HBM Host via the DERR sending unit 5024 .

[0108] S404: Take each bit signal of the multiple bit signals as a target bit signal, and determine a first circuit delay of the target bit signal on the transmission path represented by the DERR pin, wherein the first circuit delay satisfies sampling alignment between the target bit signal and the sampling clock signal.

[0109] Since the DERR signal may have a sampling offset, it is possible to calculate how to set the circuit delay on the transmission path represented by the DERR pin to correct the sampling offset and achieve sampling alignment. At the same time, since the DERR signal includes multiple bit signals, when calculating how to set the circuit delay, each bit signal in the multiple bit signals can be used as a target bit signal, and the first circuit delay of the target bit signal on the transmission path represented by the DERR pin is determined. The first circuit delay satisfies the sampling alignment between the target bit signal and the sampling clock signal.

[0110] In a possible implementation, the target bit signal and the sampling clock signal meet the sampling alignment, indicating that the rising edge (falling edge) of the sampling clock signal is aligned with the center position of the target bit signal. To this end, the circuit delay between the rising edge (falling edge) of the sampling clock signal and the rising edge and falling edge of the target bit signal can be found, so as to determine how to set the circuit delay so that the rising edge (falling edge) of the sampling clock signal is aligned with the center position of the target bit signal. In this case, the first circuit delay of the target bit signal on the transmission path represented by the DERR pin can be determined by obtaining the initial circuit delay of the target bit signal on the transmission path, and increasing the circuit delay on the basis of the initial circuit delay until the value of the collected target bit signal is 0, thereby obtaining the third circuit delay. And the circuit delay is reduced on the basis of the initial circuit delay until the value of the collected target bit signal is 0, thereby obtaining the fourth circuit delay. Then, the third circuit delay and the fourth circuit delay are averaged to obtain the first circuit delay.

[0111] During the training phase of the DERR pin, such as Figure 7As shown. The initial state of the DERR signal is phase 0. At this time, the corresponding circuit delay is the initial circuit delay, and the value of the sampled target bit signal should be 1. Adding circuit delay on the basis of the initial circuit delay, eventually, the value of the target bit signal read back should be 0, and the third circuit delay is obtained, recorded as R_CNT, and the state of the target bit signal is Phase 1.

[0112] Afterwards, the initial state of Phase1 to Phase0 is set, and the circuit delay is reduced based on the initial circuit delay. Eventually, the value of the target bit signal read back should be 0, and the fourth circuit delay is obtained, recorded as L_CNT. At this time, the state of the target bit signal is Phase2.

[0113] Then, the third circuit delay and the fourth circuit delay can be averaged, and when the second circuit delay of the target bit signal is calculated to be (R_CNT+L_CNT) / 2, the target bit signal and the sampling clock signal are sampled aligned. Next, the circuit delay of each bit signal is adjusted and calculated according to the above steps.

[0114] It should be noted that the above-mentioned circuit delays (such as the first circuit delay, the third circuit delay, and the fourth circuit delay) can be implemented by a delay circuit. Therefore, in the embodiment of the present application, a delay circuit corresponding to the target bit signal can be set on the transmission path represented by the DERR pin, that is, each bit signal included in the DERR signal has a corresponding delay circuit. The delay circuit may include multiple delay units. Figure 8 As shown, Figure 8 Taking the example that the DERR signal includes 4 bit signals (DERR 0, DERR 1, DERR 2 and DERR 3), the delay circuit is introduced. Figure 8 As can be seen in FIG. 1 , delay circuits are set for transmission paths corresponding to DERR 0, DERR 1, DERR 2 and DERR 3, respectively, and each delay circuit includes a plurality of delay elements (DE). Each delay element may include four inverter circuits.

[0115] In addition, each delay unit has a corresponding tap interface, and the delay circuit is controlled by a step counter (StepCounter). The step counter is used to control the target bit signal to be connected from the target tap interface. After the target bit signal is connected from the target tap interface, the number of delay units that the target bit signal passes through before the target tap determines the size of the circuit delay of the target bit signal on the transmission path. Therefore, by controlling the target bit signal to be connected from a certain tap interface through the step counter, the circuit delay of the DERR signal can be adjusted.

[0116] See also Fig. 9 As shown, there is a tap interface after each DE. The tap interface can be found in Fig. 9 As shown by the diagonal arrow in the dashed box. If the step counter controls the target bit signal from Fig. 9 The tap interface corresponding to the third DE is connected, which means that the circuit delay of the target bit signal on the transmission path is the circuit delay formed by three delay units.

