Physical memory distribution system

By introducing exception handling modules to uniformly process exception instructions for CB data in the physical memory allocation system, and using CB data pre-read module to improve data reading efficiency, the problems of low throughput and long-term consumption of existing systems are solved, and the more Efficient memory allocation and system performance optimization.

CN118939574APending Publication Date: 2024-11-12SHANDONG YUNHAI GUOCHUANG CLOUD COMPUTING EQUIP IND INNOVATION CENT CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410969624.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-18
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

In existing physical memory allocation systems, the CB memory allocation system has low throughput and long time, resulting in poor system performance.

Method used

A physical memory allocation system was designed. Through the combination of CB data analysis module, Entry data writing module, List data writing module and exception processing module, the exception processing module is used to uniformly process the exception instructions of multiple CB data to avoid traditional one by one. Processing method to improve system throughput and efficiency, and improve data reading efficiency through CB data pre-read module.

Benefits of technology

It significantly improves system throughput and memory allocation speed, reduces memory allocation time, enhances the overall efficiency and robustness of the system, and optimizes performance and coupling.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118939574A_ABST
    Figure CN118939574A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of physical memory allocation, and discloses a physical memory allocation system. According to the invention, the exception handling module carries out exception handling on at least one of the first exception instruction generated by the CB data analysis module, the second exception instruction generated by the Entry data writing module and the third exception instruction generated by the Entry data writing module, so that the throughput of the system can be remarkably improved, and the time consumption of memory allocation can be remarkably reduced. Meanwhile, the utilization efficiency of an interface or a bus can be enhanced by processing multiple pieces of CB data, in addition, in the CB data multi-stage pipeline processing process, due to the fact that any module can directly enter an exception processing module when determining that related information of multiple pieces of target CB data is abnormal, the overall efficiency of the system and the robustness of the system can be improved, and the reliability of the system is improved. And the overall performance of the system can be obviously optimized.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of physical memory allocation, and in particular to a physical memory allocation system. Background Art

[0002] Universal computing acceleration architecture (UAA) is a newly defined computing storage architecture. Its core idea is to optimize the division of software and hardware interfaces through micro-architecture innovation and microcode-driven approach. It not only takes into account the high performance of hardware execution, but also maintains the high flexibility and scalability of software programmability. In the UAA architecture, control pages (CP) and control blocks (CB) are defined. CP usually corresponds to an IO request or management request from the host. In a CP, there are several CBs in sequence, and these CBs correspond one by one to complete a series of operation steps required for the IO management command request.

[0003] like Figure 1A As shown in FIG. , it is a general design flow diagram of a physical memory allocation system in the related art. Figure 1B The figure shows the CB hardware analysis diagram of the physical memory allocation system. Figure 1B In the example, the input is the address of the CB. After the address is obtained at the entry, the CB pre-reading module actively initiates a read operation to read the CB back to the local. Then, the CB parsing module obtains the CB read back to the local, performs CB parsing, and completes the memory allocation of the corresponding operation according to the regulations of the CB. After completing the memory allocation, it is written out to complete, informing the peripheral control module that this CB has been completed.

[0004] Therefore, the related art uses a pipeline method to process CBs one by one in sequence, and determines whether the CB processing is abnormal after the last node completes the CB processing. The system throughput is low and the time consumption is long. Summary of the invention

[0005] In view of this, the present invention provides a physical memory allocation system to solve the problem of low throughput and long time consumption of the CB memory allocation system in the related art.

[0006] According to a first aspect, an embodiment of the present disclosure provides a physical memory allocation system, including:

[0007] The CB data parsing module is used to parse the data parsing information respectively corresponding to the multiple target CB data, and when it is determined that any target CB data among the multiple target CB data is abnormal, send a first abnormal instruction corresponding to the abnormal target CB data to the abnormal processing module;

[0008] An Entry data writing module is used to send a second exception instruction corresponding to the Entry memory pointer exception to the exception handling module when determining that the received Entry memory pointer is abnormal;

[0009] A List data writing module, configured to send a third exception instruction to the exception handling module when determining that the received List memory pointer is abnormal;

[0010] The exception handling module is used to, when receiving at least one of the first exception instruction, the second exception instruction and the third exception instruction, execute an exception handling operation corresponding to the received exception instruction according to the received exception instruction and cache the exception handling result.

[0011] In combination with the above-mentioned implementation, at least one of the first abnormal instruction generated by the CB data parsing module, the second abnormal instruction generated by the Entry data writing module, and the third abnormal instruction generated by the Entry data writing module is subjected to abnormal processing by the abnormal processing module. There is no need to process the CBs one by one in sequence in the traditional way. After processing one CB, the next CB is processed, and finally the abnormal CBs are processed uniformly. Therefore, the present invention can significantly improve the system throughput and significantly reduce the memory allocation time consumption. At the same time, processing multiple CB data can enhance the interface or bus utilization efficiency. In addition, in the process of multi-stage pipeline processing of CB data, because any module determines that the relevant information of multiple target CB data is abnormal, it can directly enter the abnormal processing module, so it can improve the overall system efficiency and system robustness, and can also significantly optimize system performance.

[0012] In an optional implementation, the physical memory allocation system in the embodiment of the present disclosure further includes:

[0013] A CB data pre-reading module, used for caching target CB data corresponding to multiple target CB addresses respectively;

[0014] The data writing execution module is used to execute a writing operation according to the received writing instruction when receiving at least one of the writing instructions transferred by the Entry data writing module, the List data writing module and the exception handling module.

[0015] In combination with the above implementation, the CB data pre-reading module is used to automatically read multiple target CB data, which can improve the data reading efficiency. In addition, a unified data write execution module is used to complete the state interaction control of the current target CB data and the peripheral modules, which reduces the coupling degree compared to the general design.

[0016] In an optional embodiment, the data parsing information corresponding to multiple target CB data respectively includes: data format information, allocation mode information, memory location information, and memory size information, and the data format information includes: data list mode and data hash mode; the Entry data writing module is further used for at least one of the following: if the target CB data is in data hash mode, after opening up the CB memory space, the write instruction is directly transferred to the data write execution module to perform the write operation; if the target CB data is in data list mode, the corresponding Entry data is written and cached to wait for the List data writing module to read it and then perform the List write operation.

[0017] In combination with the above implementation, the Entry data writing module uses a memory allocation mechanism in multiple modes (data hash mode or data list mode), which can accurately identify the current mode to terminate the pipeline in advance, thereby achieving the purpose of shortening the delay.

[0018] In an optional implementation, the exception handling module includes:

[0019] A first arbitrator, configured to receive at least one of a first abnormal instruction, a second abnormal instruction, and a third abnormal instruction in a round-robin arbitration manner;

[0020] An exception analysis unit, used to perform exception analysis on the abnormal instructions received by using the round-robin arbitration method to obtain an exception analysis result corresponding to the abnormal instructions;

[0021] An interrupt processing unit, used for interrupt processing of abnormal analysis results;

[0022] The first cache unit is used to cache the processed interrupt processing information, generate a write instruction of the interrupt processing information based on the interrupt processing information, and send the write instruction to the data write execution module.

[0023] In combination with the above implementation, the exception handling module performs exception handling on the exception sources from the CB data parsing module, the Entry data writing module and the List data writing module, so that the outside can quickly locate and handle the exception by responding to interrupts or reading registers, and it is beneficial to improve the overall efficiency of the system.

[0024] In an optional embodiment, the List data writing module includes: a List data writing unit, which is used for at least one of the following: when the read List memory pointer is abnormal, sending a third exception instruction to the exception handling module; when the read List memory pointer is normal, writing the List data in the List memory space indicated by the List memory pointer based on the allocation mode information and the memory size information; a second cache unit, which is used to cache the List data and send a corresponding write instruction for writing the List data to the data write execution module when receiving the List write response instruction.

[0025] In combination with the above implementation, when an exception occurs in the List memory pointer, the List data writing module immediately informs the exception handling module to handle it, which significantly improves the system exception handling efficiency. In addition, the use of allocation mode information is conducive to accurately identifying the current mode to terminate the pipeline in advance, thereby achieving the purpose of shortening the delay.

[0026] In an optional embodiment, the Entry data write module includes: an Entry data write unit, which is used for at least one of the following: when the read Entry memory pointer is abnormal, sending the second abnormal instruction to the exception handling module; wherein the Entry memory pointer is provided by an external host, and the Entry data write unit is specifically used to perform a CRC check on the Entry memory pointer when the Entry memory pointer enters the Entry data write unit, and if the CRC check is abnormal, forwarding the second abnormal instruction to the exception handling module to perform the abnormal handling operation; when the read Entry memory pointer is normal, based on the data hash mode and memory location information, after correspondingly opening up the CB memory space, directly forwarding the write instruction to the data write execution module to perform the write operation; when the read Entry memory pointer is normal, based on the data list mode, memory location information, memory size information and Entry memory pointer, writing the Entry data correspondingly in the Entry memory space; a third cache unit, which is used to cache the Entry data when the read Entry memory pointer is normal, and when receiving the Entry write response instruction, sending the corresponding write instruction for writing the Entry data to the data write execution module.

[0027] In combination with the above implementation, when an exception occurs in the Entry memory pointer, the Entry data writing module immediately informs the exception handling module to handle it, which significantly improves the system exception handling efficiency. In addition, the memory allocation mechanism using multiple modes (data hash mode or data list mode) can accurately identify the current mode to terminate the pipeline in advance, thereby achieving the purpose of shortening the delay.

[0028] In an optional implementation, writing data into an execution module includes:

[0029] The second arbitrator is used to receive the write instruction sent by at least one of the exception handling module, the Entry data writing module, and the List data writing module in a round-robin arbitration manner;

[0030] The data writing execution unit is used to execute a writing operation based on a writing instruction sent by at least one of the exception handling module, the Entry data writing module, and the List data writing module.

[0031] In combination with the above implementation, the exception handling module, the Entry data writing module, and the List data writing module use a unified data writing execution module to complete the state interaction control of the current target CB data and the peripheral modules, which reduces the coupling degree compared to the general design.

[0032] In an optional implementation, the allocation mode information includes: entity list mode information and shell list mode information;

[0033] The List data writing module is used for at least one of the following: if the allocation mode information is entity list mode information, the List data is written in the List memory space in serial order; if the allocation mode information is shell list mode information, the List data is written in the List memory space according to the memory size information.

[0034] In combination with the above implementation, a memory allocation mechanism using multiple allocation mode information (data hash mode or data list mode) can accurately identify the current mode to terminate the pipeline in advance, thereby achieving the purpose of shortening the delay.

[0035] In an optional implementation, the CB data parsing module includes:

[0036] The CB data parsing unit is used to parse out data parsing information corresponding to the multiple target CB data respectively, and send an exception instruction to the exception handling module when it is determined that there is abnormal target CB data among the multiple target CB data;

[0037] Data parsing register, used to optimize the timing of data parsing information corresponding to multiple target CB data;

[0038] The fourth cache unit is used to cache data parsing information corresponding to the normal target CB data when the CB data parsing unit determines that any target CB data among the multiple target CB data is normal, and send a parsing completion instruction to the Entry data writing module.

[0039] In combination with the above implementation, when an exception occurs in the data analysis information, the CB data analysis module immediately informs the exception handling module to handle it, which significantly improves the system exception handling efficiency. In addition, the CB data analysis module is encapsulated separately, thereby providing a better timing path.

[0040] In an optional implementation, the CB data pre-reading module includes:

[0041] a fifth cache unit, for reading a plurality of target CB addresses;

[0042] A sixth cache unit, used for caching target CB data corresponding to a plurality of target CB addresses respectively;

[0043] The data reading unit is used to determine in real time whether the sixth cache unit is full during the process of caching the target CB data in the sixth cache unit, and if not, continue to send and read multiple target CB addresses to the outside.