[0117] Based on the delay circuit set as above, the circuit delay is increased on the basis of the initial circuit delay until the value of the collected target bit signal is 0. The third circuit delay can be obtained by increasing the circuit delay by adjusting the value of the step counter on the basis of the initial circuit delay until the value of the collected target bit signal becomes 0, thereby obtaining the third circuit delay. In one possible implementation, the value of the step counter represents the number of delay units that the target bit signal passes through before the target tap, so the third circuit delay is the first number of delay units that the target bit signal passes through. Taking the aforementioned third circuit delay recorded as R_CNT as an example, the value of the step counter can be recorded as R_CNT.

[0118] Accordingly, the circuit delay is reduced on the basis of the initial circuit delay until the value of the target bit signal collected is 0. The fourth circuit delay can be obtained by reducing the circuit delay by adjusting the value of the step counter on the basis of the initial circuit delay until the value of the target bit signal collected becomes 0, thereby obtaining the fourth circuit delay. In one possible implementation, the value of the step counter represents the number of delay units that the target bit signal passes through before the target tap, so the fourth circuit delay is the second number of delay units that the target bit signal passes through. Taking the aforementioned fourth circuit delay recorded as L_CNT as an example, the value of the step counter can be recorded as L_CNT.

[0119] In this case, the third circuit delay and the fourth circuit delay are averaged to obtain the first circuit delay in a manner that the first number and the second number are averaged to obtain the second circuit delay.

[0120] It should be noted that if the transmission path of the DERR pin has a large initial circuit delay in extreme cases, the initial state is not Phase 0 but Figure 2In the sampling situation shown in (2), the DERR signal cannot be sampled to 1 initially, so the offset adjustment cannot be performed. In this case, in order to sample a valid DERR signal and facilitate the offset adjustment, the method for obtaining the initial circuit delay of the target bit signal on the transmission path can be to obtain the initial sub-circuit delay of the target bit signal on the transmission path, and compare the initial sub-circuit delay with the preset circuit delay threshold. If the comparison result is that the initial sub-circuit delay is greater than the preset circuit delay threshold, it means that the initial sub-circuit delay of the transmission path of the DERR pin is large, which may cause the initial state to be not Phase 0. For this reason, the operating frequency of the host that generates the data word pattern can be reduced until the adjusted initial sub-circuit delay is less than or equal to the preset circuit delay threshold, and the adjusted initial sub-circuit delay is determined as the initial circuit delay.

[0121] Normally, the clock phase-locked loop circuit can support the adjustment of the host's operating frequency within the range of 100MHz to 1800MHz. According to the actual situation, usually, the worst case is to reduce the operating frequency to 300MHz, and the DERR value of 1 can be sampled.

[0122] By lowering the operating frequency of the host, the time during which the DERR signal remains at a high level can be extended, thereby ensuring that a value of 1 of the DERR signal can be collected, thereby sampling a valid DERR signal, which facilitates offset adjustment.

[0123] S405 . Under the first circuit delay, verify the clock pulses of the plurality of bit signals that are sample-aligned with the sampling clock signal.

[0124] S406. Determine a second circuit delay of each bit signal on the transmission path according to the verification result to complete the training of the DERR pin, wherein the second circuit delay enables the multiple bit signals to meet sampling alignment with the same clock pulse of the sampling clock signal.

[0125] Since the DERR signal includes multiple bit signals, each bit signal can be sampled and aligned with the sampling clock signal through S401-S404, but it does not mean that multiple bit signals are aligned with the same clock pulse of the sampling clock signal, so it is necessary to verify the clock pulses of the sampling alignment of the multiple bit signals and the sampling clock signal under the first circuit delay. Then, the second circuit delay of each bit signal on the transmission path is determined according to the verification result, so that the multiple bit signals are aligned with the same clock pulse, thereby achieving the sampling alignment of the entire DERR signal and the sampling clock signal, and completing the training of the DERR pin.

[0126] The embodiment of the present application can train the DERR pin before the initial operation of the HBM, so that when the HBM is initially operating, if the back-end timing does not converge and a production failure of the HBM causes a DWORD sampling error, the Host can detect such an error by sampling the DERR signal and promptly take retransmission, reset, or adjust the DWORD line.

[0127] Typically, under the first circuit delay, checking the clock pulses of the sampling alignment of the multiple bit signals with the sampling clock signal is actually checking whether the clock pulses of the sampling alignment of the multiple bit signals with the sampling clock signal are the same. If the check result indicates that the clock pulses of the sampling alignment of the multiple bit signals with the sampling clock signal are not the same, the first circuit delay is further adjusted until the multiple bit signals respectively meet the sampling alignment with the same clock pulse of the sampling clock signal, and the second circuit delay is obtained. If the check result indicates that the clock pulses of the sampling alignment of the multiple bit signals with the sampling clock signal are the same, the first circuit delay is determined as the second circuit delay of the corresponding bit signal.

[0128] Through the above method, multiple bit signals included in the DERR signal can be aligned with the same clock pulse, thereby achieving sampling alignment of the entire DERR signal with the sampling clock signal, thereby improving the accuracy of DERR pin training.

[0129] It should be noted that the training process of S404 to S406 mainly involves adjusting the circuit delay on the transmission path, so as to determine the final second circuit delay by increasing or decreasing the circuit delay on the transmission path. Figure 5 The delay control circuit 5013 in the transmission path can support dynamic adjustment of the circuit delay on the transmission path.

[0130] In the embodiment of the present application, in addition to the automatic method implemented by the hardware delay circuit, the entire training process of the DERR pin can be completed by using the software configuration register method, that is, the value of the Step counter can be manually configured using the software configured register, and the data word mode sent by the HBM Host can be configured to read and verify the data word mode received by the HBM DRAM. Therefore, the software can independently initiate each step of the single-step training of the DERR pin.

[0131] The method provided in the embodiment of the present application can configure the HBM to train the DERR pin by software when the HBM detects that the PVT has obvious drift, thereby ensuring that the HBM will not cause HBM DERR signal sampling errors due to the drift of the PVT.

[0132] In addition, since the embodiment of the present application supports single-step training configured by software, the DERR pin training of the entire HBM can also be completed by software configuration alone. Since the software can complete this action when the HBM is not busy, it can ensure that the DERR pin will not have sampling errors while ensuring the efficiency of the entire HBM.

[0133] It can be seen from the above technical solution that the present application provides a training method for the DERR pin of HBM, and the training process may be to generate a data word pattern with a check error, send the data word pattern to the dynamic random access memory, and when the dynamic random access memory checks the data word pattern to determine the check error, the data word error DERR signal can be fed back to the host of the HBM through the DERR pin. Since the DERR signal may have a sampling offset, it is possible to calculate how to set the circuit delay on the transmission path represented by the DERR pin to correct the sampling offset and achieve sampling alignment. At the same time, since the DERR signal includes multiple bit signals, when calculating how to set the circuit delay, each bit signal in the multiple bit signals can be used as a target bit signal, and the first circuit delay of the target bit signal on the transmission path represented by the DERR pin is determined. The first circuit delay satisfies the sampling alignment between the target bit signal and the sampling clock signal. Then, under the first circuit delay, the clock pulses of the sampling alignment of the multiple bit signals and the sampling clock signal are checked, and then the second circuit delay of each bit signal on the transmission path is determined according to the check result. The second circuit delay makes the multiple bit signals respectively meet the sampling alignment with the same clock pulse of the sampling clock signal, and the training of the DERR pin is completed. The training process calculates that the sampling offset of the DERR signal can be corrected and the sampling alignment can be achieved by setting the second circuit delay on the transmission path represented by the DERR pin. The transmission path represented by the DERR pin used to transmit the DERR signal is fixed, so the delay circuit on the transmission path can be set based on the second circuit delay obtained in this training, so as to ensure that when the DERR pin is actually used, the offset between the DERR signal and the sampling clock signal is adjusted back through the delay circuit, and the sampling alignment is re-achieved, reducing the possibility of sampling errors of the DERR signal, thereby ensuring that the HBM host promptly detects the verification error, promptly retransmits or other processing, and reduces the possibility of errors in storing or reading data in the HBM DRAM.

[0134] After completing the training of the DERR pin, the target delay circuit on the corresponding transmission path can be configured according to the second circuit delay of each bit signal on the corresponding transmission path, so that after the actual DWORD transmission, when the target DERR signal is generated and fed back to the HBM Host, the target delay circuit can be used to offset the bit signal in the transmitted target DERR signal, so that the HBM Host receives a sample-aligned DERR signal. This can reduce the possibility of sampling errors in the target DERR signal, thereby ensuring that the HBM host detects the verification error in a timely manner, retransmits or performs other processing in a timely manner, and reduces the possibility of errors in storing or reading data in the HBM DRAM.

[0135] In the embodiment of the present application, Figure 5 The DERR receiving unit 5014 receives the target DERR signal, and the received target DERR signal is a signal that has been adjusted by the delay circuit to achieve sampling alignment.