[0044] In combination with the above implementation, multiple target CB data are read simultaneously by the CB data pre-reading module, which is beneficial to improving data reading efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1A is a general flow chart of a traditional memory allocation system according to an embodiment of the present invention;

[0047] Figure 1B is a design block diagram of a CB data pre-reading module in a conventional general design framework according to an embodiment of the present invention;

[0048] Figure 2 is a structural block diagram of a physical memory allocation system according to an embodiment of the present invention;

[0049] Figure 3 is a structural block diagram of a CB data parsing module according to an embodiment of the present invention;

[0050] Figure 4A is a schematic diagram of a data structure of a data list mode according to an embodiment of the present invention;

[0051] Figure 4B is a block diagram of the structure of PRP-Lite according to an embodiment of the present invention;

[0052] Figure 5 is a schematic diagram of a data structure of a data hashing mode according to an embodiment of the present invention;

[0053] Figure 6is a structural block diagram of another physical memory allocation system according to an embodiment of the present invention;

[0054] Figure 7 is a structural block diagram of a CB data pre-reading module according to an embodiment of the present invention;

[0055] Figure 8 is a structural block diagram of a data writing execution module according to an embodiment of the present invention;

[0056] Fig. 9 is a schematic diagram of a flow chart of a physical memory allocation system according to an embodiment of the present invention;

[0057] Fig.10 is a structural block diagram of an Entry data writing module according to an embodiment of the present invention;

[0058] Fig.11 is a structural block diagram of a List data writing module according to an embodiment of the present invention;

[0059] Fig.12 is a structural block diagram of an exception handling module according to an embodiment of the present invention;

[0060] Fig.13 is a specific structural block diagram of another physical memory allocation system according to an embodiment of the present invention;

[0061] Fig.14 It is a schematic diagram of time consumption analysis of a physical memory allocation system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0063] In related technologies, such as Figure 1A , 1B As shown, the related art uses a pipeline method to process CBs one by one in sequence. After processing one CB, the next CB is processed. After the last node is processed, it is determined whether the CB processing is abnormal, resulting in low system throughput and long time consumption.

[0064] In view of this, an embodiment of the present disclosure provides a physical memory allocation system. Before introducing the specific implementation of the embodiment of the present disclosure, the key term definitions of the related technology are explained here.

[0065] 1. FIFO, the abbreviation, English full name, First Input First Output, Chinese explanation, First in, First out;

[0066] 2. Abbreviation RRArbiter, English full name, Round-Robin Scheduling Arbiter, Chinese explanation, polling scheduling algorithm arbiter;

[0067] 3. The abbreviation CB, the full English name, Control Block, the Chinese explanation, control block, is equivalent to the concept of command;

[0068] 4. The abbreviation CP, the full name in English, Control Page, Chinese explanation, control page, equivalent to the concept of command set, has a specific data structure;

[0069] 5. NVMe protocol: NVMe protocol is short for Non-volatile memory express. The protocol was developed by the NVMe Association to provide high-speed access specifications for storage memory such as SSDs. It has become the mainstream interface standard protocol for SSD devices.

[0070] 6. PRP Entry: NVMe divides the host's memory into pages. The size of the data page can be 4KB, 8KB, 16KB...128MB. The essence of PRP Entry is a 64-bit memory physical address, but this physical address is divided into two parts: the page start address and the page offset. PRP Entry describes the start address of a continuous physical memory;

[0071] 7. PRP List: In the Nvme protocol, if you need to describe several discontinuous physical memories, you need several PRP Entries. Linking several PRP Entries together is called PRP List;

[0072] 8. PRP-Lite list: A simple and lightweight data organization format defined based on the NVMe protocol PRP list method for organizing data;

[0073] 9. PRP-Lite List entry: Similar to the role of PRP2 in the Nvme protocol, but compared to PRP2, the content contains not only the PRP List entry page address, but also the total length of the PRP-Lite list;

[0074] 10. PRP-Lite List pointer: Pointer to the start address of the next page of PRP-Lite List;

[0075] 11. Bus outstanding capability: Outstanding capability is an important parameter of the AXI bus, which indicates the number of unfinished addresses (i.e., unfinished read and write operations) that a master device can issue during data transmission. Specifically, when a master device initiates a read or write operation to a slave device, it first sends an address signal and waits for a response from the slave device. In this process, if the master device can send multiple address signals at the same time, and these address signals have not yet received a response from the slave device, then the number of these unfinished address signals is called outstanding capability.

[0076] According to an embodiment of the present invention, a physical memory allocation system embodiment is provided, such as Figure 2 As shown, the system includes: a CB data parsing module 21, an Entry data writing module 22, a List data writing module 23, and an exception handling module 24.

[0077] exist Figure 2 In the embodiment, the CB data parsing module 21 is used to parse the data parsing information corresponding to the multiple target CB data respectively, and when it is determined that any target CB data among the multiple target CB data is abnormal, send the first abnormal instruction corresponding to the abnormal target CB data to the abnormal handling module 24. The Entry data writing module 22 is used to send the second abnormal instruction corresponding to the abnormal Entry memory pointer to the abnormal handling module 24 when it is determined that the received Entry memory pointer is abnormal; the List data writing module 23 is used to send the third abnormal instruction to the abnormal handling module 24 when it is determined that the received List memory pointer is abnormal; the abnormal handling module 24 is used to perform the abnormal handling operation corresponding to the received abnormal instruction according to the received abnormal instruction when at least one of the first abnormal instruction, the second abnormal instruction and the third abnormal instruction is received and cache the abnormal handling result.

[0078] in, Figure 2 The CB data parsing module 21 is used to parse the data parsing information corresponding to the multiple target CB data respectively, and when it is determined that any target CB data among the multiple target CB data is abnormal, send the first abnormal instruction corresponding to the abnormal target CB data to the abnormal processing module 24.

[0079] Specifically, the multiple target CB data mentioned above can be different types of control instruction data, for example, order control instruction data or power generation control instruction data or charging control instruction data. The number of multiple target CB data is flexibly set according to the cache unit set in the CB data parsing module. For example, if the cache is large, the number of target CB data can be set more, and if the cache is small, the number of target CB data can be set less. The abnormality of any target CB data among the multiple target CB data includes the parameters parsed by the CB data parsing module, judging whether there are undefined values ​​in the format, and whether the allocated memory size exceeds the limit. The first abnormal instruction mentioned above can be a code instruction. Each abnormal target CB data corresponds to a first abnormal instruction.

[0080] Furthermore, the CB data parsing module can further parse the data parsing information corresponding to the multiple target CB data based on the CB data pre-reading module. The data parsing information corresponding to the multiple target CB data includes: data format information, allocation mode information, memory location information, memory size information, and the data format information includes: data list mode and data hash mode.