[0136] If the transmission path is set Figure 8 , Fig. 9 In the delay circuit shown, the second circuit delay is the number of delay units that the bit signal passes through. Then, according to the second circuit delay of each bit signal on the corresponding transmission path, the target delay circuit on the corresponding transmission path can be configured by adjusting the value of the step counter in the target delay circuit to the number of delay units that the bit signal passes through. In this way, the difficulty of configuring the target delay circuit can be reduced, and the configuration is convenient and flexible.

[0137] It is understandable that after completing a DERR pin training, the target delay circuit can be configured according to the second circuit delay obtained in the training, so that when the DERR pin is actually used, the target delay circuit can be used to offset adjust the transmitted DERR signal to achieve sampling alignment. However, due to the influence of various factors (such as environmental information, crosstalk between signals), and various factors may be non-fixed and changing, in order to avoid the target delay circuit obtained based on the second circuit delay configuration obtained in the current training being difficult to adapt to changes in various factors, the embodiment of the present application can train the DERR pin multiple times throughout the life cycle of the DERR pin to update the target delay circuit, so that the target delay circuit can have a better offset adjustment effect at different times, further reducing the possibility of DERR signal sampling errors.

[0138] In the embodiments of the present application, the training of the DERR pin can be triggered in different ways, for example, the DERR pin is trained according to a preset period, that is, a data word pattern with a checksum error is generated according to a preset period to train the DERR pin. For another example, when it is determined that the degree of change of the environmental information exceeds the preset change threshold, it can be said that the environmental information has changed significantly, and the current target delay circuit may not be able to ensure that the DERR signal and the sampling clock signal meet the sampling alignment, then the training of the DERR pin can be triggered, that is, when it is determined that the degree of change of the environmental information exceeds the preset change threshold, a data word pattern with a checksum error is generated to train the DERR pin.

[0139] The method provided in the embodiment of the present application can use the periodic training mechanism included in the hardware to regularly train the DERR pin of the HBM to ensure that the HBM will not have sampling errors caused by the offset of the DERR signal during operation.

[0140] It should be noted that, based on the implementation methods provided in the above aspects, this application can also be further combined to provide more implementation methods.

[0141] based on Figure 4 Corresponding to the training method for the DERR pin of the high bandwidth memory provided in the embodiment, the embodiment of the present application also provides a training device 1000 for the DERR pin of the high bandwidth memory. Fig.10 The training device 1000 for the DERR pin of the high bandwidth memory includes a generating unit 1001, a sending unit 1002, a transmission unit 1003, a determining unit 1004 and a checking unit 1005:

[0142] The generating unit 1001 is used to generate a data word pattern with a check error;

[0143] The sending unit 1002 is used to send the data word pattern to the dynamic random access memory;

[0144] The transmission unit 1003 is used to transmit the data word error DERR signal fed back by the dynamic random access memory through the DERR pin, wherein the DERR signal includes a plurality of bit signals;

[0145] The determining unit 1004 is configured to use each bit signal of the multiple bit signals as a target bit signal, and determine a first circuit delay of the target bit signal on a transmission path represented by the DERR pin, wherein the first circuit delay satisfies sampling alignment between the target bit signal and the sampling clock signal;

[0146] The verification unit 1005 is used to verify the clock pulses of the plurality of bit signals that are sample-aligned with the sampling clock signal under the first circuit delay;

[0147] The determination unit 1004 is further used to determine a second circuit delay of each bit signal on the transmission path according to the verification result to complete the training of the DERR pin, and the second circuit delay enables the multiple bit signals to meet sampling alignment with the same clock pulse of the sampling clock signal respectively.

[0148] In a possible implementation manner, the determining unit 1004 is specifically configured to:

[0149] Acquire an initial circuit delay of the target bit signal on the transmission path;

[0150] Adding a circuit delay on the basis of the initial circuit delay until the value of the collected target bit signal is 0, thereby obtaining a third circuit delay;

[0151] reducing the circuit delay on the basis of the initial circuit delay until the value of the collected target bit signal is 0, thereby obtaining a fourth circuit delay;

[0152] The third circuit delay and the fourth circuit delay are averaged to obtain the first circuit delay.

[0153] In one possible implementation, a delay circuit corresponding to the target bit signal is provided on the transmission path, the delay circuit includes a plurality of delay units, each delay unit is followed by a corresponding tap interface, the delay circuit is controlled by a step counter, the step counter is used to control the target bit signal to be connected from the target tap interface, after the target bit signal is connected from the target tap interface, the number of delay units through which the target bit signal passes before the target tap determines the size of the circuit delay of the target bit signal on the transmission path.