[0081] Specifically, Figure 3 The CB data parsing module in is mainly used for CB parsing. It parses and calculates commands according to the CB format and mainly obtains the following key parameters:

[0082] First information (the data format information mentioned above): for example, the data organization format currently processed by the CB is a data list mode or a data hash mode;

[0083] The second information (the allocation mode information mentioned above): allocation operator, which defines a variety of allocation modes;

[0084] The third information (the memory location information mentioned above): the entry location corresponding to the allocated memory;

[0085] The fourth information (the memory size information mentioned above): the size of the allocated memory.

[0086] Exemplarily, the CB data parsing module in the disclosed embodiment determines whether there are undefined values ​​in the format and whether the allocated memory size exceeds the limit based on the above parsed parameters. If there is an exception, the command needs to be transferred to the exception handling module. During the parsing process, the CB data parsing module writes the commands without problems into the cache and waits for the lower-level pipeline to process. If the cache is full, the CB parsing operation is back-pressured, and the operation of taking CB from the upper-level cache is further back-pressured.

[0087] When the target CB data is relatively complex, the embodiment of the present disclosure may further insert a first-level register, namely the data parsing register mentioned above, to achieve the effect of optimizing the timing.

[0088] The data organization formats under the UAA architecture include the data list mode and data hash mode mentioned above. The data list mode is as follows: Figure 4A , Figure 4B As shown, the data hashing pattern is as follows Figure 5 shown.

[0089] exist Figure 4A In the linked list part, the data organization format of the PRP-Lite list is defined based on the data organization method of the NVMe protocol PRP list.

[0090] exist Figure 4B In the PRP-Lite list entry, the total length of the list and the starting address of the list are included. The PRP-Lite list pointer only contains the starting address of the next page. The remaining shaded part is similar to the PRP Entry in the PRPList, which contains the address pointing to the data page.

[0091] exist Figure 5 In the embodiment of the present disclosure, there is also a special format CB corresponding to the memory allocation (Allocation) to complete the memory allocation of the above two data organization modes. The information in the Allocation CB mainly includes the following parameters:

[0092] 1. The data organization format currently processed by CB: data list mode or data hash mode;

[0093] 2. Allocation operator, list mode defines two allocation modes: entity list mode and shell list mode, data hash mode only defines one allocation mode: entity list mode and shell list mode, data hash mode only defines one allocation mode: entity list mode;

[0094] The difference between the data hash mode and the data list mode is that the memory space allocated by the data hash mode is all on the CP, so the space that can be opened up is smaller; the memory space allocated by the data list mode only has the entry on the CP, and the rest of the space is represented by a linked list, which may be located on other physical storages, so the space that can be opened up is larger.

[0095] 3. In data list mode, Figure 4AIn the , the entry corresponding to the allocated memory is which of the 8 PRP-Lite listentry. The data hashing mode is not controlled by this field;

[0096] 4. The size of the memory space allocated.

[0097] The disclosed embodiment uses a CB data parsing module to separately parse out data parsing information corresponding to a plurality of target CB data, which is beneficial to optimizing a longer combinational logic path.

[0098] in, Figure 2 The Entry data writing module in is used to send a second exception instruction corresponding to the Entry memory pointer exception to the exception handling module 24 when it is determined that the received Entry memory pointer is abnormal.

[0099] Specifically, in the disclosed embodiment, the Entry memory pointer can be an index of the Entry memory space to be opened provided by an external host. The Entry data writing module writes the entry to the corresponding address according to the parsed command. The entry data is written to the opened memory space. The index of the memory space to be opened is provided by the external host. When this memory pointer enters the module, a CRC check is required. If there is a problem with the CRC check, it needs to be transferred to the exception handling module. The third exception instruction mentioned above is a code instruction for the Entry memory pointer exception.

[0100] in, Figure 2 The List data writing module 23 is used to send a third exception instruction to the exception handling module 24 when it is determined that the received List memory pointer is abnormal.

[0101] Specifically, the List data writing module in the embodiment of the present disclosure accepts the data list mode command forwarded by the previous stage (i.e., accepts the data list mode command forwarded by the previous stage Entry data writing module). The List memory pointer can be an index of the List memory space to be opened provided by the external host. The third abnormal instruction in the above corresponds to the code instruction of the List memory pointer abnormality.

[0102] For example, when reading back the List memory pointer, a CRC check is performed. If an error occurs in the CRC check, the command is forwarded to the exception handler. If the List data writing module is in the CRC check process, if there is no problem in reading back the List memory pointer, the write list operation is completed, and then the command is written into the cache. After the write response is returned, the command is considered to be effective. After it is effective, the subsequent module can take the command from the cache to perform the next operation.

[0103] Among them, Figure 2The exception handling module 24 in the embodiment is used to, when receiving at least one of the first exception instruction, the second exception instruction and the third exception instruction, execute the exception handling operation corresponding to the received exception instruction according to the received exception instruction and cache the exception handling result.

[0104] Exemplarily, the exception handling module in the disclosed embodiment can use polling arbitration to process the exception parsing application of the previous module. The exception handling module has a total of 3 exception sources that need to be parsed, namely, the exception sources from the CB data parsing module, the Entry data writing module, and the List data writing module. After receiving the three different arbitration channel authorizations, the exception handling module generates different interrupt states according to the channel numbers, or sets a specific register to indicate what kind of error value has occurred in the corresponding channel (generating an interrupt and writing a register are mutually exclusive to avoid repeated processing of the same exception peripheral), so that the outside can quickly locate and handle the exception by responding to the interrupt or reading the register.

[0105] In summary, the physical memory allocation system in the disclosed embodiment performs exception handling on at least one of the first exception instruction generated by the CB data parsing module, the second exception instruction generated by the Entry data writing module, and the third exception instruction generated by the Entry data writing module through the exception handling module. There is no need to process CBs one by one in sequence in the traditional way. After processing one CB, the next CB is processed, and finally the abnormal CBs are processed uniformly. Therefore, the present invention can significantly improve the system throughput and significantly reduce the memory allocation time. At the same time, processing multiple CB data can enhance the interface or bus utilization efficiency. In addition, in the process of multi-stage pipeline processing of CB data, because any module determines that the relevant information of multiple target CB data is abnormal, it can directly enter the exception handling module, so the overall system efficiency and system robustness can be improved.