[0154] In a possible implementation manner, the determining unit 1004 is specifically configured to:

[0155] On the basis of the initial circuit delay, the circuit delay is increased by adjusting the value of the step counter until the value of the collected target bit signal becomes 0, thereby obtaining the third circuit delay, wherein the value of the step counter represents the number of delay units that the target bit signal passes through before the target tap, and the third circuit delay is the first number of delay units that the target bit signal passes through;

[0156] On the basis of the initial circuit delay, the circuit delay is reduced by adjusting the value of the step counter until the value of the collected target bit signal becomes 0, thereby obtaining the fourth circuit delay, wherein the fourth circuit delay is the second number of delay units passed by the target bit signal;

[0157] The first number and the second number are averaged to obtain the first circuit delay.

[0158] In a possible implementation manner, the determining unit 1004 is specifically configured to:

[0159] Acquire an initial sub-circuit delay of the target bit signal on the transmission path;

[0160] comparing the initial subcircuit delay with a preset circuit delay threshold;

[0161] If the comparison result is that the initial sub-circuit delay is greater than the preset circuit delay threshold, the operating frequency of the host that generates the data word pattern is reduced until the adjusted initial sub-circuit delay is less than or equal to the preset circuit delay threshold, and the adjusted initial sub-circuit delay is determined as the initial circuit delay.

[0162] In a possible implementation, the verification unit 1005 is specifically configured to:

[0163] Checking whether the clock pulses of the plurality of bit signals and the sampling clock signal sampling alignment are the same;

[0164] The determining unit 1004 is specifically configured to:

[0165] If the verification result indicates that the clock pulses of the sampling alignment of the multiple bit signals and the sampling clock signal are different, the first circuit delay is adjusted until the multiple bit signals respectively meet the sampling alignment with the same clock pulse of the sampling clock signal, thereby obtaining the second circuit delay;

[0166] If the verification result indicates that the plurality of bit signals are sampled and aligned with the sampling clock signal at the same clock pulse, the first circuit delay is determined as the second circuit delay of the corresponding bit signal.

[0167] In a possible implementation, the device further includes a configuration unit and an adjustment unit:

[0168] The configuration unit is used to configure a target delay circuit on the corresponding transmission path according to the second circuit delay of each bit signal on the corresponding transmission path;

[0169] The adjustment unit is used to perform offset adjustment on the bit signal in the transmitted target DERR signal through the target delay circuit.

[0170] In a possible implementation manner, if the second circuit delay is the number of delay units passed by the bit signal, the configuration unit is specifically configured to:

[0171] The value of the step counter in the target delay circuit is adjusted to the number of delay units passed by the bit signal.

[0172] In a possible implementation, the generating unit 1001 is specifically configured to:

[0173] generating a data word pattern with a check error according to a preset period;

[0174] Alternatively, if it is determined that the degree of change in the environmental information exceeds a preset change threshold, a data word pattern with a check error is generated.

[0175] In a possible implementation, the data word pattern includes data, a data mask signal, and a data bus flip signal, and the generating unit 1001 is specifically configured to:

[0176] Acquire data word configuration information, where the data word configuration information is used to indicate usage of the data, the data mask signal, and the data bus flip signal;

[0177] A data word pattern with a check error is generated according to the use of the data, the data mask signal and the data bus inversion signal.

[0178] In a possible implementation, the usage scenario includes any one of the following:

[0179] The data, the data mask signal and the data bus flip signal are all enabled;

[0180] The data and the data bus flip signal are both enabled, and the data mask signal is disabled;

[0181] The data is enabled, and the data mask signal and the data bus flip signal are both disabled.

[0182] It can be seen from the above technical solution that the present application provides a training method for the DERR pin of HBM, and the training process may be to generate a data word pattern with a check error, send the data word pattern to the dynamic random access memory, and when the dynamic random access memory checks the data word pattern to determine the check error, the data word error DERR signal can be fed back to the host of the HBM through the DERR pin. Since the DERR signal may have a sampling offset, it is possible to calculate how to set the circuit delay on the transmission path represented by the DERR pin to correct the sampling offset and achieve sampling alignment. At the same time, since the DERR signal includes multiple bit signals, when calculating how to set the circuit delay, each bit signal in the multiple bit signals can be used as a target bit signal, and the first circuit delay of the target bit signal on the transmission path represented by the DERR pin is determined. The first circuit delay satisfies the sampling alignment between the target bit signal and the sampling clock signal. Then, under the first circuit delay, the clock pulses of the sampling alignment of the multiple bit signals and the sampling clock signal are checked, and then the second circuit delay of each bit signal on the transmission path is determined according to the check result. The second circuit delay makes the multiple bit signals respectively meet the sampling alignment with the same clock pulse of the sampling clock signal, and the training of the DERR pin is completed. The training process calculates that the sampling offset of the DERR signal can be corrected and the sampling alignment can be achieved by setting the second circuit delay on the transmission path represented by the DERR pin. The transmission path represented by the DERR pin used to transmit the DERR signal is fixed, so the delay circuit on the transmission path can be set based on the second circuit delay obtained in this training, so as to ensure that when the DERR pin is actually used, the offset between the DERR signal and the sampling clock signal is adjusted back through the delay circuit, and the sampling alignment is re-achieved, reducing the possibility of sampling errors of the DERR signal, thereby ensuring that the HBM host promptly detects the verification error, promptly retransmits or other processing, and reduces the possibility of errors in storing or reading data in the HBM DRAM.

[0183] The embodiment of the present application also provides an electronic device, which can execute a training method for a DERR pin of a high bandwidth memory. The electronic device can be, for example, a terminal, and taking a smart phone as an example:

[0184] Fig.11 The block diagram shows a partial structure of a smart phone provided in an embodiment of the present application. Fig.11The smartphone includes: a radio frequency (RF) circuit 1110, a memory 1120, an input unit 1130, a display unit 1140, a sensor 1150, an audio circuit 1160, a wireless fidelity (WiFi) module 1170, a processor 1180, and a power supply 1190. The input unit 1130 may include a touch panel 1131 and other input devices 1132, the display unit 1140 may include a display panel 1141, and the audio circuit 1160 may include a speaker 1161 and a microphone 1162. It is understood that Fig.11 The structure of the smartphone shown in the figure does not constitute a limitation of the smartphone, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0185] The memory 1120 can be used to store software programs and modules. The processor 1180 executes various functional applications and data processing of the smartphone by running the software programs and modules stored in the memory 1120. The memory 1120 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area may store data created according to the use of the smartphone (such as audio data, a phone book, etc.), etc. In addition, the memory 1120 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0186] The processor 1180 is the control center of the smartphone, which uses various interfaces and lines to connect various parts of the entire smartphone, and executes various functions of the smartphone and processes data by running or executing software programs and / or modules stored in the memory 1120, and calling data stored in the memory 1120. Optionally, the processor 1180 may include one or more processing units; preferably, the processor 1180 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the processor 1180.

[0187] In this embodiment, the processor 1180 in the smart phone may perform the following steps:

[0188] generating a data word pattern with check errors;

[0189] sending the data word pattern to a dynamic random access memory;

[0190] Transmitting a data word error DERR signal fed back by the dynamic random access memory through the DERR pin, wherein the DERR signal includes a plurality of bit signals;

[0191] Taking each bit signal of the plurality of bit signals as a target bit signal, determining a first circuit delay of the target bit signal on a transmission path represented by the DERR pin, wherein the first circuit delay satisfies sampling alignment between the target bit signal and a sampling clock signal;

[0192] Under the first circuit delay, checking the clock pulses of the plurality of bit signals sample-aligned with the sampling clock signal;

[0193] A second circuit delay of each bit signal on the transmission path is determined according to the verification result to complete the training of the DERR pin, and the second circuit delay enables the multiple bit signals to meet sampling alignment with the same clock pulse of the sampling clock signal.

[0194] The electronic device provided in the embodiment of the present application may also be a server. Fig.12 As shown, Fig.12 The structural diagram of the server 1200 provided in the embodiment of the present application, the server 1200 may have relatively large differences due to different configurations or performances, and may include one or more central processing units (CPU) 1222 (for example, one or more processors) and memory 1232, one or more storage media 1230 (for example, one or more mass storage devices) storing application programs 1242 or data 1244. Among them, the memory 1232 and the storage medium 1230 can be short-term storage or persistent storage. The program stored in the storage medium 1230 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the server. Furthermore, the central processing unit 1222 can be configured to communicate with the storage medium 1230 and execute a series of instruction operations in the storage medium 1230 on the server 1200.

[0195] The server 1200 may also include one or more power supplies 1226, one or more wired or wireless network interfaces 1250, one or more input and output interfaces 1258, and / or one or more operating systems 1241, such as Windows Server 2000. TM , Mac OS X TM , Unix TM ,Linux TM , FreeBSD TM etc.