[0106] In the physical memory allocation system provided in the above embodiment of the present application, alternatively or additionally, as Figure 3 As shown, the CB data parsing module 21 includes:

[0107] The CB data parsing unit 210 is used to parse out data parsing information corresponding to the multiple target CB data respectively, and send an exception instruction to the exception handling module when it is determined that there is an abnormal target CB data among the multiple target CB data;

[0108] The data analysis register 211 is used to optimize the timing of the data analysis information corresponding to the multiple target CB data respectively;

[0109] The fourth cache unit 212 is used to cache the data parsing information corresponding to the normal target CB data when the CB data parsing unit 210 determines that any target CB data among the multiple target CB data is normal, and send a parsing completion instruction to the Entry data writing module. The data parsing information corresponding to the multiple target CB data respectively includes: data format information, allocation mode information, memory location information, memory size information, and the data format information includes: data list mode and data hash mode.

[0110] Specifically, the data parsing information data has been specifically introduced above and will not be repeated here. The CB data parsing unit in the disclosed embodiment determines whether there are any values ​​that are not defined in the format, and whether the size of the allocated memory exceeds the limit based on the above parsed parameters. If there is an exception, the command needs to be transferred to the exception handling module. The data parsing register mentioned above can serve the purpose of further optimizing the timing of the parsing work of the CB data parsing unit. During the parsing process, the CB data parsing unit writes commands without problems into the fourth cache unit to wait for the lower-level pipeline to process. If the cache of the fourth cache unit mentioned above is full, the CB parsing operation is back-pressured, and the operation of taking CB from the upper-level cache is further back-pressured.

[0111] The disclosed embodiment uses the CB data analysis module to immediately inform the exception handling module when an exception occurs in the data analysis information, thereby significantly improving the system exception handling efficiency. In addition, the CB data analysis module is separately encapsulated to provide a better timing path.

[0112] In the physical memory allocation system provided in the above embodiment of the present application, alternatively or additionally, as Figure 6 As shown, the physical memory allocation system in the embodiment of the present disclosure further includes: a CB data pre-reading module 20 and a data writing execution module 25. The CB data pre-reading module 20 is used to cache target CB data corresponding to multiple target CB addresses respectively; the data writing execution module 25 is used to perform a writing operation according to the received writing instruction when receiving at least one of the writing instructions transferred by the Entry data writing module 22, the List data writing module 23, and the exception handling module 24.

[0113] Exemplarily, the CB data pre-reading module has the CB address in the entry cache, that is, the pre-reading mechanism is started to read the CB, and the pre-reading will not be stopped until the exit cache is full. The data write execution module mentioned above adopts polling arbitration to receive data write execution requests from three channels: the exception handling module, the Entry data write module, and the List data write module. After receiving the command from the arbitration authorization channel, the data write execution module writes a completion signal to the corresponding field segment of the completion register according to the agreed address of the completion register to notify the peripheral control module that the memory allocation (Allocation) has completed the processing of this CB and can proceed to the next step. The data write execution module can include abnormal or normal flag information in the completion field segment to notify the status of this CB when running in the memory allocation (Allocation).

[0114] The disclosed embodiment realizes automatic reading of multiple target CB data through a CB data pre-reading module, which can improve data reading efficiency, and uses a unified data write execution module to complete the interactive control of the status of the current target CB data and the peripheral modules, which reduces the coupling degree compared with the general design.

[0115] In the physical memory allocation system provided in the above embodiment of the present application, alternatively or additionally, as Figure 7 As shown, the physical memory allocation system in the embodiment of the present disclosure further includes: a CB data pre-reading module 20, including:

[0116] A fifth cache unit 200, used for reading multiple target CB addresses;

[0117] A sixth cache unit 201, used for caching target CB data corresponding to a plurality of target CB addresses respectively;

[0118] The data reading unit 202 is used to determine in real time whether the sixth cache unit 201 is full when the sixth cache unit caches the target CB data. If not, continue to send multiple target CB addresses to the outside for reading.

[0119] Specifically, the fifth cache unit mentioned above is an entry cache unit of the CB data pre-reading module, and the sixth cache unit is an exit cache unit of the CB data pre-reading module.

[0120] For example, in Figure 7In the process, the CB address is taken from the fifth cache unit 200 to send a read request to the outside, and the fixed length of CB is read from the corresponding address and returned to the local, and written into the internal cache. The subsequent pipeline can get a valid output as long as the cache is not empty. The read control logic determines whether the sixth cache unit 201 is full every time a read request is sent out. If it is not full, the read request is sent. If it is full, the read request is sent out after the sixth cache unit has a CB read out and becomes non-full.

[0121] The disclosed embodiment simultaneously reads multiple target CB data through a CB data pre-reading module, which is beneficial to improving data reading efficiency.

[0122] In the physical memory allocation system provided in the above embodiment of the present application, alternatively or additionally, as Figure 8 As shown, the data writing execution module 25 in the embodiment of the present disclosure includes:

[0123] The second arbitrator 250 is used to receive a write instruction sent by at least one of the exception handling module, the Entry data writing module, and the List data writing module in a round-robin arbitration manner;

[0124] The data writing execution unit 251 is used to execute a writing operation based on a writing instruction sent by at least one of the exception handling module, the Entry data writing module, and the List data writing module.

[0125] For example, in Figure 6 In the embodiment of the present disclosure, the data write execution module 25 receives the data write execution requests of the three channels of the exception handling module 24, the entry data write module 22, and the list data write module 23 by means of polling arbitration. After receiving the command of the arbitration authorization channel, the data write execution module 25 writes a completion signal to the corresponding field segment of the completion register according to the agreed address of the completion register to notify the peripheral control module that the memory allocation (Allocation) has completed the processing of this CB and can proceed to the next step. The data write execution module 25 can include abnormal or normal flag information in the completion field segment to notify the status of this CB when running in the memory allocation (Allocation).

[0126] During the execution of the traditional general architecture, the lower-level nodes rely on the upper-level nodes to complete data writing, that is, the interdependence between adjacent nodes is strong, which is not conducive to the subsequent secondary development and modification of the system, that is, the logic of the hardware design framework is too coupled.

[0127] Therefore, in the embodiment of the present disclosure, the exception handling module, the Entry data writing module, and the List data writing module use a unified data writing execution module to complete the state interaction control of the current target CB data and the peripheral modules, which reduces the degree of coupling compared to the general design.

[0128] like Fig. 9 The figure is a flowchart of the work flow of the physical memory allocation system in the embodiment of the present disclosure.

[0129] In the physical memory allocation system provided in the above embodiment of the present application, alternatively or additionally, the data parsing information corresponding to the multiple target CB data respectively includes: data format information, allocation mode information, memory location information, memory size information, and the data format information includes: data list mode and data hash mode. The specific content of the data parsing information data corresponding to the multiple target CB data respectively has been introduced above and will not be repeated here.

[0130] exist Figure 6 In the example, the Entry data writing module 22 is further used for at least one of the following: if the target CB data is in data hash mode, after the CB memory space is opened, the write instruction is directly transferred to the data write execution module to execute the write operation; if the target CB data is in data list mode, the corresponding Entry data is written and cached to wait for the List data writing module to read it and then execute the List write operation.

[0131] For example, in Figure 6 In the example, if the memory pointer read back by the Entry data writing module 22 is OK, it is necessary to write entry to the corresponding address to open up memory space. The corresponding address is determined by the third information (the memory location information mentioned above). At this time, the size of the memory opened up is jointly determined by the first information (the data format information mentioned above) and the fourth information (memory size information). If the first information is a data hash mode, it is necessary to determine the size of the memory opened up according to the fourth information and then directly transfer the data to the write execution module. If the first information is a data list mode, it is only necessary to open up the entry of the PRP-Lite list, and the remaining memory opening must be completed in the write list.

[0132] Furthermore, after completing the entry write operation, the Entry data write module writes the current command into the cache, but the content in the cache does not take effect immediately. When the write response is returned, the command takes effect and becomes a state that can be read by the subsequent module to perform the operation of the next level pipeline.

[0133] The Entry data writing module in the disclosed embodiment uses a memory allocation mechanism with multiple modes (data hash mode or data list mode), which can accurately identify the current mode to terminate the pipeline in advance, thereby achieving the purpose of shortening the delay.

[0134] In the physical memory allocation system provided in the above embodiment of the present application, alternatively or additionally, as Fig.10 As shown, the Entry data writing module 22 in the embodiment of the present disclosure includes: an Entry data writing unit 220 and a third cache unit 221 . Among them, the Entry data writing unit 220 is used for at least one of the following: when the read Entry memory pointer is abnormal, sending the second abnormal instruction to the exception handling module; wherein the Entry memory pointer is provided by an external host, and the Entry data writing unit 220 is specifically used for, when the Entry memory pointer enters the Entry data writing unit 220, performing a CRC check on the Entry memory pointer, and if the CRC check is abnormal, transferring the second abnormal instruction to the exception handling module to perform the abnormal handling operation; when the read Entry memory pointer is normal, based on the data hash mode and the memory location information, after the corresponding CB memory space is opened, directly transferring the write instruction to the data write execution module to perform the write operation; when the read Entry memory pointer is normal, based on the data list mode, the memory location information, the memory size information and the Entry memory pointer, writing the Entry data in the Entry memory space accordingly; the third cache unit 221 is used for caching the Entry data when the read Entry memory pointer is normal, and sending the corresponding write instruction for writing the Entry data to the data write execution module when the Entry write response instruction is received.

[0135] Exemplarily, the Entry data writing unit in the embodiment of the present disclosure mainly completes the operation of writing the Entry to the corresponding address according to the command parsed by the CB data parsing module. The Entry data is the opened memory space, and the index of the memory space to be opened is provided externally. When this memory pointer enters the module, a CRC check is required. If there is a problem with the CRC check, it needs to be transferred to the exception handling module.

[0136] Furthermore, if the memory pointer read back by the Entry data writing unit is correct, it is necessary to write the Entry to the corresponding address to open up memory space. The corresponding address is determined by the third information (the memory location information mentioned above). At this time, the size of the memory opened up is determined by the first information (the data format information mentioned above) and the fourth information (the memory size information mentioned above). If the first information is a data hash mode, it is necessary to determine the size of the memory opened up according to the fourth information. Fig.10The selector (Demux) in the write module directly transfers the data to the write execution module. If the first information is in data list mode, it is only necessary to open an entry in the PRP-Lite list, and the remaining memory allocation must be completed in the write list.

[0137] Furthermore, after completing the write entry operation, the current command is written into the cache, but the content in the cache does not take effect immediately. When the write response is returned, the command takes effect and becomes a state that can be read by the subsequent module to perform the operation of the next level pipeline.

[0138] The Entry data writing module in the embodiment of the present disclosure uses a memory allocation mechanism with multiple modes (data hash mode or data list mode), which can accurately identify the current mode to terminate the pipeline in advance, thereby achieving the purpose of shortening the delay.

[0139] In the physical memory allocation system provided in the above embodiment of the present application, alternatively or additionally, as Fig.11 As shown, the List data writing module in the embodiment of the present disclosure includes: a List data writing unit 230 and a second cache unit 231.

[0140] The List data writing unit 230 is used for at least one of the following: when the read List memory pointer is abnormal, sending a third abnormal instruction to the abnormality handling module; when the read List memory pointer is normal, writing List data in the List memory space indicated by the List memory pointer based on the allocation mode information and the memory size information;

[0141] The second cache unit 231 is used for caching List data and sending a write instruction corresponding to writing List data to the data write execution module when receiving a List write response instruction.

[0142] In a specific example, the allocation mode information includes: entity list mode information and shell list mode information; a List data writing module is used for at least one of the following: if the allocation mode information is entity list mode information, when the List data is written in the List memory space, the memory is opened in a serial order; if the allocation mode information is shell list mode information, the CB memory space is opened according to the memory size information.

[0143] Exemplarily, the List data writing module in the disclosed embodiment accepts the list mode command forwarded by the previous stage. And according to the second information (the allocation mode information mentioned above, i.e., the allocation operator) and the fourth information (memory size information) defined by the command, the size of the memory to be opened and the designated location of the memory to be opened are determined. The second information (allocation mode information) defines 2 modes, the entity list mode and the shell list mode. The entity list mode means that the size of the space to be opened is the memory size specified by the fourth information (the memory size information mentioned above), and the location corresponding to the memory to be opened is specified by Allocation, that is, each data page requires a memory pointer; the shell list mode means that a data structure capable of storing a memory size specified by the fourth information is opened, and no real data page is opened.