[0196] In this embodiment, the central processor 1222 in the server 1200 may perform the following steps:

[0197] generating a data word pattern with check errors;

[0198] sending the data word pattern to a dynamic random access memory;

[0199] Transmitting a data word error DERR signal fed back by the dynamic random access memory through the DERR pin, wherein the DERR signal includes a plurality of bit signals;

[0200] Taking each bit signal of the plurality of bit signals as a target bit signal, determining a first circuit delay of the target bit signal on a transmission path represented by the DERR pin, wherein the first circuit delay satisfies sampling alignment between the target bit signal and a sampling clock signal;

[0201] Under the first circuit delay, checking the clock pulses of the plurality of bit signals sample-aligned with the sampling clock signal;

[0202] A second circuit delay of each bit signal on the transmission path is determined according to the verification result to complete the training of the DERR pin, and the second circuit delay enables the multiple bit signals to meet sampling alignment with the same clock pulse of the sampling clock signal.

[0203] According to one aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium is used to store program code, wherein the program code is used to execute the training method of the DERR pin of the high bandwidth memory described in the above-mentioned embodiments.

[0204] According to one aspect of the present application, a computer program product or a computer program is provided, the computer program product or the computer program including computer instructions, the computer instructions being stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the electronic device executes the method provided in various optional implementations of the above-mentioned embodiments.

[0205] The descriptions of the processes or structures corresponding to the above-mentioned figures have different emphases. For parts that are not described in detail in a certain process or structure, please refer to the relevant descriptions of other processes or structures.

[0206] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein, for example. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0207] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0208] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0209] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0210] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable an electronic device (which can be a computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store program codes.

[0211] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, ordinary technical members in the art should understand that they can still modify the technical solutions recorded in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A training method for a DERR pin of a high bandwidth memory, characterized in that: The method comprises: generating a data word pattern with check errors; sending the data word pattern to a dynamic random access memory; Transmitting a data word error DERR signal fed back by the dynamic random access memory through the DERR pin, wherein the DERR signal includes a plurality of bit signals; Taking each bit signal of the plurality of bit signals as a target bit signal, determining a first circuit delay of the target bit signal on a transmission path represented by the DERR pin, wherein the first circuit delay satisfies sampling alignment between the target bit signal and a sampling clock signal; Under the first circuit delay, checking the clock pulses of the plurality of bit signals sample-aligned with the sampling clock signal; A second circuit delay of each bit signal on the transmission path is determined according to the verification result to complete the training of the DERR pin, and the second circuit delay enables the multiple bit signals to meet sampling alignment with the same clock pulse of the sampling clock signal.

2. The method according to claim 1, characterized in that The determining a first circuit delay of the target bit signal on the transmission path represented by the DERR pin includes: Acquire an initial circuit delay of the target bit signal on the transmission path; Adding a circuit delay on the basis of the initial circuit delay until the value of the collected target bit signal is 0, thereby obtaining a third circuit delay; reducing the circuit delay on the basis of the initial circuit delay until the value of the collected target bit signal is 0, thereby obtaining a fourth circuit delay; The third circuit delay and the fourth circuit delay are averaged to obtain the first circuit delay.

3. The method according to claim 2, characterized in that A delay circuit corresponding to the target bit signal is arranged on the transmission path, and the delay circuit includes a plurality of delay units, each of which is followed by a corresponding tap interface, and the delay circuit is controlled by a step counter, and the step counter is used to control the target bit signal to be connected from the target tap interface. After the target bit signal is connected from the target tap interface, the number of delay units through which the target bit signal passes before the target tap determines the size of the circuit delay of the target bit signal on the transmission path.

4. The method according to claim 3, characterized in that The step of increasing the circuit delay on the basis of the initial circuit delay until the value of the collected target bit signal is 0 to obtain a third circuit delay includes: On the basis of the initial circuit delay, the circuit delay is increased by adjusting the value of the step counter until the value of the collected target bit signal becomes 0, thereby obtaining the third circuit delay, wherein the value of the step counter represents the number of delay units that the target bit signal passes through before the target tap, and the third circuit delay is the first number of delay units that the target bit signal passes through; The step of reducing the circuit delay on the basis of the initial circuit delay until the value of the collected target bit signal is 0 to obtain a fourth circuit delay includes: On the basis of the initial circuit delay, the circuit delay is reduced by adjusting the value of the step counter until the value of the collected target bit signal becomes 0, thereby obtaining the fourth circuit delay, wherein the fourth circuit delay is the second number of delay units passed by the target bit signal; The averaging the third circuit delay and the fourth circuit delay to obtain the first circuit delay includes: The first number and the second number are averaged to obtain the first circuit delay.