[0144] Furthermore, when the List data writing unit reads back the memory pointer, a CRC check is performed. If an error occurs in the CRC check, the command is forwarded to the exception handling. If there is no problem and the write List operation is completed, the command is then written into the cache. After the write response is returned, the command is considered to be effective. Only after it is effective can the subsequent module take the command from the cache for the next operation.

[0145] Specifically, the List data writing unit writes the list in a serial order. For example, if there are multiple linked lists, when writing the tail pointer data of the previous page linked list, the starting address of the next page linked list is actually obtained. After completing all the memory allocation of the previous page, the memory allocation of the next page is performed sequentially. In the above manner, the memory allocation of the memory size specified by the fourth information is completed.

[0146] The List data writing module in the embodiment of the present disclosure adopts a memory allocation mechanism of multiple modes and operators, which is conducive to accurately identifying the current mode to terminate the pipeline in advance, thereby achieving the purpose of shortening the delay.

[0147] In the physical memory allocation system provided in the above embodiment of the present application, alternatively or additionally, as Fig.12 As shown, the exception handling module 24 in the embodiment of the present disclosure includes: a first arbitrator 240 , an exception parsing unit 241 , an interrupt processing unit 242 , and a first cache unit 243 .

[0148] A first arbitrator 240, configured to receive at least one of the first abnormal instruction, the second abnormal instruction and the third abnormal instruction in a round-robin arbitration manner;

[0149] The exception analysis unit 241 is used to perform exception analysis on the abnormal instructions received by using the round-robin arbitration method to obtain the exception analysis result corresponding to the abnormal instructions;

[0150] An interrupt processing unit 242, used for performing interrupt processing on abnormal analysis results;

[0151] The first cache unit 243 is used to cache the processed interrupt processing information, generate a write instruction of the interrupt processing information based on the interrupt processing information, and send the write instruction to the data write execution module.

[0152] Exemplarily, the exception handling module in the disclosed embodiment uses a polling arbitration method to process the exception analysis application of the previous module. The exception handling module has a total of 3 exception sources that need to be analyzed, namely, the exception sources from the CB data analysis module, the Entry data writing module, and the List data writing module. After receiving the three different arbitration channel authorizations, the exception analysis unit generates different interrupt states according to the channel numbers, or sets a specific register to a value that can indicate what kind of error has occurred in the corresponding channel (generating an interrupt and writing a register are mutually exclusive to avoid repeated processing of the same exception peripherally), so that the outside can quickly locate and resolve the exception by responding to the interrupt or reading the register.

[0153] Furthermore, after the interrupt is triggered or the register write operation is completed, the interrupt processing unit writes the processed command into the cache, waiting for the data write execution module to process the command in the next step.

[0154] Since this exception handling module is also a level of the overall pipeline, integrating exception handling into this level of pipeline has positive significance for the entire system. The disclosed embodiment uses an exception handling module to handle any exception item, and notifies the external module whether the current CB operation status is abnormal by interrupting or writing a register, which is conducive to enhancing the robustness of the system.

[0155] In the physical memory allocation system provided in the above embodiment of the present application, alternatively or additionally, as Fig.13 As shown, the physical memory allocation system in the embodiment of the present disclosure also includes: a bus arbitration module 26, which is used to receive request instructions sent from the CB data pre-reading module 20, the Entry data writing module 22, the List data writing module 23 and the data writing execution module 25 using a round-robin arbitration method.

[0156] For example, in Fig.13In the example, the entire physical memory allocation system inputs multiple target CB addresses, and CB parsing is performed after obtaining multiple target CB addresses. If CB is found to be abnormal during the parsing process, the abnormal command is transferred. If there is no abnormality, the parsed command is placed in the cache and waits for the next level pipeline to process. The command then enters the Entry data write module 22, and a memory space is opened for the Entry information. According to whether it is a list mode or a hash mode, the next operation or all operations of this CB are selected, and when the pointer of the memory space opened (the pointer read back from the outside) is abnormal, the abnormal command is transferred. If it is a data list mode, when the List data write module 23 receives the previous level command, it will perform the operation of writing the list, and a memory space is opened for the List information, which is the same as the upper level pipeline. When the pointer of the memory space opened (the pointer read back from the outside) is abnormal, the abnormal command is transferred. The data write execution module 25 can receive three different input sources: commands forwarded by the entry data write module 22, commands forwarded by the list data write module 23, and exception commands. After receiving any of the above three commands, the data write execution module 25 writes CB completion to the register of the external specific address to notify the peripheral control module that a CB has been completed and the next operation can be performed. A cache is inserted in the middle of each module to support pipeline operation.

[0157] The disclosed embodiment uses a highly efficient access method of an external bus based on multi-channel polling arbitration, which is beneficial for not only avoiding mutual occupation of external interfaces but also improving the occupation efficiency of interfaces or buses.

[0158] In a specific example, the multiple target CB data in the embodiment of the present disclosure are analyzed by taking 4 CBs as an example:

[0159] Assume that it takes 2 seconds to fetch a CB, 5 seconds to parse a CB, 3 seconds to write an entry, 10 seconds to write a list, and 4 seconds to write data into the execution module. Then the total time to process 4 CBs is 2+5+3+10*4+4=54.

[0160] If the conventional general design is used, the next CB fetch operation can be performed only after the current CB is written, so the time consumption is (2+5+3+10+4)*4=96.

[0161] The disclosed embodiment has a performance improvement of (96-54) / 96=43.75% compared to the conventional design. Fig.14 As shown, from Fig.14From the analysis, we can conclude that, under the condition of ensuring sufficient CB, the overall pipeline throughput is only related to the processing time of the write list pipeline, and the more CBs are processed, the closer the throughput is to the number of write list operations per second. That is to say, when the system is running, the performance optimization range should be greater than 43.75% and less than (96-40) / 96=58.33%.

[0162] The physical memory allocation system in the disclosed embodiment can achieve the purpose of optimizing system performance for the following reasons:

[0163] Based on the above assumptions, if the same performance needs to be achieved, then when instantiating the original design, three original designs need to be used to achieve the throughput level of the disclosed embodiment, and the overhead that needs to be increased by the disclosed embodiment is the register inserted between each stage of the pipeline. Therefore, the use of the disclosed embodiment can significantly optimize system performance, save system area, and reduce design costs.

[0164] In addition, the physical memory allocation system in the embodiment of the present disclosure can also enhance robustness for the following reasons:

[0165] The disclosed embodiment has a related mechanism for abnormal monitoring and reporting. When an abnormality is found, it can send an interrupt or write a register to notify the peripheral module that an abnormality has occurred during the operation of the disclosed embodiment. And there is a corresponding abnormal marking signal when data writing is executed. When data writing is executed and the abnormal marking signal is detected, the peripheral module needs to query the aforementioned register to determine the type of abnormality to determine whether it is a fatal error that requires a system reset, or a common error that requires re-issuing CB and other operations. In summary, the disclosed embodiment has stronger robustness than the general design.

[0166] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A physical memory allocation system, characterized in that: The system comprises: A CB data parsing module, configured to parse data parsing information respectively corresponding to a plurality of target CB data, and when determining that any target CB data among the plurality of target CB data is abnormal, send a first abnormal instruction corresponding to the abnormal target CB data to the abnormal processing module; The Entry data writing module is used to send a second exception instruction corresponding to the Entry memory pointer exception to the exception handling module when determining that the received Entry memory pointer is abnormal; A List data writing module, configured to send a third exception instruction to the exception handling module when determining that the received List memory pointer is abnormal; The exception handling module is used to, when receiving at least one of the first exception instruction, the second exception instruction and the third exception instruction, execute an exception handling operation corresponding to the received exception instruction according to the received exception instruction and cache the exception handling result.

2. The physical memory allocation system according to claim 1, characterized in that: Also includes: A CB data pre-reading module, used for caching target CB data corresponding to multiple target CB addresses respectively; The data writing execution module is used to execute a writing operation according to the received writing instruction when receiving at least one of the writing instructions transferred by the Entry data writing module, the List data writing module and the exception handling module.

3. The physical memory allocation system according to claim 2, characterized in that: The data parsing information corresponding to the plurality of target CB data respectively includes: data format information, allocation mode information, memory location information, and memory size information, and the data format information includes: data list mode and data hash mode; The Entry data writing module is further used for at least one of the following: if the target CB data is in data hash mode, after opening up the CB memory space, directly forwarding the write instruction to the data write execution module to perform the write operation; if the target CB data is in data list mode, writing the Entry data accordingly and caching the Entry data to wait for the List data writing module to read it and then perform the List write operation.

4. The physical memory allocation system according to claim 2, characterized in that: The exception handling module includes: A first arbitrator, configured to receive at least one of the first abnormal instruction, the second abnormal instruction and the third abnormal instruction in a round-robin arbitration manner; An exception analysis unit, used to perform exception analysis on the abnormal instructions received by using the round-robin arbitration method to obtain an exception analysis result corresponding to the abnormal instructions; An interrupt processing unit, used for performing interrupt processing on the abnormal analysis result; The first cache unit is used to cache the processed interrupt processing information, generate a write instruction for the interrupt processing information based on the interrupt processing information, and send the write instruction to the data write execution module.

5. The physical memory allocation system according to claim 3, characterized in that: The List data writing module includes: The List data writing unit is used for at least one of the following: when the read List memory pointer is abnormal, sending the third abnormal instruction to the abnormal handling module; when the read List memory pointer is normal, writing List data in the List memory space indicated by the List memory pointer based on the allocation mode information and the memory size information; The second cache unit is used to cache the List data and, when receiving a List write response instruction, send a write instruction corresponding to writing the List data to the data write execution module.

6. The physical memory allocation system according to claim 3, characterized in that: The Entry data writing module includes: The Entry data writing unit is used for at least one of the following: when the read Entry memory pointer is abnormal, sending the second abnormal instruction to the abnormal handling module; The Entry memory pointer is provided by an external host, and the Entry data writing unit is specifically used to perform a CRC check on the Entry memory pointer when the Entry memory pointer enters the Entry data writing unit, and if the CRC check is abnormal, forward the second exception instruction to the exception handling module to perform an exception handling operation; When the read Entry memory pointer is normal, based on the data hash mode and the memory location information, after the CB memory space is correspondingly opened, the write instruction is directly transferred to the data write execution module to perform the write operation; When the read Entry memory pointer is normal, writing Entry data into the Entry memory space accordingly based on the data list mode, the memory location information, the memory size information and the Entry memory pointer; The third cache unit is used to cache the Entry data when the read Entry memory pointer is normal, and to send a write instruction corresponding to writing the Entry data to the data write execution module when receiving an Entry write response instruction.

7. The physical memory allocation system according to claim 2, characterized in that: The data is written into the execution module, including: A second arbitrator, configured to receive a write instruction sent by at least one of the exception handling module, the Entry data writing module, and the List data writing module in a round-robin arbitration manner; The data writing execution unit is used to execute a writing operation based on a writing instruction sent by at least one of the exception handling module, the Entry data writing module, and the List data writing module.

8. The physical memory allocation system according to claim 5, characterized in that: The allocation mode information includes: entity list mode information and shell list mode information; The List data writing module is used for at least one of the following: if the allocation mode information is entity list mode information, the List data is written in the List memory space in serial order; if the allocation mode information is shell list mode information, the List data is written in the List memory space according to the memory size information.

9. The physical memory allocation system according to claim 1, characterized in that: The CB data parsing module includes: The CB data parsing unit is used to parse out data parsing information corresponding to the multiple target CB data respectively, and send an exception instruction to the exception handling module when it is determined that there is abnormal target CB data among the multiple target CB data; Data parsing register, used to optimize the timing of data parsing information corresponding to multiple target CB data; The fourth cache unit is used to cache data parsing information corresponding to the normal target CB data when the CB data parsing unit determines that any target CB data among the multiple target CB data is normal, and send a parsing completion instruction to the Entry data writing module.

10. The physical memory allocation system according to claim 2, characterized in that: The CB data pre-reading module includes: a fifth cache unit, for reading a plurality of target CB addresses; A sixth cache unit, used for caching target CB data corresponding to a plurality of target CB addresses respectively; The data reading unit is used to determine in real time whether the sixth cache unit is full during the process of caching the target CB data in the sixth cache unit, and if not, continue to send and read the multiple target CB addresses to the outside.