5. The method according to any one of claims 2 to 4, characterized in that: The obtaining an initial circuit delay of the target bit signal on the transmission path includes: Acquire an initial sub-circuit delay of the target bit signal on the transmission path; comparing the initial sub-circuit delay with a preset circuit delay threshold; If the comparison result is that the initial sub-circuit delay is greater than the preset circuit delay threshold, the operating frequency of the host that generates the data word pattern is reduced until the adjusted initial sub-circuit delay is less than or equal to the preset circuit delay threshold, and the adjusted initial sub-circuit delay is determined as the initial circuit delay.

6. The method according to claim 1, characterized in that The step of checking the clock pulses of the plurality of bit signals sampled and aligned with the sampling clock signal under the first circuit delay comprises: Checking whether the clock pulses of the plurality of bit signals and the sampling clock signal sampling alignment are the same; The step of determining the second circuit delay of each bit signal on the transmission path according to the verification result includes: If the verification result indicates that the clock pulses of the sampling alignment of the multiple bit signals and the sampling clock signal are different, the first circuit delay is adjusted until the multiple bit signals respectively meet the sampling alignment with the same clock pulse of the sampling clock signal, thereby obtaining the second circuit delay; If the verification result indicates that the plurality of bit signals are sampled and aligned with the sampling clock signal at the same clock pulse, the first circuit delay is determined as the second circuit delay of the corresponding bit signal.

7. The method according to any one of claims 1 to 4, characterized in that: The method further comprises: According to the second circuit delay of each bit signal on the corresponding transmission path, configuring a target delay circuit on the corresponding transmission path; The target delay circuit is used to perform offset adjustment on the bit signal in the transmitted target DERR signal.

8. The method according to claim 7, characterized in that If the second circuit delay is the number of delay units passed by the bit signal, configuring the target delay circuit on the corresponding transmission path according to the second circuit delay of each bit signal on the corresponding transmission path includes: The value of the step counter in the target delay circuit is adjusted to the number of delay units passed by the bit signal.

9. The method according to any one of claims 1 to 4, characterized in that: The method of generating a data word pattern with a check error comprises: generating a data word pattern with a check error according to a preset period; Alternatively, if it is determined that the degree of change in the environmental information exceeds a preset change threshold, a data word pattern with a check error is generated.

10. The method according to any one of claims 1 to 4, characterized in that: The data word pattern includes data, a data mask signal, and a data bus flip signal. The data word pattern with a check error is generated, including: Acquire data word configuration information, where the data word configuration information is used to indicate usage of the data, the data mask signal, and the data bus flip signal; A data word pattern with a check error is generated according to the use of the data, the data mask signal and the data bus inversion signal.

11. The method according to claim 10, characterized in that The usage scenarios include any of the following: The data, the data mask signal and the data bus flip signal are all enabled; The data and the data bus flip signal are both enabled, and the data mask signal is disabled; The data is enabled, and the data mask signal and the data bus flip signal are both disabled.

12. A training device for a DERR pin of a high bandwidth memory, characterized in that: The device comprises a generating unit, a sending unit, a transmission unit, a determining unit and a checking unit: The generating unit is used to generate a data word pattern with a check error; The sending unit is used to send the data word pattern to the dynamic random access memory; The transmission unit is used to transmit a data word error DERR signal fed back by the dynamic random access memory through the DERR pin, wherein the DERR signal includes a plurality of bit signals; The determining unit is configured to use each bit signal of the multiple bit signals as a target bit signal, and determine a first circuit delay of the target bit signal on a transmission path represented by the DERR pin, wherein the first circuit delay satisfies sampling alignment between the target bit signal and the sampling clock signal; The verification unit is used to verify the clock pulses of the plurality of bit signals sampled and aligned with the sampling clock signal under the delay of the first circuit; The determination unit is further used to determine a second circuit delay of each bit signal on the transmission path according to the verification result to complete the training of the DERR pin, and the second circuit delay enables the multiple bit signals to meet sampling alignment with the same clock pulse of the sampling clock signal respectively.

13. An electronic device, characterized in that: The electronic device comprises a processor and a memory: The memory is used to store program code and transmit the program code to the processor; The processor is configured to execute the method according to any one of claims 1 to 11 according to the instructions in the program code.

14. A computer-readable storage medium, characterized in that: The computer-readable storage medium is used to store program codes, which, when executed by a processor, enable the processor to perform the method according to any one of claims 1 to 11.

15. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 11 is implemented.

Citation Information

Patent Citations

  • Training for mapping swizzled data command / address signals

    CN104903877A

  • Memory system, operation method of the same and memory controller

    CN111354390A